You are on page 1of 7

Physica E 118 (2020) 113935

Contents lists available at ScienceDirect

Physica E: Low-dimensional Systems and Nanostructures


journal homepage: www.elsevier.com/locate/physe

Fast optically controlled spin initialization of a quantum dot in the Voigt


geometry coupled to a transition metal dichalcogenide monolayer
Dionisis Stefanatos ∗, Vasilios Karanikolas, Nikos Iliopoulos, Emmanuel Paspalakis
Materials Science Department, School of Natural Sciences, University of Patras, Patras 26504, Greece

ARTICLE INFO ABSTRACT

Keywords: We analyze the problem of fast optical spin initialization for a quantum dot in the Voigt geometry by placing it
Quantum dot near a transition metal dichalcogenide (TMD) monolayer. We calculate the spontaneous emission rates of the
Spin initialization quantum dot modified by the presence of the monolayer for different materials and use the obtained results
Voigt geometry
in quantum dynamics calculations. We show that high levels of fidelity, significantly larger than in the case
Transition metal dichalcogenides
of the quantum dot in free-space, can be quickly obtained due to the anisotropy of the enhanced spontaneous
Optical pumping
Optimal control
decay rates of the quantum dot near the TMD monolayer. We use a continuous wave optical field and find
Spin dynamics the control amplitude which achieves acceptable fidelity levels in short times for different distances of the
quantum dot from the TMD monolayer and for various layer materials. We also use state of the art numerical
optimal control to find the time-dependent electric field which maximizes the final fidelity for the same short
duration and various layer materials as in the previous case. A better fidelity is obtained with this method,
while the resulting pulse is quite robust to positioning error of the quantum dot and to additive constant
control error.

1. Introduction free-space vacuum (or in an isotropic photonic environment) the decay


rates from state |4⟩ to states |1⟩ and |2⟩ are the same. The idea to
Semiconductor quantum dots (QDs), also known as ‘‘artificial accelerate the optical initialization process by utilizing a preferential
atoms’’, are promising candidates for quantum information technolo- Purcell-enhanced deexcitation towards the desired state was proposed
gies [1]. A specific QD structure that has been studied both theo- by Loo et al. [14]. By embedding the QD in a micropillar cavity, they
retically and experimentally is based on the spin states in the Voigt
showed that, due to the Purcell effect, the decay rate from state |4⟩
configuration, where a magnetic field is applied perpendicular to the
to state |2⟩ increases in comparison to the decay rate from state |4⟩
growth axis of the QD [2,3]. The spin initialization, coherent manip-
ulation, and readout in both simple QDs in the Voigt geometry and to state |1⟩, and this leads to preferential deexcitation towards state
integrated QDs in the Voigt geometry with photonic structures have |2⟩ in shorter times than without the micropillar. The idea of using
attracted significant attention in the literature [2–22]. Purcell-enhanced deexcitation towards the desired state for accelerat-
The level structure for a negatively charged QD in the Voigt ge- ing the initialization process in QDs in the Voigt geometry has also
ometry is shown in Fig. 1(a). The natural initial state of the QD is been explored by placing the QD in other photonic structures, like
an incoherent mixture with equal populations in the two electron spin a photonic crystal nanocavity [15], and metallic or metal–dielectric
states, states |1⟩ and |2⟩. A basic process in quantum information proto- nanostructures [16,17,19].
cols is the initialization [23], where, in this system, a specific electron We have recently analyzed the problem of fast optical spin initial-
spin state is created. The most common method for initialization relies
ization of a QD in the Voigt geometry near a single graphene layer
on the optical pumping process [10]. Let us consider that we aim
and found that high levels of fidelity, significantly larger than in the
to create state |2⟩. An optical field with a 𝑥-polarized electric field
is applied and couples spin state |1⟩ to the trion state |4⟩. Then, by case of the QD in free-space, can be quickly obtained due to the
spontaneous decay the population is pumped into spin state |2⟩ after a anisotropy of the Purcell-enhanced spontaneous decay rates of the QD
few decay times. During the pumping process population also returns near the graphene layer [22]. Here, we explore a different class of
back to state |1⟩ due to spontaneous decay. This population needs to two-dimensional (2D) materials for accelerating the optical pumping
be pumped again and this slows down the initialization process. In process in a QD in the Voigt geometry. Specifically, we place the QD

∗ Corresponding author.
E-mail address: dionisis@post.harvard.edu (D. Stefanatos).

https://doi.org/10.1016/j.physe.2019.113935
Received 11 November 2019; Received in revised form 17 December 2019; Accepted 25 December 2019
Available online 26 December 2019
1386-9477/© 2019 Elsevier B.V. All rights reserved.
D. Stefanatos et al. Physica E: Low-dimensional Systems and Nanostructures 118 (2020) 113935

Fig. 1. (a) Energy level diagram for a QD in the Voigt geometry. The magnetic field induces the Zeeman splitting in the upper and lower levels. The optical field resonant with
the |1⟩ ↔ |4⟩ transition is 𝑥-polarized. (b) The QD is placed a distance 𝑧 = 𝑅 above a TMD monolayer, which lies on the 𝑥-𝑦 plane at 𝑧 = 0. The growth axis of the QD is 𝑦 while
the magnetic field is applied along the 𝑥-axis.

next to a transition metal dichalcogenide (TMD) monolayer [24], as and electron spins, respectively. For the level scheme, see Fig. 1(a).
shown in Fig. 1(b). TMD semiconductors of the type MX2 , where M is Here, the vertical transitions (|1⟩ ↔ |4⟩ and |2⟩ ↔ |3⟩) give 𝑥-polarized
a transition metal atom (for example Mo or W) and X is a chalcogen dipole matrix elements and the cross transitions (|1⟩ ↔ |3⟩ and |2⟩ ↔
atom (for example S or Se), offer a promising alternative to graphene as |4⟩) give 𝑧-polarized dipole matrix elements. In this work, we aim to
2D materials for photonics, optoelectronics, and nanoelectronics appli- create state |2⟩. Thus, we take the QD to interact with a 𝑥-polarized
cations [25,26]. TMD monolayers are atomically thin, two-dimensional, laser field applied at exact resonance with the |1⟩ ↔ |4⟩ transition,
direct band gap semiconductors. They feature bandgaps in the near exploring the method of optical pumping.
infrared to the visible region, strong excitonic resonances and high In order to study the dynamics of spin initialization, we derive the
oscillator strengths. They also support exciton-polaritons [27–30] and density matrix equations of the system. The Hamiltonian describing the
can modify the spontaneous decay rates for nearby quantum emit- interaction between the optical field and the QD system, in the dipole
ters [29,31–34]. It has been shown that the spontaneous decay rates and rotating wave approximations, is
for a quantum emitter near a TMD monolayer can be anisotropic for

4
[ ]
emitter dipoles parallel and perpendicular to the layer [29]. Therefore, 𝐻= ℏ𝜔𝑛 |𝑛⟩⟨𝑛| − ℏ 𝛺(𝑡)𝑒−𝑖𝜔0 𝑡 |4⟩⟨1| + 𝛺(𝑡)𝑒−𝑖𝜔0 𝑡 |3⟩⟨2| + H.c. , (1)
a TMD monolayer has the potential for modifying the initialization 𝑛=1
process in a nearby QD in the Voigt geometry by Purcell-enhanced where ℏ𝜔𝑛 , 𝑛 = 1, 2, 3, 4, is the energy of state |𝑛⟩ and 𝛺(𝑡) is the
deexcitation process towards the desired state. generally time-dependent Rabi frequency of the applied field. The field
In this work, we initially calculate the corresponding Purcell factors is resonant with the |1⟩ ↔ |4⟩ transition, thus 𝜔0 = 𝜔4 − 𝜔1 = 𝜔41 . Using
∑4
for the QD next to a TMD monolayer, for four different semiconductor the unitary operator 𝑈 (𝑡) = 𝑒−𝑖 𝑛=1 𝑎𝑛 |𝑛⟩⟨𝑛|𝑡 , with 𝑎1 = 𝜔1 , 𝑎2 = 𝜔0 + 𝜔1 ,
materials, WS2 , MoS2 , WSe2 and MoSe2 . Then, we use them to investi- 𝑎3 = 𝜔3 + 𝜔0 − 𝜔41 , 𝑎4 = 𝜔0 + 𝜔1 , Hamiltonian (1) is transformed to the
gate the dynamics of spin initialization for the composite system and for interaction Hamiltonian
each material using optical pumping. We also use two different types of
optical pulses. First, we apply a continuous wave electromagnetic field 𝐻𝑒𝑓 𝑓 = −ℏ(𝜔0 − 𝜔21 )|2⟩⟨2| − ℏ(𝜔0 − 𝜔41 )|3⟩⟨3| − ℏ(𝜔0 − 𝜔41 )|4⟩⟨4|
[ ]
which gives a constant Rabi frequency, and find the control amplitude
that achieves acceptable fidelity levels in short times for different − ℏ 𝛺(𝑡)|4⟩⟨1| + 𝛺(𝑡)𝑒−𝑖(𝜔0 +𝜔43 )𝑡 |3⟩⟨2| + H.c. , (2)
distances of the QD from the TMD monolayer. Next, we use state of the
where 𝜔𝑛𝑚 = 𝜔𝑛 − 𝜔𝑚 . Note that ℏ𝜔21 , ℏ𝜔43 are the Zeeman splittings
art numerical optimal control to find the time-dependent electric field,
of the single-electron spin states and the heavy-hole spin trion states,
leading to a time-dependent Rabi frequency profile, which maximizes
respectively. Throughout this article we use the values ℏ𝜔21 = 0.124
the final fidelity for the same short duration as in the previous case
meV, ℏ𝜔43 = 0.078 meV [11,21].
and for different layer material qualities. A better fidelity is obtained
Using the above Hamiltonian we can easily obtain the equations of
with this method, while the resulting pulse is quite robust to positioning
motion for the density matrix elements of the QD. In the case where
error of the QD from the TMD layer and to additive constant control
the laser frequency is at exact resonance with the transition |1⟩ ↔ |4⟩,
error.
these are
The structure of the paper is as follows. In the next section we
describe the theoretical model of a QD in the Voigt geometry and the 𝜌̇ 11 = 𝛾41 𝜌44 + 𝛾31 𝜌33 − 2𝛺𝜌𝐼41 , (3)
corresponding equations for its dynamics under the interaction with an
𝜌̇ 22 = 𝛾32 𝜌33 + 𝛾42 𝜌44 + 2𝛺𝜌̃𝐼23 , (4)
optical field, while in Section 3 we calculate the Purcell factors of the
QD modified by the presence of the TMD monolayer. Then, in Section 4 𝜌̇ 33 = −(𝛾31 + 𝛾32 )𝜌33 − 2𝛺𝜌̃𝐼23 , (5)
we study the fast spin initialization using optical control with constant
𝜌̇ 44 = −(𝛾41 + 𝛾42 )𝜌44 + 2𝛺𝜌𝐼41 , (6)
and time-dependent Rabi frequencies. Finally, Section 5 summarizes
𝛾 + 𝛾42 𝐼
this work. 𝜌̇ 𝐼41 = − 41 𝜌41 + 𝛺(𝜌11 − 𝜌44 ), (7)
2
𝛾31 + 𝛾32 𝑅
2. Quantum dot in the Voigt geometry placed near a TMD mono- 𝜌̃̇ 𝑅
23
= − 𝜌̃23 + (𝜔21 + 𝜔43 )𝜌̃𝐼23 , (8)
2
layer 𝛾 + 𝛾32 𝐼
𝜌̃̇ 𝐼23 = − 31 𝜌̃23 − (𝜔21 + 𝜔43 )𝜌̃𝑅23
+ 𝛺(𝜌33 − 𝜌22 ), (9)
2
For the studied system, we consider a singly-charged self-assembled
where 𝜌̃23 = 𝜌23 𝑒−𝑖(𝜔41 +𝜔43 )𝑡 and the superscripts 𝑅, 𝐼 denote real and
QD grown along the 𝑦-axis. By applying an external magnetic field in
imaginary parts, respectively. Note that in the above equations we have
the Voigt geometry, along the 𝑥-axis, the degeneracy of electron/hole
also incorporated the spontaneous emission from the upper to the lower
levels is lifted by Zeeman splitting. Then, the ground spin levels are
energy levels, with
|1⟩ = |↓𝑥 ⟩ and |2⟩ = |↑𝑥 ⟩ and the two excited trion states are |3⟩ =
|↓𝑥 ↑𝑥 ⇑𝑥 ⟩ and |4⟩ = |↓𝑥 ↑𝑥 ⇓𝑥 ⟩. Here, ⇑ (⇓) and ↑ (↓) denote heavy hole 𝛾41 = 𝛾32 = 𝛾𝑥 = 𝐹𝑥 𝛾, 𝛾42 = 𝛾31 = 𝛾𝑧 = 𝐹𝑧 𝛾 (10)

2
D. Stefanatos et al. Physica E: Low-dimensional Systems and Nanostructures 118 (2020) 113935

being the radiative decay rates of the corresponding QD transitions Table 1


Lorentz parameters used in Eq. (20) to describe the optical response of the TMD
modified by the coupling between the QD and the TMD layer, where 𝛾
materials through the dielectric permittivity 𝜀TMD (𝜔). The values of 𝜔TO , 𝜔LO and
is the decay rate in free-space and 𝐹𝑗 , with 𝑗 = 𝑥, 𝑧, is the corresponding 𝛾𝑑 are given in eV [39].
Purcell factor. Throughout this paper the free-space relaxation rate is Material 𝜀∞ 𝜔TO 𝜔LO 𝛾𝑑
taken to be ℏ𝛾 = 1.2 μeV [10,21].
WS2 17 2.0 2.01 0.02
MoS2 21 1.86 1.89 0.07
3. Calculation of Purcell factors WSe2 15.3 1.65 1.67 0.05
MoSe2 21.3 1.54 1.56 0.05

The geometry under consideration consists of a dielectric environ-


ment with permittivity 𝜀1 and a TMD monolayer placed in it. As shown
in Fig. 1(b), the TMD monolayer lies on the 𝑥-𝑦 plane at 𝑧 = 0, while For a layer positioned at the 𝑥–𝑦 plane at 𝑧 = 0, in a homogeneous mate-
the QD is placed a distance 𝑧 = 𝑅 above the layer. The spontaneous rial with dielectric permittivity 𝜀1 , the generalized Fresnel coefficients
emission rate of a QD is proportional to the strength of its transition have the form [38],
dipole moment and the electromagnetic field strength acting on it. We −𝛼𝑘0
introduce the normalized emission rate, known as the Purcell factor, as 𝑅11
𝑀 = , (16)
𝑘𝑧1 + 𝛼𝑘0
𝛼𝑘0 𝑘𝑧1
𝛾𝑛̂ √ 6𝜋𝑐 𝑅11
𝑁 = (17)
𝐹𝑛̂ = = 𝜀1 + Im(𝐧̂ ⋅ G(𝐫, 𝐫, 𝜔) ⋅ 𝐧),
̂ (11) 𝑘21 + 𝛼𝑘0 𝑘𝑧1
𝛾 𝜔
where 𝛼 = 2𝜋𝜎∕𝑐. For a QD positioned at 𝐫QD = (0, 0, 𝑅), the Purcell
where 𝛾(𝜔) = 𝜔3 |𝐩|2 ∕3𝜋𝜖0 ℏ𝑐 3 is the free-space relaxation rate, 𝐫 is the factors corresponding to 𝑧- and 𝑥-orientations of the transition dipole
position of the QD, 𝐧̂ is the unit vector along the direction of the transi- moment are found using the above relations to be:
tion dipole moment 𝐩 of this quantum emitter, 𝜀1 is the permittivity of ( )
the host medium, and G(𝐫, 𝐫 ′ , 𝜔) is electromagnetic Green’s tensor. The √ 3𝑐
∞ 𝑘3𝑠 11 2𝑖𝑘𝑧1 𝑅
𝐹𝑧 = 𝜀1 + Im 𝑖 d𝑘𝑠 𝑅 𝑒 , (18)
latter gives the response of the geometry under consideration to a point- 2𝜔 ∫0 𝑘𝑧1 𝑘21 𝑁
like dipole excitation. The Purcell factor quantifies the enhancement or
[ ( ) ]
inhibition of the spontaneous relaxation of the QD placed closed to a √ ∞ 𝑘 𝑘2𝑧1
3𝑐
nanostructure, compared to its free space value. In the following we 𝐹𝑥 = 𝜀1 + Im 𝑖 d𝑘𝑠 𝑠 𝑅11 𝑀 + 𝑅 11
𝑒 2𝑖𝑘𝑧1 𝑅
, (19)
4𝜔 ∫0 𝑘1 𝑘21 𝑁
provide some details on how we calculate the Green’s tensor for the

TMD monolayer geometry.
where 𝑘𝑠 , 𝑘𝑧1 = 𝑘21 − 𝑘2𝑠 are the in-plane and perpendicular wavevec-
The optical response of TMD is given by its surface conductivity, √
tor components, respectively, and 𝑘1 = 𝜔𝑐 𝜀1 is the wavenumber in the
𝜎, (see more details later in this section). In order to investigate
dielectric medium.
the interaction of the QD with the TMD monolayer, the method of
In order to complete the calculation, it remains to find the in-
scattering superposition is used [35–37], where the Green’s tensor splits
plane conductivity 𝜎 which encodes the optical response of the TMD.
into two parts
The optical response of the TMD monolayer is given by a Lorentz
G(𝐫, 𝐬, 𝜔) = Gℎ (𝐫, 𝐬, 𝜔) + G𝑠 (𝐫, 𝐬, 𝜔). (12) model [39]:
( 2 )
𝜔LO − 𝜔2 − 𝑖𝛾𝑑 𝜔
The homogeneous part Gℎ (𝐫, 𝐬, 𝜔) accounts for direct interaction be- 𝜀TMD (𝜔) = 𝜀∞ , (20)
tween the source and target point at 𝐬 and 𝐫 respectively, and is 𝜔2TO − 𝜔2 − 𝑖𝛾𝑑 𝜔
non-zero when both points are in the same media and there is no where 𝜀∞ is the optical dielectric constant, 𝜔LO and 𝜔TO present the
discontinuity between them. The scattering part G𝑠 (𝐫, 𝐬, 𝜔) is always longitudinal and transverse exciton energies, respectively, and 𝛾𝑑 is
present and accounts for the multiple reflections and transmissions the exciton lifetime. We consider the interaction of a QD with WS2 ,
taking place at the interface. MoS2 , WSe2 and MoSe2 monolayers and the value of the 𝜀∞ , 𝜔LO ,
The scattering part of Green’s tensor is expanded in the plane-wave 𝜔TO and 𝛾𝑑 parameters are given in Table 1. In order to emphasize
set. The general form of the scattering part for the layer geometry is: the 2D nature of the TMDs monolayers we consider them as infinitely
1 ∑ ±(𝑖𝑗)±
thin using the surface conductivity which is given by the expression
𝑖
G(𝑖𝑗)
𝑠 (𝐫, 𝐬, 𝜔) = d2 𝑘𝑠 𝑅 𝐓(𝐤𝑠 , ±𝑘𝑧𝑖 , 𝐫) ⊗ 𝐓∗ (𝐤𝑠 , ±𝑘𝑧𝑗 , 𝐬). 𝜎2𝐷 = 𝑖𝜀0 𝜔𝑑(𝜀TMD (𝜔) − 1), where 𝑑 = 0.65 nm is the thickness of the
8𝜋 2 ∫ 𝑘𝑧𝑖 𝑘2𝑠 𝑇 𝑇
TMD monolayer.
(13) At this point we note that the external magnetic field applied to the
QD in the Voigt geometry is along the 𝑥-axis, so it is parallel to the
A summation is implied for each pair of ± indices. These indices
monolayer. This magnetic field does not modify the optical properties
show the direction of propagation of the electromagnetic modes, the
of the TMD layer. The reason is that in a strictly 2D material, as a
first index for the acceptor and the second for the donor. Also, the
monolayer TMD, the current circulation from the orbitals can only be
summation over 𝐓 is over the 𝐌 and 𝐍 modes which are connected with
within the plane; as a consequence, the corresponding orbital magnetic
the transverse electric and transverse magnetic modes, respectively.
moment points out-of-plane. Thus, it can only couple to magnetic fields
The form of 𝐌 and 𝐍 can be found in Ref. [35].
applied along the 𝑧-axis [40,41].
The boundary conditions imposed on two-dielectrics interface, Using the above relations we can calculate the Purcell factors for
where the TMD monolayer is placed, are the continuity condition and the system QD-TMD monolayer. Throughout this work we consider a
the radiation condition. The first condition is given by the continuity QD with an emission energy of ℏ𝜔0 = 1.31 eV. In Fig. 2 we display the
equations at the interface: Purcell factors 𝐹𝑥 , 𝐹𝑧 as functions of the distance 𝑅 between the QD
[ ]| and the TMD monolayer, taking that the dielectric environment is free-
𝑘̂ × G(11) (𝐫, 𝐬, 𝜔) − G(12) (𝐫, 𝐬, 𝜔) | = 0, (14)
|𝑧=0 space (𝜀1 = 1), for four different semiconductor materials in the layer,
[ ]| WS2 (blue dotted line), MoS2 (magenta dashed–dotted line), WSe2
𝑘̂ × 𝛁 × G(11) (𝐫, 𝐬, 𝜔) − 𝛁 × G(12) (𝐫, 𝐬, 𝜔) | 𝑧=0 (cyan dashed line), and MoSe2 (red solid line). In all cases there is an
|
4𝜋 enhancement of the spontaneous decay rates relevant to the decay rate
= −𝑖 𝑘0 𝜎 𝑘̂ × 𝑘̂ × G(12) (𝐫, 𝐬, 𝜔)|𝑧=0 . (15)
𝑐 in vacuum and the decay rates reduce monotonically as the distance

3
D. Stefanatos et al. Physica E: Low-dimensional Systems and Nanostructures 118 (2020) 113935

Fig. 2. Purcell factors corresponding to (a) 𝑥- and (b) 𝑧-orientations of the transition dipole moment as functions of the distance 𝑅 between the QD and the monolayer, for four
types of the semiconductor material: WS2 (blue dotted line), MoS2 (magenta dashed–dotted line), WSe2 (cyan dashed line), and MoSe2 (red solid line). The red stars highlight the
values for the four examples of Fig. 6.

between the QD and the TMD layer increases. Also, for every distance Table 2
Optimal Purcell factors corresponding to the four cases highlighted with
the decay rate for a dipole parallel to the layer (e.g. a 𝑥-oriented dipole)
a red star in Fig. 3, where the duration is set to 𝑇 = 2∕𝛾 = 1.097 ns:
is always lower than the decay rate for a dipole perpendicular to the WS2 at 𝑅 = 2.75 nm, MoS2 at 𝑅 = 4.75 nm, WSe2 at 𝑅 = 5.75 nm and
layer (e.g. a 𝑧-oriented dipole), so the necessary spontaneous emission MoSe2 at 𝑅 = 6.5 nm.
anisotropy needed for Purcell-enhanced deexcitation process towards Purcell factors WS2 MoS2 WSe2 MoSe2
the desired state is succeeded for a QD near a TMD layer. 𝐹𝑥 4.37955 4.60074 4.44001 4.43832
𝐹𝑧 7.68465 8.11449 7.82323 7.79622
4. Fast initialization with optical control

The natural initial state of the system is an equal incoherent mixture


and 𝑎 is given in Eq. (22).
of states |1⟩ and |2⟩, giving 𝜌11 (0) = 𝜌22 (0) = 1∕2, while all the
The goal is to obtain a large value of 𝜌22 in the shortest possible
other matrix elements (either for populations or coherence terms) are
time. This can be achieved if the values of the system parameters are
initially zero. Starting from this initial state, the goal is to apply the
chosen such that 𝜌22 (∞) and 𝑟 are both large. Obviously, these two
appropriate optical field in order to manage to quickly prepare the spin
requirements are competing. For example, 𝜌22 (∞) is a decreasing while
state |2⟩ with high fidelity. In the following we separately investigate
𝑟 is an increasing function of 𝛺. Additionally, 𝜌22 (∞) is a decreasing
how this can be efficiently achieved with constant control field, giving
function of 𝐹𝑧 while 𝑟 has a nonmonotonic dependence on this param-
constant Rabi frequency 𝛺, and with time-dependent control field,
eter. The optimal values of 𝐹𝑥 , 𝐹𝑧 and 𝛺 are obtained as a compromise
giving time-dependent Rabi frequency 𝛺(𝑡).
between these two competing requirements.
In Fig. 3 we show the contour plots of population 𝜌22 (𝑇 ) versus the
4.1. Constant Rabi frequency
distance 𝑅 from the monolayer and the control amplitude 𝛺, for the
four semiconductor materials and for duration 𝑇 = 2∕𝛾. These plots are
For a constant Rabi frequency 𝛺, in order to quantify the degree of
produced using the Purcell factors 𝐹𝑥 , 𝐹𝑧 displayed in Fig. 2. Observe
spin initialization as a function of the system parameters, we find first
that a large value 𝜌22 (𝑇 ) > 0.994 can be obtained for duration as short
the steady state population of level |2⟩. Equating to zero the left hand
as 𝑇 = 2∕𝛾 = 1.097 ns and for a moderate value of the constant field
sides of Eq. (2) and solving the corresponding system we obtain 𝛺 = 5.16𝛾. The optimal Purcell factors, corresponding to the cases
1 + 𝑎2 + 𝑏 2 highlighted with red stars, are listed in Table 2. Another interesting
𝜌22 (∞) = , (21)
4 + 2𝑎2 + 𝑏2 observation is that for each semiconductor material similar levels of
where 𝜌22 (𝑇 ) are obtained but at different distances from the monolayer.
𝐹 + 𝐹𝑧 𝛾 𝜔 + 𝜔43 Thus, by choosing a different material we can adjust the distance from
𝑎= 𝑥 , 𝑏 = 21 . (22) the layer where the QD should be placed in order to maximize spin
2 𝛺 𝛺
initialization for a desired duration. In Fig. 4 we display similar contour
We next obtain an estimate of the necessary time to reach this steady
plots for WSe2 only and two additional durations, 𝑇 = 1∕𝛾 = 0.548 ns
state. We follow Ref. [16] and neglect for a moment the off-resonant and 𝑇 = 3∕𝛾 = 1.645 ns. Observe that for larger durations (from left
transition |2⟩ → |3⟩. In this case level |3⟩ is decoupled from the rest and to right) the optimal value of 𝜌22 (𝑇 ) is obtained for smaller values of
we are left with a 𝛬-type system. If we use the relation 𝜌11 = 1−𝜌22 −𝜌44 𝛺 and larger values of 𝑅, the latter corresponding to smaller values of
to further eliminate 𝜌11 , we reduce the original system to the following 𝐹𝑥 + 𝐹𝑧 . This dependence is easily explained, since for larger durations
three equations the steady state value 𝜌22 (∞) is approached, which is a decreasing
𝜌̇ 22 = 𝛾42 𝜌44 , (23) function of 𝛺, 𝐹𝑥 + 𝐹𝑧 , as deduced from Eqs. (21), (22). A similar
behavior is also observed for the other materials (not shown here).
𝜌̇ 44 = −(𝛾41 + 𝛾42 )𝜌44 + 2𝛺𝜌𝐼41 , (24) For comparison with the case where no nanostructure is used, we
𝛾41 + 𝛾42 𝐼 plot in Fig. 5 the final population 𝜌22 (𝑇 ) achieved without the layer
𝜌̇ 𝐼41
= − 𝜌41 + 𝛺(1 − 𝜌22 − 2𝜌44 ). (25)
2 as a function of the control amplitude 𝛺, for the three durations
In order to obtain a rough estimate of the time required to build 𝜌22 , 𝑇 = 1∕𝛾, 2∕𝛾, 3∕𝛾 used in the previous figures. Observe that in all
we use the last two equations to eliminate adiabatically 𝜌44 , 𝜌𝐼41 , ending cases 𝜌22 (𝑇 ) is well below 0.9, thus the presence of the monolayer
up with the simple equation significantly accelerates the initialization procedure.
𝜌̇ 22 = 𝑟(1 − 𝜌22 ), (26)
4.2. Time-dependent Rabi frequency
where the rate is
𝐹𝑧 In this section we use numerical optimization to obtain time-
𝑟= 𝛾 (27) dependent controls 𝛺(𝑡) which are zero at the initial and final
2 + 𝑎2

4
D. Stefanatos et al. Physica E: Low-dimensional Systems and Nanostructures 118 (2020) 113935

Fig. 3. Contour plots of 𝜌22 (𝑇 ) for duration 𝑇 = 2∕𝛾 = 1.097 ns, versus the distance 𝑅 from the monolayer and the control field 𝛺, for the four semiconductor materials: (a) WS2 ,
(b) MoS2 , (c) WSe2 , and (d) MoSe2 .

Fig. 4. Contour plots of 𝜌22 (𝑇 ) for WSe2 , versus the distance 𝑅 from the monolayer and the control field 𝛺, for two additional durations, (a) 𝑇 = 1∕𝛾 = 0.548 ns and (b)
𝑇 = 3∕𝛾 = 1.645 ns.

allows to express the control as a polynomial function of time. Specif-


ically, we consider 𝛺 as an extra state variable, on which we impose
boundary conditions (28), while we place the polynomial control in its
derivative. The corresponding equation in normalized time 𝜏 = 𝛾𝑡 is
( )
𝑑 𝛺 ∑
4
= 𝑎𝑘 𝜏 𝑘 , (29)
𝑑𝜏 𝛾 𝑘=0

where we have used a fourth order polynomial for the derivative, so 𝛺


is of fifth order, and 𝑎𝑘 are the dimensionless unknown coefficients to
be determined by the optimization procedure. We additionally impose
the constraint
Fig. 5. Final value 𝜌22 (𝑇 ) obtained without the monolayer versus the control field 𝛺,
for three different durations 𝑇 . 0 ≤ 𝛺(𝑡)∕𝛾 ≤ 20, (30)

in order to restrict the field amplitude within experimentally feasible


values. Now we can state the optimal control problem. For system
times equations (2), which can easily be expressed using normalized time and
𝛺(0) = 𝛺(𝑇 ) = 0, (28) dimensionless variables, augmented by Eq. (29) for 𝛺, we would like
to find the coefficients 𝑎𝑘 which maximize the final value 𝜌22 (𝑇 ) for a
while maximize the final population of level |2⟩. This kind of controls specific duration 𝑇 , while satisfy the boundary condition (28) and the
are more appropriate for experimental implementation, since they
constraint (30). Having found 𝑎𝑘 we can integrate Eq. (29) and obtain
avoid the initial and final jumps of the square pulse.
the field
In order to find 𝛺(𝑡), we use the freely available optimal control
𝛺(𝑡) ∑ 𝑎𝑚−1
5
solver BOCOP [42], which can easily incorporate boundary conditions
= (𝛾𝑡)𝑚 (31)
like (28). We also exploit a feature provided by this program, which 𝛾 𝑚
𝑚=1

5
D. Stefanatos et al. Physica E: Low-dimensional Systems and Nanostructures 118 (2020) 113935

Table 3
Optimal coefficients 𝑎𝑘 , 𝑘 = 0, 1, … , 4 for the polynomial control (31) and
the example that we consider, WSe2 layer st 𝑅 = 5.75 nm. The duration
is set to 𝑇 = 2∕𝛾 = 1.097 ns.
Coefficients WSe2
𝑎0 103.6497
𝑎1 −479.0491
𝑎2 636.9199
𝑎3 −324.9254
𝑎4 55.0074

expressed in the original time 𝑡.


We consider an example corresponding to the case highlighted with
red star in Fig. 3(c), a WSe2 layer at 𝑅 = 5.75 nm. The duration is set
to 𝑇 = 2∕𝛾 = 1.097 ns. The optimal coefficients found using BOCOP are
listed in Table 3. In Fig. 6 we display the resulting control field 𝛺(𝑡) and
the corresponding evolution 𝜌22 (𝑡). The final value 𝜌22 (𝑇 ) = 0.9962 is
larger than the value 0.9945 obtained with constant control 𝛺 = 5.16𝛾
in Fig. 3(c), as expected from an optimization procedure. In Fig. 6(c)
we display contour plots of 𝜌22 (𝑇 ) with respect to a positioning error
|𝛿𝑅| ≤ 1.5 nm around the nominal value 𝑅 = 5.75 nm and an additive
constant error in the control 0 ≤ 𝛿𝛺∕𝛾 ≤ 1. Observe the robustness of
the final value with respect to the amplitude error, while large values
𝜌22 (𝑇 ) ≥ 0.994 can be obtained for a positioning error between 0.5–
1 nm. Similar results are also obtained when using the other materials
in the layer (not shown here).
At this point it is worth mentioning that control fields with poly-
nomial time dependence and coefficients optimized with respect to
an objective have been used in several recent works, for example
Refs. [43–45]. In these studies, some of the coefficient are preselected
in order to satisfy the boundary conditions at initial and final times. For
each value of the remaining free coefficients, the system under consid-
eration is simulated and the resulting objective value is recorded. The
optimal coefficients are those optimizing the objective. Usually, one or
two free coefficients are employed, since the number of simulations and
thus the computational time grows exponentially with the number of
free parameters. With the optimal control solver BOCOP we can exploit
a large number of coefficients in the optimization procedure, while
the optimal solution is obtained much faster than with the previously
described method. An analogous approach, which also does not require
the calculation of the final fidelity for each point in the parameter space Fig. 6. Example with optimized time-dependent control for WSe2 at 𝑅 = 5.75 nm and
before the optimization, can be found in the recent study [46]. We duration 𝑇 = 2∕𝛾 = 1.097 ns: (a) Optimal control 𝛺(𝑡). (b) Corresponding 𝜌22 (𝑡). The
final value 𝜌22 (𝑇 ) = 0.9962 is larger than the value 0.9945 obtained with constant
finally point out that instead of a polynomial we could have used a control in Fig. 3(c). (c) Sensitivity diagram of the final value 𝜌22 (𝑇 ) obtained with the
trigonometric expansion for the control, as in our recent work [47]. optimal field 𝛺(𝑡), versus a positioning error around the nominal value 𝑅 = 5.75 nm
and a constant control error 𝛿𝛺.

5. Summary

Declaration of competing interest


In this article we considered the problem of fast spin initialization
for a QD in the Voigt geometry placed near a TMD monolayer. We
The authors declare that they have no known competing finan-
calculated the spontaneous emission rates of the QD modified by the
cial interests or personal relationships that could have appeared to
TMD nanostructure, and showed that high levels of fidelity, signifi-
influence the work reported in this paper.
cantly larger than in the case of the QD in free-space, can be quickly
obtained due to the anisotropy of the enhanced spontaneous decay CRediT authorship contribution statement
rates of the QD near the TMD layer. We first used a continuous wave
electromagnetic field and found constant control amplitudes which Dionisis Stefanatos: Conceptualization, Methodology, Software,
achieve acceptable fidelity levels in short times for different distances Writing - original draft. Vasilios Karanikolas: Investigation, Methodol-
of the QD from the TMD monolayer. We also studied the effect of the ogy, Software, Writing - original draft. Nikos Iliopoulos: Visualization,
layer material on the fidelity of the initialization. Next, we used state Writing - review & editing. Emmanuel Paspalakis: Conceptualization,
of the art numerical optimal control to find the time-dependent electric Supervision, Writing - original draft, Writing - review & editing.
field which maximized the final fidelity for the same short duration and
various TMD materials in the layer as in the previous case. A better Acknowledgments
fidelity was obtained with this method, while the resulting pulses were
quite robust to positioning error of the QD and to additive constant The research is implemented through the Operational Program ‘‘Hu-
control error. man Resources Development, Education and Lifelong Learning’’ and is

6
D. Stefanatos et al. Physica E: Low-dimensional Systems and Nanostructures 118 (2020) 113935

co-financed by the European Union (European Social Fund) and Greek [23] D.P. DiVincenzo, Fortschr. Phys. 48 (2000) 771.
national funds (project E𝛥BM34, code MIS 5005825). [24] S. Manzeli, D. Ovchinnikov, D. Pasquier, O.V. Yazyev, A. Kis, Nat. Rev. Mater.
2 (2017) 17033.
[25] Q.H. Wang, K. Kalantar-Zadeh, A. Kis, J.N. Coleman, M.S. Strano, Nature
References Nanotechnol. 7 (2012) 699.
[26] K.F. Mak, J. Shan, Nat. Photonics 10 (2016) 216.
[1] P. Michler (Ed.), Quantum Dots for Quantum Information Technologies, Springer, [27] Y.N. Gartstein, X. Li, C.-W. Zhang, Phys. Rev. B 92 (2015) 075445.
2017. [28] D.N. Basov, M.M. Fogler, F.J.G. de Abajo, Science 354 (2016) aag1992.
[2] R.J. Warburton, Nature Mater. 12 (2013) 483. [29] V.D. Karanikolas, C.A. Marocico, P.R. Eastham, A.L. Bradley, Phys. Rev. B 94
[3] W.B. Gao, A. Imamoglu, H. Bernien, R. Hanson, Nat. Photonics 9 (2015) 363. (2016) 195418.
[4] X. Xu, Y. Wu, B. Sun, Q. Huang, J. Cheng, D.G. Steel, A.S. Bracker, D. Gammon, [30] V.D. Karanikolas, E. Paspalakis, Phys. Rev. B 96 (2017) 041404(R).
C. Emary, L.J. Sham, Phys. Rev. Lett. 99 (2007) 097401. [31] F. Prins, A.J. Goodman, W.A. Tisdale, Nano Lett. 14 (2014) 6087.
[5] D. Press, T.D. Ladd, B. Zhang, Y. Yamamoto, Nature 456 (2008) 218. [32] A. Raja, A. Montoya-Castillo, J. Zultak, X.-X. Zhang, Z. Ye, C. Roquelet, D.A.
[6] A. Delteil, Z. Sun, W.-B. Gao, E. Togan, S. Faelt, A. Imamoglu, Nat. Phys. 12 Chenet, A.M. van der Zande, P. Huang, S. Jockusch, J. Hone, D.R. Reichman,
(2016) 218. L.E. Brus, T.F. Heinz, Nano Lett. 16 (2016) 2328.
[7] K.G. Lagoudakis, P.L. McMahon, K.A. Fischer, S. Puri, K. Müller, D. Dalacu, P.J. [33] K.G. Schädler, C. Ciancico, S. Pazzagli, P. Lombardi, A. Bachtold, C. Toninelli,
Poole, M.E. Reimer, V. Zwiller, Y. Yamamoto, J. Vuckovic, New J. Phys. 18 A. Reserbat-Plantey, F.H.L. Koppens, Nano Lett. 19 (2019) 3789.
(2016) 053024. [34] H. Liu, T. Wang, C. Wang, D. Liu, J.-B. Luo, J. Chem. Phys. C 123 (2019) 10087.
[8] S. Sun, H. Kim, Z. Luo, G.S. Solomon, E. Waks, Science 361 (2018) 57. [35] W.C. Chew, Waves and Fields in Inhomogeneous Media, IEEE Press, New York,
[9] S.E. Economou, L.J. Sham, Y. Wu, D.G. Steel, Phys. Rev. B 74 (2006) 205415. NY, USA, 1995.
[10] C. Emary, X. Xu, D.G. Steel, S. Saikin, L.J. Sham, Phys. Rev. Lett. 98 (2007) [36] H.T. Dung, L. Knöll, D.-G. Welsch, Phys. Rev. A 62 (2000) 053804.
047401. [37] V.D. Karanikolas, C.A. Marocico, A.L. Bradley, Phys. Rev. B 91 (2015) 125422.
[11] S.E. Economou, T.L. Reinecke, Phys. Rev. Lett. 99 (2007) 217401. [38] G.W. Hanson, J. Appl. Phys. 103 (2008) 064302.
[12] C. Emary, L.J. Sham, J. Phys. Condens. Matter 19 (2007) 056203. [39] G.T. Papadakis, A. Davoyan, P. Yeh, H.A. Atwater, Phys. Rev. Mater. 3 (2019)
[13] H. Sun, X.-L. Feng, C. Wu, J.-M. Liu, S.-Q. Gong, C.H. Oh, Phys. Rev. B 80 (2009) 015202.
235404. [40] A. Srivastava, M. Sidler, A.V. Allain, D.S. Lembke, A. Kis, A. Imamoǧlu, Nat.
[14] V. Loo, L. Lanco, O. Krebs, P. Senellart, P. Voisin, Phys. Rev. B 83 (2011) 033301. Phys. 11 (2015) 141.
[15] A. Majumdar, P. Kaer, M. Bajcsy, E.D. Kim, K.G. Lagoudakis, A. Rundquist, J. [41] A.V. Stier, K.M. McCreary, B.T. Jonker, J. Kono, S.A. Crooker, Nature Commun.
Vuckovic, Phys. Rev. Lett. 111 (2013) 027402. 7 (2016) 10643.
[16] M.A. Antón, F. Carreño, S. Melle, O.G. Calderón, E. Cabrera-Granado, M.R. Singh, [42] F. Bonnans, P. Martinon, D. Giorgi, V. Grélard, S. Maindrault, O. Tissot, BOCOP:
Phys. Rev. B 87 (2013) 195303. An Open Source Toolbox for Optimal Control, INRIA-Saclay, Ile de France, 2017.
[17] F. Carreño, Francisco Arrieta-Yáñez, M.A. Antón, Opt. Commun. 343 (2015) 97. [43] M. Palmero, S. Wang, D. Guéry-Odelin, J.-S. Li, J.G. Muga, New J. Phys. 18
[18] P. Kumar, T. Nakajima, Phys. Rev. A 91 (2015) 023832. (2016) 043014.
[19] Y. Peng, Z.-Y. Yun, Y. Liu, T.-S. Wu, W. Zhang, H. Ye, Physica B 470–471 (2015) [44] Q. Zhang, X. Chen, D. Guéry-Odelin, Sci. Rep. 7 (2017) 15814.
1. [45] Y. Yan, Y.C. Li, A. Kinos, A. Walther, C. Shi, L. Rippe, J. Moser, S. Kröll, X.
[20] P. Kumar, T. Nakajima, Opt. Commun. 370 (2016) 103. Chen, Opt. Express 27 (2019) 8267.
[21] E. Paspalakis, S.E. Economou, F. Carreño, J. Appl. Phys. 125 (2019) 024305. [46] P.V. Pyshkin, E.Y. Sherman, J.Q. You, L.-A. Wu, New J. Phys. 20 (2018) 105006.
[22] D. Stefanatos, V. Karanikolas, N. Iliopoulos, E. Paspalakis, Physica E 117 (2020) [47] D. Stefanatos, E. Paspalakis, Phys. Rev. A 99 (2019) 022327.
113810.

You might also like