PHYSICAL REVIEW A 95, 012325 (2017)

Quantum efficiency of a single microwave photon detector based

on a semiconductor double quantum dot

Clement H. Wong
Laboratory for Physical Sciences, University of Maryland, College Park, Maryland 20740, USA

Maxim G. Vavilov
Department of Physics, University of Wisconsin, Madison, Wisconsin 53706, USA
(Received 22 December 2015; published 24 January 2017)

Motivated by recent interest in implementing circuit quantum electrodynamics with semiconducting quantum
dots, we consider a double quantum dot (DQD) capacitively coupled to a superconducting resonator that is driven
by the microwave field of a superconducting transmission line. We analyze the DQD current response using
input-output theory and show that the resonator-coupled DQD is a sensitive microwave single photon detector.
Using currently available experimental parameters of DQD-resonator coupling and dissipation, including the
effects of 1/f charge noise and phonon noise, we determine the parameter regime for which incident photons are
completely absorbed and near-unit 98% efficiency can be achieved. We show that this regime can be reached
by using very high quality resonators with quality factor Q  105 .

DOI: 10.1103/PhysRevA.95.012325

I. INTRODUCTION tunneling relative to the DQD inelastic decay rate, and does
not require strong DQD-resonator coupling relative to the
High-performance, single-photon detectors are essential
DQD dissipation rates. The photon detector utilizes only
tools in quantum optics, with applications in optical quantum
charge states of the DQD and does not require magnetic fields
information processing, communication, cryptography, and
that would degrade the quality factor of the superconducting
metrology [1]. Single-photon detection in the microwave
resonator cavities.
regime has similar applications in the emerging field of mi-
The zero-reflection regime is also relevant in the context
crowave quantum photonics, made possible by recent advances
of quantum computation, since it enables distant transmission
in implementing circuit quantum electrodynamics (cQED)
of quantum information required in quantum cryptography
with superconducting circuit technology [2,3], but are more
and communication [30,31]. We also note that when the input
difficult to achieve because microwave photons have energy
photons come from a hot thermal source, this device acts as a
five orders of magnitude less than optical photons. Besides
quantum heat engine [32,33].
photonics, microwave photon detectors have applications in
This paper is organized as follows. In Sec. II, we introduce
astronomy and cosmology, for example, in measuring the
our theoretical model for the transmission line carrying incom-
cosmic microwave background [4]. Microwave radiometers
ing photons and the photon detector. In Sec. III, we present
are commonly used in meteorological and oceanographic
the equations of motion governing the system dynamics, and
remote sensing.
in Sec. IV, we present the steady-state solution, which is used
Recent theoretical proposals and experimental develop-
to derive an analytic expression for the quantum efficiency of
ments in microwave photon detectors include those based
photon detection in Sec. V. In Sec. VI, we discuss how the
on Josephson junctions [5–10] and opto-electromechanical
reflected signal is incorporated in our model, and in Sec. VII,
systems [11–13]. At the same time, experimental progress
we find the optimal parameter regime. A brief summary of
in implementing cQED with semiconducting quantum dots
the decoherence model we used for the DQD is given in the
is showing promise [14–24]. Currently available resonator–
Appendix.
quantum dot systems already allow for some interesting
quantum optics applications such as on-chip single-emitter
masers [25,26] and tunable self-interaction and dissipation of II. MODEL
the resonator photons induced by the quantum dot [27,28].
When a quantum dot is connected to electric leads, this We consider a microwave photon detector consisting of
system provides a platform for studying the interplay between a superconducting microwave resonator coupled to a DQD,
quantum impurity physics and quantum optics [29]. which receives photons from a semi-infinite microwave
In this paper, we propose a photon detector based on transmission line, as shown in Fig. 1. The DQD is operated
photon-assisted tunneling of electrons through a double in the “pumping” configuration, with zero source-drain bias
quantum dot (DQD) and determine the quantum efficiency and near the charge transition between the charge states
of single-photon detection. We identify the parameter regime |L = |N + 1,M and |R = |N,M + 1, where |N,M de-
where reflection of input photons from the resonator vanishes, notes N (M) electrons in the left (right) dot. The electron
so that near-unit efficiency can be achieved with currently Hamiltonian governing behavior of the dot with M + N + 1
available experimental parameters. Such a high efficiency electrons is
is possible even in the presence of strong DQD dissipation τz + 2tc τx ωσ σz
because the detection process takes advantage of fast dot-lead HDQD =  = , (1)
2 2

2469-9926/2017/95(1)/012325(10) 012325-1 ©2017 American Physical Society

CLEMENT H. WONG AND MAXIM G. VAVILOV PHYSICAL REVIEW A 95, 012325 (2017)

transmission line through the DQD can be written as

bin Cκ Detector
Ṅ ωin ωa
I  = eR {pe |R|e|2 − p0 |R|g|2 }, (4)
bout
Cg
where pα = |αα| is the projection operator to state |α, . . . 
|R
g 2 /Γ2 |L
denotes the quantum mean value and time average.
|g |n + 1 |e |n tc In the absence of photon excitation of the dot and at
ΓL ΓR temperature T = 0, the DQD is in the ground state, pg  = 1
ωin S |e D
and pe  = p0  = 0, and no current flows. However, photons
κ ωa
in the microwave resonator cause transitions between the
Γ1 ωσ ground and excited states of the DQD, which enables sequen-
|g |n tial tunneling (N,M) → (N + 1,M) → (N,M + 1), resulting
|g
in a finite current. Photon arrivals can thus be detected
by measuring the DQD source-drain current. We note that,
FIG. 1. Top left: A transmission line carries a continuous wave of because the DQD is operated in the limit of relatively strong
microwave photons with flux Ṅ and frequency ωin . Right dashed box: coupling to the leads, spin is not conserved [35] and the spin
The photon detector is a single-port superconducting resonator with blockade does not occur in the proposed device. At finite
frequency ωa , capacitively coupled to a double quantum dot (DQD) temperature, the tunneling rates Eq. (3a) and Eq. (3b) should be
with a coupling g. The DQD is set to zero source-drain bias, with modified with appropriate Fermi functions for the leads. When
the lead chemical potentials set in between the ground and excited thermal broadening in the leads becomes comparable to the
state. Absorption of a resonator photon causes transitions between photon energy, there can be a back current due to electrons that
the DQD states |g and |e, resulting in current flow. Lower left tunnel from the lead to the excited state, emit a photon followed
box: Energy levels (horizontal lines) and transition processes (single by the transition to the ground state, and then tunneling from
directed arrows) in the resonator-DQD system. A single photon the ground state to the lead. However, this effect is negligible
received by the resonator (|g|n + 1) can be emitted at the rate κ at an operating temperature of 40 mK = 0.8 GHz h/kB , an
or be absorbed by the DQD at the rate g 2 / 2 , exciting the metastable order of magnitude lower than the photon energy we consider.
state |e|n and then decaying back to the system ground state |g|n The Hamiltonian of noninteracting photons and the DQD
at the inelastic decay rate 1 . in the rotating frame at the frequency of input photons ωin are
 ∞  
where τx,z are the Pauli matrices in the charge basis, |L,R, dω 1
while σz = |ee| − |gg| is the diagonal Pauli matrix in the Hfree = ωbω† bω + ab a † a + , (5)
−∞ 2π 2
basis of eigenstates of HDQD : σ b
     H̃DQD = σz , (6)
|g cos θ2 − sin θ2 |L 2
= . (2)
|e θ
sin 2 θ
cos 2 |R where ab = ωa − ωin (σ b = ωσ − ωin ) is the detuning of
the resonator frequency (DQD) ωa (ωσ ) from ωin . Here, we
We introduced the following notations:  is the DQD voltage used the mode expansions
bias across the dots, counted from the charge degeneracy
 point,
tc is the interdot tunnel coupling, and ωσ = 4tc2 +  2 is the 
∞
DQD excitation energy and cos θ = −/ωσ . dω ωZTL
VTL = (bω + bω† ), (7)
We consider zero bias across the leads, when electrons in the −∞ 2π 2

excited state can tunnel out incoherently to the leads, resulting

ωa
in the “empty” state |0 = |N,M, and then, the ground state Va = (a+a † ), (8)
|g can be loaded by electron tunneling from the leads. The 2Ca
Hamiltonian describing tunneling to the leads is given by
  for the resonator (Va ) and transmission line (VTL ) voltage
† † †
H = (k cik cik + ti cik di + ti∗ cik di ), operator at the coupling capacitor Cκ , where a is the resonator
i=L,R k photon annihilation operator and bω is the transmission
† †
line photon annihilation√ operator with frequency ω [36].
where di= |i0|, = |ki|, k is the dispersion for an
cik Furthermore, ZTL = l/c is the characteristic impedance of
electron state |k with wave vector k in the leads, and |i is a the line, l is the inductance per unit length, c is the capacitance
state in the L/R dot. The resulting incoherent tunneling rates to ground per unit length, and Ca is resonator capacitance.
from |e to |0 (0e ) and from |0 to |g (g0 ) are given by
√ the transmission line dispersion ωk = vk, where
We assume
[19,20,34] v = 1/ lc is the group velocity.
The resonator-DQD coupling is described by the Jaynes-
0e = L cos2 (θ/2) + R sin2 (θ/2), (3a) Cummings Hamiltonian:
g0 = L sin (θ/2) + R cos (θ/2),
2 2
(3b)
HJC = g(σ † a + a † σ ), (9)
where R(L) are tunneling rates to the right (left) leads.
Taking into account opposite currents due to loading from where σ = |ge| is the DQD lowering operator. Here, the
and tunneling to the leads, the time-averaged electron current interaction strength g = g0 sin θ is defined by the dipole matrix

012325-2
QUANTUM EFFICIENCY OF A SINGLE MICROWAVE . . . PHYSICAL REVIEW A 95, 012325 (2017)

element in the basis of energy eigenstates, and input field in the rotating frame is defined by the initial field

configuration as
π Za  ∞
g0 = ωa βg , dω
RQ bin (t) = bω (ti )e−iω(t−ti ) . (13)
−∞ 2π
where Za is the resonator characteristic impedance, RQ = We will consider the case where the input field is formed by
h/2e2 = 12.9 k is the resistance quantum, βg = Cg /(Cg + a continuous flux of photons in a narrow spectral band around
C ), Cg is the gate capacitance between the resonator and the ωin , corresponding to the initial state [41]
DQD, and C is the total capacitance of the DQD [24].  nb
1 †
The photon mode in the resonator is driven by photons |βi  = √ dωf (ω)bω (ti ) |0TL ,
exchange with the transmission line. Assuming a weak, local nb !
coupling capacitance Cκ between transmission line photons where |f (ω)|2 = δ(ω − ωin ) and |0TL is the vacuum state of
and the resonator, the interaction Hamiltonian in the rotating the transmission line. The flux of incoming photons is given
wave approximation is given by †
by the expectation value in |βi , Ṅ = bin (t)bin (t). In terms
 of experimental parameters, Ṅ = vnb /L = nb /ZTL cL, where
√ dω †
HTL = Cκ Va VT L =  κ [b a + a † bω ], (10a) nb is total input photon number of the waveguide and the
2π ω
transmission line impedance is ZTL = 50 .
where Using the solution Eq. (12) specified by bin leads to
ZTL Za the Heisenberg-Langevin equations for the system operators
κ = Cκ2 ωin
3
(10b) {X} = {a,σ,σz ,p0 } given by [37,39,40]
4
√ i
is the photon leakage rate from the resonator, Za = La /Ca Ẋ = [Hsys ,X] + (Dγ + D )X
is the impedance of the coplanar waveguide resonator, and 
√ † √
La is the resonator inductance. We notice that the resonator + (κ/2−i κbin )[X,a] + [a † ,X](κ/2 + i κbin ), (14)
quality factor can be expressed as
where D(A) is the Lindblad superoperator defined by
ωin 4ZTL D(A)ρ = AρA† − (A† Aρ + ρA† A)/2. Here, D is the dis-
Q= = , sipative operator describing incoherent tunneling to the leads,
κ Za (ωin τRC )2
given by
where τRC = Cκ ZTL is the recharging time of the coupling
capacitor Cκ . D = 0e D(|0e|) + g0 D(|g0|). (15)
To finalize the description of the model, we present the full Other incoherent processes in the dynamics of the DQD
Hamiltonian of the system: originate from interaction of charge degrees of freedom of
the DQD with its environment, such as the phonon system
H = Hsys + HTL + H + Hγ , (11a) and charge noise in dielectric. The DQD charge coupling to
phonons is similar to the coupling to the resonator field; see
Hsys = H̃DQD + HJ C + Hfree . (11b)
Appendix A 1 and Refs. [19,42,43]. The charge noise results
In Eq. (11a), Hγ is the Hamiltonian describing the DQD in an additional component δ(t) of the bias between the
dissipative environment, which consists of voltage fluctuations dots,  →  + δ(t). This noise typically exhibits a 1/f power
and phonons. spectrum; see Appendix A 2. The Lindblad superoperator Dγ
accounts for these incoherent processes
III. EQUATIONS OF MOTION Dγ = γ↑ ()D(σ † ) + γ↓ ()D(σ ). (16)
We will employ input-output theory [37–40] to model the Here, γ↑ () and γ↓ () are the bias-dependent excitation and
resonator–transmission line interaction. This formalism will relaxation rates, respectively. We describe contributions to
enable us to optimize the quantum efficiency of photon de- these rates from electron-phonon interaction and the high-
tection including interference effects between the microwave frequency component of the charge noise in Appendixes A 1
signals reflected by the resonator-DQD system, which is and A 2.
not captured by density matrix master equations. The key From Eq. (14) and Eq. (15), the equations of motion for the
assumptions in this formalism are the rotating wave and photon and DQD operators are [37]
Markov approximation.  
The equation of motion for the transmission line modes that κ √
ȧ = − + iab a − igσ +i κbin , (17a)
follows from the Hamiltonians in Eq. (10a) and Eq. (5) can be 2
solved analytically to yield σ̇ = −(2 + iσ b )σ +igaσz , (17b)
 t
−iω(t−t0 )
√ where
bω (t) = e bω (t0 ) + i κ dt e−iω(t−t ) a(t ), (12)
t0 γ1 + 0e
2 = (18)
which is a solution that can be specified by initial t0 = ti < t 2
or final condition t0 = tf > t, at time ti (tf ) long before (after) is the transverse relaxation rate, which includes incoherent
the transmission line photons interact with the resonator. The dynamics of the DQD between states |e and |g with the

012325-3
CLEMENT H. WONG AND MAXIM G. VAVILOV PHYSICAL REVIEW A 95, 012325 (2017)

combined rate γ1 = γ↑ + γ↓ and lifetime broadening due to Then, substituting Eq. (20) into Eq. (19a) yields effective
incoherent tunneling to the leads at the rate 0e , Eq. (3a). equations of motion for the polarization
The low-frequency component of charge 1/f noise causes
non-Markovian dynamics, which strictly speaking should be ṁz = 1 (m0 − mz ) + 2iga † σ − σ † a, (21a)
described by a nonlocal t damping kernel in Eq. (17b) of 20e g0 − γs 0e
the form σ̇ = −[2 + 0 dt (t − t )]σ (t ) + · · · [44], where 1 = γ1 + , (21b)
0e + 2g0
(t) = γφ ωl e−ωl t , ωl being the noise correlation time, and
γφ can be estimated with the pure dephasing rate given in (γs + 0e )g0
m0 = − , (21c)
Eq. (A7). However, we will only need the steady-state solution, (γ1 + 0e )g0 + 0e (γ1 − γs )/2
which is determined by the zero mode in the Laplace transform
where 1 is the effective depolarization rate and γs = γ↓ − γ↑
of the equations of motion. This zero-mode equation is
is the phonon-induced spontaneous emission rate; cf. Eq. (16).
∞Eq. (17b) with the total damping rate 2 + γφ ,
consistent with
At this point, the three-level DQD system consisting
where γφ = 0 dt(t) is the zero-mode component of the
of {|e,|g,|0} is reduced to a two-level system described
damping kernel. Thus, to take into account low-frequency
by Eqs. (17b) and (21a) with longitudinal and transverse
1/f -type noise in our proposed device, it is sufficient to
relaxation rates 1 and 2 , respectively. The equilibrium
take 2 → 2 + γφ , where γφ is given by Eq. (A7) of
polarization m0 gives the value of mz in the absence of coupling
Appendix A 2.
to resonator photons, g = 0. The effective DQD excitation,
We note that it has been established both experimentally and
relaxation, and dephasing rates are ↑ = 1 (1 + m0 )/2, ↓ =
theoretically that noise in the DQD bias  is much stronger than
1 (1 − m0 )/2, and φ = 2 − 1 /2, respectively.
tunnel-coupling noise, and in most cases, DQD decoherence
The steady-state photon and DQD operators satisfy
can be modeled by only  noise [45,46]. Furthermore, our
device operates in the regime   tc , where the DQD frequency √
a = χa ( κbin − gσ ), (22a)
depends very weakly on tc , so that any dephasing due to noise
in tc is strongly suppressed. Finally, we note that since we σ = gχσ (aσz ), (22b)
are not detecting the quantum state of the photon, details of where χa = (ab − iκ/2)−1 and χσ = (σ b − i2 )−1 are the
the DQD coherence are not of central importance. As we will resonator and DQD susceptibilities [40]. To solve these
show below in Eq. (36), the decoherence rate 2 can change equations, we apply the mean-field approximation to Eq. (22b)
the optimal parameter regime but does not necessarily limit by taking aσz → amz , and substitute Eq. (22b) into Eq. (22a)
performance, since the resonator leakage rate κ can be adjusted to find the resonator photon field
to maintain the optimal condition.
The equations of motion the DQD polarization and “empty” a = χab bin , (23a)
state projection operators are √
κ
χab = . (23b)
† †
ab − iκ/2 + g 2 mz χσ
σ̇z = 2ig(a σ − σ a)
    This solution, together with the steady-state solution of
0e 0e Eq. (21a), yields the mean-field equation
+ γ− − pσ − γ 1 + σz − p0 g0 , (19a)
2 2 m0
pσ + σz mz = , (24a)
ṗ0 = 0e − g0 p0 , (19b) 1 + Gna 
2 †
na  = |χab |2 bin bin , (24b)
where pσ = |ee| + |gg| is the projection operator into the †
where G = (4g / 1 )Imχσ and na = a a is the resonator
2
charge qubit subspace, determined by the constraint p0 + photon number operator.
pσ = |ee| + |gg| + |00| = 1, which represents conser- Due to the dependence on mz in Eq. (23b), the mean-field
vation of probability in the DQD-state space. equation (24a) is a cubic equation for mz [39]. However, a
simple estimate based on the perturbative parameter G 
g 2 / 1 2  10−4 will show that it is sufficient to take mz = m0
IV. STEADY-STATE SOLUTION in Eq. (23b), so that Eq. (24a) yields an explicit solution for
We will compute the detection efficiency for the case of mz . The photon-induced polarization is
continuous flux of photons using the steady-state solution to Gna 
the equations of motion [1]. Here, we present the steady-state mz = mz − m0 = −m0 , (25)
1 + Gna 
solution for the polarization mz ≡ σz , for a purely quantum
input field with zero classical component bin  = 0, and then which to leading order yields mz ≈ |m0 |Gna   GṄ /κ,
relate mz to the detection efficiency of the device [47]. where we kept na   Ṅ /κ to O(1). Substituting this in
We eliminate the “empty” dot operator p0 using the steady- Eq. (23b) yields the leading order correction to the photon
state solution of Eq. (19b), which yields number na   CG(Ṅ /κ)2 , where
C = g 2 / 2 κ (26)
2g0 − 0e σz
pσ = . (20) is the cooperativity. As shown in Sec. V, C = O(1), and,
0e + 2g0 since we will consider input flux Ṅ and leakage rate κ

012325-4
QUANTUM EFFICIENCY OF A SINGLE MICROWAVE . . . PHYSICAL REVIEW A 95, 012325 (2017)

(a) polarization as
 
Freq. (GHz)
eR (1 + mz ) 2 θ 2 θ
I  = − 0e cos − g0 sin . (28)
0e + 2g0 2 2
As one would expect, the current is proportional the probability
for the DQD to be excited, pe  = (1 + mz )/2. The last factor
in Eq. (28) takes into account the cancellation of the left and
right moving electrons.
(b) We will henceforth consider symmetric dot-lead tunneling
g/2π (MHz)

Γ1,2 (GHz)
at the rate l = L = R . Then 0e = g0 = l , the photon
induced current is
l cos θ
I  = −e mz , (29)
3
and Eqs. (21c) and (21b) become
2l − γs γ s + l
Cooperativity, C

(c) 1 = γ1 + , m0 = − . (30)
3 3γ1 /2 − γs /2 + l
We will be interested in the linear regime with respect to Ṅ ,
where single-photon detection occurs. This regime coincides
with the leading order expansion of Gna  in Eq. (25), where
the efficiency is given by
4 κg 2 (l / 1 )|cos θ Imχσ |
(d) η= . (31)
3 (ab − g 2 |m0 |Reχσ )2 + (κ/2 + g 2 |m0 |Imχσ )2
Δmz

mz

Next, we analyze quantities that characterize the DQD-

resonator coupling, dissipation, and response as a function of
bias; see Figs. 2(a)–2(c). As a point of reference, we show in
Fig. 2(a) the Jaynes-Cummings energy levels in the zero and
one photon subspace:
 2
DQD bias, /2π (GHz) E1,± = ωa ± 4g + 2aσ ,
2
FIG. 2. As a function of the double-dot bias , for
where g = g0 sin θ and we take the parameters
(tc ,g)/2π = (2,0.05) GHz and Ṅ = 1 MHz: (a) Jaynes-Cummings ωa /2π = 7 GHz and g0 /2π = 50 MHz, which is a value
energy levels E1± of the DQD-resonator system. (b) DQD-resonator that has been reported in experiments [16,22,48]. In Fig. 2,
coupling g = g0 sin θ for g0 /2π = 50 MHz, transverse relaxation we choose a tunnel coupling of tc /2π = 2 GHz, but we will
rate 2 and depolarization rate 1 of the effective two-level DQD. optimize this parameter below. Note that since 2  g, the
(c) Cooperativity of the DQD-resonator system C = g 2 / 2 κ; the resonances at the Jaynes-Cummings energies are destroyed;
real and imaginary parts of the DQD susceptibility χσ . (d) DQD the subspace of the resonator-DQD system with n photons is
polarization mz and photon-induced polarization mz . specified by the uncoupled basis {|g|n,|e|n − 1}, where
|e denotes the DQD excited state that is strongly broadened
with a linewidth 2 , as shown in Fig. 1 (lower left box).
in the MHz range, na  = O(G), which results in the We use DQD relaxation parameters appropriate for
subleading term of O(G2 ) in mz . The O(G) estimate for silicon DQD [43,49]: phonon noise spectral density
the induced polarization will be verified numerically below Jph /2π = 0.1 GHz [cf. Eq. (A4)], quasistatic bias noise
[see Fig. 2(d)]. The linearization of Eq. (22b) is justified by variance rms /2π = 1.2 GHz, and take dot-lead tunnel rate
the same perturbative expansion. l /2π = 1 GHz. The DQD-resonator coupling g [Fig. 2(b)]
has a broad peak centered at  = 0, due to a strong dipole
V. QUANTUM EFFICIENCY moment at the charge degeneracy point, while the transverse
relaxation rate 2 has a minimum due to a sweet spot
The detector efficiency is defined by the ratio of the steady- where dephasing is to the first order insensitive to quasistatic
state mean DQD current per input photon flux, bias noise. These effects combined lead to a broad peak
|I | around  = 0 in the cooperativity C = g 2 / 2 κ, shown in
η= , (27) Fig. 2(c), indicating strong DQD-resonator interaction. The
eṄ
strong charge dipole also increases transitions driven by
where e is the electron charge, I  = I − I0  is the mean phonons and 1/f charge noise, leading to a peak in the
polarization and current caused by absorbing photons, and I0  equilibrium polarization of m0  −0.95 at  = 0 [Fig. 2(d)]
are the dark counts due to the current at zero photon input flux. and a peak in the depolarization rate 1 [Fig. 2(b)], which
The electron current Eq. (4) can be expressed in terms of the is otherwise dominated by dot-lead tunneling. The real and

012325-5
CLEMENT H. WONG AND MAXIM G. VAVILOV PHYSICAL REVIEW A 95, 012325 (2017)

1.
(a) (b)

reflection, |r|
res
(a)

∗
I , (pA)

0.8

phase φr
Maximum efficiency,
1
10
0.6
Current,

0
10 0.4

0.2
10−1
100 101 102 103 104 0 1 2 3 4 5
(b)
Input flux, N (MHz) Dot−lead tunnel rate, l (GHz)

Pa (dBm)
na
FIG. 3. For optimal parameters κ/2π = 76 kHz and
tc /2π = 0.5 GHz, on resonance ab = 0: (a) Logarithmic
plot of the photon-induced current I  as a function of input flux
Ṅ for l = 1 GHz. (b) Maximum efficiency as a function of the
incoherent dot-lead tunnel rate l with Ṅ = 1 MHz.

imaginary parts of χσ , which modify the effective resonator DQD bias, /2π (GHz)
frequency and decay rate, respectively, are plotted in Fig. 2(c).
We conclude this section by considering the detector FIG. 4. As a function of the double-dot bias , for
response as a function of input flux Ṅ and dot-lead tunnel (tc ,g)/2π = (2,0.05) GHz and Ṅ = 1 MHz: (a) Magnitude |r| and
rate l . Nonlinear response at large flux will cause I to phase φr of the reflection coefficient. (b) Photon number na and power
saturate and sets the detector dead time [1]. At the same time, a absorbed Pa .
sufficiently large flux is necessary for the current I = ηeṄ to
be measurable. Figure 3(a) shows the photon-induced current
this case, we can introduce the output field as a counterpart of
I = 0.16ηṄ pA/MHz the input field, Eq. (13):
 ∞
as a function of Ṅ at l = 1 GHz using Eq. (37) for the dω
bout (t) = bω (tf )e−iω(t−tf ) . (32)
∗
optimal efficiency ηres found below. The response is linear −∞ 2π
up to Ṅ = 100 MHz, and saturates as the flux approaches the The equation of motion for the resonator field has a structure
∗
effective inelastic decay rate Ṅ  1 . Figure 3(b) shows ηres as similar to that of Eq. (17a), but an opposite sign in front of the
a function of the lead tunnel rate l at fixed Ṅ = 1 MHz. The term describing spontaneous emission:
optimal rate occurs near l = 1 GHz, and is determined by the  
competition between two effects: when l is too low, the elec- κ √
ȧ = − iab a − igσ +i κbout . (33)
tron relaxes back to |g, but when l is too high, the efficiency 2
suffers due to level broadening, 2 ∝ l ; see Eq. (18).
By subtracting the above equation from Eq. (17a), we obtain
Maximum photon absorption by the DQD occurs on
the relation between input and output modes [40],
resonance, as shown by the peaks at res in the photon-
√
induced DQD polarization mz plotted in Fig. 2(d), which bout (t) = bin (t) + i κa(t), (34)
are accompanied by minima in photon number with na  1,
indicating perfect photon to electron conversion. Note that where on the right-hand side, the first term is the reflection
induced polarization is very small, mz (res )  10−4  m0 , of the input field and the second is the field radiated by the
which justifies our approximation of taking mz = m0z in resonator. The reflection coefficient r is defined by bout = rbin .
Eq. (23b), and agrees with our previous estimate below Eq. (26) A system of a coupled DQD and microwave resonator can
that mz = O(G). be also used to control the output microwave field [27,50].
Figure 6(a) shows the photon detector efficiency as a Here, we briefly analyze the suppression of the reflected
function of bias for κ = 1 MHz, computed using Eqs. (25), signal from the resonator when the DQD device acts as an
(27), and (29). When the charge transition is sharp, at adjustable dissipating element. We consider the reflecting
tc /2π = 0.5 GHz, a double-peak behavior emerges: one peak signal of microwave photons for input photon frequency equal
is due to resonance and the other is due to the competition to the resonator frequency, ab = 0, and for input flux Ṅ = 1
between enhanced cooperativity and cancellation of left and MHz, which is well within the linear regime, as shown below.
right moving currents, which goes as sin2 θ cos θ . As expected, In Fig. 4(a), we plot the magnitude |r| and phase φr of the
the maximum efficiency occurs on resonance: ηres  80% at reflection coefficient, computed by using the general relation
 = res and tc /2π = 1.5 GHz. This efficiency will be further between input and output modes Eq. (34) with a given by
optimized in Sec. VII. Eq. (23b), which yields
√
r = |r|eiφr = 1 + κχab . (35)
VI. REFLECTED SIGNAL
The mean photon number na  Eq. (24b) and absorbed input
The field of the transmission line, see Eq. (12), can be power Pa = ωin (1 − |r|2 )Ṅ is plotted in Fig. 4(b). When the
described in terms of its final configuration at t0 = tf > t. In input frequency is on resonance with the DQD excitation at

012325-6
QUANTUM EFFICIENCY OF A SINGLE MICROWAVE . . . PHYSICAL REVIEW A 95, 012325 (2017)


 = res = ωin 2 − 4t 2 , reflection |r| is minimal and power
c Efficiency, res
absorption Pa is maximal.
0 0.2 0.4 0.6 0.8 1.0
VII. OPTIMAL CONDITIONS
The photon detector efficiency can be further improved 1
by reducing the reflection of input photons, which for the (a)
parameters chosen so far is nonzero even on resonance,

(GHz)
as shown in Fig. 4(a). Using Eq. (23b) and (35), we find 0.75
r(res ) = 0 when the resonator leakage rate κ matches the
DQD-mediated photon dissipation rate, 0.5
 

g0 /2
2(g0 sin θres )2 m0z 
κ= . (36) 0.25
2 (res )
The latter can be understood from the Fermi golden rule as 0
a transition from the one-resonator photon state |g|1 to the 0 0.5 1 1.5 2 2.5
broadened DQD excited state |e|∅ where |∅ denotes the
DQD tunnel coupling, tc /2 (GHz)
empty resonator state. The factor m0 takes into account that this
transition can occur only when |g (|e) is (un)occupied. These 1
transitions are illustrated in Fig. 1. We note that Eq. (36) can be (b)
expressed as the condition on the cooperativity [cf. Eq. (26)] 0.8

res
1 1
C(res ) = ≈ , 0.6
2|m0 (res )| 2 Efficiency,
so the optimal point does not require strong resonator-DQD 0.4 tc /2 (GHz)
coupling. 0.5
0.2 1
When Eq. (36) is satisfied, the efficiency on resonance is 1.5
given by
0
∗ 2l /31 0 0.2 0.4 0.6 0.8 1
ηres = |cos θres | . (37)
1 + 2Ṅ /|m0 |1 Resonator leakage rate, (MHz)
For a sufficiently fast dot-lead tunnel rate l  γ↑,↓ ,
FIG. 5. (a) Photon detector efficiency on resonance ηres as a
1  3l /2, see Eq. (30), so that ηres is limited mainly by
function of DQD-resonator coupling g0 and interdot DQD tunnel
the factor coupling tc , for κ/2π = 1 MHz. (b) Efficiency ηres as a function of
  resonator leakage rate κ.
|cos θres | = 1 − 4tc2 ωin
2. (38)

Higher maximum efficiency ηres can be thus achieved by

lowering tc , but at the cost of increasing the optimal coupling current cQED technology. We show that very high quantum
g0∗ , due to the reduction in the DQD-resonator coupling by efficiency is possible with currently achievable values of the
sin θres = 2tc /ωin . This behavior is shown in Fig. 5(a), where DQD-resonator coupling and DQD dissipation, and deter-
we plot efficiency on resonance ηres as a function of tc and g0 . mined the parameter regime for near-unit efficiency. While
The optimal regime defined by Eq. (36) can be reached by we utilized pure charge states of the DQD in this work,
tuning DQD-resonator parameters tc , g0 , and κ. At κ = 1 MHz, our theoretical model can readily be applied to spin-charge
it can satisfied at (tc ,g0∗ )  2π (0.5,0.2) GHz, but this g0∗ is hybridized DQD states, for example, the singlet and triplet
larger than presently available DQD-resonator couplings. We states of Ref. [52], which are protected from charge noise
therefore propose lowering the resonator leakage rate κ at fixed dephasing and hence have advantages for applications in
∗ quantum communication.
g0 . Figure 5(b) shows ηres plotted as a function of κ for several
values of tc . The maximum efficiency is ηmax = 98.4% and The proposed photon detector allows measurements of the
occurs at κ ∗ /2π = 76 kHz and tc /2π = 0.5 GHz. We note that input photon statistics as well, so that one could distinguish
this value of κ ∗ is well within the range of three-dimensional pure input Fock states from classical states by measuring the
superconducting resonators which can have leakage rates as second-order photon correlation [39]. However, determining
low as κ = 0.1 kHz [51]. Figure 6(b) shows the detector the exact relation between photon and electric current noise
efficiency as a function of DQD bias in the optimal regime. is beyond the scope of this paper. To address this relation,
one has to analyze the effect of fluctuations in the resonator
photon number due to fluctuations in Ṅ , which results in σz and
VIII. CONCLUSIONS AND DISCUSSION
current fluctuations and also in the backaction of the DQD on
In summary, we have theoretically proposed and optimized the resonator photons. Note also that multiphoton states with
a microwave photon detector based on a resonator-coupled k photons in the resonator can also be detected by tuning the
double quantum dot, which could readily be integrated with DQD to resonance with kωin , but result in weaker signal since

012325-7
CLEMENT H. WONG AND MAXIM G. VAVILOV PHYSICAL REVIEW A 95, 012325 (2017)

Charge relaxation rates (GHz)

1
(a) tc/2 (GHz)
0.5 0.8
0.75 1
1.5
Efficiency,

0.6
0.5 /2 =1 MHz 1 /2
0.4 s /2
/2
0.25
0.2
0
1 0.
(b) tc/2 (GHz)
0.5 − 20 − 10 0 10 20
0.75 1 DQD biax, /2 (GHz)
1.5
Efficiency,

0.5 /2 =76 kHz FIG. 7. The DQD charge relaxation rates, T1 relaxation rate γ1 ,
difference in excitation and relaxation rates due to thermal phonons
γs , and quasistatic dephasing rate due to charge noise γφ as a function
0.25
of the DQD bias .

0 dependence of 2 [cf. Eq. (18)] follows that of γφ , since the

0 5 10 15 20 dip in γφ near  = 0 is typically greater than the peak in γ1 .
Below we describe the minimal phenomenological model
DQD Bias, /2 (GHz)
we used to incorporate phonons and charge noise in the
FIG. 6. Efficiency of photon detector η as a function of DQD DQD dissipation and decoherence, and provide explicit for-
ph
bias , with the tunnel couplings tc /2π = (0.5,1,1.5) GHz and DQD- mulas for the relaxation rates γ↑,↓ , γ1ch , γs , and γφ . Using
resonator coupling g0 = 50 MHz, for (a) nominal resonator leakage these formulas, we plot γ1 , γs , and γφ as a function of
rate, κ/2π = 1 MHz, and (b) optimal resonator leakage rate, κ/2π = the DQD bias  in Fig. 7 for the following choice of
76 kHz, Eq. (36). parameters: DQD frequency ωσ /2π = 7 GHz, DQD-resonator
coupling g0 /2π = 50 MHz, phonon noise spectral density
the multiphoton absorption transition amplitudes are reduced Jph /2π = 0.1 GHz, temperature T = 50 mK, quasistatic bias
by the factor g k at weak coupling [53]. noise variance rms /2π = 1.2 GHz, and 1/f noise strength
c = 2.4 μeV.
ACKNOWLEDGMENTS
We are thankful to R. McDermott, B. Plourde, M. Schoen- 1. Relaxation rates due to phonons
dorf, F. Wilhelm, and C. Xu for fruitful discussions. M.V. was The electron-phonon interaction Hamiltonian can be writ-
supported by the Army Research Office under Contract No. ten as [54]
W911NF-14-1-0080 and NSF Grant No. DMR 0955500. C.W.  †
was supported by the Intelligence Community Postdoctoral Hph = sin θ βkλ (σ † akλ + akλ σ ), (A1)
Research Fellowship Program. kλ

where λ labels the phonon branches,

APPENDIX: DQD DEPOLARIZATION 
|φL (r)|2 − |φR (r)|2
AND DEPHASING RATES βkλ = vkλ dreik·r ,
2
In this section, we discuss dependence on the DQD bias
where φi (r) are the localized (Wannier) basis functions for the
 of the total DQD depolarization rate γ1 , the spontaneous
DQD electrons, and vkλ are coupling coefficients that depend
relaxation rate γs , and the charge noise induced pure dephasing
on phonon parameters [55].
γφ . The depolarization rate is the sum of the relaxation and
From Fermi’s golden rule, the stimulated emission and
excitation rates γ1 = γ↑ + γ↓ and originates due to a variety
absorption rates are given by
of processes including electron-phonon interactions and bias  ph 
noise contributions. Because charge noise and phonons couple  
γ↑ nph (ωσ )
to the DQD charge dipole matrix element, the associated = γ , (A2)
γ
ph s
1 + nph (ωσ )
relaxation is proportional sin2 θ = 4tc2 /ωσ2 , leading to the peak ↓
in γ1 and γs at  = 0. In contrast, bias noise induced pure where nph (ω) = (e ω/kB T
− 1)−1 is the phonon thermal dis-
dephasing is strongest at large  where the DQD frequency tribution. The phonon-induced spontaneous emission rate is
ωσ is more sensitive to bias noise δ, and smallest at  = 0 given by
where the DQD energy is first-order-insensitive to bias noise.
The physical reason for this dependence on  is that at large  γs = sin2 θ Jph (ωσ ), (A3)
the DQD eigenstates are pure charge states |L or |R, while
  ph 
at the charge transition  = 0 the √ eigenstates are fully hy- where Jph (ω) = 2π |βk |2 δ ω − ωkλ (A4)
bridized charge states (|L ± |R)/ 2. The qualitative bias kλ

012325-8
QUANTUM EFFICIENCY OF A SINGLE MICROWAVE . . . PHYSICAL REVIEW A 95, 012325 (2017)

ph by [58]
is the phonon spectral density and ωkλ are the phonon
dispersions. We assume a typical temperature of 50 mK. Note
γ1ch () S (ωσ )
that phonons do not cause pure dephasing due to the vanishing = sin2 θ . (A6)
phonon density of states at zero frequency [56]. 2π 22
The phonon spectrum is material dependent. For silicon If the subsystem that produces this noise is thermal at the
quantum dots, Ref. [43] measures charge relaxation times of low temperature, spontaneous relaxation rate γsch () due to
up to 10 ns at the DQD excitation frequency ωσ /2π = 12 GHz this noise would coincide with γ1ch (). For the case in which
with the tunnel coupling 2tc = 5.9 GHz. We take a nominal the source of the noise is a high-temperature environment, we
value of Jph (ω) = 0.1 GHz. expect γsch ()  γ1ch ().
The low-frequency part of the noise spectrum causes
2. Relaxation rates due to charge noise quasistatic fluctuations δωσ of the DQD excitation frequency.
We assume a 1/f charge noise spectrum The associated dephasing rate is given by [57]

c2    2 2
S (ω) = , ωl < |ω| < ωh , (A5) γφ ∂ωσ rms 2 2
∂ ωσ rms
|ω| = √ + . (A7)
2π ∂ 2 ∂ 2 2
where c = 2.4 μeV [57]. The depolarization rate is given

[1] R. H. Hadfield, Nat. Photonics 3, 696 (2009). [19] M. Kulkarni, O. Cotlet, and H. E. Türeci, Phys. Rev. B 90,
[2] Y. Nakamura, in Photonics Conference (IPC), 2012 IEEE (IEEE, 125402 (2014).
Piscataway, 2012), pp. 544–545. [20] C. Xu and M. G. Vavilov, Phys. Rev. B 87, 035429 (2013); 88,
[3] Y. Nakamura and T. Yamamoto, IEEE Photonics J. 5, 0701406 195307 (2013).
(2013). [21] K. D. Petersson, L. W. McFaul, M. D. Schroer, M. Jung, J. M.
[4] H. Spieler, Nucl. Instrum. Methods Phys. Res., Sect. A 531, 1 Taylor, A. A. Houck, and J. R. Petta, Nature (London) 490, 380
(2004). (2012).
[5] B. Peropadre, G. Romero, G. Johansson, C. M. Wilson, E. [22] T. Frey, P. J. Leek, M. Beck, A. Blais, T. Ihn, K. Ensslin, and A.
Solano, and J. J. Garcı́a-Ripoll, Phys. Rev. A 84, 063834 (2011). Wallraff, Phys. Rev. Lett. 108, 046807 (2012).
[6] Y.-F. Chen, D. Hover, S. Sendelbach, L. Maurer, S. T. Merkel, [23] P.-Q. Jin, M. Marthaler, J. H. Cole, A. Shnirman, and G. Schön,
E. J. Pritchett, F. K. Wilhelm, and R. McDermott, Phys. Rev. Phys. Rev. B 84, 035322 (2011).
Lett. 107, 217401 (2011). [24] L. Childress, A. S. Sørensen, and M. D. Lukin, Phys. Rev. A 69,
[7] G. Romero, J. J. Garcı́a-Ripoll, and E. Solano, Phys. Rev. Lett. 042302 (2004).
102, 173602 (2009). [25] Y.-Y. Liu, K. D. Petersson, J. Stehlik, J. M. Taylor, and J. R.
[8] K. Koshino, K. Inomata, T. Yamamoto, and Y. Nakamura, Phys. Petta, Phys. Rev. Lett. 113, 036801 (2014).
Rev. Lett. 111, 153601 (2013). [26] Y.-Y. Liu, J. Stehlik, C. Eichler, M. J. Gullans, J. M. Taylor, and
[9] K. Koshino, K. Inomata, Z. Lin, Y. Nakamura, and T. Yamamoto, J. R. Petta, Science 347, 285 (2015).
Phys. Rev. A 91, 043805 (2015). [27] M. Schiró and K. Le Hur, Phys. Rev. B 89, 195127 (2014).
[10] A. Poudel, R. McDermott, and M. G. Vavilov, Phys. Rev. B 86, [28] A. D. Greentree, C. Tahan, J. H. Cole, and L. C. L. Hollenberg,
174506 (2012). Nat. Phys. 2, 856 (2006).
[11] S. Barzanjeh, M. C. de Oliveira, and S. Pirandola, [29] K. Le Hur, L. Henriet, A. Petrescu, K. Plekhanov, G. Roux, and
arXiv:1410.4024. M. Schiró, C. R. Phys. 17, 808 (2016).
[12] S. Barzanjeh, S. Guha, C. Weedbrook, D. Vitali, J. H. Shapiro, [30] J. I. Cirac, P. Zoller, H. J. Kimble, and H. Mabuchi, Phys. Rev.
and S. Pirandola, Phys. Rev. Lett. 114, 080503 (2015). Lett. 78, 3221 (1997).
[13] K. Zhang, F. Bariani, Y. Dong, W. Zhang, and P. Meystre, Phys. [31] D. Pinotsi and A. Imamoglu, Phys. Rev. Lett. 100, 093603
Rev. Lett. 114, 113601 (2015). (2008).
[14] G.-W. Deng, L. Henriet, D. Wei, S.-X. Li, H.-O. Li, G. Cao, [32] M. G. Vavilov and A. D. Stone, Phys. Rev. Lett. 97, 216801
M. Xiao, G.-C. Guo, M. Schiro, K. Le Hur, and G.-P. Guo, (2006).
arXiv:1509.06141. [33] C. Bergenfeldt, P. Samuelsson, B. Sothmann, C. Flindt, and M.
[15] G.-W. Deng, D. Wei, S.-X. Li, J. R. Johansson, W.-C. Kong, Büttiker, Phys. Rev. Lett. 112, 076803 (2014).
H.-O. Li, G. Cao, M. Xiao, G.-C. Guo, F. Nori, H.-W. Jiang, and [34] S. A. Gurvitz and Y. S. Prager, Phys. Rev. B 53, 15932 (1996).
G.-P. Guo, Nano Lett. 15, 6620 (2015). [35] A. B. Vorontsov and M. G. Vavilov, Phys. Rev. Lett. 101, 226805
[16] G.-W. Deng, D. Wei, J. R. Johansson, M.-L. Zhang, S.-X. Li, (2008).
H.-O. Li, G. Cao, M. Xiao, T. Tu, G.-C. Guo, H.-W. Jiang, F. [36] The resonator mode a is dimensionless,
√ while the transmission
√
Nori, and G.-P. Guo, Phys. Rev. Lett. 115, 126804 (2015). line modes have units of bω ∼ time and bin (t) ∼ 1/ time.
[17] M.-L. Zhang, D. Wei, G.-W. Deng, S.-X. Li, H.-O. Li, G. Cao, [37] C. Gardiner and P. Zoller, Quantum Noise: A Handbook of
T. Tu, M. Xiao, G.-C. Guo, H.-W. Jiang, and G.-P. Guo, Appl. Markovian and Non-Markovian Quantum Stochastic Methods
Phys. Lett. 105, 073510 (2014). with Applications to Quantum Optics, Springer Series in
[18] J. Basset, D.-D. Jarausch, A. Stockklauser, T. Frey, C. Reichl, Synergetics (Springer, New York, 2004).
W. Wegscheider, T. M. Ihn, K. Ensslin, and A. Wallraff, Phys. [38] M. J. Collett and C. W. Gardiner, Phys. Rev. A 30, 1386
Rev. B 88, 125312 (2013). (1984).

012325-9
CLEMENT H. WONG AND MAXIM G. VAVILOV PHYSICAL REVIEW A 95, 012325 (2017)

[39] D. Walls and G. Milburn, Quantum Optics, Springer Study [50] M. Goldstein, M. H. Devoret, M. Houzet, and
Edition (Springer, Berlin, 2012). L. I. Glazman, Phys. Rev. Lett. 110, 017002
[40] A. A. Clerk, M. H. Devoret, S. M. Girvin, F. Marquardt, and (2013).
R. J. Schoelkopf, Rev. Mod. Phys. 82, 1155 (2010). [51] M. Reagor, H. Paik, G. Catelani, L. Sun, C. Axline, E. Holland,
[41] B. Q. Baragiola, R. L. Cook, A. M. Brańczyk, and J. Combes, I. M. Pop, N. A. Masluk, T. Brecht, L. Frunzio, M. H. Devoret, L.
Phys. Rev. A 86, 013811 (2012). Glazman, and R. J. Schoelkopf, Appl. Phys. Lett. 102, 192604
[42] X. Hu and S. Das Sarma, Phys. Rev. Lett. 96, 100501 (2006). (2013).
[43] K. Wang, Phys. Rev. Lett. 111, 046801 (2013). [52] C. H. Wong, M. A. Eriksson, S. N. Coppersmith, and M. Friesen,
[44] W. A. Coish and D. Loss, Phys. Rev. B 70, 195340 (2004). Phys. Rev. B 92, 045403 (2015).
[45] C. Buizert, F. H. L. Koppens, M. Pioro-Ladrière, H.-P. Tranitz, [53] W. G. van der Wiel, S. De Franceschi, J. M. Elzerman, T.
I. T. Vink, S. Tarucha, W. Wegscheider, and L. M. K. Vander- Fujisawa, S. Tarucha, and L. P. Kouwenhoven, Rev. Mod. Phys.
sypen, Phys. Rev. Lett. 101, 226603 (2008). 75, 1 (2002).
[46] O. E. Dial, M. D. Shulman, S. P. Harvey, H. Bluhm, V. Umansky, [54] G. Mahan, Many-Particle Physics, Physics of Solids and Liquids
and A. Yacoby, Phys. Rev. Lett. 110, 146804 (2013). (Springer, New York, NY, 2000).
[47] When bin  = 0, we have to keep the term aσz in the equation [55] M. J. Gullans, Y.-Y. Liu, J. Stehlik, J. R. Petta, and J. M. Taylor,
for σ . Phys. Rev. Lett. 114, 196802 (2015).
[48] J. J. Viennot, M. R. Delbecq, M. C. Dartiailh, A. Cottet, and T. [56] X. Hu, Phys. Rev. B 86, 035314 (2012).
Kontos, Phys. Rev. B 89, 165404 (2014). [57] C. H. Wong, Phys. Rev. B 93, 035409 (2016).
[49] K. D. Petersson, J. R. Petta, H. Lu, and A. C. Gossard, Phys. [58] Y. Makhlin, G. Schön, and A. Shnirman, Chem. Phys. 296, 315
Rev. Lett. 105, 246804 (2010). (2004).

012325-10