See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/256097511

Single-Photon Transistor in Circuit Quantum Electrodynamics

Article  in  Physical Review Letters · August 2013

DOI: 10.1103/PhysRevLett.111.063601 · Source: PubMed

CITATIONS READS
82 221

3 authors:

Lukas Neumeier Martin Leib

ICFO Institute of Photonic Sciences Heriot-Watt University
6 PUBLICATIONS   138 CITATIONS    24 PUBLICATIONS   687 CITATIONS   

SEE PROFILE SEE PROFILE

Michael J Hartmann
Heriot-Watt University
99 PUBLICATIONS   4,386 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Stationary States of Dissipative Many-Body Quantum Systems via Matrix Product Operators View project

None-equilibrium quantum many-body physics with photons View project

All content following this page was uploaded by Martin Leib on 26 May 2014.

The user has requested enhancement of the downloaded file.

 A Quantum Single Photon Transistor in Circuit Quantum Electrodynamics

Lukas Neumeier,∗ Martin Leib, and Michael J. Hartmann

Technische Universität München, Physik Department, James Franck Str., 85748 Garching, Germany
(Dated: December 3, 2012)
We introduce a circuit quantum electrodynamical setup for a quantum single photon transistor.
In our approach single photons propagate in two open transmission lines that are coupled via two
interacting transmon qubits. The interaction is such that photons are not exchanged between the two
transmission lines but a photon in one line can completely block respectively enable the propagation
of photons in the other line. High on-off ratios can be achieved for feasible experimental parameters.
Our approach is inherently scalable as all photon pulses can have the same pulse shape and carrier
frequency such that output signals of one transistor can be input signals for a consecutive transistor.

PACS numbers: 42.50.Ex,85.25.Cp,42.25.Fx,42.50.Nn

arXiv:1211.7215v1 [quant-ph] 30 Nov 2012

Photons are the most suitable carrier for transmitting
information over long distances as they are largely im-
mune to environmental perturbations, and can propa-
gate with very low loss and long-lived coherence in a
wide range of media [1]. The use of photons in infor-
mation processing however still suffers from the inability
to realize controlled, strong interactions between indi-
vidual photons. To make photons a more versatile in-
formation carrier, it is therefore of great importance to
conceive means of making photonic signals interact with
each other [2]. In vacuum, direct photon-photon inter-
actions are absent. Nonetheless, optical signals can in-
fluence each other in nonlinear media. Yet, the quantum
regime with interactions between individual photons only
becomes accessible for devices where optical nonlineari-
ties exceed incoherent and dissipative processes. Such
devices therefore require a strong coupling of the pho-
tons to the material that mediates the effective photon-
photon interactions. Since the coupling of light to matter
can be enhanced if light fields are confined to small vol-
umes in space, cavities and one-dimensional waveguides
are prime candidates for such devices.
Here we introduce a scheme for a quantum single pho-
ton transistor, a device that can be considered to form
a cornerstone of quantum optical information process-
ing. In our approach individual photons propagate in two
one-dimensional waveguides of low transverse dimension
and scatter off each other at a localized scattering center
formed by two two-level systems (qubits) that each cou-
ple to one of the waveguides, see figure 1a. The qubits
interact in such a way that no excitations can be ex-
changed between them and thus ensure that each pho-
ton remains in its initial waveguide after the scattering
event. Nonetheless, as we show below, the presence of a
single photon in one waveguide can completely block or
enable the propagation of a photon in the other waveg-
uide. Importantly, our approach considers propagating
light signals that all have the same carrier frequency and
is thus inherently scalable as the output signals of one
transistor can enter as input signals into a consecutive
transistor, c.f. figure 1c for an illustration. Such scal-
ability is questionable in previous proposals which are
based on different technological platforms [3, 4]. More-
over the device we propose is a passive element that does
not require any temporal tuning. This implies that the
arrival time of the photons at the scattering center can
be completely unknown.
A technology that is ideally suited for realizing the de-
vice we envision is provided by itinerant microwave pho-
tons in superconducting circuits [5, 6]. Here, coherent
scattering at a superconducting qubit [7–9] and entan-
glement with a qubit [10] have been demonstrated for
individual photons that propagate in open transmission
lines. Moreover precise shaping of single photons pulses
has been shown [11] very recently. An implementation of
our approach in circuit quantum electrodynamics thus
requires two superconducting qubits that are coupled
to open transmission lines. We show that the desired
qubit-qubit interaction can be realized with two trans-
mon qubits [12] that are coupled via a SQUID which can
be tuned to ensure that no excitations are exchanged be-
tween both transmons. Importantly, this coupling is not
dispersive [13] and thus strong as both transmons have
the same transition frequency. These rather unique pos-
sibilities for qubit-qubit interactions offered by supercon-
ducting circuits are very suitable for our aims. Moreover,
their robustness with respect to dephasing noise make
transmons ideal qubits for our device. Yet, alternatively
one could also use two flux qubits that are coupled via
an induction loop [14, 15].
To demonstrate the capabilities of the single photon
transistor we propose, we provide results for the photon
reflection and transmission spectra and probabilities for
both transmission lines, which depend on the incoming
single or two-photon pulses. Furthermore, we determine
the achievable strength of photon-photon interactions un-
der realistic experimental conditions, i.e. taking into ac-
count all dissipative processes in our setup.
Setup We consider two interacting qubits that each
couple to a one-dimensional waveguide in which the pho-
tons propagate. Here we focus on a setup for which we
can refer to a control and a target photon, where the

!t
a ∞
⌧t
1/⌧1
!1 subsystem 1
J tive) for left (right) going modes. The transmission line of
!c
!2
⌧c subsystem 2
1/⌧2
b c
Thursday, November 22, 12
Thursday, November 22, 12
FIG. 1: Sketch of the considered setup. a) Two photons
Thursday, November 22, 12
traveling in separate waveguides scatter off each other at a
scattering center formed by two interacting qubits. b) Since
all photons have the same frequency, multiple devices can
be concatenated to form a network. c) Level scheme of the
scattering center Hamiltonian Hsys , c.f. equation (1).
presence of the control photon influences the target pho-
ton’s direction of propagation, while the control photon’s
direction of propagation always changes. A sketch of this
setup is shown in figure 1. The control photon propagates
in the waveguide of subsystem 2, c.f. figure 1, which has
a closed end right where it couples to qubit 2. This ar-
rangement enhances the absorption of photons by qubit
2 and hence its inversion as compared to an open waveg-
uide end, a property that is not necessary but beneficial
for our aims. The control photon is thus always reflected.
The target photon in turn propagates in the waveguide of
subsystem 1. The qubit-qubit interaction is such that no
excitations are exchanged between the two qubits. This
implies that a photon in one waveguide can not tunnel
to the other waveguide and vice versa. Nonetheless one
control photon in waveguide 2 can completely block re-
spectively enable the propagation of a target photon in
waveguide 1 as we show below. The Hamiltonian of the
two coupled qubits reads,
ω1 z ω2 z
Hsys = σ + σ − Jσ1z σ2z ,
2 1 2 2
where the σiz are Pauli-operators, ω1 and ω2 the transi-
tion frequencies of the two qubits and J the strength of
their mutual interaction. This Hamiltonian can be im-
plemented with two transmon qubits that are coupled via
a SQUID, see figure 3 and supplementary material [16],
or with two inductively coupled flux qubits [14, 15].
Both waveguides respectively transmission lines have
a continuous spectrum of photonic modes and can be de-
2
R∞
scribed by the Hamiltonian [20], HT = −∞ dp p (rp† rp −
lp† lp ) + −∞ dp p b†p bp , where rp† (lp† ) creates a photon in
R
subsystem 1 which travels to the right (left) and b†p cre-
ates a photon in subsystem 2. p = vg η where vg is the
group velocity and the wave vector η is negative (posi-
subsystem 2 is cut and can be described by only one con-
tinuum of modes since the incoming and outgoing modes
of a semi-infinite transmission line can be mapped to an
infinite waveguide where photons only propagate in one
direction [17].
The dispersion relation of a transmission line is lin-
ear and the frequency integration can be extended to
±∞ since we only consider single photon pulses with a
frequency width that is much smaller than their carrier
frequency. For these narrow linewidth pulses we thus
take the photon-qubit coupling to be independent of the
photon frequency,
Z ∞
1
HI = √ dp[σ1+ (rp + lp ) + σ1− (rp† + lp† )] (2)
2πτ1 −∞
Z ∞
1
+√ dp[σ2+ bp + σ2− b†p ].
πτ2 −∞
Here τ1 and τ2 are the lifetimes of the two level systems
associated to their coupling to the transmission lines. A
further benefit√ of the semi-infinite waveguide in subsys-
tem 2 is the 2 enhancement of the coupling in HI . Fig-
ure 1b shows the level scheme of the two qubits described
by Hsys and the transitions induced by the photons.
In a realistic system, the qubits will be subject to dissi-
pation. We thus consider for each qubit a coupling to an
environment which we model as a continuum of bosonic
modes with creation operators c†p respectively d†p and the
R∞ R∞
Hamiltonian HB = −∞ dp p c†p cp + −∞ dp p d†p dp . The
couplingR ∞of the qubits to their environments reads  HBI =
γ/π −∞ dp (σ1+ cp + σ1− c†p ) + (σ2+ dp + σ2− d†p ) , where
p
γ is the relaxation rate of the qubits. The total Hamil-
tonian that includes the environment, the transmission
lines, the qubits and all couplings between these con-
stituents thus reads,
H = HT + Hsys + HI + HB + HBI . (3)
To investigate the dynamics of single photon pulses in
this setup, we combine quantum scattering theory [18]
(1) with the input-output formalism [19] of quantum optics
as in [20], see supplementary material [16] for details.
For the dynamics of the qubit operators, including dissi-
pative processes, we derive quantum Langevin equations
[21] starting from the Hamiltonian (3). The solutions of
these yield the source terms for the input-output rela-
tions which we then relate to the scattering matrix [16].
Photon-Photon Interaction To see the effect of the
photon-photon interaction most clearly, we first consider
the situation in which only one target photon but no

control photon is present. An incoming target pho-
Rton that travels to the right is described by |Ψt i =
dk αt (k) r† (k)|0i, where k labels the frequency compo-
nents. We assume for the target photon a pulse with a
√
Lorentzian frequency distribution, αt (k) = { πτt [i(ωt −
k) + τt−1 ]}−1 . Here τt is the temporal width of the pulse
and ωt its carrier frequency. A pulse of this form would
for example describe a photon that was spontaneously
emitted into the transmission line from an excited qubit
as experimentally realized in [22]. We here chose to oper-
ate the transistor such that the target photon is reflected
in the absence but unaffected in the presence of the con-
trol photon and choose ωt to be equal to the frequency
of the transition |g1 g2 i → |e1 g2 i (|gj i/|eij denotes qubit
j in the ground/excited state), i.e. ωt = ω1 + 2J. The
reverse mode of operation where the target photon is un-
affected in the absence and reflected in the presence of the
control photon can be selected by choosing ωt = ω1 − 2J
and works equally well. Without control photon evi-
dently no photon-photon interaction R can P take place and
the output state reads, |Ψout i = dp i=r,l,c βi (p)i†p |0i,
where the transmission amplitude is denoted βr (p), the
reflection amplitude βl (p) and the amplitude for the tar-
get photon being lost βc (p).R These amplitudes relate to
the initial state via βi (p) = dk αt (k) Si (k, p), where the
Si (k, p) are the S-matrix elements for the different pro-
cesses, see supplementary material [16]. The resulting
transmission probability for the target photon reads,
τ1 + τ12 γ + (τ1 γ)2 τt
pT = (4)
(1 + τ1 γ)(τ1 + τt + τt τ1 γ)
and the reflection probability pR = τt /(1 + τ1 γ)(τ1 + τt +
τt τ1 γ). We note that pT + pR < 1 because the photon
can also be lost due to qubit relaxation. Importantly, in
the regime of γ −1  τt  τ1 , the reflection probability
for the target photon approaches unity [3].
Next we consider the case of the same incident target
photon but now in the presence of an incoming control
photon. As the control photon inverts qubit 2, the scat-
tering center is in the state |g1 e2 i and the target pho-
ton can only couple to the transition |g1 e2 i → |e1 e2 i,
see figure 1b. This transition is detuned by 4J from
the target photon frequency, and thus the transmission
probability for the target photon approaches unity if J
is large compared to the linewidths of target pulse and
qubit 1, J  τ1−1 + τt−1 . Our scheme works best if
the control photon pulse is chosen such that it maxi-
mally inverts qubit 2. A suitable pulse is thus the time
reversed version of a pulse resulting from spontaneous
emission of qubit 2 into the transmission line [23] which
is often called an inverting pulse [24]. The generation
of inverting pulses and their release into a transmis-
sion line was demonstrated recently [11]. For the cut
transmission line in subsystem 2 an inverting pulse of
carrier frequency ωc and temporal width τc thus reads
√
|Ψc i = dk αc (k) b†k |0i with αc (k) = { πτc [−i(ωc −
R
3
k) + τc−1 ]}−1 . We note that our results do not change
if the target photon pulse also has the shape of an in-
verting pulse so that our scheme is indeed scalable. Due
to the coupling to vacuum, the qubit 2 is of course never
Rcompletely P The output state reads, |Ψout i =
P inverted.
dp1 dp2 i=r,l,c j=b,d βi,j (p1 , p2 )i† (p1 )j † (p2 )|0i. The
first index in the amplitudes βi,j (p1 , p2 ) refers to the tar-
get photon, which can be reflected, l, transmitted, r, or
lost, c, and the second index refers to the control photon
which can be reflected, b, or lost, d. For the probability
of the target photon being transmitted in the presence of
a control photon we thus get,
Z h i
2 2
pT C = dp1 dp2 |βr,b (p1 , p2 )| + |βr,d (p1 , p2 )| . (5)
We now quantify the performance of the single pho-
ton transistor we propose via the difference Cs and ratio
Rs between the transmission probabilities for the target
photon in the presence and absence of a control photon,
Cs = pT C − pT and Rs = pT C /pT , (6)
where pT C and pT are given in equations (4) and (5) re-
spectively. For Cs = 1 the setup would describe an ideal
quantum transistor for single photons. Figure 2 shows
the achievable transmission contrast, Cs , and on-off ratio,
Rs , for a realistic device with given qubit-qubit coupling
of J = 0.01 ω1 as a function of the qubit decay rate γ.
As the plot shows, an ideal single photon transistor can
be realized in the limit of vanishing γ/ω1 whereas very
good performance can already be expected for currently
realized values of J ∼ 50 MHz and γ/ω1 ∼ 10−6 [25–27],
where a single control photon changes the transmission
probability of the target photon by a factor 20.
The performance of the single photon transistor we
propose depends on the shapes of the target and con-
trol photon pulses and the parameters of the Hamilto-
nian (3). As expected the best choices for the carrier
frequencies of the control and target pulses are equal to
the transition frequencies of |g1 g2 i → |g1 e2 i respectively
|g1 g2 i → |e1 g2 i, i.e. ωc = ω2 + 2J and ωt = ω1 + 2J. For
a single transistor ω1 and ω2 may be chosen arbitrarily.
Yet to enable concatenation of multiple transistors, we
choose ω1 = ω2 . Moreover the interaction of the target
photon with qubit 1 should be as high as possible. We
choose τ1 = 20/ω1 which is compatible with experiments.
For a control photon which is an inverting pulse, i.e. the
time reversed version of a photon that would result from
a spontaneous emission process, the optimal choice for
its temporal width is obviously τc = τ2 . There are thus
two remaining parameters, τt and τc , which we have op-
timized numerically. The results are shown in figures 2b
and 2c, where we show the optimal choices of τt and τc
as a function of γ/ω1 .
Coupled transmons As stated above, the qubit-qubit
interaction in equation (1) can be realized with two
 4

a b 100
0.8 50
pT

pT C /pT
0.6 20
10
pT C

0.4 5

0.2 2
!8 !7 !6 !5 !4
10 10 10 10 10 0.001 10!8 10!7 10!6 10!5 10!4 0.001
c /!1 d /!1
1 " 10 4
1 " 106
5 " 105
5000
1 " 105
2000 5 " 104
⌧c ! 1
⌧t ! 1

1000 1 " 104

5000
500
1000
200
10!8 10!7 10!6 10!5 10!4 0.001 10!8 10!7 10!6 10!5 10!4 0.001
/!1 /!1 FIG. 3: Circuit model of the two coupled transmon qubits
Tuesday, November 27, 12
with Josephson energy EJ and shunting capacitance C. Both
FIG. 2: Performance of the quantum single photon transistor. transmons are coupled via a combination of capacitive and
a: Maximized contrast Cs = pT C − pT as a function of the inductive coupling realized with a SQUID arrangement with
inverse Q-factor of the qubits γ/ω1 for ωc = ω2 + 2J, ωt = Josephson energy EJm and shunting capacitance Cm . The
ω1 + 2J, ω2 = ω1 , τ1 = 20/ω1 , J = 0.01 ω1 and τc = τ2 . transmons are capacitively coupled to transmission lines.
b: On-off ratio Rs = pT C /pT for the same parameters. c:
Optimal choice of τt for a given γ/ω1 . d: Optimal choice of
τc for a given γ/ω1 .
tum single photon transistor in circuit quantum electro-
dynamics that is inherently scalable as both photons can
transmon qubits that are coupled via a SQUID. The have the same carrier frequency and pulse shape. The
circuit we consider is sketched in figure 3. and de- performance of the device might be further improved by
scribed by the Lagrangian [28], L = L1 + L2 + L12 , suppressing losses with multiple, regularly spaced qubit
where the Lagrangians of the individual qubits read, pairs [29]. Moreover, the complexity of a network built
C with such transistors could be increased further by inte-
Lj = C2 ϕ̇2j + 2g (ϕ̇j − Vj )2 + EJ cos (ϕj /ϕ0 ) and L12 =
Cm 2

ϕ1 −ϕ2
 grating directional couplers between them [30].
2 (ϕ̇1 − ϕ̇2 ) + EJm cos ϕ0 . Here ϕ0 = ~/(2e) is
Acknowledgements This work is part of the Emmy
the flux quantum divided by 2π, C and EJ are the ca-
Noether project HA 5593/1-1 and the CRC 631, both
pacitance and Josephson energy of the individual trans-
funded by the German Research Foundation, DFG.
mons which are assumed to be identical. Cm and EJm
are the capacitance and Josephson energy of the capaci-
tively shunted coupling SQUID. The Cg are the coupling
capacitances between the transmission lines and the in-
dividual transmons (also assumed to be identical) and ∗
Electronic address: lukas-neumeier@gmx.de
the Vi are the fully quantum mechanical quadratures of [1] J.L. O’Brien, A. Furusawa, and J. Vuckovic, Nat. Photon
the electric potential of the transmission line fields. All 3, 687 (2009)
Josephson energies of the setup are tunable by threading [2] H.J. Kimble, Nature (London) 453, 1023 (2008)
[3] D.E. Chang, A.S. Sørensen, E.A. Demler and M.D.
external fluxes ΦJ and ΦJm through respective SQUID
Lukin, Nat. Phys. 3, 807 (2007).
loops, c.f. figure 3. We write the corresponding Hamil- [4] F.-Y. Hong and S.-J. Xiong, Phys. Rev. A 78, 013812
tonian of the transmons in terms of creation and anni- (2008).
hilation operators a†j and aj [16]. By tuning EJ and [5] R.J. Schoelkopf and S.M. Girvin, Nature (London) 551,
EJm such that EJE+E Jm
Jm
= C+CCmm+Cg , all interactions 664 (2008)
[6] A. Wallraff et al, Nature (London) 431, 162 (2004)
of the form a1 a†2 + a†1 a2 cancel and the leading term of [7] O. Astafiev, A. M. Zagoskin, A.A. Abdumalikov Jr., Yu.
the remaining interactions reads − (C+C 2EC Cm
a† a1 a†2 a2
m +Cg ) 1
A. Pashkin, T. Yamamoto, K. Inomata, Y. Nakamura,
which is equivalent to the interaction in equation (1) with J.S. Tsai, Science 327, 840 (2010)
Cm [8] I.-C. Hoi, C.M. Wilson, G. Johansson, T. Palomaki, B.
J = 2(C+C m +Cg )
EC . Within the approximations we use
Peropadre, and P. Delsing, Phys. Rev. Lett. 107, 073601
[16] the achievable qubit-qubit coupling is J < EC /10. (2012)
In conclusion, we have introduced a scheme for a quan- [9] I.-C. Hoi, T. Palomaki, J. Lindkvist, G. Johansson, P.
 5

Delsing, and C.M. Wilson, Phys. Rev. Lett. 108, 263601 of transmission of a single photon, the S-matrix element
(2012) reads,
[10] C. Eichler, C. Lang, J. M. Fink, J. Govenius, S. Filipp,
and A. Wallraff, arXiv:1209.0441 †
Sr (k, p) = h0|rout (p)rin (k)|0i, (8)
[11] Y. Yin, Y. Chen, D. Sank, P.J.J. O’Malley, T.C. White,
R. Barends, J. Kelly, E. Lucero, M. Mariantoni, A. where
Megrant, C. Neill, A. Vainsencher, J. Wenner, A.N. Ko-
rotkov, A.N. Cleland and J.M. Martinis, arXiv:1208.2950 †
rin/out (ω) = lim eiHt0 e−iHT t0 r† (ω)eiHT t0 e−iHt0 , (9)
[12] J. Koch, T.M. Yu, J. Gambetta, A.A. Houck, D.I. Schus- t0 →∓∞
ter, J. Majer, A. Blais, M.H. Devoret, S.M. Girvin, and R∞ R∞
R.J. Schoelkopf, Phys. Rev. A 76, 042319 (2007). and HT = −∞ dp p (rp† rp − lp† lp ) + −∞ dp p b†p bp is as in
[13] G. Kirchmair, B. Vlastakis, Z. Leghtas, S.E. Nigg, H. the main text. In turn, for a two photon process, where
Paik, E. Ginossar, M. Mirrahimi, L. Frunzio, S.M. the target photon is transmitted and the control photon
Girvin, and R.J. Schoelkopf, arXiv:1211.2228
is lost in the vacuum, the S-matrix element reads,
[14] T.P. Orlando et al.,Phys. Rev. B 60, 15398 (2000)
[15] J.E. Mooij et al, Science 60, 1036 (1999) †
[16] see Supplementary Material
Sr,d (k, k 0 , p, p0 ) = h0|rout,p dout,p0 rin,k b†in,k0 |0i (10)
[17] J.T. Shen and S. Fan, Phys. Rev. A 79, 023837 (2009)
[18] J.R. Taylor, Scattering Theory: the Quantum Theory of For our system it is convenient to introduce even and odd
Nonrelativistic Collisions (Dover, New York, 2006), Sec. modes for transmission line 1,
9-e and Chap. 10
[19] C. W. Gardiner and M. J. Collet, Phys. Rev. A, 31, 6 1
a1 (p) = √ (rp + l−p ) (11)
(1985) 2
[20] S. Fan, S. E. Kocabas J.T. Shen, Phys. Rev. A 82, 063821 1
(2010) å1 (p) = √ (rp − l−p ) (12)
[21] C. W. Gardiner and P. Zoller, Quantum Noise, Springer-
2
Verlag, (2000). Due to the form of the photon-qubit coupling in equation
[22] C. Eichler, D. Bozyigit, C. Lang, L. Steffen, J. Fink, and
A. Wallraff, Phys. Rev. Lett. 106, 220503 (2011).
(2) of the main text, only even modes couple to the qubit
[23] J.I. Cirac, P. Zoller, H.J. Kimble, and H. Mabuchi, Phys. whereas odd modes completely decouple from the rest of
Rev. Lett. 78, 3221 (1997) the system and describe freely propagating photons. The
[24] E. Rephaeli, J.T. Shen, S. Fan, Phys. Rev. A, 82, 033804 coupling √ strength to the even modes is enhanced by a
(2010) factor of 2.
[25] M. Sandberg, M.R. Vissers, T. Ohki, J. Gao, J. Aumen- To calculate the desired scattering matrix elements a
tado, M. Weides, and D.P. Pappas, arXiv:1211.2017 link to equations of motion, i.e. the Langevin equations
[26] C. Rigetti, J.M. Gambetta, S. Poletto, B.L.T. Plourde,
J.M. Chow, A.D. Córcoles, J.A. Smolin, S.T. Merkel,
[19] that describe the dynamics of the scattering center
J.R. Rozen, G.A. Keefe, M.B. Rothwell, M.B. Ketchen, is exploited. To make use of the Langevin equations, it
and M. Steffen, Phys. Rev. B 86, 100506(R) (2012). is necessary to connect the scattering operators to stan-
[27] J.M. Fink, M. Göppl, M. Baur, R. Bianchetti, P.J. Leek, dard input-output operators that fulfill the input-output
A. Blais, and A. Wallraff, Nature 454, 315 (2008). relations,
[28] Yong Hu, Guo-Qin Ge, Shi Chen, Xiao-Fei Yang, You- r
Ling Chen, Phys. Rev. A 84, 012329 (2011) 2
[29] H. Zoubi and H. Ritsch, Europhys. Lett. 90, 23001 (2010) aout (t) = ain (t) − i σ− (t), (13)
[30] H. S. Ku, F. Mallet, L. R. Vale, K. D. Irwin, S. E. Russek,
τ
G. C. Hilton, and K. W. Lehnert, IEEE Transactions on where ain (t) and aout (t) are any even mode input/output
Applied Superconductivity 21, 452 (2011)
operators, defined by
1
Z
SUPPLEMENTARY MATERIAL ain/out (t) ≡ √ dωe−iω(t−t0/1 ) a0/1 (ω). (14)
2π
Scattering theory and input-output relations Here, the operator a0/1 (ω) plays the role of an initial
value in the Heisenberg picture,
In this work we calculate the output states of our de-
vice for given input states by making use of a combination a0/1 (ω) = eiHt0/1 a(ω)e−iHt0/1 (15)
of scattering theory and the input-output formalism. In
this approach, output and input states are connected via The corresponding input and output scattering operators
the scattering matrix (S-matrix), for even modes are defined as in equation (9),

|Ψout i = S|Ψin i (7) ain/out (ω) = lim eiHt0 e−iHT t0 a(ω)eiHT t0 e−iHt0 (16)
t0 →∓∞
The S-matrix elements in turn can be written in terms = lim eiωt0 eiHt0 a(ω)e−iHt0 ,
of scattering operators [18]. For example for the process t0 →∓∞
 6

transmission line reads,

C 2
Cm 2
L = ϕ̇1 + ϕ̇22 + (ϕ̇1 − ϕ̇2 ) (18)
2 2
Cg Cg
+ (ϕ̇1 − V1 )2 + (ϕ̇2 − V2 )2
2    2  
ϕ1 ϕ2
+ EJ cos + cos
ϕ0 ϕ0
ϕ1 − ϕ2
 
+ EJm cos
ϕ0
where ϕ0 = ~/(2e) is the flux quantum divided by 2π,
C and EJ are the capacitance and Josephson energy of
the individual transmons which are assumed to be iden-
tical. Cm and EJm are the capacitance and Josephson
energy of the capacitively shunted coupling SQUID. The
Cg are the coupling capacitances between the transmis-
sion lines and the individual transmons (also assumed to
be identical) and the Vi are the fully quantum mechani-
cal electric potential quadratures of the transmission line
fields. Hence, the corresponding Hamiltonian reads,
 
Cm
H = 4EC n21 + n22 + 2 n1 n2 (19)
FIG. 4: Circuit model of the pair of transmons with Joseph- C + Cg + Cm
 
son energy EJ and shunting capacitance C. Both transmons Cm
are coupled via a combination of capacitive and inductive cou- + 8EC ng,1 n1 + n2
C + Cg + Cm
pling realized with a SQUID arrangement with Josephson en-  
ergy EJm and shunting capacitance Cm . The transmons are Cm
+ 8EC ng,2 n1 + n2
capacitively coupled to transmission lines and all Josephson C + Cg + Cm
energies of the setup are tunable by threading external fluxes − EJ (cos φ1 + cos φ2 ) − EJm cos(φ1 − φ2 ),
ΦJ and ΦJm through respective SQUID loops.
where φi = ϕi /ϕ0 , ng,i = ϕ0 Cg Vi for i = 1, 2 and
~2 C+Cg +Cm
EC = 8ϕ 2 (C+C )(C+C +2C ) . The secondary capacitive
g g m
0
where we have made use of the Baker-Campbell- coupling between transmission line 1 and transmon 2 and
Hausdorff formula. We now solve equation (16) for a(ω) vice versa is small compared to the other couplings since
and plug the result into equation (14) by making use of Cm /(C + Cg + Cm )  1. We thus neglect such cou-
equation (15). Taking the limits t0 → −∞ respectively plings and obtain for the coupling between the transmis-
t1 → +∞ in equation (14) we find that the scattering sion lines and the transmons,
operator ain (ω) is the Fourier transform of the Heisen-
berg operator ain (t) in the limit t0 → −∞ and that the Hg = 8EC (ng,1 n1 + ng,2 n2 ) (20)
the scattering operator aout (ω) is the Fourier transform We furthermore separate the Josephson terms into local
of the Heisenberg operator aout (t) in the limit t1 → +∞ and nonlocal terms,
[20],
EJm cos(φ1 − φ2 ) = EJm (cos φ1 + cos φ2 ) (21)
1
Z | {z }
−iωt
ain/out (t) = √ dωain/out (ω)e (17) local
2π +EJm [(cos φ1 − 1)(cos φ2 − 1) + sin φ1 sin φ2 ],
| {z }
nonlocal
Using equation (17) we express the scattering matrix S
in terms of input and output operators which are found where we have added an irrelevant constant. This leaves
from solutions of the Langevin equations that describe us with the local Hamiltonians for each transmon (i =
the dynamics of the scattering center. 1, 2),
Hi = 4EC n2i − E J cos φi , (22)

Coupled Transmons
where E J = EJ + EJm and the coupling Hamiltonian,
Cm
H12 = 8EC n1 n2 (23)
The Lagrangian describing the two transmons that are C + Cg + Cm
mutually coupled via a SQUID and each couple to a − EJm [(cos φ1 − 1)(cos φ2 − 1) + sin φ1 sin φ2 ]
 7

We describe the transmons in the approved approxima- Here we have applied a rotating wave approximation and
tion with anharmonic oscillators [12] and introduce rais- made use of the fact that our initial states only contain
ing and lowering operators a†i and ai via one photon per transmission line which allows us to ne-
glect terms of the form a1 a1 a†2 a†2 + H.c.. Since the dy-
 1 namics of our system is, for the initial state we consider,
i EJ 4
ni = (ai − a†i ) (24) restricted to the subspace of at most one excitation per
2 2EC
 1 transmon respectively transmission line we can write our
2EC 4 Hamiltonian in terms of Pauli matrices,
φi = (ai + a†i ) (25)
EJ
Keeping only the leading nonlinear terms we thus find, H 2×2 = Hsys + Hg2×2 (30)
EC † †
q
Hi ≈ 8EC E J a†i ai − a a ai ai (26)
2 i i wherepHsys is as in equation (1) of the main text with
We are interested in a scenario where tunneling of exci- ω = 8EC E J + δω, where δω accumulates all minor
tations from one transmon to the other is strongly sup- renormalizations of the transmon frequencies due to the
pressed. The leading tunneling terms are found by ex- nonlinear terms of the cosine potentials and
panding the coupling Hamiltonian H12 to linear order in
a1 and a2 . With a rotating wave approximation we find,
Cm
J= EC . (31)
Cm 2(C + Cm + Cg )
8EC n1 n2 − EJm φ1 φ2 ≈ (27) Since Cm /(C + Cg + Cm )  1 we find the upper limit,
C + Cg + Cm
s  
2EC Cm
≈ EJ − EJm (a1 a†2 + a†1 a2 ) 1
EJ C + C g + C m J< EC , (32)
10
These terms vanish if one chooses the external fluxes that
control the values of EJ and EJm such that, for the strength of the qubit-qubit interaction. In turn
EJm Cm the couplings to the transmission lines read,
= (28)
EJ + EJm C + Cm + Cg
  14 X
The rotating wave approximation applied for deriving EJ
Hg2×2 = 4iEC ng,j (σj+ − σj− ),
equation (27) furthermore requires that EJm < 2EJ /3. 2EC j=1,2
The desired density-density interaction is contained in
the nonlocal cosine interaction terms,
which become identical to equation (2) after applying a
EJm 2 2 EJm
HJ = − φ1 φ2 ≈ −2EC a† a1 a†2 a2 (29) rotating wave approximation.
4 EJ + EJm 1

View publication stats