Superconducting Qubits and Circuits:

Artificial Atoms Coupled to Microwave Photons

Steven M. Girvin1
Yale University
Department of Physics
New Haven, CT 06520
United States of America

⃝2011
c

1
⃝S.M.
c Girvin, 2011
Lectures delivered at Ecole d’Eté Les Houches, July 2011
To be published by Oxford University Press.
ii S.M. Girvin

These lectures are based on work done in collabora-

tion with many students, postdocs and colleagues,
including particularly my friends and colleagues
Michel H. Devoret and Robert J. Schoelkopf.
Contents

0.1 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1 Introduction to Quantum Machines 3

2 Quantum Electrical Circuits 5

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Plasma Oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 Quantum LC Oscillator . . . . . . . . . . . . . . . . . . . . . . . 8
2.3.1 Coherent States . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4 Modes of Transmission Lines Resonators . . . . . . . . . . . . . . 18
2.5 ‘Black Box’ Quantization . . . . . . . . . . . . . . . . . . . . . . 23

3 Superconductivity 27

4 Superconducting Qubits 33
4.1 The Cooper Pair Box . . . . . . . . . . . . . . . . . . . . . . . . 33
4.2 Inductively Shunted Qubits . . . . . . . . . . . . . . . . . . . . . 43

5 Noise induced Decoherence in Qubit Circuits 49

6 Introduction to Cavity and Circuit QED 59

7 NOBEL SYMPOSIUM TEXT BEGINS HERE 65

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
7.2 Circuit QED . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
7.2.1 The Quantronium Qubit . . . . . . . . . . . . . . . . . . . 70
7.2.2 Transmon Qubits . . . . . . . . . . . . . . . . . . . . . . . 71
7.2.3 Fluxonium Qubits . . . . . . . . . . . . . . . . . . . . . . 74
7.3 Recent Progress and Future Directions . . . . . . . . . . . . . . . 74

8 Appendix: Semi-Infinite Transmission Lines, Dissipation and

Input/Output Theory 79
8.0.1 Quantum Input-output theory for a driven cavity . . . . . 84

Bibliography 93

iii
iv S.M. Girvin
Circuit QED 1

Key papers students may wish to consult:

• Blais et al., ‘Cavity quantum electrodynamics for superconducting elec-
trical circuits: An architecture for quantum computation,’ Phys. Rev.
A 69, 062320 (2004).

• Clerk et al., ‘Introduction to quantum noise, measurement and am-

plification,’ Rev. Mod. Phys (2008) (longer version with pedagogical
appendices available at arXiv.org:0810.4729)

0.1 Outline
1. Introduction to Quantum Machines
2. Why are there so few degrees of freedom in SC circuits?
(a) Superconducting excitation gap
(b) Strong Coulomb force
3. Brief Introduction to the Quantum to Classical Crossover –why is it hard
to make large objects that are quantum and not classical, and that have
simple spectra like atoms?
4. Elements of a quantum electrical circuit: L,C,JJ
5. Quantization of a lumped-element LC oscillator
(a) What is a photon? First quantization and second quantization views.
(b) Role of circuit admittance Y (ω)
(c) Distributed element resonators: Box modes and how to quantize
them using admittance
6. Josephson junctions
(a) Just what is superconductivity anyway?
(b) How does a Josephson junction JJ work?
7. Superconducting Qubits
(a) Brief survey of different SC qubit topologies and geometries: charge,
phase, flux, fluxonium
(b) dephasing and decay mechanisms, T1 , T2 , sweet spots
(c) strengths and weaknesses of different qubit designs
(d) The transmon: World’s simplest SC qubit: a JJ and an antenna
8. Introduction to Cavity and Circuit QED
(a) Jaynes-Cummings model for a two-level atom
2 S.M. Girvin

(b) dressed atom/polariton picture

(c) vacuum Rabi coupling and degenerate regime
(d) dispersive regime
(e) strong dispersive regime, number splitting
(f) weakly anharmonic transmon in a cavity: perturbation expansion in
the anharmonicity; α ∼ g
9. Input-Output Theory
(a) qubit driving via driving cavity
(b) quantum noise, dissipation, decay and the Purcell effect
(c) linear readout, back action, quantum Zeno
(d) non-linear readout methods
i. Josephson bifurcation amplifier
ii. Chirped JBA
iii. Using the self-non-linearity induced by qubit
10. Quantum Control and Algorithms
(a) single qubit rotations, Rabi, Ramsey
(b) two-qubit entangling operations using conditional phase shifts
(c) qubit-cavity entangling operations in the strong-dispersive regime
(d) QND measurement of photon number
(e) Bell/CSHS inequalities
(f) Grover Search and Deutsch-Josza algorithms on a two-qubit proces-
sor
(g) GHZ/Mermin inequalities
(h) Towards Quantum Error Correction
i. measurement based QEC
ii. non-measurement based QEC
11. Towards Scaling up to Many Qubits
(a) importance of large on-off coupling ratios
(b) Houck double transmon
(c) multiple-cavity architectures and communication protocols
(d) cavities as quantum memories
(e) cavities as qubits
(f) qubits as switches
(g) Many-body physics with lattice models for strongly interacting po-
laritons
Chapter 1

Introduction to Quantum
Machines

A quantum machine is a device whose degrees of freedom are intrinsically quan-

tum mechanical. While we do not yet fully understand the properties of such
devices, there is great hope (and some mathematical proof) that such machines
will have novel capabilities that are impossible to realize on classical hardware.
You might think that quantum machines have already been built. For example,
the laser and the transistor would seem to rely on quantum physics for their
operation. It is clear that the frequency of a laser cannot be computed without
the quantum theory that predicts the excitation energies of the atoms in the
laser. Similarly the optimal bias voltage of a bipolar transistor depends on the
electronic band gap energy of the material from which it is made. Nevertheless,
it is only the particular values of the operating parameters of these machines
that are determined by quantum physics. Once we know the values of these
parameters, we see that these are classical machines because their degrees of
freedom are purely classical. Indeed the light output from a laser is special
because it is exactly like the RF output of the classical oscillator that powers
a radio station’s antenna. Similarly, the currents and voltages in an ordinary
transistor circuit need not be treated as non-commuting quantum operators in
order to understand the operation of the circuit. Its degrees of freedom are, for
all intents and purposes, classical.
These lectures are devoted to understanding the basic components of quan-
tum machines that can be constructed from superconducting electrical circuits.
These circuits can be used to create resonators which store individual microwave
photons and to create superconducting quantum bits. Both of these circuit ele-
ments are intrinsically quantum mechanical. They have quantized energy levels
with spacing much greater than temperature (for low enough temperatures!)
and they can be placed into quantum superpositions of different energy states.
To properly predict their properties, the currents and voltages in such circuits
must be represented by non-commuting quantum operators.

3
4 S.M. Girvin

The beauty of electrical circuits is that they can be constructed in a modular

manner combining together a few different building blocks in a simple manner.
The wires connecting these building blocks have to be capable of carrying quan-
tum signals [1], but are still relatively simple and the problem of spatial mode
matching that occurs in optical circuits is largely eliminated. Furthermore, with
modern lithographic techniques, parallel fabrication of complex structures is rel-
atively straightforward lending hope that (someday) it will be possible to scale
up to processors with large numbers of qubits.
We will study different qubit designs and their relative merits. We will also
learn how to control and readout the quantum states of qubits and cavities, and
how to entangle different their quantum degrees of freedom. In recent decades,
we have come to understand that superposition and entanglement are features
of quantum physics which can be used as resources to make powerful quantum
computers to process information in ways that are impossible classically. What
else we can do with quantum machines is not yet fully understood. The people
who invented the laser had no idea that it would be used to play music, com-
municate over optical fiber, and perform eye surgery. We expect that similar
surprises and unexpected applications will be developed once quantum hardware
becomes routine enough to play with.
Chapter 2

Quantum Electrical Circuits

2.1 Introduction
Quantum electrodynamics is the theory of interaction between electrons (and
atoms) with electromagnetic fields. These lecture notes discuss the closely re-
lated problem of quantization of electrical circuits [1, 2]. Experimental progress
over the last decade in creating and controlling quantum coherence in supercon-
ducting electrical circuits has been truly remarkable. The quantum electrody-
namics of superconducting microwave circuits has been dubbed ‘circuit QED’
by analogy to cavity QED in quantum optics. These lecture notes will describe
the quantum optics approach to microwave circuits with superconducting qubits
playing the role of artificial atoms whose properties can be engineered. Despite
being large enough to be visible to the naked eye, these artificial atoms have
a very simple discrete set of quantized energy levels which are nearly as well
understood as those of the prototypical single-electron atom, hydrogen. Fur-
thermore it has proven possible to put these atoms into coherent superpositions
of different quantum states so that they can act as quantum bits. Through
clever engineering, the coherence times of such superposition states has risen
more than four orders of magnitude from nanoseconds for the first supercon-
ducting qubit created in 1999 [3] to several tens of microseconds today [4]. This
’Moore’s Law’ for the exponential growth of coherence time is illustrated in
Fig. (??).
In addition to being a potentially powerful engineering architecture for build-
ing a quantum computer, circuit QED opens up for us a novel new regime to
study ultra-strong coupling between ‘atoms’ and individual microwave photons.
The concept of the photon is a subtle one, but hopefully these notes will convince
the reader that microwaves, despite their name, really are particles. We will ac-
cordingly begin our study with a review of the quantization of electromagnetic
fields in circuits and cavities.
The quantization of electrical circuits has been thoroughly addressed in the
Les Houches lecture notes of my colleague, Michel Devoret [2], to which I direct

5
6 S.M. Girvin

the interested reader. The circuit elements that are available to the quantum
engineer are those familiar to classical engineers: resistors, capacitors, and in-
ductors. [Resistors cause unwanted dissipation and we will attempt to avoid
them. See however further discussion in the Appendix (8) of spontaneous emis-
sion into transmission lines.] In addition to these standard circuit elements,
there is one special element in superconducting circuits, the Josephson tunnel
junction. We will be learning more about superconductivity and Josephson
junctions later, but for now we simply note the following. With capacitors and
inductors we can build simple LC harmonic oscillators. If we can eliminate all
resistors then the harmonic oscillations will be undamped. The use of super-
conducting circuits takes us a long way towards this goal of zero dissipation,
about which more later. The essential feature of superconductivity is that elec-
trons of opposite spin pair up and condense into a special ground state with a
substantial excitation gap 2∆ needed to break one of the pairs and create an
excited state. This pair excitation gap is essential to the ability of current to
flow in a superconductor without dissipation. A closely related advantage of the
excitation gap is that it dramatically reduces the number of effective degrees of
freedom in the circuit, allowing us to construct artificial ‘atoms’ that behave like
simple single-electron atoms even though they are made up of 109 − 1012 alu-
minum atoms. The extremely powerful force of the Coulomb interactions also
plays an essential role in limiting the low energy degrees of freedom in circuits.
When the Coulomb interaction is unscreened, the gapless collective motion of
supercurrents is lifted up to the plasma frequency which is orders of magnitude
higher than any relevant frequency scale for the circuits we will consider. In the
presence of screening due to ground planes or shields, the plasma oscillations
are ‘acoustic modes’ with a linear dispersion and velocity close to the speed of
light in vacuum. When quantized, these will be our propagating photons.

2.2 Plasma Oscillations

Because the powerful effect of long-range Coulomb interactions plays a crucial
role in simplifying the spectrum of quantum electrical circuits, let us begin our
analysis by reviewing the plasma oscillations in a bulk metal. Throughout this
work we will use SI units. We will consider infinitesimal density fluctuations δn
around the mean electron number density n. In the ‘jellium’ model the mean
charge density is canceled by the ionic background so the net charge density is

ρ(⃗r) = −e δn. (2.1)

The current flowing (to zeroth order in δn) is

⃗ r, t) = −en⃗v (⃗r, t),
J(⃗ (2.2)

where the local electron mean velocity field obeys Newton’s law
∂ −e ⃗
⃗v = E, (2.3)
∂t m
Circuit QED 7

where m is the electron (effective) mass. This in turn yields

∂ ⃗ ne2 ⃗
J= E. (2.4)
∂t m
Taking the divergence of both sides of this equation and applying Gauss’s law

∇ ⃗ = ρ,
⃗ ·E (2.5)
ϵ0
and the continuity equation

⃗ · J⃗ + ∂ ρ = 0,
∇ (2.6)
∂t
yields
∂2
ρ = −ωp2 ρ (2.7)
∂t2
where the so-called ‘plasma frequency’ is given by1

ne2
ωp2 ≡ . (2.8)
mϵ0
Electromagnetic waves cannot propagate in a plasma at frequencies below the
plasma frequency [?]. In the earth’s ionosphere, the typical plasma frequency is
range of 10’s of MHz and varies between night and day, thereby affecting short-
wave radio reception. In the typical metals we will be concerned with (e.g.,
aluminum), the valence electron density is sufficiently high that the plasma
frequency is in the ultraviolet region of the optical spectrum. Hence aluminum
(whose plasma frequency ωp /(2π) ∼ 3.6×1015 Hz corresponds to a photon energy
of ∼ 15 eV) is highly reflective in the visible. Essentially the electrons are so
dense and so agile that they screen out any electric fields almost perfectly over
a very short screening distance. For frequencies far below the plasma frequency,
Maxwell’s equations yield

∇
⃗ ×∇ ⃗ ≈ −λ−2
⃗ ×E ⃗
p E, (2.9)

where the penetration depth, λp , of the electromagnetic fields is

c 1
λp = =√ , (2.10)
ωp 4πnre

where the classical radius of the electron is given by

e2 1
re = ≈ 2.818 × 10−15 m. (2.11)
4πϵ0 mc2
1 We neglect here the various details of the band structure of Al as well as the possibility

that the core electrons in the atoms of the metal contribute a dielectric constant ϵ ̸= 1 seen
by the valence electrons whose dynamics create the plasma oscillations of the metal.
8 S.M. Girvin

For Al, Eq. 2.10 yields2 λp ∼ 14nm. We will be dealing with GHz frequency
scales many orders of magnitude below the plasma frequency and centimeter
wavelength scales relative to which the penetration depth is effectively zero.
Exercise 2.1 Derive Eq. 2.9 in the limit of low frequencies and show that
it leads to exponential decay of transverse electromagnetic waves with decay
length λp .

The above jellium model yields a plasma mode which is completely dispersionless–
the mode frequency is independent of wave vector q. The frequency of the bulk
collective plasma mode is vastly higher than any microwave frequency that we
will be dealing with. From the point of view of quantum mechanics, the amount
of energy required to create a bulk plasmon is so large that we can consider these
degrees of freedom to be frozen into their quantum mechanical ground state.
Hence they can be ignored. The approximations leading to Eq. (2.2) breakdown
as short distances due to the granularity of the electron charge. At large very
wave vectors approaching the Fermi wave vector, the jellium continuous charge
picture breaks down and the plasma oscillation frequency rises and the mode be-
comes ‘Landau-damped’ due to the collective charge oscillation mode decaying
into single-particle excitations [?]. Conversely for extremely small wave vectors,
there is a cutoff associated with the finite size of any sample. This we can take
into account by considering the capacitance matrix between different lumps of
metal in the circuit we are trying to quantize. In certain circumstances, the ca-
pacitance matrix is such that there do exist collective charge oscillation modes
which are down in the microwave range. These will be the important modes
which we will quantize. Here the superconductivity is vital for gapping the
single-particle excitations so that the collective charge modes are both simple
and extremely weakly damped.

2.3 Quantum LC Oscillator

The circuit element with the simplest dynamics is the LC oscillator illustrated
schematically in Fig. (2.1a). Now that we understand that supercurrents can
flow essentially without dissipation and that the great strength of the Coulomb
interaction lifts density fluctuations up to optical frequencies, we can understand
that the LC oscillator has, to a very good approximation, only a single low-
energy degree of freedom, namely uniform divergenceless current flow in the
wire of the inductor which does not build up charge anywhere except on the
plates of the capacitor. This is a very good approximation in the ‘lumped
element’ limit where the physical size of the LC oscillator is much smaller than
than the wavelength of electromagnetic waves at the frequency of the oscillator,
λ = 2πc/Ω. [This caveat is associated with the unstated assumption in our
discussion of plasma oscillations that we neglected relativistic retardation effects.
2 The measured value of the London penetration depth in Al (at zero frequency) is somewhat

larger, λp ∼ 51.5nm. The difference is presumably due to variation in the core electron
dielectric constant with frequency which has been neglected in our model.
Circuit QED 9

That is, we effectively assumed c = ∞.] In terms of the capacitor charge Q and
the inductor current I the Lagrangian is readily written
1 2 1 Q2
L= LI − . (2.12)
2 2 C
Using charge conservation, I = +Q̇, this can be cast into the more familiar form
L 2 1 2
L= Q̇ − Q (2.13)
2 2C
which yields the Euler-Lagrange equation of motion

Q̈ = −Ω2 Q, (2.14)

where the natural oscillation frequency is

1
Ω= √ . (2.15)
LC
The momentum conjugate to the charge is the flux through the inductor
δL
Φ= = LQ̇ = LI. (2.16)
δ Q̇
Thus the Hamiltonian can be written
Φ2 1 2
H = ΦQ̇ − L = + Q . (2.17)
2L 2C
Hamilton’s equations of motion are then
∂H Φ
Q̇ = = (2.18)
∂Φ L
∂H Q
Φ̇ = − =− . (2.19)
∂Q C
In the usual way, the coordinate and its conjugate momentum can be pro-
moted to quantum operators obeying the canonical commutation relation

[Φ̂, Q̂] = −ih̄ (2.20)

and we can write the Hamiltonian

{ }
h̄Ω { † } 1
H= a a + aa† = h̄Ω a† a + , (2.21)
2 2
in terms of raising and lowering operators
1 1
a = +i √ Φ̂ + √ Q̂ (2.22)
2Lh̄Ω 2Ch̄Ω
1 1
a† = −i √ Φ̂ + √ Q̂ (2.23)
2Lh̄Ω 2Ch̄Ω
10 S.M. Girvin

Figure 2.1: Simple LC electrical oscillator analogous to a mass and spring me-
chanical oscillator. a) The coordinate is taken to be Q and the conjugate mo-
mentum is Φ. b) The coordinate is taken to be Φ and the conjugate momentum
is Q. Note the important sign change in the definition of Q in the two cases,
needed to maintain the canonical commutation relation between momentum
and coordinate.

which obey the usual relation

[a, a† ] = 1. (2.24)
In the above discussion we chose the charge Q on the capacitor as the natural
coordinate of the harmonic oscillator and found that the inductor flux Φ was the
momentum conjugate to this flux. In the picture we interpret the capacitance C
as the inverse of the ‘spring constant,’ and the inductance L as the ‘mass.’ This
seems natural from our intuitive view of the capacitance as storing the potential
energy and the inductor storing the kinetic energy (actually the kinetic energy
of the electrons makes only a small contribution (called the ‘kinetic inductance’)
to the total inductance. It is primarily the energy stored in the magnetic field
created by the current which dominates the inductance in most situations.)
When dealing with Josephson junctions we will start with this same repre-
sentation but then find that they act as non-linear inductors and it will be more
convenient to take Φ to be the coordinate rather than the momentum. In order
to get used to this alternative representation, we will practice here on the LC
oscillator. Following Devoret [2] let us define the node flux at the point show in
Fig. (2.1b) by
∫ t
Φ(t) = dτ V (τ ), (2.25)

so that V (t) = Φ̇. Then the potential energy stored on the capacitor is
1
U= C Φ̇2 (2.26)
2
and now looks like the kinetic energy with this choice of coordinate. Similarly,
using Farady’s law and the sign convention for the direction of the current
Circuit QED 11

defined in Fig. (2.1) we have

V = LI˙ = Φ̇ (2.27)
and thus see that the flux variable Φ really is the physical magnetic flux winding
through the inductor. Hence the kinetic energy stored in the inductor is
1 2
T = Φ , (2.28)
2L
which now looks like the potential energy. With this choice of coordinate the
Lagrangian becomes
1 1 2
L = C Φ̇2 − Φ , (2.29)
2 2L
and the momentum conjugate to the flux
δL
Q= = C Φ̇ (2.30)
δ Φ̇
is now the charge as defined in Fig. (2.1b). Notice the crucial minus sign relative
to the previous result. This is necessary to maintain the sign of the commutation
relation when we interchange the momentum and coordinate.
Just to be completely explicit, we now repeat the derivation of the Hamilto-
nian and its quantization for this new choice which we will be using throughout
the remainder of these notes. Thus the Hamiltonian can be written
1 2 Φ2
H = QΦ̇ − L = Q + . (2.31)
2C 2L
Hamilton’s equations of motion are then
∂H Q
Φ̇ = + =+ (2.32)
∂Q C
∂H Φ
Q̇ = − =− . (2.33)
∂Φ L
Again in the usual way, the coordinate and its conjugate momentum can be
promoted to quantum operators obeying the canonical commutation relation

[Q̂, Φ̂] = −ih̄ (2.34)

and we can write the Hamiltonian

{ }
h̄Ω { † †
} † 1
H= a a + aa = h̄Ω a a + , (2.35)
2 2
in terms of raising and lowering operators
1 1
a = +i √ Q̂ + √ Φ̂ (2.36)
2Ch̄Ω 2Lh̄Ω
1 1
a† = −i √ Q̂ + √ Φ̂ (2.37)
2Ch̄Ω 2Lh̄Ω
12 S.M. Girvin

which obey the usual relation

[a, a† ] = 1. (2.38)
The charge and flux operators can be expressed in terms of the raising and
lowering operators as
( )
Q̂ = −iQZPF a − a† (2.39)
( )
Φ̂ = ΦZPF a + a† , (2.40)

where
√ √
Ch̄Ω h̄
QZPF = = (2.41)
2 2Z
√ √
Lh̄Ω h̄Z
ΦZPF = = , (2.42)
2 2
where Z is the characteristic impedance of the oscillator
√
L
Z= . (2.43)
C
Notice that the notation has been chosen such that the quantum ground state
uncertainties in charge and flux are given by

⟨0|Q̂2 |0⟩ = Q2ZPF (2.44)

⟨0|Φ̂ |0⟩
2
= Φ2ZPF . (2.45)

Using the superconducting resistance quantum

h
RQ ≡ ≈ 6, 453.20 Ohms, (2.46)
(2e)2
we can define a dimensionless characteristic impedance

z ≡ Z/RQ , (2.47)

to obtain
√
1
QZPF = (2e) (2.48a)
4πz
√
z
ΦZPF = Φ0 , (2.48b)
4π
where
h
Φ0 ≡ (2.49)
2e
is the superconducting flux quantum. Notice that the usual uncertainty product
is obeyed.
h̄
QZPF ΦZPF = . (2.50)
2
Circuit QED 13

The voltage is the important physical variable we need to know if we are go-
ing to capacitively couple another system to our oscillator. The voltage operator
is given by

dΦ̂ i
V̂ = = [H, Φ̂]
dt h̄ √
1 h̄Ω ( ) ( )
= Q̂ = −i a − a† = −iVZPF a − a† , (2.51)
C 2C
where √
z
VZPF = ΩΦZPF = ΩΦ0 . (2.52)
4π
The superconducting flux quantum in convenient units is given by

Φ0 ≈ 2.06783367 µV/GHz (2.53)

which tells us that the vacuum fluctuations of the voltage across the capacitor
in a typical 10 GHz, Z = 100 Ohm impedance resonator circuit will be on the
scale of ∼ (1/3)µV.
How do we interpret the excitation quanta of this harmonic oscillator? We
can think of these as excitations of the collective motion of the electrons in the
wire, or we can think of them as photons of the electromagnetic field. Because
this is a lumped element resonator (as opposed to a cavity or other distributed
resonator), the electric field appears between the capacitor plates and the mag-
netic field appears in a separate place, namely within the coil of the inductor.
Nevertheless it is perfectly acceptable to think of these excitations as photons.
The coordinate of the oscillator is the flux in the coil (or in the first choice we
made, the charge on the capacitor plates which is equivalent to the electric field
in the gap between the plates. One does not normally think about photons in
the context of first quantization, but this is also useful for building up intuition
and for thinking about things like the full probability distribution of electric
field measurements. The wave function of the vacuum state is a gaussian in the
coordinate Φ as shown in Fig. (2.2)

1 − 1
Φ2
4Φ2
Ψ0 (Φ) = e ZPF . (2.54)
[2πΦ2ZPF ]1/4

If in the vacuum state we make a precise measurement of the flux the resulting
value will be random and have a gaussian probability distribution given by

P (Φ) = |Ψ0 (Φ)|2 . (2.55)

Hence the most probable value of the flux is zero. On the other hand, in the
one-photon state

Φ 1 − 1
Φ2
4Φ2
Ψ1 (Φ) = e ZPF (2.56)
ΦZPF [2πΦ2ZPF ]1/4
14 S.M. Girvin

zero flux would never be measured because the wave function has a node there.
The measured flux is still zero on average. This is true for any photon Fock
state (number eigenstate) from simple parity consideration. On the other hand
if the photon number is uncertain, for example in the coherent superposition
state
1
Ψ+ = √ (Ψ0 + Ψ1 ) , (2.57)
2
then the centroid of the probability distribution is displaced away from zero
as shown in Fig. (2.2) and the average value of the flux will be non-zero. A
similar conclusion is readily reached within the second quantized formulation of
Eq. (2.40) by noticing that the flux and charge operators are purely off-diagonal
in the photon number basis.

Y  Y¤2

F
F
Figure 2.2: LC oscillator wave function amplitude plotted vs. the coordinate
Φ. Solid: ground state, Ψ0 ; Dashed: first excited state, Ψ1 ; Dotted: linear
combination of the ground and first excited states, √12 (Ψ0 + ψ1 ).

Such superpositions of zero and one-photon states cannot be achieved by

simply weakly driving the oscillator as this produces a coherent superposition
of all photon number states (to be described further below). However they have
been achieved experimentally [5] by applying control pulses to a qubit to put it
into a superposition of the ground state |g⟩ and the excited state |e⟩

|ψinitial ⟩ = α|g⟩ + β|e⟩ (2.58)

and then allowing the qubit to spontaneously decay (if it is excited). This leaves
the qubit in the ground state and the electromagnetic field in a superposition
of zero and one photon with coefficients α and β inherited from the qubit

|ψfinal ⟩ = |g⟩ [α|0⟩ + β|1⟩] . (2.59)

This operation maps a stationary qubit onto a ‘flying qubit’ (the photon) and
is an essential step towards communicating quantum information via photons.
In the experiment of Houck et al. [5] the photons could be sent into a square
law detector to measure the photon number, or into a homodyne detector to
measure either quadrature of the electric field (equivalent to measuring Q̂ or
Circuit QED 15

Φ̂ in Eq. (2.40). The experiment directly showed that the one photon Fock
state had zero electric field on average and that the phase of the electric field
for superposition states was determined by the phase imposed initially upon
the qubit superposition state. We tend to think of spontaneous emission as an
incoherent process but the above results show that this is not entirely correct.
What we really mean by incoherent is that the decay of an atom which starts
purely in the excited state yields a photon state which varies randomly from
shot to shot and which vanishes only on average.

2.3.1 Coherent States

A simple way to achieve a superposition of different number states is to displace
the quantum oscillator with a classical external driving force so that the ground
state is mapped to a so-called ‘coherent state.’
Ψ0 (Φ) −→ Ψ0 (Φ − ∆)). (2.60)
Using the Taylor series expansion to all orders we can write the unitary trans-
formation that displaces the state as

e−∆ ∂Φ Ψ0 (Φ)
∂
Ψ∆ (Φ) = (2.61)
− h̄i ∆Q̂
= e Ψ0 (Φ), (2.62)
which illustrates the fact that the momentum Q̂ is the generator of displace-
ments of its conjugate coordinate Φ. The unitary displacement operator may
be written as †
Uα = e− h̄ ∆Q̂ = e−α(a−a ) ,
i
(2.63)
where the dimensionless displacement parameter is
∆QZPF ∆
α≡ = . (2.64)
h̄ 2ΦZPF
Now using the Feynman disentangling theorem [?] this can be normal ordered
†
Uα = e+αa e−αa e− 2 |α| .
1 2
(2.65)
Taking advantage of the fact that a|0⟩ = 0, we see that in second-quantized
notation the coherent state becomes
†
|α⟩ = e− 2 |α| eαa |0⟩
1 2
(2.66)

Exercise 2.2 Since Uα is unitary, it must be that ⟨α|α⟩ = 1. Verify this by

direct calculation from Eq. (2.66).
Coherent states have some very nice properties. For example, because they
are special coherent superpositions of all possible photon numbers, they are
eigenstates of the photon destruction operator
a|α⟩ = α|α⟩. (2.67)
16 S.M. Girvin

You can destroy a photon and still be in the same state! Curiously coherent
states are not eigenstates of a† . It is clear that a† |α⟩ has no amplitude for
zero photons and hence is linearly independent of |α⟩ (and therefore not an
eigenstate). One can reach the same conclusion by noting that a and a† do not
commute.
[a, a† ]|α⟩ = |α⟩ ̸= 0. (2.68)
On the other hand, it is true that the mean phonon number is given by

N̄ = ⟨α|a† a|α⟩ = |α|2 . (2.69)

The phonon number distribution in a coherent state is given by the standard

Poisson distribution
N̄ n −N̄
Pn = |⟨n|α⟩|2 = e . (2.70)
n!

Exercise 2.3 Derive Eq. (2.67) and Eq. (2.70).

Because a† is a raising operator for the energy, the coherent state has a very
simple time evolution even though it is itself not an energy eigenstate. The dis-
placement parameter α becomes complex and simply varies sinusoidally in time
indicating that the displacement alternates between position and momentum:
−iΩt †
|α(t)⟩ = |e−iΩt α(0)⟩ = e− 2 |α| eαe
1 2
a
|0⟩. (2.71)

This corresponds in the classical limit to the circular motion in phase space of
the simple harmonic oscillator.
Rather than working with Φ̂ and Q̂, we will find it convenient to work with
the dimensionless quadrature amplitudes
1[ ]
X̂ ≡ a + a† (2.72)
2
1[ ]
Ŷ ≡ −i a − a† . (2.73)
2

Like Φ̂ and Q̂, these are canonically conjugate with the following commutator

i
[X̂, Ŷ ] = + (2.74)
2
and for coherent states obey

⟨α|X̂|α⟩ = Real α(t) (2.75)

⟨α|Ŷ |α⟩ = Imag α(t) (2.76)
1
⟨α|[X̂ − ⟨X̂⟩]2 |α⟩ = ⟨0|[X̂]2 |0⟩ = (2.77)
4
1
⟨α|[Ŷ − ⟨Ŷ ⟩]2 |α⟩ = ⟨0|[Ŷ ] |0⟩ = .
2
(2.78)
4
Circuit QED 17

The last two equations show that there are quantum fluctuations in X̂ and
Ŷ (as there must be since they do not commute with each other). The result-
ing uncertainties in the measured values of these quantities play a central in
understanding quantum noise [6]. The energy of the oscillator (in units of h̄Ω)
is
1
ϵ̂ = X̂ 2 + Ŷ 2 = N̂ + , (2.79)
2
so the number operator is simply

1
N̂ = X̂ 2 + Ŷ 2 − . (2.80)
2
To understand the fluctuations
√ in photon number, let us consider a coherent
state with amplitude α = N̄ which is real. As illustrated in Fig. (??), fluctu-
ations in X̂ lead to photon number fluctuations while fluctuations in Ŷ lead to
fluctuations in the phase of the coherent state as measured in homodyne detec-
tion [6]. As we have seen, the coherent state is nothing more than a displaced
vacuum state
|α⟩ = Uα |0⟩. (2.81)

Instead of actively displacing the physical system, we can equivalently leave the
system alone and displace the coordinate system, transforming all operators
according to the usual rule

ã = Uα† aUα = a + α (2.82)

ㆠ= Uα† a† Uα = a† + α. (2.83)

Exercise 2.4 Derive Eqs. (2.82-2.83) by differentiating with respect to α and

solving the resulting differential equation.

Writing the X quadrature amplitude as

X̂ = α + ∆X̂, (2.84)

we see that ∆X̂ has the same statistical properties in the coherent state |α⟩ as
X̂ does in the vacuum state. The number fluctuations are therefore given by
the usual Poisson distribution result derived above
[ ]2
2 2 1
⟨α|[N̂ − N̄ ]2 |α⟩ = ⟨α| 2α∆X̂ + ∆X̂ + ∆Ŷ − |α⟩ = N̄ . (2.85)
2

Essentially the above results mean that a coherent laser or microwave beam
is as classical as possible. The fluctuations come only from the fact that the
photon detection events are discrete and the photons are sprinkled randomly
throughout the beam in an uncorrelated manner. A thermal beam has larger
fluctuations because the photons tend to bunch together [6].
18 S.M. Girvin

Fluctuations in the quadrature orthogonal to α cause uncertainty in a mea-

surement of the phase of the coherent state. For the case of α real and in the
limit |α| ≫ 1, we have
∆Ŷ
∆θ ≈ (2.86)
α
and
1
⟨α|(∆θ)2 |α⟩ = . (2.87)
4N̄
Thus we arrive at the fundamental number-phase uncertainty relation
1
⟨α|(∆θ)2 |α⟩1/2 ⟨α|(∆N̂ )2 |α⟩1/2 ≥ . (2.88)
2
Coherent states are minimum uncertainty gaussian states which satisfy this
relation as an equality. Other non-gaussian states satisfy this relation only as
an inequality.

Exercise 2.5 X̂ 2 and Ŷ 2 are clearly Hermitian operators with non-negative

eigenvalues. How then can you explain the fact that
1
⟨0|X̂ 2 Ŷ 2 |0⟩ = − (2.89)
16
is negative? Similarly how can
i
⟨0|X̂ Ŷ |0⟩ = (2.90)
4
be complex?
From the equation of motion of the free oscillator we see that the quadrature
amplitudes obey

X̂(t) = cos(Ωt)X̂(0) + sin(Ωt)Ŷ (0) (2.91)

Ŷ (t) = cos(Ωt)Ŷ (0) − sin(Ωt)X̂(0) (2.92)

In Appendix 8 we study photons traveling in transmission lines and we again find

that the traveling modes are also harmonic oscillators. The above results provide
the first hint that the sin Ωt and cos Ωt quadratures of a quantum electrical signal
are canonically conjugate and hence cannot be simultaneously measured with
perfect accuracy. Equivalently even the vacuum contains noise which will appear
in any measurement in which one attempts to measure both quadratures of the
signal. Eq. (2.78) tells us that this uncertainty gives a vacuum ‘noise energy’
(noise power per unit measurement bandwidth) of half a photon [?].

2.4 Modes of Transmission Lines Resonators

The above lengthy discussion of the simple harmonic oscillator has laid the
groundwork for our next topic which is the quantum modes of transmission
Circuit QED 19

lines. We will start with finite length transmission lines which have discrete
electromagnetic resonances, each of which will turn out to be an independent
simple harmonic oscillator. Then we will move on to the semi-infinite transmis-
sion line and discover that it can act like a dissipative bath even though every
one of its elements is non-dissipative.
Our finite length transmission line could be a length of ordinary coaxial cable
or its 2D equivalent, the coplanar waveguide which consists of a superconducting
wire evaporated on an insulating substrate and having superconducting ground
planes adjacent to it on the same surface as shown in Fig. (2.4a) Such a system
exhibits many standing wave resonances and we will soon see that each resonance
is an independent harmonic oscillator equivalent to the simple LC oscillator
just discussed. The discretized equivalent circuit for such a system is shown in
Fig. (2.4b). We will assume in this example open boundary conditions for which
the current (but not the voltage) vanishes at the ends of the resonator.

Figure 2.3: a) Micrograph of a typical coplanar waveguide resonator used in

circuit QED. b) Discretized equivalent circuit.

It is convenient to define a flux variable analogous to that used above but

now dependent on position [?]
∫ t
Φ(x, t) ≡ dτ V (x, τ ), (2.93)
−∞

where V (x, t) = ∂t Φ(x, t) is the local voltage on the transmission line at position
x and time t. Each segment of the line of length dx has inductance ℓ dx and the
voltage drop along it is −dx ∂x ∂t Φ(x, t). The flux through this inductance is
thus −dx ∂x Φ(x, t) and the local value of the current is given by the constitutive
equation
1
I(x, t) = − ∂x Φ(x, t). (2.94)
ℓ
The Lagrangian for a system of length L (L is not to be confused with some
discrete inductance)
∫ L ∫ L ( )
c 1
Lg ≡ dx L(x, t) = dx (∂t Φ) −
2 2
(∂x Φ) , (2.95)
0 0 2 2ℓ

The Euler-Lagrange equation for this Lagrangian is simply the wave equation

vp2 ∂x2 Φ − ∂t2 Φ = 0. (2.96)

The momentum conjugate to Φ(x) is simply the charge density

δLg
q(x, t) ≡ = c∂t Φ = cV (x, t) (2.97)
δ∂t Φ
20 S.M. Girvin

and so the Hamiltonian is given by

∫ L { }
1 2 1
H= dx q + (∂x Φ)2 . (2.98)
0 2c 2ℓ

Let us next proceed to consider the classical normal mode solutions of

Eq. (2.96). If we assume a sinusoidal time-dependence with angular frequency
ω,
Φ(x, t) = e−iωt ϕ(x), (2.99)
we arrive at the Schrödinger like eigenvalue problem

−∂x2 ϕ(x) = k 2 ϕ(x), (2.100)

where k = ω/vp and the mode wave velocity is vp = √1ℓc . The open boundary
(zero-current) conditions tell us that the eigenfunctions have vanishing deriva-
tive at the boundaries. We choose a particular normalization for eigenfunctions
which will keep the equations looking as close to those of the single harmonic
oscillator as possible √
ϕn (x) = 2 cos(kn x), (2.101)
where n ϵ {0, 1, 2, 3, . . .}, kn = nπ
L . Because for these boundary conditions the
operator ∂x2 is self-adjoint, and because the eigenvalues are non-degenerate the
eigenfunctions have two helpful properties
∫ L
dx ϕn (x)ϕm (x) = Lδnm (2.102)
0
∫ L
dx [∂x ϕn (x)][∂x ϕm (x)] = Lkn2 δnm . (2.103)
0

From this it follows that the Lagrangian is diagonalized by these (spatial) normal
modes. Let us parameterize the field Φ(x, t) by
∞
∑
Φ(x, t) = ξn (t)ϕn (x), (2.104)
n=0

where the ξn are arbitrary (i.e. not necessarily sinusoidal) functions of time.
Substituting into the Eq. (2.95) and using Eqs. (2.102-2.103)
∞
1 ∑
Lg = Lc [∂t ξn ]2 − ωn2 ξn2 (2.105)
2 n=0

we see that each normal mode becomes an independent simple harmonic oscil-
lator. The momentum conjugate to the normal mode amplitude ξn is

δLg
qn = = Lc∂t ξn , (2.106)
δ∂t ξn
Circuit QED 21

so the Hamiltonian is
∞ { }
1∑ 1 2
H= qn + Lcωn2 ξn2 , (2.107)
2 n=0 Lc

which we can quantize as before. Before doing so, let us note that the n = 0
mode is a ‘free particle’ rather than a harmonic oscillator because its spring
constant vanishes. This mode simple corresponds to a uniform net charge dis-
tributed evenly along the transmission line. For a free particle the momentum
(in this case charge) is a constant and the coordinate (flux) increases linearly
with time. In most situations the total charge is indeed simply a constant of the
motion (and typically vanishes) and we can ignore the zero mode altogether.
We will assume this is the case henceforth.
We end up with a set of independent normal modes with coordinate ξn and
conjugate momentum qn which when quantized can be expressed in terms of
mode raising and lowering operators in a manner analogous to Eq. (2.40)
√
h̄
ξˆn = (an + a†n ) (2.108)
2ωn Lc
√
h̄ωn Lc
q̂n = −i (an − a†n ) (2.109)
2

where the ladder operators of the different modes obey

[an , a†m ] = δnm . (2.110)

If we are coupling a qubit to a resonator at some particular position x, we

need to be able to express the flux and charge density operators at that point
in terms of the normal mode operators. Eq. (2.104) is readily extended to the
quantum operators
∑∞
Φ̂(x) = ϕn (x)ξˆn , (2.111)
n

as is Eq. (2.97)
∞
1∑
q̂(x) = ϕn (x)q̂n . (2.112)
L n

Similarly the analog of Eq. (2.51) for the voltage operator at point x is given by
∞ ∞ √
1 1∑ ∑ h̄ωn
V̂ (x) = q̂(x) = ϕn (x)q̂n = −i (an − a†n )ϕn (x). (2.113)
c L n n
2Lc

The total capacitance to ground of the resonator, Lc, enters this expression
in a way that is similar to lumped element oscillator expression in Eq. (2.51).
(Recall that L is the length of the resonator, not the inductance.)
22 S.M. Girvin

Notice that the flux and charge density operators obey the following com-
mutation relation
∞
′ 1∑
[q̂(x), Φ̂(x )] = −ih̄ ϕn (x)ϕn (x′ ). (2.114)
L n

Using the completeness relation (and recalling that the factor of L appears
because we did not normalize the eigenfunctions to unity) we end up with the
standard field theoretic relation

[q̂(x′ ), Φ̂(x)] = −ih̄δ(x − x′ ). (2.115)

Expressing the quantum Hamiltonian in Eq. (2.98) in terms of these operators,

we have simply { }
∫ L
1 2 1 2
Ĥ = dx q̂ + (∂x Φ̂) . (2.116)
0 2c 2ℓ
As a sanity check let’s look at the Hamilton equations of motion. Using com-
mutation relation in Eq. (2.115) and its extension to

[q̂(x′ ), ∂x Φ̂(x)] = −ih̄∂x δ(x − x′ ). (2.117)

we arrive at
i 1
∂t Φ̂(y) = [Ĥ, Φ̂(y)] = q̂(y) (2.118)
h̄ c
i 1 2
∂t q̂(y) = [Ĥ, q̂(y)] = ∂y Φ̂(y). (2.119)
h̄ ℓ
and hence the quantum version of the wave equation in Eq. (2.96)

vp2 ∂x2 Φ̂(x) − ∂t2 Φ̂(x) = 0. (2.120)

When we studied coherent states of a single oscillator we found that they

were simply the vacuum state displaced in either position (flux) or momentum
(charge). For a multi-mode resonator one can coherently displace a linear com-
bination of the of the normal modes. The familiar problem of solving the time
evolution of a plucked string is a good classical analog. Suppose that we wish to
displace the resonator degrees of freedom so that the local displacement obeys

⟨Φ̂(x)⟩ = ∆(x), (2.121)

where ∆ is some specified function. The analog of Eq. (2.63) is simply

∫L
U∆ = e− h̄
i
dx ∆(x)q̂(x)
0 . (2.122)

Using Eq. (2.112) this can be understood in terms of coherent displacement of

each of the normal modes
∑ ∏ i
U∆ = e− h̄ n ∆n q̂n = e− h̄ ∆n q̂n .
i
(2.123)
n
Circuit QED 23

Exercise 2.6 In analogy with Eqs. (2.82-2.83) show that

†
U∆ Φ̂(y)U∆ = Φ̂(y) + ∆(y). (2.124)

Hint: It may be useful to scale ∆(x) by an overall factor θ and differentiate

with respect to θ.

2.5 ‘Black Box’ Quantization

We have so far studied a single LC oscillator and found that its quantum exci-
tation energy h̄Ω is given directly by its classical frequency
√ Ω. We also found in
Eq. (2.48) that the characteristic impedance Z = L/C determines the size of
the zero-point fluctuations in flux and charge. The typical circuit that we will
study is more complex than a single LC oscillator and might even be a ‘black
box’ whose properties we need to determine. Suppose that we have such a black
box and we have access to one port of this structure as shown in Fig. (2.4a).
The only thing we know (or assume) is that all the elements inside the black
box are linear and purely reactive; i.e., the black box is a network of inductors
and capacitors. It might for example be a transmission line resonator such as
we studied above. We may ultimately want to connect a qubit or some measure-
ment apparatus to the port of the black box. In order to predict the quantum
properties we need to know each of the normal modes of the box and the size
of their zero point fluctuations at the port. Some modes may be localized in-
side the box and have very little connection to the port. Others may be more
strongly coupled to the port.

Figure 2.4: a) One-port black box containing an arbitrary reactive network. b)

Lumped element LC resonator. c) Admittance of the LC resonator vs. frequency.

Since the black box is linear, we can probe it by applying a sinusoidal driving
voltage at the port
V (t) = v(ω)ejωt + v ∗ (ω)e−jωt (2.125)
(where to avoid confusion
√ with the current i we follow the electrical engineering
convention j = − −1) and measuring the resulting current

I(t) = i(ω)ejωt + i∗ (ω)e−jωt . (2.126)

The linear response coefficient that relates these is known as the admittance

i(ω) = Y (ω)v(ω) (2.127)

It turns out that knowing the admittance of the box port as a function of
frequency completely characterizes the classical and the quantum properties of
24 S.M. Girvin

the black box. Because the box contains only reactive elements (assumed finite
in number) the admittance is purely imaginary. For example suppose that the
black box contains a single parallel LC oscillator as shown in Fig. (2.4b). Then
the admittance is simply
1
Y (ω) = jωC + . (2.128)
jωL

Note that this is indeed purely imaginary and further that it passes through
1
zero at the resonance frequency Ω = √LC as shown in Fig. (2.5a). The admit-
tance is zero because the inductor and capacitor have opposite admittances at
the resonance frequency. But this is precisely the condition for self-sustaining
oscillation where the currents in the inductor and capacitor are opposite to each
other and no external input is needed.
Knowing the frequency of our oscillator we can immediately write down the
quantum Hamiltonian (neglecting the zero-point energy)

H0 = h̄Ωa† a. (2.129)

This is not enough however. If we couple an external circuit to our black box
we need to know the matrix elements of the coupling Hamiltonian. For this we
need to know how to express the charge and flux in terms of a and a† and hence
must know the characteristic impedance of the resonance. Happily, the slope
with which the admittance passes through zero determines the characteristic
impedance of the resonance
( ) √
∂Y C
Ω = 2j = 2jΩL, (2.130)
∂ω Ω L

so that
2j
Z= ( ∂Y ) . (2.131)
Ω ∂ω Ω

Using Eqs. (2.41-2.42) we can then find any physical quantity we desire.
To see the generality of this result, consider the example of the lumped
element circuit in Fig. (??). If L1 + L2 = L then this has the same bare
resonance frequency Ω but clearly will have a different coupling to the port.
Use of Eq. (2.42) yields
( )2
h̄ΩL L1
Φ2ZPF = , (2.132)
2 L1 + L2

which is just what we expect from the transformer turns ratio.

Suppose for example that we couple to our black box through an inductor
Lc as shown in Fig. (??a). The coupling Hamiltonian is

1 ( )2
H1 = Φ̂ − Φ̂in . (2.133)
2Lc
Circuit QED 25

The operator Φ̂in is either a classical control field or is a quantum operator for
whatever system we hook up to our black box. Now that we know impedance
of the resonance we now how to express Φ̂ using Eq. (2.42)so that we have

1 ( )2
H1 = ΦZPF (a + a† ) − Φ̂in . (2.134)
2Lc
In a black box with more components there will be a series of such resonances.
According to Foster’s theorem [7] the admittance always passes through zero
with positive slope so therefore each zero must be separated from the next by
a pole as shown in Fig. (2.5b). These poles correspond to effective series LC
resonances. These can be important but for the particular case where nothing is
hooked up to the external port, these poles do not correspond to active degrees
of freedom. (If there is a pole at zero frequency it corresponds to the ‘free-
particle’ Hamiltonian of a capacitor. An inductor and capacitor in series cannot
oscillate on their own at non-zero frequencies.)

Figure 2.5: a) Admittance of a parallel LC resonator vs. frequency. b) Admit-

tance of a black box with multiple resonances.

For the multi-mode case the black box Hamiltonian is simply

∑
H0 = Ωm a†m am , (2.135)
m

where the summation is over the different modes and the flux operator at the
port of the black box is simply
∑ (m) ( )
Φ̂ = ΦZPF am + a†m . (2.136)
m

This ‘black box’ formalism will prove useful if it is possible to either measure
or use finite element simulations to compute the admittance as a function of
frequency. We will return to these ideas later when we model transmon qubits
coupled to multi-mode resonators.
26 S.M. Girvin
Chapter 3

Superconductivity

In these notes I will not touch upon the microscopic theory of superconductivity.
Rather I will present only the minimal phenomenology needed to understand the
Josephson effect in small circuits containing tunnel junctions. A useful overview
of Josephson tunnel junctions and various qubit circuits may be found in the
papers by Devoret and Martinis [8] and Clarke and Wilhelm [9].
An ordinary (normal-state) tunnel junction consists of two metallic elec-
trodes separated by a thin oxide barrier which allows electrons to tunnel quan-
tum mechanically from one electrode to the other. Because even in rather small
(mesoscopic but not nanoscopic) grains, the size of the grain is much larger
than the Angström scale of a typical Fermi wavelength, the ‘particle-in-a-box’
quantum level spacing is extremely tiny and the density of states is essentially
a continuum as shown in Fig. (??a). As a result the tunneling of electrons
is an incoherent process well-described by the irreversible dynamics of Fermi’s
Golden Rule. Under voltage bias V the chemical potential is higher in one grain
than the other by an amount eV . Electrons in this energy interval are able to
tunnel from the ‘cathode’ to ‘anode’ with blocking due to the Pauli exclusion
principle. As a result the tunnel current is linear in the applied voltage (on
voltage scales low compared to the tunnel barrier height) and the junction can
be characterized as a simple resistor. Because the two electrodes are separated
by such a thin barrier they also form a capacitor so the equivalent circuit is
that shown in Fig. (??b) and the incoherent relaxation of the charge through
the junction has the familiar characteristic time scale τ = RC. Obviously this
incoherent behavior is not what we seek in a qubit and so we must turn to
superconductivity to rescue the situation.
Let us begin our discussion of superconductivity by considering a small iso-
lated metallic electrode of some metallic superconductor. Because the electrode
is isolated, the number of electrons in it is fixed and well-defined. For the mo-
ment, let us assume that the electron number has even parity. The essential
physics of an ordinary superconductor is that effective attractive interaction re-
sulting from virtual phonon-exchange leads to pairing of electrons of opposite
spin into so-called Cooper pairs. If the number of electrons in the electrode is

27
28 S.M. Girvin

even, then the quantum ground state has all of the electrons paired up into a
special non-degenerate low-energy ground state. The system has a finite exci-
tation gap, 2∆, defined by the energy needed to break a Cooper pair into two
separate quasiparticles. The scale of this gap is typically several Kelvin. As
illustrated in Fig. (??), the quasiparticle states form a near continuum above
the gap. (Only if the electrode is extremely tiny (on the few nm scale) will
the single-particle level spacing be appreciable. We will not consider this limit
further.)
Recalling that a temperature of 1K corresponds to a frequency of approx-
imately 21 GHz, will be considering the limit of low temperatures and low
frequencies relative to the gap: kB T, h̄ω ≪ 2∆. Hence to a good approximation
we can say that the primary effect of superconductivity is the reduce the vast
Hilbert space of the electrons in the electrode to a single quantum state, |N ⟩,
labeled by the number of pairs which is a constant of the motion. This sim-
plification is very important and will explain how we can produce macroscopic
circuit elements whose quantum energy level spectrum is as simple as that of a
single atom. Obviously however we have overshot the mark because a quantum
system with only one energy level cannot be used to make a two-level qubit. To
repair this error, consider a system with two metallic electrodes connected by
a tunnel junction as shown in Fig. (??). We will again limit our attention to
the quantum ground state of each electrode, assuming the electrons in each are
fully paired up. Once again the total number of electron pairs in the system
is fixed to some value N = NL + NR . Unlike the case of the single electrode,
the total number of pairs no longer uniquely specifies the quantum state. We
must specify the number on each electrode. A useful notation for the set of
low-energy states is:
|m⟩ = |NL − m, NR + m⟩, (3.1)

where m defines the number of pairs that have transferred through the tun-
nel junction from the left electrode to the right electrode starting from some
reference state with pair number NL(R) on the left (right) electrode.
The two electrodes form a small capacitor, but for the moment we will ignore
the Coulomb energy that builds up as Cooper pairs are transferred from one elec-
trode to the other. In this case we then have a one-dimensional family of states
labeled by the integer m, and which are degenerate in energy. Remarkably, it
turns out that the tunnel junction permits pairs of electrons to coherently tun-
nel together from one side to the other. We will discuss the microscopic details
further below, but for now consider the phenomenological Hamiltonian

1 ∑
HT = − EJ {|m⟩⟨m + 1| + |m + 1⟩⟨m|} . (3.2)
2 m

The parameter EJ is called the Josephson coupling energy is a measure of the

ability of Cooper pairs to tunnel through the junction. We see that HT causes
m to either increase or decrease by unity corresponding to the tunneling of a
pair to the right or the left.
Circuit QED 29

We can gain some intuition by noticing that HT is identical to that of a one-

dimensional tight-binding lattice model with near-neighbor hopping amplitude
EJ . The (unnormalized) eigenfunctions are plane-wave like states labeled by a
dimensionless ‘wave vector’ φ

∑
+∞
|φ⟩ = eimφ |m⟩. (3.3)
m=−∞

Recalling the cosine dispersion of the one-dimensional tight-binding band with

near-neighbor hopping we see that

HT |φ⟩ = −EJ cos φ|φ⟩. (3.4)

Imagine a wave packet moving to the right on our tight-binding lattice. This
corresponds to a net current of Cooper pairs coherently tunneling through the
junction. The group velocity of the packet is proportional to the derivative of
the energy with respect to wave vector
1 ∂
vg (φ) = [−EJ cos φ], (3.5)
h̄ ∂φ
so the net current flowing is given by the celebrated (first) Josephson relation
2e
I(φ) = 2e vg (φ) = EJ sin φ. (3.6)
h̄
The maximum possible coherent (dissipationless) current occurs at φ = π/2 and
is called the critical current
2e
Ic = EJ . (3.7)
h̄
If more current than this is forced through the junction, the voltage rises from
zero to a high value above the excitation gap and our low-energy effective model
is no longer applicable.
As an alternative approach to the derivation of the Josephson relation for the
current, let us define the operator n̂ to be the number operator for the Cooper
pairs transferred across the junction
∑
n̂ ≡ |m⟩m⟨m|. (3.8)
m

Hamilton’s equation of motion gives for the current operator

dn̂ i e ∑
Iˆ ≡ 2e = 2e [HT , n̂] = −i EJ {|m⟩⟨m + 1| − |m + 1⟩⟨m|} . (3.9)
dt h̄ h̄ m

Next we simply note that the plane wave energy eigenfunctions are also eigen-
functions of the current operator obeying
ˆ = Ic sin φ|φ⟩
I|φ⟩ (3.10)
30 S.M. Girvin

which is of course equivalent to Eq. (3.6).

Exercise 3.1 Derive Eq. (3.10).

We continue for the moment to ignore the Coulomb interaction, but as a
first step towards including it, let us think about the situation where an external
electric field is applied and maintained in such a way that there is a fixed voltage
drop V across the tunnel junction. This adds to the Hamiltonian a term

U = −(2e)V n̂. (3.11)

Hamilton’s equation of motion yields the celebrated (second) Josephson re-

lation1
∂H
h̄∂t φ = − = 2eV, (3.12)
∂ n̂
relating the time rate of change of the ‘momentum’ h̄φ to the ‘force’ 2eV .
Equivalently, the solution of the Schrödinger equation is
i
∫t
|Ψ(t)⟩ = e+ h̄ EJ 0
dτ cos φ(τ )
|φ(t)⟩, (3.13)

where
2e
φ(t) = φ(0) + V t. (3.14)
h̄

Exercise 3.2 Verify for yourself that this does indeed solve the Schrödinger
equation:
ih̄∂t |Ψ(t)⟩ = (HT + U )|Ψ(t)⟩. (3.15)
The overall phase factor in front of the wave function is not particularly im-
portant, but the time variation of the ‘wave vector’ φ(t) is extremely important
because it leads to the ac Josephson effect, namely that dc voltage bias leads to
an ac current
⟨I(t)⟩
ˆ = Ic sin(φ(0) + ωt) (3.16)
where the ac Josephson frequency is given by
2e
ω = 2π V. (3.17)
h
The inverse flux quantum in frequency units is
2e
≈ 483.597891(12)MHz/µVolt. (3.18)
h
Since frequency is the easiest physical quantity to measure with high accuracy,
the ac Josephson effect finds great practical use in metrology to maintain (but
not define) the volt.
1 As we will discuss in Chap. (4), the fact that the time derivative of the phase variable is

proportional to the voltage means that φ is directly proportional to the flux variable introduced
in Chap. (2).
Circuit QED 31

To summarize our results, we are using here a representation in which the

phase ϕ(t) across the Josephson junction is viewed as a (dimensionless) wave
vector. The Hamiltonian of the junction with voltage bias V is

H = −EJ cos φ − 2eV n̂. (3.19)

The pair of Hamilton equations

∂ n̂ ∂H
= (3.20)
∂t ∂h̄φ
∂H
h̄∂t φ = − (3.21)
∂ n̂
correspond to the pair of Josephson relations found in Eq. (3.6) and Eq. (3.12).
32 S.M. Girvin
Chapter 4

Superconducting Qubits

So far we have studied a single isolated Josephson junction which is able to coher-
ently transfer Cooper pairs from one metallic island to another. Our discussion
of this simple structure (known as the Cooper pair box) has been incomplete
because it has neglected the fact that as current flows through the junction,
charge builds up on the islands and the Coulomb energy becomes important.
Once we include the Coulomb interaction, we will see that this structure makes
an excellent artificial atom which can be used as a superconducting qubit. The
first evidence that Josephson tunneling causes the Cooper pair box to exhibit
coherent superpositions of different charge states was obtained by Bouchiat et
al. [10]. This was followed in 1999 by the pioneering experiment of the NEC
group [3] demonstrating time-domain control of the quantum state of the CPB
using very rapid control pulses to modulate the offset charge.
A number of different qubit designs, illustrated in Fig. (4.1) and Fig. (??)
have been developed around the Josephson junction including the Cooper pair
box [3, 10–16] based on charge, the flux qubit [17–19], and the phase qubit [20,
21]. Devoret and co-workers have recently introduced the fluxonium qubit [22] in
which the small Josephson junction is shunted by a very high inductance created
from a string of larger Josephson junctions. Fig. (4.3) shows an ‘evolutionary
phylogeny’ for these different types of qubits. We will turn now to a discussion
of the Hamiltonians of these different types of qubits and subsequently to an
analysis of the relative merits in terms of their sensitivity to noise perturbations.
Once we understand the Hamiltonians we will be in a position to classify the
qubits according to their location in the ‘Mendeleev Table’ shown in Fig. (4.4).
The lectures by Daniel Esteve will discuss the different methods by which the
state of different qubits can be read out.

4.1 The Cooper Pair Box

The Cooper pair box is topologically distinct from the other designs shown in
Fig. (4.2) in that it has no wire closing the loop around the junction. Hence

33
34 S.M. Girvin

a) b)
supercon- LJ
single ducting
small Al island
JJ n `
‘antenna’
Cg CJ

second island
or ground
250 Pm

Figure 4.1: a) Cooper pair box qubit (R. Schoelkopf lab) and b) its equivalent
circuit.

Figure 4.2: Inductively shunted qubits. a) Phase qubit with flux bias circuit;
b) Fluxonium qubit. Flux bias circuit and capacitances in the long shunting
junction array not shown; c) Flux qubit. Flux bias circuit and capacitances in
the shunting junctions not shown.
Circuit QED 35

Superconducting Qubit
Evolutionary Phylogeny

RF-SQUID COOPER-PAIR-BOX

QUANTRONIUM
QUBIT
PHASE FLUX
QUBIT QUBIT

TRANSMON
QUBIT
FLUXONIUM
QUBIT

Figure 4.3: Evolutionary phylogeny of superconducting qubits. (Courtesy M.

Devoret.)

SUPERCONDUCTING ARTIFICIAL ATOMS

"MENDELEEV" TABLE
0 1 EL / E J
Cooper Pair Box inverse of
1 number of wells
Quantronium Fluxonium in potential
10
Transmon
100 Flux Qubit

1000

10000
Phase qubit
100000
charge fluctuations
relative to phase EJ / EC
fluctuations

Figure 4.4: ‘Periodic Table’ of superconducting qubits. EJ is the tunneling

Josephson energy, 4EC is the energy cost to charge the junction with one Cooper
pair, and EL /2 is the energy cost to ‘charge’ the shunt inductor with one flux
quantum. (Courtesy M. Devoret.)
36 S.M. Girvin

the number of Cooper pairs transferred through the junction is a well-defined

integer as we have already discussed in detail. The integer charge implies the
conjugate phase is compact; that is, in the phase representation, the system
obeys periodic boundary conditions. As we will see below, this implies that
charge-based qubits are sensitive to stray electric field noise, but that this can
be overcome by putting the Cooper pair box in the ‘transmon’ regime where
the Josephson tunneling energy dominates over the Coulomb charging energy.
Our previous discussion of the Josephson effect has neglected the Coulomb
interaction. Let us define the charging energy associated with the transfer of a
single electron to be
e2
EC = , (4.1)
2C
where C is the total capacitance between the island electrodes. The Coulomb
energy to transfer a Cooper pair is four times larger and so the Coulomb energy
operator is given by
Û = 4EC (n̂ − ng )2 , (4.2)
where ng is called the ‘gate charge’ or ‘offset charge’ and represents either the ef-
fect of an externally applied electric field or some microscopic junction asymme-
try which breaks the degeneracy between positive and negative charge transfers.
The number operator n̂ has integer eigenvalues, while ng is a continuous vari-
able, which may intentionally or randomly and uncontrollably fluctuate. The
effects of such fluctuations on the coherence of the qubit will be discussed in
Chapter (5).
From Eq. (3.3) we see that φ is a compact variable; that is, the state |φ+2π⟩
is identified with the state |φ⟩. Hence without loss of generality we may take the
wave vector φ to live in the first Brillouin zone φ ϵ [−π, +π], or equivalently on
a circle. In fact, because of this periodic boundary condition, it is much more
appropriate to view φ as an angular coordinate living on a circle, rather than
as a momentum as we have been doing until now. Notice further that
∂
n̂|φ⟩ = −i |φ⟩. (4.3)
∂φ
Thus we see that h̄n̂ is nothing more than the (dimensionless) angular momen-
tum conjugate to the angle φ. Because φ is compact, the conjugate (dimension-
less) angular momentum n̂ naturally has integer eigenvalues which is just the
property that we require for the operator representing the number of Cooper
pairs which have been transferred through the Josephson junction. Except for
the ng term (to which we will return shortly) the Hamiltonian becomes that of
a quantum rotor in a gravitational field (see Fig. (4.5)
2
H = 4EC (n̂ − ng ) − EJ cos φ, (4.4)
where the charging energy determines the (inverse) moment of inertia, and the
Josephson energy is a measure of the torque produced by gravity.
We know that for small amplitude oscillations the classical quantum rotor
is very nearly a simple harmonic oscillator. This can be seen by expanding the
Circuit QED 37

Figure 4.5: The Cooper pair box Hamiltonian in the phase representation is
equivalent to that of a quantum rotor. The offset charge ng is equivalent to an
Aharonov-Bohm flux which produces a Berry phase proportional to the winding
number of the rotor trajectory. Unlike other qubit circuit topologies, the rotor
wave function obeys periodic boundary conditions. From [15].

cosine to second order, an approximation that is valid in the limit EJ ≫ EC

where the zero-point fluctuations in the phase are small. Up to an irrelevant
constant in the energy (and ignoring the offset charge for the moment) we have
1
H ≈ 4EC n̂2 + EJ φ2 . (4.5)
2
It is interesting therefore to try connect the Hamiltonian of this Cooper pair box
to the quantum LC oscillator we studied previously. From the second Josephson
relation in Eq. (3.12) we find that the phase angle φ is directly proportional to
the flux variable Φ defined in Eq. (2.25) used in the quantization of the LC
oscillator.
2e Φ
φ = Φ = 2π . (4.6)
h̄ Φ0
Thus each time the flux variable changes by one flux quantum, the supercon-
ducting phase variable winds by 2π. The classical Lagrangian for the Cooper
pair box can be written
( )
1 Φ
L = C Φ̇ + EJ cos 2π
2
, (4.7)
2 Φ0
and the Hamiltonian becomes
( )
1 2 Φ
H= Q − EJ cos 2π . (4.8)
2C Φ0

Expanding the cosine term about Φ = 0 to lowest order (and dropping the
zeroth order term) we have
1 2 1 2
H≈ Q + Φ , (4.9)
2C 2LJ
38 S.M. Girvin

where the (small signal) effective inductance of the Josephson junction is given
by
( )2
h̄ 1
LJ = . (4.10)
2e EJ
In this approximation, the CPB becomes a simple harmonic oscillator with
resonant frequency (known as the Josephson plasma frequency) given by
1 1√
ΩJ ≡ √ = 8EJ EC . (4.11)
LJ C h̄
The Taylor series expansion of the cosine is justified only if we are discussing
small amplitude motions. Classically we can always choose to study this limit.
Quantum mechanically, we are forced to deal even in the ground state with
zero-point fluctuations in Φ. From Eq. (2.42) we see that (in the harmonic
approximation used above)
( )1/4
ΦZPF EC
2π = 2 , (4.12)
Φ0 EJ
so the harmonic approximation is self-consistent in the limit EJ ≫ EC . In
the quantum rotor picture, this corresponds to strong gravity and large mass
(moment of inertia).
For general flux Φ, not necessarily small, we can define the differential (in-
verse) inductance as
( )2 ( )
−1 d2 H 2π Φ
L (Φ) ≡ = EJ cos 2π , (4.13)
dΦ2 Φ0 Φ0
we see that the Josephson junction acts as a non-linear inductor. It is this key
feature which will make the energy levels of the Cooper pair box anharmonic.
If the quantum fluctuations in Φ are small, then the above non-linear in-
ductor picture can be useful, but in general we need to resort to numerical
diagonalization of the CPB Hamiltonian. Let us therefore now return to the
full Hamiltonian in Eq. (4.4). In the phase basis, the Schrödinger eigenvalue
equation is the Matthieu equation whose solutions are formally known in terms
of Matthieu functions. Numerical diagonalization is more conveniently per-
formed in the charge (number) basis where the Coulomb term is diagonal and
the Josephson term is tri-diagonal: ⟨n ± 1| cos φ|n⟩ = 12 . The basis states are
labeled by the eigenvalue m of the number operator n̂ and the Hilbert space
must be truncated at some largest |m| = mmax . If we are interested only in
first N low-lying states, the size of the Hilbert space needed can be estimated
from the zero-point fluctuations of the charge in the harmonic limit given in
Eq. (2.41)
(√ )1/4
√ QZPF √ EJ
mmax ≫ N ∼ N (4.14)
2e 32EC
which, conveniently, is never very large.
Circuit QED 39

The qubit spectrum is periodic in the offset charge ng with unit period as
can be seen in figure 4.6. Physically this simply means that the integer part of
the offset charge can allows by canceled out by transferring an integer number of
Cooper pairs from one island to the other. To understand this mathematically,
recall that the wave functions obey periodic boundary conditions in the angle
φ. Thus
U± = e±iφ (4.15)
is a ‘legal’ unitary transformation which preserves the boundary conditions (and
physically transfers one pair through the junction). Recognizing that such a
unitary transformation preserves the spectrum of the Hamiltonian and that it
shifts the angular momentum (transferred charge) by one unit

U n̂U † = n̂ ∓ 1, (4.16)

we see that the spectrum must be indeed be invariant under unit translations
of ng .

Figure 4.6: Energy spectrum of the Cooper pair box as a function of offset
charge for different values of the dimensionless ratio of Josephson energy to
charging energy. The exponential decrease in the charge dispersion is clearly
seen. From [15].

Since the offset term ng does not appear in the ordinary quantum rotor
problem we need to extend our analogy a bit. It turns out that this term can
be viewed as resulting from the rotor carrying (fake) charge and undergoing an
Aharanov-Bohm phase shift as it circles a line of (fake) magnetic flux which is
passing through the axis. To see this, let us recall that for a particle with charge
q moving in the presence of a vector potential A, ⃗ the canonical momentum is
40 S.M. Girvin

replaced by the mechanical momentum

p⃗ −→ p⃗ − q A(⃗
⃗ r). (4.17)

For our quantum rotor turning on the z axis, we are interested in the angular
momentum
Lz = (⃗r × p⃗)z −→ (⃗r × p⃗)z − q(⃗r × A)
⃗ z. (4.18)
If the magnetic field is zero everywhere except for an Aharanov-Bohm tube of
flux on the z axis, we can choose the following gauge for the vector potential

⃗ r) = ΦAB 1 ẑ × r̂,
A(⃗ (4.19)
2πr
which has the correct total flux
I
⃗ · d⃗r = ΦAB
A (4.20)

for any loop with winding number +1 around the z axis. The mechanical angular
momentum operator thus becomes
( )
q ⃗ z = h̄ −i ∂ − ΦAB
(⃗r × p⃗)z − ΦAB (⃗r × A) (4.21)
2π ∂φ Φ0

where Φ0 is the flux quantum corresponding to charge q (which since it is arbi-

trary since we can choose to be 2e). By comparison with Eq. (4.4), we see that
the real offset charge is equivalent to a fake Aharonov-Bohm flux
ΦAB
ng = . (4.22)
Φ0
The fact that the spectrum is periodic in ΦAB is simply a reflection of the
fact that a flux tube containing an integer number of flux quanta produces an
Aharanov-Bohm phase shift
In the ‘charge’ limit, EC ≫ EJ , the states of our rotator and nearly angu-
lar momentum (i.e. charge) eigenstates weakly perturbed by the gravitational
torque (Josephson coupling). As can be seen in Fig. (4.6a) when the offset
charge is half-integer, there are two adjacent charge states which have the same
Coulomb energy. This degeneracy is lifted by by the Josephson term. The
disadvantage of working in this regime is that EC is large and the excitation
energies are extremely strong functions of the offset charge ng . As we will see
in Chap. (5) environmental noise in ng will lead to severe qubit decoherence.
In the ‘transmon’ limit, EJ ≫ EC , the states of the quantum rotor are
nearly harmonic small-amplitude oscillations near the minimum in gravitational
potential energy (i.e. Josephson energy). As can be seen in Fig. (4.6d) the
qubit levels are nearly harmonic and very nearly completely insensitive to the
value of the offset charge. The reason for this is readily understood within the
Aharanov-Bohm analogy of the offset charge. Let us start our discussion with
a path integral language. We know that the only way that the system can be
Circuit QED 41

aware of the value of the Aharanov-Bohm flux is for the rotor to circle completely
around the flux tube in order to acquire the Aharnov-Bohm phase shift. The
interference between this path and the path where the rotor does not circle the
flux modifies the quantum energy. However in the limit of large EJ (strong
gravity), the rotor must tunnel through a very high energy barrier in order to
wind the phase from φ = 0 to φ = 2π. The barrier height is proportional
to EJ and the particle ‘mass’ is inversely proportional to EC . As a result the
contribution of these processes to the energy is exponentially small.
Let us return now to the language of wave functions for a more quanti-
tative discussion. Mathematically we being by performing the unitary gauge
transformation
U = e−ing φ , (4.23)
which completely removes the offset charge term from the Hamiltonian

U (n̂ − ng )U † = n̂. (4.24)

Notice that the transformed wave function U Ψ no longer obeys periodic bound-
ary conditions:
{U Ψ(φ + 2π)} = e−i2πng {U Ψ(φ)}. (4.25)
Thus, while the Hamiltonian becomes independent of ng

U HU † = H = 4EC (n̂)2 − EJ cos φ, (4.26)

the change in boundary condition with ng nevertheless changes the energy eigen-
value spectrum. On the other hand, for large EJ /EC the wave function is ex-
ponentially small at the boundary φ = ±π, so we do not expect a large change
in the spectrum due to this change in boundary condition.
To develop a better intuition for how the change in boundary condition
affects the energy eigenvalues, consider the following. Rather than viewing the
Hamiltonian as that of a quantum rotor (with compact φ living on the interval
−π ≤ φ + π) let us view this as the Hamiltonian of a phase ‘particle’ moving in
the extended cosine potential as shown in Fig. (4.7). Bloch’s theorem tells us
that the discrete translation symmetry of the potential implies that the wave
functions must be of the form

Ψnk (φ) = eikφ ψn (φ), (4.27)

where ψn obeys periodic boundary conditions φ −→ φ + 2π, n is the band index

and (because the ‘lattice constant’ of the potential is 2π) the wave vector k lives
in the first Brillouin zone − 12 ≤ k ≤ + 12 . We now simply invoke the boundary
condition in Eq. (4.25) by choosing the Bloch wave vector to be k = ng . The
use of the extended cosine potential is just a crutch to get us to Eq. (4.27).
The Bloch wave solution in the interval −π ≤ φ ≤ +φ solves the quantum
rotor problem with compact phase and boundary condition given in Eq. (4.25).
Thus we arrive at the important conclusion that the spectrum of the quantum
rotor plotted against offset charge in Fig. (4.6) is nothing more than the band
42 S.M. Girvin

Figure 4.7: Extended cosine potential U = −EJ cos(φ).

structure of a one-dimensional solid with a cosine potential in which the offset

charge defines the Bloch wavevector.
The band structure picture is useful because in the limit of large EJ we can
invoke (at least for the low-lying bands) the tight-binding approximation to the
band structure in which we envision the phase particle in a bound state within a
well being able to hop to the corresponding state an adjacent well by tunneling
through the barrier. We know that for near-neighbor hopping the tight-binding
model has a simple cosine dispersion as a function of wave vector (which in the
case is offset charge). Hence within this approximation the energy of the nth
level has the form [15]

Em (ng ) ≈ Em + ϵm cos(2πng ), (4.28)

where the tight-binding bandwidth defines the so-called ‘charge dispersion’.

Within the WKB approximation for the tight-binding hopping amplitude the
charge dispersion for the mth level is given by [15]
√ ( ) m2 + 43 √
24m+5 2 EJ
ϵm ∼ (−1) EC
m
e− 8EJ /EC
. (4.29)
m! π 2EC

The charge dispersion is a measure of the sensitivity of the energy to the offset
charge. We see that going to large EJ /EC makes the transmon qubit quite
insensitive to (low-frequency) charge noise which dramatically improves the de-
phasing time. The fact that the ‘charge dispersion’ is greater for the higher
excited states is simply the naturally larger bandwidth of the higher energy
states of the band structure. (More energetic particles tunnel through the bar-
rier more readily.)
In the limit of large EJ /EC the quantum rotor begins to approach a har-
monic oscillator. Fortunately the anharmonicity defined by

A ≡ ω12 − ω01 ∼ −EC (4.30)

goes to zero very slowly as the charging energy is reduced and can be easily
kept above 100-200MHz [15] which is adequate to prevent smooth nano-second
Circuit QED 43

control pulses from taking the qubit out of the logical subspace (the two lowest
levels) [?].
We may perturbatively estimate the anharmonicity (in the limit of negligible
charge dispersion) from Eq. (4.26) by assuming that the zero-point fluctuations
of the phase are small allowing us to expand the cosine potential to write

H ≈ H0 + V, (4.31)

where
1
H0 = 4EC n̂2 + EJ φ2 , (4.32)
2
and
1
V =− E J φ4 . (4.33)
24
Using Eq. (4.12) we find √
2EC
φ2ZPF = , (4.34)
EJ
which as noted earlier in connection with Eq. (4.12) is indeed small for large
EJ /EC, so our assumption is self-consistent in this limit.
Using
φ = φZPF (a + a† ), (4.35)
and neglecting off-diagonal terms we can write the perturbation in a simple form
useful for first order perturbation theory
1 EC † †
V =− EC (a + a† )4 ≈ − (a a aa + 2a† a). (4.36)
12 2
The second term renormalizes the harmonic oscillator frequency downward slightly
√
h̄Ω̃J = 8EJ EC − EC (4.37)

and the first introduces an anharmonicity which yields Eq. (4.30).

4.2 Inductively Shunted Qubits

We turn now to a the class of superconducting qubits illustrated in Fig. (4.2)
which are inductively shunted. In the phase qubit [?] the Josephson junction
is shunted by a lumped element inductor. In the fluxonium qubit [22], an
extremely large value of inductance is required, so large that it is impossible to
achieve with a coiled wire. This is due to parasitic capacitance in any coil and
is essentially the result of the small value of the fine structure constant. Thus
instead of a coil, a long chain of Josephson junctions provides the inductance.
The flux qubit [?] uses only two junctions in this chain and hence it turns out
that the shunting inductor is not fully linear.
As stated several times previously, the Cooper pair box is unique in its topol-
ogy because there is no connection between the two superconducting islands.
44 S.M. Girvin

Thus the only way to change the charge is to tunnel an integer number of Cooper
pairs through the Josephson junction. The number of Cooper pairs transferred
is represented in the Hamiltonian as an angular momentum operator n̂ which is
conjugate to a compact phase angle φ. We see from Eq. (3.3) that the state of
the Cooper pair box |φ+2π⟩ is not distinct from (and indeed is identical to) |φ⟩.
This is what we mean when we say that φ is a compact angular variable living
on a circle, or equivalently the wave function Ψ(φ) obeys periodic boundary
conditions. One obvious consequence is that the current flowing through the
junction I = Ic sin(φ) is (from the first Josephson relation) a periodic function
of the phase variable.
On the other hand, we know from the second Josephson relation that the
phase variable φ is equivalent to the flux variable Φ which we introduced in
quantizing the LC oscillator where we found that the energy stored in the in-
ductor is
1 2 1
U= Φ = E L φ2 (4.38)
2L 2
where ( )2 ( )2
h̄ 1 Φ0 1
EL ≡ = . (4.39)
2e L 2π L
Clearly the energy U stored in the inductor and the current flowing through it

dU Φ0
I= = φ (4.40)
dΦ 2πL
are not periodic in φ. If we shunt our Josephson junction with an external
inductor, the change in topology of the circuit has profound consequences on
the mathematical description. The Hamiltonian becomes
1
H = 4EC (n̂ − ng )2 − EJ cos φ + EL φ2 . (4.41)
2
Equally important however is the fact that charge can move onto the junction
capacitor plates continuously through the inductor so the charge variable is no
longer integer-valued but rather continuous, as expected from the fact that the
system no longer obeys periodic boundary conditions–φ is now an extended
variable and we expect vanishing boundary conditions Ψ(φ −→ ±∞) −→ 0.
Because the charge is now a continuous variable we expect on physical
grounds that a static offset charge ng can be completely screened and should not
affect the energy. Mathematically this can be seen by performing the unitary
gauge transformation discussed previously in Eq. (4.23). Notice however the im-
portant difference that the transformed wave function U Ψ still obeys the same
vanishing boundary conditions. Hence unlike the previous case, the spectrum
does not depend on the static offset charge in any way.
Let us consider the case of a shunt inductor but take the limit L −→ ∞
which means EL −→ 0. Since the inductor is present, the phase variable is
presumably no longer compact and yet it would seem that the inductance term
does not change the Hamiltonian. Physically, it seems reasonable to assume
Circuit QED 45

that the high frequency oscillations of the qubit would be unaffected by the
enormous reactance of the inductor. This is indeed the case and the spectrum
is the same as before except that now all values of the wave vector k are allowed
since the phase is non-compact. Thus there is a continuum of states instead
of a single state within each band. The only effect that offset charge has is to
shift the k states k −→ k + ng , but this has no effect on the spectrum because
all values of k are allowed and (in the extended zone scheme) the spectrum is
periodic under k −→ k + 1.
Exercise 4.1 In the limit of L −→ ∞ described above, the eigenstates are
plane-wave-like Bloch waves. Using the general periodicity properties of Bloch
waves, compute the mean charge (which can be non-zero only because of the
infinite inductance)
n̄ = ⟨Ψnk |n̂|Ψnk ⟩. (4.42)
Strictly speaking, the wave vector k is not gauge invariant. What is the correct
gauge invariant operator we should have used in this equation?
If the inductive energy EL is non-zero then Bloch’s theorem no longer ap-
plies. The quadratic potential resulting from the inductive energy leads to the
curvature illustrated in Fig. (4.8). The interplay between the quadratic term
and the Josephson cosine term allows us to create a number of different po-
tential energy well shapes and thus generate different interesting qubit spectra.
Before delving into this we need to recognize that there is one more ‘control
knob’ at our disposal, namely externally applied flux which we can view as the
inductive analog of the offset charge studied previously. If our inductor is part
of a transformer with dc current applied in the other winding then there is a
flux offset and the Hamiltonian becomes
1
H = 4EC (n̂ − ng )2 − EJ cos φ + EL (φ − φg )2 . (4.43)
2
It is convenient to translate φ and n̂ using the unitary transformation

U = eiφg n̂ e−ing φ (4.44)

which yields

U φU † = φ + φg (4.45)
U n̂U † = n̂ + ng . (4.46)

The Hamiltonian then becomes

1
H = 4EC (n̂)2 − EJ cos(φ + φg ) + EL (φ)2 . (4.47)
2
As with the case of offset charge, we see that the resulting spectrum must be
periodic in the offset flux.
The ability of control the three energy scales EJ , EC , EL as well as the offset
flux φg gives the experimentalist the opportunity to create a rich variety of level
46 S.M. Girvin

Figure 4.8: a) Extended cosine potential U = −EJ cos(φ + φg ) + EL φ2 with

EL = 0.02EJ , φg = 0. Used in the fluxonium qubit to produce a ‘∆’ level
configuration; b) Same as (a) but with φg = π. Used in the flux and fluxonium
qubits to produce a ‘Λ’ level configuration; c) Same as (a) but displaced a
distance φ = 6π. Used in the current biased phase qubit.
Circuit QED 47

structures in the inductance-shunted family of qubits. The phase qubit [?] is

shunted by a lumped element inductor and typically operates with EJ /EC ∼ 104
which makes it very nearly a harmonic oscillator. To increase the anharmonic-
ity it is operated with large flux bias on the inductor which drives a steady
state current close to the critical current through the junction. In addition to
greatly increasing the anharmonicity, this flux bias also lowers the energy level
spacing so the device is typically constructed
√ with large area junctions having a
Josephson plasma frequency ΩJ /2π = 8EJ EC /h which is very high1 , ∼ XXX
GHz.

1 If the capacitance is dominated by the junction itself, then E E is independent of junc-

J C
tion area, but when external and parasitic capacitance is included, then the plasma frequency
rises with junction area.
48 S.M. Girvin
Chapter 5

Noise induced Decoherence

in Qubit Circuits

Discuss Taylor series expansion of dependence of qubit H on various param-

eters, quantronium and CPB flux and charge sweet spots. Discuss dephasing
by cavity photons via light shift. Quasiparticles, environmental impedance,
etc.

The ideal qubit would have a completely stable transition frequency un-
perturbed by the external environment and yet would have a port open to the
outside world through which its state could be controlled. Unfortunately, things
are never ideal and it is our job as quantum engineers to design qubits for the
real world. Unlike the Cs and Rb atoms used in atomic clocks which are (lit-
erally) indistinguishable, each qubit we build is a unique individual. The good
news is that we can engineer different classes of Hamiltonians and qubit Hamil-
tonians with different parameters within the same class. The bad news is that
in reality, if we can vary the Hamiltonian parameters, they will tend to vary
on their own due to various random sources of noise. This variation of param-
eters has two effects. First it can modulate transition frequencies which leads
to dephasing of superpositions. Second, if the noise (including as we will see,
vacuum noise) is high-frequency it can cause transitions to occur which change
the state of the qubit. The phenomenological Bloch equation from NMR is a
useful starting point to understand the standard parametrization of qubit re-
laxation rates. In NMR one typically studies the dynamics of a large collection
of spin-1/2 (say) nuclei in a sample by coupling a drive/readout coil to the total
magnetic moment. In the absence of any perturbations and assuming that all
the spins have the same Zeeman energy (no ‘inhomogeneous broadening’), the
Hamiltonian is simply
Ω∑ z
N
H= σ . (5.1)
2 j=1 j

49
50 S.M. Girvin

The component of the magnetization parallel to the Zeeman field commutes

with the Hamiltonian and so is a constant of the motion.

1 ∑ z
N
mz ≡ ⟨σ ⟩. (5.2)
N j=1 j

The transverse components of the magnetization precess around the magnetic

field direction. If we go to a frame rotating at frequency Ω, these transverse
components of the magnetization are also constants of the motion. Inhomo-
geneous broadening and various fluctuating perturbations will complicate this
simple picture and cause the net magnetization to relax to its equilibrium value.
This relaxation is characterized in the phenomenological Bloch equations with
a longitudinal relaxation time, T1 , and a transverse relaxation time, T2 .
1
ṁz = − (mz − m̄z ) (5.3)
T1
1
ṁx = − mx (5.4)
T2
1
ṁy = − my , (5.5)
T2
The longitudinal relaxation rate can be shown to be given by the sum of the
transition rates from down to up and up to down
1
= Γ↑ + Γ ↓ . (5.6)
T1
The equilibrium magnetization is given by
Γ↑ − Γ↓
m̄z = . (5.7)
Γ ↑ + Γ↓

The transverse relaxation rate can be shown to be given by

1 1 1
= + . (5.8)
T2 2T1 Tφ

The first term represents the fact that if the spin makes a transition from up to
down (or vice versa) the coherent superposition of up and down is destroyed.
The second term containing T1φ represents the rate at which the transverse
magnetization decays due to random fluctuations in the spin precession rate
which cause the different spins to get out of phase with each other and thus
destroy the mean transverse polarization.
Exercise 5.1 Derive Eqs. (5.6) and (5.7).
We turn now to the microscopic noise which leads to the above phenomenol-
ogy. Suppose that the qubit Hamiltonian contains a set of n parameters λj ;
jϵ{1, ..., n}. Let us take the value of these parameters at the working point to
Circuit QED 51

be λ̄j ; jϵ{1, ..., n}, and let the deviation from the nominal working point do to
noise in the parameter values be ηj ; jϵ{1, ..., n}. The qubit Hamiltonian can
then be expressed in a Taylor series expansion. Working for the moment only
to first order we have
∑n
⃗
H[λ] = H0 + ηj Vj , (5.9)
j=1

where H0 ≡ H[⃗λ̄] and 

∂H 
Vj ≡ . (5.10)
∂λj [⃗λ̄]

The noise amplitudes ηj could represent classical random variables or quantum

operators associated with the bath to which the qubit is coupled. If Vj com-
mutes with H0 then it is diagonal in the energy eigenbasis and cannot affect the
eigenfunctions, it can only modulate the eigenvalues. Such perturbations can
dephase superpositions but cannot cause transitions between levels and hence
cannot relax the population. On the other hand if [Vj , H0 ] ̸= 0, then Vj is not
diagonal in the energy basis and can cause both dephasing and relaxation.
As a simple example, consider the Hamiltonian of a two-level system

Ω z
H= σ + ⃗η (t) · ⃗σ . (5.11)
2
Let us first ignore the transverse field fluctuations and focus on the ηz longi-
tudinal term which commutes with H0 . For simplicity we will take ηz to be a
classical random variable. The exact time evolution operator in the interaction
representation is
U (t1 , t2 ) = e− 2 θ(t1 ,t2 )σ
i z
(5.12)
where the random phase accumulated due to the longitudinal noise is given by
∫ t2
θ(t1 , t2 ) = 2 dτ ηz (τ )⟩. (5.13)
t1

Assuming ηz has zero mean then θ also has zero mean and variance given by
∫ t2
⟨θ ⟩ = 4
2
dτ1 dτ2 ⟨ηz (τ1 )ηz (τ2 )⟩. (5.14)
t1

If we assume that ηz itself is gaussian distributed then θ will be gaussian. Or

if ηz is not gaussian but its autocorrelation function appearing in Eq. (5.14) is
short-ranged in time compared to t2 − t1 , then (under rather mild assumptions)
the central limit theorem applies because θ is the sum of a large number of
independent random variables. If θ is gaussian distributed with mean zero,
then we have the exact result

⟨⟨eiθ ⟩⟩ = e− 2 ⟨⟨θ ⟩⟩
1 2
(5.15)
52 S.M. Girvin

where ⟨⟨⟩⟩ refers to the statistical ensemble average over the noise. If the auto-
correlation time of the noise is short we can write Eq. (5.14) as

⟨⟨θ2 ⟩⟩ ≈ 4tSηz ηz (0), (5.16)

where Sηz ηz (ω) is the noise spectral density at frequency ω. The full expressions
for the case where the autocorrelation time of the noise is not negligible are given
in [?]. Because it is only low frequencies that count here, it is generally safe to
treat ηz classically and ignore that fact that it typically represents a quantum
operator describing a bath variable. [This will not be true for the transverse
noise fluctuations where the high-frequency spectral density is important.]
Let us define the (pure) dephasing rate 1/Tφ via
− Ttφ
⟨⟨eiθ ⟩⟩ = e , (5.17)

from which we see that

1
= 2Sηz ηz (0). (5.18)
Tφ
It turns out that the important fluctuations of the transverse fields are the
ones near the transition frequency of the qubit because this is necessary for
energy conservation in the transitions between qubit states. For such high-
frequency fluctuations we should more properly use a full quantum theory which
involves the quantum density matrix of the qubit. In the presence of quantum
or classical noise, we desire the ensemble average over the noise of any given
physical observable. Let pj be the probability that an element of the ensemble
is in state |ψj ⟩ (or equivalently that the random noise has driven our system to
state |ψj ⟩). The expectation value of observable O is
∑
⟨⟨O⟩⟩ = pj ⟨ψj |O|ψj ⟩ = Tr ρO, (5.19)
j

where the density matrix is defined by

∑
ρ= pj |ψj ⟩⟨ψj |. (5.20)
j

Clearly the density matrix must satisfy

Tr ρ = 1 (5.21)
Tr ρ 2
≤ 1, (5.22)

the latter satisfied as an equality only for a pure state. In fact for a pure state,
the density matrix is idempotent

ρ2 = ρ. (5.23)

The von Neumann entropy is given by an expression closely analogous to the

classical one
S = −Tr {ρ log ρ} (5.24)
Circuit QED 53

Exercise 5.2 Prove Eq. (5.22) and Eq. (5.23).

The density matrix contains all the information we need to compute the
expectation value of any observable, or any moment of the distribution of some
observable, for example
⟨⟨Om ⟩⟩ = Tr ρOm (5.25)
or even the full probability distribution for the measurement results for the
observable
P (λ) = Tr {ρδ(λ − O)} . (5.26)
Since a two-level qubit or a spin-1/2 particle has only two independent quan-
tum states, the density matrix is 2 × 2. The most general such matrix which is
hermitian and has unit trace can be written
1
ρ= ⃗ · ⃗σ ) ,
(1 + m (5.27)
2
where the qubit polarization is given by

mj = Tr ρσj . (5.28)

Exercise 5.3 Use Eq. (5.27) to prove Eq. (5.28).

Having now established these statistical results we can proceed to ensemble

average the density matrix over the noise to obtain the reduced density matrix
where the noise bath has been integrated out. In the interaction representation

⟨⟨ρ(t2 )⟩⟩ = ⟨⟨U (t1 , t2 )ρ(t1 )U † (t1 , t2 )⟩⟩, (5.29)

where we now take ρ to be the full density matrix for system plus bath and
the interaction representation of the time evolution operator is given by the
time-ordered exponential of the perturbation
{ ∫ }
i t2
U (t1 , t2 ) = Tτ exp − dτ V̂ (τ ) (5.30)
h̄ t1

Notice that we cannot average the two time evolution operators separately be-
cause they are correlated–they depend on the same noise. It is convenient to do
a perturbative calculation which will be valid for short times t2 = t1 + ∆t. We
assume the perturbation is weak enough that our perturbative expansion is in
fact valid for times longer than the autocorrelation time of the noise, τc . In that
case we can safely choose a a time step ∆t ≫ τc and the random noise variables
in U (t1 , t2 ) will be essentially uncorrelated with the past noise variables that led
to ρ(t1 ). This statistical independence will allow us to make the ‘no memory’
or Markov approximation [?] which will considerably simplify matters Again,
we can only justify this for sufficiently weak noise that low-order perturbation
theory is still valid for the chosen value of ∆t.
54 S.M. Girvin

By definition we choose ⟨⟨V̂ (τ )⟩⟩ = 0 so we must (at least in the Markov

approximation) take our expansion to second order (just as we do in Fermi’s
Golden Rule for transition rates)
∫ t2 ∫ t2
i 1
U (t1 , t2 ) ≈ 1 − dτ V̂ (τ ) − dτ dτ ′ Tτ V̂ (τ )V̂ (τ ′ ). (5.31)
h̄ t1 2h̄2 t1

The time-ordering operator will prove inconvenient, but the following argument
suggests that we can remove it. We have

Tτ V̂ (τ )V̂ (τ ′ ) = V̂ (τ )V̂ (τ ′ ) + θ(τ ′ − τ )[V̂ (τ ′ ), V̂ (τ )]. (5.32)

We know that the commutator vanishes at equal times and presume that evolves
smoothly away from zero. Hence the commutator term will contribute a cor-
rection which is higher-order in ∆t and can be neglected in our perturbative
treatment. Anticipating being able to drop the terms linear in V̂ (because they
have zero mean and within the Markov approximation no correlations with the
prior noise that led to ρ(t1 )) when we trace out the bath, we are thus led to
∫ t2 {
1
ρ(t2 ) = ρ(t1 ) + dτ dτ V̂ (τ )ρ(t1 )V̂ (τ ′ )
′
h̄2 t1
}
1 ′ 1 ′
− V̂ (τ )ρ(t1 )V̂ (τ )ρ(t1 ) − ρ(t1 )V̂ (τ )ρ(t1 )V̂ (τ ) .(5.33)
2 2

In order to make further progress we now need to express the perturbation

in terms of system and bath variables and trace out the bath variables to obtain
an expression for the time evolution of the reduced density matrix describing the
system alone. Let us suppose that in the interaction picture the bath coupling
has the form
1
V̂ (τ ) = η̂z (τ )σ̂ z (τ ) + η̂x (τ )σ̂ x (τ ) (5.34)
h̄
= η̂z (τ )σ z + η̂x (τ ){eiΩτ σ + + e−iΩτ σ − }. (5.35)

To keep the discussion simple we will assume that the longitudinal and trans-
verse noises are uncorrelated and can therefore be treated separately in our
perturbative treatment. We start with the longitudinal noise and will attempt
to reproduce our previous semi-classical result in Eq. (5.18). Introducing the
notation
Tr bath ρ(t) ≡ ⟨⟨ρ(t)⟩⟩, (5.36)
we can write
∫ t2 {
⟨⟨ρ(t1 + ∆t)⟩⟩ = ⟨⟨ρ(t1 )⟩⟩ + dτ dτ ′ Tr bath η̂z (τ )σ z ρ(t1 )σ z η̂z (τ ′ )
t1
}
1 1
− η̂z (τ )η̂z (τ ′ )σ z σ z ρ(t1 ) − ρ(t1 )η̂z (τ )η̂z (τ ′ )σ z σ z . (5.37)
2 2
Circuit QED 55

Using the cyclic property of the trace in the first term, interchanging the dummy
labels τ and τ ′ in the remaining terms and invoking the Markov approximation
we obtain
∫ t2
⟨⟨ρ(t1 + ∆t)⟩⟩ = ⟨⟨ρ(t1 )⟩⟩ + dτ dτ ′ ⟨⟨η̂z (τ ′ )η̂z (τ )⟩⟩
t1
{ }
1 z z 1
σ ⟨⟨ρ(t1 )⟩⟩σ − σ σ ⟨⟨ρ(t1 )⟩⟩ − ⟨⟨ρ(t1 )⟩⟩σ z σ z .
z z
(5.38)
2 2
Of course σ z σ z = 1, but for notational symmetry, we will not use this. Invoking
the Markov approximation as in Eq. (5.16) and defining

γφ = 2Sηz ηz (0), (5.39)

we have
γφ
⟨⟨ρ(t1 + ∆t)⟩⟩ = ⟨⟨ρ(t1 )⟩⟩ + ∆t D[σ z ]⟨⟨ρ(t1 )⟩⟩, (5.40)
2
where the ‘dissipator’ D is called the Lindblad superoperator
1 1
D[O]ρ ≡ OρO† − O† Oρ − ρO† O. (5.41)
2 2
Even though ∆t is not supposed to be too small, if we are appropriately
cautious about the meaning, we can interpret this difference equation as a dif-
ferential equation for the time evolution of the reduced density matrix,
d 1
⟨⟨ρ(t)⟩⟩ = γφ D[σ z ]⟨⟨ρ(t)⟩⟩. (5.42)
dt 2
Let us now try to go from this result to the time evolution of the qubit polar-
ization. Invoking Eq. (5.27) we have
d 1 dm
⃗ 1 1
⟨⟨ρ(t)⟩⟩ = · ⃗σ = γφ D[σ z ] [1 + m
⃗ · σ]. (5.43)
dt 2 dt 2 2
Using the fact that the different Pauli matrices anticommute this can be rewrit-
ten
dm
⃗
· ⃗σ = −γφ {mx σ x + my σ y }. (5.44)
dt
From this we immediately obtain

ṁz = 0 (5.45)
1
ṁx = − mx (5.46)
Tφ
1
ṁy = − my , (5.47)
Tφ

in agreement with the phenomenological Bloch equations (5.5) with T1 = ∞.

In order to obtain a finite energy relaxation rate, we must allow for noise in
the transverse field components. Repeating the derivation above and defining
56 S.M. Girvin

the up and down transition rates (which agree with the derivation of Fermi’s
Golden Rule in [6])

Γ↓ = Sηx ηx (+Ω) (5.48)

Γ↑ = Sηx ηx (−Ω), (5.49)

Eq. (5.42) becomes the standard master equation [23] in Lindblad form

d γφ
⟨⟨ρ(t)⟩⟩ = D[σ z ]⟨⟨ρ(t)⟩⟩ + Γ↑ D[σ + ]⟨⟨ρ(t)⟩⟩ + Γ↓ D[σ − ]⟨⟨ρ(t)⟩⟩. (5.50)
dt 2
As discussed extensively in Ref. [6], the bath noise spectral density at frequency
+Ω is a measure of the ability of the bath to absorb energy and thereby de-
excite the qubit, while the spectral density at −Ω gives the ability of the bath
to emit energy at that frequency, thereby exciting the qubit. Classically there is
no distinction since the noise correlators are real-valued, but quantum mechan-
ically we must distinguish positive and negative frequency because the noise
correlators are complex-valued (which is possible since the noise operators are
hermitian but their product is not: they do not commute with each other at
different times [6]). Evaluating the above, we arrive at the full Bloch equations
in (5.5).

Exercise 5.4 Derive Eq. (5.5) in detail from Eq. (5.50).

We emphasize that the Markov approximation is not necessarily valid. For

instance the qubit could be looking at the environmental noise filtered by a
a resonator or cavity, or in the case of dephasing noise at low frequencies the
spectrum of the noise may be 1/f rather than white. The former can be handled
by adding the filter (as a harmonic oscillator) to the Hamiltonian. The latter
complicates the analysis and leads to non-exponential decay [24].

Language inherited from NMR. Homogeneous, inhomogeneous. Requires care

in interpreting when switching from giant ensemble to single qubits. Also Γ↓
much bigger than Γ↑ , unlike in NMR where temperature is much larger than
the frequency.
For generic values of the gate charge, the ground state Ψ0 and excited state
Ψ1 differ in their respective static electric ‘dipole moments’

pj ∼ 2ed⟨Ψj |n̂|Ψj ⟩ (5.51)

where d is (approximately) the distance between the two islands of the qubit.
(More precisely, the effective value of d depends in a complex way on the details
of the cavity and qubit geometry and the resulting capacitance matrix [15].)
Nakamura et al. [3] used the dependence of a certain quasi-particle tunneling
rate on pj to readout the state of the qubit. Aassime et al. [25] and Lehnert
et al. [26] developed an RF single electron transistor readout scheme for charge
based qubits.
Circuit QED 57

Unfortunately in the regime where charge based readout works, a stray elec-
tric field E causes a first-order perturbation theory shift of the qubit excitation
frequency by an amount
1
δω01 = E(p1 − p0 ). (5.52)
h̄
This leads to very rapid dephasing of quantum superpositions at rate [24]
( )2
1 1 p1 − p0
= SEE , (5.53)
Tφ 2 h̄

where SEE is the electric field spectral density at low frequencies. The total
decoherence rate is then given by
1 1 1
∗ = + . (5.54)
T2 2T1 Tφ

Recognizing this difficulty, Vion et al. [14] introduced the notion of operating
charge qubits at a ‘sweet spot’ where the transition frequency is an extremum
with respect to the voltage and magnetic field control parameters. Several such
extrema (known in atomic physics as ‘clock points’) are visible in the Cooper
pair box spectrum shown in Fig. (4.6). Here the leading term in the Taylor
series expansion presented in Eq. (5.9 vanishes and only the second order effects
of the noise contribute to the dephasing. The qubit coherence time is therefore
dramatically increased [14]. Measurements by the Devoret group have demon-
strated that these coherence lifetimes are indeed consistent with the expected
second-order effects of the noise determined by the curvature of the spectrum
at the extremum [27]. As noted in Fig. (4.6), by going to large EJ /EC one can
dramatically reduce the curvature of the spectrum at the extrema and eventu-
ally make the transition frequency exponentially close to constant independent
of the gate charge [15, 28]. Operation in this transmon regime leads to even
longer lifetimes and long-term qubit frequency stability [4, 28].
58 S.M. Girvin
Chapter 6

Introduction to Cavity and

Circuit QED

Quantum electrodynamics (QED) studies atoms and electrons coupled to the

quantum fluctuations of the electromagnetic field. Cavity QED engineers those
quantum fluctuations of the vacuum by surrounding the atoms with a resonant
cavity which supports only discrete modes of the electromagnetic field. By
adjusting the frequencies (and damping) of those resonant modes with respect
to the transition frequency of the atoms, one can dramatically alter the coupling
of the atoms to their quantum environment. Furthermore because the cavity
traps photons, they interact with the atoms repeatedly and one can enter a
regime of strong coupling in which the natural excitations of the system are
coherent superpositions of atom and photon excitations known as polaritons.
This physics of the environment is related to that of the Purcell effect [29],
first observed for a superconducting qubit by Houck et al. [30]. Here a qubit
placed inside a cavity can have its decay rate suppressed if it is far detuned from
the cavity resonance or enhanced if the qubit transition frequency is close to the
cavity resonance. The former is useful for protecting quantum superpositions.
The latter is useful for providing rapid qubit reset to the ground state. It has
also been successfully used to generate single microwave photons on demand
and enhance the fidelity of coherent quantum information transfer from a su-
perconducting qubit to a ‘flying’ photon qubit [5]. One can view the Purcell
effect as the resonator performing an impedance transformation on the external
dissipation presented by the environment to the qubit [30]. Neeley et al. [31]
have used a tunable transformer coupling to quantitatively explore the role of
environmental coupling in a phase qubit circuit over a wide range of coupling
strengths.
The coupling between the electric field in the cavity and the dipole moment
of a (single) atom at position R⃗ is given by

U = −⃗
p · E(
⃗ R).
⃗ (6.1)

59
60 S.M. Girvin

As we have seen, each discrete mode of a resonator is an independent harmonic

oscillator. Will assume for now that only a single mode (the one closest in
frequency to the atom transition) is important. (Of course it is important to
remember that this is not always a good approximation.) The electric field in the
cavity (or voltage in a circuit) can be written in terms of the mode polarization
direction ϵ̂ and the mode zero point fluctuation amplitude EZPF (computed at
⃗ and the photon operators as1
position R)

⃗ = ϵ̂EZPF (a + a† ).
E (6.2)

We assume that the atom can be approximated as a two-level system. The

dipole moment operator connects the ground and excited states of the atom
and so the interaction Hamiltonian becomes

H1 = h̄g(a + a† )σ x , (6.3)

where σ x flips the two-level system between states and the ‘vacuum Rabi cou-
pling’ g is given by
h̄g = −⟨ψ1 |⃗
p · ϵ̂|ψ0 ⟩EZPF . (6.4)
Eq. (6.3) can be rewritten as

H1 = h̄g(aσ + + a† σ − ) + h̄g(aσ − + a† σ + ). (6.5)

The second term only mixes states that are far away from each other, or equiva-
lently, it is rapidly rotating in the interaction picture. Dropping this term (again
not always a valid approximation!) is called the rotating wave approximation
(RWA).
In this simplest approximation (including the rotating wave approximation),
the system is described by the Jaynes-Cummings Hamiltonian

H = H0 + +Hdrive + Hdamping + V, (6.6)

where
h̄ω01 z
H = h̄ωc a† a +σ (6.7)
2
where the single cavity mode is described as a simple harmonic oscillator of
angular frequency ωc , the two-level atom is represented as a simple spin-1/2
with excitation frequency ω01 ,

V = h̄g(aσ + + a† σ− ) (6.8)

and the ‘vacuum Rabi coupling’, g, represents the dipole matrix element for
the process in which the atom absorbs or emits a photon. The external driving
and damping terms, not written explicitly here, which help control the electro-
magnetic state of the cavity, are treated using the input-output formalism of
1 We make a certain gauge choice here and could have used E ⃗ = −iϵ̂EZPF (a − a† ) to be
closer to the choice made in discussing the voltage of the LC oscillator in Eq. (2.51).
Circuit QED 61

quantum optics [6] described in Appendix 8. The extension of this Hamiltonian

to the case of multiple qubits is known as the Tavis-Cummings model [32].
There is a long history of cavity QED studies in the AMO community for
both alkali atoms in optical cavities [33–37] and Rydberg atoms in microwave
cavities [38–43]. In the optical case one typically monitors the effect of the
atoms on the photons transmitted through the cavity. It is not possible to
measure the state of the atoms after they have fallen through the cavity because
the spontaneous emission lifetime is on the order of nanoseconds at optical
frequencies. In the microwave experiments pioneered by the Paris group it
is difficult to directly measure the microwave photons but relatively easy to
measure the state of the Rydberg atoms with very high fidelity after they exit
the cavity since they have a lifetime of approximately 30 ms and can be probed
with state-selective ionization.
‘Circuit QED’ uses superconducting qubits as artificial atoms coupled to mi-
crowave resonators [1, 44–47] as illustrated schematically in figure 7.1. Measur-
ing the amplitude and phase of microwaves transmitted through the resonator
realizes the equivalent of optical cavity QED at microwave frequencies. In re-
cent years there were many theoretical proposals for coupling qubits to either
three-dimensional cavities or lumped element resonators [48–56] and there has
been a flurry of experiments [5, 16, 27, 30, 45, 46, 57–80] and further theoretical
proposals too numerous to list.
There exists a dual geometry in which the Josephson junction qubit is placed
in line with the center pin of the resonator and couples directly to the microwave
currents flowing in the resonator [47,81]. In this dual geometry the fine structure
constant is replaced by its inverse and the problem is engineering the circuit to
reduce the coupling to manageable levels.
The coupling g is most readily measured by tuning the qubit transition
frequency ω01 to match the cavity frequency ωc . The resulting degeneracy is
lifted by the dipole coupling term leading to the so-called vacuum Rabi splitting
2g [35–37, 42, 45, 60]. The two lowest lying excited states are coherent super-
positions (‘bonding-anti-bonding’ combinations) of photon excitation and qubit
excitation. The coupling available in circuit QED is now so strong that split-
tings of ∼ 300 line widths are easily achieved [70, 77]. The higher lying excited
states form a strongly anharmonic ladder which can be explored by either strong
driving or use of two excitation tones [69, 70, 77].
In the so-called dispersive regime where the qubit is far detuned from the
cavity (|ω01 − ωc | ≫ g), diagonalization of the Hamiltonian to lowest order in g
leads to a second-order dispersive coupling which is QND with respect to both
the photon number and the qubit energy
g2 † 1
Udispersive = h̄ [a a + ]σ z , (6.9)
∆ 2
where ∆ ≡ ω01 − ωc is the detuning of the qubit from the cavity. The disper-
sive coupling can be interpreted either as a shift in the cavity frequency which
depends on the state of the qubit, or as the ‘ac-Stark’ or ‘light’ shift (plus the
Lamb shift [44, 57, 73]) of the qubit frequency proportional to the number of
62 S.M. Girvin

(a)

Cin Cg
CJ , E J
Lr Cr CB
Vg Φ

(b)

Cin E
Cg1

CB Cg2

Figure 6.1: Circuit QED: Schematic illustration (not to scale) of a transmon

qubit embedded in a coplanar waveguide resonator. Panel (a) shows the lumped
element circuit equivalent to the distributed circuit shown in panel (b). From
[15].

photons in the cavity. The qubit-state-dependent shift of the cavity frequency

leads to changes in the amplitude and phase of photons reflected from or trans-
mitted through the cavity and is the basis of the QND readout of the qubit state
in circuit QED [44,45]. The mean value of the light shift can be used to rapidly
tune qubit transition frequencies [57,59,67,82]. The fluctuating part of the light
shift can be viewed as the quantum back action [6] of the qubit measurement.
As required by the principles of quantum measurement [6], the photon shot
noise [83] in the cavity gradually dephases the qubit superposition as informa-
tion is gained about σ z . This back action effect leads to a broadening of the
spectroscopic line width of the qubit [57, 59, 84–87]. In the so-called ‘strong-
dispersive’ regime [63], the coupling is so large that the light shift per photon
2
exceeds both the cavity line width κ and the atom line width γ: g∆ > κ, γ. In
this regime the qubit spectrum breaks up into a series of separately resolved
peaks representing the distribution of photon numbers within the driven cav-
ity [63]. This ‘photon number’ detector was used to distinguish thermal and
coherent states in the cavity and could be used to measure number-squeezed
states and other non-classical states [63]. This strong-coupling physics has been
beautifully observed in the time domain by the Paris group [38–41].
To derive Eq. (7.4) we need to find a unitary transformation
†
U = eχ ; u† = eχ = e−χ , (6.10)
Circuit QED 63

which removes the off-diagonal term in Eq. (??) which is first-order in g. Using
the Baker-Hausdorf expansion we have
1
H̃ = U HU † = H + [χ, H] + [χ, [χ, H]] + . . . (6.11)
2
In perturbation theory we write H = H0 + V where V is the off-diagnonal term
linear in g and we expect χ to therefore be of order g. Expanding to second
order in g we have
1
H̃ ≈ H0 + V + [χ, H0 ] + [χ, V ] + [χ, [χ, H0 ]]. (6.12)
2
We need to choose χ to satisfy

[χ, H0 ] = −V (6.13)

in order to remove the off-diagonal term. We are then left with only the second-
order terms which conveniently combine to
1
H̃ = H0 + [χ, V ]. (6.14)
2
It is straightforward to verify that the solution to Eq. (6.13) is
g
χ= (aσ + − a† σ − ). (6.15)
∆
Clearly the expansion in g is valid only for large enough detuning so that
g/∆ ≪ 1. (On the other hand if ∆ becomes too large the RWA fails so cau-
tion is required.) Computing the second-order term (and dropping an irrelevant
constant) we arrive at the dispersive Hamiltonian

g2 † 1
H̃ = H0 + h̄ (a a + )σ z . (6.16)
∆ 2
The rate of progress in observing novel strong coupling non-linear quantum
optics effects in superconducting electrical circuits is quite remarkable. As noted
above, Houck et al.used the Purcell effect [30] to generate non-classical photon
states in a cavity [5]. The states were a superposition of n = 0 and n = 1 Fock
states with controlled amplitude and phase. ‘Fluorescence tomography’ was per-
formed on these states using square law detection to determine the probability
of having a photon. In addition, homodyne measurements were performed to
determine the two quadratures of the electric field which are controlled by the
off-diagonal coherence between the n = 0 and n = 1 Fock states. In particular
they showed that the mean electric field of the one-photon Fock state was zero.
Higher Fock states up to n = 6 were synthesized by the UCSB group [71] who
also observed that the decay rate scaled linearly with n as expected [72]. This
same effect was seen qualitatively in the frequency domain in the experiment
of Schuster et al. [63]. The qubit spectrum showed up to 6 resolved peaks
64 S.M. Girvin

displaying the distribution of photon numbers within the driven cavity and the
line width of the peaks increased with n. In a 2009 tour-de-force, Hofheinz et
al. [79] demonstrated a remarkable method for synthesizing arbitrary photon
states (including Fock and various cat states) in a cavity and measuring their
Wigner distributions. This level of control now exceeds what has been possible
to date with atomic physics methods.
Because microwave photons have 104 to 105 times less energy than visible
photons, they are much more difficult to detect. The work of Houck et al. [5]
and Schuster et al. [63] showed that individual photons could be detected with
low efficiency and the recent work of Hofheinz et al. [79] demonstrated very high
efficiency detection of individual photons in a cavity. However a general purpose
high bandwidth ‘photomultiplier’ does not yet exist in the microwave regime.
There have been some theoretical proposals for single photon detection [88, 89]
but this remains an important open experimental problem.
Another novel new direction is construction of single artificial atom ‘lasers’
[68, 90, 91] as well as Sisyphus cooling and amplification [74] of an oscillator.
The extreme strong coupling available should permit observation of ‘photon
blockade’ effects [92], and parametric down-conversion by three-wave mixing [93,
94]. The advances in our understanding and fabrication of Josephson junction
circuits motivated by the quest for a quantum computer have led to dramatic
advances in the ability to do four-wave mixing, parametric amplification near
the quantum limit, as well as strong squeezing of the vacuum [95, 96]. These
advances will not only permit much better dispersive readout of qubits, they also
open up the possibility of continuous variable quantum information processing
[97, 98] since two-mode squeezed states are an entanglement resource.
Chapter 7

NOBEL SYMPOSIUM
TEXT BEGINS HERE

7.1 Introduction
The remarkable recent progress in creating superconducting quantum bits and
manipulating their states has been summarized in several reviews [1,8,9,99–104].
Nearly 30 years ago Leggett discussed the fundamental issues concerning the
collective degrees of freedom in superconducting electrical circuits and the fact
that they themselves can behave quantum mechanically [105]. The essential
collective variable in a Josephson junction [106] is the phase difference of the
superconducting order parameter across the junction. The first experimental
observation of the quantization of the energy levels of the phase ‘particle’ was
made by Martinis, Devoret and Clarke in 1985 [107, 108].
Caldeira and Leggett also pointed out the crucial role that quantum fluc-
tuations of the dissipative electromagnetic environment play in the quantum
coherence and dynamics of Josephson junctions [109]. These ideas were tested
in early experiments on macroscopic quantum tunneling (MQT) [110, 111]. In
a novel experiment, Turlot et al. [112] used a sliding absorber to mechanically
vary the electromagnetic impedance seen by the Josephson junction and hence
modulate the rate of MQT. It was subsequently realized that the quantum
fluctuations of the environment can play an important role even in transport
through normal metal junctions [113, 114].
This physics of the environment is related to that of the Purcell effect [29],
first observed for a superconducting qubit by Houck et al. [30]. Here a qubit
placed inside a cavity can have its decay rate suppressed if it is far detuned from
the cavity resonance or enhanced if the qubit transition frequency is close to the
cavity resonance. The former is useful for protecting quantum superpositions.
The latter is useful for providing rapid qubit reset to the ground state. It has
also been successfully used to generate single microwave photons on demand
and enhance the fidelity of coherent quantum information transfer from a su-

65
66 S.M. Girvin

perconducting qubit to a ‘flying’ photon qubit [5]. One can view the Purcell
effect as the resonator performing an impedance transformation on the external
dissipation presented by the environment to the qubit [30]. Neeley et al. [31]
have used a tunable transformer coupling to quantitatively explore the role of
environmental coupling in a phase qubit circuit over a wide range of coupling
strengths.

7.2 Circuit QED

Quantum electrodynamics (QED) is the study of the interaction of atoms with
the quantized electromagnetic field. In cavity QED one modifies the electro-
magnetic environment by placing the atoms inside a high finesse Fabry-Pérot
resonator. This not only simplifies the physics by making the spectrum of
electromagnetic modes discrete, it also gives the experimentalist control over
both the damping of the resonances and their detuning from the atomic transi-
tion frequency. Furthermore, because the photons bounce between the mirrors
many times, their coupling to the atoms can be greatly enhanced. In the sim-
plest approximation (including the rotating wave approximation), the system is
described by the Jaynes-Cummings Hamiltonian

h̄ω01 z { }
H = h̄ωc a† a + σ + h̄g aσ + + a† σ − + Hdrive + Hdamping , (7.1)
2
where the single cavity mode is described as a simple harmonic oscillator of
angular frequency ωc , the two-level atom is represented as a simple spin-1/2
with excitation frequency ω01 , and the ‘vacuum Rabi coupling’, g, represents
the dipole matrix element for the process in which the atom absorbs or emits a
photon. The external driving and damping terms, not written explicitly here,
which help control the electromagnetic state of the cavity, are treated using the
input-output formalism of quantum optics [6]. The extension of this Hamilto-
nian to the case of multiple qubits is known as the Tavis-Cummings model [32].
There is a long history of cavity QED studies in the AMO community for
both alkali atoms in optical cavities [33–37] and Rydberg atoms in microwave
cavities [38–43]. In the optical case one typically monitors the effect of the
atoms on the photons transmitted through the cavity. It is not possible to
measure the state of the atoms after they have fallen through the cavity because
the spontaneous emission lifetime is on the order of nanoseconds at optical
frequencies. In the microwave experiments pioneered by the Paris group it
is difficult to directly measure the microwave photons but relatively easy to
measure the state of the Rydberg atoms with very high fidelity after they exit
the cavity since they have a lifetime of approximately 30 ms and can be probed
with state-selective ionization.
‘Circuit QED’ uses superconducting qubits as artificial atoms coupled to mi-
crowave resonators [1,44–47] as illustrated schematically in figure 7.1. Measuring
the amplitude and phase of microwaves transmitted through the resonator re-
alizes the equivalent of optical cavity QED at microwave frequencies. In recent
Circuit QED 67

years there were many theoretical proposals for coupling qubits to either three-
dimensional cavities or lumped element resonators [48–56] and there has been a
flurry of experiments [5,16,27,30,45,46,57–80] and further theoretical proposals
too numerous to list. The beauty of coplanar waveguide resonators is that they
are quasi-one-dimensional Fabry-Pérot cavities with orders of magnitude smaller
mode volume than can be achieved with ordinary three-dimensional resonators.
Because the mode volume of these quasi-one-dimensional resonators can be as
small as 10−6 cubic wavelengths, and because the artificial atoms have tran-
sition dipoles much larger than even Rydberg atoms, the coupling strength g
between the atom and a single photon in the resonator is enormously enhanced
and becomes orders of magnitude larger than can be ordinarily achieved. In
fact the dimensionless ratio of the coupling strength to the transition frequency
approaches the limit set by the fine structure constant α [1]
√
g α
∼ , (7.2)
ω01 ϵ

where ϵ is the dielectric constant of the medium surrounding the qubit. (Strictly
speaking this limit is obtained assuming that the quantum charge fluctua-
tions are only of order one Cooper pair. In actuality they increase slowly as
(EJ /EC )1/4 so this limit can be exceeded.) For the lumped element equivalent
circuit shown in figure 7.1 the vacuum Rabi coupling is given by
√
Cg ωq ωr
g= , (7.3)
2 Cq Cr

where ωq is the qubit frequency, ωr is the resonator frequency, and Cq = CB +

2CJ is the capacitance across the the Josephson junction. The values of the
lumped circuit elements are determined by the capacitance matrix of the actual
distributed circuit elements [15].
There exists a dual geometry in which the Josephson junction qubit is placed
in line with the center pin of the resonator and couples directly to the microwave
currents flowing in the resonator [47,81]. In this dual geometry the fine structure
constant is replaced by its inverse and the problem is engineering the circuit to
reduce the coupling to manageable levels.
The coupling g is most readily measured by tuning the qubit transition
frequency ω01 to match the cavity frequency ωc . The resulting degeneracy is
lifted by the dipole coupling term leading to the so-called vacuum Rabi splitting
2g [35–37, 42, 45, 60]. The two lowest lying excited states are coherent super-
positions (‘bonding-anti-bonding’ combinations) of photon excitation and qubit
excitation. The coupling available in circuit QED is now so strong that split-
tings of ∼ 300 line widths are easily achieved [70, 77]. The higher lying excited
states form a strongly anharmonic ladder which can be explored by either strong
driving or use of two excitation tones [69, 70, 77].
In the so-called dispersive regime where the qubit is far detuned from the
cavity (|ω01 − ωc | ≫ g), diagonalization of the Hamiltonian to lowest order in g
68 S.M. Girvin

(a)

Cin Cg
CJ , E J
Lr Cr CB
Vg Φ

(b)

Cin E
Cg1

CB Cg2

Figure 7.1: Circuit QED: Schematic illustration (not to scale) of a transmon

qubit embedded in a coplanar waveguide resonator. Panel (a) shows the lumped
element circuit equivalent to the distributed circuit shown in panel (b). From
[15].

leads to a second-order dispersive coupling which is QND with respect to both

the photon number and the qubit energy

g2 † 1
V = h̄ [a a + ]σ z , (7.4)
∆ 2
where ∆ ≡ ω01 − ωc is the detuning of the qubit from the cavity. The disper-
sive coupling can be interpreted either as a shift in the cavity frequency which
depends on the state of the qubit, or as the ‘ac-Stark’ or ‘light’ shift (plus the
Lamb shift [44, 57, 73]) of the qubit frequency proportional to the number of
photons in the cavity. The qubit-state-dependent shift of the cavity frequency
leads to changes in the amplitude and phase of photons reflected from or trans-
mitted through the cavity and is the basis of the QND readout of the qubit state
in circuit QED [44,45]. The mean value of the light shift can be used to rapidly
tune qubit transition frequencies [57,59,67,82]. The fluctuating part of the light
shift can be viewed as the quantum back action [6] of the qubit measurement.
As required by the principles of quantum measurement [6], the photon shot
noise [83] in the cavity gradually dephases the qubit superposition as informa-
tion is gained about σ z . This back action effect leads to a broadening of the
spectroscopic line width of the qubit [57, 59, 84–87]. In the so-called ‘strong-
dispersive’ regime [63], the coupling is so large that the light shift per photon
2
exceeds both the cavity line width κ and the atom line width γ: g∆ > κ, γ. In
Circuit QED 69

this regime the qubit spectrum breaks up into a series of separately resolved
peaks representing the distribution of photon numbers within the driven cav-
ity [63]. This ‘photon number’ detector was used to distinguish thermal and
coherent states in the cavity and could be used to measure number-squeezed
states and other non-classical states [63]. This strong-coupling physics has been
beautifully observed in the time domain by the Paris group [38–41].
The rate of progress in observing novel strong coupling non-linear quantum
optics effects in superconducting electrical circuits is quite remarkable. As noted
above, Houck et al.used the Purcell effect [30] to generate non-classical photon
states in a cavity [5]. The states were a superposition of n = 0 and n = 1 Fock
states with controlled amplitude and phase. ‘Fluorescence tomography’ was per-
formed on these states using square law detection to determine the probability
of having a photon. In addition, homodyne measurements were performed to
determine the two quadratures of the electric field which are controlled by the
off-diagonal coherence between the n = 0 and n = 1 Fock states. In particular
they showed that the mean electric field of the one-photon Fock state was zero.
Higher Fock states up to n = 6 were synthesized by the UCSB group [71] who
also observed that the decay rate scaled linearly with n as expected [72]. This
same effect was seen qualitatively in the frequency domain in the experiment
of Schuster et al. [63]. The qubit spectrum showed up to 6 resolved peaks
displaying the distribution of photon numbers within the driven cavity and the
line width of the peaks increased with n. In a 2009 tour-de-force, Hofheinz et
al. [79] demonstrated a remarkable method for synthesizing arbitrary photon
states (including Fock and various cat states) in a cavity and measuring their
Wigner distributions. This level of control now exceeds what has been possible
to date with atomic physics methods.
Because microwave photons have 104 to 105 times less energy than visible
photons, they are much more difficult to detect. The work of Houck et al. [5]
and Schuster et al. [63] showed that individual photons could be detected with
low efficiency and the recent work of Hofheinz et al. [79] demonstrated very high
efficiency detection of individual photons in a cavity. However a general purpose
high bandwidth ‘photomultiplier’ does not yet exist in the microwave regime.
There have been some theoretical proposals for single photon detection [88, 89]
but this remains an important open experimental problem.
Another novel new direction is construction of single artificial atom ‘lasers’
[68, 90, 91] as well as Sisyphus cooling and amplification [74] of an oscillator.
The extreme strong coupling available should permit observation of ‘photon
blockade’ effects [92], and parametric down-conversion by three-wave mixing [93,
94]. The advances in our understanding and fabrication of Josephson junction
circuits motivated by the quest for a quantum computer have led to dramatic
advances in the ability to do four-wave mixing, parametric amplification near
the quantum limit, as well as strong squeezing of the vacuum [95, 96]. These
advances will not only permit much better dispersive readout of qubits, they also
open up the possibility of continuous variable quantum information processing
[97, 98] since two-mode squeezed states are an entanglement resource.
70 S.M. Girvin

7.2.1 The Quantronium Qubit

The next great advance was the first Ramsey fringe experiment in an electrical
circuit performed at Saclay [14] using a charge qubit dubbed the quantronium.
This group recognized that there is a sweet spot in offset charge at ng = 1/2
for which the ground and excited state have no difference in dipole moment:
p1 = p0 . At this bias point, the energy splitting is an extremum with respect
to offset charge. Hence the transition frequency depends on stray electric fields
only in second order

 ( )  ( )2
∂ω01  1 1 ∂ 2 ω01  1
δω01 = ng − + ng − + ... (7.5)
∂ng ng =1/2 2 2 ∂n2g ng =1/2 2

By analogy with the effects of stray magnetic fields on atomic transitions

and the optimal working point used in atomic clocks, the charge noise sweet
spot is sometimes called the ‘clock point’. Working at this point, the Saclay
group managed to extend T2∗ approximately three orders of magnitude to ∼
500ns. Subsequently at Yale, the Devoret group [27] was able to perform very
rapid sequences of Ramsey fringe experiments and show that the measured
decoherence rate was consistent with the second-order effects of the curvature
of the transition frequency in the vicinity of the extremum, given typical values
for the offset charge noise.
We now arrive at an interesting quandary. By tuning the qubit to the sweet
spot, the environment is no longer able to detect which state the qubit is in,
based on coupling to its electric dipole moment. This is why the coherence
time is so dramatically enhanced. But, if the environment cannot measure the
state of the qubit by looking at the dipole moment, neither can we! The Saclay
group recognized this and developed the concept of reading out the qubit by
measuring the state dependent susceptibility (inductance). Rather than going
into the details of this, it is easier in the context of the present discussion to
instead explain the closely related state-dependent susceptibility method based
on capacitance developed by the Yale group [44, 45]. Because the offset charge
is essentially equivalent to an applied voltage, and the potential energy of a
capacitor is 12 CV 2 , the second derivative of the transition energy with respect
to ng in (7.5) is essentially the difference in quantum capacitance [11, 115–119]
presented to an external probing field when the qubit is in the ground and
excited states. Essentially this effect was used by the Yale group in developing
the dispersive readout [44, 45] based on (7.4). Working with a low frequency
probe, precisely this effect was measured in [118,119]. The difference is that the
high-frequency dispersive probe depends on the matrix elements related to the
quantum capacitance, but as is clear from (7.4), it also depends on the detuning
of the qubit and resonator frequencies. The importance of this difference will
become clear below.
Circuit QED 71

7.2.2 Transmon Qubits

The most recent evolution of the charge qubit has been the ‘transmon’ qubit
developed by the Yale group [5,15,16,28] and schematically illustrated in figure
7.1. This is nothing more than a CPB operated in the regime of large EJ /EC .
Figure 7.2 shows both the anharmonicity of the CPB spectrum, defined as the
difference of the first two transition frequencies α ≡ ω12 − ω01 , and the relative
anharmonicity, defined as αr ≡ α/ω01 . The anharmonicity is important because
it sets the limit τp on how short qubit control pulses can be to avoid mixing in
higher energy levels outside the logical subspace [58,78,120,121]. For EJ /EC < 9
the CPB spectrum has anharmonicity which is positive. Near EJ /EC = 9 the
first two transition frequencies are too close together for the CPB to be used as a
qubit. Beyond EJ /EC = 9 the CPB has negative anharmonicity whose asymp-
totic value approaches the charging energy α → −EC /h̄. Because √ the transition
frequency approaches the Josephson
√ plasma frequency Ω pl = 8EJ EC , the rela-
tive anharmonicity αr ∼ − 8E EC
J
asymptotically approaches zero as the system
comes closer and closer to being a simple harmonic oscillator. Nevertheless the
charging energy is easily controlled using the experimental geometry and can
be conveniently set in the range of a few hundred MHz, large enough so that
all but the very shortest control pulses do not excite the system out of the sub-
space of the two lowest states. Remarkably, the ‘charge dispersion’ ϵ (overall
amplitude of the periodic variation in transition frequency with offset charge)
becomes exponentially small at large EJ /EC [11, 15, 16, 28]
√
ϵ ∼ e− 8EJ /EC , (7.6)

so that for EJ /EC greater than roughly 50, dephasing due to charge noise is
negligible. Instead of a single charge sweet spot, the levels are essential flat
independent of offset charge and every spot is sweet. Coherence times without
spin echo as large as T2∗ ≈ 3µs with Tφ ≥ 35µs have been observed for the
transmon [28].
We now arrive at our next major quandary. If the energy eigenvalues are
essentially independent of the offset charge then neither we nor the environment
can read the state of the qubit using either the dipole moment or the suscepti-
bility (quantum capacitance) since neither is dependent on the quantum state.
While this explains the even longer coherence times of the transmon, we are left
to wonder how it is that the dispersive readout still works even though quan-
tum capacitance is zero in both states. Recall that if the qubit were actually a
perfect harmonic oscillator, the transition frequencies would not respond at all
to changes in offset charge (displacement of the origin of the oscillator). It is
obvious from classical considerations that the susceptibility would be a constant
(given by the inverse of the spring constant) independent of the state. The os-
cillator is highly polarizable and responds strongly to slow variations in offset
charge by being displaced, but this displacement to a new equilibrium position
has no effect on the excitation spectrum. As noted above, the transmon comes
exponentially close to this ideal behavior and yet, the dispersive readout still
72 S.M. Girvin

Figure 7.2: a) Relative anharmonicity, αr , of the CPB as a function of the dimen-

sionless ratio of the Josephson and charging energies; b) absolute anharmonicity,
α, which asymptotically approaches the negative of the charging energy EC ; c)
(left axis) Phase coherence time T2 given by the inverse of the charge disper-
sion, and (right axis) the minimum control pulse duration τp set by the absolute
anharmonicity. The anharmonicity passes through zero at EJ /Ec = 9 and in
this vicinity the CPB cannot be used as a two-level qubit since τp diverges.
From [15].

works. This is because, while the charge dispersion falls off exponentially, the
transmon retains its anharmonicity. As can be seen from (7.4), the cavity pull
due to the virtual polarization of the qubit is strongly dependent on the detun-
ing between the qubit and cavity. For the case of the multi-level transmon, the
expression for the cavity pull has to be rederived, but the essential point is that
the detuning for the 0 → 1 transition is not the same as that for the 1 → 2
transition and so the cavity pull is state dependent, provided that the cavity
frequency is reasonably close to the qubit. For a very low frequency cavity, we
are back in the regime measuring the quantum capacitance where the effect is
small.
We now face one final quandary associated with the relatively weak (αr ∼
5−10%) anharmonicity of the transmon. In the limit of large detuning from the
cavity (∆ > EC > T1 ), the difference in cavity pulls between the qubit excited
state and ground state is well approximated by [15]

g2
χ1 − χ0 ≈ α . (7.7)
∆2
Detecting the qubit state is equivalent to detecting the cavity pull within time
T1 . A crude figure of merit for our ability to do this is the phase accumulation

ϕ = [χ1 − χ0 ]T1 . (7.8)

Circuit QED 73

The qubit decay rate due to the Purcell effect is

g2
γκ ≈ κ . (7.9)
∆2
Assuming that there is no intrinsic (non-Purcell) relaxation (i.e. that the decay
is Purcell dominated), then T1 ∼ 1/γκ and we have

α EC
ϕ∼ ∼ (7.10)
κ κ
which, remarkably, is independent of g and ∆. The problem is that moving the
qubit closer to the cavity to enhance the homodyne signal shortens the lifetime
of the qubit and reduces the allowable power in the readout beam so that the
SNR does not improve.
While ϕ can in principle be made large, we want EC to remain small to
minimize the dephasing due to charge dispersion and we want κ to be large
so that the readout is fast compared to the intrinsic (non-Purcell) relaxation
(which has been neglected in the above). Hence it would be better to have a
larger anharmonicity. The fluxonium qubit described below will prove useful in
this regard.
A more sophisticated version of the above argument that takes into account
the fact that phase resolution improves with drive power is as follows. The
signal to noise power ratio for the readout to lowest order in χ is given for a
(single-sided) cavity by [6]
[ ]2
χ1 − χ0
SNR ∼ 4 T1 n̄κ, (7.11)
κ

where n̄ is the mean photon number in the readout cavity. Assuming the photon
number is kept at the critical value beyond which the lowest order dispersive
∆2
approximation becomes invalid [82] (n̄ ∼ 4g 2 ) we see that

α2 2
EC
SNR ∼ ∼ , (7.12)
κ2 κ2
independent of g and ∆ (for g/∆ small).
The large islands of the transmon qubit make them susceptible to quasipar-
ticle poisoning, but the very small charge dispersion means that this is almost
certainly not a significant source of dephasing, though in principle quasiparticle
tunneling can contribute to relaxation [16, 122–125]. The contribution to the
relaxation rate by spontaneous photon emission via the Purcell effect for charge
qubits is now well-understood both theoretically and experimentally [30], but
residual sources of relaxation due to dielectric losses [30,126] in the substrate or
the Josephson junction oxide remain less well understood. Small-junction charge
qubits made with minimal fabrication methods seem to suffer less [16] from ma-
terials problems such as the presence of two-level glassy fluctuators [127–132].
74 S.M. Girvin

7.2.3 Fluxonium Qubits

A new qubit design [22] dubbed the ‘fluxonium’ has the closed-loop topology
of the flux qubit, but the loop contains a Josephson junction series array which
gives an inductance much larger than the simple geometric inductance. This
loop shorts out stray low frequency electric fields, but at the qubit frequency
has such a large reactance that it is effectively an open circuit reminiscent of
the charge qubit. Initial results suggest that this qubit design is indeed robust
and stable against charge noise and exhibits very long phase coherence times
reaching several microseconds. It also has the advantage that the anharmonicity
can be large. Interestingly, the state-dependent polarizability and the excitation
spectrum of this design is such that the qubit state can be measured even when
the qubit transition frequency is driven down to very low values of order 0.5
GHz.

7.3 Recent Progress and Future Directions

Progress in the field of superconducting qubits continues at an amazing pace
both in terms of fundamental non-linear quantum optics and in terms of quan-
tum information processing. Since the first experiments ten years ago, phase co-
herence times have risen approximately exponentially from immeasurably small
(1 nanosecond or less) to several microseconds. This has been achieved through
a great deal of hard work and clever quantum engineering of artificial atoms
and circuits by many groups. Building on this progress, DiCarlo et al. [80]
recently demonstrated the first quantum algorithms on a two-qubit supercon-
ducting quantum processor. Future improved qubit designs, microwave circuit
designs, and materials improvements should allow this trend to continue un-
abated. One interesting possibility to extend the coherence time of quantum
circuits would be to use high Q resonators as quantum memories or to form
cavity-stabilized qubits [133, 134]. Another potentially important future di-
rection will be hybrid systems using superconducting resonators and trapped
atoms, molecules or ions [135, 136].
A diverse and remarkable set of methods for reading out charge qubits has
been developed, beginning with the charge state (dipole moment) [3, 25] and
moving on to dispersive readout of the inductive [14] and capacitive [44, 45]
susceptibilities and the low frequency readout of the quantum capacitance [118,
119]. Protocols for optimal readout using linear dispersive detection in the pres-
ence of a finite energy relaxation rate for the qubit have been developed [137].
A novel fast latching dispersive readout using the bifurcation of a driven non-
linear oscillator has proved very effective [27,61,62,138] and fidelities as high as
70% have been achieved. The Saclay group [138] has recently reported fidelities
greater than 90% using a cavity bifurcation amplifier. Thus, fidelities for the
(approximately) QND readout [62, 139] of charge qubits are now approaching
those obtained for flux qubits and for the destructive readout of phase qubits.
The measurement back action (dephasing) required [6] by the Heisenberg un-
Circuit QED 75

certainty principle is now understood to arise from quantum fluctuations of the

light shift [57, 59, 84, 85] and has been observed. The fact that this back ac-
tion can be partially destructive and not simply dephasing has been understood
in terms of the dressed dephasing model [86, 87], although further quantitative
experimental studies of this mechanism still need to be performed. The Santa
Barbara group has observed the non-unitary measurement back action in a ‘par-
tial collapse’ experiment [140] in a phase qubit. All these results demonstrate
that fundamental aspects of quantum measurement theory are now accessible
to condensed matter experiment for the first time. An important future goal for
QND measurements is to see quantum jumps of a qubit and observe the Zeno
effect induced by continuous observation of the qubit [141].
It should be noted that while single-shot high fidelity readouts are extremely
useful, they are in principle not essential for the exponential speed-up promised
by quantum computation. It is possible to design a sequence of quantum algo-
rithms in which one reads out only a single bit of the N -bit answer with each
run. Hence, even if one only has the capability to read out a single qubit and
that only with low fidelity F , the extra computational cost (∼ N/F ) is only lin-
ear. (Of course in practice this may be a huge factor, but it is not exponential.)
High fidelity multi-qubit readout with minimal cross talk is however essential
for Bell inequality tests [142] that close the detector loophole. Using a co-planar
waveguide resonator rather than capacitive coupling, the Santa Barbara group
has recently greatly reduced the cross talk of their readout scheme for phase
qubits and closed the measurement loophole for the Bell violation [143].
An important idea in circuit QED is the understanding that dispersive cou-
pling to the cavity can be used to perform a simultaneous joint readout of multi-
ple qubits [44,144]. This joint readout was used for two-qubit state tomography
by Majer et al. [67]. In its most basic form, the idea is simply that with two
qubits, there are four possible quantum states and four different dispersive fre-
quency pulls of the cavity. If one is in the strong dispersive coupling regime and
all four frequency pulls can be reliably distinguished in a single shot, then one
has two bits of classical information and a complete projective measurement of
both qubits. Of course in the presence of qubit decay and amplifier noise, the
detector tomography can be complex [144]. A more sophisticated understand-
ing of the situation of imperfect resolution of the four peaks has been developed
recently and Filipp et al. [145] demonstrated that it is possible to still reliably
measure two-qubit correlations even in the presence of readout noise.
A simplified version of the theory for joint readout of two qubits is the
following. While the cavity pull is linear in the qubit polarizations
δωc = χ1 σ1z + χ2 σ2z , (7.13)
the corresponding homodyne (transmission) amplitude is not
{ }
z z iφ κ/2
A(σ1 , σ2 ) = Re e . (7.14)
∆ − δωc + iκ/2
Here, ∆ is the detuning of the readout tone from the bare cavity resonance, κ is
the cavity line width, and φ is the local oscillator phase. Because this can take
76 S.M. Girvin

on only four distinct values (corresponding to two classical bits of information)

this expression can always be recast in the form

A(σ1z , σ2z ) = β0 + β1 σ1z + β2 σ2z + β12 σ1z σ2z . (7.15)

The joint coefficient β12 is in general non-zero (as long as ∆ ̸= 0) and typically
on the same scale as the other coefficients. By using pre-rotations (by angle zero
or π) of each of the two qubits prior to making the measurement, it is straight-
forward to obtain any one or two-qubit correlator in the z basis. Ensemble
averaging many such measurements will reduce the statistical uncertainty to
arbitrarily low values. For example,
1
⟨σ1z σ2z ⟩ = ⟨A(σ1z , σ2z ) − A(−σ1z , σ2z ) − A(σ1z , −σ2z ) + A(σ1z , σ2z )⟩ . (7.16)
4β12

Any other arbitrary correlators (e.g. ⟨σ1x σ2y ⟩) can be achieved by pre-pending
rotations through appropriate arbitrary angles. The Yale group has recently
used this to measure values of the CHSH entanglement witness well above the
classical bound [146].
High-fidelity single qubit gates [58, 121] have been developed allowing ob-
servation of the Berry phase for spin 1/2 [66] using a charge qubit and higher
spins [147] simulated using the multilevel structure of a phase qubit. This is a
first step towards larger scale quantum simulators. State and process tomogra-
phy for one and two qubit operations [67, 142, 148] and randomized benchmark-
ing [78] are now routine. A number of methods for two-qubit gates have been
suggested [54, 65, 149–153] involving fixed capacitive couplings [21, 154–156], in-
ductive coupling [157], and virtual photon exchange via a cavity bus [65, 67].
Controlled phase gates have been proposed [158, 159] and realized using virtual
states outside the logical basis [80]. A key problem for the future is to further
increase the on-off ratio of controllable qubit couplings. One possible resource
for this would be tunable cavities [75, 160, 161] or tunable couplings via an ac-
tive element [162, 163]. It seems clear however that simply tuning a coupling
element off resonance will not produce an adequately large on-off ratio. One
may be able to do better by using a tunable interference between two coher-
ent coupling channels to actually null out the coupling at a special operating
point [164].
Given the current rate of progress, it will not be long before other two
qubit algorithms [165] and quantum information processing with more than two
superconducting qubits will be realized. Key short term goals will be to create
multi-qubit entangled states such as the GHZ and W states [144, 166–168], and
begin to execute simple error correction protocols [169, 170].
Another exciting direction involves using multiple physical qubits to realize
individual logical qubits to overcome the difficulties of maintaining stable transi-
tion frequencies. In particular, the possibility of topological protection [171–175]
is beginning to be explored in superconducting qubits [176]. The central idea
is that qubits are constructed in which the ground and excited states are de-
generate and this degeneracy is robust against local variations in Hamiltonian
Circuit QED 77

parameters. Even if the energy levels are not exactly degenerate, it would be
very useful to have a qubit with a “Lambda” energy level scheme, that is,
two nearly degenerate levels that can be coupled via stimulated Raman pulses
through a third level. This would be advantageous both as a robust qubit and
for purposes of fundamental quantum optics studies. It seems reasonably cer-
tain that this cannot be achieved without applied magnetic flux to frustrate the
Josephson couplings (as in a flux qubit or in the fluxonium qubit). Indeed the
fluxonium qubit may turn out to be quite useful as a Lambda system.
To scale up to more qubits in the circuit QED scheme, it will be necessary
to move to two cavities [177] and ultimately to cavity grids [178]. A possible
architecture for an eight-qubit processor is shown in figure 7.3.

Figure 7.3: Circuit QED: Schematic illustration of a possible architecture for

an eight-qubit processor. Four transmon qubits are embedded in each quantum
bus (a coplanar waveguide resonator). The two resonators in turn are connected
by an ‘express bus’ consisting of a tunable resonator whose resonance frequency
can be rapidly moved.

The case of large cavity arrays will be interesting not only as a quantum
computation architecture but also for fundamental quantum optics purposes.
An array of resonators each containing a qubit that induces a Kerr nonlinearity
will be a realization of the boson Hubbard model [179] which exhibits both
superfluid and Mott insulator phases. There is now a burgeoning interest in
seeing ‘quantum phase transitions of light’ [180–199]. Since the transmon qubit
is itself an anharmonic oscillator, one might imagine it would be easier to simply
use a lattice of coupled transmons to realize the boson Hubbard model (with
negative Kerr coefficient). The advantage of using a lattice of resonators is that
their resonance frequencies can be closely matched to a single fixed value. The
Kerr coefficient induced by coupling each resonator to an off-resonant qubit
will have some variation due to variations in qubit transition frequencies, but
this disorder in the Hubbard U will be more tolerable than disorder in the
78 S.M. Girvin

photon ‘site energies.’ Just as cold atom systems are now used to simulate
condensed matter models, so we may be able to use photons as interacting
strongly correlated bosons, which can be probed, measured and controlled in
ways that are impossible in ordinary condensed matter.
In summary, the future of circuit QED looks bright indeed for both practical
applications (such as quantum limited amplifiers) and fundamental new physics
with artificial atoms and microwave photons. Circuit QED is much more than
atomic physics with wires. We have a set of modular elements which are readily
connected together (spatial mode matching is easy with wires!). Hence, we
have the opportunity to assemble large scale structures from these quantum
building blocks and do some real quantum engineering. While our cold atoms
colleagues are busy trying to emulate condensed matter systems, we may be able
to use topological quantum computation and quantum error correction schemes
to realize non-abelian gauge theories of particle physics.
We are grateful for research support provided by Yale University, the Na-
tional Science Foundation, the Army Research Office, and IARPA. We are also
grateful to the numerous colleagues with whom we have had enlightening dis-
cussions and collaborations on the quantum mechanics of electrical circuits.
Chapter 8

Appendix: Semi-Infinite
Transmission Lines,
Dissipation and
Input/Output Theory

This appendix is based on material presented in [6].

Now moving on to semiinfinite transmission lines. More about running waves

than standing waves which are subtle to quantize. Again useful to start with
an understanding of the classical case where we find that a semi-infinite trans-
mission line acts like a resistor and dissipate energy even though all elements
are purely reactive. This irreversibility appears only in the correct order of
limits, size to infinity before time to infinity so that the waves launched down
the line do not ever see the other end.
So far we have studied transmision line resonators which contain discrete
standing wave resonance. We turn now to the case of infinite and semi-infinite
transmission lines where we will deal with traveling waves rather than standing
waves. In order to control the state of quantum bits we will want to send control
pulses down a transmission line to the qubit. However opening up a port into
the qubit to allow control signals in can also allow quantum and thermal noise
to also enter. Such noises can dephase the coherent superpositions of qubits and
also permit the qubits to decay by spontaneously emitting energy out the port
and into the transmission line.
Another reason for studying the case of traveling waves in semi-infinite trans-
mission lines is the following. In addition to controlling qubits by sending pulses
down transmission lines, we may also wish to measure the state of a qubit by
sending microwave pulses down a transmission line and seeing out the reflected

79
80 S.M. Girvin

pulse is modified according to the state of the qubit. Thus we also need to
understand continuum scattering theory for microwaves and how to relate the
properties of the reflected wave to those of the incident wave and of the system
from which they are reflecting.
You might think that since a transmission line is made of purely reactive
elements, it cannot dissipate energy. If however it is semi-infinite in length, then
a wave launched at one end carries away energy and never returns. As usual
in studying irreversible behavior, there is a subtle order of limits here. If time
goes to infinity before the transmission line length goes to infinity then there
is no dissipation. In the opposite order of limits, there is dissipation because
the waves travel away and never have time to reach the end of the transmission
line and reflect back. We will in fact see that a semi-infinite transmission line
acts just like
√ a resistor with resistance equal to its characteristic impedance
R = Zc = ℓ/c. Remarkably however because the transmission line is made
of purely reactive elements we know, as we have already seen, how to treat it
quantum mechanically. Hence we will be able to develop a fully quantum theory
of a dissipative resistor [?].
It is useful to start with a classical model considering as before a coaxial
transmission line modeled as a perfectly conducting wire with inductance per
unit length of ℓ and capacitance to ground per unit length c as shown in Fig. 8.1.
If the voltage at position x at time t is V (x, t), then the charge density is
q(x, t) = cV (x, t). By charge conservation the current I and the charge density
are related by the continuity equation
∂t q + ∂x I = 0. (8.1)
The constitutive relation (essentially Newton’s law) gives the acceleration of the
charges
ℓ∂t I = −∂x V. (8.2)
We can decouple Eqs. (8.1) and (8.2) by introducing left and right propagating
modes
V (x, t) = [V → + V ← ] (8.3)
1 →
I(x, t) = [V − V ← ]. (8.4)
Zc
In terms of the left and right propagating modes, Eqs. (8.1) and 8.2 become
v p ∂x V → + ∂t V → = 0 (8.5)
v p ∂x V ← − ∂t V ← = 0 (8.6)
√
where vp ≡ 1/ ℓc is the wave phase velocity. These equations have solutions
which propagate by uniform translation without changing shape since the line
is (in the present model) dispersionless
x
V → (x, t) = Vout (t − ) (8.7)
vp
x
V ← (x, t) = Vin (t + ), (8.8)
vp
Circuit QED 81

where Vin and Vout are arbitrary functions of their arguments. For an infinite
transmission line, Vout and Vin are completely independent. However for the
case of a semi-infinite line terminated at x = 0 (say) by some system S, these
two solutions are not independent, but rather related by the boundary condition
imposed by the system. We have
V (x = 0, t) = [Vout (t) + Vin (t)] (8.9)
1
I(x = 0, t) = [Vout (t) − Vin (t)], (8.10)
Zc
from which we may eliminate V (x = 0, t) to derive the crucially important
‘input-output’ relation
Vout (t) = Vin (t) + Zc I(x = 0, t). (8.11)
The first term on the RHS is simply the direct reflection of the input wave,
while the second term represents waves radiated into the transmission line by
current injected by the system S.

Figure 8.1: a) Coaxial transmission line, indicating voltages and currents as

defined in the main text. b) Lumped element representation of a transmission
line with capacitance per unit length c = C/a and inductance per unit length
ℓ = L/a. c) Discrete LC resonator terminating a transmission line.

If the system under study is just an open circuit so that I(x = 0, t) = 0, then
Vout = Vin , meaning that the outgoing wave is simply the result of the incoming
wave reflecting from the open circuit termination. In general however, there is
an additional outgoing wave radiated by the current I that is injected by the
system dynamics into the line. In the absence of an incoming wave we have
V (x = 0, t) = Zc I(x = 0, t), (8.12)
indicating that the transmission line acts as a simple resistor which, instead of
dissipating energy by Joule heating, carries the energy away from the system
as propagating waves. As noted above, the fact that the line can dissipate
energy despite containing only purely reactive elements is a consequence of its
infinite extent. One must be careful with the order of limits, taking the length
to infinity before allowing time to go to infinity. In this way the outgoing waves
never reach the far end of the transmission line and reflect back. Since this is a
conservative Hamiltonian system, we will be able to quantize these waves and
make a quantum theory of resistors [?]. The net power flow carried to the right
by the line is
1 2
P = [V (t) − Vin2 (t)]. (8.13)
Zc out
The fact that the transmission line presents a dissipative impedance to the
system means that it causes damping of the system. It also however opens up
82 S.M. Girvin

the possibility of controlling the system via the input field which partially deter-
mines the voltage driving the system. From this point of view it is convenient
to eliminate the output field by writing the voltage as

V (x = 0, t) = 2Vin (t) + Zc I(x = 0, t). (8.14)

As we will discuss in more detail below, the first term drives the system and
the second damps it. From Eq. (8.11) we see that measurement of the outgoing
field can be used to determine the current I(x = 0, t) injected by the system
into the line and hence to infer the system dynamics that results from the input
drive field.
The great benefit of Eq. (8.14) is the following. If the system S is coupled to
the transmission line through a capacitor, then the coupling Hamiltonian can
be expressed in terms of the voltage at the end V (x = 0, t). By eliminating Vout
we can see how the system is driven by Vin and damped by Zc I(x = 0, t). This
classical result will be helpful in understanding the closely analogous quantum
expressions which will be derived further below in a rather different manner.
As a simple classical example, consider the system consisting of an LC res-
onator shown in Fig. (8.1 c). This can be viewed as a simple harmonic oscillator
whose coordinate Q is the charge on the capacitor plate (on the side connected to
L0 ). The current I(x = 0, t) = Q̇ plays the role of the velocity of the oscillator.
The equation of motion for the oscillator is readily obtained from

Q = C0 [−V (x = 0+ , t) − L0 I(x
˙ = 0+ , t)]. (8.15)

Using Eq. (8.14) we obtain a harmonic oscillator damped by the transmission

line and driven by the incoming waves

2
Q̈ = −Ω20 Q − γ Q̇ − Vin (t), (8.16)
L0
√
where the resonant frequency is Ω20 ≡ 1/ L0 C0 . Note that the term Zc I(x =
0, t) in Eq. (8.14) results in the linear viscous damping rate γ ≡ Zc /L0 .
If we solve the equation of motion of the oscillator, we can predict the out-
going field. In the present instance of a simple oscillator we have a particular
example of the general case where the system responds linearly to the input
field. We can characterize any such system by a complex, frequency dependent
impedance Z[ω] defined by

V (x = 0, ω)
Z[ω] = − . (8.17)
I(x = 0, ω)

Note the peculiar minus sign which results from our definition of positive current
flowing to the right (out of the system and into the transmission line). Using
Eqs. (8.9, 8.10) and Eq. (8.17) we have

Vout [ω] = r[ω]Vin [ω], (8.18)

Circuit QED 83

where the reflection coefficient r is determined by the impedance mismatch

between the system and the line and is given by the well known result

Z[ω] − Zc
r[ω] = . (8.19)
Z[ω] + Zc

If the system is constructed from purely reactive (i.e. lossless) components,

then Z[ω] is purely imaginary and the reflection coefficient obeys |r| = 1 which
is consistent with Eq. (8.13) and the energy conservation requirement of no net
power flow into the lossless system. For example, for the series LC oscillator we
have been considering, we have

1
Z[ω] = + jωL0 , (8.20)
jωC0

where, to make contact with the usual electrical engineering sign conventions,
we have used j = −i. If the damping γ of the oscillator induced by coupling it to
the transmission line is small, the quality factor of the resonance will be high
√ and
we need only consider frequencies near the resonance frequency Ω0 ≡ 1/ L0 C0
where the impedance has a zero. In this case we may approximate

2
Z[ω] ≈ [Ω0 − ω] = 2jL0 (ω − Ω0 ) (8.21)
jC0 Ω20

which yields for the reflection coefficient

ω − Ω0 + jγ/2
r[ω] = (8.22)
ω − Ω0 − jγ/2

showing that indeed |r| = 1 and that the phase of the reflected signal winds by
2π upon passing through the resonance. 1
Turning to the more general case where the system also contains lossy ele-
ments, one finds that Z[ω] is no longer purely imaginary, but has a real part
satisfying Re Z[ω] > 0. This in turn implies via Eq. (8.19) that |r| < 1. In
the special case of impedance matching Z[ω] = Zc , all the incident power is
dissipated in the system and none is reflected. The other two limits of interest
are open circuit termination with Z = ∞ for which r = +1 and short circuit
termination Z = 0 for which r = −1.
Finally, if the system also contains an active device which has energy being
pumped into it from a separate external source, it may under the right conditions
be described by an effective negative resistance Re Z[ω] < 0 over a certain
frequency range. Eq. (8.19) then gives |r| ≥ 1, implying |Vout | > |Vin |. Our
system will thus act like the one-port reflection amplifier discussed in great
detail in Ref. [6].
1 For the case of resonant transmission through a symmetric cavity, the phase shift only

winds by π.
84 S.M. Girvin

8.0.1 Quantum Input-output theory for a driven cavity

The results from the previous section can be more formally derived in a full
quantum theory of a cavity driven by an external coherent source. The theory
relating the drive, the cavity and the outgoing waves radiated by the cavity
is known as input-output theory and the classical description was presented in
Appendix 8. The present quantum discussion closely follows standard refer-
ences on the subject [?, ?, 34]. The crucial feature that distinguishes such an
approach from many other treatments of quantum-dissipative systems is the
goal of keeping the bath modes instead of tracing them out. This is obviously
necessary for the situations we have in mind, where the output field emanating
from the cavity contains the information acquired during a measurement of the
system coupled to the cavity. As we learned from the classical treatment, we
can eliminate the outgoing waves in favor of a damping term for the system.
However we can recover the solution for the outgoing modes completely from
the solution of the equation of motion of the damped system being driven by
the incoming waves.
In order to drive the cavity we must partially open one of its ports which
exposes the cavity both to the external drive and to the vacuum noise outside
which permits energy in the cavity to leak out into the surrounding bath. We
will formally separate the degrees of freedom into internal cavity modes and
external bath modes. Strictly speaking, once the port is open, these modes are
not distinct and we only have ‘the modes of the universe’ [?, ?, ?]. However for
high Q cavities, the distinction is well-defined and we can model the decay of
the cavity in terms of a spontaneous emission process in which an internal boson
is destroyed and an external bath boson is created. We assume a single-sided
cavity. For a high Q cavity, this physics is accurately captured in the following
Hamiltonian
Ĥ = Ĥsys + Ĥbath + Ĥint . (8.23)
The bath Hamiltonian is ∑
Ĥbath = h̄ωq b̂†q b̂q (8.24)
q

where q labels the quantum numbers of the independent harmonic oscillator

bath modes obeying
[b̂q , b̂†q′ ] = δq,q′ . (8.25)
Note that since the bath terminates at the system, there is no translational
invariance, the normal modes are standing not running waves, and the quantum
numbers q are not necessarily wave vectors.
The coupling Hamiltonian is (within the rotating wave approximation)
∑[ ]
Ĥint = −ih̄ fq ↠b̂q − fq∗ b̂†q â . (8.26)
q

For the moment we will leave the system (cavity) Hamiltonian to be completely
general, specifying only that it consists of a single degree of freedom (i.e. we
Circuit QED 85

concentrate on only a single resonance of the cavity with frequency ωc ) obeying

the usual bosonic commutation relation

[â, ↠] = 1. (8.27)

(N.B. this does not imply that it is a harmonic oscillator. We will consider both
linear and non-linear cavities.) Note that the most general linear coupling to
the bath modes would include terms of the form b̂†q ↠and b̂q a but these are
neglected within the rotating wave approximation because in the interaction
representation they oscillate at high frequencies and have little effect on the
dynamics.
The Heisenberg equation of motion (EOM) for the bath variables is

˙ i
b̂q = [Ĥ, b̂q ] = −iωq b̂q + fq∗ â (8.28)
h̄
We see that this is simply the EOM of a harmonic oscillator driven by a forcing
term due to the motion of the cavity degree of freedom. Since this is a linear
system, the EOM can be solved exactly. Let t0 < t be a time in the distant past
before any wave packet launched at the cavity has reached it. The solution of
Eq. (8.28) is
∫ t
b̂q (t) = e−iωq (t−t0 ) b̂q (t0 ) + dτ e−iωq (t−τ ) fq∗ â(τ ). (8.29)
t0

The first term is simply the free evolution of the bath while the second represents
the waves radiated by the cavity into the bath.
The EOM for the cavity mode is
i ∑
â˙ = [Ĥsys , â] − fq b̂q . (8.30)
h̄ q

Substituting Eq. (8.29) into the last term above yields

∑ ∑
fq b̂q = fq e−iωq (t−t0 ) b̂q (t0 )
q q
∑ ∫ t
+ |fq |2 dτ e−i(ωq −ωc )(t−τ ) [e+iωc (τ −t) â(τ )], (8.31)
q t0

where the last term in square brackets is a slowly varying function of τ . To

simplify our result, we note that if the cavity system were a simple harmonic
oscillator of frequency ωc then the decay rate from the n = 1 single photon
excited state to the n = 0 ground state would be given by the following Fermi
Golden Rule expression
∑
κ(ωc ) = 2π |fq |2 δ(ωc − ωq ). (8.32)
q
86 S.M. Girvin

From this it follows that

∫ +∞ ∑
dν
κ(ωc + ν)e−iν(t−τ ) = |fq |2 e−i(ωq −ωc )(t−τ ) . (8.33)
−∞ 2π q

We now make the Markov approximation which assumes that κ(ν) = κ is a

constant over the range of frequencies relevant to the cavity so that Eq. (8.33)
may be represented as
∑
|fq |2 e−i(ωq −ωc )(t−τ ) = κδ(t − τ ). (8.34)
q

Using ∫ x0
1
dx δ(x − x0 ) = (8.35)
−∞ 2
we obtain for the cavity EOM
i κ ∑
â˙ = [Ĥsys , â] − â − fq e−iωq (t−t0 ) b̂q (t0 ). (8.36)
h̄ 2 q

The second term came from the part of the bath motion representing the wave
radiated by the cavity and, within the Markov approximation, has become a
simple linear damping term for the cavity mode. Note the important factor of
2. The amplitude decays at half the rate of the intensity (the energy decay rate
κ).
Within√ the spirit of the Markov approximation it is further convenient to
treat f ≡ |fq |2 as a constant and define the density of states (also taken to
be a constant) by ∑
ρ= δ(ωc − ωq ) (8.37)
q

so that the Golden Rule rate becomes

κ = 2πf 2 ρ. (8.38)

We can now define the so-called ‘input mode’

1 ∑ −iωq (t−t0 )
b̂in (t) ≡ √ e b̂q (t0 ) . (8.39)
2πρ q

For the case of a transmission line treated in Appendix ??, this coincides with
the field b̂→ moving towards the cavity [see Eq. (??)]. We finally have for the
cavity EOM
i κ √
â˙ = [Ĥsys , â] − â − κ b̂in (t). (8.40)
h̄ 2
Note that when a wave packet is launched from the bath towards the cavity,
causality prevents it from knowing about the cavity’s presence until it reaches
the cavity. Hence the input mode evolves freely as if the cavity were not present
Circuit QED 87

until the time of the collision at which point it begins to drive the cavity. Since
b̂in (t) evolves under the free bath Hamiltonian and acts as the driving term in
the cavity EOM, we interpret it physically as the input mode. Eq. (8.40) is the
quantum analog of the classical equation (8.16), for our previous example of an
LC-oscillator driven by a transmission line. The latter would also have been
first order in time if as in Eq. (??) we had worked with the complex amplitude
A instead of the coordinate Q.
Eq. (8.39) for the input mode contains a time label just as in the interaction
representation. However it is best interpreted as simply labeling the particular
linear combination of the bath modes which is coupled to the system at time t.
Some authors even like to think of the bath modes as non-propagating while the
cavity flies along the bath (taken to be 1D) at a velocity v. The system then
only interacts briefly with the local mode positioned at x = vt before moving
on and interacting with the next local bath mode. We will elaborate on this
view further at the end of this subsection.
The expression for the power Pin (energy per time) impinging on the cavity
depends on the normalization chosen in our definition of b̂in . It can be obtained,
for example, by imagining the bath modes b̂q to live on a one-dimensional waveg-
uide with propagation velocity v and length L (using periodic boundary condi-
tions). In that case we have to sum over all photons to get the average⟨ power ⟩
∑
flowing through a cross-section of the waveguide, Pin = q h̄ωq (vp /L) b̂†q b̂q .
Inserting the definition for b̂in , Eq. (8.39), the expression for the input power
carried by a monochromatic beam at frequency ω is
⟨ ⟩
Pin (t) = h̄ω b̂†in (t)b̂in (t) (8.41)

Note that this has the√correct dimensions due to our choice of normalization for
b̂in (with dimensions ω). In the general case, an integration over frequencies
is needed (as will be discussed further below). An analogous formula holds for
the power radiated by the cavity, to be discussed now.
The output mode b̂out (t) is radiated into the bath and evolves freely after
the system interacts with b̂in (t). If the cavity did not respond at all, then the
output mode would simply be the input mode reflected off the cavity mirror. If
the mirror is partially transparent then the output mode will also contain waves
radiated by the cavity (which is itself being driven by the input mode partially
transmitted into the cavity through the mirror) and hence contains information
about the internal dynamics of the cavity. To analyze this output field, let
t1 > t be a time in the distant future after the input field has interacted with
the cavity. Then we can write an alternative solution to Eq. (8.28) in terms of
the final rather than the initial condition of the bath
∫ t1
−iωq (t−t1 )
b̂q (t) = e b̂q (t1 ) − dτ e−iωq (t−τ ) fq∗ â(τ ). (8.42)
t
Note the important minus sign in the second term associated with the fact that
the time t is now the lower limit of integration rather than the upper as it was
in Eq. (8.29).
88 S.M. Girvin

Defining
1 ∑ −iωq (t−t1 )
b̂out (t) ≡ √ e b̂q (t1 ), (8.43)
2πρ q

we see that this is simply the free evolution of the bath modes from the distant
future (after they have interacted with the cavity) back to the present, indicating
that it is indeed appropriate to interpret this as the outgoing field. Proceeding
as before we obtain
i κ √
â˙ = [Ĥsys , â] + â − κ b̂out (t). (8.44)
h̄ 2
Subtracting Eq. (8.44) from Eq. (8.40) yields
√
b̂out (t) = b̂in (t) + κ â(t) (8.45)

which is consistent with our interpretation of the outgoing field as the reflected
incoming field plus the field radiated by the cavity out through the partially
reflecting mirror.
The above results are valid for any general cavity Hamiltonian. The general
procedure is to solve Eq. (8.40) for â(t) for a given input field, and then solve
Eq. (8.45) to obtain the output field. For the case of an empty cavity we can
make further progress because the cavity mode is a harmonic oscillator

Ĥsys = h̄ωc ↠â. (8.46)

In this simple case, the cavity EOM becomes

κ √
â˙ = −iωc â − â − κ b̂in (t). (8.47)
2
Eq. (8.47) can be solved by Fourier transformation, yielding
√
κ
â[ω] = − b̂in [ω] (8.48)
i(ωc − ω) + κ/2
√
= − κχc [ω − ωc ]b̂in [ω] (8.49)

and
ω − ωc − iκ/2
b̂out [ω] = b̂in [ω] (8.50)
ω − ωc + iκ/2
which is the result for the reflection coefficient quoted in Eq. (??). For brevity,
here and in the following, we will sometimes use the susceptibility of the cavity,
defined as
1
χc [ω − ωc ] ≡ (8.51)
−i(ω − ωc ) + κ/2
For the case of steady driving on resonance where ω = ωc , the above equations
yield √
κ
b̂out [ω] = â[ω]. (8.52)
2
Circuit QED 89

In steady state, the incoming power equals the outgoing power, and both are
related to the photon number inside the single-sided cavity by
⟨ ⟩ κ⟨ ⟩
P = h̄ω b̂†out (t)b̂out (t) = h̄ω ↠(t)â(t) (8.53)
4
Note that⟨ this⟩ does not coincide with the naive expectation, which would be
P = h̄ωκ ↠â . The reason for this discrepancy is the the interference between
the part of the incoming wave which is promptly reflected from the cavity and
the field radiated by the cavity. The naive expression becomes correct after the
drive has been switched off (where ignoring the effect of the incoming vacuum
√
noise, we would have b̂out = κâ). We note in passing that for a driven two-
sided cavity with coupling constants κL and κR (where κ = κL + κR ), the
incoming power sent into the left port is related to the photon number by
⟨ ⟩
P = h̄ωκ2 /(4κL ) ↠â . (8.54)
Here for κL = κR the interference effect completely eliminates the reflected
beam and we have in contrast to Eq. (8.53)
κ⟨ ⟩
P = h̄ω ↠â . (8.55)
2
Eq. (8.47) can also be solved in the time domain to obtain
â(t) = e−(iωc +κ/2)(t−t0 ) â(t0 )
∫ t
√
− κ dτ e−(iωc +κ/2)(t−τ ) b̂in (τ ). (8.56)
t0

If we take the input field to be a coherent drive at frequency ωL = ωc + ∆

so that its amplitude has a classical and a quantum part
b̂in (t) = e−iωL t [b̄in + ξ(t)]
ˆ (8.57)
and if we take the limit t0 → ∞ so that the initial transient in the cavity ampli-
tude has damped out, then the solution of Eq. (8.56) has the form postulated
in Eq. (??) with √
κ
ā = − b̄in (8.58)
−i∆ + κ/2
and (in the frame rotating at the drive frequency)
∫ t
√
ˆ =− κ
d(t) ˆ ).
dτ e+(i∆−κ/2)(t−τ ) ξ(τ (8.59)
−∞

Even in the absence of any classical drive, the input field delivers vacuum
fluctuation noise to the cavity. Notice that from Eqs. (8.39, 8.57)
[b̂in (t), b̂†in (t′ )] =ˆ ξˆ† (t′ )]
[ξ(t),
1 ∑ −i(ωq −ωL )(t−t′ )
= e
2πρ q
= δ(t − t′ ), (8.60)
90 S.M. Girvin

which is similar to Eq. (??) for a quantum transmission line. This is the op-
erator equivalent of white noise. Using Eq. (8.56) in the limit t0 → −∞ in
Eqs. (??,8.59) yields

[â(t), ↠(t)] = [d(t),

ˆ dˆ† (t)]
∫ t ∫ t
= κ dτ dτ ′ e−(−i∆+κ/2)(t−τ )
−∞ −∞
′
e−(+i∆+κ/2)(t−τ ) δ(τ − τ ′ )
= 1 (8.61)

as is required for the cavity bosonic quantum degree of freedom. We can inter-
pret this as saying that the cavity zero-point fluctuations arise from the vacuum
noise that enters through the open port. We also now have a simple physical
interpretation of the quantum noise in the number of photons in the driven
cavity in Eqs. (??,??,??). It is due to the vacuum noise which enters the cav-
ity through the same ports that bring in the classical drive. The interference
between the vacuum noise and the classical drive leads to the photon number
fluctuations in the cavity.
In thermal equilibrium, ξˆ also contains thermal radiation. If the bath is
being probed only over a narrow range of frequencies centered on ωc (which we
have assumed in making the Markov approximation) then we have to a good
approximation (consistent with the above commutation relation)

⟨ξˆ† (t)ξ(t
ˆ ′ )⟩ = N δ(t − t′ ) (8.62)
ˆ ξˆ† (t′ )⟩
⟨ξ(t) = ′
(N + 1)δ(t − t ) (8.63)

where N = nB (h̄ωc ) is the thermal equilibrium occupation number of the mode

at the frequency of interest. We can gain a better understanding of Eq. (8.62)
by Fourier transforming it to obtain the spectral density
∫ +∞
S[ω] = ˆ ′ )⟩eiω(t−t′ ) = N.
dt ⟨ξˆ† (t)ξ(t (8.64)
−∞

As mentioned previously, this dimensionless quantity is the spectral density that

would be measured by a photomultiplier: it represents the number of thermal
photons passing a given point per unit time per unit bandwidth. Equivalently
the thermally radiated power in a narrow bandwidth B is

P = h̄ωN B. (8.65)

One often hears the confusing statement that the noise added by an amplifier
is a certain number N of photons (N = 20, say for a good cryogenic HEMT
amplifier operating at 5 GHz). This means that the excess output noise (referred
back to the input by dividing by the power gain) produces a flux of N photons
per second in a 1 Hz bandwidth, or 106 N photons per second in 1 MHz of
bandwidth (see also Eq. (??)).
Circuit QED 91

We can gain further insight into input-output theory by using the following
picture. The operator b̂in (t) represents the classical drive plus vacuum fluctu-
ations which are just about to arrive at the cavity. We will be able to show
that the output field is simply the input field a short while later after it has
interacted with the cavity. Let us consider the time evolution over a short time
period ∆t which is very long compared to the inverse bandwidth of the vac-
uum noise (i.e., the frequency scale beyond which the vacuum noise cannot be
treated as constant due to some property of the environment) but very short
compared to the cavity system’s slow dynamics. In this circumstance it is useful
to introduce the quantum Wiener increment related to Eq. (??)
∫ t+∆t
c≡
dW ˆ )
dτ ξ(τ (8.66)
t

which obeys
c , dW
[dW c † ] = ∆t. (8.67)
In the interaction picture (in a displaced frame in which the classical drive
has been removed) the Hamiltonian term that couples the cavity to the quantum
noise of the environment is from Eq. (8.26)
√
V̂ = −ih̄ κ(â† ξˆ − âξˆ† ). (8.68)

Thus the time evolution operator (in the interaction picture) on the jth short
time interval [tj , tj + ∆t] is
√ c † −↠dW
c)
κ(â dW
Ûj = e (8.69)

Using this we can readily evolve the incoming temporal mode forward in time
by a small step ∆t
√
dWc ′ = Û † dW
c Û ≈ dW
c + κ∆t â. (8.70)

Recall that in input-output theory we formally defined the outgoing field

as the bath field far in the future propagated back (using the free field time
evolution) to the present, which yielded
√
b̂out = b̂in + κâ. (8.71)

Eq. (8.70) is completely equivalent to this. Thus we confirm our understanding

that the incoming field is the bath temporal mode just before it interacts with
the cavity and the outgoing field is the bath temporal mode just after it interacts
with the cavity.
This leads to the following picture which is especially useful in the quantum
trajectory approach to conditional quantum evolution of a system subject to
weak continuous measurement [?, 34]. On top of the classical drive b̄in (t), the
bath supplies to the system a continuous stream of “fresh” harmonic oscilla-
tors, each in their ground state (if T = 0). Each oscillator with its quantum
92 S.M. Girvin

fluctuation dW c interacts briefly for a period ∆t with the system and then is
disconnected to propagate freely thereafter, never interacting with the system
again. Within this picture it is useful to think of the oscillators arrayed in an
infinite stationary line and the cavity flying over them at speed vp and touching
each one for a time ∆t.
Bibliography

[1] RJ Schoelkopf and SM Girvin. Wiring up quantum systems. Nature,

451:664, 2008.

[2] M. Devoret. Quantum Fluctuations, pages 351–385. Elsevier, Amsterdam,

1997.

[3] Y. Nakamura, Yu. A. Pashkin, and J. S. Tsai. Coherent control of macro-

scopic quantum states in a single-cooper-pair box. Nature, 398:786–788,
1999.

[4] Hanhee Paik, D. I. Schuster, Lev S. Bishop, G. Kirchmair, G. Catelani,

A. P. Sears, B. R. Johnson, M. J. Reagor, L. Frunzio, L. Glazman, and
R. J. Schoelkopf. How coherent are josephson junctions? arXiv:1105.4652.

[5] A. A. Houck, D. I. Schuster, J. M. Gambetta, J. A. Schreier, B. R. John-

son, J. M. Chow, L. Frunzio, J. Majer, M. H. Devoret, S. M. Girvin,
and R. J. Schoelkopf. Generating single microwave photons in a circuit.
Nature, 449:328–331, 2007.

[6] A. A. Clerk, M. H. Devoret, S. M. Girvin, F. Marquardt, and R. J.

Schoelkopf. Introduction to quantum noise, measurement and amplifica-
tion. Rev. Mod. Phys., 2008. (Longer version with pedagogical appendices
available at: arXiv.org:0810.4729).

[7] Ronald M. Foster. A reactance theorem. Bell System Technical Journal,

3:259–267, 1924.

[8] MH Devoret and JM Martinis. Superconducting qubits. In Esteve, D and

Raimond, JM and Dalibard, J, editor, Quantum Entanglement and In-
formation Processing, volume 79 of LES HOUCHES SUMMER SCHOOL
SESSION, pages 443–485, 2004. Les Houches Session 79th on Quantum
Entanglement and Information Processing, Les Houches, FRANCE, JUN
30-JUL 25, 2003.

[9] John Clarke and Frank K. Wilhelm. Superconducting quantum bits. Na-
ture, 453:1031–1042, 2008.

93
94 S.M. Girvin

[10] V. Bouchiat, D. Vion, P. Joyez, D. Esteve, and M. H. Devoret. Phys. Scr.,

T76:165, 1998.
[11] D.V. Averin, A.B. Zorin, and K.K. Likharev. Bloch oscillations in small
josephson junctions. Sov. Phys. JETP, 61:407, 1985.
[12] M. Büttiker. Zero-current persistent potential drop across small-
capacitance josephson junctions. Phys. Rev. B, 36:3548–3555, 1987.
[13] P. Lafarge, P. Joyez, D. Esteve, C. Urbina, and M. H. Devoret. Two-
electron quantization of the charge on a superconductor. Nature, 365:422–
424, 1993.
[14] D. Vion, A. Aassime, A. Cottet, P. Joyez, H. Pothier, C. Urbina, D. Es-
teve, and M. H. Devoret. Manipulating the Quantum State of an Electrical
Circuit. Science, 296:886–889, 2002.
[15] Jens Koch, Terri M. Yu, Jay Gambetta, A. A. Houck, D. I. Schuster,
J. Majer, Alexandre Blais, M. H. Devoret, S. M. Girvin, and R. J.
Schoelkopf. Charge-insensitive qubit design derived from the Cooper pair
box. Phys. Rev. A, 76:042319, 2007.
[16] J. A. Schreier, A. A. Houck, Jens Koch, D. I. Schuster, B. R. Johnson,
J. M. Chow, J. M. Gambetta, J. Majer, L. Frunzio, M. H. Devoret, S. M.
Girvin, and R. J. Schoelkopf. Suppressing charge noise decoherence in
superconducting charge qubits. Phys. Rev. B, 77:180502, 2008.
[17] J. E. Mooij, T. P. Orlando, L. Levitov, Lin Tian, Caspar H. van der Wal,
and Seth Lloyd. Josephson Persistent-Current Qubit. Science, 285:1036–
1039, 1999.
[18] Caspar H. van der Wal, A. C. J. ter Haar, F. K. Wilhelm, R. N. Schouten,
C. J. P. M. Harmans, T. P. Orlando, Seth Lloyd, and J. E. Mooij.
Quantum superposition of macroscopic persistent-current states. Science,
290:773–777, 2000.
[19] I. Chiorescu, Y. Nakamura, C. J. P. M. Harmans, and J. E. Mooij. Co-
herent Quantum Dynamics of a Superconducting Flux Qubit. Science,
299:1869–1871, 2003.
[20] JM Martinis, S Nam, J Aumentado, and C Urbina. Rabi oscillations in a
large Josephson-junction qubit. Phys. Rev. Lett., 89:117901, 2002.
[21] A. J. Berkley, H. Xu, R. C. Ramos, M. A. Gubrud, F. W. Strauch, P. R.
Johnson, J. R. Anderson, A. J. Dragt, C. J. Lobb, and F. C. Wellstood.
Entangled Macroscopic Quantum States in Two Superconducting Qubits.
Science, 300:1548–1550, 2003.
[22] Vladimir E. Manucharyan, Jens Koch, Leonid Glazman, and Michel De-
voret. Defying the fine structure constant: single cooper pair circuit free
of charge offsets. Science, (in press):[arXiv:0906.0831v1], (2008).
Circuit QED 95

[23] HJ Carmichael. An Open Systems Approach to Quantum Optics. Springer-

Verlag, 1993.

[24] JM Martinis, S Nam, J Aumentado, KM Lang, and C Urbina. Decoher-

ence of a superconducting qubit due to bias noise. Phys. Rev. B, 67:094510,
2003.

[25] A. Aassime, G. Johansson, G. Wendin, R. J. Schoelkopf, and P. Delsing.

Radio-frequency single-electron transistor as readout device for qubits:
Charge sensitivity and backaction. Phys. Rev. Lett., 86:3376–3379, 2001.

[26] K. W. Lehnert, K. Bladh, L. F. Spietz, D. Gunnarsson, D. I. Schuster,

P. Delsing, and R. J. Schoelkopf. Measurement of the excited-state lifetime
of a microelectronic circuit. Phys. Rev. Lett., 90:027002, 2003.

[27] M. Metcalfe, E. Boaknin, V. Manucharyan, R. Vijay, I. Siddiqi, C. Rigetti,

L. Frunzio, R. J. Schoelkopf, and M. H. Devoret. Measuring the decoher-
ence of a quantronium qubit with the cavity bifurcation amplifier. Phys.
Rev. B, 76:174516, 2007.

[28] A.A. Houck, Jens Koch, M.H. Devoret, S.M. Girvin, and R.J. Schoelkopf.
Life after charge noise: recent results with transmon qubits. Quantum
Information Processing, 8:105, 2009.

[29] E. M. Purcell. Spontaneous emission probabilities at radio frequencies.

Phys. Rev., 69:681, 1946.

[30] A. A. Houck, J. A. Schreier, B. R. Johnson, J. M. Chow, Jens Koch, J. M.

Gambetta, D. I. Schuster, L. Frunzio, M. H. Devoret, S. M. Girvin, and
R. J. Schoelkopf. Controlling the spontaneous emission of a superconduct-
ing transmon qubit. Phys. Rev. Lett., 101:080502, 2008.

[31] Matthew Neeley, M. Ansmann, Radoslaw C. Bialczak, M. Hofheinz,

N. Katz, Erik Lucero, A. O’Connell, H. Wang, A. N. Cleland, and John M.
Martinis. Transformed dissipation in superconducting quantum circuits.
Phys. Rev. B, 77:180508, 2008.

[32] Michael Tavis and Frederick W. Cummings. Exact solution for an n-

molecule radiation-field hamiltonian. Phys. Rev., 170:379–384, 1968.

[33] H. Mabuchi and A. Doherty. Cavity quantum electrodynamics: Coherence

in context. Science, 298:1372, 2002.

[34] D. Walls and G. Milburn. Quantum optics. Spinger-Verlag, Berlin, 1994.

[35] R. J. Thompson, G. Rempe, and H. J. Kimble. Observation of normal-

mode splitting for an atom in an optical cavity. Phys. Rev. Lett., 68:1132–
1135, 1992.
96 S.M. Girvin

[36] A Boca, R Miller, KM Birnbaum, AD Boozer, J McKeever, and HJ Kim-

ble. Observation of the vacuum Rabi spectrum for one trapped atom.
Phys. Rev. Lett., 93:233603, 2004.

[37] I. Schuster, A. Kubanek, A. Fuhrmanek, T. Puppe, P. W. H. Pinkse,

K. Murr, and G. Rempe. Nonlinear spectroscopy of photons bound to one
atom. Nature Physics, 4:382–385, 2008.

[38] G. Nogues, A. Rauschenbeutel, S. Osnaghi, M. Brune, J. M. Raimond,

and S. Haroche. Seeing a single photon without destroying it. Nature,
400:239–242, 1999.

[39] Christine Guerlin, Julien Bernu, Samuel Deleglise, Clement Sayrin, Se-
bastien Gleyzes, Stefan Kuhr, Michel Brune, Jean-Michel Raimond,
and Serge Haroche. Progressive field-state collapse and quantum non-
demolition photon counting. Nature, 448:889–893, 2007.

[40] Sebastien Gleyzes, Stefan Kuhr, Christine Guerlin, Julien Bernu, Samuel
Deleglise, Ulrich Busk Hoff, Michel Brune, Jean-Michel Raimond, and
Serge Haroche. Quantum jumps of light recording the birth and death of
a photon in a cavity. Nature, 446:297–300, 2007.

[41] Samuel Deleglise, Igor Dotsenko, Clement Sayrin, Julien Bernu, Michel
Brune, Jean-Michel Raimond, and Serge Haroche. Reconstruction of non-
classical cavity field states with snapshots of their decoherence. Nature,
455:510–514, 2008.

[42] J. Raimond, M. Brune, and S. Haroche. Manipulating quantum entan-

glement with atoms and photons in a cavity. Rev. Mod. Phys., 73:565,
2001.

[43] Serge Haroche and Jean-Michel Raimond. Exploring the Quantum:

Atoms, Cavities and Photons. Oxford University Press, 2006.

[44] Alexandre Blais, Ren-Shou Huang, Andreas Wallraff, S. M. Girvin, and

R. J. Schoelkopf. Cavity quantum electrodynamics for superconducting
electrical circuits: an architecture for quantum computation. Phys. Rev.
A, 69:062320, 2004.

[45] A. Wallraff, D. I. Schuster, A. Blais, L. Frunzio, R.-S. Huang, J. Majer,

S. Kumar, S. M. Girvin, and R. J. Schoelkopf. Circuit quantum elec-
trodynamics: Coherent coupling of a single photon to a cooper pair box.
Nature, 431:162–167, 2004.

[46] I. Chiorescu, P. Bertet, K. Semba, Y. Nakamura, C. J. P. M. Harmans,

and J. E. Mooij. Coherent dynamics of a flux qubit coupled to a harmonic
oscillator. Nature, 431:159–162, 2004.
Circuit QED 97

[47] Michel Devoret, Steven Girvin, and Robert Schoelkopf. Circuit-QED:

How strong can the coupling between a Josephson junction atom and a
transmission line resonator be? Annalen der Physik, 16:767–779, 2007.

[48] A. Shnirman, G. Schön, and Z. Hermon. Quantum manipulations of small

josephson junctions. Phys. Rev. Lett., 79:2371, 1997.

[49] Y. Makhlin, G. Schön, and A. Shnirman. Quantum-state engineering with

josephson-junction devices. Rev. Mod. Phys., 73:357, 2001.

[50] O. Buisson and F. Hekking. Macroscopic Quantum Coherence and Quan-

tum Computing, chapter Entangled states in a Josephson charge qubit
coupled to a superconducting resonator. Kluwer, New York, 2001.

[51] F. Marquardt and C. Bruder. Superposition of two mesoscopically distinct

quantum states: Coupling a cooper pair box to a large superconducting
island. Phys. Rev. B, 63:054514, 2001.

[52] W. A. Al-Saidi and D. Stroud. Eigenstates of a small josephson junction

coupled to a resonant cavity. Phys. Rev. B, 65:014512, 2001.

[53] F. Plastina and G. Falci. Communicating josephson qubits. Phys. Rev.

B, 67:224514, 2003.

[54] A. Blais, A. Maassen van den Brink, and A. Zagoskin. Tunable coupling
of superconducting qubits. Phys. Rev. Lett., 90:127901, 2003.

[55] C.-P. Yang, S.-I. Chu, and S. Han. Possible realization of entanglement,
logical gates, and quantum-information transfer with superconducting-
quantum-interference device qubits in cavity qed. Phys. Rev. A, 67:042311,
2003.

[56] J. Q. You and F. Nori. Quantum information processing with supercon-

ducting qubits in a microwave field. Phys. Rev. B, 68:064509, 2003.

[57] DI Schuster, A Wallraff, A Blais, L Frunzio, RS Huang, J Majer,

SM Girvin, and RJ Schoelkopf. ac Stark shift and dephasing of a su-
perconducting qubit strongly coupled to a cavity field. Phys. Rev. Lett.,
94:123602, 2005.

[58] A Wallraff, DI Schuster, A Blais, L Frunzio, J Majer, MH Devoret,

SM Girvin, and RJ Schoelkopf. Approaching unit visibility for control
of a superconducting qubit with dispersive readout. Phys. Rev. Lett.,
95:060501, 2005.

[59] D. I. Schuster, A. Wallraff, A. Blais, L. Frunzio, R.-S. Huang, J. Majer,

S. M. Girvin, and R. J. Schoelkopf. Erratum: ac stark shift and dephasing
of a superconducting qubit strongly coupled to a cavity field [Phys. Rev.
Lett. 94, 123602 (2005)]. Phys. Rev. Lett., 98:049902, 2007.
98 S.M. Girvin

[60] J. Johansson, S. Saito, T. Meno, H. Nakano, M. Ueda, K. Semba, and

H. Takayanagi. Vacuum rabi oscillations in a macroscopic superconducting
qubit lc oscillator system. Phys. Rev. Lett., 96:127006, 2006.
[61] I Siddiqi, R Vijay, M Metcalfe, E Boaknin, L Frunzio, RJ Schoelkopf, and
MH Devoret. Dispersive measurements of superconducting qubit coher-
ence with a fast latching readout. Phys. Rev. B, 73:054510, 2006.
[62] N. Boulant, G. Ithier, P. Meeson, F. Nguyen, D. Vion, D. Esteve, I. Sid-
diqi, R. Vijay, C. Rigetti, F. Pierre, and M. Devoret. Quantum nonde-
molition readout using a josephson bifurcation amplifier. Phys. Rev. B,
76:014525, 2007.
[63] D.I. Schuster, A.A. Houck, J.A. Schreier, A. Wallraff, J. Gambetta,
A. Blais, L. Frunzio, B. Johnson, M.H. Devoret, S.M. Girvin, and R.J.
Schoelkopf. Resolving photon number states in a superconducting circuit.
Nature, 445:515–518, 2007.
[64] A. Wallraff, D. I. Schuster, A. Blais, J. M. Gambetta, J. Schreier, L. Frun-
zio, M. H. Devoret, S. M. Girvin, and R. J. Schoelkopf. Sideband tran-
sitions and two-tone spectroscopy of a superconducting qubit strongly
coupled to an on-chip cavity. Phys. Rev. Lett., 99:050501, 2007.
[65] Mika A. Sillanpää, Jae I. Park, and Raymond W. Simmonds. Coher-
ent quantum state storage and transfer between two phase qubits via a
resonant cavity. Nature, 449:438–442, 2007.
[66] P. J. Leek, J. M. Fink, A. Blais, R. Bianchetti, M. Goeppl, J. M. Gam-
betta, D. I. Schuster, L. Frunzio, R. J. Schoelkopf, and A. Wallraff. Ob-
servation of Berry’s phase in a solid-state qubit. Science, 318:1889–1892,
2007.
[67] J. Majer, J. M. Chow, J. M. Gambetta, Jens Koch, B. R. Johnson, J. A.
Schreier, L. Frunzio, D. I. Schuster, A. A. Houck, A. Wallraff, A. Blais,
M. H. Devoret, S. M. Girvin, and R. J. Schoelkopf. Coupling supercon-
ducting qubits via a cavity bus. Nature, 449:443–447, 2007.
[68] O. Astafiev, K. Inomata, A. O. Niskanen, T. Yamamoto, Yu. A. Pashkin,
Y. Nakamura, and J. S. Tsai. Single artificial-atom lasing. Nature,
449:588–590, 2007.
[69] Frank Deppe, Matteo Mariantoni, E. P. Menzel, A. Marx, S. Saito,
K. Kakuyanagi, H. Tanaka, T. Meno, K. Semba, H. Takayanagi, E. Solano,
and R. Gross. Two-photon probe of the Jaynes-Cummings model and
controlled symmetry breaking in circuit QED. Nature Physics, 4:686–691,
2008.
[70] J. M. Fink, M. Goeppl, M. Baur, R. Bianchetti, P. J. Leek, A. Blais, and
A. Wallraff. Climbing the Jaynes-Cummings ladder and observing its root
n nonlinearity in a cavity QED system. Nature, 454:315–318, 2008.
Circuit QED 99

[71] Max Hofheinz, E. M. Weig, M. Ansmann, Radoslaw C. Bialczak, Erik

Lucero, M. Neeley, A. D. O’Connell, H. Wang, John M. Martinis, and
A. N. Cleland. Generation of Fock states in a superconducting quantum
circuit. Nature, 454:310–314, 2008.

[72] H. Wang, M. Hofheinz, M. Ansmann, R. C. Bialczak, E. Lucero, M. Nee-

ley, A. D. O’Connell, D. Sank, J. Wenner, A. N. Cleland, and John M.
Martinis. Measurement of the Decay of Fock States in a Superconducting
Quantum Circuit. Phys. Rev. Lett., 101:240401, 2008.

[73] A. Fragner, M. Goeppl, J. M. Fink, M. Baur, R. Bianchetti, P. J. Leek,

A. Blais, and A. Wallraff. Resolving Vacuum Fluctuations in an Electrical
Circuit by Measuring the Lamb Shift. Science, 322:1357–1360, 2008.

[74] M. Grajcar, S. H. W. van der Ploeg, A. Izmalkov, E. Il’ichev, H.-G. Meyer,

A. Fedorov, A. Shnirman, and Gerd Schon. Sisyphus cooling and ampli-
fication by a superconducting qubit. Nature Physics, 4:612–616, 2008.

[75] M. Sandberg, C. M. Wilson, F. Persson, G. Johansson, V. Shumeiko,

T. Bauch, T. Duty, and P. Delsing. Fast tuning of superconducting mi-
crowave cavities. In Goan, HS and Chen, YN, editor, Solid-state Quan-
tum Computing, Proceedings, volume 1074 of AIP CONFERENCE PRO-
CEEDINGS, pages 12–21, 2008. 2nd International Workshop on Solid-
State Quantum Computing/Mini-School on Quantum Information Sci-
ence, Taipei, TAIWAN, JUN 23-27, 2008.

[76] E. Il’ichev, S. H. W. van der Ploeg, M. Grajcar, and H. G. Meyer. Weak

continuous measurements of multiqubits systems. Quantum Information
Processing, 8:133–153, 2009.

[77] Lev S. Bishop, J. M. Chow, Jens Koch, A. A. Houck, M. H. Devoret,

E. Thuneberg, S. M. Girvin, and R. J. Schoelkopf. Nonlinear response of
the vacuum Rabi resonance. Nature Physics, 5:105–109, 2009.

[78] J. M. Chow, J. M. Gambetta, L. Tornberg, Jens Koch, Lev S. Bishop,

A. A. Houck, B. R. Johnson, L. Frunzio, S. M. Girvin, and R. J.
Schoelkopf. Randomized Benchmarking and Process Tomography for Gate
Errors in a Solid-State Qubit. Phys. Rev. Lett., 102:090502, 2009.

[79] Max Hofheinz, H. Wang, M. Ansmann, Radoslaw C. Bialczak, Erik

Lucero, M. Neeley, A. D. O’Connell, D. Sank, J. Wenner, John M. Mar-
tinis, and A. N. Cleland. Synthesizing arbitrary quantum states in a
superconducting resonator. Nature, 459:546–549, 2009.

[80] L. DiCarlo, J. M. Chow, J. M. Gambetta, Lev S. Bishop, B. R. Johnson,

D. I. Schuster, J. Majer, A. Blais, L. Frunzio, S. M. Girvin, and R. J.
Schoelkopf. Demonstration of two-qubit algorithms with a superconduct-
ing quantum processor. Nature, 460:240–244, 2009.
100 S.M. Girvin

[81] Jérôme Bourassa, Jay M. Gambetta, Abdufarrukh A. Abdumalikov

Jr, Oleg Astafiev, and Alexandre Blais Yasunobu Nakamura. Ultra-
strong coupling regime of cavity qed with phase-biased flux qubits.
arXiv:0906.1383, 2009.

[82] Jay Gambetta, Alexandre Blais, D. I. Schuster, A. Wallraff, L. Frunzio,

J. Majer, M. H. Devoret, S. M. Girvin, and R. J. Schoelkopf. Qubit-photon
interactions in a cavity: Measurement-induced dephasing and number
splitting. Phys. Rev. A, 74:042318, 2006.

[83] P. Bertet, I. Chiorescu, G. Burkard, K. Semba, C. J. P. M. Harmans,

D. P. DiVincenzo, and J. E. Mooij. Dephasing of a superconducting qubit
induced by photon noise. Phys. Rev. Lett., 95:257002, 2005.

[84] G Ithier, E Collin, P Joyez, PJ Meeson, D Vion, D Esteve, F Chiarello,

A Shnirman, Y Makhlin, J Schriefl, and G Schon. Decoherence in a
superconducting quantum bit circuit. Phys. Rev. B, 72:134519, 2005.

[85] A Lupascu, CJPM Harmans, and JE Mooij. Quantum state detection of

a superconducting flux qubit using a dc-SQUID in the inductive mode.
Phys. Rev. B, 71:184506, 2005.

[86] Maxime Boissonneault, J. M. Gambetta, and Alexandre Blais. Nonlinear

dispersive regime of cavity QED: The dressed dephasing model. Phys.
Rev. A, 77:060305, 2008.

[87] Maxime Boissonneault, J. M. Gambetta, and Alexandre Blais. Dispersive

regime of circuit QED: Photon-dependent qubit dephasing and relaxation
rates. Phys. Rev. A, 79:013819, 2009.

[88] Ferdinand Helmer, Matteo Mariantoni, Enrique Solano, and Florian Mar-
quardt. Quantum nondemolition photon detection in circuit qed and the
quantum zeno effect. Phys. Rev. A, 79:052115, 2009.

[89] G. Romero, J. J. Garcia-Ripoll, and E. Solano. Microwave Photon Detec-

tor in Circuit QED. Phys. Rev. Lett., 102:173602, 2009.

[90] M. Marthaler, Gerd Schön, and Alexander Shnirman. Photon-number

squeezing in circuit quantum electrodynamics. Phys. Rev. Lett.,
101:147001, 2008.

[91] S. Ashhab, J. R. Johansson, A. M. Zagoskin, and Franco Nori. Single-

artificial-atom lasing using a voltage-biased superconducting charge qubit.
NJP, 11:023030, 2009.

[92] K. M. Birnbaum, A. Boca, R. Miller, A. D. Boozer, T. E. Northup, and

H. J. Kimble. Photon blockade in an optical cavity with one trapped
atom. Nature, 436:87–90, 2005.
Circuit QED 101

[93] K Moon and SM Girvin. Theory of microwave parametric down-conversion

and squeezing using circuit QED. Phys. Rev. Lett., 95:140504, 2005.

[94] Florian Marquardt. Efficient on-chip source of microwave photon pairs in

superconducting circuit qed. Phys. Rev. B, 76:205416, 2007.

[95] M. A. Castellanos-Beltran, K. D. Irwin, G. C. Hilton, L. R. Vale, and

K. W. Lehnert. Amplification and squeezing of quantum noise with a
tunable josephson metamaterial. Nature Physics, 4:929–931, 2008.

[96] N. Bergeal, R. Vijay, V. E. Manucharyan, I. Siddiqi, R. J. Schoelkopf,

S. M. Girvin, and M. H. Devoret. Analog information processing at the
quantum limit with a josephson ring modulator. arXiv.org:0805.3452,
2008.

[97] Daniel Gottesman, Alexei Kitaev, and John Preskill. Encoding a qubit in
an oscillator. Phys. Rev. A, 64:012310, 2001.

[98] Samuel L. Braunstein and Peter van Loock. Quantum information with
continuous variables. Rev. Mod. Phys., 77:513–577, 2005.

[99] D Esteve and D Vion. Solid state quantum bit circuits. In Bouchiat, H
and Gefen, Y and Gueron, S and Montambaux, G and Dalibard, J, editor,
Nanophysics: Coherence and Transport, volume 81 of LES HOUCHES
SUMMER SCHOOL SESSION, pages 537+, 2005. Les Houches Session
81st on Nanophysics - Coherence and Transport, Les Houches, FRANCE,
JUN 28-JUL 30, 2004.

[100] G. Wendin and V.S. Shumeiko. Handbook of Theoretical and Computa-

tional Nanotechnology, Vol. 3, chapter ‘Superconducting circuits, qubits
and computing’, pages 223–309. American Scientific Publishers, Los An-
geles, 2006.

[101] G. Wendin and V. S. Shumeiko. Quantum bits with josephson junctions.

J. Low Temperature Physics, 33:724, 2007.

[102] JQ You and F Nori. Superconducting circuits and quantum information.

Physics Today, 58:42–47, 2005.

[103] Franco Nori. Superconducting qubits: Atomic physics with a circuit.

Nature Physics, 4:589–590, 2008.

[104] Alexander N. Korotkov. Special issue on quantum computing with super-

conducting qubits. Quantum Information Processing, 8:51–54, 2009.

[105] AJ Leggett. Macroscopic quantum systems and the quantum theory of

measurement. Progress of Theoretical Physics Suppl., 69:80–100, 1980.

[106] JM Martinis, A Wallraff, and MH Devoret. Superconducting qubits: A

short review. cond-mat/0411.174, 2004.
102 S.M. Girvin

[107] JM Martinis, MH Devoret, and J Clarke. Energy level quantization in

the zero-voltage state of a current-biased josephson junction. Phys. Rev.
Lett., 55:1543 – 1546, 1985.

[108] John Clarke, Andrew N. Cleland, Michel H. Devoret, Daniel Esteve, and
John M. Martinis. Quantum Mechanics of a Macroscopic Variable: The
Phase Difference of a Josephson Junction. Science, 239:992–997, 1988.

[109] AO Caldeira and AJ Leggett. Quantum tunneling in a dissipative system.

Annals of Physics, 149:374–456, 1983.

[110] Richard F Voss and Richard A Webb. Macroscopic quantum tunneling in

1 µ-m nb josephson junctions. Phys. Rev. Lett., 47:265 – 268, 1981.

[111] Michel H Devoret, Martinis, John M, and John Clarke. Measurements of

macroscopic quantum tunneling out of the zero-voltage state of a current-
biased josephson junction. Phys. Rev. Lett., 55:1908 – 1911, 1985.

[112] Emmanuel Turlot, Daniel Esteve, Cristian Urbina, John M. Martinis,

Michel H. Devoret, Sebastian Linkwitz, and Hermann Grabert. Escape
oscillations of a josephson junction switching out of the zero-voltage state.
Phys. Rev. Lett., 62:1788–1791, 1989.

[113] M. H. Devoret, D. Esteve, H. Grabert, G.-L. Ingold, H. Pothier, and

C. Urbina. Effect of the electromagnetic environment on the coulomb
blockade in ultrasmall tunnel junctions. Phys. Rev. Lett., 64:1824–1827,
1990.

[114] S. M. Girvin, L. I. Glazman, M. Jonson, D. R. Penn, and M. D. Stiles.

Quantum fluctuations and the single-junction coulomb blockade. Phys.
Rev. Lett., 64:3183–3186, 1990.

[115] A. Widom, G. Megaloudis, T.D. Clark, J.E. Mutton, R.J. Prance, and
H. Prance. The josephson pendulum as a nonlinear capacitor. J. Low
Temp. Phys., 57:651, 1984.

[116] K.K. Likharev and A.B. Zorin. Theory of the bloch-wave oscillations in
small josephson junctions. J. Low Temp. Phys., 59:347, 1985.

[117] D. V. Averin and C. Bruder. Variable electrostatic transformer: Control-

lable coupling of two charge qubits. Phys. Rev. Lett., 91:057003, 2003.

[118] T Duty, G Johansson, K Bladh, D Gunnarsson, C Wilson, and P Delsing.

Observation of quantum capacitance in the cooper-pair transistor. Phys.
Rev. Lett., 95:206807, 2005.

[119] MA Sillanpää, T Lehtinen, A Paila, Y Makhlin, L Roschier, and PJ Hako-

nen. Direct observation of Josephson capacitance. Phys. Rev. Lett.,
95:206806, 2005.
Circuit QED 103

[120] M Steffen, JM Martinis, and IL Chuang. Accurate control of Josephson

phase qubits. Phys. Rev. B, 68:224518, 2003.

[121] Erik Lucero, M. Hofheinz, M. Ansmann, Radoslaw C. Bialczak, N. Katz,

Matthew Neeley, A. D. O’Connell, H. Wang, A. N. Cleland, and John M.
Martinis. High-fidelity gates in a single Josephson qubit. Phys. Rev. Lett.,
100:247001, 2008.

[122] R Lutchyn, L Glazman, and A Larkin. Quasiparticle decay rate of Joseph-

son charge qubit oscillations. Phys. Rev. B, 72:014517, 2005.

[123] R. M. Lutchyn, L. I. Glazman, and A. I. Larkin. Kinetics of the super-

conducting charge qubit in the presence of a quasiparticle. Phys. Rev. B,
74:064515, 2006.

[124] John M. Martinis, M. Ansmann, and J. Aumentado. Energy decay in

josephson qubits from non-equilibrium quasiparticles. arXiv:0904.2171,
2009.

[125] John M. Martinis. Quasiparticle tunneling: P(e) theory. arXiv:0904.2035,

2009.

[126] JM Martinis, KB Cooper, R McDermott, M Steffen, M Ansmann, KD Os-

born, K Cicak, S Oh, DP Pappas, RW Simmonds, and CC Yu. Decoher-
ence in Josephson qubits from dielectric loss. Phys. Rev. Lett., 95:210503,
2005.

[127] RW Simmonds, KM Lang, DA Hite, S Nam, DP Pappas, and JM Martinis.

Decoherence in Josephson phase qubits from Junction resonators. Phys.
Rev. Lett., 93:077003, 2004.

[128] KB Cooper, M Steffen, R McDermott, RW Simmonds, S Oh, DA Hite,

DP Pappas, and JM Martinis. Observation of quantum oscillations be-
tween a Josephson phase qubit and a microscopic resonator using fast
readout. Phys. Rev. Lett., 93:180401, 2004.

[129] Lara Faoro and Lev B. Ioffe. Quantum two level systems and kondo-like
traps as possible sources of decoherence in superconducting qubits. Phys.
Rev. Lett., 96:047001, 2006.

[130] Lara Faoro and Lev B. Ioffe. Microscopic origin of critical current fluc-
tuations in large, small, and ultra-small area josephson junctions. Phys.
Rev. B, 75:132505, 2007.

[131] Matthew Neeley, M. Ansmann, Radoslaw C. Bialczak, M. Hofheinz,

N. Katz, Erik Lucero, A. O’Connell, H. Wang, A. N. Cleland, and John M.
Martinis. Process tomography of quantum memory in a Josephson-phase
qubit coupled to a two-level state. Nature Physics, 4:523–526, 2008.
104 S.M. Girvin

[132] Magdalena Constantin, Clare C. Yu, and John M. Martinis. Saturation

of two-level systems and charge noise in josephson junction qubits. Phys.
Rev. B, 79:094520, 2009.
[133] R. H. Koch, J. R. Rozen, G. A. Keefe, F. M. Milliken, C. C. Tsuei, J. R.
Kirtley, and D. P. DiVincenzo. Low-bandwidth control scheme for an
oscillator-stabilized josephson qubit. Phys. Rev. B, 72:092512, 2005.
[134] R. H. Koch, G. A. Keefe, F. P. Milliken, J. R. Rozen, C. C. Tsuei, J. R.
Kirtley, and D. P. DiVincenzo. Experimental demonstration of an oscilla-
tor stabilized josephson flux qubit. Phys. Rev. Lett., 96:127001, 2006.
[135] A. Andre, D. Demille, J. M. Doyle, M. D. Lukin, S. E. Maxwell, P. Rabl,
R. J. Schoelkopf, and P. Zoller. A coherent all-electrical interface be-
tween polar molecules and mesoscopic superconducting resonators. Nature
Physics, 2:636–642, 2006.
[136] D. I. Schuster, Lev S. Bishop, I. L. Chuang, D. DeMille, and R. J.
Schoelkopf. Cavity qed in a molecular ion trap. arXiv:0903.355, 2009.
[137] Jay Gambetta, W. A. Braff, A. Wallraff, S. M. Girvin, and R. J.
Schoelkopf. Protocols for optimal readout of qubits using a continuous
quantum nondemolition measurement. Phys. Rev. A, 76:012325, 2007.
[138] F. Mallet, F. Ong, A. Palacios-Laloy, F. Nguyen, P. Bertet, D. Vion, and
D. Esteve. Single-shot qubit readout for circuit quantum electrodynamics.
Nature Physics, (in press), 2009.
[139] A. Lupascu, S. Saito, T. Picot, P. C. De Groot, C. J. P. M. Harmans, and
J. E. Mooij. Quantum non-demolition measurement of a superconducting
two-level system. Nature Physics, 3:119–123, 2007.
[140] N Katz, M Ansmann, RC Bialczak, E Lucero, R McDermott, M Neeley,
M Steffen, EM Weig, AN Cleland, JM Martinis, and AN Korotkov. Co-
herent state evolution in a superconducting qubit from partial-collapse
measurement. Science, 312:1498–1500, 2006.
[141] Jay Gambetta, Alexandre Blais, M. Boissonneault, A. A. Houck, D. I.
Schuster, and S. M. Girvin. Quantum trajectory approach to circuit qed:
Quantum jumps and the zeno effect. Physical Review A, 77:012112, 2008.
[142] Matthias Steffen, M. Ansmann, Radoslaw C. Bialczak, N. Katz, Erik
Lucero, R. McDermott, Matthew Neeley, E. M. Weig, A. N. Cleland,
and John M. Martinis. Measurement of the entanglement of two super-
conducting qubits via state tomography. Science, 313:1423–1425, 2006.
[143] Markus Ansmann, H. Wang, Radoslaw C. Bialczak, Max Hofheinz, Erik
Lucero, M. Neeley, A. D. O’Connell, D. Sank, M. Weides, J. Wenner,
A. N. Cleland, and John M. Martinis. Violation of bell’s inequality in
josephson phase qubits. Nature, 461:504–506, 2009.
Circuit QED 105

[144] Lev S Bishop, L Tornberg, D Price, E Ginossar, A Nunnenkamp, A A

Houck, J M Gambetta, Jens Koch, G Johansson, S M Girvin, and R J
Schoelkopf. Proposal for generating and detecting multi-qubit ghz states
in circuit qed. NJP, 11:073040, 2009.

[145] S. Filipp, P. Maurer, P. J. Leek, M. Baur, R. Bianchetti, J. M. Fink,

M. Göppl, L. Steffen, J. M. Gambetta, A. Blais, and A. Wallraff. Two-
qubit state tomography using a joint dispersive readout. Physical Review
Letters, 102:200402, 2009.

[146] J. M. Chow, L. DiCarlo, J. M. Gambetta, A. Nunnenkamp, Lev S. Bishop,

L. Frunzio, M. H. Devoret, S. M. Girvin, and R. J. Schoelkopf. En-
tanglement metrology using a joint readout of superconducting qubits.
(arXiv:0908.1955).

[147] Matthew Neeley, Markus Ansmann, Radoslaw C. Bialczak, Max Hofheinz,

Erik Lucero, Aaron D. O’Connell, Daniel Sank, Haohua Wang, James
Wenner, Andrew N. Cleland, Michael R. Geller, and John M. Martinis.
Emulation of a Quantum Spin with a Superconducting Phase Qudit. Sci-
ence, 325:722–725, 2009.

[148] Matthias Steffen, M. Ansmann, R. McDermott, N. Katz, Radoslaw C.

Bialczak, Erik Lucero, Matthew Neeley, E. M. Weig, A. N. Cleland, and
John M. Martinis. State tomography of capacitively shunted phase qubits
with high fidelity. Phys. Rev. Lett., 97:050502, 2006.

[149] C Rigetti, A Blais, and M Devoret. Protocol for universal gates in opti-
mally biased superconducting qubits. Phys. Rev. Lett., 94:240502, 2005.

[150] P Bertet, CJPM Harmans, and JE Mooij. Parametric coupling for super-
conducting qubits. Phys. Rev. B, 73:064512, 2006.

[151] O Gywat, F Meier, D Loss, and DD Awschalom. Dynamics of coupled

qubits interacting with an off-resonant cavity. Phys. Rev. B, 73:125336,
2006.

[152] Alexandre Blais, Jay Gambetta, A. Wallraff, D. I. Schuster, S. M. Girvin,

M. H. Devoret, and R. J. Schoelkopf. Quantum-information processing
with circuit quantum electrodynamics. Phys. Rev. A, 75:032329, 2007.

[153] J. H. Plantenberg, P. C. de Groot, C. J. P. M. Harmans, and J. E. Mooij.

Demonstration of controlled-not quantum gates on a pair of supercon-
ducting quantum bits. Nature, 447:836–839, 2007.

[154] Yu. A. Pashkin, T. Yamamoto, O. Astafiev, Y. Nakamura, D. V. Averin,

and J. S. Tsai. Quantum oscillations in two coupled charge qubits. Nature,
421:823–826, 2003.
106 S.M. Girvin

[155] T. Yamamoto, Yu. A. Pashkin, O. Astafiev, Y. Nakamura, and J. S. Tsai.

Demonstration of conditional gate operation using superconducting charge
qubits. Nature, 425:941–944, 2003.

[156] R McDermott, RW Simmonds, M Steffen, KB Cooper, K Cicak, KD Os-

born, S Oh, DP Pappas, and JM Martinis. Simultaneous state measure-
ment of coupled Josephson phase qubits. Science, 307:1299–1302, 2005.

[157] J. B. Majer, F. G. Paauw, A. C. J. ter Haar, C. J. P. M. Harmans, and

J. E. Mooij. Spectroscopy on two coupled superconducting flux qubits.
Phys. Rev. Lett., 94:090501, 2005.

[158] Frederick W. Strauch, Philip R. Johnson, Alex J. Dragt, C. J. Lobb,

J. R. Anderson, and F. C. Wellstood. Quantum logic gates for coupled
superconducting phase qubits. Phys. Rev. Lett., 91:167005, 2003.

[159] Song Ke-Hui, Zhou Zheng-Wei, and Guo Guang-Can. Implementation of a

controlled-phase gate and Deutsch-Jozsa algorithm with superconducting
charge qubits in a cavity. Comm. Theor. Phys., 47:821–825, 2007.

[160] M. Wallquist, V. S. Shumeiko, and G. Wendin. Selective coupling of su-

perconducting charge qubits mediated by a tunable stripline cavity. Phys.
Rev. B, 74:224506, 2006.

[161] A. Palacios-Laloy, F. Nguyen, F. Mallet, P. Bertet, D. Vion, and D. Esteve.

Tunable resonators for quantum circuits. J. Low Temp. Phys., 151:1034–
1042, 2008.

[162] T. Hime, P. A. Reichardt, B. L. T. Plourde, T. L. Robertson, C.-E.

Wu, A. V. Ustinov, and John Clarke. Solid-State Qubits with Current-
Controlled Coupling. Science, 314:1427–1429, 2006.

[163] A. O. Niskanen, K. Harrabi, F. Yoshihara, Y. Nakamura, S. Lloyd, and

J. S. Tsai. Quantum Coherent Tunable Coupling of Superconducting
Qubits. Science, 316:723–726, 2007.

[164] S.M. Girvin and R.J. Schoelkopf. (unpublished).

[165] M. Dobsicek, G. Johansson, V.S. Shumeiko, and G. Wendin:. Arbitrary

accuracy iterative phase estimation algorithm as a two qubit benchmark.
Phys. Rev. A, 76:030306(R), 2007.

[166] L. F. Wei, Yu-xi Liu, and Franco Nori. Generation and control
of greenberger-horne-zeilinger entanglement in superconducting circuits.
Phys. Rev. Lett., 96:246803, 2006.

[167] Dimitris I. Tsomokos, Sahel Ashhab, and Franco Nori. Fully connected
network of superconducting qubits in a cavity. NJP, 10:113020, 2008.
Circuit QED 107

[168] Andrei Galiautdinov and John M. Martinis. Maximally entangling tripar-

tite protocols for josephson phase qubits. Phys. Rev. A, 78:010305(R),
2008.

[169] I. L. Chuang and M. A. Nielsen. Quantum Information and Quantum

Computation. Cambridge University Press, 2000.

[170] L. Tornberg, M. Wallquist, G. Johansson, V.S. Shumeiko, and G. Wendin.

Implementation of the three-qubit phase-flip error correction code with
superconducting qubits. Phys. Rev. B 77, 77:214528, 2008.

[171] A.Yu. Kitaev. Fault-tolerant quantum computation by anyons. Ann.

Phys., 303:2–30, 2003.

[172] L. B. Ioffe and M. V. Feigelman. Possible realization of an ideal quantum

computer in josephson junction array. Phys. Rev. B, 66:224503, 2002.

[173] L. B. Ioffe, M. V. Feigel’man, A. Ioselevich, D. Ivanov, M. Troyer, and

G. Blatter. Topologically protected quantum bits using josephson junction
arrays. Nature, 415:503–506, 2002.

[174] B. Doucot, M. V. Feigelman, and L. B. Ioffe. Topological order in the

insulating josephson junction arrays. Phys. Rev. Lett., 90:107003, 2003.

[175] B. Doucot, M. V. Feigelman, L. B. Ioffe, and A. S. Ioselevich. Protected

qubits and chern-simons theories in josephson junction arrays. Phys. Rev.
B, 71:024505, 2005.

[176] Sergey Gladchenko, David Olaya, Eva Dupont-Ferrier, Benoit Doucot,

Lev B. Ioffe, and Michael E. Gershenson. Superconducting nanocircuits
for topologically protected qubits. Nature Physics, 5:48–53, 2009.

[177] C. D. Ogden, E. K. Irish, and M. S. Kim. Dynamics in a coupled-cavity

array. Phys. Rev. A, 78:063805, 2008.

[178] F. Helmer, M. Mariantoni, A. G. Fowler, J. von Delft, E. Solano, and

F. Marquardt. Cavity grid for scalable quantum computation with super-
conducting circuits. EPL, 85:50007, 2009.

[179] Matthew P. A. Fisher, Peter B. Weichman, G. Grinstein, and Daniel S.

Fisher. Boson localization and the superfluid-insulator transition. Phys.
Rev. B, 40:546–570, 1989.

[180] Andrew D. Greentree, Charles Tahan, Jared H. Cole, and Lloyd C. L.

Hollenberg. Quantum phase transitions of light. Nature Physics, 2:856–
861, 2006.

[181] Fabrizio Illuminati. Quantum optics: Light does matter. Nature Physics,
2:803–804, 2006.
108 S.M. Girvin

[182] Michael J. Hartmann and Martin B. Plenio. Strong photon nonlinearities

and photonic mott insulators. Phys. Rev. Lett., 99:103601, 2007.
[183] T. C. Jarrett, A. Olaya-Castro, and N. F. Johnson. Optical signatures of
quantum phase transitions in a light-matter system. EPL, 77:34001, 2007.
[184] Davide Rossini and Rosario Fazio. Mott-insulating and glassy phases of
polaritons in 1d arrays of coupled cavities. Phys. Rev. Lett., 99:186401,
2007.
[185] M J Hartmann, F G S L Brandao, and M B Plenio. A polaritonic two-
component bose-hubbard model. NJP, 10:033011 (12pp), 2008.
[186] M. I. Makin, Jared H. Cole, Charles Tahan, Lloyd C. L. Hollenberg, and
Andrew D. Greentree. Quantum phase transitions in photonic cavities
with two-level systems. Phys. Rev. A, 77:053819, 2008.
[187] Markus Aichhorn, Martin Hohenadler, Charles Tahan, and Peter B. Little-
wood. Quantum fluctuations, temperature, and detuning effects in solid-
light systems. Phys. Rev. Lett., 100:216401, 2008.
[188] Jaeyoon Cho, Dimitris G. Angelakis, and Sougato Bose. Simulation of
high-spin heisenberg models in coupled cavities. Phys. Rev. A, 78:062338,
2008.
[189] Jaeyoon Cho, Dimitris G. Angelakis, and Sougato Bose. Fractional quan-
tum hall state in coupled cavities. Phys. Rev. Lett., 101:246809, 2008.
[190] J. Zhao, A. W. Sandvik, and K. Ueda. Insulator to superfluid transition
in coupled photonic cavities in two dimensions. arXiv:0806.3603, 2008.
[191] Neil Na, Shoko Utsunomiya, Lin Tian, and Yoshihisa Yamamoto. Strongly
correlated polaritons in a two-dimensional array of photonic crystal mi-
crocavities. Phys. Rev. A, 77:031803, 2008.
[192] Soi-Chan Lei and Ray-Kuang Lee. Quantum phase transitions of light in
the dicke-bose-hubbard model. Phys. Rev. A, 77:033827, 2008.
[193] An-Chun Ji, Qing Sun, X. C. Xie, and W. M. Liu. Josephson effect for
photons in two weakly linked microcavities. Phys. Rev. Lett., 102:023602,
2009.
[194] J. Hartmann, F.G.S.L. Brandao, and M.B. Plenio. Quantum many-body
phenomena in coupled cavity arrays. Laser & Photonics Review, 2:527–
556, 2009.
[195] Michal Grochol. Quantum phase transitions in an array of coupled
nanocavity quantum dots. Phys. Rev. B, 79:205306, 2009.
[196] Denis Dalidovich and Malcolm P. Kennett. Bose-hubbard model in the
presence of ohmic dissipation. Phys. Rev. A, 79:053611, 2009.
Circuit QED 109

[197] I. Carusotto, D. Gerace, H. E. Tureci, S. De Liberato, C. Ciuti, and

A. Imamoglu. Fermionized photons in an array of driven dissipative non-
linear cavities. Phys. Rev. Lett., 103:033601, 2009.
[198] S. Schmidt and G. Blatter. Strong coupling theory for the jaynes-
cummings-hubbard model. arXiv:0905.3344, 2009.
[199] Jens Koch and Karyn Le Hur. Superfluid–mott insulator transi-
tion of light in the jaynes-cummings lattice. Phys. Rev. A, (in
press):arXiv.org:0905.4005, 2009.