Quantum Nonlinearities in Circuit QED

Circuit QED offers the possibility to realize an exceptionally strong coupling

between artificial atoms – individual superconducting qubits – and single
microwave photons in a one dimensional waveguide resonator. This new
solid state approach to investigate the matter-light interaction on the level
of individual quanta also represents a promising hardware architecture for
the realization of a scalable quantum information processor. We first
summarize and review the important theoretical and experimental
principles of this rapidely developing field and give an overview of the
relevant literature. Here the focus is to provide a detailed discussion of the
experimental ultra-low temperature measurement setup and the design,
fabrication and characterization of superconducting Josephson junction
devices. Presented key experimental results include: 1) The first clear
observation of the Jaynes-Cummings nonlinearity in the vacuum Rabi
spectrum by measuring both two- and three-photon dressed states. 2) The
observation of strong collective coupling for up to 3 qubits. 3) A
quantitative investigation of the quantum to classical transition induced by
an elevated temperature environment.

Johannes M. Fink

Johannes M. Fink
Quantum Nonlinearities in
The author is a post doctoral scientist at the Swiss
Federal Institute of Technology Zurich. He studied in
Strong Coupling Circuit QED
Vienna, Sydney and at ETH Zurich where he earned a
Ph.D. in physics. His research is focused on designing,
On-Chip Quantum Optics with Microwave Photons
fabricating and controlling coherent devices for and Superconducting Electronic Circuits
quantum optics and quantum information processing
applications.
Johannes M. Fink

978-3-8454-1971-8
Q UANTUM N ONLINEARITIES IN S TRONG
C OUPLING C IRCUIT QED

Johannes M. Fink
For Pia and Alwin
and for Stephanie
 C ONTENTS

Abstract ix

List of publications xi

I Basic Concepts 1

1 Introduction 3
1.1 QED in cavities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Outline of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Review and Theory 7

2.1 On-chip microwave cavity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.1 Coplanar waveguide resonator . . . . . . . . . . . . . . . . . . . . . 9
2.1.2 Transmission matrix model . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.3 Circuit quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Superconducting quantum bits . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.1 Charge qubits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2.2 Transmon regime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.3 Spin-1/2 notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3 Matter – light coupling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.1 Atom-field interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.2 Jaynes-Cummings model . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.3 Transmon – photon coupling . . . . . . . . . . . . . . . . . . . . . . . 20
2.3.4 Dispersive limit: Qubit control and readout . . . . . . . . . . . . . . 22

3 Conclusion 25

II Experimental Principles 27

4 Measurement Setup 31
4.1 Microchip sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.1.1 Superconducting circuit . . . . . . . . . . . . . . . . . . . . . . . . . . 33

v
vi CONTENTS

4.1.2 Printed circuit board . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

4.1.3 Wire bonds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.1.4 Sample mount, coils and shielding . . . . . . . . . . . . . . . . . . . 35
4.2 Signal synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2.1 Microwave generation, control and synchronization . . . . . . . . . 37
4.2.2 Quadrature modulation . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2.3 Generation of Gaussian noise . . . . . . . . . . . . . . . . . . . . . . 38
4.3 Cryogenic setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.3.1 Dilution refrigerator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.3.2 Microwave input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.3.3 Heat loads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.3.4 DC bias input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.3.5 Microwave output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.4 Data acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

5 Design and Fabrication of Josephson Junction Devices 49

5.1 Qubit design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.1.1 Charging energy and voltage division . . . . . . . . . . . . . . . . . 49
5.1.2 Josephson energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.2 Fabrication process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.2.1 Chip preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.2.2 Electron beam lithography . . . . . . . . . . . . . . . . . . . . . . . . 56
5.2.3 Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.2.4 Shadow evaporation and oxidation . . . . . . . . . . . . . . . . . . . 63
5.2.5 Resist stripping and liftoff . . . . . . . . . . . . . . . . . . . . . . . . 65
5.3 Room-temperature characterization . . . . . . . . . . . . . . . . . . . . . . . 66
5.3.1 DC resistance measurements . . . . . . . . . . . . . . . . . . . . . . 66
5.3.2 Junction aging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

6 Sample Characterization 71
6.1 Resonator spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
6.2 Flux periodicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
6.3 Dipole coupling strengths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
6.4 Qubit spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
6.4.1 Charging energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.4.2 Flux dependence, dressed states and number splitting . . . . . . . 77
6.4.3 AC-Stark shift photon-number calibration . . . . . . . . . . . . . . . 79
6.5 Rabi oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

7 Conclusion 83

III Main Results 85

p
8 Observation of the Jaynes-Cummings n Nonlinearity 89
CONTENTS vii

8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
8.1.1 The Jaynes-Cummings model . . . . . . . . . . . . . . . . . . . . . . 90
8.1.2 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
8.2 Coherent dressed states spectroscopy . . . . . . . . . . . . . . . . . . . . . 92
8.2.1 Vacuum Rabi splitting . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
8.2.2 Two photon vacuum Rabi splitting . . . . . . . . . . . . . . . . . . . . 94
8.3 Weak thermal excitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
8.3.1 Sample parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
8.3.2 Generalized Jaynes-Cummings model . . . . . . . . . . . . . . . . . 97
8.3.3 Thermal background field . . . . . . . . . . . . . . . . . . . . . . . . 98
8.3.4 Quasi-thermal excitation . . . . . . . . . . . . . . . . . . . . . . . . . 100
8.4 Three photon pump and probe spectroscopy of dressed states . . . . . . . 103
8.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

9 Collective Multi-Qubit Interaction 107

9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
9.1.1 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
9.1.2 Tavis-Cummings model . . . . . . . . . . . . . . . . . . . . . . . . . . 109
9.2 Collective dipole coupling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
9.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

10 Quantum-to-Classical Transition 115

10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
10.1.1 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
10.1.2 Vacuum Rabi splitting . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
10.2 Strong quasi-thermal excitation . . . . . . . . . . . . . . . . . . . . . . . . . . 117
10.2.1 The high temperature limit . . . . . . . . . . . . . . . . . . . . . . . . 117
10.2.2 Quantitative model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
10.2.3 Time domain measurements . . . . . . . . . . . . . . . . . . . . . . 120
10.2.4 Extraction of the effective field temperature . . . . . . . . . . . . . . 120
10.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

11 Conclusion and Prospects 123

IV Appendices 127

A Aspects of the Cryogenic Setup 129

A.1 Dimensioning He vacuum pumping lines . . . . . . . . . . . . . . . . . . . . 129
A.2 Mechanical vibration measurements . . . . . . . . . . . . . . . . . . . . . . . 132
A.3 Double gimbal vibration isolation . . . . . . . . . . . . . . . . . . . . . . . . . 134

B Nano Fabrication Recipes 137

B.1 Chip preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
viii CONTENTS

B.2 Electron beam lithography . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

B.3 Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
B.4 Shadow evaporation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
B.5 Resist stripping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

Bibliography I

Acknowledgements XIX
 A BSTRACT

The fundamental interaction between matter and light can be studied in cavity quantum
electrodynamics (QED). If a single atom and a single photon interact in a cavity resonator
where they are well isolated from the environment, the coherent dipole coupling can
dominate over any dissipative effects. In this strong coupling limit the atom repeatedly
absorbs and emits a single quantum of energy. The photon and the atom loose their
individual character and the new eigenstates are quantum superpositions of matter and
light – so called dressed states.
Circuit QED is a novel on-chip realization of cavity QED. It offers the possibility to real-
ize an exceptionally strong coupling between artificial atoms – individual superconducting
qubits – and single microwave photons in a one dimensional waveguide resonator. This
new solid state approach to investigate the matter-light interaction enables to carry out
novel quantum optics experiments with an unprecedented degree of control. Moreover,
multiple superconducting qubits coupled via intra-cavity photons – a quantum bus – are
a promising hardware architecture for the realization of a scalable quantum information
processor.
In this thesis we study in detail a number of important aspects of the resonant in-
teraction between microwave photons and superconducting qubits in the context of cav-
ity QED:
p
We report the long sought for spectroscopic observation of the n nonlinearity of the
Jaynes-Cummings energy ladder where n is the number of excitations in the resonantly
p
coupled matter-light system. The enhancement of the matter-light coupling by n in
the presence of additional photons was already predicted in the early nineteen sixties.
It directly reveals the quantum nature of light. In multi-photon pump and probe and in
elevated temperature experiments we controllably populate one or multi-photon / qubit
superposition states and probe the resulting vacuum Rabi transmission spectrum. We
find that the multi-level structure of the superconducting qubit renormalizes the energy
levels, which is well understood in the framework of a generalized Jaynes-Cummings
model. The observed very strong nonlinearity on the level of single or few quanta could
be used for the realization of a single photon transistor, parametric down-conversion, and
for the generation and detection of individual microwave photons.
In a dual experiment we have performed measurements with up to three indepen-
dently tunable qubits to study cavity mediated multi-qubit interactions. By tuning the
qubits in resonance with the cavity field one by one, we demonstrate the enhancement of

ix
x ABSTRACT

p
the collective dipole coupling strength by N , where N is the number of resonant atoms,
as predicted by the Tavis-Cummings model. To our knowledge this is the first observation
of this nonlinearity in a system in which the atom number can be changed one by one in
a discrete fashion. In addition, the energies of both bright and dark coupled multi-qubit
photon states are well explained by the Tavis-Cummings model over a wide range of de-
tunings. On resonance we observe all but two eigenstates to be dark states, which do
not couple to the cavity field. The bright states on the other hand are an equal superpo-
sition of a cavity photon and a multi-qubit Dicke state with an excitation equally shared
among the N qubits. The presented approach may enable novel investigations of super-
and sub-radiant states of artificial atoms and should allow for the controlled generation of
entangled states via collective interactions, not relying on individual qubit operations.
Finally, we study the continuous transition from the quantum to the classical limit of
cavity QED. In order to access the quantum and the classical regimes we control and
sense the thermal photon number in the cavity over five orders of magnitude and extract
the field temperature given by Planck’s law in one dimension using both spectroscopic
and time-resolved vacuum Rabi measurements. In the latter we observe the coherent
exchange of a quantum of energy between the qubit and a variable temperature thermal
field. In the classical limit where the photon occupation of the cavity field is large, the
quantum nonlinearity is small compared to any coupling rates to the environment. Here
the signature of quantization vanishes and the system’s response is indistinguishable
from the response of a classical harmonic oscillator. The observed transition from quan-
tum mechanics to classical physics illustrates the correspondence principle of quantum
physics as introduced by Niels Bohr. The emergence of classical physics from quantum
mechanics and the role of decoherence in this process is an important subject of current
research. In future experiments entanglement and decoherence at elevated tempera-
tures can be studied in the context of quantum information.
 L IST OF PUBLICATIONS

This thesis is based in part on the following journal articles:

1. J. M. Fink, M. Göppl, M. Baur, R. Bianchetti, P. J. Leek, A. Blais and A. Wallraff.

p
Climbing the Jaynes-Cummings ladder and observing its n nonlinearity in a cavity
QED system. Nature 454, 315-318 (2008)

2. J. M. Fink, R. Bianchetti, M. Baur, M. Göppl, L. Steffen, S. Filipp, P. J. Leek, A. Blais,

A. Wallraff. Dressed collective qubit states and the Tavis-Cummings model in circuit
QED. Physical Review Letters 103, 083601 (2009)

3. J. M. Fink, M. Baur, R. Bianchetti, S. Filipp, M. Göppl, P. J. Leek, L. Steffen, A. Blais

and A. Wallraff. Thermal excitation of multi-photon dressed states in circuit quan-
tum electrodynamics. Proceedings of the Nobel Physics Symposium on Qubits for
Future Quantum Computers. Physica Scripta T137, 014013 (2009)

4. J. M. Fink, L. Steffen, P. Studer, L. S. Bishop, M. Baur, R. Bianchetti, D. Bozyigit,

C. Lang, S. Filipp., P. J. Leek and A. Wallraff. Quantum-to-classical transition in
cavity quantum electrodynamics. Physical Review Letters 105, 163601 (2010)

In addition, contributions have been made to the following articles:

5. P. J. Leek, J. M. Fink, A. Blais, R. Bianchetti, M. Göppl, J. M. Gambetta, D. I. Schus-

ter, L. Frunzio, R. J. Schoelkopf and A. Wallraff. Observation of Berry’s phase in a
solid state qubit. Science 318, 1889 (2007)

6. M. Göppl, A. Fragner, M. Baur, R. Bianchetti, S. Filipp, J. M. Fink, P. J. Leek,

G. Puebla, L. Steffen, A. Wallraff. Coplanar waveguide resonators for circuit quan-
tum electrodynamics. Journal of Applied Physics 104, 113904 (2008)

7. A. Fragner, M. Göppl, 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 (2008)

8. S.Filipp, P. Maurer, P. J. Leek, M. Baur, R. Bianchetti, J. M. Fink, M. Göppl, L. Stef-

fen, J. M. Gambetta, A. Blais, A. Wallraff. Two-qubit state tomography using a joint
dispersive read-out. Physical Review Letters 102, 200402 (2009)

xi
xii LIST OF PUBLICATIONS

9. P. J. Leek, S. Filipp, P. Maurer, M. Baur, R. Bianchetti, J. M. Fink, M. Göppl, L. Stef-

fen, A. Wallraff. Using sideband transitions for two-qubit operations in supercon-
ducting circuits. Physical Review B (Rapid Communications) 79, 180511(R) (2009)

10. M. Baur, S. Filipp, R. Bianchetti, J. M. Fink, M. Göppl, L. Steffen, P. J. Leek, A. Blais,

A. Wallraff. Measurement of Autler-Townes and Mollow transitions in a strongly
driven superconducting qubit. Physical Review Letters 102, 243602 (2009)

11. R. Bianchetti, S. Filipp, M. Baur, J. M. Fink, M. Göppl, P. J. Leek, L. Steffen, A. Blais

and A. Wallraff. Dynamics of dispersive single qubit read-out in circuit quantum
electrodynamics. Physical Review A 80, 043840 (2009)

12. P. J. Leek, M. Baur, J. M. Fink, R. Bianchetti, L. Steffen, S. Filipp and A. Wall-

raff. Cavity QED with separate photon storage and qubit readout modes. Physical
Review Letters 104, 100504 (2010)

13. R. Bianchetti, S. Filipp, M. Baur, J. M. Fink, C. Lang, L. Steffen, M. Boissonneault,

A. Blais and A. Wallraff. Control and tomography of a three level superconducting
artificial atom. Physical Review Letters 105, 223601 (2010)

14. D. Bozyigit, C. Lang, L. Steffen, J. M. Fink, C. Eichler, M. Baur, R. Bianchetti,

P. J. Leek, S. Filipp, M. P. da Silva, A. Blais, and A. Wallraff. Correlation measure-
ments of individual microwave photons emitted from a symmetric cavity. Journal of
Physics: Conference Series 264, 012024 (2011)

15. D. Bozyigit, C. Lang, L. Steffen, J. M. Fink, C. Eichler, M. Baur, R. Bianchetti,

P. J. Leek, S. Filipp, M. P. da Silva, A. Blais, and A. Wallraff. Antibunching of
microwave frequency photons observed in correlation measurements using linear
detectors. Nature Physics 7, 154 (2011)

16. C. Eichler, D. Bozyigit, C. Lang, L. Steffen, J. M. Fink, and A. Wallraff. Experimental

tomographic state reconstruction of itinerant microwave photons. Physical Review
Letters 106, 220503 (2011)

17. C. Lang, D. Bozyigit, C. Eichler, L. Steffen, J. M. Fink, A. A. Abdumalikov Jr.,

M. Baur, S. Filipp, M. P. da Silva, A. Blais, and A. Wallraff. Observation of res-
onant photon blockade at microwave frequencies using correlation function mea-
surements. Physical Review Letters 106, 243601 (2011)

18. S. Filipp, M. Göppl, J. M. Fink, M. Baur, R. Bianchetti, L. Steffen, and A. Wall-

raff. Multimode mediated qubit-qubit coupling and dark-state symmetries in circuit
quantum electrodynamics. Physical Review A 83, 063827 (2011)
 PA R T I

BASIC C ONCEPTS

1
 PTER
1
CHA
I NTRODUCTION

A first formulation of the quantum theory of light and matter was presented by Paul
Dirac in the late 1920’s [Dirac27, Fermi32]. He described the quantized electromagnetic
field as an ensemble of harmonic oscillators and included interactions with electrically
charged particles to compute the coefficient of spontaneous emission of radiation by an
atom. Modern quantum electrodynamics (QED) has been pioneered by Richard Feyn-
man, Freeman Dyson, Julian Schwinger and Sin-Itiro Tomonaga back in the late 1940’s
[Cohen-Tannoudji89]. They brought together the ideas of quantum mechanics, classical
electrodynamics and special relativity and formulated a consistent relativistic quantum
field theory (QFT) of electrodynamics. QED mathematically describes all phenomena
involving charged particles by means of exchange of photons where both fields and par-
ticles are treated as discrete excitations of fields that are called field quanta. QED is
among the most successful general and stringently tested theories to date. The success
of QED also stimulated further successful QFTs such as quantum chromodynamics.
Modern quantum optics is the result of a fruitful union of QED and advanced ex-
perimental optical techniques [Feynman71, Gardiner91, Walls94, Mandel95, Scully97,
Yamamoto99]. A remarkable example is the observation of the Lamb shift by Willis Lamb
and Robert Curtis Retherford in 1947 [Lamb47, Bethe47]. The measured tiny difference
between two degenerate energy levels of the hydrogen atom reveals its interaction with
the virtual excitations of the vacuum field, which is a consequence of field quantization
in QED. Lamb’s observation was the precursor and stimulus for both modern QED which
uses the concept of renormalization to eliminate vacuum induced infinite level shifts and
the many spectacular quantum optics experiments that followed.
In spite of the close connection of QED and quantum optics many important phe-
nomena such as stimulated emission, resonance fluorescence, lasing, holography and

3
4 CHAPTER 1. INTRODUCTION

even the photoelectric effect [Wentzel26, Wentzel27] can be explained with a semiclas-
sical theory where matter is treated quantum mechanically and radiation is described
according to Maxwell’s equations. Even the Lamb shift, the Planck distribution of black-
body radiation and the linewidth of a laser can be explained semiclassically if vacuum
fluctuations are added stochastically to the otherwise classical field [Boyer75].
While the semiclassical understanding has remained a tempting concept in quan-
tum optics, more advanced experimental methods allowed for many intriguing obser-
vations that unavoidably require the concept of the photon. In particular the obser-
vation of photon correlations by Hanbury Brown and Twiss [Brown56] and the subse-
quent systematic theoretical treatment of the quantum and classical properties of light
[Glauber63b, Glauber63a] paved the way for modern quantum optics. In the same year,
Jaynes and Cummings as well as Paul presented solutions for the interaction of a single
atom and quantized radiation [Jaynes63, Paul63], an important theoretical model which
today is routinely being studied in the field of cavity QED.
Observations that require the full machinery of the quantum theory of radiation in
their explanation are non-classical photon correlations [Clauser74], photon antibunch-
ing [Kimble77], single photon anticorrelations [Grangier86], photon-photon interference
[Ghosh87] and quantum beats [Scully97]. Moreover, early QED experiments in cavities
showed collapse and revival of quantum coherence [Eberly80, Rempe87, Brune96] as
a result of the predicted nonlinearity of the atom-photon coupling strength. The direct
spectroscopic observation of this nonlinearity [Schuster08, Fink08] represents the first
major result of this thesis, see Chapter 8. We also just reported the first demonstration of
microwave frequency photon antibunching [Bozyigit10b].

1.1 QED in cavities

In cavity QED [Haroche89, Raimond01, Mabuchi02, Haroche06, Walther06, Ye08] atoms
are coupled to photons inside a cavity with highly reflective walls. The cavity acts like a
high finesse resonator which supports only a few narrow line-width standing wave modes.
The mirrors therefore modify the mode structure of the cavity field and very effectively
protect the intra-cavity atom from the electromagnetic environment at all off-resonant
frequencies. Spontaneous emission is suppressed at these frequencies because the
density of modes at which the atom could radiate is even inhibited for the occurrence
of vacuum fluctuations. At the resonance frequency however, the mode density is large
and the photon storage time is enhanced because an entering photon is reflected at the
cavity mirrors many times before it leaves the cavity. Cavity QED therefore represents an
open quantum system who’s coupling to the environment can be engineered by choosing
an appropriate reflectivity of the cavity mirrors. The cavity can be designed such that
the mode volume is decreased considerably compared to free space. This leads to an
increased electromagnetic field E for a given intra-cavity photon number n and in turn
to an increased single atom - photon coupling. Small-volume large-quality resonators
therefore enable to observe the coherent dynamics of a two state system, such as an
atom, coupled to individual energy quanta of a single electromagnetic field mode.
1.2. OUTLINE OF THE THESIS 5

This situation is well modeled by a spin which can only absorb and emit a sin-
gle quantum of energy coupled to a harmonic oscillator which can give and receive
an arbitrary number of quanta. In the electric dipole approximation, which is valid
for long wavelengths compared to the size of the atom, and the appropriate gauge
[Göppert-Mayer31, Cohen-Tannoudji89] the electric dipole coupling is given by g = Ed
with the dipole moment of the atom to first order d = er with charge e and dipole size r.
If g À κ, γ with κ the rate at which photons leave the cavity and γ the rate at which the
qubit looses its excitation to modes other than the resonator mode, the strong coupling
regime of cavity QED is reached. In this limit, single photon coherent matter-light dynam-
ics can be observed. The repeated coherent exchange of a field quantum between the
(microscopic) atom and the (macroscopic) cavity is called vacuum Rabi oscillations.
In circuit QED [Blais04, Wallraff04, Schoelkopf08], where the atom is replaced by a
superconducting circuit with an atom-like energy spectrum and the cavity by an electrical
microwave resonator, the dipole coupling g is spectacularly increased due to two effects.
Firstly, using an effectively one dimensional microwave resonator two of the three spatial
dimensions are made much smaller than the wavelength of the field which decreases the
mode volume dramatically and typically leads to an enhancement of up to ∼ 103 of the
electric field per photon. Secondly, the effective dipole moment of the superconducting
circuit (a transmon type quibt) can be increased by a factor of up to ∼ 103 compared to
the largest known dipole moment atoms, e.g. Hydrogen atoms prepared in large quantum
number Rydberg states, simply because of their macroscopic size.
Circuit QED allows to study in detail all facets of strong coupling cavity QED in a
very controlled fashion and offers a number of new opportunities. The properties of artifi-
cial atoms can be engineered during fabrication, some of their parameters can be in-situ
tuned and they remain fixed in space which implies a constant coupling strength. Circuit
QED therefore has become a realistic platform to engineer a quantum information pro-
cessor [Nielsen00, Ladd10] in the solid state. It solves an important task required in many
implementations of quantum computation [Tanzilli05, Spiller06, Wilk07], i.e. it provides a
fast interface between fixed solid state qubits with long coherence and quick operation
times and flying qubits (photons) which can be sent to another circuit or chip [Wallraff04,
Houck07, Majer07, Fink08, Hofheinz08, Hofheinz09, DiCarlo09, Bozyigit10c]. Moreover,
circuit QED employs only conventional lithographic fabrication methods and in principle
is based on a scalable hardware architecture that allows for an integrated design.

1.2 Outline of the thesis

We start with a general review and introduction to the basic concepts of circuit QED
in Chapter 2. The microwave resonator and the superconducting charge qubit are the
main building blocks of this on-chip cavity QED setup. The quantum description of these
electrical circuits is discussed in Sections 2.1 and 2.2. The physics of their interaction is
well described by a generalized Jaynes-Cummings Hamiltonian. This model is introduced
step by step in Section 2.3. In addition, the dispersive limit which is particularly relevant
for qubit state readout and coherent qubit state control is discussed.
6 CHAPTER 1. INTRODUCTION

After this theoretical part (Part I) we discuss in detail the required experimental tech-
niques to successfully perform state of the art circuit QED experiments in Part II. We
present the measurement setup built and used to characterize numerous resonator and
qubit samples in Chapter 4. For additional technical aspects of the cryogenic microwave-
frequency measurement setup refer to Appendix A. The design and nano-fabrication of
Josephson junction devices such as transmon qubits, tunable resonators and parametric
amplifiers is yet another important part of this thesis and will be discussed in Chapter 5.
For detailed fabrication recipes refer to Appendix B. At the end of this part, in Chapter 6,
we discuss a typical set of important circuit QED sample characterization measurements.
Before more sophisticated measurements can be envisaged, these experiments need to
be conducted very carefully to determine the individual sample parameters every time a
new sample is cooled down.
In the main part of this thesis (Part III) we quantitatively investigate the dependence of
the large single-photon single-atom dipole coupling strength g = Ed on the photon num-
ber n ∝ E2 in Chapter 8 and the number of qubits N ∝ d2 in Chapter 9. In the presented
spectroscopic vacuum Rabi splitting measurements we p observe strong nonlinearities on
p
the level of individual quanta. The observed n and N nonlinearities provide direct
evidence for the quantization of the electromagnetic field and for the collective dynamics
of up to three qubits entangled by the single photon. In Chapter 10 we study the strong
coupling regime at large effective temperatures of the resonator field. As the number of
thermal photons is increased, the signature of quantization is lost as expected from Bohr’s
correspondence principle. We quantitatively analyze this quantum-to-classical transition
and demonstrate how to sense the effective cavity field temperature over a large range
using both spectroscopic and time-resolved vacuum Rabi measurements.
 PTER
2
CHA
R EVIEW AND T HEORY

Research in circuit QED broadly makes connections to the fields of quantum optics,
atomic physics, quantum information science and solid state physics. As a result of the
remarkably rapid progress during the last decade it has evolved from a theoretical idea
to an established platform of quantum engineering.
The observation of coherent dynamics of a superconducting charge qubit by Y. Naka-
mura et al. in 1999 [Nakamura99] inspired many physicists to think about systems in
which electromagnetic oscillators and superconducting qubits are coupled to form a novel
type of solid state cavity QED system. It was suggested to use discrete LC circuits
[Buisson01, Makhlin01], large junctions [Marquardt01, Plastina03, Blais03] and three di-
mensional cavities [Al-Saidi01, Yang03, You03] to observe coherent dynamics of qubits
and photons. In 2004 A. Blais et al. [Blais04] suggested to couple a superconducting
charge qubit to a coplanar waveguide resonator in order to realize an on-chip cavity QED
system, see Fig. 2.1. It promised to be suitable to study strong coupling quantum optics
and also has the potential for the successful implementation of a solid state quantum
computer [Blais07]. Based on this proposal A. Wallraff et al. [Wallraff04] observed vac-
uum Rabi mode splitting, a hallmark experiment demonstrating strong coupling between
a single qubit and a single photon for the first time in 2004. Coherent sideband oscillations
between a flux qubit and an LC-resonator were also observed at that time [Chiorescu04].
For two introductory reviews on circuit QED see [Schoelkopf08, Wallraff08].
The observation of strong coupling was the outset for a remarkable number of
novel solid state quantum optics and quantum computing experiments. A list of
the important milestones is given here: observation of the quantum AC-Stark shift
[Schuster05], demonstration of high qubit readout fidelities [Wallraff05], observation of
vacuum Rabi oscillations [Johansson06] and photon number splitting of the qubit spec-

7
8 CHAPTER 2. REVIEW AND THEORY

trum [Schuster07b], demonstration of a single photon source [Houck07], cavity side-

band transitions [Wallraff07] and single qubit lasing [Astafiev07], observation of Berry’s
phase [Leek07], coupling of two qubits via a cavity bus [Majer07, Sillanpää07], ob-
p
servation of the n nonlinearity of the Jaynes-Cummings ladder [Fink08], observa-
tion of the Lamb shift [Fragner08], cooling and amplification with a qubit [Grajcar08],
controlled symmetry breaking in circuit QED [Deppe08], generation of Fock states
[Hofheinz08] and arbitrary superpositions of Fock states [Hofheinz09], observation of
collective states of up to 3 qubits [Fink09b], observation of Autler-Towns and Mollow
transitions [Baur09], high drive power nonlinear spectroscopy of the vacuum Rabi res-
onance [Bishop09], demonstration of two qubit entanglement using sideband transi-
tions [Leek09], demonstration of gates and basic two qubit quantum computing algo-
rithms [DiCarlo09], violation of Bell’s inequality [Ansmann09], demonstration of single
shot qubit readout [Mallet09], implementation of separate photon storage and qubit read-
out modes [Leek10], measurement of the quantum-to-classical transition and thermal
field sensing in cavity QED [Fink10], quantum non-demolition detection of single mi-
crowave photons [Johnson10], implementation of optimal qubit control pulse shaping
[Motzoi09, Chow10a, Lucero10], preparation and generation of highly entangled 2 and
3-qubit states [Chow10b, Neeley10, DiCarlo10] and the first measurement of microwave
frequency photon antibunching [Bozyigit10c, Bozyigit10b] using linear amplifiers and on-
chip beam splitters.
Similarly, strong interactions have also been observed between superconducting
qubits and freely propagating photons in microwave transmission lines. This includes
the observation of resonance fluorescence [Astafiev10a], quantum limited amplification
[Astafiev10b] and electromagnetically induced transparency [Abdumalikov10] with a sin-
gle artificial atom. The rapid advances in circuit QED furthermore inspired and enabled
the demonstration of single phonon control of a mechanical resonator passively cooled
to its quantum ground state [O´Connell10].

m
9m
L=1

Figure 2.1: Schematic of an experimental cavity QED (a) and circuit QED (b) setup. a, Optical analog of
circuit QED. A two-state atom (violet) is coupled to a cavity mode (red). b, Schematic of the investigated circuit
QED system. The coplanar waveguide resonator is shown in light blue, the transmon qubit in violet and the first
harmonic of the standing wave electric field in red. Typical dimensions are indicated.
2.1. ON-CHIP MICROWAVE CAVITY 9

We will now review the basics of circuit QED using transmon type charge qubits and
coplanar waveguide resonators.

2.1 On-chip microwave cavity

Most circuit QED setups are using 1D transmission line resonators with quality factors
reaching Q ∼ 105 − 106 as a cavity. Coplanar waveguide (CPW) resonators can be fabri-
cated with a simple single layer photo-lithographic process using gap or finger capacitors
to couple to input and output transmission lines, see Figs. 2.1 and 2.2. The coplanar
geometry resembles a coaxial line with the ground in the same plane as the center con-
ductor. CPWs allow to create well localized fields in one region of the chip, e.g. where the
qubit is positioned, and less intense fields in another region, e.g. where the dimensions
should match with the printed circuit board (PCB). In contrast to circuits based on mi-
crostrip lines, this is possible at a fixed impedance simply by maintaining the ratio of the
center conductor width to the ground to center conductor gap width, [Pozar93, Simons01].
Their distributed element character helps avoiding uncontrolled stray conductances and
inductances which allows to design high quality circuits up to well above 10 GHz. For a
detailed discussion of the properties of coplanar waveguide resonators refer to [Göppl08].

2.1.1 Coplanar waveguide resonator

The fundamental mode (m = 1) resonance frequency of a CPW of length l , capacitance
per unit length C l and inductance per unit length L l is given as νr,1 = v ph /(2l ) with the
p
phase velocity v ph = 1/ L l C l = c/ ²eff . Here is c the speed of light in vacuum and
p

²eff ∼ 5.9 the effective permittivity of the CPW which depends of the CPW geometry and
the relative permittivity ²1 , see Fig. 2.2. The resonance condition for the fundamental
standing wave harmonic mode is fulfilled at a wavelength λ1 = 2l and the characteristic
impedance is Z0 = L l /C l ∼ 50Ω. For nonmagnetic substrates (µeff = 1) and neglecting
p

kinetic inductance, L l depends (similar to C l ) on the CPW geometry only. Using con-
formal mapping techniques [Simons01], one can determine L l = µ0 K (k 00 )/(4K (k 0 )) and
C l = 4²0 ²eff K (k 0 )/K (k 00 )), where ²0 is the vacuum permittivity and K denotes the com-
q
plete elliptic integral of the first kind with the arguments k 0 = w/(w + 2s) and k 00 1 − k 02 .
The resonator is symmetrically coupled to input and output transmission lines with
capacitance C κ with typical values in the range of 10 fF to 50 fF realized with gap or
finger capacitors, see geometries in Fig. 2.2. Due to this coupling we need to distinguish
between the internal quality factor of the resonator Q int = mπ/2αl for the considered
harmonic mode m and the external quality factor Q ext = mπ/(4Z0 )(1/(C κ2 R L ω2r,m ) + R L )
obtained from a LCR-model mapping, see [Göppl08]. While the former accounts for dis-
sipative photon losses via dielectric, radiative and resistive interactions taken into account
by the attenuation constant α, the latter is related to the input and output coupling of pho-
tons which depends on C κ and the real part of the load impedance of the input and output
transmission lines R L ∼ 50 Ω.
The loaded quality factor Q L can directly be measured in a resonator transmission
measurement. It is given as a combination of the internal and external quality factors
10 CHAPTER 2. REVIEW AND THEORY

1/Q L = 1/Q int + 1/Q ext . The coupling coefficient is defined as g CPW = Q int /Q ext and the
expected deviation of the peak transmission power from unity is given by the insertion
loss IL = g CPW /(g CPW + 1), or in decibel ILdB = −10 log (g CPW /(g CPW + 1)) dB. All presented
experiments were done in the over-coupled regime where g CPW À 1, IL ≈ 1 and ILdB ≈ 0.
The rate of photon loss κ is related to the measured quality factor as Q L = ωr /κ for
the Lorentzian shaped transmission power spectrum

Pr
P (ν) = . (2.1)
ν−νr 2
³ ´
1+ κ/(4π)

The photon storage time of the considered cavity mode is simply given as

τn = 1/κ (2.2)

with κ/2π the full width at half maximum of the resonant transmission peak power P r =
IL · P in where P in is the probe power applied directly at the resonator input. The phase
shift of the transmitted microwave with respect to the incident wave can also be measured
and is given by [Schwabl02]
−1 ν − νr
µ ¶
δ(ν) = tan . (2.3)
κ/(4π)
On resonance and for g CPW À 1 the average photon number n inside the cavity is directly
proportional to the power applied at its input

P in ≈ P r ≈ nħωr κ, (2.4)

valid for a coherent input tone and a symmetrically coupled resonator without insertion
loss. In order to populate the resonator with a single photon on average we therefore
require a microwave power of only ∼ 10−18 W at the resonator input for typical values of κ.

s w s
a b
finger gap
ts

l ε1 hs

lf wg

Figure 2.2: Coplanar waveguide resonator geometry. a, Top view of a CPW resonator of length l with
finger capacitor of length l f (left) and gap capacitor of width w g (right). b, Cross section of a CPW resonator
design. Center conductor of width w and lateral ground plane (light blue) spaced by two gaps of width s . The
metallization is either evaporated Aluminum or etched Niobium of thickness t s ∼ 200 nm patterned in a standard
single layer photo-lithographic process. As a substrate 2 inch wavers of c-cut sapphire with a thickness of
h s = 500 µm and relative permittivity ²1 ∼ 11 were used (dark blue).
2.1. ON-CHIP MICROWAVE CAVITY 11

2.1.2 Transmission matrix model

In contrast to lumped element oscillators, distributed element resonators carry multiple
harmonic resonance modes. Their full spectrum can be calculated using a transmission
matrix model, see [Pozar93], where each component of a microwave network is repre-
sented by a 2 x 2 matrix. The product of these matrices gives an overall ABCD matrix
which can be used to calculate the transmission coefficient S 21 of the network. The ABCD
matrix of a symmetrically coupled transmission line resonator is defined by the product
of an input-, a transmission-, and an output matrix as
µ ¶ µ ¶µ ¶µ ¶
A B 1 Zin t 11 t 12 1 Zout
= , (2.5)
C D 0 1 t 21 t 22 0 1

with input and output impedances Zin = Zout = 1/(i ωC κ ). The transmission matrix pa-
rameters are defined as t 11 = t 22 = cosh (l (α + i β)), t 12 = Z0 sinh (l (α + i β)) and t 21 =
(1/Z0 ) sinh (l (α + i β)) with a typical attenuation constant α ∼ 2.4 · 10−4 and the phase
propagation constant β = ωm /v ph . The entire resonator transmission spectrum is then
simply calculated as
2
S 21 = , (2.6)
A + B /R L +C R L + D
where R L is the real part of the load impedance, accounting for all outer circuit compo-
nents.

2.1.3 Circuit quantization

Here we address the description of a (non-dissipative) lumped-element circuit known from
electrical engineering in the quantum regime, for details see Refs. [Devoret97, Blais04,
Burkard04, Bishop10a]. An electrical circuit is a network of nodes that are joined by
two-terminal capacitors and inductors, see e.g. Fig. 2.3. Each component a carries the
current i a (t ) and causes a voltage drop v a (t ). The corresponding charges and fluxes,
Rt
which are more convenient to derive the Hamiltonian, are given by Q a (t ) = −∞ i a (t 0 )d t 0
Rt
and Φa (t ) = −∞ v a (t 0 )d t 0 , where it is assumed that any external bias is switched on adi-
abatically from t = −∞. Most relevant devices, such as the transmission line resonator
used in this thesis, can be modeled by combinations of capacitative v a = f (Q a ) and in-
ductive i a = g (Φa ) circuit elements. The classical Hamiltonian of the entire circuit may be
derived using the Lagrangian formulation L(φ1 , φ̇1 , ...) = T − V with T the energy of the
capacitive and V the energy of the inductive components. The quantum Hamiltonian is
then retrieved by replacing the classical variables by the corresponding quantum opera-
tors obeying the commutation relation [φm , q m ] = i ħ with φm the flux and q m the charge
of the node m .
An important example is the quantization of the parallel LC oscillator, see Fig. 2.3 a,
which has only one active node and one ground. The Lagrangian is L(φ, φ̇) = C φ̇2 /2 −
φ2 /2L and by defining the charge as the conjugate momentum of the node flux q =
∂L/∂φ̇ and applying the Legendre transform H (φ, q) = φ̇q − L we obtain the Hamiltonian
H = q 2 /2C + φ2 /2L . In analogy to the Hamiltonian of a particle in a harmonic potential
12 CHAPTER 2. REVIEW AND THEORY

and the direct mappings p → q , x → φ and ω2r → 1/(LC ) we can quantize it in the usual
way as µ ¶
1
H = ħωr a † a + , (2.7)
2
p
by introducing
p the annihilation operator a = 1/ 2ħZ (φ + i Z q) obeying [a, a †p ] = 1 with
p
φ = ħZ /2(a + a † ), q = −i ħ/(2Z )(a − a † ), the characteristic impedance Z = L/C and
the angular resonance frequency ωr .
A distributed element transmission line can be treated as the continuum limit of a
chain of LC oscillators [Pozar93], see Fig. 2.3 b. The effective Lagrangian L(Φ1 , Φ̇1 , ...) =
2 2
m=1 C m Φ̇m /2 − Φm /2L m describes an infinite number of uncoupled LC oscillators with
P∞

effective capacitances C = C m = C l l /2, effective inductances L m = 2l L l /(m 2 π2 ) and res-

onance frequencies ωm = mv ph π/l . The quantum Hamiltonian of the transmission line
cavity is then given as µ ¶
† 1
H = ħ ωr,m a m
X
am + , (2.8)
m 2
where m is the harmonic mode number. In many cases it is sufficient to characterize the
behavior of the circuit only in the vicinity of a particular frequency where the Hamiltonian
reduces to the single mode Hamiltonian Eq. (2.7). Near resonance of the chosen mode
with frequency ωr the mapping to the simple single LC or single LCR model provides a
sufficient understanding of the CPW resonator.

2.2 Superconducting quantum bits

A quantum bit or qubit is a quantum system with a two-dimensional Hilbert space and
represents the unit of quantum information [Nielsen00]. In contrast to a classical bit to
which the state space is formed only by the two basis states 0 and 1, a qubit can be
prepared in any one of an infinite number of superposition states |ψ〉 = α|0〉 + β|1〉 of the
two basis states |0〉 and |1〉 with the normalization α2 + β2 = 1. It is very useful to think
of a sphere with radius 1, the Bloch sphere, where the two poles represent the two basis
states and the collection of points on its surface represent all possible pure states |ψ〉
of the qubit. The concept of quantum information promises insights to the fundamentals

a ϕ b ϕ ϕ2 ϕ3 ϕ ∞−1 ϕ∞
1

L C

Figure 2.3: LC circuits. a, The parallel LC oscillator with one active node (m = 1) and corresponding node
flux φ1 = φ at the top and the ground node at the bottom of the diagram. b, The transmission line resonator
modeled as an infinite chain of LC oscillators with open-circuit boundary conditions.
2.2. SUPERCONDUCTING QUANTUM BITS 13

of physics [Deutsch85, Landauer91, Zurek03, Lloyd05], an exponential speedup of cer-

tain complex computational tasks [Feynman82, Deutsch92, Grover96, Shor97] and has
become a substantial motivation and driving force for the research in the fields of infor-
mation theory, computer science, quantum optics, AMO physics, cavity QED, solid state
physics and nanotechnology.
The experimentalist needs to implement the idealized qubit concept in an actual phys-
ical system [DiVincenzo00, Nielsen00, Ladd10]. Like most other quantum systems that
are used to implement qubits (with exception of the spin-1/2 systems), superconducting
circuits in principle have a large number of eigenstates. If these states are sufficiently
nonlinearly distributed in energy, one can unambiguously choose two as the basis states
|g 〉 and |e〉 of the physical qubit.
In order to achieve high fidelity state preparation we need be able to reliably initial-
ize one of the two basis states, typically the ground state |g 〉. This can be achieved if
the thermal occupation of the qubit is negligibly small such that k B T ¿ hνg,e . For high
fidelity state control, the two qubit states |g 〉 and |e〉 need to be sufficiently long lived
compared to manipulation times. This implies on the one hand the need for strong cou-
pling to control and readout elements, and on the other hand close to perfect protection
from any coupling to the environment. The latter requires not only the suppression of
any dissipative loss but necessitates also an efficient protection from spontaneous emis-
sion which is triggered by vacuum fluctuations [Houck08, Reed10b]. Both can induce
unwanted transitions between the qubit levels and therefore limit the energy relaxation
time T1 = 1/γ1 of the excited state |e〉. Similarly, in order to maximize the coherence time
T2 = 1/γ = 1/(1/(2T1 ) + 1/Tφ ), with Tφ = 1/γφ the pure dephasing time, any interactions
between the qubits and its environment need to be minimized [Ithier05].

2.2.1 Charge qubits

The transmission-line shunted plasma oscillation qubit [Koch07b, Schreier08, Houck08],
in short transmon, is based on the Cooper pair box (CPB) which is the prototype of a
qubit based on superconducting electronic circuits [Büttiker87, Bouchiat98, Nakamura99,
Vion02]. The basis states of the CPB qubit are two charge states defined by the number
of charges on a small superconducting island which is coupled to a superconducting
reservoir via a Josephson tunnel junction that allows for coherent tunneling of Cooper
pairs. The Josephson tunnel junction, see Fig. 2.4 a and b, consists of two electrodes
connected by a very thin (∼ 1 nm) insulating barrier which acts like a non-dissipative
nonlinear inductor according to the Josephson effect [Josephson62, Tinkham96]. The
transmon type qubit is a CPB where the two superconductors are also capacitatively
shunted in order to decrease the sensitivity to charge noise, while maintaining a sufficient
anharmonicity for selective qubit control [Koch07b], see Fig. 2.4 c and d. Other actively
investigated superconducting qubit types include the RF-SQUID (prototype of a flux qubit)
and the current-biased junction (prototypal phase qubit), for a review see [Devoret04,
Zagoskin07, Clarke08].
The Hamiltonian of the transmon or CPB can be shown to be [Büttiker87, Devoret97,
Bouchiat98, Makhlin01]
H = 4E C (n̂ − n g )2 − E J cos ϕ̂, (2.9)
14 CHAPTER 2. REVIEW AND THEORY

where the first term denotes the energy associated with excess charges on the island
and the second term is the energy associated with the Josephson coupling between the
two islands. The latter can be understood as a measure for the overlap of the Cooper pair
wavefunctions of the two electrodes. The symbols n̂ = −q̂/(2e) and ϕ̂ = φ̂ 2e/ħ denote the
number of Cooper pairs transferred between the islands and the gauge-invariant phase
difference between the superconducting electrodes, respectively. ϕ̂ is a compact vari-
able that satisfies ψ(ϕ̂ + 2π) = ψ(ϕ̂) and the commutation relation between the conjugate
variables is given as [ϕ̂, n̂] = −i . The effective offset charge on the island in units of the
Cooper pair charge 2e may be controlled via a gate electrode (Vg ) capacitively coupled to
the island (C g ) such that n g = Q r /(2e) + C g Vg /(2e) with Q r an environment induced offset
charge, e.g. from 1/f charge noise or quasi-particle poisoning.
The charging energy
E C = e 2 /(2C Σ ) (2.10)
is the energy needed in order to charge the island with an additional electron. It solely
depends on the total capacitance C Σ = C g + C S , given as the sum of the the gate ca-
pacitance C g and the transmon specific shunt capacitance C S , see Fig. 2.4. The latter
also includes the junction capacitance C J and other relevant parasitic capacitances, see
Section 5.1 for a detailed analysis of an actual qubit design. For typical transmon qubit
designs C S is chosen such that the charging energy is reduced significantly to the range
200 MHz . E C /h . 500 MHz. Its lowest value is limited by the minimal anharmonicity of
the transmon levels required for fast single qubit gates. The upper value on the other
hand is determined by the intended suppression of charge noise sensitivity at typical
qubit transition frequencies, see Subsection 2.2.2 for details.

a c Cg
CS
S1 I S2 Vg 
I
5m 0.2m
0.2
b d
CJ IJ EJ

50m
50

Figure 2.4: Josephson junction and transmon charge qubit. a, A Josephson tunnel junction consisting of
two superconducting electrodes S 1 and S 2 connected via a thin insulating barrier I . b, Circuit representation
of the junction. The Josephson element is represented by a cross and the junction capacitance C J is taken
into account by the boxed cross. c, Circuit diagram of the transmon qubit (shown in blue) consisting of two
superconducting islands (top and bottom leads) shunted with a capacitor C S and connected by two Josephson
junctions (boxed crosses) thus forming a DC-SQUID loop. The top island can be voltage (Vg ) biased via the
gate capacitor C g . In order to induce an external flux φ in the SQUID loop a current ( I ) biased coil is used
(shown in black). d, Colorized optical image of a transmon qubit. It is made of two layers of aluminum (blue) of
thicknesses 20 nm and 80 nm on a sapphire substrate (dark green). The SQUID loop of size 4 µm by 2 µm
and one of the two Josephson junctions of size 200 nm by 300 nm (colorized SEM image) are shown on an
enlarged scale.
2.2. SUPERCONDUCTING QUANTUM BITS 15

The second characteristic energy of the circuit is the Josephson energy E J which is
the energy stored in the junction as a current passes through it, similar to the energy of
the magnetic field created by an inductor. In the case of the Josephson junction no such
field is created however and the energy is stored inside the junction. If the current through
the junction is smaller than the critical current of the junction I c , there is no associated
voltage drop rendering it the only known dissipationless and nonlinear circuit element.
By choosing a split junction design, see Fig. 2.4 c and d, the Josephson energy can be
tuned by applying a magnetic field to the circuit which threads an external magnetic flux
φ through the dc-SQUID formed by the two junctions [Tinkham96]

E J (φ) = E Jmax | cos (πφ/φ0 )| (2.11)

for the simple case of two identical junctions and φ0 = h/(2e) the magnetic flux quantum.
Both the charging energy and the maximum Josephson energy

E Jmax = φ0 I c /(2π) (2.12)

with I c the critical current of the Josephson junctions are fabrication parameters that
depend on the circuit geometry and the details of the tunnel barriers respectively, see
Chapter 5 for details.

2.2.2 Transmon regime

The qubit Hamiltonian can be solved exactly in the phase basis using Mathieu functions
[Devoret03, Cottet02], see Fig. 2.5 for the energy level diagram. For numerical simu-
lations it is equivalent to solve the Hamiltonian by exact diagonalization in a truncated
charge basis
N NX
−1
( j − n g )2 | j 〉〈 j | − E J
X
H = 4E C (| j + 1〉〈 j | + | j 〉〈 j + 1|), (2.13)
j =−N j =−N

where the number of charge basis states that need to be retained in order to obtain
an accurate result, 2N + 1, depends on the ratio E J /E C and on the transmon eigenstate
l ∈ {0, 1, 2, 3, ...} (or equivalently l ∈ {g , e, f , h, ...}) of interest. Typically, N ∼ 10 charge
basis states are sufficient to obtain a good accuracy for the lowest few energy levels in
the transmon regime where 20 . E J /E C . 100.
In this limit we can also find analytic expressions. Approximately, the eigenenergy of
the state l is given as [Koch07a]

EC
µ ¶
1
(6l 2 + 6l + 3),
p
E l w −E J + 8E J E C l + − (2.14)
2 12

valid for the first few levels l and large values of the ratio E J /E C À 1. The |g 〉 to |e〉 level
transition frequency of the transmon is therefore simply given as

νge ' (
p
8E J E C − E C )/h. (2.15)
 16 CHAPTER 2. REVIEW AND THEORY

a EJ / EC = 1 b EJ / EC = 5 c EJ / EC = 50
4 3.0
15
h
2.5
h
∼ 8 EJ E C - 3 E C
h 3
2.0 f
Energy, E l E ge

10

2 f 1.5 ∼ 8 EJ E C - 2 E C
f
e
e 1.0
5
e 1
0.5 ∼ 8 EJ E C - E C

g ∼ EJ ∼4EC g g
0 0 0.0
1 0 1 1 0 1 1 0 1
Gate charge, ng 2e Gate charge, ng 2e Gate charge, ng 2e

Figure 2.5: Energy level diagram of the Cooper pair box and the transmon. Calculated eigenenergies E l of
the first four transmon levels |g 〉, |e〉, | f 〉 and |h〉 as a function of the effective offset charge n g for different ratios
E J /E C = 1, 5, 50 in panels a, b and c. Energies are given in units of the transition energy E ge , evaluated at the
degeneracy point n g = 0.5 and the zero point in energy is chosen as the minimum of the ground state level |g 〉.

The anharmonicity of the transmon levels α ≡ E ef − E ge , which can limit the minimal qubit
manipulation time, decreases only slowly with increasing E J /E C and is approximately
given as α ' −E C , see Fig. 2.5 c. The energy dispersion of the low energy eigenstates
l with respect to charge fluctuations on the other hand ²l ≡ E l (n g = 1/2) − E l (n g = 0)
approaches zero rapidly [Koch07a]
¶l 3
24l +1 2 E J 2 + 4 −p8E J /E C
r µ
l
²l w (−1) E C e (2.16)
l! π 2E C
with the ratio E J /E C , see Fig. 2.5 a-c. A DC-gate bias Vg for qubit control, as shown in
Fig. 2.4 c, is therefore obsolete in the transmon regime. This abandonment of charge
control has dramatically improved the stability of the qubit energy levels, which in many
CPB devices is limited by 1/f charge noise and randomly occurring quasi particle tunnel-
ing events. The new design has furthermore significantly improved the dephasing times
of superconducting charge qubits [Schreier08] by effectively realizing the charge noise
insensitive ‘sweet spot’ of the CPB [Vion02] at any charge bias point. It is important to
note that, in contrast to its insensitivity to low frequency noise, the transmon matrix el-
ements for resonant level to level transitions are even increased compared to the CPB,
see Subsection 2.3.3.

2.2.3 Spin-1/2 notation

The qubit Hamiltonian Eq. (2.13) can also be rewritten in the basis of the transmon states
|l 〉 which gives
H = ħ ωl |l 〉〈l |.
X
(2.17)
l
2.3. MATTER – LIGHT COUPLING 17

Introducing the atom transition operators σi j = |i 〉〈 j |, Eq. (2.17) can be written as H =

ħ l ωl σl l . In case only two transmon levels are relevant we can make use of the notation
P

used to describe a spin-1/2 particle. With the relations ωa = ωge = ωe − ωg , σgg + σee =
1 and the Pauli matrix notation σz = σgg − σee = |g 〉〈g | − |e〉〈e| the two state transmon
Hamiltonian simplifies to the spin-1/2 particle Hamiltonian [Scully97]

1
H = ħωa σz . (2.18)
2
The transmon qubit pseudo-spin can be represented by a vector on the Bloch sphere and
its dynamics is governed by the Bloch equations [Allen87], widely used in the description
of magnetic and optical resonance phenomena.
Although the spin-1/2 model is a sufficient description for the transmon qubit in many
cases, optimal control techniques are required for short qubit control pulses with a band-
width comparable to the anharmonicity [Motzoi09, Chow10a, Lucero10]. The two-state
model is also not appropriate if the transmon is strongly coupled to a field mode, see
Section 2.3, which is occupied by more than a single photon on average n, n th & 1, see
Chapters 8 and 10.

2.3 Matter – light coupling

In this section we address the physics of superconducting circuits coupled to photons
in a microwave resonator. Before going into the details in the context of circuit QED,
we start with the description of matter-light interactions in the more general context of
quantum electrodynamics in Subsection 2.3.1. The dipole coupling Hamiltonian is then
used to introduce the famous Jaynes-Cummings model which describes atoms coupled
to cavity photons in Subsection 2.3.2. In Subsection 2.3.3 we show that a superconduct-
ing artificial atom in a microwave resonator also realizes Jaynes-Cummings physics and
introduce a generalized model taking which takes into account the multiple states of the
transmon. In Subsection 2.3.4 we address the physics of coherent qubit state control and
readout in the dispersive limit of circuit QED.

2.3.1 Atom-field interaction

The quantitative description of the interaction of matter and radiation is a central part of
quantum electrodynamics. The minimal coupling Hamiltonian

1 ¡ ¢2
Hmin = p − eA(r, t) + eU (r, t ) + V (r ), (2.19)
2m
describes an electron in an electromagnetic field with the vector and scalar potentials
A(r, t ) and U (r, t ), the canonical momentum operator p = −i ħ∇ and V (r ) an electrostatic
potential (e.g. the atomic binding potential). It can be derived from the Schroedinger
equation of a free electron
−ħ2 2 dψ
∇ ψ = iħ , (2.20)
2m dt
18 CHAPTER 2. REVIEW AND THEORY

and the additional requirement of local gauge (phase) invariance [Cohen-Tannoudji89,

Scully97], such that both the electron wave functions ψ(r, t ) and also ψ(r, t )e i χ(r,t ) are
valid solutions. Here the arbitrary phase χ(r, t ) is allowed to vary locally, i.e. it is a function
of space and time variables. While the probability density P (r, t ) = |ψ(r, t )|2 of finding an
electron at position r and time t remains unaffected by the phase change, Eq. (2.20) is
no longer satisfied and needs to be modified. It can be shown that the new Schroedinger
equation
dψ
Hmin ψ = i ħ (2.21)
dt
with the minimal coupling Hamiltonian Eq. (2.19) satisfies local phase invariance and
covers the physics of an electron in an electromagnetic field. In the dipole approximation,
valid for long wavelength compared to the size of the particle, and the radiation gauge
[Göppert-Mayer31, Scully97, Cohen-Tannoudji98, Yamamoto99, Woolley03] the minimal
coupling Hamiltonian of an electron at position r0 given in Eq. (2.19) can be simplified to
H = p 2 /(2m) + V (r ) + Hint . Here the interaction part

Hint = −erE(r0 , t ) (2.22)

represents the well known dipole coupling Hamiltonian with the dipole operator d = er.

2.3.2 Jaynes-Cummings model

The physics of cavity QED with superconducting circuits [Blais04] is very similar to the
physics of cavity QED using natural atoms [Haroche06]. By making use of the previously
found expression for the resonator field Eq. (2.7) the spin-1/2 particle Eq. (2.18) and the
electron field interaction term Eq. (2.22), we will now introduce the full quantum model for
the interaction of a two state system with quantized radiation in a cavity.
Using the atom transition operators σi j = |i 〉〈 j |, we can reexpress the dipole operator
in Eq. (2.22) as d = i , j M i j σi j with the electric-dipole transition matrix element M i j =
P

e 〈i |r| j 〉. The electric field of mode m with unit polarization vector ²̂m at the position of
†
the atom is given as E = m E m ²̂m (a m + a m
P
) with the photon creation and annihilation
†
operators a and a . When the field is confined p to a finite one-dimensional cavity with
volume V the zero point electric field is E m = ħωm /(²0 V ). In the case of just two atomic
levels |g 〉 and |e〉 and only one electromagnetic field mode the interaction Hamiltonian
reduces to
Hint = ħg (σge + σeg )(a + a † ), (2.23)
with the single photon dipole coupling strength g = g ge = −M ge ²̂k E /ħ.
By introducing the Pauli matrix notation where σ+ = σg e = |g 〉〈e| and σ− = σeg =
|e〉〈g | and by combining the interaction Hamiltonian Eq. (2.23) with the single mode cavity
Hamiltonian Eq. (2.7) and the two state qubit Hamiltonian Eq. (2.18) we get
µ ¶
1 1
H = ħωr a † a + + ħωa σz + ħg (σ+ + σ− )(a + a † ) (2.24)
2 2
fully describing all aspects of the single mode field interacting with a single two level atom
or qubit without dissipation.
2.3. MATTER – LIGHT COUPLING 19

The two energy conserving terms σ− a † (σ+ a ) describe the process where the atom is
taken from the excited to the ground state and a photon is created in the considered mode
(or vice versa). The two terms which describe a simultaneous excitation of the atom and
field mode (or simultaneous relaxation) are energy nonconserving. In particular when the
coupling strength g ¿ ωr , ωa and the two systems are close to degeneracy ωr ∼ ωa the
latter terms can be dropped, which corresponds to the rotating-wave approximation. More
specifically this approximation holds as long as the energy of adding a photon or adding
a qubit excitation is much larger than the coupling or the energy difference between them
(ωr +ωa ) À g , |ωr −ωa |. The resulting Hamiltonian is the famous Jaynes-Cummings model
µ ¶
† 1 1
H JC = ħωr a a + + ħωa σz + ħg (σ+ a + a † σ− ), (2.25)
2 2

which describes matter-field interaction in the dipole and rotating wave approximations
[Jaynes63]. It is analytically solvable and represents the starting point for many calcula-
tions in quantum optics.
Close to resonance (ωr ∼ ωa ) the photon number state |n〉 and the atom ground
and excited states |g 〉 and |e〉 are no longer eigenstates of the full Hamiltonian. The
interaction term lifts their degeneracy and the new
p
eigenstates are superpositions of qubit
and cavity states |n, ±〉 = (|g 〉|n〉 ± |e〉|n − 1〉)/ 2, where the two maximally entangled

a b
|n+
|n – 1
–
|n √ng/π |n – 1 |n

|n–

|2+
|1
|2 – |2
√2g/π |1
νr + g2/(Δ2π)
|2–

|1+ |0
Δ/2π
|1 g/π |0 |1
|1–
νr νge νr – g2/(Δ2π) νge

|0 |0
|g |g, 0 |e |g |e

Figure 2.6: Jaynes-Cummings dressed states energy level diagram. The uncoupled product states (black
lines) |g , n〉 (left) and |e, n〉 (right) are given in frequency units ν = E /h . a, The resonant dipole coupled states
p
|n±〉 (blue lines) are split in frequency by ng /π. b, In the detuned case where |∆| À g the energy levels (blue
lines) are state dependently shifted to lower (|g 〉) or larger (|e〉) frequencies by (n + 1/2)g 2 /(2π∆).
20 CHAPTER 2. REVIEW AND THEORY

p
symmetric and antisymmetric superposition states are split by n 2g ħ, see level diagram
in Fig. 2.6 a. An atom in its ground state resonantly interacting with one photon in the
cavity will therefore flip into the excited state and annihilate the photon inside the cavity
|g , 1〉 → |e, 0〉 and vice versa. This process was named vacuum Rabi oscillation and
p
occurs at a frequency n g /π. The term vacuum refers to the fact that the process also
happens with an initially empty cavity n = 0 where the vacuum fluctuations of the cavity
field trigger the relaxation of the atom. Systems that show several vacuum Rabi cycles
before either the photon decays with rate κ or the atom decays into a mode other than
the resonator mode at rate γ are said to be in the strong coupling limit of cavity QED
where g À κ, γ. In circuit QED systems it is comparatively easy to realize this limit, see
Chapter 8.
In the dispersive limit where the detuning ∆ = |ωa − ωr | À g no atomic transitions
occur. Instead, virtual photons mediate dispersive interactions which lead to level shifts
proportional to g 2 /∆ of the coupled system, see Fig. 2.6 b. The Hamiltonian in this
regime can be approximated using second order time dependent perturbation theory of
the Jaynes Cummings Hamilonian Eq. (2.25). Expanding the terms into powers of g /∆
yields [Haroche92, Gerry05]

g2
µ ¶µ ¶
† 1 ħωa
H ≈ ħ ωr + σz a a + + σz , (2.26)
∆ 2 2

which illustrates the qubit state dependent shift of the resonator with the new oscil-
lation frequency ω̃r = ωr ± g 2 /∆. This resonator frequency change is detectable in a
time-resolved cavity transmission measurement and allows to perform a quantum non-
demolition (QND) measurement of the atom state. Rearranging the terms in Eq. (2.26)
yields
2g 2 † g2
µ ¶ µ ¶
1 ħ
H ≈ ħωr a † a + + ωa + a a+ σz , (2.27)
2 2 ∆ ∆
and crosses out the dual effect of the dispersive interaction. Here the atomic transition
frequency is shifted by the photon number dependent AC-Stark shift 2g 2 a † a/∆ and the
constant Lamb shift g 2 /∆. The former can e.g. be used to perform a QND measurement
of the photon number state inside the cavity [Brune94, Gleyzes07, Guerlin07, Baur07,
Johnson10].

2.3.3 Transmon – photon coupling

In this thesis we experimentally explore the above described Jaynes-Cummings physics
by integrating a transmon qubit into a coplanar microwave cavity as shown in Fig. 2.1
b. In such a solid state setting it is more natural to express the dipole coupling in terms
of voltages instead of electric fields. As discussed in Subsection 2.2.2 the transmon is
insensitive to a change in the DC gate voltage bias Vg , see Fig. 2.5. The qubit does
however couple to an AC electric field which arises due to photons populating the cavity.
If the transmon is positioned at the maximum of a considered standing wave electric field
mode with resonance frequency ωr we can express the corresponding AC gate voltage
2.3. MATTER – LIGHT COUPLING 21

as s
ħωr
V̂g = (a + a † ) = V (a + a † ), (2.28)
2C
where we have used V̂g = q/C with q proportional to the charge operator introduced
to quantize the LC-oscillator (Subsection 2.1.3). Here C is p
the capacitance
p and Z the
characteristic impedance of the resonator such that ωr = 1/ LC and Z = L/C is fulfilled
and V denotes the rms vacuum voltage of the LC oscillator1 similar to the zero point
electric field E in Subsection 2.3.2.
The quantum gate voltage Eq. (2.28) is related to the gate charge as n̂ g = C g V̂g /(2C ).
If we substitute this in the electrostatic part of the charge qubit Hamiltonian Eq. (2.9)
and expand the square we obtain a coupling term H ∝ −4E c C g V̂g n̂/e which contains the
charge qubit state n̂ as well as the quantum field oscillator state V̂g . This relation can be
simplified as
H = 2ħg (a + a † )n̂ (2.29)
with the single qubit single photon coupling strength

C g eV
g= . (2.30)
CΣ ħ

The ratio β = C g /C Σ ∈ {0, 1} is the coupling capacitance divided by the total capacitance
of the qubit. It accounts for the division of voltage in the CPB – the fact that part of
the voltage V̂g drops e.g. from the resonator center conductor to the qubit island, see
Section 5.1 for details. 2ħg therefore represents the energy needed to move one Cooper
pair across a portion β of the rms vacuum voltage fluctuations V in the resonator. For
only two qubit states, valid for example at the charge degeneracy point of the CPB, we
can replace the charge operator with the Pauli spin operator n̂ → σx /2. If we now also
apply the rotating wave approximation, which neglects the rapidly rotating terms a † σ+
and aσ− , we recover the Jaynes-Cummings Hamiltonian for the qubit photon coupling

H = ħg (aσ+ + a † σ− ). (2.31)

More generally for a multilevel system, such as the transmon qubit, the coupling
strength between the levels i and j does also depend on the transition matrix element

g i j = 2βeV 〈i |n̂| j 〉/ħ. (2.32)

In the asymptotic limit where E J /EC À 1 the matrix elements can be examined with a
perturbative approach [Koch07b]
s
l + 1 E J 1/4
µ ¶
|〈l + 1|n̂|l 〉| ≈ , (2.33)
2 8EC
1 V can for example be derived by equating the electric field part of the zero point energy with the electro-
static energy in the resonator ħωr /4 ≡ C V 2 /2.
22 CHAPTER 2. REVIEW AND THEORY

and for all non-nearest neighbor transitions (|k| > 1) the matrix element |〈l + k|n̂|l 〉| ap-
proaches zero rapidly. Employing the rotating wave approximation we obtain a general-
ized Jaynes-Cummings Hamiltonian which takes into account multiple transmon levels l
[Koch07b]
µ ¶ Ã !
† 1 †
+ ħ ωl |l 〉〈l | + ħ g l ,l +1 |l 〉〈l + 1|a + H.c. .
X X
H JC = ħωr a a + (2.34)
2 l l

Interestingly, the maximal dimensionless qubit-photon coupling g /ωr depends only on

resonator specific geometric and dielectric constants if the transition matrix element is
neglected. There exists therefore a relation between the relative coupling strength of
a single superconducting qubit and the fine structure constant approximately given as
[Devoret07, Koch07b] p
g /ωr ∼ 4β α/²r , (2.35)
with α = e 2 /(4π²0 ħc). The maximal coupling strength of a single qubit is therefore
on the order of g /ωr ∼ 0.1 for realistic values of the dielectric constant. Note how-
ever that this limit is only valid for a half wave transmission line resonator. In addition,
an inductively coupled superconducting qubit can in principle easily exceed this bound
[Devoret07, Bourassa09, Niemczyk10, Forn-Díaz10]. The same is true if multiple qubits
are collectively coupled to the resonator field.

2.3.4 Dispersive limit: Qubit control and readout

The dispersive coupling of atoms and photons detuned by ∆ = |ωa − ωr | À g was al-
ready introduced in Subsection 2.3.2. The dispersive Hamiltonian, see Eqs. 2.26 and
2.27, in principle also applies to circuit QED. There is however an interesting quantitative
difference. Due to the large coupling strength as well as the multi-level structure of the
transmon, a substantially different (typically much larger) dispersive shift term is obtained.
This enabled not only the observation of the quantum AC-Stark shift [Schuster05],
which is very useful to obtain a calibration for the mean cavity photon number as a func-
tion of the applied microwave probe power, see Section 6.4. In circuit QED it has fur-
thermore been demonstrated that the AC-Stark shift per photon can become much larger
than the qubit line width. This allows to spectroscopically infer the cavity photon number
distribution [Gambetta06, Schuster07b]. In addition, it has been observed that the Lamb
shift can even exceed the former in some cases [Fragner08].

Dispersive shifts The dispersive frequency shift between a multilevel qubit and the
resonator (χ ∼ g 2 /∆ in Subsection 2.3.2) can be calculated as χ = χ01 − χ12 /2 where
χi j = g i2j /(ωi j − ωr ) and ωi j = ω j − ωi for multilevel circuits [Koch07b]. For large detun-
ings this shift can therefore be approximated as

g 2 E C /ħ
χ≈− (2.36)
∆(∆ − E C /ħ)
in the transmon regime. Due to the reduced anharmonicity also virtual transitions through
excited transmon states need to be taken into account. This leads to a renormalization of
 2.3. MATTER – LIGHT COUPLING 23

Π
1
Transmission, T Tmax

2
e g

Phase shift, ∆ (rad)

0.5 0 e g

Π
0 2
Ν'r + χ Ν'r Ν'r - χ Ν'r + χ Ν'r Ν'r - χ
Frequency, Ν Frequency, Ν

Figure 2.7: Calculated qubit state dependent resonator spectrum. a, Calculated transmission amplitude
according to Eq. (2.1) and b, relative phase shift of the transmitted microwave according to Eq. (2.3) for a
moderate dispersive shift | ± χ| ∼ κ/2. The solid lines represent the spectrum of the cavity for the qubit being in
the ground state |g 〉 (blue) or the excited state |e〉 (red).

0 0
both the qubit ωge = ωge +χge and cavity ωr = ωr −χef /2 frequencies due to their interaction.
0
ωr is then shifted by +χ or −χ depending on the qubit state |g 〉 or |e〉, see Fig. 2.7, and
0
ωge is shifted by +2χ per cavity photon. We can write the effective Hamiltonian as

1 0 0
H ≈ ħωge σz + (ħωr + ħχσz )a † a. (2.37)
2
For small positive detunings ∆ . E C /ħ much larger (and positive) frequency shifts can
be obtained. This so called straddling regime of circuit QED has been identified as an
interesting parameter region for an efficient single shot qubit readout [Srinivasan10].

Qubit readout For a dispersive QND qubit readout the qubit state dependent shift of the
cavity resonance frequency by +χ or −χ is detected in a time-resolved resonator trans-
mission measurement [Wallraff05, Bianchetti09], see Fig. 2.7. We apply a continuous
coherent microwave tone at a frequency ωm to the resonator starting at the time t = 0. In-
cluding the measurement drive into the dispersive Hamiltonian Eq. (2.37) and expressing
it in a frame rotating at the measurement frequency leads to
1
Hm = ħω0ge σz + (ħω0r − ħωm + ħχσz )a † a + ħ²m (t )(a † + a), (2.38)
2
where ²m (t ) is the time dependent amplitude of the measurement tone. Measuring the
radiation transmitted through the cavity by heterodyne detection we can infer a complex
valued signal p
S(t ) = Z ħωm κ 〈a(t )〉 , (2.39)
which gives us access to the time evolution of the expectation value of the annihilation
operator, where Z is the characteristic impedance of the system. This signal is different
for the qubit in the |g 〉 or in the |e〉 state, see also Section 6.5, and can be calculated by
24 CHAPTER 2. REVIEW AND THEORY

solving the cavity Bloch equations [Bianchetti09]. By comparing the measurement with
the theoretical prediction the population of the excited state

P e = (〈σz 〉 + 1)/2 (2.40)

can be extracted. The frequency of the measurement drive is chosen typically close to the
resonator frequency ωm = ω0r or the qubit ground state resonator frequency ωm = ω0r − χ,
because the effect of the dispersive shift is strongest in this frequency range. This dis-
persive readout scheme has been shown to enable full quantum state tomography of
multiple qubits [Filipp09, Leek09, DiCarlo09] in joint measurements. In addition this
scheme was extended to readout the quantum state of multiple levels of an artificial atom
[Bianchetti10b]. It was furthermore demonstrated that even single shot readout of up to
three qubits [Reed10a, DiCarlo10] is possible by measuring at the bare cavity frequency
ωm = ωr with a very high average photon number. This scheme relies on the nonlin-
earity provided by the Jaynes-Cummings energy ladder [Boissonneault10, Bishop10b].
However, in contrast to a weak dispersive readout it is likely not of the QND type.

Qubit control Coherent control of the qubit state is achieved by applying a microwave
tone at the qubit transition frequency ω0a when the qubit is strongly detuned from the res-
onator. Although most of the radiation is reflected at the resonator input port, according
to Eq. (2.1) a small part will enter the resonator and drive qubit transitions at the Rabi
p
frequency Ω/(2π) = n d g /π, where n d is the number of drive photons inside the res-
onator.2 For large detunings of the drive frequency ωd from the resonance frequency
∆d ≡ ω0r − ωd À κ/2 the average number of drive photons is given as n d ≈ (²d /∆d )2 . The
coherent driving field can be considered as a classical field with an amplitude ²d (t ) and
a frequency ωd which leads to the following effective Hamiltonian [Blais07]

1³ 0 ´ 1
Hd = (ħω0r − ħωd )a † a + ħωge − ħωd + 2ħχa † a σz + ħΩ(t )σx , (2.41)
2 2
with the Rabi frequency Ω(t ) = 2²d (t )g /∆d . This control parameter allows to rotate the
qubit state vector on the Bloch sphere around the x-axis, or by shifting the phase of
the drive signal by π/2 also around the y-axis. The overall rotation around the x-axis is
characterized by the Rabi angle θR = Ω(t )d t . By applying a pulse of suitable amplitude
R

and length such that θR = π the qubit can be excited from the ground to the excited state.
For arbitrary Rabi angles the final qubit state reads

|ψ〉 = cos (θR /2)|g 〉 + sin (θR /2)|e〉, (2.42)

which can be prepared and readout in a time-resolved Rabi oscillation experiment, see
Section 6.5.

2 Alternatively, a charge gate line can be used to provide local access to drive the qubit, see Fig. 4.2.
 PTER
3
CHA
C ONCLUSION

The physics of cavity QED with superconducting circuits is very similar to the physics
of cavity QED using natural atoms. There are however also a number of substantial
differences.
The most obvious difference is a dramatically increased single-photon single-atom
coupling strength g . This very strong coupling regime of cavity QED allows for a number
of experimental investigations of matter-light interaction effects only amendable to theo-
retical consideration up to now. Additionally, the artificial atom stays inside the resonator
under all circumstances without changing its position or coupling, also e.g. if very high
drive powers are applied. As a consequence it is much easier to investigate the large
photon number regime as well as multi-qubit interactions without the need of introducing
effective coupling constants.
In contrast to cavity QED where the dipole moment constitutes a fixed property of a
given atom, the effective dipole moment of the superconducting qubit can be adjusted
by design, fabrication and by using in-situ bias currents. If relatively harmonic qubits
like the transmon are used it turns out that the multi-level physics is in general more
relevant compared to traditional cavity QED. This brings about both a number of additional
difficulties, e.g. in high fidelity state preparation, but also new and interesting multi-level
physics.
Interestingly, even the comparatively much shorter qubit state lifetimes add to the new
experimental quality in many cases since they allow for a drastically increased repetition
rate of the experiments on the order of ∼ 100 kHz. The long time stability of charge noise
insensitive transmon qubits combined with efficient data acquisition and averaging meth-
ods then allows for long time averaging and the acquisition of low noise, high resolution
data sets.

25
 PA R T II

E XPERIMENTAL P RINCIPLES

27
PART II. EXPERIMENTAL PRINCIPLES 29

Be it a ’trapped‘ microwave photon or the position of a single electron out of hundreds of

billions in a small piece of aluminum; the topic of this part are the experimental techniques
required to explore such a single quantum in a solid-state microchip circuit. Technologi-
cal challenges include ultra low power microwave frequency measurement and control at
millikelvin temperatures, see Chapter 4 and Appendix A, as well as design and fabrication
of micro- and nanometer sized circuits discussed in Chapter 5 and Appendix B. Neces-
sary sample characterization measurements are discussed in Chapter 6 before the main
experimental results are presented in Part III.
 PTER
4
CHA
M EASUREMENT S ETUP

Typical energy level splittings of superconducting Josephson junction devices are in the
microwave frequency range. The technology required for experiments at frequencies of
a few GHz is well established due to its applications in radio-astronomy and wireless
telecommunication. However, as we intend to study coherent physical effects on the level
of a single quantum of energy, we have to efficiently control and analyze microwave sig-
nals at extremely low powers < 10−18 W. In addition, the experimental setup, see Fig. 4.1,
needs to be cooled to ultra-low temperatures < 20 mK in order to suppress any thermal
excitations and prepare the system in its quantum ground state. These requirements
imply the need for appropriate wiring, electromagnetic shielding and careful filtering at
ultra-low temperatures, as well as careful mixing techniques, high sampling rates and a
full characterization of the RF-equipment in use. For a complete circuit diagram of the
experimental setup see Fig. 4.1 a.
At the beginning of this chapter, in Section 4.1, we will introduce the superconducting
microchip sample, discuss its main design parameters and explain how it is integrated
into the measurement setup. In order to control and readout the qubit and cavity states
of such a sample three main ingredients are required. In Section 4.2 the signal synthesis
part is discussed, where DC bias and phase coherent microwave signals are generated
and pulse modulated before they enter the dilution refrigerator, see Fig. 4.1 a top left. The
cryogenic part of the circuit constitutes the second main building block and is discussed
in Section 4.3. In this part of the setup the generated signals are filtered and attenuated
before they interact with the qubit. Subsequently, the transmitted measurement tone is
amplified and the cables are routed back to room temperature, see lower part of Fig. 4.1
a. Finally, in Section 4.4, we will explain how the amplified microwave tone is down-
converted to frequencies that can be sampled, averaged and post-processed digitally,
see top right part in Fig. 4.1 a.
31
32 CHAPTER 4. MEASUREMENT SETUP

4.1 Microchip sample

A number of qubits with varying parameters and layouts have been designed and fabri-
cated for the experiments presented in this thesis and Refs. [Fink09a, Fink09b, Fink10,
Leek10, Bianchetti10b, Bozyigit10c, Bozyigit10b, Lang10]. One of the fabricated chips is
depicted as an example in Fig. 4.2 a and b.

RF1 Sn RF2 LO 300 K

SRS VLF
Spec

VLF
AWG AFS
-30dB -xdB
ZFL
VFH
VFH

Bias
SLP
Vdc AFS

ADC
RC-F

BCD

BCD

BCD

FPGA
77 K
2.1 K
-10dB

-20dB

-20dB

HEMT
800 mK
100 mK
-20dB -20dB

-20dB -20dB
VLFX VLFX

10 mK

Cc Ci Co on-chip

Resonator
Lfl Cg
EJ EC Lr Rr
Cr
Cs

Lc Transmon 20 cm
a b

Figure 4.1: Exemplary measurement & control circuit (a) and cryogenic part of the setup (b). a, The
transmon (bottom, middle) is controlled via a superconducting coil L c , an on-chip flux bias line with large
bandwidth L fl and an on-chip microwave frequency charge gate line C c (from left to right). The transmon is
also coupled via C g to the L r C r R r modeled resonator (bottom right) coupled to input and output lines via C i
and C o respectively. Different temperature stages are indicated with color. See text for details. b, Cryogen free
dilution refrigerator from Vericold, equipped with two complete measurement circuits as shown in a. Different
temperature stages are indicated.
4.1. MICROCHIP SAMPLE 33

4.1.1 Superconducting circuit

The wave-guide structures, such as the coplanar resonator (green) in Fig. 4.2, qubit
control lines (violet) and additional components such as microwave beam splitters (red)
with a typical minimal feature size of ∼ 4 µm are fabricated in a standard photo-
lithographic process. Different chip designs with varying geometries and coupling pa-
rameters are patterned on a polished 2 inch c-cut sapphire waver of thickness 500 µm.
This is done either by reactively etching a magnetron sputtered Niobium thin film of
thickness ∼ 150 nm or alternatively by evaporation of an aluminum thin film of thick-
ness ∼ 200 nm. Details of this process, which was not part of this thesis, can be
found in Refs. [Frunzio05, Schuster07a, Göppl08, Göppl09]. In contrast to thin film
aluminum resonators, circuits made of Niobium can be characterized in the supercon-
ducting state at liquid helium temperatures. In addition, their visibility in the electron
microscope is much better due to its larger atomic weight compared to aluminum. This
is relevant for the accurate alignment of qubit structures in the following nano fabrica-
tion steps. For details of microwave beam splitter design and characterization refer to
Refs. [Frey08, Henzen08, Littich09] and for details of the design of local charge gate and
flux bias lines refer to [Steffen08].
The three main design parameters of the coplanar waveguide resonator are the res-
onance frequency νr determined by the length of the resonator, the quality factor Q de-
termined by the size of the chosen gap or finger capacitors and finally the impedance of
the circuit which is determined by the ratio of the center conductor width to the size of
the gap between the center conductor and the ground planes (designed to be ∼ 50 Ω),

a B C b

Ifl D

500m F G

Figure 4.2: Fabricated exemplary microchip sample. a, The sapphire chip of size 7x4.2 mm2 with patterned
aluminum thin film was used in [Bozyigit10b, Bozyigit10a, Lang10]. Indicated is the resonator in green with
input port A and the qubit in violet (enlarged inset) equipped with control lines for DC current pulses via port B
and RF pulses via port C. Also shown is the high capacitance resonator output port (enlarged inset) which at the
same time represents the input port D of an on-chip beam splitter (red) with vacuum port E and outputs F and
G. For design of beam splitters and local qubit control lines see [Frey08, Henzen08, Littich09] and [Steffen08]
respectively. b, Printed circuit board with diameter 30 mm containing the glued and wire bonded chip shown in
panel a. Also shown is the PCB without (bottom right) and with soldered surface mount SMP connectors that
connect to ports G, F, E, A, B, C (clockwise from bottom).
34 CHAPTER 4. MEASUREMENT SETUP

see Subsection 2.1.1. In order to obtain the desired resonator frequency, the resonator is
usually shaped in the form of a meander, see Fig. 4.2 a (green), which does not harm the
low loss transmission properties as long as the curvature is much less than the gap size
of the waveguide [Simons01, Wang09]. For large external quality factors the gate capac-
itors are designed just as a gap in the center conductor. In order to get high coupling on
the other hand, the capacitance can be made almost arbitrarily large by designing finger
capacitors, where the minimum spacing is set by the minimal feature size of the optical
lithography process. The center conductor width was chosen to be 10 µm and, together
with the gap, tapered smoothly at the edges of the chip to a width of about 300 µm in or-
der to be able to connect the chip to the PCB, using wire bonds, with a minimum amount
of unwanted reflections.
After cutting the waver into chips of size 2x7, 4.2x7, or 6.4x7 mm2 with the dicing saw
DAD 321 from Disco, the qubits (violet) are fabricated. These are made using double
layer resist electron beam lithography with a typical minimal feature size of ∼ 100 nm. This
is followed by double angle shadow evaporation of two aluminum thin films including a
controlled oxidation step. Again there are three main design parameters, namely the two
characteristic energies: Josephson energy E J and charging energy E C as well as the qubit
photon coupling strength g , see Sections 2.2 and 5.1. Virtually all desired combinations
of these parameters can be implemented with an appropriate geometry and fabrication.
Fabrication and design of Josephson junction devices is covered in detail in Chapter 5
and relevant recipes are given in Appendix B.

4.1.2 Printed circuit board

The printed circuit board (PCB), see Fig. 4.2 b, provides the necessary interface between
the microchip and the rest of the measurement circuit. Its main design consideration is
the suppression of parasitic resonances due to impedance mismatches. One action is
to periodically contact the bottom ground plane with the top ground strips of the PCB
with copper vias, suppressing any modes between these two planes up to over 15 GHz.
No vias could be fabricated in the silicon substrate of the chip, however the chips were
taken to be small enough so the cutoff frequency of such modes are much higher than
the operation frequencies. In order to keep the impedance mismatch due to different
dielectric constants small, a PTFE - woven fiberglass - ceramic filled laminate AD1000 of
thickness 0.5 mm from Arlon was used as a dielectric. It has a low loss tangent of tan δ ∼
0.0023 and a large dielectric constant of ∼ 10.2 at 10 GHz, which allows for a compact
design of the conductor backed coplanar PCB wave guides. Electrodeposited copper is
used for the front and back metallization of thickness 17.5 µm and 70 µm respectively.
The two metal films are connected by a large number of copper vias of diameter 300 µm
which are arranged with a spacing of 1 mm in order to minimize stray resonances. The
boards are designed with a cutout to hold different chip sizes and fabricated at Hughes
Circuits Inc. Common minimal PCB feature sizes are on the order of 100 µm with typical
tolerances of ∼ ±75 µm. For further details of PCB design refer to [Schmidlin08].
Right angle surface mount SMP connectors 19S102-40ML5 with good performance
up to about 20 GHz from Rosenberger are heat gun soldered to the PCB using leaded
4.1. MICROCHIP SAMPLE 35

solder paste SC170 from Solder Chemistry. After the soldering, the PCB is cleaned in
an ultrasonic bath with citric acid (1:1 in H2 O ) and IPA for about ∼ 3 minutes each. The
chips are then glued and carefully aligned in the PCB cutout under an optical microscope.
A thin layer of PMMA (thinned with about 4% chlorbenzene) is used as a glue that drys
typically within ∼ 20 minutes.

4.1.3 Wire bonds

The device is contacted with aluminum wires of thickness ∼ 30 µm using the wire bonder
7476E from Westbond. About 2-3 bonds were used to contact each waveguide launcher.
All chip ground planes were contacted to the PCB ground at the circumference with about
5-10 bonds per millimeter. Typical bond wire configurations are shown in Fig. 4.3 a.
Short bonds are more difficult to achieve but generally preferred as they have smaller
inductance and therefore improved high frequency properties.
In addition, on-chip bonds were used for more complicated designs where the ground
plane is divided into several parts due to the presence of bias lines or different beam
splitter paths, see Fig. 4.3 a. In such a case superconducting interconnects between
the ground planes are needed in order to avoid spurious resonances of the circuit. One
quite successful strategy was to simply connect all on-chip ground planes by as much
superconducting bonds as easily possible on the limited space available. In particular,
multiple connections between both sides of the resonator reduce the occurrence of so
called odd modes [Simons01]. If these bonds are used on either side of a qubit however,
a closed superconducting loop around the qubit is formed. In this case a decreased
coupling to an externally applied B-field bias has been observed. This effect is probably
due to screening currents through the loop formed by the bond wires which needs to be
taken into account. A new air bridge process is under development and should replace
the sometimes difficult on-chip bonding process in the future.

4.1.4 Sample mount, coils and shielding

The main design parameter for the sample mount is once more the suppression of any
electromagnetic stray fields as well as any unwanted reflections due to impedance mis-
matches in the connections of the circuit. A sample mount made from solid copper, as
shown in Fig. 4.3, was used to protect the chip from interference with electromagnetic
fields. In order to end up with nicely sealed box and provide good contact between the
bottom ground plane of the PCB and the sample holder, suitable bullet connectors were
used to connect the semi rigid coaxial lines to the surface mount connectors. In order to
minimize the volume for parasitic modes a sample cover was used, see Fig. 4.3 b, to fill
the empty volume between PCB and sample box optimally, see Ref. [Marx09] for details.
At the outside of the copper box a milled blind hole houses two small superconducting
coils with diameter of ∼5 mm for selective magnetic field biasing of the qubits. In addition
a third larger coil with diameter of ∼18 mm was installed on top of the small ones, see
Fig. 4.3 c. The used wire was SC-T48B-M0.026mm from Supercon with an outer diame-
ter of 36 µm with ∼100 turns per layer and ∼70 layers implemented. The bias lines were
common twisted pairs with RC filters at room temperature. Suitable filters at cryogenic
36 CHAPTER 4. MEASUREMENT SETUP

b c d e

Figure 4.3: Chip and sample mounting. a, Section of the wire bonded chip. b, Sample cover and bottom
part of the sample holder. c, Closed sample holder containing three superconducting coils. d, Mounted sample
holder with shield holder (top). e, Mounted double-walled magnetic shielding.

temperatures would be an improvement for future experiments. For further details of the
superconducting coils and the used filters refer to Subsection 4.3.4 and [Bianchetti10a].
The sample holder is fixed to the base plate of the dilution refrigerator using a long
threaded rod with spanner flat. The rod is also used to height adjustably mount the dou-
ble walled magnetic shields made from tempered Cryoperm 10 of thickness 1 mm from
Vacuumschmelze, which is soft magnetic at low temperatures, see Fig. 4.3 d and e. All
copper parts below 4 K are made using oxygen-free copper in order to ensure reasonable
heat conduction and thermalization even at temperatures of only a few millikelvin. As a
side effect the vacuum properties are also improved.

4.2 Signal synthesis

Most experiments presented in Chapter 8 are based on spectroscopic measurements
where phase coherent continuous wave (CW) tones of variable power are used as input
signals. For time-resolved qubit manipulation however, see Section 6.5 and Chapter 10,
phase sensitive pulsed microwave signals are required. In both cases, a number of trigger
signals and an accurate synchronization is absolutely necessary to ensure successful
and reliable operation of the experiment.
4.2. SIGNAL SYNTHESIS 37

4.2.1 Microwave generation, control and synchronization

Microwave frequency signal generators PSG E8257C and PSG E8257D from Agilent
were used to provide classical phase coherent signals1 of up to 20 GHz, see RF1, RF2
and Spec signal sources as shown in Fig. 4.1 a top. In order to be able to modulate
the amplitude of these tones with nanosecond resolution, arbitrary waveform generators
AWG520 or AWG5014 from Tektronix are used to generate envelope pulses with a sam-
pling rate of 1 GS/s on up to four independent channels and 4 additional digital markers,
see AWG in Fig. 4.1 a top left. The digital markers are used to define and control the
sequence of each measurement by triggering the output of microwave sources and data
acquisition cards at the beginning of every measurement with typical repetition rates of 10
to 50 kHz, limited by the qubit relaxation time T1 . Pulse sequences are designed in Math-
ematica and transferred into the memory of the AWG via an ethernet connection. Apart
from the pattern file transfer, all communication between the measurement and control
computer and the discussed devices is done via their GPIB-interface where Labview soft-
ware is used to automate the measurement procedures. Long term phase stability and
synchronization is achieved by phase locking all signal generators, AWGs and acquisition
cards with an ultra low phase noise 10 MHz Rubidium frequency standard SRS FS745
from Stanford Research Systems.

4.2.2 Quadrature modulation

To apply microwave pulses phase sensitively to the qubit, e.g. via the charge bias line, an
IQ quadrature mixer IQ-0714M from Marki Microwave was used to modulate the output
N
of the spectroscopy generator, see symbol in Fig. 4.1 a top left. Such a device has a
LO (local oscillator) input, two separate IF (intermediate frequency) inputs I and Q (DC-
500 MHz) and one RF (radio frequency) output. The LO input is split with a power divider
into two signals. One of them is directly multiplied with the I input while the other is first
phase shifted by -90 degree (for RF>LO) before being multiplied with the Q input. Those
two branches are finally combined again at the RF output.
Mathematically the amplitudes of the in-phase carrier I and the quadrature phase
carrier Q , which are applied at the I and Q inputs of the mixer, are interpreted as the two
Fourier components
2
Z
I= S IF (t ) cos (ωIF t ) d t (4.1)
T T
2
Z
Q= S IF (t ) sin (ωIF t ) d t (4.2)
T T
of the IF signal

S IF (t ) = A cos (ωIF t + φ) = I cos (ωIF t ) −Q sin (ωIF t ), (4.3)

where T is any chosen period interval e.g. 0 – 2π. We can therefore represent the state
of the sine wave S IF (t ) by a vector in a complex plane with I on the real axis and Q on the
1 Phase noise in CW mode: -115 dBc/Hz at 100 kHz offset in the relevant frequency range of 3.2 GHz to
10 GHz.
38 CHAPTER 4. MEASUREMENT SETUP

imaginary axis. Amplitude and phase of the IF signal are encoded in the two quadratures
I and Q as
I
q µ ¶
A = I 2 +Q 2 and φ = tan−1 . (4.4)
Q
We can therefore adjust both the amplitude and phase of the RF signal simply by chang-
ing the voltages applied to the I and Q ports of the mixer. The angular intermediate
frequency νIF = ±(νRF − νLO ) can now be chosen to be either zero (homodyne mode or
direct modulation), or finite (heterodyne mode, or sideband modulation). In the direct
modulation scheme a DC pulse applied e.g. to the Q input port will cause the LO input
signal to be transmitted with a phase shift of 90 degrees. Using a finite intermediate fre-
quency two sidebands at frequencies νRF = νLO ± νIF are transmitted. In the ideal case
the RF amplitude scales linearly with the IF amplitude and if both IF inputs are set to
ground no signal is transmitted through the mixer.
In reality, the finite LO to RF isolation of the mixer can be a substantial problem for high
fidelity state preparation as the qubit is weakly driven even if there is no pulse applied.
In case of direct modulation the best strategy was to use an AWG marker similar to a
backup switch, gating the microwave sources shortly before and after each pulse. Here
a pin diode is used in the microwave source to switch on and off the signal with an on/off
ratio of well above 80 dB. In accordance with the specifications of the signal generators
being externally pulse modulated in such a procedure, a rise and fall time of the backup
pulses are however limited to ∼ 6 ns each. An even better strategy is to use single
sideband modulation of the microwave pulses at an intermediate frequency of typically
100 MHz. In this case the remaining leakage of the LO signal is detuned from the RF
pulse (and the qubit transition frequency) which improves the fidelity of the qubit state
preparation. In both modulation schemes it is required to apply and calibrate DC bias
voltages at the IF ports of the mixer in order to maximally suppress the leakage of the
carrier signal and guarantee phase conservation and a linear dependence of input and
output powers, see [Schmidlin09] for details of mixer calibration.

4.2.3 Generation of Gaussian noise

For the experiments presented in Chapter 10 a variable amplitude noise source with a
constant frequency spectrum in a bandwidth ∼ ±100 MHz centered at the ∼ 6.5 GHz
resonator mode is needed. The arbitrary waveform generator AWG520 provides a quasi-
random (gaussian) noise source output S n with a sampling rate limited bandwidth of 500
MHz, see top of Fig. 4.1 a. In order to up-convert this signal to the frequency of the
resonator mode, the tone is mixed with the local oscillator RF2, see Fig. 4.1. In order
to maximally suppress the coherent carrier tone leakage, the noise signal was amplified
twice by 14 dB with the 350 MHz preamplifier SR445A from Stanford Research Systems
which allows to make use of the full linear IF-input range of the mixer. Furthermore,
calibrated constant DC voltages on the I and Q ports of the mixer were applied to further
suppress carrier leakage.
The resulting tone is then attenuated with two USB programmable step attenuators
DA8-30 from Weinschel Associates with a maximal attenuation of 31.5 dB in steps of
4.3. CRYOGENIC SETUP 39

0.5 dB from DC to 8 GHz, which were operated in series. In order to restrict the band-
width of the upconverted noise tone to the required frequency range a 105 MHz low pass
filter (VLF) from Minicircuits was used before the mixer. In order to suppress any har-
monics as a result from the mixing, also a bandpass filter (VHF) from Minicircuits was
employed. Usage of 3 dB attenuators between each active component helps to reduce
reflections and improve the flatness of the noise spectrum. The noise tone is then added
to the measurement tone, RF1 in Fig. 4.1 a, with a standard resistive power splitter from
Minicircuits.

4.3 Cryogenic setup

In this subsection we discuss some important aspects of the dilution refrigerator, the part
of the measurement setup providing low operating temperatures for the circuits. The
generated signals need to be filtered, attenuated and after interaction with the sample,
amplified and routed back to the data analysis circuit at room temperature. The main
design consideration for the wiring of a cryostat as shown in Figures 4.1 and 4.4 is to
minimize the amount of heat and Johnson noise that reaches the sample, which requires
careful thermalization of all lines. At the same time the attenuation on the input side and
the attenuation on the output side as well as the amplifier noise needs to be minimized
to be able to apply large field amplitudes and ensure good signal to noise ratio during
data acquisition respectively. It is also important to ensure that the heat loads at differ-
ent stages of the cryostat do not harm its proper operation. Furthermore, high quality
microwave frequency components, such as coaxial lines, connectors and circulators are
required.

4.3.1 Dilution refrigerator

In order to perform experiments at 10 − 20 mK, three new dilution refrigerators were in-
stalled, equipped and maintained by the QuDev team during the past 4 years. These
include a conventional Kelvinox 400HA and a cryogen free Triton 200 dilution refriger-
ator from Oxford Instruments and another cryogen free system DR200 from Vericold
Technologies. All of which are specified with a base temperature of < 10 mK which was
confirmed by nuclear-orientation thermometry.
Prior to their installation sensitive vibration measurements with µg resolution by
means of geophone sensors have been carried out. The suppression of vibrations is
essential to reach ultra low and stable base temperatures, see Appendix A. Furthermore,
measurements of the electromagnetic spectrum in the laboratory up to low GHz frequen-
cies have been carried out in order to characterize the amount of parasitic e.m. fields
that could possibly interfere with the experiments implemented later on. In the first test
run, without the necessary wiring installed, we could reach a stable base temperature of
<5 mK was reached in the Kelvinox 400HA. After the cables were in place a temperature
of ∼ 10 mK was typical.
For general principles of cryostat design and operation refer to [Pobell06], [Enss05],
or [White02] and for rather technical details like pumping line layout and the necessary
40 CHAPTER 4. MEASUREMENT SETUP

vacuum techniques, see Appendix A and [Roth76]. For the development of efficient vi-
bration isolation of pumping lines [Richardson88] and [Kirk78] were very helpful, see also
Appendix A.

4.3.2 Microwave input

We generate high power and high frequency signals at room temperature, but, according
to Eq. (2.4) and typical values of κ, require only ∼ −145 dBm (∼ 10−18 W) for measurement
and typically not more than ∼ −60 dBm to drive qubit transitions with a Rabi period of a
few nanoseconds. The necessary drive power depends on the qubit to charge gate line
coupling via capacitor C c , see Fig. 4.1. If the resonator input is used for qubit driving most
of the drive tone is reflected and the necessary drive power therefore depends on the
drive to resonator frequency detuning ∆d , as discussed in Subsection 2.3.4. Typically we
therefore use a fixed attenuation of -30 dB right after the measurement signal generation,
RF in Fig. 4.1 a, as a protection and no room temperature attenuation on the qubit drive
line, Spec in Fig. 4.1 a.
After generation and initial attenuation the microwave tones are routed to the cryo-
genic part of the setup using standard 50 Ω impedance semi rigid UT85 coaxial lines from
Rosenberger Micro-Coax equipped with SMA connectors from AMP. SMA DC-blocks
8039 from Inmet are used to isolate the microwave generator device ground form the
cryostat ground, see DCB’s in Fig. 4.1 a. In fact every connection to the cryostat includ-
ing the vacuum lines are DC isolated except for one single well defined grounding cable
in order to avoid the occurrence of ground loops. Standard SMA vacuum feedthroughs
34_SMA-50-0-3/111_NE from Huber and Suhner, see Fig. 4.4 a, are then employed at
the 300 K plate of the dilution refrigerator.
Also inside the fridge semi rigid coaxial lines were used. Each of them was bent at
least once between every temperature stage such that the mechanical stress acting on
the connectors is reduced. This can be a problem for long sections of cables and in par-
ticular if the inner and outer conductors thermalize differently.2 Standard SMA connectors
from AEP (male with dielectric) and SGMC (female without dielectric) were soldered ac-
cording to the manufacturers connector mounting instructions. The cable preparation
before soldering turned out to be a crucial point in order to achieve the connector spec-
ifications with typically less than 20 dB reflection (voltage standing wave ratio (VSWR)
smaller than ∼ 1.2) in the range from DC to 20 GHz. In Fig. 4.4 b a nicely stipped UT-
85 stainless steel cable with a very smooth and flat stripped dielectric is shown. During
soldering it was important to use not too much solder flux and evaporate it entirely dur-
ing soldering, as residual solder flux can corrode the solder tin such that the connection
breaks under the stress cause e.g. by thermal expansion of the cables. In the case of
copper mantled cables even water based flux was used. This also helps to avoid electrical
shorts induced by residual flux between the inner and outer conductor.
At GHz frequencies 1/f noise is usually a minor issue. In contrast, Johnson noise
generated by the thermal motion of charge carriers in the conductor, is the dominant
noise source at temperatures above 100 mK. The voltage noise power spectral density
2 The linear thermal expansion coefficient of both stainless steel and copper is α ≈ 17 · 10−6 m/(m · K).
th
4.3. CRYOGENIC SETUP 41

a b c d e

Figure 4.4: Cabling and thermalization at cryogenic temperatures. a, Eight cables mounted to the cryostat
input port plate using vacuum feedthroughs. b, Stripped semirigid stainless steel cable prepared for connector
soldering. c, Clamped and heat sunk cables inside the cryostat. Thermalization of the outer conductor using
oxygen free copper braces. d, Thermalization of the inner conductor by heat sinking a connected attenuator
using a copper clamp and braids. e, Heat sunk twisted pair loom glued to copper post with GE-varnish.

in this regime can be estimated to be linear with the temperature T and the resistance
R as S v2 (ω) = 4k B T R . Since the required signal powers are very low, the most effective
way to suppress thermal radiation is to attenuate the signal together with the noise. In
order to reduce the room temperature thermal noise, at ∼ 300 K, down at liquid helium
temperatures, at ∼ 4 K, it is therefore sufficient to install a cryogenic attenuator providing
20 dB, this is a factor of 100, of attenuation.
For the necessary attenuation between liquid He temperature and base temperature,
the consideration with a linear temperature dependance is not valid anymore. For this
section of wire the radiation spectrum of a one dimensional black body was calculated
and analyzed, particularly the radiation in the frequency band the experiments are oper-
ated at. In order to suppress thermal radiation from temperatures higher than the base
temperature, a 20 dB attenuator at 100 mK and at base temperature each were installed,
see Fig. 4.1 a. For lower attenuation on the base plate of the cryostat slightly elevated
cavity temperatures > 70 mK were observed and characterized, see Chapter 10.
Even larger attenuations would on the other hand reduce the maximal field amplitudes
and thus the minimal qubit Rabi rotation times achievable. If this was compensated with
larger input powers, the resulting heat load due to dissipation at the base temperature at-
tenuation may become an issue. Additional frequency dependent attenuation is provided
by the coaxial cables, see Table 4.1, as well as the insertion loss of the connectors and
other components.

4.3.3 Heat loads

In order to find a good tradeoff between low heat conductance and low electric loss dif-
ferent coaxial cables were employed, see Table 4.1. The used materials for ground /
center conductor were low loss silver and tin plated copper / copper at room tempera-
ture (UT-85-TP), medium loss stainless steel / copper from 2.1 K to room temperature
42 CHAPTER 4. MEASUREMENT SETUP

(UT-85-SS), which is typically only used on the signal output side, and relatively lossy
stainless steel / stainless steel semi rigid cables (UT-85-SS-SS) for any connections be-
tween different temperature stages inside the dilution refrigerator. Connections within a
given temperature stage are always made of the low loss coaxial lines.

Frequency [GHz] UT85-TP [dB/m] UT85-SS [dB/m] UT85-SS-SS [dB/m]

0.5 -0.45 -1.02 -2.92
1 -0.64 -1.46 -4.13
5 -1.51 -3.33 -9.32
10 -2.21 -4.79 -13.26
20 -3.29 -6.94 -18.91
Table 4.1: Cable attenuation. Specified frequency dependent attenuations of one meter of semi rigid coaxial
cable used in the experimental setup.

The signal attenuation method for thermal noise reduction only works if the coaxial
lines are well thermalized with each temperature stage. In addition, every thermalization
minimizes the heat load at the next lower temperature stage. Therefore, all cables are
thermalized at every temperature with copper braces, see heat sinks indicated by symbol
a in Fig. 4.1 a. Since the teflon dielectric in the coaxial lines has extremely low thermal
conductance this only thermalizes the outer conductor. Therefore, the only points where
the center conductor can be thermally anchored to the surrounding temperature is at the
position of the attenuators, filters, amplifiers and circulators. Typical heat sinking of outer
eight conductors and an attenuator by means of copper clamps and a solder flux free
copper braid is indicated in Fig. 4.4 c and d respectively.
This cabling strategy was confirmed to be compatible with the typical maximal heat
loads of the three fridges for up to 16 coaxial lines per cryostat. The total thermal heat
load at every temperature stage was estimated by considering shield and center conduc-
tor separately, assuming every thermalization to be perfect and calculating the resulting
RT
total heat load with Q̇ = A/L T12 λ(T ) d T where A is the cross section, L the length, λ(T )
the temperature dependent heat conductivity of the material and T1 and T2 the tem-
peratures of the considered fridge stage and the next higher temperature stage without
intermediate thermalization, respectively. Instead of explicitly integrating λ(T ) one can di-
RT
rectly subtract two heat conductivity integrals Θ = 0 λ d T given in W/m and available for
example from Lakeshore Cryogenics, see also Ref. [Pobell06]. For typical combinations
of temperatures mean values of λ̄(T1 , T2 ) in W/(m · K) are also available and can be used
without the need for integration, see Table 4.2 and Ref. [White02]. At very low temper-
atures the heat conductivity of pure metallic elements highly depends on small traces of
chemical impurities and physical defects, therefore these values may vary quite a lot and
are given in brackets. Typical cross section areas of the coaxial cable shield and center
conductor were 1.6 mm2 and 0.2 mm2 , respectively.
Typical values for the cooling power available at the different temperature stages, see
Fig. 4.1 b, of the installed cryostat are ∼ 5 − 10 µW at ∼20 mK, typical ∼ 400 µW at the
100 mK stage (specified), estimated 100 - 300 mW at the 1.5, 2.1 or 4 K stages of the
4.3. CRYOGENIC SETUP 43

Kelvinox, Triton and Vericold dilution refrigerators respectively. At the 4 K stage of the
conventional fridge the cooling power is only limited by the maximal evaporation rate of
liquid Helium and in the case of cryogen free systems the cooling power provided by the
pulse tube cooler is 4 and 2 W respectively. Note that here also the twisted-pair copper
wires, typically used for DC bias between room temperature and ∼ 2.1 K, can contribute
considerably to the total heat flow Q̇ .

Material 300-77 K 300-4 K 77-4 K 4-1 K 1-0.1 K

do. copper 190 160 80 5 (1)
el. copper 410 570 980 200 (40)
st. steel 12.3 10.3 4.5 0.2 0.06
Table 4.2: Thermal conductivities. Temperature dependent mean thermal conductivities of phosphorus deox-
idized copper (typically used for bars, rods and pipes), electrolytic tough pitch copper (typically used in spools
of wire) and stainless steel (18/8) in W/(m · K), from [White02].

4.3.4 DC bias input

The split transmon qubit transition frequency can be controlled by applying a magnetic
field which threads a flux through the SQUID loop thus changing the phase difference
across the two Josephson junctions, see Section 2.2. In our setup we make use of su-
perconducting coils with large inductance to generate a constant offset flux bias and also
on-chip flux lines for rapid flux bias changes on a typical timescale of 3-5 nanoseconds.
Due to field focusing effects and the large mutual inductance between the coil with
inductance on the order of 0.8 H and qubit with loop sizes of up to 4x4 µm2 , currents
of some tens of micro amperes were sufficient to induce single flux quanta. We used a
Yokogawa voltage source and an RC-filter with a 3 dB cutoff frequency of << 50 Hz and
a total resistance of 32 kΩ at room temperature to bias the coils, see Fig. 4.1 a top left.
For details refer to [Bianchetti10a]. The voltage source output is differential and isolated
in order to keep the cryostat on just one well defined ground.
The wiring between a home made breakout box and the superconducting coils was
done with electromagnetically shielded twisted pairs and shielded connectors from Lemo
at room temperature. Standard twisted pair looms made of copper, between room tem-
perature and 2.1 K, and NbTi for temperatures below 2.1 K were employed inside the
cryostat. Connections inside the cryostat were done using micro sub-D connectors. Re-
sistive connectors are avoided in the low temperature part of the coil wiring to avoid
heating. All wires were thermally anchored at every temperature stage by wrapping them
around a copper post and fixing them with GE varnish (C5-101) from Oxford Instruments,
see Fig. 4.4 e.
The on-chip flux bias lines are operated by applying short voltage pulses with a band-
width of ∼ 500 MHz directly via one of the AWG outputs, see Fig. 4.1 a top left. The
voltage translates to a current via the characteristic impedance of the circuit which is
dominated by the 10 dB attenuator and the impedance of the transmission line. Once
the current pulse I fl reached the sample via the on-chip flux line, it passes by the qubit
44 CHAPTER 4. MEASUREMENT SETUP

SQUID loop and induces a flux. Finally it enters the ground plane where it is dissipated
on the PCB, see left inset in Fig. 4.2 a. For details of flux and charge line design refer to
[Steffen08].
The flux line cabling and thermal anchoring is similar to the one used for the mi-
crowave frequency inputs. However, since large attenuation in DC bias lines would lead
to heating, two VLFX 300 MHz low pass filters from Minicircuits, see Fig. 4.1 a, are used
to suppress thermal noise at the relevant frequencies.
Several pulse imperfections have been identified and need to be corrected. First of
all, occurring DC offsets on the order of up to ±10 mV deviation from 0 V need to be cor-
rected by adding an external DC voltage to the active port of the AWG. A pulse shaping
procedure was furthermore implemented in order to correct temporal distortions intro-
duced by the AWG itself, the low pass filters, and the semi rigid coaxial transmission line
connectors. The correct pulse modulation was inferred from a characterization measure-
ment of the signal transfer function of the signal generating system at room temperature
and without a sample. For details refer to Appendix A in [Bozyigit10a].

4.3.5 Microwave output

After interaction with the sample, signal powers corresponding to less than a single mi-
crowave photon energy per mean cavity photon lifetime are to be measured. In order to
achieve their detection in the presence of thermal photons and amplifier noise, which is
typically larger than the signal itself, any losses need to be minimized along the output
line. An additional damping of the signal before the first amplifier by say −3 dB requires 4
times more averaging to realize the same signal to noise ratio (SNR), see [Pozar93].
In order to protect the sample also from noise that could enter the resonator from
the output side e.g. input noise from the first amplifier, the signal is directed through two
circulators from Pamtech or alternatively two isolators from Raditech Inc. In the case of
the circulators one of the three ports is terminated with a well thermalized 50 Ω cryogenic
termination, which makes the circulator acting similar to a one-way microwave valve.
These devices provide an isolation of typically 20 dB each in the reverse direction while
transmitting the microwaves with only little loss of < 0.5 dB in the forward direction, see
Fig. 4.1 a.
Due to the low loss requirements the cables between sample and the first amplifier
were made of superconducting coaxial cables based on a Niobium-Titanium alloy. These
have a very low electric loss below the critical temperature comparable to the UT-85 coax-
ial cables. In contrast to the copper cables there is almost no thermal heat conductivity
through the superconductor however. Since these cables were not suitable for soldering,
appropriate SMA connectors were crimped with the Universal Compression Crimp Tool
Kit from AMP.
A good amplifier is the key hardware ingredient in order to be able to detect ultra low
power signals. Together with a large gain, an ideal amplifier also provides a low noise
temperature and a large dynamic range over a large bandwidth. High electron mobility
transistor (HEMT) amplifiers offer a superb compromise of all these requirements. The
InP HEMTs employed in our setup were manufactured by the Sander Weinreb group at
4.3. CRYOGENIC SETUP 45

Caltech and by the Low noise factory at Chalmers. These devices have a typical gain
of 30 - 40 dB, a large dynamic range with a 1 dB compression point at ∼ −10 dB and
most importantly a characteristic noise temperature of only ∼ 2.5 − 4 K in a bandwidth of
typically 4 − 8 GHz. The low noise amplifiers (LNA) are operated with a stable (< 10 mV)
and carefully adjustable voltage supply (±3 V) for drain and gate biases, which was made
in house3 , see Refs. [Puebla08, Bolesch10] for further details.
An amplifier noise temperature of Ta = P n /(k B B n ) with the noise power P n and the
relevant bandwidth B n of only 2.6 K is specified for the HEMT amplifier, a value which is
among the lowest one can obtain with conventional transistor based amplification tech-
niques. Ta can also be thought of as the number of thermal photons n a added at the
input of an otherwise perfect amplifier expressed as a temperature via Planck’s law in
one dimension n a = 1/[exp (ħω/k B Ta )−1]. At typical microwave photon frequencies of 6 -
7 GHz this corresponds to adding 6 - 9 thermal photons before the amplification of signal
plus noise by 4 orders of magnitude. For successful detection a cascade of amplifiers
with a total amplification of ∼ 14 orders of magnitude is required. The few noise photons
at the output of the first amplifier are therefore amplified by almost another 100 dB.
For a chain of amplifiers the total noise temperature is given by Frii’s law

Ta,2 Ta,3
Ta = Ta,1 + + + ... (4.5)
G1 G 1G 2

with the individual amplifier gain G i and the noise temperature Ta,i of the i ’s amplifier.
For a sufficiently large G 1 the signal to noise ratio (SNR) which depends on the total Ta
is therefore governed by the noise temperature of the first amplifier Ta,1 only. In fact,
the effective system noise temperature relative to the resonator output port has been
measured [Bozyigit10b] to be on the order of 17 K. This value is considerable larger than
expected from the thermal occupation of the resonator field, which for the latest damping
configuration is on the order of Tr < 70 mK, see Subsection 8.3.3 for a characterization of
the background resonator field temperature. The larger effective noise temperature of the
amplification chain is therefore attributed to cable and insertion losses of 5.8 dB between
the resonator output port and the low noise amplifier.
Due to Frii’s law it is not necessary to be exceptionally worried about signal loss
once the microwave tone has been amplified with the LNA. Standard lossy stainless
steel coaxial lines and in some instances the medium loss copper/stainless lines are
used to route the microwave tone back to room temperature. At the cryostat output plate
another DC block isolates the experiment from the active components that follow, i.e. two
Miteq amplifiers with a measured total gain of 59 dB. This gain includes an extra 3 dB
attenuator for the reduction of reflections between the amplifiers as well as two filters
from Minicircuits which form a bandpass that is positioned in between the two amplifiers
in order to not saturate the second amplifier with noise in unused frequency bands. In
Section 4.4 we review how the weak coherent measurement tone, which corresponds
3 Note that the DC voltage source wiring resistance should not become a notable fraction of the amplifiers
internal drain resistance in order to be able to source the intended source-drain and bias currents with the
specified voltages that were optimized to give the minimal noise temperature.
46 CHAPTER 4. MEASUREMENT SETUP

to only a few photons populating the resonator on average, is extracted from the noise
dominated signal.

4.4 Data acquisition

All measurements presented in this thesis are measurements of the cavity transmission.
Employing a heterodyne detection scheme [Walls94, Wallraff04] we can detect the mea-
surement photons which are phase coherent with a local oscillator while averaging out
all uncorrelated noise photons. In order to be able to fully analyze GHz frequency sig-
nals digitally, it is convenient to use analog down-conversion. We used the IQ mixer
IRM0408LC2Q from Miteq in which the measurement signal is mixed with a separately
generated phase coherent and phase locked local oscillator signal, see LO in Fig. 4.1 a.
The LO signal is typically offset by νIF = 10 - 25 MHz from the measurement frequency,
which allows for convenient digital sampling of the I and Q quadratures of the down con-
verted signal at frequency νIF . After another stage of amplification with a Minicircuits
amplifier (24 dB gain) the signal has an amplitude in the millivolts range. Another low-
pass filter from Minicircuits is used to reduce high frequency noise and suppress aliasing
by removing frequency components above the Nyquist frequency set by the data acqui-
sition sampling rate [Bozyigit08, Lang09, Bozyigit10a]. After that, the down-converted
measurement tone is digitized and averaged with an ADC data acquisition card, see
Fig. 4.1 a.
The PCI acquisition board from Aquiris with a maximum sampling rate of 1 GS/s on
each of two channels, has fast AD converters and an on-board FPGA hard-coded to av-
erage the acquired signals in real time. This is particularly interesting as it is extremely
large amounts of data that can be acquired in very short time at the high sampling rates
used in these experiments. When we make use of the provided averaging optimization,
which is sufficient for studying ensembles of qubit states, the data transfer time becomes
negligible compared to the acquisition time for numbers of averages &1000.The archi-
tecture of the board and the available memory of 2 Mpoints (24 bit accumulation memory
per channel), allowed for a maximum of 216 averages at the full sampling rate, i.e. 1 GS/s
per channel or 2 GS/s on one channel with a resolution of 8 bit. Alternatively an FPGA
programmable Xilinx Xtreme DSP board was used [Bozyigit08, Lang09]. It only features
a sampling period of 10 ns on two channels but can be programmed to perform arbitrary
operations on the measurement data prior to averaging. This allows to very efficiently
acquire and statistically analyze large amounts of individual single shot measurements,
required for single shot qubit readout and field tomography, see [daSilva10, Bozyigit10b].
Once the two quadratures are acquired we typically extract the phase and ampli-
tude information of the transmitted measurement signal. For the case of a homodyne
measurement, where the local oscillator is taken to be at the same frequency as the
measurement signal, the analog hardware mixed the signal down to DC already. We can
simply substitute the measured DC values of I and Q into the formulas Eq. (4.4) in order
to determine amplitude and phase of the measurement tone. Homodyne measurement is
the method of choice when high time resolution is required. The resolution is just limited
4.4. DATA ACQUISITION 47

by the sampling rate of the ADC board and the bandwidth of the IQ mixer, i.e. 500 MHz
in our case. Every nanosecond we acquired a value for I and Q providing amplitude and
phase. Such a procedure was implemented e.g. to precisely analyze the timing of our
DC modulation pulses used for the up-conversion.
Although the DC signals are easy to handle they are much more prone to 1/f noise,
slow drifts and offsets. For the regular measurements therefore, we realized a heterodyne
data acquisition procedure with an IF frequency of typically νIF = 10 − 25 MHz. One can
now think about using either one or both of the two quadratures to reconstruct amplitude
and phase digitally. In order not to be that sensitive to offsets and imbalances in the IQ
mixer or the rest of the amplification chain it turned out to be an effective method only to
sample one of the two quadratures. This goes along with some reduction of bandwidth.
While using both channels one is just limited by the bandwidth of the IQ mixer (similar to
the homodyne acquisition), using only one channel means being limited by the chosen IF
frequency. In a process called single channel digital homodyne, see [Schuster07a], we
acquire and digitally integrate a full IF period of just one IF channel and extract the two
Fourier coefficients given in Eq. (4.1) and Eq. (4.2), which in turn yields amplitude and
phase using Eq. (4.4). In our case, with νIF = 25 MHz and the ADC sampling rate being
2 GS/s, we acquire a total of 80 data points yielding only one data point of amplitude and
phase. The detection bandwidth is therefore set by the IF frequency with a period of 40
ns. This detection bandwidth is still larger than the typical cavity bandwidth ∼ κ and was
therefore found to be appropriate for the experiments implemented in this thesis.
 PTER
5
CHA
D ESIGN AND FABRICATION OF J OSEPHSON
J UNCTION D EVICES

In this section the design and the essential aspects of the thin film fabrication of super-
conducting transmon qubits are summarized. In addition, the fabrication of SQUIDS for
parametric amplifiers is briefly discussed. Reasonable values for the desired Josephson
energy are between 30 and 100 GHz in the case of qubits and above 1 THz for para-
metric amplifiers. Sub-micron ∼ 150 nm x 150 nm to micron sized junctions with a tunnel
barrier thickness of some atomic layers (1 - 2 nm) were fabricated with electron beam
lithography and shadow evaporation and oxidation of aluminum thin films.

5.1 Qubit design

In order to be able to faithfully design and predict the qubit parameters prior to their
fabrication and measurement a detailed simulation was carried out. While the charging
energy E C and the qubit / resonator coupling strength g basically depend on the geometry
of the circuit only, the maximal Josephson energy E Jmax depends exponentially on the
thickness of the tunnel barrier and therefore requires some fine tuning and feedback
from the fabrication process. In case of sufficiently small charging energies, i.e. when the
junction capacitance C J is only a small fraction of the geometric part of the transmon shunt
resistance C S , g and E C can be designed independently of the required and fabricated
E Jmax .

5.1.1 Charging energy and voltage division

Both E C and the qubit-photon coupling g depend on the network of capacitances formed
by the qubit and the resonator. We use the electrostatic solver provided in the software

49
50 CHAPTER 5. DESIGN AND FABRICATION OF JOSEPHSON JUNCTION DEVICES

package Ansoft Maxwell to simulate a 3 dimensional model of our intended qubit de-
sign, see Fig. 5.1 a. In such a model typically a box of size 500µm by 500µm by 500µm
which contains the relevant parts of the chip is considered. The dielectrics and metals
such as sapphire, vacuum and aluminum as well as different electrostatic potentials be-
tween 0 and 1 V are assigned to each independent part of the circuit. The circuit model
effectively consists of only 4 electrodes, i.e. the two ground planes shown in blue, the
resonator center conductor shown in red, the qubit reservoir shown in light blue and the
qubit island shown in violet, see Fig. 5.1 a. The electrostatic solver then tries to find the
capacitance matrix that contains all mutual capacitances C i j between the four electrodes
i , j ∈ {1, 2, 3, 4} using finite element techniques.
Typical values such as the ones extracted for the circuit shown in Fig. 5.1 a are1 C 12 =
25.8 fF, C 13 = 5.6 fF, C 23 = 27.7 fF, C 24 = 3.7 fF and C 34 = 6.7 fF. In contrast, the capacitance
between the ground planes and the resonator center conductor C 14 is typically in the low
pF range. In this limit where C 14 À C g ,C S one can verify that the reduced Cooper pair
box network as shown in Fig. 5.1 c is equivalent to the full transmon network shown in
Fig. 5.1 b by introducing the effective shunt

C S = C 23 +C J (5.1)

and gate capacitances

(C 34 +C 13 )(C 24 +C 12 )
Cg = . (5.2)
C 34 +C 13 +C 24 +C 12
The qubit charging energy is then simply given by E C = e 2 /(2C Σ ) with C Σ = C g + C S . The
voltage division

β = VΣ /Vg = C g /C Σ (5.3)
(C 12 +C 24 )(C 13 +C 34 )
= (5.4)
C 23C 24 +C 13 (C 23 +C 24 ) + (C 23 +C 24 )C 34 +C 12 (C 13 +C 23 +C 34 )

is the fraction of the voltage that drops across the junction VJ if the voltage Vg is ap-
plied between resonator and ground. For a systematic approach involving a full network
analysis see appendix A in [Koch07a].
The qubit’s ground to first excited state transition couples to a resonant photon with
the vacuum Rabi frequency

1 E J 1/4 1
r µ ¶
2g /(2π) = 4βeV (5.5)
2 8EC h

where p
V = ħmωr /(2C r ) (5.6)
is the resonator rms zero point voltage, C r = ²r π/(2ωr Z0 ) the effective resonator ca-
pacitance, νr = ωr /(2π) the fundamental resonator frequency, Z0 = 50 Ω the resonator
1 Note that the capacitance C obtained from such a simulation is not correct since only a small part of the
14
resonator and ground planes is considered.
5.1. QUBIT DESIGN 51

4 2
1

b 4 C34 3 Cg

C24 CJ, EJ EJ
C14 C23 CS
Vg Vg
C13  
1 C12 2

Figure 5.1: Simulation for qubit parameter design. a, Screen shot of an Ansoft Maxwell 3D electrostatic sim-
ulation of an exemplary qubit design. The four electrodes, i.e. ground (blue), reservoir (light blue), island (violet)
and resonator (red) are assigned to the indices i , j = 1, 2, 3 and 4. b, Equivalent circuit diagram comprising all
mutual capacitances. c, Reduced equivalent circuit diagram of the qubit, similar to the one shown in Fig. 2.4 c.

impedance, ²r ≈ 5.9 the effective dielectric constant of the resonator and E J ≈ (mνr +
E C )2 /(8E C ) the Josephson energy of the qubit when in resonance with the resonator
mode m , see also Sections 2.1 and 2.2.
We find very good agreement of this model and a number of fabricated and charac-
terized devices by taking into account a capacitance of both junctions on the order of
C J ∼ 3−10 fF. The junction capacitance depends on the size and thickness of the junction
and is therefore directly related to the chosen Josephson energy. This dependence could
be integrated into our model for future device simulations. While the prediction of the
charging energy is typically very accurate with an error of < 5 %, the predicted qubit cou-
pling strength tends to be lower by 0 − 20 % compared to the actually measured values.
The relative deviation between simulation and measurement is found to be increased for
larger couplings. Despite a number of approximations in the presented simple model,
i.e. reduced circuit model, finite element solving with a typical specified error of < 2%
in the total energy, estimated junction capacitance, finite resolution in fabrication, ana-
lytic approximations of transition matrix element and transition frequency and simplified
calculation of the resonator capacitance using an effective dielectric constant, it works
sufficiently well for most applications of designing new qubits in the typical parameter
range of the transmon and the Cooper pair box.
52 CHAPTER 5. DESIGN AND FABRICATION OF JOSEPHSON JUNCTION DEVICES

5.1.2 Josephson energy

The Josephson energy of the transmon E J is in situ tunable by threading a flux φ through
the DC-SQUID according to Eq. (2.11). The maximal Josephson energy E Jmax obtained
for integer φ/φ0 on the other hand is a junction property that very sensitively depends on
its size, thickness and material. Numerous test junctions were fabricated under varying
conditions and analyzed in order to optimize the fabrication parameters such that an
intended Josephson energy can reproducibly be realized. In this section we point out the
important relations to understand and control this qubit parameter.
According to Josephson’s first relation [Josephson62], the DC Josephson equation

I s = I c sin ϕ, (5.7)

a tunneling current I s of Cooper pairs flows through a thin insulating junction that connects
two superconductors without an externally applied voltage. In Eq. (5.7) I c is the maximal
super current that the junction can support before it turns normal conducting and ϕ =
ϕ2 − ϕ1 is the relative phase difference2 between the two electrodes containing Cooper
pairs in their lowest energy state each described by a single wave function

ψ(r) = ρ(r)e i ϕ1,2 (r) ,

p
(5.8)

with the charge density ρ and the individual condensates quantum phases ϕ1,2 . If a
voltage is built up across the junction, the relative phase ϕ evolves in time according to
the AC-Josephson relation
∂ϕ 2π
= V, (5.9)
∂t Φ0
with Φ0 being the flux quantum Φ0 = h/(2e) ≈ 2.07 · 10−15 Tm2 . After integration one
obtains the voltage-phase relation

2π
ϕ(t ) = ϕ0 + Vt (5.10)
Φ0

where ϕ0 is an integration constant. Substituting Eq. (5.10) into Eq. (5.7) one observes
that the Josephson-current is oscillating sinusoidal with amplitude I c and Josephson-
frequency
V
νJ = , (5.11)
Φ0
where 1 µV corresponds to about 500 MHz. The energy hνJ represents the energy that
is needed to transfer one Cooper pair across the junction.
Is the externally applied current lower than the critical current I c there is no voltage
drop across the junction, which means that there is in principle no energy dissipated in
the contact. Nevertheless, there is energy stored in the Josephson-junction. In order
to apply a current I s that current has to be ramped up starting from I s = 0. During this
2 In the presence of external magnetic fields the gauge invariant phase difference needs to be taken into
account here.
5.1. QUBIT DESIGN 53

time the phase ϕ changes according to Eq. (5.7) hence giving rise to a voltage across
R
the contact, Eq. (5.10). When integrating the product I s (t )V (t )d t over that time one
obtains the Josephson coupling energy

E J (ϕ) = E Jmax (1 − cos ϕ), (5.12)

which denotes the energy due to the overlap of the two wavefunctions of the Cooper pair
condensates and the Josephson energy

Φ0 I c
E Jmax = , (5.13)
2π
is directly proportional to the critical current I c .
The critical current is a measure of how strongly the two phases are coupled through
the tunnel junction. Accordingly, I c is proportional to the area of the tunnel barrier A and
it is useful to define the critical current density J c = I c /A . It can further be shown that
J c scales approximately exponentially with the insulator thickness d of a superconduc-
tor insulator superconductor (SIS) junction I c ∝ e −d . The latter dependence is used to
control the critical current in the fabrication process by changing the oxide layer thick-
ness, see Section 5.2. Typical values of the fabricated devices lie between I c ∼ 20 nA,
i.e. E Jmax /h ∼ 10 GHz for the small qubit junctions of size 200 by 250 nm2 and I c ∼ 4 µA,
i.e. E Jmax /h ∼ 2 THz for larger SQUID junctions of up to 2.5 by 1.5 µm2 used for tunable
resonators. The typical current densities are then in the range of 20 to 50 A/cm2 , for the
3 qubit experiments presented in Chapter 9 accidentally even up to 400 A/cm2 .
The temperature dependence of the critical current is given by [Ambegaokar63]

π∆(T ) ∆(T )
µ ¶
I c Rn = tanh , (5.14)
2e 2k B T

where R n denotes the junction’s electrical resistance in its normal state, T the tempera-
ture of the material and ∆(T ) the superconducting energy gap. For T → 0 the relation can
be simplified as
π∆(0)
I c Rn = , (5.15)
2e
where according to BCS theory [Bardeen57] the superconducting energy gap is related
to the critical temperature of aluminum Tc = 1.176 K as ∆(0) = 1.76k B Tc . Using Eq. (5.15)
and Eq. (5.13) we can therefore easily determine I c and E Jmax for T ¿ Tc simply by mea-
suring the electrical resistance of the junction at room temperature

R q ∆(0)
E Jmax = . (5.16)
Rn 2

Here we have introduced the resistance quantum R q = h/4e 2 . Depending on the ac-
tual critical temperature of the material in use, we can estimate the Josephson en-
ergy expressed as a frequency in GHz from the normal state resistance in kΩ as
E Jmax /h . 140 GHz kΩ/R n .
 54 CHAPTER 5. DESIGN AND FABRICATION OF JOSEPHSON JUNCTION DEVICES

60
a b
50

40
EJ,max GHz

30

20

10

0
2 4 6 8 10 0.0 0.1 0.2 0.3 0.4
Rn k 1Rn 1k

Figure 5.2: Josephson energy versus normal state resistance (a) and calculated conductance (b). Spec-
troscopically measured Josephson energies of seven qubit samples versus measured normal state resistances
of seven test chips fabricated simultaneously (red dots). Junction aging, see Fig. 5.8, between the normal state
resistance measurement and the qubit cool-down is taken into account in the shown data. The size of the
error bars (red rectangles) is determined by statistical variations of the normal state resistances and Joseph-
son energies on the same chip, imprecision of the available spectroscopy data as well as expected systematic
imprecisions of the junction aging estimation and slightly varying junction sizes due to different lithography,
development and substrate wavers.

In Fig. 5.2 we plot Josephson energy values obtained from qubit spectroscopy at
a temperature < 20 mK, see Section 6.4, as a function of room temperature resistance
(conductance) measurements obtained on nominally identical test chips fabricated si-
multaneously. A fit (red line) yields a conversion factor close to its theoretical value
E Jmax /h ≈ 120+30
−20 GHz kΩ/R n . The stated imprecision is indicated by blue lines. Exper-
imental deviations from the expected superconducting gap energy have also been re-
ported previously, see for example Refs. [Bouchiat98, Steinbach01, Court08] and refer-
ences therein.

5.2 Fabrication process

Qubit fabrication is a central cornerstone of experimental circuit QED. Reproducible de-
vice parameters and long coherence times can only be obtained with a robust and care-
fully conducted fabrication process. In this subsection we describe the fabrication of
superconducting qubits with the main steps being chip preparation, electron beam expo-
sure, development, thin film shadow evaporation, oxidation and liftoff. These steps are
indicated in Fig. 5.3 and are discussed in detail in the following. For fabrication recipes
refer to Appendix B. All described fabrication steps were conducted in the cleanroom
FIRST at ETH Zurich.

5.2.1 Chip preparation

Chip cleaning
Similar to any other clean room process, it is important to start with a clean substrate.
This is even more true for nano-fabrication of thin film Josephson junctions since their
5.2. FABRICATION PROCESS 55

critical current can strongly be influenced by the surface roughness of the substrate.
The qubit fabrication is the second major process step after photolithography and sub-
sequent dicing, therefore the sapphire substrate surface can contain resist residues and
debris which are difficult to observe optically without careful analysis, e.g. by means of
an atomic force microscope. Our first step, which was found to improve the fabrication
reproducibility substantially, is therefore to remove organic residues and photoresists in
an O 2 plasma asher. This is followed by rinsing the chip with hot acetone and isopropanol
in an ultrasonic bath, see Appendix B for details.

Resist deposition
The bilayer resist for electron beam lithography is spun onto the chip directly after the
cleaning procedure. Due to very small chip sizes of ∼ 2 mm by 7 mm and the relatively
large resist thickness required, this can be a delicate issue. Before spinning, the chip is
baked on a hot plate at 120◦ C in order to remove water molecules present in the clean-
room atmosphere. We use a pure copolymer of methyl methacrylate and methacrylic
acid, or P(MMA-MAA) in short, thinned in 4% ethyl lactate as the first resist layer of
thickness 630 ± 10 nm, see Fig. 5.3 a. This layer is very sensitive to direct, secondary
and backscattered electron deposition during lithography. It is rapidly developed in iso-
propanol (IPA) and basically serves as a spacer between the substrate and the second
resist layer allowing for a large undercut of up to 1 µm to be realized, see Fig. 5.3 a-d.
After spinning, the chip is baked for 5 minutes at 180◦ C on a hot plate.
This is followed by spinning a layer of low sensitivity but high resolution ploymethyl
methacrylate resist with a large molecular weight of 950000 (PMMA 950k) on top of the

a PMMA / MAA b PMMA / 950k c 30 kV electrons d MIBK development

resist bridge

sapphire sapphire sapphire sapphire

e aluminum f oxidation g aluminum h resist stripping

Josephson junctions

sapphire sapphire sapphire sapphire

Figure 5.3: Thin-film nano fabrication. a, Copolymer of thickness 630 nm is spun on the sapphire substrate. b,
After baking, a PMMA 950k layer of thickness 115 nm is spun on top of the first layer and post-baked. c, Electron
beam lithography of qubits, junctions and SQUIDS using 30 kV electrons. d, Development of the exposed resist
with MIBK in isopropanol (1:3). e, Evaporation of aluminum of thickness 20 nm at a growth rate of 0.5 nm/s at
an angle of typically 0◦ to −45◦ (from vertical direction). f, Aluminum oxidation with Ar − O 2 (85% - 15%) gas
mixture for 1 to 30 minutes at typically 9 Torr. g, Shadow evaporation of aluminum of thickness 80 nm at a rate
of 0.5 nm/s and at an angle of typically 20◦ to 45◦ . h, Resist liftoff in 50◦ C hot acetone for 30 to 120 minutes.
Figure adapted from [Göppl09].
56 CHAPTER 5. DESIGN AND FABRICATION OF JOSEPHSON JUNCTION DEVICES

copolymer. In order to improve the resolution, which can be limited by beam broadening
due to secondary (forward scattered) electrons, a small thickness of only 115 ± 10 nm is
used. The two layers are then baked for 15 minutes on a hotplate.
Exposed PMMA is developed using methyl isobutyl ketone (MIBK). It is not as sensi-
tive as the copolymer, which in the end limits the speed of the exposure, but provides a
much higher resolution and thus is used to define the geometry of the aluminum struc-
tures. The copolymer on the other hand is basically overexposed (and over developed)
and therefore forms a large undercut of up to 1 µm after development. The undercut size
is controlled by choosing an appropriate exposure dose. Their combination therefore
allows for the formation of free standing bridges necessary for the shadow evaporation
technique.
In rare cases it can happen that the bottom layer does not adhere to the clean and
smooth sapphire surface during spinning. To avoid this problem it turned out to be a good
strategy to wait for 5 to 10 seconds between the resist deposition and spinning. Another
solution would be slower spinning speeds and a more gentle spinner acceleration. This
would however also affect the resulting edge bead which we tried to minimize by using
relatively large spinning speeds of 3000 to 4000 rpm and short ramp up times. Typical
edge bead widths are on the order of 500 µm. Much larger or smaller width indicate that
the resist thickness is too large or too small. The substantial chip area covered by the
edge bead is not used for subsequent electron beam lithography.

5.2.2 Electron beam lithography

In a positive resist electron beam lithography (EBL) process, patterns are formed through
the photo scission of the polymer backbone by exposure to a beam of high energy elec-
trons, see 5.3 c. In the subsequent development process the exposed lower molecular
weight resist is removed chemically, see 5.3 d. In our qubit fabrication process we use the
RAITH150 electron beam writer providing electron energies of up to 30 keV. The relevant
steps of EBL and the system specific parameters are described below.

Charge layer
Electron beam lithography on an insulating substrate such as sapphire can be problem-
atic because the electrons deposited in the process of lithography can not easily escape
from the exposed surface area. This causes local electrostatic charging effects which
lead to distortions and shifts in the exposed patterns. We therefore deposit a 10 nm thick
layer of aluminum on top of the double layer resist using an electron beam evaporator3
from Plassys Bestek. This layer facilities the quick discharge of the deposited electrons
during exposure. We furthermore contact the charge layer with the sample holder from
the top using small copper clamps in order to assure a firm contact to ground. The
lithography is not noticeably affected by the additional aluminum layer, as it is basically
transparent for the 30 kV electrons used. Its use turned out to be very important in particu-
lar when sapphire chips without any previous metallization are processed. See Appendix
B for a detailed chip preparation recipe.
3 In general it is not suggested to use an electron beam evaporator for this task since electrons and x-rays
in the evaporation chamber could partly expose the resist. We did not find this an issue however.
5.2. FABRICATION PROCESS 57

File preparations
The geometry of the qubits, SQUIDS and test structures is drawn in the RAITH specific
GDSII editor software4 . A dose factor in the range of 3-5 is assigned to every square
area in the designed structures where the total dose is the multiple of the dose factor
and the chosen base dose of 100 µC/cm2 . The zero coordinate was typically taken to
be the center of the on-chip alignment marks or the middle of the chip in case of empty
test chips without a previous photolithography process. If alignment marks are available
we make use of a semi automatic beam based write-field alignment. The positions of
these marks, typically four crosses at the edges of a square of size 300µm by 300µm,
need to be indicated, the RAITH software specific layer 63 needs to be assigned to
the squares indicating these positions and finally, also the number of averages for the
automatic alignment mark scan needs to be set in the geometry file.
For samples fabricated throughout this thesis no proximity correction was imple-
mented. Instead, we make use of undercut boxes, i.e. additional areas exposed at a low
dose of typically 100 µC/cm2 , in order to increase the undercut at that specific position
if needed. The typical bridge dimensions of the fabricated transmon qubit samples are
100 nm by 200 nm which, depending on the exposure dose, leads to a single Josephson
junction width of ∼ 250 nm, see Fig. 5.4 f. The SQUID sizes were chosen between mini-
mally 1 µm by 1 µm and maximally 4.5 µm by 4.5 µm, depending on the intended inductive
coupling to coils and flux bias lines, see Fig. 5.5 d and e.

Dose tests
The ideal dose and the requirement of additional undercut boxes needs to be verified in a
dose test series. The resulting structures can then be inspected using an scanning elec-
tron microscope (SEM) either directly after development or after shadow evaporation. A
few typical examples of SEM imaged test Cooper pair box (CPB) like structures are de-
picted in Fig. 5.4. Panel a (and d) shows a developed CPB structure with nicely resolved
edges corresponding to a well chosen exposure dose of typically 400 to 450 µC/cm2 and
a correct development time. Panels b and c on the other hand indicate the effect of a too
low dose. In such a case the resist edges are not smooth which leads to rough aluminum
edges after evaporation, see panel c. Additionally large amounts of resist residues can
remain at exposed areas on the substrate. Panel d shows a nice example of developed
resist imaged under an angle of 35◦ . Here one can see the two free standing bridges of
width and thickness < 100 nm and length ∼ 300 nm. Also visible is the undercut box that
was placed above the exposed vertical bar and leads to some additional resist surface
roughness.
In case of a too large dose the undercut size increases and the bridge width shrinks
until it becomes too fragile and breaks down, which results in a shorted Josephson junc-
tion. In panel e a CPB test structure after development, evaporation and resist liftoff is
shown. The shadow, visible on the sapphire substrate in the vicinity of the evaporated
aluminum, indicates the size of the undercut and is likely due to indirect (diffusive) de-
position of aluminum during evaporation. In panel f a close up of one of the two well
4 Certainly the geometries can also be imported from a different CAD software.
58 CHAPTER 5. DESIGN AND FABRICATION OF JOSEPHSON JUNCTION DEVICES

resolved Josephson junctions of size 250 nm by 300 nm is shown. Note, in all panels
but panel e a 5 − 10 nm thin layer of aluminum was evaporated on the structures under
45◦ and rotation prior to their imaging in order to reduce electrostatic charging problems
during SEM imaging.

Exposure preparations
Every qubit fabrication, see Fig. 5.5 a and b, was accompanied by the simultaneous
fabrication of a monitor chip, see Fig. 5.5 c-e, of size 5 mm by 5 mm. This chip is used for
initial focus and write-field alignment and most importantly room temperature Josephson
junction resistance measurements to estimate the maximal Josephson energies.
Both chips are prepared identically and loaded on the waver sample holder. The
resist on the surface of the test chip is scratched a few times close to one of its corners
to facilitate focusing on the otherwise blank sapphire chip. The charge layer of both chips
are contacted to ground using top clamps close to the edge of the chips. After the load
lock is evacuated, the high voltage electron source is ramped all the way up to 30 kV
and the 10 µm aperture is chosen, the previously saved electron beam optics aperture
and stigmatism correction values are loaded and the write-field size of 500 µm2 is set
(720000 fold magnification). Next, the faraday cup positions used to measure the total

a b c

5 μm 5 μm 1 μm

d e f

2 μm 2 μm 400 nm

Figure 5.4: SEM inspection of resist and junctions. a, Developed resist exposed with appropriate dose. b,
Developed resist exposed with too small dose. c, Exposed resist after shadow evaporation exposed with to
small dose. d, Free standing bridges of the developed resist. e, Finished double layer aluminum structure after
resist stripping (only panel imaged without the use of a charge layer). f, Single Josephson junction.
5.2. FABRICATION PROCESS 59

a c d

10 μm 10 μm

b e

10 μm 50 μm 10 μm

Figure 5.5: Optical microscope inspection of resist, qubit and monitor chips. a, Imaged resist after
exposure and development of a typical qubit structure embedded in an aluminum resonator. b, Evaporated
qubit structure after liftoff, embedded in a niobium resonator. c, Typical CPB test structure contacted to DC
probing pads for 4 point resistance measurements. d, Enlarged view of the contacted CPB with a small (and
large in panel e) SQUID loop.

beam current right before exposure are corrected and the lower left corner of the test
chip is approached carefully with the laser controlled sample holder stage.
At this position the origin of the chip coordinate system is set such that (U ,V ) = (0, 0)
corresponds to the center of the monitor chip. After a rough angle correction at the
lower edge of the chip, an initial focussing step is performed on a suitable dirt particle
or scratch. Aperture, stigmatism and focus need to be adjusted iteratively until features
of size < 30 nm are easily resolvable. The edge of a larger dirt particle is a suitable
feature in order to perform a number of manual stage based write-field alignments with
increasing stage displacements and decreasing scan area sizes. In this procedure one
and the same feature is scanned at three displaced positions of the stage by deflection
of the electron beam. The resulting positions are then used to adjust the occurring shifts,
zooms and rotations of the electron beam deflection coordinate system with respect to
the stage or chip coordinates.
In order to further improve the electron beam optics the so called contamination dot
technique is employed next. The coarsely focused beam is concentrated onto the resist
surface for a short time, typically on the order of 5 seconds. Residual gas molecules
are cracked by the incoming electrons and become visible as a circular dot on the resist
60 CHAPTER 5. DESIGN AND FABRICATION OF JOSEPHSON JUNCTION DEVICES

surface. This dot is used as a reference point to further optimize focal length, astigma-
tism correction and aperture alignment until a circular dot of size ≤ 30 nm is obtained.
The contamination dot should be burned in the vicinity of the planned exposure region,
typically 200 − 300 µm away. Any sort of imaging of the exposure region or other critical
areas like resonator gaps needs to be avoided however.5
Once the beam and stage coordinate systems are aligned and the beam properties
are optimized, the exposure dose needs to be calibrated before the exposure is initialized.
This is done by focussing the beam to the closest Faraday cup and measuring the total
flux of deposited electrons. Typical beam current values are 27 pA and 5 nA for the two
used apertures of size 10 µm and 120 µm. The intended area dose D as a function of the
beam current I b is given as
IbT
D= 2 , (5.17)
S
where T is the exposure time or dwell time. The exposure area given as the square of
the step size S is typically chosen to be a single pixel, i.e. S ∼ 8 nm for the chosen write-
field of size 500 µm. The typical beam speed given as v = S/T is then on the order of
v ∼ 3.5 mm/s with a dwell time of T ∼ 10 µs. Large features such as the DC probing pads
were exposed with the large 120 µm aperture, see pads of size 1502 µm2 Fig. 5.5 c. Here
the step size was chosen to be up to 19 pixel, i.e. S ∼ 150 nm and although typically the
dwell time should not be much shorter than a microsecond and the beam speed should
not exceed about 10 mm/s, we went up to 70 mm/s of beam speed in order to minimize
the writing time of these particularly simple and coarse structures.

Exposure
The geometries prepared in the GDSII editor need to be organized in a position list where
the size of the working field is typically set to be the size of the write-field (if no stitching is
needed). Furthermore we define which layer will be exposed and at what UV coordinates
the given structure should be centered at. On the test chip we usually create a position
list with 16 evenly spaced copies of the contacted junction test patterns as shown in
Fig. 5.5 c. First the CPB patterns are exposed. This is followed by a change of aperture
and without any further focusing steps the square shaped DC probing pads are written.
Note, that the beam current needs to be measured and the dwell time as well as the step
size need to be adjusted when the aperture is switched.
Some of the exposure details have been adjusted in order to improve the stabilization
of the beam during exposure on the insulating substrate. We introduced a waiting time of
20 ms during which the beam is blanked between each primitive (each box). Additionally
a flyback factor of 0.15 was set. It defines the settling time between each written line
as multiple of the waiting time and the flyback factor. Also the dynamic compensation
5 In the presence of charging effects or in case the resist was spun already a few days prior to its exposure,
such that the out-gasing of solvents is reduced, it can be very hard to obtain a high quality contamination dot
with good visibility. In such a case it often helps to try to burn dots at various different coordinates of the chip.
If this does not help the charging layer, its grounding or the resist itself need to be checked.
5.2. FABRICATION PROCESS 61

of beam drag effects was turned on. All other exposure settings are chosen as recom-
mended in the RAITH manual. The typical writing time for the monitor chip is on the order
of an hour.
The subsequent exposure procedure of the actual qubit sample is very similar to the
one already described. One difference is that we need to carefully align the qubits with
respect to the resonator. Therefore after the rough origin and angle correction at one of
the chip corners a careful 3 point alignment on three far separated outer alignment marks
is carried out in order to align the stage and chip coordinates. In particular if the resonator
(and the alignment marks) are made of an aluminum thin film it can be very difficult to find
the first alignment mark due to the low atomic weight aluminum compared to the usually
used niobium 6 . Independent of the material it is crucial to have a good knowledge where
to find all required marks such that no critical parts of the chip are exposed accidentally.
It also helps to have determined the focus and beam correction settings already on the
test chip. These are typically not changed when switching to the resonator chip which is
of the same thickness.
Once the chip alignment is finished, the contamination dot technique is employed
again. This time it is carried out directly on one of the inner alignment marks which are
centered at the corners of a square of size 150 µm that is centered at the qubit cutout.
This is reasonable since the alignment marks are on the same focal plane as the qubit
gap (no metallization). This time typically only a few adjustments are required in order to
burn a circular dot of diameter ∼ 20 − 30 nm.
Before the exposure is initialized, an additional and more accurate beam based write-
field alignment is performed. In this version of beam calibration the electron beam scans
three different inner alignment marks which allows to adjust for remaining small rotations,
zooms and shifts between the chip coordinates and the beam deflection coordinates,
without moving the stage. In practice this last alignment step is done by duplicating the
qubit pattern in the position list two times. For the first two copies we set the RAITH
specific layer for alignment marks (layer 63) as the exposure layer. Once the exposure
is started first the alignment marks are scanned and manually corrected. This is done at
least twice and directly followed by the exposure of the actual qubit pattern.

Fabrication of large junctions

Large SQUIDs and SQUID arrays for parametric amplifiers, see Fig. 5.6, are prepared,
exposed and developed similar to the qubit and test chips presented earlier. In order to
achieve large junction areas (and larger Josephson energies) the free standing bridges
need to be longer however. This requires some extra tests in order to assure their me-
chanical stability. We have found that a bridge width to length aspect ratio of 1 to 10 is
possible but starts to become problematic. If the bridge width on the other hand is chosen
too large, typically > 400 nm, the resist directly below the bridge is not entirely cleared off
during development. Free standing bridges of up to 4.5 µm length and ∼ 300 nm width
(exposure width of 400 nm) and minimal SQUID loop sizes of 2.5 µm by 4.5 µm have been
6 Initial alignment could also be done using 10 kV electrons in case the visibility is too low. This was not
required for the work described in this thesis however.
62 CHAPTER 5. DESIGN AND FABRICATION OF JOSEPHSON JUNCTION DEVICES

fabricated, see Fig. 5.6. An extreme example is depicted in Fig. 5.6 b where the entire
support structure, i.e. the resist SQUID loop, is lowered due to a too large undercut. Such
a change in distance from the substrate in combination with the large evaporation angles
of ±45◦ used for these SQUIDs can lead to trapezoidal Josephson junctions.
The images in Fig. 5.6 b and c have been obtained using the Nomarski prism. This
enables to see also the lower aluminum film and therefore the overlap and size of the
junction area (area between the two faint vertical lines). The optimal dose was found to
be somewhat lower than for the typical qubit designs, i.e. ∼ 300 µC/cm2 in contrast to the
usually used 350 − 400 µC/cm2 , which is expected due to the proximity effect.

5.2.3 Development
After electron beam exposure the aluminum charge layer is removed using sodium hy-
droxide (NaOH) solved in water (10 %). The aluminum etch is stopped after about 30
seconds by rinsing the chip in water. After very carefully blow drying the chip with the
nitrogen gas (N2 ) it is useful to check if all aluminum is removed. Optionally one can
already check the exposed structure in the optical microscope. Here it helps to use the
Normarski prism for differential interference contrast microscopy which is most sensitive
to the slightly different optical depth of the exposed resist. In the second step the bilayer

a b

10 μm 10 μm

c d

10 μm 10 μm

Figure 5.6: Optical and SEM images of SQUIDs and SQUID arrays used for parametric amplifiers. a,
Optical image of developed resist for a DC measurement test SQUID. White areas indicate the region of the
undercut. b, SEM picture of developed resist for a very large test SQUID imaged under at an angle of only
15◦ . The two free standing bridges have the dimensions of 4.5 µm by ∼ 300 nm by ∼ 110 nm (exposure dose
300 µC/cm2 ). c and d, Differential interference contrast microscope images of a single SQUID (c) and a SQUID
array (d) contacting the short end of a λ/4 resonator, thus forming a parametric amplifier that can be operated
in reflection.
5.2. FABRICATION PROCESS 63

resist is developed using a mixture of MIBK:IPA (1:3) which provides a very high selec-
tivity, i.e. low sensitivity (high resolution, low throughput) of the PMMA layer and high
sensitivity (low resolution, high throughput) of the copolymer layer. After 50 seconds of
careful moving in the developer the exposed resist is removed and the copolymer typically
shows an undercut on the order of 500 nm or more, see Figs. 5.3 d, 5.5 a and 5.6 a.
If the undercut is too large it limits the lower bound of attainable SQUID sizes and
similarly thin structures like the gaps between the transmon finger capacitors. In such a
case it was found to be most useful to reduce the electron beam area dose in the vicinity
of these structures. The PMMA development is stopped by rinsing in IPA for 10 seconds
and blow drying with the N2 gun.

5.2.4 Shadow evaporation and oxidation

The shadow evaporation technique is an established process known and used for many
years [Dolan88]. It still remains the most suitable process to obtain nanometer sized
Josephson junctions but requires some extra care to obtain reproducible results. Figure
5.3 e-g captures the essentials of the process.

Equipment
We use an electron beam evaporator MEB550 from Plassys Besteck for our process.
If maintained properly it features a stable evaporation and deposition rate of high purity
aluminum (5N) in the range between 0.2 and at least 1 nm/s. The load lock is separated
from the main evaporation chamber which reduces pump down times considerably and
allows for a controlled oxidation without breaking the chamber vacuum. The load lock
can controllably be vented with N2 , O 2 , Ar and a gas mixture of argon and oxygen up
to a chosen constant pressure or flow rate. The sample holder can fully automatically
be rotated (360◦ from arbitrary zero) and tilted (∼ ±60◦ from load, etch or deposition
positions) during and between evaporations. Amongst others the machine also features
an ion gun source for milling oxides and cleaning substrates.

Preliminary steps
The first step is to mount the sample together with the monitor chip with the resist bridges
well aligned in the center of the specially designed sample holder. The sample holder is
then mounted in the load lock and rotated such that the resist bridges are well aligned
with the tilt axis of the sample holder stage. The rest of the process is largely automated
by writing evaporation recipes. Typically it contains the following steps.
The load lock is pumped down to pressures < 2·10−6 Torr in approximately 30 minutes.
Then the chamber and load lock are switched to process mode where the pneumatically
controlled valve between them is opened and the load lock pressure is further reduced
to < 5 · 10−7 Torr supported by the cryogenic chamber pump. The time for pump down is
limited by the amount of humidity that is acquired at the load lock surfaces. It therefore
strongly depends on the time the load lock was kept open. In order to improve the repro-
ducibility of the junction oxidation we then purge the Ar O 2 stainless steel line connecting
the pressurized gas bottle with the load lock by venting the load lock with Ar O 2 gas up to
a pressure of 10 Torr which is again followed by the same pump down procedure.
64 CHAPTER 5. DESIGN AND FABRICATION OF JOSEPHSON JUNCTION DEVICES

If an electrical contact between the resonator and the aluminum layers forming the
SQUID is required, such as in the case of the parametric amplifiers shown in Fig. 5.6 b
and c, a short (±45◦ , 4 seconds each) and intense (500 V beam voltage / 5 mA beam
current, 40 V acceleration voltage, 35 V discharge voltage) use of ion gun milling in a flow
controlled 3.8 sccm Ar atmosphere (chamber pressure 5 · 10−5 Torr) was used to partly
remove the niobium oxide layer of the resonator surface. Since the etch rate of PMMA
is substantially larger than the oxide etch rate, longer milling times have been found to
collapse the resist bridges. Although it is unlikely that a substantial part of the niobium
oxide (or resist scum) is removed during such a short etching time, this process has been
found to reliably establish a DC contact. Without the ion milling treatment on the other
hand a DC contact could not be established.
In the standard qubit fabrication process no such surface treatment was typically
used. In principle it would however be advisable to use one of the established des-
cumming techniques such as O 2 plasma ashing, UV ozone cleaning or ion milling7 at this
point in order to improve the quality and reproducibility of the Josephson junctions. It has
furthermore been shown [Koppinen07] that plasma ashing can reduce the effect of junc-
tion aging, see Subsection 5.3.2. The effect on coherence times of qubit samples would
need to be checked however after different treatments. While all of these techniques
have been tried at least once, no conclusive results could be obtained concerning their
compatibility and usefulness. A more systematic analysis needs to be done in order to
determine process parameters that reliably improve the substrate surface without harm-
ing any of the fragile and thin resist bridges. A moderate O 2 plasma ashing for 10 s at
50 W or a more intense ashing in combination with wider resist bridges > 100 nm would
be reasonable starting points for further improvements.

Deposition and oxidation

After these preliminary steps the material Alpure, this is the aluminum supply reserved
for qubit evaporations only, is selected by rotation of the crucible stack. The electron gun
is switched on and operated at a constant voltage of ∼ 9.5 kV. Then the beam current is
slowly ramped up to 100 mA where it stays for about one minute in order to slowly heat
up the aluminum and also to manually fine tune the alignment of the electron beam to
the center of the material. The chosen beam pattern is simply a single point at a medium
gain setting of 15, a number which is proportional to the beam spot size. This is most
appropriate for a stable evaporation of aluminum which melts before its evaporation. After
heating, the beam current is ramped up to typically 280 − 300 mA.
At this point we measure the deposition rate by tracking the resonance frequency
of a crystal oscillator which slightly changes its mass and resonance frequency of ∼
5 − 6 MHz during deposition. In order to obtain an accurate rate, the specific weight
of the material and the tooling factor, which depends on the specific geometry of the
crystal mount position, needs to be determined and calibrated. A software based PID
controller is employed to feedback control the rate to typically (0.5 ± 0.05) nm/s. Then
the substrate shutter is opened and a counter integrates the acquired film thickness until
7 The latter two can be done directly inside the evaporator just before evaporation.
5.2. FABRICATION PROCESS 65

the target value of typically 20 nm or 15 nm for the first aluminum layer is reached. In
the case of qubit fabrication this layer is deposited under an angle of 0◦ from the direction
perpendicular to the chip surface. In case of large junctions used for parametric amplifiers
a deposition angle of up to −45◦ and thicknesses of up to 30 nm are used to compensate
the angle dependent deposition rate, see Fig. 5.3 e.
The static oxidation is initialized by closing the valve between chamber and load lock
followed by controlled venting of the load lock with a gas mixture of 85% argon and 15%
oxygen.8 The gas inlet pressure is adjusted to about 0.5 bar at the bottle overpressure
valve and subsequently reduced with a needle valve such that the chosen load lock pres-
sure of 9 Torr is reached within about 1 minute. At higher gas inlet speeds the final
oxidation pressure is not reliably met and at even lower speeds the inlet and pump out
times become a substantial fraction of the total oxidation time. The critical current of the
fabricated junctions is then controlled by choosing oxidation times in the range between
2 minutes (parametric amplifiers) and 30 minutes (low transition frequency qubits).
It is important to note that the typically employed pirani/penning pressure gauges are
calibrated for gases of molecular mass similar to that of air, i.e. O 2 and N2 . This gives
a nonlinear dependence for other gases like Ar particularly in the range of pressures
> 1 mbar where the pirani gauge relies on the thermal conductivity which is reduced
for heavier molecules. Instead, an absolute pressure gauge based on the capacitance
measurement of a membrane, a Baratron gauge, calibrated for arbitrary gases up to 10
Torr is employed.
After the oxidation the pump down procedure is started again. This is followed by
tilting the sample holder to the opposite angle, i.e. +20◦ or +45◦ for qubits or paramps
respectively. Then the second aluminum layer of thickness 80 nm or 85 nm is deposited,
see Fig. 5.3 g. Again, in case of large evaporation angles this thickness is increased to
∼ 150 nm to compensate the angle dependent deposition rate. One reason for choosing
the first layer substantially thinner than the second is that this way the top layer thickness
does not substantially vary at the transition from substrate to aluminum overlap, see
Fig. 5.4 f. Secondly, for very thin films on the order of 15 to 10 nm, the superconducting
gap energy is expected to be modified which would enable to engineer quasi-particle
traps [Court08]. If the gap energy of the superconducting qubit island could be increased,
tunneling of quasi particles from the reservoir would be suppressed. Such an effect of
the film thickness on qubit coherence times could be studied in the future.
Directly after the second aluminum deposition the load lock is vented with pure oxygen
up to 10 Torr and the aluminum structure is oxidized for 15 minutes. This step is intended
to generate a uniformly distributed passivation layer of clean aluminumoxide on top of the
structure before the sample is exposed to atmosphere.

5.2.5 Resist stripping and liftoff

When the thin film aluminum shadow evaporation is finished the remaining resist needs
to be stripped from the sapphire chip. We used the least invasive technique based on hot
8 The reduced oxygen partial pressure of the gas mixture allows for longer oxidation times for a given
Josephson critical current. This decreases the effect of slightly variable gas inlet and pump down times and
avoids the danger connected with pumping pure oxygen.
66 CHAPTER 5. DESIGN AND FABRICATION OF JOSEPHSON JUNCTION DEVICES

(50◦ C) acetone only. Due to the large undercut and a comparably large resist thickness
this usually works quite well. In order to facilitate the aluminum layer liftoff we mount
the sample surface vertically in a teflon mount and use a stirrer with ∼ 300 rpm rotation
speed to move the acetone across the chip surface. After 1-2 hours a clean acetone filled
pipette was used to carefully blow off the aluminum layer. Ideally the film lifts off in one
piece only.
Before the chip is dried it is necessary to check that all aluminum came off, in par-
ticular it needs to be checked if the SQUID centers stripped properly. It is difficult to
remove residual metal once the chip is dry. Therefore the chip is placed in a shallow
vessel filled with some IPA and placed under the optical microscope.9 After rinsing in
IPA and blow drying, the chips are ready for inspection10 , optical microscope imaging,
bonding and measurement at cryogenic temperatures. See Figs. 5.4 e, f, 5.5 b-e and 5.6
c, d for finished Josephson junction devices and refer to Appendix B for more details of
the process. The monitor chip on the other hand is now ready for DC junction resistance
measurements in order to determine the junction critical currents and their scatter.

5.3 Room-temperature characterization

5.3.1 DC resistance measurements
We use the DC prober PM8 from Karl Süss in order to determine the room temperature
resistance of DC contacted test Josephson junction SQUIDs, see Fig. 5.5 c-e. The four
point measurement is done by carefully placing two probe needles on each probing pad of
size 150 by 150 µm2 . The current voltage characteristic is then determined by supplying
a current between 10 nA and 60 nA and measuring the voltage drop across the junctions.
From a linear model fit we extract the junction resistances R n in the low kΩ range. One
of the test structures is fabricated with shorted junctions in order to extract part of the
measured resistance which is not due to the junctions. Additionally the shorted test
structure is useful to bring the four probe needles on the same potential before the actual
junctions, which are very sensitive to voltage discharges11 , are measured. In order to
be able to measure the much smaller resistance of these shorted junctions and also of
the large critical current junctions used for parametric amplifiers, larger currents of 1 µA
to 6 µA are supplied. As a rule of tumb the measurement currents should not exceed the
critical current of the Josephson junctions.
9 Problems with this liftoff process typically indicate an issue with either the resist layer or the exposure
details. Most likely it is a too large exposure dose that leads to large undercuts such that part of the top resist
layer collapses. In case the liftoff problem is only marginal, residuals of the metal film can be removed using
short periods (1-10 s) of ultrasonic power (power level 2) in an acetone bath. Note however that this procedure
can potentially lift off parts of the qubit from the sapphire substrate.
10 While SEM inspection has been found to be a harmful for the Josephson junctions, O plasma ashing did
2
not affect the room temperature resistance of a set of fabricated test junctions. Such an additional cleaning step
could be useful to remove resist residues in the vicinity of the junctions that could possibly harm the energy
relaxation times of the qubits. There is however no final conclusion about the effectiveness of such a process
step yet.
11 In order not to damage the junctions, it is furthermore suggested to not turn on or off the prober light
source, or alternatively not to use it at all during the measurements.
 5.3. ROOM-TEMPERATURE CHARACTERIZATION 67

Figure 5.7 depicts the results of a series of resistance measurements on a typical set
of chips that have been fabricated over the period of four month. Each of these chips
contains up to 12 double junction test structures and was oxidized for a variable time
τoxi between 7 and 30 minutes. All junctions were approximately of size 0.05 µm2 each
and oxidized at a pressure of 9 Torr in the argon-oxygen gas, see Subsection 5.2.4. We
observe a monotonic increase in the normal resistance for larger oxidation times. The
error bars indicate the statistical error of the mean value of all test structures on a single
chip. The red line in Fig. 5.7 a is an empirical nonlinear model fit

R n (τoxi ) = A(τoxi + τ0 )2/3 , (5.18)

+0.06
with A = 0.534−0.03 kΩ/min2/3 and τ0 = 1 min accounts for the gas inlet and pump out
times. The accuracy and reproducibility is expected to be limited by variations in the
junction size which is directly proportional to R n and therefore A . The blue lines in Fig. 5.7
a indicate the confidence region due to changes in A . The normal resistance R n is related
to the maximal Josephson energy of the SQUID via Eq. (5.16), see Fig. 5.7 b. Using
Eq. (2.15) we show the expected qubit ground to excited state transition frequencies for
typical charging energies in Fig. 5.7 c. According to these measurements the expected
reproducibility is on the order of νge ± 500 MHz.

a b 60 c
6 14

5 50
12
EJ, max GHz

Νge GHz
RJJ k

4
40 10

3
30 8

2
6
20
1

Τoxi min Τoxi min Τoxi min

0 10 20 30 40 0 10 20 30 40 0 10 20 30 40

Figure 5.7: Transmon Josephson junction oxidation measurements. a, Measured normal state resistance
of up to 12 double junctions of approximately equal size (0.05 µm2 each) on 5 different chips and error of the
plotted mean value. Each chip was oxidized for variable time τoxi at 9 Torr using argon oxygen gas. The red
line is a nonlinear model fit to Eq. (5.18) and the blue lines indicate the region of expected reproducibility. b,
Calculated Josephson energies using Eq. (5.16) (conversion factor of 120) with fit and expected variation. c,
Calculated qubit transition frequencies using Eq. (2.15) for typical charging energies of 200, 300, 400 and 500
MHz (from bottom to top), a fit in red and region of expected reproducibility between the two blue lines.
 68 CHAPTER 5. DESIGN AND FABRICATION OF JOSEPHSON JUNCTION DEVICES

5.3.2 Junction aging

The results shown in Fig. 5.7 have been obtained by measuring the normal state junction
resistance within one hour after evaporation and liftoff. Once at atmosphere the Joseph-
son junctions are found to slowly increase their normal state resistance over time, an
effect which is called junction aging. There is at least two possible mechanisms respon-
sible for this kind of effect. First, the junctions are expected to oxidize further when they
are exposed to air, in particular at the circumference of the formed junctions. The sig-
nificance of such a process is therefore dependent on the area to circumference ratio of
the junctions. In such a scenario it would be expected that junctions with a close to sat-
urated oxidation layer, i.e. low critical current density and low Josephson energy, should
age slower compared to very thin junctions. This type of aging can easily be stopped by
storing the finished samples e.g. in an evacuated or oxygen free environment [Nahum93].
There is however a second mechanism that could be significantly contribute to the ob-
served post oxidation aging process. It is believed that the AlO x barrier that is formed dur-
ing fabrication is rich of oxygen and only slowly relaxes to the stable Al 2 O 3 by a diffusive
mass transport from the oxygen rich surface to the underlying aluminum [Jeurgens02].
Such a process would depend on the sample temperature and a stabilization of junction
resistances has indeed been achieved by annealing the junctions at temperatures above
350◦ C in vacuum [Koppinen07]. Another approach to reduce the effect of aging is to
use a double oxidation scheme [Holmqvist08]. In order to suppress changes on exist-
ing samples, storage at liquid nitrogen temperatures should stop such a diffusive aging
process.

a Τage days b Τage days c EJ,0 GHz

0 5 10 15 0 5 10 15 50 40 35 30 25 22
0.30

10
50 0.25

8 0.20
40
Αage  k  √h 
EJ GHz
RJJ k

0.15
6
30
0.10

4 20
0.05

2 10 0.00
Τage hours Τage hours RJJ,0 k
0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350 2.5 3.0 3.5 4.0 4.5 5.0 5.5

Figure 5.8: Aging of transmon Josephson junctions. a, Measured normal state resistances at time τage
after fabrication and fit to Eq. (5.19). b, Calculated Josephson energy. c, Extracted aging speed versus initial
normal state resistance and linear fit (line).
5.3. ROOM-TEMPERATURE CHARACTERIZATION 69

We also observe the effect of aging and find good agreement with a simple square
root dependence
R n (τage ) = R n,0 + αage τage ,
p
(5.19)
for the same 5 test chips, see Fig. 5.8 a and b. τage denotes the time between fabrication
and measurement and R n,0 is the initial normal resistance measured directly after the
liftoff process shown in Fig. 5.7. The longest time scale aging measurement was done 3
month after fabrication. This measurement could also be explained reasonably well using
Eq. (5.19). For even longer times we would however expect a saturation of the oxide layer
thickness and therefore a slow down of aging. The extracted αage are plotted in Fig. 5.8
c. We find that the rate of post oxidation αage increases approximately linearly with the
initial Josephsonpjunction normal
p resistance over the investigated parameter range as
αage ≈ −0.15 kΩ/ h + 0.08 R n / h.
The methods and models discussed in this section can be used as a reference and
calibration for further sample fabrication with controlled sample parameters. For a de-
tailed step by step fabrication recipe refer to Appendix B. In case further improvements,
such as an additional resist descumming step is implemented, the presented oxidation
and post oxidation dependences are likely to be affected as well.
 PTER
6
CHA
S AMPLE C HARACTERIZATION

After the long list of steps required to fabricate the device and install the measurement
setup, it is a thrilling experience when a new sample is finally cooled down in the dilution
refrigerator. While the RF chip design has typically already been tested at liquid helium
temperatures, the qubit parameters such as Josephson energies and coherence times
can only be accurately determined at ultra-low temperatures. The basic properties of the
sample need to be carefully quantified before more elaborated experiments can be done.
In the following we will briefly review a typical set of circuit QED characterization mea-
surements. Typically these experiments need to be done iteratively with an increasing
level of control and precision.

6.1 Resonator spectrum

The first step is to measure the transmission spectrum of the resonator. This includes the
determination of the resonance frequencies and quality factors of the different resonator
harmonics. Additionally, a transmission spectrum across the full detection bandwidth
should be recorded in order to identify spurious modes.1 If the sample is processed on
a new waver it is furthermore advisable to test and characterize the loss properties of
the niobium thin film. The internal quality factor can be determined by measuring the
transmission of an under-coupled resonator that is processed on the same waver, see for
example Ref. [Göppl09].
1 Note that in a transmission measurement only those modes can be identified which couple to the resonator
input and output ports. Spurious modes which couple to the qubit but not to the resonator need to be identified
in a qubit spectroscopy measurement.

71
72 CHAPTER 6. SAMPLE CHARACTERIZATION

The resonator transmission of a weakly driven and damped harmonic oscillator ex-
hibits a Lorentzian line shape according to Eq. (2.1) and an associated phase shift be-
tween the transmitted and incoming wave as given in Eq. (2.3). In order to measure the
transmission shown in Fig. 6.1 a, a constant input power of -30 dBm is applied with the
RF1 signal generator, see the measurement setup Fig. 4.1. The transmitted amplitude
is then measured versus frequency. More precisely, the I and Q voltage quadratures of
the transmitted signal are sampled at a rate of one sample per ns for 40 µs separately,
see Section 4.4 for details, and averaged on the order of 104 times for each data point.
We then calculate the transmission power P (ν) = (I 2 + Q 2 )/R , with R = 50 Ω being the
characteristic impedance of the measurement circuit.
2 2
Figure 6.1p a-d depicts the measured relative microwave transmission power T /Tmax ,
where T = I 2 +Q 2 is the measured absolute voltage transmission amplitude, in the
vicinity of the harmonic modes m = 1, 2, 3 and 4. The red line is a fit to a Lorentzian
model as given in Eq. (2.1). The quality factors and resonance frequencies of these over-
coupled harmonic resonator modes are indicated. The used sample is shown in Fig. 9.2.
From earlier measurements, see [Göppl08], we would expect a monotonic decrease of
the quality factor, which is not observed in this sample. There is no final conclusion if
the qubits, spurious modes, or another parameter is responsible for the observed quality
factors. A measurement of the transmission amplitude over the full spectral range of the
experimental setup, see Fig. 6.1 e, shows that there is only one visible spurious reso-
nance at ∼ 8.8 GHz. This resonance mode is expected to be either due to a resonance
matching condition of the sample holder geometry or, more likely, the resonance of a so
called slot line or odd mode of the coplanar waveguide, [Pozar93, Simons01]. In order
to suppress this resonance superconducting on-chip bonds connecting the two coplanar

1.
a b c d
T2/ Tmax

Q1 = 1750 Q2 = 990 Q3 = 890 Q4 = 1380

2

1 = 3.345 GHz 2 = 6.729 GHz 3 = 10.011 GHz 4 = 13.439 GHz

0.5

0.
3.3 3.35 3.4 6.7 6.75 6.8 9.95 10 10.05 13.35 13.4 13.45 13.5

5. e
T (mV)

1.

0.1
0.05

4 6 8 10 12 14
RF (GHz)

Figure 6.1: Resonator characterization. a - d, Measured power transmission spectrum of the fundamental
(a) and first three harmonic modes (b-d) of the coplanar waveguide resonator shown in Fig. 9.2 and Lorentzian
fits (red) plotted over an identical frequency range of 150 MHz. e, Measured transmission amplitude T over the
entire detection bandwidth plotted on a logarithmic scale.
6.2. FLUX PERIODICITY 73

waveguide ground planes are used in more recent samples, in the future micro fabricated
air bridges will be used.
In general during the first resonator measurements there will be dispersive or even
resonant interactions with qubits if they are present in the resonator. These interactions
renormalize the resonance frequency and also alter the transmission line-shape. The
measurements shown in Fig. 6.1 were however conducted with the qubit’s transition fre-
quencies maximally detuned from the resonator resonance frequencies. In this case the
interaction of the qubit with the resonator is quite small and the bare resonator properties
can be approximated. A more precise way would be to determine the bare resonator fre-
quency by carefully measuring the qubit induced shifts versus qubit-resonator detuning,
see for example Ref. [Bianchetti10b]. Another way to infer the bare resonator frequency is
to go to the limit of very large coherent or incoherent radiation field powers, see Chapter
10 and Ref. [Reed10a, Bishop10b, Boissonneault10].

6.2 Flux periodicity

A good strategy to quickly estimate the qubit ground to excited state transition frequency
is to investigate the dependence of the resonator frequency shift versus flux applied to
the qubit which tunes its transition frequency. In Fig. 6.2 we present two measurements
similar to the ones shown in Fig. 6.1 a-d. Here the resonator spectrum is measured for
different applied flux bias values controlled by changing a DC current through one of the
superconducting coils shown in Fig. 4.3 c. The strong coherent single qubit single photon
interaction leads to avoided crossings whenever the qubit transition frequency crosses
the resonator frequency. It is important to conduct this measurement with measurement
powers corresponding to not much more than a single intra-resonator photon on average
in order to resolve a clear avoided crossing, see Chapters 8 and 10.
For the devices fabricated and measured in the framework of this thesis it is a good
approximation to assume that the qubit transition frequency follows the simple absolute
cosine dependence of the Josephson energy of a symmetric SQUID. Using Eq. (2.15)
and Eq. (2.11) we find
³q ´
νge ' 8E Jmax | cos (π(φ + φoff )/φ0 )|E C − E C /h, (6.1)

for the ground to first excited state qubit transition frequency νge as a function of flux
φ (violet lines in Fig. 6.2 a). From this relatively simple measurement we can infer not
only the flux coupling, i.e. the flux quantum φ0 in units of the bias voltage (or bias cur-
rent) applied to the coil but also the offset flux φoff associated with that very qubit. We
furthermore can infer the approximate maximal transition frequency νge,max at φ/φ0 = 0
and using Eq. (6.1) determine the Josephson energy E J for an arbitrary φ if the charging
energy E C is known. The accuracy of determining νge,max is drastically increased if the
flux response of additional resonator modes is measured.
We would like to stress that this characterization method is particularly useful to un-
derstand the considerably more complex transmission spectrum of a resonator coupled
74 CHAPTER 6. SAMPLE CHARACTERIZATION

2 1 Flux bias 0 0 1 2
a
6.65
6.6
GHz

6.55
ΝRF

6.5
6.45

5 Flux bias V 0 5

1 Flux bias 0,A 2 3

b
7.1
ΝRF GHz

7.05

7.

0 1 2 Flux bias VA 3 4 5

Figure 6.2: Resonator flux response. a, Resonator transmission (red is high and blue is low transmission)
versus flux bias of a single qubit. The violet line indicates ground to first excited state transition frequency
according to Eq. (6.1). b, Similar measurement in the presence of three qubits (violet, red and yellow lines) as
a function of external flux bias.

to multiple qubits. In Fig. 6.2 b we have depicted a resonator measurement which shows
the flux dependent interaction of three qubits A , B and C as a function of external flux
bias φ0,A . All three qubits couple on a similar level to the large coil A that was used for
this measurement. Using Eq. (6.1) for every qubit we can explain the value of applied flux
at which each splitting is observed, see violet, yellow and red lines in Fig. 6.2 b.
There are however also a number of limitations to this convenient way of inferring
the flux couplings and qubit parameters. One problem is that the cross couplings are
typically not large enough to thread entire flux quanta through the qubit SQUID loop.
Also, the accuracy of determining the maximal transition frequency and therefore the
maximal Josephson energy is limited. Finally, the charging energy needs to be inferred
from a seperate experiment. All of these points can be solved and improved doing careful
qubit spectroscopy measurements, see Section 6.4.

6.3 Dipole coupling strengths

The qubit resonator dipole coupling strength is another important parameter to be ex-
tracted from resonator transmission measurements. By focussing on one of the avoided
crossings, see Fig. 6.3 a, similar to the ones seen in Fig. 6.2, we observe the vacuum
Rabi mode splitting spectrum when the qubit transition frequency is in resonance with the
resonator frequency, see Fig. 6.3 b. On resonance the separation of the two Lorentzian
peaks is minimal and corresponds to twice the single-photon single-atom dipole coupling
 6.3. DIPOLE COUPLING STRENGTHS 75

strength. Absorption and emission of single quantum of energy happens at this splitting
frequency 2g /(2π).
It is interesting to investigate the scaling of the coupling strength extracted from the
Rabi splitting with the harmonic mode number m of the resonator, see Fig. 6.3 c. Accord-
ing to Eq. (5.5) we expect a larger coupling to higher frequency resonator modes due to
p
the increased zero point electric field which is proportional to m , see Eq. (5.6). There is
however a second contribution which increases the coupling strength at higher frequen-
cies even further. When the qubit is tuned in resonance with higher harmonic modes
the Josephson energy and thus the effective dipole moment is changed, see Eqs. (2.33)
and (5.5). If we make use of Eq. (6.1) and consider only qubit frequencies resonant
with the resonator harmonics νge = ν1 m , we arrive at an expression that is approximately
proportional to m s s
ħmω1 mω1 + EC
ħg = E · d ≈ · 2eβ . (6.2)
2C r 16EC
Experimentally, see blue dots in Fig. 6.3 c, we find such an approximately linear depen-
dence when extracting the minimal peak separation in 4 measurements similar to the
one shown in Fig. 6.3 a and b. Comparing the analytic solution Eq. (6.2), see blue line in
Fig. 6.3 c, to an exact numerical diagonalization of the truncated Hamiltonian Eq. (2.34)
(red line) we find only a small correction at higher mode numbers.
It may be interesting to draw a comparison to cavity QED with natural atoms at this
point. The possibility to change both the transition frequency and the effective dipole

a b Νr gC Νr Νr gC c
Ν1 3.345 GHz

150
7.0 0.3
Frequency, ΝRF (GHz)

Coupling gge MHz

6.8
T Tmax

Νr 100
0.2

Νge
6.6
0.1 50

6.4

0. 0
0.466 0.468 0.47 0.472 6.4 6.6 6.8 7.0 0 1 2 3 4
Flux bias C 0 Νrf GHz Harmonic mode, m

Figure 6.3: Vacuum Rabi splitting. a, Resonator transmission (red is high and blue is low transmission) versus
flux bias of a single qubit. The white dash dotted lines indicate the qubits ground to first excited state transition
and the uncoupled resonator frequency respectively. The white dashed lines correspond to the dressed state
eigenenergies obtained by numerically solving the Jaynes-Cummings Hamiltonian. b, Cavity transmission as a
function of probe frequency for the flux value with minimal separation of the two transmission peaks (blue) and
fit to two Lorentzians (red). The frequency splitting 2g /(2π) corresponds to twice the dipole coupling energy
ħg . c, Measured (blue dots) dependence of the vacuum Rabi mode splitting on the harmonic mode index m .
Lines are calculated using Eq. (6.2) (blue) and an exact numerical diagonalization of the truncated Hamiltonian
Eq. (2.34) (red).
76 CHAPTER 6. SAMPLE CHARACTERIZATION

moment over a very large range is one aspect that is novel. It is furthermore also the
order of magnitude of the effective dipole moment which is very different to conventional
cavity QED setups. For typical single photon resonator fields on the order of 13 mV/m,
which corresponds to a voltage of 0.5 µV dropping across the 30 µm gap, and a moderate
voltage division factor β ∼ 0.25 we arrive at an effective dipole moment of d ∼ 1·10−24 C m.
This corresponds to about 2 · 106 Debye and is a ∼ 105 fold increase compared to large
dipole moment molecules and still an increase of up to ∼ 103 compared to atoms in highly
excited Rydberg states.
This new domain of cavity QED allows for the careful investigations of the photon
number and qubit number scaling of the resonant qubit - photon interaction. If on average
more than one coherent or thermal photon is present in the considered resonator mode,
the presented vacuum Rabi mode splitting measurements are strongly altered. In terms
of sample characterization this property can for example be used to calibrate the number
of photons in the resonator. These and related effects due to changes in the number of
coupled qubits are studied in Chapters 8, 9 and 10.

6.4 Qubit spectroscopy

Spectroscopic qubit measurements are an important tool in circuit QED. Spectrally re-
solving both, the ground to first and also the first to second excited state transitions of
the transmon, directly reveals the charging and Josephson energies and fully determines
the qubit spectrum. Additionally, the flux coupling including all cross couplings can be
determined with good accuracy. The photon number dependent shift of the transmon
level splitting, i.e. the AC Stark shift, furthermore represents a useful effect to calibrate
the number of photons in the resonator [Schuster05]. Qubit spectroscopy has also been
used to investigate dressed state [Schuster07b] and side band transitions [Wallraff07]
and also the Lamb shift [Fragner08]. In addition, spurious resonator modes which do
not couple to the input and output transmission lines can only be characterized via spec-
troscopy of the qubit. Finally, also the sometimes present two level fluctuators coupling to
a qubit level transition [Simmonds04, Neeley08, Lisenfeld10] and also possibly occurring
slow transition frequency fluctuations due to flux instabilities [Koch07b] can be analyzed
in detail.
For the dispersive qubit spectroscopy and readout we use a second microwave tone
νspec , see setup in Fig. 4.1. If νspec is in resonance with a qubit transition and one of
the two involved states is populated, the resonant tone induces a change in the qubit
state. This is however only noticeable if the drive rate (Rabi frequency) is comparable to
the intrinsic qubit decay [Schuster05]. The state change then induces a state dependent
shift by ±χ of the renormalized bare resonator resonance frequency, see Subsection
2.3.4. By measuring the resonator transmission at a fixed frequency, e.g. the resonator
frequency with the qubit is in its ground state, we detect this shift as an amplitude and
phase change in the measured resonator transmission. Note that in case the flux bias is
changed, the new optimal measurement frequency needs to be determined for each flux
6.4. QUBIT SPECTROSCOPY 77

setting. For further details refer to Subsection 2.3.4 and Refs. [Schuster05, Wallraff05,
Schreier08, Bianchetti09, Bianchetti10b, Reed10a, Bishop10b, Boissonneault10].
In Fig. 6.4 we present three typical spectroscopic sample characterization measure-
ments. We demonstrate the easiest way to extract the Josephson and charging energies
of the transmon in the far detuned dispersive regime in panel a. In order to increase the
precision of this measurement and in order to also infer the maximal Josephson energy a
flux dependent transmon spectroscopy is demonstrated. Working at elevated cavity pho-
ton numbers such a measurement can also reveal individual cavity photon number (Fock)
states if the qubit dipole strength is sufficiently large, see panel b. In panel c we show a
typical photon number calibration measurement conducted in the far detuned dispersive
regime.

6.4.1 Charging energy

In Fig. 6.4 a we show a measurement of the qubit state dependent phase shift of the
cavity transmission as a function of the applied spectroscopy frequency and a fit to two
Lorentzians. At intermediate spectroscopy power, i.e. the measurement which is offset by
-30 deg, the qubit |g 〉 to |e〉 transition is already broadened and there is a weakly driven
second transition visible in the spectrum. If the spectroscopy power is further increased
the second transition becomes clearly visible. The additional spectroscopic line at lower
frequency corresponds to a two photon process driving the transmon population between
the states |g 〉 and | f 〉.
By fitting the two transmission peaks with two Lorentzian functions, see [Schuster05,
Schreier08], we can extract the two transition frequencies νge and νgf /2. For typical
values of E J /E C the anharmonicity of the transmon [Koch07a] is α = νge − νgf = 2(νge −
νgf /2) ∼ 1.2 E C , which directly gives us an estimate for the charging and, using Eq. (2.15),
also the Josephson energy. For an exact value we solve the Cooper pair box hamiltonian
Eq. (2.13) exactly, e.g. analytically in the phase basis. The measured two values fully
determine the qubit Hamiltonian with the two free parameters E J and E C .
In this measurement scheme any AC Stark shifts of the one and two photon transition
are neglected. To reduce this effect, measurements like the one presented in Fig. 6.4
a should be taken in the far detuned regime and with a very low measurement power,
ideally corresponding to less than a single photon in the resonator on average. An even
better strategy is to implement pulsed qubit spectroscopy where the measurement tone
is only turned on directly after the excitation of the qubit with a long saturation pulse
at frequency νspec , see also Subsection 6.5. This scheme enables to excite the qubit
in the absence of resonator photons and AC Stark shifts. To avoid the need for large
spectroscopy powers one can alternatively use a pump and probe spectroscopy scheme.
In such a scheme low power spectroscopy tones at frequencies νge and νspec are used
to populate the |e〉 state with a pump power dependent probability and probe the single
photon |e〉 to | f 〉 transition directly.

6.4.2 Flux dependence, dressed states and number splitting

In circuit QED the qubit photon coupling strength can be large enough such that even in
the dispersive regime it is difficult to talk about uncoupled qubit or cavity states only. In
78 CHAPTER 6. SAMPLE CHARACTERIZATION

a g e b g0 e0 c navg
30 7.2 g1 e1 0 10 20 30
g f /2
g2 e2 5.35
20 7.1 g3 e3
g1 f0 5.34 10
7.0

Νspec (GHz)
10 g0 f0 2

ΝAC Stark MHz

∆ tr deg

g0 e1 2 5.33 20

Νspec GHz
6.9
0 g1 e2 2
6.8 g0 e2 3 5.32 30
10 g e
6.7
5.31 40
20 g f /2
6.6
5.30 50
30 6.5
3.7 3.8 3.9 0.10 0.05 0.00 0.05 0.10 40 20 17
Νspec GHz Flux, 0 PRF dBm

Figure 6.4: Transmon spectroscopy. a, Phase shift of the fixed frequency cavity transmission measurement
tone as a function of applied spectroscopy frequency for two different spectroscopy powers (blue) and fit to two
Lorentzians (red). In the lower power measurement (trace offset by -30 deg) only the transmon ground to first
excited state transition is clearly visible. As the spectroscopy power is increased also the two photon ground to
second excited state transition is clearly resolved. b, Amplitude of the measurement tone as a function of flux
bias and spectroscopy frequency where blue is high transmission and green is low transmission. If both the
measurement tone and the spectroscopy tone powers are increased, many transitions are observed. In the very
strong coupling regime different photon number states are clearly resolved (number splitting) as indicated in the
legend. c, Measured and fitted qubit ground to excited state transition frequency versus measurement power
in the far detuned dispersive regime. The cavity photon number dependent AC Stark shift of the qubit transition
frequency provides a photon calibration for the applied cavity input power, in this case P n=1 ≈ −32 dBm.

Fig. 6.4 b we present a spectroscopy measurement versus flux bias at elevated measure-
ment and spectroscopy powers. While such a data set is much more difficult to analyze
than one obtained with lower powers or one where both drive and measurement tones
are pulsed, it in the end contains also more useful information. For solving the flux depen-
dent eigenenergies we numerically solve the multilevel Jaynes-Cummings Hamiltonian in
the charge basis Eq. (2.34) by exact diagonalization.
In addition to the transition observed in panel a, i.e. |g 0〉 → |e0〉 where the number now
indicates a tensor product with the photon number state in the Fock basis |n〉, we also
observe the |g 1〉 → |e1〉, |g 2〉 → |e2〉 and very faint |g 3〉 → |e3〉 transitions in Fig. 6.4 b. The
observation of spectroscopic number splitting in the ’strong dispersive coupling regime‘
[Schuster07b], is an incarnation of the AC Stark shift and has only been observed in
circuit QED due to the very large coupling strength attainable. In addition we observe the
two photon side band transition |g 0〉 → |e1〉/2, its number split pendant |g 1〉 → |e2〉/2 and
the second order three photon sideband transition |g 0〉 → |e2〉/3, see also [Wallraff07,
Leek09]. All of these states are mainly dependent on the transmon ground to excited
state transition energy and the qubit photon coupling strength g ge .
States containing 2 or more quanta of energy are however shifted in frequency due
to the presence of the additional transmon states | f 〉, |h〉, ... and an accurate fit would not
be possible without including them in the model. Most sensitive to the qubit parameters
E J and E C are however transitions that directly involve higher energy transmon states
such as the |g 0〉 → | f 0〉/2 already presented in panel a and its number split single photon
6.5. RABI OSCILLATIONS 79

equivalent |g 1〉 → | f 0〉, see Fig. 6.4 b. We argue that fitting a measurement such as the
one presented in Fig. 6.4 b represents the most precise and complete way to determine
all relevant sample parameters, i.e. all free variables in the Jaynes-Cummings Hamilto-
nian Eq. (2.34), i.e. νr , g ge and νge (φ), which is a function of E C , E Jmax , φ and the flux
periodicity φ0 .

6.4.3 AC-Stark shift photon-number calibration

There is one remaining parameter that has not been discussed in enough detail yet. This
is the average number of photons that populate the cavity due to the applied resonant
and coherent measurement tone. We can calibrate the average number of photons in
the resonator n avg in a far detuned dispersive AC Stark shift measurement [Schuster05,
Gambetta06]. Far dispersive refers to the regime where the individual photon number
split qubit lines can not be resolved anymore, i.e. if the shift per photon 2χ ¿ γ2 or in
case of a strong spectroscopy drive where the qubit line is broadened 2χ ¿ ΩRabi /(2π).
The measurement as shown in Fig. 6.4 c is obtained by extracting the ground to excited
state transition frequency νspec from a Gaussian line fit for different applied measurement
powers P RF .
We observe an AC Stark shifted qubit transition frequency with a linear dependence
νge (n avg ) = νge (0) + 2χn avg on the applied measurement power (blue dots). The detuning
∆ dependent qubit shift per photon of 2χ, see Eq. (2.37), can either directly be measured
by probing the cavity frequency for different qubit drive powers, it can approximately be
calculated with Eq. (2.36), or exactly by solving the Hamiltonian Eq. (2.34) for n = 0
and n = 1. For the measurement presented in Fig. 6.4 c the relevant parameters are
∆/(2π) = −1.088 GHz, g /(2π) = 54 MHz and the calculated AC Stark shift per photon of
2χ/(2π) ≈ −1.69 MHz using the relation Eq. (2.36). From the linear fit shown in Fig. 6.4 c
we conclude that a coherent measurement power of P n=1 ≈ −32 dBm populates the cavity
with on average a single photon.
In case of a symmetrically over-coupled resonator with a negligible insertion loss, the
microwave power applied to the input of the resonator is related to the average resonator
photon number, see Subsection 2.1.1 as P RF = n avg hνr κ where 1/κ is the mean photon
lifetime in the resonator. This relation, see also Eq. (2.4), together with the AC Stark shift
calibration directly yields the attenuation along the input line of the experimental setup.
The total attenuation extracted in this way was found to be consistent with the expected
frequency dependent damping coefficients. These are typically not exactly known at
ultra-low temperatures due to a variation of attenuation with temperature.

6.5 Rabi oscillations

Time-resolved Rabi oscillations, see Fig. 6.5, are an important characterization mea-
surement which demonstrates full coherent control of a single qubit [Nakamura01,
Vion03, Wallraff05, Allen87]. Well calibrated Rabi π pulses are furthermore a neces-
sary prerequisite for the vacuum Rabi oscillation measurements presented in Chap-
ter 10, for the single photon source used in [Bozyigit10c, Bozyigit10b] and also for
80 CHAPTER 6. SAMPLE CHARACTERIZATION

a 300 b 0 d 1.0
2
250

Δ tr deg
4 0.8
Pulse length, t ns

6
200
8
0.6
10

Pe
150
c 0
t
0.4
2
100

Δ tr deg
4
Μw pulse meas. pulse 0.2
50 6
8
10 t
0 0.0

t ns
6 8 10 12 14 6 8 10 12 14 0 50 100 150 200 250
t Μs t Μs

Figure 6.5: Rabi oscillations. a, Phase shift of the pulsed measurement tone as a function of time and applied
Rabi pulse length (red is large negative phase shift). b, Maximal phase response versus time for a Rabi pulse
length corresponding to a full ground to excited state qubit rotation by π (dots) and exponential fit (red). The
fitted T1 time is 1.4 µs. c, Same measurement as in panel b for a 2π pulse. Inset shows the used pulse
sequence. d, Extracted excited state population versus applied Rabi pulse length (dots) and exponential cosine
fit (red).

any other of our experiments more closely related to quantum information processing
[Leek07, Leek09, Leek10, Bianchetti09, Bianchetti10b].
The pulse sequence of the presented Rabi oscillation measurement is depicted in the
inset of Fig. 6.5 c. We apply an amplitude modulated square shaped coherent microwave
pulse which is resonant with the qubit ground to excited state transition frequency of vari-
able length ∆t . This is followed by a long measurement pulse in resonance with the
resonator frequency at time t = 5 µs. This scheme enables to excite the qubit in the ab-
sence of resonator photons which would induce an AC Stark shift or dressed dephasing,
[Boissonneault08]. Similar to the spectroscopic experiments presented above, the trans-
mitted measurement tone experiences a phase shift depending on the prepared qubit
state and the prepared qubit state is a function of the applied Rabi pulse length or ampli-
tude, see Subsection 2.3.4.
In Fig. 6.5 a we show the measured phase response as a function of time and the ap-
plied Rabi pulse length ∆t . In this density plot red color corresponds to a large negative
phase shift of the transmitted tone and blue to the case where the transmitted tone does
not see a phase shift due to the qubit. In panels c and d the phase response correspond-
ing to a rotation of the qubit state vector by π and 2π is depicted (dots). Note that the
measured phase is not a well defined quantity before the measurement pulse is applied.
Therefore, the two field quadratures are analyzed and fitted to a full cavity Bloch model
in current experiments, see Ref. [Bianchetti09] for details.
An exponential fit to the π pulse response (panel c) yields an energy relaxation time
of T1 = 1/γ1 ≈ 1.4 µs. By taking the area below the π pulse phase response as a cal-
ibration for a qubit excited state population P e = 1 we can extract the qubit population
for all measured pulse length ∆t , see Fig. 6.5 d (dots). A similar technique has been
used to determine the population of single qubit states in Refs. [Wallraff05, Bianchetti09],
6.5. RABI OSCILLATIONS 81

entangled qubit states in Refs. [Filipp09, Leek09] and multi level system states in
Ref. [Bianchetti10b].
The Rabi oscillations are fitted (red line) using an exponentially decaying cosine with
the Rabi frequency Ω/(2π) ≈ 10 MHz and a decay constant of τRabi ≈ 1.3 µs. In the limit
of strong driving the Rabi decay is related to energy relaxation and pure dephasing rates
as 1/τRabi = (3γ1 + 2γφ )/4, see [Allen87, Bianchetti09]. Accordingly, we obtain a pure
dephasing time Tφ = 1/γφ ≈ 1.9 µs and a coherence time of T2 = 1/(1/(2T1 ) + 1/Tφ ) ≈
1.1 µs. This is a remarkably long T2 time in current experiments. In order to confirm this
number a Ramsey type experiment would be suitable.
Time-resolved quit measurements were not the main focus of this thesis, we refer the
reader to [Bianchetti09, Bianchetti10b] for details of the dispersive qubit readout. For an
explicit measurement of the qubit coherence in Ramsey and spin echo measurements
refer to [Steffen07, Fink07, Bianchetti10a]. Pulsed qubit spectroscopy is a useful tool
to measure the qubit transition frequency in the absence of resonator photons, which
avoids spurious effects such as number splitting, side-band transitions and dressed de-
phasing. It can be implemented simply by probing the resonator response after a long
(∆t À T1 , T 2) qubit saturation pulse for different frequencies νspec , see e.g. [Göppl09].
For flux-pulsed spectroscopy or time-resolved vacuum Rabi oscillations, see Chapter
10, it is furthermore important to carefully characterize the pulsed flux response, see
[Bozyigit10c].
 PTER
7
CHA
C ONCLUSION

In this chapter we have reviewed the relevant experimental techniques required to obtain
the main results presented in Chapters 8, 9 and 10 of this thesis.
While most of the characterization measurements presented in the previous section
would have been called novel only a few years ago, now they form a set of established
techniques to quickly learn about all aspects of a new circuit QED sample. Similarly, the
experimental setup has advanced e.g. by using cryogen free dilution refrigerators that are
easy to operate and offer an unprecedented amount of space for the cold experimental
setup. Another example is the development of the data acquisition system, which now
in principle allows for the implementation of microwave field tomography and real time
feedback algorithms, [Bozyigit08, Lang09, Bozyigit10a].
The field of circuit QED is developing rapidly at all ends indeed.1 Setting up a quan-
tum processor that can solve a given task more quickly than a classical computer remains
an extremely challenging goal however, and the potential scalability of quantum electronic
circuits still needs to be demonstrated. This could e.g. be done by integrating and coher-
ently controlling many qubits in a single resonator, controllable couple grids of resonators
with embedded qubits [Helmer09], or scale up the system size by making use of alterna-
tive tunable coupling techniques that do not require microwave resonators [Bialczak10].
It will be interesting to see if this task can be achieved without degrading the single qubit
performance.
Rather than tackling this ambitions goal, the measurements presented in the following
chapters aim at harnessing the new possibilities of circuit QED to study some of the most
basic quantum optical effects. This includes an investigation of the nature of matter light
1 Refer to the introduction of Chapter 2 for a detailed list of recent progress in circuit QED.

83
84 CHAPTER 7. CONCLUSION

interaction on the level of individual quanta as well as an exploration of the transition from
the quantum to the classical regime of an open quantum system.
 PA R T III

M AIN R ESULTS

85
PART III. MAIN RESULTS 87

In this part of the thesis we quantitatively investigate the strong dipole coupling strength
between individual artificial atoms and single photons. As discussed in Section 2.3, the
single-photon single-qubit interaction can be modeled by the electric dipole coupling en-
ergy g ħ = Ed of a single particle with dipole moment d subject to an electric field E.
According to quantum mechanics the energy of the electromagnetic field is quantized
in the photon number n . Because the field amplitude E is proportional to the square
p
root of its energy, the dipole coupling strength g ħ = Ed is expected to scale with n as
the number of photons is increased. In the bottom left part of the energy level diagram
in Fig. 7.1 we indicate the resonant dressed states of one qubit and up to two photons
p
|1, ±〉 and |2, ±〉 split by 2 ng ħ. In Chapter 8, see also Refs. [Fink08, Fink09a], we probe
the indicated level transitions (blue vertical arrows) between these states with two weak
coherent probe tones. The observed discrete and nonlinearly spaced excitation energies
of this strongly coupled multi-photon / qubit molecule is a direct consequence of the
quantization of the electromagnetic field.
While the quantization of matter in terms of the number of atoms, or in our case the
number of qubits, N is not under debate, it is interesting to show that multiple qubits can

n,+
n α E
n g /π
n photons

n,- classical
νr

n-1,+
νr

∼ n g /π
g = E d
n-1,-
quantum collective
2,+
2 photons

2 g /π N α d
2,-
ν1+,2+
ν1-,2-

2,1+ 3,1+
1,+
1 photon

g /π 2 gge/π 3 gge/π
1,- 2,1- 3,1-
νg0,21+

νg0,31+
νg0,1+

νg0,21-

νg0,31-
νg0,1-

g,0 g,0 g,0

1 qubit 2 qubits 3 qubits

Figure 7.1: Quantum, collective and classical regimes of cavity QED. Dipole coupled dressed states (black
lines) of n = 0, 1, 2, ... photons in resonance with N = 1, 2, 3 qubits given in frequency units. Bare qubit and cavity
states (dashed lines) and allowed transitions (colorized vertical arrows) are indicated.
88 PART III. MAIN RESULTS

be collectively coupled such that they need to be described as a single quantum system.
Quantum mechanics predicts that the magnitude of the collective dipole moment d in
such a case scales with the square root of the number of atoms. In a dual experiment
to the previous one, see Chapter 9 and Ref. [Fink09b], we observe this square root
nonlinearity by probing the level transitions indicated by green vertical arrowsp in Fig. 7.1
in the presence of one, two and three qubits. The increased splitting by N is clear
evidence for the collective interaction of up to three qubits entangled by a single photon.
It has been a long standing goal to observe these quantum nonlinearities on the level
of only a few quanta. Due to the excellent control and the large dipole coupling
p strength g
p
in our circuit QED setup, we could demonstrate the expected n and N nonlinearities
independently for up to 3 photons and up to 3 qubits.
In addition to these experiments which explore the quantum and the collective regime
of matter-light interaction we have also studied the regime of very large photon numbers
in Chapter 10 and Ref. [Fink10], see Fig. 7.1. In this limit, the relative nonlinearity be-
tween two neighboring dressed state doublets disappears, see level transitions indicated
by red vertical arrows in Fig. 7.1, and the level spectrum is that of a classical harmonic
p
oscillator. We observe this gradual disappearance of the n nonlinearity by increasing
the number of thermal photons, or equivalently the effective temperature of the resonator
field. We study this transition from the quantum to the classical regime of cavity QED with
a quantitative model and demonstrate how to sense the effective cavity field temperature
in spectroscopic and time-resolved vacuum Rabi measurements.
 PTER
8
CHA
O BSERVATION OF THE J AYNES -C UMMINGS
p
n N ONLINEARITY

The already very active field of cavity quantum electrodynamics (QED), traditionally stud-
ied in atomic systems [Raimond01, Mabuchi02, Walther06], has recently gained ad-
ditional momentum by the advent of experiments with semiconducting [Reithmaier04,
Yoshie04, Peter05, Hennessy07, Englund07] and superconducting systems [Wallraff04,
Chiorescu04, Johansson06]. In these solid state implementations, novel quantum optics
experiments are enabled by the possibility to engineer many of the characteristic parame-
ters at will. In cavity QED, the observation of the vacuum Rabi mode splitting is a hallmark
experiment aimed at probing the nature of matter-light interaction on the level of a single
quantum. However, this effect can, at least in principle, be explained classically as the
normal mode splitting of two coupled linear oscillators [Zhu90]. It has been suggested
that an observation of the scaling of the resonant atom-photon coupling strength in the
Jaynes-Cummings energy ladder [Walls94] with the square root of photon number n is
sufficient to prove that the system is quantum mechanical in nature [Carmichael96].

8.1 Introduction
In early experiments the quantization of the electromagnetic field was observed in cavity
QED with Rydberg atoms by measurements of collapse and revival [Rempe87]. Simi-
p
larly, the n scaling of the atom/photon coupling strength with the number of photons
n has been observed in the time domain by measuring n photon Rabi oscillations using
coherent states [Brune96] and Fock states [Varcoe00, Bertet02].

89
90 CHAPTER 8. OBSERVATION OF THE JAYNES-CUMMINGS NONLINEARITY

p
In the frequency domain the n scaling can be extracted from spectroscopic cavity
transmission measurements. Initial attempts employing pump and probe spectroscopy
with alkali atoms [Thompson98] were inconclusive. In a recent experiment however the
two-photon vacuum Rabi resonance was resolved using high power nonlinear spec-
troscopy [Schuster08]. At the same time, we observed this quantum nonlinearity by
measuring the spectrum of two photons and one artificial atom in circuit QED [Fink08].
In our measurements the originally proposed pump and probe spectroscopy scheme
[Thompson98] was used. Similarly the n photon Rabi mode splitting was studied using
multi-photon transitions up to n = 5 [Bishop09].
In the dispersive regime the photon number splitting of spectroscopic lines provides
similar evidence for the quantization of microwave radiation in circuit QED [Schuster07b].
Further experimental progress in the time domain has enabled the preparation and detec-
tion of coherent states [Johansson06], photon number states [Hofheinz08] and arbitrary
superpositions of photon number states [Hofheinz09] in circuits. Using Rydberg atoms
quantum jumps of light have been observed [Gleyzes07, Guerlin07] and Wigner functions
of Fock states have been reconstructed [Deleglise08].
These experiments demonstrate the quantum nature of light by measuring the non-
p
linear n scaling of the dipole coupling strength with the discrete number of photons n
in the Jaynes-Cummings model [Carmichael96].
In this thesis, we report a direct spectroscopic observation of this characteristic quan-
tum nonlinearity. Measuring the photonic degree of freedom of the coupled system,
our measurements provide unambiguous, long sought for spectroscopic evidence for the
quantum nature of the resonant atom-field interaction in cavity QED. We explore atom-
photon superposition states involving up to two photons, using a spectroscopic pump and
probe technique. The experiments have been performed in a circuit QED setup [Blais04],
in which ultra strong coupling is realized by the large dipole coupling strength and the
long coherence time of a superconducting qubit embedded in a high quality on-chip mi-
crowave cavity. Circuit QED systems also provide a natural quantum interface between
flying qubits (photons) and stationary qubits for applications in quantum information pro-
cessing and communication (QIPC) [Nielsen00].

8.1.1 The Jaynes-Cummings model

The dynamics of a two level system coupled to a single mode of an electromagnetic field
is described by the Jaynes-Cummings Hamiltonian

Hˆ0 = ħωge σ̂ee + ħωr â † â + ħg ge (σ̂†ge â + â † σ̂ge ) . (8.1)

Here, ωge is the transition frequency between the ground |g 〉 and excited state |e〉 of the
two level system, ωr is the frequency of the field and g ge is the coupling strength between
the two. â † and â are the raising and lowering operators acting on the photon number
states |n〉 of the field and σ̂ij = |i 〉〈 j | are the corresponding operators acting on the qubit
states. When the coherent coupling rate g ge is larger than the rate κ at which photons
are lost from the field and larger than the rate γ at which the two level system looses its
coherence, the strong coupling limit is realized.
8.1. INTRODUCTION 91

Figure 8.1: Level diagram of a resonant (νr = νg e ) cavity QED system. The uncoupled qubit states |g 〉, |e〉
FIG. 1:
and | f 〉 from left to right and the photon states |0〉, |1〉, ..., |n〉 from bottom to top are shown. The dipole coupled
dressed states are shown in blue and a shift due to the | f , 0〉 level is indicated in red. Pump νg 0,1− , νg 0,1+ and
probe ν1−,2− , ν1+,2+ transition frequencies are indicated accordingly.

On resonance (ωge = ωr ) and in the presence of n excitations, p

the new eigen-
states of the coupled systempare the symmetric (|g , n〉 + |e, n − 1〉)/ 2 ≡ |n+〉 and anti-
symmetric (|g , n〉 − |e, n − 1〉)/ 2 ≡ |n−〉 qubit-photon superposition states, see Fig. 8.1.
For n = 1, these states are equivalently observed spectroscopically as a vacuum Rabi
mode splitting [Thompson92, Wallraff04, Boca04, Reithmaier04, Yoshie04, Peter05,
Hennessy07, Englund07] or in time-resolved measurements as vacuum Rabi oscilla-
tions [Brune96, Varcoe00, Bertet02, Johansson06, Hofheinz08] at frequency 2g ge . The
Jaynes-Cummings model predicts a characteristic nonlinear scaling of this frequency as
p
n2g ge with the number of excitations n in the system, see Fig. 8.1. This quantum effect
is in stark contrast to the normal mode splitting of two classical coupled linear oscillators,
which is independent of the oscillator amplitude.
Since the first measurements of the vacuum Rabi mode splitting with, on average,
a single intra-cavity atom [Thompson92] it remains a major goal to clearly observe this
p
characteristic n nonlinearity spectroscopically to prove the quantum nature of the in-
teraction between the two-level system and the radiation field [Zhu90, Carmichael96,
Thompson98]. In time domain measurements of vacuum Rabi oscillations, evidence for
p
this n scaling has been found with circular Rydberg atoms [Brune96] and supercon-
ducting flux qubits [Johansson06] interacting with weak coherent fields. Related experi-
ments have been performed with one and two-photon Fock states [Varcoe00, Bertet02].
We now observe this nonlinearity directly using a scheme similar to the one suggested
in Ref. [Thompson98] by pumping the system selectively into the first doublet |1±〉 and
probing transitions to the second doublet |2±〉. This technique realizes efficient excitation
into higher doublets at small intra cavity photon numbers avoiding unwanted AC Stark
92 CHAPTER 8. OBSERVATION OF THE JAYNES-CUMMINGS NONLINEARITY

shifts occurring in high drive [Bocquillon09] and elevated temperature experiments, see
Refs. [Fink09a, Fink10] and Chapter 10.
In a different regime, when the qubit is detuned by an amount |∆| = |ωge − ωr | À g ge
from the cavity, photon number states and their distribution have recently been observed
using dispersive quantum non-demolition measurements in circuit QED [Schuster07b]
and also in Rydberg atom experiments [Guerlin07].

8.1.2 Experimental setup

In our experiments, in the resonant regime a superconducting qubit playing the role of
an artificial atom is strongly coupled to photons contained in a coplanar waveguide res-
onator in an architecture known as circuit QED [Blais04, Wallraff04]. We use a transmon
[Koch07a, Schreier08], which is a charge-insensitive superconducting qubit design de-
rived from the Cooper pair box (CPB) [Bouchiat98], as the artificial atom. Its transition
frequency is given by ωge /2π ' 8E C E J (Φ) − E C with the single electron charging en-
p

ergy E C ≈ 0.4 GHz, the flux controlled Josephson energy E J (Φ) = E J,max | cos (πΦ/Φ0 )| and
E J,max ≈ 53.5 GHz, as determined in spectroscopic measurements. The cavity is realized
as a coplanar resonator with bare resonance frequency νr ≈ 6.94 GHz and decay rate
κ/2π ≈ 0.9 MHz. Optical images of the sample are shown in Fig. 8.2 a. The large di-
mension of the qubit in the quasi one dimensional resonator provides a very large dipole
coupling strength g ge . A simplified electrical circuit diagram of the setup is shown in
Fig. 8.2 b.

8.2 Coherent dressed states spectroscopy

The system is prepared in its ground state |g , 0〉 by cooling it to temperatures below
20 mK in a dilution refrigerator. We then probe the energies of the lowest doublet |1±〉
measuring the cavity transmission spectrum T and varying the detuning between the

a b 300 K 20 mK 1.5 K 300 K

Cin Cout
ν Vin
0.3 mm
ν Cg
Qubit

Resonator

Figure 8.2: Sample and experimental setup. a, Optical images of the superconducting coplanar waveguide
resonator (top) with the transmon type superconducting qubit embedded at the position indicated. On the
bottom, the qubit with dimensions 300 × 30 µm2 close to the center conductor is shown. b, Simplified circuit
diagram of the experimental setup, similar to the one used in Ref. [Wallraff04]. The qubit is capacitively coupled
to the resonator through C g and the resonator, represented by a parallel LC circuit, is coupled to input and output
transmission lines via the capacitors C in and C out . Using ultra low noise amplifiers and a down-conversion
mixer, the transmitted microwave signal is detected and digitized.
8.2. COHERENT DRESSED STATES SPECTROSCOPY 93

qubit transition frequency νge and the cavity frequency νr by applying a magnetic flux Φ,
see Fig. 8.3 a. The measurement is performed with a weak probe of power P ≈ −137 dBm
applied to the input port of the resonator populating it with a mean photon number of
n̄ ≈ 1.6 on resonance when the qubit is maximally detuned from the resonator. P is
calibrated in a dispersive AC Stark shift measurement [Schuster05]. At half integers of
a flux quantum Φ0 , the qubit energy level separation νge approaches zero. At this point
the bare resonator spectrum peaked at the frequency νr is observed, see Fig. 8.3 b. We
use the measured maximum transmission amplitude to normalize the amplitudes in all
subsequent measurements. At all other detunings |∆| À g ge the qubit dispersively shifts
2
[Koch07a] the cavity frequency νr by χ ' −g ge E C /(∆(∆ − E C )).

8.2.1 Vacuum Rabi splitting

Measuring cavity transmission T as a function of flux bias Φ in the anti-crossing region
yields transmission maxima at frequencies corresponding to transitions to the first doublet
|1±〉 in the Jaynes-Cummings ladder as shown in Fig. 8.3 c. On resonance (∆ = 0), we

Flux bias current, I ( A)

50 100 150 200 250
a b
(GHz)

6.96
rf

6.94
Frequency,

6.92

6.9
0.5 0.0 0.5 0. 0.5 1.

c d
7.2

7.1
(GHz)

7.0
rf
Frequency,

6.9

6.8

6.7

0.598 0.602 0.606 0. 0.025 0.05

Flux bias, 0 Transm., T Tmax

Figure 8.3: Vacuum Rabi mode splitting with a single photon. a, Measured resonator transmission spectra
versus external flux Φ. Blue indicates low and red high transmission T . The solid white line shows dressed
state energies as obtained numerically and the dashed lines indicate the bare resonator frequency νr as well
as the qubit transition frequency νge . b, Resonator transmission T at Φ/Φ0 = 1/2 as indicated with arrows in
panel a, with a Lorentzian line fit in red. c, Resonator transmission T versus Φ close to degeneracy. d, Vacuum
Rabi mode splitting at degeneracy with Lorentzian line fit in red.
 94 CHAPTER 8. OBSERVATION OF THE JAYNES-CUMMINGS NONLINEARITY

extract a coupling strength of g ge /2π = 154 MHz, see Fig. 8.3 d, where the linewidth of
the individual vacuum Rabi split lines is given by δν0 ≈ 2.6 MHz. This corresponds to
a transmission peak separation g ge /π of over 100 linewidths δν0 , clearly demonstrating
that the strong coupling limit is realized [Wallraff04, Schoelkopf08]. Solid white lines in
Figs. 8.3 (and 8.4) are numerically calculated dressed state frequencies with the qubit
and resonator parameters as stated above, being in excellent agreement with the data.
For the calculation, the qubit Hamiltonian is solved exactly in the charge basis. The qubit
states |g 〉 and |e〉 and the flux dependent coupling constant g ge are then incorporated
in the Jaynes-Cummings Hamiltonian Eq. (8.1). Its numeric diagonalization yields the
dressed states of the coupled system without any fit parameters.

8.2.2 Two photon vacuum Rabi splitting

In our pump and probe scheme we first determine the exact energies of the first doublet
|1±〉 at a given flux Φ spectroscopically. We then apply a pump tone at the fixed frequency
νg 0,1− or νg 0,1+ to populate the respective first doublet state |1±〉. A probe tone of the
same power is then scanned over the frequency range of the splitting. This procedure is
repeated for different flux controlled detunings. The transmission at the pump and probe
frequencies is spectrally resolved in a heterodyne detection scheme.
Populating the symmetric state |1+〉, we observe an additional transmission peak at a
probe tone frequency that varies with flux, as shown in Fig. 8.4 a. This peak corresponds
to the transition between the symmetric doublet states |1+〉 and |2+〉 at frequency ν1+,2+ .
Similarly, in Fig. 8.4 c where the antisymmetric state |1−〉 is populated we measure a
transmission peak that corresponds to the transition between the two antisymmetric dou-
blet states |1−〉 and |2−〉 at frequency ν1−,2− . The transmission spectra displayed in
Figs. 8.4 b and d recorded at the values of flux indicated by arrows in Figs. 8.4 a and
c show that the distinct transitions between the different doublets are very well resolved
with separations of tens of linewidths. Transitions between symmetric and antisymmetric

7.2 a b c d

7.1
(GHz)

7.0
rf
Frequency,

6.9

6.8

6.7

6.6
0.598 0.602 0.606 0. 0.01 0.598 0.602 0.606 0. 0.01
Flux bias, 0 Transm., T Tmax Flux bias, 0 Transm., T Tmax

Figure 8.4: Vacuum Rabi mode splitting with two photons. a, Cavity transmission T as in Fig. 8.3 with an
additional pump tone applied to the resonator input at frequency νg 0,1+ populating the |1+〉 state. b, Spectrum
at ∆ = 0, indicated by arrows in a. c, Transmission T with a pump tone applied at νg 0,1− populating the |1−〉
state. d, Spectrum at Φ/Φ0 ≈ 0.606 as indicated by arrows in c.
8.2. COHERENT DRESSED STATES SPECTROSCOPY 95

14.2

14.

13.8

13.6

7.2
(GHz)

7.
rf
Frequency,

6.8

6.6
0.598 0.602 0.606
Flux bias, 0

Figure 8.5: Experimental dressed state energy levels. Measured dressed state energies (blue dots) re-
FIG. 5:
constructed by summing pump and probe frequencies, compared to the calculated uncoupled cavity and qubit
levels (dashed lines), the calculated dressed state energies in the qubit two-level approximation (dotted) and to
the corresponding calculation including the third qubit level (solid red lines).

doublet states are not observed in this experiment, because the flux-dependent transition
matrix elements squared are on average smaller by a factor of 10 and 100 for transitions
|1+〉 → |2−〉 and |1−〉 → |2+〉, respectively, than the corresponding matrix elements be-
tween states of the same symmetry.
The energies of the first doublet |1±〉, split by g ge /π on resonance, are in excellent
agreement with the dressed states theory (solid red lines) over the full range of flux Φ
controlled detunings, see Fig. 8.5. The absolute energies of the second doublet states
|2±〉 are obtained by adding the extracted probe tone frequencies ν1−,2− and ν1+,2+ to
the applied pump frequencies νg 0,1− or νg 0,1+ , see blue dots in Fig. 8.5. For the second
doublet, we observe
p
two peaks split by 1.34 g ge /π on resonance, a value very close to
the expected 2 ∼ 1.41. This small frequency shift can easily be understood, without any
fit parameters, by taking into account a third qubit level | f , 0〉 which is at frequency νgf '
2νge −E C for the transmon type qubit [Koch07a], just below the second doublet states |2±〉.
96 CHAPTER 8. OBSERVATION OF THE JAYNES-CUMMINGS NONLINEARITY

In order to find the energies of the dressed states in the presence of this additional level
we diagonalize the Hamiltonian Hˆ = Hˆ0 + Hˆ1 , where Hˆ1 = ħωgf σ̂ff + ħg ef (σ̂†ef â + â † σ̂ef )
and g ef /2π ≈ 210 MHz (obtained from exact diagonalization) denotes the coupling of the
|e〉 to | f 〉 transition to the cavity. The presence of the | f , 0〉 level is observed to shift
the antisymmetric state |2−〉, being closer in frequency to the | f , 0〉 state, more than
the symmetric state |2+〉 p
, see Figs. 8.1 and 8.5, leading to the small difference of the
observed splitting from 2. The | f , 0〉 state, being dressed by the states |g , 2〉 and |e, 1〉,
is also directly observed in the spectrum via the transition |1−〉 → | f , 0〉 at frequency
ν1−, f 0 , see Fig. 8.4 c. This is in excellent agreement with the dressed states model, see
p
Fig. 8.5. For comparison the dressed states split by 2g ge /π in the absence of the | f , 0〉
state are shown as dotted red lines in Fig. 8.5.
Our experiments clearly demonstrate the quantum nonlinearity of a system of one or
two photons strongly coupled to a single artificial atom in a cavity QED setting. Both sym-
metric and antisymmetric superposition states involving up to two photons are resolved
by many tens of linewidths.

8.3 Weak thermal excitation

In this section we investigate the resonant interaction between a single transmon-type
multilevel artificial atom and weak thermal fields. We explore up to three photon dressed
states of the coupled system in a linear response heterodyne transmission measurement,
similar to the one presented in Section 8.2. The results are also in good quantitative
agreement with the generalized Jaynes-Cummings model. Our data indicates that the
role of thermal fields in resonant cavity QED can be studied in detail using superconduct-
ing circuits. Note that a different sample was used for the measurements presented in
this section.

8.3.1 Sample parameters

For the experiments presented here we use a transmon type qubit, which is a charge-
insensitive superconducting qubit design derived from the Cooper pair box [Bouchiat98],
as
p
the artificial atom [Koch07a, Schreier08]. Its transition frequency is given by ωg,e /2π '
8E C E J (Φ) − E C with the single electron charging energy E C /h ≈ 0.232 GHz, the flux con-
trolled Josephson energy E J (Φ) = E J,max | cos (πΦ/Φ0 )| and E J,max /h ≈ 35.1 GHz, as deter-
mined by spectroscopic measurements. The two characteristic energies E J and E C define
the full level spectrum of the qubit where the eigenenergy of level l is approximately given
as E l ' −E J + 8E C E J (l + 1/2) − E C /12(6l 2 + 6l + 3) [Koch07a]. The cavity is realized as a
p

coplanar resonator with bare resonance frequency νr ≈ 6.44 GHz and photon decay rate
κ/2π ≈ 1.6 MHz. Details of the sample design and fabrication can be found in Chapter 5.
Optical microscope images of the sample are shown in Fig. 8.6 a. A simplified electri-
cal circuit diagram of the setup, including the qusi-thermal photon source, is shown in
Fig. 8.6 b.
8.3. WEAK THERMAL EXCITATION 97

a b 300 K 20 mK 1.5 K 300 K

νrf
300 µm
Cin Cout
νg0, 1

ν1 ,2
Cg
Qubit
I
νr Resonator Coil
Swhite
Quasi-thermal noise source

Figure 8.6: Sample and experimental setup. a, Optical images of the superconducting coplanar waveguide
resonator (top) with the transmon type superconducting qubit embedded at the position indicated. The qubit
of dimensions 300 × 30 µm2 is shown on the bottom. b, Simplified circuit diagram of the experimental setup.
The qubit is capacitively coupled to the resonator through C g and flux tuned with a current biased ( I ) external
miniature superconducting coil. The resonator, represented by a parallel LC circuit, is coupled to input and
output transmission lines via the capacitors C in and C out . A quasi thermal field can be generated by modulating
a microwave carrier tone νr with a large bandwidth pseudo-random noise spectrum S white . The noise signal is
filtered, attenuated and added to pump (νg0,1± , ν1±,2± ) and probe (νrf ) tones. The resulting microwave fields
are further attenuated and applied to the resonator input. After amplification with an ultra low noise amplifier the
transmitted probe signal νrf is downconverted with a local oscillator (LO) and digitized with an analog to digital
converter (ADC).

8.3.2 Generalized Jaynes-Cummings model

The transmon qubit is a superconducting circuit with a nonlinear energy level spectrum.
In many experiments the nonlinearity, which can be adjusted by circuit design and fab-
rication, is sufficient to correctly model it as a two level system. For the experimental
results presented in this work it is however essential to treat the qubit as a multilevel sys-
tem taking into account the coupling of all relevant transmon levels to the cavity photons.
While in Section 8.2 it was sufficient to include only one additional transmon level, here
we present a more general approach.
The physics of a multilevel artificial atom strongly coupled to a single mode of the
electromagnetic field is described by a generalized Jaynes-Cummings model with the
Hamiltonian
X ³ ´
Hˆ = ħωr â † â + ħωl σ̂l ,l + ħg l −1,l (σ̂†l −1,l â + â † σ̂l −1,l ) . (8.2)
l=e,f,h,...

Here, ħωl is the energy of the l ’th excited state |l 〉 of the multilevel artificial atom, ωr is
the frequency of the resonator field and g l −1,l is the coupling strength of the transition
l − 1 → l and one photon. â † and â are the raising and lowering operators acting on the
field with photon number n and σ̂i , j = |i 〉〈 j | are the corresponding operators acting on the
qubit states. In Fig. 8.7, a sketch of the energy level diagram of the resonantly coupled
qubit-resonator system (νr = νg,e ) is shown for up to three photons n = 0, 1, 2, 3 and the
first four transmon levels l = g , e, f and h .
Considering the first two levels of the artificial atom at zero detuning (∆ ≡ νrp−νg,e = 0)
the eigenstates of the coupled system are the symmetric (|g , n〉 + |e, n − 1〉)/ 2 ≡ |n+〉
98 CHAPTER 8. OBSERVATION OF THE JAYNES-CUMMINGS NONLINEARITY

3+
3 3 g ge/π 2

3- 1
0
2+
2 2 g ge/π 1
2- 0

1+ νgh
1 gge/π 0
1- νgf
νr νge

0
g g,0 e f h

Figure 8.7: Sketch of the energy level diagram of a resonant cavity QED system (νr = νg,e ). The uncoupled
qubit states |g 〉, |e〉, | f 〉 and |h〉 (from left to right) and the photon states |0〉, |1〉, |2〉, |3〉, ... (from bottom to top)
are shown with black and red solid lines. The dipole coupled dressed states are shown in blue and the shifts
due to the | f , 0〉, | f , 1〉 and |h, 0〉 levels are indicated by red dashed lines.

p
and antisymmetric (|g , n〉 − |e, n − 1〉)/ 2 ≡ |n−〉 qubit-photon superposition states, see
Fig. 8.7. For n = 1, the coupled one photon one atom eigenstates are split due to the
dipole interaction [Agarwal84, Mondragon83]. The Jaynes-Cummings model predicts a
p
characteristic nonlinear scaling of this frequency as n 2g g,e with the number of excita-
tions n in the system. In the general multilevel case, the higher energy atomic levels
renormalize the dipole coupled dressed state energies. This causes frequency shifts in
p
the excitation spectrum as indicated by red dashed lines in Fig. 8.7 and the simple n
scaling is slightly modified.

8.3.3 Thermal background field

Experimentally, the coupled circuit QED system is prepared in its ground state |g , 0〉 by
cooling it to temperatures below < 20 mK in a dilution refrigerator. Measuring the cav-
ity transmission spectrum T in the anti-crossing region of the qubit transition frequency
νg,e and the cavity frequency νr yields transmission maxima at frequencies correspond-
ing to transitions from |g , 0〉 to the first doublet |1±〉 of the Jaynes-Cummings ladder,
see Fig. 8.8 a. On resonance, see Fig. 8.8 b, we extract the coupling strength of
g g,e /2π = 133 MHz. This is a clear indication that the strong coupling limit g g,e À κ, γ,
with photon decay and qubit decoherence rates κ/2π, γ/2π ∼ 1 MHz, is realized. Solid
lines in Fig. 8.8 a (and Fig. 8.9 a) are numerically calculated dressed state frequencies,
see Fig. 8.8 c, with the qubit and resonator parameters as stated above. For the cal-
culation, the qubit Hamiltonian is diagonalized exactly in the charge basis. The qubit
 8.3. WEAK THERMAL EXCITATION 99

6.7 a b

νg0,1+
6.6 g,0 1+
Probe Frequency, Νrf GHz

ν1+,2+
c
1+ 2+ 2+
6.5 ν1-,2-
νr 2 1
1- 2-
2- 0
νge

ν1+,2+
6.4

ν1-,2-
ν 1-,
f0
1+
νg0,1- 1 0
6.3 1-
g,0 1-
ν1-,f0

νg0,1+
νg0,1-
0.25 0.26 0. 0.01 0.02 0
Flux Bias, 0 Transm., T/Tmax g g,0 e f

Figure 8.8: Vacuum Rabi mode splitting with a weak coherent probe tone. a, Measured resonator trans-
mission spectra versus external flux Φ close to degeneracy. Black indicates low and white high transmission
T . The solid lines show dressed state energies as obtained numerically and the dashed lines indicate the bare
resonator frequency νr and the qubit transition frequency νg,e . b, Resonator transmission T at degeneracy as
indicated with arrows in panel (a), with a line fit to four Lorentzians in red. c, Corresponding energy level spec-
trum (similar to Fig. 8.7) with observed transitions indicated by arrows and small thermal population indicated
with gray circles.

states |g 〉 and |e〉 and the flux dependent coupling constant g g,e are then incorporated
in the Jaynes-Cummings Hamiltonian Eq. (8.2). Its numerical diagonalization yields the
dressed states of the coupled system without any fit parameters.
In addition to the expected spectral lines corresponding to the transition from the
ground state |g , 0〉 to the first doublet states |1±〉, we observe three lines with very low
intensities, see Figs. 8.8 a and b. These additional transitions are visible because the
system is excited by a small thermal background field with a cavity photon number dis-
tribution given by the Bose-Einstein distribution. This thermal field is a consequence of
incomplete thermalization of the room temperature black-body radiation at the input and
output ports of the resonator. This effect can easily be avoided by adding additional cold
attenuation to the input line of the setup.
A quantitative analysis taking into account the two photon states |2±〉 and the pres-
ence of the higher energy qubit levels f and h in Eq. (8.2) yields the transition frequencies
indicated by yellow and red solid lines in Fig. 8.8. For this analysis the coupling constants
g e,f = 184 MHz and g f,h = 221 MHz of higher energy qubit levels to the cavity mode, ob-
tained from exact diagonalization of the qubit Hamiltonian, have been included. We thus
identify two of the additional spectral lines as transitions between the first |1±〉 and second
100 CHAPTER 8. OBSERVATION OF THE JAYNES-CUMMINGS NONLINEARITY

|2±〉 doublet states of the resonant Jaynes Cummings ladder, see Fig. 8.8 c. The low-
est frequency additional spectral line corresponds to a transition from the antisymmetric
doublet state with one photon |1−〉 to the qubit f level without a photon | f , 0〉.
The details of the thermally excited transmission spectrum can be used as a sensitive
probe for the cavity field temperature, see Chapter 10 and [Fink10]. Analyzing the ampli-
tudes of the Rabi splitting spectrum with a quantitative master equation model [Bishop09],
leads to an estimated cavity field temperature of Tc ' 0.2 K which corresponds to a rela-
tively high mean thermal occupation number of n th ' 0.3 photons for the data presented
in Fig. 8.8 a. Careful filtering and thermalization at the input and output ports results in
a typical cavity field temperature of < 90 mK and < 54 mK (n th < 0.03 and n th < 0.003) as
reported for example in Refs. [Fragner08] and [Bishop09], see also [Fink10] and Chap-
ter 10.

8.3.4 Quasi-thermal excitation

In order to access the three photon doublet states |3±〉 of the coupled multi-photon
multilevel-atom system we use externally applied broadband quasi-thermal fields. In this
new approach the dressed eigenstates are populated according to a thermal distribution
depending on the chosen thermal field temperature. This allows to investigate the flux
dependence of all resolvable spectral lines for a given effective resonator mode tempera-
ture Tc in a single experimental run. The spectrum of a one dimensional black body such
as the considered cavity is given as S 1D (ν) = hν/[exp (hν/k B Tc ) − 1]. At a temperature
of Tc > hνr /k B ' 300 mK this energy spectrum is flat with a variation of < 5% within a
500 MHz band centered at the cavity frequency νr ' 6.44 GHz. It is therefore a very good
approximation to make use of a white noise spectrum in the narrow frequency band of
the experimentally investigated transition frequencies centered at νr , in order to gener-
ate quasi-thermal fields of temperatures Tc > 300 mK and populate the considered cavity
mode with thermal photons.
In order to generate such a spectrum a carrier microwave tone at frequency νr is
modulated with a low frequency large bandwidth quasi-random noise spectrum S white
using a mixer, see Fig. 8.6 b. This approximately yields a microwave frequency white
noise spectrum with a bandwidth of 500 MHz centered symmetrically around the cavity
frequency νr . Using tunable attenuators, we can adjust the noise power spectral density
over a wide range of values. For this experiment we adjust it such that the thermal
population of the cavity mode is on the order of n th ∼ 0.9 corresponding to a temperature
of Tc ∼ 0.4 K. This noise spectrum constitutes a reasonable approximation of a black body
thermal noise source for the considered 1D cavity mode, temperature and frequency.
At the same time, the chosen mean thermal population n th ∼ 0.9 allows to observe all
allowed transitions between the ground state |g 〉 and the three photon doublet states |3±〉.
In the presence of the thermal field, we probe the cavity transmission spectrum as
a function of flux in the anticrossing region, see Fig. 8.9 a, with a weak coherent probe
tone νrf in the linear response limit. In this limit the weak probe tone is only a small
perturbation to the field and no multi-photon transitions are induced. In this measurement
we resolve all allowed transitions between the thermally occupied dipole coupled states
 8.3. WEAK THERMAL EXCITATION 101

a b c
3+
6.7
g,0 1+ 3 2
νg0,1+
Probe Frequency, Νrf GHz

6.6 1
ν2+,3+ ν1+,2+ 3-

ν2+,3+
1+ 2+
0

ν2-,3-
ν2-,3- 2+ 3+ 2+
ν1-,2-

νf 0,f1
6.5 1- 2-
νr 2 1
2- 3-
νf0,f1 2- 0

ν1+,2+
6.4 f,0 f,1
νge

ν1-,2-
ν 1-,
f0

g,0 1- 1+
6.3 1 0
ν1-,f0 1- f,0
νg0,1- 1-

νg0,1+
6.2

νg0,1-
0.24 0.25 0.26 0.27 0. 0.001 0
Flux Bias, 0 Transm., TTmax g g,0 e f h

Figure 8.9: Vacuum Rabi mode splitting in the presence of a thermal field. a, Cavity transmission T
as in Fig. 8.8 with an additional quasi-thermal field of temperature Tc ∼ 0.4 K applied to the resonator input
populating the |1±〉, |2±〉 and | f , 0〉 states. b, Transmission spectrum at Φ/Φ0 = 0.25, indicated by arrows in (a).
c, Corresponding energy level spectrum (similar to Fig. 8.7) with observed transitions indicated by arrows and
induced thermal population indicated with gray circles.

in the generalized Jaynes-Cummings ladder. The solid lines are again the calculated
dressed state transition energies which agree well with the observed spectral lines. In
Fig. 8.9 b, a cavity transmission measurement at flux Φ/Φ0 = 0.25 is shown. We identify 8
allowed transitions, compare with Fig. 8.9 c. It follows that the states |1±〉, |2±〉 and also
| f , 0〉 are thermally populated.
In the two-level-atom approximation, transitions between symmetric and antisymmet-
ric doublet states are forbidden at degeneracy. In the generalized Jaynes-Cummings
model the dressed state transition matrix elements are renormalized due to higher qubit
levels. Numerical diagonalization shows that the matrix elements squared, which are
related to the amplitude of the expected spectral lines, are 140 (6) times smaller for
the symmetry changing transitions |1−〉 → |2+〉 (|1+〉 → |2−〉) than for the observed sym-
metry preserving transitions |1−〉 → |2−〉 (|1+〉 → |2+〉) at degeneracy. Similarly, for the
transitions |2−〉 → |3+〉 (|2+〉 → |3−〉) the matrix elements squared are 235 (16) times
smaller than the measured transitions |2−〉 → |3−〉 (|2+〉 → |3+〉) at degeneracy. There-
fore transitions between symmetric and antisymmetric doublet states are not resolved in
our experiment. Symmetry changing transitions populating the antisymmetric states |2−〉
and |3−〉 have larger matrix elements than symmetry changing transitions populating the
symmetric states |2+〉 and |3+〉 because the former are closer in frequency to the qubit
levels | f , 0〉 and |h, 0〉. Similarly, the matrix element for the transition |1−〉 → | f , 0〉 is 34
times larger than for the transition |1+〉 → | f , 0〉 at degeneracy. The latter is therefore also
not observed in the experimental data. In addition to the transition |1−〉 → | f , 0〉, also
seen in the data presented in Fig. 8.8, we observe a transmission line which corresponds
102 CHAPTER 8. OBSERVATION OF THE JAYNES-CUMMINGS NONLINEARITY

to the transition | f , 0〉 → | f , 1〉, see Fig. 8.9 b. A numerical calculation shows that the
matrix element is 5 times larger at degeneracy than |2−〉 → | f , 1〉 and 7 times larger than
| f , 0〉 → |h, 0〉 which in principle could also have been observed. All transitions observed
in the experimental data are in qualitative agreement with the calculated matrix elements
stated above.

20

19.8 3+
Energy, E3± / h [GHz]

19.6
2νr+νg,e
19.4 3- 3 gg,e/π
3νr

19.2

19 νg,h
h,0 f,1 νr+νg,f

13.4

13.2 2+
Energy, E2± / h [GHz]

νr+νg,e
13 2 gg,e/π
2- 2νr
12.8

12.6
f,0
νg,f
12.4

7.
Energy, E1± / h [GHz]

6.8
1+
6.6
νr gg,e/π
6.4 νg,e
1-
6.2

6.

0.24 0.25 0.26 0.27

Flux Bias, 0

Figure 8.10: Measured energy level diagram of the three-photon / artificial atom system. Measured
dressed state energies (blue dots) reconstructed from extracted transition frequencies from data in Fig. 8.9
compared to calculated uncoupled cavity and qubit levels (dashed black lines), the calculated dressed state
energies in the qubit two-level approximation (dotted red lines) and to the corresponding calculation including
four qubit levels (solid red lines).
8.4. THREE PHOTON PUMP AND PROBE SPECTROSCOPY 103

In Fig. 8.10 the full level spectrum reconstructed from the measured data of the bound
photon/atom system up to the third excitation is shown. To calculate the absolute ener-
gies of the levels (blue dots) we extract the transition frequencies from data presented in
Fig. 8.9 with Lorentzian line fits and add them accordingly. For the first doublet states |1±〉
we find excellent agreement with both a simple two-level atom Jaynes-Cummings model
(dotted red lines) as well as the generalized multilevel Jaynes-Cummings model (solid red
lines). In the case of the second |2±〉 and third doublet states |3±〉 we find considerable
frequency shifts with regard to the two level model (compare dotted and solid red lines)
but excellent agreement with the generalized model taking account the additional qubit
levels, as expected from the results in Section 8.2 and Ref. [Fink08]. Furthermore it can
be seen in Fig. 8.10 that the negative anharmonicity of the transmon qubit, together with
the strong dipole coupling, causes large frequency shifts of the antisymmetric dressed
levels |2−〉 and |3−〉 since they are closer in frequency to the qubit levels | f , 0〉 and | f , 1〉,
p
|h, 0〉. This leads to a small reduction of the n nonlinearity which is in agreement with
the numerical results.
We extended our previous work presented in Section 8.2 and Ref. [Fink08] by in-
troducing thermal fields to populate the dressed eigenstates in a resonant cavity QED
system. In addition to the one and two photon/atom superposition states we report a
measurement of the three photon doublet using quasi-thermal fields. The results are in
good agreement with a generalized multilevel-atom Jaynes-Cummings Hamiltonian with-
out any fit parameters. It has been shown that cavity QED with superconducting circuits
can be a sensitive probe for thermal fields. Refer to Chapter 10 and Ref. [Fink10] for
further details on thermal field sensing.

8.4 Three photon pump and probe spectroscopy of dressed states

In order to probe the excitation spectrum of the two and three photon doublet states
we extend the pump and probe spectroscopy scheme presented in Section 8.2 and
Refs. [Thompson98, Fink08], where one of the first doublet states |1±〉 is coherently
pumped and the transition to the states |2±〉 is probed, see also [Fink09a]. This technique
avoids large intra-cavity photon numbers which are needed in high-drive and elevated
temperature experiments. The results presented in this section have been measured
using the sample introduced in Section 8.3.
In analogy to the previous experiments, we wait for the system to equilibrate with its
cold environment to prepare the ground state |g , 0〉. The qubit is then tuned close to
degeneracy where Φ/Φ0 ≈ 0.25. We weakly probe the resonator transmission spectrum
as shown in Fig. 8.11 (blue lines).
In a second step we apply a pump tone at frequency νg0,1− (νg0,1+ ) populating the
dressed state |1−〉 (|1+〉) and probe the system again, see Fig. 8.11 a (b) yellow line.
Clearly the transitions |1±〉 → |2±〉 become visible at the calculated eigenenergy which
is indicated with yellow vertical arrows in Figs. 8.11 a, b and c. We also note that the
state | f , 0〉 is populated by the probe tone via the transition |1−〉 → | f , 0〉, see red arrows
in Figs. 8.11 a and c.
104 CHAPTER 8. OBSERVATION OF THE JAYNES-CUMMINGS NONLINEARITY

Probe frequency, ΝRF GHz

c
6.3 6.4 6.5 6.6 3+
0.04 a x3
Transmission, T Tmax

3 2
0.03
2 3
3-

ν2+,3+
0.02
0

ν2-,3-
1 2
2+
1 f,0
0.01
2 1
0.
g,0 1 g,0 1 2- 0

ν1+,2+
ν1-,2-
0.04 b x3 ν f0
1-,
1+
Transmission, T Tmax

0.03 1 0
2 3 1-
0.02

νg0,1+
1 2

νg0,1-
0.01
0
0.
g,0 1 g,0 1 g g,0 e f h

Figure 8.11: Vacuum Rabi mode splitting with one probe tone and zero, one or two coherent pump tones
at flux Φ/Φ0 ≈ 0.25 close to degeneracy. a, Measured resonator transmission spectra T /Tmax without a pump
tone (blue) with one pump tone νg0,1− (yellow) and with two pump tones νg0,1− and ν1−,2− (green) applied. The
spectra are offset by 0, 0.02 and 0.03 T /Tmax and the boxed area is scaled in amplitude by a factor of 3 for
better visibility. Vertical arrows indicate numerically calculated transition frequencies. b, Similar measurement
of resonator transmission T for the case of no (blue), |1+〉 (yellow) and both |1+〉 and |2+〉 symmetric dressed
states pumped coherently. c, Energy level spectrum (similar to Fig. 8.7) with relevant transitions indicated by
arrows.

In a last step we apply two pump tones at frequencies νg0,1− and ν1−,2− , see
Fig. 8.11 a (green line), or at frequencies νg0,1+ and ν1+,2+ , see Fig. 8.11 b (green line)
respectively. The three-photon one-qubit dressed state transitions |2±〉 → |3±〉 become
visible in the spectrum, see green vertical arrows. At the same time, transitions from
the ground state are found to saturate considerably when the pump tones are turned on,
compare the amplitudes of the spectral lines at the frequency indicated by the left blue
arrow in Fig. 8.11 a, or similarly by the right blue arrow in figure Fig. 8.11 b. This is ex-
pected since the occupation probability of the ground state is reduced and the transition
starts to be saturated when the pump tones are turned on.
Again, the observed transition frequencies are in good agreement with the calculated
dressed state transition energies indicated by vertical arrows in Figs. 8.11 a and b. Ad-
ditional spectral lines with low intensity, see Fig. 8.11 a and b blue lines, occur because
of a small probability of occupation of the first doublet due to the residual thermal field,
see Section 8.3. The transition | f , 0〉 → | f , 1〉 is not observed because the level | f , 0〉 is
neither thermally, nor coherently populated here.

8.5 Conclusion
Our experiments clearly demonstrate the quantum non-linearity of a system of one, two
and three photons strongly coupled to a single artificial atom in a cavity QED setting.
8.5. CONCLUSION 105

Both symmetric and antisymmetric superposition states involving two and three photons
are resolved by tens of linewidths. In our circuit QED system, excited states |n±〉 with
n > 2 are observable by pumping the system with thermal photons, by applying coherent
drive fields in a pump and probe scheme and also by applying strong drive fields which in-
duce multi-photon transitions, see [Bishop09]. We have also observed that higher excited
states of the artificial atom induce energy shifts in the coupled atom-photon states. These
shifts should also be observable in time-resolved measurements of Rabi-oscillations with
photon number states.
The observed very strong nonlinearity on the level of single or few quanta could be
used for the realization of a single photon transistor, parametric down-conversion, and
for the generation and detection of individual microwave photons. A more detailed quan-
titative analysis of the thermally excited vacuum Rabi spectra could be of interest in the
context of environmentally induced dissipation and decoherence, thermal field sensing
and the cross-over from the quantum to the classical regime of cavity QED, see Chap-
ter 10. Related results have been reported in an atomic system [Schuster08] and similar
effects may be interesting to approach also in semiconducting cavity QED systems.
 PTER
9
CHA
C OLLECTIVE M ULTI -Q UBIT I NTERACTION

We present an ideal realization of the Tavis-Cummings model in the absence of atom

number and coupling fluctuations by embedding a discrete number of fully controllable
superconducting qubits at fixed positions into a transmission line resonator. Measuring
the vacuum Rabi mode splitting with one, two and three qubits strongly coupled to the
cavity field, we explore
p both bright and dark dressed collective multi-qubit states and ob-
serve the discrete N scaling of the collective dipole coupling strength. Our experiments
demonstrate a novel approach to explore collective states, such as the W -state, in a fully
globally and locally controllable quantum system. Our scalable approach is interesting
for solid-state quantum information processing and for fundamental multi-atom quantum
optics experiments with fixed atom numbers.

9.1 Introduction
In the early 1950’s, Dicke realized that under certain conditions a gas of radiating
molecules shows the collective behavior of a single quantum system [Dicke54]. The
idealized situation in which N two-level systems with identical dipole coupling are res-
onantly interacting with a single mode of the electromagnetic field was analyzed by
Tavis and Cummings [Tavis68].
p This model predicts the collective N -atom interaction
strength to be G N = g j N , where g j is the dipole coupling strength of each individ-
ual atom j . In fact, in first cavity QED experiments the normal mode splitting, ob-
servable in the cavity transmission spectrum [Agarwal84, Leslie04], was demonstrated
with on average N̄ > 1 atoms in optical [Raizen89, Zhu90] and microwavep [Bernardot92]
cavities to overcome the relatively weak dipole coupling g j . The N scaling has
been observed in the regime of a small mean number of atoms N̄ with dilute atomic

107
108 CHAPTER 9. COLLECTIVE MULTI-QUBIT INTERACTION

beams [Bernardot92, Childs96, Thompson98] and fountains [Münstermann00] cross-

ing a high-finesse cavity. In these experiments, spatial variations of the atom po-
sitions and Poissonian fluctuations in the atom number inherent to an atomic beam
[Childs96, Carmichael99, Leslie04] are unavoidable. In a different limit where the cav-
ity was populated with a very large number of ultra-cold 87 Rb atoms [Tuchman06] p and
more recently with Bose-Einstein condensates [Brennecke07, Colombe07] the N non-
linearity was also demonstrated. However, the number of interacting atoms is typically
only known to about ∼ 10% [Brennecke07].
Here we present an experiment in which the Tavis-Cummings model is studied for a
discrete set of fully controllable artificial atoms at fixed positions and with virtually identical
couplings to a resonant cavity mode. The investigated situation is sketched in Fig. 9.1 a,
depicting an optical analog where three two-state atoms are deterministically positioned
at electric field antinodes of a cavity mode where the coupling is maximum.

9.1.1 Experimental setup

In our circuit QED [Wallraff04, Schoelkopf08] realization of this configuration (Fig. 9.1 b),
three transmon-type [Koch07a] superconducting qubits are embedded in a microwave
resonator which contains a quantized radiation field. The cavity is realized as a copla-
nar waveguide resonator with a first harmonic full wavelength resonance frequency of
ωr /2π = 6.729 GHz and a photon decay rate of κ/2π = 6.8 MHz. The qubits are posi-
tioned at the antinodes of the first harmonic standing wave electric field. The transition
frequencyq between ground |g 〉 and first excited state |e〉 of qubit j , approximately given
by ω j ≈ 8E C j E J j (Φ j )/ħ − E C j /ħ, is controllable through the flux dependent Josephson

a
mm
19

Figure 9.1: Schematic of the experimental setup. a, Optical analog. Three two-state atoms are identically
coupled to a cavity mode with photon decay rate κ, atomic energy relaxation rate γ and collective coupling
strength G N . b, Schematic of the investigated system. The coplanar waveguide resonator is shown in light
blue, the transmon qubits A, B and C in violet and the first harmonic of the standing wave electric field in red.
9.2. COLLECTIVE DIPOLE COUPLING 109

energy E J j (Φ j ) = E J max j | cos (πΦ j /Φ0 )| [Koch07a]. Here E C j is the single electron charg-
ing energy, E J max j the maximum Josephson energy at flux Φ j = 0 and Φ0 the magnetic
flux quantum. Effectively independent flux control of each qubit is achieved by applying
magnetic fields with three external miniature current biased coils (Fig. 9.2 a) where we
take into account all cross-couplings, see Table 9.1, by inverting the full coupling matrix

ΦA 1/Φ0aA 1/Φ0bA 1/Φ0cA Va ΦoffA

      
ΦB  = 1/Φ0 1/Φ0bB 1/Φ0cB  Vb  + ΦoffB  . (9.1)
aB
ΦC 1/Φ0aC 1/Φ0bC 1/Φ0cC Vc ΦoffC

Optical images of the investigated sample are depicted in Fig. 9.2 b and c. The
resonator was fabricated employing optical lithography and aluminum evaporation tech-
niques on a Sapphire substrate. All qubits were fabricated with electron beam lithography
and standard Al/AlOx /Al shadow evaporation techniques. Table 9.2 states the individual
qubit parameters obtained from spectroscopic measurements.

9.1.2 Tavis-Cummings model

The physics of our system is described by the Tavis-Cummings Hamiltonian [Tavis68]
µ ¶
X ħ
HˆTC = ħωr â † â + ω j σ̂zj + ħg j (â † σ̂−j + σ̂+j â) , (9.2)
j 2

where g j is the coupling strength between the field and qubit j . â † and â are the creation
and annihilation operators of the field, σ̂+j and σ̂−j are the corresponding operators acting
on the qubit j , and σ̂zj is a Pauli operator. The ground state |g , g , g 〉 ⊗ |0〉 of the three-
qubit/cavity system is prepared by cooling the microchip to a temperature of 20 mK in a
dilution refrigerator.

9.2 Collective dipole coupling

First we investigate the resonant coupling of the |g 〉 to |e〉 transition of qubit A to the first
harmonic mode of the resonator. We measure the anti-crossing between qubit A (νA ) and
the cavity (νr ) by populating the resonator with much less than a photon on average. We
record the resulting transmission spectrum T versus magnetic flux ΦA controlled detuning
of qubit A (Fig. 9.3 a). Qubits B and C remain maximally detuned from the resonator at

Qubit Φ0a (V) Φ0b (V) Φ0c (V) Φoff (Φ0 )

A -6.55 7.36 -228 -0.018
B -31.8 6.67 59.0 -0.027
C 250 5.47 5.28 -0.430

Table 9.1: Coil-coupling parameters and flux offsets of three qubits coupled to three superconducting coils. The
stated values are obtained from flux dependent spectroscopy.
110 CHAPTER 9. COLLECTIVE MULTI-QUBIT INTERACTION

µ µ

Figure 9.2: Circuit diagram and false color optical images of the sample. a, Simplified electrical circuit
diagram of the experimental setup. The waveguide resonator operated at a temperature of 20 mK, indicated
as LC oscillator with frequency ωr , is coupled to input and output leads with the capacitors C in and C out .
Qubits A, B and C are controlled with external current biased coils ( I A,B,C ) and coupled to the resonator via
identical capacitors C g . A transmission measurement is performed by applying a measurement tone νrf to the
input port of the resonator, amplifying the transmitted signal and digitizing it with an analog-to-digital converter
(ADC) after down-conversion with a local oscillator (LO) in a heterodyne detection scheme. b, The coplanar
microwave resonator is shown truncated in gray (substrate in dark green) and the locations of qubits A, B and C
are indicated. c, Top, magnified view of transmon qubit B (violet) embedded between ground plane and center
conductor of the resonator. Bottom left, qubits A and C, of same dimensions as qubit B, are shown at reduced
scale. Bottom right, magnified view of SQUID loop of qubit B.

ΦB = ΦC = Φ0 /2 where they do not affect the measurement. At finite detuning (left hand
side of Fig. 9.3 a) we observe a shift of the resonator spectrum which increases with
decreasing detuning due to the dispersive interaction with qubit A.
On resonance (ω j = ωr ) and in the presence of just one two level system (N = 1),
Eq. (9.2) reduces to the Jaynes-Cummings Hamiltonian [Jaynes63]. The eigenstates
|N , n ±〉 of this system in the presence of a single excitation n = 1 are the symmetric
p
and anti-symmetric qubit-photon superpositions |1, 1±〉 = 1/ 2 (|g , 1〉 ± |e, 0〉) (Fig. 9.4 a)
where the excitation is equally shared between qubit and photon. Accordingly, we ob-
serve a clean vacuum Rabi mode splitting spectrum formed by the states |1, 1±〉 (Fig. 9.3
b). From analogous measurements performed on qubits B and C (not shown) we obtain

Qubit j E C j /h (MHz) E Jmax j /h (GHz) g j /2π(MHz)

A 283 224 83.7
B 287 226 -85.7
C 294 214 85.1
Table 9.2: Qubit and qubit-resonator coupling parameters. The single electron charging energy E C j , the max-
imum Josephson energy E Jmax j extracted from spectroscopic measurements and the coupling strengths g j
obtained from resonator transmission measurements for qubits A , B and C .
9.2. COLLECTIVE DIPOLE COUPLING 111

a b c d e f

Figure 9.3: Vacuum Rabi mode splitting with one, two and three qubits. a, Measured resonator transmis-
sion spectrum T (blue, low and red, high transmission) versus normalized external flux bias ΦA /Φ0 of qubit
A . Dash-dotted white lines indicate bare resonator νr and qubit νA frequencies and dashed white lines are
calculated transition frequencies νg0,Nn± between |g , 0〉 and |N , n ±〉. b, Resonator transmission T /Tmax at
degeneracy normalized to the maximum resonator transmission Tmax measured at ΦA,B,C = Φ0 /2 (not shown),
as indicated with arrows in panel a. Red line is a fit to two Lorentzians. c, Resonator transmission spectrum
T /Tmax versus external flux bias ΦC /Φ0 of qubit C with qubit A degenerate with the resonator (νA = νr ). d,
Transmission spectrum T /Tmax at flux as indicated in panel c. e, Transmission spectrum versus flux ΦB /Φ0
with both qubits A and C at degeneracy (νA = νC = νr ). The white dashed line at frequency νg0,31d1,2 = νr
indicates the dark state occurring at degeneracy. f, Transmission spectrum T /Tmax at flux indicated in panel e.

the single qubit coupling constants g j listed in Table 9.2. The magnitudes of the coupling
strengths are virtually identical with a scatter of only a few MHz. The strong coupling of an
individual photon and an individual two-level system has been observed in a wealth of dif-
ferent realizations of cavity QED both spectroscopically [Wallraff04, Boca04, Khitrova06]
and in time-resolved experiments [Brune96, Hofheinz08]. The regime of multiple excita-
tions n which proves field quantization in these systems has been reported both in the
time-resolved results cited above and more recently also in spectroscopic measurements
[Schuster08, Fink08, Bishop09].
In a next step, we maintain qubit A at degeneracy (νA = νr ), where we observed the
one-photon one-qubit doublet (see left part of Fig. 9.3 c). Qubit B remains far detuned
(ΦB = Φ0 /2) for the entire measurement. Qubit C is then tuned through the already cou-
pled states from lower to higher values of flux ΦC . In this case, the doublet states |1, 1±〉
of qubit A are found to be dispersively shifted due to non-resonant interaction with qubit C
(Fig. 9.3 c). When both qubits and the resonator are exactly in resonance, the transmis-
sion spectrum T (Fig. 9.3 d) shows only two distinct maxima corresponding to the doublet
p
|2, 1±〉 = 1/ 2 |g , g 〉 ⊗ |1〉 ± 1/2 (|e, g 〉 + |g , e〉) ⊗ |0〉 with eigenenergies ħ(ωr ± G 2 ). Here
a single excitation is shared between one photon, with probability 1/2, and two qubits,
with probability 1/4 each (Fig. 9.4 b). Both states have a photonic component and can
be excited from the ground state |g , g , g 〉 ⊗ |0〉 by irradiating the cavity with light. These
are thus referred to as bright states. In general wepexpect N + n = 3 eigenstates for two
qubits and one photon. The third state |2, 1d 〉 = 1/ 2(|e, g 〉 − |g , e〉) ⊗ |0〉 with energy ħωr
at degeneracy has no matrix element with a cavity excitation and is referred to as a dark
112 CHAPTER 9. COLLECTIVE MULTI-QUBIT INTERACTION

a 1,1

1,1
b 2,1
2
2,1
c 3,1
3
3,1

Figure 9.4: Level diagram representing the total energy of one (a) two (b) and three qubits (c) resonantly
coupled to a single photon. Bare energy levels of the qubits |g 〉, |e〉 and the cavity |0〉, |1〉 are shown in
black. The bright dressed energy levels |N , n ±〉, with N the number of qubits, n the number of excitations
and ± indicating the symmetry of the state, are illustrated in blue. The areas of the circles indicate the relative
population of the bare states in the eigenstates |N , n ±〉.

state. Accordingly we observe no visible population in the transmission spectrum at fre-

quency νr at degeneracy. In this regime the two qubits behave like one q effective spin with
p
the predicted [López07a] coupling strength G 2 = 2 g AC with g AC = 1/2(g A2 + g C2 ), which
is indicated by dashed black lines in Fig. 9.3 d. This prediction is in very good agreement
with our measurement.
Following the same procedure, we then flux tune qubit B through the already reso-
nantly coupled states of qubits A, C and the cavity (νA = νC = νr ), (Fig. 9.3 e). We observe
the energies of three out of N + n = 4 eigenstates, the fourth one being dark, for a range
of flux values ΦB . Starting with the dark state |2, 1d 〉 at frequency νr and the doublet
|2, 1±〉 (left part of Fig. 9.3 e), the presence of qubit B dresses these states and shifts the
doublet |2, 1±〉 down in frequency. Again one of these states turns dark as it approaches
degeneracy where it is entirelyp
mixed with qubit
p
B. At degeneracy we identify two bright
doublet states |3, 1±〉 = 1/ 2 |g , g , g 〉⊗|1〉±1/ 6 (|e, g , g 〉−|g , e, g 〉+|g , g , e〉)⊗|0〉 (Fig. 9.4
c). The part of the states |3, 1±〉 carrying the atomic excitation is a so called W -state, in
which a single excitation is equally shared among all N qubits [Dür00]. Both |3, 1±〉 states
are clearly visible in the transmission spectrum shown in Fig. p
9.3 f.
In addition,
p
there are two dark states |3, 1d 1 〉 = 1/ 2(|e, g , g 〉 − |g , g , e〉) ⊗ |0〉 and
|3, 1d 2 〉 = 1/ 2(|g , e, g 〉 + |g , g , e〉) ⊗ |0〉 which do not lead to resonances in the trans-
mission spectrum at degeneracy. In general all N + n − 2 dark states are degenerate at
energy ħωr . The symmetries of the dressed three-qubit states are determined by the
signs of the coupling constants g A ≈ −g B ≈ g C . While our measurement is not sensitive
to the sign of coupling, it is a simple consequence of the phase shift of the electric field
mode by π between the ends and the center of the resonator. Again, the observed trans-
missionp
peak frequencies are in agreement with the calculated splitting of the doublet
G 3 = 3 g ABC (dashed black lines in Fig. 9.3 f). Also at finite detunings the measured
energies of all bright states are in excellent agreement with the predictions based on
9.3. SUMMARY 113

Collective coupling, GN MHz

150

100

50

0
0 1 2 3
Qubits, N

Figure 9.5: Scaling of the collective dipole coupling strength. Measured coupling constants (blue dots)
extracted from Fig. 9.3 and nine similar data sets and theoretical scaling (red line).

the Tavis-Cummings model (dashed white lines in Fig. 9.3 a, c, e) using the measured
qubit and resonator parameters. We have also performed analogous measurements of
all twelve one, two and three qubit anti-crossings (nine are not shown) and find equally
good agreement.
In Fig. 9.5 all twelve measured coupling strengths (blue dots) for one, two and three
qubits at degeneracy are plottedp vs. N . Excellent agreement with the expected collec-
tive interaction strength G N = N g ABC (red line) is found without any fit parameters and
g ABC = 84.8 MHz.

9.3 Summary
Our spectroscopic measurements clearly demonstrate the collective interaction of a dis-
crete number of quantum two-state systems mediated by an individual photon. All results
are in good agreement with the predictions of the basic Tavis-Cummings model in the
absence of any number, position or coupling fluctuations. The presented approach may
enable novel investigations of super- and sub-radiant states of artificial atoms. Flux tuning
on nanosecond timescales should furthermore allow the controlled generation of Dicke
states [Stockton04, López07b] and fast entanglement generation via collective interac-
tions [Tessier03, Retzker07], not relying on individual qubit operations. This could be
used for quantum state engineering and an implementation of Heisenberg limited spec-
troscopy [Leibfried04] in the solid state.
 HAPTER
10
C
Q UANTUM - TO -C LASSICAL T RANSITION

The quantum properties of electromagnetic, mechanical or other harmonic oscillators can

be revealed by investigating their strong coherent coupling to a single quantum two level
system in an approach known as cavity quantum electrodynamics (QED). At tempera-
tures much lower than the characteristic energy level spacing the observation of vacuum
Rabi oscillations or mode splittings with one or a few quanta asserts the quantum na-
ture of the oscillator. Here, we study how the classical response of a quantum cavity
QED system emerges from the quantum one when its thermal occupation – or effective
temperature – is raised gradually over 5 orders of magnitude. In this way we explore in
detail the continuous quantum-to-classical cross-over in the spirit of Bohr’s correspon-
dence principle. We also demonstrate how to extract effective cavity field temperatures
from both spectroscopic and time-resolved vacuum Rabi measurements.

10.1 Introduction
Cavity QED [Haroche06] enables the study of the nature of matter light interactions in
exquisite detail. It realizes an open quantum system in which the coupling to the environ-
ment is highly controllable. In a circuit realization of cavity QED [Wallraff04], we carefully
investigate the quantum-to-classical transition of a harmonic oscillator strongly coupled
to a two level system by increasing the effective oscillator temperature. From measured
vacuum Rabi splitting spectra and from time-resolved vacuum Rabi oscillations we con-
sistently extract effective cavity field temperatures between 100 mK and a few Kelvin
using a quantum master equation model as suggested in Ref. [Rau04]. The dissipative
quantum-to-classical cross-over of a field mode coupled to a qubit was also studied theo-
retically in Ref. [Everitt09]. The emergence of classical physics from quantum mechanics

115
116 CHAPTER 10. QUANTUM-TO-CLASSICAL TRANSITION

and the role of decoherence in this process is an important subject of current research
[Schlosshauer07].

10.1.1 Experimental setup

In our experiments a superconducting artificial atom is strongly coupled to a coplanar
transmission line resonator to realize a circuit QED system [Wallraff04]. The transmon
qubit [Koch07a] consists of two superconducting aluminum thin films weakly connected
by two Josephson tunnel junctions. Its ground |g 〉 and excited state |e〉 differ by the spe-
cific charge configuration of the two superconducting islands. The transition frequency
νg,e ' 8E C E J (Φ) − E C is determined spectroscopically. Here, E C /h ≈ 0.502 GHz is the
p

single electron charging energy, E J (Φ) = E J,max | cos (πΦ/Φ0 )| the flux controlled Joseph-
son energy with E J,max /h ≈ 14.4 GHz and Φ0 is the superconducting flux quantum. The
cavity has a bare resonance frequency of νr ≈ 6.44 GHz and a coupling limited photon
decay rate of κ/(2π) ≈ 3.2 MHz. Optical microscope images of the sample are shown in
Fig. 10.1 a.
In our experimental setup, see Fig. 10.1 b, the coupled qubit/cavity system is prepared
in its ground state |g , 0〉, see Fig. 10.2 a, with close to n = 0 photons in the resonator by
cooling the sample to below 20 mK in a dilution refrigerator. Instead of increasing the
physical temperature of the sample to control the thermal occupation of the cavity, we
apply thermal radiation only at its input. We approximate the one-dimensional Planck
spectrum of the thermal field as constant with a power spectral density S n in the small
relevant bandwidth around νr . S n is controlled by applying broadband microwave fre-
quency white noise of controlled amplitude. In this case, we can assign an intra-cavity
thermal photon number n th = [exp (ħωr /k B Tc ) − 1]−1 and thus an equivalent cavity field
temperature Tc to the externally applied white noise. Owing to the high internal quality
factor of the resonator, the field does not thermalize on the chip, which allows us to con-
trol the effective temperature of the resonator field to up to Tc ∼ 100 K. In addition our

a 300 m b
RT Cin Resonator Cout RT
probe

spec D Qubit Cg

r B

Swhite I dc 20 mK
10 m

Figure 10.1: Experimental setup. a, Coplanar microwave resonator with two qubit slots and flux bias lines (top)
and a single embedded qubit (violet) shown on an enlarged scale (bottom). b, Circuit diagram of setup. Input
(left) is at room temperature (RT) with microwave sources for measurement (νprobe ), qubit spectroscopy and
qubit drive (νspec ), variable attenuation (D ) quasi thermal field source (S white ) at νr and an arbitrary waveform
generator for fast flux biasing ( I dc ) through the on-chip flux line (B ). At 20 mK the qubit is coupled via a
capacitance C g to the transmission line resonator between the capacitances C in and C out . The transmitted
microwave tone is amplified, down-converted with a local oscillator (LO) and digitized (ADC).
10.2. STRONG QUASI-THERMAL EXCITATION 117

setup allows for the phase sensitive detection of the quadrature amplitudes of a weak
coherent probe tone that populates the resonator with n probe . 0.1 photons on average
while effectively rejecting all uncorrelated thermal radiation from the detection system.

10.1.2 Vacuum Rabi splitting

At the lowest measured cavity field temperature Tc ∼ 100 mK, a clear vacuum Rabi
mode splitting is observed in a linear response cavity transmission measurement, see
Fig. 10.2 c (dark blue lines). Two Lorentzian lines characteristic for the dressed states
¡ ¢ p
|n, ±〉 = |g , n〉 ± |e, n − 1〉 / 2 with excitation number n = 1 separated by twice the dipole
coupling strength g g,e /(2π) = 54 MHz are clearly observed at the frequencies νg0,1± . This
indicates that the strong coupling regime, g g,e À κ, γ, with the qubit relaxation rate
γ/(2π) ≈ 0.6 MHz, is realized.
While the observation of the simple vacuum Rabi mode splitting for n = 1 could in
principle be interpreted as a normal mode splitting using a semiclassical model, the ob-
servation of additional transitions to higher excited dressed states with coupling increased
p
by a factor n [Schuster08, Fink08, Bishop09, Fink09a] is an unambiguous signature for
the quantum nature of the system [Agarwal86, Agarwal91, Tian92], see also Chapter 8.
In our experiment these transitions are observed at frequencies ν1±,2± , see the level dia-
gram in Fig. 10.2 a and measurements in Fig. 10.2 c (green lines) [and Fig. 10.3 a (blue
and green lines)], when slightly increasing the effective cavity temperature and thus pop-
ulating excited states of the Jaynes-Cummings ladder. In this case the weak coherent
probe tone excites transitions between higher levels.

10.2 Strong quasi-thermal excitation

At high temperatures the Jaynes-Cummings system is excited to high quantum numbers
and many transitions can be accessed by the weak probe tone [Agarwal86, Agarwal91,
Cirac91, Tian92, Rau04]. The resulting linear response transmission spectrum can be
understood as a sum of individual level to level transitions that overlap. At a mean ther-
mal photon number of n th > 1 the transitions from the ground state νg0,1± are almost satu-
rated and therefore only weakly contribute to the observed cavity transmission spectrum
(Fig. 10.2 c, yellow lines). At even larger effective temperatures all available transition
frequencies are densely spaced close to the resonator frequency, see level diagram in
Fig. 10.2 b and measurement in Fig. 10.2 c (red lines). At these temperatures all transi-
p
tions which would show the nonlinear n signature of the single qubit–photon interaction
are saturated and the transmission spectrum does not carry any information about the
intrinsic quantum nature of the system.

10.2.1 The high temperature limit

The resulting harmonic spectrum resembles that of two uncoupled linear oscillators, a
situation where not the absolute energies but the energy differences of the system are
degenerate [vandenBrink02, Rau04]. In contrast to the regime of strong coherent driving,
where non-classicality can persist at large excitation numbers [Alsing91, Armen09], the
118 CHAPTER 10. QUANTUM-TO-CLASSICAL TRANSITION

a b c Classical
0.7
response 
g0,1- g0,1+
2,+ n,+ 0.6 nth≈ 400

Transmission A2 A2max
2 gge/ n gge / Quantum
0.5
response
2,- n,-
1+,2+

r
0.4 nth≈ 0.05
1-,2- 1+,2+
1-,2-

n-1,+
r

1,+ 0.3
gge/  n gge /
1,- 0.2
n-1,-
g0,1+

0.1
g0,1-

quasi harmonic
g,0 0.0
6.35 6.40 6.45 6.50 6.55
anharmonic Νprobe (GHz)

Figure 10.2: Quantum-to-classical transition of the vacuum Rabi splitting spectrum. a, Energy level
diagram of dipole coupled dressed states |n, ±〉 for excitation numbers n = 0, 1 and 2 (black lines), uncoupled
qubit and cavity energies (grey dashed lines) and allowed transitions at frequencies νg0,1± and ν1±,2± (blue
arrows). b, Dressed state diagram for large excitation numbers n > 280 and allowed transitions at the resonator
frequency νr (red arrows). c, Measured cavity transmission A 2 /A 2max for intra-cavity thermal photon numbers
0.05 (blue) . n th . 400 (red) and a fit to a Lorentzian line (black). Measured data sets are normalized to the
transmission amplitude A max obtained when the qubit is maximally detuned. The data is offset and colored for
better visibility, see inset in Fig. 10.4 for color code.

phase noise of the thermal field used in our experiments renders the two high-excitation
dressed-state ladders indistinguishable. As a result, at high temperatures the spectrum
is that of a classical resonator and we find excellent agreement with a Lorentzian line fit
of width κeff /(2π) = 4 MHz close to the intrinsic resonator line width (Fig. 10.2 c, black
line).

10.2.2 Quantitative model

A quantitative understanding of the measured results is obtained by numerically solving
the Markovian master equation
" #
i ˆ X g l −1,l
L [ρ̂] = − [H , ρ̂] + (n th + 1)κD[â]ρ̂ + γ D σ̂l −1,l ρ̂ + n th κD[â † ]ρ̂ , (10.1)
ħ l =e,f,... g g ,e

where the thermal photon number n th or equivalently the cavity field temperature Tc is
extracted as the only fit parameter. Here the coherent dynamics is described by the
Jaynes-Cummings Hamiltonian Hˆ with l transmon levels, see [Koch07a], and without a
drive term, which is justified in the linear response limit. The three damping terms model
the loss of cavity photons at rate κ, the intrinsic relaxation of the transmon excited state
|e〉 at rate γ and the creation of cavity photons due to a thermal bath with occupation
n th . D is the usual Lindblad damping superoperator. Note that neither qubit dephasing
nor an independent qubit thermal bath has significant influence on the agreement be-
tween the theory and the presented spectroscopic results. Therefore, these two terms
 10.2. STRONG QUASI-THERMAL EXCITATION 119

have been omitted. Equation (10.1) is solved by exact diagonalization and the transmis-
sion spectrum is calculated as A/A max = (κ + γ)/2 〈a † a〉/n th , using similar methods as in
Ref. [Rau04].
We observe that the measured and calculated spectra agree very well over a large
range of applied thermal noise powers (Fig. 10.3). This allows to extract the effective
cavity field temperature (Fig. 10.4). However, for mean thermal photon numbers n th & 30
the accuracy of our numerical calculations is limited by the finite size of the Hilbert space
spanned by 200 resonator states and 6 transmon states. In the large photon number
limit the qubit only negligibly perturbs the empty cavity Lorentzian spectrum [Fig. 10.2 c],
because the cavity and cavity dissipation terms in Eq. (10.1) scale linearly with n , while
p
the qubit photon coupling term scales only with n , see also [Savage89]. In the limit
p
where the coupling is smaller than the damping ng g,e < nκ the dominant terms in
Eq. Eq. (10.1) describe a damped harmonic oscillator with a Lorentzian spectrum. A
classical harmonic oscillator is furthermore a good description of our cavity QED system
p
if the relative n nonlinearity, i.e. the frequency shift due to the qubit, is smaller than the
p p
relevant dissipation rate 2 g g,e ( n + 1 − n) < κ. For large photon numbers these two cri-
teria are equivalent and in our experiment, the classical limit is entered for mean thermal
photon numbers n th > (g g,e /κ)2 ≈ 280.

a b 0.5
2.5

0.4 722 mK
Qubit population Pe

nth= 24 7440 mK
0.3 383 mK
2.0
nth= 8.4 2750 mK
0.2 245 mK

179 mK
nth= 2.7 989 mK 0.1
Transmission A/Amax

1.5
142 mK
nth= 1.1 0.0
476 mK 0 20 40 60 80
c 1.0
νspec π
1.0 nth= 0.42 255 mK
0.8 I dc τ
Qubit population Pe

νr Measurement
nth= 0.19 170 mK 0.6 Noise
0.5
nth= 0.09 126 mK 0.4
142 mK
0.2
nth= 0.05 104 mK
0.0
0.0
6.35 6.40 6.45 6.50 6.55 0 20 40 60 80
νprobe (GHz) Pulse length, τ (ns)

Figure 10.3: Quantitative analysis of elevated temperature Rabi mode splittings and vacuum Rabi os-
cillations. a, Measured (dots) and calculated (lines) cavity transmission (same data as in Fig. 2 c) shown for
applied S n = −221.5 dBm/Hz (bottom) to −190 dBm/Hz (top) in steps of 4.5 dBm/Hz. Extracted thermal photon
numbers n th and cavity temperatures Tc are indicated. b, Measured qubit excited state population P e as a func-
tion of the resonant cavity interaction time τ (dots) and master equation simulation (lines) for S n = −214 dBm/Hz
to −202 dBm/Hz in steps of 3 dBm/Hz. c, Similar measurement as in (b) with qubit prepared in the excited state
for S n = −214 dBm/Hz. Inset: Vacuum Rabi pulse sequence.
120 CHAPTER 10. QUANTUM-TO-CLASSICAL TRANSITION

10.2.3 Time domain measurements

We have also performed time domain vacuum Rabi oscillation measurements in the
presence of thermal photons. In these experiments, we first detune the qubit from the
resonator by 0.5 GHz where the applied quasi thermal noise is both filtered and largely
reflected from the resonator and does not affect the qubit state. We then apply an ampli-
tude shaped current pulse I dc via an on-chip flux line, see sample and setup in Fig. 10.1,
to tune the qubit into resonance with the cavity field for a variable time τ, see pulse se-
quence in inset of Fig. 10.3 c. The qubit excited state population P e is then determined
in a dispersive measurement [Bianchetti09].
For the lowest cavity temperatures we observe low contrast, long coherence vacuum
Rabi oscillations induced by the weak thermal field only, see Fig. 10.3 b. When the num-
ber of thermal photons is increased the amplitude of the coherent oscillations increases
while their coherence time decreases, which is expected for a thermal distribution of
photon numbers. At long interaction times τ, the qubit population shows a temperature
dependent saturation that approaches the value of 0.5 when the qubit is strongly driven
by the thermal field at high temperatures. We have repeated the experiment preparing
the qubit initially in the excited state using a π-pulse and measuring vacuum Rabi oscil-
lations at low applied noise power, see Fig. 10.3 c. In this case high contrast oscillations
are observed albeit with a small offset in the qubit population visible at large τ caused by
the thermal field. From these data the cavity temperature has been extracted by solving
a time dependent master equation, see solid lines in Figs. 10.3 b and c. In this case 2
qubit states and up to 6 cavity states are considered and qubit dephasing is included.
The deviations at short times τ are due to imperfections in the qubit flux tuning pulses.

10.2.4 Extraction of the effective field temperature

The cavity field temperatures Tc extracted from both spectroscopic (full dots) and time-
resolved measurements (crosses) plotted on a logarithmic scale versus the applied noise
spectral density S n are in close agreement (Fig. 10.4). Both data sets are explained
consistently considering that the cavity thermal photon number n th = S n /ħω + n 0 is the
sum of the applied quasi thermal noise S n /ħω expressed as a photon number and a
small thermal background field of n 0 = 0.04 photons1 (T0 = 95 mK). The measured intra-
cavity photon number n th is consistent with the noise applied to the input of our setup
and the expected attenuation of the line and also with calibration measurements based
on the known qubit ac-Stark shift [Schuster05]. The background photon number is due
to incomplete thermalization of the input line and can easily be decreased by adding
additional cold attenuation.
The inset of Fig. 10.4 shows an extrapolation of the temperature to the largest ap-
plied noise powers and provides a calibration for the measurements shown in Fig. 10.2 c
where the same color code is used (also used in Figs. 10.3 and 10.4). Accordingly, the
1 In more recent experiments an additional attenuation of -10 dB was added on the resonator input line at
20 mK, see setup in Fig. 4.1. This reduced the background photon number by almost one order of magnitude.
In fact, in the measured transmission spectrum we cannot resolve any two photon dressed states. Using the
model Eq. (10.1) we therefore can only state an upper bound of n 0 . 0.005.
10.3. CONCLUSION 121

Attenuation, D (dB)
45 40 35 30 25 20 15 10
40
10.0 D (dB)
40 30 20 10 0
400 20
5.0 100.
100 10

Thermal photons, nth

Temperature, Tc (K)
10.
2.0

nth
10
Tc (K)
5
1.
1.0 1
2
0.1 0.05
0.5 - 220 - 200 - 180 1
Sn (dBm/Hz) 0.5
0.2
0.2
0.1
0.1 0.05
- 220 - 215 - 210 - 205 - 200 - 195 - 190
Noise power spectral density, Sn (dBm/Hz)

Figure 10.4: Temperature sensing measurement. Resonator field temperature Tc and photon number n th
extracted from measured Rabi spectra (full dots) and Rabi oscillations (crosses) versus noise power spec-
tral density S n or equivalently attenuation D . Theory (black line). Inset: extrapolation to higher Tc provides
temperature scale (color coded) for the measurements shown in Figs. 10.2 c and 10.3.

measurements at the highest temperatures were conducted with approximately ∼ 370

thermal photons in the cavity. This corresponds to an effective cavity temperature of
Tc ∼ 115 K. We clearly observe the transition to a classical harmonic resonator at
n th ≈ 370 > (g g,e /κ)2 ≈ 280 as indicated by the Lorentzian spectrum in Fig.10.2 c.

10.3 Conclusion
We have demonstrated a quantitative understanding of the transition from the quantum
to the classical response of a cavity QED system. Moreover, we have presented a viable
approach to measure the temperature of an electromagnetic field in a cavity over a wide
range. A different approach to extract the resonator temperature in the dispersive regime
of circuit QED has been proposed [Clerk07] and measurements of the qubit temperature
have been demonstrated [Palacios-Laloy09]. Related experiments have been performed
in the optical domain using a semiconductor cavity QED system [Laucht09]. In these
systems, entanglement and decoherence at elevated temperatures can be studied in
future experiments in the context of quantum information.
 HAPTER
11
C
C ONCLUSION AND P ROSPECTS

We have successfully installed and characterized the experimental infrastructure neces-

sary to perform circuit QED experiments in the Quantum Device Laboratory (QuDev) at
ETH Zurich. Various qubit samples have been fabricated and analyzed in spectroscopic
and time-resolved resonator transmission measurements.
The main focus of the presented work was the study of the resonant interaction of
individual superconducting qubits and single photons. We showed the first direct spectro-
scopic measurement of both matter-light superposition states of a single (artificial) atom
and two photons [Fink08]. We also investigated the nonlinear spectrum of three photon
superposition states with both coherent and thermal fields [Fink09a]. Furthermore, we
showed the nonlinear scaling of the collective dipole coupling strength with the number
of artificial atoms without ensemble averaging for up to three qubits [Fink09b]. Finally,
we demonstrated a quantitative understanding of the quantum-to-classical transition in
an open quantum system by controlling the system temperature. In addition, a detailed
simulation of the resonator transmission spectrum allowed the sensing of thermal fields
over more than 2 orders of magnitude in temperature [Fink10].
During the completion of this thesis contributions have also been made to other exper-
iments such as the observation of Berry’s phase [Leek07], the Lamb shift [Fragner08] and
Autler-Townes and Mollow transitions [Baur09]. In addition, fabrication of qubits for ex-
periments aimed towards the goal of quantum computing [Leek09, Leek10, Bianchetti09,
Bianchetti10b] as well as fabrication of SQUIDs for frequency tunable quantum limited
parametric amplification and squeezing was part of this thesis. More recently, qubit de-
vices used for single photon correlation measurements [Bozyigit10c, Lang10], the ob-
servation of microwave photon antibunching [Bozyigit10b], as well as for studying the
potential of time-resolved resonant three qubit quantum state synthesis, see for example
Ref. [Haack10], have been fabricated.
123
124 CHAPTER 11. CONCLUSION AND PROSPECTS

The rapid conceptual, experimental and technological progress in circuit QED opens
up a wealth of new research directions. This is also reflected by the large number of
circuit QED related proposals that are being published.
Increasing the number of qubits in our system promises to extend the domain of
ultra strong coupling [Bourassa09, Niemczyk10, Forn-Díaz10, Nataf10b] to the domain
where the Dicke superradiant quantum phase transition is predicted to occur [Chen07,
Lambert09, Nataf10a]. Using a chain of superconducting qubits coupled to light also
promises new insights in related quantum many body physics models [Larson10, Tian10,
Tomadin10]. Utilizing the recently implemented fast local qubit flux bias control enables
the study of the dynamics of few qubit collective spin states [Keeling09]. These collective
modes could potentially be used to realize quantum computing with a small qubit spin
ensemble, similar to Ref. [Wesenberg09]. More realistic is their application for resonant
quantum gates and efficient quantum state generation [Haack10], which would be useful
also for one way quantum computing. Other cooperative phenomena that should be
readily observable in our circuit QED system are superradiance [Scully09] and multiphase
stability [Delanty10] as well as the collective Lamb shift [Scully10].
A natural extension of current circuit QED experiments would be to study ar-
rays of resonators coupled to superconducting qubits. Such setups have been sug-
gested as potential candidates to realize Bose and Jaynes-Cummings Hubbard models
[Schmidt09, Silver10, Leib10] and probe the superfluid to Mott insulator transition of light
[Koch09]. It has also been pointed out that time-reversal symmetry breaking [Koch10] and
localization and self-trapping of photons in circuit QED photon lattices [Schmidt10] could
be investigated. Cavity grids would complement the research with cold atomic gases.
Circuit QED with coupled cavity arrays has also been suggested for scalable quantum
computing and quantum communication [Giampaolo09, Lin09, Helmer09].
Another interesting direction is in part already pursued in our laboratory. Making
use of the tuning capabilities and SQUIDs and qubits enables to in-situ tune the reso-
nance frequency of microwave resonators [Sandberg08a, Sandberg08b]. Rapidly tunable
resonators and qubits have been suggested [Dodonov09, De Liberato09, Johansson09,
Johansson10, Dodonov10] to be used for the parametric generation of photons and the
observation [Wilson10] of vacuum radiation in analogy to the dynamical Casimir effect.
The nonlinearities provided by SQUIDs and qubits can also be used to develop para-
metric amplifiers for quantum limited amplification and squeezing [Yurke87, Yurke89,
Castellanos-Beltran07, Yamamoto08, Castellanos-Beltran08]. A quantum limited ampli-
fier would be extremely useful not only for efficient single shot qubit readout and field
state tomography of itinerant microwave photons [Eichler10a] with a high signal to noise
ratio, it also allows to engineer one and multi-mode squeezed field states [Eichler10b].
For the future even continuous variable quantum computing may be envisaged.
A better signal to noise ratio will complement our recently implemented improve-
ments in the data acquisition hard- and software, used for single photon source anti-
bunching correlation measurements [Bozyigit10b]. Using two single photon sources it
should be possible to observe microwave photon-photon interference in a Hong-Ou-
Mandel type experiment. We have already found evidence for related quantum optics
 125

effects such as the predicted dressing of dressed states [Shamailov10, Lang10] and the
spontaneous dressed state polarization in the strong driving limit of resonant cavity QED
[Alsing91, Armen09, Delanty10]. Once the parametric amplifier is well integrated into our
data acquisition circuit one can also start to think about real-time control and feedback
of solid state quantum systems. Advanced implementations of teleportation and optimal
control techniques will benefit from these capabilities.
While basic quantum computing algorithms involving up to two qubits have already
been demonstrated in circuit QED [DiCarlo09], a demonstration of scalability and an im-
plementation of quantum error correction codes is still lacking. An in-depth analysis of
decoherence mechanisms, novel designs [Manucharyan09, Steffen10], fabrication tech-
niques, as well as clever strategies to efficiently control multi-qubit systems will be essen-
tial. Hopefully more applicable, capable and useful quantum computing algorithms will be
developed in the near future. It appears that the quantum simulation of problems which
are hard to solve by classical means, such as quantum many body physics problems,
may be a fruitful pathway to approach quantum information processing.
Another promising development is the new trend to investigate hybrid quantum sys-
tems [Rabl06, Rabl07, Wesenberg09, Imamoglu09, Marcos10, Twamley10]. These sys-
tems would have the potential to combine the fast gate operation times and the large
coupling strengths attainable in circuit QED, with the long coherence times demonstrated
for example in trapped ion and neutral atom as well as nuclear spin and color center im-
plementations of quantum computing. In addition, mechanical transducers or plasmonic
devices could potentially be used to form a link between the microwave and the opti-
cal domain such that long distance quantum communication between superconducting
circuits or other solid states qubits may be envisaged.
The circuit QED architecture clearly provides an excellent playground for curiosity
driven experimental research. There are numerous technological as well as conceptual
challenges waiting to be solved and there are a lot of interesting quantum optics, quantum
information processing and solid state physics phenomena to be unveiled on this path.
Quantum engineering with superconducting circuits will continue to test and push the
boundaries and the size of truly quantum mechanical systems further and further.
 PA R T IV

A PPENDICES

127
 APPENDIX
A
A SPECTS OF THE C RYOGENIC S ETUP

In this chapter we summarize a number of important aspects of the infrastructure required

to successfully operate a dilution refrigerator at its lowest temperatures for high sensitivity
microwave frequency measurements.

A.1 Dimensioning He vacuum pumping lines

We decided to install the vacuum pumps and compressors required to operate the dilution
refrigerators Kelvinox 400 HA from Oxford Instruments and the prototypical cryogen free
system from VeriCold not directly in the laboratory, due to the mechanical vibrations,
noise and dirt which is potentially generated. This requires longer pumping lines which
can reduce the effective pumping speed. Since the cooling power and to some extent also
the base temperature are connected to the effective 3 He and 4 He pumping speeds on the
still and 1K-pot pumping lines respectively, the question arose which tubing diameters
are necessary to compensate the extra lengths of tubing such that the nominal pumping
speeds are not significantly reduced.
The most important parameter for such a dimensioning is the actual pressure of the
gas directly before the roots (BOC EH 500), rotary vane (RV12) or turbo pump (TPH 1201
P). This pressure can vary over a large range of typically 0.01 mbar < p < 10 mbar on the
still and 1 mbar < p < 10 mbar on the 1K-pot, depending on the system and operation
mode [OxfordInstr06]. From our calculations we will see that at the higher end of this
pressure range practical tube dimensions are not limiting the nominal pumping speeds.1
1 In case the turbo pump is used, the pump itself will be the limiting factor at larger pressures since its
nominal pumping speed at 0.001 mbar is already decreased by about two orders of magnitude at a pressure of
0.1 mbar. This pressure dependence is not as dramatic in case of the roots pump and the rotary vane pump is
actually designed for optimal operation at pressures above one millibar.

129
130 APPENDIX A. ASPECTS OF THE CRYOGENIC SETUP

The interesting situation in terms of tube dimensioning is therefore the still line operated
at a low pressure which we take to be p s = 5 Pa. The typical flow or throughput Q in the
Oxford system is on the order of 125 sccm (standard cubic centimeter per minute) if the
dilution refrigerator is operated at base temperature and about 800 sccm if the cooling
power test is performed at 100 mK [OxfordInstr06]. At pressure p s this corresponds to a
required pumping speed S = Q/p s of at least 0.04 m3 /s and 0.27 m3 /s respectively.
Starting with the nominal pumping speed of the Oxford Instruments cryostat roots still
pump, i.e. S n ∼ 0.1 m3 /s (Vericold turbo pump S n ∼ 0.2 m3 /s) at p s = 5 Pa, we calculate the
reduced pumping speed at the cryostat S cryo due to the pumping line consisting of two
back to back pipes of lengths l 1 = 1 m, l 2 = 6.5 m and radii r 1 = 0.025 m, r 2 = 0.05 m as

1
S cryo = 1
(A.1)
Sn + C1tot

where
1
C tot = 1
(A.2)
C1 + C12

is the total conductance value of the two tubes with the individual conductance values
C 1,2 operated in series. The associated total pressure drop between the cryostat and the
the pump is
Q
∆p = (A.3)
C tot
with the total gas throughput Q . Additional sections of tubes and other constrictions
like bends or baffles installed in series need to be added according to Eq. (A.2). In
order to determine the two tube conductances C 1,2 we first need to determine if laminar
(R e < 1100) or turbulent flow (R e > 2100) occurs. The Reynolds number R e is defined as
the ratio of inertial forces and viscous forces which can be calculated as
2r 1
R e = ρv ≈ 1.4, (A.4)
η

with the density

M ps
ρ= ≈ 6 · 10−6 kg/m3 (A.5)
R0 T
the average speed
Sn
v= ≈ 12 m/s (A.6)
r 12 π
and the dynamic viscosity of the gas [Richardson88]
s
2 M R0 T kg
η= ≈ 1.1 · 10−5 . (A.7)
3 N A2 π3 (2r 1 )4 ms

Here, M = 0.003 kg/mol is the atomic mass of 3 He, R 0 = 8.31 J/(K mol) the gas constant,
T ≈ 300 K the temperature and N A ≈ 6.022 · 1023 1/mol the Avogadro constant. Clearly we
A.1. DIMENSIONING HE VACUUM PUMPING LINES 131

are in the limit of laminar flow where R e ¿ 1100 even by taking into account the smaller
tube diameter 2r 1 as the characteristic length scale.
In the laminar flow regime the conductance value can be dominated by internal friction
caused by particle particle collisions. This is the regime of continuous or viscous flow
where the conductance is proportional to the mean gas pressure (Kn > 110). The other
extreme is the regime of molecular flow (Kn < 1) where the conduction is dominated by
friction caused by particles colliding with the walls and constrictions of the tubes which
is independent of the gas pressure. In our case we realize the intermediate or Knudsen
flow (1 < Kn < 110) characterized by the Knudsen number given as

2r 1
Kn = ≈ 49, (A.8)
L
where 2r 1 is the characteristic length scale and

T
L = R0 p ≈ 3.85 · 10−3 m, (A.9)
2πd 2 N A p s

the mean free path2 of a helium particle with a hard sphere diameter [Kennard38] taken
to be d ∼ 2.2 · 10−10 m. We therefore need to take into account the contributions of both,
continuous flow and molecular flow.
In the regime of laminar continuous flow the conduction of a long smooth pipe can be
calculated according to the law of Hagen-Poiseuille
4
πr 1,2
C cont 1,2 = ps. (A.10)
8ηl 1,2

For the molecular flow on the other hand the conduction is given as [Roth76]
3
r 1,2 π πR 0 T
r
C mol 1,2 = , (A.11)
l 1,2 2M

see also [PfeifferVacuum10]. In the relevant intermediate regime it is given as [Roth76]

3 p
2πR 0 T 4r 1,2 1 + M /(R 0 T ) 2r 1,2 p s /η
r
C int 1,2 = C cont 1,2 + p . (A.12)
M 3l 1,2 1 + 1.24 M /(R 0 T ) 2r 1,2 p s /η

Note furthermore that there exists an upper bound for the conduction value

C max ≤ c s r 12 π ≈ 9 m3 /s, (A.13)

which is limited by the speed of sound of the fluid c s = γR 0 T /M ≈ 1.2 · 103 m/s with the
p

adiabatic index γ = c p /c v = ( f +2)/ f ≈ 1.67 for f = 3 degrees of freedom of the monatomic

3
He gas.
2 Using the mean free path and the mean particle speed of v = 8R 0 T /(M π) one can also obtain the
p
p
dynamic viscosity as η = M pv p L/(3R 0 T ).
132 APPENDIX A. ASPECTS OF THE CRYOGENIC SETUP

For the above mentioned configuration of two back to back tubes we obtain the
conductances C int 1 ≈ 0.13 m3 /s, C int 2 ≈ 0.28 m3 /s and the total conductance of C tot ≈
0.09 m3 /s. This corresponds to a pumping speed at the cryostat of S cryo ≈ 0.05 m3 /s
which relates to a pressure drop of ∆p = S cryo p s /C tot ≈ 2.8 Pa and a reduction of the nom-
inal pumping speed S n ≈ 0.1 m3 /s by about 50% due to the tubing. If our configuration is
compared to the original layout of pumping lines the loss in pumping speed is however not
as dramatic. In addition, the throughput expected when operated at base temperature,
i.e. 0.04 m3 /s at 5 Pa can still be accomplished. After installation the specified cooling
power could be demonstrated and a lowest base temperature of < 6 mK was achieved
and analyzed using a nuclear orientation thermometer [Pobell06].

A.2 Mechanical vibration measurements

Before setting up the cryostat we analyzed the mechanical vibration spectrum on 1) the
laboratory floor which has a base separated from the rest of the building and 2) the
cryostat frame which is mounted on the laboratory ceiling. We make use of 3 geophone
sensors GS-20DX-2C from Geo Space Corporation with a natural resonance frequency
of f n = 10 Hz and an internal coil resistance of r = 630 Ω. The geophones detect vibrations
longitudinal to their cases. They are firmly mounted in a solid aluminum holder in order
to sense the vibrations in the vertical and both horizontal directions. This holder is then
attached firmly to the object of interest.
The operation principle of a geophone sensor is based on the relative movement
between a coil and a magnetic field. Typically the magnetic field is generated by a per-
manent magnet held by a spring and the coil is fixed with respect to the housing of the
sensor. The relative movement of the coil in the magnetic field due to external vibra-
tions then induce a voltage at its leads. The excitation frequency f dependent voltage
response E ( f ) in V/(m/s) of the used sensor can be calculated as
³ ´2
R f
G r +R fn
E ( f ) = sµ , (A.14)
³ ´2 ¶2 ³ ´2
f 2 f
1 − fn + 4b t f n

with the intrinsic voltage sensitivity G = 34.6 V/(m/s). This is the geophone sensitivity with-
out any damping shunt resistance across the two geophone terminals R . The geophone
total damping expressed as a fraction of the critical damping is suggested to be in the
range 0.6 < b t < 0.7. In this regime the intrinsic sensitivity is only slightly lowered while the
overall voltage response is close to constant in the excitation frequency, see Fig. A.1 a.
The necessary shunt resistance to achieve this can be calculated as R = (D/(b t −b to )−r )
with D = 868 the damping constant and b to = 0.3 the open circuit damping. However, for
the presented measurements we did not use a shunt resistor and worked with the open
circuit damping of b t = 0.3 where R → ∞.
The voltage signal is directly detected and Fourier transformed using the spectrum an-
alyzer SR785 from Stanford Research Systems. The peak acceleration plotted in Fig. A.1
 A.2. MECHANICAL VIBRATION MEASUREMENTS 133

b-d is calculated from the measured peak voltage Vpk using A pk ( f ) = 2π f Vpk ( f )/(E ( f )g )
with g = 9.81 m/s2 . For the measurements we used a bandwidth of 0.25 Hz, 800 FFT
lines, BMH windowing and 10 rms averages. The spectra between 10 Hz and 100 Hz are
found to be peaked at a value below 10 µg Hz, 900 µg Hz and 1000 µg Hz for the vertical
direction, 10 µg Hz, 1000 µg Hz and 20000 µg Hz for vibrations in the horizontal longitu-
dinal direction of the cryostat support frame and 10 µg Hz, 2000 µg Hz and 50000 µg Hz
for vibrations in the horizontal transverse direction. Here, the first number is measured
on the laboratory floor (red curves in Fig. A.1), the second on the support frame (blue
curves) and the third is obtained from the measurements on the frame after it was hit by
hand (green curves). The peaked spectrum at 150 Hz in Fig. A.1 b is likely to be due to
an electrical resonance.

70
a
60
50
E Vms

40
30
20
10
0
1000 b

100
A Μgpk

10

104 c
1000
A Μgpk

100
10
1

d
104
1000
A Μgpk

100
10
1

f Hz
0 50 100 150

Figure A.1: Vibration measurements. a, Calculated voltage response spectrum for different dampings b t =
0.7, 0.6, 0.3 (yellow, red, blue). The blue spectrum is used to convert the measured voltage spectrum to a
velocity. b, Vibration spectrum measured in vertical direction on the lab floor (red), the support frame (blue) and
the support frame excited with a manual hit (green). c, Vibrations measured in horizontal longitudinal direction
of the rectangular support frame. d, Vibrations measured in horizontal transverse direction of the support frame.
134 APPENDIX A. ASPECTS OF THE CRYOGENIC SETUP

In summary we can state that the laboratory base is very well isolated in terms of me-
chanical vibrations. With the largest accelerations below 10 µg , the vibration spectrum is
comparable to measurements conducted with an accelerometer in Donald Eigler’s lab at
IBM. We furthermore observe that the situation at the ceiling is substantially worse, typi-
cally by one to two orders of magnitude. In case the support frame is excited we observe
an additional set of low frequency resonances. In the subsequent frame designs a more
stiff layout has been implemented to reduce these resonances. At higher frequencies
there is however no substantial deviation between these two cases. For practical rea-
sons the cryostats were however all mounted on the ceiling with the option to supporting
them from the floor if required.

A.3 Double gimbal vibration isolation

In order to minimize mechanical vibrations transmitted from the vacuum pumps to the
cryostat through the pumping lines, we implemented one meter of each pumping line
using flexible tubes. In the middle of this section the flexible tubes were firmly attached
to a solid wall. In order to further reduce the amount of vibrations on the still line which
is quite stiff due to its large diameter, we in addition designed a double gimbal vibration
isolation device, as proposed in Ref. [Kirk78].
The double gimbal design, see Fig. A.2, basically prevents the mechanical contraction
of the flexible bellows due to the vacuum and thus keeps them flexible. At the same time,

Figure A.2: Double gimbal vibration isolation. The pumping side (top) is connected to the cryostat still line
(right) using two flexible stainless steel bellows. The bellows are elongated to their nominal length using steel
ropes suspended at two gimbals. The two gimbals are fixed at two rings which are free to pivot around one axis
relative to the two inner rings. The inner rings themselves are also free to pivot about the orthogonal axis. The
connections were implemented with a set of 16 tight fitted axial deep grove ball bearings.
A.3. DOUBLE GIMBAL VIBRATION ISOLATION 135

the gimbals ability to pivot about two axes minimizes the restoring forces on the cryostat
side caused by small mechanical deflections on the pumping side. Such a decoupling
is possible because the resonance frequency of the construction < 1 Hz is much smaller
than the expected vibration frequency. The effect is therefore similar to that of a very
flexible pumping line only, which is not applicable at large pipe diameters of & 50 mm.
The implemented design has also been argued to have a superior isolation performance
compared to the frequently used crossed-bellow design which requires 4 flexible bellows
[Kirk78].
 APPENDIX
B
N ANO FABRICATION R ECIPES

Here we give a point by point list of process steps for the fabrication of a typical set of
superconducting qubit devices based on aluminum shadow evaporation.

B.1 Chip preparation

A careful inspection of the processed chip is strongly advised. All critical resonator fea-
tures including the alignment marks should be well resolved. Carefully rinse and dry all
glass ware before use. For every spinning layer use a new pipette. Both, the pipette and
the chip should be cleaned with the N2 gun to remove residual dust before the deposition
of resist. Try to avoid resist flowing across the chip edge. If it happened remove resist
sticking on the bottom of the chip before baking such that good thermal contact between
chip and hotplate is assured.

Process step Description Comments

Chip cleaning O 2 plasma ashing for 120 s, Dicing protection resist needs to have
150 W, 0.7 Torr been removed already

Acetone ultrasonic bath for Take care that moving chips do not
2 min, 50◦ C, US power 2 scratch each other

Isopropanol ultrasonic bath for Take care that moving chips do not
2 min, 50◦ C, US power 2 scratch each other

N2 blow dry chip carefully Inspect chip under opt. microscope

137
138 APPENDIX B. NANO FABRICATION RECIPES

Spinning Prebake on hotplate for 5 min at

120◦ C

Cool down for 5 min

Cover entire chip with P(MMA- Wait 5 − 10 sec before starting to spin
MAA) pure

Spin for 60 sec at 3000 rpm Ramp up time 3 sec

Bake on hotplate for 300 s at

180◦ C

Cool down for 5 min Thickness should be 630 ± 10 nm

Cover entire chip with PMMA Wait 5 − 10 sec before starting to spin
950k 2:1 in ethyl lactate

Spin for 60 s at 4000 rpm Ramp up time 4 sec

Bake on hotplate for 900 s at

180◦ C

Cool down for 5 min Inspect resist and edge bead. Thick-
ness of second resist layer should be
115 ± 10 nm

Charge layer Mount chips in evaporator Clamp chips from the side

Deposit 10 nm aluminum Use Alpure and check currents and rate

stability

B.2 Electron beam lithography

The typical SQUID Josephson junction geometry is defined in the CAD file by two bridges
of width 100 nm and length 200 nm. Typical exposure doses are ∼ 400 µC/cm2 for small
structures like the Josephson junctions, ∼ 350 µC/cm2 for larger structures that could
become instable such as the transmon finger capacitors and ∼ 100 µC/cm2 for undercut
boxes where only the copolymer layer should be developed.

Process step Description Comments

Chip loading Mount monitor and resonator Sketch their positions and check chip
chips on waver holder orientations

Clamp the chips from the top at Use clean in house made beryllium
one of the corners with caution copper or phosphor bronze clamps
B.2. ELECTRON BEAM LITHOGRAPHY 139

Carefully scratch the resist on Sharp tweezers or small screw driver

the monitor chip a few times on are suitable
lower left corner

N2 blow dry chip carefully Also check and clean the electrical con-
tact of the waver holder

Start EBL loading procedure

EBL setup Set 10 µm aperture and 30 kV Ramp up the voltage starting from 21 kV
gun voltage with 1 kV/min

Go to the Faraday cup (FC)

closest to the chips

Set previous stigmator and These are the values of the last person
aperture values who used the same aperture and volt-
age

Set the write-field to 500 µm corresponds to an magnification of

720k

Choose SE2 detector and

coarsely focus on FC

Correct all user positions All FC’s are scanned, their coordinates
are corrected and saved

Chip alignment Go to lower left corner of moni- No critical chip regions should be im-
tor chip aged before exposure

Coarse focus, origin correction, Set the middle of the monitor chip as
angle correction (U ,V ) = (0, 0)

Improve focus on a scratch or Needs optimization of focus, stigma-

dirt particle feature < 30 nm tism and aperture corrections, save val-
ues

Manual stage based write-field Start with large scan size and small
alignment deflection and iteratively improve accu-
racy

Burn a circular contamination Needs optimization of focus, stigma-

dot of diameter 20 − 30 nm tism and aperture corrections, save val-
ues, try at different positions

JJ exposure Measure beam current Correct dwell time, check beam speed
(∼ 28 pA) and base dose of 100 µC/cm2

Check position list, exposure Also check patterns, geometries and

layers, working field settings doses
and positions on waver map
140 APPENDIX B. NANO FABRICATION RECIPES

Go to origin and start exposure Exposure details: ∼ 8 nm step size (1

of the position list pixel), 20 ms waiting time, fly back factor
0.15, beam drag correction on

Pads exposure Set 120 µm aperture Load saved correction values

Measure beam current Correct dwell time, check beam speed

(∼ 4.9 nA) and base dose of 100 µC/cm2

Check position list, exposure Also check patterns, geometries and

layers, working field settings doses
and positions on waver map

Go to origin and start exposure Exposure details: ∼ 150 nm (19 pixel),

of the position list 20 ms waiting time, fly back factor 0.15,
beam drag correction on

Qubit exposure Set 10 µm aperture Load previous correction values

Go to lower left corner of res- Again, no critical chip regions should be

onator chip imaged before exposure

Temporary origin correction Set lower left corner as (U ,V ) = (0, 0)

Find closest qubit alignment Set center of qubit alignment marks as

mark with caution origin

Careful 3 point chip alignment Use the 3 alignment marks most sepa-
rated on the chip

Burn a circular contamination Requires only small optimization of fo-

dot of diameter 20 − 30 nm on cus, stigmatism and aperture correc-
top of one of the inner qubit tions, save values
alignment marks

Measure beam current Correct dwell time, check beam speed

(∼ 28 pA) and base dose of 100 µC/cm2

Check position list, exposure Also check patterns, geometries and

layers, working field settings doses
and positions on waver map

Go to origin and start exposure Exposure details: ∼ 8 nm step size (1

of the position list pixel), 20 ms waiting time, fly back factor
0.15, beam drag correction on

Before qubit exposure perform Center the cross hair in at least 3 of

the beam based write-field the 4 scans on the qubit alignment mark
alignment two or three times and confirm the correction
B.4. SHADOW EVAPORATION 141

B.3 Development
The stated development times should tried to be met as accurately as possible. Carefully
rinse and dry all glass ware before use. During the process steps the chip is slowly moved
inside the liquids to provide steady liquid flow across the chip surface. Note that freshly
mixed developer can have a reduced temperature which leads to a slower development
rate.

Process step Description Comments

Al removal 10 % NaOH in H2 O for 30 s Move chip steadily in the liquid

H2 O for 10 s Move chip steadily in the liquid

Blow dry extra carefully and Residual H2 O can substantially alter

make sure the chip and the the subsequent development. Check
tweezers are completely dry that all aluminum is removed.

Development MIBK in isopropanol (1:3) for Move chip in the liquid steadily. Note,
50 s freshly mixed developer can develop at
a lower rate.

Isopropanol for 10 s Move chip steadily in the liquid

Blow dry carefully Check structure under optical micro-

scope

B.4 Shadow evaporation

In the process described below we obtain Josephson energies between 60 and 20 GHz
for oxidation times between 6 and 40 minutes. This is the case for two Josephson junc-
tions with a typical size of 250 by 200 nm2 .

Process step Description Comments

Preliminaries Mount chips on special holder Rotate sample holder such that resist
with slider clamps bridges are aligned to the sample stage
tilt direction

Standard load lock pump down

Purge ArO2 gas line Vent load lock with ArO2 gas (10 Torr)

Pump down Start pump down and pause process

for at least 1 h (up to 15 h)

Evaporation 20 nm aluminum at 0.5 nm/s in Typical beam current 270-300 mA, use
perpendicular direction (0◦ ) material Alpure
142 APPENDIX B. NANO FABRICATION RECIPES

Oxidation Oxidation with 15 % O2 in Ar at Typical transmon oxidation times are in

9 Torr the range of 10 to 20 minutes, measure
pressure with baratron sensor

Evaporation 80 nm aluminum at 20◦ and 0.5 Sign of evaporation angle is typically

nm/s chosen such that the top layer junction
fingers overlap the bottom layer part of
the SQUID

Passivation Oxidation with O2 pure at 10

Torr for 15 minutes

B.5 Resist stripping

For the liftoff procedure we clamp the chips at one corner by means of teflon clamp. The
chip holder is placed on a teflon base that gives space to a magnetic stirrer below the
base. Base and holder are dipped into an acetone filled glass. This glass is heated in hot
water which also contains a magnetic stirrer.

Process step Description Comments

Liftoff Heat acetone filled glass in wa- Water and acetone glass contain a
ter to 50 degree Celsius on a magnetic stirrer
hotplate

Mount chips vertically on a

holder

Place holder on a base and dip Place holder close to the perimeter of
both in hot acetone the glass

Wait for 1-2 hours at a rotation

speed of about 300 rpm

Use a pipette to blow off the alu- Never remove the chip from the solvent
minum layer in one piece before the aluminum layer is completely
removed

Wait for another ∼10 minutes

Inspection Relocate the chips to a shallow Do this step quickly such that the chip
glass filled with isopropanol does not dry

Check liftoff under optical mi- Take a close look at the liftoff of the
croscope SQUIDs and chip edges in particular
B.5. RESIST STRIPPING 143

Remove residues with pipette If absolutely necessary apply ultrasonic

or tweezers with caution (if nec- power 2 in a bath of acetone for a short
essary) time. This should not be required and
can remove part of the qubit metalliza-
tion!

Rinse in isopropanol and blow

dry carefully

Optical imaging and DC resis-

tance measurements
 B IBLIOGRAPHY

Abdumalikov10 A. A. Abdumalikov, O. Astafiev, A. M. Zagoskin, Y. A. Pashkin, Y. Nakamura & J. S. Tsai.

Electromagnetically induced transparency on a single artificial atom. Phys. Rev. Lett., 104
193601 (2010). (Cited on page 8)

Agarwal84 G. S. Agarwal. Vacuum-field Rabi splittings in microwave absorption by Rydberg atoms in a

cavity. Phys. Rev. Lett., 53 1732 (1984). (Cited on pages 98 and 107)

Agarwal86 G. Agarwal, R. Bullougha & G. Hildred. Exact finite temperature quantum statistics of
single atom electrodynamics in a cavity of arbitrary Q. Opt. Commun., 59 23 (1986).
(Cited on page 117)

Agarwal91 G. Agarwal, R. Bullough & N. Nayak. Probing the dressed states of an atom interacting with
a quantized field. Opt. Commun., 85 202 (1991). (Cited on page 117)

Al-Saidi01 W. A. Al-Saidi & D. Stroud. Eigenstates of a small Josephson junction coupled to a resonant
cavity. Phys. Rev. B, 65 014512 (2001). (Cited on page 7)

Allen87 L. Allen & J. Eberly. Optical resonance and two-level atoms. Dover (1987).
(Cited on pages 17, 79, and 81)

Alsing91 P. Alsing & H. J. Carmichael. Spontaneous dressed-state polarization of a coupled atom

and cavity mode. Quantum Opt., 3 13 (1991). (Cited on pages 117 and 125)

Ambegaokar63 V. Ambegaokar & A. Baratoff. Tunneling between superconductors. Phys. Rev. Lett., 10 486
(1963). (Cited on page 53)

Ansmann09 M. Ansmann, H. Wang, R. C. Bialczak, M. Hofheinz, E. Lucero, M. Neeley, A. D. O’Connell,

D. Sank, M. Weides, J. Wenner, A. N. Cleland & J. M. Martinis. Violation of Bell’s inequality
in Josephson phase qubits. Nature, 461 504 (2009). (Cited on page 8)

Armen09 M. A. Armen, A. E. Miller & H. Mabuchi. Spontaneous dressed-state polarization

in the strong driving regime of cavity QED. Phys. Rev. Lett., 103 173601 (2009).
(Cited on pages 117 and 125)

Astafiev07 O. Astafiev, K. Inomata, A. O. Niskanen, T. Yamamoto, Y. A. Pashkin, Y. Nakamura & J. S.

Tsai. Single artificial-atom lasing. Nature, 449 588 (2007). (Cited on page 8)

Astafiev10a O. Astafiev, A. M. Zagoskin, J. Abdumalikov, A. A., Y. A. Pashkin, T. Yamamoto, K. Inomata,

Y. Nakamura & J. S. Tsai. Resonance fluorescence of a single artificial atom. Science, 327
840 (2010). (Cited on page 8)

Astafiev10b O. V. Astafiev, A. A. Abdumalikov, A. M. Zagoskin, Y. A. Pashkin, Y. Nakamura & J. S. Tsai.

Ultimate on-chip quantum amplifier. Phys. Rev. Lett., 104 183603 (2010). (Cited on page 8)

Bardeen57 J. Bardeen, L. Cooper & J. Schrieffer. Theory of superconductivity. Physical Review, 108
1175 (1957). (Cited on page 53)

I
II BIBLIOGRAPHY

Baur07 M. Baur. Quantum Electrodynamics with Superconducting Circuits: Measurement of the

Cavity Photon Number using Ramsey Interference. Master’s thesis, ETH Zurich (2007).
(Cited on page 20)

Baur09 M. Baur, S. Filipp, R. Bianchetti, J. M. Fink, M. Goppl, L. Steffen, P. J. Leek, A. Blais &
A. Wallraff. Measurement of Autler-Townes and Mollow transitions in a strongly driven su-
perconducting qubit. Phys. Rev. Lett., 102 243602 (2009). (Cited on pages 8
and 123)

Bernardot92 F. Bernardot, P. Nussenzveig, M. Brune, J. M. Raimond & S. Haroche. Vacuum Rabi splitting
observed on a microscopic atomic sample in a microwave cavity. Europhys. Lett., 17 33
(1992). (Cited on pages 107 and 108)

Bertet02 P. Bertet, S. Osnaghi, P. Milman, A. Auffeves, P. Maioli, M. Brune, J. M. Raimond &

S. Haroche. Generating and probing a two-photon Fock state with a single atom in a cavity.
Phys. Rev. Lett., 88 143601 (2002). (Cited on pages 89 and 91)

Bethe47 H. A. Bethe. The electromagnetic shift of energy levels. Phys. Rev., 72 339 (1947).
(Cited on page 3)

Bialczak10 R. C. Bialczak, M. Ansmann, M. Hofheinz, M. Lenander, E. Lucero, M. Neeley, A. D.

O’Connell, D. Sank, H. Wang, M. Weides, J. Wenner, T. Yamamoto, A. N. Cleland & J. M.
Martinis. Demonstration of a tuneable coupler for superconducting qubits using coherent,
time domain, two-qubit operations. arXiv:1007.2219 (2010). (Cited on page 83)

Bianchetti09 R. Bianchetti, S. Filipp, M. Baur, J. M. Fink, M. Göppl, P. J. Leek, L. Steffen, A. Blais &
A. Wallraff. Dynamics of dispersive single-qubit readout in circuit quantum electrodynamics.
Phys. Rev. A, 80 043840 (2009). (Cited on pages 23, 24, 77, 80, 81, 120, and 123)

Bianchetti10a R. Bianchetti. Control and readout of a superconducting artificial atom. Ph.D. thesis, ETH
Zurich (2010). (Cited on pages 36, 43, and 81)

Bianchetti10b R. Bianchetti, S. Filipp, M. Baur, J. M. Fink, C. Lang, L. Steffen, M. Boissonneault, A. Blais

& A. Wallraff. Control and tomography of a three level superconducting artificial atom. Phys.
Rev. Lett. (in print) (2010). (Cited on pages 24, 32, 73, 77, 80, 81, and 123)

Bishop09 L. S. Bishop, J. M. Chow, J. Koch, A. A. Houck, M. H. Devoret, E. Thuneberg, S. M. Girvin

& R. J. Schoelkopf. Nonlinear response of the vacuum Rabi resonance. Nat. Phys., 5 105
(2009). (Cited on pages 8, 90, 100, 105, 111, and 117)

Bishop10a L. S. Bishop. Circuit Quantum Electrodynamics: Volume II. Ph.D. thesis, Yale University
(2010). (Cited on page 11)

Bishop10b L. S. Bishop, E. Ginossar & S. M. Girvin. Response of the strongly driven Jaynes-Cummings
oscillator. Phys. Rev. Lett., 105 100505 (2010). (Cited on pages 24, 73, and 77)

Blais03 A. Blais, A. M. van den Brink & A. M. Zagoskin. Tunable coupling of superconducting qubits.
Phys. Rev. Lett., 90 127901 (2003). (Cited on page 7)

Blais04 A. Blais, R.-S. Huang, A. Wallraff, S. M. Girvin & R. J. Schoelkopf. Cavity quantum electro-
dynamics for superconducting electrical circuits: An architecture for quantum computation.
Phys. Rev. A, 69 062320 (2004). (Cited on pages 5, 7, 11, 18, 90, and 92)

Blais07 A. Blais, J. Gambetta, A. Wallraff, D. I. Schuster, S. M. Girvin, M. H. Devoret & R. J.

Schoelkopf. Quantum-information processing with circuit quantum electrodynamics. Phys.
Rev. A, 75 032329 (2007). (Cited on pages 7 and 24)

Boca04 A. Boca, R. Miller, K. M. Birnbaum, A. D. Boozer, J. McKeever & H. J. Kimble. Observation

of the vacuum Rabi spectrum for one trapped atom. Phys. Rev. Lett., 93 233603 (2004).
(Cited on pages 91 and 111)
BIBLIOGRAPHY III

Bocquillon09 E. Bocquillon, C. Couteau, M. Razavi, R. Laflamme & G. Weihs. Coherence measures for
heralded single-photon sources. Phys. Rev. A, 79 035801 (2009). (Cited on page 92)

Boissonneault08 M. Boissonneault, J. M. Gambetta & A. Blais. Nonlinear dispersive regime of cavity qed:
The dressed dephasing model. Phys. Rev. A, 77 305 (2008). (Cited on page 80)

Boissonneault10 M. Boissonneault, J. M. Gambetta & A. Blais. Improved superconducting qubit readout by

qubit-induced nonlinearities. Phys. Rev. Lett., 105 100504 (2010). (Cited on pages 24, 73,
and 77)

Bolesch10 C. Bolesch. Ultra low noise amplifiers and their DC wiring. Semester thesis, ETH Zurich
(2010). (Cited on page 45)

Bouchiat98 V. Bouchiat, D. Vion, P. Joyez, D. Esteve & M. H. Devoret. Quantum coherence with a single
Cooper pair. Phys. Scr., T76 165 (1998). (Cited on pages 13, 54, 92, and 96)

Bourassa09 J. Bourassa, J. M. Gambetta, J. A. A. Abdumalikov, O. Astafiev, Y. Nakamura & A. Blais.

Ultrastrong coupling regime of cavity QED with phase-biased flux qubits. Phys. Rev. A, 80
032109 (2009). (Cited on pages 22 and 124)

Boyer75 T. H. Boyer. Random electrodynamics: The theory of classical electrodynamics with classi-
cal electromagnetic zero-point radiation. Phys. Rev. D, 11 790 (1975). (Cited on page 4)

Bozyigit08 D. Bozyigit. Design and development of an high performance signal processing platform for
qubit readout (2008). (Cited on pages 46 and 83)

Bozyigit10a D. Bozyigit. Correlation function measurements of a microwave frequency single photon

source. Master’s thesis, ETH Zurich (2010). (Cited on pages 33, 44, 46, and 83)

Bozyigit10b D. Bozyigit, C. Lang, L. Steffen, J. M. Fink, M. Baur, R. Bianchetti, P. J. Leek, S. Filipp,

M. P. da Silva, A. Blais & A. Wallraff. Antibunching of microwave frequency photons ob-
served in correlation measurements using linear detectors. Nat. Phys. (in print) (2010).
(Cited on pages 4, 8, 32, 33, 45, 46, 79, 123, and 124)

Bozyigit10c D. Bozyigit, C. Lang, L. Steffen, J. M. Fink, M. Baur, R. Bianchetti, P. J. Leek, S. Filipp, M. P.

da Silva, A. Blais & A. Wallraff. Measurements of the correlation function of a microwave
frequency single photon source. arXiv:1002.3738 (2010). (Cited on pages 5, 8, 32, 79, 81,
and 123)

Brennecke07 F. Brennecke, T. Donner, S. Ritter, T. Bourdel, M. Kohl & T. Esslinger. Cavity QED with a
Bose-Einstein condensate. Nature, 450 268 (2007). (Cited on page 108)

Brown56 R. H. Brown & R. Q. Twiss. Correlation between photons in coherent beams of light. Nature,
177 27 (1956). (Cited on page 4)

Brune94 M. Brune, P. Nussenzveig, F. Schmidt-Kaler, F. Bernardot, A. Maali, J. M. Raimond &

S. Haroche. From Lamb shift to light shifts: Vacuum and subphoton cavity fields measured
by atomic phase sensitive detection. Phys. Rev. Lett., 72 3339 (1994). (Cited on page 20)

Brune96 M. Brune, F. Schmidt-Kaler, A. Maali, J. Dreyer, E. Hagley, J. M. Raimond & S. Haroche.

Quantum Rabi oscillation: A direct test of field quantization in a cavity. Phys. Rev. Lett., 76
1800 (1996). (Cited on pages 4, 89, 91, and 111)

Buisson01 O. Buisson & F. Hekking. Entangled states in a Josephson charge qubit coupled to a su-
perconducting resonator. In D. V. Averin, B. Ruggiero & P. Silvestrini, eds., Macroscopic
Quantum Coherence and Quantum Computing. Kluwer, New York (2001). (Cited on page 7)

Burkard04 G. Burkard, R. H. Koch & D. P. DiVincenzo. Multilevel quantum description of decoherence

in superconducting qubits. Phys. Rev. B, 69 064503 (2004). (Cited on page 11)

Büttiker87 M. Büttiker. Zero-current persistent potential drop across small-capacitance Josephson

junctions. Phys. Rev. B, 36 3548 (1987). (Cited on page 13)
IV BIBLIOGRAPHY

Carmichael96 H. J. Carmichael, P. Kochan & B. C. Sanders. Photon correlation spectroscopy. Phys. Rev.
Lett., 77 631 (1996). (Cited on pages 89, 90, and 91)

Carmichael99 H. J. Carmichael & B. C. Sanders. Multiatom effects in cavity QED with atomic beams. Phys.
Rev. A, 60 2497 (1999). (Cited on page 108)

Castellanos-Beltran07 M. A. Castellanos-Beltran & K. W. Lehnert. Widely tunable parametric amplifier based

on a superconducting quantum interference device array resonator. Appl. Phys. Lett., 91
083509 (2007). (Cited on page 124)

Castellanos-Beltran08 M. A. Castellanos-Beltran, K. D. Irwin, G. C. Hilton, L. R. Vale & K. W. Lehnert. Amplifi-

cation and squeezing of quantum noise with a tunable Josephson metamaterial. Nat. Phys.,
4 929 (2008). (Cited on page 124)

Chen07 G. Chen, Z. Chen & J. Liang. Simulation of the superradiant quantum phase transition
in the superconducting charge qubits inside a cavity. Phys. Rev. A, 76 055803 (2007).
(Cited on page 124)

Childs96 J. J. Childs, K. An, M. S. Otteson, R. R. Dasari & M. S. Feld. Normal-mode line

shapes for atoms in standing-wave optical resonators. Phys. Rev. Lett., 77 2901 (1996).
(Cited on page 108)

Chiorescu04 I. Chiorescu, P. Bertet, K. Semba, Y. Nakamura, C. J. P. M. Harmans & J. E. Mooij. Co-

herent dynamics of a flux qubit coupled to a harmonic oscillator. Nature, 431 159 (2004).
(Cited on pages 7 and 89)

Chow10a J. M. Chow, L. DiCarlo, J. M. Gambetta, F. Motzoi, L. Frunzio, S. M. Girvin & R. J.

Schoelkopf. Implementing optimal control pulse shaping for improved single-qubit gates
(2010). (Cited on pages 8 and 17)

Chow10b J. M. Chow, L. DiCarlo, J. M. Gambetta, A. Nunnenkamp, L. S. Bishop, L. Frunzio, M. H.

Devoret, S. M. Girvin & R. J. Schoelkopf. Detecting highly entangled states with a joint qubit
readout. Phys. Rev. A, 81 062325 (2010). (Cited on page 8)

Cirac91 J. I. Cirac, H. Ritsch & P. Zoller. Two-level system interacting with a finite-bandwidth thermal
cavity mode. Phys. Rev. A, 44 4541 (1991). (Cited on page 117)

Clarke08 J. Clarke & F. K. Wilhelm. Superconducting quantum bits. Nature, 453 1031 (2008).
(Cited on page 13)

Clauser74 J. F. Clauser. Experimental distinction between the quantum and classical field-theoretic
predictions for the photoelectric effect. Phys. Rev. D, 9 853 (1974). (Cited on page 4)

Clerk07 A. A. Clerk & D. W. Utami. Using a qubit to measure photon-number statistics of a driven
thermal oscillator. Phys. Rev. A, 75 042302 (2007). (Cited on page 121)

Cohen-Tannoudji89 C. Cohen-Tannoudji, J. Dupont-Roc & G. Grynberg. Photons & Atoms - Introduction to

Quantum Electrodynamics. Wiley (1989). (Cited on pages 3, 5, and 18)

Cohen-Tannoudji98 C. Cohen-Tannoudji, J. Dupont-Roc & G. Grynberg. Atom-Photon Interactions: Basic

Processes and Applications. Wiley Science Paperback Series (1998). (Cited on page 18)

Colombe07 Y. Colombe, T. Steinmetz, G. Dubois, F. Linke, D. Hunger & J. Reichel. Strong atom-field
coupling for Bose-Einstein condensates in an optical cavity on a chip. Nature, 450 272
(2007). (Cited on page 108)

Cottet02 A. Cottet, D. Vion, A. Aassime, P. Joyez, D. Esteve & M. H. Devoret. Implementation of

a combined charge-phase quantum bit in a superconducting circuit. Physica C, 367 197
(2002). (Cited on page 15)
BIBLIOGRAPHY V

Court08 N. A. Court, A. J. Ferguson, R. Lutchyn & R. G. Clark. Quantitative study of quasi-

particle traps using the single-cooper-pair transistor. Phys. Rev. B, 77 100501 (2008).
(Cited on pages 54 and 65)

daSilva10 M. P. da Silva, D. Bozyigit, A. Wallraff & A. Blais. Schemes for the observation of photon
correlation functions in circuit QED with linear detectors. Phys. Rev. A, 82 043804 (2010).
(Cited on page 46)

De Liberato09 S. De Liberato, D. Gerace, I. Carusotto & C. Ciuti. Extracavity quantum vacuum radiation
from a single qubit. Phys. Rev. A, 80 053810 (2009). (Cited on page 124)

Delanty10 M. Delanty, S. Rebic & J. Twamley. Superradiance and phase multistability in circuit quantum
electrodynamics. arXiv:1007.2231 (2010). (Cited on pages 124 and 125)

Deleglise08 S. Deleglise, I. Dotsenko, C. Sayrin, J. Bernu, M. Brune, J.-M. Raimond & S. Haroche. Re-
construction of non-classical cavity field states with snapshots of their decoherence. Nature,
455 510 (2008). (Cited on page 90)

Deppe08 F. Deppe, M. Mariantoni, E. P. Menzel, A. Marx, S. Saito, K. Kakuyanagi, H. Tanaka,

T. Meno, K. Semba, H. Takayanagi, E. Solano & R. Gross. Two-photon probe of the
Jaynes-Cummings model and controlled symmetry breaking in circuit QED. Nat. Phys.,
4 686 (2008). (Cited on page 8)

Deutsch85 D. Deutsch. Quantum theory, the Church-Turing principle and the universal quantum com-
puter. Proc. R. Soc. Lond. A, 400 97 (1985). (Cited on page 13)

Deutsch92 D. Deutsch & R. Jozsa. Rapid solution of problems by quantum computation. Proc. R. Soc.
Lond., 439 553 (1992). (Cited on page 13)

Devoret97 M. H. Devoret. Quantum fluctuations in electrical circuits. In S. Reynaud, E. Giacobino

& J. Zinn-Justin, eds., Quantum Fluctuations: Les Houches Session LXIII, pp. 351–386.
Elsevier (1997). (Cited on pages 11 and 13)

Devoret03 M. H. Devoret & J. M. Martinis. Quantum entanglement and information processing. In J.-M.
Raimond, J. Dalibard & D. Esteve, eds., Les Houches Session LXXIX, pp. 443–485. Elsevier
(2003). (Cited on page 15)

Devoret04 M. H. Devoret, A. Wallraff & J. M. Martinis. Superconducting qubits: A short review. cond-
mat/0411174v1 (2004). (Cited on page 13)

Devoret07 M. Devoret, S. Girvin & R. Schoelkopf. Circuit-QED: How strong can the coupling between a
Josephson junction atom and a transmission line resonator be? Ann. Phys., 16 767 (2007).
(Cited on page 22)

DiCarlo09 L. DiCarlo, J. M. Chow, J. M. Gambetta, L. S. Bishop, B. R. Johnson, D. I. Schuster, J. Majer,

A. Blais, L. Frunzio, S. M. Girvin & R. J. Schoelkopf. Demonstration of two-qubit algorithms
with a superconducting quantum processor. Nature, 460 240 (2009). (Cited on pages 5, 8,
24, and 125)

DiCarlo10 L. DiCarlo, M. D. Reed, L. Sun, B. R. Johnson, J. M. Chow, J. M. Gambetta, L. Frunzio,

S. M. Girvin, M. H. Devoret & R. J. Schoelkopf. Preparation and measurement of three-qubit
entanglement in a superconducting circuit. Nature, 467 574 (2010). (Cited on pages 8
and 24)

Dicke54 R. H. Dicke. Coherence in spontaneous radiation processes. Phys. Rev., 93 99 (1954).

(Cited on page 107)

Dirac27 P. Dirac. The quantum theory of the emission and absorption of radiation. Proc. R. Soc.
Lond., A 114 243 (1927). (Cited on page 3)
VI BIBLIOGRAPHY

DiVincenzo00 D. P. DiVincenzo. The physical implementation of quantum computation. Fortschr. Phys., 48

771 (2000). (Cited on page 13)

Dodonov09 A. V. Dodonov. Photon creation from vacuum and interactions engineering in nonstationary
circuit QED. J. Phys.: Conf. Ser., 161 012029 (2009). (Cited on page 124)

Dodonov10 V. V. Dodonov. Current status of the dynamical Casimir effect. Phys. Scr., 82 038105 (2010).
(Cited on page 124)

Dolan88 G. Dolan & J. Dunsmuir. Very small (20 nm) lithographic wires, dots, rings, and tunnel
junctions. Physica B, 152 7 (1988). (Cited on page 63)

Dür00 W. Dür, G. Vidal & J. I. Cirac. Three qubits can be entangled in two inequivalent ways. Phys.
Rev. A, 62 062314 (2000). (Cited on page 112)

Eberly80 J. H. Eberly, N. B. Narozhny & J. J. Sanchez-Mondragon. Periodic spontaneous collapse

and revival in a simple quantum model. Phys. Rev. Lett., 44 1323 (1980). (Cited on page 4)

Eichler10a C. Eichler, et al. & A. Wallraff. Field state tomography of itinerant microwave photons (2010).
In preperation. (Cited on page 124)

Eichler10b C. Eichler, et al. & A. Wallraff. Gaussian state reconstruction of a two-mode squeezed state
(2010). In preperation. (Cited on page 124)

Englund07 D. Englund, A. Faraon, I. Fushman, N. Stoltz, P. Petroff & J. Vuckovic. Controlling cavity
reflectivity with a single quantum dot. Nature, 450 857 (2007). (Cited on pages 89 and 91)

Enss05 C. Enss & S. Hunklinger. Low-Temperature Physics. Springer-Verlag (2005).

(Cited on page 39)

Everitt09 M. J. Everitt, W. J. Munro & T. P. Spiller. Quantum-classical crossover of a field mode. Phys.
Rev. A, 79 032328 (2009). (Cited on page 115)

Fermi32 E. Fermi. Quantum theory of radiation. Rev. Mod. Phys., 4 87 (1932). (Cited on page 3)

Feynman71 R. Feynman, M. Sands & R. Leighton. The Feynman Lectures on Physics Vol.3. Addison-
Wesley (1971). (Cited on page 3)

Feynman82 R. P. Feynman. Simulating physics with computers. Int. J. Theor. Phys., 21 467 (1982).
(Cited on page 13)

Filipp09 S. Filipp, P. Maurer, P. J. Leek, M. Baur, R. Bianchetti, J. M. Fink, M. Göppl, L. Steffen,

J. M. Gambetta, A. Blais & A. Wallraff. Two-qubit state tomography using a joint dispersive
readout. Phys. Rev. Lett., 102 200402 (2009). (Cited on pages 24 and 81)

Fink07 J. M. Fink. Single qubit control and observation of Berry’s phase in a superconducting quan-
tum circuit. Master’s thesis, Universität Wien and ETH Zurich (2007). (Cited on page 81)

Fink08 J. M. Fink, M. Göppl, M. Baur, R. Bianchetti, P. J. Leek, A. Blais & A. Wallraff. Climbing the
Jaynes-Cummings ladder and observing its nonlinearity in a cavity QED system. Nature,
454 315 (2008). (Cited on pages 4, 5, 8, 87, 90, 103, 111, 117, and 123)

Fink09a J. M. Fink, M. Baur, R. Bianchetti, S. Filipp, M. Göppl, P. J. Leek, L. Steffen, A. Blais &
A. Wallraff. Thermal excitation of multi-photon dressed states in circuit quantum electrody-
namics. Phys. Scr., T137 014013 (2009). (Cited on pages 32, 87, 92, 103, 117,
and 123)

Fink09b J. M. Fink, R. Bianchetti, M. Baur, M. Goppl, L. Steffen, S. Filipp, P. J. Leek, A. Blais &
A. Wallraff. Dressed collective qubit states and the Tavis-Cummings model in circuit QED.
Phys. Rev. Lett., 103 083601 (2009). (Cited on pages 8, 32, 88, and 123)
BIBLIOGRAPHY VII

Fink10 J. M. Fink, L. Steffen, P. Studer, L. S. Bishop, M. Baur, R. Bianchetti, D. Bozyigit, C. Lang,

S. Filipp, P. J. Leek & A. Wallraff. Quantum-to-classical transition in cavity quantum electro-
dynamics. Phys. Rev. Lett., 105 163601 (2010). (Cited on pages 8, 32, 88, 92, 100, 103,
and 123)

Forn-Díaz10 P. Forn-Díaz, J. Lisenfeld, D. Marcos, J. J. GarcÃa-Ripoll, E. Solano, C. J. P. M. Harmans

& J. E. Mooij. Observation of the Bloch-Siegert shift in a qubit-oscillator system in the
ultrastrong coupling regime. arXiv:1005.1559v1 (2010). (Cited on pages 22 and 124)

Fragner08 A. Fragner, M. Göppl, J. M. Fink, M. Baur, R. Bianchetti, P. J. Leek, A. Blais & A. Wallraff.
Resolving vacuum fluctuations in an electrical circuit by measuring the Lamb shift. Science,
322 1357 (2008). (Cited on pages 8, 22, 76, 100, and 123)

Frey08 T. Frey. Design of microwave beam splitters for photon correlation measurements. Semster
thesis, ETH Zurich (2008). (Cited on page 33)

Frunzio05 L. Frunzio, A. Wallraff, D. Schuster, J. Majer & R. Schoelkopf. Fabrication and character-
ization of superconducting circuit QED devices for quantum computation. IEEE T. Appl.
Supercon., 15 860 (2005). (Cited on page 33)

Gambetta06 J. Gambetta, A. Blais, D. I. Schuster, A. Wallraff, L. Frunzio, J. Majer, M. H. Devoret, S. M.

Girvin & R. J. Schoelkopf. Qubit-photon interactions in a cavity: Measurement-induced
dephasing and number splitting. Phys. Rev. A, 74 042318 (2006). (Cited on pages 22
and 79)

Gardiner91 C. W. Gardiner & P. Zoller. Quantum Noise. Springer (1991). (Cited on page 3)

Gerry05 C. Gerry & P. L. Knight. Introductory Quantum Optics. Cambridge University Press (2005).
(Cited on page 20)

Ghosh87 R. Ghosh & L. Mandel. Observation of nonclassical effects in the interference of two pho-
tons. Phys. Rev. Lett., 59 1903 (1987). (Cited on page 4)

Giampaolo09 S. M. Giampaolo & F. Illuminati. Long-distance entanglement and quantum teleportation in

coupled-cavity arrays. Phys. Rev. A, 80 050301 (2009). (Cited on page 124)

Glauber63a R. J. Glauber. Coherent and incoherent states of the radiation field. Phys. Rev., 131 2766
(1963). (Cited on page 4)

Glauber63b R. J. Glauber. The quantum theory of optical coherence. Phys. Rev., 130 2529 (1963).
(Cited on page 4)

Gleyzes07 S. Gleyzes, S. Kuhr, C. Guerlin, J. Bernu, S. Deleglise, U. Busk Hoff, M. Brune, J.-M. Rai-
mond & S. Haroche. Quantum jumps of light recording the birth and death of a photon in a
cavity. Nature, 446 297 (2007). (Cited on pages 20 and 90)

Göppert-Mayer31 M. Göppert-Mayer. Über Elementarakte mit zwei Quantensprüngen. Ann. Phys., 9 273
(1931). (Cited on pages 5 and 18)

Göppl08 M. Göppl, A. Fragner, M. Baur, R. Bianchetti, S. Filipp, J. M. Fink, P. J. Leek, G. Puebla,

L. Steffen & A. Wallraff. Coplanar waveguide resonators for circuit quantum electrodynamics.
J. Appl. Phys., 104 113904 (2008). (Cited on pages 9, 33, and 72)

Göppl09 M. Göppl. Engineering Quantum Electronic Chips - Realization and Characteriza-

tion of Circuit Quantum Electrodynamics Systems. Ph.D. thesis, ETH Zurich (2009).
(Cited on pages 33, 55, 71, and 81)

Grajcar08 M. Grajcar, S. H. W. van der Ploeg, A. Izmalkov, E. Il/’ichev, H.-G. Meyer, A. Fedorov,
A. Shnirman & G. Schon. Sisyphus cooling and amplification by a superconducting qubit.
Nat. Phys., 4 612 (2008). (Cited on page 8)
VIII BIBLIOGRAPHY

Grangier86 P. Grangier, G. Roger & A. Aspect. Experimental evidence for a photon anticorrelation effect
on a beam splitter: A new light on single-photon interferences. Europhys. Lett., 1 173 (1986).
(Cited on page 4)

Grover96 L. K. Grover. A fast quantum mechanical algorithm for database search. In Proceedings
of the twenty-eighth annual ACM symposium on theory of computing, pp. 212–219. ACM,
Philadelphia, Pennsylvania (1996). (Cited on page 13)

Guerlin07 C. Guerlin, J. Bernu, S. Deleglise, C. Sayrin, S. Gleyzes, S. Kuhr, M. Brune, J.-M. Raimond &
S. Haroche. Progressive field-state collapse and quantum non-demolition photon counting.
Nature, 448 889 (2007). (Cited on pages 20, 90, and 92)

Haack10 G. Haack, F. Helmer, M. Mariantoni, F. Marquardt & E. Solano. Resonant quantum gates in
circuit quantum electrodynamics. Phys. Rev. B, 82 024514 (2010). (Cited on pages 123
and 124)

Haroche89 S. Haroche & D. Kleppner. Cavity quantum electrodynamics. Phys. Today, 42 24 (1989).
(Cited on page 4)

Haroche92 S. Haroche. Fundamental Systems in Quantum Optics, chap. Cavity quantum electrody-
namics, p. 767. Elsevier (1992). (Cited on page 20)

Haroche06 S. Haroche & J.-M. Raimond. Exploring the Quantum: Atoms, Cavities, and Photons. OUP
Oxford (2006). (Cited on pages 4, 18, and 115)

Helmer09 F. Helmer, M. Mariantoni, A. G. Fowler, J. von Delft, E. Solano & F. Marquardt. Cavity grid
for scalable quantum computation with superconducting circuits. Europhys. Lett.), 85 50007
(2009). (Cited on pages 83 and 124)

Hennessy07 K. Hennessy, A. Badolato, M. Winger, D. Gerace, M. Atature, S. Gulde, S. Falt, E. L. Hu

& A. Imamoglu. Quantum nature of a strongly coupled single quantum dot-cavity system.
Nature, 445 896 (2007). (Cited on pages 89 and 91)

Henzen08 D. Henzen. Measurement and analysis of superconducting microwave beam splitters. Sem-
seter thesis, ETH Zurich (2008). (Cited on page 33)

Hofheinz08 M. Hofheinz, E. M. Weig, M. Ansmann, R. C. Bialczak, E. Lucero, M. Neeley, A. D. O’Connell,

H. Wang, J. M. Martinis & A. N. Cleland. Generation of Fock states in a superconducting
quantum circuit. Nature, 454 310 (2008). (Cited on pages 5, 8, 90, 91, and 111)

Hofheinz09 M. Hofheinz, H. Wang, M. Ansmann, R. C. Bialczak, E. Lucero, M. Neeley, A. D. O’Connell,

D. Sank, J. Wenner, J. M. Martinis & A. N. Cleland. Synthesizing arbitrary quantum states
in a superconducting resonator. Nature, 459 546 (2009). (Cited on pages 5, 8, and 90)

Holmqvist08 T. Holmqvist, M. Meschke & J. P. Pekola. Double oxidation scheme for tunnel junction fabri-
cation. J. Vac. Sci. Technol. B, 26 28 (2008). (Cited on page 68)

Houck07 A. Houck, D. Schuster, J. Gambetta, J. Schreier, B. Johnson, J. Chow, L. Frunzio, J. Majer,

M. Devoret, S. Girvin & R. Schoelkopf. Generating single microwave photons in a circuit.
Nature, 449 328 (2007). (Cited on pages 5 and 8)

Houck08 A. A. Houck, J. A. Schreier, B. R. Johnson, J. M. Chow, J. Koch, J. M. Gambetta, D. I.

Schuster, L. Frunzio, M. H. Devoret, S. M. Girvin & R. J. Schoelkopf. Controlling the sponta-
neous emission of a superconducting transmon qubit. Phys. Rev. Lett., 101 080502 (2008).
(Cited on page 13)

Imamoglu09 A. Imamoglu. Cavity QED based on collective magnetic dipole coupling: Spin ensembles
as hybrid two-level systems. Phys. Rev. Lett., 102 083602 (2009). (Cited on page 125)

Ithier05 G. Ithier, E. Collin, P. Joyez, P. J. Meeson, D. Vion, D. Esteve, F. Chiarello, A. Shnirman,

Y. Makhlin, J. Schriefl & G. Schon. Decoherence in a superconducting quantum bit circuit.
Phys. Rev. B, 72 134519 (2005). (Cited on page 13)
BIBLIOGRAPHY IX

Jaynes63 E. Jaynes & F. Cummings. Comparison of quantum and semiclassical radiation theories
with application to the beam maser. Proc. IEEE, 51 89 (1963). (Cited on pages 4, 19,
and 110)

Jeurgens02 L. P. H. Jeurgens, W. G. Sloof, F. D. Tichelaar & M. E. J. Growth kinetics and mechanisms

of aluminum-oxide films formed by thermal oxidation of aluminum. J. Appl. Phys., 92 1649
(2002). (Cited on page 68)

Johansson06 J. Johansson, S. Saito, T. Meno, H. Nakano, M. Ueda, K. Semba & H. Takayanagi. Vacuum
Rabi oscillations in a macroscopic superconducting qubit LC oscillator system. Phys. Rev.
Lett., 96 127006 (2006). (Cited on pages 7, 89, 90, and 91)

Johansson09 J. R. Johansson, G. Johansson, C. M. Wilson & F. Nori. Dynamical Casimir effect in a super-
conducting coplanar waveguide. Phys. Rev. Lett., 103 147003 (2009). (Cited on page 124)

Johansson10 J. R. Johansson, G. Johansson, C. M. Wilson & F. Nori. The dynamical Casimir effect in
superconducting microwave circuits. arXiv:1007.1058 (2010). (Cited on page 124)

Johnson10 B. R. Johnson, M. D. Reed, A. A. Houck, D. I. Schuster, L. S. Bishop, E. Ginossar, J. M. Gam-

betta, L. DiCarlo, L. Frunzio, S. M. Girvin & R. J. Schoelkopf. Quantum non-demolition de-
tection of single microwave photons in a circuit. Nat. Phys., 6 663 (2010). (Cited on pages 8
and 20)

Josephson62 B. D. Josephson. Possible new effects in superconductive tunnelling. Physics Letters, 1 251
(1962). (Cited on pages 13 and 52)

Keeling09 J. Keeling. Quantum corrections to the semiclassical collective dynamics in the Tavis-
Cummings model. Phys. Rev. A, 79 053825 (2009). (Cited on page 124)

Kennard38 E. H. Kennard. Kinetic Theory of Gases. Mc (1938). (Cited on page 131)

Khitrova06 G. Khitrova, H. M. Gibbs, M. Kira, S. W. Koch & A. Scherer. Vacuum Rabi splitting in
semiconductors. Nat. Phys., 2 81 (2006). (Cited on page 111)

Kimble77 H. J. Kimble, M. Dagenais & L. Mandel. Photon antibunching in resonance fluorescence.

Phys. Rev. Lett., 39 691 (1977). (Cited on page 4)

Kirk78 W. P. Kirk & M. Twerdochlib. Improved method for minimizing vibrational motion transmitted
by pumping lines. Rev. Sci. Instrum., 49 6 (1978). (Cited on pages 40, 134, and 135)

Koch07a J. Koch, T. M. Yu, J. Gambetta, A. A. Houck, D. I. Schuster, J. Majer, A. Blais, M. H. Devoret,

S. M. Girvin & R. J. Schoelkopf. Charge-insensitive qubit design derived from the Cooper
pair box. Phys. Rev. A, 76 042319 (2007). (Cited on pages 15, 16, 50, 77, 92, 93, 95, 96,
108, 109, 116, and 118)

Koch07b R. Koch, D. DiVincenzo & J. Clarke. Model for l/f flux noise in SQUIDs and qubits. Phys.
Rev. Lett., 98 267003 (2007). (Cited on pages 13, 21, 22, and 76)

Koch09 J. Koch & K. Le Hur. Superfluid – Mott-insulator transition of light in the Jaynes-Cummings
lattice. Phys. Rev. A, 80 023811 (2009). (Cited on page 124)

Koch10 J. Koch, A. A. Houck, K. L. Hur & S. M. Girvin. Time-reversal-symmetry breaking in circuit-

QED-based photon lattices. Phys. Rev. A, 82 043811 (2010). (Cited on page 124)

Koppinen07 P. J. Koppinen, L. M. Vaisto & I. J. Maasilta. Complete stabilization and improvement of

the characteristics of tunnel junctions by thermal annealing. Appl. Phys. Lett., 90 053503
(2007). (Cited on pages 64 and 68)

Ladd10 T. D. Ladd, F. Jelezko, R. Laflamme, Y. Nakamura, C. Monroe & J. L. O’Brien. Quantum

computers. Nature, 464 45 (2010). (Cited on pages 5 and 13)
X BIBLIOGRAPHY

Lamb47 W. E. Lamb & R. Retherford. Fine structure of the hydrogen atom by a microwave method.
Physical Review, 72 241 (1947). (Cited on page 3)

Lambert09 N. Lambert, Y. nan Chen, R. Johansson & F. Nori. Quantum chaos and critical behavior on
a chip. Phys. Rev. B, 80 165308 (2009). (Cited on page 124)

Landauer91 R. Landauer. Information is physical. Physics Today, 44 23 (1991). (Cited on page 13)

Lang09 C. Lang. Read-out strategies for multi-qubit states in circuit quantum electrodynamics.
Diploma thesis, LMU Munich (2009). (Cited on pages 46 and 83)

Lang10 C. Lang, et al. & A. Wallraff. Observation of photon blockade and dressing of dressed states
(2010). In preperation. (Cited on pages 32, 33, 123, and 125)

Larson10 J. Larson. Circuit QED scheme for the realization of the Lipkin-Meshkov-Glick model. Euro-
phys. Lett., 90 54001 (2010). (Cited on page 124)

Laucht09 A. Laucht, N. Hauke, J. M. Villas-Bôas, F. Hofbauer, G. Böhm, M. Kaniber & J. J. Finley.

Dephasing of exciton polaritons in photoexcited InGaAs quantum dots in GaAs nanocavities.
Phys. Rev. Lett., 103 087405 (2009). (Cited on page 121)

Leek07 P. J. Leek, J. M. Fink, A. Blais, R. Bianchetti, M. Göppl, J. M. Gambetta, D. I. Schuster,

L. Frunzio, R. J. Schoelkopf & A. Wallraff. Observation of Berry’s phase in a solid-state
qubit. Science, 318 1889 (2007). (Cited on pages 8, 80, and 123)

Leek09 P. J. Leek, S. Filipp, P. Maurer, M. Baur, R. Bianchetti, J. M. Fink, M. Göppl, L. Steffen &
A. Wallraff. Using sideband transitions for two-qubit operations in superconducting circuits.
Phys. Rev. B, 79 180511 (2009). (Cited on pages 8, 24, 78, 80, 81, and 123)

Leek10 P. J. Leek, M. Baur, J. M. Fink, R. Bianchetti, L. Steffen, S. Filipp & A. Wallraff. Cavity
quantum electrodynamics with separate photon storage and qubit readout modes. Phys.
Rev. Lett., 104 100504 (2010). (Cited on pages 8, 32, 80, and 123)

Leib10 M. Leib & M. J. Hartmann. Bose-Hubbard dynamics of polaritons in a chain of circuit quan-
tum electrodynamics cavities. New J. Phys., 12 093031 (2010). (Cited on page 124)

Leibfried04 D. Leibfried, M. D. Barrett, T. Schaetz, J. Britton, J. Chiaverini, W. M. Itano, J. D. Jost,

C. Langer & D. J. Wineland. Toward Heisenberg-limited spectroscopy with multiparticle
entangled states. Science, 304 1476 (2004). (Cited on page 113)

Leslie04 S. Leslie, N. Shenvi, K. R. Brown, D. M. Stamper-Kurn & K. B. Whaley. Transmission

spectrum of an optical cavity containing N atoms. Phys. Rev. A, 69 043805 (2004).
(Cited on pages 107 and 108)

Lin09 G. W. Lin, X. B. Zou, X. M. Lin & G. C. Guo. Scalable, high-speed one-way quantum
computer in coupled-cavity arrays. Appl. Phys. Lett., 95 224102 (2009). (Cited on page 124)

Lisenfeld10 J. Lisenfeld, C. Müller, J. H. Cole, P. Bushev, A. Lukashenko, A. Shnirman & A. V. Usti-

nov. Rabi spectroscopy of a qubit-fluctuator system. Phys. Rev. B, 81 100511 (2010).
(Cited on page 76)

Littich09 G. Littich. Superconducting Mach-Zehnder interferometers for circuit quantum electrody-

namics. Semester thesis, ETH Zurich (2009). (Cited on page 33)

Lloyd05 S. Lloyd. The computational universe: Quantum gravity from quantum computation. quant-
ph/0501135 (2005). (Cited on page 13)

López07a C. E. López, H. Christ, J. C. Retamal & E. Solano. Effective quantum dynamics of

interacting systems with inhomogeneous coupling. Phys. Rev. A, 75 033818 (2007).
(Cited on page 112)
BIBLIOGRAPHY XI

López07b C. E. López, J. C. Retamal & E. Solano. Selective control of the symmetric Dicke subspace
in trapped ions. Phys. Rev. A, 76 033413 (2007). (Cited on page 113)

Lucero10 E. Lucero, J. Kelly, R. C. Bialczak, M. Lenander, M. Mariantoni, M. Neeley, A. O´Connell,

D. Sank, H. Wang, M. Weides, J. Wenner, T. Yamamoto, A. Cleland & J. Martinis. Reduced
phase error through optimized control of a superconducting qubit. arXiv:1007.1690 (2010).
(Cited on pages 8 and 17)

Mabuchi02 H. Mabuchi & A. C. Doherty. Cavity quantum electrodynamics: Coherence in context. Sci-
ence, 298 1372 (2002). (Cited on pages 4
and 89)

Majer07 J. Majer, J. M. Chow, J. M. Gambetta, J. Koch, B. R. Johnson, J. A. Schreier, L. Frun-

zio, D. I. Schuster, A. A. Houck, A. Wallraff, A. Blais, M. H. Devoret, S. M. Girvin & R. J.
Schoelkopf. Coupling superconducting qubits via a cavity bus. Nature, 449 443 (2007).
(Cited on pages 5 and 8)

Makhlin01 Y. Makhlin, G. Schön & A. Shnirman. Quantum-state engineering with Josephson-junction

devices. Rev. Mod. Phys., 73 357 (2001). (Cited on pages 7 and 13)

Mallet09 F. Mallet, F. R. Ong, A. Palacios-Laloy, F. Nguyen, P. Bertet, D. Vion & D. Esteve.

Single-shot qubit readout in circuit quantum electrodynamics. Nat. Phys., 5 791 (2009).
(Cited on page 8)

Mandel95 L. Mandel & E. Wolf. Optical Coherence and Quantum Optics. Cambridge University Press
(1995). (Cited on page 3)

Manucharyan09 V. E. Manucharyan, J. Koch, L. I. Glazman & M. H. Devoret. Fluxonium: Single Cooper-pair

circuit free of charge offsets. Science, 326 113 (2009). (Cited on page 125)

Marcos10 D. Marcos, M. Wubs, J. M. Taylor, R. Aguado, M. D. Lukin & A. S. Sorensen. Coupling

nitrogen vacancy centers in diamond to superconducting flux qubits. Phys. Rev. Lett. (in
print) (2010). (Cited on page 125)

Marquardt01 F. Marquardt & 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). (Cited on page 7)

Marx09 S. Marx. Optimization of the microwave properties of a cryostat sample holder. Semester
thesis, ETH Zurich (2009). (Cited on page 35)

Mondragon83 J. J. S. Mondragon, N. B. Narozhny & J. H. Eberly. Theory of spontaneous-emission line

shape in an ideal cavity. Phys. Rev. Lett., 51 550 (1983). (Cited on page 98)

Motzoi09 F. Motzoi, J. M. Gambetta, P. Rebentrost & F. K. Wilhelm. Simple pulses for elimination of
leakage in weakly nonlinear qubits. Phys. Rev. Lett., 103 110501 (2009). (Cited on pages 8
and 17)

Münstermann00 P. Münstermann, T. Fischer, P. Maunz, P. W. H. Pinkse & G. Rempe. Observation of cavity-

mediated long-range light forces between strongly coupled atoms. Phys. Rev. Lett., 84 4068
(2000). (Cited on page 108)

Nahum93 M. Nahum & J. M. Martinis. Ultrasensitive-hot-electron microbolometer. Appl. Phys. Lett.,

63 3075 (1993). (Cited on page 68)

Nakamura99 Y. Nakamura, Y. A. Pashkin & J. S. Tsai. Coherent control of macroscopic quantum states
in a single-Cooper-pair box. Nature, 398 786 (1999). (Cited on pages 7 and 13)

Nakamura01 Y. Nakamura, Y. A. Pashkin & J. S. Tsai. Rabi oscillations in a Josephson-junction charge

two-level system. Phys. Rev. Lett., 8724 246601 (2001). (Cited on page 79)
XII BIBLIOGRAPHY

Nataf10a P. Nataf & C. Ciuti. No-go theorem for superradiant quantum phase transitions in cavity QED
and counter-example in circuit QED. Nat. Commun., 1 72 (2010). (Cited on page 124)

Nataf10b P. Nataf & C. Ciuti. Vacuum degeneracy of a circuit QED system in the ultrastrong coupling
regime. Phys. Rev. Lett., 104 023601 (2010). (Cited on page 124)

Neeley08 M. Neeley, M. Ansmann, R. C. Bialczak, M. Hofheinz, N. Katz, E. Lucero, A. O’Connell,

H. Wang, A. N. Cleland & J. M. Martinis. Process tomography of quantum memory
in a josephson-phase qubit coupled to a two-level state. Nat. Phys., 4 523 (2008).
(Cited on page 76)

Neeley10 M. Neeley, R. C. Bialczak, M. Lenander, E. Lucero, M. Mariantoni, A. D. O/’Connell, D. Sank,

H. Wang, M. Weides, J. Wenner, Y. Yin, T. Yamamoto, A. N. Cleland & J. M. Martinis. Gener-
ation of three-qubit entangled states using superconducting phase qubits. Nature, 467 570
(2010). (Cited on page 8)

Nielsen00 M. A. Nielsen & I. L. Chuang. Quantum Computation and Quantum Information. Cambridge
University Press (2000). (Cited on pages 5, 12, 13, and 90)

Niemczyk10 T. Niemczyk, F. Deppe, H. Huebl, E. P. Menzel, F. Hocke, M. J. Schwarz, J. J. Garcia-Ripoll,

D. Zueco, T. Hummer, E. Solano, A. Marx & R. Gross. Circuit quantum electrodynamics in
the ultrastrong-coupling regime. Nat. Phys., 6 772 (2010). (Cited on pages 22 and 124)

O´Connell10 A. D. O´Connell, M. Hofheinz, M. Ansmann, R. C. Bialczak, M. Lenander, E. Lucero, M. Nee-

ley, D. Sank, H. Wang, M. Weides, J. Wenner, J. M. Martinis & A. N. Cleland. Quantum
ground state and single-phonon control of a mechanical resonator. Nature, 464 697 (2010).
(Cited on page 8)

OxfordInstr06 OxfordInstr. Private communication with Ricardo Viana, Oxford Instruments (2006).
(Cited on pages 129 and 130)

Palacios-Laloy09 A. Palacios-Laloy, F. Mallet, F. Nguyen, F. Ong, P. Bertet, D. Vion & D. Esteve. Spectral
measurement of the thermal excitation of a superconducting qubit. Phys. Scr., T137 014015
(2009). (Cited on page 121)

Paul63 H. Paul. Induzierte Emission bei starker Einstrahlung. Ann. Phys., 466 1521 (1963).
(Cited on page 4)

Peter05 E. Peter, P. Senellart, D. Martrou, A. Lemaitre, J. Hours, J. M. Gerard & J. Bloch. Exciton-
photon strong-coupling regime for a single quantum dot embedded in a microcavity. Phys.
Rev. Lett., 95 067401 (2005). (Cited on pages 89 and 91)

PfeifferVacuum10 PfeifferVacuum. Introduction to vacuum technology, fundamentals, conductivities.

http://www.pfeiffer-vacuum.de/en/know-how/container.action (2010). (Cited on page 131)

Plastina03 F. Plastina & G. Falci. Communicating Josephson qubits. Phys. Rev. B, 67 224514 (2003).
(Cited on page 7)

Pobell06 F. Pobell. Matter and Methods at Low Temperatures. Springer, 3rd edition, (2006).
(Cited on pages 39, 42, and 132)

Pozar93 D. M. Pozar. Microwave Engineering. Addison-Wesley Publishing Company (1993).

(Cited on pages 9, 11, 12, 44, and 72)

Puebla08 G. Puebla. Dual channel power supply. Tech. rep., Quantum Device Lab, ETH Zurich (2008).
(Cited on page 45)

Rabl06 P. Rabl, D. DeMille, J. M. Doyle, M. D. Lukin, R. J. Schoelkopf & P. Zoller. Hybrid quantum
processors: Molecular ensembles as quantum memory for solid state circuits. Phys. Rev.
Lett., 97 033003 (2006). (Cited on page 125)
BIBLIOGRAPHY XIII

Rabl07 P. Rabl & P. Zoller. Molecular dipolar crystals as high-fidelity quantum memory for hybrid
quantum computing. Phys. Rev. A, 76 042308 (2007). (Cited on page 125)

Raimond01 J. M. Raimond, M. Brune & S. Haroche. Manipulating quantum entanglement with atoms
and photons in a cavity. Rev. Mod. Phys., 73 565 (2001). (Cited on pages 4 and 89)

Raizen89 M. G. Raizen, R. J. Thompson, R. J. Brecha, H. J. Kimble & H. J. Carmichael. Normal-mode

splitting and linewidth averaging for two-state atoms in an optical cavity. Phys. Rev. Lett., 63
240 (1989). (Cited on page 107)

Rau04 I. Rau, G. Johansson & A. Shnirman. Cavity quantum electrodynamics in superconduct-

ing circuits: Susceptibility at elevated temperatures. Phys. Rev. B, 70 054521 (2004).
(Cited on pages 115, 117, and 119)

Reed10a M. D. Reed, L. DiCarlo, B. R. Johnson, L. Sun, D. I. Schuster, L. Frunzio & R. J. Schoelkopf.

High-fidelity readout in circuit quantum electrodynamics using the Jaynes-Cummings non-
linearity. Phys. Rev. Lett., 105 173601 (2010). (Cited on pages 24, 73,
and 77)

Reed10b M. D. Reed, B. R. Johnson, A. A. Houck, L. DiCarlo, J. M. Chow, D. I. Schuster, L. Frunzio &

R. J. Schoelkopf. Fast reset and suppressing spontaneous emission of a superconducting
qubit (2010). (Cited on page 13)

Reithmaier04 J. P. Reithmaier, G. Sek, A. Loffler, C. Hofmann, S. Kuhn, S. Reitzenstein, L. V. Keldysh,

V. D. Kulakovskii, T. L. Reinecke & A. Forchel. Strong coupling in a single quantum dot-
semiconductor microcavity system. Nature, 432 197 (2004). (Cited on pages 89
and 91)

Rempe87 G. Rempe, H. Walther & N. Klein. Observation of quantum collapse and revival in a one-
atom maser. Phys. Rev. Lett., 58 353 (1987). (Cited on pages 4
and 89)

Retzker07 A. Retzker, E. Solano & B. Reznik. Tavis-Cummings model and collective multiqubit entan-
glement in trapped ions. Phys. Rev. A, 75 022312 (2007). (Cited on page 113)

Richardson88 R. C. Richardson & E. N. Smith. Experimental Techniques in Condensed Matter Physics at

Low Temperatures. Addison-Wesley Publishing Company (1988). (Cited on pages 40
and 130)

Roth76 A. Roth. Vacuum technology. North-Holland Publishing Company (1976).

(Cited on pages 40 and 131)

Sandberg08a M. Sandberg, C. M. Wilson, F. Persson, T. Bauch, G. Johansson, V. Shumeiko, T. Duty &

P. Delsing. Tuning the field in a microwave resonator faster than the photon lifetime. Appl.
Phys. Lett., 92 203501 (2008). (Cited on page 124)

Sandberg08b M. Sandberg, C. M. Wilson, F. Persson, G. Johansson, V. Shumeiko, T. Bauch, T. Duty &

P. Delsing. Fast tuning of superconducting microwave cavities. AIP Conf. Proc., 1074 12
(2008). (Cited on page 124)

Savage89 C. M. Savage. Resonance fluorescence spectrum of an atom strongly coupled to a cavity.

Phys. Rev. Lett., 63 1376 (1989). (Cited on page 119)

Schlosshauer07 M. Schlosshauer. Decoherence and the Quantum-to-Classical Transition. Springer (2007).

(Cited on page 116)

Schmidlin08 S. Schmidlin. Design and characterization of microwave printed circuit boards for circuit
QED experiments. Semester thesis, ETH Zurich (2008). (Cited on page 34)

Schmidlin09 S. Schmidlin. Generation of amplitude and phase controlled microwave pulses for qubit
manipulation in circuit QED. Master’s thesis, ETH Zurich (2009). (Cited on page 38)
XIV BIBLIOGRAPHY

Schmidt09 S. Schmidt & G. Blatter. Strong coupling theory for the Jaynes-Cummings-Hubbard model.
Phys. Rev. Lett., 103 086403 (2009). (Cited on page 124)

Schmidt10 S. Schmidt, D. Gerace, A. A. Houck, G. Blatter & H. E. Türeci. Nonequilibrium delocalization-

localization transition of photons in circuit quantum electrodynamics. Phys. Rev. B, 82
100507 (2010). (Cited on page 124)

Schoelkopf08 R. Schoelkopf & S. Girvin. Wiring up quantum systems. Nature, 451 664 (2008).
(Cited on pages 5, 7, 94, and 108)

Schreier08 J. Schreier, A. Houck, J. Koch, D. I. Schuster, B. Johnson, J. Chow, J. Gambetta, J. Majer,

L. Frunzio, M. Devoret, S. Girvin & R. Schoelkopf. Suppressing charge noise decoherence
in superconducting charge qubits. Phys. Rev. B, 77 180502(R) (2008). (Cited on pages 13,
16, 77, 92, and 96)

Schuster05 D. I. Schuster, A. Wallraff, A. Blais, L. Frunzio, R.-S. Huang, J. Majer, S. M. Girvin & R. J.
Schoelkopf. AC Stark shift and dephasing of a superconducting qubit strongly coupled to a
cavity field. Phys. Rev. Lett., 94 123602 (2005). (Cited on pages 7, 22, 76, 77, 79, 93,
and 120)

Schuster07a D. I. Schuster. Circuit Quantum Electrodynamics. Ph.D. thesis, Yale University (2007).
(Cited on pages 33 and 47)

Schuster07b D. I. Schuster, A. A. Houck, J. A. Schreier, A. Wallraff, J. M. Gambetta, A. Blais, L. Frunzio,

J. Majer, B. Johnson, M. H. Devoret, S. M. Girvin & R. J. Schoelkopf. Resolving photon
number states in a superconducting circuit. Nature, 445 515 (2007). (Cited on pages 8,
22, 76, 78, 90, and 92)

Schuster08 I. Schuster, A. Kubanek, A. Fuhrmanek, T. Puppe, P. W. H. Pinkse, K. Murr & G. Rempe.

Nonlinear spectroscopy of photons bound to one atom. Nat. Phys., 4 382 (2008).
(Cited on pages 4, 90, 105, 111, and 117)

Schwabl02 F. Schwabl. Quantenmechanik. Springer-Verlag, Berlin Heidelberg New York (2002).

(Cited on page 10)

Scully97 M. O. Scully & M. S. Zubairy. Quantum Optics. Cambridge University Press (1997).
(Cited on pages 3, 4, 17, and 18)

Scully09 M. O. Scully & A. A. Svidzinsky. The super of superradiance. Science, 325 1510 (2009).
(Cited on page 124)

Scully10 M. O. Scully & A. A. Svidzinsky. The Lamb shift – yesterday, today, and tomorrow. Science,
328 1239 (2010). (Cited on page 124)

Shamailov10 S. Shamailov, A. Parkins, M. Collett & H. Carmichael. Multi-photon blockade and dressing
of the dressed states. Opt. Commun., 283 766 (2010). (Cited on page 125)

Shor97 P. W. Shor. Polynomial-time algorithms for prime factorization and discrete logarithms on a
quantum computer. SIAM J. Sci. Stat. Comp., 26 1484 (1997). (Cited on page 13)

Sillanpää07 M. A. Sillanpää, J. I. Park & R. W. Simmonds. Coherent quantum state storage and transfer
between two phase qubits via a resonant cavity. Nature, 449 438 (2007). (Cited on page 8)

Silver10 A. O. Silver, M. Hohenadler, M. J. Bhaseen & B. D. Simons. Bose-Hubbard models coupled

to cavity light fields. Phys. Rev. A, 81 023617 (2010). (Cited on page 124)

Simmonds04 R. W. Simmonds, K. M. Lang, D. A. Hite, S. Nam, D. P. Pappas & J. M. Martinis. Decoherence

in Josephson phase qubits from junction resonators. Phys. Rev. Lett., 93 077003 (2004).
(Cited on page 76)
BIBLIOGRAPHY XV

Simons01 R. N. Simons. Coplanar waveguide circuits, components and systems. Wiley Series in
Microwave and Optical Engineerging. Wiley Inter-Science (2001). (Cited on pages 9, 34,
35, and 72)

Spiller06 T. P. Spiller, K. Nemoto, S. L. Braunstein, W. J. Munro, P. van Loock & G. J. Milburn. Quantum
computation by communication. New J. Phys., 8 30 (2006). (Cited on page 5)

Srinivasan10 S. Srinivasan, A. Hoffman & H. Andrew. Straddling regime of a transmon qubit. Abstract,
APS March Meeting (2010). (Cited on page 23)

Steffen07 L. Steffen. Spin echo measurements in a superconducting qubit. Semester thesis, ETH
Zurich (2007). (Cited on page 81)

Steffen08 L. Steffen. Local Qubit Control in Circuit Quantum Electrodynamics. Master’s thesis, ETH
Zurich (2008). (Cited on pages 33 and 44)

Steffen10 M. Steffen, S. Kumar, D. P. DiVincenzo, J. R. Rozen, G. A. Keefe, M. B. Rothwell & M. B.

Ketchen. High-coherence hybrid superconducting qubit. Phys. Rev. Lett., 105 100502
(2010). (Cited on page 125)

Steinbach01 A. Steinbach, P. Joyez, A. Cottet, D. Esteve, M. H. Devoret, M. E. Huber & J. M. Martinis.

Direct measurement of the Josephson supercurrent in an ultrasmall Josephson junction.
Phys. Rev. Lett., 87 137003 (2001). (Cited on page 54)

Stockton04 J. K. Stockton, R. van Handel & H. Mabuchi. Deterministic Dicke-state preparation with
continuous measurement and control. Phys. Rev. A, 70 022106 (2004). (Cited on page 113)

Tanzilli05 S. Tanzilli, W. Tittel, M. Halder, O. Alibart, P. Baldi, N. Gisin & H. Zbinden. A photonic
quantum information interface. Nature, 437 116 (2005). (Cited on page 5)

Tavis68 M. Tavis & F. W. Cummings. Exact solution for an N -molecule-radiation-field Hamiltonian.

Phys. Rev., 170 379 (1968). (Cited on pages 107 and 109)

Tessier03 T. Tessier, I. H. Deutsch, A. Delgado & I. Fuentes-Guridi. Entanglement sharing in the two-
atom Tavis-Cummings model. Phys. Rev. A, 68 062316 (2003). (Cited on page 113)

Thompson92 R. J. Thompson, G. Rempe & H. J. Kimble. Observation of normal-mode splitting for an

atom in an optical cavity. Phys. Rev. Lett., 68 1132 (1992). (Cited on page 91)

Thompson98 R. J. Thompson, Q. A. Turchette, O. Carnal & H. J. Kimble. Nonlinear spectroscopy in the

strong-coupling regime of cavity QED. Phys. Rev. A, 57 3084 (1998). (Cited on pages 90,
91, 103, and 108)

Tian92 L. Tian & H. J. Carmichael. Incoherent excitation of the Jaynes-Cummings system. Quantum
Opt., 4 131 (1992). (Cited on page 117)

Tian10 L. Tian. Circuit QED and sudden phase switching in a superconducting qubit array. Phys.
Rev. Lett., 105 167001 (2010). (Cited on page 124)

Tinkham96 M. Tinkham. Introduction to Superconductivity. McGraw-Hill International Editions (1996).

(Cited on pages 13 and 15)

Tomadin10 A. Tomadin & R. Fazio. Many-body phenomena in qed-cavity arrays. J. Opt. Soc. Am. B, 27
A130 (2010). (Cited on page 124)

Tuchman06 A. K. Tuchman, R. Long, G. Vrijsen, J. Boudet, J. Lee & M. A. Kasevich. Normal-mode split-
ting with large collective cooperativity. Phys. Rev. A, 74 053821 (2006). (Cited on page 108)

Twamley10 J. Twamley & S. D. Barrett. Superconducting cavity bus for single nitrogen-vacancy defect
centers in diamond. Phys. Rev. B, 81 241202 (2010). (Cited on page 125)
XVI BIBLIOGRAPHY

vandenBrink02 A. M. van den Brink & A. M. Zagoskin. Liouville invariance in quantum and classical me-
chanics. Quantum Inf. Process., 1 55 (2002). (Cited on page 117)

Varcoe00 B. T. H. Varcoe, S. Brattke, M. Weidinger & H. Walther. Preparing pure photon number
states of the radiation field. Nature, 403 743 (2000). (Cited on pages 89 and 91)

Vion02 D. Vion, A. Aassime, A. Cottet, P. Joyez, H. Pothier, C. Urbina, D. Esteve & M. H. De-
voret. Manipulating the quantum state of an electrical circuit. Science, 296 886 (2002).
(Cited on pages 13 and 16)

Vion03 D. Vion, A. Aassime, A. Cottet, P. Joyez, H. Pothier, C. Urbina, D. Esteve & M. H. Devoret.
Rabi oscillations, Ramsey fringes and spin echoes in an electrical circuit. Fortschr. der
Phys., 51 462 (2003). (Cited on page 79)

Wallraff04 A. Wallraff, D. I. Schuster, A. Blais, L. Frunzio, R. S. Huang, J. Majer, S. Kumar, S. M. Girvin

& R. J. Schoelkopf. Strong coupling of a single photon to a superconducting qubit using
circuit quantum electrodynamics. Nature, 431 162 (2004). (Cited on pages 5, 7, 46, 89,
91, 92, 94, 108, 111, 115, and 116)

Wallraff05 A. Wallraff, D. I. Schuster, A. Blais, L. Frunzio, J. Majer, S. M. Girvin & R. J. Schoelkopf.

Approaching unit visibility for control of a superconducting qubit with dispersive readout.
Phys. Rev. Lett., 95 060501 (2005). (Cited on pages 7, 23, 77, 79, and 80)

Wallraff07 A. Wallraff, D. I. Schuster, A. Blais, J. M. Gambetta, J. Schreier, L. Frunzio, M. H. Devoret,

S. M. Girvin & R. J. Schoelkopf. Sideband transitions and two-tone spectroscopy of a super-
conducting qubit strongly coupled to an on-chip cavity. Phys. Rev. Lett., 99 050501 (2007).
(Cited on pages 8, 76, and 78)

Wallraff08 A. Wallraff. Quantenmechanik mit Schaltkreisen: Photonen und Qubits auf einem supraleit-
enden Mikrochip. Physik Journal, 7 39 (2008). (Cited on page 7)

Walls94 D. Walls & G. Milburn. Quantum optics. Spinger-Verlag, Berlin (1994). (Cited on pages 3,
46, and 89)

Walther06 H. Walther, B. T. H. Varcoe, B.-G. Englert & T. Becker. Cavity quantum electrodynamics.
Rep. Prog. Phys., 69 1325 (2006). (Cited on pages 4 and 89)

Wang09 H. Wang, M. Hofheinz, J. Wenner, M. Ansmann, R. C. Bialczak, M. Lenander, E. Lucero,

M. Neeley, A. D. O’Connell, D. Sank, M. Weides, A. N. Cleland & J. M. Martinis. Improving
the coherence time of superconducting coplanar resonators. Appl. Phys. Lett., 95 233508
(2009). (Cited on page 34)

Wentzel26 G. Wentzel. Zur Theorie des photoelektrischen Effekts. Z. Phys., 40 574 (1926).
(Cited on page 4)

Wentzel27 G. Wentzel. Über die Richtungsverteilung der Photoelektronen. Z. Phys., 41 828 (1927).
(Cited on page 4)

Wesenberg09 J. H. Wesenberg, A. Ardavan, G. A. D. Briggs, J. J. L. Morton, R. J. Schoelkopf, D. I. Schuster

& K. Mølmer. Quantum computing with an electron spin ensemble. Phys. Rev. Lett., 103
070502 (2009). (Cited on pages 124 and 125)

White02 G. K. White & P. J. Meeson. Experimental Techniques in Low-Temperature Physics. Oxford

University Press (2002). (Cited on pages 39, 42, and 43)

Wilk07 T. Wilk, S. C. Webster, A. Kuhn & G. Rempe. Single-atom single-photon quantum interface.
Science, 317 488 (2007). (Cited on page 5)

Wilson10 C. Wilson, T. Duty, M. Sandberg, F. Persson & P. Shumeiko, V. Delsing. Photon genera-
tion in an electromagnetic cavity with a time-dependent boundary. arXiv:1006.2540 (2010).
(Cited on page 124)
BIBLIOGRAPHY XVII

Woolley03 R. G. Woolley. Handbook of Molecular Physics and Quantum Chemistry. Wiley (2003).
(Cited on page 18)

Yamamoto99 Y. Yamamoto & A. Imamoglu. Mesoscopic Quantum Optics. Wiley (1999). (Cited on pages 3
and 18)

Yamamoto08 T. Yamamoto, K. Inomata, M. Watanabe, K. Matsuba, T. Miyazaki, W. D. Oliver, Y. Nakamura

& J. S. Tsai. Flux-driven Josephson parametric amplifier. Appl. Phys. Lett., 93 042510
(2008). (Cited on page 124)

Yang03 C.-P. Yang, S.-I. Chu & 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). (Cited on page 7)

Ye08 J. Ye, H. J. Kimble & H. Katori. Quantum state engineering and precision metrology using
state-insensitive light traps. Science, 320 1734 (2008). (Cited on page 4)

Yoshie04 T. Yoshie, A. Scherer, J. Hendrickson, G. Khitrova, H. M. Gibbs, G. Rupper, C. Ell, O. B.

Shchekin & D. G. Deppe. Vacuum Rabi splitting with a single quantum dot in a photonic
crystal nanocavity. Nature, 432 200 (2004). (Cited on pages 89 and 91)

You03 J. Q. You & F. Nori. Quantum information processing with superconducting qubits in a
microwave field. Phys. Rev. B, 68 064509 (2003). (Cited on page 7)

Yurke87 B. Yurke. Squeezed-state generation using a Josephson parametric amplifier. J. Opt. Soc.
Am. B, 4 1551 (1987). (Cited on page 124)

Yurke89 B. Yurke, L. R. Corruccini, P. G. Kaminsky, L. W. Rupp, A. D. Smith, A. H. Silver, R. W.

Simon & E. A. Whittaker. Observation of parametric amplification and deamplification in a
Josephson parametric amplifier. Phys. Rev. A, 39 2519 (1989). (Cited on page 124)

Zagoskin07 A. Zagoskin & A. Blais. Physics in Canada / La Physique au Canada, vol. 63, chap. Super-
conducting qubits. CAP (2007). (Cited on page 13)

Zhu90 Y. Zhu, D. J. Gauthier, S. E. Morin, Q. Wu, H. J. Carmichael & T. W. Mossberg. Vacuum Rabi
splitting as a feature of linear-dispersion theory: Analysis and experimental observations.
Phys. Rev. Lett., 64 2499 (1990). (Cited on pages 89, 91, and 107)

Zurek03 W. H. Zurek. Decoherence, einselection, and the quantum origins of the classical. Rev.
Mod. Phys., 75 715 (2003). (Cited on page 13)
 ACKNOWLEDGEMENTS

This doctoral thesis is the result of a collaboration with many helpful and motivated people
who all contributed in different ways.
First of all, I want to express my gratitude to my advisor Prof. Andreas Wallraff who
gave me the opportunity to work on this fascinating subject. I profited a lot from his
personal mentoring, wide knowledge, detailed interest, continuous support and steady
encouragement.
All of my colleagues at the QuDev Lab contributed to this work in one way or another.
Peter Leek showed me clean room techniques and was involved in most experiments
to some degree. Romeo Bianchetti worked with me on the installation and setup of the
dilution refrigerators and Martin Göppl fabricated some of the first samples. Matthias
Baur improved the measurement software and Lars Steffen did most of the mask design
and the photolithographic part of the fabrication. Stefan Filipp was always helpful with
theory related problems and all of us profited from Gabriel Puebla’s cables that equip the
VeriCold cryostat. In particular, I want to mention Priska Studer and Simon Michels who
worked with me on the quantum-to-classical experiment.
I also wish to express my gratitude to Prof. Alexandre Blais from Université de Sher-
brooke and Lev S. Bishop from Yale University for fruitful and enjoyable collaborations.
I cannot thank enough my parents, family and friends. You encouraged my curiosity
and you supported and accompanied me anytime before and while I was absorbed in the
scientific research for this thesis.

XIX