Vol 451|7 February 2008

Wiring up quantum systems

R. J. Schoelkopf and S. M. Girvin
The emerging field of circuit quantum electrodynamics could pave the way for the design
of practical quantum computers.
In the past two decades, mechanics for macroscopic objects, is now at
scientists and engineers least a qualified ‘yes’. Pioneering work in the
in a variety of disciplines 1980s on simple superconducting circuits
have been excited by the incorporating a Josephson junction3 (see
idea of quantum informa- κ
Box 1) showed that macroscopic variables
g
tion processing1, in which such as voltages could indeed exhibit quan-
a computation is carried tum behaviour. Further work established that
HORIZONS γ
out by controlling a com- junctions could be considered as ‘atoms with
plex collection of quantum wires’, which display energy-level quantization4
objects. This idea seeks to combine two of the and interact strongly with the electromagnetic
greatest advances in science and technology of Figure 1 | Cavity quantum electrodynamics. environment5,6. It was not until the end of the
the twentieth century. Schematic representation of a cavity quantum 1990s, however, that the first evidence for
The first breakthrough is the development electrodynamics (QED) system, consisting of an coherent superpositions7 and time-domain
of quantum mechanics, with its sometimes atom with two energy levels interacting with a control of the quantum state8 in a supercon-
strange and counterintuitive rules that hold single photon mode (pink) trapped by mirrors ducting qubit was demonstrated.
(blue) to form a cavity. The blue dot is an electron
sway in the domain of atoms and single parti- The past decade has seen rapid progress in
occupying one of the energy levels. The strong
cles. The second is the technological revolution coupling regime is reached when the interaction this field. Several different ‘flavours’ of super-
that followed the invention of the integrated rate of the atom and a single photon (g) is larger conducting qubit9 (see Box 1) have now been
circuit and the advent of powerful digital than the dissipation arising from the loss of demonstrated, and two qubits have been cou-
computers, which gave rise to the current photons (at rate κ) or from emission from the pled to demonstrate the entanglement between
information age. Surprisingly, the seemingly atom into other modes at rate γ; in other words, them10 and to perform simple quantum logic
bizarre quantum-mechanical ideas of super- when g >> κ,γ. operations11. The current state-of-the-art
position and entanglement are expected to allows for superposition states that survive for
lead to a kind of natural parallel processing back and forth during the computation. several microseconds, long enough for hun-
during computations. The unlikely marriage Many different physical implementations dreds of operations on a single qubit. With
of these two revolutions could lead to incred- of quantum information processors are being improvements in superconducting qubit
ible advances in computational power, at least pursued today. Some systems comprise ‘natu- design, as well as in the materials and methods
for certain special problems. ral’ candidates, such as single atoms, ions or used for fabricating circuits, the lifetime of the
Unfortunately, the practical challenges to spins, for which the manipulation of quantum stored quantum information may be further
making a quantum information device are states has a long history and is routine in many increased and the precision of qubit control
daunting. To build a quantum computer, the laboratories. Others are based on artificial and measurement enhanced.
classical bits that store information in an ordi- systems in the solid state, such as quantum But how can we address the second question
nary computer must first be replaced with dots or superconducting circuits. These latter and realize the quantum connections between
quantum bits (qubits). These qubits can be candidates have a certain appeal as they can qubits? For communicating quantum informa-
composed of any quantum system with two dis- be designed and fabricated using techniques tion between real atoms, optical photons are
tinct states (0 and 1), but they have the special borrowed from conventional electronics. natural candidates12. They have many advan-
property that they can be placed into quantum Before making a quantum information proc- tages, including rapid propagation and the
superpositions, existing in both states at once. essor from solid-state systems such as super- ability to be guided on optical fibres for many
A computation then proceeds by combining conducting circuits, two basic questions must kilometres without being lost. Superconduct-
manipulations of the superpositions in single be addressed. First, can the qubits be made ing qubits also interact electromagnetically,
qubits (one-bit operations) and controlled from sufficiently ‘atom-like’ circuit elements, but because of their much smaller energy-level
interactions of multiple qubits (the quantum in which the macroscopic variables such as separations, the ‘photons’ they best couple with
equivalent of logic gates). But to truly exceed current and voltage can exist in controllable lie in the microwave range of the spectrum
the capabilities of conventional computers, the superpositions of distinct quantum states? And (frequencies of 3–30 GHz, or wavelengths of
quantum engineer must acquire extremely pre- second, can we connect these qubits together 1–10 cm). Several authors13–22 have speculated
cise control over the quantum domain, prevent in the required manner, perhaps using familiar that such microwave photons could be a route to
any unknown evolution that affects the quan- electrical means such as actual wires, but keep- connecting qubits, and recent experiments23–30
tum states (decoherence), and amass many ing in mind that any information transported have demonstrated qubit–photon couplings
thousands of qubits. Moreover, these qubits must remain in its intrinsically quantum form in superconducting circuits. This approach is
must then be ‘wired up’ in complex and pre- and exchanged as individual quanta? similar to the branch of atomic physics known
scribed arrangements, so that they can interact The answer to the first question, originally as cavity quantum electrodynamics (cavity
and communicate their quantum information posed2 to test the applicability of quantum QED), which studies the interaction of photons
664
NATURE|Vol 451|7 February 2008 HORIZONS

and atoms at the quantum level. The new field, Box 1 | Superconducting qubits
dubbed ‘circuit QED’, offers a tentative ‘yes’ to a b c
the second key question about whether we can
Superconductor
create quantum devices with interconnected
qubits.
Here we begin by discussing the physics of
cavity QED with real atoms, and then introduce Insulator = = LJ
C

the analogous circuit QED system, in which

microwave photons are coupled to supercon- Cooper pair
ducting qubits acting as artificial atoms. As Superconductor
we will see, the tight confinement of micro-
waves on a chip naturally leads to extremely
strong ‘atom–photon’ coupling, offering new d e f
possibilities for fundamental experiments on I
the light–matter interaction and interacting Φe
quantum systems. After reviewing various Cg
experiments in circuit QED that have been Vg
performed to date, we will point out some of
the outstanding issues and future directions for
this rapidly progressing field.

An atom meets a photon In a superconductor means that the energy levels interrupted by one or more
Cavity QED31–33 is the physicist’s prototype well below its transition can be anharmonic (not (often three, as here)
system for studying the interaction of light and temperature, electrons are regularly spaced) and, with Josephson junctions. A
matter at the quantum level. At its simplest, it strongly bound together the right circuit configuration, current through an external
consists of a single atom with just two relevant in ‘Cooper pairs’, enabling two low-lying states can be inductor generates a magnetic
quantum states, coupled to a single mode of the electrical signals to propagate obtained that are sufficiently flux (Φe) threading the loop,
electromagnetic field, defined for example by a with very low dissipation. separated from the others which induces clockwise or
pair of mirrors (Fig. 1). A photon in the cavity, Superconducting qubits are so that the junction can be anticlockwise circulating
bouncing back and forth between the mirrors, based on Josephson junctions, treated as a quantum two- supercurrents. This qubit
can be absorbed by the atom; conversely, if the which are made by two pieces level system, or a qubit. The couples to magnetic fields.
atom is excited, it can decay by emitting a pho- of superconductor separated typical energy separation The third type of qubit is the
ton into the cavity. The rate of this atom–light by an insulating layer thin is large enough to probe at phase qubit (f), consisting of
interaction (g) is proportional both to the dipole enough to allow tunnelling of millikelvin temperatures in a single Josephson junction
moment of the atom and to the electric field of Cooper pairs (a). A Josephson cryogenic refrigerators. connected to a current source.
the photon at the atom’s location (see Box 2, junction is usually denoted in Three main ‘flavours’ of Current (I) flowing through
overleaf). Unfortunately, in any real system, circuit diagrams as a box with superconducting qubit the junction alters the phase
other undesired processes can take place, such a cross (b). have been used, classified difference between the two
as a loss of photons from the cavity (at rate κ) A dissipation-free according to the variables sides of the junction.
supercurrent can then flow by which they are controlled All three flavours of qubit
resulting from imperfect mirrors, or the decay
through the junction, which and excited. The simplest have been used successfully,
of the atom (at rate γ) into other channels.
turns out to be equivalent qubit is the charge qubit and the ability to make and
The ‘strong coupling regime’ of cavity QED to a nonlinear inductor10. (d), or Cooper-pair box, control superpositions has
is obtained when the rate of absorption or The physical realization of which consists of an isolated been demonstrated. The
emission of a single photon by the atom is more the junction (c), with two Josephson junction placed typical lifetime of a quantum
rapid than any of the rates of loss (g >> κ,γ). In electrodes separated by an between the plates of a superposition is on the order
this case, an excited atom in an initially empty insulator, makes an LC circuit capacitor. Applying a voltage of a microsecond, allowing
cavity will emit one (and only one) photon, (a capacitor, C, and inductor, (Vg) to the capacitor (Cg) hundreds of single-qubit
which can then be trapped and reabsorbed LJ, in parallel), which is the induces a charge difference operations. Experiments
again (at rate 2g), a phenomenon known as electrical equivalent of a between the two sides of with two and three qubits
vacuum Rabi oscillations. The presence of harmonic oscillator. the junction. Alternatively, coupled to each other,
the cavity has made the spontaneous emis- The Josephson junction one can say that the qubit including the generation of
sion from the atom, usually an irreversible pro- is a very special oscillator, responds to electric fields. entangled states and the
cess, into a coherent and reversible oscillation. however, as it combines The second type is the operation of a conditional-
Entering this regime dramatically reveals the nonlinearity with low flux qubit (e), consisting NOT logic gate, have also been
quantum nature of the electromagnetic field, dissipation. The nonlinearity of a superconducting ring performed. R.J.S. & S.M.G.
allows the experimenter to make and measure
non-classical states of light, and makes experi- helps to have an atom that is a strong emitter, cavity loss kept small. A further difficulty is
ments in nonlinear optics possible at the level with a large dipole moment. To enhance the the placement and trapping of a single atom at
of a single photon. In the language of quan- coupling further, the photon’s energy should the desired location in the cavity.
tum computation, strong coupling means that be confined in the smallest cavity possible, so Despite the obvious technical challenges,
quantum information can be exchanged back that the corresponding electric fields are spread there are several examples of strong-coupling
and forth between the atom and the photon over the minimum volume. Equivalently, one cavity QED using real atoms. For optical pho-
many times before it is lost for ever. can imagine that the mirrors act to reflect the tons trapped between mirrors, the vacuum Rabi
The challenge for realizing strong-coupling photon past the atom repeatedly, giving many splitting, which indicates strong coupling, was
cavity QED is to maximize the vacuum Rabi chances for the interaction to take place. At first observed back in 1992 (ref. 34). Another
frequency (see Box 2) while simultaneously the same time, the atom should be as decou- approach35–38 uses ‘Rydberg atoms’, which are
minimizing the decay (κ,γ). Obviously, it pled as possible from other influences, and the highly excited atomic states that have very large
665
HORIZONS NATURE|Vol 451|7 February 2008

Box 2 | The fine-structure limit for cavity QED atoms, and then individually addressing and
A simple calculation 31,53 controlling them, make it difficult at present
shows that the coupling λ/2 to build large-scale quantum systems.
strength (the vacuum Rabi
frequency) of an atom and r Quantum optics on a chip
a photon in cavity quantum Circuit QED is a more recent attempt to bring
electrodynamics (QED) about strong coupling within an integrated
has an upper limit given by → superconducting circuit22. This approach
E
fundamental constants. A L offers the prospect of reaching an upper limit
photon excites the atom by for strong coupling. Josephson-junction qubits
moving one of its electrons (see Box 1) can play the role of the atom or the
into a larger orbit; the ‘dipole matter component, but how can we trap a pho-
moment’ (d = eL, with units frequency of the atom/cavity frequency in dimensionless ton on a chip? At the microwave frequencies
of charge × distance, where and V is the volume of the units: emitted by superconducting qubits, photons
e is the electron charge) is cavity. Thus, the field strength can exist in three-dimensional form as stand-
a measure of the size of the increases as the volume of ______ ___

冢 冣冪 冢 冣冪
g __
__ L e2
______ L 2α ing waves in a metallic box a few wavelengths
atom, and also determines the cavity is decreased and = = __ ___
ω r 2π ε0ᐜc r
2
π across, like those used in the Rydberg-atom
how strongly the atom the photon is more tightly
experiments. In the world of electrical circuits,
interacts with a given electric confined. However, a typical
field. The vacuum Rabi
however, photons can also be understood as
three-dimensional cavity used we find that the dimensionless
frequency is thus given by with real atoms will have a combination of the the quantized excitations of any electromag-
g = dE0/ᐜ, where E0 is the root- volume that is many cubic fundamental physical constants netic resonator, including the simple combi-
mean-square electric field at wavelengths. of electromagnetism, the fine- nation of an inductor and a capacitor50. Such
the location of the atom due In circuit QED, we can structure constant an electrical oscillator can in principle be
to vacuum fluctuations (ᐜ use a one-dimensional α = e2/4πε0ᐜc ~ 1/137, has much smaller than a wavelength in all dimen-
is the Planck constant). The transmission-line cavity, like appeared. The best situation sions, so that the ‘photons’ are confined very
vacuum fluctuations exist in the simple coaxial geometry is to arrange a cavity whose tightly indeed, and are effectively zero-dimen-
both electric and magnetic shown here, which must be transverse size is small enough sional. Another possibility is that photons are
fields, and have an amplitude half a wavelength long but that the atom completely fills confined in one dimension and travel along a
equal to that due to half a can be much smaller in the the transverse dimension transmission line, not unlike the coaxial cable
photon. A simple estimate transverse direction, and have (L/r ~ 1), and then the coupling used for TV transmission. A key realization22
of this electric field can be a volume, V = πr 2λ/2, much can be several per cent. In was that strong coupling might be achieved as
obtained by adding up the less than a cubic wavelength. comparison, because the a result of the tight transverse confinement,
density of energy (ε0E2/2) This leads to a greatly three-dimensional cavities in while still having a long ‘wire’ that can trans-
stored in the electric fields, enhanced field strength: either optical or microwave port signals from place to place.
which must be equal to cavities have bigger sizes An implementation of circuit QED using a
half the energy of a photon ______ and the real atoms used have
冪 ᐜω2
1 transmission-line resonator whose electric fields
(remembering that half of E0 = __ ______ smaller dipole moments, the
r 2π 2ε0c are coupled to a superconducting charge qubit
this energy is also stored in largest couplings possible so
is shown in Figure 2a. A central superconduct-
magnetic fields): far have much smaller g/ω,
ing wire running between two ground planes
where we have used the fact on the order of 10¬6. The large
ᐜω __ε
冮
___ ε defines the transmission line. Gaps in the wire,
= 0 E2dV = __0 E02 V that the wavelength λ = 2πc/ω interactions achievable in the
4 2 2 placed an integer number of half-wavelengths
and c is the speed of light. one-dimensional cavities
Multiplying this field strength of circuit QED make it easier (a few centimetres) apart, are the ‘mirrors’ used
where ε0 is the permittivity of by the dipole moment, we to attain the strong-coupling to form a cavity, which is the microwave version
free space, ω is the transition can express the vacuum Rabi limit. R.J.S. & S.M.G. of the Fabry–Pérot geometry used in optics. The
size and shape of the gaps controls the rate at
dipole moments and low energy transitions at ment. A spectacular recent achievement is the which photons enter and leave the cavity, and
~50 GHz. In this case, the photons are in the ability to perform quantum ‘non-demolition’ the entire structure can be made using conven-
microwave domain and the cavity consists of experiments, in which photons in a microwave tional microelectronic fabrication techniques.
a superconducting metallic box a few wave- cavity can be monitored without destroy- Such superconducting transmission lines have
lengths (several centimetres) across. Other ing them, revealing the progressive collapse been extensively studied in the past. But recent
efforts have focused on strong coupling with of the wavefunction under successive meas- experiments at temperatures close to abso-
semiconductor quantum dots as the emitters; urements44,45. Other efforts have developed lute zero, where they are used as detectors for
these have the potential advantage of emitting quantum ‘applications’, such as the creation of astrophysics51,52, have shown that photons can
at infrared wavelengths close to those used sources of non-classical light and single pho- make up to a million bounces before being lost
for telecommunication39–41. The coupling of tons on demand46, or the detection of single (the ‘cavity quality factor’, Q, is 106). This means
internal states of an atomic ion to its motion in atoms and the cooling and manipulation of that the losses are remarkably low — a gigahertz
a trap42,43 can also be understood as a realiza- their motion. photon travels back and forth a total distance of
tion of strong-coupling cavity QED, because The ability to control the interactions of 10 kilometres before being lost.
it contains the same essential ingredients of a atoms and single photons in a quantum- The qubit, an isolated Josephson junction,
two-level system (the ion) interacting with a mechanical way has intriguing implications for is placed between the wire and the ground
harmonic oscillator (the quantized motion of quantum computation and communication. planes, at or near an antinode of the standing
the ion, or phonons). Photons in a cavity47,48, or phonons in an ion wave of the voltage on the line, so it couples
Many beautiful experiments have been done trap49, can be used to generate entanglement to the electrical fields of the transmission line.
in the past two decades using these strongly and make a ‘quantum bus’ to communicate Exciting the qubit corresponds to transporting
coupled cavity QED systems, performing text- quantum information between multiple atoms. one or a few pairs of bound electrons (known
book demonstrations of fundamental quantum But the technical difficulties of achieving as Cooper pairs) from one electrode of the
phenomena such as decoherence and entangle- sufficiently strong coupling, trapping many junction to the other. This means that the
666
NATURE|Vol 451|7 February 2008 HORIZONS

dipole moment of this artificial atom is very and demonstrated a maser based on a single dispersive regime’ in circuit QED26, this inter-
large, often more than four orders of magni- artificial atom30. action, although roughly ten times smaller
tude greater than the typical value for an elec- Circuit QED has also been used for quan- than the resonant case, is still larger than all
tronic transition of a real atom. Because the tum communication and coupling between sources of decoherence, a situation that has
qubit’s size and shape are adjustable, the dipole qubits. A source of non-classical microwaves been accessed in only a few experiments with
coupling can also be engineered by having the has been demonstrated, for example, in which Rydberg atoms44,45. Circuit QED couplings can
atom essentially fill the transverse dimension single photons are produced on demand27. approach the limit where multiphoton effects,
of the cavity, which means that the vacuum This experiment also showed that the quantum which are usually rare, play an important role.
Rabi frequency (expressed as a fraction of the information contained in a superposition state Other new phenomena include optical bist-
photon frequency) approaches a maximum of a qubit could be mapped onto the photon ability of the cavity, in which the presence of
value53 of a few per cent, set by the fine-struc- state, demonstrating the conversion between a single atom makes the cavity oscillations
ture constant (see Box 2). In comparison, the a standing and a flying qubit, a milestone for strongly anharmonic, and causes the entangle-
best values obtained so far using real atoms in quantum computation. Finally, a cavity has ment of multi-photon states. It is also possible
either optical or microwave cavities are much been used to realize a solid-state quantum to engineer strong photon–photon nonlineari-
smaller, of the order of one part in 106. The bus, with a quantum state being transferred ties, based for example on the simultaneous
very large interactions achievable in circuit from one qubit to another using a microwave interaction of two cavities with a single qubit.
QED make it easier to attain the strong cou- photon as the intermediary. This last achieve- What is the real limit on the strength of cou-
pling limit of cavity QED. Another advantage ment was made simultaneously in experiments pling? It should be possible to push coupling
of circuit QED is that it avoids the difficulties of with phase qubits29 and charge qubits28. Taken strengths beyond the fine-structure limit dis-
cooling and trapping the atom, as the qubit can together, these experiments indicate that com- cussed above for electric fields. For instance,
be fabricated at precisely the desired location munication between small prototype systems if the current in a transmission line is passed
inside the cavity. of several qubits, wired together with photons directly through a Josephson junction53, the
Several experiments with superconduct- and cavities, is possible. The combination of relative coupling can be larger than unity (g > ω,
ing qubits in the past few years have accessed techniques and concepts from quantum optics, where ω is the transmission frequency of the
the regime of strong coupling, and have reca- in conjunction with the technology for super- atom/cavity), so the photon and the qubit cease
pitulated many classic results from quantum conducting quantum circuits, is likely to lead to be separate entities and the coupling can-
optics. Strong coupling with circuit QED was to continued rapid progress. not be considered as a perturbation. All these
first achieved in 2004 (refs 23, 24), and a device The combination of circuit QED and experi- investigations could add significantly to the
like that shown in Figure 2b has been used23 to mental advances with superconducting circuits body of knowledge on the light–matter inter-
observe vacuum Rabi splitting in a solid-state, raises many interesting questions, and next we action already gleaned from cavity QED.
artificial system. When transmission through shall discuss some possible themes and areas
the cavity was measured when the qubit was for future work. What are the limits of coherence?
tuned into resonance, two separate peaks (the Perhaps the greatest outstanding problem with
vacuum Rabi splitting) could be resolved (see New regimes of quantum optics all solid-state implementations of quantum
Fig. 3a, overleaf), corresponding to coherent As mentioned above, the relative coupling systems is how to minimize decoherence, the
superpositions of a single photon in the trans- strength in circuit QED is many orders of inevitable loss of quantum information owing
mission line and a single excitation of the qubit. magnitude greater than in the better-known to coupling to undesired degrees of freedom,
A more recent experiment54 with an optimized versions of cavity QED with real atoms. This and secure enough time to allow complex
qubit now approaches the fine-structure limit, means that less-familiar, higher-order effects manipulations. In their roughly 10 years of
with a dimensionless coupling strength of can have a noticeable influence. One exam- existence, the coherence time of supercon-
about 2.5%, yielding the large splitting shown ple is the dispersive, or off-resonant, case, in ducting qubits has increased by a factor of
in Figure 3b. Other experiments have observed which the qubit and the photon interact with- almost 1,000 (from just nanoseconds to a few
vacuum Rabi oscillations in the time domain25 out the photon being absorbed. In the ‘strong microseconds), but further improvements will

Out
a b
Transmission- ~ 1 cm
line cavity

Cooper-pair 10 µm
box atom

10 GHz in 10 µm

Figure 2 | Circuit QED devices. a, Schematic representation (adapted b, Micrograph of an actual circuit QED device that achieves the strong-
from ref. 22) of the circuit analogue of cavity quantum electrodynamics coupling limit. It consists of a superconducting niobium transmission line
(QED), where a superconducting qubit (green) interacts with the electric on a sapphire substrate with two qubits (green boxes) on either side. The
fields (pink) in a transmission line (blue), consisting of a central conductor inset shows one of the superconducting Cooper-pair box charge qubits
and two ground planes on either side. The cavity is defined by two gaps located at the ends of the cavity where the electric fields are maximal. The
(the mirrors) separated by about a wavelength. The cavity and qubit are qubit has two aluminium ‘islands’ connected by a small Josephson junction.
measured by sending microwave signals down the cable on one side of Changing the state of the qubit corresponds to moving a pair of electrons
the cavity and collecting the transmitted microwaves on the output side. from the bottom to top (shown schematically).

667
HORIZONS NATURE|Vol 451|7 February 2008

be necessary. It is not yet clear whether materi- be stored as photon superpositions. Because and molecules, or magnetic dipoles such as
als, circuit designs or other, unknown factors cavities can inherit a nonlinearity from cou- nuclear and electron spins, which each have
will ultimately be the limiting factor. pling with a qubit, it may be useful to ask what their own advantages and disadvantages. They
Three-dimensional dielectric 55,56 and the optimal amount of nonlinearity should will all interact with the electric or magnetic
superconducting microwave resonators57 be, and to imagine ‘photonic qubits’ in which fields of a photon if placed appropriately inside
that can store photons for about a second or energy is shared between linear and nonlinear a cavity, but some will interact more strongly,
more already exist. But for the miniaturized, elements in order to optimize coherence. and others will tend to have longer coherence
on-chip cavities used for circuit QED, dem- times. Several approaches for building ‘hybrid’
onstrated photon lifetimes are only about ten Wiring up elemental quantum objects quantum systems with both macroscopic,
times longer than those of a qubit, perhaps tens We have so far confined ourselves to discuss- artificial components and microscopic, indi-
of microseconds. Because superconducting ing the circuit QED interaction of supercon- vidual particle elements have been proposed
qubits are actually rather similar to electrical ducting qubits. There are, however, a large recently 59–61.
resonators (with the extra ingredient of non- variety of elemental quantum objects with What would make the most ideal ‘atom’ in a
linearity provided by a Josephson junction), microwave transitions, which could in prin- circuit QED or a quantum device? There are var-
making the lifetime of an on-chip, linear cavity ciple be coupled via transmission lines (Fig. 4). ious trade-offs, which can be viewed by arrang-
effectively infinite can be viewed as a necessary, Qubits made from fundamental systems such ing the systems in a rough hierarchy based on
but not sufficient, step for making truly robust as atoms or spins offer certain advantages, the ‘size’, or transition moment. In general terms,
qubits. Indeed, this quest may teach us how to including perfect reproducibility (identi- the larger the size, the higher the vacuum Rabi
make better junctions and qubits58. If cavity cal atoms have identical spectra) and longer frequency, and the more rapidly the qubits can
lifetimes continue to exceed those of qubits, coherence times, although they can be more communicate via the cavity. But the coherence
they might serve a useful role29 as a ‘quantum difficult to integrate together. These include times of these systems tend to vary inversely, and
memory’, where quantum information could electric-dipole-coupled systems such as atoms what counts is the number of operations possi-
ble, which is essentially the ratio of coupling and
decoherence rates. At one end of the spectrum
a
are Rydberg atoms and superconducting qubits,
which have micrometre-scale electric dipoles
Microwave transmission

0.08 that can match the size of a typical supercon-

ducting cavity and approach the fine-structure
0.06 limit with vacuum Rabi frequencies of hundreds
of megahertz, although these have coherence
0.04 times on the order of microseconds9 to milli-
seconds60. In the middle are polar molecules,
0.02 which are small compared with the cavity and
would have correspondingly slower coupling
0 rates, but their coherence times could be more
6.02 6.03 6.04 6.05 6.06 6.07 than 1,000 times longer. At the other extreme are
Frequency (GHz) spins, which can have lifetimes of seconds but
have coupling rates of around a hertz, which is
b probably too small to be practical. In many cases,
0.12 ensembles of particles could be used to increase
2g 2 the coupling strength, but at the expense of los-
0.10
2 1 ing nonlinearity.
Cavity transmission

0.08 2g Experimental efforts with these hybrid

1 0 systems are now under way in several labora-
0.06 tories62,63. Another approach to communicat-
0 ing quantum information around a chip is to
0.04 actually transport the qubits themselves. This
g e
is already being done for trapped ions based
0.02
on microfabricated traps62,64. As well as being
0.00 an approach to engineering quantum proces-
6.75 6.85 6.95 7.05
sors, all this work may lead to new ways to cool
and manipulate quantum objects, and perhaps
Frequency (GHz)
even to new kinds of spectroscopy and preci-
Figure 3 | Vacuum Rabi splitting. Observation of strong coupling and the fine-structure limit in sion measurements. Manipulating rotational
a circuit. a, Measurement of the microwave transmission of a cavity like that in Fig. 2b (adapted or hyperfine microwave transitions in atoms
from ref. 23). The appearance of two peaks in the transmission, as a result of vacuum Rabi splitting, and molecules can influence the electronic
indicates strong coupling. Without the qubit, a single transmission peak (dashed line) is observed. transitions at optical wavelengths, which are
With the qubit tuned to match the cavity frequency, the qubit–cavity interaction mixes together the accessible simultaneously. This might even-
photon and qubit states, and the new eigenstates of the system are coherent superpositions that are tually lead to the possibility of transferring
symmetric and antisymmetric combinations of atom and photon. The decay rates of these half-atom/ quantum information from a chip to an optical
half-photon superposition states is the average of the photon and atom decay rate, (g+κ)/2. Strong fibre. Such a quantum interconnect is a highly
coupling is observed by starting with the system in its lowest energy state (with no photons and the
desirable feature for quantum repeaters and
atom in the ground state) and measuring the presence of two peaks separated by 2g ~12 MHz about
the original cavity resonance. This splitting of the cavity resonance is akin to observing vacuum Rabi communication.
oscillations in the frequency domain. b, A more recent experimental result, showing a separation
of the vacuum Rabi peaks by about 2g/2π = 350 MHz, in which g/ω ~2.5%; the cavity decay rate is Making a complex quantum state
κ/2π ~800 kHz and the qubit decay rate is γ/2π ~200 kHz. This experiment approaches the fine- Through its ability to use photons to com-
structure limit (see Box 2) for the maximal value of an electric dipole coupling. municate between several qubits, circuit QED
668
NATURE|Vol 451|7 February 2008 HORIZONS

11. Yamamoto, T., Pashkin, Yu. A., Astafiev, O., Nakamura, Y. &
n Spin e Spin Atom/ Rydberg Charge qubit Tsai, J. S. Nature 425, 941–944 (2003).
molecule atom 12. Cirac, J. I., Zoller, P., Kimble, H. J. & Mabuchi, H. Phys. Rev.
Lett. 78, 3221–3224 (1997).
13. Shnirman, A., Schon, G. & Hermon, Z. Phys. Rev. Lett. 79,
2371–2374 (1997).
14. Makhlin, Y., Schon, G. & Shnirman, A. Rev. Mod. Phys. 73,
357–400 (2001).
15. Marquardt, F. & Bruder, C. Phys. Rev. B 63, 054514 (2001).
Magnetic moment Electric dipole moment 16. Buisson, O. & Hekking, F. in Macroscopic Quantum
Coherence and Quantum Computing (eds Averin, D. V.,
10–3 µB 1 µB 1 10 102 103 104 Debye Ruggiero, B. & Silvestrini, P.) (Kluwer, New York, 2001).
17. Al-Saidi, W. A. & Stroud, D. Phys. Rev. B 65, 014512
(2001).
18. Plastina, F. & Falci, G. Phys. Rev. B 67, 224514 (2003).
Small Size Large 19. Blais, A., Maassen van den Brink, A. & Zagoskin, A.
Phys. Rev. Lett. 90, 127901 (2003).
10–16 m 1Å 0.1 µm 1 µm 20. Yang, C.-P., Chu, S.-I. & Han, S. Phys. Rev. A 67, 042311
(2003).
21. You, J. Q. & Nori, F. Phys. Rev. B 68, 064509 (2003).
22. Blais, A., Huang, R.-S., Wallraff, A., Girvin, S. & Schoelkopf,
Slow Coupling rate Fast R. Phys. Rev. A 69, 062320 (2004).
23. Wallraff, A. et al. Nature 431, 162–167 (2004).
24. Chiorescu, I. et al. Nature 431, 159–162 (2004).
0.1 Hz 100 Hz 10 kHz 100 kHz 10 MHz 100 MHz 25. Johansson, J. et al. Phys. Rev. Lett. 96, 127006 (2006).
26. Schuster, D. I. et al. Nature 445, 515–518 (2007).
27. Houck, A. A. et al. Nature 449, 328–331 (2007).
28. Majer, J. et al. Nature 449, 443–447 (2007).
Long Coherence time Short 29. Sillanpaa, M. A., Park, J. I. & Simmonds, R. W. Nature 449,
438–442 (2007).
Hours µs 30. Astafiev, O. et al. Nature 449, 588–590 (2007).
31. Haroche, S. & Raimond, J. M. Exploring the Quantum:
Atoms, Cavities, and Photons (Oxford Univ. Press, 2006).
32. Walther, H. et al. Rep. Prog. Phys. 69, 1325–1382 (2006).
Figure 4 | Wiring up quantum systems. Besides a Cooper-pair box, many other quantum systems 33. Miller, T. E. et al. J. Phys. B 38, S551–S565 (2005).
34. Thompson, R. J., Rempe, G. & Kimble, H. J. Phys. Rev. Lett.
have microwave transitions that could be coupled to photons in a transmission line. What is 68, 1132–1135 (1992).
important for quantum computation is the number of operations that can be performed, so the 35. Raimond, J. M., Brune, M. & Haroche, S. Rev. Mod Phys. 73,
longer lifetimes of smaller particles can partly offset the weaker couplings. For a line with transverse 565–582 (2001).
dimensions on the order of a micrometre, the electrical field strengths for a single 5-GHz photon 36. Meschede, D., Walther, H. & Muller, G. Phys. Rev. Lett. 54,
are E0 ~1.4 V m−1 (see Box 2), and the corresponding vacuum Rabi coupling rate is g/2π ~10 kHz for 551–554 (1985).
37. Rempe, G., Walther, H. & Klein, N. Phys. Rev. Lett. 58,
a 1-debye dipole moment. Magnetic dipoles such as spins could also couple to the corresponding
353–356 (1987).
magnetic fields, B0 ~ 0.1 milligauss, with vacuum Rabi rates of about g/2π ~100 Hz per Bohr 38. Brune, M. et al. Phys. Rev. Lett. 76, 1800–1803 (1996).
magneton. Quantum systems can be compared according to their electric or magnetic dipole 39. Vahala, K. J. Nature 424, 839–846 (2003).
moments (or the magnitude of the emitter strength, top bars), the required transverse size of a 40. Reithmaier, J. P. et al. Nature 432, 197–200 (2004).
cavity to reach maximal coupling (second bar), their coupling rates (third bar) to a technologically 41. Yoshie, Y. et al. Nature 432, 200–203 (2004).
feasible (1 μm) cavity, and the expected lifetimes of coherent superpositions (bottom bar). 42. Leibfried, D. et al. Rev. Mod. Phys. 75, 281–324 (2003).
43. Gabrielse, G. & Dehmelt, H. Phys. Rev. Lett. 55, 67–70
Which quantum system is optimal depends on many details, including the ease of trapping or (1985).
fabricating in a cavity, and on the many factors in the qubit environment that can affect the 44. Gleyzes, S. et al. Nature 446, 297–300 (2007).
coherence times. 45. Guerlin, C. et al. Nature 448, 889–894 (2007).
46. Hijlkema, M. et al. Nature Phys. 3, 253–255 (2007).
47. Osnaghi, S. et al. Phys. Rev. Lett. 87, 037902 (2001).
may help to bring about more complex quan- which we have little experience with so far. We 48. Pellizari, T., Gardiner, S. A., Cirac, J. I. & Zoller, P. Phys. Rev.
tum systems with superconducting circuits. may find that there is a fundamental principle, Lett. 75, 3788–3791 (1995).
49. Cirac, J. I. & Zoller, P. Phys. Rev. Lett. 74, 4091–4094
The next step could be to demonstrate multi- which we haven’t discovered yet, that prevents (1995).
particle entanglement and develop simple their existence. Such a possibility might even 50. Devoret, M. H. in Quantum Fluctuations (eds Reynaud,
schemes for quantum error correction. mark the end of the road for quantum comput- S., Giacobino, E. & Zinn-Justin, J.) (Elsevier, Amsterdam,
1997).
Some of the most beautiful investiga- ing — but provide a new beginning for basic 51. Day, P. K., LeDuc, H. G., Mazin, B. A., Vayonakis, A. &
tions of quantum optics have shown that the science. ■ Zmuidzinas, J. Nature 425, 817–821 (2003).
most counterintuitive properties of quantum R. J. Schoelkopf and S. M. Girvin are in 52. Frunzio, L. et al. IEEE Trans. Appl. Supercond. 15, 860–863
(2005).
mechanics, such as entanglement, nonlocality the Departments of Applied Physics and 53. Devoret, M., Girvin, S. & Schoelkopf, R. Ann. Phys. 16,
and the measurement problem, are real. What Physics, Yale University, New Haven, 767–779 (2007).
might we learn by extending these tests to Connecticut 06520, USA. 54. Houck, A. A., Chow, J. M., Johnson, B. R. & Schoelkopf, R. J.
(unpublished data, 2007).
engineered, macroscopic systems? 55. Braginsky, V. B. & Panov, V. I. IEEE Trans. Magnetics 15,
To build even a small quantum informa- 1. Nielsen, M. A. & Chuang, I. L. Quantum Computation and 30–32 (1979).
tion device, we will need unprecedented con- Quantum Information (Cambridge Univ. Press, 2000). 56. Braginsky, V. B., Ilchenko, V. S. & Bagdassarov, Kh. S.
2. Leggett, A. J. Prog. Theor. Phys. 69 (suppl.), 80 Phys. Lett. A 120, 300–305 (1987).
trol over matter at the quantum level. Is the (1980). 57. Kuhr, S. et al. Appl. Phys. Lett. 90, 164101 (2007).
often-cited factoring of large numbers the 3. Devoret, M. H., Martinis, J. M. & Clarke, J. Phys. Rev. Lett. 58. Martinis, J. M. et al. Phys. Rev. Lett. 95, 210503 (2005).
only, or indeed the most interesting, way to 55, 1908–1911 (1985). 59. Sorensen, A., van der Wal, C. H., Childress, L. I. & Lukin,
4. Clarke, J., Cleland, A. N., Devoret, M. H., Esteve, D. & M. D. Phys. Rev. Lett. 92, 063601 (2004).
exploit such an amazing capability? It is pos- Martinis, J. M. Science 239, 992–997 (1988). 60. Hyafil, P. et al. Phys. Rev. Lett. 93, 103001 (2004).
sible that quantum computers of this sort will 5. Devoret, M. H. et al. in Quantum Tunnelling in Condensed 61. Andre, A. et al. Nature Phys. 2, 636–642 (2006).
simply prove too difficult to build. So finding Media (eds Kagan, Y. & Leggett, A. J.) (Elsevier, 62. Seidelin, S. et al. Phys. Rev. Lett. 96, 253003 (2006).
Amsterdam, 1992) 63. Nirrengarten, T. et al. Phys. Rev. Lett. 97, 200405 (2006).
short-term applications for smaller quantum 6. Turlot, E. et al. Phys. Rev. Lett. 62, 1788–1791 (1989). 64. Kielpinski, D., Monroe, C. & Wineland, D. J. Nature 417,
machines that justify the effort may be crucial 7. Bouchiat, V. et al. Phys. Scr. T76, 165–170 (1998). 709–711 (2002).
to the future of these endeavours. 8. Nakamura, Y., Pashkin, Yu. A. & Tsai, J. S. Nature 398,
786–788 (1999). Acknowledgements The authors acknowledge
A final point is that, during a large-scale continued support from the US Army Research Office,
9. Devoret, M. H. & Martinis, J. M. Quant. Inform. Process. 3,
quantum computation, the device will need 163–203 (2004). the National Security Agency, the National Science
to occupy devilishly complex quantum states, 10. Steffen, M. et al. Science 313, 1423–1425 (2006). Foundation and Yale University.

669