PHYSICAL REVIEW A 71, 042308 共2005兲

Measurement-induced continuous-variable quantum interactions

Radim Filip,1 Petr Marek,1 and Ulrik L. Andersen2
1
Department of Optics, Palacký University, 17. listopadu 50, 772 07 Olomouc, Czech Republic
2
Institut für Optik, Information und Photonik, Max-Planck Forschungsgruppe, Universität Erlangen-Nürnberg,
Staudtstrasse 7/B2, 91058 Erlangen, Germany
共Received 15 June 2004; published 4 April 2005兲

We propose feasible implementations of basic continuous-variable 共CV兲 interactions 共squeezer, parametric

amplifier, and quantum nondemolition interaction兲 between light modes without the requirement for in-line
nonlinear couplings in a strongly pumped optical medium. The method is based entirely on linear optics,
homodyne detection, and off-line squeezed ancillary states and therefore represents the CV analog of the
measurement-induced nonlinearity approach, previously used in single-photon qubit experiments to probabi-
listically implement a controlled-NOT gate.

DOI: 10.1103/PhysRevA.71.042308 PACS number共s兲: 03.67.⫺a, 42.50.⫺p

I. INTRODUCTION extreme care has been taken to spatially and temporally over-
Quantum optics is well suited for applications in quantum lap the fiber or cavity modes with the signal modes. Further-
communication and information processing. To achieve high more, by using cavities the impedance-matching condition
fidelity in these applications, feasible and efficient quantum can also hinder successful coupling. Another drawback of
interactions between light pulses are normally required. these traditional approaches is that the strongly pumped non-
However, a successful demonstration of complex quantum linear processes are normally not pure and they introduce
protocols is often hindered by the low efficiency and poor excess noise into the quadratures.
quality of coupling between signal modes in a nonlinear me- In this paper, we demonstrate that single-mode squeezers
and phase-insensitive amplifiers as well as QND interactions
dium. Recently, it has been shown that a basic two-qubit
can be accomplished by using only linear optical elements,
controlled-NOT gate between the individual photons, requir-
homodyne measurements, and off-line squeezed light beams.
ing nonlinear interaction, can be conditionally implemented In these schemes signal modes are coupled to ancillary
using only linear coupling between the input and ancillary modes in passive devices, the ancillary modes are subse-
photons 关1,2兴. A nonlinear interaction is induced by the mea- quently detected, and the measurement results are used either
surement of ancillary qubits and feedforward postselection to modulate the freely propagating output optical signals us-
procedure. Probabilistic gates of this kind yield the desired ing electro-optic feedforward or equivalently to postcorrect
operation only when a specific outcome from a detector the measured results after the protocol is terminated. Thus
event is obtained. Thus the interaction can be implemented strongly pumped nonlinear devices are not used in line but
only conditionally, for some fraction of events. only off line to prepare squeezed vacuum states in the ancil-
Beyond the experiments with individual photons, there is lary modes. Contrary to the experiments with individual pho-
an experimentally interesting alternative for quantum com- tons 关2兴, such measurement-induced CV Gaussian operations
munication and information processing—namely, quantum are deterministic without any postselection. The proposed
optical experiments with continuous variables 共CVs兲 共for a methods can overcome previous experimental problems in
review see 关3兴兲. CV quantum information processing is usu- achieving efficient quantum interaction between light beams
ally executed by a combination of active optical devices such and may pave the way for further experiments in CV quan-
as squeezers, amplifiers, and quantum nondemolition 共QND兲 tum information science.
interactions. They are based on strongly pumped nonlinear Some previous proposals have been made on the use of
processes in the optical crystals or optical fibers. Therefore linear optics and electro-optics feed forward loops to circum-
the efficiency of such quantum information protocols relies vent cumbersome nonlinear interactions in the CV regime. In
on the efficiency of the individual active devices. Unfortu- Ref. 关4兴 a scheme based on linear optics and measurement
nately, however, the efficient coupling of optical fields inside was employed to perform noiseless amplification of a single
a nonlinear medium poses severe practical problems render- quadrature and in Ref. 关5兴 the same techniques were used to
ing experimental implementation of these devices difficult. accomplish a quantum nondemolition measurement of a
To obtain sufficient gain of the active device, the power of single quadrature. A serious drawback of this scheme is that
the pump beam must be large and the effective interaction it only performs a part of the interaction on a single quadra-
length of the crystal or optical fiber must be long. Strong ture of the signal. Here we propose a method how to ap-
pump power can be obtained by using a very short pumping proach the interaction which will work completely for both
pulse and the interaction length can be enlarged by embed- complementary quadratures.
ding the crystal inside a cavity or using a long fiber. Such In Sec. II, we introduce the measurement-induced squeez-
enhancement techniques make it difficult to inject an un- ing operation which illustrates the basic concepts of our ap-
known quantum state into the system. Coupling of the quan- proach. In Sec. III we discuss how our method can be em-
tum state in the fiber and the cavity is a lossy process unless ployed in various implementations of the QND interactions.

1050-2947/2005/71共4兲/042308共6兲/$23.00 042308-1 ©2005 The American Physical Society

FILIP, MAREK, AND ANDERSEN PHYSICAL REVIEW A 71, 042308 共2005兲

pling at the CD approaches a squeezing operation of X1 by a

degree equal to the transmittance 冑T. To achieve a corre-
sponding amplification 共antisqueezing兲 of the P1 quadrature,
we perform a measurement of the PA⬘ quadrature using a
homodyne detection system and subsequently use the gained
information to finalize the appropriate squeezing operation.
This is done by amplifying photocurrent fluctuations id ⬀ p̄A
共where p̄A is the measured value兲 with an appropriate gain g
and imposing information onto the mode 1: P1⬘ → P1⬘ + gid.
This displacement operation 共it is not shown in the figure兲
can be performed either electro-optically onto the optical
modes or purely electronically onto the measurement results
after the experiment has been terminated.
These two displacement transformations are thus identical
FIG. 1. Schematic setup of the measurement-induced squeezing under the assumption that any Gaussian operation on the
operation. PBS, polarizing beam splitter; SQ, single-mode vacuum signal mode following this squeezing procedure is described
squeezer; PS, phase shifter; HOM, homodyne detection system; by linear transformations of the quadrature operators. To
CD, coupling device; HWP, half-wave plate; and LO, local complete the interaction by a postcorrection of the measured
oscillator. data we only need to explicitly know the Gaussian operations
that have been performed on the output of the operation. The
First we address two schemes implementing QND interac- data postcorrections remain deterministic and hence they do
tion with nonunity transfer gain where the interaction takes not rely on postselection, in contrast to the measurement-
place either locally or at a distance, and at the end of this induced operations with single photons 关2兴.
section we discuss an implementation of unity-gain QND. In After an appropriate postcorrection P⬘1 → P1⬘ + gid, the re-
Sec. IV we discuss the implementation of a phase-insensitive sults of any measurement on the mode 1 are produced by the
optical parametric amplifier. Finally, in Sec. V, the results are output quadrature operators
summarized with a discussion of the experimental viability. X1⬘ = 冑TX1 + 冑1 − TXA ,
II. SINGLE-MODE SQUEEZER
1 冑共1 − T兲共1 − ␩兲
First we discuss a basic ingredient of all the schemes pro- P⬘1 = P1 − P0 , 共2兲
posed below—a measurement-induced single-mode phase- 冑T 冑T ␩
sensitive optical parametric amplifier or equivalently a where P0 is the quadrature operator of a mode exhibiting
single-mode squeezer. The optimal squeezing operation re- vacuum fluctuations. The overall efficiency of this homodyne
duces the mean value and variance of a certain quadrature detector is given by ␩. We see that the success of the squeez-
variable of an incident optical mode at the cost of minimal ing operation of the input signal is limited by the squeezing
increase of the mean value and variance of the complemen- of the ancillary state as well as the nonunity detector effi-
tary quadrature. Previously, such an operation has been ciency of the homodyne detection system. More precisely,
implemented by injecting a signal into a nonlinear medium the uncertainty of the squeezed quadrature is enhanced by
that is intensively pumped 关6兴. 共1 − T兲VA due to finite squeezing of the ancillary state, while
Our alternative—the measurement-induced squeezing the uncertainty of the antisqueezed quadrature is enlarged by
scheme—is depicted in Fig. 1. We assume that the signal is 关共1 − T兲共1 − ␩兲 / T␩兴V0 due to imperfect detectors. Here VA
carried by a linear polarization mode 1 while the squeezed
and V0 are the variances of XA and P0, respectively. The
ancillary mode A is prepared in the orthogonal polarization
mean values of the input signal are, however, transformed
mode. To controllably couple the signal mode and squeezed
perfectly since the ancillary state is assumed to be vacuum
vacuum ancillary mode, a coupling device, CD, comprising a
squeezed. For a strongly squeezed input ancilla and high
half-wave plate sandwiched by two polarizing beam splitters,
detection efficiency we approach the ideal squeezing opera-
tion: X1 = 冑TX1, P1 = 共1 / 冑T兲P1. In this case the squeezing op-
is used. Such a device effectively controls the linear coupling
X1⬘ = 冑TX1 + 冑1 − TXA, P1⬘ = 冑TP1 + 冑1 − TPA , eration is optimal and the degrees of squeezing and anti-
squeezing are uniquely controlled only by the beam splitter
XA⬘ = 冑TXA − 冑1 − TX1, PA⬘ = 冑TPA − 冑1 − TP1 共1兲 transmittance.
We can easily control the quadrature being squeezed by
between the two input polarization modes with a beam split- simple adjustment of the relative phase between the ancillary
ting ratio given by the angle ⌰ of the HWP, where the trans- and signal modes 共by using phase shifter PS1兲. The measured
mittance T of the CD is T = cos2 ⌰ and Xi , Pi 共Xi⬘ , Pi⬘兲, i quadrature can be controlled by the phase shifter PS2 which
= 1 , A, are input 共output兲 pairs of quadrature operators asso- introduces a phase shift between the local oscillator and sig-
ciated with the light modes. nal mode 1. Therefore, an appropriate adjustment of the
If the ancillary mode A is prepared in a vacuum state that phase shifters PS1 and PS2 enables the squeezing operation
is strongly squeezed in the XA quadrature, the passive cou- to be performed on any quadrature of the signal state.

042308-2
MEASUREMENT-INDUCED CONTINUOUS-VARIABLE… PHYSICAL REVIEW A 71, 042308 共2005兲

III. QND INTERACTIONS

An optimal CV QND interaction with unit transfer gain is

a coupling between two optical modes 1 and 2 described by
X⬘1 = X1, P⬘2 = P2, X⬘2 = X2 + GX1, P1⬘ = P1 − GP2, where
X1 , P1 , X2 , P2 are the complementary quadrature operators
关7兴. After the interaction, information carried by a single
nondemolition variable X1 共P2兲 is transferred to the output
variable X2⬘ 共P1⬘兲 of the other mode. The noise inevitably
associated with such an interaction is transferred to the
complementary output variable P1⬘ 共X2⬘兲 keeping the non-
demolition variable noise-free. Previously, the complete
QND interaction was generated by coupling two optical FIG. 2. Schematic setup of the measurement-induced QND in-
modes in a nonlinear optical crystal 关8兴. We now outline teraction with nonunity transfer gains. The abbreviations are ex-
plained in the caption text of Fig. 1.
three different methods for the complete QND interaction
without the need for efficient coupling inside a nonlinear
medium. 1 冑共1 − T1兲共1 − T2兲
P1⬘ = P1 − P2 ,
冑T 1 冑T 1 T 2
A. Local QND interaction with nonunity transfer gain

X2⬘ = 冑T2X2 + 冑共1 − T1兲共1 − T2兲X1 .

A measurement-induced implementation of the QND in-
共4兲
teraction requiring only a single source of squeezing is de-
picted in Fig. 2. An ancillary linear polarization mode A is
In this limit, the QND variable X1 is not influenced by noise
mixed with two orthogonally polarized modes 1 and 2 in the
and information is perfectly transmitted to the variable X⬘2,
two coupling devices CD1 and CD2 with the transmission
but at the cost of the inevitable addition of noise in P⬘1. Note
coefficients T1 and T2, respectively. In the next step, the
that if the quadrature P2 is taken to serve as the nondemoli-
quadrature PA⬘ of the ancillary mode is measured by a homo-
tion variable, quantum optical tapping 关7,9兴 between P1 and
dyne detector with the efficiency ␩. Consequently, we obtain
P2 can be achieved without any squeezing. Both the P1 and
a photocurrent id which is proportional to the random mea-
P2 variables are amplified at the outputs and the QND gain
G = 冑共1 − T1兲共1 − T2兲 / 冑T1T2 can be arbitrarily high. On the
surement outcomes p̄A. According to these measurement out-
comes, the modes 1 and 2 can be displaced P⬘1 → P1⬘ + g1id,
other hand, taking X1 as the nondemolition variable, we have
P2⬘ → P⬘2 + g2id in the data postcorrection procedure, where g1
quantum optical tapping between X1 and X2 without the need
and g2 are suitable scaling factors. Then the scheme in Fig. 2
of a measurement. Here, however, the two outputs are at-
tenuated by the factor 冑T2. Any quadrature of the input
effectively implements the following transformation:

X⬘1 = 冑T1X1 + 冑1 − T1XA , modes can be chosen to be the nondemolition one by adjust-
ing the phase shifter PS1 with respect to the phase of the
1 冑共1 − T1兲共1 − T2兲 冑共1 − ␩兲共1 − T1兲 input modes and adjusting the phase of the local oscillator
P1⬘ = P1 − P2 − LO in the homodyne measurement accordingly 共using the
冑T1 冑T1T2 冑␩ T 1 T 2 P 0 , phase shifter PS2兲. We can easily prove that the outlined
QND scheme described by Eqs. 共4兲 approaches the optimal
X⬘2 = 冑T2X2 + 冑共1 − T1兲共1 − T2兲X1 − 冑T1共1 − T2兲XA , QND transformation with unit transfer gain by implementing
squeezing operations with gains T1 and T2 onto the two out-
1 冑共1 − ␩兲共1 − T2兲 put modes:
P2⬘ = P2 − P0 , 共3兲
冑T 2 冑␩ T 2 X1⬘ = X1, P⬘2 = PB ,
where P0 is the quadrature operator of a mode prepared in
vacuum state. The QND interaction 共3兲 transfers information 冑共1 − T1兲共1 − T2兲
about the quadrature X1 to the quadrature X2⬘ and similarly P1⬘ = P1 − P2 ,
about P2 to the P1⬘ quadrature. The efficiency of the former
冑T 2
information transfer is limited by the squeezing in the ancil-
lary mode A, whereas the efficiency of the latter information 冑共1 − T1兲共1 − T2兲
transfer is degraded only by the detector efficiency ␩. X2⬘ = X2 + X1 . 共5兲
Assuming that the ancillary mode A is prepared in a
冑T 2
strongly squeezed vacuum state and the detection efficiency
␩ is almost unity, we can approach the perfect nonunity gain This transformation is the optimal QND interaction with
QND interaction: unity transfer gain and the QND interaction gain is given by
G = 冑共1 − T1兲共1 − T2兲 / T2. Let us stress that the squeezing op-
X1⬘ = 冑T1X1,
1 erations can be implemented using the scheme discussed in
P2⬘ =
冑T 2 P 2 , Sec. II 共see also Fig. 1兲.

042308-3
FILIP, MAREK, AND ANDERSEN PHYSICAL REVIEW A 71, 042308 共2005兲

FIG. 3. Schematic setup for the measurement-induced QND in-

teraction at a distance with nonunity transfer gain. The abbrevia-
tions are explained in the caption of Fig. 1.
FIG. 4. Schematic setup for measurement-induced QND inter-
action with unity transfer gain, and the measurement-induced
B. Distant QND interaction with nonunity transfer gain phase-insensitive amplifier. The abbreviations are explained in the
The QND scheme can be modified using a CV entangled caption of Fig. 1.
state shared between two distant parties 共Alice and Bob兲.
This can be useful when two distant parties want to perform vacuum modes. As the two-mode squeezing increases
a QND coupling using only local linear operations and clas- 共meaning that the uncertainties of the joint operators XA
sical communication. The proposed scheme is depicted in − XB and PA + PB decrease兲 and the efficiency ␩ of the homo-
Fig. 3. The entanglement is carried by a two-mode squeezed dyne detectors approaches unity, the QND interaction is
vacuum state of which one half 共denoted mode A兲 is pre- achieved at a distance. However, the QND interaction is per-
served by Alice and the other half 共denoted mode B兲 is sent formed with a nonunity transfer gain. The nondemolition
to Bob via a quantum channel. The two-mode squeezed variables X1 and P2 are amplified whereas the conjugate
vacuum state can be prepared by mixing two orthogonally quadratures P1 and X2 are attenuated. As in the previous
squeezed states at a balanced beam splitter. Alice mixes the section the optimality of the QND interaction can be proved
modes 1 and A in the coupling device CDA with transmit- by implementing local single-mode squeezing operations in
tance T1, and similarly Bob couples the modes 2 and B with Alice’s and Bob’s modes. Then we obtain the optimal QND
the coupling device CDB with transmittance T2. Then Alice coupling with unity transfer gain between two distant parties
measures the quadrature XA⬘ and she obtains a photocurrent ix with the interaction gain G = 冑共1 − T1兲共1 − T2兲 / T1T2.
proportional to the result x̄A while Bob measures the conju-
gate quadrature PB⬘ which yields the photocurrent i p ⬀ p̄ P. Fi- C. QND interaction with unity transfer gain
nally, they mutually exchange these classical data by two-
way classical communication and thus they can achieve the The implementation of the measurement-induced QND
transformations coupling depicted in Fig. 2 utilized only a single squeezing
source to accomplish the basic features of a QND coupling in
1 冑共1 − ␩兲共1 − T1兲 both complementary variables. However, as already pointed
X1⬘ = X1 − X0 ,
冑T 1 冑␩ T 1 out, the nondemolition quadratures are not transferred with
unity gain. Here we show that by adding an extra squeezed
冑共1 − T1兲共1 − T2兲 冑共1 − ␩兲共1 − T1兲 vacuum ancillary mode to the scheme we can approach the
P1⬘ = 冑T1 P1 − P2 + P0 optimal QND interaction with unity transfer gain. The
冑T 2 冑␩ T 2 scheme is shown in Fig. 4. The two input signals are carried
+ 冑1 − T1共PA + PB兲,
by two orthogonally polarized modes denoted 1 and 2. These
modes are mixed by a half-wave plate HWP1 aligned at an
angle ␽ with respect to the optical axis. The degree of mix-
冑共1 − T1兲共1 − T2兲 冑共1 − ␩兲共1 − T2兲
X⬘2 = 冑T2X2 +
ing can be quantified by the transmission coefficient T1
X1 − X0
冑T 1 冑␩ T 1 = sin ␽. Subsequently, the input signals are coupled on a
polarization-insensitive beam splitter 共BS with a transmis-
− 冑1 − T2共XA − XB兲, sion coefficient of T兲 with two ancillary orthogonally polar-
ized modes denoted A and B. The two ancillary modes are
1 冑共1 − ␩兲共1 − T2兲 assumed to be squeezed in conjugate quadratures PA and XB,
P2⬘ = P2 − P0 , 共6兲
冑T 2 冑␩ T 2 respectively. After the linear coupling, the ancillary modes
are decoupled by a polarizing beam splitter 共PBS兲 and the
if the local postcorrections X⬘i → X⬘i + gXiix and Pi⬘ → P⬘i amplitude quadrature XA⬘ of mode A is measured 共with the
+ g Pii p, where gXi , g Pi 共i = 1 , 2兲 are appropriate scaling fac- result x̄A兲 while the phase quadrature PB⬘ is measured on
tors, are used on the signal data after termination of the pro- mode B 共with the result p̄B兲 by using two homodyne detec-
tocol. The quadratures X0 and P0 describe noncommuting tors with identical efficiencies. The measurements give rise

042308-4
MEASUREMENT-INDUCED CONTINUOUS-VARIABLE… PHYSICAL REVIEW A 71, 042308 共2005兲

to the photocurrents iX ⬀ x̄A and i P ⬀ p̄B which will be used in appropriate data postcorrections X⬘i → X⬘i + gXiiX, P⬘i → P⬘i
the postcorrection procedure. The signal outputs modes 1 + g Pii P, i = 1 , 2, we have implemented the following transfor-
and 2 are once more coupled by a half-wave plate 共HWP2兲 mation:
which is aligned at an angle ␽⬘ with respect to the optical
冑␭ 冑1 − T
axis corresponding to a transmission coefficient of T2 X1⬘ = 冑GX1 − 冑G − 1X2 + X0 +
= sin ␽⬘. Adjusting the coupling and transmission coefficients 冑2 冑2 X A ,
such that
冑␭ 冑1 − T
P⬘1 = 冑GP1 + 冑G − 1P2 +
T1 = 1/共1 + T兲, T2 = T/共1 + T兲, 共7兲
P0 +
冑2 冑2 PB ,
we have generated the following transformation:

X1⬘ = X1 − 冑␣X0 − 冑␤XA , 冑␭ 冑1 − T

X2⬘ = 冑GX2 − 冑G − 1X1 − X0 +
冑2 冑2 X A ,
P⬘2 = P2 − 冑␣ P0 + 冑␤ PA ,
冑␭ 冑1 − T
P⬘1 = P1 − 冉冑 冑 冊
1
T
− T P2 + 冑T␤ PA + 冑␣/TP0 ,
P⬘2 = 冑GP2 + 冑G − 1P1 −
冑2 P0 +
冑2 PB , 共10兲

where G = 41 共1 / 冑T + 冑T兲2 is the gain of the amplifier, ␭ = 共1

X2⬘ = X2 + 冉冑 冑 冊
1
T
− T X1 + 冑T␤XA − 冑␣/TX0 , 共8兲
− T兲共1 − ␩兲 / T␩, and X0 , P0 are quadrature operators of the
commuting modes prepared in vacuum state. The perfor-
mance of this device is limited, as in the previous sections,
if the appropriate postcorrections X⬘i → X⬘i + gXiiX, P⬘i → P⬘i by the detection efficiency ␩ of the homodyne detectors as
+ g Pii P, i = 1 , 2, are applied to the signal data. Here ␣ = 共1 well as by the finite squeezing of the ancillary modes. As-
− T兲共1 − ␩兲 / 共1 + T兲␩, ␤ = 共1 − T兲 / 共1 + T兲, and X0 , P0 are suming almost perfect homodyne detectors and high degrees
quadrature operators of the commuting modes exhibiting of squeezing in the ancillary modes we approach the phase-
vacuum noise. As the squeezing of the ancillary modes in- insensitive amplifier transformation
⬘ = 冑GX1,2 − 冑G − 1X2,1 ,
creases and the efficiency of the homodyne detectors ap-
proaches unity, the QND interaction with unit transfer gain, X1,2

X⬘1 = X1, P⬘2 = P2 , ⬘ = 冑GP1,2 + 冑G − 1P2,1

P1,2 共11兲

P1⬘ = P1 − 冉冑 冊
1
T
− 冑T P2 ,
with an arbitrary accuracy.

冉冑 冑 冊
V. DISCUSSION
1
X2⬘ = X2 + − T X1 , 共9兲 In this paper we have proposed schemes for
T measurement-induced parametric amplifiers and quantum
is generated with an arbitrary accuracy. Through control of nondemolition measurements without the necessity of a cum-
the transmission coefficient T, the gain G = 1 / 冑T − 冑T of the bersome interaction between signal and pump modes in an
optimal QND interaction can be freely adjusted. active medium. They avoid the difficulties mentioned in the
Introduction. Linear coupling of two modes in a linear de-
IV. PHASE-INSENSITIVE OPTICAL vice 共like a beam splitter兲 is relatively easy and can be per-
PARAMETRIC AMPLIFIER formed very efficiently by carefully overlapping the modes.
Therefore the proposed schemes based on linear optics and
A phase-insensitive optical parametric amplifier based on measurements represent a remarkable practical simplification
off-line squeezers, linear optics, and measurements can also of the traditional two-mode nonlinear interaction schemes. In
be implemented. The phase-insensitive amplifier optimally addition to the practical simplifications, the measurement-
amplifies any quadrature of the signal with the same gain at induced interaction schemes have some distinct features.
the cost of the addition of excess noise. To date, parametric First, they are deterministic schemes that employ every input
amplification has been achieved by merging signal and idler state. This is in contrast to the measurement-induced nonlin-
modes with a strong pump inside a nonlinear optical crystal earities on single-photon level where events are probabilisti-
关10兴. Our implementation of the optical parametric amplifier cally postselected which renders these protocols rather inef-
without the necessity to couple the signal modes within a ficient 关2兴. In our approaches the states are not postselected
nonlinear medium is similar to the setup depicted in Fig. 4. but only deterministically postcorrected 共displaced兲 accord-
Now, however, the wave plates HWP1 and HWP2 must be ing to data from the measurement on some ancillary modes.
oriented such that a balanced coupling between the polariza- However, similar to experiments with qubits, we can perform
tion modes is achieved, i.e., sin ␽ = sin ␽⬘ = 1 / 冑2. Then by this displacement simply on the measured data after the ex-
controlling the transmittance T of the beam splitter and after ecution of the optical experiment. Finally, the various trans-

042308-5
FILIP, MAREK, AND ANDERSEN PHYSICAL REVIEW A 71, 042308 共2005兲

formations presented above are independent of the anti- introduces noise in the correction procedure. A first genera-
squeezed quadrature of the ancillary states. This means that tion of such detectors was used in previous feedforward ex-
the eventual excess noise in the antisqueezed quadrature of periments 关4,5兴 and the development of the next generation
the ancilla is not transferred to the output signal state. Hence, of detectors with low electronic noise is one recent experi-
a minimum uncertainty operation can be approached even mental task.
with a nonminimum uncertainty squeezed ancillary state. Note added in proof. Similar methods are used in the re-
The performance of our schemes is ultimately hampered cent experimental demonstration of quantum cloning of light
by imperfections. For a successful demonstration, strongly to accomplish a phase insensitive interaction 关13兴.
squeezed ancillary modes and good homodyne detectors are
required. Strongly squeezed states can be prepared using ACKNOWLEDGMENTS
present technology for a wide range of the frequencies based
on either optical parametric oscillation 关11兴 or the optical We would like to thank Ladislav Mišta, Jr., and J.
Kerr effect in fibers 关12兴 in both the continuous-wave regime Fiurášek for stimulating and helpful discussions. R.F. ac-
and the pulse regime. In both cases a suppression to −6 dB is knowledges the hospitality of Professor G. Leuchs from Uni-
reliably achievable. A squeezing near −10 dB could be versität Erlangen-Nürnberg. The work was supported by
achieved once all the components and material are opti- Project No. 202/03/D239 of The Grant Agency of the Czech
mized. Efficient implementation of the feedforward loops Republic, the Projects No. LN00A015, No. CEZ:J14/98, and
can be attained by employing high-quantum-efficiency pho- No. MSM6198959213 and “Measurement and Information
todiodes in the detectors and simultaneously using a circuitry in Optics” of the Ministry of Education of Czech Republic,
with low electronic dark noise. Electronic noise is taken by and by the COVAQUIAL Grant No. FP6-511004 of the sixth
the detector as classical noise of the ancillary modes, which framework program of the European Union. U.L.A acknowl-
is uncorrelated with the output signal modes and therefore edges the Alexander von Humboldt Foundation for support.

关1兴 E. Knill et al., Nature 共London兲 409, 46 共2001兲; M. Koashi et 关8兴 S. F. Pereira et al., Phys. Rev. Lett. 72, 214 共1994兲; K.
al., Phys. Rev. A 63, 030301 共2001兲; T. B. Pittman et al., ibid. Bencheikh et al., ibid. 75, 3422 共1995兲.
64, 062311 共2001兲; 68, 032316 共2003兲. 关9兴 E. Goobar et al., Phys. Rev. Lett. 71, 2002 共1993兲; K.
关2兴 T. B. Pittman et al., Phys. Rev. Lett. 88, 257902 共2002兲; S. Bencheikh et al., ibid. 78, 34 共1997兲.
Gasparoni et al., ibid. 93, 020504 共2004兲; G. J. Pryde et al., 关10兴 M. D. Levenson et al., Phys. Rev. Lett. 70, 267 共1993兲; Z. Y.
ibid. 92, 190402 共2004兲; J. L. O’Brien et al., ibid. 93, 080502 Ou et al., ibid. 70, 3239 共1993兲; R. Bruckmeier et al., ibid. 78,
共2004兲. 1243 共1997兲.
关3兴 For example, in Quantum Information Theory with Continuous 关11兴 L.-A. Wu et al., Phys. Rev. Lett. 57, 2520 共1986兲; E. S. Polzik
Variables, edited by Samuel L. Braunstein and Arun K. Pati et al., Appl. Phys. B: Photophys. Laser Chem. 55, 279 共1992兲;
共Kluwer Academic, Dordrecht, 2002兲. C. Kim and P. Kumar, Phys. Rev. Lett. 73, 1605 共1994兲; G.
关4兴 P. K. Lam et al., Phys. Rev. Lett. 79, 1471 共1997兲; E. H. Breitenbach et al., J. Opt. Soc. Am. B 12, 2304 共1995兲; P. K.
Huntington et al., Opt. Lett. 23, 7 共1998兲; B. C. Buchler et al., Lam et al., J. Opt. B: Quantum Semiclassical Opt. 1, 469
Phys. Rev. A 60, 4943 共1999兲. 共1999兲; J. Laurat et al., e-print quant-ph/0410081.
关5兴 R. Bruckmeier et al., Phys. Rev. Lett. 79, 43 共1997兲; B. 关12兴 M. D. Levenson et al., Opt. Lett. 10, 514 共1985兲; K. Bergman
Buchler et al., Phys. Rev. A 60, 4943 共1999兲; 65, 011803 and H. A. Haus, ibid. 16, 663 共1991兲; H. Haus, J. Opt. Soc.
共2001兲; U. L. Andersen et al., J. Opt. B: Quantum Semiclassi- Am. B 12, 2019 共1995兲; S. Schmidt et al., Phys. Rev. Lett. 81,
cal Opt. 4, S229 共2002兲. 2446 共1998兲; D. Krylov and K. Bergman, Opt. Lett. 23, 1390
关6兴 J. A. Levenson et al., J. Opt. Soc. Am. B 10, 2233 共1993兲; 共1998兲; Ch. Silberhorn et al., Phys. Rev. Lett. 86, 4267 共2001兲;
Phys. Rev. Lett. 70, 267 共1993兲; J. P. Poizat and P. Grangier, J. Heersink et al., e-print quant-ph/0410142.
ibid. 70, 271 共1993兲. 关13兴 U. L. Andersen, V. Josse, and G. Leuchs, e-print quant-ph/
关7兴 P. Grangier et al., Nature 共London兲 396, 537 共1998兲. 0501005.

042308-6