|Views: 37
|Likes: 1

Published by sid_senadheera

See more

See less

Precision measurement scheme using a quantum interferometer

Taesoo Kim

Department of Physics, University of Ulsan, Ulsan 680-749, Korea

Jacob Dunningham and Keith Burnett

Clarendon Laboratory, Department of Physics, University of Oxford, Oxford OX1 3PU, United Kingdom

Received 10 August 2005; published 1 November 2005

We present a scheme that makes use of quantum mechanical entanglement to signiﬁcantly enhance themeasurement precision accessible by interferometers. This is achieved by making use of a special type of interferometer composed of two “quantum beam splitters,” which each transform the input particles into amacroscopic superposition at the outputs. We show that this technique not only enables phase measurements tobe made with Heisenberg limited precision, but also overcomes the major practical problems of detectorinefﬁciencies and imperfect beam splitters. It may provide a promising route to implementing sub-shot-noiselimited measurement schemes in the laboratory.DOI:10.1103/PhysRevA.72.055801PACS number

s

: 42.50.St, 03.67.Mn, 42.50.Dv, 03.75.Gg

The ability to make increasingly precise measurements of physical quantities has long been an important challenge inphysics. One of the key developments in the ﬁeld of opticalmeasurements was the interferometer, which enabled pathlength differences to be detected through phase shifts withunprecedented accuracy. Further advances in metrology wereproposed with the discovery of nonclassical

e.g., squeezed

states of light

1–5

. These included a proposal for enhancingthe precision that could be achieved in an interferometer byusing light with reduced phase ﬂuctuations as the input

6,7

.Since then, considerable effort has been devoted to the de-tection of phase shifts below the “shot noise limit,” wherethe phase uncertainty scales as,

1/

N

and

N

is the totalnumber of particles involved

8

. Here we demonstrate apromising route for achieving this in the laboratory that isremarkably immune to the effects of imperfections in theparticle detectors. This may be of importance in a number of areas of physics including the detection of gravitationalwaves and in a range of quantum information schemes.Standard interferometers are limited by shot noise, whichresults from using a stream of uncorrelated particles to makea measurement. It has, however, been argued that by makinguse of “cooperative” effects between the particles, i.e., quan-tum mechanical entanglement, it should be possible to reachthe Heisenberg limit, where the measurement accuracy scalesinversely with

N

.In 1993, Holland and Burnett demonstrated how theHeisenberg limit could be achieved in an interferometer byusing dual Fock states at the input, i.e., when each input portof the interferometer has precisely the same number of par-ticles

9

. This paper stimulated considerable interest in sucha scheme, particularly with regard to its detailed implemen-tation. Kim

et al.

showed that the number correlated lightfrom an optical parametric oscillator

OPO

or ampliﬁer

OPA

would be a practical alternative to the dual Fock stateinput, by analyzing the effects of decorrelation and the sta-tistics of the input photon pairs

10

. In these schemes, how-ever, the phase information cannot be determined simply bymeasuring the population imbalance at the output ports, as isthe case for standard interferometry. Instead we need to mea-sure coincidences or correlations between the particles

10

.Unfortunately, such measurements are extremely sensitive toany deviation from unit detector efﬁciency, which suggeststhat these schemes are likely to be impractical

11

. Dun-ningham, Burnett, and Barnett demonstrated how this sub-stantial problem can be overcome with an atomic analogueof this scheme, which involves disentangling the atomic statebefore measurements are made on it

12

. Another promisingscheme which overcomes the problem of detector efﬁcien-cies involves measuring the collapses and revivals of thevisibility of interference fringes for Bose-Einstein conden-sates

13

. In the optical regime, recent experiments haveprovided evidence for Heisenberg-limited interferometrywith ultrastable twin beams

14

.In this paper, we wish to explore a different route toachieving sub-shot-noise limited measurements that involvescreating maximally entangled states inside the interferom-eter. These so-called NOON states

15

have the form,

=1

2

i

N

,0

+

0,

N

,

1

where

k

,

l

represents the number of particles on each of thetwo paths. The NOON state is a macroscopic superpositionof all the particles being on one path of the interferometerand all on the other. Bollinger

et al.

proposed the idea of using such maximally entangled states to make precise mea-surements of the frequency of atomic transitions

16

. Theyshowed that the resolution that could be achieved by thistechnique scales inversely with the total number of particles.Various proposals have been made for producing NOONstates in the laboratory. These include the use of Fredkingates

17–20

, quantum switching

17,21

, and coupling aquantum superposition state to a beam splitter

22

. Experi-ments have successfully created NOON states with threephotons

23

, and four

9

Be

+

ions

24

, and could in principlebe scaled up to larger numbers. A particularly promising the-oretical proposal involves the making use of ordinary beamsplitters and nonlinear unitary evolution to produce states of

PHYSICAL REVIEW A

72

, 055801

2005

1050-2947/2005/72

5

/055801

4

/$23.00 ©2005 The American Physical Society055801-1

the form of

1

25

. The latter type of evolution could beimplemented with nonlinear crystals for photons or by ex-ploiting the interactions between atoms in Bose-Einsteincondensates. We shall, in any case, show that we do not needto create perfect NOON states to realize much of the associ-ated advantage in interferometry.We shall use the general term “quantum beam splitter” todenote any scheme or device that creates a NOON state fromparticles that are not initially entangled. The special propertyof these devices is that if we feed particles into one input portand a vacuum into the other, the output is a superposition of all the particles at one port and all the particles at the other.In contrast, if we passed a stream of uncorrelated particlesthrough an ordinary 50:50 beam splitter, we would obtain abinomial distribution of particles at the outputs. The opera-tion of a quantum beam splitter can be seen to be equivalentto that of an ordinary beam splitter if all the particles weresomehow “stuck together.” Throughout this paper, we shalltreat the quantum beam splitter as a “black box” to preservethe generality of the scheme and to avoid unnecessary de-tails. However, as outlined above, it should be noted thatthere are realistic practical schemes for implementing thesedevices.The two terms of

1

will acquire different phases depend-ing on the length of the path they follow inside the interfer-ometer. Other authors have made suggestions for how thisphase could be read out, which include passing the statethrough an ordinary 50:50 beam splitter and making intensitycorrelation measurements on the outputs. Here we show thatthese read-out schemes impose prohibitively strict conditionson the detector efﬁciencies when

N

is large. We then showhow we can overcome this problem by proposing a “quan-tum interferometer” composed of two quantum beam split-ters. An analysis of this scheme shows that it should enableus to achieve Heisenberg-limited resolution for phase mea-surements and depends only weakly on the detector efﬁcien-cies. This suggests a promising route for making sub-shot-noise measurements in the laboratory.We begin by considering a Mach-Zehnder interferometerfed with a number state

N

at one input port and a vacuumstate

0

at the other

see Fig. 1

. The ﬁrst element

QBS1

inthis interferometer is a quantum beam splitter. This acts onthe input state to give the maximally entangled NOON state

1

. The path length difference between the two arms of theinterferometer gives rise to a phase factor

e

−

iN

between theterms of the superposition and the state incident on the sec-ond beam splitter is

II

=1

2

i

N

,0

+

e

−

iN

0,

N

.

2

In this scheme, the second beam splitter

BS2

is simply anordinary 50:50 one.We now come to the crucial limitation of the setup de-picted in Fig. 1, which is that one has to measure a sufﬁ-ciently high order correlation function to see the effects of the phase shift. These correlation functions have the generalform,

:

nˆ

r

:

=

:

nˆ

x

1

nˆ

x

2

¯

nˆ

x

r

:

,

3

where

nˆ

is the photon number operator,

r

is the order of thecorrelation function, and

x

i

are the positions of each mea-surement

15,26,27

.Let us suppose that we measure the

r

th order correlationfunction at one of the two outputs of BS2. If we label the twooutputs 3 and 4, then the annihilation operators correspond-ing to these modes can be written as

aˆ

3

=1

2

iaˆ

1

+

aˆ

2

,

4

aˆ

4

=1

2

aˆ

1

+

iaˆ

2

.

5

This enables us to write the correlation function for mode 3as

:

nˆ

3

r

:

=

II

aˆ

3†

r

aˆ

3

r

II

, where

II

denotes the state just before BS2. When we substitute

2

into this expression,all but four terms vanish:

aˆ

1†

r

aˆ

1

r

,

aˆ

2†

r

aˆ

2

r

,

aˆ

1†

r

aˆ

2

r

,

aˆ

2†

r

aˆ

1

r

.For the case of

r

N

, the correlation function

:

nˆ

3

r

:

is com-pletely independent of

and so there are no interferencefringes. For the case

r

=

N

, we get

:

nˆ

3

r

:

=

N

!2

N

1 −

− 1

N

/2

sin

N

for even

N

,

6

:

nˆ

3

r

:

=

N

!2

N

1 +

− 1

N

+1

/2

cos

N

for odd

N

.

7

A similar calculation holds for

:

nˆ

4

r

:

. The fact that the orderof correlation must be the same as the number of input par-ticles to see any effect of the phase shift means that we mustdetect every particle. This imposes severe constraints on thescheme. In particular, the efﬁciency,

, of the detectors mustsatisfy

1−1/

N

, which renders the scheme impractical forany more than a few particles.We would now like to consider how we can overcome thisserious impediment by considering the case of a Mach-Zehnder interferometer with two quantum beam splitters

seeFig. 2

. In this case, the output from the second quantumbeam splitter

QBS2

with input

2

is

FIG. 1.

Color online

A schematic of a Mach-Zehnder interfer-ometer with one quantum beam splitter

QBS1

and an ordinary50:50 beam splitter

BS2

. The quantum beam splitter has the prop-erty of creating a superposition of the form of

1

with all theparticles in one output port and all in the other. There is a phaseshift,

, on one arm that we wish to measure.BRIEF REPORTS PHYSICAL REVIEW A

72

, 055801

2005

055801-2

III

=12

− 1 +

e

−

iN

N

,0

+

i

1 +

e

−

iN

0,

N

.

8

The mean number of particles detected at output ports 3 and4 are now given by

III

nˆ

3

III

=

N

1 − cos

N

/2

9

and

III

nˆ

4

III

=

N

1 + cos

N

/2,

10

respectively, and the number variance at each port is

n

3

2

=

n

4

2

=

N

2

sin

2

N

/4. We should note that, foreach measurement, we detect all

N

particles at one port or all

N

at the other. It is the relative probability of these twoevents occuring that enables us to measure the phase,

. Acalculation of the square of the phase uncertainty for detec-tors with perfect efﬁciency gives

2

=

n

3

2

nˆ

3

III

2

=1

N

2

,

11

which shows that this scheme allows us to reach the Heisen-berg limit.We would now like to consider how imperfect detectorsdo not prevent us from getting close to the Heisenberg limit.Let us suppose that we measure the phase shift

with adetector with efﬁciency

, where 0

1. According to themodel of nonideal photodetection, the detected ﬁeld mode isdescribed by a photon annihilation operator,

aˆ

3

=

aˆ

3

+

1−

v

ˆ

3

, where

v

ˆ

3

is the annihilation operator for thevacuum state mode

28,29

. The number operator for thedetected photons is then given by,

nˆ

3

=

aˆ

3

†

aˆ

3

. This allows usto obtain the relations,

nˆ

3

III

=

nˆ

3

III

and

n

3

2

=

2

n

3

2

+

1−

nˆ

3

III

. Using these relations we can ob-tain an expression for the phase uncertainty,

2

=

n

3

2

nˆ

3

III

2

=1

N

2

+

1 −

1

N

3

cos

2

N

/2

.

12

The ﬁrst term on the right-hand side, which is independent of the detector efﬁciency, represents the ideal case. The secondterm accounts for detector imperfections and hence vanishesfor perfect detectors,

=1. The key point is that this secondterm scales as 1/

N

3

, for values of

N

not too close to

1+2

p

, where

p

is an integer. This means that the destructiveeffects of realistic detectors are negligible for NOON stateswith large

N

. This is a remarkable result as detector efﬁcien-cies are a major obstacle to beating the standard quantumlimit in other precision measurement schemes. Moreover, forlarge

N

, the signal-to-noise ratio

SNR

, is given bySNR=

nˆ

3

n

3

=

tan

N

/2

,

13

which means that we should obtain a clear signal when

N

lies in the interval

/2+2

p

, 3

/2+2

p

, and

p

is aninteger.An indication of the robustness of this scheme to imper-fections in the beam splitting process can be obtained byconsidering the more general case where QBS2 splits theincoming photons into a superposition of

N

−

m

and

m

0

m

N

/2

photons at the outputs

25

. In this case, theoutput state from the interferometer is given by,

III

=12

m

=0

N

/2

C

m

− 1 +

e

−

iN

N

−

m

,

m

+

i

1 +

e

−

iN

m

,

N

−

m

,

14

where

m

C

m

2

=1. Following a similar calculation to above,the square of the phase uncertainty,

02

, is given by

02

=1

N

2

+4

m

2

N

N

− 2

m

¯

sin

N

2

,

15

where

m

2

m

¯

2

−

m

¯

2

,

m

¯

m

C

m

2

m

, and

m

¯

2

m

C

m

2

m

2

.In the case that

m

¯

N

/2 and

m

¯

2

N

2

/4, Eq.

15

can bewritten as

02

1

N

2

+4

m

2

N

4

sin

2

N

.

16

This reduces to

02

1/

N

2

so long as

N

is not too closeto an integer multiple of

. We can, therefore, still expect toapproach the Heisenberg limit in this more general case.Similar results hold when both beam splitters are imperfect.This is because, if the ﬁrst quantum beam splitter is slightlyimperfect, the state inside the interferometer is very close toa NOON state and should allow measurements resolutionsclose to the Heisenberg limit. Since the output beam splitteronly marginally degrades the signal, it should still be pos-sible to obtain enhanced measurement resolution when bothbeam splitters are imperfect.

FIG. 2.

Color online

A schematic of a Mach-Zehnder quantuminterferometer composed of two quantum beam splitters

QBS1 andQBS2

. By feeding

N

particles into one input and 0 into the otherand then dectecting the particles at the two outputs

correspondingto the annihilation operators

aˆ

3

and

aˆ

4

, Heisenberg limited mea-surements can be made of the phase shift,

.BRIEF REPORTS PHYSICAL REVIEW A

72

, 055801

2005

055801-3