## Ajp journal

**Can wave–particle duality be based on the uncertainty relation?**
Stephan Du¨rr and Gerhard Rempea)

*Max-Planck-Institut fu¨r Quantenoptik, Hans-Kopfermann-Strasse 1, 85748 Garching, Germany*
͑Received 28 July 1999; accepted 27 January 2000͒
Wave and particle properties of a quantum object cannot be observed simultaneously. In particular,the fringe visibility in an interferometer is limited by the amount of which-way information whichcan be obtained. This limit is set by the recently discovered duality relation. So far, all derivationsof the duality relation are independent of Heisenberg’s uncertainty relation. Here we demonstratethat it is alternatively possible to derive the duality relation in the form of an uncertainty relation forsome suitably chosen observables.

*2000 American Association of Physics Teachers.*
**I. INTRODUCTION**
This result was confirmed experimentally in Refs. 21 and 22.

None of the derivations of Eqs. ͑1͒ and ͑2͒ cited above
Wave–particle duality refers to the fact that a quantum
involves any form of the uncertainty relation. It therefore
object can exhibit either wave or particle properties, depend-
seems that ‘‘the duality relation is logically independent of
ing on the experimental situation. In a double-slit experi-
the uncertainty relation.’’ 13 In this article, we will show,
ment, for example, the object must pass through both slits
however, that for arbitrary which-way schemes, Eqs. ͑1͒ and
simultaneously in order to create an interference pattern.

͑2͒ can always be derived in the form of a Heisenberg–
This testifies to the object’s wave nature. On the other hand,
Robertson uncertainty relation for some suitably chosen ob-
performing a which-way experiment reveals which of the
servables ͑which will turn out to be different from position
slits each object passes through, manifesting its particle na-
ture. However, performing a which-way experiment un-avoidably destroys the interference pattern.

This was illustrated in various gedanken experiments,

**II. PREDICTABILITY**
such as Einstein’s recoiling slit1 or Feynman’s light
In this section, we consider a two-beam interferometer
microscope.2 In order to explain the loss of interference in
without a which-way marker, as shown in Fig. 1. Let ͉ϩ͘
which-way experiments, one usually invokes Heisenberg’sposition–momentum uncertainty relation. This has been ana-
and ͉Ϫ͘ denote the state vectors corresponding to the two
lyzed in great detail by, e.g., Wiseman

*et al.*3 However,
ways along which the object can pass through the interfer-
Scully, Englert, and Walther4 pointed out that such an expla-
ometer. After passing the first beam splitter, the density ma-
nation need not always be possible, but that the entanglement
trix in a representation with respect to the basis ͕͉ϩ͘,͉Ϫ͖͘
between the which-way marker and the interfering quantum
object can always explain the loss of interference. Severalexperiments support this point of view.5–11
This entanglement need not always be perfect. In general,
a measurement performed on the which-way marker yields
only incomplete which-way knowledge. In order to quantify
ϩ and

*w*Ϫ that the object moves along
one way or the other, respectively, fulfill Tr͕͖ϭ

*w*
how much which-way information is available from such a
measurement, one typically uses the ‘‘distinguishability,’’

*D*.

1. The magnitude of the difference between these prob-
With incomplete which-way information stored, one obtains
interference fringes with a reduced visibility,

*V*, which is
limited by the so-called duality relation
which is obviously determined by the reflectivity of the first
beam splitter.

*P *quantifies how much which-way knowledgewe have. For

*P*ϭ0, corresponding to a 50:50 beam splitter,
This fundamental limit was recently discovered by Jaeger,
we have no which-way knowledge, whereas for

*P*ϭ1, we
Shimony, and Vaidman,12 and independently by Englert.13 It
know precisely which way the object takes.

can be regarded as a quantitative statement about wave–
Without loss of generality, we assume that the second
particle duality. In the special case, where full which-way
beam splitter is a 50:50 beam splitter. Taking into account
information is stored,

*D*ϭ1, it implies that the interference
the phase shift between the two interferometer arms, the
fringes are lost completely,

*V*ϭ0. The first experimental tests
upper output beam corresponds to the state vector ͉

*u*͘
of the duality relation have been performed recently.14,15
ϭ(͉ϩ͘ϩ

*ei*͉Ϫ͘)/&. The intensity in this beam is
Incomplete which-way information can alternatively be
obtained without a which-way marker by setting up the in-

*u*͉͉

*u*͘ϭ 1 1ϩ2͉
terferometer such that the particle fluxes along the two ways
differ. In this case, the which-way knowledge is expressed in
Ϯϭ͉Ϯ͉

*ei*0. The visibility of this interference pattern
the form of the so-called ‘‘predictability,’’

*P*, which is lim-
Am. J. Phys.

**68 **͑11͒, November 2000

2000 American Association of Physics Teachers

**III. DUALITY RELATION**
Let us now add a second quantum system ͑called which-
way marker͒ to the interferometer. When an object is passingthrough the interferometer, a suitable interaction shall
Fig. 1. Scheme of a typical two-beam interferometer. The incoming beam
change the quantum state of the which-way marker depend-
͑left͒ is split into two beams, denoted ͉ϩ͘ and ͉Ϫ͘. After reflection from
ing on the way the object took. This creates an entanglement
mirrors, the phase of one of the beams is shifted by . Next, the two beams
between the which-way marker and the way of the object. A
are recombined on a second beam splitter. Due to interference, the intensi-
later measurement on the which-way marker can then reveal
ties of the two outgoing beams vary as a function of the phase shift .

which way the object took. In other words, which-way infor-mation is now stored in the which-way marker. For simplic-ity, we assume that the which-way marker does not sufferfrom decoherence25 ͑at least as long as we do not couple the
where

*I*max and

*I*min denote the maximum and minimum in-
marker to a macroscopic ‘‘needle’’͒.

tensities. The relation, Eq. ͑2͒, limiting visibility and predict-
ability can easily be derived from Tr͕2͖ϭ

*w*2
tot denote the density matrix of the total system ͑ob-
ject plus which-way marker͒ after the interaction ͑but before
ϩ2͉Ϯ͉2ϭ͕1ϩ

*P*2ϩ

*V*2͖/2р1.

the phase shifter and the second beam splitter͒. Again, we
We will now show that this inequality can alternatively be
denote the pseudospin corresponding to the ways by

*x *,

*y *,
obtained in the form of a Heisenberg–Robertson uncertainty

*z *. And again, we choose the relative phase between
states ͉ϩ͘ and ͉Ϫ͘ such that ͗ϩ͉Tr ͕ ͖͉Ϫ͘
⌬

*A*⌬

*B*у 1͉͓͗

*A*,

*B*͔͉͘,
Tr

*M *denotes the trace over the which-way marker. Thus wereproduce the above results, in particular,
which applies to each pair of Hermitian operators

*A *and

*B*,with the expectation values and standard deviations of opera-
tors defined as ͗

*A*͘ϭTr͕

*A*͖ and ⌬

*A*ϭͱ͗

*A*2͘Ϫ͗

*A*͘2, re-
In order to read out the which-way information, we mea-
sure an observable

*W *of the which-way marker with eigen-
In order to find suitable operators

*A *and

*B*, we investigate
1 ,

*w *2 ,. ͖ and an orthonormal basis of eigenstates
1 , ͉

*w *2 ,. ͖ . Let

*p *( Ϯ ,

*w i*) denote the joint probability

*i *is found and that the object moves along way ͉Ϯ͘. If

*wi *is found, the best guess one can make about the way is toopt for way ͉ϩ͘ if

*p*(ϩ,

*wi*)у

*p*(Ϫ,

*wi*), and for way ͉Ϫ͘
otherwise. This yields the ‘‘likelihood for guessing the way

*w*ϩϪ

*w*Ϫ . Obviously, ͗

*z *reflects
our which-way knowledge, whereas ͗ ͘
max͕

*p*͑ϩ,

*wi *,

*p*͑Ϫ,

*wi *.

Without loss of generality, we choose the relative phase be-

*W *can vary between 1/2 and 1, it is natural to scale
this quantity by defining the ‘‘which-way knowledge’’ 26
tween states ͉ϩ͘ and ͉Ϫ͘ such that Ϯ is real, i.e., ϭ
2

*LW *1ϭ ͚

*p*͑ϩ,

*wi p*͑Ϫ,

*wi*
so that 0р

*KW *1. Obviously,

*KW *depends on the choice of
wave character and particle character of the ensemble, re-
the measured observable

*W*. In order to quantify how much
spectively. The standard deviations of these observables,
which-way information is actually stored, the arbitrariness ofthe read-out process can be eliminated by defining the
are easily obtained, because 2ϭ2ϭ2ϭ

**1**. Using the

2

*i *͚

*l jkl l *, we can now evaluate the
uncertainty relation, Eq. ͑7͒, for all possible pairs of the
which is the maximum value of

*KW *that is obtained for the
best choice of

*W*. The distinguishability is limited by the
duality relation, Eq. ͑1͒, which has been derived in Refs. 12
1Ϫ

*V*2ϭ⌬ ⌬ у͉͗ ͉͘ϭ
and 13 without using the uncertainty relation.

ͱ1Ϫ

*P*2ϭ⌬ ⌬ у͉͗ ͉͘ϭ
We will now show that the duality relation—just as Eq.

͑2͒—can alternatively be derived in the form of a
Heisenberg–Robertson uncertainty relation for some suitably
chosen observables. For that purpose, let
Equation ͑14͒ yields a trivial result, because standard devia-tions are non-negative by definition. However, Eqs. ͑12͒ and
͑13͒ are equivalent to the desired relation, Eq. ͑2͒. Hence, for
the case without a which-way marker, Eq. ͑2͒ can be derivedin the form of an uncertainty relation for the components of
denote which way to bet on if the eigenstate ͉

*w *͘
Am. J. Phys., Vol. 68, No. 11, November 2000
Second, we note that for the case without a which-way

*K *ϭ ͚ ⑀ ͑͗ ϩ͉ ͉ ϩ͘Ϫ͗ Ϫ͉ ͉ Ϫ͒͘
marker, Eq. ͑2͒ is

*equivalent *to the uncertainty relations for

*z *, Eqs. ͑12͒ and ͑13͒. This equivalence
can be read in both directions: In one direction, as discussed
above, the uncertainty relation implies Eq. ͑2͒. In the other
direction, Eq. ͑2͒ implies the uncertainty relation for these
where we used ϭ͉ϩ͗͘ϩ͉Ϫ͉Ϫ͗͘Ϫ͉
the trace over the total system. Let us define the observable
Third, we would like to draw attention to the fact that the
result. This is somewhat surprising, because from Eq. ͑10͒
we concluded that

*x *represents the wave character, whereas
⑀ ϭ

**1 **and ͓

*z *represents the particle character. Since we are investigat-

*x *,

*W *⑀͔ ϭ ͓

*y *,

*W *⑀͔
ing the limit for the simultaneous presence of wave character

*z *,

*W *⑀͔ ϭ 0. Inserting

*W *⑀ into Eq. ͑21͒, we obtain
and particle character, one might have guessed that the un-
Note that we are considering a joint observable of the total
ever, this is not the case. Instead, ⌬

*y *is employed in our
system ͑object plus which-way marker͒ here, which is
calculation. An intuitive interpretation of

*y *in terms of a
clearly necessary to explore the correlations between the
wave picture or a particle picture is not obvious.

which-way marker and the way taken by the object.

Next, we would like to mention that the observables
whose uncertainty relations we evaluate in Eqs. ͑12͒ and ͑13͒
maximized. For simplicity, we will denote the corresponding
depend on the density matrix, . In the presentation in Sec.

observable defined by Eq. ͑22͒ by

*W *͑
II, this fact is somewhat hidden in our choice of the relative
phase of states ͉ϩ͘ and ͉Ϫ͘, i.e., ϭ
becomes more obvious, if we consider arbitrary values of
0 . In this case, we can define the observables
It is easy to see that

*zW*0 is Hermitian and that (

*zW*0)2
ϭ

**1**, so that its standard deviation is

Additionally, let us consider the observable
fulfills (

*yW*0)2ϭ

**1**. As it is also Hermitian, its expectation

*x *,

*y *, and

*z *in our above presen-
tation. Obviously, these observables depend on via 0 . As
the commutation relations of the ⌺’s and ’s are the same,Eq. ͑2͒ can be derived analogously. The situation is similar
Using the commutator ͓(

*yW*0),(

*zW*0)͔ϭ2

*i*

*x *, we can
now write down the corresponding uncertainty relation. In
Finally, we will discuss whether either correlations ͑i.e.,
combination with Eqs. ͑15͒, ͑25͒, and ͑26͒, we obtain
entanglement͒ or uncertainty relations are more closely con-
nected to wave–particle duality. For that purpose, we will
investigate all the explanations for the loss of interference
This directly yields the duality relation, Eq. ͑1͒. Alterna-
fringes, referenced in Sec. I. We will sort these explanations
tively, the commutator ͓

*x *,(

*yW*0)͔ϭ2

*i*

*zW*0 can be used
into three categories, depending on whether they employ
which again yields the duality relation.

͑3͒ correlations and some uncertainty relation.

To summarize, we have demonstrated here that in an ar-
bitrary which-way scheme, the duality relation can be ex-
The textbook explanations for Einstein’s recoiling slit in Ref.

pressed in the form of a Heisenberg–Robertson uncertainty
1 and Feynman’s light microscope in Ref. 2 are based on the
relation for some suitably chosen observables.

Englert–Walther explanation4 as well as the derivations ofthe duality relation in Refs. 12 and 13 are based on the cor-

**IV. DISCUSSION**
relations. Our derivation as well as the discussion of Wise-man

*et al.*3 make use of both the correlations and some un-
The above calculation reveals a new aspect of the connec-
certainty relation. This is because these calculations involve
tion between wave–particle duality and the uncertainty rela-
the density matrix for the total system, consisting of the ob-
tion. We would like to add a few comments concerning the
ject plus the which-way marker. Consequently, the full quan-
tum correlations between these subsystems are embodied in
Let us first point out that the uncertainty relation used in
our calculation is not the position–momentum uncertainty
The above categorization reveals a crucial point: The ex-
relation. This is obvious, because, for example, the observ-
planations for the loss of interference fringes involving

*only*
ables considered here have only two eigenvalues, namely
the uncertainty relation are ͑so far͒ limited to a few special
Ϯ1, whereas position and momentum have a continuous
schemes. In other words: There are several other schemes for
which no such explanation is known, see, e.g., Refs. 4 and
Am. J. Phys., Vol. 68, No. 11, November 2000
11. In the language of Ref. 3, the loss of interference in these
12G. Jaeger, A. Shimony, and L. Vaidman, ‘‘Two interferometric comple-
schemes cannot be explained in terms of ‘‘classical momen-
mentarities,’’ Phys. Rev. A

**51**, 54–67 ͑1995͒.

tum transfer.’’ On the other hand, explanations involving
B.-G. Englert, ‘‘Fringe visibility and which-way information: An inequal-

*only *correlations apply to all which-way schemes known so
ity,’’ Phys. Rev. Lett.

**77**, 2154–2157 ͑1996͒.

far. This leads us to the conclusion that wave–particle dual-
S. Du¨rr, T. Nonn, and G. Rempe, ‘‘Fringe visibility and which-way infor-

mation in an atom interferometer,’’ Phys. Rev. Lett.

**81**, 5705–5709

ity is connected to correlations more closely than to the un-
15P. D. D. Schwindt, P. G. Kwiat, and B.-G. Englert, ‘‘Quantitative wave–
particle duality and nonerasing quantum erasure,’’ Phys. Rev. A

**60**, 4285–

a͒Electronic mail:

[email protected]
N. Bohr, ‘‘Discussion with Einstein on epistemological problems in
16W. K. Wootters and W. H. Zurek, ‘‘Complementarity in the double-slit
atomic physics,’’ in

*Albert Einstein: Philosopher-Scientist*, edited by P. A.

Schilpp ͑Library of Living Philosophers, Evanston, 1949͒, pp. 200–241,
experiment: Quantum nonseparability and a quantitative statement of
reprinted in J. A. Wheeler and W. H. Zurek,

*Quantum Theory and Mea-*
Bohr’s principle,’’ Phys. Rev. D

**19**, 473–484 ͑1979͒.

*surement *͑Princeton U.P., Princeton, 1983͒, pp. 9–49.

L. S. Bartell, ‘‘Complementarity in the double-slit experiment: On simple
2R. P. Feynman, R. B. Leighton, and M. Sands,

*The Feynman Lectures on*
realizable systems for observing intermediate particle–wave behavior,’’

*Physics *͑Addison–Wesley, Reading, MA, 1965͒, Vol. III, Chap. 1.

Phys. Rev. D

**21**, 1698–1699 ͑1980͒.

3H. M. Wiseman, F. E. Harrison, M. J. Collett, S. M. Tan, D. F. Walls, and
18R. J. Glauber, ‘‘Amplifiers, attenuators, and Schro¨dinger’s cat,’’ Ann.

R. B. Killip, ‘‘Nonlocal momentum transfer in welcher weg measure-
͑N.Y.͒ Acad. Sci.

**480**, 336–372 ͑1986͒.

ments,’’ Phys. Rev. A

**56**, 55–75 ͑1997͒.

19D. M. Greenberger and A. Yasin, ‘‘Simultaneous wave and particle
4M. O. Scully, B.-G. Englert, and H. Walther, ‘‘Quantum optical tests of
knowledge in a neutron interferometer,’’ Phys. Lett. A

**128**, 391–394

complementarity,’’ Nature ͑London͒

**351**, 111–116 ͑1991͒.

S. Haroche, ‘‘Quantum beats and time-resolved fluorescence spectros-
20L. Mandel, ‘‘Coherence and indistinguishability,’’ Opt. Lett.

**16**, 1882–

copy’’, in

*High-resolution Laser Spectroscopy*, edited by K. Shimoda,Topics in Applied Physics Vol. 13 ͑Springer, New York, 1976͒, pp. 253–
J. Summhammer, H. Rauch, and D. Tuppinger, ‘‘Stochastic and determin-
6X. Y. Zou, L. J. Wang, and L. Mandel, ‘‘Induced coherence and indistin-
istic absorption in neutron-interference experiments,’’ Phys. Rev. A

**36**,

guishability in optical interference,’’ Phys. Rev. Lett.

**67**, 318–321 ͑1991͒.

7T. J. Herzog, P. G. Kwiat, H. Weinfurter, and A. Zeilinger, ‘‘Complemen-
22P. Mittelstaedt, A. Prieur, and R. Schieder, ‘‘Unsharp particle–wave du-
tarity and the quantum eraser,’’ Phys. Rev. Lett.

**75**, 3034–3037 ͑1995͒.

ality in a photon split-beam experiment,’’ Found. Phys.

**17**, 891–903

8U. Eichmann, J. C. Bergquist, J. J. Bollinger, J. M. Gilliban, W. M. Itano,
D. J. Wineland, and M. G. Raizen, ‘‘Young’s interference experiment with
23W. Heisenberg, ‘‘U¨ber den anschaulichen Inhalt der quantentheoretischen
light scattered from two atoms,’’ Phys. Rev. Lett.

**70**, 2359–2362 ͑1993͒.

Kinematik und Mechanik,’’ Z. Phys.

**43**, 172–198 ͑1927͒; for an English

9H. Rauch, A. Zeilinger, G. Badurek, A. Wilfing, W. Bauspiess, and U.

translation see J. A. Wheeler and W. H. Zurek in Ref. 1, pp. 62–84.

Bonse, ‘‘Verification of coherent spinor rotation of fermions,’’ Phys. Lett.

24H. P. Robertson, ‘‘The uncertainty principle,’’ Phys. Rev.

**34**, 163–164

A

**54**, 425–427 ͑1975͒.

G. Badurek, H. Rauch, and D. Tuppinger, ‘‘Neutron interferometric
double-resonance experiment,’’ Phys. Rev. A

**34**, 2600–2608 ͑1986͒.

W. H. Zurek, ‘‘Decoherence and the transition from quantum to classi-
11S. Du¨rr, T. Nonn, and G. Rempe, ‘‘Origin of quantum-mechanical comple-
cal,’’ Phys. Today

**44**, 36–44 ͑October 1991͒.

mentarity probed by a ‘which-way’ experiment in an atom interferom-
26B.-G. Englert, ‘‘Remarks on some basic issues in quantum mechanics,’’ Z.

eter,’’ Nature ͑London͒

**395**, 33–37 ͑1998͒.

Naturforsch., A: Phys. Sci.

**54**, 11–32 ͑1999͒.

Am. J. Phys., Vol. 68, No. 11, November 2000

Source: http://www.quantum3000.narod.ru/papers/edu/AmJPhys68p1021_Duerr.pdf

This is to keep you informed about recent events, various issues and current developments about Drug Action and campaigns with which we have been associated. To make our interventions more meaningful, we need your active interest, time and commitment. Please write to Anant Phadke at our rational drug policy cell, 50 LIC quarters, University Road, Pune 411016 if you would like to participate in an

Tra rito celebrato e sacramento permanente Agosto 2002 Cammina. Senza sosta cammina. Va qui e poi là. Trascorre la propria vita su circa sessanta chilometri di lunghezza, trenta di larghezza. E cammina. Senza sosta. Si direbbe che il riposo gli è vietato. Quello che si sa di lui lo si deve a un libro. Se avessimo un orecchio un po' più fine, potremmo fare a meno di quel libro e rice