Article
Reference
Quantum Correlations with Spacelike Separated Beam Splitters in
Motion: Experimental Test of Multisimultaneity
STEFANOV, André, et al.
Abstract
Multisimultaneity is a causal model of relativistic quantum physics which assigns a real time
ordering to any set of events, much in the spirit of the pilot-wave picture. Contrary to standard
quantum mechanics, it predicts a disappearance of the correlations in a Bell-type experiment
when both analyzers are in relative motion such that each one, in its own inertial reference
frame, is first to select the output of the photons. We tested this prediction using acousto-optic
modulators as moving beam splitters and interferometers separated by 55 m. We did not
observe any disappearance of the correlations, in agreement with quantum mechanics.
STEFANOV, André, et al. Quantum Correlations with Spacelike Separated Beam Splitters in
Motion: Experimental Test of Multisimultaneity. Physical review letters, 2002, vol. 88, no. 12
DOI : 10.1103/PhysRevLett.88.120404
PMID : 11909434
Available at:
http://archive-ouverte.unige.ch/unige:36819
Disclaimer: layout of this document may differ from the published version.
1 / 1
VOLUME 88, N
UMBER 12 PHYSICAL REVIEW LETTERS 25M
ARCH 2002
Quantum Correlations with Spacelike Separated Beam Splitters in Motion: Experimental
Test of Multisimultaneity
André Stefanov, Hugo Zbinden, and Nicolas Gisin
Group of Applied Physics, University of Geneva, 1211 Geneva 4, Switzerland
Antoine Suarez
Center for Quantum Philosophy, P.O. Box 304, CH-8044 Zurich, Switzerland
(Received 22 October 2001; published 7 March 2002)
Multisimultaneity is a causal model of relativistic quantum physics which assigns a real time ordering
to any set of events, much in the spirit of the pilot-wave picture. Contrary to standard quantum mechan-
ics, it predicts a disappearance of the correlations in a Bell-type experiment when both analyzers are in
relative motion such that each one, in its own inertial reference frame, is first to select the output of the
photons. We tested this prediction using acousto-optic modulators as moving beam splitters and inter-
ferometers separated by 55 m. We did not observe any disappearance of the correlations, in agreement
with quantum mechanics.
DOI: 10.1103/PhysRevLett.88.120404 PACS numbers: 03.65.Ud, 03.65.Ta, 42.50.Ar
Many experiments have demonstrated quantum corre-
lations between spatially separated measurements, under
several conditions [1], in perfect concordance with quan-
tum mechanical predictions. The most striking feature of
quantum correlations being the violation of Bell’s inequali-
ties [2].
In this Letter we confront quantum correlations with
a natural alternative model, called multisimultaneity [3].
First, we summarize multisimultaneity, stressing its close
relation to the famous pilot-wave model of de Broglie and
Bohm [4]. Next, we oppose the predictions of quantum
mechanics and of multisimultaneity in the situation where
two entangled particles are analyzed by two beam splitters
moving apart in such a way that each beam splitter in its
own inertial reference frame analyzes his particle before
the other. We argue that multisimultaneity is the natural
application of the pilot-wave intuition to this configuration.
Finally, we present an experimental test based on two-
photon interferences.
Within Newtonian physics, where time is absolute, it is
possible to describe quantum correlations at a distance in
a causal time-ordered way [5]. One class of examples as-
sumes that the collapse of the state vector is a real physical
phenomenon [6–8]: the first measurement produces a col-
lapse, and the second measurement happens on a system
in the collapsed state.
Another explicit example, closer in spirit to the subject
of this Letter, is provided by the pilot-wave model of de
Broglie and Bohm. There the particle and the wave always
coexist, the wave guiding the particle and the particle trig-
gering the detectors. The two slit experiment is then not
more difficult to understand than the evolution of a cork
floating in a river: if an island separates the river in two
over a certain length, then the cork passes on one (and only
one) side of the island, but its subsequent evolution is also
affected by the water that passes around the other side.
When this model is applied to two entangled particles, the
model is less intuitive (the “wave flows” in configuration
space), but it still provides a well defined, “mechanistic”
description, of how quantum correlations build up: the
measurement on one side modifies the “wave” which in
turn guides the distant particle (the model is local in con-
figuration space, but nonlocal in real space). This model
reproduces all quantum predictions. Further, if one as-
sumes that a privileged reference frame (e.g., defined by
the cosmic microwave background radiation) determines
the time ordering, then this model is self-consistent. How-
ever, when time is relative, as in special relativity, it is
ambiguous. Indeed, it is then no longer defined which
measurement modifies the wave first and which particle is
then guided.
Multisimultaneity is an attempt to set the pilot-wave in-
tuition in a relativistic context. For this, one of the authors
of this Letter, Suarez, together with Scarani, proposed a
model in which a causal temporal order is defined for each
measurement [3]. The basic idea is that the relevant refer-
ence frame for each measurement is the inertial frame of
the massive apparatus. More specifically, multisimultane-
ity assumes that the relevant frame is determined by the
analyzer’s inertial frame (e.g., a polarizer or a beam split-
ter in our case). Paraphrasing Bohr, one could say that the
relevant frame, hence the relevant time ordering, depends
on the very condition of the experiment [9]. In multisimul-
taneity, as in the pilot-wave model, each particle emerging
from a beam splitter follows one (and only one) outgoing
mode, hence particles are always localized, although the
guiding wave (i.e., the usual quantum state
c) follows all
paths, in accordance with the usual Schrödinger equation.
When all beam splitters are at relative rest, this model re-
duces to the pilot-wave model and has thus precisely the
same predictions as quantum mechanics. However, when
two beam splitters move apart, then there are several (i.e.,
two) relevant reference frames, each defining a time or-
dering, hence the name of multisimultaneity. In such a
120404-1 0031-9007兾02兾 88(12)兾120404(4)$20.00 © 2002 The American Physical Society 120404-1
VOLUME 88, N
UMBER 12 PHYSICAL REVIEW LETTERS 25M
ARCH 2002
configuration it is possible to arrange the experiment in
such a way that each of the two beam splitters in its own
reference frame analyzes a particle from an entangled pair
before the other. Each particle has then to “decide” where
to go before its twin particle makes its choice (even be-
fore the twin is forced to make a choice). Multisimultane-
ity predicts that in such a
before-before configuration, the
correlations disappear, contrary to the quantum prediction.
Let us emphasize that the model of multisimultaneity,
although conceptually quite foreign both to quantum me-
chanics and to relativity, is not in contradiction with any
existing experimental data. Furthermore, it has the nice
feature that it can be tested using existing technology.
Since it would have been very difficult to put conventional
beam splitters in motion, we used traveling acoustic waves
as beam splitters to realize a before-before configuration.
It has been shown that the state of motion of the moving
acoustical wave defines the rest frame of the beam splitters
[10]. We stress that a before-before experiment using de-
tectors in motion has already been performed confirming
quantum mechanics, i.e., the correlations did not disappear
[11,12].
Energy-time entanglement can be demonstrated by two-
photon interference experiments [13]. In this experiment
(Fig. 1) we use the Franson configuration [14]. Each pho-
ton from an energy-time entangled photon pair source is
sent to an analyzer. Each one consists of an unbalanced
interferometer, the difference between the long and short
AOM
I
I
short
long
short
long
1
ω+Ω
ω
1
1
ω
2
ω −Ω
2
AOM
2
2500 m/s
2500 m/s
Laser
656 nm
Source
φ
APD
1
APD
2
&
ω
1
ω
2
L
1
L
2
55 m
C
2
C
1
∆ L
PPLN
IF
RG1000
1
2
FIG. 1. Schematic of the experiment. Each photon of one pair
created by the PPLN waveguide is coupled into an optical fiber
(L
1
and L
2
). An RG1000 filter blocks the pump laser and an
11 nm interference filter (IF) narrows the photon bandwidth. L
2
can be changed of DL by pulling on a fiber. Each photon is sent
in an interferometer (I
1
and I
2
), which uses an acousto-optic
modulator (AOM) as beam splitter. The AOMs are 55 m apart
and oriented such that the acoustic waves propagate in opposite
directions. Optical circulators C
1
and C
2
guide the photons
coming out from the interferometers into avalanche photodiodes
detectors (APD). The detection signals are sent to a coincidence
circuit. As the frequency shifts are compensated, the total energy
共v
1
1v
2
兲 when both photons take both the short arms or both
the long ones is equal. Two-photon interference fringes are
observed by scanning the phase f with a moving mirror.
arms being much longer than the coherence length of a
single photon. The coincidence events when both photons
take the short arms or both the long ones are indistinguish-
able because the emission time is undetermined, due to the
long coherence time of the pump laser. If we select only
those events, the coincidence rate between the two outputs
is proportional to
1 1 V cos共f
1
1f
2
兲 ,
where V is the visibility and f
1
and f
2
are the phase dif-
ferences between the long and the short arms in interfer-
ometers 1 and 2, respectively. Experimentally the visibility
is always lower than 1 due, e.g., to detector noise and to
partial distinguishability of the interfering paths. Never-
theless, it can be larger than 0.71, the maximal visibility
compatible with any local theory [2].
We use a recently developed periodically poled lithium
niobate (PPLN) waveguide source of energy-time entan-
gled photons [15]. It features very high efficiency, so
we can register interferences in shorter measuring times.
Violation of Bell inequality has already been demonstrated
with this source [16].
We built two unbalanced bulk Michelson interferometers
using AOMs (acousto-optic modulators, Brimrose AMF-
100-1.3-2mm) as beam splitters. An AOM can be seen
as a realization of a moving beam splitter: the traveling
acoustic wave inside the material changes the refractive
index, thus creating a traveling diffraction grating. The
reflection coefficient is maximal at the Bragg angle u
B
:
2l
s
sinu
B
苷 l兾n , (1)
where l
s
is the sound wavelength, l is the light wave-
length in vacuum, and n the refractive index of the mate-
rial. The reflection coefficient is given by R 苷 sin
2
共
p
aI 兲
[17], where I is the acoustic power and a depends on
the AOM size and material, and on the light wavelength.
Hence, the acoustic power was set, such that the beam
splitting ratio is 50兾50. The AOM ends by a skew cut
termination; thus the wave is damped rather than reflected.
Hence it is effectively traveling and not stationary. A point
which gives us confidence in using AOM [10] is that the
reflection on a moving mirror produces a frequency change
of the light, due to the Doppler effect, given by
Dn 苷
2ny sinu
c
n , (2)
where y is the mirror velocity and u the angle between
the incident light and the plane of reflection. Within an
AOM the reflected light is also frequency shifted and the
frequency shift is equal to the acoustic wave frequency
(100 MHz in our case):
Dn 苷 n
s
. (3)
Using l
s
n
s
苷 y
s
for the sound wave and Eqs. (1) and (3)
we found that the frequency shift induced by the AOM
is the same as the one induced by a mechanical grating
traveling at speed y
s
.
120404-2 120404-2
VOLUME 88, N
UMBER 12 PHYSICAL REVIEW LETTERS 25M
ARCH 2002
The frequency shift induced by the AOMs on the re-
flected beam (200 MHz because it passes twice through
the modulator) forces us to use an AOM in each interfero-
meter; otherwise, the total energy of both photons would
not be the same for the short-short and the long-long paths,
leading to distinguishability between the two paths. There-
fore both interferometers were made such that the shifts
compensate, and the total energy remains the same. More-
over, both frequencies have to be exactly identical to avoid
beatings. This requires that the frequency of both AOM rf
drivers is synchronized [18].
Because of the small deviation angle (about 5
±
) of the
interferometer arms, we collect only the light coming out
from the input port by using a fiber optical circulator (see
Fig. 1). As the deviation angle depends on the light wave-
length, an AOM acts as a bandpass filter for the reflected
beam with a bandwidth of about 30 nm. Therefore we have
to ensure that the bandwidth of the photons is smaller by
placing a spectral filter after the source. The transmission
through each interferometer is about 0.25.
When the phase
f is changed by slightly moving back
and forth one of the mirrors with a piezoelectric actuator,
we observe two-photon interference fringes with a visibil-
ity up to 97% 6 5% after substraction of the accidental
coincidences (Fig. 2). We observed that the visibility is
very sensitive to a small difference in the electric spectra
of the rf signals driving the two AOMs [18].
In order to test multisimultaneity, both interferometers
have to be far away, and both photons have to reach the
moving beam splitters at the same time. The criterion
given by special relativity for the change in time ordering
of two events in two reference frames counterpropagating
at speed y is
jDtj ,
y
c
2
d , (4)
where Dt and d are, respectively, the time difference and
distance between the two events in the laboratory frame
[3]. This criterion is much more stringent than the space-
0 50 100 150 200 250 300
100
150
200
250
300
350
400
450
Noise level
Coincidences per 10 s
Phase [arb. units]
FIG. 2. Two-photons interference fringes. According to a si-
nusoidal fit, the visibility is 97% 6 5% after substraction of the
accidental coincidences.
like separation condition jDtj ,
d
c
. Because of the high
speed of the acoustic wave (2500 m兾s, specified by the
manufacturer and computed from the mechanical proper-
ties of amorphous material transmitting IR [19]), a distance
of 55 m between the interferometers is sufficient, allowing
us to realize the experiment inside our building. The per-
mitted discrepancy on the time of arrival of the photons
at the AOMs is then, according to (4), 61.53 ps, corre-
sponding to a distance of 60.46 mm in air. The fiber path
length can be measured with a precision of 0.1 mm using
a low coherence interferometry method. The error on the
interferometers’ path lengths is measured manually with a
precision smaller than 0.5 mm. To be sure that we have
set the lengths as required we scan the path length differ-
ence by pulling on a 1 m long fiber. The scan steps are of
0.12 mm. Simultaneously we keep scanning the phase to
observe interferences.
It is not sufficient to precisely equalize the path lengths,
we also have to ensure that the coherence length of the
photons is smaller than the permitted discrepancy. This
is the case because, with the interference filter placed af-
ter the photon pair source, the photons coherence length
is about 0.14 mm. Another effect to take into account is
the chromatic dispersion in the fibers. This will spread
the photon wave packet. However, thanks to the energy
correlation, the dispersion can almost be canceled. The re-
quirement for the two-photon dispersion cancellation [20]
is that the center frequency of the two photon is equal to
the zero dispersion frequency of the fibers. We measured
this value on a 2 km fiber with a commercial apparatus
(EG&G) which uses the phase shift method. We found
a value of 1313.2 nm for l
0
. Then we used 100 m of
the same fiber, assuming that the dispersion is equally dis-
tributed. We set the laser wavelength at half this value.
The pulse spreading over 100 m, if we conservatively as-
sume a 1 nm difference between the laser wavelength and
l
0
, is 0.2 ps [11] corresponding to a length of 0.06 mm in
air, which is much smaller than the permitted discrepancy.
According to the multisimultaneity model, the correla-
tions should disappear on a range of about five scanning
steps. We scanned the path difference over a range of
63 mm (Fig. 3), around the equilibrium point. Hence the
time order of the events passes from a before-after situ-
ation, where there is a defined time ordering in all reference
frames to the ambiguous before-before condition and then
back to an after-before situation. However, several scans
like the one presented in Fig. 3 show no effects on the visi-
bility. A similar measure with both acoustic waves travel-
ing against each other (
after-after condition [3]) does not
show any change of the correlations either. We can add
that given the distance and the uncertainty on the time of
arrival, we can fix a lower bound of the speed in the labo-
ratory frame of any hypothetical quantum influence to be
4.6 3 10
5
c.
Classical correlations are correlations between events;
either the events have a common cause or one event has
120404-3 120404-3
VOLUME 88, N
UMBER 12 PHYSICAL REVIEW LETTERS 25M
ARCH 2002
-960 -720 -480 -240 0 240 480 720 960
100
200
300
400
Noise level
scanning st
ep
Coincidences per 20 s
Path Difference
∆
L [
µ
m]
FIG. 3. Two-photon interference fringes. The total scan was
done on a range of 63 mm, but for more clarity the figure shows
only a range of 61 mm around the estimated equilibrium point.
Each vertical line grid corresponds to a step of the scan. No
disappearance of the correlation is observed.
a direct influence on the other(s). That is, classical cor-
relations is a secondary concept, built upon the primary
concept of event: the cause of ordinary correlations can
be reduced to the cause of the events. As for quantum
correlations, the violation of Bell’s inequality rules out
the common cause explanation, and correlations between
spacelike separated events exclude influences propagating
slower than the speed of light.
Multisimultaneity is an alternative model to standard
quantum mechanics in which several reference frames, de-
termined by the local physical devices and apparatuses,
each define a time-ordered causality with faster than light
influences (the influences being not under human control,
they cannot be used for signaling). In all situations where
the different components of the measuring apparatuses are
at relative rest, multisimultaneity has the same prediction
as quantum mechanics. However, in the intriguing situ-
ation where entangled particles are analyzed by two beam
splitters in relative motion such that each one analyzes
“his” particle before the other, multisimultaneity predicts
that the quantum correlations disappear. Since in the re-
ported experiment the correlations did not disappear, mul-
tisimultaneity is refuted. Recall that a model assuming
that the detectors determine the relevant frames has already
been refuted [11,12].
These results stress the oddness of quantum correlations.
Not only are they independent of the distance, but also
it seems impossible to cast them in any real time order-
ing. Hence one cannot maintain any causal explanation in
which an earlier event influences a later one by arbitrarily
fast communication. In this sense, quantum correlations
are a basic (i.e., primary) concept, not a secondary con-
cept reducible to that of causality between events: Quan-
tum correlations are directly caused by the quantum state in
such a way that one event cannot be considered the “cause”
and the other the “effect.”
This work would not have been possible without the
financial support of the “Fondation Odier de psycho-
physique” and the Swiss National Science Foundation.
We thank Valerio Scarani and Wolfgang Tittel for very
stimulating discussions and François Cochet from Alcatel
Cable Suisse SA for the chromatic dispersion measure-
ment apparatus.
[1] A. Aspect, J. Dalibard, and G. Roger, Phys. Rev. Lett.
49
, 1804–1807 (1982); G. Weihs et al., Phys. Rev. Lett.
81
,
5039 (1998); W. Tittel, J. Brendel, H. Zbinden, and
N. Gisin, Phys. Rev. Lett. 81
,
3563–3566 (1998); M. A.
Rowe et al., Nature (London) 409
,
791 (2001).
[2] J. S. Bell, Physics 1
,
195–200 (1964); Speakable and Un-
speakable in Quantum Mechanics (Cambridge University
Press, Cambridge, U.K., 1987).
[3] A. Suarez and V. Scarani, Phys. Lett. A 232
,
9 (1997);
A. Suarez, Phys. Lett. A 236
, 383 (1997).
[4] D. Bohm, Phys. Rev. 85
, 166–193 (1952); D. Bohm
and B. J. Hilley, The Undivided Universe (Routledge,
New York, 1993).
[5] Ph. H. Eberhard, in Quantum Theory and Pictures of Real-
ity, edited by W. Schommers (Springer, New York, 1989),
pp. 169–215.
[6] Ph. Pearle, Phys. Rev. D 13
, 857 (1976).
[7] G. C. Ghirardi, A. Rimini, and T. Weber, Phys. Rev. D 34
,
470 (1986).
[8] I. Percival, Quantum State Diffusion (Cambridge Univer-
sity Press, Cambridge, U.K., 1998).
[9] N. Bohr, Phys. Rev. 48
, 696– 702 (1935).
[10] A. Suarez, Phys. Lett. A 269
, 293 (2000).
[11] H. Zbinden, J. Brendel, N. Gisin, and W. Tittel, Phys. Rev.
A 63
, 022111 (2001).
[12] N. Gisin, V. Scarani, W. Tittel, and H. Zbinden, Ann. Phys.
9
, 831–842 (2000).
[13] W. Tittel and G. Weihs, Quantum Information and Com-
putation 1
,
3–56 (2001).
[14] J. D. Franson, Phys. Rev. Lett. 62
, 2205– 2208 (1989).
[15] S. Tanzilli et al., Electron. Lett. 37
, 26 (2001).
[16] S. Tanzilli et al., Eur. Phys. J. D 18
, 155–160 (2002).
[17] A. Yariv, Optical Electronics (Saunders College Publish-
ing, Philadelphia, 1991), 4th ed.
[18] Full length article in preparation.
[19] Handbook of Optics, edited by M. Bass (McGraw-Hill,
New York, 1995), 2nd ed., Vol. 2
, p. 33.51.
[20] A. M. Steinberg, P. G. Kwiat, and R. Y. Chiao, Phys. Rev.
Lett. 68
, 2421 (1992); T. S. Larchuk, M. C. Teich, and
B. E. A. Saleh, Phys. Rev. A 52
, 4145– 4154 (1995).
120404-4 120404-4
