used acousto-optic modulators with trav- eling sound waves, taking care that the frequency shifts in both paths did not re- veal any information about the particular path taken. Our experiments are the first of a new kind of test of “quantum nonlocality.” On the basis of our results, there is no hidden communication. Of course, correlations still cry out for explanation. References 1. J.S. Bell, Speakable and Unspeakable in Quantum Me- chanics, Cambridge Univ. Press, Cambridge, 1987. 2. H. Zbinden, J. Brendel, N. Gisin and W.Tittel, Phys. Rev.A, 63, 022111 (2001);V. Scarani,W.Tittel, H. Zbinden and N. Gisin, Phys. Lett.A 276, 1 (2000). 3. A. Stefanov, H. Zbinden, N. Gisin and A. Suarez, Phys. Rev. Lett. 88, 120404 (2002). 4. A. Suarez and V. Scarani, Phys. Lett.A 232, 9 (1997). 5. G. C. Ghirardi,A. Rimini and T.Weber, Phys. Rev. D 34, 470 (1986). N. Gisin,V. Scarani,A. Stefanov,A. Suarez,W.Tittel and H. Zbinden are with the Applied Physics Group, Uni- versity of Geneva, Geneva, Switzerland. A. Suarez is also with the Center for Quantum Philosophy,Zurich, Switzerland. QUANTUM OPTICS December 2002 ■ Optics & Photonics News 51 QUANTUM OPTICS Quantum Correlations With Moving Observers N. Gisin,V. Scarani, A. Stefanov, A. Suarez, W.Tittel and H. Zbinden I t is not really an exaggeration to say that science is the study of correlations and that scientific theories explain observed correlations. Quantum correlations differ radically from classical ones in that they can violate Bell’s inequalities between spa- tially separated locations 1 : quantum corre- lations can’t be explained by theories based solely on local variables and finite- speed classical communication. The experimental data, including ob- served quantum correlations, agree to an extraordinary degree of precision with quantum theory. Quantum correlation (or “entanglement”) raises great hopes of improving our computational power, providing secure communications and deepening our understanding of what constitutes “information.” Yet “Correlations cry out for explana- tion!” 1 : how can these spatially separated locations “know” what happens else- where? Is it all predetermined at the source? Do they somehow communicate? The first assumption, that it is all pre- determined at the source, has been ruled out by numerous Bell tests. During the past three years, we have performed a se- ries of experiments to test the second as- sumption. If some sort of hidden com- munication does exist, it must propagate faster than light, and hence be defined in some privileged reference frame deter- mined either by the very condition of the experiment itself, or by cosmology. In one type of experiment, 2 we repro- duced one of our earlier long distance quantum correlation experiments, taking care to ensure that the two detection events, which took place 10 km apart, oc- curred at the same time (± 5 ps, limited by second-order chromatic dispersion) in the Geneva (i.e., laboratory) frame. We used a parametric down-conversion two- photon source, installed telecommunica- tions fibers and two similarly unbalanced interferometers in the Franson configura- tion. Since the two-photon interference fringes were clearly visible, we conclude that the hypothetical hidden communica- tion must propagate at least 10 7 times faster than the speed of light! If this were not the case, the hidden communication would not have reached the other location on time. When the same experiment was analyzed in the cosmic background radia- tion frame, a limit of 10 4 was found. In the second type of experiment, 3 we assumed 4 that the relevant reference frames were determined by the inertial frames of the devices in which the ran- dom choices took place. An intriguing consequence was that one could set the “choice-devices” in relative motion in such a way that each, in its own frame, an- alyzed its own particle first! In the context of dynamical collapse models, 5 the “choice-devices” are the detectors; in the spirit of the De Broglie-Bohm pilot-wave mode, 1 they are the beam splitters. We tested both assumptions, each time con- firming the quantum predictions. In the “moving detector” experiment, 2 we actually used moving absorbers, argu- ing that the “photon absorbed/not ab- sorbed” result could be read off the other port of the interferometer, where a pho- ton counter had been positioned in the absolute future of the absorber. In the moving beam splitter experiment, 3 we Figure 1.General scheme of a Bell experiment with moving “choice-devices.” Each photon of an entan- gled pair is sent through optical fiber to space-like separated analyzers.To test the quantum correlations, the analyzers should give binary results (“red” or “blue” lights) and possess free parameters (a, b and a ' , b'). Depending on the model, we assume the analyzer is either the detectors (a), or the beam splitters (b). We set d 1 , d 2 and L such that each “choice-device” selects its results before the other in its own reference frame (|d 1 -d 2 |<vL/c).We always observe correlations,even in this peculiar situation.