Ann. Phys. Fr.
1 (2008)
c EDP Sciences, Les Ulis
arXiv:0710.2597v1 [quant-ph] 13 Oct 2007
Wheelers delayed-choice thought experiment: Experimental realization and theoretical analysis
V. Jacques 1 , E Wu 1, 2 , F. Grosshans 1 , F. Treussart 1 , A. Aspect 3 , Ph. Grangier 3 and J.-F. Roch 1
Laboratoire de Photonique Quantique et Molculaire, UMR CNRS 8537, Ecole e Normale Suprieure de Cachan, France e 2 Key Laboratory of Optical and Magnetic Resonance Spectroscopy, East Chinan Normal University, Shanghai, Chine 3 Laboratoire Charles Fabry de lInstitut dOptique, UMR CNRS 8501, Palaiseau, France
1
Abstract
Wheeler has strikingly illustrated the wave-particle duality by the delayedchoice thought experiment, in which the conguration of a 2-path interferometer is chosen after a single-photon light-pulsed has entered it. We present a quantitative theoretical analysis of an experimental realization of Wheelers proposal.
INTRODUCTION
Wheeler has strikingly illustrated the wave-particle duality when he proposed the delayed-choice thought-experiment, where the output beamsplitter of a MachZehnder interferometer (MZI) is inserted or removed after a single-photon lightpulse has entered it [1]. Since the wave-like or particle-like behavior of the lightpulse can be shown with these complementary experimental setups, this waveparticle duality seems consistent with a light choosing to be a wave or a particle, according to the experimental setup. Wheelers delayed-choice experiment challenges such a naive interpretation by delaying the experimental setup choice, so that the light-pulse can only learn this choice after having entered the setup. 2 EXPERIMENTAL REALIZATION
We recently realized almost ideally this thought experiment [2]. The light pulses are true single-photons, emitted by a NV color center in a diamond nanocrystal [3]. The photon number of the pulse is checked by observing an anticorrelation parameter = 12% < 1 on the output ports of a beamsplitter [4].
ANNALES DE PHYSIQUE
Figure 1. Left: sketch of the experimental setup. Right: possible theories compatible with the parameters V and . corresponds to the experimental results.
As shown in gure 1, we have built a Jamin polarization interferometer, followed by an electro-optical modulator (EOM) and a Wollaston prism (WP). This interferometer is equivalent to a MZI where the output beamsplitter reection coecient can be tuned from 1/2 (closed conguration) to 0 (open conguration) in 40 ns. The two paths of this interferometer are spatially separated by 5 mm and propagate over 48 m, which corresponds to a propagation time of 160 ns. The conguration choice is performed by a quantum random number generator (QRNG) placed at the interferometer output. The timing ensures that any information about the conguration choice should travel at 4 times the speed of light to inuence the photon behavior at the entrance [2]. We have observed interference with a visibility V = 94% as well as an anticorrelation parameter = 12%, both in the delayed-choice mode, in full agreement with quantum mechanics predictions. 3 THEORETICAL ANALYSIS
As stated above, the goal of the experiment is to disprove theories where the light-pulse behaves either like a wave or a particle, depending on the experimental setup. We present here a quantitative model of this behavior, and confront it to the experimental results reported above. Note that such experiment can be perfectly explained by other theories exhibiting a wave-particle duality, like the de Broglie-Bohm pilot-wave theory [5, 6]. Wheelers delayed-choice experiment tests against particle-OR-wave theories, as opposed to particle-AND-wave theories, like quantum mechanics or the pilot-wave theory. In a particle-or-wave theory, the behavior of the light-pulse is quantied by the probability pw to behave like a wave, and the complementary probability pp = 1 pw to behave like a particle. The dependence of this behavior on the interferometer conguration is characterized by the probability pw|o (resp. pw|c) of a wave-like behavior when the interferometer is open (resp. closed). When the interferometer conguration choice is random and relativistically separated from the entrance of the photon in the interferometer, this choice cannot inuence the photon behavior, imposing pw|o = pw|c .
WILL BE SET BY THE PUBLISHER
The interference visibility V quanties the wave aspect of the light-pulse when the interferometer is closed. It depends of the optical pathlength dierence between the two arms, which can only inuence something (e.g. a wave) which simultaneously travels along bthe two arms. A single classical particle, on the contrary, has to choose its path and will not be sensitive to this path dierence, leading to V = 0 when the light-pulse behaves like a particle. Therefore, one has V pw|c . The anticorrelation parameter obeys 1 for any theory where the light is described by a wave, but decreases to zero for classical single particles. It is therefore a measure of the particle-like behavior [4], with the constraint pw|o . These inequalities show that a wave-or-particle theory can only explain experimental results in the delayed-choice regime if V . On the other hand, if V > , as in our experiment where V = 94% and = 12%, this can only be explained by either a wave-and-particle theory, like quantum mechanics, or by an information leakage from the QRNG to the light-pulse. To quantify the amount of information leakage needed to explain the observed values of the parameters V and , we need to compute a lower bound for the probability of the light-pulse to correctly guess the interferometer conguration. This probability is constrained by the conditional probability pc|w (resp. po|p ) for the interferometer to be closed (resp. open) knowing the light behaves like a wave (resp. like a particle). Applying Bayes theorem, one has pc|w = 1 pw|o pw|c V 1 > 88%; po|p = > 93%. pw|o + pw|c V + 2 pw|o pw|c 2V
Assuming a symmetrical probability for the light to correctly guess the QRNG output (i.e. pc|w = po|p ), this probability is therefore greater than 93%, which is hardly compatible with the estimated 52% predictability of the QRNG [2]. In conclusion, the experimental results can only be explained if the light-pulse is described by either (1) a wave-and-particle theory, like quantum mechanics; (2) a wave-or-particle theory where information on the setup can travel at 4 times the speed of light; (3) a wave-or-particle theory and our QRNG exhibits undetected correlationstaht could be detected with probability a higher than 93%. Needless to say, the most reasonable option seems to be the rst one! References
[1] A. Wheeler in Quantum Theory and Measurement, J. A. Wheeler and W. H. Zurek edit., (Princeton University Press, 1984). [2] V. Jacques, E Wu, F. Grosshans, F. Treussart, P. Grangier, A. Aspect, and J.-F. Roch, Science 315, 966 (2007). [3] A. Beveratos, R. Brouri, Th. Gacoin, J.-Ph. Poizat, and Ph. Grangier, Phys. Rev. A 64 061802(R) (2001). [4] Ph. Grangier, G. Roger, and A. Aspect, Europhys. Lett. 1, 173 (1986). [5] D. J. Bohm, C. Dewdney, and B. H. Kiley, Nature 315 294 (1985). [6] J. S. Bell, Speakable and Unspeakable in Quantum Mechanics (Cambridge Univ. Press, Cambridge, 1987).
