REPORTS

nuclear magnetic resonance (NMR) implemen- tations (19, 20). In order to overcome this issue, we then experimentally demonstrated a quantum A Quantum Delayed-Choice Experiment delayed-choice scheme based on Bell’s inequality (21),whichallowedustotestthemostgeneral Alberto Peruzzo,1* Peter Shadbolt,1* Nicolas Brunner,2† Sandu Popescu,2 Jeremy L. O’Brien1‡ classical model. The main conceptual novelty of this scheme is that the temporal arrangement Quantum systems exhibit particle- or wavelike behavior depending on the experimental apparatus of Wheeler’s original proposal—the delayed they are confronted by. This wave-particle duality is at the heart of . Its choice of closing the interferometer or not—is paradoxical nature is best captured in the delayed-choice thought experiment, in which a is not necessary anymore. Instead, we certify the forced to choose a behavior before the observer decides what to measure. Here, we report on quantum nature of the photon’s behavior by a quantum delayed-choice experiment in which both particle and wave behaviors are investigated observing the violation of a Bell inequality. This simultaneously. The genuinely quantum nature of the photon’s behavior is certified via nonlocality, demonstrates in a device-independent way— which here replaces the delayed choice of the observer in the original experiment. We observed that is, without making assumptions about the strong nonlocal correlations, which show that the photon must simultaneously behave both as a functioning of the devices—that no local hidden particle and as a wave. variable model can reproduce the quantum pre- dictions. In other words, no model in which uantum mechanics predicts with remark- the two arms of the interferometer can be ad- the photon decided in advance which behavior able accuracy the result of experiments justed so that the particle will emerge in output to exhibit—knowing in advance the measurement Qinvolving small objects, such as atoms D′ with certainty. That is, the interference is setup—can account for the observed statistics. and . However, when looking fully constructive in output D′ and fully de- In our experiment, we achieve strong Bell in- more closely at these predictions we are forced structive in output D″. This measurement thus equality violations, hence giving an experimen- to admit that they defy our intuition. Indeed, clearly highlights the wave aspect of the quan- tal refutation to such hidden variable models, up quantum mechanics tells us that a single particle tum particle. However, the observer performing to a few additional assumptions about the imple- can be in several places at the same time, and the experiment has the choice of modifying mentation that are regularly used in experimen-

that distant entangled particles behave as a sin- the above experiment, in particular by remov- talBelltests. on January 15, 2015 gle physical object no matter how far apart they ing the second beamsplitter of the interferom- Our scheme is presented in Fig. 1B. A single are (1). eter. In this case, he will perform a which-path photon (our system) is sent through an inter- In trying to grasp the basic principles of the measurement. The photon will be detected in ferometer. At the first beamsplitter, the photon theory—in particular, to understand more intu- each mode with probability one half, thus ex- evolves into a superposition of the two spatial itively the behavior of quantum particles—the hibiting particle-like behavior. The main point modes, represented by two orthogonal quantum notion of wave-particle duality was introduced is that the experimentalist is free to choose which states |0〉s and |1〉s. Formally, this first beam- (2). A quantum system—for instance, a photon— experiment to perform (interference or which- may behave either as a particle or a wave. How- path, thus testing the wave or the particle as-

ever, the way in which it behaves depends on the pect) once the particle is already inside the www.sciencemag.org kind of experimental apparatus with which it is interferometer. Thus, the particle could not have measured. Hence, both aspects, particle and wave, known in advance (for instance via a hidden which appear to be incompatible, are never ob- variable) the kind of experiment with which it served simultaneously (3). This is the notion of will be confronted because this choice was complementarity in quantum mechanics (4–7), simply not made when the particle entered which is central in the standard Copenhagen the interferometer. Wheeler’s experiment has interpretation and has been intensely debated been implemented experimentally by using var-

in the past. ious systems, all confirming quantum predictions Downloaded from In an effort to reconcile quantum predictions (11–15). In a recent experiment with single pho- and common sense, it was suggested that quan- tons, a spacelike separation between the choice tum particles may in fact know in advance to of measurement and the moment the photon which experiment they will be confronted, via enters the interferometer was achieved (16). a hidden variable, and could thus decide which We explored a conceptually different take behavior to exhibit. This simplistic argument was, on Wheeler’s experiment. Our starting point is a however, challenged by Wheeler in his elegant recent theoretical proposal (17) of a delayed- “delayed choice” arrangement (8–10). In this choice experiment based on a quantum-controlled gedanken experiment, as shown in Fig. 1A, a beamsplitter, which can be in a superposition Fig. 1. Quantum delayed-choice experiment. (A) Schematic of Wheeler’s original delayed-choice quantum particle is sent toward a Mach-Zender of present and absent. Hence, the interferom- experiment. A photon is sent into a Mach-Zehnder interferometer. The relative phase ϕ between eter can be simultaneously closed and open, interferometer and split into a superposition across thus testing both the wave and the particle be- both paths at the first beamsplitter (solid blue havior of the photon at the same time. Using a 1Centre for Quantum Photonics, H. H. Wills Laboratory line). By inserting (or not) the second beamsplit- and Department of Electrical and Electronic Engineering reconfigurable integrated quantum photonic cir- ter (dashed blue line), wave (or particle) behavior 2 18 University of Bristol, Bristol BS8 1UB, UK. H. H. Wills Physics cuit ( ), we implemented an interferometer fea- can be observed at detectors D′ or D″.(B)Schematic Laboratory, University of Bristol, Tyndall Avenue, Bristol BS8 turing such a quantum beamsplitter, observing of the quantum delayed-choice experiment. The 1TL, UK. continuous morphing between wave and particle second beamsplitter is now a quantum beamsplit- *These authors contributed equally to this work. behavior (17). However, this morphing behavior ter (represented by a controlled-Hadamard opera- †Present address: Département de Physique Théorique, Uni- versité de Genève, 1211 Genève, Switzerland. can be reproduced by a simple classical model, tion), which can be set in a superposition of present ‡To whom correspondence should be addressed. E-mail: and this loophole also plagues both the theoret- and absent by controlling the state of an ancilla [email protected] ical proposal of (17) as well as two of its recent photon |y〉a.

634 2 NOVEMBER 2012 VOL 338 SCIENCE www.sciencemag.org REPORTS

2 splitter is represented by a Hadamard operation took. The measured intensities are ID′ =cos(ϕ/2) data shows interference fringes. Our results are 2 (22), which transforms the initial photonpffiffiffi state and ID″ =sin(ϕ/2). in excellent agreement with theoretical predic- |0〉s into the superposition ðj0〉s þj1〉sÞ= 2.A The main feature of this quantum controlled tions (Fig. 3). phase shifter then modifies the relative phase beamsplitter is that it can be put in a superpo- To achieve our main goal—to refute models between the two modes,pffiffiffi resulting in the state sition of being present and absent. Indeed, if the in which the photon knows in advance with iϕ jy〉s ¼ðj0〉s þ e j1〉sÞ= 2. Both modes are ancilla photon is initially in a superposition—for which setup it will be confronted—we must go then recombined on a second beamsplitter before instance, in the state |y〉a =cosa|0〉a +sina|1〉a— one step further. Indeed, the result of Fig. 3 does a final measurement in the logical ({|0〉s,|1〉s}) then the global state of the system evolves into not refute such models. Although we have in- basis. In the standard delayed-choice experiment, serted the ancilla photon in a superposition, hence the presence of this second beamsplitter is con- testing both wave and particle aspects at the same |Yf (a,ϕ)〉 = cosa|y〉s |0〉a + trolled by the observer (see Fig. 1A). For a closed ,particle time, we have in fact not checked the quantum (3) interferometer, the statistics of the measurements sina| y〉s,wave|1〉a nature of this superposition. This is because the at detectors D′ and D″ will depend on the phase final measurement of the ancilla photon was made ϕ, revealing the wave nature of the photon. For The system and ancilla photons now become en- in the logical ({|0〉a,|1〉a}) basis. Therefore, we can- an open interferometer, both detectors will click tangled, when 0 < a < p/2. not exclude the fact that the ancilla may have been 2 with equal probability, revealing the particle na- The measured intensity at detector D′ is then in a statistical mixture of the form cos a|0〉〈0|a + 2 ture of the photon. given by sin a|1〉〈1|a, which would lead to the same mea- Here, on the contrary, the presence of the sec- sured statistics. Hence, the data can be explained 2 2 ond beamsplitter depends on the state of an an- ID′ (ϕ,a) ¼ Iparticle(ϕ)cos a þ Iwave(ϕ)sin a by a classical model, in which the state of the cillary photon. If the ancilla photon is prepared ancilla represents a classical variable (a classical in the state |0〉a, no beamsplitter is present; hence, 1 ϕ bit) indicating which measurement, particle or ¼ cos2a þ cos2 sin2 a ð4Þ the interferometer is left open. Formally, this cor- 2 2 wave, will be performed. Because the state of the responds to the identity operator acting on |y〉s, ancilla may have been known to the system pho- resulting in the state whereas intensity at D″ is ID″(ϕ, a)=1− ID′(ϕ, a). toninadvance—indeed, here no delayed choice We fabricated the quantum circuit shown in is performed by the observer—no conclusion can p1ffiffiffi iϕ Fig. 2 in a silica-on-silicon photonic chip (18). be drawn from this experiment. This loophole jy〉s,particle ¼ ðj0〉s þ e j1〉sÞð1Þ 2 The Hadamard operation is implemented by a also plagues the recent theoretical proposal of directional coupler of reflectivity 1/2, which is (17), as well as two of its NMR implementations The final measurement (in the {|0〉s,|1〉s}basis) equivalent to a 50/50 beamsplitter. The controlled- (19, 20). indicates which path the photon took, revealing Hadamard (CH) is based on a nondeterministic In order to show that the measurement choice the particle nature of the photon. The measured control-phase gate (23, 24). The system and an- could not have been known in advance, we must intensities in both output modes are equal and cilla photon pairs are generated at 808 nm via ensure that our quantum controlled beamsplitter phase-independent, ID′ = ID″ =1/2. parametric down conversion and detected with behaves in a genuine quantum way. In particular, If, however, the ancilla photon is prepared in silicon avalanche photodiodes at the circuit’s we must ensure that it creates entanglement be- the state |1〉a, the beamsplitter is present, and the output. tween the system and ancilla photons, which is the interferometer is therefore closed. Formally, this We first characterized the behavior of our clear signature of a quantum process. The global corresponds to applying the Hadamard opera- setup for various quantum states of the an- state of the system and ancilla photons, given in tionto|y〉s, resulting in the state cilla photon. We measured the output intensities Eq. 3, is entangled for all values 0 < a < p/2. ID′(ϕ, a)andID″(ϕ, a)fora ∈ [0, p/2], and ϕ ∈ Because 〈yparticle|ywave〉 ∼ cosϕ, the degree of ϕ ϕ [−p/2, 3 p/2]. In particular, by increasing the entanglement depends on ϕ and a; in particular, jy〉s,wave ¼ cos j0〉s − i sin j1〉s ð2Þ 2 2 value of a we observe the morphing between for a = p/4 and ϕ = p/2 the state in Eq. 3 is a particle measurement (a = 0) and a wave mea- maximally entangled. The final measurement gives information about surement (a = p/2). For a = 0 (no beamsplitter), In order to certify the presence of this entan- the phase ϕ that was applied in the interferome- the measured intensities are independent of ϕ. glement, we tested the Clauser-Horne-Shimony- ter, but indeed not about which path the photon For a = p/2, the beamsplitter is present, and the Holt (CHSH) Bell inequality (25), the violation

Fig. 2. Implementation of the quantum delayed- choice experiment on a reconfigurable integrated photonic device. Non- entangled photon pairs are generated by using type I parametric down- conversion and injected into the chip by using maintaining fibers (not shown). The system photon (s), in the lower part of the circuit, enters the interferometer at the Hadamard gate (H). A relative phase ϕ is state of the second beamsplitter—a superposition of present and absent. applied between the two modes of the interferometer. Then, the controlled- Last,thelocalmeasurementsfortheBell test are performed through single- Hadamard (CH) is implemented by a nondeterministic CZ gate with two qubit rotations (UA and UB) followed by APDs. The circuit is composed of additional MZ interferometers. The ancilla photon (a), in the top part of the directional couplers of reflectivity 1/2 (dc1−5 and dc9−13)and1/3(dc6−8)and circuit, is controlled by the phase shifter a, which determines the quantum resistive heaters (orange rectangles) that implement the phase shifters (25).

www.sciencemag.org SCIENCE VOL 338 2 NOVEMBER 2012 635 REPORTS

of which would imply in a device-independent surement operators of Alice and Bob [adjusting making further assumptions, a loophole-free Bell way that the measured data could not have been phase shifters 5, 6, and 8 (26)] for the maxi- inequality violation is required. This is not the produced by a classical model. In the CHSH Bell mally entangled state |Yf (a = p/4, ϕ = p/2)〉. case in our experiment, as in all optical Bell tests scenario, each party (here, Alice holds the sys- Hence, for this state we expect the maximal pos- performed so far, which forces us to make a few tem photon while Bob holds the ancilla photon) sible violation of the CHSH inequalitypffiffiffi in quan- additional assumptions. We make the standard chooses among two possible measurement set- tum mechanics—namely, S ¼ 2 2 (27). The fair-sampling assumption (allowing us to dis- tings, denoted x = 0,1 for Alice and y =0,1for choice of apparatus in Wheeler’s original setup card inconclusive results and postselect only co- Bob. Each measurement is dichotomic, giving a is here, in some sense, replaced by the choice of incidence events), which must here be slightly binary result Ax = T1andBy = T1. The CHSH measurement settings for the . The latter strengthened because of the nondeterministic inequality then reads choice is nevertheless conceptually different from implementation of the controlled Hadamard op- the former, in that it can be performed after the eration. We must also assume independence be- S 〈A B 〉 〈A B 〉 〈A B 〉 − 〈A B 〉 ≤ = 0 0 + 0 1 + 1 0 1 1 2 (5) photon left the interferometer. tween the photon source and the choice of Experimentally, we observed a maximal vio- measurement setting used in the Bell inequality This represents a Bell inequality in the sense lation of S =2.45T 0.03 for a = p/4 and ϕ = p/2, test. As usual, if the photons could know in ad- that any local model must satisfy it. which is in good agreement with theoretical pre- vance the choice of measurement setting in the Indeed, this inequality can be violated by dictions (Fig. 4). Therefore, our data could not Bell test, then a local model can mimic Bell in- making judiciously chosen local measurements have been accounted for by any model in which equality violations. It would be interesting to per- on certain entangled states. We measured S for the system photon would have known in ad- form a more refined experiment in which these the output state |Yf (a, ϕ)〉 for a ∈ [0, p/2] and vance whether to behave as a particle or as a assumptions could be relaxed (28, 29). ϕ ∈ [−p/2, 3p/2]. We tailored the local mea- wave. However, for this claim to hold without We have reported on a quantum delayed- choice experiment, giving a novel demonstration of wave-particle duality, Feynman's “one real mystery” in quantum mechanics. In our experi- ment, the delayed choice of Wheeler’s proposal is replaced by a quantum controlled beamsplit- ter followed by a Bell inequality test. In this way, we demonstrate genuine quantum behav- ior of single photons. The demonstration of a quantum controlled beamsplitter shows that a single measurement device can continuously tune between particle and wave measurements, hence pointing toward a more refined notion of complementarity in quantum mechanics (17, 30–32). Note added in proof. We note a related work Fig. 3. Characterization of the continuous transition between wave and particle behavior. (A)Measured of Kaiser et al.(33), who performed a similar and (B) simulated intensity at detector D′ when continuously tuning the state of the ancilla photon |y〉a. quantum delayed-choice experiment. The experimental data (white dots) were fitted by using Eq. 4. The data shows excellent agreement with theoretical predictions. Error bars due to Poissonian noise are smaller than the data points; hence, they References and Notes arenotdrawn.Thediscrepancybetweentheexperimental and theoretical results is not due to statistical 1. J. S. Bell, Speakable and Unspeakable in Quantum fluctuations but to imperfection in the device calibration. Mechanics (Cambridge Univ. Press, Cambridge, 2004). 2. R. P. Feynman, R. B. Leighton, M. L. Sands, Lecture Notes on Physics (Addison-Wesley, Reading, MA, 1965). 3. N. Bohr, in Quantum Theory and Measurement, J. A. Wheeler, W. H. Zurek, Eds. (Princeton Univ. Press, Princeton, NJ, 1984), pp. 9–49. 4. M. O. Scully, B.-G. Englert, H. Walther, Nature 351, 111 (1991). 5. B.-G. Englert, Phys. Rev. Lett. 77, 2154 (1996). 6. W. K. Wootters, W. H. Zurek, Phys. Rev. D Part. Fields 19, 473 (1979). 7. V. Jacques et al., Phys. Rev. Lett. 100, 220402 (2008). 8. J. A. Wheeler, in Mathematical Foundations of Quantum Mechanics, A. R. Marlow, Ed. (Academic, New York, 1978), pp. 9–48. 9. J. A. Wheeler, in Quantum Theory and Measurement, J. A. Wheeler, W. H. Zurek, Eds. (Princeton Univ. Press, Princeton, NJ, 1984), pp. 182–213. 10. A. J. Leggett, in Compendium of Quantum Physics, D. Greenberger, K. Hentschel, F. Weinert, Eds. (Springer, Fig. 4. Experimental Bell-CHSH inequality test. (A)Measuredand(B) simulated Bell-CHSH parameter Berlin, 2009), pp. 161–166. S (Eq. 1). When the CHSH inequality is violated—when S > 2 [yellow dots in (A) and yellow circle in 11. T. Hellmut, H. Walther, A. G. Zajonc, W. Schleich, (B)]—no local hidden variable model can explain the observed data, hence demonstrating genuine Phys. Rev. A 35, 2532 (1987). quantum behavior. The maximal experimental violation (S =2.45T 0.03) is achieved for a = p/4 and 12. B. J. Lawson-Daku et al., Phys. Rev. A 54, 5042 ϕ = p/2, as expected. The data are in excellent agreement with theoretical predictions. Error bars due (1996). 13. Y.-H. Kim, R. Yu, S. P. Kulik, Y. Shih, M. O. Scully, to Poissonian noise are smaller than the data points; hence, they are not drawn. The discrepancy Phys. Rev. Lett. 84, 1 (2000). between the experimental and theoretical results is not due to statistical fluctuations but to imper- 14. A. Zeilinger, G. Weihs, T. Jennewein, M. Aspelmeyer, fection in the device calibration. Nature 433, 230 (2005).

636 2 NOVEMBER 2012 VOL 338 SCIENCE www.sciencemag.org REPORTS

15. A. Zeilinger, G. Weihs, T. Jennewein, M. Aspelmeyer, 26. Materials and methods are available as supplementary European Research Council (ERC), the Quantum Integrated Nature 446, 342 (2007). materials on Science Online. Photonics (QUANTIP) project, A Toolbox for Photon Orbital 16. V. Jacques et al., Science 315, 966 (2007). 27. B. S. Cirel’son, Lett. Math. Phys. 4, 93 (1980). Angular Momentum Technology (PHORBITECH) project, the 17. R. Ionicioiu, D. R. Terno, Phys. Rev. Lett. 107, 230406 28. A. Aspect, J. Dalibard, G. Roger, Phys. Rev. Lett. 49, 1804 Quantum InterfacES, SENsors, the Communication based on (2011). (1982). Entanglement (Q-ESSENCE) integrating project, Nokia, the 18. P. J. Shadbolt et al., Nat. Photonics 6, 45 (2012). 29. G. Weihs, T. Jennewein, C. Simon, H. Weinfurter, A. Zeilinger, Centre for Nanoscience and Quantum Information (NSQI), the 19. S. S. Roy, A. Shukla, T. S. Mahesh, Phys. Rev. A 85, Phys. Rev. Lett. 81, 5039 (1998). Templeton Foundation, and the European Union Union 022109 (2012). 30. J.-S. Tang, Y.-L. Li, C.-F. Li, G.-C. Guo, Nat. Photonics Device-Independent Quantum Information Processing (DIQIP) 20. R. Auccaise et al., Phys. Rev. A 85, 032121 (2012). 6, 602 (2012). project. J.L.O. and S.P. acknowledge a Royal Society Wolfson 21. J. S. Bell, Physics 1, 195 (1964). 31. T. Qureshi, Quantum Phys., arXiv:1205.2207. Merit Award. A.P. holds a Royal Academy of Engineering 22. M. A. Nielsen, I. L. Chuang, Quantum Computation 32. X.-s. Ma, Quantum Phys., arXiv:1206.6578. Research Fellowship. and Quantum Information (Cambridge Univ. Press, 33. F. Kaiser, T. Coudreau, P. Milman, D. B. Ostrowsky, Cambridge, MA, 2000). S. Tanzilli, Science 338, 637 (2012). Supplementary Materials 23. T. C. Ralph, N. K. Langford, T. B. Bell, A. G. White, www.sciencemag.org/cgi/content/full/338/6107/634/DC1 Materials and Methods Phys. Rev. A 65, 062324 (2002). Acknowledgments: We thank R. Ionicioiu, S. Pironio, 24. H. F. Hofmann, S. Takeuchi, Phys.Rev.A66, 024308 (2002). T. Rudolph, N. Sangouard, and D. R. Terno for useful Fig. S1 25. J. F. Clauser, M. Horne, A. Shimony, R. A. Holt, Phys. Rev. discussions, and acknowledge financial support from the UK 28 June 2012; accepted 18 September 2012 Lett. 23, 880 (1969). Engineering and Physical Sciences Research Council (EPSRC), 10.1126/science.1226719

photon sent to the interferometer and the other Entanglement-Enabled one as a corroborative photon. Here, as opposed to previous experiments (8, 11), the state of the Delayed-Choice Experiment interferometer remains unknown, as does the wave or particle behavior of the test photon, until

1 2 2,3 1 1 we detect the corroborative photon. By continuous- Florian Kaiser, Thomas Coudreau, Pérola Milman, Daniel B. Ostrowsky, Sébastien Tanzilli * ly modifying the type of measurement performed on the corroborative photon, we can morph the Wave-particle complementarity is one of the most intriguing features of quantum physics. To test photon from wave to particle behavior, even emphasize this measurement apparatus–dependent nature, experiments have been performed after the test photon was detected. To exclude in which the output beam splitter of a Mach-Zehnder interferometer is inserted or removed after interpretations based on either mixed states, as- a photon has already entered the device. A recent extension suggested using a quantum beam sociated with preexisting state information (15), splitter at the interferometer’s output; we achieve this using pairs of polarization-entangled or potential communication between the two pho- photons. One photon is tested in the interferometer and is detected, whereas the other allows tons, the presence of entanglement is verified via us to determine whether wave, particle, or intermediate behaviors have been observed. Furthermore, the violation of the Bell inequalities with a space- this experiment allows us to continuously morph the tested photon’s behavior from wavelike to like separation (16–18). particle-like, which illustrates the inadequacy of a naive wave or particle description of . The QBS is based on the idea that when a photon in an arbitrary polarization state enters lthough the predictions of quantum me- photon’s path can be known, and consequently, no an interferometer that is open for |H〉 (horizon- chanics have been verified with marked interference occurs. Particle behavior is said to be tally polarized) and closed for |V〉 (vertically Aprecision, subtle questions arise when observed, and the detection probabilities at Da polarized) photons, the states of the interferom- attempting to describe quantum phenomena in and Db are equal to ½, independent of the value eter and the photon become correlated. Our ap- classical terms (1, 2). For example, a single quan- of q (Fig. 1C). In other words, these two different paratus, shown in the right-hand side of Fig. 2 tum object can behave as a wave or as a particle. configurations—BS2 present or absent—give dif- and detailed in fig. S1, therefore reveals a particle This concept is illustrated by Bohr’scomplemen- ferent experimental results. Recently, Jacques et al. behavior for the |H〉 component of the photon tarity principle (3) which states that, depending have shown that, even when performing Wheeler’s state and a wave behavior for the |V〉 compo- on the measurement apparatus, either wave or original gedanken experiment (7)inwhichthe nent. Note that such an experiment has been particle behavior is observed (4, 5). This is dem- configuration for BS2 is chosen only after the realized with the use of single photons prepared onstrated by sending single photons into a Mach- photon has passed the entrance beam splitter BS1, in a coherent superposition of |H〉 and |V〉 (12). Zehnder interferometer (MZI) followed by two Bohr’s complementarity principle is still obeyed However, we take this idea a step further by detectors (Fig. 1A) (6). If the MZI is closed [that (8). Intermediate cases, in which BS2 is only par- achieving genuine quantum behavior for the out- is, if the paths of the interferometer are recom- tially present, have been considered in theory and put beam splitter by exploiting an intrinsically bined at the output beam splitter (BS2)], the prob- led to a more general description of Bohr’scom- quantum resource, entanglement. This allows us abilities for a photon to exit at detectors Da and plementarity principle expressed by an inequality to entangle the quantum beam splitter and test Db depend on the phase difference q between limiting the simultaneously available amount of photon system with the corroborative photon. Thus, the two arms. The which-path information remains interference (signature of wavelike behavior) and measurement of the corroborative photon enables unknown, and wavelike intensity interference pat- which-path information (particle-like behavior) us to project the test photon–QBS system into an terns are observed (Fig. 1B). On the other hand, (9, 10). This inequality has also been confirmed arbitrary coherent wave-particle superposition, which if the MZI is open (i.e., if BS2 is removed), each experimentally in delayed-choice configurations is a purely quantum object. In other words, our (11, 12). QBS is measured by another quantum object, ’ 1Laboratoire de Physique de la Matière Condensée, CNRS UMR We take Wheeler s experiment one step fur- which projects it into a particular superposition 7336, Université de Nice–Sophia Antipolis, Parc Valrose, 06108 ther by replacing the output beam splitter by a of present and absent states. More precisely, we Nice Cedex 2, France. 2Laboratoire Matériaux et Phénomènes quantum beam splitter (QBS), as theoretically pro- use as a test photon one of the photons from Quantiques, Université Paris Diderot, Sorbonne Paris Cité, CNRS, posed of late (13, 14). In our experiment (Fig. 2), the maximally polarization-entangled Bell state UMR 7162, 75013 Paris, France. 3Institut de Sciences Moléc- † † † † ’ we exploit polarization entanglement as a re- þ p1ffiffi ulaires d Orsay (CNRS) Bâtiment 210, Université Paris Sud 11, jF 〉 ¼ ðcH tH þ cV tV Þjvac〉, produced at the Campus d’Orsay, 91405, Orsay Cedex, France. source for two reasons. First, doing so permits 2 *To whom correspondence should be addressed. E-mail: implementing the QBS. Second, it allows us of 1560 nm using the source de- [email protected] to use one of the entangled photons as a test scribed in (19). Here, using the notation of Fig. 2,

www.sciencemag.org SCIENCE VOL 338 2 NOVEMBER 2012 637 A Quantum Delayed-Choice Experiment Alberto Peruzzo et al. Science 338, 634 (2012); DOI: 10.1126/science.1226719

This copy is for your personal, non-commercial use only.

If you wish to distribute this article to others, you can order high-quality copies for your colleagues, clients, or customers by clicking here. Permission to republish or repurpose articles or portions of articles can be obtained by following the guidelines here.

The following resources related to this article are available online at www.sciencemag.org (this information is current as of January 15, 2015 ):

Updated information and services, including high-resolution figures, can be found in the online version of this article at: on January 15, 2015 http://www.sciencemag.org/content/338/6107/634.full.html Supporting Online Material can be found at: http://www.sciencemag.org/content/suppl/2012/10/31/338.6107.634.DC1.html http://www.sciencemag.org/content/suppl/2012/10/31/338.6107.634.DC2.html A list of selected additional articles on the Science Web sites related to this article can be found at: http://www.sciencemag.org/content/338/6107/634.full.html#related www.sciencemag.org This article cites 23 articles, 2 of which can be accessed free: http://www.sciencemag.org/content/338/6107/634.full.html#ref-list-1 This article has been cited by 6 articles hosted by HighWire Press; see: http://www.sciencemag.org/content/338/6107/634.full.html#related-urls This article appears in the following subject collections:

Physics Downloaded from http://www.sciencemag.org/cgi/collection/physics

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. Copyright 2012 by the American Association for the Advancement of Science; all rights reserved. The title Science is a registered trademark of AAAS.