scientific correspondence (for example, by cascading conventional detectors) could provide an effective rem- edy. Appropriate selection could be imple- mented with a polarization-independent S quantum non-demolition measurement of the total photon number at Bob's end. Alternatively, pre-selection could be imple- Alice mented by enhancing the coupling betwee f2 f1 Classical 8n modes (2, 3) relative to modes (1, 4). Despite our comments, we believe that the experiment of Bouwmeester et al. is a information significant achievement in demonstrating the non-local structure of teleportation. Samuel L. Braunstein SEECS, University of Wales, 1 2 Bangor LL57 1UT, UK H. J. Kimble Norman Bridge Laboratory of Physics 12­33, SI SII d1 California Institute of Technology, Sources Pasadena, California 91125, USA 4 3 1. Bouwmeester, D. et al. Nature 390, 575­579 (1997). d2 2. Bennett, C. H. et al. Phys. Rev. Lett. 70, 1895­1899 (1993). p PBS Bouwmeester et al. reply - Braunstein and Bob Kimble observe correctly that, in the Inns- ? bruck experiment, one does not always observe a teleported photon conditioned on a coincidence recording at the Bell-state analyser. In their opinion, this affects the FFiigguurree 11 The teleportation set-up of ref. 1. PBS, polarizing beam splitter. fidelity of the experiment, but we believe, in contrast, that it has no significance, and polarization state of a single-photon wave- field 4 at detector p, which projects the field that when a teleported photon appears, it packet (in beam 1) from Alice's sending in beam 1 accordingly. Joint detection at has all the properties required by the tele- station to Bob's receiving station (in beam (f1, f2) then provides threefold coincidence portation protocol. These properties can 3). Statistics consistent with teleportation with p, yielding a statistical mixture for the never be achieved by "abandoning telepor- are obtained for events with a fourfold field 3 arriving at Bob's station. The fraction tation altogether and transmitting ran- coincidence (from detectors f1 and f2, d1 of the state in this mixture gives F. To domly selected polarization states" as or d2, and p). We ask whether the detec- the lowest order in the down-converter Braunstein and Kimble suggest. The fact tion of all four quanta is essential for tele- coupling strength, the Bouwmeester et al. that there will be events where no tele- portation in this scheme. To answer this scheme yields a 50:50 mixture of the vacu- ported photons are created merely affects question, we calculated the teleportation um state 0 and the desired state , with the efficiency of the experiment. This sug- fidelity, F, when the coincidence condi- F 1/2, so that there is never a physical state gests that the measure of fidelity used by tion is relaxed to exclude detection at Bob's with high teleportation fidelity. Indeed, Bob Braunstein and Kimble is unsuitable for station (d1, d2). could achieve this same fidelity, F 1/2, by our experiment. Under relaxed conditions, requiring abandoning teleportation altogether and During the detection of the teleported only threefold coincidence of detectors p transmitting randomly selected polariza- photons, no selection was performed based and (f1, f2), teleportation is achieved when tion states. Faced with this state of affairs, on the properties of these photons. There- the fields of beams 1 and 3 match with suf- the experiment of ref. 1 obtains a surrogate fore, no a posteriori measurement in the ficiently high fidelity. In the simplest for high fidelity by destructively recording usual sense as a selective measurement was approximation, type II parametric down- the field 3 at (d1, d2). performed. The detection of the teleported conversion of modes (i, j) generates We emphasize that the nature of the photon could have been avoided altogether wavepacket states as follows: mixture containing the vacuum state has if we had used a more expensive detector, p, definite physical implications, which can be that could distinguish between one- and A0 0 ij + A1 i,j + A2 i,j +..., (1) verified by more general measurements two-photon absorption. The inability of than photon counting (for example, by our teleportation experiment to perform where A0, A1 and A2 are the coefficients for quantum-state tomography). Moreover, the such refined detections does not, however, obtaining no (vacuum), one and two freedom of a potential consumer of the out- imply that "a teleported state can never down-converted pairs, respectively, and put from Bob's receiving station to select emerge as a freely propagating state...". (i, j) (1, 4) (2, 3). Of these terms, only alternative detection strategies means that Braunstein and Kimble do not, therefore, states corresponding to the second term are classical analogies fail. reveal a principal flaw in our teleportation selected by fourfold coincidence, as speci- To achieve conventional a priori telepor- procedure, but merely address a non-trivial fied by equations (2) and (3) of ref. 1. How- tation, the set-up in ref. 1 would have to be practical question. ever, anything less than complete modified to eliminate the vacuum from the D. Bouwmeester, J.-W. Pan, M. Daniell, destruction of the output 3 necessarily mixture. Because the vacuum appears when H. Weinfurter, M. Zukowski, A. Zeilinger leaves undesirable terms that reduce F. two pairs of (1, 4) photons are created, we Institut für Experimentalphysik, The initial input state to Alice's station, might seek to resolve one- and two-photon Universität, Innsbruck, Technikerstrasse 25, , is prepared by detecting the state in detection events at p. Upgraded detection A-6020 Innsbruck, Austria NATURE | VOL 394 | 27 AUGUST 1998 841 Nature © Macmillan Publishers Ltd 1998