Experimental wavelength division multiplexed photon pair distribution
Joe Ghalbouni, Imad Agha, Robert Frey, Eleni Diamanti, Isabelle Zaquine
aa r X i v : . [ qu a n t - ph ] O c t Experimental wavelength division multiplexed photon pair distribution
Joe Ghalbouni, Imad Agha, Robert Frey, Eleni Diamanti and Isabelle Zaquine
Institut T´el´ecom/T´el´ecom Paristech, CNRS-LTCI 46 rue Barrault, 75013 Paris, France ∗ Corresponding author : [email protected]
Compiled 23 juin 2018We have experimentally implemented the distribution of photon pairs produced by spontaneous parametricdown conversion through telecom dense wavelength division multiplexing filters. Using the measured countsand coincidences between symmetric channels, we evaluate the maximum fringe visibility that can be obtainedwith polarization entangled photons and compare different filter technologies. c (cid:13)
OCIS codes:
In order to be truly useful for technological applica-tions, future quantum communication networks will re-quire a large number of high quality entangled photonpair sources. An attractive way to reduce the associa-ted cost is to use the wide spectrum produced by spon-taneous parametric down conversion (SPDC) in combi-nation with standard off-the-shelf dense wavelength di-vision multiplexing (DWDM) filters to distribute non-degenerate photon pairs to a large number of users. It isthen an important task to test photon pair sources in aWDM environment with various types of filters, in orderto unveil filter characteristics that are necessary to ob-tain sources featuring high visibility while maintaininga useful brightness. Among previous works based on theidea of DWDM distribution of photon pairs [1–4], Limet al [1] performed a proof-of-principle experiment usinga dichroic mirror to separate the two photons of the pairand tunable filters on each channel to simulate the de-multiplexing operation. Furthermore, a wavelength selec-tive switch, based on arrayed waveguide grating techno-logy, was tested in [2]. Frequency dependent losses weretaken into account to explain the experimental coinci-dence probabilities, however the system required a dif-ferent tuning of the pump for distribution over the va-rious channel pairs, hence it could not be simultaneouslymulti-user in that configuration. In this work, we com-pare various DWDM filters with respect to the qualityof a photon pair source with the ultimate goal of de-termining the filter performances required for quantumcommunication applications.The WDM photon pair distribution device is based onthe energy conservation condition of spontaneous para-metric down conversion : if ω p is the pump frequency,then the frequencies of the signal and idler photons ω s and ω i are symmetric with respect to ω p /
2. The ideathen is to tune the degeneracy frequency ω p / ω A ) and Bob ( ω B ), wouldmean tuning the pump frequency to ω A + ω B .Commercial optical demultiplexing filters are based onthree main technologies [5] : (a) dielectric thin-film filters(DTF), consisting of Fabry-Perot cavities and quarterwavelength layers ; these transmission bandpass filtersare cascaded in order to separate the different channels,(b) arrayed-waveguide gratings (AWG) that are planarlightwave circuits and are based on multi-beam interfe-rence ; the constructive interference condition in the out-put coupling region is satisfied at different focal pointsfor the different wavelengths that can then be coupledinto different output waveguides, (c) diffraction gratings(DG), which are free space diffraction gratings used incombination with imaging optics ; due to the wavelength-dependent diffraction angle, diffracted beams with dif-ferent wavelengths, are focused into different locationsand then coupled into output fibers. For all technolo-gies, the shape of the transmission curves can either beclose to a Gaussian one or have a flat-top structure.When such filters are used in a photon pair distribu-tion device, the main performance limitation is expectedto arise from the loss of coherence induced by variousimperfections of the demultiplexer that splits the pho-ton pairs. More specifically, the relevant filter parametersare the insertion loss, the stability and uniformity of thechannel bandwidth, the precision of the channel centerwavelength, as well as chromatic dispersion and groupdelay, characteristics that can greatly vary for differentfilter technologies. These parameters have a direct effecton the quality of an entangled photon source, which istypically evaluated using the visibility and the brightnessas the most important figures of merit.The experimental setup is depicted in Fig 1. The pumpis generated by a continuous-wave distributed feedback(DFB) laser at 779 nm, with a 20 mW power, and isfocused in a 2 cm long MgO-doped periodically poled li-thium niobate (PPLN) bulk crystal. The phase-matchingcondition is satisfied by adjusting the crystal tempera-ture (typically 65 o C) and pump, signal and idler waveshave the same polarization in order to use the highestnonlinear coefficient of the crystal d . A halfwave plate1 igure
1. Experimental setup.is used to change the pump polarization, in order tocontrol the efficiency of the nonlinear process and conse-quently the pair generation probability. After the crys-tal, all pump photons are filtered using dichroic mirrorsand colored glass filters. The photon pairs are then cou-pled into a single-mode fiber and go through the DWDMfilter, which features a channel separation of 100 GHzand a channel width of 100 GHz. In order to measuresingle counts at the signal and idler channels and coinci-dences, InGaAs avalanche diode single-photon detectorsare connected either to two symmetric channels of thefilter to obtain a deterministic splitting of the pairs orto the two outputs of a balanced fiber beamsplitter toobtain a statistical splitting. The detectors are triggeredat 2 MHz. Detector noise is 500 counts/s for a 20 ns gateand detector dead time is 10 µ s.We have shown in previous work [6] that using onlysingle count and coincidence measurements, in combina-tion with detector noise count data, the maximum fringevisibility that can be obtained when entanglement is im-plemented with such a photon pair source can be predic-ted accurately. The calculation takes into account thevisibility degradation due to double pair generation inSPDC, and examines the dependence of the visibility onthe channel insertion losses and the shape of the filtertransmission. In particular, we use single count, coin-cidence and noise count measurements performed withour setup to determine the ratio of true to accidentalcoincidences P TC /P AC , and thus evaluate the maximumattainable visibility V max of the source as follows [6] : V max = 11 + 2 P AC P TC (1)To characterize the influence of the DWDM filters, wemeasure their transmission using a wideband infraredsource in order to determine the following integrals : I (channel i ) = Z dω π T ( ω − ω F i ) (2) I (channel pair ij ) = Z dω s π T ( ω s − ω F ) T ( ω p − ω s − ω F )where T ( ω ) is the normalized intensity transmission ofthe filter, ω F i the center frequency of channel i and ω F the center frequency of all channel pairs. The single countprobabilities measured at the output of channel i are proportional to I (channel i ), while the coincidence pro-bability between channel i and channel j is proportionalto I , which is maximum when the center frequency ofthe filter is equal to the SPDC degeneracy frequency ω p / ω F . Using the values of I and I for the consi-dered channel or channel pair, we can then determine thedown conversion probability within the filter bandwidthor inband pair emission probability p ( I ) :statistical splitting p ( I ) = I I P AC P TC (3)deterministic splitting p ( I ) = I I P AC P TC (4)In the deterministic splitting case, where signal andidler frequencies are different, true coincidences could beobserved only when delaying the detection of one of thetwo filter channels. This is due to the demultiplexer tech-nologies, where the wavelength separation is associatedto a variable group delay. This delay was measured fordifferent channel pairs (see Table 1) within the availabledelay range of the detectors (0-25 ns). For the AWG fil-ter channel pair 21 −
27, the delay was too large to bemeasured. For a given pump frequency, the diffractiongrating technology gives rise to a constant delay for allchannel pairs where as the dielectric thin film techno-logy exhibits a delay varying from -2.5 to 22.5 ns. If thepump laser wavelength is fixed, then the distribution al-lows only fixed pairs of users. In this case, the delay canbe compensated for each channel pair. If the pump lasercan be tuned, in order to change the pairs of users thatreceive the two photons of one pair, a variable delay lineshould be used and a preliminary optimization of thecoincidences would have to be implemented before thepair distribution can start.
Table
1. Group delays for different channel pairs for fourdifferent filters. The channel numbers correspond to theInternational Telecommunication Union (ITU) grid.Filter type Channel numbers Delay (ns)23-25 15DTF (Flat-top) 22-26 22.521-27 -2.5AWG (Flat-top) 23-25 12.522-26 10DG (Flat-top) 23-25, 22-26, 21-27 10DG (Gaussian) 23-25, 22-26, 21-27 10Fig. 2 shows the maximum visibility as a function ofinband pair emission probability p ( I ), using determinis-tic or statistical splitting, for the three different flat-topfilters. The inband pair emission probability was variedbetween 0.01 and 0.18 to include a wide range of visi-bility values with a lower bound of √ /
2, below whichno entanglement is possible. Note that lower pair emis-sion probabilities can yield higher visibility, but at theexpense of impractical values for the source brightness,2
Inband pair emission probability M a x i m u m v i s i b ili t y DTF channel 24 + 50/50 couplerDTF channel pair 23−25AWG channel 24 + 50/50 couplerAWG channel pair 23−25DG Flat−top channel 24 + 50/50 couplerDG Flat−top channel pair 23−25
Figure
2. Maximum possible visibility as a function ofinband pair emission probability p ( I ) for the channelpair 23-25 and for the channel 24 using a 50/50 beam-splitter, with three different flat top filters.which corresponds to the number of true coincidencesper second. As we observe in Fig. 2, the deterministicsplitting gives better results than the statistical one [7].With respect to the various filter technologies, the AWGgives slightly lower visibility, but the difference is notlarge. The difference is clearly more significant when thesource brightness is considered, as we can see in Fig. 3.On the one hand, a large dispersion of this parameteris observed between different channel pairs of the dielec-tric thin film demultiplexer that cannot be explained so-lely by the insertion loss dispersion (the maximum chan-nel transmission varies from 0.69 to 0.82). On the otherhand, all the channel pairs of this demultiplexer give hi-gher brightness than any of the arrayed-waveguide anddiffraction grating demultiplexers.Fig. 4 shows both the good uniformity of the bright-ness for the different channel pairs of the flat-top andGaussian diffraction grating filters and the impact of thetransmission curve shape. Although we have confirmedthat the visibility of the Gaussian filter is only slightlylower than that of the flat-top, the corresponding bright-ness shows a very significant difference in favor of theGaussian shape, due its higher transmission efficiency.We have shown experimentally that commercialDWDM filters can be used to distribute photon pairsto a large number of users, taking advantage of the largebandwidth of spontaneous parametric down conversionin a PPLN crystal. We studied the impact of the de-multiplexer technology on the important figures of meritof the entanglement multi-user distribution, and showedthat, for a given visibility, the dielectric thin film tech-nology gives large brightness but also large inter-channeldispersion, whereas a good uniformity is provided by thediffraction grating filters, at the expense of a brightnessthat is two to four times smaller. Such a DWDM photonpair distribution device can provide an economical solu-tion for the development of quantum communication net-works as long as the requirements on filter performanceare satisfied, depending on the application. The trade-off Inband pair emission probability S ou r c e b r i gh t ne ss ( t r ue c o i n c i den c e s / s ) DTF channel pair 23−25DTF channel pair 22−26DTF channel pair 21−27AWG channel pair 23−25AWG channel pair 22−26DG Flat−top channel pair 23−25DG Flat−top channel pair 22−26DG Flat−top channel pair 21−27
Figure
3. Number of true coincidences per second as afunction of inband pair emission probability p ( I ) for allchannel pairs of the three flat top demultiplexers. Inband pair emission probability S ou r c e b r i gh t ne ss ( t r ue c o i n c i den c e s / s ) DG Flat−top channel pair 23−25DG Flat−top channel pair 22−26DG Flat−top channel pair 21−27DG Gaussian channel pair 23−25DG Gaussian channel pair 22−26DG Gaussian channel pair 21−27
Figure
4. Number of true coincidences per second asa function of inband pair emission probability for thediffraction grating filter, comparing Gaussian and flat-top shapes for the channel pairs 23-25, 22-26, and 21-27.between source visibility and brightness is clearly one ofthe main issues that have to be considered in such anoptimization process.
R´ef´erences
1. H. C. Lim, A. Yoshizawa, H. Tsuchida, and K. Kikuchi,Optical Fiber Technology , 225 (2010).2. J. Oh, C. Antonelli, and M. Brodsky, Journ. of Light-wave techn., , 3 (2011).3. I. Herbauts, B. Blauensteiner, A. Poppe, T. Jennewein,H. H¨ubel, CLEO-Europe, Munich 2010.4. S. Sauge, M. Swillo, S. Albert-Seifried, G. Xavier, J.Waldeback, M. Tengner, D. Ljunggren, and A. Karlsson,Opt. Express 15, 6926-6933 (2007).5. Y. Clavin Si and Y. Cheng, “Optical Multi-plexer/Demultiplexer : Discrete”, in ”WDM Tech-nologies, Passive optical components”, ed. A. K. Dutta,N. K. Dutta, M. Fujiwara, (Academic, 2003) pp 39-78.6. J.-L. Smirr, S. Guilbaud, J. Ghalbouni, R. Frey, E. Dia-manti, R. All´eaume, and I. Zaquine, Opt. Express ,616 (2011).7. J.-L. Smirr, R. Frey, E. Diamanti, R. All´eaume, and I.Zaquine, J. Opt. Soc. Am. B , 1 (2011). nformational Fourth Page In this section, please provide full versions of citationsto assist reviewers and editors (OL publishes a short formof citations) or any other information that would aid thepeer-review process.
R´ef´erences
1. H. C. Lim, A. Yoshizawa, H. Tsuchida, and K. Ki-kuchi, “Wavelength multiplexed entaglement distribu-tion”, Optical Fiber Technology , 225 (2010).2. J. Oh, C. Antonelli, and M. Brodsky, “Coincidence ratesfor photon pairs in WDM environment”, Journ. of Light-wave techn., , 3 (2011).3. I. Herbauts, B. Blauensteiner, A. Poppe, T. Jennewein,and H. H¨ubel, “Waveguide source for an on-demand en-tanglement distribution network”, CLEO-Europe, Mu-nich 2010.4. S. Sauge, M. Swillo, S. Albert-Seifried, G. Xavier, J.Waldeback, M. Tengner, D. Ljunggren, and A. Karls-son, “Narrowband polarization-entangled photon pairsdistributed over a WDM link for qubit networks,” Opt.Express 15, 6926-6933 (2007).5. Y. Clavin Si and Y. Cheng, “Optical Multi-plexer/Demultiplexer : Discrete”, in ”WDM Tech-nologies, Passive optical components”, ed. A. K. Dutta,N. K. Dutta, M. Fujiwara, (Academic, 2003) pp 39-78.6. J.-L. Smirr, S. Guilbaud, J. Ghalbouni, R. Frey, E.Diamanti, R. All´eaume, and I. Zaquine, “Simple per-formance evaluation of pulsed spontaneous parametricdown-conversion sources for quantum communications”,Opt. Express , 616 (2011).7. J.-L. Smirr, R. Frey, E. Diamanti, R. All´eaume, and I.Zaquine, “Intrinsic limitations to the quality of pulsedspontaneaous parametric down conversion sources forquantum information applications”, J. Opt. Soc. Am.B , 1 (2011)., 1 (2011).