Implementation of an efficient linear-optical quantum router

For several decades, scientists have been aware of significant benefits allowing quantum information processing technologies to surpass their classical counterparts. Recent technological development allows these benefits to be tested experimentally and in some cases also implemented in practical devices. So far the majority of experimental quantum networks was limited to peer-to-peer communications between two parties. Practical implementation of quantum communications networks, however, needs to address the problem of scalability to serve large numbers of users. Similarly to classical computer networks, their quantum counterparts would require routing protocols to direct the signal from its source to destination. Devices implementing these routing protocols are called quantum routers and have recently been subject of an intense research. In this paper, we report on experimental implementation of a linear-optical quantum router. Our device allows single-photon polarization-encoded qubits to be routed coherently into two spatial output modes depending on the state of two identical control qubits. The polarization qubit state of the routed photon is maintained during the routing operation. The success probability of our scheme can be increased up to 25% making it the most efficient linear-optical quantum router developed to this date.


I. THREE-PHOTON SOURCE
In this experiment, we used a typical three-photon source based on spontaneous parametric down-conversion (SPDC) and attenuated coherent state (see Fig. 1). We use femtosecond laser system Mira (Coherent) to generate pulses with repetition rate 80 MHz, 800 mW mean power, central wavelength 826 nm and spectral width 10 nm (FWHM). These pulses are frequency doubled in the process of collinear second harmonics generation (SHG). The upconverted light beam in separated on a dichroic mirror. The depleted fundamental mode is attenuated by neutral density filter (NDF) to single-photon level (approximately 0.00125 photons per pulse). The generated second harmonics with central wavelength of 413 nm is filtered spectrally by a pinhole in 4F system. Remaining 100 mW of mean optical power pump nonlinear crystal BBO to produce photon pairs in the Type-I process of spontaneous parametric downconversion (SPDC). Approximate rate of photon pairs is 2 000 per second. All three optical modes -attenuated fundamental used as a signal (S IN ) and down-conversion used as two controls (C1 IN ,C2 IN ) -are spectrally filtered by narrow-band filters with 3 nm FWHM. Subsequently the modes are coupled into single-mode optical fibers leading to three optical inputs of the main experimental setup -linear-optical quantum router.
The main problem of every three photon source of this type is noise caused by multiphoton contributions. In all three modes there is a nonzero probability of having generated two or more photons per laser pulse. And if the spatial modes are mixed on beam splitters as in our router than the multiphoton events can cause false three fold coincidence. These false coincidences have lower probability than the right ones but still they can affect the measurement results. Their rate can be easily found out by sequential blocking of each mode, see analysis in Ref. [1].

II. MEASUREMENT PROCEDURES AND OBTAINED DATA
Typical rate of three fold coincidence counts (two controls and one of the signal outputs) was 1-2 per minute. This rate depends on the polarization projection on the signal outputs. To have low errors we have typically accumulated the data for 300 minutes for each setting of the router. Typical probability of accidental coincidences caused by multiple photons was 20%. Due to the polarization projection the effective rate of noisy coincidences was ten times lower then right ones.
We present raw data without correction on false three fold coincidences and the corrected data for routing probabilities in the second output port in Table I. Uncorrected routing probabilities are visualized in Fig. 2. Mean contrast of the routing in first output mode without correction is (5.7 ± 0.9) : 1, with correction then (15.7 ± 4.6) : 1. In the second output port we obtain raw contrast (5.8 ± 0.6) : 1 and corrected one (41.8 ± 19.7) : 1.
In the table II, we show measured uncorrected and corrected output state fidelities on the first (control OFF) and the second (control ON) output port. The mean values for both cases are in the last row of the table. The uncorrected fidelities are also shown in Fig. 3.
The coherence between signal outputs OUT1 and OUT2 was tested without mirror M1 and beam displacer BD3 only for one input polarization. The relative phase