Up-conversion, or hybrid, detectors have been investigated in quantum communication experiments to replace Indium-Gallium-Arsenide avalanche photodiodes (InGaAs-APD) for the detection of infrared and telecom single photons. Those detectors are based on the supposedly noise-free process of frequency up-conversion, also called sum-frequency generation (SFG), using a second order (χ2) non-linear crystal. Powered by an intense pump laser, this process permits transposing with a certain probability the single photons at telecom wavelengths to the visible range where silicon APDs (Si-APD) operate with a much better performance than InGaAs detectors. To date, the literature reports up-conversion detectors having efficiency and noise figures comparable to that of the best commercially available IngaAs-APDs. However, in all of these previous realizations, a pump-induced noise is always observed which was initially expected to be as low as the dark count level of the Si-APDs. Although this additional noise represents a problem for the detection, up-conversion detectors have advantageously replaced InGaAs-APDs in various long-distance quantum cryptography schemes since they offer a continuous regime operation mode instead of a gated mode necessary for InGaAs-APDs, and the possibility of much higher counting rates. Despite attempted explanations, no detailed nor conclusive study of this noise has been reported.

The aim of this paper is to offer a definitive explanation for this noise.We first give a review of the state of the art by describing already demonstrated up-conversion detectors. We discuss these realizations especially regarding the choices made for the material, in bulk or guided configurations, the single photon wavelengths, and the pump scheme. Then we describe an original device made of waveguides integrated on periodically poled lithium niobate (PPLN)or on single-domain lithium niobate aimed at investigating the origin of the additional pump-induced noise. The poled waveguides are designed to up-convert single photons at 1550 nm to 600 nm when a 980 nm diode laser is used as pump. We obtain an overall efficiency of about 0.6% for a noise level of about 8 × 103 counts/s. This overall efficiency includes both insertion and propagation losses, and internal up-conversion and quantum detection (Si-APD) efficiencies. Despite a low efficiency value compared to what has been obtained so far by other groups, the efficiency/noise ratio is still comparable which still allows us investigating the noise issue.

From the spectrum obtained in both poled and non-poled waveguides we conclude that the noise comes from an alternative phase-matching scheme which permits creating paired photons at 1550 and 2700 nm wavelength by down-conversion of the 980 nm pump laser. Knowing that 1550 nm corresponds to the input signal wavelength, up-conversion of actual signal or pump-induced photons at this particular wavelength cannot be discriminated, therefore contributing to the noise at the final wavelength of 600 nm. We believe that this process of down-conversion of the pump laser to the signal wavelength (plus complementary wavelength) is responsible for the unexpected noise level reported in all the up-conversion detector realizations.