We study the properties of a photodetector that has a number-resolving capability. In the absence of dark counts, due to its finite quantum efficiency, photodetection with such a detector can only eliminate the possibility that the incident field corresponds to a number of photons less than the detected photon number. We show that such a non-photon number-discriminating detector, however, provides a useful tool in the reconstruction of the photon number distribution of the incident field even in the presence of dark counts.
I. THE PROBLEMWith the recent advent of linear optical quantum computation (LOQC), interest in photon number-resolving detectors has been growing. The ability to discriminate the number of incoming photons plays an essential role in the realization of nonlinear quantum gates in LOQC [1,2,3,4] as well as in quantum state preparation [5,6,7,8,9]. Such detectors are used to post-select particular quantum states of a superposition and consequently produce the desired nonlinear interactions of LOQC. In addition, the most probing attacks an eavesdropper can launch against a typical quantum cryptography system exploit photon-number resolving capability [10,11]. Recent efforts in the development of such photon number-resolving detectors include the visible light photon counter [12,13], fiber-loop detectors [14,15,16,17,18], and superconducting transition edge sensors [19]. Standard photodetectors can measure only the presence or absence of light (single-photon sensitivity), and generally do not have the capability of discriminating the number of incoming photons (single-photon number resolution). There have been suggestions of accomplishing single-photon resolution using many singlephoton-sensitive detectors arranged in a detector array or detector cascade [20,21,22,23]. For example, the VLPC (visible light photon counter) is based on a confined avalanche breakdown in a small portion of the total detection area and, hence, can be modeled as a detector cascade [24]. Then again, fiber-loop detectors may be regarded as a detector cascade in the time domain.Let us suppose a number-resolving detector detects two photons in a given time interval. If the quantum efficiency of the detector is one, we can be certain that two photons came from the incident light during that time interval. If the quantum efficiency is, say, 0.2, what can we say about the incident light? In the absence of dark counts, we can safely say only that there were at least two photons in the incident beam. In this case, photodetection rather has a non-photon number discriminating feature, since the conditional probability of not having zero, or one photon in the incident pulse is zero. Can we say more than that? For example, what is the probability that the incident pulse actually corresponds to two photons, or three? In this paper we attempt to answer these questions and discuss the effect of quantum efficiency on the photon-number resolving capability.The semiclassical treatment of photon counting statistics was first derived by Mand...