Quantum Optics Seminar by Jelmer Renema

The hardness of boson sampling with imperfections 

A major goal in quantum information is demonstrating a task which can be performed faster in real time by a quantum machine than by a classical computer. One candidate for such a demonstration is boson sampling, a problem in quantum interferometry. In boson sampling, the task is to give a sample from the output of a multimode interferometer fed with single photons. Boson sampling does not follow the normal models of quantum error correction; it is therefore an open question how errors affect the quantum hardness of such a device. The errors in a boson sampler are those which occur in linear optics such as photon losses, photon distinguishability, dark counts, and so on. The question of how boson sampling responds to various imperfections is particularly pressing because there are several platforms for constructing boson samplers, which have different performance characteristics on each imperfection. 

In this talk, I will provide a partial solution to this problem. I will demonstrate a classical algorithm which uses imperfections to classically simulate a boson sampler. The existence of such an algorithm rules out a quantum advantage in any region of the parameter space where it is efficient. 

In particular, I will show how to solve the long-standing problem of simulating a boson sampler with linear loss. I will show an algorithm that can take advantage of linear loss and distinguishability simultaneously, which makes it a useful metric for comparing photon sources. I will compare various single photon sources (including the quantum dots from the Lodahl group) using this algorithm.