Boson sampler
Realizes: sampling from the permanent of a unitary matrix (classically #P-hard)
Identical single photons enter an m-mode linear optical network (beam splitters and phase shifters implementing a unitary U). Detectors at the outputs sample from a distribution whose probabilities are proportional to |Perm(U_S)|² ā the squared permanent of submatrices of U ā a quantity believed to be classically intractable to compute. Aaronson & Arkhipov (2011) proved that an efficient classical simulation would collapse the polynomial hierarchy. The device does not solve a user-defined optimization problem; rather, it demonstrates quantum advantage on a specific sampling task. Gaussian boson sampling (GBS) variants use squeezed-light inputs and have been demonstrated at scale (Jiuzhang, 2020). Speed: nanoseconds per sample (photon transit time through chip). Capacity: 53+ photons demonstrated (Jiuzhang); quantum advantage claimed for nā„50.
Examples
The Computational Complexity of Linear Optics ā Aaronson & Arkhipov (2011)
Original proof that efficient classical simulation of boson sampling would collapse the polynomial hierarchy