Photonic Quantum Computing

Photonic machines store qubits in single particles of light, called photons. For example, a qubit can be which of two paths a photon takes. They compute with beam splitters, phase shifters and detectors. Photons are fast and do not need a super-cold chip. But photons do not naturally interact, and they are easy to lose. So two-qubit logic only works some of the time, and fighting loss is most of the engineering.

How can light be a qubit?

A photon is the smallest packet of light. Light comes in countable lumps, and one lump is a photon. To make a qubit, give one photon a two-way choice.

The most common way is called dual-rail. Two waveguides run side by side. A waveguide is a tiny channel cut into a chip that guides light, like an optical fibre. If the photon travels in the top guide, that means 0. The bottom guide means 1. The photon can also have its amplitude split across both guides at once. (An amplitude is the number that says how strongly it leans toward each answer.) Another common choice is polarisation, the direction the light wave wiggles in.

Picture a train at a fork with two tracks. Unlike a train, a single photon's amplitude can be spread over both tracks until a detector clicks on one of them.

Each separate channel light can use (a path, a polarisation, a time slot) is called an optical mode. Photonic spec sheets often count modes and photons separately. That is one reason their numbers cannot be compared with a superconducting qubit count until you read the definitions.

Here is the deepest difference from every platform so far. Photons do not sit still waiting for pulses. They fly at the speed of light. The computing happens while they stream through the optical circuit. The gates are physical objects the photon passes through. They are not signals sent to a qubit that stays put.