A superconducting nanowire cooled deep into its superconducting phase is biased with a current Ibias just below its critical current Ic, and carries that current with exactly zero resistance. When a single photon lands on the wire, its energy breaks apart a cluster of Cooper pairs, opening a tiny normal-conducting "hotspot" in an otherwise superconducting strip.
hotspot forms -> local resistance R > 0
bias current detours around it -> current density spikes at the hotspot's edges
edge current density > Jc -> hotspot boundary itself goes normal -> hotspot expands
expansion spans the wire's full width -> full cross-section resistive -> V = I_bias * R(t)
- Nucleation & expansion — because the wire is only tens of nanometres wide, a hotspot needs only a few picoseconds to force the diverted current above Jc at its own edges, so it snaps across the entire cross-section almost instantly — that sharp transition is what produces an easily measured voltage pulse rather than a soft blip.
- Dead time — the resistive segment dissipates heat into the substrate and the hotspot cools back below Tc within nanoseconds; until it does, the wire can't register another photon.
- Too low a bias — the detour current at the hotspot's edge never reaches Jc, so the hotspot just re-condenses without ever spanning the wire: a real photon lands and nothing is recorded (a missed count).
- Too high a bias — the bias current is already close to Ic everywhere, so an ordinary thermal phase-slip (no photon needed) is enough to tip a random segment resistive: a dark count. Push Ibias to or past Ic and the wire latches resistive permanently.
This nucleation-and-expansion mechanism is why SNSPDs combine near-unity detection efficiency with extremely low dark-count rates — but only inside a narrow bias window between those two failure modes.