A strong control laser makes the atomic ensemble transparent to the weak signal photon — Electromagnetically Induced Transparency (EIT) opens a narrow transmission window inside an otherwise opaque absorption line. While the control beam is on and the photon enters, its quantum state (amplitude + phase) maps adiabatically onto a collective spin-wave excitation shared across all the atoms — a "dark-state polariton" that slows down and stops as the control beam is ramped off.
Switching the control beam off completes the write: the photon is gone, and its state is parked as a stationary atomic excitation. The medium is now optically opaque again — nothing gets in or out. Switching the control beam back on reverses the process, converting the stored spin wave back into an outgoing photon: the read-out.
Real ensembles are never perfectly isolated. Magnetic-field fluctuations, residual atomic motion and collisions dephase the stored superposition, so retrieval fidelity decays with storage time:
F(t) = 1/2 + 1/2 · e^(−t / T₂)
At t = 0 the read-out is perfect (F = 1). As t grows past a few T₂, the stored state has fully dephased and F settles at 1/2 — no better than a random guess. This decay curve is exactly what sets the practical storage-time limit of a real EIT quantum memory.
- θ, φ — the input photon's polarization state as a point on the Bloch sphere.
- Target storage time — how long the excitation is held before read-out is triggered (or press "Read out now" to interrupt storage early).
- Measured fidelity — computed from the actual (slightly noisy) read-out Bloch vector, so it scatters around the theoretical curve rather than sitting exactly on it.