Coherence collapse
The chaotic state a semiconductor laser falls into under moderate optical feedback: linewidth explodes from MHz to GHz, RIN rises tens of dB, and the light is useless for coherent or analog purposes. The reason transmitter packages contain isolators.
Send a little of a semiconductor laser's own light back into its cavity and the response is not a little perturbation. Because the gain medium couples amplitude to phase (the α-factor) and the external path adds delay, feedback turns the laser into a delayed-feedback nonlinear oscillator, with everything that implies.
The behavior organizes into the five classic regimes (Tkach and Chraplyvy's map, measured on 1.5 µm DFBs; boundaries shift with device, distance, and bias, so treat the dB values as landmarks rather than fences):
| Regime | Feedback (approx.) | Behavior |
|---|---|---|
| I | below −80 dB | Linewidth narrows or broadens depending on feedback phase |
| II | −80 to −45 dB | Mode hopping between external-cavity modes; spectral splitting |
| III | −45 to −39 dB | A narrow window of stable, narrowed single-mode operation |
| IV | −39 to −8 dB | Coherence collapse: chaotic dynamics, GHz-broad line, RIN up tens of dB |
| V | above −8 dB (AR facet) | Strong-feedback external-cavity operation, stable again |
Regime IV is the destroyer. The optical spectrum broadens from megahertz to gigahertz, RIN develops huge structure at external-cavity harmonics and the relaxation resonance, and BER floors appear in links that measured clean moments earlier. The onset threshold drops as grows and as damping falls, which is why low-α, strongly damped designs (and quantum-dot lasers, with their small effective α) tolerate feedback that would collapse a standard quantum-well DFB, and why 'isolator-free' operation is a headline claim when a vendor can make it.
The engineering translations are blunt. A PC connector's −14 dB open-air reflection or even a −30 dB mated interface can sit squarely in regime IV territory for a sensitive DFB, hence isolators inside transmitter packages, APC connectors in analog and PON plants, and return-loss budgets on everything. Regime III's stable window and regime V's strong-feedback branch are not curiosities either: the first is exploited in self-injection-locked narrow-linewidth sources (a high-Q microresonator's controlled backscatter), the second is simply the external-cavity laser working as designed.
The regime boundaries in dB, the conversion from a reflector's return loss to the feedback level the laser sees, and the Helms and Petermann expression for where collapse begins on a given device are collected in optical feedback regimes.