Distinguishing dynamical regimes in a semiconductor laser with optical feedback by using event-detection methods

In semiconductor lasers, optical feedback generates different types of oscillations in the time trace of the emitted intensity. Depending on the laser and feedback parameters, the emission regimes can be of three main types: stable emission, Low Frequency Fluctuations (LFFs), and Coherence Collapse (CC). In the stable regime, the intensity signal shows noisy fluctuations, while in the LFF regime, these fluctuations are suddenly interrupted by large drops, during which the laser almost turns off. These reset-type events are separated by time intervals of stable emission, where the intensity displays only small fluctuations. In the CC regime, the intensity signal shows highly irregular fluctuations. We analyze experimental intensity time traces, and, by combining three event-detection methods that detect different numbers of events, and using a clustering algorithm, we show that time traces can be qualitatively distinguished within these regimes. We also find that the event-detection methods capture the gradual or abrupt nature of the transitions between the different regimes. Importantly, these methods offer potential to advance the long-standing challenge of distinguishing high-dimensional chaos from noise, since we can differentiate time traces recorded below the laser threshold, where the dynamics is dominated by noise, and well above threshold, where the dynamics is dominated by nonlinearity and feedback. Our results are also relevant to the scientific community working with stochastic point process data (cardiac beats, neuronal discharges, seismic data, extreme precipitation, price spikes, etc.), because we show that event-detection methods that detect different types of events can provide complementary information and can be useful for differentiating or discovering new features of the data.