Build up of Laser Action in Random Fibre Lasers

The feedback mechanism in Fibre Random Lasers (RFL) is the Rayleigh Backscattering along the fibre, which is amplified within the active section of the laser. Amplification can be distributed along the same fibre that provides the random feedback or can be lumped in a restricted section of the open cavity. Raman and Brillouin amplification are examples of distributed-gain RFLs, whereas a pumped Erbium-doped fibre section or a semiconductor optical amplifier (SOA) are examples of lumped-gain RFLs. A fully open cavity requires a double scattering mechanism to account for the feedback, which imposes high pump thresholds for laser action. However, a single mirror or a Bragg Grating can be used at one side so that the single backscattering in the fibre plus the reflection provide the round-trip feedback in a half-open cavity, which still keeps the required randomness with much lower threshold pump powers.

The half-open SOA-based RFL, where the length of the gain section is negligible, is the simplest system to investigate the build-up of lasing action in RFLs and their statistical properties. The RFL used here consisted of an SOA connected to a Fibre Bragg Grating on one side and an 8-km single mode fibre (SMF) on the other. The output of the SMF was connected to a fast photodetector via an optical isolator and a passband optical filter, tuned to the FBG wavelength in order to eliminate the ASE contribution. Two FBGs with bandwidths 0.48 and 1.1 nm were used for comparison. Near threshold this FRL operates in pulsed mode due to random fluctuations in the backscattered light intensity [1], whereas for higher currents it starts to operate in a multimode condition with random intensity fluctuations due to the randomness of mode competition and feedback properties [2]. Here we present measurements performed in time domain to investigate the statistical properties and build up processes near and above the threshold of this SOA-based RFL.

By modulating the bias current with square pulses above the threshold current the laser was set above threshold and the firing of a laser pulse could be registered. The statistics of the time delay between the start of the current pulse and the laser firing could be measured for different levels of excess gain above threshold, generated by the current pulses. For small excess gain values, it was observed that the cumulative distribution of time-to-fire grows linearly with time, consistent with a uniform distribution of occurrence of backscattered light spikes along time. The growth rate was observed to be proportional to the spectral bandwidth of the FBG, which defines the probability of occurrence of a Rayleigh backscattered spike. However, a minimum time-to-fire was observed, which was observed to decrease with increasing excess gain values being the same for both FBGs. This time lag is consistent with the number of round trips needed to deplete the gain and achieve laser action, the greater the gain the smaller the number of round trips for lasing. For higher excess gain levels, the laser approaches the multimode quasi-continuum regime and the statistics of the firing time can be described by the gain depletion behaviour of a long cavity fibre laser where the average cavity length is given by half the fibre length used for random feedback.

In conclusion, we observed that the length of RFL, together with the number of roundtrips needed to achieve lasing action, are the responsible for the minimum time-to-fire in the low excess gain condition. They are also responsible for the time lag between the current pulse and the build-up of the laser pulse at high gain. The fluctuations in the Rayleigh backscattered intensity generated by the feedback fibre give rise to a uniform distribution of spike probabilities which are responsible for the randomness of the time-to-fire distribution of the RFL in pulsed regime as well as the randomness of intensity fluctuations, together with mode competition in multimode operation.