1. Introduction

Gain switching is a convenient pulse generation technique allowing the generation of pulses with pulse durations spanning at least 6 orders of magnitude. As opposed to Q-switching which involves modulating the fractional power loss per round trip, switching the gain of laser transition, a traditional pulsing technique going back to the first demonstrations of lasers, involves directly modulating the population in the energy levels comprising the population inversion independent of the loss of the cavity. By modulating the injection current of semiconductor diode lasers [1] for example, short pulses, down to a few ps duration, can be generated [2] and more recently, using optical pump pulses to excite fiber lasers [3] the generation pulses of a few ns duration [4] has been demonstrated.

Cascade lasing of rare earth ion transitions allows the generation of a number of distinct emission wavelengths simultaneously, which may be an effective method to create power-scaled mid-infrared fiber lasers. Cascading lasing at 2.81 µm and 1.62 µm from Er³⁺ crystalline lasers [5] for example relies on host materials that support less energetic phonons so that adjacent electronic transitions have sufficient luminescence efficiencies to allow a lasing threshold for each transition at a comparable level of pump power. To date, cascade lasing of fiber lasers has relied on low phonon energy fluoride glasses which are capable of supporting fluorescence from the near infrared [6,7] to the mid-infrared [8–10] and cascade lasing on up to three transitions has been demonstrated [11,12]. Cascade lasing reduces the thermal load by creating large photon conversion efficiencies which is particularly important in low phonon glasses which typically have relatively poor thermo-mechanical characteristics [13].

In this paper, we demonstrate a new gain-switching process whereby modulation of the population inversion of a transition is produced from the forced modulation of an adjacent transition of a two transition cascade. We show using cascade lasing of a Ho³⁺-doped fluoride fiber laser at 3 µm and 2.1 µm, that by Q-switching one transition of the cascade, the energy levels of the adjacent transition are sufficiently modulated to cause gain-switched pulsing. Gain-switched pulses of a similar duration to the Q-switched pulses can be produced with a time delay between pulses that is dependent on the pump power and repetition rate of the Q-switch.

2. Experiment setup

The experimental setup is shown in Fig. 1 . Commercial high power at 1.15 µm diode lasers (Ferdinand-Braun-Institut für Höchstfrequenztechnik, Berlin) were used to pump each end of the double clad Ho³⁺-doped fluoride fiber after polarization multiplexing using a standard polarizing beam cube and focusing using two ZnSe objective lenses (Innovation Photonics, LFO-5-6-5, 0.25 NA) which also collimated the ⁵I₆ → ⁵I₇ and ⁵I₇ → ⁵I₈ laser transition outputs from the fiber core. Two highly pump-transmitting dichroic mirrors with 60% reflectivity between 2.0 µm and 2.1µm and >98% reflectivity between 2.5 µm and 3.2 µm were each positioned between the polarizing beam splitter and focusing lens at angle of 15° with respect to the pump beam. One dichroic mirror was used to steer the laser output onto a TeO₂ acousto-optic modulator (AOM) that had a rise of 115 ns; the other mirror was used to outcouple the fiber laser output. The AOM with 89% transmission in the 2.1 µm region and 86% transmission in the 3 µm region was placed into the external cavity of the fiber laser in a first-order diffraction mode arrangement. The measured diffraction efficiencies were 82% and 83% for the 2.1 µm and 3 µm emissions, respectively. To produce ⁵I₆→⁵I₇ output or ⁵I₇→⁵I₈ output simultaneously when its cascade laser operates in the Q-switching state, the fiber end with respect to the AOM was cleaved perpendicularly to form a Fresnel-reflection-based cavity. The Ho³⁺-doped double clad fluoride fiber (FiberLabs, Japan) was identical to the fiber used in our previous free-running experiment [14] and had a D-shaped cladding with a diameter of 125 µm and a numerical aperture (NA) of 0.50. The core was 10 µm in diameter, had an NA of 0.16 and was doped with 1.2 mol% Ho³⁺; the fiber length of 11.0 m provided 98% pump absorption efficiency. A gold-coated plane ruled grating (600 lines per mm, blaze angle θ_B = 17.5°) was used to separate the two cascaded transitions. Two InAs photodiodes (response time of ~10 ns) and a 100 MHz oscilloscope (Tektronix TDS1012) were used to measure the pulse characteristics. An optical spectrum analyzer (Yokogawa AQ6375, Japan) was used to measure the spectrum at 2.1 µm and the spectrum at 3 µm was measured using a calibrated monochromator with 0.3 nm resolution.

 

Fig. 1 Schematic of the experimental setup. D1-D4 represents the pump diodes, pbs the polarizing beam splitter and PD1 and PD2 represent the photo detector. Included is a simplified energy level diagram of the Ho³⁺ ion showing the gain-switching transition driven by the adjacent Q-switched laser transition.

Download Full Size | PDF

3. Results and discussion

The first-order diffracted light from the ⁵I₆ → ⁵I₇ transition was first fed back into the cavity and, by adjusting the repetition frequency of the AOM and pump power, stable pulses arising from the ⁵I₆→⁵I₇ transition were generated, see Fig. 2 , which was measured at the repetition rate of 30 kHz for the maximum launched pump power of 7.4 W. The Q-switched fiber laser operated at a slope efficiency of 12.7% and threshold of 300 mW. The pulse had a duration of 350 ns (FWHM), 21.7 µJ pulse energy and peak power of 62 W. After a time delay of ~2.2 µs, a gain-switched pulse train from the ⁵I₇ → ⁵I₈ transition was generated. The total gain-switched output from both fiber ends operated at a slope efficiency of 4.8% and threshold of 3.4 W and the pulse duration, pulse energy and peak power at maximum launched pump power were 460 ns, 6.4 µJ and 13.9 W, respectively. The gain-switched pulse was approximately 31% longer than the Q-switched pulse which indicates that the mechanism for gain switching is directly related to the Q-switched pulse; the population in the ⁵I₇ level was modulated at the Q-switched pulse generation rate. The gain-switched pulse quickly de-populated the ⁵I₇ level excitation replenishing ground states and therefore ground state absorption (GSA). The pulse-to-pulse stability for both the Q-switched and gain-switched pulses was approximately ± 10%.

 

Fig. 2 Synchronized output pulse trains for the Q-switched ⁵I₆→⁵I₇ pulse and the gain-switched ⁵I₇→⁵I₈ pulse at the repetition rate of 30 kHz for the maximum launched pump power of 7.4 W. The inset shows the temporal pulse waveform for ⁵I₆→⁵I₇ and ⁵I₇→⁵I₈ transitions.

Download Full Size | PDF

Figure 3 shows the repetition rates providing stable pulse generation and the corresponding pulse widths of the pulses for each transition at three launched pump powers. The gain-switched pulses were generated at the same repetition rate as the Q-switched pulses; multiple Q-switch and gain-switch pulsing occurred when the repetition rate of the AOM was lower than 30 kHz; unstable pulsing occurred when the repetition rate of the AOM was larger than 70 kHz. The width of the Q-switched and gain-switched pulses shortened with increasing pump power and lengthened significantly with increasing repetition rate. The inset to Fig. 3 shows the time delay, Δt, between the Q-switched and gain-switched pulses as a function of repetition rate and launched pump power. The time delay decreases and increases near linearly with the launched pump power and repetition rate, respectively. The shortest time delay of 2.2 µs occurred at a repetition rate of 30 kHz and at maximum launched pump power.

 

Fig. 3 Measured pulse width, P_w, of the ⁵I₆ → ⁵I₇ and ⁵I₇ → ⁵I₈ transitions as a function of the repetition rate, R_p at three launched pump powers. The inset shows the time delay, Δt, between ⁵I₆ → ⁵I₇ and ⁵I₇ → ⁵I₈ pulses as a function of R_p at three launched pump powers.

Download Full Size | PDF

Figure 4 shows the output pulses for first-order Q-switching of the ⁵I₇ → ⁵I₈ transition and the resultant gain-switched pulses generated from the ⁵I₆ → ⁵I₇ transition; the inset to Fig. 4 shows the characteristics of each pulse at a repetition rate of 30 kHz and at maximum launched pump power. The Q-switched (⁵I₇ → ⁵I₈) pulse operated at a slope efficiency of 4.0% and threshold of 2.8 W had pulse duration of 300 ns, pulse energy of 6.1 µJ and a peak power of 20.4 W. After a time delay of 9.8 µs, multiple gain-switched (⁵I₆ → ⁵I₇) pulses were measured for all values of the repetition rate of the AOM. The gain-switched pulse durations were 720 ns, nearly twice the duration of the Q-switched pulse. The repetition rate providing stable Q-switched (⁵I₇ → ⁵I₈) pulses was 17 kHz to 47 kHz. The number of gain-switched pulses and time between them decreased with increasing repetition rate. The duration of the gain-switched and Q-switched pulses shortened with increasing pump power and increased with increasing repetition rate; the narrowest Q-switched and gain-switched pulse widths were 270 ns and 640 ns, respectively which was achieved at 17 kHz and 7.4 W launched pump power. The characteristics of the time delay between Q-switched and gain switched pulses was identical to the alternative switching arrangement; the shortest time delay of 7.6 µs was observed at 17 kHz and 7.4 W launched pump power.

 

Fig. 4 Synchronized output pulse trains for the Q-switched ⁵I₇ → ⁵I₈ pulse and gain-switched ⁵I₆ → ⁵I₇ pulse for the maximum launched pump power of 7.4 W. The inset shows the temporal pulse waveform for ⁵I₆ → ⁵I₇ and ⁵I₇ → ⁵I₈ transitions.

Download Full Size | PDF

The optical spectrum of the output at the maximum launched pump power for the Q-switched pulses and the corresponding gain-switched pulses is shown in Fig. 5 . The Q-switched pulses arising from ⁵I₆ → ⁵I₇ (⁵I₇ → ⁵I₈) transition operated with a center wavelength of 3.002 µm (2.074 µm) and bandwidth of 16 nm (4.5 nm), while the gain-switched pulses arising from ⁵I₆ → ⁵I₆ (⁵I₇ → ⁵I₈) transition operated with a center wavelength of 2.986 µm (2.072 µm) and bandwidth of 13 nm (4 nm). The centre wavelength of the Q-switched pulse of a transition is longer than the corresponding gain-switched pulse of the same transition; the spectra of the Q-switched pulses are also broader. The gain-switched pulses were generated from cavities using feedback of approximately 4% Fresnel reflection that forces a higher threshold than the corresponding cavity generating Q-switched pulses. Thus the gain-switched transitions have a larger population inversion at threshold which allows the terminating Stark sub-level to be deeper in the lower laser level thus creating shorter emission wavelengths.

 

Fig. 5 Measured spectrum of the ⁵I₆→⁵I₇ laser transition at maximum pump power. The inset shows the measured spectrum for the ⁵I₇→⁵I₈ laser transition at maximum pump power.

Download Full Size | PDF

A gain-switched pulse has been produced by a Q-switched pulse of the adjacent transition of a two transition cascade. The gain-switched pulses from the lower transition rely on the excitation of the upper laser level from the emission of the Q-switched higher transition pulse. This process is fast which results in a single gain-switched pulse of similar duration to the Q-switched pulse. On the other hand, when the lower transition is Q-switched, the process of gain-switching the higher transition is more complicated. The upper laser level of the higher transition is continuously excited from the diode pump source; however the de-excitation of the lower laser level relies on the generation of the Q-switched pulse. After the generation of the lower transition Q-switched pulse, GSA is increased this creates greater localized excitation of the ⁵I₆ level feeding the gain-switched pulse. There exists, therefore, a complex interplay between the excitation of the energy levels and pulse generation; we are currently constructing a numerical model to understand and further optimize this pulse generation technique. This mechanism can be transferred to passively Q-switched and mode-locked cascade lasers.

4. Conclusion

In this paper we have demonstrated the Q-switched pulse induced gain switching in a two-transition cascade laser. Using a Ho³⁺-doped ZBLAN fibre laser that was pumped with high power CW diode lasers emitting at 1.15 µm and operated cascade on two adjacent transitions emitting at 3 µm and 2.1 µm, gain-switched pulses as short as 460 ns at 2.07 µm were created from 350 ns Q-switched pulses at 3.002 µm. With optimized fibre and cavity parameters in conjunction with increased pumping, shorter pulses are envisaged. The pulse generation mechanism has potential applications in mid-infrared photonics and nonlinear optics.