Introduction

Pulsed lasers are extremely useful for applications like frequency conversion, or marking of plastics because of the higher peak power compared to cw-lasers [1]. Additionally for these applications a good beam quality is important that can be accomplished by light propagation in single-mode fibers. Therefore, it is very promising to combine the Q-switching technique with the advantages of fiber lasers.

Active Q-switching is typically achieved by inserting an acousto-optic or an electro-optic modulator into the cavity [2]. However, such a set-up is relatively complicated and expensive due to the active optical switch and the related electronics. In contrast to active methods, the passive Q-switching by insertion of a bulk saturable absorber is relatively simple and shows a high reliability. Commonly, saturable absorbers on the basis of Cr⁴⁺-doped crystals or ceramics were used for Q-switching of Nd³⁺- or Yb³⁺-doped infrared lasers with emission wavelengths around 1 µm [3]. However, for fiber-based laser systems the use of such bulk elements results in a larger complexity in contradiction to a needed compact and reliable integrated set-up. Therefore, the advantages of all-fiber laser systems like compact set-up, low adjustment requirements and low sensitivity to thermal effects and mechanical perturbations can not fully be exploited in such laser systems. Recently a Cr⁴⁺-doped fiber was produced the first time. By splicing this fiber to a Nd³⁺-gain fiber a compact passive Q-switched fiber laser emitting at 1.084 µm was realized [4].

In this work we present for the first time a Q-switched wavelength-tunable narrow-linewidth diode-pumped Yb³⁺-fiber laser. This set-up overcomes the previous mentioned restrictions by using a Tm³⁺-codoping in the gain fiber as a saturable absorber. By this way higher integrated Q-switched laser set-ups can be realized in the future. For our system the influence of the pump power, the fiber length and the outcoupling ratio on repetition rate, pulse duration and pulse energy was investigated. Additionally, this system was widely tunable in wavelength and showed a narrow emission bandwidth. This makes this system very useful for quasi-cw pumping of optical parametric oscillators (OPO), since this enables increased peak power in combination with a small linewidth and a wide wavelength tunability [5].

2. Experimental set-up

Figure 1 shows the detailed laser set-up, which consists of the following main components: a Yb³⁺-doped double-clad fiber with Tm³⁺-codoping that provides the gain and the saturable absorption, quarter- (QWP) and half-wave-plates (HWP), a Faraday rotator (FR) for unidirectional laser operation and a grating pair (1200 lines/mm, gold coated) in Littrow-Littman configuration for tunable narrow-linewidth operation. The polarization beam splitter (PBS2) at the input of the FR acted as a variable output coupler in combination with HWP2. In addition, a polarization beam splitter (PBS1) was placed in front of this HWP2 to fix the polarization state and therefore the outcoupling factor at the PBS1. Otherwise the cavity loss could be reduced by nonlinear polarization effects at high outcoupling factors which would favor polarization rotation mode-locking. The gain fiber with a dopant concentration of 1500 ppm Yb₂0₃ and about 5 ppm Tm₂O₃ had a pump and laser core diameter of 250 µm and 7 µm, respectively. For this fiber with a cut-off wavelength below 1,0 µm a D-shaped pump core was chosen for improved pump light absorption.

 

Fig. 1. Set-up of the tunable Q-switched fiber laser system.

Download Full Size | PDF

The light from the fiber was steered by mirrors M1 and M2 through a polarization beam splitter and to the variable output coupler consisting of HWP2 and the PBS2 in front of the FR. Behind the FR, the signal propagated towards the grating pair. The wavelength of the back reflected signal was determined by adjustment of the incident angle on the Littrow tuning grating. For the back-reflected signal the FR in combination with the PBS2 acted as an optical circulator and the beam was then coupled into the Yb³⁺-fiber via mirror M3. The polarization at the fiber input was determined by the settings of the zeroth-order (1064 nm) wave-plates (HWP1 + QWP).

3. Results and discussion

Using a set-up with a 20 m long gain fiber and an output coupling factor adjusted to 65% stable pulsed operation started at 5.6 W launched pump power for appropriate waveplate settings. At this threshold the laser showed a repetition rate of 20.1 kHz and an output power of 33 mW while the wavelength was kept fixed to 1065 nm. At maximum pump power of 21.9 W an output power of 1.44 W at a repetition rate of 98.1 kHz was achieved. The output power and the repetition rate versus pump power are plotted in Fig. 2(a). The output power did not increase linearly with pump power but showed a S-shaped curve. This was due to a thermally induced increase of the pump diode wavelength with pump power. While the laser diode case was mounted on a heat sink with constant temperature the laser emitter temperature, and therefore the emission wavelength, increased with pump power, because of the increased heat generation. In combination with the relatively narrow absorption peak of Yb³⁺ at 975 nm, the pump wavelength moves closer to the absorption peak with increasing pump power resulting in an improved pump light absorption. This was confirmed by measurements of the transmitted pump power which showed that the pump light absorption increased with pump power especially at medium power levels.

Figure 2(b) shows the pulse energy as a function of the pump power. The pulse energy was calculated from the measured output power and repetition rate. The pulse energy increased nearly linear from 1.6 µJ at 5.6 W pump power to a maximum of 15.1 µJ at 19.7 W pump power. For even higher pump powers the pulse energy decreased slightly with pump power.

Furthermore, the pulse width (FWHM) as a function of the pump power, measured with a 2 GHz photodiode and a 500 MHz oscilloscope, is shown in Fig. 2(b). The pulse width decreased asymptotically from 6.5 µs to 1.1 µs with increasing pump power. Typically the Q-switched pulses showed a substructure above the µs pulse pedestal with a periodicity equal to the cavity round trip time (left diagram in figure 3). The width of this substructure pulses in the oscilloscope trace was roughly equal to the resolution limit of the oscilloscope of about 2 ns. The pulse width of the µs pulse pedestal was determined by reducing the detection bandwidth electronically to about 20 MHz and therefore smoothing the signal. It can be seen from Fig. 3 (red curve) that the 20 MHz trace nearly coincides with the 500 MHz resolution curve, except for the ns peaks. This trace show, that the contribution of the ns peaks to the total energy of the µs pulses is quite small (<5%). It should be noted, that the amplitude of the substructure depended on waveplate settings and the coiling (twist) of the fiber and that also Q-switched operation without such a substructure was observed. A typical photodiode trace of the output signal on a larger time scale is shown in the right diagram of Fig. 3. The variations of the pulse amplitude in this diagram are attributable to the discrete sampling of the digital oscilloscope in combination with the pulse substructure. No variations of the peak amplitude were observed on an analog oscilloscope with 20 MHz bandwidth.

 

Fig. 2. Output power, repetition rate, pulse width and pulse energy as a function of the pump power.

Download Full Size | PDF

 

Fig 3. Left diagram: 500 MHz Oscilloscope trace (black) of a Q-switched pulse which shows the nssubstructure above the µs pulse pedestal. The red curve shows the averaged time signal (averaging time 5 ns). Right diagram: Typical “long-time” photodiode trace of the Q-switched fiber laser output signal.

Download Full Size | PDF

Variations of the output coupling (OC) from 0 to 90% resulted in up to 20% increase of the pulse width relative to the pulse width minimum which was observed for OC’s between 50% and 65%. Furthermore, the output power increased roughly linearly with the OC factor, while its influence on the repetition rate was below 5%. Only unstable Q-switching was achieved for extremely high output coupling factors (i.e., >90% at 15 W pump power).

In order to investigate the influence of a larger fiber length on the laser performance a second set-up with a 32 m long gain fiber was realized. The results for this set-up were qualitatively the same as for the previous set-up. The main output data of both set-ups are shown in Table 1. It can be seen from this table, that for the longer fiber the lasing threshold was nearly two times lower and the slope efficiency was about 60% higher due to an improved pump light absorption. Furthermore, for the 32 m set-up the maximum output power was 67% and the pulse energy up to 44% higher while the minimum pulse duration was about 20% longer than for the 20 m gain fiber set-up. It should be noted, that in combination with the 32 m fiber the maximum pulse energy of 21.8 µJ was observed at a pump power of 15 W. A further increase of pump power to 16.2 W caused a strong increase of the repetition rate and therefore a decrease of the pulse energy which could be seen from the laser data shown in the last row of Table 1. For even higher pump powers the Tm³⁺-codoping caused only a periodic intensity modulation and no on/off Q-switching was observed.

Using a Yb³⁺-doped fiber without an additional Tm³⁺-doping in the set-up, we could achieve only a cw output or random fluctuations but no stable pulsed operation. This indicated, that the pulsed operation is due to a saturable absorption effect of the codoped Tm³⁺-ions. In addition, during infrared laser operation, the codoped gain fiber shows a blue fluorescence. The brightness of this spontaneous emission at about 480 nm increased with signal intensity and was much stronger than for conventional Yb³⁺-doped fibers.

The blue emission of the Tm³⁺-ions, pumped at 1.1 µm is caused by a three step upconversion process [6]. Therefore it showed that the signal is partially absorbed by the Tm³⁺-ions. In combination with the observed pulsed operation this indicated that the Tm³⁺-codoping act as a saturable absorber. This assumption was strongly supported by numerical calculations, which take the energy structure and the various transitions of the Tm³⁺-ions into account (Fig. 4 left). The calculated time dependence of the absorption curve, which clearly shows saturable absorption, and the corresponding signal in the cavity are shown in the right diagram of Fig 4. For these calculations a circulating signal with 20 µJ pulse energy, 1.5 µs pulse width and 65 kHz repetition rate was assumed in the 32 m fiber. These simulations also showed, that the saturable absorption is mainly due to the ³F₄→³F_2,3 transition. This is due to the strong absorption of this transition, the long lifetime of the ³F₄ level and to the decay time from the upper to the lower level of about 15 µs, which is of the same magnitude as the time interval between subsequent pulses. The main effect of the ³H₆→³H₅ transition is the population of the ³F₄ level and therefore activation of the previous mentioned transition. In contrast, the ³H₄→¹G₄ transition reduces the population of the ³H₄ level which also results in a lower population of the ³H₅ level. By this way it decreases the absorption by the ³F₄→³F_2,3 transition. In these simulations of the Tm³⁺ population and the absorption dynamics a (unsaturated) ground state absorption of the Tm³⁺-ions of 0.5 dB was assumed for the 32 m fiber. This estimation was based on absorption measurements on a Tm³⁺-doped fiber. Furthermore, the ratio between the absorption cross section of the first and the second transition was estimated to 1:30 based on the absorption spectra of a Tm³⁺-fiber and on the information that the oscillator strength of the ³F₄→³F_2,3 transition is similar to the ³H₆→³H₄ transition [7]. For the ³H₄→¹G₄ transition it was roughly estimated that it is about three times higher than the ground state absorption. Variations of these ratios showed that the qualitative behavior of the absorption curve was quite insensitive to these ratios. Therefore a saturable absorption effect of the Tm³⁺-ions is confirmed by these simulations despite of uncertainties for the absorption cross sections.

Table 1. Laser data for set-ups with a 20 m and a 32 m long gain fiber

View Table

The effect of the Tm³⁺-ions on laser operation of the Yb³⁺-fiber is similar to the saturable absorption effect of Er³⁺-clusters in heavily doped Er³⁺-fibers which also favors self-pulsing operation [8]. For these systems the loss modulation of the Tm³⁺-ions or Er³⁺-ion clusters, respectively, is weak relative to fiber gain. Hence the passive modulation amplitude is relative small relative to the fiber gain which is also indicated by the above mentioned simulations (~3% modulation in Fig. 4(b). However, the pulse duration in the µs range and the observed high duty cycle were an indication that laser operation was also influenced by relaxation oscillations. Calculation of the relaxation oscillation frequency based on standard relaxation theory showed a good agreement between the calculated and the observed repetition rate as a function of the pump power for both set-ups [11]. For these calculations a cavity round trip loss of 90% was assumed and the lifetime of the upper Yb³⁺-level was calculated by taking into account the emitted signal power, the pump and signal intensities and the absorption and emission cross sections. The deviations between calculated and measured repetition rates were below 10%. Such a relation was also previously reported for heavily doped Er³⁺-fibers [12]. This correspondence and the above mentioned properties of the Tm³⁺-codoping leads to the conclusion that the saturable absorption of the Tm³⁺-ions initiates and sustains relaxation oscillations in the Yb³⁺-fiber. Consequently repetition rate and pulse build up can be calculated by standard relaxation theory for different doping concentrations and fiber lengths.

 

Fig. 4. Energy level diagram for Tm³⁺-ions (left). Pump transitions and fluorescence are indicated by vertical dashed arrows. Nonradiative decays which are relevant for the upconversion process (solid) or which are taken into account for the calculations (dashed) are indicated by diagonal arrows. For the pump transitions the wavelength of the absorption peaks are shown. Lifetimes which were determined by calculation of the nonradiative decay rates are in brackets (same relation between energy gap and nonradiative decay rate as in Er³⁺-silica fibers assumed [10]). Lifetime of the ¹G₄ level was determined from the exponential decay of the blue fluorescence. Right diagram shows the calculated Tm³⁺-absorption (upper curve) of the 32 m fiber. The lower curve shows the corresponding intra-cavity signal (20 µJ, 1.5 µs, 65 kHz).

Download Full Size | PDF

For both set-ups the Q-switched operation was influenced by the waveplate settings which is due to the following effects. On the one hand the resonator loss depends on the birefringence of the fiber and the waveplates. On the other hand, the polarization evolution in the fiber depends on the signal intensity for an elliptically polarized signal due to the Kerr-effect, which can be used for mode locking [13]. In standard gain fibers the birefringence shows random fluctuations which results in an elliptical polarization state somewhere in the fiber. Hence the typically observed ns substructure of the µs pulse is likely due to nonlinear polarization effects in the fiber. Both effects could be avoided by a polarization maintaining (PM) gain fiber. For such a fiber no birefringence compensation by waveplates is required and the linear polarization along the fiber avoids nonlinear polarization rotation effects.

By rotating the tuning grating the signal wavelength was tunable between 1055 and 1085 nm (1059 and 1090 nm) for the 20 m (32 m) set-up. This was measured at a pump power of 16.8 W (13.7 W). The spectra, measured with a resolution of 0.5 nm, showed also that the spectral intensity of the amplified spontaneous emission was at least 37 dB lower than the main laser radiation. If the tuning grating position was adjusted for a emission wavelength outside the above mentioned tuning limits only unstable Q-switching was observed. In the wings of the tuning range the repetition rate and the pulse width increased typically up to a factor of two. The maximum output power was measured at about 1065 nm and decreased only slightly (< 15%) if the wavelength was detuned. However at the long wavelength limit for the 32 m fiber a power drop of up to 35% was observed.

In order to determine the linewidth of the laser, a confocal scanning Fabry-Perot interferometer with 10 GHz free spectral range and a finesse of 150 was used. The measured linewidth increased for the pump power ranges mentioned in table 1 from 0.17 GHz to 2.3 GHz (from 0.35 GHz to 4 GHz) with pump power for the 20 m (32 m) set-up. This increase was nearly linear at low pump powers and showed an asymptotic behaviour at higher pump power levels. Regarding the linewidth it should be noted that for cw-operation of a similar set-up with a conventional 20 m long Yb³⁺-fiber nearly the same linewidth (<2.5 GHz) was observed [14].

4. Conclusion

To our knowledge we realized the first tunable Yb³⁺-fiber laser passively Q-switched by a Tm³⁺-codoping of the gain fiber. Q-switched operation (up to 21.8 µJ, 140 kHz, 1.1 µs), in combination with narrow linewidth operation (<4 GHz) and wavelength tunability between 1055 and 1090 nm was demonstrated. It was shown, that the pulsed output of this system is due to relaxation oscillations which were sustained by saturable absorption of the Tm³⁺-ions. Detailed simulations of the Tm³⁺-absorption dynamics and variation of the concentration and the axial distribution of the Yb³⁺- and the Tm³⁺-doping for further optimization of the laser’s Q-switching performance are in progress.