1. Introduction

The generation of mid-infrared (mid-IR) supercontinuum (SC) radiation from 2 to 5 $\mu$m is getting much attention in recent years owing to its many applications in the field of free space optical communication, range finding, counter measures or remote chemical sensing [1–3]. The strong water absorption in this wavelength region offers also interesting applications in the medical and biomedical field like tissue ablation [4], optical coherence tomography [5] or CARS microscopy [6]. SC generation is nowadays widely investigated to target wavelengths in the mid-IR. It gained additional interest with the developing of optical fibers. Owing to the guiding properties of fibers, it is possible to maintain a high intensity interaction over the entire length of the fiber. The non-linear interaction is therefore not limited to few millimeters (as in the case of gases, liquid cells or crystals). This puts optical fibers in the position of best candidates for supercontinuum generation, offering to the entire system robustness and compactness. To address the spectrum window of 2 - 5 $\mu$m, the use of soft-glass fibers is mandatory because of their transparency at longer wavelengths compared to standard silica which is transparent up to 2.6 $\mu$m. In particular, fluoride glass materials such as ZBLAN (a composition of ZrF$_4$, BaF$_2$, LaF$_3$, AlF$_3$ and NaF) and InF$_3$ are used.

The conversion efficiency during SC generation depends sensitively in which dispersion regime the nonlinear medium is pumped. In the anomalous dispersion region, meaning with a pump wavelength longer than the zero dispersion wavelength (ZDW) of the material, soliton dynamics dominate the nonlinear broadening and superior conversion efficiency can be reached [7]. The zero dispersion wavelength of fluoride fibers usually lies between 1.6 $\mu$m and 2 $\mu$m, dependent on the fiber design making the thulium-doped silica fiber laser emitting at 2 $\mu$m suitable for this application. As a pump source for high-power mid-IR SC generation, pure Q-switching (QS) as pulse generating method is not suitable due to insufficient peak power levels; thus, seed lasers with amplifier chains are very often the preferred choice. Passively mode-locked (ML) fiber lasers with pulse widths in the picosecond range have been demonstrated [8–10]. In the case of this solution, at least two amplifiers are necessary for high-power SC generation, making such setups relatively expensive and complex.

In this research study, a single-oscillator cavity design is chosen to bypass the need of multiple amplifiers and to realize the high-peak-power operation with a single oscillator. An actively Q-switched and mode-locked (QML) Tm$^{3+}$-doped double-clad silica fiber laser is used to pump an InF$_3$ fiber for generating mid-IR high average power SC radiation. In QML mode, the peak power and the pulse energy levels of the ML sub-pulses below the QS envelope are orders of magnitude higher compared to single ML systems making this approach very interesting. The InF$_3$ fiber has been especially selected for high output power levels and efficient SC generation. To the best of our knowledge, this is the first experimental demonstration of a Watt-level range SC generation in an InF$_3$ fiber pumped by a single-oscillator laser system.

2. Experimental setup

The block diagram of the Tm$^{3+}$-doped fiber laser setup is shown in Fig. 1. The active fiber is pumped by two pump diodes. The diode set emits at a wavelength of 792 nm providing a total maximum power of 150 W per module. The light is reflected by two dichroic mirrors which are transparent for the laser wavelength and highly reflecting for the light of the pump diodes under an incident angle of 45°. The pump radiation is then launched into the active thulium-doped fiber by two AR-coated (at the pump and laser wavelength) lenses with a focal length of 15 mm. The thulium-doped fiber length is 2.2 m, with a doping level of 2.8 wt.%. These values ensure a pump light absorption around 10 dB. It has a core diameter of 25 $\mu$m with an NA of 0.08. The fiber is placed in a steel container which is filled with water cooled at a temperature of 19 °C. Two end-caps are fusion spliced onto the tips of the active fiber, in order to reduce the risk of thermal damages. Without a fiber core to confine the beam, the mode field diameter of the beam will diverge within the end-cap; the large diameter reduces the power density and improves the tolerance to end face defects [11]. The fiber has a bend radius of around 10 cm inside the steel container. The bend losses have been estimated to around 0.006 $\frac {dB}{m}$ at a wavelength of 2 $\mu$m. For active ML and active QS, two free-space acousto-optic modulators (AOMs) are used inside the cavity. We chose this configuration, as passive modelockers can lead to QS instabilities and the QS repetition rate in QML mode would be close to the relaxation time of the laser system, so we opted for an active modelocker. The addition of the second active modulator allows for controlling the QS repetition rate. It is possible to obtain a QML system by simply using one single active QS modulator. The necessary mode-locked modulation is then provided by the modulator working with a radio-frequency that corresponds exactly to the free spectral range of the laser resonator. This approach would be the best in terms of compactness and ease of implementation, however is not practically convenient for fiber lasers, as the cavity length has always to be adapted and optimized. The data sheet from the manufacturer states a diffraction efficiency of the AOM for QS of more than 90%, when operated at a fixed radio-frequency of 40 MHz with a build-up time of 500 ns. The AOM for ML provides a diffraction efficiency of at least 10%, depending on the power of the radio-frequency. A diffraction grating with 450 grooves per millimeter and 70% of reflectivity for the $0^{th}$ order at 2 $\mu$m has been used as end cavity mirror. This implementation allows for wavelength tuning, which is obtained by changing the grating orientation around the vertical axis.

 

Fig. 1. Schematic setup of the thulium-doped fiber laser.

Download Full Size | PDF

A block diagram of the setup for SC experiments is shown in Fig. 2. It consists of a first stage to condition and attenuate the 2 $\mu$m free space propagating beam before coupling into the nonlinear fiber. A polarization stage, made by a half wave plate and a thin film polarizer, reflects one polarization of the laser beam and transmits the other one for pumping the nonlinear fiber. The actual attenuation stage allows the setting of the pump power level to inject into the fluoride fiber so that the thulium fiber laser can work at stable high power operation with constant parameters. An afocal zoom optic, made by two AR-coated aspheric silica lenses with focal lengths of 100 mm and 130 mm, respectively, increases the beam diameter of the pump beam for mode field adaptation to the fluoride fiber. A third lens is used to couple the laser beam into the fiber. The 15 m long InF$_3$ fiber has a core diameter of 7.5 $\mu$m and an NA of 0.3. Before coupling the light into the InF$_3$ fiber, an intermediate 26 $\mu$m core and 1 m long single mode ZBLAN fiber is used. The large core of this fiber allows for stable and reliable air-to-fiber coupling. This step has the advantage to overcome the delicateness of direct injection into the InF$_3$ fiber. Indeed, InF$_3$ fibers are less mature from a technological point of view and more sensitive to light propagation in the cladding compared to ZBLAN fibers. The radiation from the ZBLAN fiber is then coupled into the core of the InF$_3$ fiber by means of a commercial fiber-to-fiber coupler allowing to interconnect two FC/APC connectorized fibers. It is made by two collimators with a system of lenses to reduce optical aberrations. The fibers can be aligned and the laser radiation can be focused with high, low coupling losses and high stability during all the experiment. To ensure efficient heat dissipation, the InF$_3$ fiber is placed on an aluminum plate water-cooled at 20 °C. An uncoated calcium fluoride (CaF$_2$) lens with a focal length of 20 mm collects the supercontinuum output radiation before the measurement stage. A long wave pass filter with 3 dB at a cut-off wavelength around 2 $\mu$m can be inserted to measure the supercontinuum radiation excluding the pump radiation. The spectral power distribution is detected with a liquid-nitrogen cooled indium-antimonide (InSb) detector, which is sensitive in the wavelength range between 1 $\mu$m and 5.5 $\mu$m. This detector is placed at the output slit of a 320 mm focal length monochromator with a grating with 300 lines/mm blazed at 2 $\mu$m. The signal from the detector is treated by a preamplifier and a lock-in amplifier. The supercontinuum output radiation reaches the monochromator by means of a 500 $\mu$m diameter hollow-core fiber with an inner coating made by silver for total internal reflection.

 

Fig. 2. Schematic of the experimental setup used to generate SC radiation.

Download Full Size | PDF

3. Experimental results and discussion

The output power of the fiber laser versus the pump power is shown in Fig. 3 in continuous wave (CW) and QML operation. The fiber laser provides an output power level of 45 W in CW operation with a slope efficiency of 58%. In the QML regime, the maximum average output power level is 40 W, for a QS repetition rate of 150 kHz. The curves in QML operation, compared to the CW performance, indicate a rollover behavior at certain pump power levels. The reasons are most likely nonlinear effects of the silica fiber stimulated by the high peak intensities of the QML pulses. For instance, the Raman scattering transfers optical power to a wavelength around 2.2 $\mu$m [12], which has been experimentally verified.

 

Fig. 3. Output power versus launched pump power in QML operation for different QS repetition rates.

Download Full Size | PDF

For the SC experiment the pump laser operated at an output power level of 15 W with a QS repetition rate 60 kHz for stable operation, resulting in a total pulse energy of 220 $\mu$J. Figure 4 reports the QML pulse train. At this output power level and QS repetition rate, about 10 ML pulses are enveloped by the QS modulation and the QS envelope duration is around 90 ns. The repetition frequency of the ML subpulses has been measured and equals 37.9 MHz which corresponds to an optical cavity length of 3.9 m. The ML pulse duration has been measured to about 30 ps. The QML pulse train and the single ML pulse have been measured with a photodetector and an oscilloscope with an electrical bandwidth of 22 GHz and 50 GHz, respectively. The results of the SC experiment with the InF$_3$ fiber are shown in Fig. 5. In total, 7 W of SC output power have been reached, with 2 W beyond the 2.15 $\mu$m optical long wave pass filter. The output spectra at different launched pump power levels are displayed in the right part of Fig. 5. At the highest launched pump power (10 W), a long wavelength edge of 4.7 $\mu$m has been measured. The mechanism of SC generation pumped by picosecond pulses in the anomalous dispersion region is started with self-phase modulation and modulation instability. The pump pulse is broken into multiple femtosecond sub-pulses. Then, these ultrafast sub-pulses are red-shifted towards the longer wavelength region via complicated soliton dynamics such as soliton fission and Raman soliton self-frequency shift. The spectral distribution measured during the experiment is not flat as expected. It is very difficult to compare the achieved efficiency with other results reported in literature because of the different pump laser systems. There are no reports so far of QML laser systems used for supercontinuum generation in this material. A possible limitation of the conversion efficiency in the here presented work could be that in our setup, the supercontinuum output radiation is a superposition of the different spectra generated by all the different ML pulses. The pulses at the temporal edges of the QS envelope are less energetic and intense than the ones close to the QS envelope peak, making the overall SC generation efficiency lower.

 

Fig. 4. QML pulse train measured at an output power of 15 W and a QS repetition rate of 60 kHz.

Download Full Size | PDF

 

Fig. 5. Left: SC output power versus launched pump power in all bands and with two optical long wave pass filters; Right: Output spectra for three different launched pump power levels.

Download Full Size | PDF

4. Conclusion

In conclusion, mid-IR SC generation has been demonstrated in an InF$_3$ fiber using a QML Tm$^{3+}$-doped silica fiber laser single-oscillator as pump source. A new design of injection system, consisting of a large core diameter ZBLAN optical fiber and a commercial fiber-to-fiber coupler, allowed to enhance the thermo-mechanical stability of the fiber. The SC radiation generated in InF$_3$ showed an output spectrum spanning up to around 4.7 $\mu$m with an output power level of 7 W in all spectral bands.