Broadband laser radiation covering the 3–5 μm spectral range is finding applications in a wide range of disciplines. In spectroscopy, the mid-infrared (mid-IR) region where most gaseous chemical species have fundamental ro-vibrational absorption lines [1] is of great interest. The shorter wavelengths, from 3–4 μm, include the fundamental C-H stretching resonant frequency found in all hydrocarbons [2]. At wavelengths longer than 4 μm, carbon oxides possess strong absorption bands (i.e., CO between 4.5–4.9 μm and ${\mathrm{CO}}_2$ near 4.2 μm). In addition, the 3–5 μm spectral region also overlaps with an atmospheric transmission window, which enables remote sensing applications such as Lidar spectroscopy [3]. Furthermore, mid-IR lasers operating in this atmospheric window are key components in the development of infrared countermeasure systems (defense and security) [4].

Supercontinuum (SC) generation in soft glass (e.g., fluoride or chalcogenide) fibers can efficiently produce a broad mid-IR spectrum. While several demonstrations of fluoride-based SC sources have been reported in the last few years, there is now a growing need to simplify their designs and improve their reliability. This can be achieved by limiting the number of free-space components and by avoiding complex pump setups such as solid state laser pumped optical parametric amplifier (OPA) systems. One successful approach is based on the amplification of picosecond/nanosecond pulses in doped fiber gain media. The seed pulses can be generated either with $Q$-switched fiber lasers, pulsed laser diodes, or optical parametric generation (OPG) sources. Seed pulses are then launched in an anomalous dispersion fiber in which they are split into multiple shorter pulses by modulation instabilities and then broadened in the mid-IR through Raman self-frequency shift (RSFS). Among the best results, a 21.8 W average power supercontinuum spanning 1.9–3.8 μm was generated by launching 16 ps pulses amplified with a thulium-doped fiber amplifier at 2 μm [5]. Using a similar approach, a SC extending to 4.75 μm was also reported [6]. In another demonstration, 1.55 μm pulses were sequentially amplified in erbium- and thulium-doped amplifiers before being launched in a fluoride fiber to generate a spectrum from 1.9 to 4.5 μm [7]. More recently, a 2.6–4.1 μm supercontinuum was entirely generated in an erbium-doped fluoride fiber amplifier (at 2.75 μm) when seeded with 400 ps pulses from a diode-pumped OPG source [8].

Since these demonstrations were done with fluorozirconate $({\mathrm{ZrF}}_4)$ glass fibers such as ZBLAN, the maximum broadening was limited to $<4.75\mathrm{\mu m}$. To overcome this limitation and broaden the spectrum over 5 μm, fluoroindate $({\mathrm{InF}}_3)$ glass has emerged as a very promising alternative. In fact, these low phonon glasses, i.e., with a maximum phonon frequency of $\sim 510{\mathrm{cm}}^{-1}$ [9], compared to $\sim 579{\mathrm{cm}}^{-1}$ for ${\mathrm{ZrF}}_4$ based glasses [10], have a transparency window that reaches a loss of 1 dB/m at 5 μm [11], which is roughly 1 μm above ${\mathrm{ZrF}}_4$-based glasses. To date, however, the previous demonstrations were not able to benefit from the full potential of ${\mathrm{InF}}_3$-based glasses, with measured spectrum barely extending over the transparency window of ${\mathrm{ZrF}}_4$ glasses [12–14]. One of the best results obtained in an ${\mathrm{InF}}_3$ fiber was the generation of a supercontinuum from 2.7–4.7 μm while pumping with 3.4 μm femtosecond pulses from a complex OPA system [14]. Extending SC beyond 4.5 μm is important not only for direct light generation, but also to pump directly chalcogenide-based glass fibers near their zero dispersion wavelength (ZDW) to extend the SC spectral coverage further in the mid-IR [15].

In this Letter, we report a 2.4–5.4 μm SC source consisting of an erbium-doped fiber amplifier directly spliced to a low-loss fluoroindate fiber seeded by a 2.75 μm OPG source. It has a maximum average power $\sim 10\mathrm{mW}$ with up to 82% of the power located in the mid-IR $(>3\mathrm{\mu m})$. To the best of our knowledge, this is the first demonstration of a compact ${\mathrm{InF}}_3$-based SC source having a spectrum extending well beyond 5 μm.

Figure 1(a) depicts the supercontinuum generation setup which consists of an erbium-doped fluoride fiber amplifier spliced to a low-loss fluoroindate fiber. To benefit from a fusion-spliced interface between the fibers, a co-directional (forward) pumping configuration was selected, even though it typically provides a reduced gain compared to a counter-propagating (backward) scheme [16]. The amplifier is pumped by a 970 nm multimode diode (6 W continuous wave (CW), Alfalight, model AM6-976B-10-604) and seeded with a 2.75 μm OPG source (2 mW avg power, 400 ps, 2 kHz, Light Matter Interaction, Inc., model PIRL-VLP-2800) which consists of a diode-pumped passively $Q$-switched microchip laser followed by a PPLN nonlinear crystal [17]. Considering the large FWHM bandwidth of the seed (Fig. 3, ${\mathrm{P}}_{970\mathrm{nm}}\text{:}0\mathrm{mW}$), the 400 ps pulses are clearly not Fourier transform limited. Both seed and pump beams are coupled in the doped fiber through a ZnSe aspheric lens (L2, $\mathrm{f}=12.7\mathrm{mm}$) with respective launching efficiencies of 15% and 50%, including lens losses, DM losses, and mode coupling efficiency.

 

Fig. 1. (a) Experimental setup of the ${\mathrm{InF}}_3$-based SC source pumped by an ${\mathrm{Er}}^{3+}\text{:}{\mathrm{ZrF}}_4$ fiber amplifier. (b) Image of the fusion splice between the ${\mathrm{Er}}^{3+}$-doped ${\mathrm{ZrF}}_4$ fiber and the ${\mathrm{InF}}_3$ fiber (L1, L2, coupling lenses; DM, dichroic mirror).

Download Full Size | PDF

The amplifier is made of a 1.25 m long ${\mathrm{Er}}^{3+}\text{:}{\mathrm{ZrF}}_4$ double-clad fiber (Le Verre Fluoré). Its core has a 16 μm diameter (numerical aperture $(\mathrm{NA})=0.12)$ and an erbium concentration of 7 mol. %. The 260 μm cladding has a circular symmetry broken by two parallel flats and is coated with a low-index fluoroacrylate polymer to enable multimode pump propagation $(\mathrm{NA}>0.46)$. A short amplifier length was selected to raise the parasitic CW lasing threshold [8] and to limit spectral broadening in the amplifier, therefore avoiding the loss band of the ${\mathrm{ZrF}}_4$-based fiber (i.e., material and confinement losses).

The doped fiber is then spliced [Fig. 1(b)] to a piece of fluoroindate fiber (Le Verre Fluoré) for further spectral broadening of the amplified pulses. This ${\mathrm{InF}}_3$ fiber has a 100 μm diameter cladding and an average core diameter size of 13.5 μm (increasing from 12.5 to 14.5 μm along the 31 m fiber). With a NA of 0.3, it is slightly multimode for most of the wavelength range covered by the supercontinuum (the average cutoff wavelength is $\sim 5.3\mathrm{\mu m}$). Its attenuation spectrum is shown in Fig. 2, along with the spectrum of a standard ${\mathrm{ZrF}}_4$ fiber (Le Verre Fluoré, 6.5/125 μm). The fluoroindate fiber used in the current study presents an attenuation coefficient below 30 dB/km for all wavelengths between 2.2 and 4.2 μm, with a minimum value of 12 dB/km at 3.8 μm and a low OH absorption peak of 28 dB/km. Such values are significantly lower than the previous recent reports (for example, a minimum absorption coefficient of 640 dB/km at 4.1 μm was reported in [12–14] with an OH absorption peak of 3700 dB/km), which allows us to use significantly longer fiber lengths (15 versus $\sim 0.5\mathrm{m}$ in [12–14]) to enhance the SC spectral coverage.

 

Fig. 2. Attenuation spectrum of the 13.5/100 μm ${\mathrm{InF}}_3$ fiber compared to a standard 6.5/125 μm ${\mathrm{ZrF}}_4$ fiber from Le Verre Fluoré.

Download Full Size | PDF

The fusion splice between the ${\mathrm{ZrF}}_4$ and ${\mathrm{InF}}_3$ fibers is made using a filament-based splicer (Vytran, model GPX-3000). It should be noted that these two fluoride glass compositions have a similar transition temperature $({\mathrm{T}}_g\approx 300°\mathrm{C})$ [17]. Nevertheless, the splicing point (i.e., center of the filament) was offset toward the ${\mathrm{ZrF}}_4$ fiber to account for the significant difference in fiber diameters. The splice losses, estimated at 20%, are mainly attributed to the mode mismatch between the fibers. To reduce feedback in the amplifier, input and output fiber tips were angle cleaved at respective angles of 5° and 14°, considering the different NA of the two fibers.

A grating-based monochromator (Digikrom, model DK480) coupled to a liquid nitrogen-cooled InSb detector (Judson, model J10D-M204-R01M-60) is used to measure the output SC at a resolution of 4 nm. A broadband AR-coated germanium window is placed before the detection system to filter out the residual signal below 1.8 μm. A long-pass dichroic mirror suppressing all spectral components below 3.2 μm is also added when the signal is broadened beyond 5 μm. In this way, we prevent the second-order diffraction at shorter wavelengths from overlapping with the long wavelength side of the SC. The spectral response of the spectrometer was evaluated with a blackbody source (Ocean Optics, model Cool Red) and was used to correct all spectral measurements. This correction was required since the detector’s response slowly decreases at wavelengths above 4.5 μm. Finally, the SC average output power was monitored with a thermopile detector (Gentec, model EO XLP12-3S-H2) to calibrate the power spectral density curves.

The amplifier output spectrum was measured before performing the splice with the ${\mathrm{InF}}_3$ fiber. As shown in Fig. 3, the erbium-doped fiber generates some spectral broadening up to 3.1 μm ($-20\mathrm{dB}$ level) at maximum pump power. In fact, spectral broadening appears to start as peak amplification stops for a 970 nm pump power level beyond $\sim 550\mathrm{mW}$.

 

Fig. 3. Spectrum of the 1.25 m ${\mathrm{Er}}^{3+}\text{:}{\mathrm{ZrF}}_4$ amplifier at different launched pump powers from 0 to 978 mW.

Download Full Size | PDF

Spectral broadening in the fluoroindate fiber was then investigated as a function of both pump power and fiber length (31, 15, and 1 m). Resulting spectra for the two shortest lengths are presented in Figs 4(a) and 4(b). As can be seen, a low pump power of a few hundred milliwatts is generally enough to generate a significant spectral broadening. It is also clear that the SC builds up mostly in the first tens of centimeters as the 1 m ${\mathrm{InF}}_3$ fiber already produces a spectrum extending from 2.5 to almost 5 μm. The broadest spectra, however, were achieved with the 15 m fiber. At a pump power of 978 mW, a SC spanning more than one octave is measured from 2.4 to 5.4 μm ($-20\mathrm{dB}$ level), with 82.3% of its energy beyond 3 μm. We also observe a large peak in the 4.1–4.2 μm area that grows stronger when increasing either fiber length or input power. For the maximum pump level, more than half of the total output power is contained within this peak. At wavelengths longer than 5.2 μm, the steep decrease of the spectral density is a result of the ${\mathrm{InF}}_3$ fiber transparency limit, where material losses sharply increase (see Fig. 2). This was confirmed when testing the 31 m long ${\mathrm{InF}}_3$ fiber, for which the output SC was essentially identical (to the 15 m experiment), except for a lower energy level above 5 μm. It is worth noting that the steep signal drop observed at 4.2 μm is caused by the absorption lines of atmospheric ${\mathrm{CO}}_2$ in the spectrometer. Since this absorption band hides a portion of the high energy peak, the missing energy had to be accounted for in the power spectral density calibration.

 

Fig. 4. Spectrum after propagation in (a) 1 m and (b) 15 m ${\mathrm{InF}}_3$ fibers at different launched pump powers from 77 to 978 mW.

Download Full Size | PDF

The SC output power as a function of the launched pump power is presented in Fig. 5 for the different ${\mathrm{InF}}_3$ fiber lengths. The output power of the amplifier itself is also shown for comparison. We observe a 29 mW/W amplifier efficiency relative to the launched pump power which is comparable to previously reported results in a counter-directional pump configuration [8]. It is worth noting that the very low seeding power (i.e., $\sim 300\mathrm{\mu W}$ launched avg power) is partly responsible for such a low efficiency, due to competing spontaneous processes. At the maximum pump power, the launched peak power in the ${\mathrm{InF}}_3$ fiber is estimated at 13.5 kW for the pulse envelope, considering 550–600 ps pulse duration [8] and 20% splice losses. A lower slope efficiency is observed after propagation in the fluoroindate fiber: respectively, 19.4 mW/W at 1 m, 5.9 mW/W at 15 m, and 3.4 mW/W at 31 m. As expected, a longer propagation distance results in a lower average output power due to fiber losses. In addition, a slight saturation of the power curves is observed for longer fiber lengths and high power levels since the SC extends to the high loss edge of the ${\mathrm{InF}}_3$ fiber (beyond 5 μm).

 

Fig. 5. Average output power as a function of the launched pump power and ${\mathrm{InF}}_3$ fiber length. The black dots represent the experimental data points.

Download Full Size | PDF

It appears that several mechanisms play a role in the supercontinuum formation. Since relatively long pulses are used to pump the ${\mathrm{InF}}_3$ fiber in its anomalous dispersion region, we believe they first follow the standard scenario with a modulation-instability induced pulse breakup into multiple fundamental solitons [18]. Then, these shorter pulses shift to longer wavelengths with further propagation as a result of intrapulse Raman scattering (Raman self-frequency shift). The very asymmetric broadening process, i.e., almost exclusively toward the long wavelengths, can be explained by the fact that the 2.75 μm seed wavelength is far from the ZDW, which is known to be near 1.7 μm in such glass compositions [13]. On the other hand, it should be noted that the prominent peak near 4.1–4.2 μm does not appear to be driven by the mechanisms described above. In fact, it suddenly emerges at a fixed wavelength, which is not influenced by the input power or by the fiber length. Second-order Raman Stokes generation from the seed was first discarded as it would appear near 3.8–3.9 μm, considering the Raman gain peak is $\sim 510{\mathrm{cm}}^{-1}$ in such glass [9]. A possible explanation for this prominent peak would rather be related to the dispersion properties of the fiber, namely to the presence of a second ZDW near 4 μm. Unlike silica glass, the dispersion characteristics of fluoride glasses can be much more complex due to their multi-component nature. There have been several reports of fluoride fibers with a group velocity dispersion remaining near zero on an extended wavelength range, sometimes displaying two or more ZDWs [19,20]. Based on this hypothesis, the emergence of the 4.2 μm peak could be accounted for by the formation of dispersive waves (DW) [20,21]. In addition, the excitation of higher-order modes during the spectral broadening process could lead to a similar feature [22,23]. Further investigation, namely a thorough analysis of the dispersive properties of the fiber, are planned to clarify the underlying mechanisms.

At this point, the main limitation of this SC source in terms of average output power is the onset of the parasitic lasing threshold of the erbium fiber amplifier, which prevents increasing the 970 nm pump power. We also observed that the threshold was reduced after splicing the ${\mathrm{InF}}_3$ fiber. Considering that the output fiber tip $({\mathrm{InF}}_3)$ was cleaved at a very high angle of 14°, the most probable feedback source appears to be a slight reflection at the fusion splice. We expect that it will be possible to push away this lasing threshold by improving the splicing process to reduce the feedback. Increasing the repetition rate of the seed source and improving the seed launching efficiency should not only push away the CW lasing threshold, but also increase the optical/optical efficiency of the amplifier and, thus, raise the average output power without having to increase the pump level. We are also investigating the possibility of adding an isolator in the seed’s path to reduce the feedback from the entrance.

A thorough optimization of the amplifier parameters (fiber length and geometry) could also allow us to increase the system’s efficiency to reach the targeted watt-level average output power. Furthermore, we are also considering an ${\mathrm{Er}}^{3+}$-doped ${\mathrm{InF}}_3$ fiber as a replacement for the ${\mathrm{ZrF}}_4$ fiber, thus enabling the generation of the entire SC in the fiber amplifier to reduce the source’s complexity.

In summary, we have demonstrated a mid-IR SC source by cascading an erbium-doped fluoride fiber amplifier and a low-loss ${\mathrm{InF}}_3$ fiber. With relatively low pump power requirements, the spectrum broadened from 2.4 to 5.4 μm, i.e., up to the long wavelength edge of the ${\mathrm{InF}}_3$ glass transparency window. These results show the potential of ${\mathrm{InF}}_3$-based fibers and are a major step toward compact and reliable all-fiber SC sources covering the 3–5 μm band.