There is currently a growing demand for lasers emitting in the mid-infrared (MIR), driven by applications in the medical, military, and environmental sectors, among others [1–3]. The 3–5 $\mathrm {\mu }$m range, covering one of the transmission windows of the atmosphere, is particularly interesting for the remote sensing of greenhouse gases and atmospheric pollutants, but also for material processing, directional infrared countermeasures, and free-space communications [4–6]. In the last decade, the output power of single-mode rare-earth-doped fiber lasers in this region has increased significantly, with demonstrations of 41 W near 2825 nm, 15 W near 3550 nm, and 2 W near 3800 nm reported in fluorozirconate (ZrF₄) fibers [7–9]. This downward trend in achieved output power versus wavelength, especially above 3500 nm, is explained in part by the phonon energy of ZrF₄ ($\sim$580 cm⁻¹) that increases both the non-radiative decay rate of the energy levels and the attenuation of the fiber. In fact, Schneider et al. have shown in 1997 that the phonon energy of ZBLAN was too high to achieve lasing at 3920 nm in a Ho³⁺:ZrF₄ fiber at room temperature, but it could be achieved at cryogenic temperature with a cascade lasing scheme [10]. In 2015, Berrou et al. showed that by using a fluoroindate host (InF₃), which has a lower phonon energy ($\sim$510 cm⁻¹) than ZrF₄, a pulsed laser operation could be achieved at room temperature in a bulk [11]. In 2018, Maes et al. used a 100 $\mathrm {\mu }$m diameter 10 mol. % Ho³⁺:InF₃ fiber to demonstrate the first room-temperature fluoride-based fiber laser emitting at 3920 nm in continuous-wave operation [12]. The short laser cavity of 23 cm was bounded by two butt-coupled dichroic mirrors (DM) and emitted 200 mW near 3920 nm when pumped at 888 nm in free-space, until it got damaged at the entrance butt-coupled end.

It was shown afterward that fiber Bragg gratings (FBGs), a key component for all-fiber laser cavities, could be written in InF₃ fibers using the femtosecond (fs) scanning phase-mask writing technique, with refractive index modulations even larger than that in ZrF₄ fibers, and also showing a better thermal stability [13]. It was shown in parallel that high reflectivity FBGs could also be written with a line-by-line writing setup near 4 $\mathrm {\mu }$m in an undoped InF₃ fiber [14] and with a two-beam phase-mask interferometer writing setup near 2 $\mathrm {\mu }$m in a similar fiber [15]. Recently, a line-by-line written FBG with a reflectivity of 98% at 3.2 $\mathrm {\mu }$m has been used as the input coupler of a Dy³⁺:InF₃ fiber laser, reaching a total output power of 160 mW when a butt-coupled 50% DM was used as the output coupler and free-space pumped at 2825 nm [16]. To the best of our knowledge, no other InF₃-based fiber lasers using FBGs were reported so far, nor any further experimental demonstration of fiber lasers operating near 3920 nm with the Ho³⁺:InF₃ system since Maes et al. [12].

Nonetheless, based on the experimental results of Maes et al., several numerical models have been proposed in the meantime to improve the emission at 3920 nm by various means [17–23]. The simplest lasing scheme for the 3920 nm transition (${}^{5}\textrm {I}_{5}\to {}^{5}\textrm {I}_{6}$) is to pump its upper level directly from the ground state (⁵I₈) using a laser diode at 888 nm. At low Ho³⁺ concentration, the ⁵I₅ level, strongly quenched by non-radiative multiphonon decays, is short-lived in fluoride glasses ($\sim$135 $\mathrm {\mu }$s) and rapidly decay to the long-lived ⁵I₆ level ($\sim$6.2 ms), making the transition self-terminated. At high Ho³⁺ concentration (a few mol. % at least), energy transfer upconversion (ETU) and cross-relaxation (CR) processes start significantly affecting the laser dynamic and either deplete the ⁵I₆ level or repopulate the ⁵I₅ level. Numerical models were reported to propose the use of a dual-wavelength pumping (DWP) scheme with a second pump wavelength at 962 nm (${}^{5}\textrm {I}_{7}\to {}^{5}\textrm {F}_{5}$, ${}^{5}\textrm {I}_{6}\to {}^{5}\textrm {S}_{2}$) [17], 1660 nm (${}^{5}\textrm {I}_{7}\to {}^{5}\textrm {I}_{5}$) [18], or 2100 nm (${}^{5}\textrm {I}_{6}\to {}^{5}\textrm {I}_{4}$) [19], while another one has proposed a cascade lasing (CL) scheme at 2920 nm (${}^{5}\textrm {I}_{6}\to {}^{5}\textrm {I}_{7}$) [20]. The effect of co-doping the Ho³⁺-doped fiber with another rare-earth ion, such as europium [21], neodymium [22], or terbium [23], has also been studied and has shown great potential to improve the 3920 nm laser efficiency.

In this Letter, experimental results showing a monolithic all-fiber laser emitting a single-mode output of 1.7 W at 3920 nm in a Ho³⁺:InF₃ fiber are reported. This was achieved by optimizing the active fiber geometry, specifically 1) reducing its core diameter to ensure a single-mode operation at 3.9 $\mathrm {\mu }$m; 2) increasing its cladding to about 250 $\mathrm {\mu }$m to reduce the pump absorption and heat load, thereby enabling the use of longer fiber lengths and more powerful pump sources; and 3) splicing the pump fiber and writing intra-core FBGs to allow for a robust all-fiber design.

The experimental setup of the fiber laser is shown in Fig. 1(a). The active fiber was manufactured by Le Verre Fluoré, and its design follows on from the one used in Maes et al. [12]. The Ho³⁺ concentration is kept at 10 mol. % since its effect on the lasing efficiency, by means of the ETU and CR processes, was found beneficial, although its precise effect is not yet fully understood. The core diameter and its numerical aperture (NA) were slightly decreased from 16 to 15 $\mathrm {\mu }$m and from 0.20 to 0.18, respectively, to ensure a single-mode propagation at 3.9 $\mathrm {\mu }$m ($\lambda _c \approx 3.5$ $\mathrm {\mu }$m). One key improvement over the previous design has been to increase the double-D-shaped cladding diameter of the fiber from $90 \times 100$ to $240 \times 260$ $\mathrm {\mu }$m, in order for it to match the 200/220 silica fiber output of the pump laser diode. Such size enlargement significantly improved the fiber mechanical strength compared to the 100 $\mathrm {\mu }$m fiber, making it easier to handle, cleave, and splice while also reducing the cladding pump absorption by a factor of $\sim 7$ (the ratio of the claddings’ areas), allowing the design of a longer cavity. Views from the side and cross section of the Ho³⁺:InF₃ fiber are shown in Figs. 1(c) and 1(d). Measured by cutback, the attenuation in the core at 3920 nm is only 0.03 dB/m, while being 3.75 dB/m at 888 nm in the cladding, as shown in Fig. 2. It is apparent from the figure that the attenuation at 888 nm originates from two distinct phenomena. The first is the GSA of Ho³⁺ ions, which accounts only for 1.1 dB/m of the 3.75 dB/m that is measured. The remaining 2.65 dB/m is a broadband attenuation peak spanning several hundreds of nanometers. While its origin has not been confirmed, its spectral shape is similar to the attenuation spectrum caused by surface crystallization (cf. [24]) and could be reduced with further improvements in the manufacturing process of the fiber.

 

Fig. 1. (a) Laser cavity layout. (b) Silica-to-ZrF₄ and (c) ZrF₄-to-InF₃ splices. (d) Cross section of the Ho³⁺:InF₃ fiber. PF, pump filter.

Download Full Size | PDF

 

Fig. 2. Attenuation of the Ho³⁺:InF₃ fiber measured by cutback, in the (a) cladding and the (b) core.

Download Full Size | PDF

The optimal cavity length and output coupler (OC) reflectivity were estimated prior to the experiment by simulations, which are described in more details in Supplement 1. This was done by implementing the parameters and rate equations related only to the single pumping at 888 nm from the DWP scheme at 888 and 962 nm [17], from the CL scheme at 3920 and 2920 nm [20], and the DWP scheme at 888 and 1660 nm [18]. All three models accurately reproduce the experimental results obtained by Maes et al. through modeling [12] and can be applied to our fiber by simply adjusting the fiber geometry and attenuation. While the three models predict different output powers (from 4 to 8 W) at 120 W of pump power and different slope efficiencies (from 5 to 10%), they all predict a pump threshold near 40 W and around the same optimal length and output coupler reflectivity. For example, the simulation done with the DWP scheme at 888 and 962 nm [17] predicts an optimal cavity length of 1.7 m with an output coupler reflectivity of 65% that emits 4.3 W at 3920 nm. The threshold is at 47.9 W of pump power and the slope efficiency is at 5.9%. According to the simulations, more than 90% of the maximum output power (i.e., more than 4 W) can still be reached if the length of the cavity is between 1.1 and 2.7 m, and if the OC reflectivity is between 30 and 80%. It is worth noting that the length of the cavity is limited by the high pump attenuation in the cladding. Indeed, by lowering it to a value of 0.035 dB/m that is common for Er³⁺:ZrF₄ [25], the model predicts a longer cavity length of 5.4 m that emits 13.8 W with a slope efficiency of 19%.

The FBGs bounding the laser cavity were written through the coating of a 1.7-m-long segment of the Ho³⁺:InF₃ fibers with the fs scanning phase-mask writing technique. Details of our writing setup can be found in [13,26]. The writing parameters are similar to those used to write FBGs in doped ZrF₄ fibers [27] and were set to correspond to a type I regime, ensuring that the FBGs’ losses remain as low as possible. The HR-FBG is 20 mm long and was written with a uniform pitch phase mask ($\Lambda _{\text {PM}} = 2675.5$ nm). Its transmission spectrum was measured with an optical spectrum analyzer (AQ6377, Yokogawa) with an in-house built supercontinuum source, as shown in Fig. 3(a), following a thermal annealing at 120$^{\circ }$C for 10 min to improve its long-term stability. The aged HR-FBG retained a reflectivity of $\sim 98$% near 3921.5 nm while showing no appreciable loss within the resolution limit of our characterization setup. To ensure a good spectral overlap with the narrow HR-FBG, while simultaneously reaching a sufficiently high reflectivity, the LR-FBG was written with a chirped phase mask ($\bar {\Lambda }_{\text {PM}} = 2674.7$ nm, $C_{\text {PM}}$ = 1.2 nm/cm) over 25 mm. A dynamic apodization technique, where the translation stage is moving faster on the edges of the FBG than in its center, was used to reduce the modulation in its spectrum, which is typical in non-apodized chirped FBGs. The LR-FBG retained its initial reflectivity of $\sim 80$% after being thermally annealed at 120$^{\circ }$C for 10 min. Its transmission spectrum is shown in Fig. 3(a). As additional information, the annealing behavior at high temperature of a shorter uniform HR-FBG written with similar parameters in this fiber is shown in Figs. 3(b) and 3(c), where it is annealed successively for 10 min at each temperature step. The FBG’s reflectivity is shown to remain stable up to 250$^{\circ }$C, as seen in Fig. 3(b), but quickly goes to zero as the annealing temperature approaches the glass transition temperature. The FBG’s central wavelength appears to be less stable as it progressively decreases with temperature, shifting by $-6$ nm when annealed at 275$^{\circ }$C (cf. Fig. 3(c)).

 

Fig. 3. (a) Transmission spectrum of the FBGs bounding the laser cavity. The thermal annealing behavior of FBGs in the Ho³⁺:InF₃ fibers is shown in (b) and (c).

Download Full Size | PDF

The Ho³⁺:InF₃ fiber was spliced in two steps to the 200/220 $\mathrm {\mu }$m silica fiber output of the 120 W laser diode at 888 nm (K888BN2RN-120.0W, BWT) to obtain the monolithic cavity design shown in Fig. 1(a). The two splices were made with a Vytran GPX series system that was set up with an iridium filament, as described in more detail in [8]. While a direct silica-to-Ho³⁺:InF₃ splice would have been possible, the effect of the high heat load generated at the input face of the pumped Ho³⁺:InF₃ fiber was a matter of concern. It was thus deemed safer to add a segment of an undoped fluoride fiber, which shifts the high heat load to a stronger fluoride-to-fluoride splice. Due to fiber availability, a segment of an undoped ZrF₄ fiber was used instead of one made from InF₃. The silica-to-ZrF₄ and the ZrF₄-to-InF₃ splices are shown in Figs. 1(b) and 1(c), respectively. They were recoated on an aluminum spool using a low-index UV-cured fluoroacrylate polymer, and the fiber was fixed on the spool with a high thermal conductivity silicone to improve the thermal transfer and to ease future removal. Both the aluminum spool and the pump diode were installed on a water-cooled aluminum plate that was kept at a constant temperature of 20$^{\circ }$C. The fiber tip was fixed inside a V-groove on a copper block with the same UV-cured polymer, and a constant stream of dry air was blown on the tip during operation. Given the long operating wavelength that does not significantly overlap with OH absorption from atmosphere, no endcap was required at this power [28]. A dichroic filter was used to remove the residual pump power ($T=-32$ dB at 888 nm), and the output power at 3920 nm was measured with a thermopile power detector (XLP12-3S-H2, Gentec-EO).

The output power of the 3920 nm laser as a function of the launched pump power is shown in Fig. 4(a). The cavity emits 1.7 W at 3921.5 nm when 120 W of pump power at 888 nm is launched. According to the curve on Fig. 4(a), the threshold is around 32 W, and the mean slope efficiency is 2.1%, gradually increasing from 1.8% at low power up to 2.2% at high power. It is important to note that only 27.1 W of the incident 120 W is estimated to be absorbed by holmium ions, according to Fig. 2. Another 65.3 W is estimated to be absorbed or diffused in the cladding of the fiber and does not contribute to pumping. Finally, the residual pump power, estimated to be 27.7 W but experimentally measured at 24.4 W, is filtered by the pump filter (24.4 W–32 dB = 15.4 mW). Considering this, the laser efficiency according to the absorbed pump power is 9.2%. The laser output spectra, shown at 0.5, 1.0, and 1.7 W in Fig. 4(b), have a linewidth of $0.7\pm 0.1$ nm at FWHM, which is expected from the bandwidth of the HR-FBG.

 

Fig. 4. (a) Output power of the 3920 nm laser with respect to both the launched pump power at 888 nm and the power absorbed by Ho³⁺ ions. (b) Corresponding output spectrum at different output powers.

Download Full Size | PDF

The output power stability of the laser cavity is shown in Fig. 5(a) over 15 min when turned on near full power. The power fluctuates as much as 6% in the first 2 min, which is attributed to the cavity warming up. It increases the non-radiative decay rate of the upper level of the laser transition, slightly reducing its efficiency. Once the fiber and the aluminum spool have reached thermal equilibrium, the output power is much more stable, slowly decreasing by 2% over the next 13 min. It is worth noting that the rapid power fluctuations in the beginning are not observed if the cavity is already warm, e.g., when the laser cavity has already been operating for a few minutes. The change in the output power for the first minute after the cavity is turned on is shown in Fig. 5(b) for different output powers. It is interesting to note that below 200 mW of output power, the laser is not stable and stops lasing after a few seconds to a few tens of seconds. The holmium-doped fiber reaches nearly 60$^{\circ }$C at 120 W of the launched pump power in the current configuration, i.e., an increase of 0.34$^{\circ }$C/W. It is expected that a further 60 W (180 W in total) could be injected without compromising the laser stability.

 

Fig. 5. Output power stability (a) near full power over 15 min and (b) for different power levels over 1 min.

Download Full Size | PDF

At this point, we have not been able to reproduce our experimental results through simulations with a set of parameters that can also reproduce the results obtained by Maes et al. [12]. The problem lies in the low slope efficiency of our system, which can only be simulated by adding either 14.8% ($-0.70$ dB) of losses at each FBGs, or an additional 0.69 dB/m in the fiber core attenuation at 3920 nm. However, the transmission of the segment of the fiber forming our cavity was re-measured after the experiment and compared to a shorter segment of the same fiber batch without FBGs, and no additional losses were observed in the segment used. It is believed that the low slope efficiency may be due to the reabsorption of the 3920 nm signal inside the cavity because the last part of the fiber is not pumped high enough to reach population inversion.

In order to continue power scaling further, a better understanding of the dynamic of this laser is necessary. This should rely on a thorough investigation of the spectroscopic properties of the active fiber, which will be used to assess the validity of the models. In this demonstration, the FBGs were written at 3920 nm since lasing was previously observed at this wavelength. However, our longer cavity might increase reabsorption at shorter wavelengths, such that lasing at longer wavelengths (toward 4 $\mathrm {\mu }$m) might help in increasing the output power. To our knowledge, the impact of the lasing wavelength on the laser performances of this transition in a Ho³⁺:InF₃ fiber laser has not been investigated by a numerical model yet.

It is worth noting that even though the three models presented in Supplement 1 all reproduce the results obtained by Maes et al. [12] for the simple case of single pumping at 888 nm and single lasing at 3920 nm, their predictions differ from each other for different fiber lengths, output coupler reflectivities, or cladding diameters. This raises the question of how valid the models are to predict more complex lasing scheme such as DWP [17,18] or CL [20] and how well they will work experimentally. Their experimental demonstrations will definitely be made soon, since they predicted several benefits over the single pumping at 888 nm, such as higher output power and lower heat load. Co-doping also shows great potential to quench the lower level of the laser transition while keeping the laser cavity simple, but it necessitates the manufacturing of other fibers.

In conclusion, a 1.7 W monolithic fiber laser at 3920 nm was demonstrated in a Ho³⁺:InF₃ fiber. This was achieved by recent improvements in the manufacturing of rare-earth-doped low-loss fluoroindate fibers, as well as in the writing of FBGs in these fibers and in high-quality silica-to-fluoride splices. This work paves the way for monolithic high-power fiber lasers emitting near and beyond 4 $\mathrm {\mu }$m. It is believed that further optimization of the laser cavity, a better understanding of the laser dynamic, and most notably a reduction of the cladding losses will lead to an all-fiber laser cavity exceeding 10 W in the future.