1. Introduction

The search for powerful, compact and stable fiber-based sources emitting in the mid-IR is motivated by their numerous applications in domains such as laser surgery [1], frequency conversion [2] and spectroscopy [3]. For laser surgery applications, output powers of 1-10W at wavelengths of 2.7 to 3 μm are required for the efficient processing of high moisture content materials such as biological tissues [4]. Recently, several demonstrations of fiber lasers emitting around 3 μm based on fluoride glass (ZBLAN) fibers were reported using active ions such as erbium [5,6], holmium [7] and more recently dysprosium [8]. The lack of economical high-power diodes in the 1100nm and 1300nm pump bands for Ho and Dy doped ZBLAN lasers, respectively, limits the achievable output power. On the other hand, for the 970-980 nm pump band of erbium, high power diodes are readily available at affordable prices. Since the first demonstration of a 1-W diode-pumped Er-doped ZBLAN fiber laser in 1999 [9], several experimental demonstrations of fiber laser operation at 2.7-2.8 μm in both the CW [10–12] and pulsed [13] regimes have been reported. The output power was recently scaled up to >9 W [14] by using a highly-doped Er:ZBLAN fiber pumped with a 975nm diode setup. However, the authors also reported severe output power fluctuations at high output powers (≥3W) that eventually lead to the destruction of the fiber end-face. The output spectrum was also reported to shift towards longer wavelengths with increasing pump powers from 2.71 to 2.79μm. The reported slope efficiency of 21% with respect to the launched pump power was less than half of what can be expected from the highly efficient pump energy recycling process [15]. In all previous demonstrations of laser emission in the 3 μm region, the laser efficiency, robustness and compactness were compromised by the use of bulk optical components to form the laser cavity. Recently, our group reported the capability of writing highly reflective fiber Bragg gratings (FBGs) in ZBLAN fibers using ultra-short laser pulses [16]. Such FBGs proved to be suitable for use as high reflectors in a Tm-doped, all-fiber laser, emitting at 1480nm [17], with output powers reaching as high as 2.3W [18].

In this paper, we report the operation of an all-fiber, highly stable, efficient erbium-doped fluoride fiber laser. The maximum output power of 5W at a stable wavelength of 2824nm was enabled by the use of a high reflector (HR), which is a FBG written directly in the highly-doped fluoride fiber core. The optical-to-optical conversion efficiency of 32% is significantly higher than previously reported. The peak-to-peak power fluctuations were observed to be less than ± 0.3%.

2. Experiment

The fiber used in the experiment is a 6.42m, 7 mol% singly Er³⁺-doped fluoride fiber provided by Le Verre Fluoré. The standard ZBLAN composition was modified to permit this level of doping. The pump core of the double-clad fiber had a 160x135 μm D-shaped geometry and a numerical aperture (NA) >0.46. The 8 μm diameter, 0.24 NA fiber core ensured single-mode operation of the fiber laser. The pump absorption at 976 nm was measured at 1.65dB/m through a cutback experiment, while the background losses for the laser signal at 2.8μm were measured to be 0.18 dB/m. The fiber laser cavity configuration used to test the laser operation is presented in Fig. 1 .

 

Fig. 1 Erbium-doped all fiber laser configuration with a FBG acting as the high reflector.

Download Full Size | PDF

As shown in Fig. 1, the doped fiber was pumped through the HR using the combined output of 6 pump diodes (Alfalight AM6-976B-10-604) coupled through a 7x1 multimode pump combiner (ITF Labs MMC07011011). This setup provided up to 36W of available pump power at the output of the 125 μm diameter, 0.46 NA fiber of the pump combiner. The pump power was launched into the doped fiber by butt-coupling. A launch efficiency of 84% was inferred from the results of the cutback experiment. To minimize thermal effects at the launch end, a water-cooled fiber mount was used to hold both the launch fiber end and the HR, although active air cooling would have likely been sufficient given the high launch efficiency. The length of the segment between the pump fiber end and the FBG was roughly 7 cm. Both fiber ends were carefully polished to a perpendicular, smooth finish. The ~4% Fresnel reflection provided the low reflectivity output reflector of the fiber laser cavity. During our early experiments, we systematically observed catastrophic optical damage (COD) of the output fiber end-face for output powers of about 2 W. The cause of this COD was attributed to the presence of moisture at the ZBLAN-air interface. To prevent this, we found it necessary to purge the output fiber end-face from moisture, which was accomplished by blowing pressurized nitrogen on the fiber tip. A dichroic mirror (≥95% reflectivity at 976nm) was used to remove the ~8% residual pump power from the collimated output beam. The output spectrum of the fiber laser was measured with a Digikrom DK480 monochromator and a PbSe detector, with a spectral resolution of about 0.1nm, while the laser output power was monitored using a thermopile power meter (Gentec UP19K-15S-H5).

The HR was written using the method described in [16] based on the interaction of femtosecond laser pulses at a central wavelength of 806 nm. The polymer coating of the fiber was removed prior to writing the FBG and a low index polyacrylate was subsequently applied to ensure low pump guiding losses and to protect the fiber. The phase mask used in the experiment had a uniform pitch of 1903.5 nm, giving a first-order Bragg reflection in the Er-doped fluoride fiber at 2822.5nm. The transmission spectrum of the FBG was monitored using a ZBLAN-based super-continuum light source developed in-house which covered a bandwidth extending from 1000nm to 3750nm. After the inscription process, the FBG was thermally annealed so as to stabilize the FBG’s reflectivity. The DK480 monochromator was used to measure the transmission spectrum of the grating with a spectral resolution of 0.4nm. Figure 2 shows the transmission spectrum of the annealed FBG, which was written in the highly-doped fluoride fiber.

 

Fig. 2 Measured transmission spectrum of the FBG written in the Er³⁺-doped fluoride fiber (exposure time: 15s, grating length: 8mm).

Download Full Size | PDF

The maximum reflectivity of the grating occurs at 2822.5 nm with a full width at half maximum (FWHM) of 2.4 nm. The throughput losses of less than 5% were evaluated separately through a cut-back measurement outside of the FBG’s bandwidth at 2830 nm, so as not to interfere with the FBG transmission dip. These measurements imply a peak reflectivity of at least 95% at 2822.5 nm. The pump transmission through the highly reflective FBG was measured to be 97%.

3. Results and discussion

Figure 3 shows the output power of the laser with respect to the launched pump power.

 

Fig. 3 Laser output power vs launched pump power.

Download Full Size | PDF

A maximum of 5W of output power was measured for a launched pump power of 20.5W, indicating an overall laser efficiency of 24%. For higher pump powers, the pump fiber end was subject to thermally-induced damage. The laser threshold could not be measured directly due to the unstable operation of the pump source near threshold, but a value of about 60 mW was inferred from a linear extrapolation of the results at low pump powers. For output powers <2.5W, the laser slope efficiency reached 29% with respect to the launched pump power. By accounting for the residual pump power of ~8% of the launched pump power, the optical-to-optical conversion efficiency was 32%. This is close to the Stokes efficiency of 34% but far from the theoretical efficiency of >50% inherent to the pump energy recycling scheme in such a highly doped (7 mol.%) fiber [15]. Our calculations, based on a model similar to Pollnau’s, suggest that this discrepancy could be due to lower energy transfer parameters in our fiber, which would arise from a reduced amount of clusters. Also, there is a noticeable decrease of the slope efficiency at output powers >2.5W. The origin of this is unclear but it might be due to thermal effects, pump ESA from the upper laser level or to a variation of the overall gain with increasing pump power, which has been reported before [12]. We should note that a similar saturation behavior has been reported in the past [19] and was attributed to simultaneous lasing at 850nm from the ⁴S_3/2→⁴I_13/2 transition. However, pumping in the 980 nm has been reported to prevent this from occurring.

The emission spectrum of the laser shown in Fig. 4 was measured for an output power of 3.24W and the inset shows a blow-up measured with a 0.1nm resolution.

 

Fig. 4 Laser output spectrum for an output power of 3.24W. The inset shows the corresponding narrowband laser emission spectrum measured with a resolution of 0.1nm.

Download Full Size | PDF

The FWHM and the central wavelength were 0.22nm and 2823.9nm, respectively. We attribute the 1.4nm offset between the peak emission wavelength and the Bragg wavelength of 2822.5nm to the thermal expansion of the fiber leading to an increase of the Bragg wavelength. In fact, measurements of the output spectrum as a function of the output power indicated a spectral drift of the Bragg wavelength of about 0.4nm per watt of laser emission.

The temporal stability of the fiber laser output was monitored using a thermopile detector (Spectra Physics, 407A) with a 0.5s response time. Figure 5 shows the output power stability of the laser over an observation of 100s at different output powers ranging from 0.85W up to 5W.

 

Fig. 5 Output power stability for different output powers ranging from 0.85W up to 5W

Download Full Size | PDF

Peak-to-peak fluctuations of less than ± 0.3% are measured at 5W of output power. The largest peak-to-peak fluctuations ( ± 0.8%) were observed at output power of 0.85W. A liquid-nitrogen-cooled InSb photodetector (Judson J10D) was also used to confirm the laser stability on a time-scale of 0.25s (acquisition rate of 125 kHz) in a similar laser cavity and no rapid fluctuations larger than those observed in Fig. 5 has been noted.

4. Conclusion

In summary, we demonstrated the efficient, highly stable, 5 W CW laser operation at 2.8 μm of a diode-pumped erbium-doped all-fiber laser employing an intracore fiber Bragg grating high reflector. The optical-to-optical conversion efficiency (32%), the temporal stability and the spectral stability of the laser output all represent significant improvements from previous reports. Our results indicate that higher output powers must be within reach by further optimizing the laser cavity parameters to increase the slope efficiency.