1. Introduction

Power-scaling of ytterbium (Yb³⁺)-doped single-fiber lasers (YDFL’s) [1–10] has now led to output powers beyond 1 kW [9, 10]. This rapid progress has been fueled both by the exceptional ability of fibers to carry high optical power and the availability of powerful diode lasers for use as pumps. These in turn are due mainly to improvements in fiber design and fabrication, and in multi-emitter lasers diodes, diode bars, and diode stacks. For the highest powers, silica-based Yb³⁺-doped fibers (YDF’s) emitting in the 1–1.1 µm wavelength range are the leading contenders because of their excellent power conversion efficiency (>80%) which results from the small quantum defect with the pump wavelength of ~975 nm. As with other rare-earth doped silica fibers, YDFL’s can be made tunable over several tens of nanometers, while one can access a wide range of wavelengths from the near infrared to the so-called eye-safe region using other dopants, in particular neodymium (Nd³⁺) [11], erbium (Er³⁺), and thulium (Tm³⁺). As a result, there are many applications that benefit from the many advantages of fiber lasers. These include material processing, marking, medicine, range finding, free space communication, and security.

We show that at present the output power of a well-designed YDFL is limited only by the available pump power. Other potential constraints on power scaling of fiber lasers include thermal effects, optical damage of fiber materials, and nonlinear scattering (especially stimulated Raman scattering). Fortunately, the high conversion efficiencies possible with YDFL’s mean that the heat dissipation can be as low as 20 W/m with a 10 m long fiber generating 1 kW of output power. With regard to optical damage and stimulated Raman scattering, a large-core design is preferred while maintaining acceptable beam quality. These large cores typically can sustain a number of modes and it is therefore possible for the fiber to lase on higher-order modes, which may also lead to temporal and spatial instabilities in the output beam. Nonetheless, several groups have reported stable and nearly diffraction-limited output can be generated with core diameters up to ~30 µm [6–8]. For further power-scaling, even larger cores with diameters of ~40 µm are helpful to mitigate nonlinear scattering and optical damage, while maintaining the high efficiency [9–10]. However, this approach tends to degrade the beam quality and requires mode selection methods, such as modal bend-loss filtering [6], to restore it; however, this requires a sufficiently large modal bend-loss differential. In all cases, a low numerical aperture (NA) was required for good beam quality.

In this paper, we demonstrate further power-scaling of YDFL’s, using a fiber with a further reduced NA, thus reducing the number of modes which can propagate. We reached an output power of 1.36 kW in a near single-mode beam in a highly efficient configuration. To achieve this, we designed and fabricated a double-clad YDF with a low-NA large-diameter core and pumped it with two 975-nm diode-stack sources with 1.8 kW of total pump power in a double-ended pumping scheme. In comparison with our previous work with a higher-NA fiber [9], we achieved a significant improvement in the beam quality as well as even higher output power. The output power is still limited only by available pump power, without any roll-over even at the highest power.

2. Experiments and results

A double-clad YDF was designed to meet the requirements for power-scaling beyond 1 kW. It was drawn from a preform that was fabricated in-house by the modified chemical-vapor-deposition and solution-doping technique. The fiber had a 40-µm diameter Yb³⁺-doped aluminosilicate-based glass core with an NA below 0.05, centered in the preform. The mode-field area for the fundamental mode (based on the formula of Ref. [12]) is estimated to ~0.9×10³ µm². The D-shaped inner cladding had a 650/600-µm diameter for the longer/shorter axis. This diameter was chosen to enable efficient coupling of the large beams from high power diode pump sources. The fiber was coated with a low-refractive-index polymer outer cladding which provided a nominal inner-cladding NA of 0.48. Compared with our previous YDF from which we obtained 1 kW of output power [9], we have modified the core design and achieved a significantly lower NA. This reduces the V-parameter and the number of guided modes. In addition, because of the weaker guiding with this lower NA, the core diameter could be reduced by about 7% while maintaining a similar effective mode area for the fundamental mode. Consequently, the resultant V-parameter of this fiber becomes ~5.7, which is approximately half the value from our previous design [9].

Furthermore, the lower core NA facilitates the use of bend-induced modal losses to higher-order modes. Under the assumption of a single-clad step-index fiber [13], we theoretically estimate that the bend-induced loss to the fundamental mode (LP₀₁) is still negligible with a moderate bend radius, e.g., less than ~0.04 dB/m for a bend radius of greater than 10 cm while the losses to the higher-order modes can be significantly higher: The bend-loss to the first higher-order mode to be rejected (LP₁₁) is estimated to ~1 dB/m for a bend radius of 12 cm. Such a difference in bend-loss would be enough to ensure single-mode operation. By contrast, with higher NAs of, say, 0.1, the bend-loss to the LP11-mode is negligible for practical bend radii. On the other hand, fiber bending leads to mode coupling between modes [14], which works against the desired mode filtering, and also modifies the bend-loss of a given mode. The presence of the inner cladding and the gain medium further modifies the bend-loss, which in any case is very sensitive to parameter values. Thus, our simple bend-loss estimates, made under the assumption of a simple single-clad passive fiber and neglecting coupling between guided modes, are still uncertain. Nevertheless, we conclude that bend-loss can be an important factor in this fiber.

The small-signal absorption rate in the inner cladding was ~1.5 dB/m at the pump wavelength of 975 nm. This corresponds to an Yb³⁺-concentration of ~6000 ppm by weight. The fiber length used in the laser experiments was 12 m.

 

Fig. 1. YDFL experimental arrangement with two diode-stack pump sources. HR: high reflectivity, HT: high transmission.

Download Full Size | PDF

The experimental setup is shown in Fig. 1. We used a double-ended pumping scheme, with two pump sources launched into opposite ends of the fiber. Two diode-laser-stacks were used, one emitting 1.2 kW and the other emitting 0.6 kW of power at 975 nm. The pump beams were coupled into the active fiber via collimating and focusing lenses and a pair of folding dichroic mirrors. We obtained a coupling efficiency of over 90%. Both ends of the fiber were cleaved perpendicularly to the fiber axis and remained uncoated. No special measures such as anti-reflection coatings were used to reduce the pump reflection at the fiber facet. At one end of the laser cavity, high-reflectivity feedback was provided by a pair of dichroic mirrors having high transmission at the pump wavelength and high reflection at the signal wavelength. The mirrors were external to the fiber and coupled to it via a lens. The laser output coupler was formed by the 4% reflecting flat perpendicular cleave at the other end of the fiber. The signal was separated from the pump beam using another dichroic mirror. Both ends of the fiber were held in temperature-controlled metallic V-grooves designed to prevent possible thermal damage to the fiber coating by any non-guided pump or signal power, or by the heat generated in the laser itself.

The laser output power characteristics are shown in Fig. 2, together with the output spectrum at full output power. The laser gave a single-ended output of 1.36 kW for a launched pump power of just over 1.6 kW, representing excellent slope efficiency of 83% and quantum efficiency close to 95%. The pump throughput was estimated to be below 1.2%. The standard deviation of the temporal power spectrum was <1.5%, measured with a 5 GHz photo-detector and a 400 MHz bandwidth oscilloscope. We measured a beam quality factor (M²) of 1.4, showing that the beam was nearly diffraction-limited although we did not take special measures, such as well-controlled coiling or tapering of the fiber [6, 15], to suppress the lasing on higher-order modes beyond the design of the core. Among the possibilities that led to such good beam quality with a multi-mode core, modal gain differentials and bend-induced mode filtering may have helped to improve the output beam quality. Our bend diameter was in the range of ~20 cm, which agrees with the range for which bend-loss filtering would be effective, based on our rough estimate. However, the bend-radius was not adjusted for the best beam quality, and the influence of the bend radius was not investigated in the experiments. We must also reiterate the sensitivity of the bend-loss to parameter values, and in particular the index step. This point is emphasized by the limited precision in our refractive index profile measurements at this low NA (index step less than 10^-3). This makes it difficult to correlate our data with theoretical estimates of differential modal bend-loss or gain. However, in the past we have found that in addition to having a low NA, the elimination of the frequently occurring dip in the refractive index profile in the center of the core was important for obtaining good beam quality from large-core fibers.

 

Fig. 2. (a) Fiber laser output power vs. launched pump power. (b) Laser output spectrum at full power.

Download Full Size | PDF

The laser output increased linearly with launched pump power and showed no evidence of roll-over even at the highest output power, which was limited only by available pump power. There was no evidence of any power limitation due to nonlinear scattering, nor was any stimulated Raman scattering observed. Assuming a Raman gain coefficient of 1×10^-13 m/W [12], we estimated the Raman gain of our YDFL to be below 4 dB, which is far below the single-pass laser gain threshold of ~14 dB.

There was no sign of damage to the end faces during the operation. The maximum power density was ~1.5 W/µm². Compared with the power density reported in other work, e.g. ~2 W/µm² in a YDF and ~6 W/µm² in a passive fiber in [5], and still higher in bulk silica in a pulsed regime [16], our power density was still low. This suggests that our fiber is far from the damage threshold even at the maximum output power. On this basis, we estimate that single fiber lasers could be power-scaled to beyond multi-kW if sufficient pump power is available, for example, via a more powerful pump source or with additional wavelength-multiplexed pump sources.

3. Conclusion

We have demonstrated a highly efficient (83%), high Yb³⁺-concentration (~6000 ppm by weight), double-clad Yb³⁺-doped large-core fiber laser with a cw output power of 1.36 kW at 1.1 µm in a near single-mode beam (M²=1.4). The laser used on 12 m long single fiber and showed no evidence of roll-over in laser output power even at the highest launched pump powers (~1.6 kW). We expect to achieve diffraction-limited beam quality with a comparable or higher output power in the near future through more advanced fiber design, combined with more powerful pump sources. Power-scaling beyond 10 kW in a single-fiber configuration looks entirely feasible with our fiber laser design.