1. Introduction

All current interferometric gravitational wave detectors (GWDs) are based on 1.064 μm lasers. The next generation of gravitational wave detectors will probably also use laser sources at 1.5 μm and suitable mirror substrates, to lower the noise level of the interferometer [1]. Due to the special requirements of such detectors, the used laser source must be single-frequency, be linearly polarized, have diffraction limited beam quality, and deliver high output power (depending on the final design up to 1 kW). A promising way to realize such a source is to use a fiber amplifier, in which a low power single-frequency laser with a diffraction limited beam quality gets amplified to higher output levels. While pure Er³⁺-doped fibers are able to amplify wavelengths in the 1.5 μm region, they typically suffer from low absorption, if pumped at 976 nm. Additionally, the doping concentration of Er³⁺-ions is limited by the so called quenching [2], so that long fiber lengths are needed, if high output power levels are required. In amplifiers with a broad bandwidth seed signal, this is typically not a problem. But if the signal has a narrow linewidth, additional nonlinear effects like stimulated Brillouin scattering (SBS) can limit the achievable output power. Co-doping with Yb³⁺-ions leads to acceptable levels of pump light absorption and high output power levels around 150 W for MOPA amplifier setups [3] and around 300 W for laser setups [4] have been demonstrated. Nonetheless, Er³⁺:Yb³⁺ co-doped fiber amplifiers typically suffer from the increasing build-up of Yb-ASE at high pump power, first leading to a lower amplifier efficiency at 1.5 μm and then finally resulting in instabilities due to parasitic lasing processes at 1.0 μm [3]. To overcome this problem, several approaches have been proposed [5–11]. One promising method is the so called co-seeding method, in which a second seed laser at 1.0 μm is used to achieve stable amplification at 1.0 μm and to suppress by this any parasitic processes of the Yb-ASE [12]. In addition, if the absorption cross-section for the co-seeding wavelength is non-zero, if the fiber is long enough, and if the co-seed is co-propagating with the pump radiation, the co-seed can extract excess energy mainly from the pump launch end of the fiber and can act as a second pump source mainly towards the other end of the fiber. Although this method of applying a second external seed at 1.0 μm was investigated experimentally [12–14] and theoretically [15, 16] in detail, no real power scaling to output power levels above 10 W was demonstrated, so far.

The current GWDs require a pure TEM₀₀ mode, thus, corresponding laser sources must have an excellent beam quality. In traditional step-index fibers the beam quality is given by the V-Parameter V = (2π/λ) · r_core · NA, which depends on the core NA and the core radius r_core. If one wants to amplify a single-frequency signal with such a fiber, it is desirable to have a large core radius and therefore a large mode field diameter of the fundamental mode to increase the threshold of nonlinear effects (e.g. SBS). Therefore, an Er³⁺:Yb³⁺ fiber with a large core diameter and a small core NA is a crucial part for the realization of a single-frequency fiber amplifier at 1.5 μm with high output power and good beam quality. Unfortunately, especially in Er³⁺:Yb³⁺ co-doped fibers the core NA cannot be lowered to arbitrary levels, because the required co-doping with phosphor (suppressing the energy back-transfer from Er³⁺ to Yb³⁺) leads to a very pronounced difference of the index of refraction of the core and the cladding. Lowering both the doping concentration of Yb³⁺ and phosphor opens the possibility of having a lower core NA, but leads to a lower pump light absorption per unit length as well. On the other hand, lowering only the phosphor concentration leads to a lower amplifier efficiency at 1.5 μm. Er³⁺:Yb³⁺ co-doped fibers with core diameters around 20 μm, a core NA of less then 0.1, a sufficient pump light absorption at 976 nm and the possibility to set up amplifiers at 1.5 μm with an efficiency around 30 %, are therefore not commercially available today. In this paper we report on the development of such an Er³⁺:Yb³⁺ fiber and on the power scaling of a single frequency fiber amplifier based on that fiber. Additionally, we report on beam quality measurements to determine the overlap of the amplifier output beam with the TEM₀₀ mode.

2. Fiber design and fabrication

The Er³⁺:Yb³⁺ triple clad fiber was designed and developed at INO (Québec City, QC, Canada). The fiber was drawn from a silica preform where the third cladding low refractive index polymer material was applied on-line during the fiber drawing process. A secondary coating, designed to ensure robustness of the fiber, was applied over the third cladding. The silica preform consisting of a silica substrate tube, depressed first clad, and a core co-doped with P₂O₅, Er₂O₃, and Yb₂O₃ was fabricated by the modified chemical vapor deposition (MCVD) process and solution doping [17]. The depressed first cladding was deposited by the MCVD process out of GeO₂ and P₂O₅ co-doped silica [18]. Porous white silica soot core layers were then deposited into the substrate silica tube. Subsequently, the soot layers were impregnated with an aqueous solution of H₃PO₄ and YbCl₃×6H₂O, dried, sintered, and finally collapsed into a solid mother preform. The well-known core index dip caused by the evaporation of the P₂O₅ during the collapse step was eliminated by taking special measures during the MCVD process. The final preform was obtained by sleeving a pure silica substrate tube over the mother preform. The second silica clad was etched to obtain the desired pump guide to core diameters ratio and then hexagonally-shaped for ensuring efficient optical pumping of the core. A schematic of the geometry and refractive index profile of the triple-clad fiber is shown in Fig. 1. A measured refractive index profile of the resulting Er³⁺:Yb³⁺ triple clad fiber is shown in Fig. 2(a).

 

Fig. 1 Schematic of the geometry and refractive index profile of the triple-clad fiber.

Download Full Size | PDF

 

Fig. 2 (a): Refractive index profile of the Er³⁺:Yb³⁺ triple-clad fibre. (b): Simulated LP₀₁ amplitude distribution.

Download Full Size | PDF

The core NA, defined with respect to the depressed-index level of the first cladding, was lowered to ∼0.09 by adding the first doped silica cladding. With a diameter of ∼23 μm the core would normally support higher-order modes but the W-index profile makes it effectively single-mode. Still, higher-order modes can propagate within the first cladding, but with a first-clad-to-core ratio of over 6 in diameters, these cladding modes extend largely out of the core and their overlap with the Er³⁺:Yb³⁺ dopants is very small, thus ensuring a good output beam quality. Figure 2(b) shows the amplitude distribution of the LP₀₁ fundamental mode calculated from the measured refractive index distribution of Fig. 2(a). The calculated effective mode area is ∼290 μm² and by fitting a TEM₀₀ mode to the calculated LP₀₁ amplitude distribution the overlap factor of the LP₀₁ mode with the TEM₀₀ mode was determined to be 99.7 %.

3. Amplifier setup

The experimental amplifier setup used for all measurements is shown in Fig. 3. A 22 m long piece of the Er³⁺:Yb³⁺ triple clad fiber with a core NA of 0.09, a core diameter of 23 μm and a pump light absorption of 1.5 dB/m at 976 nm was used in a counter pumped master oscillator power amplifier (MOPA) setup. Both fiber ends were angle cleaved in order to suppress back reflections at 1.0 μm and 1.5 μm. The 976 nm pump source (Laserline LDM500) delivered up to 500 W of power out of a fiber with 200 μm core diameter and a NA of 0.22. The corresponding multimode light was launched after collimation into the pump cladding of the triple clad fiber with a single aspherical f=15 mm lens. The 1.5 μm seed source was a commercial preamplified single-frequency fiber oscillator (TheRock from NP Photonics) with a kHz linewidth, a center wavelength of 1556 nm and an output power of 2 W. The corresponding seed light was mode matched to the fundamental mode of the triple clad fiber with a spherical f=150 mm lens and an aspherical f=8 mm lens. The co-seed source was a self-made tunable Yb-fiber ring laser, which was tunable from 1030 nm to 1060 nm and delivered up to 2 W of power for all wavelengths. The co-seed light was launched into the triple clad fiber with the same f=15 mm lens, which was used for pump light coupling. Therefore, the mode matching of the co-seed into the fiber core was not perfect, but by beam walking and monitoring the output power at 1.0 μm it was ensured that as much power as possible was launched into the core of the fiber. Measurements without any pump light revealed that the coupling efficiency for the co-seed light into the fiber core was around 80 %. Dichroic mirrors at both ends of the active fiber were used to combine and separate the pump, seed and co-seed radiation. In order to minimize problems with the heating of the fiber during any high power experiment, we placed the whole fiber on a conductive heatsink. Additionally, the pump launch end of the fiber was held by a water cooled copper block, directly connected to the heatsink.

 

Fig. 3 Schematic overview of the used experimental setup. The Er³⁺:Yb³⁺ triple clad fiber was placed on an additional conductive heatsink.

Download Full Size | PDF

4. Power scaling experiments

The co-seeding method can be used to suppress parasitic processes at 1.0 μm, but can also be used (if the fiber is long enough) to raise the efficiency at 1.5 μm by reabsorption of the amplified co-seed. In order to investigate which co-seeding parameters are the best to get the maximum out of these both advantages, we performed some preliminary experiments prior to any power scaling experiments. Note that all these experiments were performed with both seeds turned on. First we measured a fraction of the (with respect to the pump light direction) backward propagating Yb-ASE in dependency of the co-seeding wavelength and power for a fixed pump power of 25 W. Because parasitic processes at 1.0 μm are directly related to the power of the Yb-ASE, the possibility of parasitic processes is minimized for a maximized suppression of the Yb-ASE. The corresponding results can be seen in Fig. 4. As shown in Fig. 4(a) the best suppression of the Yb-ASE was achieved for co-seeding wavelengths between 1045 nm and 1060 nm, depending on the actual co-seed power. For a maximum co-seeding power of 2 W the best suppression of the Yb-ASE was achieved for a co-seeding wavelength of 1050 nm, see Fig. 4(b). It is worth to point out that, to the best of our knowledge, this was the first time that the influence of the co-seeding wavelength and power directly on the power of the Yb-ASE was investigated experimentally.

 

Fig. 4 (a): Backward propagating Yb-ASE in dependency of co-seeding power and wavelength (25 W pump power). (b): Yb-ASE in dependency of co-seeding wavelength for a fixed co-seeding power of 2 W and a pump power of 25 W. (c): Output power at 1.5 μm at 25 W of pump power for different co-seeding power levels and wavelengths. (d): Amplifier slopes at 1.0 μm and 1.5 μm for different co-seeding wavelengths with 2 W of co-seeding power.

Download Full Size | PDF

Next, we analyzed how the amplifier efficiency at 1.5 μm changed with the co-seeding wavelength and power. For this we monitored the amplified output power at 1.5 μm at a given pump power of 25 W for different co-seeding parameters, see Fig. 4(c). The optimal co-seeding wavelength was around 1040 nm with a corresponding power of 2 W. The slopes of the amplified 1.5 μm seed and 1.0 μm co-seed, see Fig. 4(d), for a fixed co-seeding power of 2 W also suggest that a co-seeding wavelength around 1040 nm is the best to achieve as much efficiency as possible at 1.5 μm. Although the influence of the co-seeding wavelength on the output power and slope efficiency at 1.5 μm was studied in detail [13, 14, 16], only one experimental study of the influence of the co-seeding power on the 1.5 μm signal has been carried out [14]. The corresponding authors reported only a small influence of the co-seeding power on the output power at 1.5 μm. Also, they observed that for some co-seeding wavelengths an increase in co-seeding power led to a decrease of power at 1.5 μm. Such a behavior did not occur during our measurements and we also observed, see Fig. 4(a), a more pronounced dependency of the power at 1.5 μm on the co-seeding power. Indeed, this can be explained by the fact that we performed our experiment at much higher co-seeding and output power levels.

Although a co-seeding wavelength at 1040 nm seemed to be the best to achieve as much power as possible at 1.5 μm we decided to use the co-seed laser at 1050 nm with an output power of 2 W for all subsequent power scaling experiments in order to achieve the highest possible suppression of the Yb-ASE and parasitic processes at 1.0 μm. Figure 5(a) shows the final result of our power scaling experiments. At a maximum absorbed pump power of 210 W we obtained 61 W of amplified seed power at 1.5 μm. The corresponding slope efficiency with respect to the absorbed pump power was around 30%. Nonetheless, we still measured a maximum output power of 40 W at 1.0 μm with a slope efficiency of around 23%. This means that our fiber was still not long enough to reabsorb all the amplified co-seed power. The optical spectrum of the amplified seed signal at the highest achieved output power, see Fig. 5(b), showed a peak-to-ASE suppression of around 50 dB (0.5 nm RBW). In addition, the inset in Fig. 5(b) shows the spectrum (40dB peak-to-ASE suppression, 0.5 nm RBW) of the amplified co-seed at an absorbed pump power of 200 W. Unfortunately, during all measurements, especially at pump power above 150 W, we had to deal with problems regarding the stability of the coupling of the co-seed light into the pump end of the fiber. Although the mounting of the fiber at the pump and co-seed launch end was carefully designed, thermal misaligments of the fiber tip led to the mentioned instabilities regarding the co-seed coupling. Subsequently, this limited the final power scaling due to the onset of spiking and self-lasing at around 1.05 μm. However, as we strongly believe that this behavior is only due to thermal misaligments, we are confident that in an all-fiber system, for example, such problems can be avoided.

 

Fig. 5 (a): Amplifier slopes at 1.0 μm and 1.5 μm. (b): Spectrum at 1.5 μm at the highest amplifier output power. Inset: Spectrum at 1.0 μm at an absorbed pump power of 200 W.

Download Full Size | PDF

5. Beam quality measurements

In order to analyze the beam quality of our amplified seed-signal, we used a non-confocal scanning ring cavity, see Fig. 6. Because GWDs are working with a true TEM₀₀ mode, M² measurements as a characterization of the beam quality are not sufficient in this special case. For a detailed description of the mode of operation of such a non-confocal scanning ring cavity, see [19]. In principle one uses the fact that the eigenmodes of the cavity are by design the TEM modes. Thus, a decomposition into the set of TEM modes can be achieved by scanning the cavity and monitoring the transmitted signal. Because the whole procedure only works with a low-power (around 100 mW) and a linearly polarized beam, we sampled a fraction of our amplified seed signal and adjusted the polarization with a quarter-wave plate, a half-wave plate and a polarizing beamsplitter cube and the power with an additional half-wave plate and a second polarizing beamsplitter cube.

 

Fig. 6 Schematic overview of the non-confocal scanning ring cavity that was used for the decomposition of the amplifier signal into the TEM modes. PZT: Piezoelectric transducer.

Download Full Size | PDF

This measurement was performed for different output power levels of the amplifier and the corresponding result is shown in Fig. 7. The TEM₀₀ content was above 90 % for all amplifier output power levels and, even at the highest output power of 61 W, the fundamental mode content was still 91 %. Nevertheless, at high output power one observes a slightly decreasing trend of the fundamental mode content. It is worth comparing our results with those obtained in previous mode scanning measurements on Er³⁺:Yb³⁺ co-doped amplifiers [20] and pure Er³⁺ amplifiers [21]. In [20] mode scans on a commercially available step-index triple clad Er³⁺:Yb³⁺ fiber with a 23 μm core and a V-parameter of >5 and on a specially designed multifilament-core (MFC) fiber with a hexagonal 31/28 μm core were performed. While the triple clad fiber delivered around 20 W of output power with a fundamental mode content of only 82 %, the MFC fiber only delivered a maximum output power of 8.5 W but with a corresponding fundamental mode content of 95 %. In [21], an Er³⁺ doped PCF was used in a MOPA setup and a maximum output power of 70 W was achieved. However, due to pointing instabilities mode scans could only performed up to an output power of 50W and the corresponding fundamental mode content was as low as 80 %. In our experiments we were not only able to achieve output power levels similar to those obtained with the Er³⁺ doped PCF in [21], but we also measured good fundamental mode content above 90 %, only slightly lower than the results obtained with the MFC fiber in [20].

 

Fig. 7 Fundamental mode content in dependency of the amplifier output power at 1.5 μm.

Download Full Size | PDF

6. Conclusion

We reported on the design and fabrication of an Er³⁺:Yb³⁺ co-doped triple clad fiber with a core diameter of 23 μm, a core NA lower of 0.09 and a pump light absorption of 1.5 dB/m. A 22 m long piece of the fabricated fiber was used in a single-frequency counter pumped MOPA setup. The co-seeding method was applied to suppress parasitic processes at 1.0 μm and to increase the amplifier efficiency at 1.5 μm. A maximum output power of 61 W at 1.5 μm and 40 W at 1.0 μm was achieved with a maximum absorbed pump power of 210 W at 976 nm. The slope efficiencies were about 30 % at 1.5 μm and 23 % at 1.0 μm. Nevertheless, due to thermal misaligments of the free space coupling of the co-seed at high pump power, parasitic processes at 1.0 μm finally limited our power scaling experiments. Mode scanning experiments with a non-confocal scanning ring cavity revealed that for all measurements over 90 % of the output power at 1.5 μm was in the TEM₀₀ mode. In comparison with previous mode-scanning results on high power fiber amplifiers [20,21] our MOPA setup not only delivered a comparable output power, but also due to the chosen fiber design a higher slope efficiency and a better beam quality at high output power levels.