1. Introduction

Phosphate glass is a unique host for lanthanide ion doping: high concentrations of rare earth (RE) ions can be incorporated into phosphate glass, the high multi-phonon relaxation rate of ⁴I_11/2 Er³⁺ level in the phosphate glass host provides for efficient and irreversible energy transfer from Yb³⁺ ions to the ⁴I_13/2 Er³⁺ long-living upper laser level [1] and Yb-doped phosphate glass is highly resistant against photodarkening even at high Yb³⁺ concentration [2]. The RE doping levels of phosphate glass are, on average, an order of magnitude higher that achievable in Al-doped and P-doped silica, which is typically used as a core material of silica fibers, so the optical fibers from phosphate glass are more heavily doped than silica fibers [3]. High doping levels of phosphate fibers make it possible to obtain high gain [4] and high output power [5] per unit length, thus reducing the length of active fibers in comparison to silica based fibers.

Length reduction of active fibers for lasers and amplifiers is beneficial for reducing undesirable nonlinear interactions that occur for high-intensity laser radiation in the fiber core. The nonlinear effects such as self-phase modulation, stimulated Raman scattering, and stimulated Brillouin scattering limit the achievable parameters of fiber systems [6]. These effects are well pronounced in long waveguides [7], while reducing the length of the fiber lasers and amplifiers is a reasonable way to suppress them.

Phosphate optical fibers with high gain per unit length are efficient for designing single-frequency fiber lasers with high output power (of the order of hundreds of milliwatts) [8, 9]. High concentrations of RE-ions in the phosphate core permit high output power using small fiber length, a requirement for large mode spacing.

A substantial drawback of the use of phosphate glasses as a material for fiber fabrication in comparison to silica is their low stability; exposure to air moisture causes degradation of the phosphate fibers over time. Moreover, due to the sharp difference in their physical properties, splicing of phosphate and silica fibers is difficult.

The development of a composite fiber with a Nd-doped phosphate core and silica cladding, which would unite the advantages and negate the drawbacks of phosphate and silica glasses, was first proposed in 2006 [10]. High RE-ion concentrations in the phosphate core of the composite fiber allow fiber length reduction in comparison with silica fibers. The silica cladding provided high mechanical strength and protected the phosphate core from air moisture while making it easier to splice with silica fibers.

In this paper we describe a composite optical fiber with Er/Yb co-doped phosphate glass core in a silica glass cladding as well as cladding pumped fiber laser. Fabrication of a fiber is discussed along with its optical properties. The lasing efficiency of the composite fiber is found to be significant when using cladding pumping.

2. Fiber fabrication and testing

The laser glass selected for the fiber core [11], is composed of the basic glassformer P₂O₅ combined with boron oxide and alumina: 65% mol P₂O₅, 7% mol Al₂O₃, 12% mol B₂O₃, 9% mol Li₂O and 7% mol RE₂O₃. The absolute concentrations of the active ions were: 1.7 × 10²¹ cm⁻³ Yb and 1.3 × 10²⁰ cm⁻³ Er. The glass was sintered in high frequency heated Pt crucible.

The fiber blank was fabricated using a rod-in-tube technique. Using a tubular drill, cylindrical rods with diameter of 4.4 mm were drilled out of the glass block. The surfaces of the rods were etched by hot phosphoric acid in order to remove the contaminated surface layer. The prepared 4.4 mm phosphate glass rods were inserted into a sealed silica glass (Heraeus F300) tube, then the blank was consolidated in a fiber drawing furnace and drawn into canes with an outer diameter of 2 mm. The diameter of the phosphate core at this stage was about 1 mm. One more silica tube has been jacketed to one of the drawn cane using a glass-working lathe with a propane-oxygen burner to achieve desirable core-to-cladding diameter ratio. The outer diameter of the fiber blank at this stage was 13.5 mm while the phosphate core diameter remained about 1 mm. The blank was then ground and polished to a square cross-section in order to provide efficient pump absorption by the core in case of cladding pumping. Finally, the blank was drawn into a fiber with a low refractive index polymer coating to form pump cladding. All the described operations were performed at the same temperature regimes usually used for silica fiber fabrication.

Measured core-cladding refractive index difference in the fiber blank at 633 nm using P-104 preform analyzer (Photon Kinetics) was 0.05. The refractive indexes in bulk silica and phosphate glass were measured at 589 nm. Measured refractive indexes were 1.458 and 1.537 in silica F300 (Heraeus) and phosphate glass, respectively. The refractive index difference for bulk glasses (0.079) was higher than in the fiber blank. The origin of refractive index reduction could be diffusion as well as refractive index changes generated by stress-optical effects; such effects may have appeared as a result of large difference in transitions temperatures T_g and thermal expansion coefficients of the core and cladding.

The electron microscope image of a cleaved fiber end is shown in Fig. 1(a). The side of the nearly squared silica cladding cross-section is about 100 μm and the core diameter is about 13.5 μm. The glass component distribution across the fiber cross-section (Fig. 1(b)), measured using EDS, show mutual diffusion of the core and cladding glasses. As a result, P₂O₅ concentration reduces to 30 mol % comparing to initial bulk glass and SiO₂ penetrates into the core. Concentration of SiO₂ in the center of the core is 50 mol %, Yb₂O₃ - 4-5 mol %.

 

Fig. 1 Electron microscopy image of a cleaved fiber end (a) and distribution of glass components across the fiber core (b).

Download Full Size | PDF

The numeric aperture of the fiber at 633 nm was measured to be 0.32. The refractive index difference between the core and cladding in the fiber, calculated using measured NA, is 0.035. At the core diameter of 13.5 µm, this corresponds to a normalized frequency V = 9 at λ = 1.55μm. The reduction of the refractive index difference in the fiber, in comparison to the fiber blank, is mainly due to significant diffusion of glasses.

Manufactured fiber is multimode, but for lasers and amplifiers it is desirable to use single mode fiber. One way to make proposed fiber single mode is to reduce core diameter to approximately 3 µm. Such fiber could be useful for designing single frequency lasers. Another way is to decrease the core-cladding index difference by increasing the refractive index of the cladding. It could be done by making the fiber cladding from Ge-doped silica. Doped silica could be manufactured, for example, in MCVD process.

The optical loss spectrum of the fiber is presented by Fig. 2(a). Optical loss in the wavelength range 1.2-1.3 μm, where there is no absorption of Yb³⁺ and Er³⁺ ions, is 2 dB/m. High optical losses is due to the absorption of traces of transition and lanthanide metals in combination with OH-groups absorption. In the bulk laser glass used for the fiber fabrication, the OH-groups absorption at λ = 3.33 μm was 475 dB/m. The ratio of absorption coefficients at this wavelength and at the laser wavelength 1.54 μm in phosphate glasses [12] makes it possible to estimate the OH-groups absorption in the fiber at the lasing wavelength as 0.5 dB/m.

 

Fig. 2 Optical loss spectrum (a) and cladding absorption (b) of composite fiber.

Download Full Size | PDF

We believe that the losses at the lasing wavelength can be lowered several times by using higher purity platinum crucibles and starting chemicals as well as by deeper dehydration of the glass melt during its synthesis. Nevertheless due to small fiber length the indicated relatively high losses were not fatal for efficient lasing.

The small-signal absorption of Yb³⁺ ions, measured with a white light source launched to inner cladding, were 0.65 dB/cm at λ = 975 nm and 0.15 dB/cm at λ = 914 nm (Fig. 2(b)). The measured erbium core absorption at the 1535 nm wavelength was 1.5 dB/cm (Fig. 3(a)).

 

Fig. 3 Erbium core absorption (a) spectrum of composite fiber and luminescence spectra of erbium in bulk glass (black line) and composite fiber core (red dashed line) (b).

Download Full Size | PDF

The shape of the Er³⁺ ion emission band in the composite fiber differed slightly from that of the initial bulk glass (Fig. 3(b)). This is obviously caused by the changes in Er³⁺ ions’ neighborhood due to silica diffusion into the core during the fiber drawing process. The relaxation decay of Er³⁺ ions in both the initial bulk glass and the composite fiber was exponential with lifetime of 7.6 ms.

The manufactured optical fiber was mechanically strong enough for handling. Splicing the composite fiber with silica fibers proved possible without bubble formation in the core. The measured splice losses between composite fiber and single mode silica fiber with a Ge-doped core (measured using light launching into single mode fiber) in the range 1.2-1.3 µm was lower than 1 dB.

3. Fiber laser

The laser tests of the fabricated fiber were held in the arrangement shown in Fig. 4. The laser resonator was formed by the cleaved fiber ends, each providing 4% reflections. A semiconductor laser diode array was used as a pump source. The linewidth of the pump spectrum was approximately 4 nm (full-width at half-maximum); the center wavelength shifted red with increasing pump power. The pump was launched into Er/Yb-doped fiber via a lens. In order to separate the pump and laser radiation the appropriate dielectric mirrors were used. To measure laser output parameters accurately we considered losses at the dielectric mirrors and lenses. These losses were measured separately using 980 nm and 1550 nm laser sources. The output power of the laser from both fiber ends was measured separately and summarized. Considering measured losses, the values of output power from both fiber ends were nearly the same.

 

Fig. 4 Laser set-up.

Download Full Size | PDF

Launched pump power was measured by cutting the fiber. The length of cutback was 1-2 cm. The pump absorption and back reflection at the output face of the short fiber piece was taken into account to determine the launched pump power. Figure 5 and 6 present the dependencies of the laser output power at various fiber lengths versus launched and absorbed pump power respectively. With increasing pump power the center wavelength of pump spectrum increases, so pumping was in the range 962-971 nm. In this spectral range small-signal absorption remains nearly constant: 0.13 dB/cm, this is approximately four times lower than maximum absorption at 975 nm.

 

Fig. 5 Output power versus launched pump power for various fiber lengths.

Download Full Size | PDF

 

Fig. 6 Output power versus absorbed pump power for various fiber lengths.

Download Full Size | PDF

The maximum slope efficiency with respect to the launched pump power (27-28%) was observed at fiber lengths of 43-57 cm. The threshold in this case was about 200 mW. At smaller fiber lengths, the efficiency with respect to the launched pump power decreased due to insufficient pump absorption and at larger fiber lengths, due to increasing propagation losses.

A slope efficiency of 28% is lower than slope efficiency of 34% achieved in phosphate fiber using cladding pumping with respect to coupled pump power (taking into account the coupling loss of the pump) [5]. This slope efficiency was obtained at noticeably smaller fiber lengths than in this work - 7 cm. This is probably due to higher pump absorption per unit length owing to better matching of the pump wavelength to the Yb³⁺ absorption peak and higher core to cladding area ratio.

The maximum slope efficiency with respect to the absorbed pump power (39%) was observed at a fiber length of 24 cm. At larger fiber lengths, the slope decreases due to excess propagation losses. Maximum slope efficiency of composite fiber (39%) is the same as that obtained from a 7-cm-long phosphate fiber and close to the efficiency obtained from a 1.4-m-long P-doped silica fiber laser with 980 nm pumping (38%) [13].

At a fiber length of 43 cm, the laser output spectrum has several peaks around 1535 nm with total linewidths of 3 nm and around 1545 nm with the same total linewidth. At smaller fiber lengths, the laser action predominantly shifted to 1535 nm, and at larger fiber lengths, to the longer-wavelength region, due to rising Er³⁺ reabsorption. No significant bleaching at the pump wavelength has been observed. No lasing has been observed near 1.03 µm even at maximum launched pump power of 7 W.

Laser efficiency was also measured with pump spectrum better matched to the Yb³⁺ absorption peak. It was obtained then pump source worked at higher output power. Only part of the total pump power, equal to 6.5 W, was launched to the Er/Yb co-doped fiber. Total pump power in this case could not be launched to the fiber because of degradation of low index polymer near the input end of the fiber. The center wavelength was 973 nm. Pump absorption was only two times higher than for pumping in the range 962-971 nm. Laser efficiency of 30% versus launched pump power was obtained with the fiber length of 30-50 cm.

4. Conclusion

A composite fiber with Er/Yb co-doped phosphate glass core in silica glass cladding was manufactured and tested. The silica cladding of the composite fiber provided its high mechanical strength and protected the phosphate core from interaction with air moisture. It also permits easy splicing of these types of fibers with silica fibers. We were able to demonstrate, for the first time, suitable laser efficiency at 1.55 µm using these fibers and cladding pumping. The slope efficiency with diode cladding pumping in the vicinity of 980 nm reached 28% with respect to the launched pump power and 39% with respect to the absorbed pump power. Due to high doping levels of the phosphate core, even without exactly matching the pump wavelength with Yb³⁺ absorption peak, the optimal fiber length was about 50 cm, which is several times less than the optimal length of a similar silica fiber [13].