High-power fiber lasers have made their impact on numerous industrial and defense-related applications [1], and multi-kW fiber lasers have been realized with good beam quality [2]. However, as the demand for higher laser power continues to grow, thermal management becomes an increasingly important issue. The figure-of-merit quantifying heat generation in an active fiber is the quantum defect (QD), defined as $\mathrm{QD}=1-{\lambda }_p/{\lambda }_s$, where ${\lambda }_p$ and ${\lambda }_s$ are the pump and lasing wavelengths, respectively. Since the pump wavelength is shorter than the lasing wavelength, the QD takes on a value between 0 and 1, and represents the fraction of pump power (in the quantum limit) lost to heat. For example, an aluminosilicate fiber [3,4], which typically is pumped at a wavelength of about 976 nm and lases at about 1030 nm, exhibits a QD of $\sim 5\%$. This indicates a 500 W thermal load on an active fiber operating at 10 kW.

Such QD-related heating can lead to problems ranging from the catastrophic [5–7] to the parasitic [8,9]. Catastrophic failure of optical fiber clearly represents an upper bound to power scaling, while parasitic effects, such as transverse mode instability (TMI), can have a major impact on beam quality. Therefore, if the QD can be reduced, for example to 1%, the thermal load for the same 10 kW power level would be reduced to 100 W, clearly offering significant mitigation of the aforementioned problems.

Approaches to decreasing the QD in fiber lasers come in several forms. They include judicious selection of host glass, such as the use of phosphosilicates [2,10], tandem pumping [11], or simply using brute-force methods to bring the pump and signal wavelengths closer together [12,13]. The latter is particularly challenging as amplified spontaneous emission (ASE) originating from spontaneous emission near a local maximum of the gain curve ($\sim 1030\mathrm{nm}$) can have a significant deleterious impact on performance, particularly on slope efficiency.

Considering a material approach, it was recently shown [14,15] that with sufficient fluorine (F) co-doping into a multicomponent silicate glass, the ytterbium (Yb) emission spectrum is nearly identical to that of fluoride glasses. However, as silicates, fibers made from these materials retain many desirable features, such as high strength and the ability to be fusion spliced to conventional pump fibers. Importantly, as will be shown, the emission spectrum from these fibers is significantly blue-shifted relative to more conventional aluminosilicate glasses, with a local emission maximum near 1000 nm. Hence, it stands to reason that by using these fibers, short-wavelength ($<1000\mathrm{nm}$) lasers should be feasible to realize. Such glasses may also find utility in laser cooling applications [16,17].

Here, low QD ($<1\%$), core-pumped laser operation is investigated in these fluorosilicate core optical fibers. Results indicate that slope efficiencies did not deteriorate due to the proximity of the pump and laser wavelengths, but rather to intracavity splices and background active fiber loss that both contributed to the total (unrecoverable) cavity loss. This low-QD result, coupled with the demonstrated reduction of dn/dT in these glasses [14], could potentially raise the TMI thresholds by up to 10 dB, clearly justifying continued investigation.

The optical fiber investigated herein was fabricated using the molten core method [18]. A precursor material in the form of a powder mixture of $5{\mathrm{YbF}}_3\text{-}71.25{\mathrm{SrF}}_2\text{-}23.75{\mathrm{Al}}_2{\mathrm{O}}_3$ (molar percentage) composition was inserted into a silica capillary tube preform (3 mm inner and 30 mm outer diameterss). This preform then was heated to $\sim 2000°\mathrm{C}$, whereby the core precursor mixture is molten and the pure silica cladding draws into fiber. As the fiber cools, the molten core quenches to a glass state. The resultant optical fiber had a cladding diameter of 125 μm and was coated during the draw with a conventional acrylate coating. Approximately 800 m of fiber length was collected. As a result of the inherent reactivity between the molten core and the softened cladding, silica is incorporated into the core. Consequently, the fabricated fiber is a silicate containing the precursor compounds clad in pure silica. Compositional and refractive index profiles, as well as a scanning electron microscope (SEM) micrograph of the fiber are provided in Fig. 1. The fiber possesses a relatively high $\mathrm{\Delta }n$, and thus high numerical aperture (NA), which may compromise single- or few-mode operation in a large-mode area fiber. However, approaches such as the use of a cladding pedestal design or high-index cladding materials (such as the ${\mathrm{SiO}}_2\text{-}{\mathrm{Al}}_2{\mathrm{O}}_3\text{-}{\mathrm{La}}_2{\mathrm{O}}_3$ glass system) could be used to reduce fiber NA [19]. Greater insights into the reactivity of these fluorosilicate systems can be found in Ref. [14]. For completeness, in said reference, the fiber presented herein is labeled “YbF-SrAlSiF A.” It was also from this fiber that the normalized absorption and emission cross section spectra in Fig. 2 were measured. The upper state lifetime is measured to be around 1270 μs, which is somewhat longer than that found in typical aluminosilicate fibers, but also results in relatively lower cross section values [14].

 

Fig. 1. (a) Compositional and refractive index profiles and (b) SEM image of the fiber. Not shown for reasons of clarity in (a) is the oxygen concentration, %O. However, %O (at. %) = 100 − [%F + %Sr + %Al + %Si + %Yb].

Download Full Size | PDF

 

Fig. 2. Normalized (a) emission cross section (${\sigma }_e$) and (b) absorption cross section (${\sigma }_a$) spectra for Yb-doped commercial aluminosilicate fiber and the fluorosilicate fiber of this study. Normalized data are provided in order to show the effect of F on the spectroscopic properties. The absolute value of absorption (emission) cross section at the peak for aluminosilicate fiber and fluorosilicate fiber is $2.53\times {10}^{-24}{\mathrm{m}}^2$ and $1.48\times {10}^{-24}{\mathrm{m}}^2$, respectively [14].

Download Full Size | PDF

In this Letter, two low-QD cases are studied: (1) pumping at 976.6 nm and lasing at 985.7 nm, and (2) pumping at 981 nm and lasing at 989.8 nm, indicating QDs of 0.92% and 0.90%, respectively. These cases were selected as examples of pumping near the peak of and on the red side of the zero phonon line, respectively. The experimental configuration for both are similar, with a block diagram provided in Fig. 3. Commercial, fiber Bragg grating (FBG)-stabilized, fiber-coupled, single-mode diode lasers were used as pumps. For 976.6 nm pumping, an S31-7602 model laser (Lumentum Operations LLC) was used directly. For 981 nm pumping, the FBG stabilizer of the source (FOL0908A45-H17-977.6, Furukawa Electric Co., Ltd.) was strained in order to tune the pumping wavelength from 977.6 to 981 nm. Matched pairs of FBGs (99.18% reflectivity at 989.77 nm and 39.61% reflectivity at 989.77 nm; 99.02% reflectivity at 985.74 nm and 38.34% reflectivity at 985.77 nm; O/E-Land Inc.) were used to construct the cavity. An isolator was placed between the pump and the cavity in order to avoid having reflections from cavity FBGs destabilize the pumping wavelength, which, as will be shown, is validated in Figs. 4(a) and 4(b). There, the spectra indicate pump leakage to be at the correct wavelength of 976.6 nm.

 

Fig. 3. Experimental setup (an “$\times$” indicates a fusion splice).

Download Full Size | PDF

 

Fig. 4. Output spectra with 976.6 nm pump and Yb-doped fiber length of (a) 19.1 cm, 17.1 cm and (b) 15.4 cm, 11.4 cm showing the evolution of ASE with active fiber length. These plots were measured with a pump power of 624 mW.

Download Full Size | PDF

In order to optimize the lasing conditions and slope efficiencies, experiments started with a relatively long piece of Yb-doped fiber (around 20 cm), which then was shortened in roughly 1.5 cm increments to a final length of $\sim 10\mathrm{cm}$. For each increment, the output spectrum was recorded using an optical spectrum analyzer (OSA), and the laser output power versus pump power was measured using a calibrated optical power meter. Figure 4 is a representative example of the spectral measurements for 976.6 nm pumping using four different active fiber lengths. A similar trend (described in the next paragraph) of spectral variation was also observed for the 981 nm pumping case, and plots are, therefore, not shown here. Figure 5(a) provides the slope efficiency measured for the different fiber lengths, whereas Fig. 5(b) shows lasing data (versus launched pump power) at the maximum slope efficiency, including a comparison to theoretical results. For convenience, in Fig. 5(b) the 981 and 976.6 nm pumping scales were offset for visual clarity. Pump leakage was subtracted from all the measured power data, and the model utilized for this study can be found in Ref. [20].

 

Fig. 5. (a) Slope efficiency with different Yb-doped fiber lengths for the two low-QD cases (the curves are provided only as visual aids to the reader) and (b) theoretical and experimental output power versus launched pump power at near the optimal fiber length (11.4 cm for 976.6 nm pumping and 15.6 cm for 981 nm pumping).

Download Full Size | PDF

The spectra and slope efficiencies vary significantly with fiber length. With a very short length (for example, 8 cm), the pump power is not absorbed completely and much of it leaks from the cavity, therefore degrading lasing efficiency. When the fiber is longer than optimal, near-complete pump absorption occurs. However, the noninverted length of fiber imparts reabsorption to the signal wavelength, which leads to a greater likelihood of ASE, thereby also results in a degradation of the slope efficiency. With longer fiber lengths, ASE ultimately dominates the lasing process, and self-oscillation occurs near the ASE peak wavelength ($\sim 1022\mathrm{nm}$). This process is exemplified in Fig. 4. First, Fig. 4(a) shows the emission spectra when a longer-than-optimal fiber length is employed. At 19.1 cm, self-oscillation is observed, while for 17.1 cm, significant ASE is produced. Figure 4(b) provides spectra near the optimal fiber length. With relative lasing power set to the same level (65.7 dB), a fiber length of 15.4 cm leads to a leaked pump power level of $\sim 12\mathrm{dB}$ lower but with an ASE level that is $\sim 4.4\mathrm{dB}$ higher than when using 11.4 cm of fiber.

It is clear that greater pump absorption will increase the slope efficiency, while greater ASE will decrease the slope efficiency. Therefore, with decreasing length from $\sim 20\mathrm{cm}$, the slope efficiency is expected to increase at first as the ASE level is reduced, reach an optimal point, and then decline as incomplete pump absorption takes place. This is precisely the trend observed in Fig. 5(a). The power data for the near-optimal fiber lengths (11.4 and 15.6 cm for 976.6 and 981 nm pumping, respectively) for the two cases are shown in Fig. 5(b). Corresponding slope efficiencies are measured to be 62.1% and 56.8%. Note that the output power was pump-limited when lasing at 989.8 nm.

The slope efficiency does not reach the theoretical quantum limit ($>99\%$) mainly because of the splice losses between the Yb-doped fiber and the cavity FBGs (written into 1060-nm-type single-mode fiber) as well as background loss. Background loss ($\sim 1.36\mathrm{dB}/\mathrm{m}$) is mainly due to scattering and impurity absorption, while splice loss ($\sim 0.16$ and $\sim 0.48\mathrm{dB}/\text{splice}$ for the 976.6 and 981 nm pumping cases, respectively) is influenced mainly by spatial mode competition within the cavity, with the observation that the output power may change significantly when bending or twisting the fiber. This, coupled with some variation of splice loss each time, leads to the data observed in Fig. 5(a). Pumping at 981 nm indicated a somewhat lower maximum slope efficiency relative to pumping at 976.6 nm. This can be explained by its longer optimal length and therefore larger total background loss, in addition to greater splice loss. Work is currently underway to reduce the background losses in these fibers. Additionally, a reduction in the fiber numerical aperture should render improved splice quality. Modeling results indicate that if the background loss can be largely eliminated, along with attaining a 0.16 dB loss per splice for both cases, the achievable slope efficiencies would be 80.7% for 981 nm pumping and 68.3% for 976.6 nm pumping. Modeling results also indicate that with further improvements to splice quality, optimization of FBG reflectivity, and concomitant fiber length, slope efficiencies approaching the quantum limit are feasible. However, low-QD operation is of much more significance in any subsequent power amplifier stages.

Finally, it also is observed that a higher pump leakage occurs for the 981 nm pump case for the same length of fiber. This is shown in Fig. 6, where the difference between signal output and pump leakage powers as a function of fiber length is provided. This observation can be explained by the difference in the cross sections at the various wavelengths. The absorption cross section at 976.6 nm is approximately twice that at 981 nm, which leads to greater pump absorption for given active fiber length. At the same time, the emission cross section at 985.7 nm also is slightly larger than at 989.8 nm, which results in somewhat more effectual stimulated emission. These two processes, taken together, bring about the results shown in Fig. 6, where 976.6 nm pumping not only suggests a larger power difference for a given length, but also a stronger dependence on change in that length. The uncertainty associated with the measurements in Fig. 6 are roughly $\pm 0.5\mathrm{dB}$ due to the resolution of the OSA.

 

Fig. 6. Difference between pump leakage and lasing powers at different Yb-doped fiber lengths for the two low-QD cases. These were obtained using 265 mW of pump power. The lines are simple fits to the data meant as visual aids to the reader.

Download Full Size | PDF

In conclusion, demonstrated here were two less-than-1% QD fiber laser configurations based on an Yb-doped multicomponent fluorosilicate optical fiber, having reached a maximum slope efficiency of 62.1%. Higher slope efficiencies can be expected with less splice and background losses. Modeling results indicate that with additional laser optimization (FBG reflectivity and active fiber length), slope efficiencies approaching the quantum limit should ultimately be possible in a core-pumped configuration. Since the fluorosilicate glass host possesses properties that may be effective in reducing the thermal load in active fiber, thereby improving the laser system, further work is justified. As such, a double-clad version of this fiber is currently under development and is intended for a master oscillator power amplifier configuration, to be seeded by the sources demonstrated here. Results are forthcoming.