We report on the generation of 94 W continuous wave output power at 980 nm using an Yb-doped fiber laser. This is achieved using an ultra large-mode-area rod-type photonic crystal fiber pumped at 915 nm. To the best of our knowledge this is the highest output power close to diffraction-limited beam quality (M2 about 2.2) achieved in this wavelength range from fibers so far. The experimental results are supported by detailed numerical simulations that provide a deeper understanding of the laser process, in particular the competition with the 1030 nm emission.
©2008 Optical Society of America
High brightness sources at 980 nm are very interesting for many different applications such as pumping of high power core-pumping of Ytterbium- and Erbium-doped fiber oscillators and amplifiers lasers. Moreover, with frequency-conversion to 488-490 nm blue radiation, such sources can also replace bulky inefficient argon ion lasers.
The possibility but as well the challenges to achieve laser operation of Yb-doped fibers at wavelengths around 980 nm arise from the spectral emission and absorption characteristics of trivalent Yb-ions in a silica host (see spectra in ). Around the 980 nm emission peak the absorption and emission cross sections are roughly equal, which implies that more than 50% of the Yb-ion population has to be excited to overcome re-absorption and create gain. Additionally, the transparency level at the second emission peak (around 1030 nm) is much lower (about 5% excitation), thus producing strong undesired gain or amplified spontaneous emission (ASE) in this wavelength region. For conventional (step-index) double clad fibers typically used for efficient brightness conversion of high power low brightness pump sources, this parasitic gain limits the utilizable fiber length. This fact, together with the small overlap of the pump field with the doped area in such fibers, hampers pump light absorption and, therefore, the efficiency of the process at 980nm.
One way to overcome this problem is to carry out spectral gain discrimination using, for example, ring doping [2-4]. This technique reduces the overlap of the laser field with the doped area, thus reducing gain at all wavelengths. A possible higher population inversion then increases the gain around 980 nm relative to other wavelengths because of the larger emission cross-section. The drawback of this method, however, is its limitation to relatively small intrinsically single-mode cores, as higher order modes would potentially exhibit a stronger overlap with the doped ring and preferably oscillate at 1030 nm. Once again this restricts the dimensions of the pump core and, with it, the power scalability of the technique. Other techniques to suppress the unwanted emission at longer wavelengths include the incorporation of long period gratings  or solid-core photonic bandgap fibers  as in-line ASE filters, or the use of a distributed feedback laser configuration . So far, all experimental results reporting fiber laser sources at 980 nm were limited to a few W.
In this contribution, we analyze the viability of a 980nm high power fiber laser using an extremely large-mode-area photonic crystal fiber similar to the one presented in . As will be shown, thanks to its unique geometrical dimensions, this fiber allows the existence of preferential gain at 980 nm without further sophisticated parasitic gain suppression schemes. In order to model and gain a deeper insight into the laser process, a numerical simulation solving the steady-state rate equations for the forward and backward propagating signals including pump, laser and ASE radiation has been developed. The results support the possibility of efficiently converting 915 nm pump sources to high brightness, high power 980 nm radiation in such a photonic crystal fiber laser.
Recently, we have demonstrated power scaling in the 980 nm wavelength range to 45 W . Here we report on the experimental generation of close to diffraction-limited 94 W output power in the 980 nm wavelength region using the above mentioned rod-type photonic crystal fiber in a laser configuration pumped at 915 nm with a slope efficiency of ~63%. To the best of our knowledge, this is the highest brightness achieved in this wavelength range so far, being more than an order of magnitude higher than previously published results.
2. Viability of rod-type fiber as 980 nm laser source
2.1 Basic considerations
This section is devoted to provide a more quantitative discussion of the competing emissions at 980nm and 1030 nm under 915 nm pumping. Following ref. , the undesired gain at 1030 nm can be expressed in terms of the gain at two other wavelengths, namely the wavelength of maximum emission cross-section (975 nm) and the pump wavelength (915 nm). Using the cross-sections given in , the expression becomes (in decibels) G1030=0.25G975+0.72βαP, where β approximately equals the ratio of pump core area to effective mode field area of the laser mode (Aclad/Aeff) and αP denotes the pump absorption. In a typical double clad fiber suitable for high power operation the pump and signal core sizes are 30 µm and 250 µm respectively; this gives a value of β near 70. Thus, leaving the contribution of the laser gain G975 aside, it can be seen that for every decibel in pump absorption the gain at 1030 nm increases ~50 dB. This strong gain increase implies that the amplified spontaneous emission (ASE) will dominate the emission even for 1 dB (20%) pump absorption. Therefore, in this type of fiber it is not possible to obtain an efficient laser operation at 980 nm.
In contrast, the Yb-doped rod-type photonic crystal fiber used in this experiment possesses an 80 µm diameter core, supporting a few transverse modes, with an ytterbium-doped zone diameter of 61µm surrounded by high-NA 200 µm diameter pump cladding. Thanks to this extremely large mode area, the ratio β in this fiber is less than 8. Thus, the gain at 1030 nm only increases by 5.7 dB for each dB of pump absorption. If a manageable gain of 50 dB before the onset of strong ASE increase is assumed (as stated in , a pump light absorption up to ~9 dB (87%) is feasible and thus efficient pump conversion at high average powers is possible.
2.2 Simulation of laser process
Further insight can be gained by appropriately simulating the laser process. In order to do that we have developed a numerical tool that solves the steady-state rate equations for the forward and backward propagating signals (pump, laser and ASE radiation). The numerical tool is based on a model that can be regarded as a spectrally resolved extension of that presented in . The set of equations (1) governing the steady state interactions between all the signals and the population inversion along the fiber length is:
In these equations N is the Yb-ion concentration, N2 is the population density of the excited state, σa(λ) and σe(λ) are the wavelength dependent absorption and emission cross-sections of Yb-ions respectively. Ipump is the intensity at the pump wavelength, and Isi are the intensities of the ASE at the rest of the M wavelengths in which the simulation range has been divided. Δλs is the spectral resolution of the simulation. The superscript ± indicates either forward (+) or backward (-) propagation of the signals. λj represent wavelength (with j=pump or s, to denote pump signal or ASE). τ is the spontaneous emission lifetime, A is the doped area of the fibre, and Aeff and Aeff pump are the effective modal area in the signal and pump core respectively. Γpump and Γs are the overlapping factors between the doped area and the pump or the ASE modes respectively. αpump and αs are the attenuation factors of the fibre for the pump and ASE wavelengths. Finally, h is the Planck constant and c the speed of light in vacuum.
These equations together with the appropriate boundary conditions allow us to simulate different cavity configurations. Thus, the boundary conditions used to model our experimental conditions are (2):
Ppump input and Ppump2 input are the pump powers launched at each end of the fiber (z=0 and z=L respectively). This allows the analysis of bidirectional pumping configurations. η1 and η2 are the pump coupling efficiencies. Finally, R1(λ) and R2(λ) are the wavelength dependent reflectivities at each of the fiber ends. Thus, it can be seen that dichroic mirrors (of capital importance for the experiments described inhere) can be easily incorporated in the model.
As can be interpreted from the equations above, this model considers the pump as a monochromatic signal. Besides, this model considers the fundamental transverse mode only and the longitudinal modes of the laser cavity are neglected. However, these assumptions are reasonable for the experimental configurations and effects we want to analyze in this paper, as demonstrated by the good agreement between theory and experiments.
This model allows us to simulate in detail the complex interplay between the ASE, the inversion level, the fiber cross-sections and the pump and laser power along the fiber in a user defined spectral window. Thus, the numerical tool provides us with the means to theoretically analyze all the experiments reported in the following sections.
The parameters used for the simulations (that match our experimental conditions) are: fiber length: 1.2 m, doped diameter: 61 µm, pump diameter: 200 µm, Yb3+ ion concentration: 3.2×1025 m-3. The absorption and emission cross-section data of Yb-ions is obtained from . The 915 nm pump radiation is coupled into one side of the fiber with an efficiency of 85%. Additionally, at first the feedback of both fiber ends is set to zero in order to observe the development of the amplified spontaneous emission (ASE) in both the forward and backward propagating directions (with respect to the pump). The results of the simulation for different pump powers are shown in Fig. 1(a)-1(c).
A preferential emission at the wavelength peak around 976 nm with increasing pump power can be seen for both ASE directions (Fig. 1(a) and (b)). The slight difference between the forward and backward ASE spectra is caused by the fact that the inversion naturally reaches the transparency level first on the pump side of the fiber. This means that more net gain at 976 nm is available for the backward ASE. At the same time the forward ASE is subjected to stronger re-absorption. This can be better understood by observing the developing inversion distribution along the fiber (Fig. 1(c)). The inversion exceeds the transparency level (~50%) at an arbitrary fiber position as soon as the local pump power reaches the corresponding transparency power, which is around 9.4 W in our case. This is accomplished everywhere along the fiber for a launched pump power of about 20 W, which explains the dominating 1030 nm peak for powers lower than this threshold. Furthermore, using the cross-sections of the fiber it can be calculated that a gain at 976 nm higher than that at 1030 nm is generated above an inversion level of 55%. As can be seen in the graphs this leads to a higher 976 nm net gain for pump powers of 25 W and above.
However, some amount of feedback is needed to efficiently convert the pump light into 980 nm radiation. In order to study this, we introduce in the simulation a back coupling of the forward ASE with 80% reflection around 976 nm (owing typical coupling losses with free space optics) together with a steep 30 dB reflectivity drop for wavelengths above 1000 nm (i.e. we introduced a dichroic mirror in the model). The results are displayed in Figs. 2 and 3.
The simulations predict a strong signal output at a central wavelength of 978 nm with a full width half maximum of about 2 nm. The 30 dB reflection suppression of the undesired ASE around 1030 nm prevents significant signal build-up in this wavelength range and thus keeps the inversion level well above 50% all along the fiber (see Fig. 3, blue solid curve). Since there is no feedback introduced at the pump side, virtually all of the radiation emerges at this fiber end (see Fig. 2, red curve). Furthermore, the simulated total output power is insensitive to small changes in the reflectivity ratio due to the negligible forward radiation power. This means that our simulations should be accurate enough for the range of the experimental backcoupling uncertainty. With a maximum experimentally available pump power of 200 W and 85% coupling efficiency the simulation predicts a total of 98 W output power, most of which (97.7 W) is concentrated in a 6nm range around the 978 nm peak. This results in a total conversion efficiency of 58%. The reason for this is that only 71% (5.4 dB) of the pump power is absorbed along the fiber, while about 49 W remain unabsorbed (see Fig. 3, black solid curve). This indicates that for better pump absorption a longer fiber should be used, but experimentally the fiber length available was limited to 1.2 m. Another possibility to increase absorption is to operate the laser in a double pass pump configuration, i.e. re-image the forward pump back into the fiber. With an assumed reflectivity of 80% for the remaining pump power, the simulation then predicts a total 122W output power with virtually the same spectral distribution as for the single pass configuration. The slightly different pump and inversion distributions are shown in Fig. 3 (dashed curves). The overall pump absorption increases to 9.1 dB and the conversion efficiency amounts to 72%.
It has to be noted, though, that the simulation results (especially the power ratio of the peaks at 980 nm and 1030 nm) strongly depend on the reflectivity drop for the back-coupled radiation. At a certain reflectivity ratio the back-coupled ASE around 1030 nm is strong enough to deplete the inversion and exhibit the highest net gain, thus dominating the lasing process. In the simulation this point is reached for a reflectivity ratio R980/R1030 of about 10dB. This highlights the importance of a proper feedback control in the following laser experiments.
3. Experimental results in laser configuration
The experimental laser configuration is depicted in Fig. 4. The 1.2 m long rod-type photonic crystal fiber described above is pumped on one side with a high power fiber-coupled pump diode. The 200 µm diameter delivery fiber emits highly multimode radiation at a pump wavelength around 915 nm, which is imaged into the pump core with a coupling efficiency of 85%. A dichroic mirror thereby separates any backward radiation above 950 nm from the pump light, for which it is highly transmittive. As already stated before, suppressing lasing at 1030 nm is of capital importance in this configuration in order to maintain a high inversion level along the fiber and avoid re-absorption of the 976 nm signal. Therefore, both fiber ends have been angle polished to suppress undesired feedback. In order to couple the pump and only the desired part of the signal spectrum back into the fiber, the emerging forward radiation is collimated and filtered using three dichroic mirrors. The first one separates the pump radiation, which in the double pass configuration is coupled back with a highly reflective mirror. The other mirrors efficiently eliminate the ASE around 1030 nm from the signal feedback.
In a first measurement the fiber is pumped at different powers without any feedback in order to investigate the development of the ASE and compare the results with the numerical simulations. Fig. 5 shows the measured spectral formation of the backward ASE signal with increasing pump power. At a certain pump power level (~25W) the transparency intensity for 976 nm is exceeded throughout the fiber. This is indicated by a stronger growth of the 976 nm peak relative to the 1030 nm ASE level. The spectral evolution of the ASE agrees well with the results of the numerical simulation (compare with Fig. 1.b).
The introduction of the tailored feedback for the forward ASE radiation with help of the dichroic mirrors leads to the desired lasing operation around 980 nm. The spectrum of the backward fiber output in double pass pump configuration at the highest available pump power of 200 W is shown in Fig. 6. Lasing is observed in a narrow wavelength range of about 5 nm (FWHM) around 979 nm, where numerous competing longitudinal modes, characteristic of a free running fiber laser, can be observed (see inset of Fig. 6). The suppression of the parasitic ASE around 1030 nm is better than 35 dB in both single and double pass configuration. It can be pointed out that the measured spectral characteristics are in excellent agreement with the numerical simulations (see Fig. 2).
Figure 7 shows the output characteristics of the laser for both pump configurations. The maximum achieved output power of 76 W and 95 W respectively is pump power limited, as can be seen by the linear increase of the output powers. With the help of another dichroic mirror splitting the spectrum at a wavelength of about 1000 nm, a maximum of 1% (0.7 W and 1 W respectively) output power was measured to be contained in the broad ASE peak around 1030 nm. Integration and comparison of the spectral power fractions in both the desired and undesired spectral regions suggests that an even smaller portion of the total power (0.5%) is actually contained in the spectral region around 1030nm. This small discrepancy could arise from the non-perfect transmission characteristics of the dichroic mirror used to split the spectrum.
The quantitative difference between the simulated and measured output powers (98 W versus 75 W and 122 W versus 94 W) can be explained by additional cavity losses, i.e. internal propagation losses in the rod, which are not included in the simulation model. In contrast to the output power discrepancy the simulation prediction (49 W) and measurement (52 W) of the remaining pump at maximum power in single pass pump configuration agree very well.
The evolution of the output beam quality with increasing output power is shown in Fig. 8. Up to 25 W output power the beam remains almost diffraction-limited, having an M2 factor of about 1.2. Further increases in the output power, with the feedback mirror adjusted to obtain maximum power, lead to a linear degradation of the beam quality up to an M2 of about 2.2 at 75 W output power. At even higher powers the beam quality stays constant. Thus, at the highest output power of 95W the M2 has a value of about 2.2. This progressive degradation of the beam quality is due to the excitation of a higher order mode, which coexists with the fundamental mode of the fiber (see intensity profile evolution in Fig. 9). On the other hand, it is also possible to adjust the feedback mirror to obtain a better beam quality. However, this beam quality improvement comes at the cost of output power. For example, the M2 of 1.7 at 47 W can be improved to 1.25, but with the output power dropping to 36 W. It appears that in the present configuration the overlap of the fundamental mode alone with the doped cross-section is not sufficient to extract the inversion of the few-mode fiber at elevated pump powers.
4. Conclusion and outlook
In conclusion, we have discussed and shown the utility of an 80/200 rod-type photonic crystal fiber as a high brightness high power laser source at wavelengths around 980 nm. By pumping this fiber at 915 nm we obtained a maximum of 75 W continuous wave using a single pass pump configuration output power in a nearly diffraction-limited beam. In order to achieve this we used dichroic mirrors to tailor the laser feedback, so we could achieve more than 35dB suppression better of the parasitic ASE around 1030 nm. Furthermore, we developed a numerical tool to adequately simulate the emission process. The results from these simulations agree well with the experimental data and give further insight into the fiber laser dynamics.
Even higher output powers may be achieved by optimizing the feedback conditions and the fiber length for the given fiber and pump source parameters. Thus, as has been shown, the conversion efficiency can be increased by using a double pass configuration for the pump radiation. Using this pump configuration we have managed to obtain 94 W output power. Pulsed operation at 980 nm can even be possible by inclusion of a q-switching element inside the cavity. This would open the door to externally frequency-doubled radiation in the blue.
References and links
1. R. Paschotta, J. Nilsson, A. C. Tropper, and D. C. Hanna, “Ytterbium-Doped Fiber Amplifiers,” IEEE J. Quantum Electron. 33, 1049–1056 (1997). [CrossRef]
2. J. Nilsson, J. D. Minelly, R. Paschotta, A. C. Tropper, and D. C. Hanna, “Ring-doped cladding-pumped single-mode three-level fiber laser,” Opt. Lett. 23, 355–357 (1998) [CrossRef]
3. D. B. S. Soh, C. Codemard, J. K. Sahu, J. Nilsson, V. Philippov, C. Alegria, and Y. Jeong, “A 4.3 W 977 nm ytterbium-doped jacketed-air-clad fiber amplifier,” in Advanced Solid-State Photonics, OSA Technical Digest (Optical Society of America, 2004), paper MA3.
4. R. Selvas, J. K. Sahu, L. B. Fu, J. N. Jang, J. Nilsson, A. B. Grudinin, K. H. Ylä-Jarkko, S. A. Alam, P. W. Turner, and J. Moore, “High-power, low-noise, Yb-doped, cladding-pumped, three-level fiber sources at 980nm,” Opt. Lett. 28, 1093–1095 (2003). [CrossRef]
5. L. A. Zenteno, J. D. Minelly, M. Dejneka, and S. Crigler, “0.65 W single-mode Yb-fiber laser at 980 nm pumped by 1.1 W Nd:YAG,” in Advanced Solid State Lasers, OSA Technical Digest Series (Optical Society of America, 2000), paper MD7.
6. V. Pureur, L. Bigot, G. Bouwmans, Y. Quiquempois, M. Douay, and Y. Jaouen, “Ytterbium-doped solid core photonic bandgap fiber for laser operation around 980 nm,” Appl. Phys. Lett. 92, 061113 (2008). [CrossRef]
7. L. B. Fu, M. Ibsen, D. J. Richardson, and D. N. Payne, “Three-level fiber DFB laser at 980 nm,” in Proceedings of the Optical Fiber Communication Conference (2004).
8. J. Limpert, O. Schmidt, J. Rothhardt, F. Röser, T. Schreiber, A. Tünnermann, S. Ermeneux, P. Yvernault, and F. Salin, “Extended single-mode photonic crystal fiber lasers,” Opt. Express 14, 2715–2720 (2006). [CrossRef] [PubMed]
9. F. Röser, D. N. Schimpf, C. Jauregui, J. Limpert, and A. Tünnermann, “45 W 980 nm single transverse mode Yb-doped fiber laser,” in 3rd EPS-QEOD Europhoton Conference, paper TUoC.2 (Paris, 2008).
10. Y. Wang, C.Q. Xu, and H. Po, “Analysis of Raman and thermal effects in kilowatt fiber lasers,” Opt. Commun. 242, 487–502 (2004). [CrossRef]