Thermal effects and output power characteristics of kilowatt all-fiber master-oscillator power amplifier (MOPA) are investigated. Proper designs for cooling apparatus are proposed and demonstrated experimentally, for the purpose of minimizing splice heating which is critical for the reliability of high power operation. By using these optimized methods, a thermal damage-free, highly efficient ytterbium-doped double-clad fiber MOPA operating at 1080 nm with 1.17 kW output was obtained. The maximum surface temperature at the pump light launching end splice of the booster amplifier was 345 K, and the temperature rise for this key splice was 0.052 K/W.
©2011 Optical Society of America
Due to high efficiency, compactness, and outstanding beam quality, fiber lasers are now competing with their bulk solid-sate counterparts in various scientific and industrial applications [1–3], such as material processing, defense, remote sensing, free-space communication, display, etc. The rising in output powers from ytterbium-doped double-clad fiber (YDCF) sources, via the use of cladding-pumped fiber architectures in combination with high-power and high-brightness diode pump sources, has been dramatic recently and they have matured to the point where the average power reaches the kilowatt (kW) level and beyond [4–6]. Thanks to the long and thin fiber geometries, stress fracture and beam distortion, which are major problems for bulk solid-state lasers, can both be alleviated strongly in fiber lasers . But thermal management is still one of the most critical issues for scaling higher output powers from YDCF sources. Although the intrinsic large surface-to-active-volume ratio of fibers brings many direct and indirect advantages to their heat dissipations, the use of shorter fibers with higher thermal loading densities in kW fiber lasers indeed require careful thermal management [7–9]. In practice, low index polymer coatings of conventional double-clad fibers are always so sensitive to a high thermal load that it will cause thermal damage when the temperature is approaching 150~200 °C (long-term reliability may require operation below 80 °C) , so it need to be controlled not to extend the safe range though the core temperature is always below the melting point of quartz (1982 K).
Transverse and longitudinal temperature distributions in fibers have been calculated by solving thermal conduction equations [11–13]. Based on the convective heat transfer coefficient h, heat dissipations of fiber lasers or amplifiers are analyzed and discussed in most of these studies. However, it’s difficult to analyze in this way for a single point on the fiber under different conduction conditions. In this paper, thermal contact resistance is introduced, and the thermal effects of active fibers and fiber-to-fiber splices of an all-fiber master-oscillator power amplifier (MOPA) are discussed in detail. For the purpose of figuring out heat dissipation requirements for different parts of the fibers and providing effective cooling ways, the temperature distribution of the active fiber in booster amplifier stage is analyzed through a numerical modeling. Three kinds of cooling apparatus were discussed theoretically and verified experimentally. At last, a 1.17 kW output, thermal damage-free, YDCF MOPA operating at 1080 nm with a total optical-to-optical conversion efficiency of 82.4% was developed.
2. Thermal AnalysisEq. (1), subject to Eq. (2) and other boundary conditions, results in the following expressions for the temperature in Regions I, II and III
In the practical situation of a strong pumping, one can assume that either the pump or the signal is large compared to the ASE. In this limit, the time-dependent rate equations can be solved analytically , and then the pump power distributions along the fiber can be calculated. If the quantum defect heating is assumed to be the only factor to cause heating, we could then get how the heat power density varies along the fiber. As shown in Fig. 2 , we assume a single-stage MOPA system with 1.22 kW of pump power and 1.8 dB/m of pump absorption at 975nm in YDCF, which is equivalent to a heat load of 40 W/m . By taking k 1=k 2= 1.38 W/mK , k 3= 0.2 W/mK, b= 400 μm, c= 500 μm, and h= 100 W/m2K which is appropriate to a fan-blowing air cooling condition, the temperature distribution in the fiber center along a 9 m-long YDCF is achieved and shown in Fig. 3 . It varies along the fiber, decaying approximately exponentially. We can see directly that the position with the highest temperature is at z=0 m, which is just the fusion splice point generating the maximum heat of the MOPA, and the high-temperature region mainly focuses on the first two meters.
Without considering fusion splice losses, the calculated temperature distribution as a function of r for three different h values at z=0 m is shown in Fig. 4 . The temperature depends strongly on the convective heat transfer coefficient, and for a certain value of r, a higher h value leads to a lower T value. Then, for a certain value of h, the total temperature difference in the fiber radial coordinate is about 25 K. When the heat-sinking temperature is taken as 290 K with h=100 W/m2K, the corresponding coating surface temperature is nearly 600 K which is much higher than the coatings’ long-term reliability temperature of 353 K (80 °C). Figure 5 shows the temperatures of fiber center and coating surface as a function of h at z=0 m. The fiber center temperature should be below 378 K to ensure the coatings’ safety (below 353 K), and this is corresponding to h ≈450 W/m2K. If the heating caused by fusion splice losses is considered, much higher h value is needed. That is to say, some more effective cooling methods at this splice point of fiber other than merely air cooling must be taken.
As free convection is an order of magnitude less efficient than conduction for dissipating heat from an optical fiber , we introduce a copper heat sink with V-grooves to cool the fiber. The thermal contact resistance per unit surface (m2K/W) between the fiber and the heat sink can be expressed as Equation (4) suggests the concept for the treatment of the heat flow through a fiber layer in analogy to the diffusion of electrical charge , where the temperature difference is analog to the electrical voltage which drives the heat flow through a thermal resistance. For an active fiber, the heat generation q 0 (W/m3), the heat load (W/m) and the pump absorption α (dB/m) are related to each other
Take one point on the fiber for example, Fig. 6 shows the two contact ways between the fiber and the heat sink. Due to the limited machining precision, the grooves cannot perfectly match the fibers within a period of length, and so the interstitial air between them may have a significant thickness in some place as shown in Fig. 6(a). In order to improve the contact closeness, the fiber could be pressured and fixed properly by thin copper tape, as shown in Fig. 6(b). Since copper has good ductility and conductivity, the copper tape could contact well with the fiber and has almost the same good thermal conductivity with the copper heat sink and thus the same temperature.
In order to obtain an approximation of as shown in Fig. 6(a) with Eq. (4), some assumptions are to be considered according to Ref . First, heat transfer from fiber to air is small enough to consider air as a thermal insulator. Second, the measured surface temperature exposed to air is close to the actual temperature of the fiber which is uniform in the radial direction. Third, is invariant with temperature. Finally, the temperature of the heat sink T∞ is considered uniform and, without any heat load, T∞ should be equal to T s. With all these assumptions, the heat load can be obtained from Eq. (5) and then divided by the perimeter, which is the junction between the heat sink and the fiber, to get the heat flux . For convenience, the perimeter is calculated from the groove geometry.
Figure 6(b) could be looked as being connected in parallel by the copper heat sink and the copper tape, see Fig. 7(b) and 7(c), and the thermal resistance of the geometry of Fig. 6(b) is found by calculating the thermal resistances of the equivalent half-space geometries separately and combining these. Then, the results of the experiments and numerical calculations for calculating are shown in Table 1 .It can be seen clearly that thermal contact resistance for Fig. 6(b) is much smaller than that of Fig. 6(a), which means that better cooling can be achieved by Fig. 6(b). Therefore, the thermal contact resistance varies greatly depending on the designs. The closer contact between the heat sink and the fiber, the smaller value of . However, there are many factors limiting the contact closeness in actual operations. For example, the pressure put on the fiber, the surface roughness of the heat sink, the uniformity of fibers radial thickness, the matching degree between the recoated spliced fiber and the grooves, etc. Generally speaking, the more perfect matching between the grooves and the fiber, the better.
We had designed an experiment to test the cooling effects of different methods for the same splice. By injecting pump light into a passive fiber (20/400 μm, circular-shaped inner cladding) which was spliced to a 8 dB/m absorption coefficient active fiber (20/400 μm, octagonal-shaped inner cladding), the surface temperature rises of the recoated splice by using three kinds of cooling methods were shown in Fig. 8 , where T1, T2, and T3 are corresponding to the cooling conditions of fan-blowing, Fig. 6(a), and Fig. 6(b), respectively. Surface temperature rises of them are 0.5, 0.6 and 0.06 K/W, respectively. It means that the cooling apparatus shown in Fig. 6(b) is the best. All the experimental measurements were taken by a high resolution thermo tracer (NEC Avio Infrared Technologies Co., Ltd.).
An improved method of cooling is to fill with some thermal interface material (TIM) with high thermal conductivity between the fiber and the heat sink when machining surface roughness and flatness can be improved. An actual TIM will not be able to completely fill the air cavities, and small air pockets may exist on both sides of the TIM, as shown in Fig. 9 . Then, the total thermal contact resistance is composed of three resistances in series and can be expressed as follows18]. R cond-TIM is proportional to L, and is inversely proportional to the product of k TIM and A. Therefore, for a certain setup, a lower L and higher k TIM and A will lead to a lower R cond-TIM, and hence a lower . TIM is often used between the central processor unit (CPU) and the heat sink, such as Arctic Silver thermal grease, Thermax HF-60110-BT phase change material, etc. However, the use of such materials to achieve fiber cooling is rarely reported. Currently, proper TIM with high thermal conductivity which is suitable for fiber isn’t available in our lab. By using TIM to realize better heat dissipation is one of our key research directions at higher power level fiber laser systems in the future.
In order to test the cooling capability of our novel thermal management shown in Fig. 6(b), a kW-level all-fiber MOPA configuration was built. The experimental setup was depicted in Fig. 10 . A laser oscillator was followed by a booster amplifier. All the components were connected by fusion splice, which made the system more compact and reliable. In order to be protected, all the fusion splices were recoated. The laser oscillator consisted of a pair of FBGs, YDCF, and a (6+1)×1 combiner (combiner1). A section of 50 m-long YDCF manufactured by Nufern Inc. was employed as the gain medium, whose absorption coefficient is 0.4 dB/m at 915 nm. The diameters of the fiber core and inner cladding are 20 μm and 400 μm, respectively. The laser cavity was composed of a pair of FBGs, where one was used as the high reflector (R>99%) and the other one was used as the output coupler (R≈10%). The signal delivery fiber of combiner1 was used as the monitoring port. Six 70 W at 915 nm fiber-pigtailed laser diodes were used as pump sources.
The power amplifier stage comprised a 9 m-long large-mode-area (LMA) YDCF and a (6+1)×1 combiner (combiner2) with the signal input fiber spliced to the delivery fiber of the laser oscillator. The gain fiber manufactured by Nufern Inc. had a 25 μm-diameter core and a 400 μm-diameter octagonal-shaped inner cladding. The absorption coefficient was 1.8 dB/m at 975 nm. This power amplifier was pumped by six laser diodes of 975 nm. A 400 µm core-less endcap with a length of 1.5 mm was spliced to the output end of the amplifier, and it was cleaved at an angle of 6° to minimize back-reflection into the amplifier and avoid the surface damage.
4. Results and Discussion
According to the analysis of section 2, temperature will reach the highest at the pump light launching end splice joint for a running all-fiber MOPA system, see Fig. 3. This is because the gain fiber has the strongest pump absorption here, and what’s more, fiber fusion splice will inevitably induce non-guided pump or signal power loss, and hence cause dramatic temperature rise. It is very important to realize highly efficient thermal management for this splice. Any improper splicing parameters or cladding geometry mismatch will increase the splice loss which has a significant impact on the performance of fiber laser. Therefore, many splice parameters need to be optimized, including the fiber cleaving angle, tension, the discharge time and intensity, etc. When splice is recoated, heating is mainly due to the absorption in the coating material of a fraction of the pump power scattered by distortions of the waveguide. Micro-cavities, which might be generated by any surface or interface imperfection, will significantly increase the photon path length and will tremendously increase the probability of absorption in a relatively small volume . Since surface quality will influence thermal contact resistance and light absorption in the coating, better fiber coating stripping and recoating methodologies must be applied. In our experiment, dichloromethane organic solvents were chosen to strip the coatings, and Vytran LDC-200 automate fiber cleaver, GPX-3000 large diameter fiber splicer, and Vytran PTR 200 manual recoater were chosen to cleave, splice and recoat fibers. In order to get a low signal insertion loss, fiber cores should be aligned well prior to cladding when splicing. A micrograph of a good fusion splice completed by Vytran splicer between 20/400 μm circle inner cladding passive fiber and 25/400 μm octagonal inner cladding LMA YDCF was shown in Fig. 11 .
The laser oscillator achieved an output power of 160 W under a pump power of 340 W, which is equivalent to a 4.5 W/m of heat load. This heat could be dissipated by simply wrapping the fibers around a metal disc. When it comes to the main amplifier, the maximum heat load on the active fiber has reached 40 W/m, so the fiber has to be wrapped in a metallic disc which is engraved with V-grooves and water-cooled to be a constant temperature of 20 °C. Since there would be a serious temperature rise at the pump light launching end splice joint, this key splice was specially held in temperature-controlled straight metallic V-grooves as depicted in Fig. 6(b) to achieve a lower thermal contact resistance and prevent possible thermal damage to the fiber coating.
The characteristics of the MOPA output power and splice surface temperature on the booster amplifier stage were shown in Fig. 12 . The maximum laser power of this system was up to 1.17 kW at 1080 nm with the total coupled pump power of 1.22 kW and signal power of 160 W. The slope efficiency was 85.3% and the maximum optical-to-optical conversion efficiency was around 82.4%. The splice surface temperature increased with the growth of the pump power. Thanks to the good cooling design, the maximum temperature was only 345 K (72 °C) which was below the long-term reliability requirements, and it is equivalent to a surface temperature rise of 0.052 K/W. It is noteworthy that the laser output showed no evidence of roll-over even at the highest output power, which means that the system was thermal damage-free running.
The stability of output powers and the corresponding fiber surface temperatures at the splice joint marked out with dashed box in Fig. 10 within 10 minutes running time were shown in Fig. 13 . At a pump power of 1.1 kW, the laser power was around 1.04 kW and the corresponding temperature was around 338.5 K (65.5 °C). The total laser power fluctuation was about 1.2% corresponding to a temperature fluctuation of 0.4%.
With the MOPA system being in running, there were no fiber or coating damages due to the thermal load by the absorbed pump, and no failure of splices or components was observed for any of power level. The highest output power was limited only by available pump power.
The laser spectrum of the system was shown in Fig. 14 . The emission was centered at 1080 nm. The full-width at half-maximum (FWHM) of the laser spectrum was about 1.5 nm. We did not observe any nonlinear signatures. The diameter of the metallic disc used to coil fiber is 25 cm. The M2 factor was measured as M2 X=1.81, M2 Y=1.89 without other mode-selecting technologies.
The heat dissipation of the active fiber and fiber-to-fiber splices in an YDCF MOPA was analyzed through a numerical model. Based on thermal contact resistances, thermal properties of different cooling configurations were discussed and verified experimentally. Proper cooling apparatus was then applied to the active fiber and splices of a kW-class all-fiber MOPA system. An output power of 1.17 kW was achieved under a coupled pump power of 1.22 kW, corresponding to an optical-to-optical conversion efficiency of around 82.4% and a slope efficiency of 85.3%. When the pump power was 1.22 kW, the surface temperature for the pump light launching end splice was about 345 K, and the surface temperature rise was only 0.052 K/W. The MOPA system worked stably at all power levels, and there were no nonlinear effects or thermal damages observed.
This work was supported by a grant from the National Science and Technology Major Project (No. 2010ZX04013), the National High Technology Research and Development Programs of China (863 Program) (No. 2008AA03Z405 and No. 2011AA030201), the Shanghai Rising-Star Program (No. 09QB1401700), and the Natural Science Foundation (No. 60908011).
References and links
1. S. Yin, P. Yan, and M. Gong, “End-pumped 300 W continuous-wave ytterbium-doped all-fiber laser with master oscillator multi-stage power amplifiers configuration,” Opt. Express 16(22), 17864–17869 (2008), http://www.opticsinfobase.org/abstract.cfm?URI=oe-16-22-17864. [CrossRef] [PubMed]
2. D. J. Richardson, J. Nilsson, and W. A. Clarkson, “High power fiber lasers: current status and future perspectives [Invited],” J. Opt. Soc. Am. B 27(11), B63–B92 (2010), http://www.opticsinfobase.org/abstract.cfm?URI=josab-27-11-B63. [CrossRef]
3. Y. Jeong, J. Sahu, D. Payne, and J. Nilsson, “Ytterbium-doped large-core fiber laser with 1.36 kW continuous-wave output power,” Opt. Express 12(25), 6088–6092 (2004), http://www.opticsinfobase.org/abstract.cfm?URI=oe-12-25-6088. [CrossRef] [PubMed]
4. E. Stiles, “New developments in IPG fiber laser technology,” in Proceedings of the 5th International Workshop on Fiber Lasers (2009).
5. B. He, J. Zhou, Q. Lou, Y. Xue, Z. Li, W. Wang, J. Dong, Y. Wei, and W. Chen, “1.75 killowatt continuous-wave output fiber laser using homemade ytterbium-doped large-core fiber,” Microw. Opt. Technol. Lett. 52(7), 1668–1671 (2010). [CrossRef]
6. C. Wirth, O. Schmidt, I. Tsybin, T. Schreiber, T. Peschel, F. Brückner, T. Clausnitzer, J. Limpert, R. Eberhardt, A. Tünnermann, M. Gowin, E. ten Have, K. Ludewigt, and M. Jung, “2 kW incoherent beam combining of four narrow-linewidth photonic crystal fiber amplifiers,” Opt. Express 17(3), 1178–1183 (2009), http://www.opticsinfobase.org/abstract.cfm?URI=oe-17-3-1178. [CrossRef] [PubMed]
7. D. C. Brown and H. J. Hoffman, “Thermal, stress, and thermal-optic effects in high average power double-clad silica fiber lasers,” IEEE J. Sel. Top. Quantum Electron. 37(2), 207–217 (2001). [CrossRef]
8. Y. Wang, C. Q. Xu, and H. Po, “Thermal effects in kilowatt fiber lasers,” IEEE Photon. Technol. Lett. 16(1), 63–65 (2004). [CrossRef]
9. N. A. Brilliant and K. Lagonik, “Thermal effects in a dual-clad ytterbium fiber laser,” Opt. Lett. 26(21), 1669–1671 (2001), http://www.opticsinfobase.org/abstract.cfm?URI=ol-26-21-1669. [CrossRef] [PubMed]
10. M.-A. Lapointe, S. Chatigny, M. Piché, M. Cain-Skaff, and J.-N. Maran, “Thermal effects in high power CW fiber lasers,” Proc. SPIE 7195, 719511 (2009).
11. Y. Wang, “Heat dissipation in Kilowatt fiber power amplifiers,” IEEE J. Quantum Electron. 40(6), 731–740 (2004). [CrossRef]
12. L. Li, H. Li, T. Qiu, V. L. Temyanko, M. M. Morrell, A. Schülzgen, A. Mafi, J. V. Moloney, and N. Peyghambarian, “3-Dimensional thermal analysis and active cooling of short-length high-power fiber lasers,” Opt. Express 13(9), 3420–3428 (2005), http://www.opticsinfobase.org/abstract.cfm?URI=oe-13-9-3420. [CrossRef] [PubMed]
13. P. Li, C. Zhu, S. Zou, H. Zhao, D. Jiang, G. Li, and M. Chen, “Theoretical and experimental investigation of thermal effects in a high power Yb3+-doped double-clad fiber laser,” Opt. Laser Technol. 40(2), 360–364 (2008). [CrossRef]
14. A. Hardy and R. Oron, “Signal amplification in strongly pumped fiber amplifiers,” IEEE J. Quantum Electron. 33(3), 307–313 (1997). [CrossRef]
15. D. E. Gray, American Institute of Physics Handbook, 3rd ed. (McGraw-Hill, 1972).
16. P. Yan, A. Xu, and M. Gong, “Numerical analysis of temperature distributions in Yb-doped double-clad fiber lasers with consideration of radiative heat transfer,” Opt. Eng. 45(12), 124201 (2006). [CrossRef]
17. B. Zintzen, T. Langer, J. Geiger, D. Hoffmann, and P. Loosen, “Heat transport in solid and air-clad fibers for high-power fiber lasers,” Opt. Express 15(25), 16787–16793 (2007), http://www.opticsinfobase.org/abstract.cfm?URI=oe-15-25-16787. [CrossRef] [PubMed]
18. J. P. Gwinn and R. L. Webb, “Performance and testing of thermal interface materials,” Microelectron. J. 34(3), 215–222 (2003). [CrossRef]