1. Introduction

Yb-doped fiber amplifiers have attracted extensive attention attributed to their compactness, excellent beam quality, good heat dissipation and high conversion efficiency [1–3]. They have become the dominant force in the laser industry and are widely applied in industrial processing, military, medical treatment and other fields [4–7]. As high-power Yb-doped fiber amplifiers have wider and further applications, expectations are high for them to improve in brightness, near diffraction-limit beam quality and long-term output stability [8–10]. In the early years, the development of Yb-doped fiber amplifiers was mainly hindered by the nonlinear effects (NLEs) [3], and large mode area (LMA) fibers were considered to be the most direct and effective solution. However, as LMA fibers essentially support a great many modes, it is found that if the modes of such fibers are not effectively controlled, it is highly probable that mode instability (MI) will occur before NLEs [11]. MI was observed as the sudden and threshold-like deterioration of output beam quality with increasing average power [11–13], which seriously affected the short-term output stability of fiber amplifiers [14]. The discovery of MI has sparked a widespread interest in conducting theoretical and experimental studies [15–33]. Jauregui et al. [20] and Ward et al. [13] believed that MI was generated by the power coupling between the fundamental mode (FM) and the high order modes (HOMs), following the longitudinal periodic refractive index modulation caused by the thermal effect in the gain fiber. It is known that the formation of the phase-shift is the key to the power transfer between modes [21]. At present, there is no definite reason for its formation. While the moving grating [22] and the non-adiabatic waveguide change [23] are generally two different assumptions that gain acceptance. Various approaches have been proposed to improve MI threshold [24], such as weakening thermal refractive index grating [25], enhancing gain saturation effect [26–29], adopting more single mode LMA fibers through various mode selection means [30–32] or designing the fiber structure [18,33], and so on. Photodarkening (PD) [34] refers to the phenomenon that the active fiber losses increase and the output power decreases as the fiber lasers and amplifiers run longer - a phenomenon which if not effectively inhibited, will reduce the long-term working stability and shorten the working life of fiber laser systems [14]. Laurila et al. [35] have observed for the first time that the MI threshold reduced after multiple repeat of experiments and recovered after bleaching with blue irradiation. Later, in 2015, Otto et al. [36] pointed out that PD was the second heat source in active fibers in addition to quantum defect, and indicated in simulation results that with a negligible PD-induced power loss, the thermal load of active fiber increased significantly while the MI threshold decreased. It can thus be seen that NLEs, MI, and PD are closely related. How to simultaneously suppress NLEs and PD, improve MI threshold, and realize the further power scaling and long-term operation stability of Yb-doped fiber lasers is an urgent problem to be solved. One of the promising solutions is to use effectively single-mode LMA fibers with strong PD resistance. Yb/Ce codoped gain-tailored fibers can be regarded as the representative of such fibers.

Gain tailoring [37], also called gain filtering [38,39], confined doping [40,41] or preferential gain [4,17], works by tailoring the diameter of gain dopant across the fiber core, making FM dominant in transverse mode competition, while suppressing HOM, to enable modal discrimination in LMA fibers. Simulations [38] showed that gain tailoring could realize robust single mode in multikilowatt fiber amplifiers with 100-μm-diameter cores regardless of input beam quality. As early as MI was first observed, researchers have predicted that gain-tailored fibers, prized for their preferential gain effect and the reduced transversal inversion gradient, are one possible way to improve the threshold [11]. Soon after, this prediction has been verified in theoretical simulations [15,16], which confirmed that gain tailoring can effectively reduce the coupling constants of the LP₀₁ and LP₁₁ modes, thus suppressing MI.

Although a wealth of theoretical studies have shown that people are bullish on the use of gain-tailored fibers to achieve single-mode operation and increase MI threshold in Yb-doped LMA fiber amplifiers, there are relatively few experimental reports on gain-tailored fibers [18,39–42]. This is due to the difficulty in the fabrication of gain-tailored fibers whose refractive index profile (RIP) is close to a step-index without any obvious center dip or peak. Therefrom, different composition sections of the gain-tailored fiber core should realize an ideal index match. Restricted by the difficulties, only a few reports have been reported on the fabrication of Yb-doped gain-tailored fibers. Ye et al. [40] have reported the fabrication of a gain-tailored LMA fiber via a direct nanoparticle deposition (DND) process, which is capable to tailor the dopant profile and the RIP independently and automatically. However, DND process is not as popular as the modified chemical deposition (MCVD) process due to strict equipment and environmental requirements. Recently, Liao et al. [41] used the MCVD process to fabricate a gain-tailored fiber in which the passive section (the section of the outer core without gain ions) was achieved by vapor phase deposition. It is reported that as early as in 2016, Fujikura Ltd. realized 2 kW effectively single-mode output from a fiber oscillator with high SRS suppression by utilizing their homemade Yb-doped gain-tailored fiber [4]. Recently, the company has boosted the near diffraction-limited output to up to 5 kW with a gain-tailored fiber that has a larger core size; both SRS and MI were well suppressed [5]. And in their reports, gain-tailored fiber oscillators were successfully applied to laser processing. Nevertheless, the fabrication process of the gain-tailored fibers was not described [4,5]. As the fabrication process of Yb-doped LMA fibers by MCVD combined with SDT has been well developed and commercialized by far, it will prove very instructive and commercially viable if the approach is used to produce gain-tailored fibers.

On the other hand, while the development of fiber amplifiers has seen a significant increase in output power from a few watts to several kilowatts [2,3], experiments comparing gain-tailored fibers and conventional fibers are still restricted to a few watts output in the oscillator or the ASE setup [39–41], and thus offer limited insights. Recently, a 4 kW tandem-pump fiber amplifier employing gain-tailored fiber was reported [42], in which, the beam quality evolution of gain-tailored fiber amplifier and conventional fiber amplifier was compared. Still, there is a lack of more comprehensive and direct comparison between the two types of fibers used in kW-level output amplifiers.

In this paper, the challenges of fabricating gain-tailored Yb-doped fibers through MCVD and SDT are discussed. A gain-tailored Yb/Ce codoped aluminosilicate fiber whose RIP is close to a step-index without any center dip or peak and a conventional Yb-doped fiber with similar parameters are fabricated by MCVD combined with SDT. To prove the effect of gain filtering, a series of direct and comprehensive measurements contrasting the two fibers are performed in a kW-level MOPA setup when the total pump absorption is kept the same. It is showed that the efficiency of the gain-tailored fiber amplifier is 86.75%, comparable with that of the conventional fiber amplifier. The output beam quality of the gain-tailored fiber amplifier is ~1.43 at 1.2 kW, and the MI threshold is 1.25 kW, 1.7 times higher than that of the conventional fiber amplifier. Furthermore, to characterize the PD effect, the gain-tailored fiber amplifier is stabilized at 1.1 kW for 15 hours, after which the MI threshold remains nearly unchanged. These results demonstrate the promising prospect of applying Yb/Ce codoped gain-tailored fibers fabricated by MCVD combined with SDT to ensure further power scaling and long-term operation stability of Yb-doped fiber lasers and amplifiers.

2. Fiber fabrication and characterization

The fabrication of active fiber preforms by MCVD combined with SDT can be divided into the following steps [43]: deposition and pre-sintering of soot layers, solution soaking, dehydration, vitrification and collapsing into preforms. After that, the fabricated preform is milled into an octagonal shape, and then drawn and coated with polymer coatings on a drawing tower to form a double-clad fiber.

The difficulty in fabricating the gain-tailored fibers is that the passive section without Yb³⁺-doping needs to be index-matched with the active section doped with Yb³⁺, so as to obtain a refractive index profile close to a step-index [40]. When reflected in the fabrication process, the challenges are various. We consider the critical steps lie in the temperature regulation inside the tube and the solution doping itself. The deposition temperature and pre-sintering temperature of the soot layers will affect pore size and uniformity, which will affect their adsorption capacity to the doping ions in the subsequent solution soaking process, and finally the RIP. A F300 fused silica tube with diameter of 24/28 mm was used as the substrate, we successively prepared the outer soot layer and the inner soot layer as the passive layer and the active layer, respectively. Parameters such as pressure, gas flow, gas velocity and burner moving speed were explored through a lot of trials and were kept the same during the deposition of the two soot layers. Considering the slight increase in the wall thickness of the substrate tube after the deposition of the outer soot layer, the deposition temperature of the inner soot layer is preferable to be adjusted. In our fabrication process, the deposition temperature of the outer layer was controlled between 1550 and 1650 °C, and the pre-sintering temperature was within the range of 1400-1600 °C. The inner soot deposition temperature and the pre-sintering temperature were only about 10 °C higher than the corresponding steps of the outer soot layer.

In terms of the solution doping process, the concentration of the doping components in SiO₂ matrix directly affects the refractive index. Usually, the doping components of the active section of the gain-tailored fiber core are determined first. In order to obtain sufficient pump absorption while prevent excessive Yb³⁺ concentration from causing Yb³⁺ clustering [44], the doping concentration of Yb₂O₃ was determined to be 0.15 mol%. At the same time, Al₂O₃ was doped into the SiO₂ matrix. Al³⁺ ions are well known for improving the solubility of rare earth ions and eliminating crystallization [44,45]. Further, because Al³⁺ ions have low volatility, the formation of a center dip in the RIP can be prevented. It is reported that Al³⁺ doping also has a certain inhibitory effect on PD [10]. In order to inhibit the PD effect more effectively, Ce³⁺ ions were introduced into the SiO₂ matrix [46]. It is found that Ce codoping can greatly improve the PD resistivity without obvious adverse influence on the laser performance [47]. The above doping components will increase the numerical aperture (NA) of the fiber for their positive refractive index contributions. To keep the core NA at a low level, SF₆ was introduced into the F300 tube during the fabrication of soot layers, and the reaction between SF₆ and SiO₂ could generate a negative refractive index material SiF₄ [48]. According to the reference [49], the refractive index of the doped fiber can be determined by the refractive index of SiO_{2 ${\text{n}}_{{\text{SiO}}_{\text{2}}}$} and the molar concentration $\text{f(mol\%)}$ of each doping components in the SiO₂ matrix and the corresponding generated refractive index increment ${\text{I}}_{\text{n}}$, which can be expressed as $\text{n}={\text{n}}_{{\text{SiO}}_{\text{2}}}+\sum \text{f(mol\%)}\cdot {\text{I}}_{\text{n}}$. The refractive index increments per mol% of Yb₂O₃, Al₂O₃, Ce₂O₃, and SiF₄ are $+67\cdot {10}^{-4}$, $+23\cdot {10}^{-4}$, $+67\cdot {10}^{-4}$ and $-50\cdot {10}^{-4}$, respectively [49,50]. Hence, the refractive index increment of the active section of the gain-tailored fiber core can be expressed by Eq. (1). In order to achieve refractive index matching, it is feasible to dope any passive material with positive index contribution in the passive core section, such as GeO₂, P₂O₅, Al₂O₃, Ce₂O₃, etc. to make the refractive index increment in the passive section equal or close to that in the active section. In [41], the passive section of the gain-tailored fiber was fabricated by MCVD combined with vapor deposition of GeO₂. It can be seen from the core cross section that there is a clear boundary between the active section and the passive section. In [40], the gain-tailored fiber with a 41 μm diameter and NA 0.073 core was fabricated by DND technology. The composition of the core doping components was not described in this report, but we also found a clear interface between the active section and the passive section. However, it is likely that the interface formed between different compositional sections will induce strong light-scattering and cause an undesirable increase in fiber background loss [51]. In order to prevent from forming the clear interface as much as possible, according to our current understanding, we replaced GeO₂ (absent in the active section) with Al₂O₃ (abundant in the active section) as the main doping component of the passive section for the first time. Therefore, in our design, the refractive index increments of the active and passive sections of the gain-tailored fiber core can be expressed as:

 

Fig. 1 Designed refractive index profile of a gain-tailored Yb/Ce codoped aluminosilicate fiber.

Download Full Size | PDF

Several gain-tailored fiber preforms were fabricated with the above considerations by MCVD in conjunction with SDT. One of the gain-tailored fiber preforms was milled into octogen shape and subsequently drawn and double coated with the low-index polymer into a double-clad gain-tailored fiber with a dimension of 33 μm core and 395 μm inner cladding. The inner cladding NA is 0.46. The RIP of the fiber measured by SHR-1602 fiber analyzer is demonstrated in Fig. 2(a) by the red line. A flat profile for core without any middle dip and an ideal index match between two sections are achieved. The NA was calculated to be ~0.06, corresponding to a refractive index difference of 0.00124. Figure 2(b) presents the cross section of the octagonal-shaped inner cladding. Inset of Fig. 2(b) shows the microscope image of the core region. Compared with the obvious interfaces in the core regions in [40,41], the boundary between the active section and the passive section is nearly absent in the demonstrated fiber.

 

Fig. 2 (a) Refractive index profile of the gain-tailored fiber (red line) and the conventional fiber (black line); (b) the microscope image of the cross section of the gain-tailored fiber and the core region (inset).

Download Full Size | PDF

The concentration distribution of Yb³⁺, Al³⁺, Ce³⁺, and F^- across the core region of the gain-tailored fiber was characterized by electron probe microanalysis (EPMA) and shown in Fig. 3(a). Yb³⁺ and Ce³⁺ ions are uniformly distributed in the active section and almost not exist in the passive section. The doping diameter of Yb³⁺ ion is 23 μm, accounting for about 70% of the entire core diameter. The distribution of Al³⁺ ions is basically consistent with the design. F^- ions are uniformly distributed throughout the core. The elemental distribution together with the perfect RIP indicates that the tube temperature was properly regulated in the whole fabrication process and the solution doping steps were reasonably designed.

 

Fig. 3 Elemental distribution of the (a) gain-tailored fiber and the (b) conventional fiber.

Download Full Size | PDF

In order to characterize the effect of gain filtering, we fabricated a conventional fiber [30] with very close parameters by means of MCVD with SDT for comparison. The RIP of the conventional fiber which is similar to that of the gain-tailored fiber is shown by the black line in Fig. 2(a). The distribution of each dopant in the fiber core is shown in Fig. 3(b). Unlike the distribution of Yb³⁺ in gain-tailored fiber, the Yb³⁺ ions in the conventional fiber are uniformly distributed across the core. The pump absorption of the conventional fiber and the gain-tailored fiber were measured to be 1.37 dB/m and 0.77 dB/m at 976 nm, respectively. The parameters of the two fibers are listed in Tab. 1.

Table 1. Fiber parameters

View Table

3. Experimental setup

An all-fiber master oscillator power amplifier (MOPA) was built up as shown in Fig. 4, which has been described in our previous publication [30]. The MO consisting of a linear-cavity oscillator which is composed of a pair of gratings and an active fiber is pumped by six fiber-pigtailed 915-nm laser diodes (LDs) from DILAS. The pump laser is coupled into the active fiber through a 7 × 1 pump combiner (PC) whose center arm is cut with an oblique angle in case of reflections. The high reflectivity fiber Bragg grating (HR) and the output coupler fiber Bragg grating (OC) provide a refractivity of 99.9% and 11.2% at a wavelength of ~1080 nm, respectively, and their 3 dB bandwidth are ~2 nm and ~1 nm, respectively. The active fiber is a 45 m home-made 20/400 ytterbium-doped fiber with a core and cladding NA of 0.065 and 0.46, respectively. The active fiber is coiled in a circle spiral shape on a water-cooled plate with a minimum bending diameter of 11 cm. A cladding light stripper (CLS) is applied to remove excess pump laser and cladding light, which is beneficial for subsequent amplification. Up to 110 W seed laser with a center wavelength of 1080 nm can be provided from this MO. The M² factor at the maximum output is measured to be ~1.4. The seed light and the 976-nm pump laser are launched into the power amplifier (PA) through a pump and signal combiner (P/S C). The fabricated Yb-doped fibers under test serve as the amplifier fiber. Another CLS is applied to filter out the cladding laser and the remaining pump light in this stage. Besides, an endcap (EC) is helping to prevent harmful end-face reflections.

 

Fig. 4 Scheme of the experimental setup. LD, laser diode; PC, pump combiner; HR, high reflectivity fiber Bragg grating; OC, output coupler fiber Bragg grating; CLS, cladding light stripper; MFA, mode field adaptor; P/S C, pump and signal combiner; PD, photodetector; EC, endcap; CM, collimating mirror; BS, beam splitter; PM, power meter; DM, dichroic mirror; BD, beam dump; PBS, polarization beam splitter; NDF, neutral density filter.

Download Full Size | PDF

To directly characterize the laser performance of the gain-tailored Yb-doped fiber, and compare it with the conventional Yb-doped fiber. 27 m gain-tailored fiber and 15 m conventional fiber are selected to ensure the same total pump absorption in the MOPA setup. For each measurement, they are coiled in the spiral groove of a water-cooled plate as shown in Fig. 4. The racetrack spiral shaped groove consists of two straight tracks 15 cm in length and two semicircles with a 2-mm increment of bend radius. For both fibers, the seed laser injects into the coil with a bending diameter of 11 cm. The bending diameter of the gain-tailored fiber ranges from 11 cm to 23.5 cm, and that of the conventional fiber ranges from 11 cm to 17.5 cm, as shown in Fig. 5.

 

Fig. 5 Schematic of gain fiber coils in the main amplifier.

Download Full Size | PDF

Measurements of the laser performance included power, beam quality, and MI threshold. The output laser is collimated by the collimating mirror (CM) and split into two paths by a water-cooled high-reflectivity beam splitter (BS). More than 99.95% of the light is used for power measurement, and a small amount of light is used for beam analysis. A dichroic mirror (DM) is applied to further eliminate the residual pump. The polarization beam splitter (PBS) and a set of neutral density filter (NDF) serve to attenuate the beam power. Then the beam quality evolution of fibers is recorded by operating the beam propagation analyzer (M²-200S manufactured by Spiricon) under different output power levels. To track the MI process, an InGaAs photodetector (150 MHz, 700 nm- 1800 nm) mounted with a small aperture diaphragm on its receiving head is employed to collecting the light leakage from the CLS.

4. Results and discussion

Contrast of output power characteristics: The output power versus pump power of the fiber amplifiers were recorded and the results are shown in Fig. 6(a). With a comparable laser efficiency as the conventional fiber amplifier, the slope efficiency of the gain-tailored fiber amplifier is as high as 86.75%. The high laser efficiencies also indicate that each fabricated fiber possesses a low background loss. However, it is found that the conventional fiber amplifier suffers a power roll-over after the output power exceeds ~750 W. In contrast, the output power of the gain-tailored fiber presents linear growth without obvious power roll-over. This can be attributed to the difference of HOM bending loss between the two fibers. With the same total pump absorption, the gain-tailored fiber is nearly twice the length of the conventional fiber. Due to the fact that they are coiled in a spiral shape, the longer fiber length of the gain-tailored fiber leads to the increased number of coils. With the increased diameter of the outer coils, the HOM loss decreases. The output spectra of the gain-tailored fiber amplifier are shown in Fig. 6(c). No SRS component was observed at maximum output power of 1.39 kW.

 

Fig. 6 (a) Output power and (b) M² factor (averaged of the x and y direction) as a function of pump power for the conventional fiber and the gain-tailored fiber, (c) seed laser spectrum and output laser spectrum at output power of 1.39 kW.

Download Full Size | PDF

Contrast of the evolution of output beam quality: The M² factors of the two fibers versus pump power are illustrated in Fig. 6(b). Each M² data were averaged in the x and y directions. Different from the reported case [39,42] where the beam quality of the conventional fiber oscillators seriously degraded with the increasing pump power, as seen, for this conventional fiber amplifier, the output beam quality only degrades slightly from M²~1.53 to M²~1.57 as the pump power gradually increases to 750 W, but then it suddenly degrades obviously. As reported in [52], bend loss of each guided mode greatly affects the transverse mode competition in fiber amplifiers. Due to the small bending diameter, the high HOM losses result in less gain obtained by HOMs and insufficient amplification. Thus, the beam quality keeps relatively good at first. In terms of the gain-tailored fiber amplifier, as the pump power increases, the beam quality steadily improved to M²~1.43 at an output power of 1.2 kW. Even though the maximum improvement of M² factor is less than 0.1. It can be explained by the role of preferential gain effect that the relative HOM content is continuously reduced, as the case in [17]. Nevertheless, the beam quality also suffers an obvious deterioration as the pump power further increases.

Contrast of the MI thresholds: The sudden deteriorations in the beam quality of both fiber amplifiers are caused by MI. As shown in Figs. 7(a) and (c), the photodetector signal stays stable below the MI threshold for power of 703 W and 1.22 kW was obtained in the conventional fiber amplifier and the gain-tailored fiber amplifier, respectively. Above the threshold of 717 W and 1.25 kW, respectively, the time traces show periodical fluctuations with the fluctuation frequencies in the range of 0-3 kHz appearing in its Fourier spectrum, as shown in Figs. 7(b) and (d). The onsets of MI are reflected in sharp increases of the standard deviation (STD) as shown in Fig. 7(e). We believe that three factors together lead to a 1.74 times higher MI threshold of gain-tailored fiber amplifier than that of conventional fiber amplifier. Firstly, the spatial overlap of HOM with the dopant profile is reduced, meanwhile the preferential gain principle makes the gain-tailored fiber closer to single-mode operation. Secondly, a better distribution of the heat load is achieved by the longer absorption length, and the gain saturation in the active region is strong due to the decrease of ytterbium doped area [26]. Finally, the HOM bending loss of the two fibers is different. There is no doubt that the bending loss of HOM affects the MI threshold [31,32]. In the gain-tailored fiber amplifier, the higher bending loss of HOM combined with gain-filtering effect makes HOM more suppressed. Overall, the employment of gain-tailored fiber has greatly increased the MI threshold of fiber amplifiers.

 

Fig. 7 (a) Photodiode intensity traces and (b) frequency characteristics of the conventional fiber amplifier and (c) photodiode intensity traces and (d) frequency characteristics of the gain-tailored fiber amplifier, respectively, (c)standard deviation with respect to the output power in different fiber amplifiers. Additionally, the MI thresholds are marked in green.

Download Full Size | PDF

Characterization of long-term operation stability: The long-term operation stability of a laser system can be evaluated by monitoring the temporal evolution of the output power at kW-level for several hours. The gain-tailored fiber amplifier was operated at 1.1 kW and kept running for over 15 hours. The data were collected every 4 minutes. Figure 8(a) depicts that the output power presents a power fluctuation within the range from 1111 W to 1083 W. Actually, the power fluctuation was mainly affected by the cycling of the water chiller on the power meter [53], and the output power could reach almost the same maximum value in each cooling cycle of the water chiller. In addition, after the 15-hour continuous operation, we tested the MI threshold of the gain-tailored fiber amplifier again. As shown in Fig. 8(b), it exhibits nearly the same MI thresholds. Nevertheless, as depicts in [29,35], PD-induced loss will lead to an obvious reduction in MI threshold, as it is a significant heat source [36], so the almost equal MI thresholds after the long-term high-power operation indirectly indicate that the gain-tailored fiber amplifier is PD-free. It is proved that the fabricated gain-tailored fiber possesses strong resistance to PD.

 

Fig. 8 (a) Output stability evaluation at 1.1 kW, (b) standard deviation with respect to the output power before and after the 15-hour operation.

Download Full Size | PDF

It has been shown that adopting the gain-tailored Yb/Ce codoped aluminosilicate fiber can significantly improve both the short-term and long-term laser stability without sacrifice of laser efficiency. However, one major drawback of using gain-tailored fiber is the reduction of pump absorption compared to the conventional fiber. To achieve efficient pump absorption, longer fiber length is required, which inevitably leads to a lowered threshold of NLEs, especially the SRS. Recently, studies have even revealed that MI is associated with the occurrence of SRS [54]. Therefore, further power scaling of the gain-tailored fiber amplifier to multi-kW level requires suppressing MI in optimizing system configuration to prevent SRS. Some methods could be employed, such as shortening the passive fiber length in the main amplifier [54] and adjusting the seed power [55].

5. Conclusion

In conclusion, we report on the fabrication of gain-tailored Ge-free Yb/Ce codoped Aluminosilicate fiber by MCVD combined with solution doping technique. The laser performance of the in-house gain-tailored fiber was proved in contrast with conventional fiber in a kW-level MOPA setup. Compared with the conventional fiber amplifier with the same total pump absorption, the gain-tailored fiber amplifier offers better beam quality. The MI threshold of the gain-tailored fiber amplifier is 1.25 kW, 1.74 times as much as that of the conventional fiber amplifier. The slope efficiency of the gain-tailored fiber amplifier is 86.75%. After a 15-hour continuous operation at an output power of 1.1 kW, the MI threshold does not decrease. The results indicate that gain-tailored fiber fabricated by MCVD combined with solution doping method has a great prospect in power scaling of fiber lasers and amplifiers with noticeable long-term output stability.