Featuring large optical gain, high efficiency, compact size, and high mechanical stability and reliability, fiber lasers have surpassed traditional solid-state lasers in popularity and have emerged as reliable sources across diverse fields, including industrial manufacturing^[1], laser marking and engraving, precise spectroscopy^[2], environmental monitoring, and pollution control^[3]. Notably, in the realm of gravitational wave detection, high-power, low-noise, and narrow-linewidth lasers constitute the most critical components of the detection system, ensuring adequate sensitivity amid significant losses incurred over long transmission distances^[4]. As the light source for the first-generation gravitational wave detector (GWD), 1 µm fiber lasers based on the Yb-doped fiber have attracted considerable attention in recent years^[5–8]. However, 1 µm lasers could not pass through silicon substrates, which exhibited lower thermal noise when operating at cryogenic temperatures. Thus, researchers have turned their focus toward high-power low-noise single-frequency lasers operating in 1.5 and 2 µm bands, aiming at improving the gravitational wave detection sensitivity^[9–12].

Well-established erbium-ytterbium co-doped fibers (EYDFs) are adopted for high-power low-noise amplifications at the C band, and the output power of single-frequency fiber lasers, which have ever-expanding applications, has continued to grow. In traditional EYDF amplifier schemes utilizing 976 nm laser diodes, Yb band amplified spontaneous emission (ASE) is detrimental, prompting the development of pumping schemes that are off-resonant in the Yb3+ absorption band. In 2021, Darwich et al. reported a 10 W all-fiber laser system with an ultralow-intensity noise of −160dBc/Hz beyond 200 kHz^[13]. They pointed out that the pump had a major impact on intensity noise at low frequencies, and they employed a 940 nm pump to suppress ASE for a low noise operation. Chen et al. achieved a 59.1 W single-frequency output with 90% efficiency at 1560 nm using 1480 nm pump^[14]. Varona et al. reported a 111 W 1.5 µm single-frequency fiber laser with a polarization extinction ratio of 13 dB^[15]. Similarly, Creeden demonstrated an average output power of 207 W, accompanied by a linewidth of 540 Hz and a slope efficiency of 50.5% in a large-mode-area (LMA) Er-Yb fiber^[16]. Although no ASE was observed on the optical spectrometer, the SBS limited the further enhancements.

Unfortunately, the polarization-maintaining (PM) gain fibers exhibit lower efficiency and nonlinear effect threshold compared to ordinary single-mode (SM) fiber^[17], resulting in a much lower output power for PM fiber amplifiers compared to the works mentioned above. Bai et al. reported an all-fiber single-frequency master oscillator power amplifier (MOPA) at 1550 nm with an output power of 56.4 W and a slope efficiency of only 37% in 2015^[18]. Booker et al. presented a 110 W 2-stage single-frequency Er-Yb fiber amplifier at 1556 nm^[19]. The relative intensity noise (RIN) was 10−3/Hz−1/2 at 1 Hz and decreased to 10−6/Hz−1/2 at 100 kHz. The performance of noise is also affected by the pump when the output power is above 50 W. In 2022, Huang et al. demonstrated an all-PM-fiber single-frequency narrow-linewidth laser operating at 1560 nm with 102 W output power and 40.6% slope efficiency^[20]. This performance is still inadequate for applications such as gravitational wave detection. To meet the escalating detection requirements, addressing the issue of SBS during power amplification is paramount, particularly in all-PM fiber systems.

To alleviate the SBS effect, previous works typically focused on optimizing the length of the gain fiber while neglecting the impact of the passive fiber. In this Letter, we present the design of a PM all-fiber C-band laser amplifier based on a MOPA configuration. Guided by mathematical simulations, we investigate the physical mechanisms of ASE and SBS. Experimentally, we increase the threshold power of SBS by shortening the length of the passive fiber pigtail following the gain fiber. Eventually, we decrease the SBS power by 10 dB and achieve a 202 W all-PM single-frequency fiber amplifier.

Regarding laser power amplification, ASE has consistently emerged as a pivotal challenge in mitigating power enhancement. Furthermore, in single-frequency laser systems, SBS is equally, if not more, crucial than ASE^[21]. In order to mitigate these nonlinear effects, we conducted a detailed analysis of each of them and proposed corresponding strategies.

It is well-known that the absorption capacity of Er ions is weak for pump light, and a significant concentration of doped particles often tends to cluster. This issue can be effectively addressed through Er-Yb co-doping. Figure 1 illustrates the energy level diagram of the co-doped fibers. The Yb ions can efficiently absorb the pump energy and transfer it to the Er ions via cross relaxation. The evolution of power can be mathematically represented by Eqs. (1)–(3), while the process of particle transfer is delineated by Eqs. (4)–(8)^[22]: ±dPp±dz=[ΓpNYb(n6σ65Yb−n5σ56Yb)−ΓsNErn1σ13Er−αp]Pp±,(1)±dPs±dz=[ΓsNEr(n2σ21Er−n1σ12Er)−αs]Ps±,(2)±dPASE,λ±dz=[ΓsNYb(n6σ65,λYb−n5σ56,λYb)−αλ]PASE,λ±,(3)∂n2∂t=−n2τ21+n3τ32+W12n1−W21n2−2CupNErn22,(4)∂n3∂t=−n3τ32+W13n1+R61NYbn6n1−R35NYbn3n5+CupNErn22,(5)∂n6∂t=−n6τ65+W56n5−W65n6−R61NErn6n1+R35NErn3n5,(6)n1=1−n2−n3,(7)n5=1−n6,(8)where Pp is the pump power, Ps is the seed power, and PASE,λ is the ASE power in the Yb3+ band at wavelength λ. ni is the population density in the ith energy level, Wij is the stimulated transfer rate from the energy level i to the energy level j, Rij is the energy transfer between Er3+ and Yb3+ ions, and Cup is the upconversion coefficient. Γ, σ, and α are, respectively, the overlap factor, cross section, and loss of pump and seed. τij is the lifetime of the ith energy level with respect to the transition in the jth energy level, t is the time, and z is the distance. NEr and NYb are the ion concentrations of erbium and ytterbium, respectively.

Given the finite speed of energy transfer, scenarios where the signal power is insufficient or the pump absorption rate is excessively high can lead to the accumulation of Yb3+ ions in the upper level, ultimately producing ASE and potentially inducing parasitic oscillations. To suppress ASE, we adopted the off-resonant pumping approach, utilizing a pump diode with a wavelength of 915 nm, which exhibits a lower absorption coefficient compared to 976 nm. Figure 2(a) presents the simulation results of Yb-ASE along the gain fiber at various pump wavelengths based on the aforementioned equations. It is evident that the ASE levels are reduced when pumping at 915 nm. Furthermore, we investigated the impact of pumping direction. When employing the counterpropagation pumping scheme, the power distribution of both the pump and signal remains consistent along the fiber. As the signal power increases, more particles are consumed, thereby effectively suppressing the generation of ASE. The simulation results for different pump directions are depicted in Fig. 2(b).

When high-power light propagates through the fiber, it experiences an increase in refractive index, leading to an electro strictive effect. Consequently, a significant portion of the transmitted light is converted into the backward-scattered light. The equation of threshold power can be mathematically expressed as Pth=GAKgBLΔvp+ΔvBΔvB,(9)where G is the coefficient of threshold gain. Its value equals 21 in the Smith theory model and 19 in the Küng model^[23]. K is the polarization-dependent factor, which takes 1 and 2 for linear and random polarization. A and L are the effective area and propagation length of the fiber. gB is the peak value of Brillouin scattering. Δνp is the linewidth of the laser, and ΔνB is the gain bandwidth of SBS. When the linewidth is much less than the gain bandwidth, the rightmost portion of the equation can be simplified to 1.

The simulation results for various parameters of the threshold power are presented in Fig. 3. From Figs. 3(a) and 3(b), it is evident that the threshold power increases as the length decreases or the area increases. Based on this observation, the core diameter of the gain fibers we selected is 25 µm. Additionally, to balance the pump absorption, typically 2.6 dB/m at 915 nm, with SBS suppression, a 5 m gain fiber in the amplifier is deemed an appropriate choice.

Simultaneously, we noticed that despite the gain fiber being longer than the following passive fiber, the light power progressively increased along the gain fiber. This implies that the corresponding propagation length for higher power is shorter. But in the passive fiber section, the light power reaches its maximum and remains essentially constant, exhibiting a lower threshold power. We could optimize the amplifier configuration by shortening the passive fiber in the output section. It is noteworthy that the cladding power stripper would be placed before the gain fiber in the counterpropagation pumping scheme (as illustrated in Fig. 4). This configuration is advantageous as it allows for the use of the shortest pigtail behind the gain fiber, thereby enhancing the SBS threshold. Therefore, the counterpropagation pumping scheme we selected serves the dual purpose of inhibiting both ASE and SBS simultaneously. Considering that the change of the temperature gradient of the fiber is also effective in suppressing SBS^[24] and the main amplifier is mounted on a cooled plate, we could reduce the SBS gain by adjusting the plate temperature to alter the temperature gradient, while ensuring the normal operation of the device.

Following the simulation and analysis, we have developed a single-frequency all-fiber amplifier, comprising a seed laser and two amplification stages. The schematic of the MOPA setup is shown in Fig. 4. The homemade seed laser delivers 14 mW at a wavelength of 1541 nm. Its linewidth is approximately 4 kHz measured by the delayed self-heterodyne method. This seed is then amplified to 7 W by a commercial preamplifier module. An optical isolator is incorporated to mitigate possible damage caused by the backward light. To monitor the backward light, including normal end-face reflection and SBS, emanating from the main amplification stage, a coupler with a coupling ratio of 1:99 is reversely connected to the system. A mode field adapter is spliced after the coupler to match the pigtails of subsequent devices. In the main amplification stage, a 915 nm multimode laser diode serves as the pump source, and the residual pump is removed by a cladding light stripper. The gain fiber utilized is PLMA-EYDF-25/300, with a length of 5 m. Ultimately, the amplified signal light exits the fiber end-face with an angle of 8° to reduce the Fresnel reflection. The entire main amplification stage is temperature-controlled using a cooled plate to prevent heat accumulation-induced damage.

Since the power range of optical instruments is limited, direct analysis of the output laser is infeasible except for the power meter. Therefore, a laser beam splitting-coupling system was designed and constructed for measurement purposes. We employed the end-face reflection of a wedge prism for low-power splitting and carefully adjusted the angle of the fiber collimator to achieve coupling of space light into the fiber. Due to the large divergence angle of the direct laser output, a flat-convex lens is positioned in front of the wedge prism to collimate the beam and enhance the coupling efficiency. Both the wedge prism and lens are made of fused quartz, and the latter is coated with an anti-reflection film (AR at 1000–1600 nm) to minimize light reflection. Subsequent experimental results indicate that light at the 10 mW level can be collected using this method, fulfilling the requirements for measurement and analysis.

The output power of the preamplifier is 7 W, but the actual input is only 5 W due to insertion loss in the fiber devices. When amplifying the laser without implementing any SBS suppression measures, the spectrum of backward light, as shown in Fig. 5, reveals a spike in the longer wavelength band when the output exceeds 30 W. The frequency difference between the two peaks is about 11 GHz, which corresponds to the frequency shift of Brillouin scattering. As the output power gradually increases to 130 W, the SBS peak rapidly rises and surpasses the signal peak by 20 dB, with an estimated actual SBS of about 2 W accordingly. This nonlinear increase in SBS becomes the primary limiting factor for further power amplification.

To alleviate this issue, we first investigated the influence of cooled plate temperature on SBS gain by comparing the backward optical power at the same fiber length and pump but varying temperatures. The results indicate that the SBS effect weakens as the temperature decreases above 16°C. The curves of backward optical power are plotted in Fig. 6. Considering that excessively low temperature could lead to additional energy consumption and potentially impact the performance of the pump diode, we determined that the optimal operating temperature for the cooled plate is 16°C. Subsequently, based on the simulations conducted in Sec. 2, we shortened the passive fiber behind the gain fiber. We progressively decreased the length of the fiber pigtail and recorded the backward power (including the reflection of the fuse end and SBS) at different output powers. The backward power measured directly at the monitor end (representing only 1% of the total backward power) is shown in Fig. 7.

When the fiber was reduced from 1.2 to 1 m, the threshold power of SBS increased, accompanied by a rapid decrease in backward power from 1.7 to 0.3 mW. Upon further shortening the fiber by another 10 cm, the SBS further decreased, resulting in a backward power of only 0.15 mW. If the threshold power increases sufficiently, the SBS effect will not decrease significantly, explaining the near-identical curves observed for fiber lengths of 0.9 and 0.8 m. At this point, it is plausible to assume that the backward light primarily consists of the reflected signal light.

With effective SBS suppression, the amplifier power could be further increased without significant SBS interference. Finally, we achieved an output power of 202 W at a pump power of 481 W, yielding an optical-to-optical efficiency of 42%. The linewidth was well-maintained to be about 4 kHz during the amplification, and no obvious broadening was observed. The total backward power was measured to be less than 0.2 W, indicating that the SBS was not a limiting factor in this scenario. The corresponding plots of output power versus pump power and efficiency are depicted in Fig. 8(a). The signal-to-noise ratio (SNR) we observed using the optical spectral analyzer (OSA) [as shown in Fig. 8(b)] was 23 dB, which is constrained by the SNR of the preamplifier and the sensitivity of the OSA. This SNR could be further enhanced through future optimizations of the preamplifier design, such as adopting a narrowband filter and increasing the preamplifier power. In addition, we noticed a sharp increase in ASE in Yb-band, as illustrated in Fig. 8(c). This suggests that the ASE could induce parasitic oscillations and potentially damage the system with a further increase in pump power. This issue might be alleviated by enhancing the preamplifier power and improving splicing quality. Alternatively, splicing a section of ytterbium-doped fiber before or after the gain fiber section to absorb ASE in the 1 µm band also presents a promising solution^[25]. However, the length and splicing quality of this section must be optimized.

In this work, we demonstrated an all-fiber PM fiber laser amplifier operating at the C band. Off-resonant and counterpropagation pumping schemes were employed to reduce the ASE level in the Yb band by utilizing a 915 nm diode. Notably, we found that the length of passive fiber following the gain fiber played a crucial role in SBS suppression, thereby limiting the maximum amplifier output power. By reducing the passive pigtail length to approximately 0.4 m, we achieved a single-frequency amplification output power of about 202 W at 1541 nm, with an optical-to-optical efficiency of 42%, and an SNR exceeding 20 dB, currently limited by the SNR of the commercial preamplifier. Therefore, we believe that PM single-frequency fiber laser amplifiers with a significantly higher SNR can be realized by improving the performance of the preamplifier via narrowband filtering and necessary optimization. With the help of noise suppression techniques, such as active feedback and second-harmonic generation in passive optical cavities^[26,27], a high-power low-noise single-frequency fiber amplifier would be very promising in applications in high-precision interferometry and gravitational wave detection.