The fiber laser has undergone rapid development, with outstanding properties in terms of power output, brightness and optical efficiency^[1^,2^]. Remarkable progress has further promoted diverse applications in the fields of manufacturing industry, medical science, remote sensing and communications. At the time of writing, the most common implementation is based on active fibers such as Yb-doped fiber, boosting the record for power scaling to greater than 10 kW^[3^–5^]. Another way to obtain high-power laser output is the Raman fiber laser (RFL), which employs the stimulated Raman scattering (SRS) effect in a piece of passive fiber without any rare-earth dopants^[6^–8^]. Compared with Yb-doped fiber lasers (YDFLs), the most attractive feature of the RFL is wavelength flexibility. It can emit across almost the whole spectrum of transparency of a silica window, in particular beyond the emission band of rare-earth-doped fiber lasers^[9^–19^].

The power scaling of continuous-wave RFLs has recently reached 2 kW, with a best beam quality factor M² of ~8.9^[20^]. It is, however, challenging to achieve high beam quality at this power level, especially when the enlarged fiber core is employed to suppress second-order Stokes light. In fact, the output beam quality of high-power RFLs can also be good enough to meet practical requirements, based on brightness enhancement (BE) employing either cladding-pumping double-clad or multi-clad fibers^[21^–23^], or core-pumping graded-index (GRIN) fibers^[12^,18^–20^]. In the former scheme, the coupled pump light in the cladding is converted into the Stokes laser transmitted through the core of fiber to obtain BE. The power record for cladding-pumped RFL is 1.2 kW, and the corresponding M² is 2.75 with a BE of ~7^[23^]. The manufacture of these unique Raman fibers is usually complicated and costly, however, the cladding-to-core area ratio required is small (< 8) to suppress high-order Stokes light, imposing a limitation upon BE capacity based on the currently available pump brightness^[24^].

As an economical substitute, core-pumping GRIN fiber has an advantage over beam clean-up effects for enhancing the output brightness without complex design of the fiber structure^[25^,26^]. With the rapid advancement of high-power laser diodes (LDs), it was feasible to directly apply LD pumping in the GRIN RFLs for BE. Yao et al. proposed an LD-pumped RFL in 2015, obtaining a 20 W laser from GRIN fiber with an M² that was improved from 22 to 5, and a BE of 5.2^[27^]. In 2016, Glick et al. reported an LD-pumped RFL based on GRIN fiber^[28^], and the maximum power output reached 154 W with an M² of 8 and a BE of 3 through optimizing the fiber length^[29^]. Zlobina et al. also studied RFL based on GRIN fiber, and the beam quality was quite good with an M² of ~1.2 and a BE of ~40, and the demonstrated power was 10 W^[18^].

The fabrication of fiber Bragg gratings (FBGs) directly on the GRIN fiber enables the selection on the fundamental mode, further boosting the BE based on all-fiberized RFLs^[30^]. The maximum power of LD-pumped GRIN RFL reached 62 W, and the M² was 3 with a BE of 25^[31^]. Using the customized FBGs, the all-fiber GRIN RFL pumped by fiber lasers obtained 135 W power with an M² of 2.5 and a BE of 5.6^[32^]. In general, the employment of FBGs written on GRIN fiber enabled all-fiber RFL with not only improved efficiency but also higher BE, with an M² improved to 2–3. The process for manufacturing FBGs with multimode (MM) GRIN fibers is relatively complex, however, it is challenging to further scale the power output with good beam quality through the resonant cavity.

To obtain power boosting without the critical requirements of MM GRIN FBGs, one can use the master oscillator power amplifier (MOPA) system, where a relatively low-power oscillator with stable properties is applied as the seed source. The MOPA system has shown significant potential for high-power and high-brightness laser generation based on rare-earth dopants. Furthermore, the abilities of Raman amplification based on MOPA have been proved in recent years^[20^,33^–35^]. In 2018, our group employed a high-power combiner and proposed an all-fiber Raman fiber amplifier (RFA) based on GRIN fiber, whose power output was 528.8 W with an M² of ~4.2 and a BE of 3.8^[33^]. Not long after, the power output was improved to 1002.3 W through optimizing the fiber length, with an M² of 5.06 and a BE of 2.6^[34^]. Most recently, the power output of the RFA was improved to 2 kW, while the poor input seed brightness and serious thermal effects resulted in apparent beam spot distortion, preventing the BE process, and the best beam quality M² was only 8.9^[20^]. The power scaling capacity of the RFA has been proven, but more research and discussion are still required to simultaneously achieve high BE and good beam quality at this power level.

To explore efficient BE in high-power RFAs employing GRIN passive fiber, the optimization of seed injection was adopted. We demonstrated an RFA with an optimized input seed beam quality M² of 3.37, which is ~4.7 times higher than that in previous work^[20^]. The RFA employs a MOPA, and the maximum power is 2.034 kW with a conversion efficiency of 79.35%. The beam quality parameter M² at maximum power output is ~2.8, with a BE of ~11.2, which is the best beam quality for GRIN RFLs of over 150 W. Thanks to the improved input and output brightness with fewer transverse modes, the beam fluctuation caused by thermal effects has also been greatly suppressed. Meanwhile, this system gives the highest BE parameter for high-power Raman lasers with over 100 W power.

In order to ensure high-power transmission efficiency as well as conservation of the injected seed beam quality at the same time, a (6 + 1) × 1 pump and signal combiner (PSC) was optimally fabricated, employing the same GRIN fiber in the amplifier stage as the output fiber. The core and cladding diameters of all seven input fibers are 20 μm and 130 μm, respectively. The output fiber of the PSC has core and cladding diameters of 62.5 μm and 125 μm, respectively, and the numerical aperture (NA) of the core is 0.275. The cross-section of the input bundles and the fusion point are shown schematically in Figure 1(b). The conventional PSC, which has been widely applied in cladding-pumped fiber lasers and amplifiers, utilizes the pump and signal laser beams injected separately into the cladding and core of double-clad fiber^[36^–38^]. The PSC in our work, however, is designed to combine the pump and signal lasers together into the core of the GRIN fiber. Thus, to maximize preservation of the seed beam quality, the central fiber of the tapered bundles, which has been accurately fixed at the center of the bundles, is employed for seed injection. The parameters for tapering and splicing during the PSC fabrication process were simulated and optimized to achieve precise alignment and to minimize the mismatching of mode fields of lower-order modes. The powers of the seed and pump lasers after the PSC are measured individually by an optical power meter (OPM) with high power endurance. The insertion losses of the PSC for the seed (153 W) and pump (3 kW) are separately measured to be less than 5%.

Another piece of GRIN fiber was then spliced with the pigtail GRIN fiber of the PSC, these two fibers having identical parameters. The total length of the GRIN fiber in the RFA is ~20 m, which has been optimized to suppress the higher-order Raman light. At the far end of the GRIN fiber, an end cap is employed to ensure safe lasing and the elimination of backward reflection. For removal of the thermal load, the fibers are coiled around a metal barrel whose temperature is maintained at ~20°C.

The beam qualities of the signal laser and pump laser with respect to the output signal power are measured by a beam quality analyzer (BQA), and are depicted in Figure 4. After the PSC, the M² of the launched pump laser at varying power levels remains between 10.1 and 10.8. The M² of the seed laser after the PSC is ~3.37, which is much better than the seeding beam quality of ~15.84 in the previous demonstration^[20^]. This improvement can be attributed to the optimization of seed injection accomplished by the optimized PSC. Note that there is a fluctuation in the M² curve of signal laser at a power lower than 1 kW. According to the power endurance of the BQA, different sets of lenses are applied to attenuate the incremental signal laser to an appropriate power level for measurement. At a power lower than 1 kW, multiple adjustments are applied to the free-space optical path according to the increasing power of the signal laser, inevitably resulting in measurement error. With power scaling of the amplifier, the M² of the signal laser shows an overall tendency toward improvement, thanks to the BE process, and finally reaches ~2.8 at the maximum power. As far as we know, this is the best beam quality for a 2 kW-level Raman laser in any kind of configuration with BE. In Figure 4, the M² of the residual pump light degrades gradually from 11 to ~14.8 with respect to the incremental signal laser power. This implies that the Raman conversion is accomplished by the portion of pump laser with brighter beam quality.

Laser brightness (B) is defined as the laser power per unit area and per unit solid angle, and it can be expressed as^[2^]where P is the laser power and λ is the wavelength. In the amplifier, the value of BE represents the ratio of the signal output brightness to the launched pump brightness, as shown in Figure 5(a). When the signal laser power increases from 299 W to 645 W, the BE value decreases from 8.8 to 6.3. This is due to insufficient Raman conversion at the beginning of amplification. With a constantly increasing low-brightness pump power, the power of the high-brightness signal laser increases at a slower rate, resulting in a decrease of BE, according to Equation (1). This can also be proved by the relatively low slope efficiency in Figure 2. When more pump laser is launched, the Stokes shifting is more efficient, and the high-brightness signal laser power grows rapidly. Correspondingly, the BE value increases to ~11.2 at the maximum power output. This BE value is higher than in any previous high-power BE-enabled Raman lasers (over 100 W power output). Figure 5(b) shows the beam shapes at the focal spot in the BQA. Before the PSC, the beam spots of the pump and seed lasers have regular Gaussian-like profiles. After the PSC, the beam shape of the seed laser remains, whereas those of the launched and residual pump laser degrade. At maximum power, the beam spot of the amplified signal laser is also Gaussian-like and symmetric, indicating that the optimization of the launched seed laser is effective to guide the amplification of the high-brightness laser.

In the experiment we achieved a significant improvement of output beam quality in a 2 kW RFA compared with the previous result^[20^]. This can be attributed to the optimal design of the power combination and seed injection. In the amplifier, a homemade 7 × 1 combiner is employed, whose output fibers is a piece of step-index fiber (SIF). The core and cladding diameters of the input and output fibers are 20 μm and 400 μm, and 50 μm and 360 μm, respectively, with a core NA of the output fiber of ~0.22. The pigtailed step-index fiber is incompatible with the GRIN fiber serving as a Raman fiber, however. The mode field mismatching goes against achieving good beam quality of the seed laser, and a good beam quality of the seed laser is hard to maintain. Thus, the SIF pigtailed power combiner without additional design of seeding beam quality optimization cannot ensure superior BE in the amplifier.

By utilizing GRIN fiber as the output fiber of the combiner, the mismatching of gain fiber and the combiner is eliminated, which is also employed in the resonator in Ref. [31]. Nevertheless, there is no specific design for seed injection, and thus it is not suitable for the MOPA system. Accordingly, in this demonstration, the original power combiner is replaced by a (6 + 1) × 1 PSC. On one hand, the parameters of the tapered fused bundles are optimized with an adiabatic shape to satisfy the brightness conservation criterion and ensure high transmittance of the pump and seed lasers. On the other hand, the mode field matching between the input signal fiber and the GRIN fiber is carefully optimized. The central fiber of the input fibers used for seed injection is accurately fixed at the center of the bundle during the fabrication process. The bundles are then spliced directly with the GRIN fiber rather than the unmatched step-index fiber. Consequently, low insertion loss and superior beam quality preservation of the seed laser are realized by the aid of the optimized PSC. Finally, the ultimate beam quality of the RFA is effectively improved when compared with previous results. We achieve the best beam quality for GRIN RFLs of over 150 W. Meanwhile, it is the highest BE value among all high-power BE-enabled Raman lasers with over 100 W power.

Through utilizing the specially-designed PSC to optimize the injection of a seed laser, we achieve a BE-enabled RFA based on GRIN passive fiber with the best beam quality at 2 kW power level. The GRIN fiber-pigtailed PSC well preserves the seeding beam quality, and significantly helps enhance the brightness of the amplified laser. The maximum power output of the 1130 nm signal laser reaches 2.034 kW, with a high conversion efficiency of 79.35%. This is the highest conversion efficiency obtained for any GRIN Raman laser systems. The beam quality parameter M² remains ~2.8 at the maximum power output, and the corresponding BE reaches as high as ~11.2. We achieve the best beam quality for GRIN RFLs of over 150 W. It is also the highest BE value among high-power RFLs with over 100 W power.