1. Introduction
High-power femtosecond lasers are extensively employed within various scientific and industrial applications [
1,
2,
3]. Ultrafast lasers are in high demand because they can generate high-order harmonic [
4] or terahertz radiation [
5], which opens up novel and ambitious fields, including laser particle acceleration [
6]. To create high pulse energies and ultrashort pulse durations at high repetition rates, such applications often benefit significantly from large mean powers, which improve both the measuring and processing rates [
7]. Most femtosecond lasers today are chirped pulse amplification (CPA) systems [
8] founded on flexible gain fibers or bulk solid crystals. Traditional bulk solid-state laser schemes excel when producing pulses with a high peak power [
9,
10,
11] and short pulse durations [
12]. However, they are limited in terms of the repetition rate and average power due to beam distortion produced by thermo-optical phenomena [
13]. Ultrafast fiber lasers have significant advantages when weighed up against traditional ultrafast solid-state lasers in achieving high-mean-power scaling and exceptional beam quality due to their superior heat dissipation, low quantum-defect heating, and specific fiber geometry.
Furthermore, fiber lasers are recognized as especially promising and feasible tools in various applications because of their flexible propagation route, large single-pass gain characteristics, and high-power cladding-pumping technique. Then, due to the large surface-to-volume ratio relating to the fiber geometry, the thermal-induced beam distortion is minimized when under high-power pumping conditions, ensuring outstanding spatial mode quality using the inherent waveguide features. On the one hand, fiber femtosecond lasers have received widespread praise for their large mean power capabilities, simplicity of experiential configuration, and outstanding beam quality. In contrast, mixing the chirped pulse amplification (CPA) and the extensive mode area (LMA) fiber systems may considerably increase performance. Over the last few decades, their promise has been fully realized. The kilowatt-level average power of chirped pulse amplification-based femtosecond fiber lasers has been demonstrated [
14,
15]. However, nonlinear effects, such as Raman scattering or self-phase modulation (SPM), restrict pulse peak power growth in the fiber amplifiers [
16]. Moreover, the transverse mode instability limits additional power scaling in the single-fiber chirped pulse amplifiers [
17].
A PCF-based amplifier is a technical approach that may significantly boost the average power and pulse energy. PCF has a larger mode area than the double-cladding fiber and can operate reliably in the single mode. Because of the larger mode area, nonlinear effects can be effectively reduced. The rod-type PCF has been successfully employed in high-power and high-pulse energy systems [
18,
19,
20,
21,
22,
23]. Moreover, a high pulse energy output of 1 mJ has recently been obtained employing rod-type PCF [
21]. A pre-chirp managed amplification was demonstrated in a double-pass rod-type Yb-fiber amplifier containing an output power of 102 W and pulse duration of 55 fs [
22]. The NKT Photonics A/S employs a cascade of two-stage rod-type PCFs to decrease nonlinear effects while achieving an average output power of 248 W [
23]. At present, the transverse mode instability is the main limitation for further power scaling by using rod-type fiber [
24].
Coherent beam combination (CBC) is a revolutionary technology to improve average power and pulse energy. Because passive-combining optical components provide higher peak and average power than gain fiber, CBC enables power scaling by orders of magnitude compared to a single amplifier. The additional cost is the requirement for phase stabilization in order to preserve constructive interference. This power scaling concept was initially used in continuous light [
25,
26,
27,
28] and long pulse amplification [
29,
30] systems. Coherent beam combining of ultrashort pulses has been demonstrated in recent years [
31,
32]. The 10.4 kW average output power [
33] and 10 mJ single pulse energy [
7] of the multi-channel coherent beam combination system based on active phase stabilization of the spatial structure have been successfully demonstrated. All-fiber-structure pulse coherent beam combining technology has also been developed. The fiber structure has been more compact and stable than the spatial structure in recent years. In 2018, Zhou et al. developed high-power coherent beams that are a combination of dual-channel femtosecond pulses utilizing a fiber-coupled LiNbO
3 integrated phase modulator and a piezo-driven fiber stretcher based on an all-fiber active control system [
34]. However, the kind of fiber employed in the all-fiber coherent beam combining structure is restricted, and the mode field area of the utilized fiber is lower than that of photonic crystal fiber. While the all-fiber structure can achieve high power output, due to the limits of the fiber mode field area and the employment of numerous passive fibers, it is hard to produce high-energy pulse output during the pulse amplification process because of nonlinear effects. The key to achieving high-power and high-energy ultrashort pulses is to reduce the concentration of nonlinear B-integrals throughout the amplification process. To reduce the B-integral, previous research has often used spatial light modulators (SLMs) to compensate for nonlinear phase shifts caused by processes such as self-phase modulation [
7,
33]. However, the installation of SLM will result in significant losses and will need complex feedback management to accomplish spectral chirp pre-regulation.
In this article, we present a two-channel femtosecond pulse high-power active-phase-stability coherent beam combination system based on a rod-type fiber amplifier. To limit the utilization of passive fibers in the system and hence the buildup of B-integrals, we designed a simple fiber-integrated device to reduce the use of passive fibers and mitigate the accumulation of nonlinear B-integrals during the amplification process. The drift that arises for the optical path is compensated for automatically with a piezo-driven space delay line founded upon Hänsch–Couillaud (HC) stabilization technology [
35]. The coherently combined average power is 165 W, with a combining efficiency of 91.6% and a compressed pulse duration of 330 fs without using an SLM. The output pulse has a shorter duration than a two-channel coherent combining system employing standard fiber optic devices. The capacity of rod-type photonic crystal fibers to beam coherently that is combined to overcome the power limitations of single-fiber amplifiers is demonstrated in this laser system. The power of the laser system can be enhanced when the amplifier’s pump power grows. When involved with electro-optical controlled-phase-modulation distribution amplification, the system’s average power and single pulse energy may be enhanced even further, allowing for kW-level average power and an mJ-level single pulse.
2. Experimental Setup
Figure 1 shows a representation of the system. The optical fiber and its components used here employ a polarization-maintaining (PM) structure, which is used to enable a linearly polarized output. A broadband ytterbium-based oscillator, generating femtosecond pulses at a repetition rate of 40 MHz, serves as the seed source. A nonlinear fiber amplifier, based on PM 6 μm core fibers and their components, permits the broadening of the output spectrum of the seed source to 20 nm. The oscillator and nonlinear amplifier serve as the system’s front-end seed.
We coupled 10% of the nonlinear amplifier light to generate the trigger of the digital delay signal for the subsequent pulse picker. The pulses were stretched to the nanosecond regime using a dual CFBG with 17.6 nm bandwidth. It delivered the third-order dispersion (TOD) and group velocity dispersion (GVD) of −1.11 ps3 and 73.53 ps2, respectively. We constructed a PM SM amplifier with 3 m long, low-concentration Yb-doped fiber to counteract the loss of the CFBG and optical circulator. Then, using a fiber acousto-optic modulator (AOM), the repetition rate was decreased to 1 MHz. The following two-stage fiber boost amplifier increased the pulse energy. A 9 W, 976 nm multimode laser diode (MMLD) pumped a 1.5 m long gain fiber (PLMA-YDF-10/125-HA) in the first boost amplifier.
The B-integral is commonly used to estimate the nonlinear phase of the amplification process in ultrashort pulse fiber amplifiers [
36]. When the pulse is transmitted in both the active and passive fibers, a corresponding nonlinear B-integral is accumulated [
37]. The longer the passive fiber is used, the more nonlinear phase shift is introduced. Shortening the passive fiber can effectively slow the nonlinear B-integral. The passive fiber of 0.5–1.0 m will still be maintained after the gain fiber in the classic all-fiber-construction double-clad amplifier. This passive fiber will introduce a B-integral of 0.2–1 rad or more when the repetition frequency is 1 MHz or below. To reduce the passive fiber, we integrated the combiner with the output collimator and a WDM filter, which enabled the pump laser to couple into cladding in the opposite direction of the signal. This integrated fiber component utilized the pre-taper processing, which prevented the spectral modulation problem caused by the combiner. A fiber stripper was added to the double-clad preamplifier to help the remaining pump light leak into the fiber cladding. To improve the beam quality, we coiled the fiber to a diameter of 6 cm. In the second boost amplifier, a single-polarization double-clad Yb-doped extensive-mode-area PCF (DC-200-40-PZ-Yb-03, NKT) with a 40 μm core diameter was employed; this was pumped by use of a 25 W multimode laser diode. Both ends of the PCF were collapsed and clipped at an angle of 8°, preventing the high energy density from breaking the fiber end face and suppressing parasitic oscillation. The fiber was coiled to a diameter of 25 cm to enhance the loss of the higher-order mode. An optical isolator was inserted after each fiber amplifier to prevent the returned light. The beam was spatially separated by employing polarizing beam splitters (PBSs) and half-wave plates (HWPs) to seed the next main amplifier stage.
The main amplifier stage consisted of two parallel Yb-doped rod-type amplifiers. Integrated water-cooled rod-type PCFs manufactured via NKT were used, which contain a core diameter of 85 µm, a core numerical aperture (NA) of 0.03, a length of 0.805 m, a cladding absorption coefficient of 15 dB/m at 976 nm, and a cladding diameter of 260 µm. For pumping, a 200 W fiber-coupled laser diode containing an NA of 0.22 and a core diameter of 105 μm was utilized. The average power could be boosted by more than 100 W in the main amplifier stage. Collimated by the lens was the output light from the PCF amplifier, and two dichroïc mirrors (DMs) were used to couple the signal into the main amplifier. The DM utilized in the research had an extensive reflection at 1030 nm and a large transmittance close to 976 nm, which coupled the signal into the fiber while effectively filtering out the remaining pump light. The separate output polarized beams were combined using thin-film polarizers (TFPs). After the combining process, a non-polarizing beam splitter with a split ratio of 1:99 was used to provide the weak beam sample for Hänsch–Couillaud polarization detection. The PCF amplifier’s output pulse was spatially separated into two beams. One of the optical paths was installed with a piezo-mounted mirror for interferometric measurement and phase stabilization. To reduce the loss in the combination process, we expanded the spot size to 3 mm, and then a TFP was employed for coherent beam combination in space. High-diffraction-efficiency gratings, containing a line density of 1760 line/mm, were utilized as the compressor, for which the diffraction efficiency at 1030 nm was as high as 98.4% when incident at the Littrow angle. The compressor is operated at an incidence angle of 70° with a diffraction efficiency of about 95% to decrease high-order dispersion mismatch and increase pulse quality. Despite four passes across the dielectric gratings, such a compressor scheme still had a large compression efficiency of 81% due to the gratings’ excellent diffraction efficiency.
3. Results and Discussion
This work was aimed at designing a reliable high-power ultrashort pulse coherent beam combination system, while improving the pulse peak power of the CPA output system without pulse shaping at 1 MHz. This CPA fiber amplifier could function safely and securely despite any optically induced damage at this repetition frequency. The fundamental precondition of achieving femtosecond pulse CBC was to strictly assure the phase synchronization of pulses in separate channels, which also requires adequate coherence between pulses. The optical path difference in this experiment was mostly produced through the mismatch of free space optical path lengths between various channels and the optical path difference drift generated from external environment interference throughout the amplification process. To stabilize the entire system, based on a piezo-driven scheme, we employed an adaptive control configuration for the optical path drift, which provided an optical path drift adjustment range of ±30 λ.
Figure 2 depicts the total output energy and average power at all stages. The pulse stretcher and circulator’s partial loss produces a slight decrease in the stretched pulse energy. The main amplifier’s output power is about 90 W after four stages of fiber amplification, corresponding to a single pulse energy of 90 μJ. The total efficiency of the coherent beam combination is 91.6%, resulting in 165 W combined average power, which relates to a single pulse energy of 165 μJ.
Figure 2c depicts an accumulation of the nonlinear B-integral at various stages of the system. Benefitting from highly integrated fiber components in each stage, we have significantly reduced the number of passive optical fibers, reducing the accumulated nonlinear B-integral. The B-integral of the entire system is 1.37 rad by simple calculation, which is implemented to estimate the nonlinear phase shift accumulated during amplification. The output power is 132 W after a pair of gratings compressors whose compression efficiency is 81%.
Figure 3 depicts the spectral evolution procedure during the amplification procedure, which is measured via OSA (YOLOGAWA AQ6370). The adopted CFBG cut the seed pulse’s spectral bandwidth to 17 nm, and then gain narrowing caused it to narrow further to 8.5 nm after four amplification stages, resulting in a 180 fs Fourier-transform-limit pulse duration. This compressor gratings’ vertical spacing was adjusted to 700 mm with an incident angle of 70° during the experiment, offering the GVD and TOD of −72.8342 ps
2 and 0.9145 ps
3, respectively. This system’s total fiber length was nearly 25 m, and the introduced dispersion was 0.575 ps
2 and 1.098 × 10
−3 ps
3. Without considering nonlinear effects, the total dispersion of the system is entirely matched, and the CFBG should provide 73.4092 ps
2 and −0.9156 ps
3 of dispersion. However, at this point, the dispersion of CFBG is not the best compression setting because of the influence of the SPM effect during amplification. Thanks to temperature control, the dispersion provided from CFBG can be precisely adjusted to compensate for the dispersion produced via some nonlinear effect. Introducing suitable positive dispersion can compensate for nonlinear effects, such as the SPM in the CPA system. To ensure the same B-integral for each amplifier during the amplification, we adjusted the pump current of each amplifier to a different value in which the duration and autocorrelation (AC) trace overlap, which indicates identical nonlinear phases. The next steps of the combination optimization approach must be completed once. For the seed side of the main amplifiers, i.e., underneath the piezo-driven mirrors, electronically controlled micrometer translation stages must first be used to match the optical path lengths of the two primary amplifier channels. The spatial superposition of two beam spots in the far and near fields was accomplished by employing high-speed CCD. Finally, to obtain the best combination efficiency, we rotated the HWP before the TPF to adjust the polarization state of the spot.
Optimizing these degrees of freedom produced a coherent combining and fully compressed average power of 132 W, corresponding to a pulse energy of 0.13 mJ. At this operation point, the combining efficiency, defined as the ratio of the combined average power to the total of all output powers from the individual main amplifiers, was 91.6%, showing that all previously specified parameters were well matched.
Figure 4 indicates that the pulse shape and duration were not substantially altered after the combination, which had minimal effect on the compressed pulse quality. It is difficult to improve the efficiency of coherent beam combining further because of the slight difference between the output spectra of channel 1 and channel 2; the pulse duration after beam combining is also limited, resulting in a Gaussian fitting pulse duration of 330 fs after beam combining. Furthermore, the M
2 was calculated using the 4σ method and exhibited a nearly diffraction-limited beam quality with an M
2 < 1.35 in both axes, as shown in
Figure 5b. The combined beam maintained high beam quality during transmission. We employed a pair of lenses to extend the beam during the beam-combining procedure. The output spot M
2x increased because the lenses employed were not aspheric lenses.
The residual phase error of the optical path difference during active phase control is shown in
Figure 6a over a period of 20 s. The determined RMS value of 0.12 rad relates to an λ/50 in optical path length fluctuation. The radio frequency spectrum of such a fluctuation is found by calculating the Fourier transform of the optical path difference over time, as illustrated in
Figure 6b for the open-loop and active phase-locking conditions. These oscillations are caused by the thermal expansions of the materials that transmit the laser beam or environmental vibrations. As a result, most of the fluctuation occurs in a frequency range approaching 100 Hz, where the stabilizing mechanism considerably lowers the variations in the optical path difference between the two interferometer arms.
The present system is constrained by two primary factors that restrict the expansion of combined channels and combined power: the limitation of front-end output power and the total B-integral. In order to attain high amplification efficiency for the main amplifier, it is necessary for the input power to achieve its saturation power. Consequently, increasing the number of system channels requires a corresponding amplification of the output power of the front-end amplifier. Nevertheless, the front-end amplifier can be classified as a single-fiber amplifier, with its output power limited by the phenomenon known as mode instability. Simultaneously, increasing the output power of the front-end amplifier will result in an elevation of the overall B-integral of the system, influencing the quality of the final output pulse. In the present structure, the primary amplifier necessitates a minimum incidence power of 3 W or above. Through the optimization of the front-end preamplifier structure, it is capable of delivering a preamplified output of approximately 90 W and allowing the integration of roughly 30 channels’ coherent combining. One additional limitation that impacts the number of combination channels is the accumulation of B-integrals. The increase in preamplification output power will result in significant amplification of front-end B-integrals, thus impacting the final quality of the output pulse. Simultaneously, the output pulse is influenced by the phenomenon of gain narrowing that occurs during the amplification process. The limited spectral width will also have an impact on the quality of the pulse. Currently, the prevailing approach to address the effects of the B-integral is to employ an SLM for phase modulation of the pulse before it enters the preamplifier. The SLM is also utilized for pre-compensation of the high-order dispersion mismatch and nonlinear phase induced by each component of the amplifier. This is performed to guarantee the desired quality of the output pulse.
The consideration of phase control technology has significance while expanding the number of combined beam channels within the system. Employing HC phase detection technology often involves a combination of beam channels that follows an expanding exponential structure, with the number of channels being a power of 2. As the quantity of channels increases, the phase detector also experiences linear growth, resulting in an increase in the overall cost of the system. Hence, the development of simplified locking structures for Stochastic Parallel Gradient Descent (SPGD) locking technology [
38] or Locking of Optical Coherence via Single-detector Electronic-frequency Tagging (LOCSET) phase locking technology [
39] is a crucial approach to enhance the system’s combined beam channels and output power, as well as to improve the system’s compactness.
The enhancement of the peak power of the seed source and the power scaling ratio of the front-end preamplifier are crucial factors in increasing the number of combined channels inside the system, increasing the output power of the system, and enhancing the compactness of the system structure. The pure-quartic soliton laser offers a novel approach to the generation of high-energy pulses [
40,
41]. In contrast with traditional soliton fiber lasers, it has the capability to deliver increased energy output. The power scaling ratio of the optical fiber preamplifier is an important factor. In contrast to the single-pass amplification configuration, the double-pass amplifier has a greater power scaling ratio. A single-stage amplifier has the capability to successfully amplify the average power pulse from the milliwatt level to the hundred-watt level [
22]. By reducing the number of preamplifiers, significant cost savings can be achieved in the system. Nevertheless, it should be noted that the double-pass amplification structure shows more susceptibility to mode instability compared to conventional amplification. Specifically, the mode instability threshold of the double-pass amplification structure is only half that of conventional amplification [
24].