We present a coherently combined laser amplifier with 16 channels from a multicore fiber in a proof-of-principle demonstration. Filled-aperture beam splitting and combination, together with temporal phasing, is realized in a compact and low-component-count setup. Combined average power of up to 70 W with 40 ps pulses is achieved with combination efficiencies around 80%.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
The utilization of coherent combination has drastically increased performance figures such as average power, pulse energy, and peak power of fiber laser systems. This could be realized by distributing power scaling challenges across multiple emitters to overcome the different physical limitations of these values. For example, limitations include average power-related effects such as mode instabilities [1–3], but also nonlinear effects for parameters such as pulse energy and peak power. Today, single amplifier channels have been outperformed by more than one order of magnitude by systems employing spatial combination of multiple amplifiers , temporal combination of multiple pulse replicas , or even both . Notable results include a system emitting 12 mJ femtosecond pulses at 700 W average power realized with eight parallel channels and four temporal pulse replicas  or a 4 kW continuous-wave system with eight channels . According to theoretical calculations, systems using even greater parallelization appear to be viable, since the expected total combination efficiency converges with an increasing number of channels N [9,10]. This offers the potential to move fiber laser systems towards performance parameters (e.g., joule-class pulse energies at high repetition rates) required for demanding applications in high-field physics such as laser particle acceleration.
However, in most of the previously demonstrated laser systems based on coherent combination, the component count and, thus, the complexity of the system basically grows linearly with the number of channels for spatial combination. In alleviated terms, this statement also holds true for temporal pulse combination. Therefore, it is necessary to employ integrated multi-channel components in order to decouple the component count from the channel count. The integration of multiple amplification channels into a multicore fiber has already been investigated for a tiled-aperture combination in a low-power experiment . In this case, a spatial light modulator is used to generate a beam pattern matching the core positions in the multicore fiber and to set the phase of the specific beams. 860 fs pulses at a 100 kHz repetition rate were achieved. However, due to the limited fill factor of the tiled-aperture combination scheme, the power in the central feature is theoretically limited to 76%, and a value of 49% was realized experimentally. Using diffractive optical elements, the combination efficiency can be increased by implementing the filled-aperture technique . However, the diffractive nature of the combination element makes this approach unsuitable for femtosecond sources without a pre-compensation of the spatial chirp for each incident beam. Another method is to use multiple coupled cores , but here the power scalability especially regarding the mode instability threshold and the possibility to use them as amplifiers remains unclear. Additionally, the spectral beam combination approach has been demonstrated with multicore fibers [14,15].
In this Letter, we present the realization of a filled-aperture combination setup with a 16-core fiber. The filled-aperture scheme allows for high combination efficiencies and a near-diffraction-limited output beam quality. Additionally, the nondiffractive nature of the employed components allows for the future application to broadband sources, i.e., ultrashort-pulse operation.
The setup, shown schematically in Fig. 1, consists of multiple components that feature a multi-channel design. A laser with 40 ps pulse duration, 20 MHz repetition rate, and 300 mW average power operating at a wavelength of 1032 nm is used as a seed system. The emitted beam first passes through a segmented mirror splitter (SMS). This component consists of one high-reflective mirror and one element with zones of different reflectivity. For this 16-channel setup, there are four zones with reflectivities of 0% (AR-coated), 50%, 66%, and 75%. By placing both elements parallel to each other, the input beam is split into four parallel equidistant beams (beam splitter SMS 1 in Fig. 1). A transmissive beam splitter based on a similar design was presented in . This splitting process is repeated again with a second SMS (beam splitter SMS 2) rotated by 90°. Thus, each of the four beams is split into four beams again, resulting in a two-dimensional output beam array with a 4 mm pitch between the beams. The beams are reflected at a polarization beam splitter cube (PBS) towards a piezo array with mirrors attached to each piezo actuator. Due to a required pitch of 9 mm of the piezo array, the beam array is magnified by a telescope. Each piezo has a piston range of 20 μm, resulting in an effective value of 40 μm in the presented double-pass configuration. While great care has been taken when attaching the mirrors to the piezos, a small static tilt is still present, in addition to a dynamic tilt that occurs during actuator movement. This would result in a distortion of the beam array. Therefore, a lens array with a 100 mm focal length is placed in front of the piezo array, placing the piezo mounted mirrors into the Fourier domain of the beams. This configuration transfers the tilt into a transversal beam offset, which is small compared to the beam sizes, hence alleviating the beam array distortion. Additionally, spherical aberrations of the telescope are reduced due to the horizontal and vertical flipping of each beam in the double-pass configuration. The beam array is directly imaged to the end-facet of the employed multicore fiber with an optical ensemble in a 4f-configuration after passing through the PBS.
The in-house designed fiber consists of 16 step-index ytterbium-doped cores in a configuration shown in Fig. 2. Each core has a diameter of 19 μm with a pitch of 55 μm between the core centers and a numerical aperture of 0.06. Based on previous research on a four-core fiber regarding the average power handling , this ratio between core size and pitch is sufficient to suppress optical and thermal coupling between the cores and, therefore, avoid a detrimental impact on the mode instability threshold. Thus, the multicore fiber can be treated like multiple spatially separated fibers. As can be seen in Fig. 2, the core positions in the manufactured fiber show small deviations in the range of a few micrometers from the perfectly rectangular pattern. The consequences of this issue will be discussed later in this Letter. The multicore fiber has an outer diameter of 340 μm and a low-index polymer coating acting as the shared pump cladding. Additionally, the fiber was given a D-shape in order to guarantee mode mixing of the pump light coming from a 200 μm fiber-coupled 976 nm pump diode. A fiber length of 5 m was chosen for sufficient pump absorption ( for 190 W pump power), and the end-facets were angle-polished to avoid back-reflections. After amplification, the output beam array is again magnified with optics in -configuration to the initial pitch of 4 mm (corresponding to beam diameters of 1.4 mm) before passing through another two SMS elements for beam combination. The distance between the two elements of each SMS at this combination stage is matched to the splitting stage to provide equal path lengths for each beam. At the first SMS (beam combiner SMS 1 in Fig. 1), the 16 incoming beams are combined into four output beams by combining four columns of four beams each in vertical direction. In each of these four columns, this results in three beam interference steps and, thus, in three beams containing the parts that do not interfere constructively. Therefore, the first rejection port is formed by a beam array, which represents the losses of the beam combination process at this stage. A small fraction of this array is guided towards a photodiode array connected to the active phase stabilization system, while the bulk of the power is dumped. At the second SMS (beam combiner SMS 2 in Fig. 1), the four remaining beams are further combined into a single output beam in horizontal direction, thus resulting in a beam array emitted from the second rejection port. Again, a fraction of the rejected light is guided towards a photodiode array. It should be noted that during the propagation through the SMS, the high-power beam is always reflected, which increases the average power handling capabilities. Actually, this setup is very similar to cavity-enhancement setups, where the handling of milliwatt-level average powers has already been demonstrated .
The required active path length stabilization is realized by applying small sinusoidal phase modulations to the piezo array. The top-left actuator remains steady, allowing us to use the corresponding beam as the reference signal. A modulation frequency of 6 kHz is applied to the three other actuators in the first row, while a frequency of 4 kHz is chosen for the 12 actuators in the remaining three rows. Additionally, a phase offset of is added between the phase modulations of adjacent actuators with the same frequency, which reduces residual phase fluctuations in the combined beam. The demodulated error signal was evaluated with a bandwidth of 1 kHz. This multi-detector scheme allows us to stabilize the 16 beams of the setup, even with a limited total bandwidth of the piezo actuators of around 10 kHz by reusing the same two modulation frequencies for multiple actuators .
The fiber does not contain any structures to preserve the polarization state and, thus, polarization changes for different cores were observed, even in a straight piece of fiber. In the experimental setup, the fiber was bent to a radius of about 30 cm. Therefore, static polarization control was necessary for the cores realized by an array of quartz wave plates. This allowed us to have 85% of power in p-polarization after amplification for all average powers.
The most prominent way to determine the quality of the coherent combination process is to measure the combination efficiency. In this case, to distinguish between depolarization in the amplifying fiber and the performance at the SMS beam combiner, it is defined as the output power in the combined output over the total amplified power after the polarizer. In Fig. 3, the combination efficiency is shown depending on the combined power of up to 70 W, limited by the power of the employed pump diode. As can be seen, the results are in a range of a few percentage points around 80%, with a small reduction for higher average powers. Both aspects, the upper limit of the efficiency, as well as this reduction, are investigated in the following.
Considering the overall efficiency, losses during the combination process can be observed by investigating the noncombining parts of the beams. Fortunately, this is easily realizable in the presented setup by directing a fraction of the power from the two rejection ports to cameras (same beams as for the photodiode arrays). In Fig. 4(a), the image from the first rejection port is shown with high exposure time. The noncombining parts have mostly higher-order mode shapes, which can be explained by the presence of varying higher-order mode content in the different signal cores and residual beam alignment errors. The stabilization system optimizes for maximum combined power, meaning the system is optimized for constructive interference of the dominant fundamental mode, while the amplitude and phase of the higher-order modes is random. Therefore, higher-order mode content is suppressed in the combined beams and appears as losses at the rejection port. Supporting this explanation, changing the coupling of the beams changes the observed pattern. The previously mentioned core position deviations in the fiber result in an increasing coupling into higher-order modes of the cores, instead of the fundamental mode and, therefore, increase these losses. Hence, improved production processes with better core positioning have the potential to improve the combination efficiency.
The small reduction of efficiency at higher average powers can be explained by measuring the optical spectrum of the output. In this experiment, spectral broadening caused by self-phase modulation from 130 to 260 pm at the highest average power was observed. Therefore, differences between the input and output powers in each core result in core-dependent spectral width and phases, which reduce the combination efficiency . This effect was observed by an increased sensitivity at higher powers regarding the beam coupling into the multicore fiber.
In order to specify the quality of the combined beam [Fig. 4(b)], which is an advantage of the filled-aperture scheme, an measurement device was employed. At full power, a near diffraction-limited value of was measured for the beam caustic shown in Fig. 5. This value also compares favorably to the observed beam quality of single channels, where values of up to 1.4 were measured, supporting the described beam-cleaning effect.
The power stability of the combined output was determined by connecting a photodiode through a 240 kHz low-pass filter (to suppress the fundamental repetition rate) to a high-resolution oscilloscope and measuring the output signal with a sampling rate of 1 MSp/s. In Fig. 6, the integrated amplitude noise is shown for the combined beam and for the background noise for a frequency range of 1 Hz to 100 kHz. The step-like increase of the integrated noise at 4 kHz and 6 kHz, and their harmonics are due to the applied phase modulations. A relative intensity noise (RIN) of 0.14% was calculated for the combined signal and a value of 0.01% for the background noise. This very small integrated noise demonstrates that the stabilization successfully compensates for phase perturbations of the interferometer channels.
In conclusion, we have demonstrated a scheme for the implementation of a filled-aperture coherent combination with a multicore fiber. Up to 70 W of pump limited average power with near-diffraction limited beam quality could be achieved at combination efficiencies of around 80%. The combination is realized by integrated multi-channel elements implementing beam splitting, combination, and phasing. Hence, an increase of the number of channels can be achieved without increasing the number of discrete optical components. In future research, the combination of femtosecond pulses in such a setup has to be investigated. This requires a matching of the path lengths in the different channels to a few wavelengths in order to guarantee the temporal overlap of such short pulses. Additionally, the average power handling regarding the mode instability threshold will have to be further investigated for the presented two-dimensional core layout. Finally, techniques to manufacture large-area-mode multicore fibers  (e.g., micro-structured fibers) have to be developed, and the number of cores in the fiber has to be increased. The presented concept can then scale the performance of high-repetition-rate femtosecond fiber-based laser systems to a new level such as moving from the current millijoule level pulse energies to the joule level, which will enable a variety of new applications.
H2020 European Research Council (ERC) MIMAS (670557); Free State of Thuringia PARALLAS (2015FE9158); Fraunhofer research cluster “Advanced Photon Sources.”
The authors thank the IPHT Jena for drawing our preform to fibers. M. Müller acknowledges financial support by the Carl Zeiss Stiftung Foundation.
1. T. Eidam, C. Wirth, C. Jauregui, F. Stutzki, F. Jansen, H.-J. Otto, O. Schmidt, T. Schreiber, J. Limpert, and A. Tünnermann, Opt. Express 19, 13218 (2011). [CrossRef]
2. A. V. Smith and J. J. Smith, Opt. Express 19, 10180 (2011). [CrossRef]
3. B. Ward, C. Robin, and I. Dajani, Opt. Express 20, 11407 (2012). [CrossRef]
4. T. Y. Fan, IEEE J. Sel. Top. Quantum Electron. 11, 567 (2005). [CrossRef]
5. S. Zhou, F. W. Wise, and D. G. Ouzounov, Opt. Lett. 32, 871 (2007). [CrossRef]
6. M. Kienel, M. Müller, A. Klenke, T. Eidam, J. Limpert, and A. Tünnermann, Opt. Lett. 40, 522 (2015). [CrossRef]
7. M. Kienel, M. Müller, A. Klenke, J. Limpert, and A. Tünnermann, Opt. Lett. 41, 3343 (2016). [CrossRef]
8. C. X. Yu, S. J. Augst, S. M. Redmond, K. C. Goldizen, D. V. Murphy, A. Sanchez, and T. Y. Fan, Opt. Lett. 36, 2686 (2011). [CrossRef]
9. G. D. Goodno, C.-C. Shih, and J. E. Rothenberg, Opt. Express 18, 25403 (2010). [CrossRef]
10. A. Klenke, E. Seise, J. Limpert, and A. Tünnermann, Opt. Express 19, 25379 (2011). [CrossRef]
11. L. P. Ramirez, M. Hanna, G. Bouwmans, H. El Hamzaoui, M. Bouazaoui, D. Labat, K. Delplace, J. Pouysegur, F. Guichard, P. Rigaud, V. Kermène, A. Desfarges-Berthelemot, A. Barthélémy, F. Prévost, L. Lombard, Y. Zaouter, F. Druon, and P. Georges, Opt. Express 23, 5406 (2015). [CrossRef]
12. F. Prevost, L. Lombard, J. Primot, L. P. Ramirez, L. Bigot, G. Bouwmans, and M. Hanna, Opt. Express 25, 9528 (2017). [CrossRef]
13. A. Shirakawa, in Conference on Lasers and Electro-Optics (CLEO) (Optical Society of America, 2014), paper STh4N.1.
14. P. Rigaud, V. Kermene, G. Bouwmans, L. Bigot, A. Desfarges-Berthelemot, D. Labat, A. Le Rouge, T. Mansuryan, and A. Barthélémy, Opt. Express 21, 13555 (2013). [CrossRef]
15. P. Rigaud, V. Kermene, G. Bouwmans, L. Bigot, A. Desfarges-Berthelemot, and A. Barthélémy, Opt. Express 23, 27448 (2015). [CrossRef]
16. L. Shimshi, A. A. Ishaaya, V. Ekhouse, N. Davidson, and A. A. Friesem, Appl. Phys. Lett. 88, 41103 (2006). [CrossRef]
17. H.-J. Otto, A. Klenke, C. Jauregui, F. Stutzki, J. Limpert, and A. Tünnermann, Opt. Lett. 39, 2680 (2014). [CrossRef]
18. H. Carstens, N. Lilienfein, S. Holzberger, C. Jocher, T. Eidam, J. Limpert, A. Tünnermann, J. Weitenberg, D. C. Yost, A. Alghamdi, Z. Alahmed, A. Azzeer, A. Apolonski, E. Fill, F. Krausz, and I. Pupeza, Opt. Lett. 39, 2595 (2014). [CrossRef]
19. A. Klenke, M. Müller, H. Stark, A. Tünnermann, and J. Limpert, Opt. Express (2018) (submitted).