Due to its excellent beam quality, compactness, and flexibility, the high-power fiber laser has gained increasing attention in a wide range of applications, such as laser processing, scientific research, and laser additive manufacturing^[1-3]. Among these applications, laser additive manufacturing is particularly important for the fiber laser, as it has been widely adopted in the automotive industry due to its high manufacturing efficiency and material utilization^[4,5]. Recently, the output power of a fiber laser has been scaled up to over 10 kW^[6–10], providing a viable laser source for the laser additive manufacturing system. However, practical applications may encounter issues due to the lack of multidimensional regulation of the laser^[5]. Generally, a Gaussian beam is mostly applied as a laser source in a laser additive manufacturing system. Since the Gaussian beam has a particular energy distribution, the laser energy mostly focuses on the center, which may affect the manufacturing efficiency of complex parts^[11]. Moreover, a high thermal gradient will be caused due to its energy distribution, which impacts the processing quality^[11–13]. To address the above problems, researchers have been inspired to explore beam-shaping techniques to obtain high-power beam-shaped lasers. By controlling the phase wavefront of the input beam, the output beam can be shaped to have the presupposed energy distribution and propagation characteristics^[14,15]. Thus, the beam-shaping technique can make up for the deficiencies of the Gaussian beam and significantly improve manufacturing efficiency^[16]. Recently, the beam-shaping technique has been developed rapidly. There are many practical methods for beam shaping. As a classical method, the aspheric lens group shaping method can adjust the mapping function between the output and the input rays to obtain the desired energy distribution^[17–19]. This method can be applied in a high-power laser system due to its high power damage threshold. Similarly, the microlens array^[20] and diffractive optical-element-shaping techniques^[21,22] can shape lasers with arbitrary energy distributions. However, these methods generally refer to static beam-shaping techniques, making it difficult to achieve dynamic beam shaping due to physical limitations^[5]. In comparison, dynamic beam-shaping techniques using liquid crystal spatial light modulators can flexibly produce lasers with various energy distributions^[23–26], significantly improving processing efficiency and meeting the demands of structurally complex parts. The ability to switch energy distributions in real time by adjusting the driving voltage of the liquid crystal has demonstrated excellent performance in laser additive manufacturing. However, the efficiency of this shaping technique is easily influenced by the input beam^[5]. Additionally, liquid crystals typically have a low power damage threshold, which poses a challenge for their use in high-power laser additive manufacturing systems. Therefore, obtaining a high-power dynamic beam-shaping laser source is essential and urgent for promoting further practical applications.

As a viable method for obtaining high-power lasers, the coherent beam combining (CBC)-technique can increase the output power while maintaining excellent beam quality^[27–37], thereby providing a laser source for high-power dynamic beam shaping. On the one hand, this technique has undergone significant advancements in recent years, with combined fiber lasers achieving an output power of over 10 kW^[38–40], thus offering a mature high-power laser source. On the other hand, by flexibly controlling the individual phase of each laser beamlet, it becomes possible to manipulate the combined interference pattern in the far field^[16,41]. This enables the generation of various types of beams with different intensity distributions^[42–48], which is advantageous for dynamic beam shaping. However, there are significant challenges associated with achieving high-power dynamic beam shaping. First, current high-power CBC systems typically employ large-aperture optical devices that are bulky and pose difficulties in practical implementation^[49]. Second, phase locking in a high-power CBC system is challenging. In particular, flexibly controlling the individual phase of each laser beamlet to achieve the desired value is difficult. Fortunately, internal phase-sensing techniques can aid in constructing compact CBC systems^[50]. The flexible phase control can be achieved by employing liquid crystal phase devices^[50,51], which may allow real-time switching of energy distribution within the combined beam.

In this paper, we constructed an internal phase-sensing tiled-aperture CBC system with seven beam elements for dynamic beam shaping. This system could be regarded as a digital laser, where each laser beam operated as an individual laser pixel. The amplitude and phase of each pixel can be independently adjusted in real time. In the experiment, the beam-shaping system was operated in three states: triangular, pentagonal, and hexagonal laser arrays were modulated to emit into free space, with each laser pixel modulated by a different piston phase of $n\pi$ (where $n$ is an integer). We presented various beam-shaping patterns based on the system with output power scaling over 1 kW. The results demonstrated that our compact dynamic beam-shaped laser exhibited excellent performance in both dynamic beam-shaping and power-scaling capability. This work has potential applications in various fields related to beam shaping.

In a tiled-aperture CBC system, each laser beamlet can be treated as an individual element, allowing for independent adjustment of the amplitude, phase, and output power of each pixel according to specific requirements. Consequently, the energy distribution of the combined beam in the far field can be shaped based on the principles of interference^[16,41]. For instance, assume there are $N$ beamlets in a laser array. Figure 1 illustrates their spatial distribution at the source plane with subbeam centers separated by $R$ and a beam waist of $w_0$. Additionally, we assume that all beamlets have identical polarization states and amplitudes denoted as $A$. The complex amplitude of the $i$th beamlet at its coordinate $(xi,yi)$ at the source plane can then be expressed as $$E_i(x,y,0)=A\exp [-\frac{{(x-x_i)}^2+{(y-y_i)}^2}{w_0^2}]T_i(x,y,0)\exp (i{\varphi }_i),$$(1)where ${\varphi }_i$ represents the piston phase and $T_i(x,y,0)$ represents the truncation function, which can be further expanded as $$T_i(x,y,0)=\sum _{t=1}^p{Q}_t\exp \{-\frac{K_t}{w_0^2}[{(x-x_i)}^2+{(y-y_i)}^2]\},$$(2)where $Q_t$ and $K_t$ represent its expansion coefficient and corresponding order, respectively. Thus, the complex amplitude of the laser array at the source plane can be expressed as $$E_{\text{array}}(x,y,0)=\sum _{i=1}^{N}A\exp [-\frac{{(x-x_i)}^2+{(y-y_i)}^2}{w_0^2}]T_i(x,y,0)\exp (i{\varphi }_i).$$(3)

According to the principle of the Fraunhofer diffraction integral, the diffractive optical field of the laser array in the far field is obtained by taking the Fourier transform of the near-field optical field^[43]. In practice, we can achieve a combined beam in the far field by focusing the laser array using a lens with a focal length of $f$. Therefore, expressing it in polar coordinates, we can represent the complex amplitude of the combined optical field as follows: $$U_{\text{array}}(\rho ,\theta )=\frac{k}{i2\pi f}{\int }_0^{\infty }{\int }_0^{2\pi }{E}_{\text{array}}(r,\phi )\exp [(-\frac{ik\rho }{f})r\cos (\theta -\phi )]r\mathrm{d}r\mathrm{d}\phi .$$(4)

According to Eq. (4), the intensity distribution of the combined beam in the far field can be calculated. Similarly, we assume that $N$ is equal to 19, and the laser array is arranged in a hexagonal pattern. The laser wavelength is 1064 nm, the beam waist is 38 mm, and the truncated aperture of the laser array is 0.9 times the beam waist. Additionally, an equivalent far-field pattern is formed by focusing through a lens with a focal length of 2 m. The simulated results are presented in Fig. 2. Figures 2(a1)–2(d1) represent the piston phase distribution of each beamlet at the source plane, respectively. The blue color represents the phase of 0, while the dark red represents the phase of $\pi$. Correspondingly, Figs. 2(a2)–2(d2) depict the intensity distributions of these combined beams in the far field, respectively. It can be observed from these simulation results that different piston phase modulations effectively alter patterns within their respective far-field intensity distributions.

Based on the above simulation results, it is evident that effective switching of the intensity distribution in the far field can be achieved by modulating different beamlets with a $\pi$ phase individually. Real-time phase control of each beamlet can enable dynamic beam shaping. However, practical implementation faces technical challenges. First, in high-power CBC systems, random noise causes fluctuations in the piston phase of the laser array, making it difficult to achieve reliable phase locking. Second, most current control methods focus on maximizing power in the bucket (PIB) of the combined beam in the far field^[50], resulting in a single-phase state for the piston phase distribution and limiting the flexible formation of different piston phase distributions^[52]. Currently, there are few reports on experimental demonstrations of high-power dynamic beam shaping. Addressing this challenge, the internal phase-locking technology allows for flexible control of the piston phase through a two-step process. First, by maximizing the PIB value of the sampled laser array, phase noise is compensated to achieve phase locking of the laser array. Second, the liquid crystal phase devices are controlled to shift a specific piston phase value that compensates for the external phase errors of the emitted laser array and achieves precise beam shaping.

An experiment was conducted to demonstrate dynamic beam shaping, as illustrated in Fig. 3. The seed laser (SL) is a linearly polarized single-frequency ytterbium-doped fiber laser at 1064 nm and a linewidth of 20 kHz. Initially, the SL’s 30 mW power is coupled into an electro-optic modulator (EOM) to broaden the spectral linewidth to approximately 10 GHz. Subsequently, the EOM output is directed through a fiber pre-amplifier (PA) and splitter (FS), which can generate seven laser channels. Each channel then passes through a ${\mathrm{LiNbO}}_3$ phase modulator (PM) and electric variable delay line (VDL), where the PMs are used for phase noise locking and the VDLs are employed for compensating for the optical path differences among different beamlets. Finally, each laser channel from the VDL enters a fiber amplifier (FA) and collimator (CO). Each channel’s output power was scaled to 160 W. It should be noted that the output power of the FAs was limited by the pump power, and the beam quality would degrade after the power was further scaled. The CO array collimates the laser beams into a regular hexagonal array with adjacent beamlet centers spaced by 40 mm; each collimated beamlet has a diameter of 10 mm.

Generally, in a CBC system, the phase control signal is typically obtained by sampling the emitted laser array, resulting in a bulky system. Therefore, to achieve a compact CBC system, the phase control is implemented based on an internal phase-sensing technique^[50,51]. Here, the collimated laser array is divided into two parts using a beam splitter (BS) array with an approximate reflectivity of 99.9%. Specifically, the smaller portion of the laser passes through a liquid crystal (LC) array and is then focused by a lens (L1) with a focal length of 1 m. Finally, the focused combined beam is detected by a photodetector (PD), which is equipped with a pinhole aperture measuring 100 µm. The purpose of this pinhole aperture is to truncate the PIB of the focused combined beams, which can provide the feedback signal to the field-programmable gate array (FPGA) controller for actively locking the phase. Meanwhile, most of the laser from BS reflects into the laser-emitting system, where each laser beam’s diameter expands to 38 mm using a beam expander (BE). Due to the distance of the adjacent beamlet center being 40 mm, we achieve an impressive fill factor of 95% with 38 mm/40 mm^[53]. The lasers from this emitting system are observed for their beam-shaping performance: First, a beam splitter high reflectivity mirror (HRM) with a high splitting ratio of 99.9/0.1 directs most power toward the power collector (PC) for combined power measurement purposes. Second, after passing through the HRM, these lasers are focused by another lens (L2) onto a charge-coupled device (CCD), which resides at the focal plane and captures images representing various states of the beam shaping.

The phase-control performance is illustrated in Fig. 4. When the FPGA controller was turned off (open loop), it can be observed that the feedback signal exhibited significantly and random fluctuations over time due to thermal and environmental vibration noise, as depicted in Fig. 4(a). Based on our calculations, the normalized average value was merely 0.23. Subsequently, the FPGA controller (closed loop) was activated and the stochastic parallel gradient descent (SPGD) algorithm was implemented for phase locking based on the feedback signal. The main frequency of the FPGA controller is 50 MHz, while the SPGD algorithm iteration frequency is $\sim 1\mathrm{MHz}$. The mathematical principle of the SPGD algorithm can be found in Ref. [54]. Stable locking was achieved with nearly maximum intensity of the feedback signal. The normalized average value was 0.94, while the residual phase error can be calculated to be $\lambda /32$^[55]. This indicates effective compensation of phase noise and enables emission of the laser array with locked phases.

Subsequently, we turned off four of the beams to test a three-beam CBC system, as illustrated in Fig. 5(a1). Furthermore, utilizing the LCs to shift the emitting-laser array’s phases in real time can dynamically shape the intensity distribution of the far-field’s interference pattern^[44]. The shifted phase structures of the emitting-laser array at the source plane are illustrated in Figs. 5(a1)–5(d1), respectively. The corresponding far-field intensity distributions of the combined beams detected by the CCD are presented in Figs. 5(a3)–5(d3), while the simulated results are shown in Figs. 5(a2)–5(d2). The experimental results were in very good agreement with the simulated results, indicating an excellent performance of this system. In this beam-shaping system, the LCs were used to set the phase shift on each emitted laser beam indirectly. In detail, in the closed loop, the phase differences of the detected laser beams that passed through the LCs were always compensated to be zero. When we impose a phase shift on an LC, the feedback loop will automatically compensate for this phase. As a result, the phase of the emitted laser beam will be stabilized as the conjugate value of the phase provided by the LC. Thus, the far-field’s interference pattern of combined beams can be customized based on phase shifts modulation^[44]. This capability is advantageous for various applications requiring customized manipulation of laser energy distribution.

Later, we turned off two of the beams to test a five-beam CBC system, as depicted in Fig. 6(a). Similarly, the LCs were manipulated to induce phase shifting. The phase structures of the emitted laser beams are illustrated in Figs. 6(a1)–6(f1), while the corresponding experimental intensity distributions of the far-field’s interference pattern can be observed in Figs. 6(a3)–6(f3). More intricate optical fields could be generated due to an increased number of laser channels. Notably, there existed a fascinating phenomenon wherein high-brightness energy points could be flexibly varied, as demonstrated in Figs. 6(d3)–6(f3). Additionally, the quantity of Gaussian bright spots within the combined optical field could also be customized according to specific requirements, as shown in Figs. 6(a3)–6(c3). Undoubtedly, this holds significant application prospects for individuals seeking complex device processing without necessitating alterations to their laser source, and it also substantially reduces system complexity and costs.

According to the aforementioned experimental results, it can be concluded that this system exhibited excellent performance in dynamic beam shaping. In the above experiment, the output power of each laser beam was about 1 W. However, to meet the requirements of practical applications, it is necessary to present the power-scaling capability of the system. Therefore, we turned on seven beams for testing a seven-beam CBC experiment. The emitted power of each laser beam was scaled to be about 160 W, with a total combined power of up to 1135 W. The corresponding results are illustrated in Fig. 7. Figures 7(a1)–7(d1) depict the phase structures of the laser array, while Figs. 7(a3)–7(d3) show the intensity distributions of their far-field interference patterns. It can be observed that even under 1.1 kW laser power, the system maintained an exceptional performance for dynamic beam shaping while generating desired intensity distributions. The experimental results were still maintained to a high degree with the simulated results, as shown in Figs. 7(a2)–7(d2). However, the experimental beams were imperfect, which was caused both by the beam misalignment and beam quality degradation^[56]. We should note that the switching frequency of beam shaping was about 200 Hz, which was limited by the LCs^[57].

Based on the experimental results, the dynamic beam-shaping capability of the system has been demonstrated. By controlling both the FAs and LCs, it was possible to customize both the emitted beam arrangement and its phase structure simultaneously. This enabled the generation of desired energy distributions in the far field and facilitated varied efficient energy. Furthermore, the output power of this system can be scaled up to exceed 1 kW, making it suitable for practical applications, such as laser processing of hard and brittle materials^[58–62] and laser ablation^[63], which have been applied with the pulse laser. In future developments, we aim to present more intricate beam-shaping techniques by incorporating additional laser channels, which would allow us to further scale up the output power beyond 10 kW. Additionally, we plan to explore more complex phase structures, which may achieve nonmechanical axial focus tuning^[64].

In summary, an internal phase-sensing tiled-aperture CBC system with seven beam elements was constructed for dynamic beam shaping. This system could be considered a digital laser, where each laser beam functioned as an individual pixel that could have its amplitude and phase adjusted independently in real time. In our experiment, triangular, pentagonal, and hexagonal laser arrays were constructed, and each laser pixel was modulated with a different piston phase of $n\pi$ (where $n$ is an integer). Various beam-shaping patterns were presented based on the system, and the high-brightness energy points could be flexibly varied and customized according to specific requirements. Additionally, the output power was scaled to over 1 kW. These results indicated that our compact dynamic beam-shaped laser had excellent performance in both dynamic beam-shaping and power-scaling capability. This work can benefit various applications related to beam shaping.