Multimode cavity lasers, such as multimode fiber lasers, are attractive for the opportunity they offer to generate high energy pulsed lasers, provided that one can achieve spatiotemporal mode-locking. However, it can be complicated to control the laser properties as 1) spatiotemporal dispersion, nonlinearity, gain and loss can nonlinearly interact, and 2) dispersion and mode coupling in such a system are difficult to predict or control. In a typical wavefront shaping experiment, one modulates the output of a laser beam, which can come at the cost of a significant energy loss, and only allows to control the spatial profile of the beam. In this paper, the authors use a spatial light modulator, but inside the laser cavity to modulate its boundary conditions. Using a genetic algorithm, they are able to efficiently control the laser properties, namely the output power, the output mode profile, the optical spectrum, and mode-locking.
The idea of controlling the property of a complex laser using wavefront shaping has been investigated in the context of random lasers, where one can control the pump profile to control the lasing modes (see for example our highlight from 2013 on Control of random lasing by wavefront shaping of the pump). The approach of the paper is different, using a complex multimode laser cavity and inserting a spatial modulator inside the cavity itself, one can control at will the boundary conditions of the cavity. While it seems difficult to predict a priori the effect of a given modulation, one can iteratively change the boundary conditions to enhance or enforce a chosen behavior of the laser.
The setup used is presented in Fig. 1. It consists of a ring cavity, with a double-cladding ytterbium-doped fiber as gain medium fused to a multimode graded-index fiber supporting hundreds of transverse modes. A spatial light modulator modulating the phase of the reflected beam is inserted inside the cavity and a portion of the light is extracted with a beam splitter to monitor its properties.
Figure 1. Image from [X. Wei et al., Light Sci. Appl., 9 (2020)]
Using a genetic algorithm, the phase mask displayed on the modulator is sequentially modified to optimized a chose property of the output laser beam. The first optimization consists in maximizing the total output power, which is enhanced by a factor of 2. The results are shown in Fig 2.
Figure 2. Enhancement of the output lasing power. Image from [X. Wei et al., Light Sci. Appl., 9 (2020)]
One issue with multimode laser systems, particularly in multimode fibers, concerns the quality of the beam profile, which is typically larger than for a single transverse mode system and can even be quite disordered. In a second experiment, the authors chose as a target function the intensity in a small spatial area at the output of the laser. The results are displayed in Fig 3. They show a dramatic improvement of the laser beam quality that has a close to a gaussian beam profile.
Figure 3. Optimization of the spatial beam quality. Image from [X. Wei et al., Light Sci. Appl., 9 (2020)]
In a third experiment, the authors use the system to control the spectrum of the laser output. In Fig 4a and b, is presented an example of such an optimization where the power at one particular frequency is chosen as the function to maximize. They also demonstrate that they can suppress a chosen wavelength. In Fig. 4c, are represented the results of various wavelength optimizations, showing that they can cover the entire operational bandwidth of ytterbium-doped gain media, typically from 1030 to 1070 nm.
Figure 4. Wavelength tunability. (a) Spectrum along the optimization process and (b) the resulting final spectrum. (c) results for various optimizations for target wavelengths inside the bandwidth of the ytterbium-doped gain medium. Image from [X. Wei et al., Light Sci. Appl., 9 (2020)]
Finally, the authors use their system to find a mode-locking condition. First, they set the system out of the mode-locking condition and use the genetic algorithm to optimize the intensity of the Fourier component corresponding to the repetition rate (actually, the 10th harmonic, but it does not matter much). A typical temporal output signal before and after optimization is shown in Fig 5.