Deep neural networks (DNNs) are used to classify and reconstruct the input images from the intensity of the speckle patterns that result after the inputs are propagated through multimode fiber (MMF). We were able to demonstrate this result for fibers up to 1 km long by training the DNNs with a database of 16,000 handwritten digits. Better recognition accuracy was obtained when the DNNs were trained to first reconstruct the input and then classify based on the recovered image. We observed remarkable robustness against environmental instabilities and tolerance to deviations of the input pattern from the patterns with which the DNN was originally trained.

Spatiotemporal mode-locked fiber lasers are a type of ultrafast lasers for which the longitudinal and transverse modes of the multimode fiber cavities are locked via nonlinear interaction in the cavity. Here we report the experimental realization of a spatiotemporally mode-locked fiber laser with self-similar pulse evolution. The multimode fiber oscillator generates parabolic pulses at 1030 nm with 90 mW average power, a near-Gaussian beam quality ($M^2\le 1.4$), and 2.3 ps (192 fs externally dechirped) pulse duration. Numerical simulations confirm the experimental observations of self-similar pulse propagation. These results will enable further investigation of nonlinear dynamics in spatiotemporal mode-locked fiber lasers.

Spatiotemporal nonlinear interactions in multimode fibers are of interest for beam shaping and frequency conversion by exploiting the nonlinear interaction of different pump modes from quasi-continuous wave to ultrashort pulses centered around visible to infrared pump wavelengths. The nonlinear effects in multi-mode fibers depend strongly on the excitation condition; however, relatively little work has been reported on this subject. Here, we present a machine learning approach to learn and control nonlinear frequency conversion inside multimode fibers. We experimentally show that the spectrum of the light at the output of the fiber can be tailored by a trained deep neural network. The network was trained with experimental data to learn the relation between the input spatial beam profile of the pump pulse and the spectrum of the light at the output of the multimode fiber. For a user-defined target spectrum, the network computes the spatial beam profile to be applied at the input of the fiber. The physical processes involved in the creation of new optical frequencies are cascaded stimulated Raman scattering as well as supercontinuum generation. We show experimentally that these processes are very sensitive to the spatial shape of the excitation and that a deep neural network is able to learn the relation between the spatial excitation at the input and the spectrum at its output. The method is limited to spectral shapes within the achievable nonlinear effects supported by the test setup, but the demonstrated method can be implemented to learn and control other spatiotemporal nonlinear effects.

We demonstrate the first all-fiber multimode spatiotemporally mode-locked laser. The oscillator generates dissipative soliton pulses at 1036 nm with 12 mW average power, 6.24 ps duration, and 24.3 MHz repetition rate. The reported pulse energy (0.5 nJ) represents ∼4 times improvement over the previously reported single-mode all-normal dispersion mode-locked lasers with multimode interference-based filtering. Numerical simulations are performed to investigate the cavity and spatiotemporal mode-locking dynamics. The all-fiber oscillator we present shows promise for practical use since it can be fabricated simply.

We demonstrate the ability of a multimode fiber probe to provide two-photon fluorescence (TPF) imaging feedback that guides the femtosecond laser ablation (FLA) in biological samples for highly selective modifications. We implement the system through the propagation of high power femtosecond pulses through a graded-index (GRIN) multimode fiber and we investigate the limitations posed by the high laser peak intensities required for laser ablation. We demonstrate that the GRIN fiber probe can deliver laser intensities up to 1.5x10¹³ W/cm², sufficient for the ablation of a wide range of materials, including biological samples. Wavefront shaping through an ultrathin probe of around 400 μm in diameter is used for diffraction limited focusing and digital scanning of the focus spot. Selective FLA of cochlear hair cells is performed based on the TPF images obtained through the same multimode fiber probe.

Deep neural networks (DNNs) are employed to recover information after its propagation through a multimode fiber (MMF) in the presence of wavelength drift. The intensity distribution of the speckle patterns generated at the output of an MMF when an input wavefront propagates along its length is highly sensitive to wavelength changes. We use a tunable laser to implement a wavelength drift with a controlled bandwidth, aiming to estimate the DNN’s performance in different cases and identify the limitations. We find that when the DNNs are trained with a dataset which includes the noise induced by wavelength changes, successful classification of a speckle pattern can be performed even for a large wavelength bandwidth drift. A single training step is found to be sufficient for high classification accuracy, removing the need for time-consuming recalibration at each wavelength.