In this work we present the results obtained with an artificial neural network (ANN) which we trained to predict the expected output of high-order harmonic generation (HHG) process, while exploring a multi-dimensional parameter space. We argue on the utility and efficiency of the ANN model and demonstrate its ability to predict the outcome of HHG simulations. In this case study we present the results for a loose focusing HHG beamline, where the changing parameters are: the laser pulse energy, gas pressure, gas cell position relative to focus and medium length. The physical quantity which we predict here using ANN is directly related to the total harmonic yield in a specified spectral domain (20–40 eV). We discuss the versatility and adaptability of the presented method.
We report on the reconstruction of ultrashort laser pulses from computer-simulated and experimental second harmonic generation-frequency resolved optical gating (SHG-FROG) spectrograms. In order to retrieve the spectral amplitude and phase we use a convolutional neural network trained on simulated SHG-FROG spectrograms and the corresponding spectral-domain fields employed as labels for the network, which is a complex field encompassing the full information about the amplitude and phase. Our results show excellent retrieval capabilities of the neural network in case of the simulated pulses. Although trained only on computer generated data, the method shows promising results regarding experimentally measured pulses.
The propagation of femtosecond pulses in guided structures is a matter of both fundamental and practical interest in nonlinear optics. In particular, hollow-core waveguides (HCWs) filled with a gas medium are fabricated and used as devices for the generation of attosecond pulses from high-order harmonics. In this process, the configuration of the laser field (intensity and phase) inside the waveguide is of crucial importance for enhancing the (well-known, low) efficiency of high-order harmonic generation (HHG). Here, we present numerical calculations which demonstrate the main features of the propagation process in fabricated HCWs. We consider a variety of experimental parameters like gas pressure, waveguide size, laser wavelength, and pulse energy and duration. In particular, the beam profile at the fiber input is found to be a sensitive parameter which influences the whole evolution of the laser field along the propagation. Our model is based on a split-step method modified to account for propagation in ionized media and is validated against experimental and theoretical data from the literature. Our results contribute to the description of the main features of beam propagation in HCWs and provide guiding directions for designing efficient configurations for HHG.
We numerically investigate generation of harmonics in the water window region (down to 2.8 nm) by 2 μm femtosecond pulses propagating in hollow core waveguides filled with high pressure He. Numerical calculations are based on a three dimensional macroscopic model, which solves the pulse propagation by a split-step method, uses the strong field approximation to evaluate the single atom response, and integrates it coherently to obtain the harmonic field. Two configurations for the waveguides are considered: the standard one with a constant diameter of 70 μm and a conical one with a decreasing diameter from 70 to 50 μm. We demonstrate that harmonic field enhancement can be obtained in spectral domains of great practical interest, from 2.8 to 20 nm, and identify quasi-phase matching induced by multimode beating as the mechanism responsible for this enhancement.
Ultrafast laser pulse characterization is extremely important in applications like high-harmonic spectroscopy, ultrafast imaging or time resolved measurements. The most practical, reliable and well developed method for ultrashort pulse characterization is frequency resolved optical gating (FROG). This method uses different algorithms to retrieve the ultrashort pulse amplitude and phase from a FROG spectrogram. Convolutional neural networks can be an alternative to the classically used algorithms due to their high computational speed and superior performance in processing images. In this article, two convolutional neural networks are developed in order to accurately predict the real and imaginary part of the spectral electric field based on FROG images that do not need a pre-processing step. The data on which the neural network is trained, validated and tested is generated using parameters of the chirped pulse amplification system at the ELI-NP facility, which are chosen to be in an experimentally viable range and are used to calculate the phase of the laser pulse. The low errors between the original and retrieved spectral electric field confirm the excellent prediction capabilities of these two fast neural networks.