Machine learning for laser-induced electron diffraction imaging of molecular structures

Abstract

Ultrafast diffraction imaging is a powerful tool to retrieve the geometric structure of gas-phase molecules with combined picometre spatial and attosecond temporal resolution. However, structural retrieval becomes progressively difficult with increasing structural complexity, given that a global extremum must be found in a multi-dimensional solution space. Worse, pre-calculating many thousands of molecular configurations for all orientations becomes simply intractable. As a remedy, here, we propose a machine learning algorithm with a convolutional neural network which can be trained with a limited set of molecular configurations. We demonstrate structural retrieval of a complex and large molecule, Fenchone (C10H16O), from laser-induced electron diffraction (LIED) data without fitting algorithms or ab initio calculations. Retrieval of such a large molecular structure is not possible with other variants of LIED or ultrafast electron diffraction. Combining electron diffraction with machine learning presents new opportunities to image complex and larger molecules in static and time-resolved studies.

Trends in angle-resolved molecular photoelectron spectroscopy

“…fixed-angle broadband laser driven electron scattering (FABLES), 439 or through implementation of more general retrieval methods, 440 as recently demonstrated combining LIED with machine learning to image molecules such as Fenchone. 441…”

“…fixed-angle broadband laser driven electron scattering (FABLES), 439 or through implementation of more general retrieval methods, 440 as recently demonstrated combining LIED with machine learning to image molecules such as Fenchone. 441 If pump-probe femtosecond experiments where LIED acts as a probe of ultrafast dynamics launched by a coherent pump pulse has not been yet fully demonstrated, recent time-resolved results in the picosecond range have been reported based on the analysis of PMDs acquired with a VMI spectrometer. 442,443 LIED and photoelectron holography were investigated simultaneously through the PMDs for SFI of nitrogen monoxide NO, recorded at different delays after launching a rotational wave packet in NO, tracking the coupling between valence-shell electronic and rotational dynamics.…”

Quantum interference and imaging using intense laser fields

“…LIED's extension to the more advantageous mid-infrared (MIR; i.e. λ 2 µm) wavelength range has enabled the direct retrieval of many diatomic and more complex molecular structures [155,118,119,156,8,6,143,48,94,5,17,125,93]. In fact, not only is the deBroglie wavelength λ B of the rescattering electron significantly smaller for a driving laser at λ = 3 µm (λ B ∼ 0.75 Å) than at λ = 0.8 µm (λ B ∼ 2.75 Å).…”

Femtosecond pulse parameter estimation from photoelectron momenta using machine learning

“…Most studies, however, focus on a proof of principle, using only theoretical data. Notable exceptions are, reference [26], where CNNs were used to extract molecular structure parameters from experimental laser-induced electron diffraction images, and reference [27], where deep neural networks were applied to streaking traces for parameter extraction and prediction of uncertainties. Unfortunately, the analytical power of ML-assisted imaging is limited if the laser pulse parameters cannot be accurately measured.…”