Classification of grazing-incidence small-angle X-ray scattering patterns by convolutional neural network

Abstract

<jats:p>Grazing-incidence small-angle X-ray scattering (GISAXS) patterns have multiple superimposed contributions from the shape of the nanoscale structure, the coupling between the particles, the partial pair correlation, and the layer geometry. Therefore, it is not easy to identify the model manually from the huge amounts of combinations. The convolutional neural network (CNN), which is one of the artificial neural networks, can find regularities to classify patterns from large amounts of combinations. CNN was applied to classify GISAXS patterns, focusing on the shape of the nanoparticles. The network found regularities from the GISAXS patterns and showed a success rate of about 90% for the classification. This method can efficiently classify a large amount of experimental GISAXS patterns according to a set of model shapes and their combinations.</jats:p>

GISAXS: A versatile tool to assess structure and self‐assembly kinetics in block copolymer thin films

“…With powerful computer simulations of BCP dynamics such as DPD 81,100,101 and highly flexible software for scattering simulations such as HipGISAXS 42 and Bor-nAgain 43 to calculate GISAXS patterns directly based on DPD output 82 or to fit GISAXS data in delicate detail, a much more detailed understanding of BCP rearrangement processes is evolving. In addition machinelearning approaches are in development to optimize probing the parameter space in autonomous experiments 89 and for automatic identification of scattering features [102][103][104] which will be of significant importance as more complex BCPs with new functionalities are investigated, such as in the study of a pentablock quaterpolymer thin film by Jung et al 105 Not much exploited for polymer science so far is the combination of GISAXS and anomalous scattering. This is due to the fact that there are no relevant absorption edges in the hard X-ray regime (5-20 keV), with the exception of the bromine edge.…”

Amortized Bayesian Inference of GISAXS Data with Normalizing Flows

“…Afterwards, we approximate posterior distribution p(y|c) over object parameters with normalizing flows (D), yielding fast inference that allows accelerating feedback during experiments (C). et al [5] and Liu et al [6] use convolutional neural networks for the one-step classification of experimental images. At the same time, Van Herck et al [7] infer the rotation distribution of nanoparticle arrangements.…”

Band-gap assessment from X-ray powder diffraction using artificial intelligence

“…Liu et al (2022) developed BraggNN, which is a CNN designed to determine peak positions more quickly and accurately than conventional pseudo-Voigt peak fittings in high-energy diffraction microscopy. Ikemoto et al (2020) developed a CNN to classify grazing-incidence small-angle X-ray scattering patterns in terms of the shape of nanoparticles with a 90% success rate. Banko et al (2021) used variational auto-encoders to expose latent information of simulated and experimental X-ray diffraction patterns.…”