Retrieving the quantitative chemical information at nanoscale from SEM EDX measurements by Machine Learning

📝 Abstract

The quantitative composition of metal alloy nanowires on InSb(001) semiconductor surface and gold nanostructures on germanium surface is determined by blind source separation (BSS) machine learning (ML) method using non negative matrix factorization (NMF) from energy dispersive X-ray spectroscopy (EDX) spectrum image maps measured in a scanning electron microscope (SEM). The BSS method blindly decomposes the collected EDX spectrum image into three source components, which correspond directly to the X-ray signals coming from the supported metal nanostructures, bulk semiconductor signal and carbon background. The recovered quantitative composition is validated by detailed Monte Carlo simulations and is confirmed by separate cross-sectional TEM EDX measurements of the nanostructures. This shows that SEM EDX measurements together with machine learning blind source separation processing could be successfully used for the nanostructures quantitative chemical composition determination.

💡 Analysis

The quantitative composition of metal alloy nanowires on InSb(001) semiconductor surface and gold nanostructures on germanium surface is determined by blind source separation (BSS) machine learning (ML) method using non negative matrix factorization (NMF) from energy dispersive X-ray spectroscopy (EDX) spectrum image maps measured in a scanning electron microscope (SEM). The BSS method blindly decomposes the collected EDX spectrum image into three source components, which correspond directly to the X-ray signals coming from the supported metal nanostructures, bulk semiconductor signal and carbon background. The recovered quantitative composition is validated by detailed Monte Carlo simulations and is confirmed by separate cross-sectional TEM EDX measurements of the nanostructures. This shows that SEM EDX measurements together with machine learning blind source separation processing could be successfully used for the nanostructures quantitative chemical composition determination.

📄 Content

Retrieving the quantitative chemical information at nanoscale from SEM EDX measurements by Machine Learning B.R. Jany*, A. Janas, F. Krok Marian Smoluchowski Institute of Physics Jagiellonian University, Lojasiewicza 11, 30-348 Krakow, Poland Abstract The quantitative composition of metal alloy nanowires on InSb(001) semiconductor surface and gold nanostructures on germanium surface is determined by blind source separation (BSS) machine learning (ML) method using non negative matrix factorization (NMF) from energy dispersive X-ray spectroscopy (EDX) spectrum image maps measured in a scanning electron microscope (SEM). The BSS method blindly decomposes the collected EDX spectrum image into three source components, which correspond directly to the X-ray signals coming from the supported metal nanostructures, bulk semiconductor signal and carbon background. The recovered quantitative composition is validated by detailed Monte Carlo simulations and is confirmed by separate cross-sectional TEM EDX measurements of the nanostructures. This shows that SEM EDX measurements together with machine learning blind source separation processing could be successfully used for the nanostructures quantitative chemical composition determination. Keywords: SEM, EDX, Machine Learning, BSS * corresponding author e-mail: benedykt.jany@uj.edu.pl 1/14 The Scanning Electron Microscope (SEM) with a Field Emitter Gun (FEG) electron source became a popular tool for the nanoscience[1]. It can deliver the information at the nanoscale on the sample topography by collecting the secondary electrons (SE) and relative sample composition by the backscattered electrons (BSE), which emission is related to the mean atomic number. It is also very common that a SEM is equipped with energy dispersive X-ray spectroscopy (EDX) system. Nowadays such a EDX system usually consist of high efficient Silicon Drift Detector (SDD) capable of recording high count rates. The spatial resolution in the SEM EDX mapping is related to the interaction volume of primary electron beam and consequently X-ray generation volume. Careful optimization of the X- ray depth distribution and spatial radial distribution by adjusting the electron beam energy and size (the beam current) leads to the acquisition of high spatial resolution X-ray maps at nanoscale[2; 3]. However the quantification of the recorded SEM EDX from nanostructures is challenging due to the mixing of the signals from different depths of the sample, resulted from X-ray generation depth. This is very similar as for the TEM EDX for the heterogeneous volumes, where there is a spatial overlap of the different phases in the beam path[4] . For the separation of the components from the mixture Machine Learning (ML), methods such as blind source separation (BSS) using independent component analysis (ICA)[5] and non negative matrix factorization[6] (NMF), are successfully applied. As shown it work for the TEM EELS measurements[7; 8] and recently for TEM EDX measurements of multicomponent signal unmixing of nanoheterostructures[4; 9]. The idea of BSS method is to statistically decompose the mixed signal into separate sources, without any external information. These methods are widely used also in the different fields of science[10-12]. Here we apply the BSS decomposition using NMF to SEM EDX spectrum image maps of metal alloy nanowires grown on AIIIBV semiconductor surface. The number of decomposition components is provided by principal component analysis (PCA). The quantitative composition of nanowires is recovered, the results of the quantification are additionally verified by detailed Monte Carlo simulations. The nanowires composition is confirmed by separate cross-sectional TEM EDX measurements.
The AuIn2 metal alloy nanowires on InSb(001) (AIIIBV semiconductor) surface were prepared by molecular beam epitaxy (MBE) deposition of 2 mono-layers (ML) of gold on atomically clean reconstructed InSb(001) surface at temperature of 330C in ultra high vacuum conditions (UHV). Such a perpetration conditions results in the formation of AuIn2 metal alloy nanowires on the surface in the process of thermally induced self-assembly[13] . The AIIIBV semiconductors since their unique 2/14 properties are seriously considered for future electronic devices especially that the technology to integrate the AIIIBV at the nanoscale with silicone [14; 15] was developed. The gold-rich nanostructures on AIIIBV semiconductors are widely used as a catalyst to grow standing arrays of vertically aligned AIII-BV nanowires[16; 17] for many applications as for example efficient water reduction[18] or nano light emitting diodes (LED) with high brightness[19] . They also have a potential usage as nanoelectrodes and ohmic contacts[20] . Similarly the Au hcp nanostructures, gold of rare and unique hexagonal structure, were prepared on Ge(001) surface as we recently shown[21; 22] . These have a p

This content is AI-processed based on ArXiv data.