On the quantitativeness of EDS STEM

Lugg, Nathan; Kothleitner, Gerald; Shibata, Naoya; Ikuhara, Y.

doi:10.1016/j.ultramic.2014.11.029

“…Moreover, figure 1 shows that the enhancement effect reduces with increasing tilt: by a tilt of 2°, the signal from each atom along the column is much closer to 1, suggesting small tilt as a strategy for robust counting of atoms in a column in elemental crystals. Similar results have been shown by Lugg et al for unit cell averages in SrTiO 3 [3].…”

supporting

confidence: 91%

“…Moreover, figure 1 shows that the enhancement effect reduces with increasing tilt: by a tilt of 2°, the signal from each atom along the column is much closer to 1, suggesting small tilt as a strategy for robust counting of atoms in a column in elemental crystals. Similar results have been shown by Lugg et al for unit cell averages in SrTiO 3 [3].Determining composition in mixed-species columns presents a similar challenge. Focusing specifically on the simplest composition case of a binary system we investigate the effect of ordering on the X-ray signal.…”

supporting

confidence: 85%

Probing the Effects of Electron Channelling on EDX Quantification

MacArthur

¹

,

Brown

²

,

Findlay

³

et al. 2017

View full text Add to dashboard Cite

A myriad of advances in microscope stability, aberration correction and improved detectors now make it routinely possible to achieve atomic resolution energy dispersive X-ray (EDX) mapping [1], showing atomic-scale qualitative detail as to where the elements reside within a sample. Whenever we analyse crystals down a low-order zone axis we are exploiting electron channelling to provide us with atomic resolution. The aligned columns act like miniature lenses, providing a focusing effect on the electron beam. This is beneficial because it causes our convergent beam to stay more focused onto one atomic column, providing increased signal from an aligned column of atoms than we would expect just by summing the intensity of each of the component atoms [2]. However, the focusing effect further means the contribution from atoms at different depths differs, a phenomenon often referred to as the 'TopBottom Effect', complicating the quantitative interpretation. Figure 1 plots the X-ray signal from a single column in a single element crystal, as a function of both the column thickness and of sample tilt away from the on-axis condition, normalized by the X-ray signal from a single atom. Therefore, a value of 1 indicates a linear signal and an increase of the fractional Xray intensities with increasing crystal thickness show that the 'lighting-up', i.e. the enhancement of the signal resulting from the focusing/channeling effect observed in HAADF imaging, also occurs in the EDX case. Because this effect is dependent on atom type, it will also have an effect on quantification. Moreover, figure 1 shows that the enhancement effect reduces with increasing tilt: by a tilt of 2°, the signal from each atom along the column is much closer to 1, suggesting small tilt as a strategy for robust counting of atoms in a column in elemental crystals. Similar results have been shown by Lugg et al. for unit cell averages in SrTiO 3 [3].Determining composition in mixed-species columns presents a similar challenge. Focusing specifically on the simplest composition case of a binary system we investigate the effect of ordering on the X-ray signal. Taking the scenario of a bimetallic column containing 5 Pt atoms and 2 Ni atoms (7 atoms in total) there are 21 different possible configurations. This range of configurations produces a range different X-ray signals which most likely translates to an uncertainty in determination of sample composition, if unknown. Figure 2, shows how the X-ray signal from the different configurations changes with sample tilt away from a <110> zone axis. The further one tilts away from a low-order zone-axis the closer the signal gets to the non-channelling case and the spread in values between different configurations decreases. Therefore, the more we intentionally tilt the sample the more accurate our composition determination will be, at least for small sample tilts. At larger tilts, the tilted columns may overlap in projection and the signals of individual columns can no longer be separated. Between these two competing ...

show abstract

“…In (a) the heavier element (Mn) alone appears bright. These data are more difficult to analyze, because the inner shell excitation is delocalized by a certain distance and the neighboring atomic columns simultaneously contribute to the spectrum intensity at a sampling point due to electron channeling effects [25] and orbital hybridization between the elements. For all datasets, even after the above pre-processing, a few elements in X had small negative values due to background removal.…”

Section: Methodsmentioning

confidence: 99%

“…We subsequently focus on the additional third component, the spatial distribution of which was found to be localized on the projected oxygen atom positions. This localization was attributed to the electron channeling effect [25], which is responsible for propagating the incident electron wave function along the neighboring Mn columns for a sample exceeding a certain thickness when the electron probe is placed on the oxygen column. The resolved spectrum of the third component actually exhibits a spectral profile characteristic of the weighted average of the other two components, because each oxygen atom is coordinated with trivalent Mn Oct and divalent Mn Tet atoms.…”

Section: Atomic Resolution Stem-eels Of Mn 3 Omentioning

confidence: 99%

“…We adopted the modified alternating least-square (MALS) fitting algorism of NMF [14] to map the different phases in the degradation of Li battery cathodes [15][16][17][18][19] and the chemical states of nitrogen in nitrogen-doped TiO 2 [20,21]. We also successfully applied NMF to a series of EELS datasets for the extraction of atom site-specific core-loss spectra, where the relative excitation probabilities of the spectra varied with the diffraction condition because of the electron channeling effects [22][23][24][25]. In these applied data analyses, the nonnegative constraint of the elements of extracted basis spectra and spatial intensity distributions were effective, and the resulting spectra extracted by NMF were consistent with the computational results obtained by first principles calculations [15][16][17][18][19][22][23][24][25].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

High Spatial Resolution Hyperspectral Imaging with Machine-Learning Techniques

Shiga

¹

,

Muto

²

2018

View full text Add to dashboard Cite

Recent advances in scanning transmission electron microscopy (STEM) techniques have enabled us to obtain spectroscopic datasets such as those generated by electron energy-loss (EELS)/energy-dispersive X-ray (EDX) spectroscopy measurements in a PC-controlled way from a specified region of interest (ROI) even at atomic scale resolution, also known as hyperspectral imaging (HSI). Instead of conventional analytical procedures, in which the potential constituent chemical components are manually identified and the chemical state of each spectral component is successively determined, a statistical machine-learning approach, which is known to be more effective and efficient for the automatic resolution and extraction of the underlying chemical components stored in a huge three-dimensional array of an observed HSI dataset, is used. Among the statistical approaches suitable for processing HSI datasets, methods based on matrix factorization such as principal component analysis (PCA), multivariate curve resolution (MCR), and nonnegative matrix factorization (NMF) are useful to find an essential low-dimensional data subspace hidden in the HSI dataset. This chapter describes our developed NMF method, which has two additional terms in the objective function, and which is particularly effective for analyzing STEM-EELS/EDX HSI datasets: (i) a soft orthogonal penalty, which clearly resolves partially overlapped spectral components in their spatial distributions and (ii) an automatic relevance determination (ARD) prior, which optimizes the number of components involved in the observed

show abstract

X ‐ray emission generation constant from mono‐layer graphene measured using STEM‐EDS map detected with highly sensitive EDS system

Jimbo

¹

,

Sasaki

²

,

Sawada

³

et al. 2016

European Microscopy Congress 2016: Proceedings

0

View full text Add to dashboard Cite

X‐ray generation constant of an atom by an electron irradiated is essential information for quantification of sample in X‐ray fluorescence spectroscopy. The generated X‐ray intensity ( I ) from a sample is described as I = 1/A ∫ A ∫ t P( x , z)I k ( x ) dzd x , where A is electron beam irradiated area, t is sample thickness, P is electron density and I k is X‐ray intensity excited by an electron. To evaluate I, it is requested the accurate value of I k . Then, to have accurate I k , the comparison between I k values from theoretical calculation ( I k theory ) and experimental result ( I k exp ) is important. However, it is exacting to measure the I k exp accurately using a multi‐layer specimen because of two difficulties, especially for low energy X‐rays. One is self absorption of emitted X‐ray, which can be judged from the 3D sample shape. And the other is counting a number of atoms included in an analysis area. Namely, it is difficult to grasp the 3D sample shape. The shapes of single atomic chain or single layer specimen are simple and ideal for the aimed experiment in this paper. In this paper, we report the way and results of directly measured X‐ray generation constant of carbon atoms, using a mono‐layered graphene sample, which is easy to estimate 3D shape from electron microscope images. For experiment, highly sensitive detector system is requested, since the X‐ray signal from this sample is so thin and emit a little of X‐ray. We used a multiple silicon drift detector (SDD) system for the X‐ray measurement, since the SDD has become highly sensitive recently due to design flexibility of its shape and size to fit busy space around the sample of TEM. The microscope we used was an aberration corrected microscope (JEM‐ARM300F). For the new detection system with two SDD , a pair of objective lens pole pieces and a sample holder are re‐designed to make a distance between sample and detectors as short as possible, and each detector size is enlarged to be 100 mm 2 , resulting in the total detection solid angle of 1.6 sr. Figure 1(c) shows the highly magnified ADF image of mono‐layer graphene from the square indicated by dashed line shown in Fig. 1(a). From the image we can recognize the area of mono‐layer graphens by clear lattice image and hole of the sample. An X‐ray elemental map shown in Fig. 1(b) was obtained simultaneously with ADF image shown in Fig. 1(a). The difference of X‐ray intensity between the hole and the mono‐layer graphene is apparent in Fig. 1(b). This leads that the X‐ray signal from mono layer with dual SDD detection system is well detectable under the dose, at which mono‐layer graphene maintains its structure. The experimental conditions for this experiment were followings: accelerating voltage = 80 kV, probe current = 169 pA, number of pixels = 128 × 128 and total acquisition time = 300 sec. The I k exp of carbon is estimated to be 3.29×10 ‐23 photons*cm 2 /electrons, using the experimental conditions and X‐ray intensity from the mono‐layer graphene area on Fig. 1(b). This is described as I k exp ≈ 4π I total /( D carbon *S * D electron * W ), where I total is total X‐ray counts (photons) from scan area of mono‐layer graphene, D carbon is density of carbon atoms (atoms/cm 2 ) in a analyzed area, S is scan area of graphene (cm 2 ), W is a solid angle of X‐ray detection and D electron is density of electrons (electrons/cm 2 ). D carbon was estimated to be 3.82×10 15 atoms/cm 2 from typical atomic density for a unit area. S is 1.38×10 ‐13 cm 2 . The amount of X‐ray counts ( I total ) was measured to be 155 photons accumulated over the area of mono‐layer graphene. The density of electrons ( D electron ) is simply calculated to be 7.22×10 22 electrons/cm 2 from experimental conditions. The theoretical generation constant ( I k theory ) was calculated to be 1.85×10 ‐23 photons*cm 2 , which was calculated as a product of ionization cross section (ionization atoms*cm 2 /electrons) by Bethe [2] and X‐ray fluorescence yield (photons/ionization atoms) by Burhop [3], which is shown as I k theory ≈ σ k W k , where σ k is ionization cross section and calculated to be 2.70×10 ‐20 ionization atoms*cm 2 for carbon, W k is x‐ray fluorescence yield and calculated to be 7.11×10 ‐4 photons/ionization atoms for C Ka. The experimental value ( I k exp ) do not differ from one by pure calculation ( I k theory ) so much. The difference may be due to error in the pure calculation, since the calculation contains considerable approximate expressions. In conclusion, with the highly sensitive detection system and 2D sheet sample, we could measure a physical constant with considerable accuracy. The similar 2D samples such as BN and MoS 2 sheets should be useful for measuring X‐ray generation constants.

show abstract

On the quantitativeness of EDS STEM

Cited by 71 publications

References 42 publications

Probing the Effects of Electron Channelling on EDX Quantification

Probing the Effects of Electron Channelling on EDX Quantification

High Spatial Resolution Hyperspectral Imaging with Machine-Learning Techniques

X ‐ray emission generation constant from mono‐layer graphene measured using STEM‐EDS map detected with highly sensitive EDS system

Contact Info

Product

Resources

About