Secondary electron hyperspectral imaging: Nanostructure and chemical analysis for the LV-SEM

Abstract

Materials engineers have increasing control over nanoscale chemical composition and nanostructures. The latter can be easily analysed by the scanning electron microscope (SEM) yet nanoscale chemical analysis in the SEM poses challenges. Nevertheless, nanoscale chemical analysis capabilities have been developed in the LV-SEM by secondary electron hyperspectral imaging (SEHI). In this article, the insights that SEHI has provided into well characterised semiconductor, photovoltaic, polymer and hard carbon materials are reviewed. Instrument and experimental considerations for the obtainment of secondary electron hyper-spectra are discussed and recommendations for future analyses in the LV-SEM by SEHI are made.

Keywords

Scanning electron microscopy low-voltage scanning electron microscopy secondary electron imaging secondary electron emission spectroscopy secondary electron hyperspectral imaging surface analysis energy dispersive X-ray analysis SEM LVSEM SEES SEHI EDX EDS WDX WDS

Introduction

The scanning electron microscope (SEM) is known to be a workhorse of materials characterisation [1]. An SEM interrogates a specimen by scanning a beam of electrons over the specimen surface and detecting electrons or radiation emitted from the surface. The energy, intensity and angular distribution of emitted electrons/radiation can provide topographic, morphological and chemical information about the surface. The most widely exploited signal for topographical images is the secondary electron (SE) signal, although difficulties arise for the inspection of insulating and easy to damage materials.

For such materials, low voltage SEM (LV-SEM) instrumentation has developed to the point where high-resolution SE imaging of the topography is possible [2 -4] [5 -8]. Previously, image quality was severely impacted due to charge build-up, deposited contamination and radiation damage to the specimen associated with high electron beam energy, E₀ [9].

LV-SEM provides highest quality topographic information but faces challenges in obtaining chemical information [10,6,11] If it can be obtained, most chemical analysis in the SEM provides only elemental information through energy-dispersive X-ray spectroscopy (abbreviated as EDS/EDX) and wavelength dispersive X-ray spectroscopy (WDS/WDX), rather than functional chemical information which is critical in many materials engineering applications and may be obtained by methods such as ultraviolet and X-ray induced photoelectron spectroscopy (UPS; XPS) and nano-Fourier transform infrared spectroscopy (nano-FTIR) [12,13].

An alternative to EDX was suggested for the SEM in the semiconductor industry, where non-destructive evaluation of devices on the production line is routine. Strict parameters are imposed on SEM probes so as not to damage the electronic devices, and these parameters are continuously tightened as devices become smaller and more vulnerable to damage by the electron beam used to interrogate them [14,15]. EDX and backscattered electron (BSE) analyses are still standard in semiconductor device manufacture [16], but a movement towards higher resolution SE derived chemical information for non-destructive device-specific assays and chemical analysis has taken hold [17 -21,11] and has now begun to be applied in other areas of materials science, with great effect (see Section ‎6).

The purpose of this literature review is to examine first how the gap between topographical and chemical information in the LV-SEM has developed – as a result of improvements in SE imaging and the physical limitations of EDX and WDX analysis. This is followed by an examination of how this information gap may be closed by the use of secondary electron hyperspectral imaging (SEHI) and the opportunities that lie therein.

Chemical analysis in the SEM

X-ray analysis: Background and limitations

The emission of X-rays from a target specimen induced by incident electrons is described in detail elsewhere [1,22,23], and summarised as follows: incident electrons are called primary electrons (PEs) before interaction with the specimen. PEs may transfer kinetic energy to an inner-shell electron by an inelastic collision with the inner-shell electron. The electron may have gained enough energy to be removed from the atom, a process called ionisation. The relaxation of an electron from higher energy level to the newly unoccupied energy level has an associated energy loss. Since atomic energy levels have characteristic energies, the energy loss produces a high energy photon (X-ray) of characteristic wavelength, X-rays are generated along the high energy electron paths until electrons are below the ionisation energy, thus X-rays are generated from most of the interaction depth [22,24]. Information depth may be reduced to ∼1–3 µm for E₀ ∼15–25 keV [25] and to sub-micron range with E₀ ∼5 keV as long as X-rays are produced with sufficient intensity to be detected [26]. In the SEM, X-rays may be detected by a number of methods, yielding both elemental and chemical information.

The most common method of X-ray analysis is by direct detection of a full spectrum of X-ray energies and intensities (energy dispersive X-ray analysis – EDX). The spectrum gives elemental information as the inner-shell electrons from which the X-rays are derived are least affected by neighbouring atoms. EDX analysis may provide qualitative information about the elemental composition of a specimen since emitted X-ray intensity is related to the probability that an atom will be ionised by a PE with a given energy. When SE imaging in conjunction with EDX analysis, care must be taken that SEs do not produce X-rays external to the specimen – for example at the chamber walls [27]. Quantitative analysis is also possible by EDX analysis, although measurements are conventionally corrected by mean atomic number in the specimen (Z), depth distribution of absorption (A) and fluorescence (F) (the ZAF correction [28]) and specimens must have a flat polished top surface [22,29].

Routine EDX analysis in the SEM has a lateral resolution ∼1 µm and a detection limit of 1 wt-% [24]. Improvements in spatial resolution with lowered E₀ is possible by increasing the sensitivity and count rate of X-ray detectors [30]. For example, the large-area silicon drift detector (SDD) for detection of low energy X-rays has enabled E₀ to be reduced to around 3 keV giving elemental maps with ∼20 nm spatial resolution [31]. Spectral imaging techniques (see Section ‎5.1) have also been applied to EDX produce nanoscale elemental maps [32].

The additional capability of wavelength dispersive X-ray spectroscopy (WDX) is to make trace element analyses (>0.1 wt-%) by detection of wavelength-filtered X-rays at greater energy resolution. The filtering of X-rays allows overlapping peaks to be distinguished and is achieved by reflecting the emitted X-rays in an analytical crystal of given lattice spacing before detection. The technique has ∼1 μm lateral resolution since beam requirements are similar to those required for EDX analysis [33].

As well as elemental analysis, X-ray analysis in the SEM may yield information about surface chemical species. The soft X-ray emission spectrometer (SXES) was first applied to the SEM by Terauchi et al. (2014) and was used to detect X-rays over the 35–210 eV range allowing for the observation of density-of-states (DOS) of both the conduction and valence bands as well as characteristic X-ray emission from elements down to lithium (Li Kα-emission at 52 eV [34]) [35,34]. Beam currents required are ∼50 nA and the acceleration voltage used is from 5 keV and largely above the LV-SEM operating region [35], calling in to question the application of SXES to non-conductive and beam sensitive specimens.

The generation of characteristic X-rays with sufficient intensity requires PEs with kinetic energy greater than the ionisation energy of the elements under investigation (E₀ > 1.3 times the characteristic ionisation energy) and beam currents in the 10s nA range (versus 10s pA for LV-SEM), resulting in physical limitations to the application of X-ray analysis in the LV-SEM [26,27,11].

Auger electron analysis

Auger electron (AE) ejection is another relaxation mechanism by which the energy associated with ionisation of an inner-shell electron and subsequent relaxation of an electron to the inner-shell vacancy is transferred [36]. In this case, the energy is emitted not as a high energy photon (X-ray) but a relatively low energy electron. This relaxation process was first observed by L. Meitner in 1922 [37] but it was P. Auger (1923) who observed that emitted electrons had energies independent of the energies of incident X-rays used to ionise atoms [38,39]. The electron ejected by the inner-shell ionisation is of characteristic energy, just above the SE energy range. Despite PEs in the range 1–25 keV being used to ionise the specimen, overlapping with the PE energy range used in EDX analysis, detected AEs have an escape depth of only a few layers of atoms owing to the probability that AEs produced within the bulk specimen will not reach the detector due to scattering events and absorption within the bulk specimen [40,41]. In order to produce sufficient AEs, large beam currents tend to be used (∼10⁻⁹–10⁻⁶A) [42,43].

Auger electron spectroscopy (AES) has long been the only means of analysing the surface of a bulk material with PEs owing to the information depth of AEs. AES is established as a technique for measuring surface contamination on materials [44], and can be used to relate surface composition to performance, for example, carbon fibres [45]. More recent nanotechnology roadmaps identify AES as an essential tool for the characterisation of nanoparticles and nanomaterials [46,47] in areas such as nano-toxicology and nano-medicine where characterisation is fraught with difficulties, in particular poor reproducibility [48,49].

AEs are routinely used for imaging (scanning Auger microscopy, SAM) [50,51,43], not just for spectroscopy, with SAM first demonstrated by Macdonald and Waldrop in 1971 [52]. It was noted that these images of elemental distribution have a resolution proportional to the electron beam spot size, not the interaction volume as in EDX [53,54]. SAM can produce elemental maps with 15 nm resolution [51,43], thus exceeding the detail of high-resolution EDX elemental maps for decades [55,31]. AES is a quantitative technique, as the background electron emission spectrum can be removed from the AE signal, followed by combined empirical and theoretical calculation of atomic per cent from relative elemental emission intensities [56].

Chemical and topographic information from backscattered electrons

The kinetic energy and angular deflection of emitted BSEs convey both chemical and topographic information. BSEs are produced by the interaction of PEs with the positively charged nucleus of an atom. The deflection of incident electrons is large enough for electrons to re-emerge from the specimen surface, where they are detected over a wide range of angles. Elemental contrast is obtained due to the intensity of the BSE emission, with heavier elements having a larger scattering cross-section (see Figure 3 for a comparison of BSE yield η, between C and Au), with high-Z regions appearing bright and low-Z regions appearing dark [23].

BSEs are conventionally defined to have an energy, E_BSE > 50 eV and exist over a broad range of energies, up to the PE energy, E₀. BSEs have a large energy distribution due to the large variation in the path length of electrons within the specimen [23]. Numerous detectors are used to image over the E_BSE distribution: the Everhart–Thornley detector (ETD) (for low energy BSEs) as well as dedicated overhead BSE detectors for BSEs with large-angle deflections [57]. Low-angle BSEs collected by stage tilting have additional information about crystal structure, orientation and epitaxy as well as strain within crystals [58,59,60].

The escape depth of BSEs is larger than SEs at roughly half the interaction depth [22]. Since longer path lengths in the specimen result in more scattering interactions and energy loss, the energy of a BSE may be related to escape depth. The BSE energy distribution is often classified into two regions: BSE1 for high energy BSEs and low energy BSE2 [61] (first named BS1 and BS2 by Joy, 1991 [62]). A cut-off energy may be applied to filter BSE1 for the low-loss electrons (LLEs) scattered by the very surface of the specimen [61]. The LLEs may be used for high spatial resolution imaging with Z contrast, but information from low-angle of escape LLEs may be lost due to asperity shadowing [62].

Chemical and topographic information from secondary electrons

The Auger and backscattered electrons are derived from inner-shell and nucleus interactions, and have energies defined to be >50 eV [63]. Below this 50 eV threshold the emitted SEs are assumed to originate from inelastic scattering events of PEs and BSEs with solid-state electrons [64]. Most SEs exist <10 eV and SEs are further categorised into SE1-3 depending on the interactions by which they are produced [64]. SE1 are ejected due to inelastic scatting by the initial PE-specimen interaction, whereas SE2 and SE3 are produced by a cascade of electron interactions which reduce in energy, further removed spatially and in time from the initial PE-specimen interaction. SE2 are produced by inelastic scattering of BSEs near the specimen surface and SE3 are derived from inelastic scattering on the surfaces of the SEM itself [64,65].

For high-resolution images, SE emission (SEE) must be localised to the PE spot and so SE1s are used for high-resolution topographical imaging alongside BSEs up to atomic resolution [3] although best practice typically yields sub-nanometre resolution, even for soft and insulating materials [66]. High topographic contrast is obtained from SEs at asperity edges, which appear brighter than a surface perpendicular to the PE beam since there is a larger surface area from which SEs are emitted [62]. The high spatial resolution of SE images is due to the information depth being shallow and the SE1 emission being extremely localised to the initial PE interaction in space and time, and therefore detection of SE2 and SE3 decreases image resolution. The development of the through-lens-detector (TLD) has improved the detection of the SE1 signal and exclusion of SE2 and SE3 [4]. SE1s make up only 65% of the SEs detected by the in-chamber wide-band ETD [67,68,64]. Furthermore, the proportion of SEs detected may be increased by negative specimen bias voltage [67] and increased TLD extraction voltage [18]. It has been proposed that SEE information may be stratified by depth according to the energy of emitted electrons [10] and this principle has been shown to differentiate between the bulk and surface contamination on a doped silicon wafer [69,20].

While SEs are collected to image the topography of the specimen surface, the SE emission energy spectra (SE spectra) have been of interest to microscopists for chemical analysis for as long as they have been collected. The first collection of detailed SE spectra was for pure molybdenum in 1935 [70] and alloys followed [71]. A need for high-resolution microanalysis of insulating or beam-sensitive materials, as well as instrument development, may be attributed with wider interest in secondary electron emission spectroscopy (SEES) and its application in the SEM in the 1990s–2000s [72,11,73,74]. While the physical formation of SEE is not yet well understood, SEES is still a valuable technique for differentiating between surface chemistries on a specimen [75]. Here, the basis of SE chemical signal is discussed in terms of specimen high DOS energy bands, comparison to UPS and Raman spectroscopy, removal of topographic SE signal from SE spectra and models of the inelastic scattering of low energy electrons before emission from the specimen.

SEES analysis must rely on there being a link between characteristic SE spectra and the electronic structure of the specimen, and efforts have been made to link features in SE spectra to the electronic configuration of surface chemical species. As a result, SE spectra have been related to the band structure of materials, notably in assigning SEE peaks to unoccupied high DOS energy bands [76,77]. Hoffman et al. (1992) also made comparisons of ‘signatures’ in SE spectra obtained from allotropes of carbon [78]. Furthermore, links between surface chemical species and SE spectra have been made by SEHI for a few well-studied materials. Abrams et al. relate the shifting of sp³-carbon like peaks to the modification of the surface energy barrier by contamination, reduced by amorphous carbon species (a-CH) and increased by hydroxy groups [79].

SE spectra obtained by LV-SEM may be compared to spectra obtained by UPS – corresponding emission peaks were observed in both spectra from a specimen of methylammonium lead triiodide (MAPbI₃) [80]. SE spectra from HOPG produced by He⁺ ion bombardment and ultraviolet (UV) photons also had corresponding peaks which were related to high DOS energy bands [81]. Similarly, LV-SEM obtained SE spectra may be compared to vibrational spectroscopies. Farr et al. correlated the increase in intensity of a 3.6 eV SEE peak (related to –CH; assigned Abrams et al. [79]) with increased cross linking of poly(glycerol sebacate)-methacrylate (PGS-M) polymer and a similar increase in the 2950 cm⁻¹ peak (C–H vibration) intensity in the Raman spectra [82].

In order to establish the link between SE spectra and electronic structure, the influence of surface topography on SEE intensity must be ruled out, at least in a SEE energy range of interest. In the case of analysis of specimens with large topographic contrast such as particles and fibres, the SE chemical signal may be isolated by SE energy filtering experiments. Kumar et al. varied the particle size of TiO₂, to determine that chemistry makes the largest contribution to SEE below a determined SE energy [83]. SE spectra produced by He⁺ ion bombardment and UV photons show the relation of electron emission intensity to high DOS energy bands below <14 eV, thus verifying the point at which chemical contributions dominate in LV-SEM obtained SE spectra [81]. Meanwhile, SE spectra taken from flat specimens such as poly(3-hexylthiophene) (P3HT) and PGS-M polymer films show variation between amorphous and semi-crystalline regions despite the lack of topographic SEE contrast, with peaks in the SE spectra correlated to polymer molecular order [13,82].

Established Monte Carlo (MC) models do not reproduce realistic inelastic mean free path (IMFP) values for electrons with low energy, which results in a breakdown of models in the SE spectral region of interest [84,85]. The CASINO V3 model used in Section ‎3.5 is only accurate for electrons >500 eV since it neglects inelastic scattering events and instead applies a continuous energy loss function [86,87]. Lately, MC models have been developed to include the contributions of inelastic scattering events in the cascade of energy-loss interactions. Dapor (2012) describes interactions of electrons with solid-state electrons, plasmons, phonons and polarons as contributing to the energy loss of low energy electrons, all of which are inputs to the calculated IMFP [84]. More accurate productions of SE spectra by MC models with empirical and theoretical inputs have followed [88,89,13].

Information depth and spatial resolution of secondary electrons

Given the information that may be obtained in the SEM from X-rays, AEs and BSEs is dependent on the E₀ and interaction volume within the specimen, it is imperative that the influence of these factors on SEE is well understood. First, the information depth of SEs is related to electron energy and the electron stopping distance of a material. Second, the effect of E₀ on the spatial resolution of SE imaging is discussed in terms of the emission of SE1 and SE2. Finally, the derivation of highly localised SE1 from PE interactions is reviewed.

SEs are produced throughout the PE interaction depth as well as along BSE trajectories, but only SEs produced within the SE escape depth can be detected, since there is a probability that SEs will be absorbed by the bulk material [65]. The information depth of SEs is low compared to BSEs and X-rays since the kinetic energy of SEs is dissipated by inelastic scattering. Ono and Kanaya calculated the material-characteristic probable SE escape depth with theoretical inputs: Z, first ionisation energy and BSE yield for elements lithium to thorium giving 4.8 nm for carbon, 10.7 nm for organic carbon and 1.4 nm for gold [90]. It has been shown that the information depth is also dependent on the energy of SEs. Figure 1 shows the experimental electron stopping power of carbon, with a maximum stopping power for electrons with a kinetic energy of ∼90 eV. This gives a stopping distance of 1.2 nm for 90 eV electrons in carbon and 61 nm for electrons with a kinetic energy of 10 eV. The increased information depth of low energy SEs has been exploited in energy-filtered SEM to increase voltage contrast and image beneath contamination layers [6,69].

Here, E₀ is discussed in terms of interaction depth since this may be easily related to the spatial resolution of SE topographical information by established MC models [87,86]. The results of the simulations are presented in Figure 1, showing MC model derived graphics of the interaction of PEs with the HOPG specimen (coloured grey). Bulk highly-orientated-pyrolytic-graphite (HOPG) has been chosen as the model specimen given its use to date in SE spectroscopy. In this case, the specimen is modelled by average density 2.26 gcm⁻³ and work function, Φ^HOPG = 4.8 eV [91]. As interaction depth increases from (a) <20 nm to (c) >200 nm the yield of SEs is decreased, illustrating the absorption of SEs within the bulk material. The SE2 signal distribution is the limit to SE spatial resolution [92] and the generation of SE2 in (b), (c) up to 40 nm away from the initial PE interaction by BSEs is depicted.

The CASINO MC model is best understood by the assumptions made in calculating electron trajectory. The material in which elastic and inelastic collisions occur is modelled by atomic weight, density and cross-section [87,93]. Thus, the electron means free path between collisions does not consider lattice position and electronic configuration of atoms in the specimen.

As coarsely shown in Figure 2, the spatial resolution of SE images in some cases can be determined by the emission of SE1 within the region of the initial PE interaction. It has been proposed that the higher the inelastic scattering energy loss event, the more localised the SE1 signal and high-resolution SE imaging and SE coincidence measurements have shown that inelastic scattering of PEs and emission of SE1 may dissipate several kilo-electron volts [3,94].

Low voltage SEM

Comparison of SEM to LV-SEM

SEM routinely uses E₀ < 50 keV and LV-SEM is defined by E₀ < 5 keV [7]. Low E₀ increases the spatial resolution of SE imaging by decreasing the emission of SE2 removed in space and time from the initial PE interaction (illustrated by the MC simulations in the LV-SEM range in Figure 2). The influence of these factors on image resolution was known from the outset of SEM development, with the first SEM built by M. Knoll (1935) utilising E₀ in the range 0.5–4 keV [95]. Lower E₀ also exposes the specimen to less overall power input from radiation. Radiation damage in electron microscopes is reviewed by Egerton et al. (2004) in terms of specimen heating, displacement of atoms on the lattice, mass loss and hydrocarbon deposition, but a need for better understanding between radiation dose and SEM specific parameters is acknowledged [9].

As well as low E₀ and radiation dose, investigating low Z elements in LV-SEM conditions may produce a larger SE yield, δ (Equation (1)), compared to the BSE yield (Equation (2)) resulting in a SE signal to noise ratio that is larger than the BSE signal to noise ratio at a given E₀. For an as-inserted carbon specimen the maximum δ at E₀ = 0.35 keV is δ = 0.96 while the BSE yield at the same E₀ is η = 0.10 (Figure 3 ). In this case, the total emitted electron yield, σ (Equation (3)), is greater than the PE yield, σ = 1.06. For the high Z as-inserted gold specimen, the SE yield is greater than BSE yield up to E₀ = 4.5 keV (Figure 3). (1) (2) (3)

where N(SE) and N(BSE) are the numbers of SEs and BSEs emitted and N(PE) is the number of incident PEs [64].

LV-SEM has the ability to image the surface of bulk insulating materials, where the ground current is zero, since the beam current may be dissipated by SEs and BSEs at a rate which does not cause catastrophic specimen charging [96]. For example the E₀ may be selected such that σ = 1, (for carbon E₀ = 0.25, 0.55 keV, Figure 3).

Instrumentation of LV-SEM

Alongside the need for high resolution and chemical analysis for insulating and beam sensitive materials, technological advances have been as important in enabling LV-SEM. Discussed in brief here are the electron gun and vacuum technologies required for LV-SEM (developments in SE detection are discussed in Section ‎3.4).

Perhaps the main enabling technology for LV-SEM was the increase in electron gun brightness, with the adoption of the field emission gun (FEG). The ideal electron gun is as bright as possible, while producing monochromatic electrons from as small a source size as possible [23]. In combination with lenses, the electron gun brightness and energy spread define the minimum probe diameter to which, for a given E₀, image resolution is directly related [23]. Comparison of three electron sources and their brightness's, energy spreads and probe diameters is shown in Table 1.

Table 1

Properties of the electron source compared to minimum probe diameter. Reproduced from Suga et al. [61].

Electron source	Brightness / A·cm⁻²·sr⁻¹	Energy spread / eV	Probe diameter / nm
Thermionic	10⁵–10⁶	1–3	5000–100,000
Schottky-FEG	10⁸	0.3–1	15–30
Cold-FEG	10⁸	0.3	<5

Throughout the SEM, high vacuums are required to avoid scattering of electrons by gas particles. There are further requirements for the vacuum in the electron gun and the specimen chamber. The FEG requires a high vacuum (>10⁻⁹Pa) in order to maintain the high electric field around the tip and to keep the source free from surface contamination. In the specimen chamber, chemical species that may be deposited onto the specimen surface by the electron beam must be removed before SE spectra are collected [9]. Today's instruments achieve chamber pressures ∼10⁻⁹Pa and the user can obtain images and spectra within minutes of mounting the specimen. Contrast this with the efforts of L. Haworth, who spent 146 days producing a vacuum before being able to obtain an SE spectra for molybdenum [70], and it is clear that EM microscopists owe much to modern vacuum technology.

Consequences of LV-SEM on conventional SEM-based chemical analysis

The uptake of LV-SEM has consequences for the viability of chemical analysis by X-rays, AEs and BSEs. In the first two instances, the emission of characteristic X-rays and AEs is an inherently high energy process, as incident E₀ must be greater than the ionisation of the elements to be analysed. LV-SEM may produce X-rays and AEs with insufficient intensity to be detected or, below the ionisation energy, none at all. For example, Joy et al. (2004) give the emission of SEs to be 0.1–1 electrons/steradian/electron versus fluorescent X-rays at 10⁻⁵ photons/steradian/electron [73].

While no such cut-off energy exists for BSEs, low BSE yield, η, from low atomic weight specimens in the LV-SEM E₀ range compared to δ gives SE chemical analysis a clear advantage in terms of resolution and signal to noise ratio (Figure 3(a)). Additionally, the highest energy LLEs which have the highest spatial resolution of the BSE spectrum are emitted with high deflection angles and may be shadowed from the detector by specimen surface asperities [62]. The SE1 signal is detected by the TLD directly over the interaction volume, and thus asperity shadowing is limited.

Secondary electron hyperspectral imaging in LV-SEM

Rational of multi- and hyper-spectral imaging

Spectral imaging (SI) is a means of assigning higher dimensions to an image with spatial dimensions (x, y), for example, a third spectral dimension, z. For example, in satellite images where many SI techniques were first developed, a pixel may be addressed by its position in x and y and also over a large part of the electromagnetic spectrum, λ, resulting in dimensions (x, y, λ) [99]. Thus the image may be displayed as a three-dimensional stack of images composed of defined reflected energy bands (a data volume) as shown in Figure 4 [100]. Multi- and hyper-spectral images differ only in the detail in which z is collected. Hyperspectral images are collected over a contiguous (neighbouring band) range in z, while multispectral images are composed of a few wide bands in z. Of particular use to remote sensing applications is the means by which source contributions to the image can be isolated by analysis of the z-spectra. Source separation may be done by machine learning algorithms, in which no theoretical inputs are made to aid the algorithm in pattern recognition [100]. Non-negative matrix factorisation (NMF), an algorithm for separating contributing sources from z has been applied to EDX analysis in order to map elements to ∼20 nm resolution [32].

In the context of LV-SEM, HI may be used to obtain chemical information from SEs at the same resolution as topographic information, thus closing the resolution gap between topographic and chemical information which has existed in the LV-SEM [6].

Multi- and hyper-spectral imaging by electron energy loss spectroscopy

Spectral imaging by spatially resolved electron energy loss spectroscopy (EELS) in the transmission electron microscope (TEM) is a mature technique [101], first utilised by Bonnet et al. (1999) to map distributions at an SiO₂–TiO₂ interface [102]. EELS produces an electron energy spectrum by measuring the energy of electrons transmitted through a specimen and compares transmitted electron energy to the E₀. In contrast to LV-SEM, arduous specimen preparation is required to thin the specimen to a width of nanometres and the E₀ required is generally in excess of 100 keV [9]. EELS in TEM and scanning-TEM (STEM) systems may generate spectra over a range of hundreds of electron volts with <1 eV resolution per pixel in a 100 × 100 array [103,104]. HI by STEM may achieve elemental mapping with sub-nanometre resolution for insulating material, but the technique has limitations in terms of specimen thinning and radiation damage to the specimen [103,9].

Working principal and instrumentation for SEHI with the LV-SEM

Characteristic SE energy spectra have been collected for many decades (see Section ‎3.4), yet it is still widely assumed that SE micrographs show only topographic contrast. However, recent progress in energy-filtered SE detection and data analysis enables the reconstruction of spectral images over defined SE ranges which display chemical and functional information while supressing topographic contrast [83,10]. The principal for hyperspectral imaging (HI) in the LV-SEM is shown in Figure 3. Energy spectra are obtained serially per pixel over a user-defined energy range. Thus, the data volume may be constructed in (x, y, E) where E is the energy of detected SEs. The hyperspectral data volume may be further analysed by pattern recognition machine learning algorithms to produce composite images of functional chemistry (SE chemical maps).

SEHI requires LV-SEM instrumentation. The probe diameter must be of the order of the chemical information to be resolved - generally material engineering on the nano- and sub-nanometre scales [17,6,105,83,13] – and a high yield of SE1s must be obtained by using low E₀ [7]. To construct the hyperspectral data volume, the integrated SE1 emission is collected, while excluding the detection of SE2-3 and BSEs [4,10]. The electron dose received by the specimen during spectrum acquisition is minimised by selecting the maximum contiguous energy bandwidth that does not smooth the characteristic SE spectrum peaks to be analysed (∼0.2–0.5 eV, FEI Nova NanoSEM). Highly spatially resolved SE detection is made with the TLD, and SE energy filtering is achieved by the electrostatic deflector electrodes (Figure 5). The lower pair of electrodes perform the energy filtering of SEs by progressively deflecting more of the SE signal onto the TLD as images are serially acquired. The upper electrodes correct the deflection of the PE beam caused by the lower energy filtering pair of electrodes [106].

It should be noted that the TLD is designed for SE imaging and not spectroscopy. The angular dependence of SE detection is system dependent. A TLD may detect all angles of SEs only for E_SE < 10 eV at a 3 mm working distance (WD) (FEI Sirion/Philips XL30 SEM) [106] while detection of all SEs occurs E_SE <3 eV at 1 mm WD (FEI Magellan 400) [107] (see ‘acceptance diagrams’ of angle of emission versus SE energy for the TLD [106,107]). SE spectra obtained from P3HT films with both the FEI Sirion and Nova LV-SEMs show the effect of WD on SE spectra shape, independent of E₀ [89].

Energy filtering experiments have been devised to relieve the impact of detector design such as shadowing the SE signal. In order to reduce the effect of detector shadowing, Stehling et al. (2019) took two spectra either side of a 180^o stage rotation to produce a single averaged spectra in which effects of fibre geometry and detector shadowing are cancelled [108]. Developments in energy filtering for SE detection continue to be made to the benefit of SEHI, and alternative deflectors arguably more appropriate for SEHI exist. For example, the parallel plate analyser (PPA) deflects all SE energies onto the detector throughout spectra acquisition and does not require serial acquisition of integrated SE emissions to produce spectra [10,109].

Challenges in the collection of SEHI

Instrument automation has made fast acquisition of SEHI possible. Nevertheless, SEHI presents particular challenges to the electron microscopist, namely instrument drift, specimen charging and SE spectra calibration. Furthermore, if quantitative SEHI is to be realised, great care must be taken in optimising spectra collection in order to produce coherent and comparable datasets between instruments [18,10].

Instrument drift can result in an image translation in (x, y) over the z-range of spectral imaging. In other areas of HI application, the translation of features at each measured z-band is called pixel drift [100]. In any case, pixels must be matched throughout the spectral dimension in order to create a coherent z-spectrum which can be fed into the machine learning algorithm.

Specimen charging must be alleviated by the optimisation of instrument and scanning parameters. In the SEM, as well as E₀ (discussed in the context of specimen charging in Section ‎4.1), user-defined instrument parameters that affect specimen charging include electron beam current, I₀, WD, field of view, specimen angle and specimen bias, as well as scanning parameters: scan interlace, dwell time, image integration. Additionally, there are parameters to be defined in the collection of SE spectra: the width of contiguous energy bands and the range of the energy spectrum while being detailed and wide enough to be able to differentiate individual material-characteristic peaks. Given the multitude and hierarchy of factors influencing the collected spectra, it is likely that a design-of-experiments approach to optimisation will be of some use to SEHI, until more rigorous instrument and scanning parameter models are evolved. Such efforts are justified by the range of applications of SEHI in materials science.

Applications of SEHI

Having described the limitations of X-ray, AE and BSE analysis when applied to nanoscale chemical characterisation in the LV-SEM and the working principal of HI, this section serves to provide an overview of the application of SEHI to nanomaterials engineering to date. In the fields identified, SEHI has provided information about surface chemical species that has aided materials development and manufacture where other characterisation techniques fall short.

Semiconductors

Acute need in semiconductor materials engineering for higher resolution chemical analysis than was possible with EDX analysis and at a lower E₀ and beam current than required by both EDX and AE analysis was created by device miniaturisation. Given these requirements, device-specific assays were developed followed by quantitative chemical analysis by SEHI [17,18]. Voltage contrast (VC) for p-type (bright) and n-type (dark) doping is simply described by the potential energy difference between the Fermi and vacuum energy levels for each dopant region. Specifically for clean silicon surfaces, VC is related to two effects, ‘surface band bending’ and ‘patch field’ effect, whereas for a specimen in LV-SEM conditions, a layer of a-CH may build forming a Schottky contact on n-type and ohmic contact on p-type doped silicon [110]. Hashimoto et al. employed a high bandpass filter in order to observe the same contrast effect [111]. Depth-stratified SEHI has been demonstrated by selecting SEs produced beneath a contamination layer on a doped silicon wafer, owing to the large IMFP of low energy electrons, and thus the migration of these electrons through the contamination layer with little effect [69].

Photovoltaics

One class of photovoltaic material is the bulk heterojunction organic blend. For high-efficiency energy conversion, dispersions of fullerenes in the polymer must be engineered at the nanoscale, and sub-optimal distribution has a severe impact on photovoltaic performance. Before the application of SEHI to map regions of a poly(3-hexylthiophene): [6,6]-phenyl C61butyric acid methyl ester (P3HT:PCBM) blend [6], organic photovoltaic blends were imaged by EELS assisted energy-filtered TEM [112], however this approach has limitations given high E₀ and specimen preparation considerations (discussed in Section ‎5.2). The identification of these regions was made by empirical comparison of intra-specimen SE spectra variation between P3HT and PCBM in the 1–10 eV range. This empirical approach demonstrates the value of insights provided by SEHI, at a point where detailed knowledge of links between SE spectra and electronic structure was unavailable.

Success in mapping previously unresolved bromide distributions in perovskite-based photovoltaic material (MAPbI3) followed, by identifying characteristic SE spectra for the doped perovskite constituents. The challenge to the application of EM techniques to these materials is the generation of gaseous halides by the high E₀, which simultaneously changes the chemistry and composition of the specimen. Nevertheless, by using the LV-SEM to obtain SE spectra, disruption to the chemistry was minimised compared to high E₀ and beam current methods and clear conclusions as to the distribution of bromide within the polycrystalline perovskite grain structure were made to aid the synthesis of these materials [83]. In order to exclude topographic SEE contrast, spectra from two TiO₂ particle diameters were compared, showing little variation in emission intensity below <4 eV for the same composition of the particle.

Polymers

There has been rapid progress in the ability of SEHI to resolve crystallinity [113,105,89,13] and functionalisation including crosslinking [82] in polymers, which as well as developing LV-SEM into a versatile characterisation tool, has provided new insights into comparatively well characterised polymers.

Masters et al. (2019) empirically compared SE spectra from amorphous and crystalline P3HT to spectra obtained by nano-Fourier transform infrared spectroscopy (nano-FTIR) of the chemistry. Differentiation between regions was by identification of three peaks <4 eV characteristic to the semi-crystalline P3HT regions. Results have also been verified by new SEE MC models [89,13]. The conductivity of the P3HT polymer is related to the interconnectivity of its crystalline domains, and so the capability to map polymer order is of use to developing organic electronics.

Farr et al. (2019) correlated an increase in the 3.6 eV –CH associated SEE peak intensity with the increase in 2950 cm⁻¹ peak intensity in Raman spectra as crosslinking of PGS-M was increased [82]. In this case, SEHI provides insights into regions of cross-linking in PGS-M at nano- to microscale which was lacking in existing characterisation by nuclear magnetic resonance (NMR) and Raman spectroscopy (at the molecular- and microscales respectively). The production of maps of lipid and protein distribution through coating to core spider silk regions, as well as the identification of nanostructures within the fibre matrix also demonstrates the potential for SEHI to provide multiscale SEHI perspectives to aid materials engineering [108].

Hard carbon

SEE contrast has been used to assess the layer-dependence of Φ^Graphene with nanometre resolution and SE spectra have been produced for sp² carbon [114]. Work on highly ordered pyrolytic graphite (HOPG) amorphous carbon surface contamination has isolated the sp² and sp³ carbon peaks in the SE spectrum, which has provided a greater understanding of electron beam deposition within the SEM and its implications for further SEHI applications [79]. While ultraviolet photoelectron and X-ray photoelectron spectroscopies may also detect the surface contamination of HOPG, these techniques lack the spatial resolution of SEHI [12]. SE spectra have been produced by graphite specific MC models, which account for the crystallographic effects on scattering interactions [88]. Robust characterisation and nanoscale mapping of contamination on HOPG will be of use to hard carbon manufacturers, where the formation of oxide layers has a severe impact on device performance.

Future SEHI analyses

The complex cascade of inelastic scattering events between low energy electrons and solid-state electrons, plasmons, phonons and polarons involved in the formation of SEEs have been exploited by SEHI in LV-SEM conditions to differentiate between regions of surface chemistry for a number of material classes, including highly beam sensitive materials (i.e. MAPbI₃). Despite the contribution of these inelastic scattering events to the reflection of high DOS energy band associated peaks in SE spectra being unknown, reference and intra-specimen composition comparisons can be made. However, in order for SEHI to be a truly quantitative technique, future development of the understanding of the formation of SEE in complex materials such as polymers and alloys is required.

Conclusions

The information gap between SE topographic information and chemical analysis by SEM can be closed by SEHI. SEHI equips the LV-SEM with a characterisation technique applicable to organic and inorganic functional materials. SEHI has clear advantages in terms of resolution and specimen damage compared to established SEM and TEM chemical analysis techniques. New insights into well characterised materials have been produced by SEHI, and these results have been verified by MC models. However, the influence of instrument and scanning parameters for the collection of SEHI requires to be better understood to enable a quantitative chemical analysis of different materials with a range of microscopes. Additionally, insufficient knowledge of the formation of SE spectra for material systems beyond pure elements means that for now, SEHI remains largely an empirical technique.

Footnotes

Acknowledgement

Thanks to my supervisor Dr Cornelia Rodenburg for her guidance during the writing of this literature review.

Disclosure statement

No potential conflict of interest was reported by the author(s).

References

Goldstein

Newbury

Michael

Scanning electron microscopy and X-ray microanalysis. 4th ed. New York ( NY): Springer; 2018. p. 550.

Müllerová

Konvalina

Collection of secondary electrons in scanning electron microscopes. J Microsc. 12 2009; 236(3):203–210.

Zhu

Inada

Nakamura

Imaging single atoms using secondary electrons with an aberration-corrected electron microscope. Nat Mater. 2009; 8(10):808–812.

Griffin

BJ.

A comparison of conventional Everhart-Thornley style and in-lens secondary electron detectors-a further variable in scanning electron microscopy. Scanning. 5 2011; 33(3):162–173.

Asahina

Togashi

Terasaki

High-resolution low-voltage scanning electron microscope study of nanostructured materials. Microsc Anal. 2012; 26(7):12–14.

Masters

Pearson

Glen

Sub-nanometre resolution imaging of polymer-fullerene photovoltaic blends using energy-filtered scanning electron microscopy. Nat Commun. 4 2015; 6(22).

Joy

CS.

Low voltage scanning electron microscopy. Micron. 6 1996; 27(3–4):247–263.

Joy

DC.

Scanning electron microscopy: second best no more. Nat Mater. 2009; 8(10):776–777.

Egerton

Malac

Radiation damage in the TEM and SEM. Micron. 2004.

10.

Stehling

Masters

Zhou

New perspectives on nano-engineering by secondary electron spectroscopy in the helium ion and scanning electron microscope. MRS Commun. 23 6 2018; 8(02):226–240.

11.

Joy

Khanna

Braski

Secondary electron spectroscopy for microanalysis and defect review. Metrol, Insp, Process Control Microlithogr XIII. 1999.

12.

Salim

Hurst

Montgomery

Airborne contamination of graphite as analyzed by ultra-violet photoelectron spectroscopy. J Electron Spectros Relat Phenomena. 1 8 2019; 235:8–15.

13.

Masters

Stehling

Abrams

Mapping polymer molecular order in the SEM with secondary electron hyperspectral imaging. Adv Sci. 6 3 2019; 6(5).

14.

Lorusso

Ohashi

Yamaguchi

“Enabling CD SEM metrology for 5 nm technology node and beyond,” in Metrology, inspection, and process control for microlithography XXXI. San Jose ( CA); 2017.

15.

Labbé

, Faouzi

, Belissard

and Hazart

, Limits of model-based CD-SEM metrology. 34th European Mask and Lithography Conference. Grenoble; 2018.

16.

Cunningham

Inline SEM electron beam review accelerates LTPS display yield ramps. Nanochip Fab Solutions. 2016; 11(1):32–35.

17.

Schönjahn

Humphreys

Glick

Energy-filtered imaging in a field-emission scanning electron microscope for dopant mapping in semiconductors. J Appl Phys. 2002; 92(12):7667–7671.

18.

Schönjahn

Broom

Humphreys

Optimizing and quantifying dopant mapping using a scanning electron microscope with a through-the-lens detector. Appl Phys Lett. 14 7 2003; 83(2):293–295.

19.

Kazemian

Mentink

SAM

Rodenburg

Quantitative secondary electron energy filtering in a scanning electron microscope and its applications. Ultramicroscopy. 2 2007; 107(2–3):140–150.

20.

Rodenburg

Jepson

Inkson

Energy filtered scanning electron microscopy: applications to characterisation of semiconductors. J Phy: Conference Series. 2010.

21.

Orji

Badaroglu

Barnes

Metrology for the next generation of semiconductor devices. Nat Electron. 2018; 1:532–547.

22.

Garrat-Reed

A. J.

and Bell

D. C.

, Energy dispersive X-ray analysis in the electron microscope, 1st ed., Rainforth

, editor, London: Garland Science; 2003.

23.

Reimer

Scanning electron microscopy: physics of image formation and microanalysis. 2nd ed., Berlin: Springer; 1998. p. 527.

24.

Friel

Lyman

CE.

X-ray mapping in electron-beam instruments. Microsc Microanal. 2006; 12:2–25.

25.

Willich

Bethke

“Practical aspects and applications of EPMA at low electron energies,” in Microbeam and nanobeam analysis. Vienna: Springer; 1996. p. 631–638.

26.

Boyes

ED.

Analytical potential of EDS at low voltages. Mikrochim Acta. 2002; 138:225–234.

27.

Belhaj

Jbara

Odof

An anomalous contrast in scanning electron microscopy of insulators: the pseudo-mirror effect. Scanning. 1 1 2000; 22(6):352–356.

28.

Castaing

Application of electron probes to local chemical and crystalographic analysis [PhD Thesis]. Paris: University of Paris; 1951.

29.

Newbury

Ritchie

NW.

Is scanning electron microscopy/energy dispersive X-ray spectrometry (SEM/EDS) quantitative?

Scanning. 5 2013; 35(3):141–168.

30.

Newbury

DE.

Castaing's electron microprobe and its impact on materials science. Microsc Microanal. 2001; 7(2):178–192.

31.

Burgess

Holland

“High spatial resolution energy dispersive X-ray spectrometry in the SEM and the detection of light elements including lithium,” 2013.

32.

Jany

Janas

Krok

Retrieving the quantitative chemical information at nanoscale from scanning electron microscope energy dispersive X-ray measurements by machine learning. Nano Lett. 2017; 17(11):6520–6525.

33.

Friel

Lyman

CE.

Tutorial review: X-ray mapping in electron-beam instruments. Microsc Microanal. 2006; 12:2–25.

34.

Rinaldi

Llovet

Electron probe microanalysis: a review of the past, present, and future. Microsc Microanal. 2015; 21:1053–1069.

35.

Terauchi

Koshiya

Satoh

Chemical state information of bulk specimens obtained by SEM-based soft-X-ray emission spectrometry. Microsc Microanal. 2014; 20:692–697.

36.

Cazaux

Auger microscopy and electron probe microanalysis. In: Boekestein

Pavićević

, editors. Electron microbeam analysis. Mikrochimica Acta. vol. 12, Vienna: Springer; 1992. p. 37–52.

37.

Meitner

Uber den Zusammenhang zwischen β- und γ-Strahlen. L. Z. Physik. 1922; 9(1):145–152.

38.

Auger

Sur les rayons β secondaires produits dans un gaz par des rayons X. C.R.A.S. 1923; 177:169–171.

39.

Auger

The Auger effect. Surf Sci. 1 3 1975; 48(1):1–8.

40.

Palmberg

PW.

Optimisation of Auger electron spectroscopy in LEED systems. Appl Phys Lett. 1968; 13:183.

41.

Chang

CC.

Auger electron spectroscopy. Surf Sci. 3 1971; 25(1):53–79.

42.

Bargeron

CB.

High vacuum scanning electron microscopy as a tool in surface analysis. Laurel ( MD); 1980.

43.

Rades

Hodoroaba

Salge

High-resolution imaging with SEM/T-SEM, EDX and SAM as a combined methodical approach for morphological and elemental analyses of single engineered nanoparticles. RSC Adv. 2014; 4(91):49577–49587.

44.

Taylor

NJ.

Auger electron spectrometer as a tool for surface analysis (contamination monitor). Citation: J Vac Sci Technol. 1969; 6(241).

45.

Willis

Fitton

Skinner

DK.

Study of carbon-fiber surfaces using Auger and secondary electron emission spectroscopy. J Appl Phys. 1972; 43(11):4412–4419.

46.

Baer

Engelhard

Johnson

Surface characterization of nanomaterials and nanoparticles: important needs and challenging opportunities. J Vac Sci Technol A: Vac Surf Films. 9 2013; 31(5):050820.

47.

Powell

CJ.

Growth and trends in Auger-electron spectroscopy and x-ray photoelectron spectroscopy for surface analysis. Citation: J Vac Sci Technol A. 2003; 21:42.

48.

Cressey

Tiny traits cause big headaches. vol. 467, Nature Publishing Group; 2010. pp. 264–265.

49.

Stefaniak

Hackley

Roebben

Nanoscale reference materials for environmental, health and safety measurements: needs, gaps and opportunities. Nanotoxicology. 7 12 2013; 7(8):1325–1337.

50.

Turner

Schreifels

JA.

Surface analysis: X-ray photoelectron spectroscopy and Auger electron spectroscopy. Anal Chem. 1992; 64(12):302–320.

51.

Webb

Michalak

Biteen

High-resolution soft X-ray photoelectron spectroscopic studies and scanning Auger microscopy studies of the air oxidation of alkylated silicon (111) surfaces, 2006.

52.

Macdonald

Waldrop

JR.

Auger electron spectroscopy in the scanning electron microscope: Auger electron images. Appl Phys Lett. 1971; 19:315.

53.

El Gomati

Prutton

Monte Carlo calculations of the spatial resolution in a scanning Auger electron microscope. Surf Sci. 1 4 1978; 72(3):485–494.

54.

Holm

Reinfandt

Auger microanalysis in a conventional scanning electron microscope. Scanning. 1978; 1(1):42–57.

55.

Jacka

Scanning Auger microscopy: recent progress in data analysis and instrumentation. J Electron Spectros Relat Phenomena. 2001; 114-116:277–282.

56.

Hofmann

In: Ertl

, Lüth

, Mills

, editors. Auger- and x-ray photoelectron spectroscopy in materials science: a user-oriented guide,. Berlin; Heidelberg: Springer; 2013. p. 528.

57.

Schwarzer

Hjelen

Backscattered electron imaging with an EBSD detector. Micros Today. 1 2015; 23(1):12–17.

58.

Joy

Newbury

Davidson

DL.

Electron channeling patterns in the scanning electron microscope. J Appl Phys. 1982; 53(8):81–122.

59.

Henini

Scanning electron microscopy: an introduction. III-Vs Rev. 2000; 13(4):40–44.

60.

Venables

Harland

CJ.

Electron back-scattering patterns – a new technique for obtaining crystallographic information in the scanning electron microscope. Philos Mag. 1973; 27(5):1193–1200.

61.

Suga

Asahina

Sakuda

Recent progress in scanning electron microscopy for the characterization of fine structural details of nano materials. Prog Solid State Chem. 1 5 2014; 42(1–2):1–21.

62.

Joy

DC.

Contrast in high-resolution scanning electron microscope images. J Microsc. 2 1991; 161(2):343–355.

63.

Mckay

KG.

Secondary electron emission. Adv Electron Electron Phys. 1948; 1(C):65–130.

64.

Seiler

Secondary electron emission in the scanning electron microscope. J Appl Phys. 1983; 54(11):R1–R18.

65.

Kanaya

Kawakatsu

Secondary electron emission due to primary and backscattered electrons. J Phys D: Appl Phys. 1972; 5:1727–1742.

66.

Joens

Huynh

Kasuboski

Helium Ion microscopy (HIM) for the imaging of biological samples at sub-nanometer resolution. Sci Rep. 17 12 2013; 3:3514.

67.

Everhart

Wells

Oatley

CW.

Factors affecting contrast and resolution in the scanning electron microscope. J Electron Control. 1 8 1959; 7(2):97–111.

68.

Everhart

Thornley

RFM.

Wide-band detector for micro-microampere low-energy electron currents. J Sci Instrum. 1960; 37(7):246–248.

69.

Rodenburg

Jepson

Bosch

Energy selective scanning electron microscopy to reduce the effect of contamination layers on scanning electron microscope dopant mapping. Ultramicroscopy. 8 2010; 110(9):1185–1191.

70.

Haworth

LJ.

The energy distribution of secondary electrons from molybdenum. Phys Rev. 1 7 1935; 48(1):88–95.

71.

Kollath

Zur Energieverteilung der Sekundärelektronen II. Meßergebnisse und Diskussion. Ann Phys. 1947; 436(7-8):357–380.

72.

Röder

Secondary electron microanalysis. Mikrochim Acta. 1992; 107:105–116.

73.

Joy

Prasad

Meyer

HM.

Experimental secondary electron spectra under SEM conditions. J Microsc. 7 2004; 215(1):77–85.

74.

Mil'shteinand

Joy

DC.

Microanalysis using secondary electrons in scanning electron microscopy. Scanning. 7 12 2006; 23(5):295–297.

75.

Bellissimo

Pierantozzi

Ruocco

Secondary electron generation mechanisms in carbon allotropes at low impact electron energies. J Electron Spectros Relat Phenomena. 2019.

76.

Ueno

Sugita

Seki

“Low-energy electron transmission and secondary-electron emission experiments on crystalline and molten long-chain alkanes,” 1986.

77.

Hoffman

Nybergs

Prawert

“High-energy angle-resolved secondary-electron emission spectroscopy of highly oriented pyrolytic graphite,” 1990.

78.

Hoffman

Prawer

Kalish

Structural transformation of diamond induced by 1-keV Ar-ion irradiation as studied by Auger and secondary-electron spectroscopies and total-secondary-electron-yield measurements. Phys Rev B Condens Matter. 1992; 45(22):12736–12745.

79.

Abrams

Dapor

Stehling

Making sense of complex carbon and metal/carbon systems by secondary electron hyperspectral imaging. Adv Sci. 7 10 2019; 6(19):1900719.

80.

Kumar

Schmidt

Schileo

Supplementary information: nano-scale mapping of bromide segregation on cross-sections of complex hybrid Perovskite photovoltaic films using secondary electron hyperspectral imaging in the Scanning Electron Microscope,” 2017.

81.

Ferrón

Vidal

Bajales

Role of HOPG density of empty electronic states above vacuum on electron emission spectra induced by ions and UV photons. Surf Sci. 1 7 2014; 622:83–86.

82.

Farr

Pashneh-Tala

Claeyssens

Characterising cross-linking within polymeric biomaterials in the SEM by secondary electron hyperspectral imaging. (tbc) ACCEPTED. 2019;20.

83.

Kumar

Schmidt

Schileo

Nanoscale mapping of bromide segregation on the cross sections of complex hybrid perovskite photovoltaic films using secondary electron hyperspectral imaging in a scanning electron microscope. ACS Omega. 31 5 2017; 2(5):2126–2133.

84.

Dapor

Monte Carlo simulation of secondary electron emission from dielectric targets. J Phys: Conf Ser. 2012.

85.

Ridzel

Astašauskas

Werner

WS.

Low energy (1–100 eV) electron inelastic mean free path (IMFP) values determined from analysis of secondary electron yields (SEY) in the incident energy range of 0.1–10 keV. J Electron Spectros Relat Phenomena. 2 2019; 12.

86.

Demers

Poirier-Demers

Couture

Three-dimensional electron microscopy simulation with the CASINO Monte Carlo software. Scanning. 5 2011; 33(3):135–146.

87.

Drouin

Couture

Joly

CASINO v2.42 – a fast and easy-to-use modeling tool for scanning electron microscopy and microanalysis users. Scanning. 5 2007; 29(3):92–101.

88.

Azzolini

Morresi

Abrams

Anisotropic approach for simulating electron transport in layered materials: computational and experimental study of highly oriented pyrolitic graphite. J Phys Chem. 2018; 122:10159–10166.

89.

Dapor

Masters

Ross

Secondary electron spectra of semi-crystalline polymers – a novel polymer characterisation tool?

J Electron Spectros Relat Phenomena. 1 1 2018; 222:95–105.

90.

Ono

Kanaya

The energy dependence of secondary emission based on the range-energy retardation power formula. J Phys D: Appl Phys. 1979; 12:619–632.

91.

Shiraishi

Ata

Work function of carbon nanotubes. Carbon N Y. 2001; 39(12):1913–1917.

92.

Joy

DC.

Resolution in low voltage scanning electron microscopy. J Microsc. 12 1985; 140(3):283–292.

93.

Drouin

Hovington

Gauvin

CASINO: a new monte carlo code in C language for electron beam interactions-part II: tabulated values of the mott cross section. Scanning. 7 12 1997; 19(1):20–28.

94.

Mullejans

Bleloch

Howie

Secondary electron coincidence detection and time of flight spectroscopy. Ultramicroscopy. 1993; 52(3-4):360–368.

95.

Knoll

Aufladepotentiel und sekundäremission elektronenbestrahlter körper. Zeitschrift für Technische Physik. 1935; 16:467–475.

96.

Joy

DC.

Control of charging in low-voltage SEM. Scanning. 1989; 11(1):1–4.

97.

Walker

El-Gomati

Assa'd

The secondary electron emission yield for 24 solid elements excited by primary electrons in the range 250-5000 ev: a theory/experiment comparison. Scanning. 9 2008; 30(5):365–380.

98.

Gomati

MMEL

Walker

CGH

Assa'd

AMD

Theory experiment comparison of the electron backscattering factor from solids at low electron energy (250–5,000 eV). Scanning. 1 2008; 30(1):2–15.

99.

Smaragdis

Fevotte

Mysore

Static and dynamic source separation using nonnegative factorizations: a unified view. IEEE Signal Process Mag. 2014; 31(3):66–75.

100.

Gowen

Marini

Esquerre

Time series hyperspectral chemical imaging data: challenges, solutions and applications. Anal Chim Acta. 2011; 705:272–282.

101.

Dobigeon

Brun

Spectral mixture analysis of EELS spectrum-images. Ultramicroscopy. 9 2012; 120:25–34.

102.

Bonnet

Brun

Colliex

Extracting information from sequences of spatially resolved EELS spectra using multivariate statistical analysis. Ultramicroscopy. 7 1999; 77(3-4):97–112.

103.

Muto

Yoshida

Tatsumi

Diagnostic nano-analysis of materials properties by multivariate curve resolution applied to spectrum images by S/TEM-EELS. Mater Trans. 2009; 50(5):964–969.

104.

Shiga

Tatsumi

Muto

Sparse modeling of EELS and EDX spectral imaging data by nonnegative matrix factorization. Ultramicroscopy. 1 11 2016; 170:43–59.

105.

Wan

Abrams

Masters

Mapping nanostructural variations in silk by secondary electron hyperspectral imaging. Adv Mater. 2017; 29:1703510.

106.

Rodenburg

Jepson

MAE

Inkson

Energy filtered scanning electron microscopy: applications to dopant contrast. J Phys Conf Ser. 2010; 209:012053.

107.

Konvalina

Mika

Krátký

In-lens band-pass filter for secondary electrons in ultrahigh resolution SEM. Mater (Basel). 19 7 2019; 12(14):2307.

108.

Stehling

Abrams

Holland

Revealing spider silk's 3D nanostructure through low temperature plasma etching and advanced low-voltage SEM. Frontiers Mater. 1 2019; 5(25).

109.

Khursheed

Scanning electron microscope optics and spectrometers. World Scientific Publishing; 2011. p. 402.

110.

El-Gomati

Zaggout

Jayacody

Why is it possible to detect doped regions of semiconductors in low voltage SEM: a review and update. Surf Interface Anal. 1 11 2005; 37(11):901–911.

111.

Hashimoto

Takeuchi

Sunaoshi

Voltage contrast imaging with energy filtered signal in a field-emission scanning electron microscope. Ultramicroscopy. 2 2020; 209(1).

112.

Herzing

Richter

Anderson

IM.

3D nanoscale characterization of thin-Film organic photovoltaic device structures via spectroscopic contrast in the TEM. J Phys Chemistry C. 21 10 2010; 114(41):17501–17508.

113.

Abrams

Wan

Stehling

Nanoscale mapping of semi-crystalline polypropylene. Phys Status Solidi (C) Current Topics Solid State Phys. 1 12 2017; 14(12).

114.

Zhou

Fox

Maguire

Quantitative secondary electron imaging for work function extraction at atomic level and layer identification of graphene. Sci Rep. 2 2016; 6(16).