Quantitative analysis of scanning electron micrographs

Scanning electron microscopy (SEM) provides a considerably greater depth of field than optical microscopy at the same magnification. It is therefore particularly suitable for the examination of specimens with irregular surfaces. This review of procedures for quantitative measurements on SEM images falls into two parts. The first deals with mensuration; that is, with the determination of the lengths, dihedral angles and other properties of specific features in the image. Particular emphasis is given to means of correcting for non-linearity and other distortions caused by instrumental defects. The second part considers the analysis for bulk properties such as volume fraction and boundary area per unit volume. In the case where observations are made on a planar section, these properties can be determined by use of the standard stereological relationships after making a computable correction (where necessary) for the tilt of the specimen and non-linearities in the scan. However, for irregular surfaces (for example, those produced by fracture) rigorous estimations are not possible without certain assumptions which, unfortunately, are in many cases unrealistic.     

Quantitative comparison of image contrast and pattern between experimental and simulated high-resolution transmission electron micrographs

Aiming to determine the contrast mismatch factor i.e. the Stobbs factor between the experimental and simulated high-resolution transmission electron micrographs, we have systematically compared the experimental images and simulations of a cleaved silicon sample for a series of focal settings and specimen thicknesses. For zero-loss energy filtered images, a mismatch factor of about 1.5-2.3 is measured for the image contrast, where the mismatch factor is focal dependent and higher mismatch appears around the focus value of 10nm. Attention is also given to the effects of the sample vibration and drift to the image contrast and pattern of the high-resolution micrographs.     

Quantitative and densitometric considerations in bas-relief electron micrographs

Two methods for bas-relief electron micrographs are described, one to produce transparencies for projection and one to produce prints for publication. The basis of these methods is a controlled misregistration of a positive image and a negative image, both on film, to increase contrast and produce a “bas-relief” effect. This gives the final image a three-dimensional appearance. No apparent loss of resolution results and no artificial amendment is required. To produce transparencies, a contact film positive is produced from the negative and the two are sandwiched in a slide mount with the appropriate amount of lateral displacement to give the effect. To produce prints, the film positive and the film negative are used to sequentially expose the same sheet of photographic paper prior to development to realize the effect. The bas-relief induced by the misregistration of the film plates causes apparent shadows and highlights that produce the apparent height that is seen in the enhanced images.     

Exposure level and image quality in electron micrographs

Single electrons of the energies used in transmission electron micrography can render developable one or more emulsion grains. From this fact the following deductions can be made.

• (a). Photographic contrast in an electron micrograph is a function of the mean density only.
• (b). The signal/noise ratio in an electron micrograph improves with increasing exposure, irrespective of the means adopted to enable an increased exposure to be used.
• (c). If the exposure of the specimen is limited to some maximum then the visibility of noise-limited detail in prints representing a given total magnification is independent of the negative material used in the microscope.

Experimental tests based largely on electron micrographs are presented which provide good support for these deductions. It is generally concluded that the behaviour patterns examined in this work should always be considered when limitations are being encountered in the image quality of electron micrographs.     

Quantitative image and electron microprobe analysis

Image analysis in the scanning electron microprobe analyser by use of a suitable digital scan drive/computer interface to an LSI 11/23 computer and dual 5.2 Mbyte hard disc system is described. The image processing of 1024 × 1024 pixel and 256 grey level resolution is performed to obtain size and particle coordinate positions. At these particle coordinates, x-ray analysis is automatically performed by suitable interfacing to an EDAX PV 9100 EDX system. Software has been developed for electron microprobe analysis to interface with the spacial image analysis information. By use of the video range for binary image detections and correct choice of the elemental analysis parameters, composition specific image analysis of second phase features in microstructures can be obtained. The composition specific information can be expressed in terms of size related to composition range, or composition related to size range for up to 10 elements simultaneously.     

Quantitative analysis of backscattered-electron contrast in scanning electron microscopy

Backscattered-electron scanning electron microscopy (BSE-SEM) imaging is a valuable technique for materials characterisation because it provides information about the homogeneity of the material in the analysed specimen and is therefore an important technique in modern electron microscopy. However, the information contained in BSE-SEM images is up to now rarely quantitatively evaluated. The main challenge of quantitative BSE-SEM imaging is to relate the measured BSE intensity to the backscattering coefficient η and the (average) atomic number Z to derive chemical information from the BSE-SEM image. We propose a quantitative BSE-SEM method, which is based on the comparison of Monte–Carlo (MC) simulated and measured BSE intensities acquired from wedge-shaped electron-transparent specimens with known thickness profile. The new method also includes measures to improve and validate the agreement of the MC simulations with experimental data. Two different challenging samples (ZnS/Zn(O_xS_1–x)/ZnO/Si-multilayer and PTB7/PC₇₁BM-multilayer systems) are quantitatively analysed, which demonstrates the validity of the proposed method and emphasises the importance of realistic MC simulations for quantitative BSE-SEM analysis. Moreover, MC simulations can be used to optimise the imaging parameters (electron energy, detection-angle range) in advance to avoid tedious experimental trial and error optimisation. Under optimised imaging conditions pre-determined by MC simulations, the BSE-SEM technique is capable of distinguishing materials with small composition differences.     

Spatial filtering of electron micrographs of negatively stained alpha-amylase crystals

Images of the alpha-amylase molecule from pig pancreas have been obtained in two projections by the application of spatial filtering techniques to electron micrographs of negatively stained microcrystals of the enzyme. The Fourier filtering was performed on a PPD 11/40 digital computer after microdensitometry on an Optronics P-1000 photoscan system. The set of procedures and programs we employed are described herein. We compared these images with equivalent images obtained by three-dimensional X-ray diffraction analysis using conventional isomorphous replacement at approximately twice the resolution. We find that there is quite good agreement between the two kinds of images, and that a number of gross structural features compare quite well. We conclude that the combination of electron microscopy with digital spatial filtering when applied to protein microcrystals yields very respectable results.     

A method for increasing the photographic contrast of electron micrographs

Emulsion-developer combinations used regularly in conventional electron micrography yield an approximately linear relationship between density (D) and exposure (E) up to quite high densities (> 2.0). This is due to the fact that over the corresponding exposure range the density yield per electron is approximately constant. A practical consequence is that the slope of the D/log₁₀ curve, termed the contrast, is a function of D only, i.e. it is independent of the emulsion-developer combination. To achieve higher than normal contrast requires that the density yield per electron should increase with exposure over some exposure range, at least. This effect may be obtained by using what are known commercially as ‘lith’ films and a system of infectious development. The high contrast of such films results from an acceleration of development rate when the developer oxidation products reach a critical concentration. Sensitometric data are presented for electron exposures which show that for average densities of ～ 1.0 in a micrograph a six-fold increase of contrast can be obtained with a ‘lith’ system. It is also shown that though the special development effect gives higher granularity than that obtained at the same exposure with a conventional system, the signal/noise ratio is not adversely affected. Practical demonstrations of the effectiveness of the ‘lith’ system are presented.     

A minicomputer system for image processing of electron micrographs

This paper describes how a PDP 11/34A computer system was set up to handle research problems in computer image processing of electron micrographs. This system is referred to as CIPRES, for Computer Image Processing and Reconstruction System (pronounced the same as cypress, the name of an evergreen tree). The RSX-11M operating system is used. Various execution speeds are given. For instance, a disk-based fast Fourier transform or FFT (complex to complex, 128 X 128 pixels) runs in three minutes. This report may be useful to anyone wishing to set up a similar computer system.     

Quantitative analysis of image contrast in phase contrast STEM for low dose imaging

This work quantitatively evaluates the contrast in phase contrast images of thin vermiculite crystals recorded by TEM and aberration-corrected bright-field STEM. Specimen movement induced by electron irradiation remains a major problem limiting the phase contrast in TEM images of radiation-sensitive specimens. While spot scanning improves the contrast, it does not eliminate the problem. One possibility is to utilise aberration-corrected scanning transmission electron microscopy (STEM) with an Ångstrom-sized probe to illuminate the sample, and thus further reduce irradiation-induced specimen movement. Vermiculite is relatively radiation insensitive in TEM to electron fluences below 100,000 e⁻/Ų and this is likely to be similar for STEM although different damage mechanisms could occur. We compare the performance of a TEM with a thermally assisted field emission electron gun (FEG) and charge coupled device (CCD) image capture to the performance of STEMs with spherical aberration correction, cold field emission electron sources and photomultiplier tube image capture at a range of electron fluences and similar illumination areas. We show that the absolute contrast of the phase contrast images obtained by aberration-corrected STEM is better than that obtained by TEM. Although the STEM contrast is higher, the efficiency of collection of electrons in bright field STEM is still much less than that in bright field TEM (where for thin samples virtually all the electrons contribute to the image), and the SNR of equivalent STEM images is three times lower. This is better than expected, probably due to the absence of a frequency dependent modulation transfer function in the STEM detection system. With optimisation of the STEM bright field collection angles, the efficiency may approach that of bright field TEM, and if reductions in beam-induced specimen movement are found, STEM could surpass the overall performance of TEM.     

Use of spot-scan procedure for recording low-dose micrographs of beam-sensitive specimens

Electron micrographs of monolayer crystals of paraffin have been recorded with a spot-scan mode of imaging which uses a small 50 nm diameter moving beam. In comparison with normal stationary beam images using 5 μm illuminating beams, the spot-scan micrographs show a higher and more consistent contrast from the 3.8–4.2 Å hydrocarbon chain spacings. On average the improvement in contrast is twofold, but this still leaves scope for further improvement: the best spot-scan images still do not quite reach the level of contrast calculated theoretically from electron diffraction. We believe that the cause of the low contrast in paraffin images must be specimen motion caused by radiation damage rather than a charging effect, for two reasons. First, it does not occur in control images of vermiculite, a non-beam-sensitive mineral, when treated identically. Secondly, although charging might still be a problem with paraffin, when images are taken with the objective aperture in the column, a procedure which is normally expected to reduce charging, no improvement in contrast is found. Thus we think that the use of the small beam minimizes the effect of specimen motion on image contrast by minimizing the specimen area exposed at each instant and therefore the resultant image blurring. Further improvements with even smaller illumination beam diameters might be expected.     

Quantitative analysis of micrographs by computer graphics.

A computer-aided graphical analysis system is presented which can be used to quantify several types of information from micrographs of cross-sectioned peripheral nerves. A simple input consisting of eight points, corresponding to two diameters of each axon and two for each axon plus myelin, is used to compute statistical and numerical estimates of these diameters, the myelin thickness being expressed as a function of axon diameter, and both the spatial position and distribution of nerve fibres within the bundle. The computer programs presented and summarized here describe the procedure for digitizing the input data and the subsequent computations for editing and analysis. The general applicability of this approach for automated analyses of other anatomical areas is also presented.     

Computer image processing of electron micrographs of biological structures with helical symmetry

Methods are described for the analysis of electron micrographs of biological objects with helical symmetry and for the production of three-dimensional models of these structures using computer image reconstruction methods. Fourier-based processing of one- and two-dimensionally ordered planar arrays is described by way of introduction, before analysing the special properties of helices and their transforms. Conceiving helical objects as a sum of helical waves (analogous to the sum of planar waves used to describe a planar crystal) is shown to facilitate analysis and enable three-dimensional models to be produced, often from a single view of the object. The corresponding Fourier transform of such a sum of helical waves consists of a sum of Bessel function terms along layer lines. Special problems deriving from the overlapping along layer lines of terms of different Bessel order are discussed, and methods to separate these terms, based on analysing a number of different azimuthal views of the object by least squares, are described. Corrections to alleviate many technical and specimen-related problems are discussed in conjunction with a consideration of the computer methods used to actually process an image. A range of examples of helical objects, including viruses, microtubules, flagella, actin, and myosin filaments, are discussed to illustrate the range of problems that can be addressed by computer reconstruction methods.     

Intensity gradient techniques for orientation analysis of electron micrographs

The variation in intensity between neighbouring picture points on an electron micrograph may be used to define an intensity gradient vector. This vector is related to the orientation of the feature at that point, and may be compared with similar vectors at all other points within the micrograph to obtain a quantitative measure of anisotropy. Alternatively, the results may be displayed graphically as a rosette diagram. This paper compares the results from using a simple ‘5-point’, i.e. the nearest neighbours with those from alternative formulations for the intensity gradient vector. Generally, the increased number of points improves the shape of the rosette diagram by minimizing the likelihood of pronounced ‘spikes’ at certain critical angles which is a feature of the ‘5-point’ method. The improved analysis requires an extra 60% computing time, which may be of significance if direct on-line work is contemplated. Some examples of micrographs of soil microfabric are presented to illustrate the different methods.