NAN/PHY/MSE 546 (Venables) Sect 3.5

Notes for NAN/PHY/MSE 546 Sect 3.5 (Venables)

Arizona Board of Regents for Arizona State University and John A. Venables

Click to download this document in Microsoft Word Format


Lecture notes by John A. Venables, revised for Spring 2009. Latest version 26 August 2012, reformatted.

3.5 Microscopy-Spectroscopy: SEM, SAM, SPM, etc

Refs: Review and specialist articles referred to in the text. This section has been expanded in my book, section 3.5, where more references are given. More recent references to the ASU group's work are given at http://venables.asu.edu/research/pub_2000.html

  • Scanning Electron and Auger Microscopy
  • Scanning Electron Microscopy is now a routine tool which has been extensively commercialised; it is almost essential for all experimental scientists to have access to a high performance, but non-UHV, version. Indeed, the size of new devices is such that, without SEM, the structures which have been produced cannot even be seen. Electron beam lithography is an important fabrication technique, particularly for mask manufacture, and SEM is used routinely in quality control of the devices produced in this (and other) ways.

    In a UHV, clean surface environment, there are several SEM-based techniques which are surface sensitive at the ML and sub-ML level. The main SEM signal is based on collecting secondary electrons, which typically form a large proportion of the emitted electrons (see diagram 70). In several papers, we have shown that the low energy secondary electron signal is very sensitive to work-function and other surface-related changes; by biassing the sample negatively we can be sure to collect these Biassed Secondary Electron Images (b-SEI), and ML and multilayer deposits can be readily visualised and distinguished (see e.g K. Yagi chapter 13 in Buseck et al (1988), M. Futamoto et al, Surface Sci 150 (1985) 430). UHV-SEM is particularly useful when combined with AES and RHEED (J.A. Venables et al, Phil. Trans. Roy. Soc. Lond A318 (1986) 243).

    Scanning Auger Microscopy (SAM) is the child of AES and SEM. A fine primary beam is used, scanned sequentially across the sample at positions (x,y) as in SEM, and the emitted electrons are energy analysed as in AES. We can now (attempt to) perform various types of analysis, such as a Spot mode analysis (scan the Energy E at fixed (x0, y0)), an Energy-selected line scan (scan x at fixed y0 and analysis energy EA), or an Energy-selected map or image (scan x and y at analysis energy EA). These attempts are subject to having a long enough data collection time and high enough beam current to achieve a satisfactory SNR. Some examples are shown in diagrams 117-119. In particular, you can notice that the SNR of the b-SE line-scans is very good, and the b-SE images are reasonably clear; next good are the energy-selected line scans, and most difficult/ time-consuming are the energy-selected images.

    Following on the discussion in section 3.4, we can think about the extraction of Auger data from energy-selected line scans and images, and the quantification of such information. We need to use ratio techniques, for several reasons. First, typical samples are not flat, and may be extremely rough (diagrams 117-118), or can involve changes in backscattering factor. This leads to variations in (rsecθo.T); such changes in the Auger peak channel (A) can often go in the opposite direction from what is expected, for example in diagrams 117 and 119.

    Second, one needs to have Auger information which (if possible) is independent of changes in the background spectrum, and of beam current fluctuations. The first of these is not entirely possible, but one can make a good attempt. By taking line scans or images at one or more energies in the background above the peak (B and maybe C), then various difference techniques can be tried to extract true Auger information. The ratio (A-B)/(A+B) is most commonly used; this is a quasi-logarithmic measure of Auger intensity, being a first approximation to (E.N(E))-1d(E(N(E))/dE, in the fixed retard ratio mode.

    The simplest linear measure is based on an extrapolation of the background from C to B to A. Assuming these channels are equally spaced, then the Peak to Background Ratio (P/B) = (A - 2B + C)/(2B - C). We can see from diagrams 117 and 119 that this measure is usually noiser that the ratio based on the two channels A and B. This is because one is in effect measuring the background slope at each pixel, as well as the peak height. If you are unhappy about the noise level in images, you can always trade SNR for image resolution by digital image processing, as in diagrams 118 and 123. The result is not always very pleasing, nor even necessary, since the eye does this for you anyway. If is amazing how well our eyes/brain are able to extract feature information, eg. the straight lines in Fig 118(b), from very noisy data.

  • Image Analysis of 'Real World' Samples
  • A particular problem faced by analysts in the ‘real world’ is that their samples contain many different elements; they may have rough surfaces, and this may interfere with quantitative analysis. However, they may not be so concerned about quantitative information at every point in the image; association of specific types of qualitative information with each point may be more informative. This type of ratio technique has been developed by several groups, and an illustration is given in diagram 120 (R. Browning et al. J. Vac. Sci. Tech. A2 (1984) 1453). Prutton’s group has furthered these techniques, originally developed for satellite imaging by NASA/JPL, as described in several books and papers referenced in my book, section 3.5.2.

    In such an approach an SEM picture is taken of the whole field of view (diagram 120, fig 1), and a survey spectrum is taken from this area. This shows many peaks, some of them very small (fig 2). The spectrum is used to indentify energies (channels) at which information will be recorded at each point on the image, typically a peak and background channel at higher energy, for each of the elements of interest. This information is collected and stored digitally. Before images are made with this information, scatter diagrams, such as figs 3, 4 and 5 are constructed. These show that the ratio data cluster in well defined regions of the scatter diagram, and it is easy for the analyst to indentify the clusters as particular phases, at least tentatively.

    At this stage, one can put a ‘mask’ over the data and use all the data which fall within this mask to form an image. In the case of diagram 120, an unknown phase was identified which contained Ti, Si, some S and also P. By setting limits on the various ratios, an image can now be produced from the stored data set, which shows the spatial distribution of this particular range of compositions. In this case (fig 6) it was shown that the ‘phase’ was formed in the reaction zone between the SiC fiber and the Ti alloy which made up the composite material.

    These ‘associative’ or pattern recognition techniques are very powerful, but they do require that a lot of effort be expended on a particular small area. This can result in radiation or other forms of damage, and it is always possible that you could have got the answer you really needed faster by another technique. At high spatial resolution there are various artefacts associated with sharp edges, essentially because part of the information comes from back-scattered electrons.

  • Towards the highest spatial resolution: SEM/STEM and Scanned Probe Microscopy/ Spectroscopy
  • The development of SEM/STEM and AES/SAM at the highest resolution has been pursued at ASU in what has become known as the MIDAS project, a Microscope for Imaging, Diffraction and Analysis of Surfaces. (The thought of turning base metals into gold is not exactly original, and many other projects on totally different topics have used this acronym!). Diagrams 121-123 illustrate this project (see G.G. Hembree and J.A. Venables, Ultramicroscopy 47 (1992) 109 and refs quoted). More recently, Dr Liu, I and other coworkers have reviewed this work in three articles (see http://venables.asu.edu/research/pub_2000.html that feature the MIDAS project.

    The innovation as regards electron optics is to use the spiralling of the low energy electrons in the high magnetic field of the objective lens to contain the secondary and Auger electrons close to the microscope axis. These electrons are further controlled by auxiliary magnetic fields (parallelizers on diagram 121) in the bores of the lens, and by biassing the sample negatively. A special combination Wien filter/ deflector is then used to deflect the low energy electrons off axis through a right angle, while keeping the 100 keV beam electrons on axis. The low energy electrons then enter a commercial CHA. Because of the spiralling properties in the high B field, quite a large proportion of the emitted electrons can be collected. This higher collection angle compensates for the lower yield at higher beam energy, and the smaller current available in the fine probe.

    The quality of the spectra obtained is relatively high, both with respect to energy resolution (diagram 122 a) and to sensitivity (diagram 122 b). Auger mapping is obtained by taking images A and B and using ratios (A-B)/(A+B) as in section (a) above. Diagram 123 shows the comparison of the b-SE image, with good SNR, and the Auger image, with relatively poor SNR, even after smoothing. For imaging, we have to be clear about distinctions between 'image' and 'analytical' resolution, as discussed in the reference given. This is because of the non-local nature of the Auger signal, firstly from backscattered electrons as discussed already, but also at high spatial resolution because of the finite Auger attenuation length, and non-local excitation. Image resolutions below 3nm have been obtained on bulk samples in this instrument.

    In the case of thin film substrates, we can almost eliminate the backscattering contribution, so that the image and analytical resolutions converge on the image resolution, which in practice may well be limited by the probe size. Such high resolutions are of interest in small particle research, particularly in catalysis. Here, the old joke used to be that if you can see the particle in an electron microscope, then it was already too large to be a useful catalyst. Analysis of such a particle is even harder, especially if one is interested in minority elements. Work on such samples has been pursued by J. Liu et al, Ultramicroscopy 52 (1993) 369 and refs quoted.

    Here it is not so clear what the quantification routine ought to be, and in practice Auger information has been portrayed using the raw A and B images for different elements. Even for small Ag particles, backscattering effects can be seen in the intensity of carbon Auger peak images. Even more interesting is that, for particles of size at or below the Auger attenuation length, the number of atoms in the cluster is measured by the integrated intensity of the particle, rather than the image size of the particle. On this basis, it was concluded that particles containing less than 10 Ag atoms had been observed.

    We should note that, because of the high yield for Ag MNN Auger electrons, this is a favorable case; we are still quite a way from detecting arbritary minority species on such small particles. Moreover, we are much more likely to be able to detect them first with a high SNR, qualitative, technique, such as b-SEI, than with low SNR, quantitative AES/SAM. There are more recent examples of this coming from MIDAS. For example, Oxygen KLL at 505 eV has a relatively low Auger yield. Small oxide particles on copper can be seen very readily in high resolution b-SE images, and indeed in the shape of the (secondary electron) background, whereas wide beam Auger declares the surface to be clean (K.R. Heim et al, J. Appl. Phys. 74 (1993) 7422).

    The use of spectroscopy in STM (scanned probe) instruments is being discussed in a student talk. In principle, such spectroscopic information can allow one to identify surface atomic species in favorable cases, if not in general. This is because the STM/STS techniques probe the valence and conduction bands, which are not chemical specific in the same sense as AES/SAM. This is not unlike the SEM/SAM distinction. The STM may well be able to make 'chemical' identifiction possible out of a range of possibilities, due to a combination of atomic resolution and changes of contrast due to electronic/ size effects, and in particular due to a high SNR. One of the many amazing positive features of STM is that the probing current is also the signal, which may be between 1nA and 1pA. In AES/SAM used on a microscopic scale, the probing current may be between 10nA and 10pA, but the collected current is down to maybe 100,000 times smaller than the probe current, which does not do good things for the SNR. Thus one typically has to think very carefully about what information is wanted and is practicable to obtain. Some of the examples given are close to the current technical limits.


    Return to Lecture list

    Return to Course Home Page