TOF‐SIMS Data: Image Analysis - mikee9265/SIMS-Wiki GitHub Wiki

Image Processing

The world of image manipulation and analysis is a wide one (Russ 2015), more of it applicable to TOF-SIMS images than is in common use. Features can be counted and sized. Borders between regions can be defined. Noise can be reduced.

Sophisticated techniques can be used to define regions of interest (ROIs). Intensity histograms can reveal multimodal distributions among pixels that in turn can also be used to define ROI. Edges can be sharpened. Repeat distances in sample patterns can be quantified. In addition, there is a great deal of prior art in image representation for best presentation of nonoptical images.

Image Fusion

TOF-SIMS does indeed image chemical species on surfaces, but the lateral resolution of the method is hampered by instrumental limitations and by the static SIMS limit. Imaging using an ion beam that does not damage the sample is limited by the available spot sizes of those ion beams. Imaging using ion beams with the best spot sizes is limited by the damage they cause and the resulting static SIMS limit. Both are limited by ion yields of the species of interest, which means that even with an ion beam that does not leave a damaged surface, one can end up consuming a significant portion of the sample and be far from surface sensitive in the creation of a single image.

Image fusion is the trick of using information from multiple sources to construct an image. An example of this from the past was the fusion of X-ray photoelectron spectroscopy (XPS) results with atomic force microscopy (AFM) data (Artyushkova, Farrar, and Fulghum 2009). More to the current point, the higher lateral resolution of a secondary electron microscope (SEM) image has been used to sharpen the features of a relatively blurry ion image from a TOF-SIMS analysis (Milillo et al. 2015).

The trick is to use a multivariate statistical method to find the covariance in the data sets, and thus to associate the chemical features in one data set with the sharper contrast of the other. It is not clear that this will be easy to do on a regular basis going forward. There are two significant problems. First, SEM images or AFM images do not necessarily have contrast mechanisms that correspond to those in the secondary ion images. The techniques are after all measuring quite different properties of the surfaces in question. The applicability of the technique is liable to be sample specific. Second, it is entirely possible to create artifacts in the resulting images. Artifacts have been documented in the fusion of test data sets (Tyler 2015). Finally, a great effort must be made to line up the images prior to the attempt to fuse the images. As noted above, the speed with which calculations can be performed greatly increases its usefulness. As long as image fusion remains a difficult and time-consuming process, it will have little use. On the other hand, improvements in computing speeds and processes may lead to the automation of the most difficult parts of the image fusion process.