Large-Area Quantitative Phase Mapping in the Scanning Electron Microscope
Keith Thompson T ermo Fisher Scientifi c , 5225 Verona Rd. Madison , WI 53711
Keith.thompson@thermofi
sher.com
Abstract: Silicon-drift-based energy-dispersive X-ray spectrometers and spectral imaging data storage have expanded the possibilities for collecting high information-content X-ray data in a modern scanning electron microscope. The implementation of principal component analysis for phase determination combined with an automated stage and the capability to stitch together a series of sequentially collected images enables the collection and analysis of large-area quantitative phase maps for detailed materials characterization on practical time scales. The new approach provides signifi cant improvements, particularly in speed and completeness of analysis. Examples show detailed quantitative analyses of a mineralogical sample and a complex ceramic composite.
Introduction
Scanning electron microscopy (SEM) has long been a mainstay scientifi c instrument for the study of morphology and composition on the micro- and nano-scale. Energy- dispersive X-ray spectrometry (EDS) systems provided the means to determine the phases present (via inference from elemental X-ray analysis) in materials examined in the SEM. Each advance in EDS technology has supported a concurrent expansion of the information available to the materials analyst. T e advent of silicon driſt detector (SDD) systems and spectral imaging data storage facilitated EDS acquisitions to produce element maps and other data [ 1 ]. T e spectrum image stores an entire spectrum at each pixel in the image, and various types of soſt ware are used to extract useful information from this data cube. T e addition of a motorized stage enabled multiple scan rasters to extend small-scale analysis over relatively large areas of the material (length scales of several millimeters). T e next logical step involved automated techniques that use electron image capture followed by EDS analysis in regions of signifi - cance selected from the contrast within the grayscale electron image. In this way, large-area electron imaging followed by EDS chemical typing gave rise to automated quality control and materials inspection tools. T e most prominent example may be in mining applications [ 2 , 3 ].
This article reviews the concept of large-area quanti- tative phase mapping using a multivariate statistical approach known as principal component analysis (PCA) [ 4 – 6 ]. In this case the term “large area” refers to an analysis area on the order of 1 mm 2 or greater. Phase mapping has the advantage of greatly reduced analytical ambiguity because the phase identifi cations are developed by a user-independent algorithm residing within the “engine” of the EDS system. Typically, the number of phases will not exceed a maximum of 15 to 20 within a material, and more normally the number is in the 5 to 8 range. Even when the results contain up to 20 phases, each with a well-defi ned chemistry, the data are presented in a compact enough form for direct interpretation. A set of X-ray element maps via EDS, by
36
contrast, may contain up to 20 elements spread over a million pixels and across many phases. T is presents a more daunting analytical interpretation: which elements are bound to which other elements in which phases.
The concept of large-area phase mapping is not new
[ 7 – 9 ]. However, previous techniques employed to create phase maps render large-area phase mapping impractical. Traditional element-based phase mapping requires at least an order of magnitude increase in the number of X-rays per pixel as compared to straightforward EDS element mapping, where roughly 5,000 X-rays per pixel—the sum of all X-rays collected in the spectral imaging data set divided by the total number of pixels—are needed for a robust element map. Additionally, every element, including those at trace levels, should be properly identifi ed for the validity of this technique to be optimized. Similarly whole spectrum-based phase mapping [ 10 ] has the same limiting requirements because several thousand X-rays per pixel are required to properly compare spectra from adjoining pixels with statistical certainty.
Fortunately, the principal component-analysis technique described in this article requires far fewer X-rays per pixel than that required for basic EDS X-ray element mapping. It is well-established that PCA requires only 25 to 150 X-ray counts per pixel to yield good results, and the implemen- tation described here involves—actually requires—no user interaction. As with most spectrum imaging systems, it also has the additional advantage of requiring no prior knowledge of the elements present.
Figure 1 : BSE image at 15 kV of crushed rock “grains” mounted in bakelite showing several mineral phases. Total fi eld width is 1.03 mm. Red box shows a single “grain” similar to that analyzed in later fi gures.
doi: 10.1017/S1551929517000153
www.microscopy-today.com • 2017 March
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76