DOC PREVIEW
UW-Madison G 777 - Automated Analysis of SEM X-Ray Spectral Images

This preview shows page 1-2-3-4-5-6 out of 17 pages.

Save
View full document
View full document
Premium Document
Do you want full access? Go Premium and unlock all 17 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 17 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 17 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 17 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 17 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 17 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 17 pages.
Access to all documents
Download any document
Ad free experience

Unformatted text preview:

Automated Analysis of SEM X-Ray Spectral Images:A Powerful New Microanalysis ToolPaul G. Kotula,* Michael R. Keenan, and Joseph R. MichaelMaterials Characterization Department, Sandia National Laboratories, P.O. Box 5800, MS 0886,Albuquerque, NM 87185-0886, USAAbstract: Spectral imaging in the scanning electron microscope ~SEM! equipped with an energy-dispersiveX-ray ~EDX! analyzer has the potential to be a powerful tool for chemical phase identification, but the large datasets have, in the past, proved too large to efficiently analyze. In the present work, we describe the application ofa new automated, unbiased, multivariate statistical analysis technique to very large X-ray spectral image datasets. The method, based in part on principal components analysis, returns physically accurate ~all positive!component spectra and images in a few minutes on a standard personal computer. The efficacy of the techniquefor microanalysis is illustrated by the analysis of complex multi-phase materials, particulates, a diffusion couple,and a single-pixel-detection problem.Key words: spectrum image, spectral image, statistical analysis, principal components analysis, X-ray microanal-ysis, information extraction, multivariate statistical analysisINTRODUCTIONIt is a conundrum that faces microanalysts every day: Howdo you perform a complete and comprehensive survey ofthe chemistry of a microstructure in a reasonable time?Traditional solutions in the scanning electron microscope~SEM! would include taking an image and then collectingX-ray spectra from a series of points. This method reliesheavily on the operator’s ability to identify chemically dis-tinct regions. In addition, for a complex microstructure, thiscould take a considerable amount of time to do correctly.On the positive side, if care is taken, these “point spectra”can be quantified. An alternative method that has been usedwidely is X-ray mapping ~Goldstein et al., 1992!, where awindow about a range of X-ray energies is integrated anddisplayed as an image. This has the advantage that qualita-tive elemental distributions from an area of a microstruc-ture can be visualized. The disadvantages are that maps aregenerally not quantitative, can be susceptible to artifacts~Newbury, 1997; Newbury and Bright, 1999!,relyonfore-knowledge of the elements to map, and cannot discernelemental correlations ~e.g., Al versus Al2O3!. More recently,with increases in computer CPU speed, memory, and stor-age space, X-ray spectrum-imaging systems have becomeavailable. A spectral image is a two-dimensional array ofpoints in the microstructure with a complete X-ray spec-trum from each point ~Legge and Hammond, 1979; Jeanguil-laume and Colliex, 1989; Mott et al., 1995; Anderson, 1998!.Spectral images potentially overcome the shortcomings ofpoint analyses ~e.g., how to pick the points! by being animaging technique that allows the analyst to qualitativelyanalyze and perhaps quantify the spectra. It should be notedthat for the spectral image to have as much spatially re-Received November 6, 2001; accepted June 5, 2002.*Corresponding author. E-mail: [email protected]. Microanal. 9, 1–17, 2003DOI: 10.1017/S1431927603030058MicroscopyANDMicroanalysis© MICROSCOPY SOCIETY OF AMERICA 2003solved chemical information as possible, the pixel size shouldbe somewhat finer than the X-ray generation volume ~Ly-man, 1986; Goldstein et al., 1992!. In addition, spectralimages have potential to overcome the problems with map-ping ~e.g., choosing the elements to map! because there is acomplete spectrum from each pixel. The problems withspectrum imaging up until recently were that there were notools for extracting the chemically relevant informationfrom the massive amount of data ~.65 million individualdata points for a 256 3 256-pixel 3 1024-channel spectralimage!. The tools that have been available for analyzingspectral images amount to ones that map after the fact andthen allow summing of spectra that have been thresholdedfrom maps ~Mott et al., 1995; Mott and Friel, 1999!. Addi-tionally, it can take a considerable amount of time to collectspectral images with individual spectra containing sufficientcounts to perform even a qualitative, let alone quantitative,analysis of a given spectrum. It is for these reasons that newanalysis methods are needed to objectively analyze spectralimages ~Anderson, 1999, 2000; Kotula et al., 1999; Kotulaand Keenan, 2000!.One inherent problem with spectrum imaging is thatthe data sets cannot readily be visualized in their entirety.This point can be appreciated by looking at the three-dimensional representation of the spectrum image data setin Figure 1. Each element of the cube has associated with itthree dimensions: x-position, y-position, and X-ray energychannel. Further complicating the complete analysis is thefact that much of the spectrum image is redundant. Thiswould include regions of the spectra where no peaks arefound as well as pixels from the same chemical phase.Redundancy is, in fact, absolutely necessary for the analysismethod described below to work. The single greatest prob-lem is that for a robust unbiased analysis, no assumptionscan be made as to what may or may not be present. In otherwords, you cannot blindly ignore parts of the spectra just incase there really is a peak there.The key requirements of a robust system for the auto-mated analysis of spectral image data sets are: no assump-tions about the absence or presence of any constituent; theability to handle noisy data; the ability to handle significantspectral overlap; spectral image processing times of thesame order or much less than the acquisition time for thedata; and quantitative agreement between raw and recon-structed data. The system developed in the present workmeets all of the above requirements.MATERIALS AND METHODSMultivariate Statistical AnalysisMultivariate statistical analysis ~MSA! describes a generalset of techniques that are utilized for analyzing data setssuch as opinion polls, new-drug trials, and series of spectra~Harman, 1976; Malinowski, 1991; Geladi and Grahn, 1996!.For all of these applications of MSA there are from severalto thousands of variables in the data sets, and the goal is toexplore the variation in multiple dependent variables simul-taneously. For X-ray spectral image data, the variables arethe energy channels and will typically number from 1000 to4000. The end product of MSA is a transformation from theraw data, with all its inherent


View Full Document

UW-Madison G 777 - Automated Analysis of SEM X-Ray Spectral Images

Documents in this Course
Load more
Download Automated Analysis of SEM X-Ray Spectral Images
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Automated Analysis of SEM X-Ray Spectral Images and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Automated Analysis of SEM X-Ray Spectral Images 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?