Applications of Multivariate Statistical Analysis (MSA) in Microanalysis Page: 1 of 4
This article is part of the collection entitled: Office of Scientific & Technical Information Technical Reports and was provided to UNT Digital Library by the UNT Libraries Government Documents Department.
Extracted Text
The following text was automatically extracted from the image on this page using optical character recognition software:
ORNL/CP-101036
APPLICATIONS OF MULTIVARIATE STATISTICAL ANALYSIS
(MSA) IN MICROANALYSIS
Ian M. Anderson
Oak Ridge National Laboratory, PO Box 2008, Oak Ridge, TN 37831-6376 USA
andersonim@ornl.gov
Recent improvements in computer hardware and software for the acquisition,
storage and analysis of series of spectra and images allow for a change in strategy
for quantitative microanalysis. For example, in the area of X-ray microanalysis,
whereas compositional analysis and elemental distributions have been traditionally
performed using point microanalyses and simple intensity mapping from a ROI,
respectively, the two tasks are now routinely performed simultaneously through X-
ray spectrum-imaging, where full spectra are acquired from pixels in a two-
dimensional array of points on the specimen. Commercially available software now
allows for the acquisition and storage of such spectrum-images, perhaps comprising
as much as 100 MBytes of data or more.' A variety of post-acquisition processing
tools are provided by the developer to allow the extraction of both X-ray intensity
maps, with or without rudimentary background subtraction, or full spectra from
pixels of interest. In order to maximize the extraction of information from these
large data sets, a number of linear and nonlinear methods are currently being
explored that identify statistically significant variations among the series of spectra
without a priori assumptions about the content of the data set.2 Among these
methods, linear multivariate statistical analysis (MSA) has a number of significant
advantages, including its comprehensiveness, since all spectral variations distinct
from the Poisson noise level are identified, and its broad applicability to a variety of
microanalytical techniques.34 MSA also preserves the integrity of the raw data,
since the results of the analysis are not contingent upon input from the analyst
there is a one-to-one correspondence between the raw data and the MSA results.
An application of MSA to a low-voltage EDS spectrum image is shown in FigureMAR 0 1 J
A preliminary analysis has been performed on this semiconductor chip specimen,
where images were acquired using only a portion (1.2 - 2.2 keV) of the X-ray 0 84 1
spectrum.s Data acquisition was performed with a Philips XL30/FEG SEM
equipped with an Oxford super-ATW detector and XP3 pulse processor, and an
EMiSPEC Vision integrated acquisition system. The XL30 was operated at 4 kV
with a 30 pm final aperture. EDS spectra were acquired with a 350 takeoff angle at
-1500 input counts per second and -20% dead time. The 200 x 150 pixel spectrum
image was acquired with 20 nm per pixel, 10 eV per channel, and a 250 ms dwell.
MSA was performed on a 200 channel selection (0.12 - 2.11 keV) of the spectrum
image that includes all relevant characteristic X-ray peaks. A normalized eigenvalue
plot of the first 50 MSA components is .shown in Fig. 1 a. The linear variation of the
higher order (>10) eigenvalues on the semilog plot is consistent with the variations
of these spectral components arising solely from Poisson statistics. The first nine
eigenvalues have an information value that is distinct from the Poission noise limit.
Fig. 1 shows MSA component spectra (b-d) and images (e-g) corresponding to the
first three eigenvalues. Bright (dark) features in the component images correspond
to strongly positive (negative) characteristic peaks in the corresponding spectra. For
example, the brightest areas in Fig. If come from the SiO2 dielectric (corresponding
to the strong O-K peak at -0.5 keV), the darkest areas come from the Si substrate
and the W plugs (with negative spectral features between -1.6 and -2.0 keV), and
an intermediate grey level is displayed by the Al lines (weakly positive peak at -1.5
keV). There is excellent discrimination between the Si substrate and the W plugs in
Fig. 1g, despite the unresolved W-M and Si-K X-rays (separated by -35 eV),
which give rise to the first-derivitive-type feature in Fig. Id. Also evident in Fig. ld
is the higher continuum X-ray intensity excited in the high-Z W plugs.6
The submitted manuscript has been authored by a
contractor of the U. S. Government under contract No.
DE-AGO5-96OR22464. Accordingly, the U.S.
Government retains a nonexclusive, royallytree license
to publish or reproduce the published form of this
contribution, or allow others to do so. for U.S.
Government purposes."
Upcoming Pages
Here’s what’s next.
Search Inside
This article can be searched. Note: Results may vary based on the legibility of text within the document.
Tools / Downloads
Get a copy of this page or view the extracted text.
Citing and Sharing
Basic information for referencing this web page. We also provide extended guidance on usage rights, references, copying or embedding.
Reference the current page of this Article.
Anderson, I. M. Applications of Multivariate Statistical Analysis (MSA) in Microanalysis, article, February 16, 1999; Tennessee. (https://digital.library.unt.edu/ark:/67531/metadc675630/m1/1/: accessed April 24, 2024), University of North Texas Libraries, UNT Digital Library, https://digital.library.unt.edu; crediting UNT Libraries Government Documents Department.