SCA_author checklist--prelim.pdf8852Mistakes Encountered During Automatic Peak Identification of Minorand Trace Constituents in Electron-Excited Energy Dispersive X-RayMicroanalysisAQ1DALE E. NEWBURYNational Institute of Standards and Technology, Gaithersburg, MarylandSummary: Automated peak identification i n electronbeam-excited X-ray microanalysis with energy dis-persive X-ray spectrometry has been shown to be sub-ject to occasional mistakes even on well-separated, high-intensity peaks arising from major constituents (arbi-trarily defined as a concentration, C, which exceeds amass fraction of 0.1). The peak identification problembecomes even more problematic for constituents presentat minor (0.01rCr0.1) and trace (Co0.01) levels.‘‘Problem elements’’ subjec t to misidentification as ma-jor constituents are even more vulnerable to mis-identification when present at low concentrations in theminor and trace ranges. Additional misidentificationsattributed to trace elements include minor X-ray familymembers associated with major constituents but notassigned properly, escape and coincidence peaks asso-ciated with major constituents, and false peaks owing tochance groupings of counts in spectra with poorcounting statistics. A strategy for robust identification ofminor and trace elements can be based on application ofautomatic peak identification with careful inspection ofthe results followed by multiple linear least-squares peakfitting with complete peak references to systematicallyremove each identified major element from the spec-trum before attempting to assign remaining peaks tominor and trace constit uents. SCANNING 31: 1–11,2009. r 2009 Wiley Periodicals, Inc.Key words: energy dispersive X-ray spectrometry,automatic peak identification, peak fitting, qualita-tive analysis, scanni ng electron microscopy, traceelements, X-ray microanalysisIntroductionElectron-excited energy dispersive X-ray spectro-metry (EDS) performed in the scanning electronmicroscope (SEM) or other electron beam platformshas become established as a core compositionalcharacterization method for a broad range of physi-cal and biological sciences as well as many branchesof engineering and technology (Goldstein et al. 2003).Commercial SEM/EDS instrumentation has maturedwith the development of extensive software resourcesfor computer-controlled microscope operation andcomputer-assisted X-ray analysis. Advanced softwarecontrols EDS X-ray spectrum acquisition and spec-trum analysis, including automatic qualitative ana-lysis (X-ray peak identification) and automaticquantitative analysis that can be standards-based orthat does not require local standardization (‘‘stan-dardless analysis’’). Perusal of recent EDS advertis-ing shows an increasing tendency in software systemsto simplify the operation of this complex in-strumentation by minimizing the need for theoperator’s input to the analysis. Indeed, advertisingclaims are frequently found for ‘‘one button’’ analy-sis, whereby the entire process of operation is en-capsulated in an automatic software routine thatrequires the operator only to select the desired beamlocation or image region where the analysis is to beperformed and initiate the spectrum collection, withthe software doing the rest of the task, directlyleading to a final report of analysis entry for thatspecimen location that identifies the elemental con-stituents present and assigns concentration values.Although this article is written from the perspectiveand experience of the SEM-EDS analytical commu-nity, it has been noted by a colleague of the authorthat these considerations apply equally well to thetechnique of energy dispersive X-ray fluorescence(EDXRF) analysisAQ2(Sieber 2009).Correctly identifying the elements present in thebeam-excited volume during the qualitative analysisThis will be your only chance to review these proofs. Please note that once your corrected article is posted online,it is considered legally published, and CANNOT be removed from the Web site for further corrections.Published online in Wiley InterScience (www.interscience.wiley.com)DOI 10.1002/sca.20151Received 1 April 2009; Accepted with revision 2 May 2009Address for reprints: Dale E. Newbury, National Institute ofStandards and Technology, 100 Bureau Drive, Gaithersburg, MD 20899.E-mail: [email protected] VOL. 31, 1–11 (2009)& Wiley Periodicals, Inc.procedure is obviousl y the required foundation, ifthe subsequent quantitative analysis is to make anysense. The skeptical analyst will carefully test au-tomated analysis systems, starting with the auto-matic peak identification. Two earli er articlesexamined the occurrence of mistakes that sometimesoccur during automatic peak identification withcommercial software for major constituents (arbi-trarily, a ‘‘major’’ consti tuent will be defined as anelement present at a mass fraction40.1; see furtherdefinitions below) under ‘‘conventional’’ electron-excited EDS microanalysis conditions (Newbury2005a) and under low-beam-energy microanalys isconditions (Newbury 2007). ‘‘Conventional’’ ana-lysis conditions are generally regarded as involvingselection of the incident beam energy in the range15–30 keV, a choice that provides excitation of atleast one readily measurable characteristic X-raypeak family (i.e. peaks from one atomic shell) for allelements of the periodic table with atomic numbergreater than or equal to 4 (beryllium). Those pre-vious studies considered only the simplest peakidentification problem, namely the identification ofelemental constituents present at sufficiently highconcentration to produce high-inte nsity character-istic peaks with a high peak-to-background (char-acteristic to continuum) ratio. Moreover, only thosespectrometric situations were considered wherethere were no significant interelement peak overlapsand no peaks below 0.9 keV photon energy, whichare subject to large self-absorption in the specimen.Despite the apparent simplicity of this challenge, thestudy of automatic peak identification under con-ventional analysis conditions found that mis-identifications of major constituent peaks occurredin approximately 3–5% of the cases tested(Newbury 2005a) when specimens contained ele-ments selected throughout the periodic table, butexcluding Be to Ne, which could only be analyzedwith low-photon-energy, K-shell peaks below0.9 keV. Moreover, the peak misidentifications
View Full Document
Unlocking...