Scanning Electron Microscopy and X-Ray Microanalysis

319 20

20.3.3 Analysis of a Minor Constituent

with Peak Overlap From a Major

Constituent

The problem of accurately recovering peak intensities when

overlaps occur is exacerbated when the concentration ratio of

the elements producing the overlapping peaks is large, for

example, a major constituent (C > 0.1 mass fraction) interfering

with a minor (0.01 ≤ C ≤ 0.1) constituent. The high throughput

(>100  kHz output count rate) of SDD-EDS enables collection

of high count EDS spectra in modest collection time (e.g.,

10 million counts in 100 s). Moreover, the high throughput of

SDD-EDS is achieved with stability in both the peak position

(i.e., calibration) and the peak shape (i.e., resolution) across the

entire input count rate range. In simultaneous WDS-EDS

measurements, this SDD-EDS performance been demon-

strated to the spectrum measurement capabilities necessary for

robust MLLS peak-fitting to achieve accurate measurement of

the interfering peak intensities equal to that of WDS on the

spectroscopically resolved peaks (Ritchie et al. 2012 ).

20.3.4 Ba-Ti Interference in BaTiSi 3 O

BaTiSi 3 O 9 (benitoite) provides an example of severe interfer-

ence between two constituents of identical atomic concentra-

tion but with a mass concentration ratio of Ba/Ti = 2.9—Ti

K-L2,3 (4.510  keV) and Ba L3-M4,5 (4.466  keV)—which are

separated by 44 eV, as shown in. Fig. 20.7. DTSA II analysis

of benitoite with Ti and sanbornite (BaSi 2 O 5 ) as fitting refer-

ences and standards is given in. Table 20.7. Note that in this

analysis, O has been directly analyzed with the k-ratio/matrix

corrections protocol and not by the method of assumed stoi-

chiometry. The analytical results are seen to closely match the

stoichiometry of the ideal mineral formula.

20.3.5 Ba-Ti Interference: Major/Minor Constituent Interference in K

Constituent Interference in K2496

Microanalysis Glass

NIST microanalysis research material K2496 glass contains

these same elements, but with Ba as a major constituent

(C = 0.4299 mass fraction) and Ti as a minor constituent

(C = 0.01799 mass fraction), giving an elemental ratio of Ba/

Ti = 23.9.. Figure 20.8a shows the SDD-EDS spectrum and

residual after peak fitting, and. Table 20.8 contains the

results of the analysis. Despite the severe overlap and the

large elemental ratio, the concentration for Ti is measured

with reasonable accuracy. A reasonable question that the

analyst might ask is, If it was not known that the Ti was pres-

ent, could it be detected?. Figure 20.8b shows the fitting

residual for an analysis protocol in which Ti was not fit. The

peaks for Ti K-L2,3 and Ti K-M 3 are revealed in the residual

spectrum.

20.4 The Need for an Iterative Qualitative

and Quantitative Analysis Strategy

The analysis of NIST glass K2496 demonstrates that rigorous

analysis requires an iterative qualitative analysis–quantita-

tive analysis approach. When analyzing an unknown

material, it is likely that some constituents at the minor and

trace level will not be obvious when the first qualitative anal-

ysis is performed due to peak interference from constituents

at higher concentrations. An alternating qualitative–quanti-

tative analytical strategy is required to discover possibly hid-

den minor and trace constituents. In the initial qualitative

analysis, the EDS spectrum is evaluated to identify the major

and minor elemental constituents whose peaks are readily

identifiable. The k-ratio/matrix correction protocol is then

applied with appropriate choices for elemental peak-fitting

references and for standards, and the “residual” spectrum is

constructed that contains the intensity remaining after the

fitted peaks have been subtracted. If all constituents have

been accounted for, this residual spectrum should only con-

sist of the continuum background and possibly also artifact

peaks such as escape and coincidence peaks. However,

because of the relative poor energy resolution of EDS, the

analyst must perform a second qualitative analysis of the

residual spectrum for the presence of previously unrecog-

nized peaks that are associated with constituents that suffer

interference from the higher intensity peaks. If such peaks

are discovered and assigned to an element(s) not previously

recognized, the quantitative analysis must then be repeated

with this element(s) included in the peak-fitting and

. Table 20.5 Analysis of PbS at E 0 = 10 keV with CuS and PbSe
as fitting references and standards; Integrated spectrum count,
0.1–10 keV = 5,482,000; uncertainties expressed in mass fraction.
Analysis performed with Pb M 5 -N6,7 and S K-L2,3

S Pb

Cav (atom frac) 0.4938 0.5062

Z-correction 1.31 0.983

A-correction 1.028 1.056

F-correction 1 1

σ (7 replicates) 0.000953 0.000953

σRel (%) 0.19 % 0.19 %

RDEV (%) −1.20 % 1.2

C (mass frac, single analysis) 0.1306 0.8651

Counting error, std 0.0001 0.0009

Counting error, unk 0.0003 0.001

A-factor error 0.0002 0.0017

Z-factor error 1.50×10–5 0.0001

Combined errors 0.0004 0.0022

20.4 · The Need for an Iterative Qualitative and Quantitative Analysis Strategy