Академический Документы
Профессиональный Документы
Культура Документы
EPMA
Mod 1/10/2016
Whats the point?
Whats the minimum detection limit for a
particular element or said otherwise, at what
point can be be sure that a small inflection above
the surrounding background really is a peak?
What kind of confidence level should be place on
such a number?
Definitions
Generally, WDS can achieve limits of
detection of 100 ppm in favorable cases, Major >10 wt%
with 10 ppm in ideal situations where there
are no peak interferences and negligible Minor 1-10 wt%
matrix absorption.
Trace <1 wt%
(Goldstein et al., p. 341)
The USNM olivine standard above (San Carlos, Mg.9 Fe.1SiO4) has a
published Ca content of <0.04 wt% (= 400 ppm).
This scan was acquired at 20 keV, 30 nA, with 10 seconds per channel.
Clearly there is a peak at the Ca Ka position (24 cps), somewhat above
the background (~10 cps). At what point can we say with 99%
confidence that there is a statistically significant peak?
MDL Equations - 1
The key concept here is minimum detection limit (MDL), i.e., what is the
lowest concentration of the element present that is statistically above the
background continuum level by 3 sigma (commonly accepted level).
There are (at least) two equations used to define the MDL:
the first* uses the Student t test values: a
CS 21/ 2 (t 1-
n-1 )SC
CMDL =
N S - NSB n1/ 2
where the detection limit CMDL is in wt%, CS=std wt %, bar NS=ave. peak
cts on std, bar NSB=bkg cts on std, SC= std dev of measured values and
n=number of data points
the second, which is probably more wider used, was developed by
Ziebold (1967): (3.29) a
CMDL =
(nTP PB )
1/ 2
bkgcts unk
Ca ZAFCunk std
KCa-mdl = 3 std a
std CCa
(pkcts - bkgcts)C a ZAFC a
and multiply by 3 (for 3 sigma, 99% confidence) and the ZAF of each and
then by the composition C of the standard.The mdl will be in whatever
units C is in.
Deriving the MDL equation-4
ol
ZAFC a = 1.1087
a = 1.080
ZAFCstd
ol90 1.1087
MDL = 3 Ca 38.6 = .017wt%Ca
66250 1.080
This is virtually the same result as the single line detection limit provided by
Probe for Windows (0.015 wt%, shown on next slide), derived from the Ziebold
equation.
It would appear that the Ziebold equation is not exactly correct, for we must
really be concerned with the background precision of the unknown, and the
background level of the standard could be several times higher or lower. Going
back and re-reading Ziebold, we find two interesting statements: that the
equation gives a measure of the detectability limit and there is more than one
way to define a detectability limit. Both are correct, and yes, the equation gives
an approximation of the detection limit -- but not the limit per se.
MDL in olivine - single line
Good totals
Excellent
stoichometry
P Ka peak = red line position. But n=2 Ca Kb1,3 also very close!
n=1 Si Kb1,3
n=1 P Ka1,2
But
WDS is clearly the better method for acquiring trace element data, by
an order of magnitude or so compared to EDS.