Clay Minerals (1984) 19, 243-247
NOTE
ELECTRON
MICROPROBE
ANALYSIS
OF
CLAY
MINERALS
The electron microprobe has been used in petrographic research for almost 20 years and
more recently as a tool for investigations of clay minerals. However, to the author's
knowledge no information has been published concerning the reliability of such
analyses--especially those of the alkali elements. Since the alkali contents of clays are
often critical, e.g. in determining the smectite content of mixed-layered minerals, it is
important to have some idea of the reliability of an analysis made using an electron
microprobe.
Experimental
Three variables can be controlled in electron microprobe analysis--time, beam-current
intensity and beam size (the diameter of the beam which excites the sample). Exaggerated
conditions of any one of these variables tend to diminish the signal received by the
detectors for the alkali elements, especially sodium. As an example, the effect of varying
two of these parameters on the intensity of sodium radiation for s o d i u m - a l u m i n u m silicon glass is shown in Fig. 1. The signal decreases (% loss Na) over a 200 s counting
time. Beam intensity and beam diameter were varied in this example. The intensity range
% loss
Na
30
20
I
I
pm
FIG. 1. Representation of measurements of sodium X-radiation intensity for an aluminosilicate
glass containing 8% Na20. Variables are % loss of radiation and beam diameter in pm. Values
are given for three beam intensities, indicated in nA. The acceleration voltage was 15kV,
counting time 200 s.
0
5
10
15
20
9 1984 The Mineralogical Society
25
244
Note
Ik
1.0
ML
0.9
0.8
0.7
0.6
DM
I
0
I
500
counting
1000
time
sec
Fz6. 2. Variation of potassium radiation intensity at 2.15 nA and a beam diameter of 2/Lm.
Relative stability of potassium signal is shown as a function of time for a mixed-layer
illite-smectite (95% illite) ML, detrital muscovite DM, and orthoclase = OR.
chosen was 1-3 nA as the lower limit of normal operation for most electron microprobes is in this range. The beam diameter of 5-20 r
was chosen because the
smallest spot size commonly used in back-reflection X-ray geometry is 2 pm. For 15 kV
acceleration, the excitation volume can be considered to be near 6 ~tm (Theisen, 1965).
Since clay-sized minerals are defined as those < 2/tin, the clay mineralogist will wish to use
a beam size as small as possible in order to obtain a single grain sample. It can be seen
from Fig. 1 that it is desirable to keep the beam current at a minimum and to use as large a
spot as possible.
The instruments used in this study were an MS46 and C A M E B A X microprobe
manufactured by CAMECA, France.
The effect of prolonged counting time for different minerals is shown in Fig. 2. It is
apparent that the decrease in alkali (potassium in these examples) signal varies depending
on the minerals analysed. It should be noted that the intensities of the aluminum and silicon
radiation varied by 1 to 6% during the time periods used. There was a definite differential
loss of alkali signal under extreme conditions.
In the author's experience the minimal analysis conditions of 2 ~m beam diameter and
1-2 nA current do not change the intensity of K, Si or A1 signals beyond counting statistics
errors for the following clays after a period of 200-400 s: glauconite, illite, or ISII (90%
illite) mixed-layer mineral. Metamorphic and magmatic biotites and muscovites are also
stable, but when they are found as detrital minerals in a sedimentary rock they tend to lose
alkali signals easily (Fig. 2).
Mixed-layer clays from K-bentonites (Velde & Brusewitz, 1982) were subjected to these
conditions and the results are shown in Fig. 3. Here beam-current intensity was held at a
low level (1 nA), since it was evident that it would affect the alkalis readily, and beam
diameter was varied. It is important to note that the intensity of the silica radiation varied
in much the same way as that of potassium. For this reason these data were not recorded
in the Figure. It can be seen that counting times of 100-200 s did not change the radiation
intensities significantly for all beam sizes used. This was true even for the sample
245
Note
|
countin 9
120
360
time
600
840
sec
Ik
3% K 2 0
1.0
0.9
0.8
counting time
120
i
Ik
sec
360
600
840
i
I
i
6% K20
I.O
0.9
0.8
FIG. 3. Variation in potassium signal for two mixed-layer clays with 3% K20 (a) and 6% K20
(b). Spot size varied from 2-10/~m as indicated. Beam intensity was one nA. Measurements
were made at intervals indicated on the coordinates of total seconds of analysis time elapsed.
containing only 3% K20, i.e. with 70% smectite layers. It is thus apparent that even highly
hydrous clays can be analysed using a small beam size, provided that the counting times do
not exceed 200 s.
The obvious question is how precise are such analyses using beam currents as low as one
nA. In order to assess this variable (which is related to the low counting rates at low beam
currents) it is necessary to compare two different types of detection systems which are
currently available on more recent electron microprobes, i.e. energy-dispersive (siliconlithium diode detector) and wave-length dispersive. These are commonly called EDS and
WDS detectors in the literature. Tests made in the present investigations used the TAP
crystal on the C A M E B A X Microbeam machine (WDS) and the ORTEC EDS system.
The element analysed was sodium as it is the element of lowest atomic number normally
analysed in clays and it gives the lowest count rate per unit amount of element in a sample.
It was found that the WDS system has a perfectly linear response for background-corrected counts as a function of the variables NazO concentration, beam current
and counting time. Thus simple counting statistics can be used to simulate variations in
current or concentration. Tests on feldspars containing 0.45 and 11.56% Na20 were
compared at 18, 5 and 1 nA current intensities using a beam diameter of 20 pm.
The energy-dispersive system is a little more difficult to manipulate and as a result the
following method was used: counts were measured 10 times for a 100 s accumulation on
Note
246
error
in %
I
15
/EDS
I
6
2
4
i
I
6
8
I
wt% Na20
FIG. 4. Comparison of EDS (energy-dispersive) and WDS (wave-length dispersive systems with
respect to their statistical reproducibility. Variables are percent error (~/N for EDS and ,~/N
for WDS) vs. Na20 concentration in the sample.
feldspar and pyroxene samples containing 0.3, 1.3, 3.3 and 8-3% Na. Large beam
diameters (20 pm) were used with a current intensity of 1.5 nA. These measurements are
presented in Fig. 4. The statistical error is expressed as an approximation of sigma for the
WDS system (v~/N) or sigma over the average values of ten different EDS measurements
on the same sample. This gives a value of approximate error as a function of the amount of
Na20 present in the sample. From the Figure it can be seen that the errors in counting are
comparable for the systems at Na20 values >4%, the WDS system is more precise in the
range 0 . 5 - 4 % , and both are poor at < 0 . 5 % Na20 under the operating conditions used.
It is important to note that the EDS system accumulates data for all of the elements
simultaneously so that the complete analysis is effected in the time necessary to obtain the
sodium values, while WDS systems usually have four spectrometers and therefore need
two counting cycles to obtain a full clay mineral analysis. Further, WDS systems should
count both background and radiation maxima in order to give correct results, especially
when count rates are low. In this manner, the WDS system used here takes ~200 s
counting time to effect a full nine-element analysis, counting 20-30 s for intensity maxima
on all elements. The total counting time will obviously be much longer when counting for
100 s for each element. Since the beam cannot remain on the same spot for more than
200 s without the loss of some alkali signal, it is evident that an EDS system should be used
for clays. Even though the precision is less for the light elements, the element ratios should
be more correct using an EDS system at low beam currents because of time restrictions
imposed by potential alkali loss.
Conclusions
One may surmise from the discussion above that clay mineral analysis by electron
microprobe should easily be possible if care is taken to use the correct operating conditions.
Note
247
Low b e a m intensity is necessary, one to two n A is suggested. Counting times should not
exceed 200 s for all o f the elements present. A spot size of 5 g m will avoid most alkali loss.
B. VELDE
L a b o r a t o i r e de G6ologie,
E R 224 C N R S ,
Ecole N o r m a l e Superi6ure,
46 rue d'Ulm,
75230 Paris,
France
2 D e c e m b e r 1983
REFERENCES
THEISENR. (1965) Quantitative Electron Microprobe Analysis. Springer Verlag, Berlin, 170 pp.
VELDE B. & BRUSEWITZA.M. (1982) Metasomatic and non-metasomatic low grade metamorphism of
Ordovocian meta-bentonites in Sweden. Geoehim. Cosmochim. Acta 46, 447-452.