as a PDF

Canadian Mineralogist
Vol. 23, pp.263-271 (1985)
ELEMENTS:
ELECTRON.MICROPROBE
ANALYSIS OF MINERALSFOR RARE.E^ARTH
USE OF CALCULATEDPEAK-OVERLAPCORRECTIONS
PETER L. ROEDER
Departmentof GeologicalSciences,
Queen'sUniversity,Kingston,OntarioK7L 3N6
ABSTRACT
The relative intensitiesof the lcr1, Lgy L/z, LB3, and
.L71lines were measuredusing ten R.EEphosphates.These
data were used to calculate peak-overlap corrections for
the Lal and Lp1 lines of RE'8. Wavelength-dispersionand
energy-dispersionspecfia of the Z lines of fourteen R.EE
compare favorably with the spectrausing the peak-overlap
corrections.Detection of REE by microprobe analysisis
best accomplished by first collecting a wavelengthdispersion spectrum to identify interfering elementsand
locate the best wavelengthsfor background measurements,
The ,REf phosphates were used as standards for
wavelength-dispersionanalysis of four silicate glassesprepared by Drake & Weill (1972),three silicateglassesprepared in the.presentstudy and someminerals previously
analyzedfor REE by wet+hemicalmetlod. The microprobe
resultsof thesesamplesaxeconsistentwith the known levels
of REE in the samples.
Keywords: rare-earth analysis, REE microprobe analysis,
microprobe standards.
Souuernn
On mesurel'intensitd relative desraies,Io1, L/y Lgz,
Lfu, et Lll dansdix phosphatesde terresrares (7R). Ces
donn6esservent a calculer les corrections d'empidtement
de raies pour les raieslo1 et ZB1 des ZR. Les spectresZ
de quatorze 7R, obtenuspar dispersionde longeursd'onde
et par dispersiond'6nergie,concordentavecles spectrescalcul6scompte tenu de cescorrections. La d6termination des
terresrares i la microsondeest facilit6e par l'obtention pr6alable d'un spectrepar dispersionde longueursd'onde, qui
rdvBleles 6l6mentsen interf6rence et les longueurs d'onde
i choisir pour mesurerIe bruit de fond. Les phosphates
de 7R ont servi d'dtalonsdansI'analysepar dispersionde
longueurs d'onde de quatre verres silicat6s pr6pards par
Drake et Weill (192), trois verresprepar€spour cette6tude,
et quelques mindraux dans lesquels on avait dosd les In
par voie humide. Les r6sultats que I'on obtient d Ia microsondeconcordent avecles teneursconnuesdes 7R dansces
€chantillons.
(Iraduit par la Rddaction)
Mots-clds:d&erninatton desterresrarc, analysei la microsonde6lectronique,6talonspour microsonde.
INTRoDUcTIoN
Electron-microprobe analysis of silicate minerals
has become routine because of the relatively simple
K seriesof X-ray spectrausedfor the major elements
in silicates.The microprobe analysisof samplescontaining rare-earthelementsis much more complicated
to usethe lower intensityand
becauseit is necessary
more complexZ seriesof X-ray spectra.Thus the
I line for one rare-earth element (REE) may interfere with the measurementof another element. As
an example,Z spectrafor the phosphatesof lanthanum, cerium and praseodymium are shown in
Figuresla, b, c, run under the conditionsdescribed
in the next section. Thirteen peaks for lanthanum
are identified in Figure la and illustrate the possible
complexity for a samplecontaining more than one
REE. The spectrum shown in Figure ld peltains to
a glassmicroprobe standard synthesizedby Drake
& Weill (1972); it contains almost equal concentrations (Table l) of lanthanum (La), cerium (Ce) and
praseodymium@r). The chemicalsymbol for the
individual rare-earth elements will be used in the
remainder of this paper. The lines for LaZal and
CeZculin Figure ld are almost of equal intensity,
but the PrZal is at almost the samewavelenglhas
theLaLPt line, and thus the measuredX-ray intensity at this wavelengthis due to both La and Pr. The
peak overlap is a particularly seriousproblem for the
evaluationof Pr, Eu, Gd, Ho, Er, Tm and Lu concentrationsby microprobe. The two methodsused
to aveid this problem are to chooseX-ray lines showing minimum interference@xley 198p)or to usecorrection factors for peak overlap (Amli & Griffin
1975,Haggerty1983,Segalstad& Larsen 1978).The
correction factors for overlapping peaks of differert kEE have been measured by Amli & Griffin
(1975),but the onesthat they presentedare dependent on the changein resolution and efficiency, over
the wavelengthregion, of the spectrometerusedin
their microprobe. Other analystswho wish to usethe
results of Amli & Griffin have to assumethat their
spectrometershows the same resolution and efficiencychangewith spectrometersetting.The amount
of peak overlapis very sensitiveto the resolutionof
the spectrometer.Thus it is best fo measuredirectly
the overlap correction, but suitablestandardsare not
alwaysavailable.The major purposeof the present
paper is to presentdata neededto calculatethe peak
overlap for a spectrometerof any resolution or efficiencyand to presentmicroprobedata for standard
material.
The first part of this paper describesthe measurement of the intensity of the major lines (2a1, ZB1,
263
264
THE CANADIAN MINERALOGIST
LBz, LBy L7) for ten REE; a method is presented
to usethesepeak intensitiesto calculatethe overlap
correction for each R.EE These data are then used
to calcirlatea spectrumthat is comparedto a measured spectrum. The second section describesthe
microprobe determination of fourteen REE in some
glasssamplespreparedin the presentstudy and the
Drake & Weill glasses,which are usedasR"EEstandards by many microprobe analysts. Microprobe
data are presentedfor someREEminerals and compared to wet-chemicalresults on theseminerals.
EXPERIMENTAL
PRocEDURE
All microprobe data were collectedwith an ARL
SEMQ microprobe (take-off angle of 52.5') at an
excitation voltage of 15 kV or 25 kV. The
(WD) data werecollectedwith
wavelength-dispersion
a curved lithium fluoride crystal and an argon-filled
(o) LoPOo
detector.The energy-dispersion
@D) data werecollectedwith a Princeton Gamma-Techdetectorhaving a resolution@WIIM) of 155eV for manganese.
The WD dataat 15kV werecollectedat a beamcurrent of 22 nA measured with a Faraday cage in a
brass mount.
The usefulnessofthe proposedpeak-overlapcor-'
rection procedure dependsupon how well other
analysts can measurethe resolution and efficiency
of their ED or WD spectrometers
relativeto the one
used in the presentstudy. Thi$ comparisonis best
made by measuringthe peak width and count rate
at the peakposition for the Klines of common metals
over the wavelength region used for the REE. The
resolution or peak width at one standard deviation
(60.68/oheight of the peak) for the WD spectrometer usedin the presentstudy is shown for the KB line
of somecommon metalsin Fieure 2. The peak width
in Sngstrdm units is plotted against the wavelength
of the linesaslistedin White & Johnson(1979).Also
shown in Figure 2 is the variation in count rate (only
correctedfor background)for the Kcr lines of some
common metals. The matrix correctionsof the intensity measurementswere made using a ZAF correction program supplied by Tracor Northern and
modeledafter the MAGIC programof Colby (1968).
The samplesusedfor the measurementof relative
peak-heightofthe individual REE area seriesof l0
REE phosphatesprepared and supplied by L.A.
Mn
r"f*x
o
o
Cr
Ni
./
}.rcoo
o
o
(d) REE3
2000
<
.008
3 .006
!
o
.oo4
r.6
27
?.3
1.9
spectraof R.E^Ecoltected
Ftc. l. Wavelength-dispersion
with an LiF crystal from a )r of 2.7 to 1.9 A at 15 kV:
a) LaPOa, b) CePOa,c) PrPOa, d) glassfrom Drake
& weill (1972).
r.8
2.O
2.2
2.4
X (A)
2.6
2.8
FIc.2. a) Peak count-ratecorrectedfor background for
the Ka line of the listed metals as a function of
wavelength.b) The peak width in dngstrcimsat one
standarddeviation (60.690height) for the,I(B line of
the listed metals as a function of wavelength.
26s
ELECTRON-MICROPROBE ANALYSIS OF MINERALS FOR RARE-EARTH ELEMENTS
Boatner to A.N. Mariano. These samples are
describedbyBoatner et al. (1980).The relative intensity of one Z line relative to another,L line for any
one R"EEwas checkedusing someREE metals supplied by Applied ResearchLaboratories and on
oxidesprovidedby A.N. Mariano. The R,EEglasses
preparedand distributed by Drake & Weill (1972)
wereusedas standardsfor Gd, Tb, Yb and Lu since
no phosphatesof theseelementswere available ar
the time this study wasconducted.Threeglasseswere
preparedin the presentstudy in order to checkthe
accruacyofthe analytical and correction techniques.
Theseglasses(5-236, 5-253, 5-254) wereprepared
by carefully weighing predried CaCO. (600"C),
Al2O3 (600'C) and silicic acid (1000'C). Predried
CeO, and LarO3 were used in 3-236, whereas
predried Spexmix 1032supplied by SPEX Industries
was used in preparing samples5-253 and 5-254.
Spexmix 1032 is listed as containing scandium,
yttrium and fourteen REE elementsat the 5.28
el.wt.9o level, and the glassespreparedusing this mix
were made such that they have a nominal concenTABI.E 1.
I,A
Ce
Pra
PTB
Nd
Sm
Euo
Eug
Gdc
Tb
Dy
Eoa
EoB
Era
ETB
Ttoa
tng
Yb
Lu
Ca
A1
3.55
3.42
3.79
3.79
3.65
3.67
3.80
3.80
3.87
3.78
3.80
3.85
3.85
3.81
3.81
3.81
3.81
3.74
3.75
Probe
15kv(n=3)
3.60
3.33
3.7L(4.r2)
3.6L
3.6s
3.77
X.76
3.78
3.73
3.74
3.65
3.7L
3.57
3.7L
3.46(4.78)
sl
IIAAr
Glass
Non.*
3.91
3.67
3.93
3.93
3.60
3.72
3.90
3.90
IN'rsNsrry MEASUREMENT
ett
CaLcuLerroN
The first information neededto calculatethe overlap correctionis the value of the intensity of anyREE
peaksrelative to the intensity of the major peak of
that same REE in a sample that contains no other
REE. For example, if it is known that the ZBr peak
for La is 57Voof the height of the Za, peak, then
the La contribution to the PrZa, peak shown in
CEEI{ICAL COMPOSITION OF GLASS STANDARDS*
Four GLasses of Dtake & I{elllNom.*
tration of 1.A40/ofor each REE (3-254) or 0.0890
for eachREE (5-253). Each samplewas carefully
ground and mixed in an agatemortar and then fused
twice at 1500'C for six hours with an intermediate
grinding in a tungsten carbide ball-mill. The glass
sampleswere analyzedfor REE by two laboratories
using neutron activation. Theseresults,togglheqwillt
the nominal valuesfor Ca, Al and Si, are shown in
Table I . The homogeneityof the glasseswas checked
by usingthe homogeneityindex of Boyd et ol. (1967).
Limited amounts of these three glass samplesare
available from the author.
1.04
1.04
1.04
1.04
1.04
1.04
1.04
1.04
1.04
3,96
1.04
1.04
3.81
1.04
3.81
1.04
1.04
1.04
1.04
L.O4
3.s8
1.04
3.75
L.O4
16.90
12.90
15.09
5-254
?robe
15kV(5) 25kV(3)
1.04
1.06
0.98
0.94
1.03
1.01
1.06
1.09
1.09
1.15
1.04
0.96
1.16
L.O2
L.tz
L.Lz
L.Lz
0.98
1.15
1.05
1.06
1.01
0.94
1.10
0.99
1.04
1.00
1.03
1.04
1.00
1.00
1.01
0.94
0.97
0.99
0.95
0.97
0.98
Glasg
l{AAz
NAAr
o.92 0.82
0.91 0.76
0.78
0.96
0.85
0.85
0.51
0.70
0.70
0.70
o.94
0.81
1.10
1.10
0.73
o.82
0.82
0.84 0.63
0.91 0.69
Noa.*
0.08
0.08
0.08
0.08
0.08
0.08
0.08
0.08
0.08
0.08
0.08
0.08
0.08
0.08
0.08
0.08
0.08
0.08
0.08
20.80
L6.26
18.09
5-253
Probe
15LV(3) 25LV(5)
0.07
0.08
0.07
0.06
0.06
0.06
0.07
0.06
0.10
0.04
0.06
0.02
0.13
0.09
0.10
0.06
0.06
0.09
0
0.06
0.06
0.08
0.05
0.08
0.06
0.08
0.07
0.08
0.07
0.06
0.07
0.10
0.06
0.06
0.05
0.08
0.05
0.06
NAA2
I{AAI
0.090 0.100
0.079 0.100
0.067
0.087
0.077
0.077
0.095
0.080
0.080
0.097
0.097
0.100
0.097
0.088
0.088
0.088
0.077
0.077
0.080 0.088
0.080 0.086
Glass 5-235
probe
I tlon.* 15kV l{AAs NAll8
la
11
| 10.56 Lo.26 11.1
10.30 10.7
10
Ce I ro.27
Ca | 1 5 . s 4
A1 I 10.43
st I 1 5 . 6 9
*CoEpoBltl.on
r - Neutf,on
2 o Neutron
8 o Neutrol
expressed
actlvatlon
actl.vatLotr
actlvatlo[
ttr !rt. Z
aml)rses
analyees
aaalyses
eleoellts,
Non. = NoELnaL vt.. Z fron lniti.al
weLghlng.
by Drake & 8e111 (L972).
t'y CII{A
Iln{.ver6ita
de Montr6a1,
}lontraal,
Quebec.
by Neutron ActLvatLon
Serelceg,
i{c}taster
Nuciear Reactor,
Eadlton,
Oat.
266
TI{E CANADIAN MINERALOGIST
Figure ld can be calculated.It is possibleto usethe
relative intensity of the peaksgiven in White & Johnson (1979);however,thesevaluesare given a nom!
nal value(L& 50,LBzX, Lfu 6, L115) for all the
REE The peak heightsof the five major peaksfor
the l0 REE phosphateswere measuredby using a
LiF crystal. The peak centroid was found by using
an automatic peak-searchroutine that is part of the
TASK softwaresuppliedby Tracor Northern. Each
measurementinvolved three countson different spots
of 30 secondsduration (total 90 seconds).Background was measuredat three different wavelengths
and plotted as a function ofwavelength. The intensity of eachpeak, minus the background,wasdivided
by the intensity of the La1 peak and multiplied by
l@ to give percentintensity. The intensity ofthe various peaks relative to the Zo1 peak is plotted as a
function of atomic number in Figure 3. The measured values for the IB, line are shown by the
crossesat the top on Figure 3 and show some
irregularity (are too high) in the vicinity of Nd. This
irregularity is becauseof LBoand LBt overlap; the
L$a peak shift$ in wavelength with respect to the
ZBr peak as the atomic number changesfrom La to
Lu. There is completeoverlap of thesepeaksfor Nd.
Thus, in Figure la, the LBo peak is shown at a
lower wavelength than (to the right of) the LB, for
La, and the two peakscan be almost completely discriminated with the $pectrometerset at the ZB, peak
position. As the atomic number increases@ig. lb,
c) there is greateroverlap. The contributionof. LBo
to the.LB, peak can be calculatedueing the peak
positionsgiven by White & Johnson(1979)and the
relative intensity of the LBn given by White & Johnson. The calculaledcontribution of. the LP4to LBI
has beensubtractedand is shownin Figure 3 by the
position of the solid dots. Thus the crossesrepresent
the measuredvaluesat tbe LBl peak postition, and
the dots represent the corrected ZB1 iptensity. The
corrected ZBr intensities show an approximately
linear relation to atomic number. The LBr, L$, atrd
271 peaks have no interfering peaks, and the
changeof intensity relative to Zo, as a function of
atomic number is also approximately linear. The calculation of peak-overlapcorrection assumesthat the
relative intensity of the peaks as shown in Figure 3
remains the sameno matter what the material is or
what other elementsare in the matrix. The relative
peak-intensityhas beenmeasuredin somemetalsand
simple glasseSand found to be consistentwith the
intensitiesshown in the phosphates;the data for Sm
are consistentwith the data publishedby Bolon et
al. (1979). Samplescontaining other elementswith
peaks in lhe La1 to 27, region for any particular
REE may affect the relative intensities becauseof
secondary fluorescenceand absorption, but this
effect will usually be small. For example, the K
absorption edge for iron lies betweenthe DyZcrl
and DyLB, peaksand thus, if iron is presentin the
sample,the intensity of DyZBI will be lessrelative
to DyZor. The differential absorption of thesetwo
Dy linesby iron was calculatedusingthe procedure
of Corlett & MacDonald (1974) ald the equations
presentedby Reed (1975).The intensity of DyZB,
relative to DyLal is changedfrom Mslo (from Fig.
3) in a samplewith no iron to 42.50/oin a sample
with 50 fi.90 iron, 50 wt.Vo dysprosium. This is
within the experimentalerror of measurementsmade
in the presentstudy. There is also the possibility of
the detector gas (e.g., xenon) causing absorption
r6Pr
'b
edgesbetweenZ lines (Smith & Reed 1981).Thus
Dy Ho
caution is recommendedin using the relative intenATOMIG ito.
sitiesfor calculatinglarge corrections.Other analysts
Ftc. 3. The measuredintensitiesof the ZB1,LB2, L11and,
who wish to use the relative intensities shown in
lB3 lines of the R.6Erelativeto the Zo1 line as a funcFigure
3 have to determine whether the change in
tion of the atomic number of the REE. Thecrossesfor
the IBt line representmeasuredintensities,whereasthe spectrometerresponseover the wavelength interval
dots representthe intensities corrected for ZBa overlap is approximately the same as that (Fig. 2) in the
presentstudy, If it is different, the relative intensity
(seetext).
F.Ju-,
267
ELECTRON.MICROPROBE ANALYSIS OF MINBRALS FOR RARE-EARTH ELEMENTS
of the peak relative to the Zoq peak will haveto be
changed accordingly.
The calculation of the contribution of X-ray
counts of an interfering (D peak to the measurement
of the X-ray counts at the desired(d) peak-position
requires knowledge of the intensity of the interfering peak, the accurate peak-position of both the
interfering peak (\) and the desired peak (\) and
the peak width (W) of each of the peaks. The contribution of one peak to another can be calculated
assuming a Galrssian distribution of X-ray counts
on either side of the position of the interfering peak.
A Gaussiandistribution was found to be a reasonable approximation to the shapeof the peaks,except
for a "tailing" on either side ofthe peaksthat can
only be representedby a much more complicated
function (Suortti 1980). The following equations
from Bolz & Tuve (1976) can be used to calculate
the fractional contribution (F) of one peak to the
measurementof another peak. The function X
describesthe amount of overlap expectedbetween
the desiredpeak and interfering peak. In the following relationships,\a is the wavelengthof desired
peak, \, the wavelengthof interfering peak, W, the
width ofpeak at one standarddeviation (60.690of
peak height), and .{, the fraction of countsof interfering peak at the position of the desiredpeak.
x:
(\-\)
w/2
,::@
(l)
a)
If the two peaks are very close in wavelength or if
the peaks are broad Qarge W), the value of X will
be low, and the peakswill overlap. It is thus possible to calculatethe amount of peak overlap between
theLaLh and the PrZol peaksshown in Figure ld.
The calculatedvalue for.Fis found to be 0.22 using
a )taof 2.4630@rLo1),? \ of 2.4589(LaLPr),and
a value of W of 0.0047A. The fractional contribution of LaZBI is then multiplied by the intensity of
this peak (5790from Fig.3) in order to get an overlap correctionof L2.4Vo.Anli & Griffin (1975)measured a number of peak-overlapcorrectionsdirectly
on their ARL microprobe; for the LaLfu interferencewith PrZculthey measured12.70/0.Thus if the
Pr peak in the rare-earth glass standard shown in
Figure ld was measuredat the PrZol peak position,
the actual number of counts due to Pr would be
found by subtractingl2.7Voof the net countsin the
LaLolpeak position from thosein the PrZal peak
position. The percentagecorrection can be calculated
for a spectrometerof any resolution. For example,
we have calculated the fraction overlao for our
TABLE 2.
Deslred
Lltre
taLBt
PrIol
SoIo 1
EuIcy
EurBr
GdIcl
G<Ilp1
TbtBl
Dyrcr
EoIcl
Eo.tBl
E!trqt
Er-tP1
XbIol
IdlBr
i!.!cr
ibr9r
LuIcl
LurBr
OVERIiP
CORRECTIONS lOR.[o1
calculateal
Z ovellap
Peak
@d luterfelLtg
22.0
12.4
O.2
2L,7
9.9
L.0
9.5
87 .5
11.7
Plrq1
laLqr
CeI-B2
9tL92,
DyIaz;
1,aL'!2,
ce.i11
Bolol I
ErId!
L${ES
AND i81
lleasured
LSINC WDg
OverLap
L2.7(L3.6, L6,2)
2 0 . 3 P r ( 2 3 . 3 ,1 9 . 8 )
L.2 NillB3
0.3 DYIoI
0.2 ta-t13
0.3
2.8
r-r.8 (10.9)
Eola2
0.1 Eurpr
40.8 cd;'grr 0.2 GaliBq
0.3 Eur'fr
5.2 rbtBl 4.2'tbL$\
3.5 cd-[11, 0.7 EurY3
9.3 SoI11; 1.5 DyI9q
1.0 Gilt'rt, O.8 @L72
7 .9 TbLyt
0.3 TbIB2l O.2 DYLBs
I.0 \bLtz, 0.7 Tbtr'rg
5,7 DyL\\, 0.3 ToIB3
LL.I DyLBzt 6.5 EoLB2
L.O DyL'12,0.2 DyIYg
2.8 AbLB2. 2.4 D.oL'lr
tot Ia1
iE dec€aary
No correctLou
for ,81 of Ce' Pt' Nal, SB dal DY.
42.L
3.0
8.7 8B(9.1, 8.8)
9.0 Dy(12.2)
of Ia'
cer
Nal ed
Tb,
@d
energy-dispersion spectrometer and found that
becauseof its much poorer resolution, there is almost
complete overlap of LaLBt and Prldt. Thus F
equalsl, and the overlapcorrectionis equalto 5790.
The analysisof a small amount of Pr in the presence
of a large amount of La is very difficult, and small
errors in the overlap correction will lead to large analytical errors.
The peak positions for all kEE Z lines listed in
white & Johnson (1979) have been included in a
computerprogram written in BASIC that can be used
to calculatethe peak-overlapcorrection for all major
peaks.The intensitiesof the.LB1,LBz, L& and L11
peaks relative to La1 used in the program were
those taken from the best-fit lines given in Figure
3. The intensitiesof all other lines (e.9., Ll+, L/q)
were taken from White & Johnson.The correction
factors, calculated with this program for the peak
width shown in Figure 2, are given in Table 2 for
all the Zch and LBt peaks of the REE that have a
correction greaterthan 0. I Vo.The correction factors
for both the .Lo1 and L& have been calculated
becausesomeanalysts(a9., Exley 1980)use the LB,
peak if the tra1 peak showsa large^correction.The
measured overlap-corrections of Amli & Griffin
(1975)are also shown in Table 2, and are reasonably close to the values calculated and measured
(shown in parentheses)in the presentstudy. The best
spectral line for R.EE measurement on the
microprobe would be one with as few interferences
as possibleand one wherethe correctionfactors are
small and the interfering element is in low concentration. For example,Gd showsmajor interferences
for both La1 and ZBt, but Lfu may be preferable
becausethe major interferenceis from Ho, and this
elementis usually presentat very low concentrations.
Table2 contains correctionsdue only to overlapping
268
THE CANADIAN MINERALOGIST
peaksfor the REE; no considerationis givento peaks
due to elementssuchas Mn and Fe that overlap some
REE lines and can cause$evereinterference.The
overlap correction for theseelementscan be calculated using the samespectrometer-dataand the relative intensitiesfrom White & Johnson (1979), but
it would be preferableto measurethe overlap correction. It should be stressedthat careful electronmicroprobe analysisfor REE involvesa considerable investment of time and energy if the analytical
levels are below l9o.
Spectral scan
The complexity of the R EE X-ray spectrumis such
that it is wiseto conducta WD spectralscanof any
RZZ-containing mineral before attempting an analysis. The WD spectraof the glasses5-254 and 5-253
are shown in Figures4b and 4c. The spectrumin
Figure 4b refers to a sample containing REE attbe
nominal 1.0490level; the samplefor Figure 4c containsREE'at the nominal levelof 0.0890.Thesesamples are unusual, comparedto minerals, in the number and quantity of REE, but they demonstratethe
potential complexitiesof the REE spectmmand the
difficulty in measuring the background levels for
heavy REE. A useful method for calculating the
backgroundin sampleshaving a complexspectrum
was describedby Smith & Reed(1981).The spectrum
in Figure 4c demonstratesthe sensitivity of a spectral scan,evenat levelsof 800 ppm. A spectrumof
a substanceof known R.EEconcentrationcan be calculated given a) the relative intensity of the Za,
lines of eachRZE in the sample, as calculated from
a ZAF procedure,and b) the changein peakresolution over the spectralrange. The peak-overlapcorrections (Fig. 3) and data from White & Johnson
(1979)were usedwith equationsl) and 2) to calculate (BASIC program) the spectrumshown in Figure
4a. This spectrumwas calculatedassuminga sample with equal intensity of the Zo, of the R.EE; it
compiles favorably with the measuredscan shown
in Figure 4b. The measuredscanhas a variable background and shows changesin the efficiency of Xray generationand of the spectrometeras a function of wavelength. The underlined REE in Figure
4 are thosethat have a peak-overlapcorrectionless
than 190 for the Zor line; thus the calculatedZcr,
peaksfor theseelementsare at the sameheight. The
similarity betweenthe calculatedand measuredscan
givesthe analyst someconfidencein the useof these
corrections for analytical work.
The effect of changingspectrometerresolutionis
shown quite dramatically by comparing the WD
spectrum in Figure 4b with the ED spectrum for the
sameglasssamplein Figure 5. The major peaksin
Figure 5c are due to aluminum, silicon and calcium,
whereasthe minor peaksare due to yttrium, scan-
G
2.6
2,O
t.4
Frc. 4. Calculatedand measuredWD spectrafor the wavelengthregion of 2.7 to l.a A. a) Calculatedspectrum,with
the elementsymbol above the lcyl line; underlined symbolsfor elementswhere.Lal needsno overlap correction.
b) Measured spectrum for glass 5-254, with ea.chREE at a nominal I wt.Vo, c) Measured spectrum for glass 5-253
withe eachREE at a nominal 0.08 wt.Vo.
ELECTRON-MICROPROBE ANALYSIS OF MINERALS FOR RARE-EARTH ELEMENTS
I
z
I
4
269
I
KeV
Frc. 5. Calculatedand measuredED spectrafor the glasssamplei5-254. a) Calculatedspectmmwith ldl peaksidentified. b) MeasuredED spectrumexaggeratedeight times that in Figure 5c. c) MeasuredED spectrum(0-9 kV) showing
major peaks of Al, Si, Ca and minor peaks for Y, CaKP, Sc and REE
dium and the fourteen REE The spectralregion that
includes the REE is shown separately in Figure 5b
with an eight-fold vertical exaggeration.The measured ED spectrumwas run for l@0 secondsin order
to try to smooth someof the statisticalvariation that
is apparent for a samplethat contains eachREE at
only a l9o level. An ED spectrum(Fig. 5a) was calculatedusing the peak-width variation shown by our
ED spectrometerand the samedata as for the WD
spectrum calculation. The calculated (Frg. 5a) and
measured(Fig. 5b) ED spectraare very similar except
for the changingbackground shown in the measured
$pectrum.The calculatedWD spectrum(Fig. 4a) and
the ED spectrum(Fie. 5a) cannot be exactly matched
becausethe horizontal scaleupedon the WD spectrum is in wavelength units (A), whereasthe scare
usedfor the ED spectrumis in energyunits (eV). The
LaLal and Cel,rrl peaks are completely resolved
using the WD spectrometer,but there is considerable overlap using the ED spectrometer.The calculated ED spectrumshould show the Lal peaksof all
the rare earths at approximately the sameheight as
theLaLal and Ce.Lculpeaksif the overlap corrections were low. Only La and Ce have small corrections by EDS, whereassevenREE(La, Ce, Nd, Sm,
Tb, Dy, Yb) have a correction of lessthen 190 by
WDS. It is obvious that REEanalysis by EDS is very
difficult or impossible unless one uses a multiple
least-squarestechnique such as that developedby
Tracor Northern. This has beenusedquite successfully by Robinson (1979)for REE ar levelsgreater
than 0.5V0in zircon and monazite.
ANALYSIS
OF MINERALS
FOR THE REU
The microprobe analysisof minerals for the REE
is difficult becauseof the scarcityof suitablestandards, the problem of peak overlap, the difficulty
of measuringthe backgroundin somesamples,and
matrix effects. The ten RZ? phosphatesdescribed
earlier have been used as standards to analyze for
REE in the four Drake & Weill glassesand also the
glassespreparedin the presentstudy. Sinceno REE
phosphatesfor Gd, Tb, Eu or Lu were availableat
the time of this study, the Drake & Weill glasseswere
usedas a primary standardfor theseelements.Drake
& Weill (1972) gavethe original weighed (nominal)
concentration of kEE in these glasses and also
presentedresults of neutron-activation analysis for
some of the REE. Both these values are listed in
Table I togetherwith the microprobedata from the
present study. The analytical results for all four
270
THE CANADIAN MINERALOGIST
glassesof Drake & Weill are shown in one column.
The Zo1 line for the elementsPr, Eu, Ho, Er and
Tm generallyneedsan overlap correction, so some
analysts @xley 1980) use the lower-intensity ZB,
line. Both lines for theseelementswere usedin the
presentstudy in order to comparethe results. The
analytical results shown in Tables I and 3 were
obtained during two microprobe sessions,one session at 15 kV and one at 25 kV, and only the normal precautionswere taken for analyzing major elements.Each microprobeanalysiswas correctedfor
matrix effects and, if necessary,for peak overlap
usingthe calculatedcorrectionsgivenin Table 2. Two
examplesare given in parenthesesClable l) ofvalues
in weight 9o before peak-overlapcorrections.Thus
the nominal weight 9o Pr of the Drake & Weill glass
is 3.79 and the overlap-uncorrectedPr value using
the PrIo, line is 4.12, whereasthe amount corrected for peak overlap is 3.71. The microprobe
procedureusedin the presentstudy may not justify
listing the seconddecimalplacein Table l, but it was
thought useful to include it. Each microprobe analysis representsan averageof a number (z shown in
parentheses
after kV) of SO-second
countsat different positionson the sample.The glasses5-253 and
5-254 were analyzed a sufficient number of times
to justify calculating the homogeneityindex of Boyd
et al. (1967). Nl REE in these glasses have a
homogeneityindex lessthan 3, and most are lessthan
2. Samples5-254 and 5-253 were also analyzedat
25 kV becauseof the low REE concentration,and
more attention waspaid to the measurementof background during the sessionat 25 kY. The background
was measured at seven different positions (i.e.,
wavelengths)which, by calculation, showedthe least
interference by kEE peaks, and the best-fitting
curved line through thesepoints was used for the
background. The glassesprepared in the present
study (5-236, 253,254) rrere sent to two laboratories for neutron-activation analysis. These results
TABI,E 3.
Batueslter
et.
La
Ce
Pro
Pro
Nat
SE
Eua
EoB
Gdo
Z
carF
Tb
Dt
Prob€
gh@
zX.X
29.7
2.1<1.8'
2.L
6.7
0.8
0.1(0.6)
0
0 (3.6)
0.1
00
00
22.7
30.0
2.6
2.6
6.5
0.3
0.1
0.1
0.1
0.1
(NAA) are listed in Table l, and no attempt hasbeen
made to reconcile these values with the nominal
valuesor results of microprobe analyses.
The true test of any microprobe procedurelies in
its suitability for minerals. The samplesshown in
Table 3 wereanalyzedat 15kY during the samesession as for the data in Table 1. The only difference
in procedurewas the addition of a GdZol measurement, which was not done on the glass samples
becauseof the high content of Ho. All the samples
in Table 3 were analyzed by wet-chemical analysis
of mineral separates.Many of the mineral grains
contain inclusions(e.g., biotite), which meansthat
the microprobe data will not be directly comparable with the wet-chemical analytical data. The correction factors for peak overlap can be appreciable
in natural samples becausethere is often a large
amount of one or two of the light kEE, suchas La
or Ce, and thesemajor elementsmake a significant
contribution at the peak positions for someof the
heavier trace REE.. A good example is the high
for
uncorrectedGd content (shownin parentheses)
the bastnaesite,parisite and monazite becauseof the
high La and Ce content. Thus large amountsof La
and Ce make it alrnost impossible to analyze for
minor amounts of Gd with any confidenceusing the
Zol line. The Gd.LB1line is probably much better
at very low concentrations,but only if there is no
Ho present.It should be emphasizedthat the overlap correctionsusedin the presentstudy may be quite
different from thoseusedby other investigators,and
depend upon the resolution and efficiency of the
spectrometers.
The microprobe data given in Tables I and 3 show
that reasonableanalysesof minerals for the REE are
possibleat levelsgreaterthan I Voif adequatepreparation and careare taken. The microprobedetermination of the REE in the Drake & Weill glassstandards agxeeswell with the results given by Drake &
Weill (1972).This agreementimplies that the factors
CONCENTRAIIONOF RARE-AARIE ELE}'ENIS IN SSLEC1TDMIMRALS
Parl.al,tsr
Probe
Ch@
l{o@lter
Frob€
Ch@
AlleLtst
Ptobs
10.8
22.7
2.6(3.8)
2.6
10.1
t.7
0.3(0.9)
0.1(0.2)
1.4(4.0)
1.3
0.1
0.1
16.6
16.1
2E.6
27.O
2.6<4.4) 2.6
2.4
2.6
8.3
7.6
1.1(1.r)
0.6
0.2(0.8)
0.1
0
0.1
0.2(3.6)
o.4
0.1
0.4
o
0.1
oo
4.3
6.9
0.9 (1.4)
0.6
2.O
0.3(0.3)
0 (0.2)
0.4(0.4)
0.1(0.9)
0.1
00
0.3
11.7
2L.6
2.1
2.4
8.6
1.3
0.r
0.1
1.1
l.X
0.1
0.3
Ch6
3.5
5.7
0.6
0.6
L.7
0.2
0
o
0.1
0.1
Apatl.t€r
Probe
Cb@
ADatl'tet
Probe
cbeD
0.4 .
0.4
0(0.1)
0.4
0.5
0.1
o.2
0.E
0.4
1.0
0.1
0.2
0
r) uoutet!
Pass' Cellfonla:
Provlatsd tt A.tr. t{arle
(cho)
and @1yzed
by set-ch@lqa1
rthode
tDl !t. Aholat of
r) Stroblrd lll.ae, lloDtam, U.S.A.
!) prospect mrtb
uolycorp loc.' LowLEra' colorado, u.s.A.
Aoalyate ty u. Araolil.
r) cano de t{ercado, t}uralgo, l.lad.co. young et al.
of Tldne,
ontarlo,
camata. Analysls by u. Arnolil.
(fiOS).
r) Eurttlaraftelil Tmhtp,
(L974').
at al.
Qu€b€c. trzctoaLt
ELECTRON-MICROPROBE ANALYSS
OF MINERALS FOR RARE-EARTH ELEMENTS
271
involved in the ZAF correction procedure have Colsv, T.W. (1968):MAGIC ComputerProgramfor
worked well to correct for the differencein the matrix
Quantitative Electron Microprobe Anolysis. Bell
TelephoneLaboratories, Allentown, Pa.
of the phosphatestandardsand silicateglasses.At
levelsof0.19o to l9o, the analyticalresultsare still
reasonablygood, but the problemsof overlap cor- Conr,Brr,M.I. & McDoNeLo,M. (1974):Quantitative
analysisof sulphidesandsulphosaltsusingan energy
rection and background measurement become
dispersivespectrometer.MicrobeamAnalysis Soc.'
increasingly important at the lower levels. The
9th Ann. Conf., 234-231.
problem of accuratemeasurementof background
(Smith & Reed l98l) and overlap correction for con- Dnar<s,
M.J. & Wnu.l,D.F. (1912):Newrareearthele'
centrationslessthan 0.190are very significant, and
ment standardsfor electronmicroprobeanalysis.
results at this level should be viewed with a healthv
Chem.Geol.10, 179-181.
scepticism.
R.A. (1980):Microprobestudiesof REE-rich
E><r.sv,
accessoryminerals: implications for Skye granite
ACKNowLEDGEMENTS
petrogenesis
and REE mobility in hydrothermalsystems.Eafth Planet.Sci.Lett.48, 97-110.
The author gratefully acknowledgesthe help and
stimulation of A.N. Mariano, rrho also provided
Haccsnrv,S.E. (1983):The mineralchemistryof new
many of the samples.L.A. Boatner, Oak Ridge
titanatesfrom the Jagersfonteinkimberlite, South
National Laboratory, Oak Ridge, Tennessee,is
Africa: implicationsfor metasomatismin the upper
thanked for the REE phosphatesamples,as is M.
Acta 47, 1833-1854.
mantle.Geochim.Cosmochim.
Arnold, Molycorp Inc., Louviers, Colorado, for
allowing the results of the wet-chemical analysesto Rsro, S.J.B. (1975):Electron Microprobe Analysis.
be published. Considerable assistancewas also
CambridgeUniversityPress,Cambridge,England.
provided by D. Kempson, C. Peck and J. Jeffrey.
G.W, (1979):The Occurrenceof RareEarth
M. Corlett, Queen'sUniversity, G. Pringle, Geolog- Rosnrsoi.r,
Elementsin Zircon. Ph.D. thesis,Queen'sUniverical Surveyof Canadaand D.G.W. Smith, University, Kingston,Ontaqio.
sity of Alberta, are thanked for their comments on
the manuscript. Financial support provided by the
T.V. & Lenssx,A.O. (1978):GadoliniteNatural Sciencesand Engineering ResearchCoun- SsceLstAD,
Osloregion,Norway.
(Ce)from Skien,southwestern
cil of Canada is very gratefully acknowledged.
Amer. Mineral. 63, 188-195.
D.G.W. & Rnno,S.J.B.(1981):The calculation
Sr',rrrn,
'
electron
of background in wavelength-dispersive
Avu, n. & Gnrrnr, W.L. (1975):Microprobeanalymicroprobeanatysis.X-Roy Spectrom.t0' 198'202.
sisof REEminerals usingempiricalcorrectionfactors. Amer. Mineral, 6O, 599-606.
Suonru, P. (1980):Componentsof the total X-ray
scattering. Proc. Symp. Accuracy in Powder
BoanvBn,
L.A., BeALr,G.W., Annarnu,M.M., FrNcH,
Diffraction (S.Block & C.R. Hubbard,eds.).Nal.
C.8., Hunev,P.G. & Reppaz,M. (1980):Monazite
Bur, StandordsSpec.Publ,567.
and other lanthnnideorthophosphates
as alternate
actinidewasteforms. .In ScientificBasisfor Nuclear TnzcrcNsrt,W.E., Jn., PBnneurr,G. & HEsenr,P.
(1974):A noteon apatitefrom HuddersfieldTownWasteManagement2 (C.J.M. Northrup Jr., ed.).
PlenumPublishingCorp., New York.
ship, Quebec.Can. Mineral, 12,289-291.
REFERENCES
G.G" Jn.(1979):X-ray emisBor-oN,R.8., McCoNwsr,r,
M.D. & Gnq M.E. (1979): Wnrre,E.W. & JonNsoN,
Determinationof X-ray WDS efficiencyfor absosion and absorption edgewavelengthsand interlute and relative intensity measurements.In
changesettingsfor LiF gearedcurvedcrystal specMicrobeamAnalysis- 1979(D.E. Newbury,ed.).
trometer.Earth Minerol. Sci.Spec.Pub. l'70'The
SanFranciscoPressInc., SanFrancisco,Calif.
PennsylvaniaStateUniv., UniversityPark, Pa.
E'L. & Conrr.IN,
A.T., MuNsoN,
Borz,R,E. & Tuvr, G.L. (1976):CRCHandbookof YouNc,8.J.,MvBRs,
TabtesofAppliedEngineeringScience.
CRCPress,
N.M. (1969): Mineralogy and geochemistryof
Cleveland,Ohio.
fluorapatite from Cerro de Mercado, Durang,o,
Mexico.II.S. Geol, Surv.Prof. Pap.650'D,8493.
Bovo,F.R., FNcgr, L.W. & Cneyss,F. (1967):Computer reductionof electron-microprobe
data.Car- ReceivedAugust13,1984,revisedmanuscriptaccepted
negieInst. Wash.YearBook 61,21U215.
December5, 1984.