American Mineralogist, Volume 86, pages 473–484, 2001

Cathodoluminescence study of apatite crystals

JOCELYN BARBARAND1,* AND MAURICE PAGEL2

1
CRPG-CNRS, B.P. 20, 54501 Vandoeuvre-lès-Nancy Cedex, France
2
Département des Sciences de la Terre, UMR Orsayterre, Université Paris-XI, 91405 Orsay Cedex, France

ABSTRACT
Cathodoluminescence (CL) spectrometry represents a promising technique for the analysis of
trace-element concentrations and distributions in minerals. However, a higher precision and a stan-
dardization of the recording conditions are required to use CL spectral data quantitatively. A signifi-
cant step towards a more quantitative treatment of CL spectra is presented in this study.
A procedure to correct the spectra for the various efficiencies as a function of the wavelength of
the CL detector is proposed using low-pressure mercury-vapor and quartz-iodine lamps. CL spectra
presented in this study are thus corrected for the system response. Apatite CL spectra, which are
commonly composed of two broad bands centered at 3.5 and 2.2 eV, are deconvoluted to isolate
component bands and determine their areas. The crystallographic control by prismatic or basal sec-
tions of apatite on spectral intensities is significant and only prismatic sections should be used.
Signal decrease associated with electron bombardment (electron beam aging) is exponential and
appears drastic in the first hundred seconds but continues even after 15 minutes of beam bombard-
ment.
All observed CL bands could be correlated with a specific activator [rare earth elements (REE)
or manganese]. The 3.5 eV band is composed of three bands at 3.59 eV, 3.29 eV, and 2.87 eV. Ion
microprobe results and comparison between CL and photoluminescence data support Ce3+ activa-
tion for the origin of these bands.
The relationship between CL band intensity and REE concentration measured by ion microprobe
analysis demonstrates that CL also can provide semi-quantitative data for Gd3+, Ce3+, Dy3+, and Sm3+
when recording conditions are strictly controlled.

INTRODUCTION numerous overlapping bands. Less-intense bands are not ob-

Cathodoluminescence spectroscopy has been used for a long
time in petrographic studies to determine the provenance of
mineral grains or to distinguish generations of growth within
minerals (Marshall 1988 and references therein; Machel et al.
1991). Recently, CL detectors have been attached to scanning
electron microprobes allowing a better control of the electron
beam flux and general instrument stability, and permitting
higher magnification and spectrometric measurements in the
UV and visible ranges.
However, CL spectrometry is still not used as a microprobe
to reveal the presence of activators in the minerals or their con-
centrations because of several limitations. (1) Spectra are
strongly dependent of the recording system used. This depen-
dence is related to the variation of efficiency as a function of
wavelength of the diffraction gratings and photomultipliers.
This system response implies modifications of the spectra in
intensity and shape of the bands between facilities (Townsend
and Rowlands 2000). (2) As an example, the luminescence of
apatite under electron bombardment is caused by the presence
of Mn and rare earth elements (REE) that substitute for Ca
(Mariano 1988; Blanc et al. 1995). Spectra are composed of
served while the true intensity of obvious bands is only esti-
mated. (3) Variation in luminescence can be linked with
crystallographic controls or electron beam aging (Marshall
1988; Murray and Oreskes 1997).
The purpose of this study is to provide methodological and
analytical data leading to an improvement in the precision of
the acquisition of CL spectra, and to a standardization of the
mechanism of recording. A procedure using reference lamps is
presented to correct for transmission variations with wave-
lengths of the detector. CL spectra are deconvoluted to indi-
vidualize most of the bands present and to obtain quantitative
data on these bands (intensity and area). Importance of elec-
tron beam aging and the variation of the luminescence con-
trolled by crystallographic orientations are investigated for some
apatite samples. To test the use of CL spectrometry as a micro-
probe, data acquired from CL spectra on various REE activa-
tors are compared with compositions determined by ion
microprobe analysis. Finally, the results of a CL study of apa-
tite crystals are presented to reveal chemical composition and
associated growing conditions.
INSTRUMENTATION

Methodology
* Current address: Department of Geological Sciences, Uni-
versity College London, Gower Street, London WC1E 6BT, CL images and spectra were acquired with an Oxford In-
U.K. E-mail: j.barbarand@ucl.ac.uk strument MonoCL2 system attached to a Philips XL30 scan-

0003-004X/01/0004–473$05.00 473
474 BARBARAND AND PAGEL: CL STUDY OF APATITE CRYSTALS
ning electron microscope (SEM). The MonoCL optical con-
figuration has been described by Kearsley and Wright (1988).
The recording system consists of the following: a fully retract-
able parabolic aluminum mirror acting as light collector, placed
immediately above the sample surface; a silica window that
allows the passage of UV emissions; a scanning grating mono-
chromator, with 1200 grooves per millimeter; a PA3 amplifier;
and a photomultiplier tube (PMT) as detector. The grating
monochromator is characterized by high wavelength resolu-
tion and linear scattering resulting, in turn, in an important loss
of intensity (Rémond et al. 1992). The PA3 amplifier provides
a complete set of support electronics including power supplies
and signal amplification. Photon pulses (photon counting mode)
are counted by the preamplifier. This arrangement provides a
better signal to noise ratio than the measure of average total
current (linear mode).
The mirror system is retractable to simplify the use of the
SEM for secondary electron (SE) and back-scattered electron
(BSE) imaging. The mirror is mounted symmetrically about
the electron optic axis allowing collection of up to 80% of
emitted photons. The minimum magnification for the system
is constrained by the aperture size of the mirror and is approxi-
mately 100×.
The spectral coverage is 200–840 nanometers (nm) and the
spectral dispersion is 2.7 nm per millimeter of slit width. The
width of the entrance and exit slits is usually 1 mm and, there-
fore, the spectral dispersion is 2.7 nm for the majority of the
spectra presented. Reducing the slits allows a better resolution
but produces an important decrease in the measured intensity.
CL spectra (X-axis) can be expressed in wavelength (nm)
or in energy (electron volt, eV) by the relation:
hc
E(eV) = or E(eV) = 1239.85
λ(nm) λ(nm)
where h is the Planck constant and c the speed of light. The
intensity (Y-axis) reports individual photon pulses.
For this study, CL spectra were acquired at an accelerating
voltage of 25 kV, a 1–5 nA beam current, and entrance and exit
slit widths of 1 mm. The analyzed zone corresponds to a rect-
angle of 40 × 30 µm (beam rastered at 3000×). Counting times of
500 to 1000 milliseconds (ms) were chosen to avoid noisy spec-
tra. The analytical procedure is a sequential acquisition and there-
fore the recording time varies between 7 and 15 minutes depending
on the counting time and the analytical step. All spectra were re-
corded from the lower to the higher wavelengths.
The photomultiplier voltage can be monitored by the soft-
ware interface. The response curve of the PMT has been deter-
mined for a low-pressure mercury vapor lamp. The linear
response domain represents a small interval between 79 and
84% of high voltage (HV) of the PMT. The same analysis was
carried out for an apatite crystal (Balazuc drill-core, Ardèche,
France) and the linear domain was found to be between 82 and
87% HV PMT. All the spectra presented in this study were there-
fore acquired within that range.
Panchromatic images were also obtained with an optical
CL Technosyn Mark II microscope. The components include a
small vacuum chamber with windows, an X-Y stage movement,
and a cold cathode electron gun centered on the optical axis of
the microscope. Analytical conditions were 15–20 kV with an
electron beam current between 200 and 400 mA. Photomicro-
graphs were taken with 1600 ISO films.
Apatite crystals were arranged in rows and columns under
a binocular microscope allowing easy location of grains. The
samples were then mounted in epoxy resin, ground using 400-
and 600-grade silicon carbide papers, polished using 3 and 1
µm diamond pastes, and coated with carbon. Apatite crystals
sampled from different localities and geological environments
were chosen for this CL study. The samples are described in
Appendix 1.
Major-element analyses of apatite were obtained using a
Cameca SX 50 microprobe and the PAP data-reduction pack-
age. Operating conditions were 15 kV, a beam current of 10
nA, and a beam diameter of 5 µm. Four spectrometers are at-
tached to the microprobe: three are equipped with PET, TAP,
and LIF crystals, respectively, for the analysis of Cl-K-Ca-Ba,
Na-Mg-Si-P-Sr, and Mn-Fe, whereas the fourth has a PC1 crys-
tal for the analysis of F. Natural or synthetic standards were
used to determine the element abundance: albite (Na), Mg ox-
ide (Mg), orthoclase (Si, K), hematite (Fe), Sr oxide (Sr), Mn
oxide (Mn), synthetic chlorapatite (P, Ca, Cl), and topaz (F).
Counting time was 30 seconds for F, 20 for Mg, Si, Cl, Sr, Mn,
and Fe, and 10 for Na, P, K, and Ca. Three replicate analyses
were averaged for each grain analyzed. Analyses in which the
sum of the different oxides was lower than 97% were rejected.
Trace-element abundances were determined with a Cameca
IMS 3f ion microprobe on gold-coated sections. The samples
were sputtered using a 10 mm diameter primary O– beam rang-
ing in intensity from 2 to 5 nA. Positive secondary ions of REE
were analyzed at low mass resolution (M/DM = 400) and with
an energy filtering of –60 ± 10V. The ionic intensities, mea-
sured for each isotope, were normalized to that of 43Ca. Oxide
interferences remaining after energy filtering were eliminated
using deconvolution techniques applied to the measurements
of 31 atomic masses in the range 138 to 180 (Fahey et al. 1987).
The absolute atomic ratios normalized to Ca were then obtained
from the deconvoluted ratios and the ion yields of each REE.
These ion yields were determined from the analysis of the
Durango standard apatite (Young et al. 1969; Reed 1986).
Photoluminescence (PL) spectra were measured on some
apatite crystals to study their UV luminescence and compare
these results with those of CL. Luminescence spectra were in-
vestigated with a UV laser (337 nm) that delivered pulses of
10 ns duration and 0.1 cm–1 spectral width. The pulse energy
was maintained at about 10 mJ. The luminescence was observed
with an incident angle of 90°, analyzed by INSTASPEC equip-
ment enabling time-resolving spectra acquisition and detected
by an intensified CCD matrix. Delay times and strobe pulse
duration are between 20 ns and 9 ms; spectral detection ranges
between 350–900 nm. The system is combined with an optical
microscope, enabling spectra acquisition at the microscopic
scale. Band assignment was done from the data of Tarashchan
(1978).
Correction of the CL spectra
Figure 1 illustrates the necessity of correcting the spectra
for the system response by comparing spectra of the same
 BARBARAND AND PAGEL: CL STUDY OF APATITE CRYSTALS
FIGURE 1. CL spectra of Durango apatite acquired by sev-
eral authors with different recording systems: (a) Roeder et al.
1987; (b) Murray and Oreskes 1997; (c) and (d) this study.
Note that the Y-axis scale is not the same for the different spec-
tra. Observation with a system efficient in the UV range (c)
shows that the maximum of intensity is reached for Ce3+ and
not for Eu2+ as indicated by (a). Analytical conditions: (a) Nu-
clide Luminoscope—spectra were recorded with a NM-3 mono-
chromator system including an SC-1 scanning control, a D-43
extended spectral range (250–930 nm) side-looking photomul-
tiplier detector, and model DR-1 digital radiometer; (b) Patco
ELM3R Luminoscope—spectra were recorded with a
SPISUARC spectrometer (Acton Research Corporation)
mounted with a Princeton Instruments CCD camera and an
Olympus microscope (10 kV, 0.5 mA, 45 seconds exposure
time); (c) and (d) MonoCL2, 25 kV, 1.3 nA, magnification of
1000, slits width of 1 mm, 85% of high voltage of the photo-
multiplier tube (HV PMT), exposure time of 500 ms per nm.
sample acquired with different CL devices. Differences among
the three spectra presented in Figure 1 are associated with the
CL apparatus efficiency as a function of wavelength. For scan-
ning gratings, the maximum efficiency is controlled by the
“blaze” applied during manufacture of the grating (Townsend
and Rowlands 2000). In the same way, PMT are strongly wave-
length dependent. Spectral response of the PMT used in this
study displays maximum efficiency at 260 nm and is reduced
by a factor of 10 at 700 nm. Consequently, luminescence of a
given CL apparatus represents biased information about the
original luminescence of the sample. The recorded signal dif-
fers greatly from that of the original emission. Determination
of the system response of the CL detector is essential to correct
spectra for discrepancies due to the method of acquisition so
that they may be compared quantitatively among laboratories.
Parts of the spectrum from a given CL apparatus where record-
ing efficiency is weak could not, however, be compared.
475
The calibration of wavelength and intensity was achieved
with reference lamps whose emission spectra are known. These
lamps work well with this type of SEM because the vacuum
chamber door can be opened wide, allowing the passage of
light. The grating monochromator is calibrated with a low-pres-
sure mercury vapor lamp with well-defined emission lines at
546.073, 576.960, and 579.066 nm. The monochromator can
be driven mechanically to the appropriate positions, which en-
ables precise acquisition of data. The calibration of the appara-
tus can thus be easily controlled.
To cover the entire range of the spectrum, two lamps whose
spectra are broad bands with no line are needed. For the UV
range (200–400 nm), a deuterium lamp whose spectrum corre-
sponds to a broad band between 160 and 380 nm was used
(Fig. 2a). For the visible and near IR portions of the spectrum
(400–750 nm), a quartz-iodine lamp was selected (Fig. 2b).
For both lamps, the system response (Fig. 2c and 2d) corre-
sponds to the intensity of the reference spectrum divided by
the intensity of the measured spectrum. Correction of a CL spec-
trum requires multiplication of the raw spectrum by the system
response. All spectra presented in this study are corrected for
the system response.
Deconvolution of the CL spectra
Using calcite, El Ali (1989) has demonstrated that CL spec-
tra can be deconvoluted into gaussian curves (Kopp et al. 1995;
Habermann et al. 1996; Townsend and Rowlands 2000). For
deconvolution of broad emission spectra into component bands,
luminescence spectra were plotted as a function of energy. This
conversion needs the transposition of the Y-axis by a factor λ2/hc
(Townsend and Rowlands 2000). The intensity is then expressed
in arbitrary units of photon counts.
Deconvolution was made with a least-squares fitting algo-
rithm that minimizes the difference between the experimental
spectrum and the sum of the gaussian curves. IGOR software
(Interactive Graphic Operation for Research) from Wavemetrics
Inc. was used for deconvolution. Based on presumed band num-
bers and wavelengths, iterative calculations give the band po-
sitions that correspond to the best fit between the spectrum and
the sum of calculated bands. The usual procedure is to start
with a one or two bands and to increase the band numbers until
the deconvolution does not significantly improve. Band posi-
tions can be constrained or assumed for calculation.
Deconvolution of the CL spectra can reveal different acti-
vators and establish their relative importance, particularly where
some activators are overlapped by others with broader or more
intense emission bands. It permits precise measurement of the
band parameters, such as the area or the width at half height, a
first step to quantification. Examples of deconvolution of CL
spectra are presented in Figure 3.
Deconvolution must be carried out with care and the fol-
lowing limitations must be taken into account. (1) Define the
level at which no additional band needs to be added. (2)
Deconvolution of 10 bands in one spectrum represents the limit
beyond which the results given by the algorithm are not valid.
(3) Deconvolution is difficult in the 2.5–4 eV range of the spec-
trum where most of the activators are unknown. (4) For a simi-
lar correlation between the sum of the gaussian and the
476 BARBARAND AND PAGEL: CL STUDY OF APATITE CRYSTALS
FIGURE 3. Deconvolution treatment of two CL spectra. The
bold curve corresponds to the sum of the gaussian curves. (a)
Arendal apatite spectrum characterized by REE and Mn bands;
(b) Ødegårdens Verk apatite spectrum with a high Ce band but
lacking the Mn band.
FIGURE 2. CL spectra of ref-
erence lamps and correction
curve for the MonoCL2 installa-
tion at the University of Nancy I.
(a) reference spectrum supplied
by the manufacturer (1), and
spectrum acquired with the
MonoCL2 (2) for the deuterium
lamp (CL conditions = 83% HV
PMT, slit widths of 1 mm); (b)
reference spectrum (Mazères
1997) (3) and spectrum acquired
with the MonoCL2 (4) for the
quartz-iodine lamp (CL condi-
tions = 83% HV PMT, slit widths
of 1 mm); (c) system response of
the MonoCL2 obtained with the
two previous lamps; and (d) de-
tail of system response for the
range 300–700 nm.
spectrum, two bands may lead to an ambiguous interpretation.
The addition of a third band refines the two possibilities (Fig.
4). (5) Finally, it appears difficult to superimpose two bands or
to discriminate bands whose energies are very close. An ex-
ample is the Sm3+ band at 2.19 eV, that cannot be isolated by
deconvolution (other than by inference with other Sm3+ bands
that are present at 2.07 and 1.89 eV) because of the presence of
the Mn2+ band at 2.18 eV. These limitations illustrate that not
all bands present in a CL spectrum may be assigned to a spe-
cific activator by deconvolution into gaussian curves.
CONTROLS ON CL EMISSION
Importance of the recording conditions
Instrumental conditions are important in the recording of
CL emission. Variation of beam intensity strongly modify the
CL spectra, the relative intensity, and the ratio between the dif-
ferent CL bands. Figure 5 shows spectra from an apatite grain
from the 1245 m Balazuc sandstone determined using differ-
ent beam currents. For a current <1 nA, only the band at 3.97
eV can be seen, whereas for a current >3 nA, the broad band
centered at 3.29 eV appears. A smaller band around 2 eV can
also be discriminated (2.07 and 1.91 eV). Operating conditions
must therefore be specified precisely for any CL spectra.
Activator assignment
Band assignment was based principally on data from syn-
thetic chlorapatite crystals doped with various REE, Mn2+, and
U4+ (Blanc et al. 1995). These data are in good agreement with
previous work (Mariano and Ring 1975; Roeder et al. 1987;
Mariano 1988). CL band energies associated with the different
activators presented in this study are the mean values of the
BARBARAND AND PAGEL: CL STUDY OF APATITE CRYSTALS 477

bands from numerous analyses carried out on the various

samples. The error for these energies is the standard deviation
of the mean when more than five spectra were used.
Two broad bands are usually present in the apatite crystal
spectra—one centered at 3.5 eV and the other at 2.2 eV—in
addition to several narrower bands that correspond to REE ac-
tivators. Trivalent rare earth ions have the electron configura-
tion 4f(k)[5s25p6], with the exceptions of La, Ce, Gd, and Lu.
The 5s25p6 electrons form two completely filled shells, which
partially shield the 4f subshell. These electron configurations
give rise to characteristic narrow luminescence bands through
an f → f transition (Yang et al. 1995). The interaction between
the 4f electron and the crystal field is weak and, therefore, the
emission band wavelengths vary little among different miner-
als (Rémond et al. 1992). Luminescence emission can also take
place in the 5d electron shell that interacts strongly with the
crystal field. The transition f → d yields a broad band (La, Ce,
Lu). The bands of Gd3+ (3.97 ± 0.006 eV), Dy3+ (bands at 2.57
± 0.006, 2.15 ± 0.005, and 1.86 ± 0.012 eV), Er3+ (3.08 eV),
Eu2+ (2.75 eV), Sm3+ (2.19 ± 0.006, 2.07 ± 0.003, and 1.91 ±
0.003 eV), and Tb3+ (3.25, 2.98, 2.83 ± 0.005, 2.53, and 2.27 ±
0.006 eV) can be identified by deconvolution of the CL spectra.
An Mn2+ band (2.15 ± 0.02 eV) is present as a broad band
that overlaps with the Dy3+, Sm3+, and Tb3+ bands. Mn2+ is a
transition metal ion that has an electron configuration of 3d5
and interacts strongly with the crystal field (d → d transition).
▼

FIGURE 4. Deconvolution steps of a CL spectra. The

deconvolution into one or two bands does not provide a good
fit between the experimental spectra and the sum of the gaussian
peaks. Introduction of two peaks gives two contrasting solu-
tions with differing degrees of confidence. The addition of a
third peak markedly improves the result and leads to a repro-
ducible unique solution.

▼ FIGURE 5. The same apatite crystal from the Balazuc

sample analyzed at different beam currents: (a) 0.3 nA, (b) 1.3
nA, and (c) 4.9 nA. The intensity of bands increases with the
current density, and the band ratios (for example, Gd/Er) evolve
for the different analytical conditions (CL conditions = 25 kV,
slit widths of 1 mm).
478 BARBARAND AND PAGEL: CL STUDY OF APATITE CRYSTALS
The center and width of the Mn2+ band are very sensitive to
their nearest neighbor atom surroundings in the given crystals
(Mariano 1988; Yang et al. 1995). The difference between the
position of Mn2+ band in this study and that proposed by Blanc
et al. (1995), 2.19 eV, may be explained by the fact that the
spectra presented here are expressed in eV as opposed to nm in
Blanc et al. work. This shifts the maximum of broad band to
lower energy as spectra are multiplied by K/wavelength (where
K = 1239.85). This shift is found only for broad bands because
their maximum is defined by several points.
After deconvolution, the broad band centered at 3.5 eV is
composed of three bands at 3.59 ± 0.08 eV, 3.29 ± 0.06 eV, and
2.87 ± 0.09 eV, respectively. This band is often the most in-
tense spectral band, and its maximum intensity could be as-
sociated with either one of the bands at 3.59 or 3.29 eV. The widths
at half-height of these two bands range between 0.3 and 0.6 eV.
The origin of these three bands (3.59, 3.29, and 2.87 eV) is
still controversial and two causes are usually proposed: intrin-
sic luminescence or Ce3+ activation (Barbin and Schvoerer
1997). Although Tb3+ and Tm3+ have bands in this range, their
activation can be dismissed because their intensity and width
characteristics do not match the observed spectra.
The term “intrinsic luminescence” is used improperly
(Bresse et al. 1995), but refers in this study to emission in rela-
tionship to lattice defects such as charge recombination cen-
ters, defects originally present in the crystal, or defects induced
by electron bombardment. This type of luminescence has been
proposed for both silicate and carbonate minerals. Iacconi and
Caruba (1980) showed that in silicate minerals, intrinsic lumi-
nescence forms a broad band localized in the near UV or vio-
let-blue; in carbonates, Barbin and Schvoerer (1997) considered
that blue or near UV luminescence is associated, at least partly,
with lattice point defects because no correlation had been es-
tablished between an activator (Pb2+, Eu2+, or Ce3+) and this
luminescence.
Ce3+ activation has three bands at 3.54, 3.26, and 2.71 eV
(Blanc et al. 1995). Ce3+ is characterized by an f → f transition
and also by an f → d transition located in the near UV (Cesbron
et al. 1993). This implies that the emission bands are broad and
are, therefore, compatible with the 3.59 ± 0.08, 3.29 ± 0.06,
and 2.87 ± 0.09 eV bands.
This study suggests that these three bands are related to Ce3+
activation. The three Ce3+ bands have been revealed by spec-
tral deconvolution. The relationship between the Ce3+ concen-
tration and the 3.6–3.3 eV band areas is discussed below (see
“Semi-quantitative determination of some REE contents with
CL spectrometry” and Figs. 8c and 8d). This relationship dis-
plays quite a good correlation indicating that UV activation is
controlled by the presence and concentration of Ce3+ ions.
Samples with a very low Ce3+ concentration appear to be char-
acterized by a weak UV emission.
Comparison of CL and PL spectra shows that when UV
emission is high in the CL spectra it is also high in the PL
spectra. PL spectra are time-resolved, allowing discrimination
of the different activators. UV emission is associated mainly
with Ce3+ and also with Eu2+, each being characterized by short
decay times. The Ce3+ band wavelength is shifted to a higher
value (395–400 nm) because the laser used (excitation wave-
length of 337 nm) has its optimal output near 400 nm. Spectra
acquired with lasers having a lower excitation wavelength (266
nm) are characterized by Ce3+ bands at 365 and 382 nm
(Tarashchan 1978). These positions are similar to those found
by CL. Combination of CL and PL data allows intrinsic lumi-
nescence for the UV band activation to be eliminated because
photon energy is insufficient to cause near band gap radiation.
These results argue in favor of Ce3+ activation for UV emission
in CL spectra.
Introduction of Ce3+ ions in the apatite structure is claimed
to produce vacancies or defects, and therefore to be associated
with intrinsic luminescence (Barbin and Schvoerer 1997). This
hypothesis is not retained here because the introduction of La3+
and Pr3+ which occur in the same Ca site [Ca(2); Hughes et al.
1991], should result in high UV emission as well (Blanc et al.
1995).
The intensities or area ratios between the two bands at 3.3
and 3.6 eV are, however, variable among the samples, and this
variation is not easily associated with a single activation.
Luminescence variations
Crystallographic control on the CL signal. To study the
crystallographic dependence of CL in apatite crystals, CL spec-
tra were acquired for various samples (Appendix 1) on pris-
matic and basal sections or on prismatic sections but with a
90° rotation of the mount before new acquisition. These spec-
tra were obtained from different zones to avoid electron beam
aging but from the same sector as revealed by spectroscopic
images.
A centimeter-long apatite crystal from Durango, Mexico,
was cut perpendicular and parallel to the c-axis. Numerous
spectra were acquired for each section using the same analyti-
cal conditions (Fig. 6a and 6b). The shapes of the spectra are
the same for both crystallographic orientations. The major dif-
ference in the spectra is their intensity, with the mean intensity
for the basal section lower than that for the prismatic face.
For sample Gun-F (Gunlock, Utah), spectra were acquired
before and after a 90° rotation of the slide. Normal orientation
acquisitions do not display significant difference (Fig. 6c).
However, another area from the same sector of the crystal ana-
lyzed parallel to the c-axis (prismatic section) yields a strongly
different spectrum.
For a Lestrezes (Margeride, France) granite apatite, normal
rotation of the crystal led to different results (Fig. 6d). Spectra
are composed mainly of the Ce3+ and Mn2+ bands. The 90° ro-
tation results in an inversion of the relative intensities of the
two bands: the Ce3+ band is significantly higher for the zone
analyzed in a section perpendicular to the c-axis but the Mn2+
band is significantly higher for the zone analyzed in a section
parallel to the c-axis (Fig. 6d).
These results show that in a perspective of quantitative
measurement of the activator content, crystallographic orien-
tation must be carefully controlled. Analysis should be done
on prismatic sections for maximum efficiency. Differences
between prismatic and basal sections may be explained by the
diffusion mechanism of elements along the c-axis, as already
demonstrated for electron microprobe analysis for F and Cl
(Stormer et al. 1993). Diffusion rate appears higher along the
 BARBARAND AND PAGEL: CL STUDY OF APATITE CRYSTALS 479

FIGURE 6. The effect of crystal orientation on the CL spectra. Prismatic (a) and basal (b) sections of a Durango apatite crystal
(CL conditions = 25 kV, 1.3 nA, 50 × 50 µm analyzing zone, 85% HV PMT); (c) normal acquisition of three spectra from a Gun-
F apatite crystal. Spectrum 2 was acquired after a 90º rotation of the slide. Spectrum 3 represents another zone in the same sector
as 2; and (d) normal acquisition of a Lestrezes apatite crystal. Spectrum 2 was acquired after a 90º rotation of the slide. (CL
conditions = 25 kV, 4,9 nA, slits width of 1mm, 83% HV PMT, and acquisition time of 1000 ms).

c-axis, and could imply an important element loss recorded by
lower intensity for spectra obtained on basal sections (Fig. 6a
and b). Acquisition following a 90° rotation gives various results,
with significant changes in the spectra in some cases (Fig. 6d).
Electron beam aging. CL emission displays time depen-
dent changes in luminescent color and intensity. The decrease
of luminescence in some minerals such as quartz is referred as
electron beam aging (Marshall 1988). Possible mechanisms for
the intensity decrease in synthetic phosphors include precipi-
tation of metal from the host and destruction of luminescence
centers either by heating or oxygen loss (Marshall 1988). This
phenomenon of fading is the consequence of electronic bom-
bardment (rise in temperature, creation of an electric field).
Ramseyer et al. (1989), however, demonstrated for quartz, that
thermal effects are not responsible for alteration of the lumi-
nescence.
To examine this phenomenon in apatite, time intensity plots
were determined for different wavelengths of three activators
Ce3+ (3.59 and 3.29 eV), Dy3+ (2.57 eV) and Mn2+ (2.15 eV), in
the Gun-F, Fish Canyon, and Arendal apatites. Measurements
were done with a voltage of 25 kV, a beam current of 4.9 nA.
Sample location and SEM focusing were done in different parts
of the mount with a beam current of 0.1 nA to avoid exposing
the studied zone to the beam. Acquisition times of 500 to 1200
seconds were selected to examine signal variations for the rou-
tine analysis (7–15 minutes).
Luminescence fades rapidly during electron beam bombard-
ment (Fig. 7). Fading is usually composed of three parts: (1) a
rapid decrease in the first 100 seconds, corresponding to an
intensity reduction of 40 to 80% depending on the activator;
(2) a much smaller loss intensity follows (100–400 seconds),
and (3) a period with little change (400–1200). Luminescence
decrease seems to reach a plateau after 800–1000 seconds. Af-
ter 500 seconds, intensity is reduced to 50–85%. Consequently,
the first 100 seconds are critical for luminescence decrease.
The evolution of intensity with time is not the same for the
different activators. Mn2+ band intensity displays a drastic fall
and then a very small decrease. Dy3+ and Ce3+ bands intensities
decrease more gradually.
Rapid decrease of luminescence has been demonstrated in
480 BARBARAND AND PAGEL: CL STUDY OF APATITE CRYSTALS

FIGURE 7. Time vs. intensity plots for different activators. (a) Ce3+ band (at 3.59 eV) in a Gun F apatite crystal; (b) Ce3+ band
(at 3.59 eV) in a Fish Canyon apatite crystal; (c) Dy3+ band (at 2.57 eV) in an Arendal apatite crystal; (d) Mn2+ band (at 2.15 eV)
in a Gun F apatite crystal. The 2.15 eV band represents exactly the Mn2+ luminescence because the Arendal grain displays very
small bands for Dy3+ (2.15 eV) and Sm3+ (2.19–2.07 eV). For the three bands, CL signal fades rapidly during electron bombard-
ment (CL conditions = 25 kV, 4.9 nA, 83% HV PMT, and slit width of 1 mm).

α-quartz (Ramseyer et al. 1988), and the decrease along an
exponential decay curve has been reported for apatite crystals
(Murray and Oreskes 1997). Under their operating conditions
(10 kV, 0.5 mA), aging was observed for more than 45 minutes
of exposure. Time of exposure to the electron beam is also a
critical factor during electron microprobe analysis (Stormer et
al. 1993).
These experiments indicate that measured CL spectra may
correspond to compromised information. CL bands may dis-
play a reduced intensity after a specific period of time by elec-
tron bombardment, rather than the real intensity. Great care is
thus required when comparing different spectra.
SEMI-QUANTITATIVE DETERMINATION OF SOME REE
CONTENTS WITH CL SPECTROMETRY
Correlation between the areas beneath Gd3+ and Dy3+ CL
bands and concentration, as determined by ion microprobe
analysis, has been reported for apatite crystals recovered from
a North Sea basin sandstone (Barbarand et al. 1996). In the
present study, the relationships between CL band areas and REE
concentrations, also determined by ion microprobe analysis,
have been investigated further for apatite grains recovered from
sandstones from a drill hole in the South of the Paris Basin and
for apatite crystals from several different settings (Appendix
1). CL spectra were acquired first, followed by ion microprobe
analyses on the same areas of each apatite crystal. CL spectra
were obtained for Paris Basin samples only in the range 2.5–4
eV to study Gd3+, Ce3+, and Dy3+; spectra from the other samples
cover the full range. Band area was calculated from the gaussian
curve.
For Gd3+, the correlation between CL and concentration is
quite good if grains with more than 150 ppm Gd3+ are excluded
(Fig. 8a). Gd3+ determination by CL spectrometry is, however,
difficult because the broader Ce3+ band tends to mask the Gd3+
band when its intensity is great. The detection limit for Gd in
samples with such a high Ce3+ band is, therefore, poorer.
The correlation for Dy3+ is more problematic because the
2.15 eV and 2.57 eV bands overlap respectively the Mn2+ band
and the Er3+ and Tb3+ bands. Deconvolution needs absolutely
to be carried on for the 2–2.7 eV range of the spectrum to char-
 BARBARAND AND PAGEL: CL STUDY OF APATITE CRYSTALS 481

FIGURE 8. Relationships between element concentration, as determined by ion microprobe analysis, and band area for vari-
ous activators in apatite crystals of different settings (solid circles) and detrital apatite grains from Paris Basin sandstones (open
circles). (a) Gd3+ at 3.97 eV (correlation coefficent, r = 0.76); (b) Ce3+ at 3.59 eV (r = 0.75); (c) Ce3+ at 3.29 eV (r = 0.39); (d) Dy3+
at 2.57 eV (r = 0.74); (e) Sm3+ at 2.07 eV (r= 0.82), and (f) Sm3+ at 1.91 eV (r = 0.88). (CL conditions = 25 kV, 4,9 nA, slit widths
of 1 mm).

acterize the distinctive area of each band (Fig. 8b).
The Ce3+ content varies widely among the analyzed grains
so that a better comparison can be made between concentra-
tion and CL band area. For the 3.59 eV band, the relationship
appears linear up to 1000 ppm but is less clear for higher con-
centrations (Fig. 8c). For the 3.29 eV band, the relationship is
linear for the whole range of concentration if the Paris Basin
apatites, or those from the different settings are considered sepa-
rately (Fig. 8d). This difference could be associated with dif-
ferent analytical conditions (such as variation in the vacuum or
in the sample’s position respect to the mirror). This difference
is recorded also in the minimum amount of Ce3+ necessary to
reveal the presence of the Ce3+ band: 400 ppm for the Paris
Basin samples, but only a few parts per million for the others.
For Sm3+, the relation appears essentially linear for the two
bands (2.07 and 1.91 eV) (Fig. 8e and 8f). The 1.91 band is
outside the Mn2+ band but the 2.07 band is inside its influence
and the fact that the same result was obtained for both bands
482 BARBARAND AND PAGEL: CL STUDY OF APATITE CRYSTALS
proves the validity of the deconvolution procedures.
The relationships between the CL band area and activator
content appear very poor for grains that have higher REE con-
tents and, especially higher amounts of the HREE (which rep-
resent the majority of CL activators). HREE may act as
quenchers for other REE when a certain concentration thresh-
old is crossed. All energy is not transmitted integrally to radia-
tive energy, and the CL band intensity appears lower than might
be expected from the activator content. Such an hypothesis has
been already proposed by Townsend and Rowlands (2000).
CL spectrometry detection limits are therefore very low with
the analytical conditions used, less than 50 ppm for Gd3+, Dy3+,
Tb3+, or Sm3+, and roughly 400 ppm for Ce3+. These values are
much lower than electron microprobe or BSE analyses. This
limit is, however, dependent on the chemical environment: ac-
curacy is higher for Tb3+ at 2.27 eV or Dy3+ at 2.15 eV when
the Mn2+ band (2.15 eV) is absent from CL spectra.
USE OF CL SPECTROSCOPY AND SPECTROMETRY FOR
THE STUDY OF APATITE CRYSTALS
The association of CL spectroscopy and spectrometry is a
powerful petrographic tool to differentiate multiple fluid events
as well as to characterize elemental chemistry. An example is
for apatite grains recovered from the Balazuc BA1 drill hole
FIGURE 9. Different images of a single apatite grain from a Balazuc Triassic sandstone. The images represent (a) Technosyn,
(b) SE, (c) BSE, and (d) CL-SEM at 350 nm.
located on the southeast border of the French Massif Central
near the Uzer fault (Pagel et al. 1996). Apatite grains were sepa-
rated from Upper Triassic sandstone at 1245 meters depth,
mounted in resin, and then polished. When viewed with an
optical microscope, the apatite grains appear homogeneous with
well-developed euhedral faces. These features are surprising
because, for most sandstones, the majority of detrital grains
should be rounded. Secondary electron and backscattered elec-
tron images do not show any crystal zoning, although they do
reveal an array of small cavities (probably fluid inclusions)
(Figs. 9b and 9c). Panchromatic and monochromatic CL im-
ages, on the other hand, do show that apatite grains are com-
posed of detrital rounded cores and euhedral overgrowths (Figs.
9a and 9d). Some grains are cut by fractures filled with bright
pink diagenetic apatite. Our CL study shows that these two
portions of composite grains have different compositions. The
yellow cores are dominated by the Mn2+ band with only very
small bands for Gd3+, Ce3+, Dy3+, and Sm3+ (Fig. 10). Spectra of
the bright pink overgrowths are free of the Mn2+ band and have
higher concentrations of Gd3+, Ce3+, Eu2+, and Sm3+. Color
changes in the grains are thus linked to activation by Mn2+ (yel-
low at 2.15 eV) in the cores and by Eu2+ (pink at 2.75 eV) at the
rims.
These results suggest that the diagenetic fluid that formed
 BARBARAND AND PAGEL: CL STUDY OF APATITE CRYSTALS
FIGURE 10. CL spectra in a detrital core and an overgrowth in apatite grains from a Balazuc BA1 sandstone (CL conditions
= 25 kV, 1.3 nA, analyzed zone of 20 × 20 µm).
the apatite rims was enriched in REE compared with the origi-
nal fluid from which the cores crystallized and also had a lower
concentration of Mn2+. The mobility of REE during diagenesis
in the Balazuc Triassic sandstone has also been shown by the
presence of REE in diagenetic anhydrite crystals (Blanc et al.
1994). The low Mn2+ content indicates a reducing environment,
an interpretation supported by the absence of U in the apatite
overgrowth as evidenced by fission track mapping.
Correction for the system response of the detector,
deconvolution of the spectra into component bands, assignment
of all activators, and control of the recording conditions (i.e.,
analysis on prismatic sections with the shortest exposure to the
beam and using the same voltage and intensity), allow CL spec-
trometry to obtain semi-quantitative data on some elements and
their distributions. CL spectra associated with CL spectroscopy
can be used to recognize and characterize heterogeneities in
apatite crystals. This approach is a highly useful way to test
the sample for homogeneity and to reveal different generations
of materials, which is particularly valuable in deciding on apa-
tite standards for use in chemical or geochemical purposes. This
fast and inexpensive technique should be used routinely in geo-
chronological (e.g., U/Pb for example) or thermochronometric
(fission-track analysis) studies to properly characterize the
material.
ACKNOWLEDGMENTS
This work was undertaken with the support of Elf and the GPF program.
J.B. was supported by a grant from the Ministère de l’Enseignement et de la
Recherche and by an Elf contract. The authors thank Philippe Blanc (Université
Paris VI) for the introduction to CL, Marc Chaussidon (CRPG-CNRS) for the
ion microprobe analysis, Gérard Panczer (LCPML, Université Lyon I) for the
photoluminescence spectra and Jeanne Paquette and Darryl Henry for construc-
tive comments on the manuscript and Walter E. Trzcienski for his editorial work.
REFERENCES CITED
Barbin, V. and Schvoerer, M. (1997) Cathodoluminescence et géosciences. Compte
Rendu de l’Académie des Sciences, Paris, 325, IIa, 157–169.
Barbarand, J. (1999) Cinétique de cicatrisation des traces de fission dans les cristaux
483
d’apatite et histoire thermique de la bordure sud-est du Massif Central, 492 p.
Ph.D. dissertation, Université Henri Poincaré, Nancy.
Barbarand, J., Pagel, M., Blanc, Ph., Braun, J.-J., Chaussidon, M., Vetil, C., and
Walgenwitz, F. (1996) Combined cathodoluminescence spectra, fission-track
age and chemistry of detrital apatite grains: towards a better understanding of
fission-track annealing. International conference on cathodoluminescence and
related techniques in geosciences and geomaterials, September 2–4, Nancy,
France. Book of Abstracts, 19–20.
Blanc, Ph., Arbey, F., Cros, P., Cesbron, F., and Ohnenstetter, D. (1994) Applica-
tions de la microscopie électronique à balayage et de la cathodoluminescence à
des matériaux géologiques (sulfates, carbonates, silicates). Bulletin de la Société
Géologique de France, 165, 4, 341–352.
Blanc, Ph., Baumer, A., Cesbron, F., and Ohnenstetter, D. (1995) Les activateurs de
cathodoluminescence dans des chlorapatites préparées par synthèse
hydrothermale. Compte Rendu de l’Académie des Sciences, Paris, 321, IIa,
1119–1126.
Bresse, J.F., Remond, G., and Akamatsu, B. (1995) Cathodoluminescence micros-
copy and spectroscopy of semiconductors and wide bandgap insulating materi-
als. 4th European workshop on modern developments and applications in mi-
crobeam analysis, May 14-19, St Malo, France, Book of Abstracts, 213–254.
Cesbron, F., Ohnenstetter, D., Blanc, Ph., Rouer, O., and Sichere, M.-C. (1993)
Incorporation de terres rares dans des zircons de synthèse: étude par
cathodoluminescence. Compte Rendu de l’Académie des Sciences, Paris, 316,
IIa, 1231–1238.
El Ali, A. (1989) Etude des roches carbonatées de réservoirs d’hydrocarbures par
résonance paramagnétique électronique et cathodoluminescence, 190 p. Ph.D.
dissertation, Université de Paris VII, Paris.
Fahey, A.J., Goswami, J.N., McKeegan, K.D., and Zinner, E. (1987) 26Al, 244Pu,
50
Ti, REE, and trace elements abundances in hibonite grains from CM and CV
meteorites. Geochimica and Cosmochimica Acta, 51, 329–350.
Habermann, D., Neuser, R.D., and Richter, D.K. (1996) REE-activated
cathodoluminescence of calcite and dolomite: high-resolution spectrometric
analysis of Cl emission (HRS-CL). Sedimentary Geology, 101, 1–7.
Hughes, J.M., Cameron, M., and Mariano, A.N. (1991) Rare-earth-element order-
ing and structural variations in natural rare-earth-bearing apatites. American
Mineralogist, 76, 1165–1173.
Iacconi, P. and Caruba, R. (1980) Trapping and emission centres in X-irradiated
zircon (III): influence of trivalent rare earth impurities. Physical State Solide a,
62, 589–596.
Kearsley, A. and Wright, P. (1988) Geological applications of scanning
cathodoluminescence imagery. Microscopy and Analysis, September, 49–51.
Kopp, O.C., Fuller, E.L., and Owen, M.R. (1995) Interpretation of cathodo-lumi-
nescence spectra obtained from dolomite and calcite gangue minerals, and
dolostone breccias in the central Tennessee Zinc District (U.S.A.). Scanning
Microscopy Supplement 9, 211–223.
Machel, H.G., Mason, R.A., Mariano, A.N., and Mucci, A. (1991) Causes and emis-
sion of luminescence in calcite and dolomite. In C.E. Barker and O.C. Kopp,
Eds., Luminescence microscopy: quantitative and qualitative aspects, SEPM,
9–25.
484 BARBARAND AND PAGEL: CL STUDY OF APATITE CRYSTALS

Mariano, A.N. (1988) Some further geological applications of cathodoluminescence.
In D.J. Marshall, Eds., Cathodoluminescence of geological materials, p. 94–
123. Unwin Hyman, Boston.
Mariano, A.N. and Ring, P.J. (1975) Europium-activated cathodoluminescence in
minerals. Geochimica et Cosmochimica Acta, 39, 649–660.
Marshall, D.J. (1988) Cathodoluminescence of geological materials. Unwin Hyman,
Boston, 146 p.
Mazères, S. (1997) Mise en oeuvre d’un microspectrofluorimètre pour l’étude de
microéchantillons en fluorescence stationnaire et résolue dans le temps, 200 p.
Ph.D. dissertation, Université Paul Sabatier, Toulouse.
Murray, J.R. and Oreskes, N. (1997) Uses and limitations of cathodoluminescence
in the study of apatite paragenesis. Economic Geology, 92, 368–376.
Pagel, M., Braun, J.-J., Disnar, J.R., Martinez, L., Renac, C., and Vasseur, G. (1996)
Thermal history constraints from combined organic matter, clay minerals, fluid
inclusions, apatite fission tracks and stable isotopes studies at the Ardeche
palaeomargin (BA1 drill hole, GPF program). Journal of Sedimentary Research,
67, 235–245.
Ramseyer, K., Baumann, J., Matter, A., and Mullis, J. (1988) Cathodoluminescence
colours of α-quartz. Mineralogical Magazine, 52, 669–677.
Ramseyer, K., Fischer, J., Matter, A., Eberhardt, P., and Geiss, J. (1989) A
cathodoluminescence microscope for low intensity luminescence. Journal of
Sedimentary Petrology, 59, 619–622.
Reed, S.J.B. (1986) Ion microprobe determination of rare earth elements in acces-
sory minerals. Mineralogical Magazine, 50, 3–15.
Rémond, G., Cesbron, F., Chapoulie, R., Ohnenstetter, D., Roques-Carmes, C., and
Schvoerer, M. (1992) Cathodoluminescence applied to the microcharacterization
of mineral materials: a present status in experimentation and interpretation.
Scanning Microscopy, 6, 1, 23–68.
Roeder, P.L., MacArthur, D., Ma, X.P., Palmer, G.R., and Mariano, A.N. (1987)
Cathodoluminescence and microprobe study of rare earth elements in apatite.
American Mineralogist, 72, 801–811.
Steven, T.A., Mehnert, H.H., and Obradovich, J.D. (1967) Age of volcanic activity
in the San Juan Mountains, Colorado. United State Geological Survey Profes-
sional Papers 575-D, D47–D55.
Stormer, J.C., Pierson, M.L., and Tacker, R.C. (1993) Variation of F and Cl X-ray
intensity due to anisotropic diffusion in apatite during electron microprobe analy-
sis. American Mineralogist, 78, 641–648.
Tarashchan, A.N. (1978) Luminescence of minerals. Kiev, Naukova dumka.
Townsend, P.D. and Rowlands, A.P. (2000) Information encoded in cathodo-lumi-
nescence emission spectra. In M. Pagel, V. Barbin, Ph. Blanc, and D.
Ohnenstetter, Eds., Cathodoluminescence in Geoscience, p. 41–57. Springer,
Berlin.
Yang, C., Homman, N.P.-O., Malmqvist, K.G., Johansson, L., Halden, N.M., and
Barbin, V. (1995) Ionoluminescence: a new tool for nuclear microprobes in
geology. Scanning Microscopy, 9, 1, 43–62.
Young, E.J., Myers, A.T., Munson, E.L., and Conklin, N.M. (1969) Mineralogy and
geochemistry of fluorapatite from Cerro de Mercado, Durango, Mexico. U.S.
Geological Survey Professional Paper, 650-D, D84–D93.
MANUSCRIPT RECEIVED DECEMBER 21, 1999
MANUSCRIPT ACCEPTED DECEMBER 23, 2000
MANUSCRIPT HANDLED BY WALTER E. TRZCIENSKI JR.
APPENDIX 1: BRIEF DESCRIPTIONS OF APATITE
SAMPLES
1. Durango apatite, Cerro de Mercado, Mexico (Young et
al. 1969), a fission-track age determination standard, is reported
to be generated by hydrothermal reworking of a primary mag-
netite-apatite igneous body.
2. Fish Canyon apatite, San Juan Mountains, Colorado
(Steven et al. 1967), is a fission-track age determination stan-
dard from a welded tuff.
3. Apatite crystals—Arendal, Bamble, Ødegårdens Verk,
Skaaland Mine (Norway), Faraday (Ontario, Canada), and Gun-F
(Utah)—of different environments and characterized by various
Cl/F ratio. A complete description is given in Barbarand (1999).
Samples courtesy of A. Hurford and R. Siddall.
4. Lestrezes apatites, recovered from a sample of the
Margeride granite from the French Variscan orogen.
5. Paris Basin apatites, separated from three Triassic
(Keuper) sandstones from the same borehole located in the
South of the Paris Basin (France), and sampled respectively at
2104 (BP1), 2110 (BP2), and 2104m (BP3). Samples courtesy
of F. Walgenwitz (Elf).