You are on page 1of 13

Contrib Mineral Petrol (2005)

DOI 10.1007/s00410-005-0037-4

O R I GI N A L P A P E R

Roland Stalder Æ Stephan Klemme Æ Thomas Ludwig


Henrik Skogby

Hydrogen incorporation in orthopyroxene: interaction


of different trivalent cations

Received: 12 July 2005 / Accepted: 14 September 2005


 Springer-Verlag 2005

Abstract The incorporation of hydrogen into ferrosilite,


Fe-bearing enstatite and orthopyroxene containing dif-
Introduction
ferent trivalent cations (Cr3+ and Al3+, Cr3+ and Fe3+)
The knowledge of the water content of the Earth’s upper
was investigated experimentally at 25 kbar. Hydrogen
mantle is crucial for understanding many upper mantle
concentration was determined by FTIR-spectroscopy on
processes such as mantle melting and physical properties
oriented crystal sections and by secondary ion mass
such as rheology (Mackwell et al. 1985; Mei and Kohl-
spectroscopy, whereas Mößbauer spectroscopy and
stedt 2000). It is now widely accepted that many nomi-
optical spectroscopy were used to characterise the va-
nally anhydrous minerals dissolve significant amounts of
lence state of Fe in orthopyroxene. Results suggest that
hydrogen under upper mantle conditions (Skogby and
hydrogen incorporation in ferrosilite is achieved by a
Rossman 1989; Bell and Rossman 1992; Ingrin and
similar mechanism as in pure enstatite. In Cr-bearing
Skogby 2000). Orthopyroxene, next to olivine the most
samples, however, hydrogen incorporation is reduced by
abundant constituent of the Earth’s upper mantle, is
the presence of other trivalent cations by an increased
able to incorporate several hundred ppm (wt) H2O
tendency to form Tschermaks substitutions, e.g. Si4+T +
(Skogby et al. 1990; Bell and Rossman 1992; Rossman
Mg2+M1 , Al 3+
T + Cr 3+
M1 . Thus, hydrogen solubility in
1996; Peslier et al. 2002; Rauch and Keppler 2002;
natural orthopyroxenes from the Earth’s mantle, con-
Stalder 2004), and might therefore be a major host for
taining significant amounts of Cr3+, Al3+, and Fe3+,
water in the Earth’s upper mantle. Water solubility in
may be much more limited than expected from their
both olivine (Bai and Kohlstedt 1992; Kohlstedt et al.
trivalent cation content, as a large fraction of the tri-
1996) and orthopyroxene (Rauch and Keppler 2002) is
valent cations does not participate in H-incorporating
strongly dependent on pressure by generating increasing
reactions as 2 Mg2+ 3+
M1 , MM1 + VM1 + Hi .
+
amounts of point defects in the lattice; in contrast to
olivine, orthopyroxene is able to incorporate trivalent
Keywords Orthopyroxene Æ Hydrogen incorporation Æ
cations on a much larger scale, thus opening the possi-
Trivalent cation Æ High pressure Æ Spectroscopy
bility for incorporation of additional water on a large
scale by other types of point defects (Rauch and Keppler
2002; Stalder and Skogby 2002; Stalder 2004), e.g.
2 Mg2+ 3+
M1 , M M1 + VM1 + Hi , where M
+ 3+
is a tri-
3+ 3+ 3+
valent cation as Al , Cr and Fe , and M1 and i are
Communicated by J. Hoefs M1-position and interstitial position in the lattice,
R. Stalder (&)
respectively. Actually, the position of the proton is not
Abteilung Angewandte und Experimentelle Mineralogie, very well defined; in some cases it could even be close to
Geowissenschaftliches Zentrum der Universität Göttingen, M1, but usually it is rather charge balancing this va-
Goldschmidtstraße 1, 37077 Göttingen, Germany cancy than filling it. In any case, H is small enough to be
E-mail: rstalde@gwdg.de in the ‘‘interstitial’’ position, i.e. not associated with any
S. Klemme Æ T. Ludwig particular vacancy, but it is always located very close to
Mineralogisches Institut der Universität Heidelberg, an oxygen, therefore also referred to as ( OH )O (e.g.
Im Neuenheimer Feld 236, 69120 Heidelberg, Germany Wright and Catlow 1994), meaning an (OH) on an
oxygen position.
H. Skogby
Naturhistoriska riksmuseet, Sektionen för mineralogi, However, the hydrogen budget of orthopyroxenes
Box 50007, 10405 Stockholm, Sweden may not just be a simple function of the total amount of
trivalent cations; for example, as demonstrated by Kle- talc-pyrex cells with alumina spacers in an end-loaded
mme and O’Neill (2000), Cr influences the Al-incorpo- piston cylinder apparatus at 25 kbar by slow cooling
ration in orthopyroxene, and therefore the presence of (11–13/h) from 1,400C to 1,150C (1,300C to 1,050C
different trivalent cations may influence the defect for ferrosilite synthesis). In synthesis runs RS084 and
chemistry of orthopyroxene significantly, which is ex- RS091 higher cooling rates (36 and 40/h, respectively)
pected to have a large impact on the extent of coupled were applied, leading to the same phase assemblage
substitutions involving hydrogen. In this study we report and slightly smaller crystals (i.e. £ 1 mm3). Tempera-
experimental results on water solubilities in ferrosilite tures were measured by S-type thermocouples; no pres-
and enstatite single crystals doped with Fe, Fe+Cr, and sure correction has been applied to the EMF of the
Al+Cr in order to assess the interaction of different thermocouple. The pressure was calibrated against the
trivalent cations in processes producing OH-defects. reactions anorthite + gehlenite + corundum = Ca-Al-
Even if it is widely accepted that hydrogen (or more pyroxene (Hays 1966) and quartz–coesite (Bose and
correctly protons) is incorporated by point defects in Ganguly 1995). To account for internal friction of the
nominally anhydrous minerals, the terms hydrogen talc-pyrex assembly, 10% were added to the nominal
incorporation and water incorporation are used synony- load; pressures are estimated to be correct within
mously throughout this article, as oxygen is the charge ±1 kbar.
balancing anion in the crystal. However, water content After the run, all charges were checked for leakage. In
and hydrogen content (wt ppm) differ by a factor of 9. all successful runs, run products consisted of large
orthopyroxene crystals (approximately 1 mm3), amor-
phous quench material and water. Hematite occurred as
Experimental and analytical procedure additional phase in runs RS064, RS072 and RS084. In
run RS054 ferrosilite was accompanied by quartz and a
Orthopyroxene crystals were synthesised by applying the greenish quench. Charge RS072 revealed a systematic
method described by Stalder (2002). The starting mix- zoning resulting in two groups of differently coloured
tures (corresponding to orthopyroxene + quartz in the orthopyroxene crystals; one group (listed as RS072a in
weight proportions of 3:1) were prepared from oxides Table 1) exhibited a yellowish-brownish colour with a
(99.999 wt% purity), to which different amounts of significant pleochroic behaviour (absorptions were most
Fe2O3, 57Fe2O3 (57Fe/Fetot = 0.96), Cr2O3 and Al2O3 intense for E//c); from the second group only two crys-
were added. Approximately, 55 mg oxide mixtures and tals (RS072b and RS072c) could be prepared success-
14 mg H2O were sealed in a Pt-capsule with an outer fully; they appeared greenish in unpolarised light, but
(inner) diameter of 4.0 (3.6) mm. For the ferrosilite showed strong pleochroic behaviour in polarised light
synthesis fayalite was pre-synthesised at 1,000C and (E//c: reddish, E//a,b: greenish), similar to Cr-doped
one atmosphere ðfO2 ¼ IW þ 0:5Þ: A mixture of fayalite, enstatites (Stalder 2004). As expected, crystals from the
quartz and water (corresponding to a 3:1:1 mixture (by first group turned out to be more Fe-rich and poorer in
weight) of ferrosilite, quartz, and water) was sealed in a Cr (Table 1). Apart from a chemically zoned rim
Pt-capsule lined with a cylinder of pure iron with an (approximately 100 lm wide) with increasing Fe- and
outer (inner) diameter of 3.5 (3.0) mm in order to min- decreasing Cr-contents towards the edge of the crystal
imise Fe-loss and to control the oxygen fugacity during (which also is reflected in the shape of the IR-spectra),
the run. All syntheses were conducted in half inch individual crystals from run RS072 exhibit a homoge-

Table 1 Chemical composition of the synthetic orthopyroxenes determined by electron microprobe

Doped cation SiO2 wt% MgO wt% Al2O3 wt% Cr2O3 wt% FeO* wt%

RS054 Fs 46.00 (0.16) 0.09 (0.02) 53.68 (0.40)


57
RS063 En Fe 59.21 (0.26) 39.44 (0.37) 0.98 (0.41)
57
RS064 En Fe 58.71 (0.74) 37.96 (1.59) 3.18 (2.11)
57
RS084 En Fe 57.75 (0.45) 35.03 (0.88) 7.24 (1.20)
RS088 En Al, Cr 60.04 (0.22) 39.47 (0.17) 0.17 (0.06) 0.68 (0.13)
RS067 En Al, Cr 59.28 (0.22) 39.80 (0.19) 0.17 (0.08) 0.46 (0.12)
RS065 En Al, Cr 59.09 (0.24) 39.82 (0.11) 0.18 (0.05) 0.39 (0.04)
RS085 En Al, Cr 58.94 (0.35) 38.95 (0.18) 0.75 (0.29) 1.05 (0.16)
RS053 En Al, Cr 58.72 (0.27) 39.60 (0.21) 0.63 (0.28) 0.47 (0.11)
RS091 En Al, Cr 58.89 (0.50) 38.98 (0.39) 1.50 (0.80) 0.90 (0.21)
RS087 En Al, Cr 59.19 (0.39) 39.32 (0.34) 1.30 (0.69) 0.36 (0.09)
RS072a En Fe, Cr 58.82 (0.20) 36.98 (0.22) 0.56 (0.04) 3.76 (0.08)
RS072b En Fe, Cr 59.25 (0.07) 37.72 (0.17) 0.68 (0.14) 2.56 (0.05)
RS072c En Fe, Cr 59.32 (0.25) 37.31 (0.20) 1.21 (0.12) 2.51 (0.10)

Standard deviations are given in brackets


*Total Fe
Fs Ferrosilite, En Enstatite
neous composition; as all FTIR-spectra were recorded at measured both at room and liquid-nitrogen tempera-
the centre of each individual crystal, chemical compo- tures using a conventional spectrometer system operated
sitions given in Table 1 represent the core compositions in constant acceleration mode, with a 57 Co source in Rh
of the respective crystals. matrix. Absorbers were prepared by pressing 1–5 mg
From each experimental charge several clear crystals finely ground sample material mixed with an acrylic
were handpicked and orientated in a thermoplastic resin resin (transoptic powder) to self-supporting 12 mm
according to the crystallographic axes with help of their discs. To minimise texture effects, the absorbers were
optical properties (birefringence and angle between both held at an angle of 54.7 to the incident beam. Spectra
optical axes), and ground and polished on both sides. collected over 1,024 channels were velocity-calibrated
Wafers oriented parallel (100) and wafers parallel (010) against a-Fe, folded, and further reduced to 256 chan-
were produced in this way. After preparation specimens nels before fitting using a least-squares fitting program
were between 30 and 360 lm thick. Polarised IR-spectra (Jernberg and Sundqvist 1983) assuming Lorentzian
of each sample were recorded parallel to the main peak shapes. All spectra were fitted with two doublets
refractive indices a, b and c in transmitted light using a assigned to Fe2+ in the M1 and M2 positions; spectra
Bruker IFS 88 FTIR-spectrometer coupled to a Bruker of the Fe3+-containing samples were fitted with two
microscope IR scope II at the University of Hannover. additional doublets assigned to Fe3+ in octahedral and
Measurements were taken using a MIR globar light tetrahedral coordination. No constraints of spectral
source, a KBr beamsplitter and a wire-grid polariser. parameters were applied during the fitting procedure.
For each spectrum 64 scans in the range between 400 To estimate the maximum ferric iron content, in
and 6,000 cm1 were acquired with a spectral resolution addition to Mößbauer spectroscopy a thin wafer of
of 2 cm1. Background corrections were usually carried one crystal from run RS054 (ferrosilite) and RS063
out by adjusting a linear baseline at 2,500–2,600 cm1 (57Fe-doped enstatite) were investigated by optical
and 3,700–4,000 cm1. Total absorption was obtained absorption spectroscopy at the Swedish Museum of
by integration over the wavenumber range from Natural History in Stockholm. Polarised spectra were
3,000 cm1 to 3,640 cm1. In order to get more infor- measured in the 5,000–27,000 cm1 wavenumber range
mation about the nature of the defects, additional IR- with a Zeiss MPM800 single-beam spectrometer, using
spectra were recorded from polished crystals of ferrosi- Zeiss Ultrafluar10x lenses as objective and condenser.
lite and an iron-doped enstatite after thermal treatment For the UV/VIS spectral range a 75-W Xenon arc lamp
in air and/or hydrogen atmosphere at 1 bar. was used as source and a photo multiplier as detector,
As the determination of water content by FTIR- whereas a 100-W halogen lamp and a photoconductive
spectroscopy is an indirect method and recent studies PbS cell detector was used for the NIR range (5,000–
(e.g. Bell et al. 2003) have shown that independent 12,500 cm1). Measurements were performed with the
methods for water analysis in nominally anhydrous beam collimated with a circular aperture (100 lm in
phases deviate significantly from FTIR-results, some diameter) on the sample, and spectral resolutions of
selected crystals previously analysed by FTIR (Stalder 2.5 nm and 10 nm for the UV/VIS and NIR ranges,
2004) were analysed by secondary ion mass spectrome- respectively.
try (SIMS) at the University of Heidelberg. Hydrogen Out of each charge several crystals were selected for
analyses were performed on a Cameca ims 3f at low- microprobe analysis. In case of charge RS072, the same
mass resolution (m/D m=400), at an offset of 75 eV and crystals as prepared for FTIR-analyses were used.
with the energy window set to 40 eV. The primary beam Crystals were embedded in epoxy and were analysed
was formed by 16O-ions at a net energy of 14.5 keV and with a JEOL JXA 8900 RL electron microprobe using
with beam currents ranging from 5 nA to 50 nA. In situ an accelerating voltage of 15 kV and a sample current of
contamination with water was minimised with a liquid 15 nA. For Fe, Cr and Al the analysis time lasted 30 s
nitrogen cooling finger in the sample chamber. The H- on the peak and 15 s on the background, for Si and Mg
background was further minimised using the method the analysis time lasted 15 s on the peak and 5 s on the
described in Ludwig et al. (in preparation) and is esti- background. The detection limits for Fe, Cr and Al were
mated to be <10 ppm (wt) H. Quantification was done below 150 ppm, no elements other than Mg, Si, Fe, Cr,
using relative ion yields (reference isotope 30Si). Three Al and O were detected. Usually, 10 points were mea-
different tourmalines were used as standard (Dyar et al. sured on each crystal. Analytical results are given in
2001). The range of relative ion yields for these tour- Table 1.
malines (0.08–0.12) suggests an accuracy of approxi-
mately ±20% for the SIMS analyses of H.
The valence state and structural distribution of Fe in Results
samples RS063, RS064, RS084 (doped with 57Fe2O3)
and RS054 (ferrosilite) was determined by Mößbauer All IR-spectra show several absorption bands in the
spectroscopy at 298 K and 77 K; due to the high spectral region between 3,000 and 3,600 cm1 (Fig. 1),
enrichment in 57Fe or the high concentration in iron which have previously been assigned to OH-stretching
1 mg of each sample turned out to be sufficient material modes; H2O-contents were calculated by adding the
for a high-quality spectrum. Mößbauer spectra were averaged a-, b- and c-absorbances measured by FTIR
Fig. 1 a FTIR-spectra (E//c) of ferrosilite and Fe-bearing enstatite; and Al(Al+Cr) = 1 are taken from Stalder (2004) from synthetic
the spectrum for pure enstatite is from Stalder (2004), and enstatites containing 0.5 wt% Cr2O3 or 1.2 wt% Al2O3, respec-
En(+Fe)hyd is taken from a Fe-bearing enstatite (2 wt% FeO, tively; c FTIR-spectra (E//c) of Fe+Cr-doped enstatites; the
total iron expressed as FeO) from Stalder (2004) which was spectrum labelled Fe3+/Cr = 0 and Cr/Fe3+ = 0 are taken from
thermally treated in H2-atmosphere; b FTIR-spectra (E//c) of Stalder (2004) from synthetic enstatites containing 0.7 wt% Cr2O3
Cr+Al-doped enstatites; the spectrum labelled Al(Al+Cr) = 0 or 2.0 wt% FeO (total iron as FeO), respectively

(Table 2) on 2–5 individual crystals from one charge enstatite crystals from RS063, RS064 and RS084 exhibit
using the calibration of Libowitzky and Rossman two new bands at 3,328 cm1 and 3,461 cm1, their
(1997). Generally, the c-component contributed more intensity being dependent on the Fe3+-content of the
than 60% to the absolute absorbance. H2O-contents sample. After thermal treatment in hydrogen atmo-
measured by SIMS are given in Table 3. sphere, the latter two peaks disappeared ((En(+Fe)hyd),
In addition to the observed OH-absorption bands Fig. 1a); the remaining spectrum shows nearly identical
at 3,070 and 3,361 cm1 in pure enstatite, Fe-bearing OH-band positions as pure enstatite, suggesting a
Table 2 Hydrogen incorporation in doped orthopyroxenes

FTIR Haexcess Al ppm Cr ppm Fe ppm Fe3+/Fe M3+ Al/(Cr+Al) Hexcess/M3+


H2O ppm ppm (wt) (wt) (wt) (wt) (at)c (at)c corr.b (at)c
(wt)

RS054 336 8 417,500 <0.002 <15 >0.5


RS063 300 4 7650 <0.01 <1.4 >2.8
RS064 454 21 24,830 0.09 40 0.74
RS084 580 35 56,530 0.10 101 0.49
RS088 787 58 900 4,653 123 0.27 0.86
RS067 533 30 900 3,148 94 0.36 0.63
RS065 482 24 953 2,668 87 0.41 0.59
RS085 955 77 3,971 7,184 285 0.52 0.66
RS053 622 40 3,335 3,216 185 0.67 0.68
RS091 825 62 7,941 6,158 413 0.71 0.52
RS087 566 34 6,882 2,463 302 0.84 0.52
RS072a 1,323* 103 3,830 29,240 n.d.d 99 1.04
RS072b 1305* 101 4,650 19,910 n.d.d 107 0.95
RS072c 1,746* 150 8280 19,520 n.d.d 176 0.85
a
Compared to pure enstatite (Stalder 2004), containing 264 wt ppm H2O
b
H contents determined by FTIR are corrected by a factor of 1.4; only 1/6 of Al participates in H incorporation (see text)
c
=ppm (wt)/molar mass
d
Fe3+ contributing to H incorporation was estimated to be similar to RS084
*
H-content determined by SIMS

complete reduction of the ferric iron to the ferrous state. (Fig. 3), and based on the absorption intensity and
In comparison, the IR spectrum of pure ferrosilite shows Fe3+-content of the (Fe3+, Al)-doped sample (Steffen
one major absorption band, which is broader than for et al. 1988) we infer that Fe3+/Fetot < 0.002 for sample
pure enstatite and shifted to slightly lower wavenum- RS054. Optical spectroscopy was not useful for Fe3+-
bers. Thermal treatment of ferrosilite in both air and detection in case of the Fe-doped enstatites, as the
hydrogen resulted in nearly identical spectra, exhibiting intensity of the charge transfer band is dependent on the
a narrower main absorption band (Fig. 1a) and another product of the Fe2+- and the Fe3+-concentration
very small absorption band close to 3,450 cm1. (Steffen et al. 1988).
Interestingly, infrared spectra from enstatites con- H-concentrations determined on crystals from run
taining both Cr and Al only exhibit OH-bands diag- RS072 are listed in Table 2. Results from all SIMS-
nostic for Cr-bearing enstatites (i.e. at 3,449 cm1, measurements suggest 40–60% higher water concentra-
Stalder 2004), if the Al/(Al+Cr) atomic ratio was tions than FTIR (Fig. 4). This finding is in accord with
smaller than about 0.6; on the other hand OH-bands the results of experimental water incorporation in Cr-
typical for Al-bearing enstatite (i.e. between 3,500 and doped enstatites (Stalder 2004), where the water con-
3,600 cm1) are visible in all studied enstatites contain- centration determined by FTIR using the calibration of
ing Al (Fig. 1b). For crystals containing both Cr Libowitzky and Rossman (1997) or Bell et al. (1995) was
and Fe3+ the positions of OH-bands inherent from 40% lower than expected for the most probable substi-
Cr-defects or Fe3+-defects are not easy to distinguish, as tution mechanism Cr3+ + H+ = 2 Mg2+. Figure 5a
the presence of Fe3+ and Cr in enstatite leads to very shows a hydrogen concentration profile along the crys-
similar spectra. However, the continuous shift of
the 3,330 cm1 and the 3,550 cm1-bands with varying Table 3 Hydrogen contents: SIMS versus FTIR
Cr/Fe-ratio (Fig. 1c) indicate that both types of defects
are present. H2O H (wt ppm) H (wt ppm)
Mößbauer spectra for pure ferrosilite and 57Fe-doped (wt ppm) FTIR SIMSa
FTIR
orthopyroxenes are shown in Fig. 2 and site occupancies
are listed in Table 4. No ferric iron could be observed in *En 264 29 44
ferrosilite and in RS063, suggesting Fe3+/Fetot < 0.01. *Opx Al 5 309 34 65
However, due to the high Fe-content of ferrosilite, even *Opx Al 9 543 60 87
still smaller Fe3+/Fetot-ratios could lead to significant *Opx Al 17 1443 160 308
*Opx Cr 0.1 377 42 76
Fe3+-defects in ferrosilite; therefore one ferrosilite wafer *Opx Cr 0.4 708 79 133
was analysed by optical spectroscopy (Fig. 3). In optical *Opx Fe 2 288 32 44
absorption spectra of a (Fe3+, Al)-doped ferrosilite, *Opx Fe 9 581 65 103
Steffen et al. (1988) observed a band at 12,500 cm1 *Opx Fe 18 1206 134 173
RS072a 779 87 147
polarised in the c-direction that they assigned to Fe2+–
Fe3+ intervalence charge transfer. Such a band is not *Samples described in Stalder (2004)
a
present in the c-spectrum of our ferrosilite sample RS054 Analytical accuracy is estimated to be 20%
Fig. 2 Mößbauer spectra of ferrosilite a and 57Fe-doped orthopyroxenes b–d recorded at 77 K. Assignment of absorption doublets:
dashed line: Fe2+ (M1), dashed-dotted line: Fe2+ (M2), dotted line: VI Fe3+, solid line: IV Fe3+. Summed spectrum indicated by thicker
solid line (quantitative values see Table 4)

tallographic c-axis of RS072c. If a similar zonation in all an ongoing discussion (Grant et al. 2005), suggesting
directions of the crystal is assumed, the significantly that point defect OH may be systematically lower in
lower H-content close to the rim correlates well with the orthopyroxene. However, both results from secondary
amount of Me3+-related OH-defects. The chemical ion mass spectrometry (SIMS) and previous FTIR-data
zonation of RS072c is illustrated in Fig. 5b using from a series of synthetic enstatites doped with Cr
an electron microprobe data profile parallel to the (Stalder 2004) suggest that OH-concentrations deter-
crystallographic b-axis, calculating the amount of mined by FTIR are underestimated. For the majority of
Me3+-related OH-defects assuming coupled substitution samples in this study, where OH was only determined by
of Cr3+ and H+, and Fe3+ and H+; for the latter it FTIR we will assume in the following discussion that
was taken into account that Fe3+ is partly incorporated water contents determined by FTIR are underestimated
by other substitutions; the fraction of Me3+ coupled to by a factor of 1.4. In addition, taking into account that
H+ (which is equal to the ratio (H+ incorporated due to pure enstatite also incorporates hydrogen by creating
Me3+)/Me3+) is given in Table 2 (last column). Mg-vacancies, we define Hexcess (Table 2) as the amount
of H incorporation due to coupled substitution with a
trivalent cation. H incorporation in different orthopy-
Discussion roxenes will be discussed separately.

Quantification of FTIR spectra of nominally anhydrous


minerals is not straightforward. There are uncertainties Ferrosilite
related to background subtractions, extinction coeffi-
cients and absorption from non-point defect OH. Even if Mößbauer spectroscopy and optical spectros-
Especially, concerning background subtraction there is copy cannot exclude that ferric iron contributes signifi-
Fig. 3 Optical absorption
spectrum of ferrosilite (RS054)
polarised in c-direction.
Spectrum obtained on a 47 lm
thick wafer parallel to (100).
Peaks at 8,000–11,000 cm1
are caused by spin-allowed
Fe2+ d-d transitions; peaks
near 20,000 cm1 are caused
by spin-forbidden Fe2+
transitions. Absence of a
Fe2+–Fe3+ charge transfer
band at 12,500 cm1 (cf. Steffen
et al. 1988) indicate a very low
Fe3+ content (Fe3+ /
Fetot<0.002 corresponding to
Fe2O3 £ 0.10 wt-%)

cantly to hydrogen incorporation, IR spectra re- H2 O þ Fe2 O3 þ 2SixT ¼ 2SiO2 þ 2Fe0T þ 2Hi ; ð3Þ
veal—even after thermal treatment—only one major
defect associated with hydrogen. The fact that thermal where MgxM1 and SixT
indicate Mg on a M1-site and Si on
treatment in both air and hydrogen atmosphere furnish a tetrahedral site; FeM1 and Fe0T indicate ferric iron
the same spectra suggests that almost no hydrogen is on a M1 and on a tetrahedral site, and V00M1 and Hi a
associated to Fe3+. Thus Fe2+ -vacancies M1-vacancy and a hydrogen interstitial (Kröger and
Vink, 1956). Combination of (2) and (3) furnishs a
H2 O þ FexM1 ¼ V00M1 þ FeO þ 2Hi ; ð1Þ Fe-equivalent of the Tschermaks substitution:
are the most probable defect, analogous to the point
Fe2 O3 þ MgxM1 þ SixT ¼ FeM1 þ Fe0T þ MgO þ SiO2 :
defect in pure enstatite.
ð4Þ

Fe-bearing enstatite As the M2-site in orthopyroxene is slightly larger than


the M1-site, ferric iron is preferentially incorporated on
For Fe-bearing enstatite different types of defects the M1-site and ferrous iron on the M2-site (Shi et al.
involving hydrogen and iron can be encountered, 1992). If only Eqs. 2, 3, 4 are considered, hydrogen
incorporation by one single mechanism seems improbable
H2 O þ Fe2 O3 þ 4MgxM1 ¼ 4MgO þ 2FeM1 þ 2V00M1 on the basis of the present data; if all ferric iron is assumed
þ 2Hi ; ð2Þ to participate in hydrogen incorporation (Eqs. 2 and 3),

Table 4 Mößbauer parameters


for synthetic orthopyroxene Assignment RS054 RS063 RS064 RS084 RS054 RS063 RS064 RS084
obtained at room temperature RT RT RT RT 77 K 77 K 77 K 77 K
and 77 K. Hyperfine
parameters are given in mms1 Fe2+ (M1)
Int 51.7 18.6 16.3 16.2 50.1 16.9 15.9 16.4
fwhm 0.27 0.26 0.25 0.28 0.25 0.24 0.26 0.29
cs 1.18 1.16 1.15 1.14 1.29 1.26 1.26 1.27
dq 2.48 2.47 2.47 2.48 3.09 3.02 2.99 3.00
Fe2+ (M2)
Int 48.3 81.4 74.1 73.3 49.9 83.1 75.9 74.2
fwhm 0.27 0.24 0.25 0.27 0.25 0.25 0.27 0.29
cs 1.12 1.14 1.14 1.13 1.23 1.24 1.24 1.25
dq 1.92 2.10 2.10 2.10 1.98 2.14 2.14 2.13
Fe3+ (oct)
int 6.0 6.8 5.0 4.9
fwhm 0.27 0.27 0.30 0.29
cs 0.45 0.47 0.48 0.45
dq 0.68 0.63 0.79 0.82
Fe3+ (tetr)
int intensities given in % of int 3.6 3.8 3.2 4.6
total absorption area, fwhm full fwhm 0.33 0.32 0.30 0.30
width at half maximum, cs cs 0.26 0.29 0.35 0.35
centroid shift, dq quadrupole dq 1.47 1.44 1.43 1.38
splitting
Fig. 4 Hydrogen contents (in wt ppm H) determined by FTIR crystal (i.e. only one orientation) and therefore quantitative
compared to those measured by SIMS; measured samples are a H-determination by FTIR was not possible. The error given for
suite of synthetic doped orthopyroxenes (black squares) from the SIMS analyses represents the estimated accuracy (±20%);
Stalder (2004) and one of the Fe-Cr-doped enstatite crystals from the error given for FTIR analyses represents the average
this study (RS072a, open circle). Samples RS072b and RS072c are standard deviation for measurements on several crystals of the
not shown in this diagram as they consisted both of one single same charge

hydrogen contents would be higher than observed (Fig. 6, associated with a hydrogen defect (Fig. 1b), it is possi-
black triangles); if all tetrahedral Fe3+ is assumed to ble to deduce the nature of the defects in multiple doped
participate in Fe-Tschermaks-substitution (Eq. 3), samples. In the case of enstatite containing both Al3+
hydrogen contents would be lower than observed (Fig. 6, and Cr3+, Al-related hydrogen defects can be identified
grey circles). Octahedrally coordinated Fe3+ show the in all studied samples containing Al. In contrast, only
best correlation to the observed hydrogen content (Fig. 6, samples with Al/(Al+Cr) smaller than about 0.6 exhibit
open circles), but in order to achieve charge balance for Cr-related OH-peaks. It is therefore concluded that in
the tetrahedrally coordinated Fe3+, the most realistic samples with Al/(Al+Cr) > 0.6, the incorporated Cr3+
explanation is a combination of (2), (3) and (4). is not directly associated with hydrogen. Therefore the
only possible mechanism is substitution (9). If
Al/(Al+Cr) > 0.6, at least three types of associated
Multiple doped enstatite
defects are present, namely {AlM1 + Al0T }, {CrM1 +
Al0T } and {AlM1 + V00M1 + Hi }, and maybe also {Al0T
Water concentrations in multiple doped enstatites are
+ Hi }. If Al/(Al+Cr) < 0.6, {CrM1 + V00M1 + Hi },
lower than expected from the concentration of trivalent
{CrM1 + Al0T } and {AlM1 + V00M1 + Hi } (and maybe
cations (Table 2, last column). In enstatites doped with
also {Al0T + Hi }) are present. If Al/(Al+Cr)  0.6,
both Al and Cr several reactions creating defects may be
all Cr is related to {CrM1 + Al0T } defects, and some Al is
formulated:
linked to hydrogen. In order to model the defect chem-
H2 O þ Al2 O3 þ 4MgxM1 ¼ 4MgO þ 2AlM1 þ 2V00M1 istry defect–defect interactions can be formulated as
þ 2Hi ; ð5Þ
fAlM1 þ V00M1 þ Hi g þ fAl0T þ Hi g
H2 O þ Al2 O3 þ 2SixT ¼ 2SiO2 þ 2Al0T þ 2Hi ; ð6Þ ¼ fAl0T þ AlM1 g þ H2 ð10Þ
H2 O þ Cr2 O3 þ 4MgxM1 ¼ 4MgO þ 2CrM1 þ 2V00M1 fCrM1 þ V00M1 þ Hi g þ fAl0T þ Hi g
þ 2Hi ; ð7Þ ¼ fAl0T þ CrM1 g þ H2 ð11Þ
Al2 O3 þ MgxM1 þ SixT ¼ AlM1 þ Al0T þ MgO þ SiO2 ; fCrM1 þ V00M1 þ Hi g þ fAl0T þ AlM1 g
ð8Þ ¼ fAlM1 þ V00M1 þ Hi g þ fAl0T þ CrM1 g ð12Þ

Al2 O3 þ Cr2 O3 þ 2MgxM1 þ 2SixT with the equilibrium constants


¼ 2CrM1 þ 2Al0T þ 2MgO þ 2SiO2 : ð9Þ
 
K10 fAl0T þ AlM1 g
As IR spectra from single doped enstatites show ¼    ð13Þ
specific OH-absorption bands for each trivalent cation fH2 fAl0T þ Hi g  fAlM1 þ V00M1 þ Hi g
 
K11 fAl0T þ CrM1 g here, the equilibrium lays on the right side of Eq. (12);
¼    ð14Þ previous results from a series of Al-doped enstatites
fH2 fAl0T þ Hi g  fCrM1 þ V00M1 þ Hi g (Stalder, 2004) suggests that K10/fH2>7.5 (as one out of
 0 
   00
 8.4 Al is coupled to an H) for enstatites with very high
AlT þ CrM1  AlM1 þ VM1 þ Hi Al/(Al+Cr), but a much lower value is expected for Cr-
K12 ¼  0     ð15Þ
AlT þ AlM1  CrM1 00 þ H 
þ VM1 i rich crystals. For crystals with a Al/(Al+Cr) < 0.6 we
can derive the respective defect concentrations according
As all syntheses were carried out under quasi identi- to Eq. (14)
cal conditions and no solid buffer was present, hydrogen  3þ      
fugacity probably is given by the talc-pyrex cells (Dunn Cr ¼ CrM1 þ V00M1 þ Hi þ CrM1 þ Al0T
1983) and can be regarded as constant over the whole ð16Þ
series. The defect concentrations can be calculated from
the water contents; for our calculations we will use the     
Al 3þ
¼ Al0T þ Hi þ CrM1 þ Al0T ð17Þ
water contents determined by FTIR corrected by factor
of 1.4 (referring to the SIMS analyses and the Cr-H-     
CrM1 þ Al0T ¼ K CrM1 þ V

00
M1 þ Hi

correlation in Cr-doped enstatites). According to Stalder
 AlT þ Hi ð18Þ
(2004), then only one out of 8.4 Al (6*1.4) and one out
of 1.4 Cr is coupled to an H in single doped enstatites. Hexcess as a function of K, [Al3+] and [Cr3+] can be
We then use the above definition for Hexcess as the calculated taking into account that one out of 8.4 Al3+
amount of H incorporation due to coupled substitution
with a trivalent cation, and define K=K11/fH2. If we can
quantify the above equations, a model for hydrogen
incorporation in Al+Cr-doped enstatite can be derived.
Equilibrium constant for reaction (10) and (12) can be
estimated; from the spectroscopic data it can be con-
cluded that K12 1 for Al/(Al+Cr) > 0.6; K12 may be
a function of Al/(Al+Cr), but for all samples studied

Fig. 6 Amount of ferric iron that participates in hydrogen


incorporation versus hydrogen content (determined by FTIR) of
57
Fe-doped orthopyroxenes; ppm (at) refers to ppm (wt)/molar
mass. Ferric iron participating in hydrogen incorporation was
assumed to be either merely octahedral Fe3+ (open circles),
tetrahedral and octahedral Fe3+ (black triangles), or octahedral
minus tetrahedral Fe3+ (grey circles); the last case accounts for Fe-
Fig. 5 a H-concentration profile (//c) on RS072c determined by Tschermaks-substitution. The solid line shows the expected
SIMS; b H-incorporation capacity calculated on the basis of the correlation if ferric iron is incorporated by coupled substitution
Fe3+- and Cr-concentrations determined by electron microprobe with hydrogen; the dashed line shows how the correlation would fit
over a profile (//b) on RS072c. Note that both profiles are if 40% higher hydrogen contents are assumed (owing to the results
measured in different directions and have different scale (1 mm //c, from SIMS analyses). The arrow indicates that concentration for
0.3 mm //b), but show the same principle zonation feature Fe3+ may be lower
Fig. 7 Model for defect interactions; ppm (at) refer to ppm (wt)/ ppm H) for each sample is shown inside the symbol. The difference
molar mass. Solid lines represent theoretical isolines for hydrogen between model a and b is explained in the text. Al* and Cr* denote
solubility (in wt ppm H). Black symbols represent enstatites the effective concentration of Al and Cr involved in OH-bonding.
containing both Cr- and Al-specific OH-defects, grey symbols K denotes K11/ fH2
exhibit only Al-specific OH-defects. Hydrogen concentration (in wt

and one out of 1.4 Cr3+ is coupled to an H in single participating in (11), the Al/(Al+Cr) from model (a)
doped enstatites; in model (a) (Fig. 7a) [Al3+] and was recalculated dividing [Al] by a factor of 8.4 and [Cr]
[Cr3+] are identical to the real concentrations deter- by 1.4, respectively (referred to as Al* and Cr* in
mined by microprobe divided by their respective molar Fig. 7); results are shown in Fig. 7b. Comparison of the
mass. For crystals with a Al/(Al+Cr) > 0.6 other de- experimental results to model (a) show the highest cor-
fects than in reaction (11) are introduced and two dif- respondence for K=2 (Fig. 7a); samples containing
ferent Tschermaks pairs (Al3+ + Al3+ and Al3+ + {CrM1 + V00M1 + Hi } defects (shown as black circles in
Cr3+) are present, rendering the model more compli- Fig. 7) do fit better to model (a) than those without
cated. In order to account for the additional Al not {CrM1 + V00M1 + Hi } defects (shown as grey circles). In
Fig. 7b (Contd.)

contrast, samples without {CrM1 + V00M1 + Hi } defects model between (a) and (b), depending on Al/(Al+Cr);
show a higher correspondence to model (b) (Fig. 7b), so the K for model (b) does not have a real meaning. As
with a higher value for K (between 3 and 8). It is not K=K11/ fH2 we can express the free enthalpy for Eq.
surprising that data of samples with a high (Al/ (11) for model (a)
(Al+Cr)) fit better to model (b), as in Al-rich samples
much of the Al will form {AlM1 + Al0T } defect associ- DG11 ¼ R  T  ln K11 ¼ R  T  ln K  R  T  ln fH2
ates and therefore the effective concentration of Al in- ð19Þ
volved in OH-bonding will be reduced. Whereas water
concentrations for samples containing {CrM1 + V00M1 + The only unknown in Eq. (19) is the hydrogen fugacity,
Hi } can be described rather well with model (a), Al-rich which can be estimated assuming that the oxygen
samples would probably be described best by a hybrid fugacity was externally buffered by the cell assembly and
therefore close to the C-CO-CO2-O2 buffer (Dunn 1983);
using the equation of state of Saxena and Fei (1987) and References
Belonoshko and Saxena (1992), a hydrogen fugacity of
Bai Q, Kohlstedt DL (1992) Substantial hydrogen solubility in
103 bar is obtained. Inserting R (8.3143 Jmol1 K1), T olivine and implications for water storage in the mantle. Nature
(1423 K) and K (2) in Eq. (19) we receive a DG11= 357:672–674
90 kJ/mol. As model (b) is strongly dependent on Al/ Bell DR, Rossman GR (1992) Water in the Earth’s mantle: the role
(Al+Cr) and other defects are present which are not of nominally anhydrous minerals. Science 255:1391–1397
taken into account, K does not give quantitative infor- Bell DR, Ihinger PD, Rossman GR (1995) Quantitative analysis of
trace OH in garnet and pyroxenes. Amer Mineral 80:465–474
mation concerning the energetics of equilibrium (11). It Bell DR, Rossman GR, Maldener J, Endisch D, Rauch F (2003)
also has to be emphasised that Eqs. 10, 11, 13, 14 and 15 Hydroxide in olivine: a quantitative determination of the
could alternatively be formulated with fH2 O (instead of absolute amount and calibration of the IR spectrum. J Geophys
fH2 ) controlling the equilibrium. The procedure would Res 108:2105
Belonoshko AB, Saxena SK (1992) Equations of state of fluids at
be similar as described above; which alternative is more high temperature and pressure (water, carbon dioxide, methane,
appropriate could be tested by experiments at different carbon monoxide, oxygen, and hydrogen). In: Saxena SK (ed)
fO2 and different total pressure. Thermodynamic data. Systematics and estimation, Springer, pp
The model described above is in principal also 80–97
Bose K, Ganguly J (1995) Quartz-coesite transition revisited: re-
applicable to orthopyroxenes containing a combination versed experimental determination at 500–1200C and retrieved
of other trivalent cations. However, samples from thermochemical properties. Amer Mineral 80:231–238
RS072, containing both Cr3+ and Fe3+ exhibit only Boyd FR, Mertzman SA (1987) Composition and structure of the
small deviations of Hexcess/M3+ from unity (Table 2), Kaapvaal lithosphere, southern Africa. In: Mysen BO (Ed)
and an increased tendency to form Tschermaks-like Magmatic processes: physicochemical principles. Geochem Soc
Spec Publ 1:13–24
defects cannot be proved. Dunn T (1983) Oxygen chemical diffusion in three basaltic liquids
at elevated temperatures and pressures. Geochim Cosmochim
Acta 47:1923–1930
Dyar MD, Wiedenbeck M, Robertson D, Cross LR, Delaney JS,
Implications for natural mantle derived orthopyroxenes Ferguson K, Francis CA, Grew ES, Guidotti CV, Hervig RL,
Hughes JM, Husler J, Leeman W, McGuire AV, Rhede D,
IR-spectra taken from natural orthopyroxenes from Rothe H, Paul RL, Richards I, Yates M (2001) Reference
materials for the microanalysis of light elements. Geostandards
upper mantle xenoliths (Bell et al. 1995, Rossman 1996, Newsletter 25:441–463
Peslier et al. 2002, Stalder 2004) all show OH-absorption Frey FA, Prinz M (1978) Ultramafic inclusions from San Carlos,
bands specific to a local cation arrangement containing Arizona: petrologic and geochemical data bearing on their
Al, whereas OH-bands specific for Cr are absent (Stalder petrogenesis. Earth Planet Sci Lett 38:129–176
Grant KJ, Kohn SC, Brooker RA (2005) Solubility and parti-
2004). The model presented above explains this behav- tioning of water in synthetic forsterite and enstatite in the sys-
iour satisfactorily, as all {CrM1 + V00M1 + Hi } defect tem MgO-SiO2-H2O±Al2O3. Contrib Mineral Petrol (in press)
associates are destroyed by reaction (11) and (12) if the Hays JF (1966) Lime-Alumina-Silica. Carnegie Inst Wash Yearb
molar ratio of Al/(Al+Cr) is larger than 0.6 (which is 65:234–239
Ingrin J, Skogby H (2000) Hydrogen in nominally anhydrous up-
generally the case for upper mantle material). per-mantle minerals: concentration levels and implications. Eur
The tendency of OH-defect destruction in orthopy- J Mineral 12:543–570
roxenes containing different trivalent cations has Jernberg P, Sundqvist T (1983) A versatile Mößbauer analysis
important consequences for hydrogen storage in ortho- program. Uppsala University, Institute of Physics (UUIP-1090)
pyroxene. In contrast to previous suggestions that Klemme S, O’Neill HStC (2000) The effect of Cr on the solubility
of Al in orthopyroxene: experiments and thermodynamic
hydrogen solubility is directly proportional to the con- modelling. Contrib Mineral Petrol 140:84–98
centration of trivalent cations (Rauch and Keppler 2002, Kohlstedt DL, Keppler H, Rubie DC (1996) Solubility of water in
Stalder 2004), results from this study suggest much the a, b and c phases of (Mg,Fe)2SiO4. Contrib Mineral Petrol
lower hydrogen solubilities. Natural orthopyroxenes 123:345–357
Kröger FA, Vink HJ (1956) Relations between the concentration of
from the Earth’s mantle show a broad range of trivalent imperfection in crystalline solids. In: Seitz F, Turnball D (eds)
cation concentrations and are derived from different Solid state physics 3. Academic Press, New York, pp 367–435
depths and temperatures (e.g. Frey and Prinz 1978, Libowitzky E, Rossman GR (1997) An IR absorption calibration
Boyd and Mertzman 1987, Zipfel and Wörner 1992) for water in minerals. Amer Mineral 82:1111–1115
Mackwell SJ, Kohlstedt DL, Paterson MS (1985) The role of water
affecting the equilibrium constants of the defect reac- in the deformation of olivine single crystals. J Geoph Res
tions, a general statement concerning the hydrogen sol- 90:11319–11333
ubility is not possible. Further studies investigating the Mei S, Kohlstedt DL (2000) Influence of water on plastic defor-
influence of pressure and hydrogen fugacity are neces- mation of olivine aggregates 1. Diffusion creep regime. J Geo-
sary. phys Res 105:21457–21469
Peslier AH, Luhr JF, Post J (2002) Low water contents in pyrox-
enes from spinel- peridotites of the oxidized, sub-arc mantle
Acknowledgements Andreas Kronz (Universität Göttingen) is wedge. Earth Planet Sci Lett 201:69–86
thanked for the microprobe analyses, Harald Behrens (Universität Rauch M, Keppler H (2002) Water solubility in orthopyroxene.
Hannover) is thanked for valuable discussions and for kindly Contrib Mineral Petrol 143:525–536
providing access to the IR-spectroscope. Constructive reviews by Rossman GR (1996) Studies of OH in nominally anhydrous min-
Sergei Matveev and Simon Kohn are kindly acknowledged. erals. Phys Chem Minerals 23:299–304
Saxena SK, Fei Y. (1987) High pressure and high temperature fluid Stalder R, Skogby H (2002) Hydrogen incorporation in enstatite.
fugacities. Geochim Cosmochim Acta 51:783–791 Eur J Mineral 14:1139–1144
Shi P, Saxena SK, Sundman B (1992) Sublattice solid solution Steffen G, Langer K, Seifert F (1988) Polarized electronic
model and its application to orthopyroxene (Mg,Fe)2Si2O6. absorption spectra of synthetic (Mg,Fe)-orthopyroxenes, fer-
Phys Chem Minerals 18:393–405 rosilite and Fe3+-bearing ferrosilite. Phys Chem Minerals
Skogby H, Rossman GR (1989) OH in pyroxene: an experimental 16:120–129
study of incorporation mechanisms and stability. Amer Mineral Wright K, Catlow CRA (1994) A computer simulation study of
74:1059–1069 (OH) defects in olivine. Phys Chem Minerals 20:515–518
Skogby H, Bell DR, Rossman GR (1990) Hydroxide in pyroxene: Zipfel J, Wörner G (1992) Four- and five-phase peridotites from a
variations in the natural environment. Amer Mineral 75:764–774 continental rift system: evidence for upper mantle uplift and
Stalder R (2002) Synthesis of enstatite at high pressure. Eur J cooling at the Ross Sea margin (Antarctica). Contrib Mineral
Mineral 14:637–640 Petrol 111:24–36
Stalder R (2004) Influence of Fe, Cr and Al on hydrogen incor-
poration in orthopyroxene. Eur J Mineral 16:703–711

You might also like