Mineralogical Magazine, June 2005, Vol. 69(3), pp. 345±358
Hydrogen and minor element incorporation in synthetic rutile
1,2,
G. D. BROMILEY
1
2
3
*
AND
3
N. HILAIRET
Bayerisches Geoinstitut, Universita
Èt Bayreuth, 95440 Bayreuth, Germany
Department of Earth Sciences, University of Cambridge, Downing Street, Cambridge CB2 3EQ, UK
Laboratoire de Sciences de la Terre, ENS-Lyon, 46 Allee
 d'Italie, 69364 Lyon, Cedex 07, France
ABSTRACT
The solubility and incorporation mechanisms of H and various trivalent and divalent cations in
synthetic rutile have been investigated. Experiments performed using different bulk Fe 2O3 contents
demonstrate that Fe
3+
substitutes onto the main Ti site, charge-balanced by oxygen vacancies. Under
more reducing conditions in Fe-poor systems, the concentration of Ti interstitials in rutile is increased,
resulting in a decrease in H solubility. Variation in the solubility of different oxides in rutile as a
function of ionic radius implies substitution onto the main Ti site, probably charge-balanced by oxygen
vacancies. To a lesser degree, substitution of trivalent and divalent cations is locally charge-balanced
by H incorporation. Variation in OH-stretching frequencies in infrared spectra as a function of
composition implies that octahedral defects and structurally-incorporated H are coupled. However, in
all samples, some of the H is also decoupled from substitutional impurities, as is evident from an
OH-absorption band at 3279 cm
ÿ1.
This band corresponds to the main OH band seen in spectra of
many natural rutiles, implying that in most rutiles, H defects are decoupled from substitutional defects.
K EYWORDS : rutile, hydrogen, substitution, solubility, spectroscopy.
Introduction
(Johnson et al., 1968). Studies of natural rutiles
RUTILE is the most common, naturally occurring
from
form of titanium oxide (TiO2), and is an accessory
demonstrated
mineral in many igneous and metamorphic rocks.
hydrogen incorporation, and rutile is one of the
In high-pressure and ultra-high-pressure eclogites,
most
modal percentages of rutile may reach sever-
(NAMs) so far identi®ed (Rossman and Smyth,
al wt.%. Rutile has a simple tetragonal structure,
1990; Vlassopoulos et al., 1993).
with each Ti
the
4+
corners
octahedron,
4+
three Ti
ion surrounded by six oxygens at
of
a
and
slightly
each
distorted,
oxygen
regular
surrounded
by
ions lying in a plane at the corners of
various
geological
the
`hydrous'
great
terrains
af®nity
have
of
also
rutile
nominally-anhydrous
for
minerals
Rutile is also a useful solid material because of
its
diverse
physical
photochemical
optical
properties
function.
properties
of
The
reduced,
and
unique
electronic
synthetic
and
rutile
an approximately equilateral triangle (Deer et al.,
and its defect structures have been the subject of
1992). Rutile can contain variable amounts of
a considerable amount of research in the ®elds of
5+
5+
3+
3+
,
applied physics, solid-state chemistry and mate-
)
rials science (Lu et al., 2001). Rutile, therefore,
cations. Coupled substitution of pentavalent and
provides a useful model system for investigating
pentavalent (Nb
Cr
3+
, Ta
), trivalent (Fe
) and, to a lesser degree divalent (Mg
trivalent cations onto Ti
neutrality.
valence
In
addition,
cations
may
4+
, Al
2+
2+
, Ca
sites maintains charge
substitution
be
of
lower-
charge-balanced
by
incorporation of hydrogen in the rutile structure
the
incorporation
of
coupled
lower-valence
cations and hydrogen in NAMs. In addition, the
system TiO2 provides a lower-pressure analogue
for
understanding
the
SiO2
system:
rutile
is
isostructural with stishovite, and investigation of
hydroxyl incorporation in rutile could provide
* E-mail: [email protected]
useful constraints on the importance of hydroxyl
DOI: 10.1180/0026461056930256
incorporation in stishovite (Smyth et al., 1995).
#
2005 The Mineralogical Society
G. D. BROMILEY AND N. HILAIRET
Experimental method
determined
(EMPA).
Sample synthesis
mounts
by
electron
Samples
and
microprobe
were
analysed
using
a
analysis
as
grain
Cameca
SX50
Several samples of rutile doped with various
electron
trivalent and divalent cations were synthesized at
15 nA, using the following standards: MnTiO 3
2.0 GPa, 1100ëC under water-saturated conditions
(Ti), gallium arsenide (Ga), chromium metal (Cr),
to
and
wollastonite (Ca), spinel (Al), enstatite (Mg),
were
hematite (Fe). A count time of 20 s was used for
TiO 2
each element.
investigate
coupled
lower-valence
prepared
cations.
from
substitution
Starting
high-purity
of
H
mixes
oxides:
microscope
prepared
operating
at
15 kV
and
(99.998%), Fe2O3 (99.998%), Al2O3 (99.99%),
Cr 2 O 3
(99.999%),
(99.99%),
and
Ga 2 O 3
CaO
(99.99%),
(99.9%).
Two
MgO
series
of
Infrared spectroscopy
synthesis experiments were performed. In the
Near-infrared (NIR) spectroscopy was used to
®rst series of experiments, starting mixes of pure
investigate
TiO2, and TiO2 with 1.0, 2.0 and 5.0 wt.% Fe2O3
synthetic
samples.
were prepared by weighing out 3 g of the starting
obtained
using
mixture, and then grinding the mixture under
resolution
ethanol in an automated pestle and mortar for at
Bruker IR microscope containing all-re¯ecting
least 2 h. In the second series of experiments,
Cassegranian optics. The spectrometer contains a
starting mixes were prepared from TiO2 with
permanently aligned Michelson-type interferom-
5 wt.% Al2O3, Cr2O3, Ga2O3, MgO and CaO
eter with a 30ë angle of incidence on the beam-
using the same method. Starting mixes for all
splitter.
experiments were loaded into 10 mm long, 5 mm
tungsten light source, a Si-coated CaF2 beams-
diameter
Pt
capsules
a
in
Bruker
IFS
120
HR
spectrometer coupled
Measurements
were
taken
the
were
high-
with
using
a
a
water.
plitter and a narrow-band MCT detector. Up to
400 scans were acquired for each spectrum with a
sleeves and loaded into Ý inch talc-pyrex piston-
resolution
cylinder sample assemblies. These sample assem-
recorded using a wire-strip polarizer on a KRS-5
blies
substrate. The optics of the microscope were
an
internal,
10 wt.%
FTIR
incorporation
Infrared (IR) spectra
Capsules were welded shut, placed in pyrophyllite
contained
with
hydroxyl
tapered,
graphite
of
1 cm
ÿ1.
Polarized
spectra
were
resistance furnace to ensure minimal temperature
purged with H2O- and CO2-free, puri®ed air
gradients along the length of the capsule. Full
during
details of pressure and temperature calibration
spectrometer were kept under vacuum to prevent
and accuracy are given in Bromiley and Keppler
absorption bands caused by water vapour. The
(2004). Experiments were pressurized to within
FTIR absorption spectra were only obtained from
90% of the desired run pressure and then heated at
optically clear areas of crystals, using a variable
100ëC/min to 1100ëC, before being fully pressur-
rectangular aperture in the rear focal plane of the
ized. Pressure and temperature were continually
15
monitored and maintained throughout the runs.
placed on CaF2 plates and immersed in poly-
Runs were isobarically quenched by turning off
trichloro¯uoroethylene oil to prevent the occur-
power to the heating circuit whilst maintaining
rence of interference fringes. Crystal sizes for
run pressure. Recovered capsules were weighed,
samples ranged from 30 to 500
pierced and placed in a drying oven, and then
samples, polarized spectra could be obtained from
reweighed to check for the presence of water. In
optically aligned crystals. However, for other
all experiments water was seen escaping from
samples, crystals were too small to be aligned.
capsules when pierced.
Therefore, for all samples, water contents were
6
measurements,
Cassegranian
and
the
objective.
m
optics
of
Samples
the
were
m, and for some
determined using unpolarized radiation. The OH
absorption bands in rutile, as in most NAMs, are
Sample analysis
highly anisotropic, with greatest absorption when
Phases were identi®ed by micro-Raman spectro-
the electric vector of the source radiation is
scopy using published spectra for rutile (Goresy et
polarized perpendicular to the crystallographic c
al., 2001) and by powder XRD. Single crystals of
axis (Vlassopoulos et al., 1993). Use of unpolar-
all samples were placed on glass slides and
ized radiation could, therefore, lead to inaccura-
spectra obtained using a Dilor labram Raman
cies in calculated water contents. To avoid this
spectrometer equipped with a 20 mW HeNe laser
possibility, spectra were obtained from a large
and a microscope. Compositions of samples were
number of crystals (typically ~20) in each sample,
346
HYDROGEN AND MINOR ELEMENT INCORPORATION IN RUTILE
TABLE 1. Experimental details, results and composition of run products analysed by EMPA.
Sample
Bulk
Run products
Colour
composition
Wt.%
Wt.%
TiO2
additional
Total
oxide
R1
TiO2
Rutile
Dark blue
RFe1
TiO2 + 1.0
Rutile +
Dark yellow-
wt.% Fe2O3
ilmenite
brown
RFe2
TiO2 + 2.0
Rutile +
Yellow-
wt.% Fe2O3
ilmenite
brown
TiO2 + 5
Rutile +
Bright
wt.% Fe2O3
ilmenite
brown
TiO2 + 5
Rutile +
wt.% Al2O3
corundum
TiO2 + 5
RFe3
RAl
RMg
ÿ
ÿ
.
ÿ
.
.
99.88(10)
0.21(2)
100.09(21)
99.72(11)
0.33(3)
100.06(29)
97.22 (56)
2.69(9)
99.91(79)
Pale blue
97.93(68)
1.96(15)
99.92(79)
Rutile + geikielite
Colourless
98.68(32)
0.82(2)
99.50(31)
Rutile
Red-brown
94.42(111)
4.92(82)
99.38(62)
TiO2 + 10
Carmichaelite
Dark red-
88.90(65)
10.88(42)
99.78(44)
wt.% Cr2O3
(no rutile!)
brown
TiO2 +
Rutile +
Colourless
96.18(28)
3.92(12) 100.11(28)
5 wt.% Ga2O3
unknown phase*
Colourless
99.61(11)
0.08(3)
wt.% MgO
RCr
TiO2 + 5
wt.% Cr2O3
RCr2
RGa
RCa
TiO2 + 5
Rutile +
wt.% CaO
CaTiO3 perovskite
99.69(13)
* XRD pattern of unknown (Ga-Ti oxide?) phase has peaks at d spacings of 3.604, 1.87 and 1.81.
and
water
contents
calculated
from
Table 2. Samples R1 to RFe3 were synthesized
averaged
spectra. Water contents were calculated from
from
integrated IR spectra using the Beer-Lambert law
ranging from 0 to 5 wt.%. Several interesting
starting
mixtures
with
Fe 2O 3
on species absorption, and known IR integral
features are noted from these experiments. Firstly,
absorption coef®cients for rutile. Three different
in all samples containing iron, additional amounts
contents
integral absorption coef®cients for rutile have
been
determined:
(Johnson
et
(Hammer
38,000
al.,
15,100Ô500 Lmol
1968),
and
ÿ1 ÿ2
Lmol cm
3270 Lmol
ÿ1 cm ÿ2
ÿ1 cm ÿ2
Beran,
1991)
(Maldener
et
al.,
TABLE 2. H contents (expressed as water contents) for
synthetic rutile determined from NIR analysis.
and
2001).
The large disparities between these values are
Sample
largely the result of the different techniques used
to independently calibrate absorption coef®cients
and
the
nature
of
the
samples
used
for
the
calibration. In this study, we use the calibration
Integrated
absorbance (cm
ÿ2)
Water content
(ppm wt. H2O)
R1
642
90Ô3
RFe1
951
134Ô4
302Ô10
RFe2
2152
of Johnson et al. (1973) because it was derived
RFe3
6418
900Ô31
from a study of synthetic rutile and because it
RAl
12671
1777Ô57
gives calculated water contents for rutile samples
RMg
closer to values derived using the general sample-
RCr
independent calibrations of Paterson (1982) and
RCr2
Libowitzky and Rossman (1997).
Fe3+ incorporation in synthetic rutile
5470
767Ô26
14078
RGa
ÿ
1975Ô67
4946
694Ô23
RCa
0
0
1
ÿ
Calculated using the integral absorption coef®cients
The compositions of all run products determined
for rutile of Johnson et al. (1973). Error on calculated
by EMPA are listed in Table 1. Water contents
value derived from error for integral absorption
calculated
coef®cient.
from
FTIR
spectra
are
listed
in
347
1
G. D. BROMILEY AND N. HILAIRET
of
an
Fe-rich
phase,
ilmenite
(FeTiO 3 )
are
present. Iron was added to starting mixes as
hematite
(i.e.
Fe
3+
),
thereby
implying
partial
3+
Fe
substitutes for Ti
octahedral Ti
sample
4+
RFe3
4+
; i.e. Fe
3+
is present on the
site. The light colour of rutile in
is,
therefore,
probably
due
to
reduction of Fe during the experiments. The Fe
electronic transitions in octahedrally co-ordinated
content
of
increasing
Sample
the
Fe
rutile
samples
content
RFe3,
of
synthesized
3+
increases
with
Fe
starting
mix.
from sample RFe3 (Fig. 1) contain no broad
the
from
a
starting
(Burns, 1993). The NIR spectra obtained
absorption bands, implying an absence of Ti
4+
3+
-
mixture with 5 wt.% Fe2O3, has the highest Fe
Ti
content. The Fe content of this sample is, within
RFe2
error, the same as the sample synthesized under
absorption due to Ti
the same P-T-H2O conditions in the study by
rally coordinated Fe
Bromiley et al. (2004a). The only difference in
these
synthesis between these samples is that Bromiley
6000 cm
et al. (2004a) used Pt capsules lined with Re foil
for sample R1. In fact, there appears to be a
to prevent Fe-loss to the capsules during the
marked
experiment. No Re lining was used in the present
absorption due to IVCT: increasing Fe content
study to minimize differences in fO2 conditions
in the synthetic rutile reduces the amount of Ti
between different experiments. However, compar-
Ti
ison of the present results with those of Bromiley
in industry, where small amounts of trivalent
et al. (2004a) suggests that the use of Re foil has
cation substitution (mainly Al
an undetectable effect on Fe contents of rutile
rutile
samples. Iron contents of the samples RFe1 and
giving rutile its desired brilliant white colour. It
RFe2 (synthesized from starting mixes with 1.0
would
and 2.0 wt.% Fe2O3, respectively) are consider-
substitution has a similar effect. This effect can
ably lower than sample RFe3 and also much
be most easily explained by considering mechan-
lower than the Fe content of the starting mixes for
isms for trivalent cation substitution in rutile and
these experiments. This is an interesting observa-
the synthesis conditions of rutile samples in the
tion because it implies that these samples are
present study. Previous investigations of rutile
under-saturated with respect to Fe. Long run times
have demonstrated a number of mechanisms for
were
charge-balancing
used
during
all
experiments
to
attain
4+
IVCT. Rutile crystals in samples RFe1 and
are
yellow-brown
samples
ÿ1
3+
3+
show
in
-Ti
4+
colour,
implying
IVCT and octahed-
. The NIR spectra from
some
absorption
around
(Fig. 1), although markedly less than
correlation
between
Fe
content
and
3+
-
IVCT. A similar phenomenon is well known
is
used
to
appear
suppress
that
at
3+
) in nano-phase
optical
high
trivalent
absorption,
pressure,
cation
Fe
3+
substitutions
chemical equilibrium. During EMPA, no signs
(Gesenhues and Rentschler, 1999). In natural
of chemical zoning were noted in any rutile
rutiles, substitution of trivalent cations can be
samples, and so we assume that the Fe contents of
compensated
all samples represent equilibrium values.
cations for Ti
Another
interesting
feature
of
Fe-bearing
by
4+
substitution
(e.g. Fe
3+
of
>
pentavalent
5+
+ Nb
4+
2Ti
).
However, in the present study, such mechanisms
3+
samples is the colour of the rutile crystals. In
cannot operate. Alternatively, Fe
sample R1 (Fe-free), rutile crystals are dark blue.
could be charge-balanced by H
This colour is characteristic of rutile samples
The NIR spectra for samples shown in Fig. 1
grown or annealed under reducing conditions. In a
contain sharp bands over the OH-stretching range
detailed spectroscopic study, Khomenko et al.
(4000
(1998) demonstrated that the colour of reduced,
some structurally-incorporated H. However, as
TiO2ÿx rutile is mainly due to intervalence charge
can be seen in Tables 1 and 2, H contents of
transfer
(IVCT)
on
Fe-doped rutile are several orders of magnitude
adjacent
interstitial
The
lower than Fe contents, implying that another
IVCT in rutile is evident from broad absorption
mechanism is needed to compensate most of the
in
This
Fe substitution in the samples. Alternatively, Fe
absorption is seen in NIR spectra obtained from
substitution could either be charge-balanced by
rutile as a characteristic broad absorption band
the coupled substitution of trivalent cations (either
the
optical
between
to
and
NIR
centred around 6000 cm
Ti
3+
and
octahedral
spectral
ÿ1.
Ti
4+
sites.
range.
Figure 1 shows NIR
Fe
3+
ÿ2500 cmÿ ),
or Ti
1
3+
implying
+
the
substitution
incorporation.
presence
of
) onto interstitial sites or by oxygen
spectra of samples R1 to RFe3. In contrast to
vacancies. A wealth of information is available on
sample R1, sample RFe3 is light yellow in colour.
defects present in reduced rutile (Lu et al., 2001),
Mo
È ssbauer
rutile
and there is abundant evidence indicating that
(Bromiley et al., 2004a) showed that all Fe in
both oxygen vacancies and Ti intersitials are
spectra
synthetic rutile is Fe
from
3+
Fe -doped
, and demonstrated that
348
present as stable defects. The experimental setup
HYDROGEN AND MINOR ELEMENT INCORPORATION IN RUTILE
FIG. 1. NIR spectra of rutile doped with increasing amounts of Fe 2O3 synthesized under water-saturated conditions at
2.0 GPa, 1100ëC. A horizontal baseline over the spectral range shown has been subtracted. Spectra offset vertically
with increasing Fe content for clarity. Large and small divisions on the vertical scale represent 10 and 5 cm
ÿ 1,
respectively.
used in this study produces slightly reducing
Bromiley et al. (2004a) suggested that this was
conditions for sample synthesis (presumably due
the most likely mechanism for charge-balancing
to the large graphite furnace used in the run
Fe incorporation in rutile and its high-pressure
assembly). Oxygen fugacity for the experimental
form TiO2 (II) because Mo
È ssbauer spectra from
setup is estimated to be slightly lower than NNO
both phases consisted of a single quadrupole
(Bromiley et al., 2004b), as long as reaction of the
doublet (due to octahedrally-coordinated Fe
3+
3+
),
starting materials has only a limited in¯uence. Fe
with no evidence noted for Fe
substitution in rutile charge-balanced by oxygen
more than one site. This observation is important,
vacancies
because although the main interstitial site in rutile
would,
therefore,
appear
possible.
349
occupancy of
G. D. BROMILEY AND N. HILAIRET
is octahedral, it is considerably more distorted
than the regular Ti
4+
contents must be more reducing, because the
_ defects in Fe-poor rutile implies
presence of Tii
site, and occupancy of both
octahedral and interstitial sites by Fe
3+
should be
TiO2 reduction. According to reaction 1, more
detectable by Mo
È ssbauer spectroscopy. However,
reducing conditions would result in lower Fe2O3
this would not completely exclude the possibility
contents in rutile, which is exactly what is seen in
that a small amount of Fe
3+
in the synthetic rutile
our experiments. This situation is probably further
could also be present on the main interstitial site.
complicated by the additional, minor effect of H
Interstitial Ti
provide
3+
an
in samples RFe1 and RFe2 could
incorporation in rutile, which implies breakdown
additional
of H2O. However, the overall effect of bulk Fe2O3
charge-balancing
mechanism for Fe substitution. However, this is
content on the Fe2O3 content of rutile and the
a somewhat unsatisfactory conclusion to draw
concentration of Tii
because it does not fully explain why there is no
suggests
interstitial Ti
3+
_
that
the
defects in rutile strongly
controlling
factor
in
all
in sample R3, and is inconsistent
experiments is the fO2, which can be buffered at
with the observed decrease in IVCT absorption
higher values in Fe-rich systems relative to Fe-
with increasing Fe content. To explain some of
poor systems, because Fe2O3 in the starting mix
the unusual features noted in these experiments,
acts as a source for O2. In turn, this implies that
we need ®rst to summarize what we know about
Fe
stable
rutile
charge-balanced by VOÇÇ rather than by interstitial
samples, and what reactions must occur during
cations. Unfortunately, Fe2O3 contents of samples
the synthesis of Fe-bearing rutile.
RFe1 and RFe2 were too low for investigation of
defects
present
in
the
reduced
Firstly, it has been demonstrated that reduced
3+
substitution for Ti
4+
in rutile is mainly
_
Fe incorporation using Mo
È ssbauer spectroscopy,
in the Kro
È ger-Vink notation), presumably some
isms for Fe incorporation between samples RFe1
rutile (pure TiO2) contains interstitial Ti
octahedral
Ti
3+
/
(TiTi )
and
oxygen
3+
which might reveal any differences in mechan-
(Tii
and RFe2 compared to RFe3. Further attempts to
vacancies
_) as stable defects. For the present discussion
(VO
synthesize
we will ignore the minor effect that H interstitials
contents
Fe-undersaturated
>0.3 wt.%
using
rutile
various
with
Fe
additional
(HiÇ) could have. In the Fe-doped samples we also
solid oxygen buffers have been unsuccessful,
have FeTi defects and possibly small amounts of
probably due to the dominant in¯uence of the
interstitial Fe (Fei
bulk composition on prevailing oxygen fugacity
/
_).
_
Tii
are only present in
during synthesis.
samples RFe1 and RFe2, which also have lower
Fe
contents
than
might
be
expected.
If
we
conclude that the main mechanism for chargebecause Fe is in octahedral coordination), then we
H incorporation in rutile as a function of Fe2O3
content
can assume that the Fe content of Fe-saturated
As previously noted, all NIR spectra shown in
rutile is a function of fO2. Therefore, the reducing
Fig. 1
conditions of the experimental synthesis would
implying the presence of structurally-incorporated
/
balancing FeTi
are VOÇÇ (as decoupled defects,
effectively govern the amount of Fe
3+
contain
sharp
OH-absorption
bands,
incorpo-
H. These OH-absorption bands are shown in more
rated. However, in the different synthesis experi-
detail in Fig. 2. Spectra obtained from R1 (pure
ments, several competing reactions must occur,
TiO2) contain one absorption band at 3279 cm
and these probably affect fO in turn. Presumably,
implying the presence of H in one position in the
Fe2O3 in the starting mixture reacts with TiO2 to
rutile structure. Due to the absence of other
form Fe-doped rutile, but must also be partly
substitutional impurities, H incorporation in R1 is
reduced and react with TiO2 to form ilmenite. We
probably
can
Swope et al. (1995) re®ned the H position in a
2
write
a
simple
equation
to
describe
the
equilibrium condition:
by
Ti
3+
reduction.
natural, H-bearing rutile from neutron diffraction
2FeTiO3 + ÝO2 = Fe2O3 + 2TiO2
ilmenite
charge-balanced
ÿ1,
rutiless
(1)
rutiless
data, and demonstrated that H is incorporated in
rutile close to (ÝÝ0), just off the shared O-O edge
in
where ss indicates solid solution of the compo-
the
octahedral
chain.
This
assignment
is
consistent with published polarized IR spectra
nents in rutile. In Fe-rich systems, reaction 1
from synthetic and natural rutile (Johnson et al.,
probably controls the fO2 in the capsule, buffering
1968; Vlassopoulos et al., 1993), where vibration
it to a higher value than expected. In contrast,
of the OH dipole is almost completely perpendi-
experiments performed using low bulk Fe2O3
cular to the crystallographic c axis. In contrast to
350
HYDROGEN AND MINOR ELEMENT INCORPORATION IN RUTILE
FIG. 2. OH-stretching region of the NIR spectra shown in Fig. 1 (vertical offset rescaled). Large and small divisions
on the vertical scale represent 10 and 5 cm
the R1 spectrum shown in Fig. 2, IR spectra
obtained
from
saturated
synthetic,
rutile
by
high-pressure
Khomenko
et
al.
at 3279
respectively.
associated with trivalent or divalent cations on the
4+
water-
Ti
(1998)
implies that all the H present in sample R1 is
contain two OH-absorption bands, a large band
ÿ1
cm
ÿ 1,
octahedral site. If correct, this assignment
decoupled from any Ti
3+
defects.
(i.e. similar to the band in the
Spectra obtained from rutile in samples R1 and
spectrum from R1), and a smaller, additional band
R2 contain two overlapping absorption bands, one
at 3324 cm
at 3279 cm
ÿ1.
This second absorption band has
(similar to R1) and an additional
rutile (e.g. Hammer and Beran, 1991) and shows
probably due to H associated with Fe
the same anisotropy as noted for the band at
rities. Regardless of whether we assign the OH
3279 cm
band at 3279 cm
Based on the behaviour of these two
3295 cm
ÿ1.
band
ÿ1.
at
ÿ1
also been noted in some other studies of synthetic
This
additional
band
3+
is
impu-
ÿ1 to H associated with Ti3+ or H
bands during annealing, Khomenko et al. (1998)
unassociated with compositional impurities, the
suggested that the band at 3324 cm
shift of the additional (3295 cm
H related to Ti
3379 cm
ÿ1
3+
ÿ1
was due to
/
(i.e. TiTi ), whilst the band at
was due to H unrelated to any Ti
3+
lower
wavenumbers
strongly
effects. That is, H and Fe
3+
ÿ1)
OH band to
implies
NNN
probably form a
defects. This assignment is in accordance with
defect cluster, with the proton associated with an
spectroscopic
O surrounding a Ti octahedral site occupied by
(Johnson
et
investigation
al.,
1968;
of
natural
Vlassopoulos
rutile
et
al.,
Fe
3+
/
(i.e. FeTi ± HiÇ). From this we can conclude
1993); IR spectra of natural rutiles often contain
that in the Fe-doped samples, a certain amount of
a sharp absorption band at 3279 cm
structurally incorporated H becomes coupled to
ÿ1,
and more
/
rarely, additional absorption bands at higher or
Fe Ti
lower
decoupled, or
wavenumbers.
Johnson
suggested that the band at 3279
to
H
not
associated
with
any
et
al.
ÿ1
cm
(1968)
was due
compositional
defects, and that other OH bands were due to H
351
defects,
with
the
remaining
conceivably coupled
H
with
either
/
TiTi
defects. Spectra obtained from sample R3 also
contain
3295
the
ÿ1
cm .
Fe Ti
/
±
H iÇ
absorption
band
at
In the R3 spectrum in Fig. 2 this
G. D. BROMILEY AND N. HILAIRET
slightly
amount of Cr2O3 as the starting mixture, possibly
asymmetric. By comparing all the spectra in
implying undersaturation. To determine whether
Fig. 2, it can be seen that the 3295 cm
band
is
broad
and
appears
to
be
ÿ1
band in
this was the case, we performed an additional
this spectrum probably contains a shoulder at
experiment, RCr2, using a starting mixture with
3279 cm
10 wt.% Cr2O3. In this experiment, the phase
ÿ1,
implying
the
presence
of
H
not
/
/
associated with FeTi , or H associated with TiTi .
carmichaelite was synthesized, and no rutile was
In the ®rst instance this would imply that even in
present
Fe-rich rutile, a considerable proportion of the H
[MO2ÿx(OH)x, where M includes Ti, Cr, Fe, Mg
remains decoupled from substitutional impurities,
and
which could have an important in¯uence on H
described
mobility and H diffusion in rutile. In the second
2000). Wang et al. (2000) describe an occurrence
instance, the presence of an additional band would
of carmichaelite in Garnet Ridge, Arizona, where
imply
it precipitated with its pyrope host and other
the
presence
of
reduced
Ti
only
on
in
Al]
the
is
a
run
products.
hydrous
(Wang
et
Carmichaelite,
titanate
al.,
1999;
only
recently
Wang
et
ÿ750ëC,
al.,
octahedral sites and not on interstitial sites (i.e.
titanates
no absorption due to IVCT) in the Fe-rich sample,
pressures of 2.0
and would imply that H solubility in the Fe-rich
carmichaelite
rutile sample is too great only to be compensated
pressure and high Cr contents. This appears to
/
at
temperatures
ÿ2.5 GPa.
is
of
650
and
They concluded that
probably
stabilized
by
high
by FeTi defects. This last conclusion is somewhat
be consistent with the results of experiments RCr
problematic, because the Fe
content of this
and RCr2, with carmichaelite only stabilized with
sample is several orders of magnitude higher than
Cr contents above 5 wt.%. The temperature of
3+
/
the H content, implying that isolated FeTi defects
synthesis of experiment RCr2 demonstrates a high
(which must be decoupled from VOÇÇ due to
thermal stability for this phase. To our knowl-
evidence
are
edge, this is the only reported synthesis of this
intrinsically more stable than coupled FeTi ± HiÇ
phase, and more work is required to determine its
defects.
stability
from
Mo
È ssbauer
spectroscopy)
/
One striking feature of the spectra shown in
and
possible
signi®cance
for
mantle
geochemistry.
Fig. 2 is that the height of all OH absorption
From Table 1 it can be seen that the solubility
bands increases signi®cantly with increasing Fe
of additional oxides in the rutile samples varies
content of the samples. This implies that water
considerably as a function of composition. For Fe-
contents of the samples increase with increasing
bearing samples a complex relationship between
Fe content. Lower water contents in Fe-poor
bulk composition and fO appears to have an effect
samples are probably related to the presence of
on Fe2O3 solubility in the samples. However, this
interstitial Ti
3+
, which could quite conceivably
2
situation is not expected to be as complex for
3+
2+
3+
3+
2+
reduce H af®nity in the samples due to an increase
Al
in
that
cations only have one common oxidation state
increasing H contents with Fe content do not
in rock-forming minerals (except Cr, which may
re¯ect the relative H af®nity of rutile due to
be present as Cr
differences between the H incorporation mechan-
stances). The NIR spectra from all rutile samples
O
overbonding.
It
should
be
noted
/
/
,
Mg
isms (i.e. FeTi ± HiÇ compared to HiÇ or TiTi ± HiÇ)
except
because both the bands at 3279 and 3295 cm
absorption
ÿ1
,
R1,
Cr
,
Ga
2+
RFe1
bands
,
Ca
because
these
under exceptional circumand
due
RFe2
to
show
Ti
3+
-Ti
no
4+
broad
IVCT.
show an increase in peak height with increasing
However, in sample RAl, rutile crystals are very
Fe content.
pale blue. The colour of these crystals cannot be
due
details,
run
products
and
a
crystal
®eld
effect
involving
Al
3+
substitution because it has no electrons in d
Solubility of minor elements in synthetic rutile
Experimental
to
orbitals. More likely, this slight colouration is also
the
due to Ti
3+
-Ti
4+
IVCT, presumably below the
compositions of rutile are listed in Table 1. In
detection limit of the spectrometer. Likewise, Ga-,
all experiments except RCr2, rutile was the main
Ca- and Mg-bearing rutile crystals are colourless
3+
run product. In all experiments except R1 (pure
due to a complete absence of Ti
TiO2) and RCr, small amounts of an additional
bearing rutile is dark red-brown, probably due to a
phase were also present, implying saturation of
crystal ®eld splitting effect involving Cr
the rutile with respect to the additional oxides. In
cation site (Burns, 1993), presumably Cr Ti .
experiment RCr, no additional phase was present,
and the rutile contains, within error, the same
352
. In contrast, Cr3+
on a
/
Figure 3 shows a plot of difference in ionic
radii between various cations from that of Ti
4+
HYDROGEN AND MINOR ELEMENT INCORPORATION IN RUTILE
ally put onto interstitial sites. However, the most
important mechanism for incorporation of lowervalence cations in synthetic rutile is probably
direct substitution for Ti
4+
(i.e. direct substitution
onto the octahedral site in the rutile structure),
charge-balanced
by
proton
incorporation
or
oxygen vacancies. It is possible that to a much
lesser degree trivalent cations are also present on
interstitial sites. Lack of a noticeable deviation for
Fe
3+
from the trend in Fig. 3 could imply that any
change in Fe2O3 solubility due to self-buffering of
fO is not large enough to be detected. Likewise,
2
´
the presence of Tii in rutile in sample RAl does
not alter Al solubility by a large enough degree to
FIG. 3. Plot of difference in ionic radius (6-coordinated,
octahedral sites) between cations used as additional
oxides in the present study and Ti
4+
measured
solubility
in
synthetic
rutile. A clear trend is noted in the plot. With
plot
that
with
number
to
include
of
in
Fig. 3.
extension
the
trivalent
However,
of
the
solubility
and
of
divalent
it
is
present
a
larger
cations
in
rutile, such deviations could be observed.
1100ëC. Values of ionic radius are listed in Table 2.
their
the
possible
dataset
against solubility of
oxides in the synthetic, water-saturated rutile at 2.0 GPa,
against
in¯uence
Coupled H and minor element incorporation
in synthetic rutile
increasing difference in ionic radii, the solubility
The NIR spectra obtained from synthetic rutile
of the cation decreases. Such a relationship would
doped with 5 wt.% additional oxides are shown in
be expected for direct substitution of various
Fig. 4. All spectra, except those obtained from
4+
. However, this plot masks some
sample RCa (not shown), contain OH-stretching
obvious discrepancies. Firstly, we have seen that
absorption bands over the wavenumber range
Fe2O3 reduction probably has an in¯uence on
4000
solubility of Fe2O3 in rutile in sample RFe3.
structurally-incorporated
Secondly,
sample
from
saturated,
although,
cations for Ti
RCr
is
possibly
Cr-under-
ÿ2500 cmÿ ,
1
RCa1
implying
contained
H.
no
the
presence
Spectra
features
of
obtained
over
this
Fe-
wavenumber range, implying an absence of H in
bearing series of experiments, the lack of IVCT
the structure. Polarized spectra were obtained
may indicate that rutile in this sample is close to,
from crystals in samples RFe, RCr and RGa, and
or at, saturation. Thirdly, and to a lesser degree,
show that the main absorption bands in these
by
inference
we can infer that in addition to Al
also contains Ti
3+
3+
to
the
, sample RAl
defects. However, the broad
spectra
have
a
perpendicular
to
95%
the
component
of
crystallographic
vibration
c
axis,
trend that we see in Fig. 3 probably indicates a
consistent with the H incorporation mechanism
similar incorporation mechanism for the various
proposed by Swope et al. (1995). The water
trivalent and divalent cations in rutile. As noted
contents for all samples were calculated from
above,
unpolarized IR spectra and are listed in Table 2.
Mo
È ssbauer
spectra
obtained
from
Fe-doped rutile and TiO2(II) indicate that Fe
3+
4+
Spectra from all samples consist of one main
)
absorption band. In Fe-, Al-, Cr- and Ga-bearing
site (Bromiley et al., 2004a). Computer simula-
rutile, this absorption band is at 3295, 3308, 3257
tions of defect structures in rutile have also
and 3314 cm
in both phases is present on the octahedral (Ti
demonstrated that Cr
preferentially
octahedral
3+
, Al
3+
incorporated
site,
3+
, Ga
in
and Fe
rutile
charge-balanced
3+
onto
by
ÿ1,
respectively. Variation in the
are
wavenumber of the main OH band with composi-
the
tion is consistent with the formation of coupled
oxygen
defects
/
involving
trivalent
cations
and
H,
vacancies (Sayle et al., 1995). In a study of
(M Ti ± H i Ç).
synthetic
absorption band, these spectra also contain an
Al-doped
Rentschler
pressure, Al
(1999)
3+
rutile,
showed
Gesenhues
that
at
4+
and
ambient
In
addition
additional band at ~3279 cm
to
the
main
OH-
ÿ1, usually only seen
is charge-
as a slight shoulder on the main OH-absorption
balanced by oxygen vacancies, but that with
band. As noted for the Fe-bearing samples, this
substitution for Ti
increasing Al2O3 contents, some Al
3+
is addition-
353
band is probably due to H decoupled from any
G. D. BROMILEY AND N. HILAIRET
FIG. 4. Unpolarized average NIR spectra (over the OH-stretching range) from synthetic, water-saturated rutile grown
at 2.0 GPa, 1100ëC. (a) R1, (b) RFe3, (c) RAl, (d) RMg, (e) RCr, (f) RGa.
compositional defects, but could also be due to H
associated with Ti
3+
. Spectra from sample RAl
incorporation
Unfortunately
related
no
data
to
interstitial
are
available
Al
on
3+
.
the
also contain a small OH absorption band at much
anisotropic nature of this small band, so direct
higher wavenumbers, 3551 cm
assignment
ÿ1,
which is not
is
not
possible.
However,
several
seen in the spectra of any other samples, and
potential positions for H incorporation in the
suggests that there is another mechanism for H
rutile structure have been identi®ed (Johnson et
incorporation in the Al-bearing rutile. Gesenhues
al., 1968), and the band at 3551 cm
and Rentschler (1999) noted that synthetic rutile
due to H
grown
at
amounts
ambient
of
pressure
Al 2 O 3
doped
contains
Al
with
3+
on
large
both
octahedral and interstitial sites. The additional
band at 3551 cm
ÿ1
could possibly be due to H
+
354
+
Spectra
contain
ÿ1
could be
in a Ý00 position.
an
obtained
from
asymmetric
Mg-bearing
band
at
rutile
3349 cm
ÿ1 ,
which can actually be resolved into three separate
bands, a large band at 3349 cm
ÿ1,
a shoulder at
HYDROGEN AND MINOR ELEMENT INCORPORATION IN RUTILE
3325 cm
ÿ1 and another small band at 3279
cm
ÿ1
amount of Ca
2+
, in contrast to the other trivalent
(i.e. the OH band noted in all spectra). This
and divalent cations considered, occupies inter-
interpretation
H
stitial sites in the rutile structure. From Figs 1 and
slightly
2 it can be seen that the presence of interstitial Ti
more favourable. The OH-stretching frequencies
markedly reduces the af®nity of rutile for H.
of these two absorption bands are similar, and no
Perhaps a similar effect is noted in the Ca-doped
variation in relative peak height was noted as a
sample due to interstitial Ca. The presence of
function
distinct
CaiÇÇ defects could be charge-balanced by the
difference in absorption band anisotropy. The H
coupled substitution of Ca onto the octahedral Ti
positions
implies
associated
of
that
there
with
orientation,
Mg
2+
are
,
one
implying
no
two
//
4+
/
incorporation mechanism proposed by Swope et
site (CaTi ) or by reduction of Ti
al. (1995) actually describes four equivalent H
absence of broad absorption bands in NIR spectra
positions about a symmetrical, regular octahe-
from RCa over the range 700 to 4000 cm
dron. Local distortion of the octahedral site in
indeed of any blue colouration of the Ca-doped
rutile due to Mg
result
in
2+
slight
frequencies
for
substitution could potentially
differences
H
+
in
OH-stretching
(TiTi ). The
crystals, excludes the possibility of Ti
3+
ÿ1,
or
inter-
stitials being present, but not of reduced Ti
4+
on
different
the main Ti site. The use of Ca-doping of rutile to
positions. If H incorporation in the MgO-saturated
prevent H incorporation could, potentially, have
located
in
these
rutile is related to Mg substutition for Ti
+
4+
, this
some important industrial applications.
are incorporated for
An interesting feature in IR spectra obtained
. Differences in the peak height of the
from the synthetic rutile samples is the signi®cant
absorption bands could indicate that, as a result of
variation in the width of the OH absorption bands
site distortion, the four equivalent H positions are
as a function of composition. Mostly, this effect is
split into at least two pairs with slightly differing
due to overlapping of the main OH absorption
OH-stretching
band, which has a different frequency in each
would imply that two H
every Mg
2+
frequencies,
with
the
slightly
ÿH bond being more favourable.
composition, with the smaller absorption band at
longer O
It
is
interesting
absorption-band
to
note,
on
3279 cm
ÿ1.
However, there could also be some
obtained from Ca-doped rutile (RCa) contain no
as a function of composition, possibly due to
OH
H
NNN effects: the presence of small amounts of
is not stable.
interstitial trivalent cations, or, in the example of
This
IR
of
variation in the width of the main absorption band
bands.
that
basis
spectra
absorption
assignment,
the
implies
incorporation coupled with CaTi
//
that
Indeed, decoupled H interstitials, or H interstitials
coupled with Ti
3+
defects are also not present in
this sample. As noted in Table 3, the difference in
ionic radii between Ti
4+
2+
RCr,
increased
likelihood
of trivalent
being on adjacent octahedral (Ti
4+
cations
) sites due to
higher solubility of the minor element. Signi®cant
is quite high
variation in water solubility in the synthetic rutile
(51.5 pm). Lack of structurally incorporated H in
is also noted as a function of composition. The
sample
considerable
highest water content measured is for the Cr-
distortion of the Ti site in the rutile structure
doped sample. However, no systematic variation
due to Ca occupancy. However, this does not
in water content as a function of oxide solubility
suf®ciently explain the absence of the 3279 cm
or ionic radius is seen. Water solubility probably
band.
RCa
This
could
could
be
and Ca
due
indicate
to
that
a
ÿ1
signi®cant
varies partly as a function of the change in unitcell volume due to H incorporation, and should,
TABLE 3. Ionic radii of cations in 6-coordinated
octahedral environments.
therefore,
be
distortion
due
partly
to
dependent
MTi
/
defects.
upon
relative
However,
we
have seen that in all samples some H is also
Cation
Ti
4+
3+
Fe
3+
Al
Mg
2+
3+
Ga
3+
Cr
Ca
2+
present which is either decoupled to composi-
Ionic radius (pm)
tional defects, or related to the presence of Ti
3+
.
Furthermore, at least one of the samples, RAl,
74.5
78.4
contains
67.5
probably leads to a reduction in H solubility.
86.0
Unfortunately, it is not possible to determine
76.0
accurately the areas of overlapping absorption
75.5
bands in the IR spectra, and therefore the amount
126.0
cations
on
interstitial
sites,
which
of H coupled to (or decoupled from) different
defects. Peak ®tting is made dif®cult by the fact
355
G. D. BROMILEY AND N. HILAIRET
that: (1) the signal to noise ratio is not ideal,
synthetic
mainly due to the small sample size and low
explanation appears to be more consistent with
concentration of H, (2) most absorption bands are
the similarities between IR spectra obtained from
broad and the amount of overlap is large, and
natural rutile samples with different compositions.
(3) background subtraction can lead to changes in
Furthermore, this assertion supports the argument
relative peak heights. Calculated water contents
that the 3279 cm
may
comparable
spectra in Figs 2 and 4 is due to H unrelated to
anyway, because of variation in OH-stretching
compositional impurities, because natural rutiles
frequencies
only
not,
however,
as
a
be
directly
function
of
composition.
rutile
contain
in
ÿ1
Ti
the
present
study.
This
OH-absorption band seen in
3+
under
extremely
limiting
Nakamoto et al. (1955) noted that there is a
circumstances, such as in meteoritic or lunar
signi®cant dependence of H concentration on
samples
OH-stretching frequency, which would imply a
crystallization was extremely low.
for
which
oxygen
fugacity
during
systematic overestimation of water contents in
Vlassopoulos et al. (1993) noted that rutiles
samples where the main OH absorption band is
from ma®c pegmatites contained an additional
shifted to higher wavenumbers.
a bs o r pt i o n
b a nd
a r o u nd
33 6 0 c m
ÿ1 ,
and
suggested that this may be due to H association
with Mg
Comparison with natural samples
Johnson
et
al.
(1968)
®rst
suggested
that
H
rutiles
contain
3349 cm
cation substitutions, but they were not able to
also
resolve
3324 cm
OH
. This correlates reasonably well with
data from the present study, where Mg-saturated
incorporation in rutile was linked to trivalent
impurity-associated
2+
bands
in
IR
ÿ1.
suggest
ÿ1
a
large
absorption
band
at
However, Vlassopoulos et al. (1993)
that
a
small
satellite
peak
at
seen in some rutile spectra, is due to
3+
spectra from rutiles doped with various oxides
association with Al
and
consistent with the results of the present study.
so
concluded
that
either
there
was
no
. This suggestion is not
association of OH groups with cations, or that
Spectra obtained from synthetic rutile samples
resulting shifts in OH-stretching bands were too
doped with various trivalent and divalent cations
small
replicate many of the features seen in spectra of
to
be
resolved.
However,
in
detailed
investigations of coupled H and minor element
natural
substitution
measured
in
natural
rutiles
from
various
geological terrains, Hammer and Beran (1991)
samples.
Vlassopoulos
integrated
OH-stretching
range
in
spectra
to
12300 cm
and Vlassopoulos et al. (1993) noted systematic
samples
from
1800
shifts in OH-stretching frequencies, and attempted
obtained
from
the
to correlate these with variation in composition
study range from 640 cm
and
from
geological
many
environment.
natural
rutiles
Spectra
obtained
commonly
contain
one main absorption band at 3280 to 3390 cm
ÿ1
(e.g. samples R-6, R-7, R-8, JAG83-30-3, JAG852 of Vlassopoulos et al., 1993). Fe
3+
is the major
14100 cm
ÿ2.
et
al.
absorbances
synthetic
ÿ2
(1993)
over
from
ÿ2 .
samples
the
natural
Values
in
this
(for pure TiO2) to
The largest values measured by
Vlassopoulos et al. (1993) were for rutiles from
ultra-high pressure environments, and are in very
good
agreement
with
data
from
our
study.
Interestingly, the greatest water content noted by
trivalent cation in many natural rutiles. Spectra
Vlassopoulos et al. (1993) was for a Cr-rich
from
RFe3,
sample, and the greatest water content of any
the
synthetic
Fe-saturated
rutile,
ÿ1,
which
sample in this study was measured for the Cr-
may be comparable to the band seen in spectra
doped sample. However, results from the present
from natural
study
contain an absorption band at 3295 cm
rutiles. Samples
R-6 to
R-8 of
indicate
that
many
of
the
correlations
Vlassopoulos et al. (1993) contain appreciable
between OH-stretching frequencies and composi-
amounts of Fe. In contrast, samples JAG83-30-3
tions previously reported are overstated. Most
and JAG85-2 contain much larger amounts of Cr.
likely, a signi®cant proportion of H in natural
Spectra from sample RCr contain an absorption
rutiles is not coupled to substitutional impurities.
band
wavenumbers
If so, this implies that small variations in OH-
and are not comparable to spectra
stretching frequencies between different rutile
obtained from these natural Cr-rich rutiles. An
samples is probably due to some variation in the
alternative explanation is that the main absorption
size and/or distortion of the octahedral (Ti
band in IR spectra from natural rutiles is unrelated
in the structure, which should be evident from
to compositional impurities, and is comparable to
detailed structural re®nements. Furthermore, H
the band at 3279 cm
incorporation in natural rutiles could be linked to
at
(3257 cm
signi®cantly
ÿ1),
lower
ÿ1 seen in all IR spectra from
356
4+
) site
HYDROGEN AND MINOR ELEMENT INCORPORATION IN RUTILE
4+
on
Johnson et al. (1973) and the sample independent
octahedral sites in the rutile structure. Natural
calibrations of Paterson (1982) and Libowitzky
rutiles often contain appreciable Nb and Ta, and
(1999).
Ti
substitution
for
pentavalent
cations
so this could provide an ef®cient mechanism for
H incorporation, and would also result in the
formation of H interstitials unrelated to other
Acknowledgements
The authors would like to thank Heinz Fischer,
compositional impurities.
Defect structures in synthetic and natural
rutile
Georg
Herrmannsdo
È rfer,
Hubert
Schulze
for
Detlef
technical
Krauûe
and
assistance,
and
Fabrice Gaillard and Fiona Bromiley for help
with preparation of the manuscript. This manu-
The main mechanism for H incorporation in the
script was improved by the comments of an
synthetic rutile samples is association with lower-
anonymous reviewer.
valence cation substitution onto octahedral sites.
However, the amount of water incorporated into
rutile (Table 3) is at least one order of magnitude
References
lower than the solubility of lower-valence cations
Bromiley, G. and Keppler, H. (2004) An experimental
(Table 1), implying that the dominant mechanism
investigation of hydroxyl solubility in jadeite and
for trivalent and divalent cation substitution into
Na-rich pyroxenes. Contributions to Mineralogy and
Petrology,
the synthetic rutile does not involve H incorpora-
147,
189
ÿ200.
tion. We have argued that the dominant mechanism
Bromiley, G., Hilairet, N. and McCammon, C. (2004a)
for trivalent and divalent cation substitution in
Solubility of hydrogen and ferric iron in rutile and
synthetic rutile probably involves oxygen vacan-
TiO2 (II): Implications for phase assemblages during
cies, and to a lesser degree, could also involve
ultrahigh-pressure metamorphism and for the stabi-
cations
lity
on
interstitial
sites.
In
natural
rutile,
trivalent cation substitutions can also be chargebalanced by coupled substitutions with pentavalent
3+
cations (i.e. M
5+
+ M
charge-balance trivalent cation substitutions is not
speciation
conditions expected in many geological environ-
al., 2004b). Vlassopoulos et al. (1993) noted that
Field
pentavalent cation content, and suggested that this
content
in
many
rutiles
favour
of
using
the
integral
mantle.
gem-quality
89,
941
ÿ949.
Cr-diopside.
Cambridge
University
Press,
to
the
Rock-Forming
Minerals.
Longmans, Essex, UK, 696 pp.
Gesenhues, U. and Rentschler, T. (1999) Crystal growth
and defect structure of Al
Solid State Chemistry,
3+
143,
-doped rutile. Journal of
210
ÿ218.
Goresy, A., Chen, M., Gillet, P., Dubrovinsky, L.,
Graup, G. and Ahuja, R. (2001) A natural shockinduced dense polymorph of rutile with alpha-PbO 2
structure in the suevite from the Ries crater in
is
Germany. Earth and Planetary Science Letters,
considerable, and has been used as an argument
in
natural,
Theory.
Introduction
amount of H needed to charge balance the excess
cation
lower
L04610.
Deer, W., Howie, R.A. and Zussman, J. (1992) An
was due to large amounts of H incorporation. The
trivalent
the
31,
Cambridge, UK, 551 pp.
the total trivalent cation content of a suite of rutile
samples was systematically higher than the total
in
Burns, R. (1993) Mineralogical Applications of Crystal
ments where rutile is an important accessory
mineral, such as subducting slabs (Bromiley et
in
American Mineralogist,
known. However, redox conditions in the pistoncylinder assembly probably closely resemble redox
polymorphs
F. and Jacobsen, S. (2004b) Hydrogen solubility and
). The extent to
which natural high-pressure rutiles are reduced,
and the extent to which oxygen vacancies can
silica
Bromiley, G., Keppler, H., McCammon, C., Bromiley,
4+
= 2 Ti
of
Geophysical Research Letters,
485
ÿ495.
192,
absorption
Hammer, V. and Beran, A. (1991) Variations in the OH
coef®cient determined by Hammer and Beran
concentrations of rutile from different geological
(1991), which gives unrealistically high water
contents.
However,
the
presence
of
environments. Mineralogy and Petrology,
45,
1
ÿ9.
oxygen
Johnson, O., Ohlsen, W. and Kingsbury, P., Jr. (1968)
vacancies in natural rutile, or the presence of
Defects in rutile (iii). Optical and electrical proper-
interstitial cations could also account for a large
ties
amount of excess trivalent cation content, and
Review,
would imply that the Hammer and Beran (1991)
of
Johnson,
impurities
175,
O.,
and
charge
ÿ1108.
carriers.
Physical
1102
DeFord,
J.
and
Shaner,
J.
(1973)
calibration signi®cantly overestimates the water
Experimental technique for the precise determination
content of rutile relative to the calibration of
of H and D concentration in rutile (TiO2). Journal of
357
G. D. BROMILEY AND N. HILAIRET
44,
Applied Physics,
3008
ÿ3012.
Sayle, D., Catlow, C., Perrin, M.-A. and Nortier, P.
Khomenko, V., Langer, K., Rager, H. and Fett, A.
(1998)
Electronic
absorption
by
Ti
3+
ions
(1995) Computer simulation study of the defect
and
chemistry of rutile TiO2. Journal of the Physics and
electronic delocalization in synthetic blue rutile.
Physics and Chemistry of Minerals,
25,
338
ÿ346.
Chemistry of Solids,
Libowitzky, E. (1999) Correlation of O-H stretching
type
frequencies and O-H´´´O hydrogen bond lengths in
minerals. Monatshefte fu
Èr Chemie,
Libowitzky,
E.
absorption
and
Rossman,
calibration
American Mineralogist,
for
82,
G.
130, 1047ÿ1059.
(1997)
water
1111
in
ÿ1115.
An
IR
minerals.
(2001) Defects in the reduced rutile single crystal.
304,
147
ÿ151.
rutile
compounds:
II.
in
H-bearing
Mineralogist,
80,
454
ÿ456.
compounds:
I.
diffraction
study
of
Al
American
Mineralogist,
80,
of
448
H
ÿ453.
neutron
in
rutile.
and
X-ray
American
Coupled substitution of H and minor elements in
rutile and the implications of high OH contents in
and
Nb-
American Mineralogist,
ÿ29.
chemistry
stishovite.
Single-crystal
and FTIR spectroscopy. Mineralogy and Petrology,
21
Crystal
substitution
cassiterite deduced from nuclear reaction analysis
71,
ÿ805.
Vlassopoulos, D., Rossman, G. and Haggerty, S. (1993)
Maldener, J., Rauch, F., Gavranic, M. and Beran, A.
(2001) OH absorption coef®cients of
799
Swope, R., Smyth, J. and Larson, A. (1995) H in rutile
Lu, T.-C., Wu, S.-Y., Lin, L.-B. and Zheng, W.-C.
Physica B,
56,
Smyth, J., Swope, R. and Pawley, A. (1995) H in rutile-
and
Cr-rich
rutile
from
78,
the
1181
upper
ÿ1191.
mantle.
Wang, L., Essene, E. and Zhang, Y. (1999) Mineral
Nakamoto, K., Margoshes, M. and Rundle, R. (1955)
inclusions in pyrope crystals from garnet ridge,
Stretching frequencies as a function of distances in
Arizona, USA: Implications for processes in the
hydrogen bonds. Journal of the American Chemical
upper
Society,
Petrology,
77,
6480
ÿ6486.
mantle.
135,
Contributions
164
ÿ178.
to
Mineralogy
and
Paterson, M. (1982) The determination of hydroxyl by
Wang, L., Rouse, R., Essene, E., Peacor, D. and Zhang,
infrared absorption in quartz, silicate glasses and
Y. (2000) Carmichaelite, a new hydroxyl-bearing
similar
ÿ29.
materials.
Bulletin
of
Mineralogy,
105,
titanite
20
from
Mineralogist,
Rossman, G. and Smyth, J. (1990) Hydroxyl contents of
garnet
85,
792
ridge,
ÿ800.
Arizona.
accessory minerals in mantle eclogites and related
[Manuscript received 27 July 2004:
rocks. American Mineralogist,
revised 22 April 2005]
75,
775
ÿ780.
358
American