American Mineralogist, Volume 87, pages 603–608, 2002

Hydroxyl in MgSiO3 akimotoite: A polarized and high-pressure IR study

NATHALIE BOLFAN-CASANOVA,* HANS KEPPLER,† AND DAVID C. RUBIE

Bayerisches Geoinstitut, Universität Bayreuth, D-95440 Bayreuth, Germany

ABSTRACT
Hydrous MgSiO3 akimotoite (ilmenite-type structure), containing 350 ppm wt H2O as hydroxyl,
was synthesized at 19 GPa and 1300 ∞C in a multi-anvil apparatus. Polarized Fourier-transform
infrared spectra show bands at 3300 and 3320 cm–1 due to hydroxyl groups oriented nearly perpen-
dicular to the c axis as well as a strong band at 3390 cm–1 caused by OH molecules aligned nearly
parallel to c. Unpolarized infrared measurements performed up to 13.8 GPa in a diamond anvil cell
show that the bands shift to lower frequencies with increasing pressure, which is consistent with
hydrogen bonding. At ~9 GPa, the bands initially at 3300 and 3320 cm–1 merge, suggesting that these
two bands represent the same type of hydroxyl group. The compression behavior of the infrared
bands shows that the pressure derivative of the frequency, dn/dP, decreases with increasing initial
wavenumber, nio, and this decrease is not linear. These observations are consistent with the compres-
sion mechanism of akimotoite, as previously published. The data indicate that the relative compress-
ibility of the hydrogen bonds within a structure (i.e., the sign of dn/dP vs. nio) is governed by a
combination of the directionality of the hydrogen bonds and the anisotropy of compression, whereas
the rate of compression (i.e., the slope of dn/dP vs. nio) is a function of the bulk modulus.

INTRODUCTION HIGH-PRESSURE EXPERIMENTAL METHODS

Trace amounts of water are a common constituent of most
nominally anhydrous minerals (NAMs) of the Earth’s mantle,
and hydrogen species are ubiquitously found in mantle xeno-
liths (Bell and Rossman 1992) as well as in mantle phases syn-
thesized at high pressure and temperature (Kohlstedt et al.
1996). Such hydrogen, dissolved as hydroxyl point defects,
plays an important role in the physical and chemical properties
of mantle rocks, even at very low concentrations (Hirth and
Kohlstedt 1996). These properties include electrical conduc-
tivity, ionic diffusivity, and rheology. It is therefore important
to characterize hydrogen species in NAMs and their incorpo-
ration mechanisms in order to understand the dynamics of the
Earth’s mantle.
Fourier transform infrared (FTIR) spectroscopy is a sensi-
tive tool for detecting trace amounts of water in minerals, melts,
and glasses. This technique makes it possible to distinguish
the various species of hydrogen (present in the form of hy-
droxyl or water molecules) and, with the use of polarized ra-
diation, IR also provides information about the orientation of
OH dipoles in crystal structures. In the case of MgSiO 3
akimotoite, we have combined polarized IR data with in situ
measurements of infrared spectra at high pressures in order to
constrain possible proton sites, because sites with different
compressibilities can be distinguished by differences in the
shifts of the OH stretching frequency with pressure.
* Present address: Laboratoire Magmas et Volcans, UMR 6524,
Université Blaise Pascal, 5 rue Kessler, 63038 Clermont-
Ferrand, France. E-mail: N.Bolfan@opgc.univ-bpclermont.fr
† Present address: Mineralogisches Institut, Universität
Tübingen, Wilhelmstr. 56, D-72074 Tübingen, Germany.
0003-004X/02/0506–603$05.00 603
The single crystals of MgSiO3 akimotoite (ilmenite-type
structure), used for room-pressure polarized IR measurements
and the high-pressure unpolarized IR study, were synthesized
at 19 GPa and 1300 ∞C in a multi-anvil apparatus from a mix-
ture of oxides having MgSiO3 + 5 wt% H2O + excess silica
bulk composition (see Bolfan-Casanova et al. 2000 for details).
The sample was prepared as a doubly polished slab 47 mm thick.
This akimotoite contains about 3900 H/106 Si (~350 ppm wt
H2O) as determined from polarized IR measurements (Bolfan-
Casanova et al. 2000), based on the calibration of the extinc-
tion coefficient by Paterson (1982) (see Table 1).
The crystal sections used for both the room-pressure and
the high-pressure investigations were oriented parallel to the
c-direction as determined by optical microscopy using
conoscopic light. The directions [100], [010], and [110] are
randomly distributed within the crystal plate. High-pressure
IR measurements were carried out using a Merryl-Basset type
diamond anvil cell (DAC) equipped with type I diamonds. A
250 mm thick rhenium gasket was pre-indented to 130 mm and
a 150 mm diameter hole was used as the sample chamber. Cryo-
genically loaded N2 was used as a hydrostatic pressure-trans-
mitting medium. Pressure was determined from the shift in the
fluorescence peak of two ruby chips located in the sample cham-
ber (Mao et al. 1986). Infrared spectra were measured using a
Bruker IFS 120 HR high-resolution FTIR spectrometer and
polarized radiation was generated using a wire-strip polarizer
on a KRS-5 substrate (for more details see Bolfan-Casanova et
al. 2000). The background spectra for the high-pressure IR
measurements were obtained in air because blank spectra mea-
sured through the DAC displayed interference fringes. The IR
spectra were recorded with 1 cm–1 resolution and 1000 to 1500
604 BOLFAN-CASANOVA ET AL.: HIGH-P IR STUDY OF H IN AKIMOTOITE

TABLE 1. OH concentrations (in mol H/L) in MgSiO3 akimotoite for each of the OH stretching bands, as determined with polarized IR
radiation
n–OH (cm–1) [OH]e * [OH]w * [OH]total = Percentage of absorption Percentage of absorption Percentage of [OH]total
[OH]e + 2 [OH]w in one direction (band area) in one direction (peak height)
3390 0.0663 0.0049 0.0761 87% along e 93% along e 51
3320 0.0071 0.0202 0.0475 85% along w 75% along w 31.8
3300 0.0010 0.0078 0.0166 93% along w 90% along w 11.1
3260 – 0.0019 0.0019 100% along w 100% along w 1.3
3050 0.0072 – 0.0072 100% along e 100% along e 4.8
S 0.0816 0.0348 0.1493 mol H/L 100
352 ppm wt H2O
6
3933 H/10 Si
–
Note: [OH] was calculated by integrating the band area using the extinction coefficient of Paterson (1982), i.e., 150 ¥ (3780–n ), in cm–1/(mol/L).
* e and w denote directions parallel and perpendicular to the c axis respectively.

scans. Baseline corrections of absorbance spectra were carried

out with a spline fit defined by points outside the OH-stretch-
ing region.

RESULTS AND DISCUSSION

Room-pressure polarized IR data

The polarized IR spectra of MgSiO3-akimotoite are shown
in Figure 1. The spectra consist of five pleochroic bands due to
structurally bound OH groups: three sharp, strong bands at 3300,
3320, and 3390 cm–1, and two weak bands at 3260 and 3050
cm–1. Since the MgSiO3-ilmenite structure has only one type of
O atom (Horiuchi et al. 1982) the various OH stretching fre-
quencies are due to proton sites that differ only in orientation
(and therefore also the geometry and/or the length) of the hy-
drogen bond. The band at 3300 cm–1 is strongly polarized per-
pendicular to the c axis and is consistent with hydroxyl (OH1a)
oriented nearly parallel to the plane of the shared face between
two octahedra, as shown in Figure 2. The pleochroism of the
band at 3320 cm–1 is similar to that of the 3300 cm–1 band,
except that it has only 75% absorption (expressed here as peak
height) perpendicular to the c axis instead of 90% in the case
of the 3300 cm–1 band (see Table 1). The dipole, OH1b, corre-
sponding to the 3320 cm–1 band thus makes a larger angle with
the c axis than OH1a. The very weak band at 3260 cm–1 is also
strongly polarized perpendicular to c. The orientation of the
OH dipoles within planes parallel to c is not constrained. How-
ever, if H is incorporated via a cation vacancy mechanism (in
contrast to an interstitialcy mechanism), then it is expected that
the proton sits on an octahedral face in order to efficiently com-
pensate for the excess of negative charge at the O atoms. In
this case, the hydrogen bond is likely to have a bifurcated ge-
ometry. Note that, due to the trigonal symmetry, there are three
equivalent directions for the OH dipoles. The bands at 3050
and 3390 cm–1 have maximum intensities parallel to the c axis:
the 3390 cm–1 band has a small oblique component (~7% per-
pendicular to c), whereas the 3050 cm–1 band has virtually no
component perpendicular to the c axis. Such a pleochroism is
consistent with OH vectors pointing into a tetrahedral void,
possibly implying trifurcated hydrogen bonds (see Fig. 2).
However, due to the large difference in vibrational energy be-
tween the 3390 and the 3050 cm–1 bands, these are probably
not due to the same type of hydroxyl ion. Therefore, in Figure
2, the OH2 vector refers exclusively to the dipole responsible
for the 3390 cm–1 band.
FIGURE 1. Polarized FTIR spectra of MgSiO3 akimotoite containing
3900 H/106 Si (~350 ppm wt H2O). The 47 mm thick doubly polished
section was cut parallel to the c axis.
Table 2 lists the O-O bond lengths encountered in the Mg
and Si sites of magnesium silicate ilmenite (Horiuchi et al.
1982), and Table 3 displays the OH◊◊◊O distances correspond-
ing to the observed hydroxyl bands, estimated from the recent
distance-frequency correlation of Libowitzky (1999). The lat-
ter distances range between 2.66 and 2.80 Å, with the three
intermediate distances clustering around 2.74 ± 0.01 Å. Com-
parison of the values in Table 2 and 3 shows that the OH◊◊◊O
distances inferred from the IR data do not match the O-O bond
lengths in the Mg and Si sites of the akimotoite structure (with
one exception). Such a mismatch suggests that the hydrogen
bonds are not straight, because the low OH content (3900 H/
106 Si) is not likely to significantly affect the bond lengths. For
example, the polarized spectra indicate that the band at 3300
cm–1 corresponds to a dipole (OH1a) oriented parallel to the plane
of the shared face between two octahedra. In this case, there
are three crystallographic O-O bonds corresponding to the
shared face of Mg and Si octahedra (at ~2.50 Å) and the oppo-
site unshared faces of the Mg (at ~3.08 Å) and Si (at ~2.65 Å)
octahedra. The length of the hydrogen bond for this dipole,
inferred from the distance-frequency correlation, gives an
 BOLFAN-CASANOVA ET AL.: HIGH-P IR STUDY OF H IN AKIMOTOITE 605

TABLE 3. Hydroxyl stretching frequencies and corresponding

OH…O bond distances in MgSiO3 akimotoite, as deter-
mined using the distance-frequency correlation from
Libowitzky (1999)
OH stretching frequency (cm–1) OH…O bond distance (Å)
H2 3390 2.80
3320 2.75
Mg 3300 2.74
H1 3260 2.73
3050 2.66

Si
OH◊◊◊O distance of 2.74 Å (see Table 3). This difference sug-
z
y gests that the H bridge is not straight. Bent bonds constantly
x
yield higher wavenumbers from the distance-frequency corre-
lation (Libowitzky 1999), therefore implying overestimated
inferred OH◊◊◊O distances, but bi- and trifurcated bonds are still
poorly documented. Perpendicular to c, the O atoms are re-
lated by a threefold axis, suggesting that the hydrogen bond
Si
could at least have a bifurcated geometry. In this case, the dis-
Si tance-frequency correlation is of limited value for locating the
H1
proton and independent information is required.
H2
H1 Mg The small splitting between the 3300 and 3320 cm–1 bands
H2 Mg is interpreted to result from positional ordering when quench-
ing from high pressure and temperature. Indeed, the high-pres-
sure IR data show the merging of these two bands at ~9 GPa
(see below). Moreover, these two bands have a similar pleoch-
roic behavior. Based on these observations, it is plausible that
z z
these two bands are due to the same type of hydroxyl ion. Also,
the bands at 3300, 3320, and 3390 cm–1 account for about 95%
x
y y of the OH concentration and thus represent the major incorpo-
ration mechanisms of H in MgSiO3 akimotoite (see Table 1). It
F IGURE 2. Model of the OH dipole orientation in MgSiO 3- thus appears that there are two main types of OH groups in
akimotoite inferred from the polarized FTIR data. The H atoms occupy
akimotoite, associated with the 3390 and 3300 + 3320 cm–1
vacant cationic sites (here a vacancy of Mg, shown as an empty square);
their sites are represented by black lines, and the hydrogen bonds by
bands. In addition, the hydroxyl concentration associated with
dashed lines. Top and bottom left figures: structure viewed respectively the 3390 cm–1 band almost equals that associated with the 3300
down the c axis and perpendicular to c. The direction of OH1 vectors is and 3320 cm–1 bands together, suggesting that hydrogen is
inferred from the strong intensity of the IR bands at 3300 and 3320 mainly incorporated in akimotoite as proton pairs, possibly
cm–1 perpendicular to the c axis. In this case, the OH bond has a substituting for Mg.
difurcated geometry. These OH1 are not exactly located on the triangular
face of the vacant octahedron, but make an angle with that plane, an Comparison with the IR spectrum of hydroxyl in
approximation of which is shown in the bottom left figure. The OH2 corundum
vector is associated with the band at 3390 cm–1, which is strongly Isomorphic structures can be used as analogues for assign-
polarized parallel to the c axis. This H atom is sitting in a tetrahedral
ing hydroxyl bands in high-pressure NAMs. The IR spectrum
environment, implying a trifurcated OH bond, as shown in the bottom
right figure. Note that due to the threefold axis there should be three
of colorless Verneuil-grown corundum measured by Beran
equivalent OH directions. (1991) is very similar to that of MgSiO3-akimotoite measured
in this study. The hydroxyl bands of corundum, at 3310, 3280,
and 3230 cm–1, are shifted to lower frequencies compared to
the bands of MgSiO3-akimotoite, at 3320, 3300, and 3260 cm–1.
This shift is consistent with the average O-O bond length be-
ing shorter in Al2O3 than in MgSiO3-akimotoite. In particular,
the O-O bond lengths of the AlO6 octahedron in corundum are
TABLE 2. Interatomic O-O distances (Å) in MgSiO3-akimotoite (il- intermediate between the O-O bond lengths of the SiO6 and
menite-type structure), from Horiuchi et al. (1982)
MgO6 octahedra in MgSiO3-akimotoite (Maslen et al. 1993;
SiO6 octahedron MgO6 octahedron
Horiuchi et al. 1982). In addition, Beran (1991) found that in
2.331 [3] eSi 2.497 [3] fSi
2.497 [3] fMg 2.941 [3] corundum all OH bands are strongly polarized perpendicular
2.653 [3] 3.052 [3] eMg to the c axis, which is also the case for the analogous bands in
2.682 [3] 3.076 [3] MgSiO3-akimotoite. A fourth band at 3185 cm–1 is only ob-
mean 2.541 mean 2.892
Note: Each octahedron shares three edges (e) with an octahedron of the
served in corundum. In contrast, the bands in MgSiO 3 -
same type and one face (f) with an octahedron of the other type. akimotoite that are polarized parallel to the c axis, namely those
606 BOLFAN-CASANOVA ET AL.: HIGH-P IR STUDY OF H IN AKIMOTOITE

at 3050 and 3390 cm–1, do not have analogues in the IR spec- shift to lower wavenumbers upon compression, consistent with
trum of corundum. hydroxyl groups involved in hydrogen bonding (Nakamoto et
The highest intensity band in corundum, at 3230 cm–1, gives al. 1955). At 8.7 GPa, the two bands initially at 3320 and 3300
an inferred OH◊◊◊O distance of about 2.72 Å. Beran (1991) in- cm–1 overlap completely, but the two components could still be
terpreted the pleochroism of absorption as an H ion lying in fitted with Lorentzian functions. However, above 9.56 GPa and
the plane of the shared face between two neighboring octahe- up to 13.8 GPa only one band could be resolved at low fre-
dra (with a crystallographic O-O distance of 2.88 Å, Maslen et
al. 1993), implying a bifurcated bond. Thus, in the case of co- TABLE 4. Peak positions, band widths and their pressure deriva-
rundum, the “observed” OH◊◊◊O distance is smaller than the tives as a function of pressure
crystallographic O-O distance, with a discrepancy DlO-O ~0.2 Pressure Peak positions FWHH (cm–1)‡
Å. If the analogy is true for akimotoite, then the inferred dis- (GPa)* (cm–1)† no = 3300 no = 3320 no = 3390
tances for the bands at 3320 and 3300 cm–1 (of ~2.74–2.75 Å) 10–4 3300 3320 3388 8.8 9.2 28.5
0.84 (08) 3295 3315 3382 6.3 (1) 7.9 (4) 26 (2)
should correspond to a larger crystallographic O-O bond length. 1.64 (07) 3292 3311 3378 6.8 (0) 8.5 (0) 28 (2)
This line of reasoning points to the O-O distance of 3.08 Å 2.55 (08) 3288 3305 3372 8.0 (5) 8.8 (1) 28 (1)
(with DlO-O ~0.3 Å), suggesting that the OH1 bands observed in 3.61 (10)
4.59 (10)
3283 3300 3366
3279 3294 3361
8.9 (8)
8.3 (2)
9.6 (3)
12.0 (1)
28 (1)
31 (2)
akimotoite are due to hydrogen substituting in the Mg site. 5.57 (17) 3274 3289 3356 8.4 (6) 12.8 (5) 34 (5)
6.50 (08) 3270 3284 3350 8.9 (6) 13.3 (9) 30 (3)
High-pressure IR data 7.48 (11) 3266 3277 3345 8.1 (4) 14 (1) 38 (1)
8.70 (09) 3260 3270 3336 12 (1) 14.3 (7) 40 (0)
Some representative unpolarized IR spectra recorded dur- 9.56 (01) 3259 3267 3331 16 (3) 13 (3) 39 (5)
ing compression are shown in Figure 3 after baseline correc- 10.7 (1) 3254 3323
11.7 (2) 3249 3320
tion and smoothing. The spectrum at 1 atm was taken outside 12.6 (1) 3242 3307
the DAC. The bands at 3050 and 3260 cm–1 were not observed 13.8 (2) 3235 3305
at high pressure because their initial intensity was too low. The dn/dP (cm–1/GPa) § d(Dn)/dP (cm–1/GPa) ||
bands were fitted with a Lorentzian function in order to extract –5.6 –4.8 –4.1 0.34 0.81 1.41
the peak positions and the band widths as a function of pres- * Numbers in parentheses indicate the uncertainty in the last digit(s), as
sure (see Table 4). The differences between the fitted and the determined from fluorescence measurements on two ruby ships located
observed spectra are very small, and the residuals are attrib- in the sample chamber.
† The uncertainty in frequency determination is less than 1 cm–1, as ob-
uted to an effect of the baseline correction. The peak param- served by performing two series of fits with two independent baseline
eters are shown in Figures 4 and 5. corrections.
‡ Numbers in parentheses indicate the uncertainty in the last digit, as
As shown in Figures 3 and 4, the OH stretching bands all determined from two series of fits with two independent baseline correc-
tions.
§ As determined by a second order polynomial fit.
|| As determined by a linear fit.

FIGURE 3. Representative unpolarized FTIR spectra of MgSiO3-
akimotoite recorded during compression (only half of the available
spectra are shown for clarity), after baseline correction and smoothing.
Sample thickness is 47 mm. (*) indicates a band arising from the
diamonds.
FIGURE 4. OH-stretching frequencies of MgSiO3-akimotoite as a
function of pressure. The numbers indicated are the values for the initial
pressure derivatives of the frequency, dn/dP (in cm –1/GPa), as
determined by a polynomial fit to the data (dashed line). The uncertainty
in the determination of the frequencies is smaller than the symbols.
 BOLFAN-CASANOVA ET AL.: HIGH-P IR STUDY OF H IN AKIMOTOITE
FIGURE 5. Evolution of the full width at half height of the OH-
stretching bands in MgSiO3-akimotoite as a function of pressure. The
numbers shown are the pressure derivatives, d(Dn)/dP (in cm–1/GPa),
of the band widths determined by a linear fit to the data (shown by the
dashed line).
quency, and there is no evidence of a crossing over between
the two bands. In addition, the intensity of the bands strongly
decreases with increasing pressure, concomitant with a band
broadening (see Fig. 5), especially for the 3390 cm–1 band. Such
broadening and decrease in intensity of the OH bands has al-
ready been observed at high pressures for other OH stretches
(Winkler et al. 1989; Duffy et al. 1995). Hofmeister et al. (1999)
proposed that the ubiquitous increase of band width with pres-
sure for OH stretching bands is due to an increased distribu-
tion of OH◊◊◊O angles, this effect being larger for bent bonds
than for straight bonds. As shown in Figure 5, at any given
pressure the widths of the akimotoite OH bands increase with
the wavenumber of the stretching vibration. This suggests that
the configuration of the hydrogen bond for these three bands is
different. In particular, it is expected that the broader band is
due to a proton sitting in a more disordered environment. This
interpretation is consistent with the model of the hydrogen lo-
cation inferred from the polarized data, which indicates that
this proton is occupying a tetrahedral site. This implies a tri-
furcated hydrogen bond, probably with a band width larger than
that which would be expected from a straight hydrogen bond.
Also, the increase of band width as a function of pressure is
larger for the band at 3390 cm–1, and decreases with the initial
frequency of the bands.
Interpretation of the frequency shifts
Figure 4 shows the evolution of frequency as a function of
pressure. A second order polynomial fit to the data gives initial
dn/dP values of –5.6, –4.8, and –4.1 cm–1/GPa for the 3390,
3320, and 3300 cm–1 bands, respectively. A linear function fails
607
to reproduce the apparent bending of the data for the 3320
cm–1 band at intermediate pressures, in a way that we judged
unacceptable given the fact that the relative uncertainty in fre-
quency determination is less than 1 cm–1. According to the data
of Libowitzky (1999), the correlation between OH stretching
frequency and OH◊◊◊O distance is not linear, the slope depend-
ing on the frequency range. However, since the OH bands in
akimotoite are located at very similar frequency, the relation-
ship between frequency and OH◊◊◊O distance in this narrow
range is practically linear. The present data thus show that the
absolute value of the pressure derivative of the OH stretching
frequency, |dv/dP|, increases with increasing initial wavenum-
ber. Considering that the O-H distance is practically indepen-
dent of pressure (Nelmes et al. 1993), the decrease in OH
stretching frequency (i.e., the weakening of the force constant)
is attributed to a strengthening of the hydrogen bond as a con-
sequence of the decrease in O◊◊◊O distance. Thus, the trend of
|dv/dP| observed in this study implies that, according to the
correlation between O-H◊◊◊O distances and OH stretching fre-
quencies, the proton site corresponding to the 3390 cm–1 band
is more compressible than the proton site corresponding to the
3300 cm–1 band. This is in agreement with both the compres-
sion mechanism of magnesium silicate ilmenite and the orien-
tation of the OH dipoles within the structure discussed in the
previous section. Reynard et al. (1996) found the compression
of MgSiO3 akimotoite to be anisotropic with the c axis being
twice as compressible as the a axis. This is due to the alternat-
ing stacking of MgO6 and SiO6 octahedral layers normal to the
c axis, which causes the compression along the c axis to be
achieved through flattening of the more compressible MgO6
octahedra, whereas compression along the a axis is controlled
by the less compressible SiO6 octahedra. Thus, a higher |dv/dP|
for the 3390 cm–1 band is consistent with a higher compress-
ibility along the c direction, whereas a lower |dv/dP| for the
3300 cm–1 band agrees with a lower compressibility perpen-
dicular to c. The band at 3320 cm–1 shows an intermediate value
of |dv/dP| consistent with its mixed component of absorption
in directions parallel and perpendicular to c. The ratio of |dv/dP|
between the 3390 and 3300 cm–1 band is not equal to the 2:1
ratio in compressibilities between the c and a axis of the
akimotoite structure, because the axial compressibilities rep-
resent net effects that involve the compression of several octa-
hedral units, some of which are vacant. Instead, |dv/dP| here
reflects the compressibility of the individual proton sites.
Our results contrast with the previous report of Hofmeister
et al. (1999) who found that the pressure derivative of the OH
stretching frequency, dn/dP, increases linearly as a function of
the initial wavenumber, nio, for several different phases. In the
present study the opposite trend is observed and dn/dP decreases
with increasing nio. Also, this decrease is not linear. However,
when the absolute value of the pressure derivative is consid-
ered, |dv/dP|, our data are in agreement with the observation of
Hofmeister et al. (1999) that the trend of |dv/dP| vs. nio is steep
for compressible phases and shallow for relatively incompress-
ible phases. Accordingly, the authors concluded that the slope
of |dv/dP| vs. nio is correlated with the bulk modulus K0T. This
is illustrated in Figure 6 for akimotoite, wadsleyite, and Phase
B. Phase B is softer than hydrous wadsleyite, with a bulk modu-
608 BOLFAN-CASANOVA ET AL.: HIGH-P IR STUDY OF H IN AKIMOTOITE
FIGURE 6. Pressure derivative of the frequency as a function of
initial wavenumber for MgSiO3 akimotoite containing ~350 ppm H2O
(filled circles, this study), (Mg,Fe)2SiO4 wadsleyite containing about
0.2 wt% H2O (open triangles, Cynn and Hofmeister 1994) and phase
B, Mg12Si4O19(OH)2 (filled squares, Hofmeister et al. 1999). Also
indicated are the bulk moduli for each phase (in GPa), based on values
from Reynard et al. (1996) for akimotoite, Hazen et al. (1990) for
anhydrous wadsleyite, and Crichton et al. (1999) for phase B (see text).
lus of ~151 (Crichton et al. 1999). Ross et al. (1998) found that
the bulk modulus of hydrous wadsleyite containing ~2.5 wt%
H2O, is about 10% lower than that of the dry phase (i.e., 169
GPa, Hazen et al. 1990). The specimen analyzed by Cynn and
Hofmeister (1994) contained about 0.2 wt% H2O. Thus, the
bulk modulus of this wadsleyite must be close to the value for
the anhydrous specimen, i.e., 169 GPa. Because the sample of
MgSiO3 akimotoite contained only trace amounts of water
(~350 ppm H2O), its bulk modulus should be analytically in-
distinguishable from that of the anhydrous phase, i.e., K0T ~212
GPa (Reynard et al. 1996).
The results of the present compression study also show that
the bulk modulus is not the only factor controlling the com-
pressibility of the hydrogen bonds. As shown in Figure 6, the
trends of dn/dP vs. nio have different signs for wadsleyite and
akimotoite. This is attributed to the strong pleochroism of the
OH bands in akimotoite combined with the high anisotropic
compressibility of akimotoite compared to that of wadsleyite.
Indeed, the c axis is twice as compressible as the a axis in the
akimotoite structure, whereas the a and b axes in the wadsleyite
structure are only 30% less compressible than the c axis (Hazen
et al. 1990). Therefore, the data indicate that the relative com-
pressibility of the hydrogen bonds within a structure (i.e., the
sign of dn/dP vs. nio) is governed by a combination of the di-
rectionality of the hydrogen bonds and the anisotropy of com-
pression, whereas the rate of compression (i.e., the slope of dn/
dP vs. nio) is a function of the bulk modulus.
ACKNOWLEDGMENTS
The authors thank M. Koch-Müller and an anonymous reviewer for their
helpful comments. We also appreciated the help of B. Devouard. This work was
supported by the German Science Foundation (DFG: Ke 501/2) and by the EU
TMR—Large Scale Facilities programme (contract no. ERBFMGECT980111
to D.C.R.).
REFERENCES CITED
Bell, D.R. and Rossman, G.R. (1992) Water in Earth’s mantle: the role of nominally
anhydrous minerals. Science, 255, 1391–1397.
Beran, A. (1991) Trace hydrogen in Verneuil-grown corundum and its colour variet-
ies, an IR spectroscopic study. European Journal of Mineralogy, 3, 971–975.
Bolfan-Casanova, N., Keppler, H., and Rubie, D.C. (2000) Water partitioning be-
tween nominally anhydrous minerals in the MgO-SiO2-H2O system up to 24
GPa: implications for the distribution of water in the Earth’s mantle. Earth and
Planetary Science Letters, 182, 209–221.
Crichton, W.A., Ross, N.L., and Gasparik, T. (1999) Equations of state of magne-
sium silicates anhydrous B and superhydrous B. Physics and Chemistry of Min-
erals, 26, 570–575.
Cynn, H. and Hofmeister, A.M. (1994) High-pressure spectra of lattice modes and
OH vibrations in Fe-bearing wadsleyite. Journal of Geophysical Research, 99,
17,717–17,727.
Duffy, T.S., Meade, C., Fei, Y., Mao, H.K., and Hemley R.J. (1995) High-pressure
phase transition in brucite, Mg(OH)2. American Mineralogist, 80, 222–230.
Hazen, R.M., Zhang, J., and Ko, J. (1990) Effects of Fe/Mg on the compressibility
of synthetic wadsleyite: b-(Mg1-xFex)2SiO4 (x £ 0.25). Physics and Chemistry
of Minerals, 17, 416–419.
Hirth, G. and Kohlstedt, D.L. (1996) Water in the oceanic upper mantle: implica-
tions for rheology, melt extraction and the evolution of the lithosphere. Earth
and Planetary Science Letters, 144, 93–108.
Hofmeister, A.M., Cynn, H., Burnley, P.C., and Meade, C. (1999) Vibrational spec-
tra of dense, hydrous magnesium silicates at high pressure: Importance of the
hydrogen bond angle. American Mineralogist, 84, 454–464.
Horiuchi, H., Hirano, M., Ito, E., and Matsui Y. (1982) MgSiO3 (ilmenite-type):
Single crystal X-ray diffraction study. American Mineralogist, 67, 788–793.
Kohlstedt, D.L., Keppler, H., and Rubie, D.C. (1996) Solubility of water in the a, b
and g phases of (Mg,Fe)2SiO4. Contributions to Mineral Petrology, 123, 345–
357.
Libowitzky, E. (1999) Correlation of O-H stretching frequencies and O-HºO hy-
drogen bond lengths in minerals. Monatshefte für Chemie, 130, 1047–1059.
Mao, H.K., Xu, J., and Bell, P.M. (1986) Calibration of the ruby pressure gauge to
800 kbar under quasi-hydrostatic conditions. Journal of Geophysical Research,
91, 4673–4676.
Maslen, E.N., Streltsov, V.A., Streltsova, N.R., Ishizawa, N., and Satow, Y. (1993)
Synchrotron X-ray study of the electron density in alpha-Al2O3. Acta
Crystallographica, B39, 973–980.
Nakamoto, K., Margoshes, M., and Rundle, R E. (1955) Stretching frequencies as a
function of distances in hydrogen bonds. Journal of the American Chemical
Society, 77, 6480–6486.
Nelmes, R.J., Loveday, R.M., Wilson, J.M., Besson, J.M., Pruzan, P., Klotz, G.,
Hamel, G., and Hull, S. (1993) Neutron diffraction study of the structure of
deuterated ice VIII to 10 GPa. Physical Review Letters, 71, 1192–1195.
Paterson, M.S. (1982) The determination of hydroxyl by infrared absorption in quartz,
silicate glasses and similar materials. Bulletin of Mineralogy, 105, 20–29.
Reynard, B., Fiquet, G., Itie, J.P., and Rubie, D.C. (1996) High-pressure X-ray dif-
fraction study and equation of state of MgSiO3 ilmenite. American Mineralo-
gist, 81, 45–50.
Ross, N.L., Crichton, W.A., Frost, D., and Kung, J. (1998) Equations of state of
hydrous wadsleyites determined by high-pressure single crystal X-ray diffrac-
tion. EOS, 80, F861 (abstract).
Winkler, B., Langer, K., and Johannsen, P.G. (1989) The influence of pressure on
the OH valence vibration of zoisite: an infrared spectroscopic study. Physics
and Chemistry of Minerals, 16, 668–671.
MANUSCRIPT RECEIVED APRIL 30, 2001
MANUSCRIPT ACCEPTED JANUARY 8, 2002
MANUSCRIPT HANDLED BY BJOERN WINKLER