The effect of melt composition on mineral-melt partition coefficients: The case of beryllium A.D. Burnham, H. St C. O’Neill PII: DOI: Reference:
S0009-2541(16)30481-8 doi:10.1016/j.chemgeo.2016.09.012 CHEMGE 18066
To appear in:
Chemical Geology
Received date: Revised date: Accepted date:
31 May 2016 9 September 2016 13 September 2016
Please cite this article as: Burnham, A.D., O’Neill, H. St C., The effect of melt composition on mineral-melt partition coefficients: The case of beryllium, Chemical Geology (2016), doi:10.1016/j.chemgeo.2016.09.012
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT The effect of melt composition on mineral-melt partition coefficients: the case of beryllium
T
A. D. Burnham1*, H. St C. O’Neill1
RI P
1 – Research School of Earth Sciences, Australian National University, Acton ACT 2601, Australia (* = corresponding author
[email protected])
SC
Key words: melt structure; trace elements; olivine; beryllium; partitioning Abstract
MA NU
The divalent cation Be2+ is considerably smaller than other divalent cations (Mg2+, Fe2+, Ca2+, et cetera), leading to a strong preference for tetrahedral coordination in minerals. Its thermodynamic properties in silicate melts may accordingly differ from these other divalent
ED
cations, potentially distinguishing its mineral/melt partition coefficients. In order to investigate this possibility, the partitioning of Be between silicate melt and forsterite was
PT
examined for 16 melt compositions in the systems CMAS (C = CaO, M = MgO, A = Al2O3, S = SiO2) at 1400 °C with additional experiments to investigate the effect of added Na2O and
CE
TiO2, and temperature at 1300 °C. The relative activity coefficient of BeO in the melts
AC
decreases with increasing CaO and NaO1.5. The results are compared to Mg and Ca partitioning in the same experiments, and to the partitioning of other divalent cations (Ni, Co, Mn) from the literature. While the partition coefficient of the latter correlate positively with the Mg partition coefficient, Be shows only a weak negative correlation. Compared to Ca, Be partitions less strongly into forsterite when the melt has high Na and/or Ca. Partition coefficients for Na, Al and Ti are also reported.
1. Introduction Beryllium is one of the refractory lithophile elements, which are commonly assumed to have chondritic relative abundances in the Bulk Silicate Earth (e.g., Palme and O’Neill 2014),
ACCEPTED MANUSCRIPT although it should be remembered that this is an assumption (O’Neill and Palme 2008, Campbell and O’Neill 2012). Empirically, Be appears to behave geochemically like Nd,
T
another refractory lithophile element, during the petrogenesis of ocean floor basalts (OFBs),
RI P
which have Be/Nd of 0.049 ± 0.007 (n = 616; Jenner and O’Neill 2012), close to the chondritic ratio of 0.046 (Palme et al., 2014). Given that the geochemical behaviour of Be is
SC
determined by its occurrence in magmatic systems as a 2+ cation with a very small ionic radius (0.27 Å and 0.45 Å in IV-fold and VI-fold coordination respectively; Shannon, 1976),
MA NU
while Nd occurs as a 3+ cation with a large ionic radius (0.98 Å in VI-fold coordination; Shannon, 1976), the observed constancy in OFBs and the lack of fractionation relative to the chondritic ratio potentially constrains not only the OFB petrogenetic process, but also the
ED
processes that have configured the OFB source (a.k.a. the depleted MORB mantle). We note that the average Be/Nd ratio estimated for the upper continental crust is also at present too
PT
poorly constrained, at 0.078±0.032 (Rudnick and Gao, 2014), to tell whether it is different. The significance of these constraints is somewhat obscure, because so little is known about
AC
CE
the behaviour of Be in igneous systems at present.
This is understandable given the historical analytical difficulties. The natural abundance of Be is low - in CI chondrites, its abundance of 21.9 ppb is lower than that of all the other refractory lithophile elements except Ta and U (Palme et al. 2014). Moreover, Be has too low an atomic number (Z=4) to be amenable to analytical methods based on characteristic X-rays such as X-ray fluorescence or electron probe microanalysis (EPMA). However, modern analytical techniques based on mass spectrometry now allow Be to be determined precisely in typical natural concentrations (e.g., Jenner and O’Neill 2012; (Jollands et al., 2016). Such analytical methods have allowed the substitution mechanism of Be in olivine to be determined (Jollands et al., 2016), preparing the way for a means of investigating the effect of
ACCEPTED MANUSCRIPT melt composition on the thermodynamic properties of BeO in silicate melts, which will advance the use of an element with a uniquely distinctive combination of chemical properties
RI P
T
in basalt studies.
One factor determining the behaviour of Be in basalt petrogenesis is the effect of melt
SC
composition on the thermodynamic properties, which, because of the small size of the Be2+ cation, might be expected to differ from other divalent cations (e.g. Sen and Yu, 2005), and
MA NU
from that of Nd3+. The most direct way to determine the effect of melt composition on the thermodynamic properties of a component of a silicate melt would be to use an equilibrium of the component between the melt and a simple, compositionally invariant phase insoluble in
ED
the silicate melt, allowing the composition of the melt alone to be varied experimentally with all other components of the equilibrium kept constant. In the case of a handful of elements
PT
(e.g. Ni, Co, Mo, W, the platinum group metals) this may be achieved by reduction of the relevant oxide component in the silicate melt to a metal phase (e.g. O’Neill and Eggins, 2002;
CE
O’Neill et al., 2008; Borisov and Danyushevsky, 2011), but for refractory lithophile elements
AC
this metal option is at present experimentally impractical, and for such elements an alternative approach using partitioning into a compositionally invariant crystalline phase appearing on the liquidus of a range of melt compositions is the best choice available. This approach has been used, for example, by Miller et al. (2007) for studying divalent cations (Mg, Sr, Ba) with anorthite (CaAl2Si2O8) as the crystalline phase. Compared to the metalredox method, this approach is restricted, obviously, to those combinations of melt composition and temperature and pressure that have the relevant phase on the liquidus. For forsterite (Mg2SiO4) at low pressures, this restriction still allows a wide range of melt compositions to be investigated. Consequently, the effect of melt composition on several divalent cations (Mn, Fe2+, Co, Ni and Zn) have been studied by this method (e.g. Watson,
ACCEPTED MANUSCRIPT 1977; Hirschmann and Ghiorso, 1994; Kohn and Schofield, 1994; Toplis, 2005; Mysen and Dubinsky, 2004; Mysen, 2007; Matzen et al., 2013), as have trace elements with other
RI P
T
valences (for example, REE, Y, Sc, Zr, Hf and Al by Evans et al. 2008).
In this study we report forsterite/melt partition coefficients for Be and Ca at 1300 and
SC
1400 °C for a wide range of melt compositions in the system CMAS (CaO-MgO-Al2O3-SiO2) and also investigate the effects of adding Na2O and TiO2, in order to understand the
constant Be/Nd ratio of OFBs.
ED
2. Methods
MA NU
contrasting geochemistry of Be and Nd and hopefully shed light on the significance of the
Melt compositions were selected by reference to phase diagrams, where available.
PT
Compositional sets AB212, AB213 and AB214, and AB215, AB216 and AB217 are defined by addition of varying amounts of Al2O3 and TiO2 (respectively) to base composition AB211,
CE
which corresponds to composition DFAa of Evans et al. (2008). The bulk composition for
AC
each experiment was designed to produce ~ 20% forsterite crystals. Reagent grade SiO2, Al2O3, MgO (all dried at 1000 °C) and CaCO3, TiO2 and Na2CO3 (all dried at 300 °C) were weighed out in the desired proportions; Be was added as a 1000 ppm solution in HNO3 using a syringe, or as BeO for composition AB230b. These mixtures were ground under acetone in an agate mortar until dry. For Na-free mixtures, the resulting powders were mixed with polyethylene oxide solution to produce a paste that was affixed to loops of 0.3 mm diameter Pt wire and subsequently air-dried. The Na-bearing mixtures were briefly fired at 900 °C to decarbonate them, then pressed into 3.5 mm Pt capsules that were crimped and welded, leaving a minuscule outlet to prevent pressure build-up at high temperatures whilst retaining Na. The wire loops and capsules were suspended from a chandelier in the hotspot of a 1 atm
ACCEPTED MANUSCRIPT vertical furnace, allowing up to 6 melt compositions to be run simultaneously. The furnace was heated to the run temperature and the samples were held there in air for ~ 7 days before
T
drop-quenching into water (past experience indicates that olivine nucleation is rapid; the long
RI P
experimental duration allows olivine growth by Ostwald ripening); the temperature was monitored using a Type B thermocouple and is considered accurate to ± 1 °C. The beads
SC
attached to the loops were fragmented to expose a larger area before mounting in epoxy and polishing; the capsules were mounted in epoxy and sectioned with a saw blade (to double the
MA NU
area available for analysis whilst preserving spatial relationships) before polishing. A brief soak in concentrated Citranox® (Evans et al., 2008; Tuff and O’Neill, 2010) increased the
ED
contrast between olivine and glass in reflected light images.
EPMA of the matrix glasses was performed using a Cameca SX-100 electron probe with an
PT
accelerating voltage of 15 kV and a beam current of 20 nA. The standards used were: wollastonite for Si, CaAl2O4 for Ca and Al, periclase for Mg, rutile for Ti and anorthoclase
CE
for Na. To prevent Na mobilisation under the electron beam, the three Na-bearing
AC
compositions were analysed using a 20 micron diameter beam. As an additional test of this protocol, an expedited analysis (in which counting times were reduced from 20 to 10 s for Na and 20 to 15 s for Mg and Al) was performed on one Na-rich glass.
Trace element analysis of the resulting samples was carried out using a LambdaPhysik Compex 110 Eximer 193 nm laser with a HelEX ablation chamber coupled to an Agilent 7700 series ICP-MS. The carrier gas was He–Ar, fluence was maintained at ~50 mJ and pulse rate was set to 5 Hz; the spot size was 22 or 28 µm. The isotopes analysed were 9Be, 23Na, 27
Al, 29Si, 43Ca, 45Sc and 47Ti. NIST610 glass (Jochum et al., 2011) was used as the external
standard; Si (by stoichiometry for forsterite; determined by EPMA for the glasses) was used
ACCEPTED MANUSCRIPT as the internal standard. For the three Na-rich compositions, analysis locations were restricted to near the base of the capsule because of the risk of Na loss from the tops of the capsules.
T
The data were reduced using an in-house Excel spreadsheet that includes a linear drift
RI P
correction. Occasional spikes in the time-resolved data, mostly attributed to melt inclusions because of correlation between incompatible elements, were removed prior to integration of
SC
the signal.
MA NU
In calculating partition coefficients, where an element was present in the glass at > 1 % oxide, the EPMA data for the glass were used in preference to the LA-ICP-MS data.
ED
3. Results
The melts quenched to give a homogenous glass and forsterite was the only liquidus phase in
PT
these experiments, occurring as equant crystals mostly 40 – 60 µm (and occasionally up to 100 µm) in size. Melt compositions are reported in Tables 1 and 2. The two analytical
CE
protocols for Na by EPMA gave results within error of each other and so Na mobilisation is
AC
believed not to have occurred. Forsterite compositions are reported in Table 3. Apparent Sc concentrations were averaged 2.6 ppm in forsterite, which is close to the magnitude of the 29
SiO+ interference; hence the data for this element are not reported. The LA-ICP-MS data
for Ca are ~ 10 % lower relative to those obtained by EPMA for the glasses and ~ 20 % lower for the forsterite crystals. The LA-ICP-MS data for Al are in excellent agreement up to 11 % Al2O3 but are commonly lower by 6 – 9 % relative to the EPMA data for glasses with > 13 % Al2O3. Partition coefficients for Li, Be, Na, Al, Ca and Ti are reported in Table 4.
ACCEPTED MANUSCRIPT Be is highly incompatible in olivine, but is slightly more compatible than Na (Borisov et al., 2008; this work) and significantly more compatible than the REE (Beattie 1994; Evans et al.,
RI P
T
2008). DBe varies by a factor of 2 across the range of compositions studied.
4. Discussion
SC
In silicate glasses, Be occupies a mixture of distorted and undistorted tetrahedral sites (Sen and Yu, 2005). The relative proportion of Be in distorted sites is higher at lower
MA NU
concentrations, which would probably correspond to a majority of the Be at the levels employed in the present experiments. A change in coordination environment as a function of concentration could result in deviations from Henry’s Law. At 1300 °C the observed DBe for
ED
compositions AB230 and AB230b, which differ only in Be content, are within error of each other, and therefore it appears any concentration dependence of Be partitioning is
PT
insignificant over the range studied.
CE
Although substitution of Ca into forsterite affects the unit cell volume, this can be excluded
AC
as a cause of the observed variations in partitioning of Be. The increase in unit cell volume from pure forsterite to the most calcic olivine studied is equivalent to that for an olivine of composition fo88fa12 (Louisnathan and Smith, 1968; Adams and Bishop, 1985), and it has been demonstrated that the maximum solubility of Be in olivine does not change over the range fo90fa10 to fo100 (Jollands et al., 2015). Therefore, melt structure or composition is the only remaining variable that could affect partitioning in the 1400 °C series.
The structure or composition of silicate melts is commonly parameterised as NBO/T, the ratio of non-bridging oxygens (NBO) to tetrahedrally-coordinated (T) cations (Virgo et al., 1981; Mills, 1993): high NBO/T denotes a depolymerised melt, and conversely low NBO/T denotes
ACCEPTED MANUSCRIPT a mostly polymerised melt. NBO/T provides only a simple description of silicate melt structure that treats cations in a binary fashion: cations are either network formers (Si, Al, Ti,
T
most Fe3+) or network modifiers (Na, K, Ca, Mg, Fe2+ etc.) with no further refinement within
RI P
these categories. It provides a convenient way to compare physical and chemical properties of glasses and melts in related chemical systems (e.g. the oxidation state of Fe in Na2O-SiO2,
SC
K2O-SiO2, SrO-SiO2 and BaO-SiO2 binary glasses; Virgo et al., 1981). Optical basicity is another single-value melt parameter. This is based on a measurable quantity, the frequency of
MA NU
the 1S0 → 3P1 absorption band of a probe ion (e.g. Tl+, Pb2+), which depends on the degree of electron donation by the oxide anions in the melt or glass (Duffy, 1993; Mills, 1993). The amount of electron donation depends on the identities of the cations present (e.g. Na+ > Ca2+
ED
> Mg2+ > Al3+ > Si4+), and the optical basicity of a composition can be calculated from the summed contributions of each oxide in the melt. In simple systems, e.g. CMAS or alkali
PT
silicates, optical basicity correlates with activity coefficients (γ), e.g. Mo6+ (O’Neill and Eggins, 2002), W6+ (O’Neill et al., 2008) and redox ratios e.g. Ce4+/Ce3+ (Burnham et al.,
CE
2014) and Fe3+/Fe2+ (Duffy, 1993), though many studies have demonstrated only weak or
AC
non-linear correlations with activity coefficients, demonstrating that this parameter is not universally applicable (e.g. O’Neill and Berry, 2006; Borisov and McCammon, 2010). Other workers have had success modelling activity coefficients by their own summations of contributions from each oxide component (e.g. Kress and Carmichael, 1991; Jayasuriya et al., 2004). More complex parameterisations that include quadratic terms (e.g. Wade and Wood, 2013) result in a large number of potential coefficients; elimination of the statistically insignificant terms depends strongly on the spread of data and may exclude physically significant contributions that are only apparent in a small number of experiments. For example, the model of Wade and Wood (2013) indicates that γNiO is invariant in the system
ACCEPTED MANUSCRIPT CAS, in contrast to the observed range of values that span almost half the total observed
T
variability of γNiO in the system CMAS (O’Neill and Berry, 2006).
RI P
4.1 Beryllium
For the 1400 °C experiments the observed DBe have only a moderate positive correlation with
SC
NBO/T (R2 = 0.51, weighted by the uncertainties on the data; Fig. 1b). This is unsurprising as previous work has demonstrated its lack of predictive power: for example, the Al/Si ratio also
MA NU
has a significant effect on transition metal partitioning into olivine (Mysen, 2007). Moreover, γNiO shows no correlation with NBO/T for a wider range of melt compositions (O’Neill and Eggins, 2002; O’Neill and Berry, 2006). Part of this failure may be because the NBO/T value treats the effects of Mg and Ca as identical when they often show markedly different
ED
chemical properties (e.g. O’Neill et al., 2008). Another is that NBO/T is an average that does
PT
not accurately reflect speciation in the melt: in a melt with NBO/T = 1, 10% of the tetrahedral cations could have two NBOs and 10% might have none, with the other 80% having one
AC
Fig. 1b).
CE
NBO (e.g. Maekawa et al., 1991). Optical basicity is more strongly correlated (R2 = 0.81;
Considering only the 1400 °C experiments in the system CMAST, DBe is highly anticorrelated (R2 = 0.93) with XCaO, the cation fraction of Ca in the melt (Fig. 1c). The anticorrelations with optical basicity and NBO/T are stronger in this system (R2 = 0.82 and 0.61 respectively) and there is a slight positive correlation (R2 = 0.36) with XMg. Optical basicity and XCaO are highly correlated in the system CMAST because the basicity coefficient for CaO is markedly different from that of SiO2, TiO2, Al2O3 and MgO, so it is unsurprising that DBe has similar relationships with both variables. Because of the superiority of the correlation with XCaO we make the following parameterisation:
ACCEPTED MANUSCRIPT DBe = [ a XCaO + b ]×10-3
(1)
where a = -5.8(3) and b = 2.95(6). This reproduces the observed partition coefficients with a
T
mean deviation of 3%, which is within the error of the analytical technique. Equation 1 does
RI P
not accurately predict DBe when Na is present at weight percent levels in the melt, though more calcic melts still have lower DBe (Fig. 1c); however if the controlling variable is taken
SC
* instead to be X CaO (= XCaO + n·XNaO0.5, where n is a value in the range 0.5 – 1) the Na-
bearing and Na-free data fall into closer agreement. There are insufficient data to resolve the
MA NU
effect of Na on DBe fully, though it appears similar to that of Ca; given the paucity of data it may be preferable to use the relationship between DBe and optical basicity to predict
ED
partitioning in melts with high XNaO0.5.
The partitioning data at 1300 °C, although limited to two compositions, indicate a similar
PT
dependence on melt composition but offset to lower values of DBe. Spandler and O’Neill
CE
(2009) reported DBe = 1.9×10-3 for San Carlos olivine (Mg# = 90.6) in equilibrium with an Fe-bearing melt with XCaO = 0.084; the data obtained in the present study would predict DBe =
AC
1.97×10-3 when the dependence on XCaO at 1300 °C is fitted to the limited data, or 2.08×10-3 when constrained to be the same as at 1400 °C. Comparison to the data of Brenan et al. (1998) is complicated by the presence of Na in their experimental melts. Only the partitioning data for their melt with Na2O = 9.83 % and CaO = 0.04% cannot be reproduced within ~ 15% when the effect of Na is modelled in the crude way outlined above: the prediction underestimates the observed value by between 20 and 40 %, depending on the weighting of * , further highlighting the need for more extensive work in alkali-bearing systems. Na in X CaO
Other data on Be partitioning into olivine are sparse and of variable quality. Dunn and Sen (1994) reported values of 0.46 and 0.035 for natural phenocryst/matrix pairs estimated to
ACCEPTED MANUSCRIPT have formed at 1185 ± 10 °C and 1150 ± 25 °C respectively. Both values seem likely to be compromised by the bulk analysis technique used. Adam and Green (2006) reported values of
T
0.003±0.001 and 0.001 at ~ 1100 °C and 1 – 2 GPa respectively, of a similar magnitude to
RI P
our results, but the presence of 1.7 % P2O5, 2 % K2O and 6 – 11 % H2O in their melt makes detailed comparison to the present study (which was conducted at markedly different
SC
pressures and temperatures) challenging.
MA NU
Chemographic experiments in the system BeO-MgO-SiO2 (Jollands et al., 2016) show that Be sits on (or adjacent to) M-sites rather than T-sites. As a consequence, the partitioning of Be should have the same dependence on silica activity as other M-site cations (Ni, Fe, Ca
ED
etc.). The activity model of Berman (1983) can be used to calculate major element oxide activities in the system CMAS. The observed melt compositions obey the relation
PT
∂ln(aSiO2)/∂ln(aMgO) = -1.93(3), which is close to the value of -2 expected from the stoichiometry of forsterite and suggests the model produces reliable activities over the range
CE
of compositions studied. The correlation of DBe with aSiO2 (Fig. 1d) is rather weak (R2 =
AC
0.25) compared to the empirical relationship with XCaO.
4.2 Calcium
Libourel (1999) compiled and generated new data on Ca partitioning in Fe-free systems. For the system CMAST (and, unsurprisingly given its low solubility in silicate melts, for CMAS+Cr3+) the CaO content of olivine was reported to be controlled almost entirely by the CaO content of the melt, and parameterised as an exponential function. The literature data for the system CMAST are mostly (95 %) from the temperature range 1235 – 1352 °C, and the agreement of the new 1400 °C data is excellent (Fig. 2a). It was suggested that olivine
ACCEPTED MANUSCRIPT growing from Al-free melts generally had higher CaO (Libourel, 1999) but no such effect is
T
evident from the present data set.
RI P
Applying the CMAS model of Berman (1983) to the experimental melt compositions of the present study produces a good correlation between the calculated values of aCaO and the
SC
measured CaO content of olivine (Fig. 2b). (As noted by Miller, 2007, the Margules parameters for the SSAM interaction were reported incorrectly in Table XI of Berman, 1983,
MA NU
and should be WH = 652384.49 and WS = 397.38.)
The presence of Na in the melt significantly enhances partitioning of Ca into forsterite, or
ED
rather, destabilises the presence of Ca in the melt (as is particularly evident for composition AB220, for which Ca in olivine is almost twice that predicted from the relationship in
PT
CMAST; Fig. 2a). The idea of a competition between Na and Ca for similar sites in the melt is corroborated by the increase in the activity coefficient of NaO0.5 in silicate melts with XCaO
CE
(O’Neill, 2005); the lack of correlation of γNaO0.5 with XAlO1.5 suggests that the effect is not
AC
ascribable to the relative stabilities of hypothetical Na2Al2O4 versus CaAl2O4 melt species. For predicting Ca partitioning in the system NCMAST, a modified compositional parameter * (i.e. assuming a is required, as in the case of Be partitioning. Utilising the parameter X CaO
linear effect of Na) gives a best fit value of ~ 0.58 for n for the compiled literature data, * though this relationship underestimates the Ca content of olivine at low X CaO . The non-
ideality of Ca-Na mixing in silicate melts (Neuville, 2006) means that a more complex ** treatment may be preferable; a quadratic expression X CaO (= XCaO + 2.54·XCaOX NaO0.5 +
2.07(XNaO0.5)2), allowing for non-ideal Ca-Na interactions, reproduces the literature data
ACCEPTED MANUSCRIPT better (Fig. 3), though a more systematic experimental treatment of the effect of Na (and K)
T
on Ca partitioning into olivine would be desirable.
RI P
The data for the 1300 °C experiments fall below the canonical Ca partitioning field on the diagram of XCa(melt) against CaO in olivine (Fig. 2a); this is unlikely to be an analytical error
SC
because the offset is evident in both the LA-ICP-MS and EPMA data for the olivines. That these data contradict an extensive literature indicating a negligible temperature effect (with
MA NU
the exception of a study by Shejwalkar and Coogan, 2013, in which a temperature effect was discernible) may indicate that equilibrium was not fully attained. Using the approximation √(Dt) for the length scale of diffusive transport indicates that at 1400 °C, Be and Ca should
ED
travel ~ 250 and 23 µm respectively (Morioka, 1981; Jollands et al., 2016), resulting in homogenisation of the forsterite crystals for both elements, whereas at 1300 °C, these
PT
distances are 150 and 6 µm respectively, which could conceivably allow Ca to remain at
CE
concentrations slightly below the expected values after Be has fully equilibrated.
AC
4.3 Other elements
Lithium is moderately incompatible in olivine, and partitions independently of the presence of Al, in agreement with the findings of Grant and Wood (2010).
Sodium is highly incompatible in olivine. The values reported from nominally Na-free experiments are often very close to the limits of detection (variable but typically 3-5 ppm) and should be regarded as indicative only. The three Na-bearing experiments are better constrained and for compositions AB218 and AB220 the Na content of the olivine has a relative standard deviation of < 5%. All three compositions have lower DNa than the study of Borisov et al. (2008), which included Fe as a major element. Ca is more compatible in Fe-
ACCEPTED MANUSCRIPT bearing olivine (Libourel, 1999) and the same effect may occur for Na, which is of a similar ionic radius. The compatibility of Na in olivine increases significantly with pressure (Taura et
T
al., 1998; Imai et al., 2012) and thus care must be taken when selecting an appropriate value
RI P
for petrological modelling.
SC
Although Al is highly incompatible in olivine, it has attracted attention because its concentration appears to constrain the crystallisation temperature of olivine. The Al-in-
MA NU
olivine thermometer is only calibrated for spinel-saturated melts (Wan et al., 2008), in keeping with the coupled substitution MgMgSiO4 ↔ MgAlAlO4. The samples from the present study are all spinel-free (except for occasional inclusions in olivine that probably
ED
nucleated metastably and were preserved from resorption by their hosts) and therefore cannot be used to calibrate the thermometer. They do, however, provide insights into the systematics
PT
of Al substitution: Grant and Wood (2010) found Al to obey Henry’s Law, whereas Evans et al. (2008) observed Alolivine proportional to (XAl)2, which implies short-range order between
CE
the M-site and T-site Al ions. The latter relationship is observed for the CMAST
AC
compositions, though AB208 has anomalously high Al (Fig. 4); the NCMAS and the 1300 °C CMAS olivines appear to have slightly lower Al than the trend defined in CMAST, but there are insufficient data to generalise further.
The solubility of Ti in olivine has been demonstrated to depend strongly on aSiO2: at 1400 °C the maximum Ti content of olivine at high aSiO2 is less than a fifth of that at low aSiO2 (Hermann et al., 2005). The Ti-rich compositional series studied here has a relatively high aSiO2, being almost saturated in enstatite. Composition AB217 of the present study resulted in a melt with 16.4 % TiO2 coexisting with forsterite containing only 284 ppm Ti, which is only 25% of the value achievable at TiO2 saturation. Non-ideality of Ti in the melt at
ACCEPTED MANUSCRIPT these high concentrations seems unlikely to explain the discrepancy because DTi scarcely varies over the range of XTi studied. The sample mixes to which Ti was not added
T
intentionally all have higher DTi, with the three least titaniferous experiments yielding the
RI P
highest partition coefficients. Composition AB215, with 6.4 % TiO2 in the glass, is closest in Ti concentration to MORB (Jenner and O’Neill, 2012) and therefore provides the most
SC
relevant DTi.
MA NU
Both Al and Ti can be sector-zoned in experimental samples, in contrast to Ca, though the effect is minor (Pack and Palme, 2003); the volume-averaging nature of LA-ICP-MS means no such effect was evident in the present study. It is notable that Ca is found only on the M-
ED
sites of olivine, whereas Al and Ti both partition partially onto the Si site (Hermann et al., 2005; Wan et al., 2008), but the petrological significance of sector zoning remains to be
PT
established.
CE
4.4 Beryllium in other minerals of mafic rocks
AC
Beryllium partitioning into orthopyroxene was studied by van Kan Parker et al. (2010) in the temperature range 1326 – 1390 °C and varied from 0.011 to 0.014. Their 1390 °C experiment fo/melt we can predict also contained crystals of forsterite, and although they did not measure DBe
with confidence that it would have been 0.0021±0.0001, i.e. orthopyroxene accommodates 5 – 7 times as much Be as olivine. Other workers find similar or larger enrichments for orthopyroxene (Brenan et al., 1998; Adam and Green, 2006), and Be appears to be even more compatible in diopside (Monaghan et al., 1988; Brenan et al., 1998; Adam and Green, 2006). Be is only mildly incompatible in plagioclase, with reported partition coefficients ranging from 0.13 to 0.56 (Monaghan et al., 1988; Bindeman et al., 1998; Miller, 2007). In the system anorthite/melt varies by a factor of 2 with melt composition (Miller, 2007), and as with CMAS, DBe
ACCEPTED MANUSCRIPT the case of forsterite the behaviour is antithetical to that of Ca. Thus, the behaviour observed
T
in the present study may apply to a wider range of silicate melt compositions.
RI P
5. Summary
On the basis that the exchange M2+melt + Mg2+olivine = M2+olivine + Mg2+melt implies a
SC
dependence on the activity of Mg2+ in the melt, several studies have parameterised on the lt basis of DMg or used the exchange coefficient K Dolivine/me ( M Mg ) (Colson et al., 1988; Beattie et al.,
MA NU
1991). The success of this approach relies on M2+ behaving in a similar way to Mg2+. This appears to be the case for Fe2+, Ni, Co, Cr2+ and Mn both in Na-free and Na-bearing systems (Watson, 1977; Takahashi, 1978; Jones, 1984; Hanson and Jones, 1998; O’Neill and Berry,
ED
2008; Mysen, 2007; Balta et al., 2011). Zn may also fall into this category because DMg and DZn are highly positively correlated in the only systematic experimental data (Kohn and
PT
Schofield, 1994); the few available data suggest that Pd2+ may also behave somewhat
CE
similarly (Borisov and Danyushevsky, 2011). Figure 5a illustrates the positive correlations between three such elements (M = Ni, Co and Mn) and Mg. Although the data available to
AC
Beattie et al. (1991) suggested that Ca would also belong in this group of elements, the wider range of melt compositions encompassed by Libourel (1999) and the present study indicates that no correlation exists between DMg and DCa, as illustrated in Fig. 5b. Presumably the larger size of Ca and its higher coordination number of 5-7 (Petkov et al., 2000; Cormier and Neuville, 2004) is a principal factor in this difference: all the other cations seem to be primarily tetrahedrally coordinated in silicate melts (Brown et al., 1995). The negative correlation of DBe with DMg (Fig. 5b) is anomalous compared to this large group of tetrahedrally-coordinated divalent cations (Mg, Cr, Mn, Fe, Co, Ni, Zn), which is intriguing and more akin to the chemistry of Mo and W, both of which are more soluble in melts with
ACCEPTED MANUSCRIPT high optical basicity. Be, Mo and W all have high Z/r (or Z/r2, where Z is charge and r is
T
ionic radius), which may explain the similarity.
RI P
6. Acknowledgements
We are grateful to Jung-Woo Park, Mike Jollands and Pete Tollan for assistance with the LA-
SC
ICP-MS analyses, and to Bob Rapp for assistance with the electron probe. Brian Harrold and Sarah Miller are thanked for assistance with implementing the thermodynamic model of
MA NU
Berman (1983). We thank Sascha Borisov, David London and Andreas Pack for their reviews and suggestions, and Don Dingwell for his editorial handling. This work was supported by Australian Research Council Laureate Fellowship FL130100066 to HO’N.
AC
CE
PT
ED
References Adam, J. and Green, T. (2006). Trace element partitioning between mica- and amphibole-bearing garnet lherzolite and hydrous basanitic melt: 1. Experimental results and the investigation of controls on partitioning behaviour. Contributions to Mineralogy and Petrology 152, 1-17. Adams, G.E. and Bishop, F.C. (1985). An experimental investigation of thermodynamic mixing properties and unit-cell parameters of forsterite-monticellite solid solutions. American Mineralogist 70, 714-722. Balta, J.B., Asimow, P.D. and Mosenfelder, J.L. (2011). Manganese partitioning during hydrous melting of peridotite. Geochimica et Cosmochimica Acta 75, 5819-5833. Beattie, P. (1994). Systematics and energetics of trace-element partitioning between olivine and silicate melts: Implications for the nature of mineral/melt partitioning. Chemical Geology 117, 5771. Beattie, P., Ford, C. and Russell, D. (1991). Partition coefficients for olivine-melt and orthopyroxenemelt systems. Contributions to Mineralogy and Petrology 109, 212-224. Bédard, J.H. (2005). Partitioning coefficients between olivine and silicate melts. Lithos 83, 394-419. Berman, R.G. (1983). A thermodynamic model for multicomponent melts, with application to the system CaO-MgO-Al2O3-SiO2. PhD thesis, University of British Columbia, 179 pp. Biggar, G.M. (1988) Protoenstatite compositions from 1 bar to 5 kb. Chemical Geology 70, 3. Blundy, J. and Wood, B. (1994). Prediction of crystal-melt partition coefficients from elastic moduli. Nature 372, 452-454. Blundy, J. and Wood, B. (2005). Mineral-Melt Partitioning of Uranium, Thorium and Their Daughters. In: Bourdon, B., Henderson, G.M., Lundstrom, C.C. and Turner, S.P. (Eds.). Reviews in Mineralogy & Geochemistry Volume 52: Uranium-series Geochemistry. Borisov, A. and Danyushevsky, L. (2011). The effect of silica contents on Pd, Pt and Rh solubilities in silicate melts: an experimental study. European Journal of Mineralogy 23, 355-367. Borisov, A. and McCammon, C. (2010). The effect of silica on ferric/ferrous ratio in silicate melts: An experimental investigation using Mössbauer spectroscopy. American Mineralogist 95, 545555.
ACCEPTED MANUSCRIPT
AC
CE
PT
ED
MA NU
SC
RI P
T
Borisov, A., Pack, A., Kropf, A. and Palme, H. (2008). Partitioning of Na between olivine and melt: An experimental study with application to the formation of meteoritic Na2O-rich chondrule glass and refractory forsterite grains. Geochimica et Cosmochimica Acta 72, 5558-5573. Brenan, J.M., Neroda, E., Lundstrom, C.C., Shaw, H.F., Ryerson, F.J. and Phinney, D.L. (1998). Behaviour of boron, beryllium, and lithium during melting and crystallization: Constraints from mineral-melt partitioning experiments. Geochimica et Cosmochimica Acta 62, 2129-2141. Brown, G.E., Fargees, F., Calas, G., 1995. X-ray scattering and X-ray spectroscopy studies of silicate melts. Reviews in Mineralogy Volume 32, 317-410. Burnham, A.D. and Berry, A.J. (2014). The effect of oxygen fugacity, melt composition, temperature and pressure on the oxidation state of cerium in silicate melts. Chemical Geology 366, 52-60. Campbell, I.H. and O’Neill, H. StC. (2012). Evidence against a chondritic Earth. Nature 483, 553558. Chapman, A., Arendorf, J., Castella, T., Thompson, P, Willis, P., Tercero Espinoza, L., Klug, S. and Wichmann, E. (2013). Study on Critical Raw Materials at EU Level Final Report. Oakendene Hollins, Aylesbury, UK. 158 pp. Cormier, L. and Neuville, D.R. (2004). Ca and Na environments in Na2O–CaO–Al2O3–SiO2 glasses: influence of cation mixing and cation-network interactions. Chemical Geology 213, 103-113. Duffy, J.A. (1989). Optical basicity of titanium (IV) oxide and zirconium (IV) oxide. Journal of the American Ceramic Society 72, 2012-2013. Duffy, J.A. (1993). A review of optical basicity and its application to oxidic systems. Geochimica et Cosmochimica Acta 57, 3961-3970. Dunn, T. and Sen, C. (1994). Mineral/matrix partition coefficients for orthopyroxene, plagioclase, and olivine in basaltic to andesitic systems: A combined analytical and experimental study. Geochimica et Cosmochimica Acta 58, 717-733. Evans, T.M., O’Neill, H. St C. and Tuff, J. (2008). The influence of melt composition on the partitioning of REEs, Y, Sc, Zr and Al between forsterite and melt in the system CMAS. Geochimica et Cosmochimica Acta 72, 5708-5721. Grant, K.J. and Wood, B.J. (2010). Experimental study of the incorporation of Li, Sc, Al and other trace elements into olivine. Geochimica et Cosmochimica Acta 74, 2412-2428. Hanson, B. and Jones, J.H. (1998). The systematics of Cr3+ and Cr2+ partitioning between olivine and liquid in the presence of spinel. American Mineralogist 83, 669-684. Hirschmann, M.M. and Ghioso, M.S. (1994). Activities of nickel, cobalt, and manganese silicates in magmatic liquids and applications to olivine /liquid and to silicate /metal partitioning. Geochimica et Cosmochimica Acta 58, 4109-4126. Imai, T., Takahashi, E., Suzuki, T. and Hirata, T. (2012). Element partitioning between olivine and melt up to 10 GPa: Implications for the effect of pressure. Physics of the Earth and Planetary Interiors 212-213, 64-75. Jayasuriya, K.D., O’Neill, H.StC., Berry, A.J. and Campbell, S.J. (2004). A Mössbauer study of the oxidation state of Fe in silicate melts. American Mineralogist 89, 1597-1609. Jenner, F.E. and O’Neill, H.StC. (2012). Analysis of 60 elements in 616 ocean floor basaltic glasses. Geochemistry Geophysics Geosystems 13, Q02005. Jochum, K.P., Weis, U., Stoll, B., Kuzmin, D., Yang, Q., Raczek, I., Jacob, D.E., Stracke, A., Birbaum, K., Frick, D.A., Günther, D., Enzweiler, J. (2011). Determination of reference values for NIST SRM 610–617 glasses following ISO guidelines. Geostandards and Geoanalytical Research 35, 397-429. Jollands, M.C., Burnham, A.D., O’Neill, H.StC., Hermann, J., Qian, Q. (2016). Beryllium diffusion in olivine: A new tool to investigate timescales of magmatic processes. Earth and Planetary Science Letters 450, 71-82.
ACCEPTED MANUSCRIPT
AC
CE
PT
ED
MA NU
SC
RI P
T
Jones, J.H. (1984). Temperature and pressure-independent correlations of olivine/liquid partition coefficients and their application to trace element partitioning. Contributions to Mineralogy and Petrology 88, 126-132. Jurewicz, A.J.G. and Watson, E.B. (1988). Cations in olivines. 1. Calcium partitioning and calciummagnesium distribution between olivines and coexisting melts, with petrologic applications. Contributions to Mineralogy and Petrology 99, 176-185. Kohn, S.C. and Schofield, P.F. (1994). The importance of melt composition in controlling traceelement behaviour: an experimental study of Mn and Zn partitioning between forsterite and silicate melts. Chemical Geology 117, 73-87. Kress, V.C. and Carmichael, I.S.E. (1991). The compressibility of silicate liquids containing Fe2O3 and the effect of composition, temperature, oxygen fugacity and pressure on their redox states. Contributions to Mineralogy and Petrology 108, 82-92. Libourel, G. (1999). Systematics of calcium partitioning between olivine and silicate melt: implications for melt structure and calcium content of magmatic olivines. Contributions to Mineralogy and Petrology 136, 63-80. Louisnathan, S.J. and Smith, J.V. (1968). Cell dimensions of olivine. Mineralogical Magazine 36, 1123-1134. Maekawa, H., Maeakawa, T., Kawamura, K. and Yokokawa, T. (1991). The structural groups of alkali silicate glasses determined from 29Si MAS-NMR. Journal of Non-Crystalline Solids 127, 53-64. Matzen, A.K., Baker, M.B., Beckett, J.R. and Stolper, E.M. (2013). The temperature and pressure dependence of nickel partitioning between olivine and silicate melt. Journal of Petrology 54, 2521-2545. Miller, S.A. (2007). Alkaline earth element partitioning in simplified magmatic systems (PhD thesis). California Institute of Technology, Pasadena, California. Miller, S.A., Asimow, P.D. and Burnett, D.S. (2006). Determination of melt influence on divalent element partitioning between anorthite and CMAS melts. Geochimica et Cosmochimica Acta 70, 4258-4274. Mills, K.C. (1993). The influence of structure on the physico-chemical properties of slags. ISIJ International 33, 148-155. Monaghan MC, Klein J, and Measures CI (1988). The origin of 10Be in island-arc volcanic rocks. Earth and Planetary Science Letters 89, 288-298. Morioka, M. (1981). Cation diffusion in olivine - II. Ni-Mg, Mn-Mg, Mg and Ca Geochimica et Cosmochimica Acta 45, 1573-1580. Mysen, B.O. (2007). Olivine/melt transition metal partitioning, melt composition, and melt structure – influence of Al3+ for Si4+ substitution in the tetrahedral network of silicate melts. Geochimica et Cosmochimica Acta 71, 5500-5513. Mysen, B.O. and Dubinsky, E.V. (2004). Melt structural control on olivine/melt element partitioning of Ca and Mn. Geochimica et Cosmochimica Acta 68, 1617-1633. Neuville, D.R. (2006). Viscosity, structure and mixing in (Ca, Na) silicate melts. Chemical Geology 229, 28-41. O’Neill, H.StC. (2005). A method for controlling alkali-metal oxide activities in one-atmosphere experiments and its application to measuring the relative activity coefficients of NaO0.5 in silicate melts. American Mineralogist 90, 497-501. O’Neill, H.StC. and Berry, A.J. (2006). Activity coefficients at low dilution of CrO, NiO and CoO in melts in the system CaO–MgO–Al2O3–SiO2 at 1400°C: Using the thermodynamic behaviour of transition metal oxides in silicate melts to probe their structure. Chemical Geology 231, 77-89.
ACCEPTED MANUSCRIPT
AC
CE
PT
ED
MA NU
SC
RI P
T
O’Neill, H.StC, Berry, A.J. and Eggins, S.M. (2008). The solubility and oxidation state of tungsten in silicate melts: Implications for the comparative chemistry of W and Mo in planetary differentiation processes. Chemical Geology 255, 346-359. O’Neill, H.StC. and Eggins, S.M. (2002). The effect of melt composition on trace element partitioning: an experimental investigation of the activity coefficients of FeO, NiO, CoO, MoO2 and MoO3 in silicate melts. Chemical Geology 186, 151-181. O'Neill, H.StC. and Palme, H. (2008). Collisional erosion and the non-chondritic composition of the terrestrial planets. Philosophical Transactions of the Royal Society of London A 366, 4205-4238. Pack, A. and Palme, H. (2003). Partitioning of Ca and Al between forsterite and silicate melt in dynamic systems with implications for the origin of Ca, Al-rich forsterites in primitive meteorites. Meteoritics and Planetary Science 38, 1263-1281. Palme, H., Lodders, K. and Jones, A. (2014). Solar system abundances of the elements. In: Holland, H.D. and Turekian, K.K. (Eds.) Treatise on Geochemistry (Second Edition). Volume 2, pp. 15-36. Elsevier. Palme, H. and O’Neill, H.StC. (2014). Cosmochemical estimates of mantle composition. In: Holland, H.D. and Turekian, K.K. (Eds.) Treatise on Geochemistry (Second Edition). Volume 3, pp. 1-39. Elsevier. Petkov, V., Billinge, S.J.L., Shastri, S.D. and Himmel, B. (2000). Polyhedral units and network connectivity in calcium aluminosilicate glasses from high-energy X-ray diffraction. Physical Review Letters 85, 3436-3439. Prowatke, S. and Klemme, S. (2005). Effect of melt composition on the partitioning of trace elements between titanite and silicate melt. Geochimica et Cosmochimica Acta 69, 695-709. Raudsepp, M. (1995). Recent advances in the electron-probe micro-analysis of minerals for the light elements. The Canadian Mineralogist 33, 203-218. Rudnick, R.L. and Gao, S. (2014). Composition of the continental crust. In: Holland, H.D. and Turekian, K.K. (eds.) Treatise on Geochemistry (Second Edition) Volume 4, pp. 1-51. Elsevier. Sen, S. and Yu, P. (2005). Observation of a stuffed unmodified network in beryllium silicate glasses with multinuclear NMR spectroscopy. Physical Review B 72, 132203. Shannon, R.T. (1976). Revised effective ionic radii and systematic studies of interatomic distances in halides and chalcogenides. Acta Crystallographica A 32, 751-767. Shejwalkar, A. and Coogan, L.A. (2005). Experimental calibration of the roles of temperature and composition in the Ca-in-olivine geothermometer at 0.1 MPa. Lithos 177, 54-60. Soulard, H., Boivin, P. and Libourel, G. (1994). Liquid-forsterite-anorthite-spinel assemblage at 1 bar in the CMAS system: implications for low-pressure evolution of high-Al and high-Mg magmas. European Journal of Mineralogy 6, 633-646. Spandler, C. and O’Neill, H.StC. (2009). Diffusion and partition coefficients of minor and trace elements in San Carlos olivine at 1,300°C with some geochemical implications. Contributions to Mineralogy and Petrology 159, 791-818. Takahashi, E. (1978). Partitioning of Ni2+, Co2+, Fe2+, Mn2+ and Mg2+ between olivine and silicate melts: compositional dependence of partition coefficient. Geochimica et Cosmochimica Acta 42, 1829-1844. Taura H., Yurimoto H., Kurita K. and Sueno S. (1998). Pressure dependence on partition coefficients for trace elements between olivine and the coexisting melts. Physics and Chemistry of Minerals 25, 469-484. Toplis, M.J. (2005). The thermodynamics of iron and magnesium partitioning between olivine and liquid: criteria for assessing and predicting equilibrium in natural and experimental systems. Contributions to Mineralogy and Petrology 149, 22-39.
ACCEPTED MANUSCRIPT
AC
CE
PT
ED
MA NU
SC
RI P
T
Tuff, J. and O’Neill, H. St C. (2010). The effect of sulfur on the partitioning of Ni and other first-row transition elements between olivine and silicate melt. Geochimica et Cosmochimica Acta 74, 61806205. Virgo, D., Mysen, B.O. and Seifert, F. (1981). Relationship between the oxidation state of iron and the structure of silicate melts. Carnegie Institute of Washington Yearbook 80, 308-311. Wan, Z., Coogan, L.A. and Canil, D. (2008). Experimental calibration of aluminum partitioning between olivine and spinel as a geothermometer. American Mineralogist 93, 1142-1147. Watson, E.B. (1977). Partitioning of manganese between forsterite and silicate liquid. Geochimica et Cosmochimica Acta 41, 1363-1374. Watson, E.B. (1979). Calcium content of forsterite coexisting with silicate liquid in the system Na2OCaO-MgO-Al2O3-SiO2. American Mineralogist 64, 824-829. Wood, B.J. and Wade, J. (2013). Activities and volatilities of trace components in silicate melts: a novel use of metal–silicate partitioning data. Contributions to Mineralogy and Petrology 166, 911921.
ACCEPTED MANUSCRIPT Figure Captions
RI P
T
Fig. 1 DBe as a function of a) NBO/T, b) optical basicity, c) XCaO and d) aSiO2 at 1400 °C (circles) and 1300 °C (squares) in the systems CMAST (solid symbols) and NCMAS (open symbols). The line in (c) is the best fit to the 1400 °C CMAST data. Values are multiplied by 103 for clarity. Error bars are 1σ of replicate measurements.
SC
Fig. 2 a) CaO in olivine as a function of a) XCaO at 1400 °C (circles) and 1300 °C (squares) in the systems CMAST (solid symbols) and NCMAS (open symbols). Literature data are shown in grey for comparison (sources: Libourel, 1999; Shejwalkar and Coogan, 2005). b) CaO in olivine as a function of aCaO at 1400 °C.
MA NU
* (open symbols) Fig. 3 A comparison of the predicted CaO content of olivine using X CaO ** and X CaO (solid symbols) against the observed concentration. The line is the 1:1 relationship.
Fig. 4 Al in olivine as a function of XAlO1.5 at 1400 °C (circles) and 1300 °C (squares) in the
ED
systems CMAST (solid symbols) and NCMAS (open symbols). The line is the best fit to the 1400 °C CMAST data (excluding the datum for AB208 at 0.106, 407), demonstrating the dependence on (XAlO1.5)2. Error bars are 1σ of replicate measurements.
AC
CE
PT
Fig. 5 Comparison of partition coefficients to DMg in Fe-free systems for a) the transition metals Ni, Co and Mn and b) Be and Ca. Data in (a) cover the temperature range 1250 – 1550 °C and include both Na-free and Na-bearing compositions; data in (b) cover the range 1300 – 1450 °C, with Na-free systems (solid symbols) and Na-bearing systems (open symbols) plotted separately. Sources: the present study, Biggar (1988), Kohn and Schofield (1994), Mysen (2007), Shejwalkar and Coogan (2005), Soulard et al. (1994), Tuff and O’Neill (2010) and Watson (1977).
ACCEPTED MANUSCRIPT
AB205 AB206 AB207 AB208 AB209 AB210 AB211 AB212 AB213 AB214 AB215 AB216 AB217 AB224 AB237 AB218 AB219 AB220
b.d.l. b.d.l. b.d.l. b.d.l. b.d.l. b.d.l. b.d.l.
8.63(7) 18.03(16 ) 14.53(26 ) 9.43(12) 10.26(13 )
b.d.l.
4.70(6)
b.d.l.
5.06(4)
b.d.l.
0.10(1)
b.d.l.
5.49(7) 10.61(11 ) 16.05(14 )
b.d.l. b.d.l. 6.41(08 ) 11.34(2 3) 16.38(3 8)
0.07(1) 0.06(2)
b.d.l.
0.05(1) 13.88(12 )
b.d.l.
4.97(4)
-
CaO
19.22(7)
Na2O 0.19(1 ) 0.39(3 ) 0.13(1 ) 0.11(1 )
Total 100.0 5 100.2 5 100.2 8 100.0 5
0.13(0)
-
6.39(3)
-
99.47 100.7 7
11.77(3)
-
27.93(9)
-
18.98(5)
-
29.06(6) 16.78(12 )
-
21.97(3)
SC
0.04(1) 19.16(17 ) 15.88(13 )
MgO 23.48(1 3) 20.71(1 4) 20.12(0 8) 20.91(1 8) 26.51(1 6) 22.45(2 1) 20.75(1 0) 18.91(2 7) 22.77(1 6) 20.19(1 5) 23.02(2 6) 22.55(0 7) 21.16(1 4) 20.52(1 1) 23.48(1 3) 23.45(1 2) 23.53(1 2) 22.55(1 4) 21.07(1 2) 20.66(1 1) 15.54(1 3) 17.08(0 6)
14.21(3)
MA NU
AB204
b.d.l.
ED
AB203
Al2O3
PT
AB202
TiO2
CE
AB201
SiO2 54.37(3 3) 45.78(2 1) 47.80(2 3) 51.19(3 4) 54.50(4 9) 57.16(3 4) 56.88(4 8) 43.43(2 7) 54.50(1 9) 46.15(3 1) 59.21(3 7) 54.46(3 1) 51.34(4 1) 47.67(4 6) 54.94(5 0) 49.93(5 1) 47.06(5 9) 55.66(1 9) 51.17(3 0) 51.87(2 5) 61.55(2 7) 48.69(2 5)
AC
Compositi on
RI P
T
Table 1. Major element compositions (wt %) of the matrix glasses as determined by EPMA. All experiments were performed at 1400 °C except those denoted by asterisks (T = 1300 °C). NBO/T and Λ (optical basicity) were calculated including Na as determined by LA-ICP-MS, where not measured by EPMA, and using the Ti basicity coefficient of Mills (1993) rather than that of Duffy (1989). The number of analyses ranged from 7 – 9 per sample. For some samples elements were not analysed (-) or were below the limit of detection (b.d.l.; ~ 0.02 %). Uncertainties are 1σ of replicate measurements.
-
17.72(9) 15.36(10 )
-
17.92(8)
16.34(4)
17.46(5) 16.41(4) 15.44(7) 15.78(4) 14.77(4) 13.66(2) 6.76(5) 21.79(5) 7.21(3) 1.06(4) 8.34(1)
99.02 100.5 5 100.9 9 100.4 6
0.33(3 ) 0.36(3 ) 0.34(2 ) 0.14(2 ) 0.78(4 ) 0.07(1 )
99.47 100.2 9
0.39(1 ) 2.31(4 ) 5.73(9 ) 7.04(1 0)
98.96
99.88 100.0 2 100.8 2 100.3 4 100.7 5
99.39 99.77 99.25 99.07
NBO/ T 2.16( 1) 1.03( 1) 1.15( 1) 1.52( 1) 0.77( 1) 0.86( 1) 1.12( 1) 1.88( 2) 1.72( 1) 2.24( 2) 1.77( 2) 1.62( 1) 1.35( 1) 1.14( 1) 1.74( 1) 1.76( 2) 1.67( 2) 0.91( 1) 1.83( 1) 0.83( 1) 0.52( 1) 0.88( 1)
Λ 0.6165( 6) 0.6050( 4) 0.6058( 4) 0.6096( 6) 0.5682( 6) 0.5733( 5) 0.5832( 5) 0.6407( 6) 0.6070( 4) 0.6419( 7) 0.5962( 7) 0.6038( 6) 0.6035( 6) 0.6050( 8) 0.6003( 6) 0.6073( 8) 0.6069( 9) 0.5754( 3) 0.6176( 6) 0.5853( 4) 0.5606( 4) 0.6013( 5)
ACCEPTED MANUSCRIPT
b.d.l. b.d.l.
14.50(1 1) 14.95(0 9) 15.55(1 0) 16.32(0 4)
24.65(5)
b.d.l.
99.49
23.89(9)
b.d.l. 0.12(1 ) 0.11(1 )
97.96
16.07(6) 15.31(5)
MA NU ED PT CE
99.97
T
b.d.l.
14.24(16 ) 13.78(22 ) 15.37(13 ) 14.51(20 )
97.52
RI P
b.d.l.
SC
46.07(1 9) 45.30(2 5) 52.85(4 9) 51.28(2 4)
AC
AB230 * AB230 b* AB231 * AB231 b*
1.26( 1) 1.29( 1) 0.89( 1) 0.94( 1)
0.6207( 5) 0.6209( 6) 0.5911( 7) 0.5922( 4)
ACCEPTED MANUSCRIPT Table 2. Glass compositions (weight % except where indicated) as determined by LA-ICPMS. n = 6 for all compositions except AB201, AB218 and AB220 (n=8), AB235 (n=7) and AB230b (n=4). Uncertainties are 1σ of replicate measurements.
PT
T
CaO 20.17(33) 11.95(16) 14.52(08) 17.53(09) 0.113(3) 5.82(05) 11.30(04) 26.83(11) 17.40(07) 26.91(25) 16.95(06) 16.08(13) 14.87(13) 13.62(07) 15.63(16) 13.98(16) 13.03(06) 6.05(03) 19.58(21) 5.92(07) 0.87(02) 7.04(07) 20.84(10) 20.22(17) 13.21(14) 12.55(08)
SC
RI P
Al2O3 0.0349(7) 17.59(19) 15.85(6) 8.67(5) 17.41(45) 13.86(13) 9.99(8) 10.15(7) 4.57(4) 4.92(9) 0.0949(5) 5.62(5) 10.67(7) 15.66(13) 0.0776(8) 0.0579(4) 0.0570(2) 13.51(8) 4.70(6) 16.54(22) 14.73(25) 16.98(22) 13.17(11) 12.75(8) 14.00(12) 13.31(11)
MA NU
Na2O 0.216(4) 0.431(5) 0.139(1) 0.110(0) 0.293(4) 0.238(3) 0.204(1) 0.022(0) 0.041(0) 0.009(0) 0.349(1) 0.345(3) 0.384(3) 0.355(2) 0.046(1) 0.843(8) 0.049(0) 0.105(1) 0.411(3) 2.425(3) 5.962(8) 7.226(3) 0.037(0) 0.032(0) 0.130(2) 0.085(24)
ED
Be (ppm) 1380(34) 1069(9) 1280(5) 1286(9) 1211(28) 1362(16) 1482(10) 1352(11) 1304(12) 1267(14) 1724(4) 1598(15) 1240(15) 1309(15) 1560(13) 1001(13) 1214(8) 1249(10) 1280(17) 1154(11) 1341(13) 1431(11) 2963(33) 8867(61) 2694(44) 8100(140)
CE
Li (ppm) 2.71(11) 3.09(8) 4.18(8) 3.93(10) 9.68(17) 7.88(22) 6.23(11) 1.56(3) 2.16(10) 0.93(6) 3.54(12) 3.35(11) 3.18(4) 3.03(7) 2.49(5) 2.32(3) 2.48(5) 3.91(12) -
AC
Composition AB201 AB202 AB203 AB204 AB205 AB206 AB207 AB208 AB209 AB210 AB211 AB212 AB213 AB214 AB215 AB216 AB217 AB224 AB237 AB218 AB219 AB220 AB230* AB230b* AB231* AB231b*
Ti (ppm) 6.9(1.4) 10.1(7) 5.5(4) 52.2(1.1) 51.0(7) 54.9(1.1) 57.2(2) 46.5(4) 133.4(1.4) 52.4(1.1) 55.2(1.5) 52.8(5) 61.2(1.1) 47.2(1.0) 32800(300) 56500(600) 81800(300) 62.1(9) 49.9(9) 36.6(1.1) 36.2(1.2) 51.1(2.0) 50.1(1.8)
ACCEPTED MANUSCRIPT Table 3. Forsterite compositions (ppm except where indicated) as determined by LA-ICPMS. Li and Ti were not analysed in all samples. Uncertainties are 1σ of replicate measurements.
CE
T
CaO (%) Ti 0.698(19) 0.9(3) 0.318(4) 4.3(9) 0.394(6) 1.0(1) 0.537(6) 1.1(2) < 0.006 1.3(1) 0.074(3) 1.2(1) 0.210(6) 1.2(2) 1.887(33) 1.5(2) 0.500(9) 1.9(2) 1.829(26) 1.7(3) 0.444(12) 1.1(5) 0.436(6) 0.9(1) 0.401(9) 1.0(1) 0.356(13) 1.1(2) 0.367(4) 116(2) 0.331(8) 208(4) 0.270(5) 284(15) 0.083(3) 1.6(2) 0.672(12) 0.9(1) 0.090(2) 0.016(3) 0.221(5) 0.771(19) 1.1(1) 0.690(12) 1.0(2) 0.239(10) 1.0(3) 0.199(5) <4
RI P
Al 0.7(2) 721(19) 426(18) 134(3) 486(22) 274(12) 146(13) 407(18) 61(3) 116(8) 1.6(5) 74(3) 189(9) 438(20) 1.2(1) 1.1(1) 1.0(1) 270(11) 66(2) 381(9) 223(9) 428(11) 242(9) 279(7) 166(6) 180(7)
SC
Na 7.0(1.7) 7.5(1.3) 4.5(1.1) 4.0(8) 6(3) 4.0(9) 5(3) <3 3.1(1) <3 6.1(1.4) 7.9(1.6) 8.2(1.4) 8.3(2.2) 2.0(1) 12(1) 3.1(7) <3 11.8(5) 26(1) 50(11) 85(4) <3 <3 <3 < 39
MA NU
Be 2.75(26) 2.36(13) 2.63(9) 2.49(10) 3.55(21) 3.56(24) 3.27(27) 1.80(9) 2.45(11) 1.87(10) 3.40(14) 3.19(12) 2.59(17) 2.84(14) 3.16(8) 2.12(7) 2.85(12) 3.23(14) 2.21(3) 2.60(6) 3.66(12) 2.57(12) 3.70(24) 10.86(55) 4.57(19) 12.72(62)
ED
10 12 11 13 10 13 10 11 13 14 10 12 11 10 14 13 14 14 11 10 11 14 11 9 9 8
Li 0.40(4) 0.41(4) 0.64(5) 0.75(5) 1.14(6) 1.08(4) 1.02(5) 0.18(2) 0.37(4) 0.21(5) 0.39(6) 0.58(5) 0.54(7) 0.44(3) 0.31(4) 0.23(5) 0.32(6) 0.55(6) -
PT
n
AC
Composition AB201 AB202 AB203 AB204 AB205 AB206 AB207 AB208 AB209 AB210 AB211 AB212 AB213 AB214 AB215 AB216 AB217 AB224 AB237 AB218 AB219 AB220 AB230* AB230b* AB231* AB231b*
ACCEPTED MANUSCRIPT Table 4. Forsterite/melt partition coefficients for Li, Be, Na, Al, Ca and Ti. Uncertainties are 1σ of replicate measurements.
PT
T
DCa 0.0318(9) 0.0224(3) 0.0241(4) 0.0279(3) <0.045 0.0116(5) 0.0179(5) 0.0676(12) 0.0264(5) 0.0629(9) 0.0265(7) 0.0250(4) 0.0244(6) 0.0230(9) 0.0233(2) 0.0224(5) 0.0198(4) 0.0122(4) 0.0309(6) 0.0124(3) 0.0152(24) 0.0265(6) 0.0313(8) 0.0289(5) 0.0148(6) 0.0130(4)
SC
RI P
DAl 0.0037(12) 0.0071(2) 0.0051(2) 0.0029(1) 0.0051(2) 0.0036(2) 0.0029(3) 0.0075(3) 0.0024(1) 0.0043(3) 0.0032(10) 0.0025(1) 0.0034(2) 0.0052(2) 0.0029(2) 0.0036(3) 0.0034(3) 0.0037(2) 0.0025(1) 0.0041(1) 0.0027(1) 0.0045(1) 0.0032(1) 0.0038(1) 0.0020(1) 0.0023(1)
MA NU
DNa 0.004(1) 0.0023(4) 0.004(1) 0.0050(1) 0.003(1) 0.0022(5) 0.003(2) < 0.02 0.0100(4) < 0.04 0.0023(6) 0.0031(6) 0.0029(5) 0.0032(8) 0.0059(3) 0.0019(2) 0.009(2) < 0.0048 0.0039(2) 0.0014(1) 0.0011(3) 0.0016(1) < 0.01 < 0.01 < 0.003 < 0.06
ED
DBe 0.0020(2) 0.0022(1) 0.0021(1) 0.0019(1) 0.0029(2) 0.0026(2) 0.0022(2) 0.0013(1) 0.0019(1) 0.0015(1) 0.0020(1) 0.0020(1) 0.0021(1) 0.0022(1) 0.0020(1) 0.0021(1) 0.0023(1) 0.0026(1) 0.0017(0) 0.0023(1) 0.0027(1) 0.0018(1) 0.0012(1) 0.0012(1) 0.0017(1) 0.0016(1)
CE
DLi 0.15(2) 0.13(1) 0.15(1) 0.19(1) 0.12(1) 0.14(1) 0.16(1) 0.12(2) 0.17(2) 0.23(5) 0.11(2) 0.17(2) 0.17(2) 0.15(1) 0.12(2) 0.10(2) 0.13(3) 0.14(2) -
AC
Composition AB201 AB202 AB203 AB204 AB205 AB206 AB207 AB208 AB209 AB210 AB211 AB212 AB213 AB214 AB215 AB216 AB217 AB224 AB237 AB218 AB219 AB220 AB230* AB230b* AB231* AB231b*
DTi 0.13(5) 0.43(9) 0.18(3) 0.020(3) 0.025(3) 0.022(2) 0.021(3) 0.033(5) 0.015(1) 0.033(5) 0.02(1) 0.017(2) 0.017(2) 0.024(4) 0.0030(1) 0.0031(1) 0.0029(2) 0.025(4) 0.018(3) 0.031(4) 0.026(6) 0.019(6) < 0.08
AC
CE
PT
ED
MA NU
SC
RI P
T
ACCEPTED MANUSCRIPT
Figure 1
AC
Figure 2
CE
PT
ED
MA NU
SC
RI P
T
ACCEPTED MANUSCRIPT
AC
Figure 3
CE
PT
ED
MA NU
SC
RI P
T
ACCEPTED MANUSCRIPT
ED
MA NU
SC
RI P
T
ACCEPTED MANUSCRIPT
AC
CE
PT
Figure 4
AC
CE
PT
ED
MA NU
SC
RI P
T
ACCEPTED MANUSCRIPT
Figure 5
ACCEPTED MANUSCRIPT
AC
CE
PT
ED
MA NU
SC
RI P
T
Highlights: Be is incompatible in olivine; DBe varies by a factor of 2 with melt composition beryllium exhibits unique geochemical behaviour among the divalent cations activity in melt may be controlled by coordination number