New experimental determination of Li and B partition coefficients during upper mantle partial melting

Abstract

Despite the growing interest for Li and B as geochemical tracers, especially for material transfer from subducting slabs to overlying peridotites, little is known about the behaviour of these two elements during partial melting of mantle sources. In particular, mineral/melt partition coefficients for B and to a lesser extent Li are still a matter of debate. In this work, we re-equilibrated a synthetic basalt doped with ~10 ppm B and ~6 ppm Li with an olivine powder from a spinel lherzolite xenolith at 1 GPa–1,330°C, and we analyzed Li and B in the run products by secondary ion mass spectrometry (SIMS). In our experiment, B behaved as a highly incompatible element with mineral/melt partition coefficients of the order of 10⁻² (D ^ol/melt = 0.008 (0.004–0.013); D ^opx/melt = 0.024 (0.015–0.033); D ^cpx/melt = 0.041 (0.021–0.061)), and Li as a moderately incompatible element (D ^ol/melt = 0.427 (0.418–0.436); D ^opx/melt = 0.211 (0.167–0.256); D ^cpx/melt = 0.246 (0.229–0.264)). Our partition coefficients for Li are in good agreement with previous determinations. In the case of B, our partition coefficients are equal within error to those reported by Brenan et al. (1998) for all the mineral phases analyzed, but are lower than other coefficients from literature for some of the phases (up to 5 times for cpx). Our measurements complement the data set of Ds for modelling partial melting of the upper mantle and basalt generation, and confirm that, in this context, B is more incompatible than previously anticipated.

Appendices

Appendix 1:Run information

Experiment LiB1 was run in a non end-loaded, 3/4 inch piston-cylinder apparatus (for all technical details, see Laporte et al. 2004). We used a double container made of a graphite crucible fitted into a platinum capsule (outer diameter: 5 mm; wall thickness: 0.2 mm). The graphite container was loaded with basalt and olivine as described above, and then put into the platinum capsule and covered with a graphite lid. The platinum capsule was dried in an oven at 150°C for 40 h and then rapidly welded shut while still hot. The location of the capsule in the piston-cylinder assembly was such that the bottom of the sample chamber was located at the hot spot of the furnace and the temperature at the top of the sample chamber was equal to the temperature at the thermocouple tip (the temperature difference across the sample chamber is <10°C).

From the outside to the inside, the piston-cylinder assembly consists of a NaCl cell wrapped in lead foil, an outer pyrex cylinder, a graphite furnace, an inner pyrex cylinder, and inner pieces of crushable MgO. To reduce the amount of adsorbed water, the assembly was placed 48 h in an oven at 150°C before being loaded in the piston-cylinder apparatus. Experiment LiB1 was made at 1 GPa–1,330°C. Temperature was controlled to within ~1°C of the set-point using a calibrated W95Re5/W74Re26 thermocouple. After 66.5 h at 1,330°C, we terminated the experiment by shutting off the power to the apparatus; the quench rate was ~50°C/s. After unloading, sample LiB1 was mounted in epoxy, sectioned lengthwise, and polished with diamond suspensions (6, 3, and ¼ μm).

Appendix 2: Electron and ion microprobe analysis

Major element concentrations in ol, opx, cpx, sp, and glass were analyzed with a Cameca SX-100 electron microprobe at Laboratoire Magmas et Volcans, Clermont-Ferrand. For crystalline phases, a 15 kV accelerating voltage, a 15 nA beam current, counting times of 10 s, and a focussed beam were used. For glass analyses, the beam current was lowered to 8 nA and a beam size of 5 μm was used. Standards were as follows: albite for Si and Na; olivine for Mg; aluminium oxide for Al; hematite for Fe; manganese (II) titanium oxide, MnTiO₃, for Mn and Ti; wollastonite for Ca; orthoclase for K; and chromium (III) oxide, Cr₂O₃, for Cr.

Li and B were measured using the Cameca IMS 4f ion microprobe installed at C.N.R.-IGG, Pavia on mineral grains (cpx, opx, ol and sp) in the Pt-coated Bri 5 thin section. Samples were sputtered with a 12.5 kV ¹⁶O⁻ primary beam at ~9.5 nA intensity and 10–15 μm Ø, following methods described in Ottolini et al. (2004). We used our working curves for silicates to determine Li and B contents in ol, cpx and opx; several alumino-silicate standards containing Li and B were employed for spinel calibration. We also applied an empirical correction procedure that accounts for residual variations in the relative-to-Si ion yield for Li [IY(Li/Si)] in silicates as a function of their silica content (Ottolini et al. 1993). To reduce B contamination related to sample preparation, we cleaned the sample surface before each analysis by rastering the primary ion beam over an area with an edge of a few tens of micrometers (Ottolini and McDonough 1996). The cleaning treatment, which took ca. 25 min for a 50 × 50 μm² area was effective in removing almost all detectable residual B contamination induced by the polishing procedure. After cleaning and before starting the true analysis, the primary ion beam was left to sputter the microspot area for 360 s in order to achieve steady-state sputtering conditions. After that time, secondary ions were collected for analytical work. All SIMS spots were examined under an optical microscope and by SEM after sputtering to confirm the absence of inclusions, cracks and interstices. The same procedure was then followed for the measurement of Li an B in minerals (cpx, opx and ol) and glass in the run product LiB1. We confirm that our present detection limits for B at 6σ level of the instrumental background are on the order of 10 ppb, comparable to those in Ottolini et al. (2004).

Rare Earth and other trace elements in cpx of spinel lherzolite Bri 5 were investigated in separate analytical sessions, but at spots as close as possible to the spots selected for light element analysis. The energy filtering technique was applied to remove molecular ion interferences. In the present set-up a voltage offset of −100 V was applied to the secondary ion accelerating voltage (+4,500 V), with an energy band width of ±25 eV. Secondary ions were extracted and focused under an ion-image field of 25 μm, a contrast diaphragm of 400-μm Ø and a field aperture of 1,800 μm Ø. The entrance and exit slits of the mass spectrometer were left entirely open. Analyses were performed with a ¹⁶O⁻ primary ion beam at 9.5 nA current intensity. Clinopyroxene in Bri 5 were analyzed for REEs by measuring the signals from one isotope of each investigated REE, i.e. ¹³⁹La, ¹⁴⁰Ce, ¹⁴⁶Nd, ¹⁴⁹Sm, ¹⁵³Eu, ¹⁵⁸Gd, ¹⁶³Dy, ¹⁶⁷Er and ¹⁷⁴Yb. ³⁰Si⁺ ion signal was selected as the isotope of the reference element (Si) for these matrices. Quantification of ion signals was carried out by means of the empirical approach of relative sensitivity factors -RSFs- that were derived from international cpx standards: KH1 (Kilbourne Hole; Irving and Frey 1984) and Kakanui augite (Mason and Allen 1973) and basaltic glass JDF-D2. Concerning the other trace elements, we measured signals from the following isotopes: ⁴⁵Sc, ⁴⁷Ti, ⁵¹V, ⁵²Cr, ⁸⁸Sr, ⁸⁹Y and ⁹⁰Zr. Hf was detected through two isotopes: ¹⁷⁸Hf⁺ and ¹⁸⁰Hf⁺ to check for the presence of residual interferences. Correction of the ion current at mass 45 (amu) was done to discriminate ²⁹Si¹⁶O⁺ interference from ⁴⁵Sc⁺ signal, and at mass 52 (amu) to discriminate ²⁴Mg²⁸Si⁺ contribution from the ⁵²Cr⁺ signal. All of these trace elements including REE were detected in the same analytical run. Analytical precision from counting statistics is on the order of 10% at ppm level. Similar values pertain to accuracy.