In Search of Fluid - Melt Partitioning Values; Development of the Laser Ablation ICPM-S Method

Geoffrey T. Nichols, Trevor H. Green, Norman J. Pearson

GEMOC, School of Earth Sciences, Macquarie University, NSW 2109

INTRODUCTION

Many geochemical constraints recently determined either from experimental studies or natural samples, indicate that the genesis of subduction-zone magmas, and particularly those of calc-alkaline compositions, is in some way related to the transport and interaction of source mantle with hydrous fluids. Such water-rich fluids, probably produced by the dehydration of minerals in the subducted slab, are identifiable because they impart specific trace-element signatures that are not produced by other transporting agents such as silicate melts. In some subduction zones for example, sediment-derived silicate melts contribute characteristic trace-element signals and their input is either inferred from experimental constraints (Nichols et al., 1994) or is evidenced by distinctive isotopic signatures (Kay et al., 1978).

The geochemical attributes of hydrous fluids or vapour are poorly known, and it is therefore important to determine the partitioning behaviour of trace-elements between vapour and mantle melts, as well as between vapour and mantle minerals. In all but the most recent published work, mineral-melt partition coefficients and mineral-vapour values have been determined independently, and then the vapour-melt partition values have been calculated. In these experiments we have developed a new analytical technique that permits us to measure the trace element composition of quenched vapour, as well as quenched melt.

EXPERIMENTAL METHODS

Experiments were performed with a natural starting material of basanite+H2O, at temperatures between 825-1220°C, and at pressures of 20-30 kbar; most data however, were obtained from experiments at 1200°C and 25 kbar. The basanite from Mt Leura, Queensland, was enriched in the trace elements Rb, Y, Nb, Cs, La, Sm, Lu, Hf, Ta, Th, U which total 1.045 wt%. Experiments were conducted using an end-loaded 12.7 mm piston-cylinder apparatus with either Ag50Pd50 or Ag70Pd30 capsules.

ANALYTICAL TECHNIQUES AND EQUIPMENT

The first step in the analytical process involved the release and immediate analysis of the vapour phase using the Laser Ablation Microprobe (LAM) attached to an Inductively Coupled Plasma - Mass Spectrometer ICP-MS. A detailed description of the LAM ICP-MS instrumentation is given in Norman et al. (in press). Briefly, the laser is a Q-switched, frequency quadrupled Nd:YAG laser, operating at 266 nm (UV). All analyses were performed using a repetition rate of 4 Hz and energy of ~1 mJ/pulse. These conditions produced a sampling area ~30 µm in diameter. The sample is housed in a sealed cell and the ablated material is transported in a stream of high purity Ar to the ICP-MS. The sample cell is mounted on a petrographic microscope fitted with CCD-TV, enabling viewing of the ablation process. Data acquisition was monitored in a real-time graphics display. The laser was used to drill through the Ag-Pd capsule to release the vapour ± liquid, at which time time the laser was turned off. The vapour signals were transient, typically ² 40 seconds duration. The following trace elements were analysed Be, Ti, Ni, Rb, Sr, Y, Zr, Nb, Ba, La, Ce, 147Sm, 152Sm, Ho, Yb, Lu, Hf, Ta, Pb, Th, U with dwell times of 20 ms, using peak hopping mode and with one sweep across the mass-range per reading. After vapour analysis the metal capsules were sectioned and the experimental products exposed by polishing. Major element analyses of primary minerals, spherules and glass-matrix were determined using a Cameca SX50 electron microprobe.

The trace element compositions of sufficiently large minerals, spherules, and glass matrix were analysed by LAM ICP-MS under the same conditions described for the vapour analysis, except that a longer dwell time of 50 ms was employed. In these determinations the signals were essentially steady-state, displaying nearly constant counts per second for analysis times between 120 to 185 seconds.

In all cases raw counts were quantified using the NIST610 glass as an external calibration standard and 44Ca as an internal standard using the electron microprobe data for CaO. To successfully compare signals for the vapour + H2O signal it was necessary to select a normalisation factor, as we were not able to provide an absolute concentration for any particular element. In this case we selected a somewhat arbitrary value of Sr=100 ppm, which was based on the concentration of Sr in spherules analysed using the proton-microprobe. Selecting other normalisation values, or other normalising elements, does not change the relative shape of the trace-element patterns, but translates the signal either up or down such that the "absolute" concentrations vary.

The accuracy and precision of the quantitative analysis by LAM ICP-MS is discussed in Norman et al. (in press). In this work as an example, we have propagated uncertainties on an individual analysis of a spherule (analysis sph29, run 1627), and as a relative % errors range from ~14 to ~20% (e.g., Lu, 1150 ± 170 ppm; Pb, 2.54 ± 0.50 ppm). These values include ICP-MS counting statistics as well as the error on the internal standard (CaO) determined by electron microprobe. In this example, the standard deviation on CaO (5.09 ± 0.71 wt%; 13.9% rsd) is the main contribution to the propagated uncertainty on the trace element concentrations. The magnitude of the standard deviation may reflect true compositional variation but may also in part be due to variable Na loss under the electron beam depending on the size and composition of individual spherules.

EXPERIMENTAL PRODUCTS

Depending on run temperature experiments produced some combination of a quenched-matrix and primary crystals of clinopyroxene, olivine, ± amphibole, in addition to vapour components. The components that represent the vapour, after experiments are quenched, are a complex mixture that include spheroidal shaped solids that are usually ~20 µm in diameter (between 5 and 50 µm). We have called these "spherules" and other workers (e.g. Brenan et al., 1995) have observed similar quench products in crystal-vapour experiments and termed them "fish-roe". The vapour component also includes a water-rich solvent, gas vapour (probably a C-O-N mixture), and acicular micron-sized amphiboles or pyroxenes. Unfortunately, these vapour components are the modified products formed during the quenching processes, and so are not the original single vapour phase that was in equilibrium with the basanite during the experiment. The unmixing of experimental vapour ensures that the volumes or proportions of components are poorly constrained, and this provides one of the major complexities in the complete analysis of the equilibrium vapour (Veq) phase.

The complexity of the vapour system can be considered using a mathematical representation of the interactions that occur both, during, and after an experiment. First, during an experiment, the total vapour (V) can be expressed as

V = Veq + Vdis

where Veq is the equilibrium vapour phase, and Vdis represents the vapour component dissolved in the melt. During quench Veq unmixes and Vdis exsolves from the melt. The equilibrium vapour, Veq, unmixes into a number of components and may be written as,

Veq = Sph + H2O + gas + Xtal

where Sph are spherules formed from quenched Veq, gas is a C-O-N vapour, and Xtal represents acicular micro-scale amphibole or pyroxene. The exsolution of vapour dissolved in the melt, Vdis, during quench, may be expressed as

Vdis Å Vex + SphD

where Vex is the exsolved vapour component, and SphD are spherules formed from vapour that exsolves from the melt. Thus the total vapour in the system after an experiment is

V = (Sph + H2O + gas + Xtal) + (Vex + SphD)

DISCUSSION

DECIPHERING THE VAPOUR COMPONENTS

Our earlier experiments utilized the laser as a piercing-tool, and the ICP-MS as an accurate analytical device. The results indicated that the vapour removed from the capsules did not change composition significantly across the P-T range of the experiments. This result was unexpected given that some runs were either below or on the basanite-solidus, whereas other experiments were superliquidus. At that time we believed that because subsolidus runs had vapour signals similar to those of superliquidus experiments, we must be measuring a vapour signal dominated by the equilibrium vapour. This vapour therefore, was largely unaffected by the quenching process that could contribute to the measured signal as spherules formed, and as vapour exsolved from melt. Although we did not detect any compositional variation across the temperature range of our runs, we were able to observe a physical change in the quenched vapour. Subsolidus runs have rare spherules and acicular amphiboles, whereas superliquidus experiments have more spherules dispersed throughout the quenched melt. In one experiment, a sphere of melt was encapsulated by a 2 µm thick feldspathic rim, very close in composition to that of the spherules. Spherules occur adjacent to the edge of the 275 µm diameter ball of melt, but are predominantly clustered next to three radial cracks in the quenched melt-sphere. This evidence suggests that some spherules form directly from the vapour that exsolves from the melt. Although some spherules must also form entirely from vapour at subsolidus conditions, we have not measured any distinct compositional groups of spherules that may correspond to formation either directly from the vapour, or exsolved from the melt.

In order to understand the quenching processes further, and ultimately to constrain the equilibrium vapour composition, we made two modifications to our experiments. First, we added an oxide mix to the capsule as a layer, designed to produce Mg-phlogopite, and this was separated from the basanite+H2O by a layer of diamond aggregate. These runs were partly successful. They crystallized phlogopite as well as pyrope garnet, but the low viscosity of the basanite allowed it to pass through the permeable diamond aggregate and to mix with the phlogopite "trap". In these runs we were able to analyse the matrix and coarse phlogopites with the LAM ICP-MS and thereby determine basanite-phlogopite Kds.

In other runs we added vitreous carbon spheres. They behave much like diamond aggregate, providing porosity between adjacent carbon-spheres, and like diamond are relatively unreactive. The vitreous carbon also has an advantage over diamond aggregate in that it absorbs vapour, and is easy to section or polish. In our initial runs, the analysed carbon-spheres show trace-element patterns almost identical to those of the spherules we analysed in separate runs using identical LAM ICP-MS methods. Because the trace-element patterns of spherules and carbon-spheres are identical (Fig. 1), no matter which region of the carbon-spheres we analysed, we infer that this signal must represent that close to the equilibrium vapour, since it could not have formed and penetrated the carbon-spheres during the available milliseconds of quench. We conclude that although some spherules must form during the quench process (based on textural evidence cited above), others must represent quenched solute components from primary equilibrium vapour.

Figure 1. Displays trace element patterns of spherules (Sph) compared with those measured in carbon-spheres (C sphere), determined by LAM ICP-MS, but from different experiments run under similar conditions. The absolute trace element abundances are not well constrained for the carbon-spheres, because there is no reliable internal standard. Also shown is the trace element signal for spherules measured with the proton-microprobe (p-probe). The spherule data, measured with either the proton-probe or LAM ICP-MS, compares favourably with the data from the carbon-spheres strongly suggesting that the carbon-spheres and the spherules record the equilibrium vapour composition.

These patterns are dominated by Pb and Ni depletions and relative enrichments in Ti, and Rb. Nb is weakly preferred over Ta, and Hf over Zr. In contrast, the signals we have measured from the vapour component (which equals quenched gas + H2O + Vex), display almost mirror image trace-element signals in comparison with the spherule or carbon-sphere patterns (Fig. 2). The vapour signals are strongly enriched in Pb and Rb. Nb is variably enriched over Ta, and Zr over Hf. La, Sm, Ti, Y and Lu display variable depletions, and in some experiments were below analytical detection. Figure 2 compares vapour signals with those determined by Keppler (1996). Data from our experiments are broadly similar, showing both Pb and Rb enrichments however, our data have larger Pb/Rb than that of Keppler.

Figure 2. Displays vapour signals acquired by capsule-piercing, and the key lists experimental pressure in kbar followed by run temperature. Data for one experiment that had an additional Mg-phlogopite mix (phlog mix), are also displayed. Vapour signals are normalized so that Sr=100 and then divided by the known starting composition. These data are compared with Keppler's (1996) data, which are plotted as signal only. Although the absolute values of the two signals vary, they have broadly similar profiles.

SUMMARY

Our results indicate the following: (1) that the vapour component in equilibrium with melt does not quench as a single phase, but produces at least three discrete phases that include spherules, acicular micron-sized amphibole or pyroxene, and vapour+H2O, (2) that spherules are produced in subsolidus as well as liquidus and superliquidus experiments, (3) that the composition of spherules and vitreous carbon spheres is essentially identical, suggesting that both spherules and the signal from carbon-spheres may closely represent the solute composition of equilibrium vapour.

REFERENCES

Adam, J., Green, T.H., Sie, S.H., Ryan, C.G., Subm. Trace element partitioning between aqueous fluids, silicate melts and minerals. European Journal of Mineralogy.

Brenan, J.M., Shaw, H.F., Ryerson, F.J., Phinney, D.L., 1995. Mineral-aqueous fluid partitioning of trace elements at 900°C and 2.0 GPa: Constraints on the trace element chemistry of mantle and deep crustal fluids. Geochimica et Cosmochimica Acta, 59, 3331-3350.

Kay, R.W., Sun, S.-S., Lee-Hu, C.-N., 1978. Pb and Sr isotopes in volcanic rocks from the Aleutian Islands and Pribilof Islands, Alaska. Geochimica et Cosmochimica Acta, 42, 263-273.

Keppler, H., 1996. Constraints from partitioning experiments on the composition of subduction-zone fluids. Nature, 380, 237-240.

Nichols, G.T., Green, T.H., Pearson, N., Sharma, A., 1996. A new analytical technique for measuring element partitioning between experimental vapour and melt. Geological Society of Australia, Abstracts 41, 317.

Nichols, G.T., Wyllie, P.J., Stern, C.R., 1994. Subduction zone melting of pelagic sediments constrained by melting experiments. Nature, 371, 785-788.

Norman, M.D., Pearson, N.J., Sharma, A., Griffin, W.L., 1996. Quantitative analysis of trace elements in geological materials by laser ablation ICPMS: instrumental operating conditions and calibration of NIST glasses. Geostandards Newletter (in press).

Acknowledgements: We thank A. Sharma for her assistance in operating the Macquarie University LAM ICP-MS system; D. Draper for beneficial comments on this abstract.

Back to the GEMOC Abstract Titles Page