Geochrmm a Cosmochimica Acta Vol. 56, pp. 3733-3743 Copyright 0 1992 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA Pergamon Press Ltd. Printed in U.S.A. 0016.7037/92/$5.00 + .oO The effect of oxygen fugacity on the partitioning of nickel and cobalt between olivine, silicate melt, and metal KARIN EHLERS,TIMOTHY L. GROVE, THOMASW. SISSON, STEVEN I. RECCA,and DEBORAH A. ZERVAS Department of Earth, Atmospheric, and Planetary Sciences, Massachusetts Institute of Technology, Cambridge. MA 02 139, USA zyxwvutsrqponm (Received April 10, 199 1; accepted in revised ,form May I, 1992 ) Abstract-Partitioning of nickel and cobalt between olivine, silicate melt, and metal has been determined in the system olivine-albite-anorthite-CaO at 1350°C and 1 atm over the oxygen fugacity (Jo,) range 10-5.5 to 1O-‘2.6 atm. This range of Jo,, from one log unit above Nickel-Nickel Oxide (NNO + 1) to two log units below the Iron-Quartz-Fayalite (IQF-2) buffers, spans the range relevant for crystal/liquid processes in terrestrial planets and meteorite parent bodies. There is no effect of oxygen fugacity on D$/@” over the range of fo, from 10-5.5 to 10 m9.4. Over the range of j& from 10-9.4 to 10 m’“.3, D$/@” shows an apparent decrease from a value of 6.8 at log fo, = -9.4 to a value of 5.5 at log fo, = - 10.3, a factor of 1.2. At the lowest value of fo, = 1O-12.6 the value for DNi o’m is 3 4 a factor of 2 lower than that measured under oxidizing conditions. Experimentally produced melts are metal saturated over the range of fo, from 10-‘“.3 to 10-‘2.6. Partition coefficients for Co are constant ( D~~‘g’” = 2.8) over the range of fo, from 1O-66 to 10-‘06. The Co partition coefficient decreases to a value of 1.65 at log fo, = - 12.2. The variations in Dg’,‘“‘” are examined using equilibrium constants that account for the effects of variable silicate melt composition on Ni partitioning. Measured values of Dg?*” and values estimated from equi- librium constants are identical for the experiments carried out under oxidizing conditions. The predicted values of D$\‘*” for the low fo, experiments are higher than the measured values. These changes in Ni and Co partitioning behavior at low so, may indicate that Ni and Co are present as Ni*+ and Ni” and Co2+ and Coo in the silicate melt. We estimate that approximately 30% of the Ni may be present as Ni” in the low fo, experiments. INTRODUCTION ACCURATE KNOWLEDGE OF the partitioning behavior of si- derophile elements between silicates, silicate melt, and metal is a prerequisite for developing realistic models of the chemical evolution of planetary interiors. The majority of experimental studies on partitioning of Ni and Co between olivine and silicate melt have been concerned with the effect of silicate melt composition on partitioning behavior, and there are several models that account for the effect of melt (or olivine) compositional variation on partitioning (HART and DAVIS, 1978; NIELSEN, 1985; COLSONet al., 1988; KINZLER et al., 1990). An effect of variable oxidation state on the partitioning of Ni between olivine and melt has been proposed on the basis of Ni variations observed in the olivines in terrestrial layered intrusions (MORSE, 1979) and in terrestrial lavas (MORSE et al., 199 1). Several investigations have examined the effects of variable zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA fo, on Ni olivine/silicate melt parti- tioning (MYSEN and KUSHIRO, 1976; SEIFERTet al., 1988; EHLERSand GROVE, 1990; STEELEet al., 1991). The early study of MYSEN and KUSHIRO ( 1976) showed an increase of D$‘*” as a function of decreasing fo,. At the most reducing conditions (IQF-0.5 at 1275°C) Dgl’g’s = 18 and at the most oxidizing conditions (NNO-2.4 at 1275°C) DE’,‘g’s = 4. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGF EH- LERS and GROVE ( 1990) and STEELE et al. ( 1991) show the opposite effect: a decrease in D$‘“‘” as a function ofdecreasing Jo,. The reason for the discrepancy in MYSENand KUSHIRO ( 1976) is that these authors used short experimental durations ( 12- 16 h), Pt wire loops as containers, and a beta-track method for determination of Ni concentrations. It is likely that Ni exchange between Pt and silicate melt impeded the establishment of exchange equilibrium between olivine and silicate melt in the low Jo, experiment of MYSEN and KU- SHIRO(1976). COLSON(1990, 1992) has proposed that Ni is present in a mixed valence state in silicate melts from ac- tivity variations observed in metal-silicate melt equilibria. Nickel partitioning has been used to identify primary ba- saltic magmas, both terrestrial (HART and DAVIS, 1978; EL- THON and RIDLEY, 1979; CLARKE and O’HARA, 1979) or extraterrestrial lavas (SEIFERT et al., 1988) that may be in equilibrium with a mantle source residue, as a geothermom- eter in basaltic rocks ( LEEMAN, 1974) and to model early differentiation processes of the Earth (SCHMITTet al., 1989). Several experimental studies have been performed both on natural materials and analog synthetic systems (LEEMAN, 1974; IRVINE and KUSHIRO, 1976; MYSEN and KUSHIRO, 1976; DUKE, 1976; WATSON, 1977; HART and DAVIS, 1978; LEEMAN and LINDSTROM, 1978; TAKAHASHI 1978; NABA- LEK, 1980). HART and DAVIS ( 1978) designed a series of experiments in the simple system Forsterite (Fo)-Albite (Ab)- Anorthite + CaO (“An”) and investigated systematically the temperature and melt composition effects on Dg!,‘@“. They found that D$!/*” is strongly dependent on the melt com- position, decreasing with increasing MgO. KINZLER et al. ( 1990) followed their approach and extended the HART and DAVIS( 1978) experiments to include Fe0 in the simple sys- tem. KINZLER et al. ( 1990) also developed two equilibrium constants to predict D$/*” over a wide range of compositions and temperatures. This work extends the KINZLER et al. ( 1990) experimental study to include the effects of oxygen fugacity on D$pl”. We have conducted a series of experiments in the simple system 3733