RECEIVED MARCH 31, 1997; REVISED TYPESCRIPT ACCEPTED DECEMBER 2, 1997

Oceanic island basalts have a large range in [delta]¹⁸O values (4·5-7·5%^o) compared with the assumed primordial mantle values (5·5-6·0%^o) and with mid-ocean ridge basalts (5·7 ± 0·2%^o). Some Hawaiian tholeiitic basalts have low ¹⁸O values (4·6-5·2%^o), which have been interpreted to be either a primary source feature or caused by crustal contamination. This study was undertaken to evaluate the cause of low [delta]¹⁸O values in Hawaiian tholeiitic basalts. We determined the [delta]¹⁸O values of glassy matrix material and coexisting olivines from pristine basalts produced during the current, 14-year-old Puu Oo eruption of Kilauea Volcano. Our results show that the Puu Oo eruption lavas have significant ranges in matrix (0·7%^o) and olivine [delta]¹⁸O values (0·5%^o) which do not correlate consistently with other geochemical parameters and that many of the lavas are out of oxygen isotopic equilibrium. These features probably reflect partial assimilation of and oxygen exchange with metamophosed Kilauea rocks during the magma's 19 km transit through the volcano's east rift zone. The parental magmas for Puu Oo lavas had a [delta]¹⁸O value of at least 5·2%^o and perhaps as high as 5·6%^o. Thus, Puu Oo lavas do not give a clear indication of the [delta]¹⁸O value of Kilauea's mantle source but they do indicate that the oxygen in these otherwise pristine basalts has undergone significant modification by interaction with crustal rocks.

Keywords: Kilauea Volcano; Hawaii; oxygen isotopes; Puu Oo eruption; crustal contamination

INTRODUCTION

A 1996 conference on `shallow level processes in ocean island magmatism' drew attention to the controversy that is developing on the possible extent of crustal contamination for oceanic island basalts (Bohrson et al., 1997). This controversy is important because most studies of oceanic island basalts have assumed that these lavas provide a direct indication of the geochemical and isotopic composition of mantle (e.g. Zindler & Hart, 1986). This assumption has been questioned in several recent studies (e.g. Eiler et al., 1996a; Thirlwall et al., 1997) and it has been suggested that even the most primitive basalts from some oceanic islands were contaminated by crustal materials (e.g. Hemond et al., 1993). Resolution of this controversy will have fundamental implications for our understanding of the magmatic history of oceanic island basalts and for the composition of the mantle. Oxygen isotopes are a potentially powerful tool to help resolve this debate because [delta]¹⁸O values of magma can be substantially modified by assimilation of rocks that have interacted with the hydrosphere (e.g. Taylor, 1974).

Many oceanic island basalts have [delta]¹⁸O[dagger] values somewhat below the assumed mantle range, based on studies of mantle xenoliths and lunar rocks (4·6-5·3%^o vs 5·5-6·0%^o; e.g. Taylor & Sheppard, 1986; Mattey et al., 1994; Harmon & Hoefs, 1995) and mid-ocean ridge basalts (MORB; 5·7 ± 0·2%^o; Ito et al., 1987). Many of these oceanic island basalts are mafic (>6·5 wt % MgO) and display no obvious signs of crustal contamination (Garcia et al., 1989b). Thus, it has been argued that the mantle is heterogeneous in [delta]¹⁸O (Harmon & Hoefs, 1995), which would make recognition of crustal contamination more difficult. Correlations have been noted between low [delta]¹⁸O values of olivine and whole-rock ratios of ²⁰⁶Pb/²⁰⁴Pb in Hawaiian basalts, which have been interpreted to reflect assimilation of hydrothermally altered oceanic crust (Eiler et al., 1996a). The same correlation for clinopyroxene and whole rocks with low [delta]¹⁸O values from the Canary Islands has been related to either recycling of ancient altered oceanic crust or crustal contamination (Thirlwall et al., 1997). Therefore, the role of crustal contamination in oceanic island magma petrogenesis is equivocal based on these previous studies.

Hawaiian tholeiitic glasses have yielded a considerable range of [delta]¹⁸O values (4·6-5·8%^o; Kyser et al., 1982; Garcia et al., 1989b, 1993), which overlaps with the range for MORB. Some of the low [delta]¹⁸O values are for submarine Hawaiian tholeiites that apparently assimilated hydrothermally altered crust because these rocks have elevated [delta]D values (-33 to -52%^o vs mantle values of -60%^o to -96%^o; Kyser & O'Neil, 1984), boron concentrations and [delta]¹¹B values (Chaussidon & Jambon, 1994). These lavas also have obvious petrographic signs of assimilation and complicated magmatic histories (Clague et al., 1995). There are, however, submarine Hawaiian tholeiitic lavas with a range of [delta]¹⁸O values (4·9-5·8%^o) and mantle-like [delta]D (-61 to -88%^o; Kyser & O'Neil, 1984; Garcia et al., 1989b). Thus, the large variation in [delta]¹⁸O values for these Hawaiian tholeiitic glasses apparently supports the interpretation from other isotopic work on Hawaiian tholeiites (e.g. West et al., 1987) that at least two isotopically distinct reservoirs are present in the source for Hawaiian magmas.

We embarked on this study to evaluate whether the oxygen isotope variation in fresh Hawaiian tholeiitic basalts was indeed caused by mantle source heterogeneity or crustal contamination. To simplify this effort, we focused our attention on the pristine basalts from the current Puu Oo eruption of Kilauea Volcano on the island of Hawaii. This eruption is the best studied, longest lived (14 years and continuing), most voluminous (~1·5 km³) and one of the most compositionally variable (5·7-10·1 wt % MgO) historical eruptions of Kilauea Volcano. Unlike most previous studies on oceanic island basalts, our approach was to determine [delta]¹⁸O values of both glassy matrix material (95 to >99 vol. % of these rocks) and olivine (0·1-3 vol. %) for all the compositional variants from this eruption and to check for temporal variation. Olivine analyses are a critical component of this study because olivine is the liquidus mineral in Hawaiian tholeiitic magmas (Wright, 1971) and it is thought to be resistant to oxygen exchange, even at magmatic temperatures (Cole & Ohmoto, 1986). Thus, the olivine analyses should provide a good record of the early oxygen isotope history of Puu Oo magmas.

Our results for Puu Oo lavas show a temporal variation in matrix [delta]¹⁸O values (4·56-5·25%^o) that correlates with changes in vent location and eruption style. A sharp increase in matrix [delta]¹⁸O (~0·4%^o) occurred following such a change but the [delta]¹⁸O of the coexisting olivines decreased slightly. The small difference between [delta]¹⁸O for olivines and host matrix [[Delta](ol-m)] in many Puu Oo lavas (>-0·4%^o) is indicative of disequilibrium. Thus, although the Puu Oo lavas are pristine and have no other apparent geochemical evidence of crustal assimilation, the [delta]¹⁸O disequilibrium records the effects of crustal contamination and oxygen exchange, which occurred after the crystallization of some of the olivine crystals. Our results indicate that caution should be exercised in interpreting low [delta]¹⁸O values for basalts (<5·0%^o), especially without oxygen isotope data for coexisting minerals, and that time series studies are important for interpreting geochemical processes.

BRIEF HISTORY OF THE PUU OO ERUPTION

The Puu Oo eruption started in January 1983 (episode 1) and is continuing vigorously with typical eruption rates of (0·3-0·5) * 10⁶ m³/day (Kauahikaua et al., 1996). Between 1983 and the end of 1996, the eruption produced ~1·5 km³ of lava (dense rock equivalent). The initial episode of the eruption formed a `curtain of fire' along an 8-km-long, discontinuous fissure system which intermittently produced lava for 20 days (Wolfe et al., 1987). During the next 3·5 years (episodes 2-47), eruptive activity was episodic with repose periods of 8-65 days during which magma accumulated and fractionated in a shallow reservoir under the Puu Oo vent (Fig. 1). These eruptive episodes were usually short (5-100 h) with lava fountains of variable height (tens of meters to 400 m high).

Figure 1. Hypothetical cross-section of Kilauea's summit and east rift zone showing the inferred magma reservoir system for the Puu Oo eruption based on seismic and petrologic data (after Garcia et al., 1996). The [delta]¹⁸O values are shown for the matrix from Puu Oo eruption lavas (Table 1), HGP-A geothermal well cores (Smith & Thomas, 1990), the 1977 eruption magma (as represented by sample 1-39 erupted in 1983), and the matrix from six summit lavas erupted since 1924 (E. Ito, unpublished data, 1997). Groundwater and summit condensate [delta]¹⁸O