shift from LREE-enriched augites in basaltic glass, to. REE-depleted Cr-diopside in highly silicic glass. Esti- mates indicate that the most silicic glasses represent.
Contrib Mineral Petrol (1999) 137: 59 ± 82
Ó Springer-Verlag 1999
E. Wul-Pedersen á E.-R. Neumann á R. Vannucci P. Bottazzi á L. Ottolini
Silicic melts produced by reaction between peridotite and in®ltrating basaltic melts: ion probe data on glasses and minerals in veined xenoliths from La Palma, Canary Islands
Received: 15 November 1998 / Accepted: 17 May 1999
Abstract Mantle xenoliths hosted by the historic Volcan de San Antonio, La Palma, Canary Islands include veined spinel harzburgites and spinel dunites. Glasses and associated minerals in the vein system of veined xenoliths show a gradual transition in composition from broad veins to narrow veinlets. Broad veins contain alkali basaltic glass with semi-linear trace element patterns enriched in strongly incompatible elements. As the veins become narrower, the SiO2-contents in glass increase (46 ® 67 wt% SiO2 in harzburgite, 43 ® 58 wt% in dunite) and the trace element patterns change gradually to concave patterns depleted in moderately incompatible elements (e.g. HREE, Zr, Ti) relative to highly incompatible ones. The highest SiO2-contents (ca. 68% SiO2, low Ti-Fe-Mg-Ca-contents) and most extreme concave trace element patterns are exhibited by glass in unveined peridotite xenoliths. Clinopyroxenes shift from LREE-enriched augites in basaltic glass, to REE-depleted Cr-diopside in highly silicic glass. Estimates indicate that the most silicic glasses represent melts in, or near, equilibrium with their host peridotites. The observed trace element changes are compatible with formation of the silicic melts by processes involving in®ltration of basaltic melts into mantle peridotite followed by reactions and crystallization. The Fe-Mg interdiusion pro®les in olivine porphyroclasts adjacent E. Wul-Pedersen á E.-R. Neumann (&) Mineralogical-Geological Museum, University of Oslo, Sarsgt. 1, N-0562 Oslo, Norway R. Vannucci Dipartimento di Scienze della Terra, UniversitaÁ di Pavia, via Ferrata 1, I-27100 Pavia, Italy R. Vannucci á P. Bottazzi á L. Ottolini CNR-Centro di Studio per la Cristallochimica e la Cristallogra®a, via Ferrata 1, I-27100 Pavia, Italy Editorial responsibility: J. Touret
to the veins indicate a minimum period of diusion of 600 years, implying that the reaction processes have taken place in situ in the upper mantle. The CaO-TiO2-La/ Nd relationships of mantle rocks may be used to discriminate between metasomatism caused by carbonatitic and silicic melts. Unveined mantle xenoliths from La Palma and Hierro (Canary Islands) show a wide range in La/Nd ratios with relatively constant, low-CaO contents which is compatible with metasomatism of ``normal'' abyssal peridotite by silicic melts. Peridotite xenoliths from Tenerife show somewhat higher CaO and TiO2 contents than those from the other islands and may have been aected by basaltic or carbonatitic melts. The observed trace element signatures of ultrama®c xenoliths from La Palma and other Canary Islands may be accounted for by addition of small amounts (1±7%) of highly silicic melt to unmetasomatized peridotite. Also ultrama®c xenoliths from other localities, e.g. eastern Australia, show CaO-TiO2-La/Nd relationships compatible with metasomatism by silicic melts. These results suggest that silicic melts may represent important metasomatic agents.
Introduction Metasomatism is generally recognized as a very important process in the upper mantle. However, the transport agents causing dierent types of mantle metasomatism have been the subject of considerable debate over the years. H2O- and CO2-rich ¯uids as well as basaltic and carbonatitic melts have been proposed as media capable of causing cryptic metasomatism in mantle peridotites (e.g. Francis 1976; Harte 1983; Dawson 1984; Eggler 1987; Menzies et al. 1987; Wilshire 1987; Nielson and Noller 1987). Lately it has been recognized that in addition to such ¯uids and melts, silicic melts (60±72 wt% SiO2) are frequently present in the upper mantle. Silicic melts are observed as glass inclusions and interstitial glass pockets in spinel-bearing harzburgite, lherzolite and
60
dunite xenoliths in numerous localities around the world (e.g. Frey and Green 1974; Francis 1976, 1987; Jones et al. 1983; Siena et al. 1991; Ionov et al. 1994; Schiano et al. 1992, 1994; Schiano and Clocchiatti 1994; Neumann and Wul-Pedersen 1995; Schiano et al. 1995; Zinngrebe and Foley 1995; Wul-Pedersen et al. 1996a). It has also been proposed that silicic melts may act as metasomatic transport agents (Zinngrebe and Foley 1995; WulPedersen et al. 1996a; Draper and Green 1997). Mantle xenoliths from dierent Canary Islands show extensive evidence of pervasive metasomatism in the underlying lithospheric mantle, including enrichment in incompatible elements (cryptic metasomatism) (Neumann 1991; Neumann et al. 1995; Whitehouse and Neumann 1995; Wul-Pedersen et al. 1996a; Neumann and Wul-Pedersen 1997; E.-R. Neumann, unpublished data). These peridotite xenoliths also frequently exhibit basaltic to silicic glasses, representing trapped melts (Neumann and Wul-Pedersen 1997). Silicic glasses are far more abundant than basaltic ones. Wul-Pedersen et al. (1996a) presented major element data on minerals and glasses in veined spinel peridotites from La Palma, Canary Islands. The glasses in these xenoliths show a gradual transition from alkali basaltic (ca. 43 wt% SiO2, Ti-Fe-rich, TiO2/Al2O3 > 0.15) in broad veins, to highly silicic (ca. 67% SiO2, Ti-Fe-Mg-Ca-poor, TiO2/ Al2O3< 0.08) in narrow veinlets penetrating harzburgite fragments. The mineral-melt relations and contact relations between glass and peridotite minerals found in the vein systems of veined xenoliths closely resemble those found for unveined Canary Islands xenoliths. The observed shift in glass composition and accompanying changes in phase assemblage and mineral compositions were interpreted as the results of reactions between in®ltrating TiO2-rich alkali basaltic melts and refractory peridotite wall-rock (IRC-processes = in®ltration ± reaction ± crystallization). Neumann and Wul-Pedersen (1997) concluded that IRC-processes represent the general mode of formation of silicic melts in the upper mantle under the Canary Islands, and that these melts may represent important metasomatic agents. The IRCprocesses are thus clearly widespread and important in the upper mantle under the Canary Islands, and may also represent important mantle processes in general. The veined xenoliths are believed to demonstrate on a small scale general reaction processes which take place when TiO2-rich alkali basaltic melts in®ltrate and react with refractory peridotite mantle wall-rock. The petrography and major element chemistry of the veined xenoliths, and the sequence of reactions taking place between in®ltrating melts and mantle wall-rocks were described in detail by Wul-Pedersen et al. (1996a). That paper also included major and trace element whole-rock data and major element mineral data on unveined mantle xenoliths from La Palma. Vannucci et al. (1998) discussed the partitioning of REE (rare earth elements), Y, Sr, Zr and Ti between clinopyroxene and silicate melts of dierent compositions in veined and unveined mantle xenoliths from La Palma. However,
neither of those papers presented a general discussion of the trace element compositions of glasses and coexisting minerals in mantle xenoliths from the Canary Islands, nor their implications for IRC-processes and mantle metasomatism. The present investigation was undertaken in order to: (1) describe the change in trace element compositions of basaltic to highly silicic glasses and associated minerals in the vein system of the veined xenoliths from La Palma; (2) use this information to obtain information about the eect of IRC-processes on trace elements; (3) compare the trace element compositions of glasses in the vein systems of veined xenoliths with those of glasses in unveined xenoliths; (4) compare the expected metasomatic eect of silicic melts with that of other transport agents believed to cause mantle metasomatism; (5) discuss the transport agent(s) causing metasomatic enrichment in the upper mantle under La Palma.
Analytical methods Minerals and glasses in polished thin sections were analysed for major elements using an automatic wavelength-dispersive CAMECA electron microprobe ®tted with a LINK energy-dispersive system at the Mineralogical-geological Museum in Oslo. Analyses were run with an acceleration voltage of 15 keV, sample currents of 20 nA for Na-poor (olivine, pyroxene, spinel) and 10 nA for Na-rich phases (glass), and counting times of 100 seconds. Oxides and natural and synthetic minerals were used as standards. Matrix corrections were performed by the PAP-procedure in the CAMECA software. Analytical precision and accuracy (2r) evaluated by repeated analysis of individual grains and standards are better than 1% for elements in concentrations of > 20 wt% oxide, better than 2% for elements in the range 10± 20 wt% oxide, better than 5% for elements in the range 2±10 wt% oxide, and better than 10% for elements in the range 0.5±2 wt% oxide. The low sums obtained for some of the glasses (Table 1) cannot be accounted for by their H2O-contents. Trace elements (K, Sc, Ti, V, Cr, Rb, Sr, Y, Zr, Nb, Cs, Ba, and 9 selected REE), and F and H2O content in minerals and glasses were determined using the Cameca IMS 4f ion microprobe at CSCC, Pavia. The so-called ``energy-®ltering'' technique (Shimizu et al. 1978) was applied by analysing high-energy ions consisting mostly of monoatomic species (see Bottazzi et al. 1994, for details). Residual BaO+ and NdO+ interferences on Eu+ and Gd+, respectively, were removed using the peak-stripping procedures described in Bottazzi et al. (1994). Both H2O and F were acquired under the same measurement conditions and quanti®ed according to the analytical protocol reported in Ottolini et al. (1995). Conversion of ion intensities into concentrations was accomplished by using both a major element as internal standard (Si) and several silicate reference samples as external standards. Accuracy of ion probe data is better than 10% for concentrations greater than about 2 times the chondritic value of the element, with the exceptions F, Rb, and Nb (accuracy 25%). Below this concentration level, precision and hence accuracy are limited by counting statistics; for instance, the standard deviation is 30% at 0.2 times the chondritic value. For H2O, accuracy is 15% for samples with SiO2
