Skip Navigation


Journal of Petrology Advance Access originally published online on April 16, 2007
Journal of Petrology 2007 48(5):853-883; doi:10.1093/petrology/egm004
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
48/5/853    most recent
egm004v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (2)
Right arrowRequest Permissions
Google Scholar
Right arrow Articles by Zajacz, Z.
Right arrow Articles by Pettke, T.
Right arrow Search for Related Content
GeoRef
Right arrow GeoRef Citation
Social Bookmarking
 Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

© The Author 2007. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oxfordjournals.org

Evolution of Mafic Alkaline Melts Crystallized in the Uppermost Lithospheric Mantle: a Melt Inclusion Study of Olivine-Clinopyroxenite Xenoliths, Northern Hungary

Zoltán Zajacz1,*, István Kovács2, Csaba Szabó3, Werner Halter1 and Thomas Pettke4

1Department of Earth Sciences, Institute of Isotope Geochemistry and Mineral Resources, Eth Zürich, 8092 Zürich, Switzerland
2Research School of Earth Sciences, Building 61 Mills Road, the Australian National University, Canberra, Act 0200, Australia
3Lithosphere Fluid Research Laboratory, Institute of Geography and Earth Sciences, Eötvös University Budapest (Elte), Pazmany Setany 1/c, Budapest, Hungary, H-1117
4University of Bern, Institute of Geological Sciences, Baltzerstrasse 1–3, Ch-3012 Bern, Switzerland

RECEIVED APRIL 7, 2006; ACCEPTED FEBRUARY 2, 2007


    ABSTRACT
 TOP
 ABSTRACT
 INTRODUCTION
 GEOLOGICAL SETTING
 PETROGRAPHY OF THE CUMULATE...
 ANALYTICAL TECHNIQUES
 COMPOSITIONS OF THE ROCK-FORMING...
 HOMOGENIZATION EXPERIMENTS AND...
 DISCUSSION
 CONCLUSIONS
 APPENDIX
 REFERENCES
 
Olivine-clinopyroxenite xenoliths exhumed in alkali basalts (sensu lato) in the Nógrád–Gömör Volcanic Field (NGVF), northern Hungary, contain abundant silicate melt inclusions. Geothermobarometric calculations indicate that these xenoliths crystallized as cumulates in the upper mantle near the Moho. These cumulate xenoliths are considered to represent a period of Moho underplating by mafic alkaline magmas prior to the onset of Late Tertiary alkaline volcanism in the Carpathian–Pannonian region. The major and trace element compositions of silicate melt inclusions in olivine display an evolutionary trend characterized by a strong decrease in CaO/Al2O3. The parental melt of the cumulates was a basanite formed by low-degree (~ 2%) partial melting of a garnet peridotite source. The compositional trend of the silicate melt inclusions, textural features, and modelling with pMELTS show that the parental melt evolved by major clinopyroxene and minor olivine crystallization followed by the appearance of amphibole simultaneously with significant resorption of the earlier clinopyroxene and olivine. The resulting residual melt was highly enriched in Al2O3, alkalis and most incompatible trace elements. This type of melt is likely to infiltrate and react with surrounding mantle peridotite as a metasomatic agent. It might also form high-pressure pegmatite-like bodies in the mantle that might be the source of the amphibole and sanidine megacrysts also found in the alkali basalts of the NGVF. Preferential remelting of the later-formed (i.e. lower temperature) mineral assemblage (amphibole, sanidine, residual glass) might have significantly contaminated the host alkaline mafic lavas, increasing their Al2O3 and total alkali contents and, therefore, reducing their MgO, FeO and CaO content.

KEY WORDS: silicate melt inclusions; geochemistry; petrogenesis; Nógrád–Gömör Volcanic Field; Pannonian Basin


    INTRODUCTION
 TOP
 ABSTRACT
 INTRODUCTION
 GEOLOGICAL SETTING
 PETROGRAPHY OF THE CUMULATE...
 ANALYTICAL TECHNIQUES
 COMPOSITIONS OF THE ROCK-FORMING...
 HOMOGENIZATION EXPERIMENTS AND...
 DISCUSSION
 CONCLUSIONS
 APPENDIX
 REFERENCES
 
Olivine-clinopyroxenite and clinopyroxenite xenoliths, and clinopyroxene and amphibole megacrysts are common within the alkali basalts in the Pannonian Basin (Embey-Isztin et al., 1990Go, 1993Go; Downes et al., 1995Go; Dobosi & Jenner, 1999Go; Dobosi et al., 2003Go; Török et al., 2003Go; Kovács et al., 2004Go) and worldwide (Frey & Prinz, 1978Go; Irving & Frey, 1984Go; Wilson & Downes, 1991Go; Shaw & Eyzaguirre, 2000Go; Chen et al., 2001Go; Bondi et al., 2002Go; Riley & Bailey, 2003Go; Liu et al., 2005Go). In most studies they have been interpreted as high-pressure crystallization products of melts similar in composition to the host alkaline basalts, based on thermobarometric calculations, isotopic ratios and estimated equilibrium melt compositions (Irving, 1974Go; Stuckless & Irving, 1976Go; Frey & Prinz, 1978Go; Shaw & Eyzaguirre, 2000Go; Chen et al., 2001Go; Bondi et al., 2002Go; Dobosi et al., 2003Go; Riley & Bailey, 2003Go; Downes et al., 2004Go; Kovács et al., 2004Go; Shaw et al., 2005Go). Alternatively, the olivine-clinopyroxenite xenoliths have been interpreted as the products of metamorphic segregation during anatexis of the surrounding mantle region (Chen et al., 2001Go). Alkali feldspar phenocrysts have also been reported in a number of alkali basalts (Frey & Prinz, 1978Go; Aspen et al., 1990Go; Riley & Bailey, 2003Go), but the origin of the siliceous melt from which they crystallized has not yet been identified. Only Irving (1974Go) suggested that they might form as near-solidus crystallization products of primitive alkaline melts in the mantle.

In the case of crystallization of basaltic melts, separation of the mafic minerals forms cumulus ultramafic rocks composed of clinopyroxene, olivine and usually Al-spinel; such rocks are not representative of the parental melt, thus mineral–melt equilibria are needed to constrain the geochemical character of the parental melt.

Here we study silicate melt inclusions (SMIs) from olivine-clinopyroxenite xenoliths hosted in alkali basalts of the Nógrád–Gömör Volcanic Field (NGVF), northern Pannonian Basin (Hungary), to directly obtain the composition of the parental melt of the xenoliths. Most previous studies of SMIs have focused on melt inclusions in phenocrysts in volcanic rocks (e.g. Nielsen et al., 1995Go; Kamenetsky & Clocchiatti, 1996Go; Kamenetsky et al., 1997Go; Xu et al., 1999Go; Norman et al., 2002Go; Bertagnini et al., 2003Go; Cervantes & Wallace, 2003Go; Danyushevsky et al., 2003Go; Maclennan et al., 2003Go; Schiano, 2003Go; Schiano et al., 2004Go). Here we present data on SMIs within the constituent minerals of cumulate xenoliths derived by fragmentation of a magma body that crystallized slowly near the Moho.

The major goals of this paper are fourfold: (1) to provide additional information on the genesis of Type-II (mostly clinopyroxenite and olivine-clinopyroxenite) xenoliths; (2) to obtain a better understanding of the evolution of alkali basaltic systems; (3) to test of the representative nature of olivine-hosted melt inclusions in these xenoliths; (4) to provide new information on the evolution of the Carpathian–Pannonian region. We reproduced the original composition and the chemical evolution of the parental melt based on electron microprobe analyses (EPMA) of homogenized melt inclusions and laser ablation inductively coupled plasma mass spectrometry (LA-ICPMS) analysis of unheated multiphase inclusions in olivine. Petrogenetic models, based on the major and trace element composition of the SMIs and the rock-forming minerals, are presented and used to constrain the formation and evolution of the parental melt.


    GEOLOGICAL SETTING
 TOP
 ABSTRACT
 INTRODUCTION
 GEOLOGICAL SETTING
 PETROGRAPHY OF THE CUMULATE...
 ANALYTICAL TECHNIQUES
 COMPOSITIONS OF THE ROCK-FORMING...
 HOMOGENIZATION EXPERIMENTS AND...
 DISCUSSION
 CONCLUSIONS
 APPENDIX
 REFERENCES
 
The Nógrád–Gömör Volcanic Field (NGVF) at the northern margin of the Pannonian Basin (Northern Hungary–Southern Slovakia) (Fig. 1) is mostly composed of Plio-Pleistocene alkali basalts (sensu lato). It is surrounded by the Miocene, subduction-related, Inner Carpathian calc-alkaline volcanic belt. The Pannonian Basin was formed as a result of lithospheric stretching in two major extensional episodes: (1) an Early to Middle Miocene passive rifting phase driven by subduction rollback (Royden et al., 1983aGo, 1983bGo; Csontos, 1995Go; Fodor et al., 1999Go); (2) a Late Miocene syn- to post-rift phase accompanied by asthenospheric upwelling and thinning of the mantle lithosphere (Royden et al., 1983aGo, 1983bGo; Huismans et al., 2001Go). The alkali basalts were erupted in the post-extensional phase and sampled the lithospheric section beneath the region, transporting a large quantity of peridotite (Type-I), ultramafic cumulate (Type-II) and crustal xenoliths to the surface (Embey-Isztin et al., 1993Go; Szabó & Taylor, 1994Go; Embey-Isztin & Dobosi, 1995Go; Zajacz & Szabó, 2003Go; Kovács et al., 2004Go; Seghedi et al., 2004Go) (Fig. 1). The xenoliths investigated in this study were collected from the Nagy-Salgó volcano. The locality is a well-preserved basanite volcanic neck containing a high abundance of olivine-clinopyroxenite and wehrlite xenoliths. The host rock is Pliocene (5·27 Ma) in age (Szabó et al., unpublished data).


Figure 1
View larger version (34K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 1. Map of the Nógrád–Gömör Volcanic Field. The most important alkaline basalt (sensu lato) occurrences are shown; the studied xenoliths were collected from Nagy-Salgó.

 

    PETROGRAPHY OF THE CUMULATE XENOLITHS AND SILICATE MELT INCLUSIONS
 TOP
 ABSTRACT
 INTRODUCTION
 GEOLOGICAL SETTING
 PETROGRAPHY OF THE CUMULATE...
 ANALYTICAL TECHNIQUES
 COMPOSITIONS OF THE ROCK-FORMING...
 HOMOGENIZATION EXPERIMENTS AND...
 DISCUSSION
 CONCLUSIONS
 APPENDIX
 REFERENCES
 
In the NGVF several eruptive centres contain clinopyroxene- and olivine-rich xenoliths. In some locations amphibole is also typical and has been interpreted as a metasomatic alteration product of the clinopyroxene (Kovács et al., 2004Go). The xenoliths selected for a detailed melt inclusion study are mostly olivine-clinopyroxenites, and subordinately wehrlites. Regardless of variations in the clinopyroxene to olivine ratio, the xenoliths are texturally similar. The volume proportion of olivine ranges from 20% to 40% (Table 1). Both the clinopyroxene and the olivine are euhedral or subhedral and their grain sizes are between 0·5 and 5 mm (Fig. 2a). The edges of the olivine crystals are resorbed, whereas the clinopyroxenes typically display a narrow reaction rim (Fig. 2b and c). In addition to the major rock-forming clinopyroxene and olivine, two types of interstitial phase assemblages are present: (1) a light-coloured assemblage composed of euhedral alkali feldspar, clinopyroxene, Ti-magnetite, apatite and residual glass (Fig. 2d and e); (2) a dark-coloured assemblage of partially decomposed amphibole displaying evidence of breakdown to rhönite, nepheline, clinopyroxene, olivine and Ti-magnetite. In some xenoliths the amphibole is completely decomposed and only the residual minerals are present (Fig. 2f). Interstitial phase assemblage (1) is typical of all the xenoliths whereas interstitial phase assemblage (2) is observed in small quantities only in a few xenoliths (Table 1). Despite the uniform texture and mineralogy there is a significant variation in the melt and fluid inclusion content of the xenoliths. Out of five large (3–5 cm in diameter) unaltered xenoliths selected for investigation, only two contained SMIs in olivine large enough for analysis (> 10 µm), whereas clinopyroxenes are generally rich in SMIs and CO2 inclusions (Table 1). As the main focus of our study is on olivine-hosted SMIs, these two xenoliths (NNS07 and NNS08) have been selected for detailed analytical work.


Figure 2
View larger version (172K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 2. Petrography of the xenoliths. (a) Texture of a typical olivine-clinopyroxenite xenolith (NNS08); (b) mode of occurrence of the intercumulus phase (NNS07); (c) BSE image of a typical clinopyroxene crystal in the cumulate with Ti-rich late-stage rim (NNS08); (d) BSE image of the intercumulus phase (NNS07); (e) image of the intercumulus phase in reflected light (NNS07); (f) breakdown assemblage after amphibole in transmitted light (NNS08). cpx, clinopyroxene; ol, olivine; ic, intercumulus phase; san, sanidine; gl, glass; Ti-mt, Ti-magnetite; rh, rhönite; ne, nepheline.

 

View this table:
[in this window]
[in a new window]

 
Table 1: Modal composition (vol. %) and inclusion content of the studied xenoliths

 
Olivines contain SMIs either as single inclusions or in clusters. They are generally absent from the cores of large olivine crystals. The inclusions are always rounded and almost isometric (Fig. 3a–c), ranging from 8 to 40 µm in diameter. They occur as partially recrystallized multiphase assemblages of clinopyroxene (20–30 vol. %), Al-spinel (1–3 vol. %), rhönite (10–20 vol. %) and/or Ti-rich pargasitic amphibole (5–15 vol. %), together with residual glass (40–60 vol. %) and a CO2 bubble (4–8 vol. %). Small sulphide blebs are also present in several inclusions (Fig. 3c).


Figure 3
View larger version (119K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 3. Petrography of the melt inclusions. (a) Multiphase SMIs in olivine (NNS07); (b) close-up of a non-exposed multiphase melt inclusion in olivine (NNS07); (c) melt inclusion containing sulphide bleb in olivine (RL) (NNS07); (d) clinopyroxene crystal packed with elongated and oriented melt inclusions and dark CO2 fluid inclusions (NNS07).

 
Clinopyroxene crystals are packed with elongated and oriented SMIs. Their size ranges from submicroscopic to 30 µm in diameter (Fig. 3d). They are usually present from grain cores to rims. SMI size distribution sometimes shows periodic zonation with alternating zones of small (submicroscopic to 5 µm in diameter) and large inclusions (from 5 µm to 30 µm in diameter) following the growth direction of the clinopyroxene. They are partially recrystallized and composed of clinopyroxene (0–20 vol. %), Ti-rich pargasite (0–10 vol. %), Al-spinel (1–3 vol. %), Ti-magnetite (0–2 vol. %), residual glass (65–95 vol. %) and a CO2 bubble. The CO2 fluid to melt ratio is variable, suggesting accidental entrapment of a previously exsolved fluid phase along with the melt (heterogeneous entrapment) (Fig. 3d). Individual CO2 inclusions have also been identified in clinopyroxene. Most of the CO2 inclusions have leaked; however, some smaller ones (generally <8 µm) preserve two phases or liquid CO2 at room temperature. The presence of CO2 was determined by Raman spectroscopy; significant H2O was not detected in the fluid phase.

The compositions of the daughter phases in the SMIs in both olivine and clinopyroxene are given in Table 2.


View this table:
[in this window]
[in a new window]

 
Table 2: Average composition of the daughter minerals and the residual glass in olivine- and clinopyroxene-hosted silicate melt inclusions in xenolith NNS07 (in wt %)

 

    ANALYTICAL TECHNIQUES
 TOP
 ABSTRACT
 INTRODUCTION
 GEOLOGICAL SETTING
 PETROGRAPHY OF THE CUMULATE...
 ANALYTICAL TECHNIQUES
 COMPOSITIONS OF THE ROCK-FORMING...
 HOMOGENIZATION EXPERIMENTS AND...
 DISCUSSION
 CONCLUSIONS
 APPENDIX
 REFERENCES
 
SMIs were reheated using a Linkam TS1500 stage at the Lithosphere Fluid Research Laboratory, Eötvös University, Budapest. Heating was carried out in a nitrogen atmosphere on separated and doubly polished olivine grains. The initial high (30°C/min) heating rate was progressively decreased to 5°C/min approaching the melting temperature of the daughter phases. The stage allowed a cooling rate of ~300°C/min in the high-temperature range (above the glass transition temperature).

Homogenized SMIs and the daughter phases of the SMIs were analyzed at the University of Vienna on a Cameca SX-100 electron microprobe using a 15 kV acceleration voltage, 20 nA beam current, 20 s peak and 10 s background count time for each element. On all heated SMIs, at least two analyses were carried out to test their homogeneity. Compositions of the rock-forming minerals and interstitial phase assemblages were determined at ETH Zürich using a JEOL SUPERPROBE JXA-8200 electron microprobe. A 15 kV acceleration voltage and 15 nA beam current were used and a 20 s counting time was applied for each element. For mineral analysis a 2 µm beam diameter was used; for glasses, the beam diameter was enlarged (to 5 or 10 µm) and the beam current was reduced to 10 nA. The following standards were used: wollastonite for SiO2 and CaO; forsterite for MgO; aegirine for Na2O; sanidine for K2O; apatite for P2O5; natural oxides for FeO, TiO2 and MnO.

LA-ICPMS analysis of the SMIs and their host clinopyroxene and olivine was carried out at ETH Zürich, using an ArF excimer UV (193 nm) laser source at 70mJ output energy and a repetition rate of 10 Hz. The beam diameter varied between 20 and 60 µm for SMI measurements, whereas for bulk measurement of the interstitial phase assemblages a beam diameter of 110 µm was used. SMIs were analyzed and the data reduction was performed using the method of Halter et al. (2002Go) (see Appendix for further details). Only unexposed SMIs were measured. A beam diameter slightly larger than the actual size of the inclusion was used to avoid incomplete sampling of the heterogeneous SMIs. Melt inclusions deeper than 30 µm beneath the section surface were not measured. Inclusion signals shorter than 15 readings were not considered. In the case of SMI measurements, a 10 ms dwell time was used for all elements. However, for the clinopyroxene host measurements we used an extended (20 ms) dwell time for Yb and Lu because of their low concentrations. Quantification of the interstitial phases was based on analysis of all the major elements and normalization to 100% without application of any internal standard (Halter et al., 2002Go). For external standardization the NBS SRM 610 glass standard was used.

Uncertainties are variable for each single analysis, but for SMIs are mainly in the range 5–20 relative % for concentrations of large ion lithophile elements (LILE), high field strength elements (HFSE) and light rare earth elements (LREE), and 15–50 relative % for heavy REE (HREE), which are present in very low concentrations. In the case of clinopyroxene, amphibole, and bulk intercumulus phase analyses uncertainties are below 15 relative% for most elements.


    COMPOSITIONS OF THE ROCK-FORMING MINERAL PHASES
 TOP
 ABSTRACT
 INTRODUCTION
 GEOLOGICAL SETTING
 PETROGRAPHY OF THE CUMULATE...
 ANALYTICAL TECHNIQUES
 COMPOSITIONS OF THE ROCK-FORMING...
 HOMOGENIZATION EXPERIMENTS AND...
 DISCUSSION
 CONCLUSIONS
 APPENDIX
 REFERENCES
 
Major mineral constituents
The clinopyroxene is diopsidic and displays almost homogeneous cores surrounded by reaction rims (Fig. 2c). The compositional transition between the cores and the rims is not continuous and it shows clustering of mg-numbers around 82–85 (core) and 78–80 (rim), respectively (Table 2). Both the jadeite component and TiO2 content increase from core to rim (Table 3). The reaction rim of the clinopyroxene can be distinguished on back-scattered electron (BSE) images (Fig. 2c).


View this table:
[in this window]
[in a new window]

 
Table 3: Composition of selected xenolith-forming clinopyroxenes (in wt %) and results of geothermobarometry calculations using the method of Putirka et al. (2003Go)

 
Trace element concentrations (Table 4) show no obvious trends as a function of mg-number; however, rim measurements display the lowest mg-number and the highest incompatible trace element concentrations. It should be noted that these analyses are not purely representative of the composition of narrow reaction rims, as the ablation pits are larger than the width of the reaction rim. This is also shown by the transitional major element (especially TiO2) content compared with EMPA data for clinopyroxene cores and rims. Furthermore, if we plot an incompatible trace element concentration against compatible and incompatible elements, clear trends can be observed even in the grain cores that appear to be almost homogeneous with respect to major elements (Fig. 4 and Table 3). The data from the rims fall at the evolved end of these trends. Most incompatible elements display a positive correlation with each other; however, for Y, Yb and Lu this correlation is not so obvious (Fig. 4). More precisely, the concentration of Y shows no systematic change, and the variation of Yb and Lu vs Ce shows a U-shaped pattern, with the lowest values at intermediate Ce contents. In terms of compatible trace elements, Ni shows a clear negative correlation with Ce, whereas Sc and V display a less defined distribution. The Cu content decreases exponentially with increasing Ce concentration (Fig. 4).


View this table:
[in this window]
[in a new window]

 
Table 4: Trace element composition of the clinopyroxenes (in ppm)

 

Figure 4
View larger version (11K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 4. Trace element variation in the rock-forming clinopyroxene as a function of a selected incompatible trace element, Ce.

 
The olivine grains are homogeneous in composition in each xenolith, but their compositions differ between xenoliths. Their mg-number ranges from 79 to 85 and the Ni content is between 650 and 1200 ppm. The CaO content ranges from 0·04 to 0·19 wt % with no systematic variation.

The amphibole compositions plot between pargasite and kaersutite with 4·2–4·8 wt % TiO2 (Table 5). The mg-number is between 74 and 77. Their trace element patterns normalized to primitive mantle show positive HFSE and LILE and characteristic negative Pb anomalies (Fig. 5). The breakdown assemblage of amphibole consists of rhönite, Ti-rich diopsidic clinopyroxene, nepheline (Na/K ~ 5·0), olivine (mg-number ~ 84·0) and ilmenite (Table 5, Fig. 2f). The bulk composition of the breakdown assemblage is in good agreement with the amphibole composition in terms of both major and trace elements (Table 5).


Figure 5
View larger version (23K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 5. Primitive normalized trace element diagram showing the composition of the xenolith-forming amphiboles. Metasomatic amphiboles found in peridotite xenoliths of the region (Filakovo, Maskova, Bárna) are also shown (Szabó et al., unpublished data).

 

View this table:
[in this window]
[in a new window]

 
Table 5: Average composition of the mineral phases forming the breakdown assemblage of amphibole in xenolith NNS08

 
Intercumulus phase assemblages
The alkali feldspar in the intercumulus phase is sanidine with a Na/K ratio of 0·6, the clinopyroxene is diopsidic and the residual glass is highly enriched in Al2O3 and alkali oxides (Table 6). Relative to the parental melt of the cumulates, the bulk composition of the intercumulus phase assemblage has a slightly elevated SiO2 content (between 50·9 and 55·3 wt %) and very strong enrichment in alkali elements (9·0–15·0 wt % Na2O + K2O), Al2O3 (20·8–25·2 wt %) and all incompatible trace elements (Table 7), except for Y and HREE. The CaO/Al2O3 ratio is very low (0·12–0·37). The REE pattern is remarkably steep, with La/Yb ranging from 21 to 32 (Table 7). The FeO and MgO contents are low, whereas the mg-numbers are relatively high (49–67).


View this table:
[in this window]
[in a new window]

 
Table 6: Average composition of mineral phases and glass (in wt %) in the intercumulus phase assemblage in xenolith NNS07

 

View this table:
[in this window]
[in a new window]

 
Table 7: Bulk composition of the intercumulus phase assemblages determined by LA-ICPMS in xenolith NNS07

 
Downes et al. (2004Go) have studied glass-bearing hornblendite and clinopyroxenite xenoliths entrained in intraplate alkali basalts from Germany and Israel. The bulk compositions of the intercumulus phases compared with the composition of their interstitial glasses have similar total alkali content, but higher concentrations of SiO2 and Al2O3, and lower concentrations of MgO, CaO, FeO and incompatible trace elements.


    HOMOGENIZATION EXPERIMENTS AND RESULTS
 TOP
 ABSTRACT
 INTRODUCTION
 GEOLOGICAL SETTING
 PETROGRAPHY OF THE CUMULATE...
 ANALYTICAL TECHNIQUES
 COMPOSITIONS OF THE ROCK-FORMING...
 HOMOGENIZATION EXPERIMENTS AND...
 DISCUSSION
 CONCLUSIONS
 APPENDIX
 REFERENCES
 
The melting temperatures of the residual glass in the SMIs hosted in olivine range from 1090 to 1170°C. Observation of the onset of glass melting was rather difficult; thus, these values are maximum estimates. Final melting of clinopyroxene and rhönite in SMIs occurred between 1190 and 1250°C, whereas spinel persisted to 1290–1340°C in several cases. Complete dissolution of the gas bubble did not occur in a significant proportion of the studied SMIs, which indicates heterogeneous gas entrapment or partial pressure relaxation in the post-entrapment history of the SMIs. However, decrepitation features were never observed. Despite the relatively slow cooling rate (~ 300°C/min), partial recrystallization of the homogenized SMIs could not be observed, neither at microscopic scale nor by scanning electron microscope. Homogenization of clinopyroxene-hosted inclusions was not possible, because melting of the host clinopyroxene began before complete homogenization of the SMIs at atmospheric pressure.

Eighteen reheated SMIs were exposed and polished after homogenization and analyzed by electron microprobe. All are at least 15 µm in diameter and were analyzed with at least two or three points to check homogeneity. Twelve inclusions appeared to be completely homogeneous and only data obtained from these were used for interpretation.

Melt inclusion compositions were corrected by adding olivine to achieve a KdFe–Mg value of 0·32 that is appropriate for olivines crystallized from basaltic magmas at shallow mantle depths (Ulmer, 1989Go). This correction accounts for the post-entrapment crystallization of the inclusions (i.e. overgrowth on the inclusions’ wall). The normalized compositions of the SMIs are shown in Table 8. They show a trend that represents an interval of parental melt evolution. Relationships between the composition and intragranular textural position (core or rim) could not be evaluated, because in most crystals only one inclusion was measurable. CaO concentration shows a strong negative linear correlation with Al2O3 concentration; thus, the CaO/Al2O3 ratio is a good tool to describe melt evolution and major element abundances are plotted as a function of CaO/Al2O3 (Fig. 6).


View this table:
[in this window]
[in a new window]

 
Table 8: Major element composition of the homogenized melt inclusions in olivine of xenolith NNS07 analyzed by EPMA (in wt %)

 

Figure 6
View larger version (14K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 6. Major element variation diagrams for homogenized SMIs ({square}) as a function of CaO/Al2O3 ratio. The bulk composition of the analyzed intercumulus phase (+) is also shown, forming a continuous trend towards low CaO/Al2O3 ratios.

 
The results of the LA-ICPMS analyses of single SMIs in olivine are shown in Table 9 and Fig. 7. The trace element content of the SMIs shows a trend similar to that of the major elements. Concentrations of most incompatible trace elements increase strongly with decreasing CaO/Al2O3 ratio, resulting in a 2–2·5 times enrichment (Fig. 7). Concentrations of Sc and V decrease, and, interestingly, Y is also strongly depleted throughout the melt evolution. In most analyses, HREE were below detection limits. The primitive mantle normalized average composition of the most primitive and most evolved SMIs is shown in Fig. 8. Even the most primitive melt composition shows considerable enrichment in all incompatible elements and a characteristic steep REE pattern.


View this table:
[in this window]
[in a new window]

 
Table 9: Results of LA-ICPMS analyses of individual unheated melt inclusions in olivine

 

Figure 7
View larger version (13K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 7. Trace element variation diagrams for the olivine-hosted SMIs as a function of the CaO/Al2O3 ratio. The bulk composition of the analyzed intercumulus phases is also shown (+), forming continuous trends towards low CaO/Al2O3 ratios.

 

Figure 8
View larger version (10K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 8. Primitive mantle normalized trace element composition of the primitive and the evolved end of the melt compositions shown by the SMIs. For comparison the average composition of the xenolith-hosting alkaline basalts of the region is also shown (Dobosi and Jenner, 1999Go).

 

    DISCUSSION
 TOP
 ABSTRACT
 INTRODUCTION
 GEOLOGICAL SETTING
 PETROGRAPHY OF THE CUMULATE...
 ANALYTICAL TECHNIQUES
 COMPOSITIONS OF THE ROCK-FORMING...
 HOMOGENIZATION EXPERIMENTS AND...
 DISCUSSION
 CONCLUSIONS
 APPENDIX
 REFERENCES
 
Crystallization conditions of the cumulates
The primary goal of this study is to provide an insight into the formation of the olivine-clinopyroxenite cumulate xenoliths and the high-pressure evolution of the parental mafic melts. Therefore, it is important to constrain the PT conditions of these processes and their location in the lithospheric mantle.

Estimation of the pressure and temperature conditions of cumulate formation is possible using our SMI data and the chemical compositions of the xenolith-forming minerals (Table 3). Assuming that the melting temperatures of the daughter phases in the SMIs have not been modified significantly by volatile loss after entrapment, a first-order estimation of crystallization temperature can be achieved based on the heating experiments. The maximum final melting temperatures suggest that clinopyroxene saturation occurred at 1250°C. However, because addition of olivine to the reheated SMIs was necessary to obtain equilibrium melt composition (see above), olivine was probably the first phase to crystallize on the liquidus, implying higher liquidus temperatures. We interpret the presence of the glass (Fig. 3) in the SMIs as a result of incomplete crystallization at depth or melting upon reheating by the host basalt during transport of the xenoliths and quenching after eruption. Initial glass melting temperatures indicate a minimum of 1090°C for the solidus temperature; however, this might be overestimated because of potential H loss from the melt inclusions (Danyushevsky et al., 2002Go).

Clinopyroxene compositions and the SMI data for olivines (Table 8) allow the application of the clinopyroxene–liquid geothermobarometer of Putirka et al. (2003Go), assuming that the phases represent equilibrium compositions, which is likely because of the slow crystallization and is indicated by the homogeneous major element distributions. The equilibrium melt composition was determined from the homogenized olivine-hosted SMIs. A CaO/Al2O3 ratio of 0·7 measured in the most primitive melt inclusions suggests that crystallization of significant amounts of clinopyroxene did not happen before their entrapment (see below). Thus the primitive melts (average composition of the three most primitive olivine-hosted homogenized SMIs with the highest CaO/Al2O3 ratios) were assumed to be in equilibrium with clinopyroxene cores, whereas the most evolved compositions (average composition of the olivine-hosted homogenized SMIs with CaO/Al2O3 ≤ 0·45) were thought to be in equilibrium with reaction rims of the clinopyroxenes. This assumption has some uncertainties, especially for the estimates based on the clinopyroxene core compositions, as these probably re-equilibrated to some extent with later residual melts via diffusion. The method is, however, more sensitive to variation in the jadeite component of clinopyroxene than to the variability in the equilibrium melt composition. For the calculations of the clinopyroxene components, the method of Putirka (1999Go) was used. Calculated pressure and temperature values are shown in Table 3 along with the clinopyroxene compositions. The average uncertainty of the method is ± 33°C and ± 1·7 kbar (Putirka et al., 2003Go). This might be higher for the PT estimations based on clinopyroxene cores, because of the uncertainty in the equilibrium melt composition. The estimated crystallization temperature for clinopyroxene cores ranges from 1185 to 1200°C. This is in good agreement with the temperature data for clinopyroxene final melting in the olivine-hosted SMIs but lower than the clinopyroxene saturation temperatures obtained by pMELTS modelling (see below). The calculated crystallization pressures for clinopyroxene cores are between 10 and 12 kbar. The reaction rims indicate lower crystallization temperatures (1120–1150°C) than the cores; however, the estimated pressure values are indistinguishable within the error of the method (8·7–12·0 kbar). These values suggest that crystallization of the cumulate bodies happened at a depth of ~ 29–38 km, in the uppermost lithospheric mantle close to the Moho (Huismans et al., 2001Go) without significant ascent in the recorded cooling and crystallization interval.

We estimated the oxygen fugacity during crystallization using the method of Ballhaus et al. (1991Go) based on the composition of small Cr-spinel inclusions in olivine. The values scatter between QFM + 1· 0 and QFM + 1·23 log units (where QFM is the quartz–fayalite–magnetite buffer).

Do the melt inclusions represent the composition of the parental melt?
The physical conditions of crystallization of the cumulate bodies near the Moho and their cooling rates are different from those of basaltic volcanic systems from which most previous studies of SMIs in olivine have been reported. Most of these earlier studies were carried out on compositionally zoned olivine phenocrysts formed by fractional crystallization in volcanic systems. In our case the homogeneous composition of the host olivine crystals indicates equilibrium crystallization accompanied by slow cooling characteristic of underplated igneous bodies. Iron loss, suggested for melt inclusions hosted in magmatic olivine phenocrysts (Danyushevsky et al., 2000Go), may not be significant here because of the equilibrium crystallization of the host olivine. On the other hand, the FeO/MgO ratio in the SMIs is considered to be identical to the latest melt composition, which was in equilibrium with the host olivine. In this scenario, it is assumed that the SMIs maintain equilibrium with the olivine host composition, which maintains equilibrium with the residual melt. Non-representative remelting of olivine crystallized on the inclusion walls during the reheating experiments or crystallization on the walls upon quenching can also induce significant modifications in the melt compositions. However, these can be properly corrected for, using Fe/Mg exchange equilibria between the host olivine and the entrapped melt (Table 8). In contrast, all those major and trace elements that are incompatible in the host olivine and have low diffusion coefficients are likely to be representative of the original melt composition. As data from inclusions smaller than 15 µm were not considered and crystallization should be slow, boundary layer effects are unlikely to play a significant role (Kuzmin & Sobolev, 2004Go).

Composition of the parental melt
We estimated the composition of the parental melt by fitting trend lines to the SMI compositions on major and trace element plots and taking the values at a CaO/Al2O3 ratio of 0·7, which is the characteristic value for the most primitive SMIs. The calculated composition of the most primitive melt preserved in the SMIs is given in Table 10 and Fig. 8. As SMIs are absent from the cores of the large olivine grains, it is likely that the most primitive melt represented by the SMIs had already undergone some olivine and potentially clinopyroxene fractionation. This agrees well with the fact that this melt composition is relatively high in Al2O3 and low in MgO compared with most experimental melts produced by partial melting of garnet lherzolite (Table 10). Crystallization of a significant amount of clinopyroxene before entrapment of the most primitive SMIs is unlikely, as melting of clinopyroxene in the SMIs terminates before remelting of an adequate quantity of olivine from the inclusion walls. Addition of 15 wt % olivine increases the mg-number of the melt to 74, a value expected of a primitive mantle melt in equilibrium with olivine having a mg-number of 91 at 3 GPa using the Fe–Mg exchange coefficient (Kd = 0·36 at 3 GPa) of Ulmer (1989Go). Before the addition of olivine we calculated the characteristic Fe2O3/FeO ratio of the melt using the equation of Kress & Carmichael (1991Go) and used only the FeO content for the mg-number calculation. The resulting primitive parental melt is basanitic (43·6 wt % SiO2, mg-number 74) with relatively low total alkali concentrations (4·91 wt % Na2O + K2O) (Table 10). Similar results can be achieved by adding both olivine and clinopyroxene; however, in this case the CaO concentration becomes too high, whereas the MgO content remains too low compared with the experimental melts of Walter (1998Go).


View this table:
[in this window]
[in a new window]

 
Table 10: Composition estimations for the parental melt of the cumulate xenoliths

 
Measurement of the water content of the parent melt is not possible because of the slow crystallization, long residence time at high temperatures and reheating during transportation to the surface, which could have allowed significant H loss from the SMIs by diffusion through the host mineral (Danyushevsky et al., 2002Go). However, crystallization of amphibole in the SMIs and as a cumulate xenolith-forming phase indicates that a significant amount of water was present in the melt. Compatibility of H2O during mantle melting is similar to that of P, Ce and Nd (Aubaud et al., 2004Go; Bolfan-Casanova, 2005Go), which would allow us to calculate the H2O content of the melt if the H2O content of the mantle source was known. However, there are limited data available on the water content of the subcontinental asthenosphere. Embey-Isztin et al. (1993Go) and Dobosi et al. (1995Go) have shown that Plio-Pleistocene alkaline basalts of the Pannonian–Carpathian region are derived from an ocean-island basalt (OIB)-like asthenospheric mantle source. Michael (1988Go) pointed out that H2O/(Ce + Nd) ratios for mid-ocean ridge basalt (MORB) and OIB vary in a narrow range of 95 ± 30. For OIB from Loihi this ratio varies between 107 and 125 (Michael, 1988Go). Simons et al. (2002Go) drew similar conclusions, suggesting a typical H2O/Ce ratio of 210 ± 20 for a mantle plume and 150 ± 10 for a depleted MORB mantle (DMM) component. Minor variations in the above-mentioned elemental ratios between contrasting mantle sources allow their use to estimate the water content of the parental melt of the cumulates assuming an OIB-like mantle plume component as the source. The estimated water content is between 0·73 and 0·85 wt % based on the H2O/(Ce + Nd) ratios of Michael (1988Go) and between 0·74 and 0·9 wt % using H2O/Ce ratios of Simons et al. (2002Go).

The presence of CO2 fluid inclusions and heterogeneous entrapment of silicate melt and CO2 in clinopyroxenes (Fig. 3d) indicates CO2 saturation during crystallization. Considering that primary CO2 inclusions are present in crystal cores, CO2 saturation must have occurred close to the temperature of clinopyroxene saturation. The presence of a large amount of heterogeneously entrapped CO2 fluid–silicate melt inclusions in clinopyroxene and lack of CO2 fluid inclusions in olivine indicates that nucleation of the exsolving CO2 happened on the surface of the crystallizing clinopyroxene. According to the volatile saturation model of Dixon (1997Go) for alkali basalts with 0·8 wt % water, a CO2/H2O ratio of approximately three is required to reach volatile saturation at the estimated crystallization pressure of the cumulates. This implies CO2 contents between 2·20 and 2·55 wt % for the parental melt.

Source of the parental melt
Conditions of parental melt formation can be estimated based on the composition of the SMIs. We modelled the trace element composition of partial melts of a garnet lherzolite and a spinel lherzolite produced by variable degree of partial melting. We selected Nb, Y, Zr and Nd for the model because the analytical uncertainty for these elements is relatively small in the SMI measurements (Table 9). Furthermore, these elements are thought to be insensitive to any potential fluid metasomatism in the source. The asthenospheric upper mantle is inferred to be fertile (i.e. not DMM), based on the isotopic composition of the alkali basalts of the region (Embey-Isztin et al., 1993Go; Dobosi et al., 1995Go; Szabó et al., unpublished data). We used the primitive mantle composition of McDonough & Sun (1995Go) to represent the bulk composition of the source and an internally consistent mineral–melt trace element partitioning dataset from garnet lherzolite melting experiments at pressures between 2·8 and 3·2 GPa (Salters et al., 2002Go) to model the trace element concentrations in the partial melts. We assumed equilibrium partial melting. Other parameters and results of the model are shown in Fig. 9. Yttrium is compatible in garnet and slightly incompatible in clinopyroxene; consequently, the Y content of the melt is a good indicator of the presence of garnet in the source, but is less affected by the degree of partial melting. Niobium is highly incompatible in all the source minerals, thus it is a good indicator of the degree of partial melting; however, it is not sensitive to the phase proportions in the source. We plotted the calculated primitive parent melt composition on a Nb vs Y diagram (Fig. 9). This shows the degree of partial melting and the garnet/clinopyroxene ratio in the source. To eliminate the uncertainty on the absolute concentrations caused by olivine crystallization before SMI entrapment, we plotted the Nb/Y ratio as a function of the Nd/Zr ratio. As a consequence of the high incompatibility of these elements in olivine, its crystallization does not influence their ratios in the residual melt.


Figure 9
View larger version (18K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 9. Petrogenetic model calculations. Modelled compositions of partial melts of hypothetical spinel and garnet lherzolites characterized by primitive mantle compositions are shown as a function of the degree of partial melting and the clinopyroxene/garnet ratio in the source. Values in per cent show the degree of partial melting; italic numbers indicate the clinopyroxene/garnet ratio in the hypothetical source. The compositions of the parental melt of the cumulates and the host alkali basalts are also plotted to allow comparison with the modelled partial melt compositions. Arrows in the Nb vs Y diagram (a) show the primitive composition of the host basalt.

 
The model indicates partial melting of a garnet lherzolite source with a clinopyroxene/garnet ratio between three and seven. The estimated degree of partial melting is ~ 2%.

We compared the major element composition of the melt with the experimental data of Walter (1998Go) for melting of garnet peridotite. The major element concentrations in the parental melt best resemble experimental melts produced by garnet lherzolite melting close to 3 GPa (Table 10). Melting at higher pressures would result in higher MgO and FeO, but significantly lower Al2O3 and slightly lower CaO concentrations because of the increasing proportion of olivine in the melting mode at the expense of clinopyroxene and garnet. The MgO content is still slightly lower and Al2O3 content slightly higher in our modelled parental melt composition than in the experimental melt produced at 3 GPa. This might indicate that formation of the melt occurred at even lower pressures, but still in the garnet stability field. The lowest melt fractions produced by Walter (1998Go) are around 10%, which is a much higher degree of partial melting than that proposed here. This explains the significantly lower incompatible major element (K, Na, Ti) concentrations in the experimental melts, but does not affect significantly the concentrations of Al2O3, FeO, and MgO, which are buffered by the equilibrium mineral assemblage.

Evolution of the melt
Based on the textural observations (Fig. 2a) and the heating experiments on the SMIs, the first phase to appear on the liquidus is olivine, shortly followed by cotectic crystallization of clinopyroxene. The first melt inclusion entrapment may have occurred when ~ 15% of the parental magma had crystallized, as 15% olivine fractionation is required to produce the composition of the most primitive melt inclusions from a primitive mantle melt (Table 10). At this stage, olivine was still the dominant phase in the crystallizing mineral assemblage. Subsequently, cotectic equilibrium crystallization of clinopyroxene and olivine dominated the melt evolution until a late stage when clinopyroxene started to react with the residual liquid to form the characteristic TiO2-rich rims and olivine started to be resorbed (Fig. 2b and c). The clinopyroxene/olivine ratio during crystallization along the olivine–clinopyroxene cotectic must have been high, as the modal proportion of olivine ranges from 20 to 40 vol. % and 15 vol. % olivine must have been crystallized before the onset of clinopyroxene crystallization. Because the most evolved SMIs in olivine have a CaO/Al2O3 ratio of 0·3 and are approximately two times enriched in the most incompatible trace elements (Nb, Ta, Rb, Ba) (Fig. 7), we suggest that olivine began to react with the residual melt only after ~60 vol. % of the melt had crystallized. Amphibole is an interstitial phase in the xenoliths (Fig. 2f); therefore, its saturation must have happened at a late stage of the melt evolution. In the final stages of crystallization, alkali feldspar, Ti-rich clinopyroxene, apatite and titanomagnetite crystallized from the residual melt (Fig. 2e), leading to an extremely alkaline and Al2O3-rich residual melt. The presence of compositionally homogeneous glass in the intercumulus phase (Fig. 2d and Table 6) suggests that either there was some melt still present at the time of sampling by the host alkali basalt or the xenoliths were partially melted by the host basalts.

We modelled the evolution of the estimated primitive parental melt using the pMELTS algorithm of Ghiorso et al. (2002Go), assuming equilibrium crystallization at 1 GPa and a value of 0·8 wt % for the initial water content. According to the model olivine appears on the liquidus at 1335°C, and clinopyroxene saturation occurs at 1280°C. Spinel appears at 1220°C. The interval between 1280°C and 1040°C is dominated by clinopyroxene crystallization, simultaneous with precipitation of a minor amount of olivine. At 1040°C Ti-rich clinopyroxene starts to crystallize, followed by amphibole at 1020°C. During amphibole and Ti-rich clinopyroxene crystallization, the amount of olivine decreases slightly and the amounts of normal Ti-poor clinopyroxene and spinel decrease significantly via reaction with the residual melt (Fig. 10). The pMELTS model predicts biotite saturation before the appearance of amphibole; however, biotite was not found in the xenoliths. Thus, we excluded it from the modelled stable phase assemblage. Phase compositions predicted by pMELTS for clinopyroxene have a jadeite component higher than the clinopyroxene cores in the cumulates (Table 11). The low-temperature clinopyroxene has a much higher Ti content than the Ti-rich reaction rims of cumulate clinopyroxene. Assuming equilibrium crystallization, both modelled olivine and clinopyroxene have a lower mg-number close to the solidus than those in the cumulate xenoliths. Furthermore, the clinopyroxene/olivine ratio seems to be slightly overestimated in the crystallizing phase assemblage along the clinopyroxene–olivine cotectic. The high mg-number of olivine and clinopyroxene suggests that they were not in equilibrium with the residual melt throughout the crystallization sequence, but might have been spatially separated from the main body of the residual liquid at some stage. This and potential errors in the model may explain the other discrepancies. The modelled residual melt at 80 vol. % crystallization (reached at 980°C) is characterized by extremely high Al2O3 (21·4 wt %) and total alkali oxide (16·3 wt %) and high SiO2 (54·7 wt %) similar to the bulk composition of the intercumulus phase assemblage, but is depleted in CaO, MgO and FeO to an even greater extent (Table 11). Therefore, it represents the separated residual melt, except for compatible elements, which could have been buffered by the surrounding peridotite in the main residual melt body. Re-equilibration of the residual melt with peridotite could also be a possible reason for the higher mg-number of the ferromagnesian phases than the values predicted by closed-system crystallization modelling. Below 980°C the program fails to calculate equilibrium. We conclude that the major trends of melt evolution provided by pMELTS, assuming equilibrium crystallization, are generally in good agreement with the textural–mineralogical observations; however, clinopyroxene seems to be slightly over-stabilized and there are significant differences in the predicted mineral chemistry. Although the model reproduces the major characteristics of the analytically determined evolutionary trends in melt chemistry, our model assumptions of a closed system and complete equilibrium, and imperfections of the thermodynamic model, may result in significant uncertainties in the numerical values.


Figure 10
View larger version (13K):
[in this window]
[in a new window]
[Download PowerPoint slide]
 
Fig. 10. Results of modelling with pMELTS showing the quantity of the minerals crystallized from 100 g of initial melt as a function of temperature (clinopyroxene1, Ti-poor cores; clinopyroxene2, Ti-rich reaction rims).

 

View this table:
[in this window]
[in a new window]

 
Table 11: Phase compositions at 1020°C obtained from pMelts modelling

 
Garnet precipitation from the parental melt?
Depletion in Y and HREE displayed by both the evolutionary trends in the melt inclusions (Fig. 7) and the trace element zonation in the clinopyroxene (Fig. 4) suggests crystallization of a mineral phase in which these elements are compatible. The most obvious explanation would be the primary magmatic crystallization of garnet. Experimental studies in alkaline basaltic melts, however, have reported stability of garnet only at pressures higher than the estimated pressure conditions at which these cumulates were formed (Green & Ringwood, 1964Go; Green, 1969Go, 1973Go). Amphibole could be a potential host for middle REE (MREE) and HREE, yet most reported partition coefficients in the literature are below one even for these elements (Downes et al., 2004Go). Furthermore, amphibole appears only at a late stage of crystallization and does not display elevated HREE and Y contents (Fig. 2f and Table 5). Liu et al. (2005Go) reported the presence of garnet-pyroxenite veins from the lithospheric mantle of the North China craton. They interpreted the garnet-pyroxenites to have formed through reaction of an incompatible element-rich silicic melt and mantle lherzolite. A similar scenario appears plausible in the present case, although relicts of garnet-bearing reaction products were not sampled by the host alkaline basalts or have not been found yet. Decreasing HREE and Y content could potentially also be explained by the interaction with garnet-bearing lower crustal granulites, the former (probably pre-Middle-Miocene) presence of which has been reported by Kovács & Szabó (2005Go) from the same volcanic field. However, Hermann et al. (2001Go) did not observe depletion of HREE and Y in basaltic melts intruded into lower crustal granulites. Furthermore, geobarometric calculations and the absence of plagioclase in the mineral assemblage of the clinopyroxenites indicate higher pressures than those characteristic for the lowermost crust of the study region during and after basin extension in Middle Miocene–Pliocene (~0·8–0·9 GPa) (Horvath, 1993Go).

The fate of the residual melt
A major question concerns the fate of the residual melt left behind after cumulate formation. The most likely possibilities are consumption by reaction with the surrounding rocks, in situ crystallization, or migration to higher levels of the lithosphere. These will now be considered in turn.

Interaction between host rock and evolved alkaline melts as metasomatic agent
SiO2-undersaturated alkaline mafic magmas intruded into shallow levels of the upper mantle close to the Moho are not in equilibrium with their spinel peridotite host rock. Experimental studies have shown that melts rich in alkalis tend to have high SiO2 contents as a result of the decreasing activity coefficient of silica in the melt at buffered (opx–ol assemblage) silica activity (Baker et al., 1995Go; Hirschmann et al., 1998Go). Thus, the first reaction to take place is the consumption of some orthopyroxene in the mantle peridotite. A similar metasomatic interaction between peridotite and basanitic melt, involving dissolution of orthopyroxene simultaneous with clinopyroxene and olivine crystallization, has been described by Shaw (1999Go) and Shaw et al. (2005Go). This reaction is probably most intense during the early stages of interaction, resulting in an increasing SiO2 content in the residual melt of the cumulates, regardless of the crystallizing mineral assemblage (i.e. clinopyroxene/olivine ratio). Rapid Fe–Mg exchange between mantle olivine and the melt (Gaetani & Watson, 2002Go; Shaw, 2004Go) might partially buffer the MgO/FeO ratio in the residual melt, concurrently generating less magnesian olivine in the mantle peridotite. The extent of melt–mantle peridotite interaction depends on the unknown spatial dimensions of the intruded melt bodies (surface/volume ratio); however, enrichment of SiO2 in the early stage of melt evolution appears somewhat faster than expected from closed-system equilibrium crystallization modelling, and the mg-number of the ferromagnesian phases is slightly higher than expected (Fig. 6, Table 11). As shown above, with an increasing degree of crystallization, an extremely alkaline and Al2O3-rich, but CaO-, MgO- and FeO-poor melt is generated as a result of the absence of plagioclase in the mineral assemblage. As a consequence of early volatile saturation, the H2O/CO2 ratio and the total amount of volatiles dissolved in the melt also increased. These late-stage residual melts probably migrate through the host peridotite and may be effective metasomatic agents, probably forming amphiboles and potentially alkali feldspar. Amphibole-bearing lherzolite xenoliths are commonly found in alkali basalts and were also found within the NGVF region (Szabó & Taylor, 1994Go; Szabó et al., 1996Go; Kovács et al., 2004Go). Amphiboles in peridotite xenoliths from Fil’akovo (NFL) show the same trace element pattern as amphiboles in the cumulate xenoliths (Fig. 5). Thus, metasomatism of the source of the peridotite xenolith is likely to be caused by a residual melt of a crystallizing basanitic melt body similar to that considered here. This suggests that amphibole can crystallize directly from the parental melt of the cumulates during its late-stage crystallization, at the expense of previously crystallized clinopyroxene (autometasomatism), and also that late-stage residual melts of underplated mafic alkaline magmas can act as amphibole-forming metasomat