Journal of Petrology Advance Access originally published online on September 25, 2006
Journal of Petrology 2006 47(12):2405-2431; doi:10.1093/petrology/egl049
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Phase Equilibria of the Lyngdal Granodiorite (Norway): Implications for the Origin of Metaluminous Ferroan Granitoids
1 U.R. PÉTROLOGIE ET GÉOCHIMIE ENDOGÈNE (B20), DÉPARTEMENT DE GÉOLOGIE, UNIVERSITÉ DE LIÈGE B-4000 SART TILMAN, BELGIUM
2 ISTO-CNRS UMR 6113, 1A RUE DE LA FÉROLLERIE, ORLÉANS, 45071 CEDEX 2, FRANCE
RECEIVED JANUARY 31, 2005; ACCEPTED AUGUST 16, 2006
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONGEOLOGICAL BACKGROUNDEXPERIMENTAL AND ANALYTICAL...RESULTSCRYSTALLIZATION CONDITIONS OF...COMPARISON WITH OTHER...DISCUSSIONCONCLUSIONSREFERENCES |
|---|
The Proterozoic (950 Ma) Lyngdal granodiorite of southern Norway belongs to a series of hornblende–biotite metaluminous ferroan granitoids (HBG suite) coeval with the post-collisional Rogaland Anorthosite–Mangerite–Charnockite (AMC) suite. This granitoid massif shares many geochemical characteristics with rapakivi granitoids, yet granodiorites dominate over granites. To constrain both crystallization (P, T, fO2, H2O in melt) and magma generation conditions, we performed crystallization experiments on two samples of the Lyngdal granodiorite (with 60 and 65 wt % SiO2) at 4–2 kbar, mainly at fO2 of NNO (nickel–nickel oxide) to NNO + 1, and under fluid-saturated conditions with various H2O–CO2 ratios for each temperature. Comparison between experimental phase equilibria and the mineral assemblage in the Lyngdal granodiorite indicates that it crystallized between 4 and 2 kbar, from a magma with 5–6 wt % H2O at an fO2 of NNO to NNO + 1. These oxidized and wet conditions sharply contrast with the dry and reduced conditions inferred for the petrogenesis of the AMC suite and many other rapakivi granites worldwide. The high liquidus temperature and H2O content of the Lyngdal granodiorite imply that it is not a primary magma produced by the partial melting of the crust but is derived by the fractionation of a mafic magma. Lyngdal-type magmas appear to have volcanic equivalents in the geological record. In particular, our results show that oxidized high-silica rhyolites, such as the Bishop Tuff, could be derived via fractionation of oxidized intermediate magmas and do not necessarily represent primary crustal melts. This study underlines the great variability of crystallization conditions (from anhydrous to hydrous and reduced to oxidized) and petrogenetic processes among the metaluminous ferroan magmas of intermediate compositions (granodiorites, quartz mangerites, quartz latites), suggesting that there is not a single model to explain these rocks.
KEY WORDS: ferroan granitoids; crystallization conditions; experiments; Norway; Sveconorwegian; Bishop Tuff
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGICAL BACKGROUNDEXPERIMENTAL AND ANALYTICAL...RESULTSCRYSTALLIZATION CONDITIONS OF...COMPARISON WITH OTHER...DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Metaluminous ferroan granitoids [using the terminology of Frost et al. (2001)
] have been reported from many different areas and geological epochs, but are particularly common in mid-Proterozoic terranes. In particular, the magmatic association characterized by massif anorthosites and granitoids is restricted to the Proterozoic and is commonly referred to as an AMCG complex (acronym of Anorthosite–Mangerite–Charnockite–Granite). This study focuses on such a petrogenetic association. The majority of granitoids associated with massif anorthosites can be classified into two main types: (1) hypersthene-dominant granitoids, i.e. quartz mangerites (Opx monzonites)–charnockites (Opx granites); (2) hornblende–biotite granitoids. Some of these latter plutons locally display rapakivi textures and then are called rapakivi granites. Transitional facies, with orthopyroxene rimmed by hornblende, exist in a number of complexes (e.g.Emslie et al., 1994). Both types of plutons are metaluminous ferroan granitoids but show a large variation in their MALI index [(Na2O + K2O) – CaO in wt %], from calc-alkalic to alkalic as indicated in Fig. 1. All the granitoids and volcanic rocks discussed in this paper (Fig. 1) are metaluminous with ASI 
1 [molar ratio of Al2O3/(CaO – 3·3P2O5 + Na2O + K2O)] and agpaitic index AI < 1 [molar ratio (K2O + Na2O)/Al2O3]. Ferroan peralkaline intrusions (e.g. the sodic series of the Pikes Peak batholith: Smith et al., 1999) are not considered here.
|
In many AMCG complexes, ilmenite is the dominant or sole Fe–Ti oxide and mineral equilibria studies suggest that both types of pluton crystallized from low fO2 and H2O-poor magmas (e.g. Frost & Frost, 1997
). Two main models have been proposed to explain the link between anorthosite massifs and associated granitic plutons in the AMCG complexes from North America (e.g. Laramie anorthositic complex, Nain Province). Emslie et al. (1994) proposed that the anorthosites ultimately formed from a mantle-derived magma differentiating in the lower crust, whereas the granitic magmas were produced by melting of the lower crust. In their model, there is no parent–daughter relationship between anorthosite and granite magmas and a wide range of felsic magmas associated with the anorthosite massifs is expected. The second model relies, in part, on the observation that granitoids from AMCG complexes usually have similar bulk-rock compositions and crystallization conditions (Frost & Frost, 1997; Frost et al., 2002; Anderson et al., 2003). According to this model the reduced and H2O-poor metaluminous ferroan granitoids are ultimately derived from tholeiitic magmas, from which the massif anorthosites are proposed to crystallize (the so-called ‘tholeiite connection’); the acid magmas are considered to be either differentiation products of parental basalts, or, and more likely, partial melts resulting from the anatexis of the differentiated products of such underplated basalts.
In both models, the implicit assumption is that the elevated iron content of the melt is due to the low fO2 prevailing in the source, which prevents Fe–Ti oxide crystallization during the early stages of differentiation as in arc magmas series (Martel et al., 1999). However, there exist a number of metaluminous ferroan granitoids that, despite having elevated iron contents, display a significantly higher redox state than those ferroan granitoids of clear tholeiitic affinity (e.g. Dall'Agnol et al., 1999). In the tholeiite model, one way to produce the oxidized granitoids contemporaneous with the reduced ones is through contamination of the parental magma by oxidized crust (e.g. Frost & Frost, 1997; Anderson et al., 2003). However, Anderson & Morrison (2005) have observed variations of fO2 among the ferroan granitoids at the scale of a continent, which they related to the diversity in fO2 of their source in the lower crust. Such a large-scale trend suggests that oxidized, metaluminous ferroan granitoids may be disconnected from reduced granitoids.
We base our work on the well-known geological setting of SW Norway, in which both types of metaluminous ferroan granitoids (opx- and hornblende–biotite granitoids) and anorthosite massifs have been documented within the Precambrian basement (e.g. Duchesne et al., 1985; Duchesne & Wilmart, 1997; Vander Auwera et al., 2003). These magmatic rocks were emplaced during the post-collisional stage of the orogeny between 
960 and 920 Ma. Previous experimental studies relevant to this area (Vander Auwera & Longhi, 1994; Vander Auwera et al., 1998; Longhi et al., 1999) have only considered the origin of the opx-bearing rocks, i.e. the anorthosites and their potential derivatives such as charnockites (hereafter referred to as the AMC suite for Anorthosite–Mangerite–Charnockite). These studies have shown conclusively that in southern Norway the petrogenesis of the AMC suite can be modelled by fractional crystallization of a dry jotunitic (opx-monzodiorite) parent magma evolving under low fO2 conditions.
As noted above, the hornblende–biotite granitoids, although compositionally similar to charnockites, exhibit significant differences in mineralogy, most notably in the abundance of amphibole and magnetite (Bogaerts et al., 2003a; Vander Auwera et al., 2003). The main goal of our experimental work is to constrain the crystallization conditions (P, T, fO2, H2O in melt) of these hornblende–biotite granitoids. Such constraints are important to evaluate petrogenetic models for these granitoids in southern Norway.
We have performed phase equilibrium experiments on two representative samples from the Lyngdal granodiorite, a metaluminous ferroan hornblende–biotite granitoid associated with the Rogaland–Vest Agder anorthosite massif in southwestern Norway. Our results, combined with previous studies on the AMC suite (Vander Auwera & Longhi, 1994; Duchesne & Wilmart, 1997; Vander Auwera et al., 1998; Bolle et al., 2003), suggest that the metaluminous ferroan granitoids of southwestern Norway may be derived from a variety of sources with contrasted redox states and H2O contents.
Besides their regional implications, our experimental data show that some present-day oxidized high-silica rhyolites can be produced via crystal fractionation of intermediate metaluminous ferroan magmas. The data also provide a framework for a comparison with the voluminous quartz latite ash flows associated with flood basalt events such as those of the Etendeka–Paraná Province, which exhibit a number of similarities to the intermediate ferroan magmas of the AMCG complexes (Table 1, Fig. 1). Our results confirm the complexity of petrogenetic processes at work in producing ferroan magmas.
|
|
GEOLOGICAL BACKGROUND |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGICAL BACKGROUNDEXPERIMENTAL AND ANALYTICAL...RESULTSCRYSTALLIZATION CONDITIONS OF...COMPARISON WITH OTHER...DISCUSSIONCONCLUSIONSREFERENCES |
|---|
The Sveconorwegian orogen (1200–900 Ma), which forms the southwestern part of the Baltic Shield, is made up of several Paleo-Mesoproterozoic lithotectonic domains separated by major lineaments (Bingen et al., 2001
). From east to west (inset of Fig. 2), these domains are the parautochthonous segment (the Eastern Segment) and two allochthonous terranes: the Idefjorden and Telemarkia terranes; the Bamble–Kongsberg (B–K) sector is interpreted as an early Sveconorwegian collision zone between these two terranes (Bingen et al., 2005). Assemblage and stacking of these terranes occurred during the main phase of the Sveconorwegian orogeny accompanied by high-grade regional metamorphism (amphibolite to granulite facies) between 1030 and 970 Ma (Bingen & van Breemen, 1998; Bingen & Stein, 2003) in the SW of the Telemarkia terrane (the Rogaland–Vest Agder sector; Bingen et al., 2005) and between 980 and 970 Ma in the Eastern Segment (Möller, 1998; Johansson et al., 2001). This regional metamorphism was followed by extensive post-collisional magmatism represented by numerous plutons, which preserve their original magmatic paragenesis (no metamorphic overprint or alteration) and can be divided on a petrographic basis into two groups: (1) hornblende–biotite granitoids (Vander Auwera et al., 2003); (2) plutons belonging to the charnockitic series (e.g. Duchesne & Wilmart, 1997). The former group is present in all the allochthonous terranes, whereas the latter group is restricted to the southwestern outcropping part of the Telemarkia terrane forming the Rogaland anorthosite province (Duchesne et al., 1985). In this study, we restrict our discussion to the Sveconorwegian post-collisional magmatism of the Rogaland anorthositic province and to the hornblende–biotite plutons cropping out in the Telemarkia terrane (Lyngdal, Tranevåg, Holum, Svöfjell, Valle, Rustfjellet, Verhuskjerringi; Fig. 2). Magmatic rocks from the Rogaland anorthositic province (Fig. 1) form the AMC suite (Anorthosite–Mangerite–Charnockite), whereas the associated hornblende–biotite granitoids are termed the HBG suite (Hornblende–Biotite Granitoids), following Vander Auwera et al. (2003). The large hornblende–biotite intrusions of the HBG suite define a differentiation trend from 56 to 77 wt % SiO2 (Vander Auwera et al., 2003). Small (<5 km2), less differentiated intrusions (Handeland–Tveit and Åseral quartz diorites–monzodiorites: SiO2 between 50 and 61 wt %) were also included by Vander Auwera et al. (2003) in the HBG suite (Fig. 1).
|
The anorthosite complex suite was emplaced at c. 930 Ma (Schärer et al., 1996
) whereas the Lyngdal granodiorite was emplaced at 950 ± 5 Ma (Pasteels et al., 1979). Zircon dating suggests that the entire HBG suite was emplaced between 960 and 920 Ma (Andersen et al., 2002; Bingen et al., 2006). The Farsund granitoid, dated at 936 Ma, crops out between the AMC suite and the Lyngdal massif, and shows mingling relations between the AMC and HBG suites (Dupont & Vander Auwera, 2002; Dupont et al., 2005, and unpublished data); thus the AMC and HBG suites can be considered coeval.
The Lyngdal granodiorite
The Lyngdal massif is a large (300 km2) granodioritic intrusion (SiO2 60–65 wt %: Bogaerts et al., 2003a) and is the nearest HBG intrusion to the exposures of the AMC suite (Fig. 2). Its main minerals are plagioclase, K-feldspar, quartz, amphibole, biotite and clinopyroxene (found as inclusions in amphibole and plagioclase). Accessory minerals are Fe–Ti oxides, apatite, zircon, titanite and allanite. The rocks are usually porphyritic, with plagioclase and K-feldspar phenocrysts (average length 1 cm, with a maximum of 2 cm in a few samples) included in a coarse-grained groundmass (average mineral size 2–3 mm; Fig. 3a), consisting of all other minerals, as well as plagioclase and K-feldspar. Plagioclase phenocrysts (anhedral to subhedral) have the same composition as plagioclase in the groundmass and contain inclusions of Fe–Ti oxides and rare clinopyroxenes. They are weakly zoned from An35 (andesine core) to An22 (oligoclase rim). Myrmekite is commonly developed along the boundaries between plagioclase and K-feldspar. The K-feldspar (in the groundmass or as a phenocryst) is perthitic orthoclase or microcline. K-feldspar phenocrysts (Fig. 3c and d) can contain all other mineral phases as inclusions, suggesting their late crystallization. This is a common feature in granitoids, as discussed by Vernon (1986); the large size of the K-feldspars is probably due to their difficulty to nucleate.
|
Quartz forms large amoeboid grains around the other minerals. Amphibole is the dominant mafic mineral in Lyngdal but its modal proportion relative to biotite decreases with increasing bulk-rock SiO2 content. Amphibole appears as anhedral poikilitic grains, containing Fe–Ti oxides and accessory minerals (apatite, zircon, allanite, titanite) or as sub-to euhedral grains (Fig. 3a), sometimes included in plagioclase, and in some samples amphibole contains a core of clinopyroxene (Fig. 3b). All amphiboles have (Ca + Na)M4 > 1·34 and (Na)M4 < 0·67 and are thus calcic amphiboles (Leake, 1978
; Leake et al., 1997). Their composition is relatively homogeneous in a given sample but exhibits some small variations between samples (edenitic hornblendes and magnesio-hastingites). The cationic ratio XFe [Fe/(Fe + Mg)] of the amphiboles varies between 0·47 and 0·58 and slightly increases with the SiO2 content of the whole-rock. The Al2O3 content varies between 9·6 and 11·1 wt %. Biotite is anhedral to euhedral and scarce in the less differentiated samples (60 wt % SiO2) but becomes more abundant with increasing SiO2 content. Its XFe varies between 0·47 and 0·51. These XFe values are similar to those in biotite in other oxidized Proterozoic ferroan granitoids (e.g. Jamon granite, Dall'Agnol et al., 1999). Apatite is abundant and forms large euhedral crystals (up to 5 mm); it is generally included in all other minerals. Zircon is also abundant, forming large euhedral crystals (up to 8 mm), and occurs as inclusions in all the main minerals. Titanite forms anhedral to subhedral grains and is commonly found as coronas around Fe–Ti oxides. Allanite is generally a metamict phase in biotite.
The petrographic observations suggest the following sequence of crystallization: apatite, ilmenite, magnetite and clinopyroxene are probably the liquidus minerals and are followed by amphibole and plagioclase, which are interpreted as crystallizing nearly together with zircon. Biotite appears later and is followed by quartz, potassic feldspar and titanite. Pressures calculated using the Al-in-hornblende geobarometer (Johnson & Rutherford, 1989a) are between 3·6 and 5·1 kbar.
|
EXPERIMENTAL AND ANALYTICAL PROCEDURES |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGICAL BACKGROUNDEXPERIMENTAL AND ANALYTICAL...RESULTSCRYSTALLIZATION CONDITIONS OF...COMPARISON WITH OTHER...DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Two samples from the Lyngdal granodiorite (Table 1) were selected for the experimental study, 98N50 (
60 wt % SiO2) and 98N06 (65 wt % SiO2), which are the least and most differentiated facies of the pluton, respectively (Bogaerts et al., 2003a). Several lines of evidence suggest that these samples can be considered as representative of the original magmas. First, their rare earth element (REE) patterns display slight negative Eu anomalies, an observation that is not in favour of plagioclase accumulation. Accumulation of early crystallized apatite/zircon would have enriched the magma in P2O5/Zr and produced unrealistic high apatite–zircon saturation temperatures. These saturation temperatures are in the range of 1000–800°C, close to the liquidus temperatures of ferroan granites (e.g. Dall'Agnol et al., 1999), precluding significant apatite and zircon accumulation (Bogaerts et al., 2003a; Vander Auwera et al., 2003). Second, despite its porphyritic character, the Lyngdal massif does not show structural evidence of crystal accumulation and layering, as observed in some other plutons (e.g. Barbey et al., 2001). On the contrary, fine-grained samples collected near the margins of the pluton have compositions similar to the more porphyritic ones (sample MB99N32, Table 1). Sample 98N06 is similar in composition to silicic volcanic rocks (e.g. quartz latites from the Etendeka–Paraná Province: Table 1 and discussion below), which are considered to represent liquid compositions (Ewart et al., 1998). This observation does not definitively prove that sample 98N06 is representative of the Lyngdal parent magma but at least shows that it is a plausible composition for a silicate liquid.
The starting products for the crystallization experiments were glasses obtained after two fusions (each at 1400°C and 1 atm for 4 h) of the powdered rock samples. Cleaned Au capsules were successively loaded with weighed amounts of deionized H2O, silver oxalate (Ag2C2O4) as a source of CO2, and glass powder. Au capsules were chosen to prevent Fe loss to the container (e.g. Sisson & Grove, 1993). Upon heating, the silver oxalate decomposes to blebs of silver and CO2 and a C–O–H fluid is formed, in which almost all the C forms the CO2 species under the investigated fO2 (see Scaillet et al., 1995). For each investigated temperature and each bulk composition, 4–6 capsules were prepared, each with a different initial fluid composition; i.e. the molar H2O/(H2O + CO2) ratio in the fluid, XH2Oin, was varied between one [considering the H2O-saturated melt at P–T as the standard state: aH2O (melt) = aH2O (fluid) = 1] and 0·5. Fluid-saturated conditions at 4 kbar were achieved by keeping a mass ratio of 15% fluid for 85% silicate powder. In two runs (4 kbar and 1000°C; 2 kbar and 850°C), H2O was the sole added volatile with the mass ratio H2O/silicate calculated to yield H2O-undersaturated or H2O-saturated melts. The capsules were then welded shut and checked for leaks in an oil-bath at 120°C.
Experiments were mainly performed in an internally heated pressure vessel (IHPV ‘Gros Bleu’) pressurized with Ar–H2 mixtures. The vessel was fitted with an H2-membrane (Scaillet et al., 1992) for temperatures below 950°C. One experiment at 2 kbar and 850°C was performed in another IHPV, named ‘Basset’ (Tables 2 and 3). This vessel can be pressurized only by Ar and imposes oxidized conditions (see below). Temperature in the reaction zone was read by two sheathed type-K thermocouples to check for thermal gradients. The precisions on pressure and temperature are estimated to be ±0·02 kbar and ±5°C, respectively. High-temperature experiments (1000 and 950°C) were performed with a fast-quench device (Roux & Lefèvre, 1992) to prevent quench crystallization. The fO2 of H2O-saturated charges was calculated using fH2 from the membrane, assuming that the fluid was pure H2O, and using the H2O fugacity tables of Burnham et al. (1969) together with standard thermodynamic properties from Robie et al. (1979) for the dissociation reaction of H2O at 1 bar. Knowing the fH2 and the XH2Oin, the fO2 of the H2O-undersaturated charges was calculated using the following equation: log fO2 = log fO2(aH2O = 1) + 2 log XH2Oin (see Scaillet & Evans, 1999). The quality of fO2 control was evaluated by comparing membrane-derived fO2 values with those calculated using Fe–Ti oxide equilibrium, when good analyses were available. For the 880°C run, the formulation of Ghiorso & Sack (1991) gives 880°C and 
NNO = + 0·26 (where NNO is the nickel–nickel oxide buffer), which is close to the NNO + 0·41 given by the membrane. In membrane monitored runs, we thus consider fO2 to be known within 0·15 log units. In runs at 4 kbar and 950–1000°C, fO2 can only be estimated via Fe–Ti oxides. Although the formulation of Ghiorso & Sack (1991) is not calibrated for high fO2, a graphical estimation (Ghiorso & Sack, 1991, fig. 4) suggests that fO2 was above NNO + 2·5 in the high-temperature runs as well as one experiment at 2 kbar and 850°C performed in the IHPV ‘Basset’.
|
|
|
Mounted charges were examined using a petrographic microscope and were then carefully analysed using a scanning electron microscope equipped with EDS (JEOL-JSM 6400) at ESEM (Université d'Orléans) to identify the mineral assemblages. Quantitative analyses of the experimental glasses (defocused beam between 5 and 10 µm) and minerals (focused beam of 1 µm) were obtained using a CAMECA SX-50 electron microprobe at the Service commun BRGM–CNRS d'Orléans, operating at 15 kV and 6 nA with a counting time of 10 s. During each analytical session, dacite to rhyolite glass standards with known H2O contents (0–9 wt %) were analysed to correct for alkali migration and to determine the H2Oglass of the charges using the difference to 100% method. The precision of this method is ±0·5 wt % H2O (Devine et al., 1995
).
Several lines of evidence suggest that the experiments closely approached equilibrium (e.g. Clemens et al., 1986; Dall'Agnol et al., 1999): (1) melts and minerals with euhedral shape are homogeneously distributed across the charges; (2) their compositions are homogeneous both at grain–melt pool and charge scale, and show progressive variations with experimental parameters; (3) T–fO2 values calculated with the experimental Fe–Ti oxides using the calibration of Ghiorso & Sack (1991) agree with the conditions imposed during the experiments (see ‘Experimental and Analytical Procedures’); (4) calculated concentration ratios between coexisting phases [(Fe/Mg)clinopyroxene/(Fe/Mg)melt, (Ca/Na)plagioclase/(Ca/Na)melt] are similar to those established in other experimental studies, as detailed below.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGICAL BACKGROUNDEXPERIMENTAL AND ANALYTICAL...RESULTSCRYSTALLIZATION CONDITIONS OF...COMPARISON WITH OTHER...DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Phase diagram at 4 kbar
Experimental results at 4 kbar are given in Tables 2 (sample 98N50) and 3 (sample 98N06). The related phase diagrams are illustrated in Fig. 4. In addition to glass and fluid phases (except in CO2-free charges undersaturated in H2O), the following minerals were observed: clinopyroxene, low-Ca pyroxene (either orthopyroxene or pigeonite), amphibole, biotite, plagioclase, magnetite, ilmenite, apatite. Anhedral zircon was locally observed but its habit suggests it is actually a refractory mineral incompletely melted during the preparation of the starting material. The solidus shown corresponds to that of the haplogranite system as residual liquids trend toward this composition (Holtz & Johannes, 1994
). H2O solubility varies between 8·1 and 8·8 wt % for 98N06 and between 8·2 and 9·7 wt % for 98N50. These results are in good agreement with the H2O solubility calculated using the method of Burnham (1979), except at high temperature for sample 98N50 (dotted line in Fig. 4). The liquidus was not reached in our experiments but is slightly over 1000°C at aH2O = 1, as indicated by the low mineral proportions at this temperature and H2O content (Tables 2 and 3: 96% of glass in 98N06 and 92% of glass in 98N50).
Magnetite, ilmenite and apatite are the liquidus minerals and crystallize at temperatures slightly above 1000°C at H2O-saturation. The slope of the saturation curve for the Fe–Ti oxides is not constrained by our experiments but inferred from other experimental studies (e.g. Dall'Agnol et al., 1999). Clinopyroxene is the first silicate to crystallize (T > 950°C) and is present in all charges except at low temperature (T < 850°C at aH2O = 1, Cpx-out curve) for the bulk composition 98N50, in which it reacts with melt to produce amphibole. Its crystallization temperature decreases as aH2O increases. Plagioclase was the only tectosilicate obtained under the investigated conditions. Its crystallization temperature is very sensitive to H2Omelt: it is a near-liquidus mineral for low H2O content in the melt (H2Omelt < 3 wt %), and its crystallization temperature is lower than 750°C for H2O-saturated conditions. Quartz and alkali feldspar were not observed in any of our experimental charges. This is not surprising, as experimental studies (e.g. Klimm et al., 2003) have shown that these minerals crystallize close to the solidus whereas our charges equilibrated at our lowest investigated temperature (775°C) and with the lowest H2O content in melt still contain 51 and 59 wt % glass (charges 50-31 and 06-32, respectively; Tables 2 and 3). Quartz and alkali feldspar certainly crystallize at lower temperatures and/or H2O contents than those investigated here, in agreement with petrographic observations (see above). The Low-Ca pyroxene (Lpx) stability field encompasses a large interval of temperature but is restricted to H2Omelt < 6·5 wt % for 98N50 and < 5·2 wt % for 98N06. Because of lack of data for T < 830°C at low H2Omelt, we cannot specify whether Lpx is stable at lower temperature or if it is in reaction relation with the liquid as temperature decreases (e.g. Lpx + liquid1 
Biotite + liquid2), as suggested by the dashed line in Fig. 4. The maximum crystallization temperature for amphibole is probably slightly above 880°C and its stability field is restricted to H2Omelt > 5 wt % for sample 98N50 and H2Omelt > 6 wt % for sample 98N06. This topology has previously been observed in experimental studies of granodioritic and granitic compositions (e.g. Naney, 1983; Dall'Agnol et al., 1999; Costa et al., 2004). The biotite stability field is very sensitive to the magma composition: its saturation curve has a weak positive slope in the phase diagram of sample 98N06, with an upper T stability above 880°C. In contrast, in sample 98N50, biotite crystallizes at 860°C close to the solidus, down to below 775°C at H2O saturation; the biotite-in curve thus has a negative curvature in the T–H2O projection (Fig. 4).
Experiments at 2 kbar
Two additional experiments were performed at 2 kbar and 850°C at two fO2 values [ NNO + 1·5 (Gros Bleu pressure vessel) and > NNO + 2·5 (Basset pressure vessel), Tables 2 and 3]. The aim of these experiments was to investigate the effect of pressure on the mineral stability fields, in particular that of plagioclase (see below), and the liquid line of descent. We observed the same minerals as at 4 kbar, except that no low-Ca pyroxene crystallized. The lack of low-Ca pyroxenes is probably linked to the more oxidized conditions imposed in these experiments at 2 kbar. Indeed, the experiments of Dall'Agnol et al. (1999) and Martel et al. (1999) at different fO2 showed the shrinking of the low-Ca pyroxene stability field as fO2 increases. Another difference between the two investigated pressures is the absence of amphibole at 2 kbar in the 98N06 composition. Either the amphibole is no longer stable in this composition at 2 kbar or it crystallizes below 850°C. In 98N50, below 3·4 wt % H2O in the melt, amphibole is replaced by clinopyroxene. Finally, the results for the 98N06 sample show that the crystallization temperature for clinopyroxene decreases with increasing fO2: it is absent in the experiment at fO2 > NNO + 2·5 whereas it crystallized at fO2 = NNO + 1·5.
Proportions and compositions of minerals
Individual mineral phases in our crystallization experiments are homogeneous and the reported compositions (Tables 4–9) are considered representative of those present in a given experimental charge. Phase proportions were determined by mass balance, using the least-squares method (Tables 2 and 3).The sums of the squares of residuals are generally low (
r2 < 0·35).
|
Fe–Ti oxides
Fe–Ti oxides obtained in our experiments are titanomagnetites and hemoilmenites. Representative analyses of experimental Fe–Ti oxides are given in Table 4. The fractions of ulvöspinel in magnetite (XUsp) and of ilmenite in hemoilmenite (XIlm) were calculated after Stormer (1983)
. Magnetite and ilmenite form large cubic–octahedral and elongated crystals in H2O-saturated charges, respectively, but their sizes quickly decrease with decreasing H2Omelt. Ilmenite analyses are often contaminated by glass.
For runs at 4 kbar between 880 and 775°C, XUsp ranges between 0·21 and 0·43 and XIlm ranges between 0·87 and 0·95. The XUsp increases slightly with decreasing H2Omelt (e.g. at 880°C, XUsp varies between 0·36 to 0·44 for H2Omelt between 8·2 and 4·6 wt %, respectively) as a result of the lowering of fO2 with lowering fH2O. XUsp also decreases with decreasing temperature as expected (Verhoogen, 1962; Ghiorso & Sack, 1991). At 950–1000°C, the ilmenite and ulvöspinel contents are markedly lower (XUsp <0·15 and XIlm <0·50), because of the more oxidizing conditions.
Plagioclase
The compositions of experimental and natural plagioclases are given in Table 5. Plagioclase crystallizing in both starting materials is similar at a given T and H2O. Its anorthite content varies between An51 and An28, decreasing with decreasing T and H2Omelt. It partly overlaps the range displayed by the natural plagioclases but the latter are slightly more sodic (An34–22). This slight discrepancy is interpreted as due to subsolidus re-equilibration of the natural feldspars as textural observations suggest (e.g. presence of myrmekite).
|
Ca–Na exchange between plagioclase and melt (Kd) varies between 1·45 and 3·19 in the investigated pressure range, without any obvious correlation with H2Omelt (which varies between 3·4 and 7·8 wt %). These observations are in agreement with calculated Kd values for other magmas of intermediate compositions under similar P–T–fO2 conditions (2–4 kbar, NNO or NNO + 2·5). For example, the Kd calculated for the Pinatubo dacite (Prouteau, 1999
; Scaillet & Evans, 1999) ranges between 1·76 and 3·94 (H2Omelt 5·0–8·5 wt %) and that for the Mt. Pelée andesites (Martel et al., 1999) ranges between 2·05 and 4·71 (H2Omelt 4·8–7·6 wt %). As in our Lyngdal experiments, for both Pinatubo and Mt. Pelée compositions, no clear correlation between the H2O content in the silicate melt and the Kd is observed. This contrasts with the experiments of Sisson & Grove (1993) on high-Al basalts and low-SiO2 rhyolites, in which Kd increases from 1·7 to 5·5 for 2–6 wt % H2Omelt, respectively. We suggest that the Kd sensitivity to H2Omelt decreases with the lowering of the Ca/Na ratio in the melt. Indeed, the majority of experimental melts from Sisson & Grove (1993) have a molar ratio of (Ca/Na)melt > 1 whereas the experiments on intermediate rocks have a (Ca/Na)melt < 1. Lastly, it is important to note that the difference in experimental plagioclase compositions between the Lyngdal granodiorite (An51–28) and Pinatubo (An62–30) or Mt. Pelée (An81–55) is not due to different H2O contents or bulk Ca/Na ratios (0·98 for the Mt. Pelée andesite and 0·94 for sample 98N50; 0·59 for sample 98N06 and 0·60 for the Pinatubo dacite). It rather reflects the important differences in plagioclase crystallization temperature in these compositions (see below).
Amphibole and biotite
In Tables 6 and 7, experimental amphiboles and biotites are compared with the natural mineral compositions. The experimental amphiboles are calcic (hornblendes sensu lato) and usually occur as very large euhedral crystals (up to 300 µm) containing small Fe–Ti oxide inclusions. Amphiboles that crystallized from the 98N50 and 98N06 starting products are very similar in composition for a given H2Omelt and T. The Al content is relatively constant: it varies between 1·67 to 2·03 cations p.f.u. at 4 kbar and between 1·38 and 1·75 cations p.f.u. at 2 kbar. In contrast, XFe varies with temperature and H2Omelt. At H2O-saturated conditions, XFe increases from 0·38 at 880°C to 0·58 at 775°C. Variations in H2Omelt have a smaller effect on XFe; for example, at 830°C, the XFe of amphibole crystallizing from 98N50 varies from 0·48 (8·3 wt % H2O) to 0·62 (6·5 wt % H2O). This evolution of XFe is mainly due to the lowering of the mg-number of the melt with decreasing T and H2Omelt. The influence of fO2 on XFe is examined below.
|
|
As for the amphiboles, the XFe in experimental biotites ranges between 0·39 and 0·63 at 4 kbar, as a result of the effect of H2Omelt and temperature. Table 7 shows that the compositions of the natural biotites are well reproduced by our experiments, in terms of their Fe/Mg ratio. The experimental biotites are, however, richer in Al2O3 and TiO2 and poorer in K2O. The difference in the halogen contents between the experimental and natural biotites and amphiboles could be due to their partial volatilization during atmospheric melting of the samples or to the presence of an H2O–CO2 fluid in our experiments.
Pyroxenes
The compositions of the experimental clinopyroxenes are given in Table 8 and can be compared with the composition of the natural clinopyroxenes. The calculated KdFe–Mg = [Fe/Mg]cpx/[Fe/Mg]melt (with all Fe as Fe2+) for the H2O-saturated charge (98N50) at 950°C is 0·23, which is close to the calculated Kd for the Mt. Pelée andesite (Kd = 0·25 at aH2O = 1, 900°C and 2 kbar; Martel et al., 1999).
|
As for amphibole, the experimental clinopyroxene composition is similar in both starting materials for a given temperature and H2Omelt. It is mainly a diopside–augite [after the nomenclature of Morimoto et al. (1988)
], trending toward hedenbergite in 98N06. Here too, the XFe ratio is a function of temperature and H2Omelt. Under H2O-saturation, XFe increases from 0·38 at 880°C to 0·59 at 775°C, and at 850°C XFe varies from 0·33 for 8·8 wt % H2Omelt to 0·56 for 3·6 wt % H2Omelt. At low H2Omelt, a low-Ca pyroxene (pigeonite or orthopyroxene; Table 9) crystallizes with the augite. Orthopyroxene crystallizes with a Ca-rich augite, whereas augite crystallizing with pigeonite has a lower Ca content. Calculated temperatures for coexisting pyroxenes with the QUILF program (Andersen et al., 1993) are between 57 and 259°C above the experimental temperatures. Such discrepancies are common in experimental studies (e.g. Vander Auwera & Longhi, 1994; Villiger et al., 2004) and could be partially due to the calculation procedures used to project pyroxenes with relatively high contents of, for example, Al2O3 or TiO2, in the wollastonite–enstatite–ferrosilite compositional space. Another problem is the presence of pigeonite in some runs, up to 100°C below the minimum temperature of its stability field determined by Lindsley (1983) in the system enstatite–ferrosilite–wollastonite. As we have only few data on coexisting pyroxenes, we cannot explore in details the effects of pyroxene and melt compositions on the differences between the experimental and calculated temperatures, and the possibility that the pyroxenes did not reach their equilibrium compositions in the experimental charges with low H2O contents cannot be excluded. However, the stability field for the low-Ca pyroxenes (Fig. 4) is likely to represent equilibrium conditions as it is consistent over a large temperature range (950–800°C) in our experiments and its T–H2Omelt relationships are similar to those in other experimental studies (e.g. Naney, 1983; Costa et al., 2004).
|
The natural clinopyroxenes are well reproduced in our experiments with high H2O contents, with the exception of the higher Al2O3 content in the experimental clinopyroxenes (between 5·3 and 1·2 wt % Al2O3 in the experimental clinopyroxenes, compared with between 1·5 and 1·2 wt % in the natural ones). A possible explanation for the high Al2O3 content in some experimental clinopyroxenes arises from the observation that their Al2O3 content is strongly dependent on the activity of Al2O3 in the melt (Villiger et al., 2004
). A general co-variation between the Al2O3 molar fraction (XAl2O3) in the melt and the Al2O3 content of the experimental clinopyroxenes is observed (Fig. 5). Actually, as we show below, our experimental melts are generally richer in Al2O3 than the natural liquid line of descent of the HBG suite. In charges where the XAl2O3 values of the experimental melts are similar to those of the HBG rocks, the Al2O3 content of the experimental clinopyroxenes (1·7–1·2 wt % Al2O3) is similar to that of the natural ones.
|
Glasses
Representative glass compositions are listed in Table 10. With increasing crystallization (lowering T and H2Omelt) the residual liquid becomes impoverished in MgO, FeO, CaO, TiO2 and Al2O3, whereas SiO2 and alkalis increase, in response to the precipitation of the plagioclase–pyroxenes–oxides (± amphibole) assemblage. The glass SiO2 content for the 98N50 composition varies between 59 and 76 wt % at 1000°C and 800°C, respectively, and between 67 and 76 wt % for the 98N06 composition. The more evolved liquids trend toward the minimum of the haplogranite system and closely resemble metaluminous to slightly peraluminous high-silica rhyolites erupted in extensional continental settings such as the Bishop Tuff (Table 10) or the Yellowstone rhyolites (Hildreth, 1979
; Hildreth et al., 1991).
|
|
CRYSTALLIZATION CONDITIONS OF THE LYNGDAL MAGMA |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGICAL BACKGROUNDEXPERIMENTAL AND ANALYTICAL...RESULTSCRYSTALLIZATION CONDITIONS OF...COMPARISON WITH OTHER...DISCUSSIONCONCLUSIONSREFERENCES |
|---|
H2O content
At a given pressure and fO2, the crystallization sequence depends on the H2O content in the melt (Fig. 4). The following lines of evidence indicate that the natural sequence of crystallization can be reproduced experimentally only if high H2Omelt prevailed during the early stages of crystallization. First, the low-Ca pyroxene stability field is restricted to low H2Omelt (H2Omelt < 6·5 wt % for 98N50, H2Omelt < 5 wt % for 98N06) whereas amphibole crystallizes for H2Omelt > 5 wt %. Petrographic study of the Lyngdal granodiorite samples shows that amphibole is stable in both compositions and is an early crystallizing mineral, whereas neither orthopyroxene nor pigeonite was observed, nor were the latter two minerals observed in any other HBG pluton studied by Vander Auwera et al. (2003)
. It appears that the H2O content of the melt was probably above 6 wt % during the crystallization of the Lyngdal granodiorite. Even if we suppose that a low-Ca pyroxene had crystallized in the Lyngdal magma and was later reacted out, another constraint on H2Omelt is provided by the scarcity of biotite in the least differentiated Lyngdal rocks (60 wt % SiO2) and its late crystallization relative to amphibole. Figure 4 shows that biotite crystallizes after amphibole when H2Omelt is above 5 wt % at 850°C and this is a minimum estimate. Indeed, biotite is more abundant than amphibole when they coexist with a low-Ca pyroxene, i.e. for H2O contents between 5·5 and 6·5% at 830°C (see Table 2 for phase proportions in the experimental charges), in disagreement with the petrographic observations. It is also worth noting that high H2O contents are supported by the abundance and large size of pegmatites (outcrops up to several hundred square meters with crystals reaching the metre scale) in Lyngdal intrusion and in other members of the HBG suite (Svedrup, 1960; Nogarède, 2001). For this elevated H2O content, near-liquidus phases are Fe–Ti oxides, clinopyroxene and apatite, in accordance with petrographic observations, which also suggest that plagioclase crystallizes together with or only slightly before amphibole. The crystallization path for a bulk-rock H2O content of 5 wt %, calculated assuming closed-system behaviour and using experimental phase proportions, is shown by the thick grey curve in Fig. 4. Clearly, under these conditions, the natural sequence of crystallization is well reproduced.
Oxygen fugacity
Oxygen fugacity exerts a strong control on the mineral composition (in particular, Fe–Ti oxide compositions and the XFe of the ferromagnesian minerals), and comparison between the experimental products and the natural minerals allows us to estimate this parameter. However, natural Fe–Ti oxides typically re-equilibrate under subsolidus conditions and, therefore, cannot be used to constrain fO2. In contrast, ferromagnesian silicates better preserve the redox conditions prevailing in the magma, even if they re-equilibrated down temperature to some extent. The following observations suggest that both clinopyroxene and amphibole compositions preserve magmatic conditions: (1) clinopyroxenes found as inclusions in amphibole and plagioclase are similar to each other; (2) amphibole in plagioclase has a composition close to that in the matrix. It is unlikely that minerals included in plagioclase would easily re-equilibrate by exchanging Fe and Mg with other minerals in the matrix. Tables 6–8 show that the compositions of the ferromagnesian minerals are well reproduced by our experiments at 4 kbar and NNO or NNO + 1. Admittedly, our experiments do not cover a sufficiently wide range in fO2 to allow explicit constraints on the effect of this parameter on mineral compositions. Some insight can be gained, however, from the work of Dall'Agnol et al. (1999), who performed experiments on a compositionally similar, yet more evolved composition (Jamon granite, 71 wt % SiO2; Fig. 1). In these experiments the XFe of amphibole is 0·32 ± 0·03 between 850 and 775°C at NNO + 2·5 and 0·7 at 750°C and NNO – 1·5. Interestingly, the XFe of amphiboles synthesized in our experiments at NNO or NNO + 1 broadly fills the compositional gap of amphiboles obtained at NNO + 2·5 and NNO – 1·5 by Dall'Agnol et al. (1999), suggesting that, at fixed T, fO2 exerts a dominant control on the amphibole Fe/Mg ratio relative to the bulk composition. The fact that XFe values as high as 0·70 are not found in Lyngdal amphiboles suggests an fO2 higher than NNO – 1·5 for this pluton. The Al content of our experimental amphiboles is plotted against XFe in Fig. 6. Amphibole crystallizing at NNO + 1·5 has a lower XFe than its natural counterpart, whereas that obtained at NNO or NNO + 1 reproduces this ratio. Similarly, experimental clinopyroxene and biotite at NNO or NNO + 1 also reproduce fairly well the natural compositions (Tables 7 and 8). Altogether, these observations suggest that the Lyngdal magma crystallized at an fO2 between NNO and NNO + 1.
|
Pressure
The large hornblende–biotite intrusions of the HBG suite (Fig. 2) define a differentiation trend from 56 to 77 wt % SiO2 (Vander Auwera et al., 2003
). The experimental glasses at 4 kbar reproduce the natural HBG for several elements except for CaO and Al2O3, whose content are systematically higher in the experimental glasses, as illustrated for CaO in a SiO2–CaO–(MgO + FeO + TiO2) diagram (Fig. 7). As the HBG trend is thought to reflect fractionation of Fe–Ti oxides, clinopyroxene, plagioclase and apatite, with or without amphibole, from a magma similar in composition to sample 98N50 (Bogaerts et al., 2003a), this misfit requires further investigation. Two different processes could explain such a misfit: either the starting materials represent liquids slightly laden with liquidus minerals (oxides and/or pyroxene) or the experimental intensive parameters do not fully match the natural conditions. As previously discussed, accumulation of plagioclase, apatite and zircon is not supported by the available geochemical data (Bogaerts et al., 2003a, and section ‘Experimental and analytical procedure’) and a liquidus temperature near 1000°C for intermediate compositions is compatible with the high liquidus temperature of ferroan granites (900°C for a bulk composition with SiO2 = 71 wt %; Dall'Agnol et al., 1999). Moreover, the presence of accumulated liquidus minerals in the starting material would increase their proportions in the experimental charges, but not greatly modify the composition of the melt in equilibrium with plagioclase, clinopyroxene, Fe–Ti oxides, apatite and amphibole. A misfit related to an intensive parameter is thus favoured and examined below.
|
|
The modal proportions of representative experimental mineral assemblages from Tables 2 and 3 are normalized to 100% in Table 11 and, as the composition of the individual minerals has been measured, the chemical compositions of these assemblages can be calculated. The experimental mineral assemblages and individual mineral are plotted (Fig. 7) in the compositional space SiO2–CaO–(MgO + FeO + TiO2). Based on major element modelling by the least-squares method on natural rocks, Bogaerts et al. (2003a)
calculated the mineral assemblages that have to be fractionated to produce the natural trend from quartz monzodiorite to granite in the HBG suite. This trend has been modelled in two steps, with step 1 (two possible models) starting from the least differentiated sample (56 wt % SiO2) to an intermediate composition (62 wt % SiO2), and step 2 starting from this latter composition to give a residual melt of 72 wt % SiO2. The modal proportions of the fractionated mineral assemblages (hereafter referred to as modelled mineral assemblages) are given in Table 11 and plotted in Fig. 7 in the compositional space SiO2–CaO–(MgO + FeO + TiO2) for comparison with the experimental mineral assemblages. The modal proportions of the experimental mineral assemblages (Table 11) first show that the proportion of oxides and clinopyroxene in experiments at approximately NNO and NNO + 1 is similar to the modelled mineral assemblages. This again supports our observation that, if present, the proportion of accumulated minerals is negligible. Second, the misfit between the experimental and modelled mineral assemblages at 4 kbar is due to a higher amphibole/plagioclase ratio in the former. A higher proportion of crystallized plagioclase should lower the CaO and Al2O3 content in the experimental glasses to fit with the HBG trend. The experiments conducted at 2 kbar show that plagioclase crystallizes at 850°C for aH2O = 1, which is not the case at 4 kbar. However, despite this higher proportion of plagioclase at 2 kbar, the experimental glasses are still richer in Al2O3 and CaO than the natural HBG suite. Figure 7 shows that the CaO content of residual liquids is very sensitive to the plagioclase/Fe–Ti oxides ratio, which depends on pressure (mainly affecting plagioclase) and fO2 (mainly affecting the Fe–Ti oxides). In the experimental mineral a