Journal of Petrology Advance Access originally published online on November 22, 2007
Journal of Petrology 2008 49(1):25-45; doi:10.1093/petrology/egm068
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Depths of Partial Crystallization of H2O-bearing MORB: Phase Equilibria Simulations of Basalts at the MAR near Ascension Island (7–11°S)
1Institute of Mineralogy, Leibniz University Of Hannover, Callinstrasse 3, 30167, Germany
2Department of Earth Sciences, University of Aarhus, C. F. Møllers Allé 110, DK-8000 Aarhus C, Denmark
3leibniz Institute For Marine Sciences (IFM-GEOMAR), WISCHHOFSTRASSE. 1–3, D-24148 Kiel, Germany
RECEIVED MAY 3, 2006; ACCEPTED OCTOBER 10, 2007
 |
ABSTRACT |
|---|
Phase equilibria simulations were performed on naturally quenched basaltic glasses to determine crystallization conditions prior to eruption of magmas at the Mid-Atlantic Ridge (MAR) east of Ascension Island (7–11°S). The results indicate that mid-ocean ridge basalt (MORB) magmas beneath different segments of the MAR have crystallized over a wide range of pressures (100–900 MPa). However, each segment seems to have a specific crystallization history. Nearly isobaric crystallization conditions (100–300 MPa) were obtained for the geochemically enriched MORB magmas of the central segments, whereas normal (N)-MORB magmas of the bounding segments are characterized by polybaric crystallization conditions (200–900 MPa). In addition, our results demonstrate close to anhydrous crystallization conditions of N-MORBs, whereas geochemically enriched MORBs were successfully modeled in the presence of 0·4–1 wt% H2O in the parental melts. These estimates are in agreement with direct (Fourier transform IR) measurements of H2O abundances in basaltic glasses and melt inclusions for selected samples. Water contents determined in the parental melts are in the range 0·04–0·09 and 0·30–0·55 wt% H2O for depleted and enriched MORBs, respectively. Our results are in general agreement (within ±200 MPa) with previous approaches used to evaluate pressure estimates in MORB. However, the determination of pre-eruptive conditions of MORBs, including temperature and water content in addition to pressure, requires the improvement of magma crystallization models to simulate liquid lines of descent in the presence of small amounts of water.
KEY WORDS: MORB; Mid-Atlantic Ridge; depth of crystallization; water abundances; phase equilibria calculations; cotectic crystallization; pressure estimates; polybaric fractionation
|
INTRODUCTION |
|---|
The chemical compositions of basaltic glasses recovered from mid-oceanic ridges commonly show characteristics that are believed to be the result of various processes occurring at the stage of (1) primary magma generation and (2) subsequent modification of the parental magmas in the course of fractional crystallization, magma mixing and wall-rock assimilation en route to the ocean floor. Mantle source heterogeneity or/and different extents of partial melting, and in some cases the influence of deep mantle plumes, are generally advocated to explain differences in trace element abundances and ratios as well as in isotopic compositions of MORB glasses. Geochemical data, in particular large variations in trace element ratios for most MORB suites, clearly demonstrate that the trace element abundances in basalts (including K2O) are principally controlled by processes occurring in the source region (composition of the source, melt fraction), although crystal fractionation also plays a role. In contrast, major element variations in a majority of MORB, as a first-order approximation, can be directly related to the continuous evolution of the parental basaltic melts along Ol + Pl and Ol + Pl + Cpx cotectics (Mineral abbreviations: Ol-olivine, Pl-plagioclase, Cpx-clinopyroxene). This observation is supported by a number of experimental studies on phase equilibria in MORB-like systems in which liquid lines of descent fairly well reproduced natural petrochemical trends. Another important observation of the experimental studies was to highlight a pronounced dependence of the clinopyroxene saturation temperature on pressure (Bender et al., 1978
): at higher pressure Cpx crystallizes earlier (at higher temperature), resulting in a compositional trend showing a decreasing CaO concentration and CaO/Al2O3 ratio in residual liquids with decreasing MgO concentration. This property of MORB systems allows estimation of the pressure (depth) at which the basaltic magmas last equilibrated with the Ol + Pl + Cpx mineral assemblage before eruption. Several semi-empirical techniques use this property to evaluate partial crystallization pressures (e.g. Ariskin et al., 1992; Grove et al., 1992; Danyushevsky et al., 1996; Yang et al., 1996; Herzberg, 2004; Villiger et al., 2007). These methods are based mostly on the results of melting experiments on MORB compositions. The pressure can be determined within a precision estimated to be ±100–200 MPa. However, all these methods, as has been pointed out by Michael & Cornell (1998), have two main limitations: the models require that (1) the system is saturated with respect to Cpx (in addition to Ol + Pl) and (2) the system is assumed to be anhydrous. The first of these limitations can be accounted for by making pressure estimates only for lavas containing less than 8 wt% MgO, as this is the value at which the CaO concentrations of most MORB compositions start to decrease with decreasing MgO content. The second limitation (anhydrous crystallization) is related to the database used for the calibration of the models, which were constructed from phase equilibria experiments conducted at nominally dry conditions. Two sources of error arise: (1) high-pressure experiments performed in a piston cylinder apparatus are never absolutely water-free, although their water activities are generally not determined (Hirschmann et al., 1998; Holtz et al., 2001; Kagi et al., 2005); (2) even the small amounts of water in typical MORBs [less than 0·6 wt% H2O; see, for example, compilations of Michael (1995) and Danyushevsky (2001)] will have discernible effects on mineral cotectics and petrochemical trends (e.g. Michael & Chase, 1987; Danyushevsky, 2001; Asimow et al., 2004) and so need to be taken into account when simulating magma crystallization processes.
In this paper we present a method to constrain the pressure at which crystallization occurred in hydrous MORB. The effects of pressure and of small amounts of H2O on liquid lines of descent were simulated using the COMAGMAT (version 3.57) program (Ariskin et al., 1993; Ariskin & Barmina, 2004) (hereinafter referred to as COMAGMAT). The method was applied to estimate pre-eruptive conditions for basaltic magmas from four segments (A1–A4) of the Mid-Atlantic Ridge (MAR) between 7 and 11°S (Moeller, 2002). Two approaches, simulating (1) fractional crystallization and (2) equilibrium crystallization, were applied. The results obtained with the two methods are in a good agreement and we find polybaric crystallization conditions (200–900 MPa) for the normal (N)-MORB magmas beneath segments A1, A2 and A4. In contrast, the geochemically enriched MORB magmas from segment A3 apparently experienced their last equilibration with the Ol + Pl + Cpx mineral association at nearly isobaric crystallization conditions and at lower pressures (
100–300 MPa).
|
GEOLOGICAL BACKGROUND AND SAMPLE LOCATIONS |
|---|
 TOPABSTRACTINTRODUCTION GEOLOGICAL BACKGROUND AND SAMPLE... METHODS AND DATA TREATMENTSRESULTSRESULTS FROM FORWARD AND...DISCUSSIONCONCLUSION: TOWARDS AN ACCURATE...SUPPLEMENTARY DATAREFERENCES |
|---|
The 200 km long portion of the Southern Mid-Atlantic Ridge axis in the vicinity of Ascension is bounded to the north and south by the Ascension and Bode Verde Fracture zones, respectively (Fig. 1). Between these fracture zones, the slow-spreading (
3 cm/year) mid-ocean ridge is divided into four second-order segments of contrasting character and magma output by three non-transform offsets. The two central segments (A2 and A3) have shallow depths and rift axial highs whereas the two marginal segments (A1 and A4) are characterized by deep axial valleys (Fig. 1). Seismic studies have shown segments A2 and A3 to be characterized by a relatively thick crust of 11 km, whereas segments A1 and A4 have crustal thicknesses of 5 km (Minshull et al., 1998; Bruguier et al., 2003). The shallow depth and anomalous axial morphology (for a slow-spreading segment) of segment A3, associated with the spatial proximity of numerous seamounts (Ascension, Circe, Grattan, seamount D), the presence of a mantle Bouguer gravity anomaly (Minshull et al., 1998, 2003), as well as geochemical and isotopic data (Schilling et al., 1985; Hanan et al., 1986; Graham et al., 1992; Fontignie & Schilling, 1996; Bourdon & Hemond, 2001) have been used to suggest a localized anomaly in mantle composition (not necessarily related to a thermal mantle plume but responsible for excess mantle melting) in this region (Minshull et al., 1998; Bruguier et al., 2003) or the influence of a hot mantle plume (Schilling et al., 1985; Bourdon & Hemond, 2001) on the convective system of the MAR in the vicinity of the segment A3.
|
).In 1998 the Mid-Atlantic Ridge spreading axis between 7 and 11°S was dredged during cruise M41/2 of the German research vessel Meteor (see Fig. 1 for dredge locations). The dataset of chemical compositions, including major and trace elements and Sr, Nd and Pb isotopic ratios, of the recovered basaltic glasses has been presented by Moeller (2002
). Selected samples from this set were used here to provide input data for the geochemical modeling. |
METHODS AND DATA TREATMENTS |
|---|
Analytical techniques
Electron microprobe
Major element compositions of selected quench glasses, as well as melt inclusions hosted in olivine phenocrysts, were analyzed using a Cameca SX100 electron microprobe at the University of Hannover, at 15 kV acceleration potential. For major and minor elements, the beam current was 4 nA and the counting time was set to 6 s for Na and K and 12 s for the other elements. These analytical conditions are considered as the most appropriate for the analysis of experimental hydrous basaltic to silicic glasses. All glass analyses were performed using a defocused beam of 20 µm, except for some small glass inclusions, for which we used a beam size of 5 µm. Each reported analysis (Table 1) is the average of
10 points for the quenched glasses (except one glass with 175 analyses; see below), and fewer than four points for the glass inclusions. Cl, P and Cr in MORB glasses were measured with a beam current of 30 nA, and counting times varied between 30 and 120 s.
|
Infrared (IR) spectroscopy
Doubly polished plates of natural quenched glass (2–3 mm in diameter) and olivine crystals with glass inclusions (80–120 µm in size) with thickness of 60–80 µm were prepared for Fourier transform IR (FTIR) spectroscopy to investigate the H2O abundances. The thickness of each sample plate was measured with a digital micrometer (Mitutoyo; precision ±2 µm). The H2O concentrations were determined using a Bruker IFS88 FTIR spectrometer coupled with an IR-Scope II optical microscope (operation conditions: MCT narrow range detector, globar light source and KBr beamsplitter). H2O contents were analyzed using the main OH-stretching peak of OH groups and molecular H2O at
3500 cm–1. Typically 50 scans were used for IR measurements. The spot size applied was 100 µm x 100 µm for the glasses and 60 µm x 60 µm for the glass inclusions. The analyzed area was checked optically before IR measurement to avoid the presence of microlites, fluid bubbles, cracks or impurities. The glass density was assumed to be 2800 g/l. Molar absorptivity used for all glasses was 67 l/mol per cm for 
3500 (Stolper, 1982).
Estimation of crystallization conditions: forward and inverse modeling
We developed and applied a new methodological approach to estimate intensive variables of crystallization for MORB glasses. Following the terminology of Myers & Johnston (1996), we performed numerical forward and inverse experiments to identify the conditions under which the given MORB composition could be generated. In our forward modeling we assumed a genetic relationship between all MORB lavas within a given segment. We checked if the compositions within one segment could be produced by fractional crystallization of the most primitive sample of this segment by varying pressure and initial water content. In the inverse approach we performed equilibrium crystallization calculations for each MORB composition, with the aim of determining the pressure and temperature at which multiple saturation (Ol + Pl + Cpx) occurs.
As a model to simulate phase equilibria we used the COMAGMAT program (Ariskin & Barmina, 2004), as the effects of water and pressure on liquid lines of descent (LLD) can be modeled simultaneously. Comparisons with experimental data on dry MORBs show that this program gives consistent results with respect to mineral crystallization sequences, mineral proportions, and melt and mineral compositions (Yang et al., 1996; Ariskin, 1999; Almeev et al., 2004). Although COMAGMAT utilizes a simplified approach (Almeev & Ariskin, 1996; Ariskin, 1999) to quantify the effect of H2O on the crystallization temperatures of minerals, it has been shown that the phase equilibria in H2O-saturated high-alumina basalt studied experimentally by Sisson & Grove (1993) can be reliably predicted (Almeev & Ariskin, 1996).
In our calculations the oxygen fugacity was assumed to be buffered by the quartz–fayalite–magnetite (QFM) assemblage. These conditions are slightly more oxidizing than those measured in MORB glasses. In recent determinations of Fe3+/
Fe in MORB glasses, Bezos & Humler (2005) showed that the average ratio is 0·12, indicating that the oxygen fugacity of most MORB is 0·4 log units below the QFM buffer.
Forward modeling: fractional crystallization (FC) calculations
Figure 2 illustrates an example of the forward modeling approach. Assuming that most of the glasses within one segment (e.g. A1) are genetically related to each other through differentiation of the same parental melt (e.g. sample 140DS1, one of the most primitive samples of segment A1), we performed a set of fractional crystallization calculations for this composition with initial H2O contents of 0, 0·2, 0·5, 0·7 and 1·0 wt% H2O. Calculations were performed up to 60% crystallization, in the pressure range 0·1 MPa–1 GPa with a small pressure increment of 10 MPa. For each pressure, COMAGMAT was used to identify the liquidus phase(s) (using a bulk crystallization increment of 1 wt%) and the composition of the residual liquids. As a result, we obtained 100 isobaric LLDs for each initial water content (in total 500 LLDs because five initial H2O contents were tested). The full dataset of modeled MORBs consisted of 30 000 liquid compositions for each parental melt (500 LLDs x 1 composition for each of 60 crystallization increments per LLD). A typical dataset of the calculated MORB melts (only dry glasses are presented) is shown in Fig. 2a. In a second step, we performed a systematic comparison of the calculated residual melt compositions with each natural MORB glass. The following procedure was used to determine the modeled liquid that has the same composition as the natural MORB in terms of all major oxides. First, we selected glasses that have the same MgO content (as a proxy of degree of differentiation) and the same CaO/Al2O3 ratio (as a proxy of location on the same isobaric mineral cotectic). Then, from these compositions, we selected the modeled liquids that are as close as possible to the natural composition in terms of other major oxides (e.g. CaO, FeO, SiO2, Al2O3, Na2O). If the correspondence between natural and modeled MORB was within the analytical precision of electron microprobe microanalyses (2
) for each oxide, the natural glass was considered to be successfully modeled, and the calculated intensive parameters of crystallization (e.g. temperature, pressure) and H2O in the melt (H2O that was accumulated in the melt as a result of its incompatibility), as well as mineral assemblage and mineral proportions were assigned as appropriate to produce the given natural MORB from the chosen starting composition (Fig. 2b–f). For example, in Fig. 2 the modeled sample (evolved MORB) is best reproduced assuming a LLD at 420 MPa with 0·2 wt% H2O in the parental melt (black diamond on the 420 MPa LLD in Fig. 2b–d).
|
, 420 MPa and 0·2 wt% H2O; 
, 400 MPa and 0·5 wt% H2O; 
, 410 MPa and 0·7 wt% H2O; +, 390 MPa and 1 wt% H2O. Filled symbols indicate the modeled compositions, which have the same MgO and CaO/Al2O3 ratio as the evolved natural MORB (139DS2). In this example, the natural evolved MORB is best reproduced at 420 MPa from the parental composition 140DS1 with 0·2 wt% H2O. Grey crosses in (b)–(f) are results of 175 replicate microprobe measurements of the sample 140DS1 (see text for further details).This search procedure was applied to each natural MORB glass from segments A1–A4, although not all natural MORB samples were successfully reproduced (see below). As a result we obtained a representative set of crystallization parameters for MORB magmas from segments A1–A4 that are interpreted as the conditions existing in partly molten systems just before ascent of the magmas to the ocean floor. It is essential to appreciate that the method outlined above does not require a knowledge of the H2O concentration in the system; H2O concentration is determined in the course of the calculations. This requires that the crystallization model can correctly predict the role of small amounts of water on the crystallization temperatures of olivine, plagioclase and clinopyroxene.
Inverse modeling: equilibrium crystallization (EC) calculations
The second approach is based on the assumption that all MORB melts are (multiply) saturated with respect to Ol + Pl + Cpx prior to eruption. It is known that pressure and aH2O are the most important variables that significantly affect cotectic crystallization in the MORB system. Therefore, if the first parameter is known (pressure or H2O), the second parameter can be obtained for a given MORB composition if all three minerals (Ol + Pl + Cpx) are in equilibrium with this MORB liquid. In practice, the pressure of partial crystallization is unknown, but the H2O contents of the glass can be measured or estimated from the H2O–Ce or H2O–K2O covariation (Michael, 1995). Then, for a given MORB composition (with known H2O), calculations simulating equilibrium crystallization can be performed to identify the pressure at which all three minerals Ol + Pl + Cpx coexist with a melt having the composition of the natural glass sample (the condition at which the natural glass composition is on the Ol + Pl + Cpx cotectic; Fig. 3). Thus, in these inverse calculations the exact H2O content is a required parameter for the modeling. In this study we used H2O contents estimated from the H2O–K2O relationship (see below).
|
|
RESULTS |
|---|
Natural dataset: basaltic glasses from the MAR (7–11°S)
The chemical variations of lavas erupted along the spreading segments of the MAR (7–11°S) are shown in Fig. 4 [data of Moeller (2002
) and this study]. Roughly, four slightly overlapping compositional clusters can be noted: (1) MORBs of segments A1 and the majority of samples from segment A4; (2) MORBs of segment A2; (3) MORBs of segments A3; (4) evolved MORBs of segment A4. Basaltic glasses from segments A1 and most samples from A4 are relatively primitive (MgO contents between 6·7 and 9·7 wt% and 7 and 9·1 wt%, respectively) and exhibit typical N-MORB characteristics (0·03 < K/Ti < 0·2). Al2O3 decreases with decreasing MgO; however, CaO does not vary systematically. The CaO/Al2O3 ratio increases from 0·65 to 0·85 with decreasing MgO. Similar to A1 and A4 MORB, the glasses of segment A2 have a wide range of MgO contents (from 6 to 9·6 wt%). However, in contrast to A1 and A4 MORBs, for the same MgO content, they are characterized by lower CaO and tend to have slightly higher Al2O3 contents (Fig. 4). Thus, the CaO/Al2O3 ratio of A2 basalts is systematically lower than that of A1 and A4 MORBs. Segment A3 MORBs are more fractionated (MgO ranges from 4·4 to 7·1 wt%) and are uniformly enriched in incompatible elements (0·2 < K/Ti < 0 4). Although Al2O3 and Na2O are slightly scattered, these lavas show an apparent trend in which CaO and CaO/Al2O3 decrease with decreasing MgO. The last compositional cluster is formed by the samples from two dredge stations collected within segment A4. These samples are characterized by lower CaO and MgO contents and higher Na2O content. They are also geochemically enriched (K/Ti 0·3). However, they are distinguishable from the A3 basaltic group in having lower CaO content at the same MgO content (Fig. 4).
|
Although the range of K2O/TiO2 illustrated in Fig. 4 and the range in trace element ratios reported by Moeller (2002
) probably reflect mantle source variability, most of the major element variations can be explained by fractional crystallization processes. The Mg-number of all MORB glasses (Table 1) is below the value (70) that is assumed to be typical for primary melts of spinel lherzolite. This suggests that all MORBs have experienced significant amounts of fractional crystallization. The FeO content increases and Al2O3 content decreases with decreasing MgO content, indicating that melt evolution is controlled by fractionation along the Ol + Pl cotectic. The scatter in CaO at a given MgO content observed in A1, A2 and A4 MORB may be related to differences in the pressure at which Cpx begins to crystallize (see also CaO/Al2O3 ratio in the range 0·65–0·85; Fig. 4). In contrast, the fairly linear positive dependence observed in the CaO vs MgO and the CaO/Al2O3 vs MgO diagrams for segment A3 MORBs (Fig. 4) definitely reflects fractionation of the Ol + Pl + Cpx-bearing mineral assemblage.
Petrochemical trends and experimental liquid lines of descent (A1 and A3 MORB as an example)
The comparison of petrochemical trends from natural lavas with the compositions of experimental glasses is a widely used approach to characterize the conditions at which magma differentiation occurred. Compositional similarity between petrochemical trends and experimental LLDs implies that conditions simulated in the experiments may be similar to those prevailing in nature [see review by Myers & Johnston (1996)]. Melting–crystallization relations in MORB-systems have been extensively studied experimentally (Bender et al., 1978; Walker et al., 1979; Fisk & Bence, 1980; Fisk et al., 1980; Stolper, 1980; Fujii & Bougault, 1983; Grove & Bryan, 1983; Fujii & Scarfe, 1985; Tormey et al., 1987; Juster et al., 1989; Grove et al., 1990, 1992; Kinzler & Grove, 1992; Gaetani et al., 1994; Thy & Lofgren, 1994; Yang et al., 1996; Thy et al., 1998, 1999; Berndt et al., 2005). However, only a few starting compositions were found to be suitable analogues for parental magmas relevant for segment A1 lavas and more evolved basalts from segment A3. These two segments are discussed in detail below, considering that they show the most contrasting compositions.
Phase relations in systems relevant to the A1 most primitive composition were studied under anhydrous conditions at 0·1 MPa (composition 70-002; Yang et al., 1996) and 800 MPa (composition ALV-2004-3-1; Grove et al., 1992), and in a water-bearing system at 200 MPa (synthetic MORB B1; Berndt et al., 2005). Compositions similar to the most primitive A3 segment magmas have been investigated at 0·1 MPa (East Pacific Rise basalt PROTEA-61-002; Yang et al., 1996), and three nominally dry isobaric crystallization sequences (at 0·1, 200 and 800 MPa) were obtained on Serocki volcano tholeiitic basalt (sample ALV-1690-20; Grove et al., 1990, 1992). A hydrous tholeiite system was examined at 500 MPa (composition OB93, Kerguelen Plateau; Freise, 2004). All experimental glasses produced from these starting basalts are plotted together with natural quenched glasses of the A1 and A3 segments on the variation diagrams in Fig. 5.
|
As shown in Fig. 5, the majority of A1 basaltic glasses are located within the compositional space defined by 0·1 MPa and 800 MPa Ol + Pl and Ol + Pl + Cpx cotectics at dry conditions. These cotectics probably bracket the possible pressure range of magma evolution beneath segment A1, as is shown on the CaO vs MgO diagram (Fig. 5a). Some of the A1 glasses, however, exhibit slightly Al2O3-enriched compositions, pointing to the possible role of H2O dissolved in the magmas. Their compositions are located between the anhydrous and water-bearing LLDs determined by Berndt et al. (2005
) at 200 MPa (Fig. 5). However, the H2O contents of the experimental glasses produced by Berndt et al. (2005) are in the range 1–4 wt%, a value that is too high to be realistic for magmas along the MAR to explain the observed difference in Al2O3 content between natural lavas and the experimental LLD. Thus, the characteristics of most glasses from segment A1 are probably consistent with the nearly anhydrous experimental LLDs in the 0·1–800 MPa range.
The basaltic glasses of segment A3 exhibit a well-developed trend on a CaO vs MgO plot (Fig. 5d). This trend is nearly parallel to the isobaric LLDs produced experimentally by Grove et al. (1992) and Yang et al. (1996), in contrast to A1 lavas (Fig. 5a). This observation may indicate crystallization at nearly isobaric conditions, although the estimation of absolute pressure values is not possible from this limited experimental dataset. Figure 5e and f shows that crystallization must have occurred in a H2O-bearing system for the A3 segment lavas. As shown in Figure 5e and f, the trend of A3 lavas is out of the experimental range obtained for dry MORB systems (0·1–800 MPa). The experimental results from anhydrous systems show that, with increasing pressure of crystallization, both FeO and Al2O3 increase in residual melts with the same MgO content. Thus, the natural compositions cannot be reproduced by pressure variations only. Assuming that the anhydrous 500 MPa LLD trend is intermediate between the 0·1 MPa and the 800 MPa trends, and using the experimental data for hydrous conditions (Freise, 2004), the effect of increasing water activity at constant pressure can be estimated. An increase in water activity causes a decrease of FeO and an increase of Al2O3 in residual melts at a given MgO content (see below). Thus, the evolution of natural compositions can be explained only if water is present in the magmas. However, the experimental dataset of Freise (2004) cannot be used to estimate the exact water contents of basalts from segment A3 because the water contents in this study (2·3–9·3 wt% of H2O in the melts) are significantly higher than the maximum H2O abundances measured in MORB (<1%; Danyushevsky, 2001).
H2O and chlorine in MORBs
H2O and chlorine concentrations were determined in 12 selected natural glasses, specifically chosen to represent the whole compositional range from the most primitive to the most evolved compositions for each of the segments A1, A2 and A3 (see all black symbols in Fig. 4). In addition, to characterize the H2O abundances in the parental magmas, we also measured H2O in several glass inclusions in olivines from the most primitive samples of segment A1 (sample 140DS1) and segment A3 (169DS2) (these samples were chosen as starting compositions in our forward calculations). Measured water and chlorine contents of MORB (7–11°S) glasses are given in Table 1.
The H2O concentrations range from 0·05 to 1·01 wt%. This H2O range is in agreement with those found in other studies: typical water contents measured in MORB glasses so far vary in the range of 0·05–0·6 wt% (Michael, 1995; Danyushevsky, 2001) and usually rarely exceed values higher than 1 wt%. Enriched MORB glasses (segment A3) have the highest H2O contents (0·43–1·01 wt%). N-MORB glasses from segment A1 are within a narrow range of H2O concentrations (0·1–0·24 wt%). Segment A2 glasses cluster within these two groups: a few A2 samples have H2O contents similar to those in the primitive samples from segment A3 (0·44–0·53 wt%), and the remaining samples have even lower H2O contents (0·05–0·12 wt%) than the A1 glasses. H2O contents determined in glass inclusions in olivines from the most primitive A1 and A3 samples also cluster within these two groups (grey symbols in Fig. 6). Water contents in glass inclusions (A1: 0·04–0·09 wt% H2O; A3: 0·32–0·55 wt% H2O) are slightly lower than the water contents determined in primitive quenched glass samples (Fig. 6). They may represent the water contents in the parental magmas.
|
, +, Figure 6a demonstrates that H2O is positively correlated with the K2O content of the basaltic glasses. Excluding one sample, the compositional trend of the quenched glasses can be described by a simple linear equation
|
| (1) |
The calculated water abundances from the correlation between water in glasses and K2O contents [equation (1)] are assumed to represent ‘magmatic H2O’. In principle, however, the regression may be affected by assimilation of seawater, because the water contents are determined in erupted MORB glasses (and not in glass inclusions, which may be trapped before contamination occurs). The presence of a seawater component is apparent from the Cl contents determined in A1–A3 glasses (Table 1). The Cl/K ratio is useful to trace the presence of hydrothermally altered material (Michael & Cornell, 1998) and the overall Cl/K in the studied samples ranges from 0·07 to 0·17 (Fig. 6c, Table 1). These values are similar to or higher than the upper limit of the Cl/K range (0·01–0·07) proposed for MORBs that are unaffected by seawater contamination (Michael & Cornell, 1998). In addition, the Cl contents in differentiated A3 lavas significantly exceed the range of Cl enrichment allowed by crystal fractionation (Fig. 6d).
The effect of seawater contamination on the estimation of ‘magmatic H2O’ using equation (1) is difficult to quantify. If the seawater contamination effect was strong, the predicted water contents at a given magmatic K2O content would be higher than those from glass inclusions in the same sample—this is not observed (Fig. 6a). The good correlation between H2O and K2O as well as between H2O and melt Mg-number for different geochemical groups (e.g. segment A1, segment A3) is a further indication that variations in H2O concentrations are inherently magmatic and more probably due to source heterogeneity rather than seawater assimilation.
|
RESULTS FROM FORWARD AND INVERSE MODELING |
|---|
Forward modeling at anhydrous conditions
In our fractional crystallization calculations (forward modeling) we used four starting compositions, representing the parental melts (most magnesian samples) of the segments A1 (140DS1), A2 (161DS1), A3 (169DS2) and A4 (199DS2). The compositions of these starting MORBs are given in the Electronic Appendix (Table S1), which may be downoaded from http://www.petrology.oxfordjournals.org. The compositions of the modeled MORB liquids that have similar composition to their natural counterparts are also given in this table, together with conditions [P, T, H2O (wt%)] at which they can best be produced from the parental melt (H2O in the parental melts is also given). The compositions of the natural counterparts (MORB glasses) have been given by Moeller (2002
). Sample numbers in this study are identical to those of Moeller (2002). Figure 7 shows the results of anhydrous fractional crystallization calculations at different pressures for MORB glasses from segments A1 and A3. The ‘pressure-sensitive’ CaO/Al2O3–MgO diagram is widely used to discriminate pressure of differentiation, as Cpx crystallization occurs earlier at higher pressure when compared with plagioclase and olivine. This affects the CaO/Al2O3 of the residual melts. Figure 7a indicates that A1 glasses are the products of differentiation occurring over a wide pressure range. Other diagrams, such as CaO–MgO or Al2O3–MgO plots (Fig. 7a–c), also indicate a wide range of pressure, varying from 200 MPa to 1 GPa. It should be noted that all calculated isobars are within the uncertainty of the calculations on the FeO–MgO plot (Fig. 7d) and that this diagram can, therefore, not be used for pressure estimations. However, natural A1 glasses have slightly lower FeO concentrations than would be predicted from modeled anhydrous LLDs.
|
The compositional evolution of the A3 glasses indicates nearly isobaric crystallization, considering that the compositional trend is parallel to the calculated isobaric LLDs on all plots (Fig. 7e–h). However, the absolute values of the modeled pressure vary, depending on the major oxide considered. Natural compositions cluster within the range 200–400 MPa on the CaO/Al2O3–MgO plot (Fig. 7e), and along the 100 MPa and 600 MPa isobars on the CaO–MgO and Al2O3–MgO plots, respectively (Fig. 7f and g). The reason for these discrepancies has been formulated by Michael & Chase (1987
) and has been thoroughly discussed by Danyushevsky et al. (1996) and Danyushevsky (2001), who addressed the problem of the effects of small amounts of H2O dissolved in MORB magmas. Although H2O is present in trace abundances, it can significantly suppress the plagioclase crystallization relative to olivine and clinopyroxene (Danyushevsky, 2001). This implies that if the water content in the system is higher than 0·3–0·4 wt% H2O, it should be taken into account to model MORB magma differentiation accurately.
The delay in Pl crystallization means that the strong FeO enrichment and Al2O3 depletion of the residual melts, which is observed in dry tholeiitic systems in which plagioclase predominates over the Fe–Mg silicates in the crystallizing mineral assemblage (e.g. Toplis & Carroll, 1995; Yang et al., 1996), does not occur (see also Figs 5e,f and 7d,h).
To summarize, the results of these preliminary fractional crystallization calculations at anhydrous conditions are acceptable for most of the H2O-poor MORBs of segment A1 (Table 1) but are not satisfactory for H2O-rich MORBs from segment A3. The crystallization model should lead to consistent pressure estimates (pressure intervals) for all major oxide (Fig. 7f–h), which is not the case for segment A3.
Forward modeling: fractional crystallization with different initial H2O contents
General remarks
Following the procedure described above, we assumed that, within each segment, a single parental magma could give rise, by fractionation, to the entire suite of MORB. For segments A1, A2 and A3, approximately 60% of the basaltic glasses could be numerically reproduced from one parental melt, implying that the chemical diversity of most glasses can be explained by a fractional crystallization process occurring at various pressures, temperatures and initial melt H2O content. The calculations have been successful for 25 glasses from segment A1 (from a total of 40), 26 glasses from segment A2 (from a total of 40), and 20 glasses from segment A3 (from a total of 32). For the remaining basaltic glass compositions, which could not be modeled by crystallization of the chosen parental liquid, we emphasize that only small changes in the starting parental melt composition would lead to an overlap between natural and modeled derivative compositions.
Although most of the basalts from segment A4 are very similar to A1 basalts (Fig. 4), only 12 glasses (glasses with MgO ranging from 7·5 to 8·5 wt%) from a total of 32 could be reproduced from the chosen parental melt (199DS2). This is not surprising, as segment A4 samples do not show clear compositional trends. For example, A4 samples with MgO >7·4 wt% show no negative correlation between Na2O and MgO contents (Fig. 4), contrary to what would be expected for melts related by a process of crystal differentiation. Instead, the samples appear to fall into two compositional clusters with lower Na2O and MgO and higher Na2O and MgO (see Fig. 4). Within these clusters, the Na2O and MgO contents of the glasses are positively correlated. All these features probably indicate a variety of parental magmas, as does the existence of groups of A4 magmas with low and high K/Ti ratios (Fig. 4). The high K/Ti segment A4 glasses display small compositional variations and cannot be modeled using the forward approach without reliable determination of the parental melt.
Comparison between modeled and natural MORBs
The compositions of the MORBs and the modeled residual liquids that best reproduce these natural compositions are compared in Fig. 8. The correlation between MgO and CaO/Al2O3 is not shown in this figure, as these values are identical by definition (see also the ‘Methods and data treatment’ section). There is a good agreement for CaO, Na2O, Al2O3 and SiO2, whereas systematic inconsistencies are observed for FeO and TiO2. This is because COMAGMAT systematically produces slightly FeO-enriched and TiO2-depleted compositions, especially in highly differentiated samples. In previous studies (Ariskin, 1999; Ariskin & Barmina, 2004), a similar behavior of FeO in modeled melts was attributed to the lack of parameterization for spinel crystallization in the COMAGMAT model. The difference in K2O and TiO2 between calculated and natural melts may be attributed to variations of these oxides in the parental melts as discussed above. Both elements are not expected to influence the calculated LLDs significantly. Although Yang et al. (1996) noted that the combined effect of TiO2 and K2O on Ol + Pl + Cpx-saturated melts shifts the composition of melts coexisting with Ol + Pl + Cpx to higher Al and lower Ca and Fe contents, we emphasize that these differences (<0·1 wt% for FeO, <0·02 wt% for CaO and Al2O3) are significantly lower than the analytical uncertainties of electron probe microanalysis (see below). Yang et al. (1996) noted that the effect of TiO2 and K2O on calculated LLD is definitely more pronounced in Na-rich alkaline basalts, but such compositions were not considered in the course of this study.
|
Crystallization conditions
Figure 9a demonstrates that basalts from the northern A1 segment experienced crystallization over a range of pressures varying from 600 to 200 MPa and in the temperature interval 1250–1170°C. Almost all ‘numerically’ reproduced A1 basalts have been modeled at anhydrous conditions. It was necessary to assume water-bearing conditions (0·2 wt% H2O) for a few samples (Electronic Appendix, Table S1) to obtain chemical similarity to the natural samples. The degree of fractionation for the majority of A1 MORB varied from 11 to 38 wt%.
|
Segment A2 MORBs show the largest range of calculated pressures (900–200 MPa) and temperatures (1250–1130°C). According to our calculations, 30% of modeled MORB can be reproduced at dry conditions (H2O in parental melt = 0) and 70% in the presence of small amounts of water. Among these latter samples, the best fits are obtained with 0·2 wt% H2O (for 30% of MORB) or 0·5 wt% H2O (for 70% of MORB) in the parental melt. The degree of fractionation varies from 2 to 40 wt%.
Twelve samples from the southern segment A4 appear to have crystallized in the pressure range 480–230 MPa and temperature range 1220–1175°C (Fig. 9a). Only two samples have been modeled in dry conditions; the remaining MORB compositions can be simulated assuming 0·2 wt% of H2O in the parental melt (Electronic Appendix, Table S1). The degree of fractionation varies in the range from 7 to 20%.
The calculations suggest that the A3 MORBs have the highest parental magma H2O contents (0·5–1·0 wt%; Electronic Appendix, Table S1). Crystal fractionation of between 7 and 40% leads to residual melts with water contents up to 1·6 wt% H2O. Crystallization appears to have occurred at nearly isobaric conditions within a pressure interval between 300 and 200 MPa (Fig. 9a and b). The basaltic melts evolved in the temperature interval 1170–1070°C, which is significantly lower than that calculated for A1, A2 and A4 MORBs. The low temperatures obtained for A3 MORBs result from the combined effects of lower crystallization pressures, more evolved character of the parental melt (7 wt% MgO instead of 8–9 wt% MgO in A1 and A4) and the relatively high amount of H2O.
Inverse modeling: equilibrium crystallization with known H2O contents
In our equilibrium crystallization calculations (inverse modeling) all natural A1 to A4 MORB glasses were considered. The conditions [P, T, H2O (wt%)] at which each MORB composition is in equilibrium with Ol + Cpx + Plag (conditions of multiple saturation) are presented in the Electronic Appendix (Table S2).
General remarks
The option of equilibrium crystallization in COMAGMAT was implemented to simulate the conditions of multiple saturation (Ol + Pl + Cpx) for the studied basaltic glasses. Calculations were carried out in two steps. First, for the given MORB composition with the estimated H2O content [see equation (1)], calculations were performed in the pressure range 0·1–900 MPa with a pressure increment of 100 MPa to check for the stable mineral assemblage within the first 3–5% of crystallization. These preliminary calculations allowed us to bracket a pressure interval in which the mineral crystallization sequence was changing, and in which Ol, Pl and Cpx crystallize simultaneously (see Fig. 3). In the next step, the pressure increment was reduced to 10 MPa, and calculations were performed over a narrower pressure range. This second step allowed us to constrain the conditions (temperature and pressure; see Fig. 3 for a typical example) at which Ol, Pl and Cpx would crystallize simultaneously (within a crystallization degree of 1%). The inverse calculations have been successful for more than 80% of studied samples from segments A1 to A3: for 34 glasses from segment A1 (from a total of 40), 37 glasses from segment A2 (from a total of 40), and 28 glasses from segment A3 (from a total of 32). Amongst the N-MORB glasses of segment A4 (20 samples) only 14 samples were modeled. Calculations for the enriched MORBs of segment A4 were successful for all samples from the dredge station 196 (four samples; Moeller, 2002) and failed for all samples from the dredge station 197 (seven samples; Moeller, 2002). For the samples that could not be modeled by the inverse approach (20% of total dataset), the main problem was related to the difficulty in achieving simultaneous crystallization of Pl with the other two phases Ol and Cpx. These glasses did not represent a cotectic composition within the investigated range of pressures and were either ‘plagioclase-oversaturated’ or ‘plagioclase-undersaturated’. Either such MORB compositions are affected by processes such as magma mixing or crystal accumulation and dissolution (Danyushevsky, 2001), or the estimation of the melt water content following equation (1) is not appropriate for these samples—they were excluded from further consideration.
Crystallization conditions
Pressures and temperatures of multiple saturation obtained for the basaltic glasses are shown in Fig. 9c. This dataset is in good general agreement with the results of our forward modeling (Fig. 9a), which was successful for only a limited number of samples (see above). The most important difference is observed for the estimation of pressure in segment A3. The data from inverse modeling indicate a wider pressure range (300–100 MPa) than that obtained from forward modeling (300–200 MPa). In the following discussion, the values of P–T estimates obtained in the course of inverse modeling are used to discuss the pre-eruptive conditions of MORBs.
There are two important observations arising from the results of our modeling. First, Fig. 9a and Fig. 9c demonstrate contrasting crystallization conditions between depleted N-MORBs (A1, A2 and A4) and geochemically enriched A3 MORB magmas. Within the A1, A2 and A4 segments, crystallization occurs over a wide range of pressures (900–200 MPa). This polybaric evolution yields compositions ranging from 9 to 7 wt% MgO (Fig. 9b). The absence of differentiated samples within these segments may indicate the lack of significant magma reservoirs in which differentiation could proceed. In contrast, enriched A3 MORB magmas display the lowest pressures of partial crystallization, as well as the smallest pressure interval (300–100 MPa) (Fig. 9c). This low pressure interval and the lack of primitive samples in segment A3 point to the possible existence of magma reservoirs cooling at shallow depths. Such magma chambers may be stable for longer periods of time, allowing primitive mantle-derived compositions to stagnate and differentiate to the erupted evolved ferro-basaltic compositions. It should be emphasized that the low pressures obtained for the A3 MORB do not necessarily mean that this segment is dominated by low-pressure isobaric crystallization only. Probably the earlier (high-pressure) crystallization history was simply overprinted by later equilibration at low pressures.
The second observation is related to the basaltic melts of segments A1, A2 and A4, which also show systematic differences in pre-eruptive conditions. Each segment forms a distinct polybaric path of magmatic evolution (Fig. 9). For example, at the same MgO content (index of differentiation) segment A2 MORBs show higher and most segment A4 MORBs show lower pressures of partial crystallization than those of segment A1 (Fig. 9b and d). The difference between A1 and A2 MORBs is more evident on the P–T plot (Fig. 9c), where A2 samples indicate crystallization temperatures that are systematically lower (10°C) than those of segment A1. The A4 N-MORBs, which show the lowest pressures for a given MgO content, tend to have the highest temperatures (Fig. 9c). We note that the pre-eruptive conditions of enriched (high K/Ti ratio) MORBs from segment A4 differ from those of the A4 N-MORBs and are similar to those of A2 samples.
These observations may indicate either that there are compositional differences between the parental magmas feeding the magmatic systems beneath each segment or, assuming similar parental melts for each segment, that crystallization and differentiation follow different P–T paths for each segment (Fig. 9).
|
DISCUSSION |
|---|
Uncertainties of the calculations
Forward modeling
As stated above, the natural glass was considered to be successfully modeled when the correspondence between its composition and that of the simulated residual liquid was within the analytical precision (2
) of electron microprobe analysis. The values of 2 (listed in wt%) were determined by 175 replicate measurements of the basaltic glass 140DS1 using a Cameca SX100 electron microprobe at the University of Hannover as follows: 0·73 (SiO2), 0·37 (Na2O), 0·48 (CaO), 0·44 (Al2O3), 0·31 (MgO) and 0·79 (FeO). For comparison, the compositional range of the 175 analyses of glass 140DS1 is shown in Fig. 2 (grey crosses) together with an average microprobe composition of 140DS1 (black star in a circle) obtained by Moeller (2002) at the University of Kiel using a JEOL Superprobe (star in the circle). Interlaboratory differences are evident in Fig. 2. Considering these interlaboratory differences, the heterogeneity of natural quenched glasses (Langmuir et al., 1992), and the analytical uncertainties of the electron microprobe itself, the resulting 2 values used in this study are higher than the typical analytical uncertainty of electron microprobe analyses recorded in most studies on natural MORBs. However, we prefer to apply these large 2 values, because the overall uncertainty defines the precision in predicting the absolute values of crystallization pressure. We can quantify the effects of these errors on pressure estimates. In a dry system using 140DS1 as the starting composition, a change of 0·96 wt% in the CaO content of the starting melt (variations within ±2, from 11·71 to 12·67 wt% CaO) results in calculated pressure variations of 80 MPa. An MgO variation within 2 values (±0·3 wt% MgO) leads to a maximal variation in calculated pressure of 110 MPa. These values are lower than those assumed for the accuracy of mineral–melt geothermobarometers (±150 MPa) in the COMAGMAT model (Ariskin & Barmina, 2004). The uncertainty on the evaluation of water contents in the parental melt is illustrated in Fig. 2c and d. Al2O3 and FeO are the most sensitive oxides to constrain the water content. The example of Fig. 2c and d shows that the natural glass is best reproduced with an H2O content of 0·2 wt% in the initial parental melt. However, modeled compositions with 0·0, 0·5 and 0·7 wt% H2O in the parental melt are also within the error bars. Thus, the typical uncertainty on absolute water content estimations, using our forward approach, is estimated to be up to 0·7 wt% H2O.
Inverse modeling
Although the inverse approach seems to be more effective in evaluating the conditions of MORB partial crystallization (if the H2O content is known), it is also limited by the ability of the model to predict simultaneous (three-phase mineral association) crystallization with sufficient precision. Our calculations demonstrate that the stability curves of Ol, Pl and Cpx intersect within ±1°C (e.g. see Fig. 3) for 50% of the modeled MORB compositions, allowing pressure to be determined within ±20 MPa. For the other samples, pressure can be modeled with a precision of ±100 MPa or less, assuming minor temperature corrections to reach multiple saturation (Ol + Pl + Cpx). These corrections did not exceed 10°C and average values are ±4·7°C for Pl, ±3·5°C for Ol and ±2·2°C for Cpx (see Electronic Appendix, Table S2). It should be noted that these temperature corrections are always lower than the uncertainty of the mineral–melt geothermometer (±10°C at 0·1 MPa and 15–20°C at elevated pressures) reported for the COMAGMAT model (Ariskin & Barmina, 2004).
Pressure estimates—comparison with other models
Pressure estimates obtained in this study (inverse modeling) are compared with crystallization pressures derived from the models of Danyushevsky et al. (1996), Yang et al. (1996), Herzberg (2004) and Villiger et al. (2007) in Fig. 10. All models are roughly consistent and give similar pressure estimates varying within the uncertainties of the calculations (depending on the studies, these are between 100 and 200 MPa). For segments A1, A2 and A4, a large pressure range is observed up to 1 GPa. In contrast, for the relatively ‘water-rich’ segment A3, a narrow, low-pressure range is obtained (0·1–300 MPa; Fig. 10). Our pressure estimates show the best agreement with the model of Yang et al. (1996) (Fig. 10a). The absolute values of pressure are almost identical in the interval 400–800 MPa, but the lowest pressures are slightly higher than those predicted by the method of Yang et al. (1996). Pressures calculated using the models of Danyushevsky et al. (1996) and Herzberg (2004) are very close to pressure estimates obtained in our study for the water-rich MORBs of segment A3. However, our calculated pressures for depleted MORBs with low water contents are systematically 150–200 MPa and 200–250 MPa higher than those calculated by the model of Danyushevsky et al. (1996) and Herzberg (2004), respectively (Fig. 10c and d). Pressure estimates derived from the recent empirical equation of Villiger et al. (2007) scatter around the one-to-one line (Fig. 10b) but are within the uncertainty (150–200 MPa).
|
In a previous comparative study, Michael & Cornell (1998
) emphasized the differences in calculated crystallization pressures derived by using the models of Danyushevsky et al. (1996) and Yang et al. (1996). For a large number of slow- to fast-spreading ridge MORB suites Michael & Cornell (1998) showed that, although these two models are qualitatively consistent, systematic pressure deviations up to 300 MPa are observed between the models for some MORB suites [see also fig. 1 of Michael & Cornell (1998)]. The same type of deviation is observed between our inverse approach and the models of Danyushevsky et al. (1996) and of Herzberg (2004) for samples from segments A1–A4 (Fig. 10c and d). As already noted by Michael & Cornell (1998), it is difficult to evaluate which model leads to the most accurate result.
The effect of small amounts of H2O on pressure estimates for MORB
In previous studies it has been argued that the effect of H2O on estimates of the pressure of partial crystallization in MORB is not significant (Herzberg, 2004) or is unsystematic (Villiger et al., 2007), because the H2O contents in MORBs are too low to produce an apparent displacement of the mineral cotectics in the pseudoternary projections. Calculations accounting for the effect of H2O on the LLDs performed in this study clearly show that the presence of water results in a decrease in the pressure and temperature at which multiple saturation occurs. However, the effect of water on pressure estimates is small compared with the precision of the calculations. Figure 3 demonstrates that in the presence of 0·3 wt% H2O the condition of multiple saturation in the investigated MORB composition occurs at 1192°C and 540 MPa. Under anhydrous conditions, multiple saturation for the same MORB composition occurs at higher pressure (600 MPa) and temperature (1215°C). The same effect can be seen in results from forward modeling (Fig. 2b). The black symbols on the calculated LLDs represent liquids that have the same MgO and CaO/Al2O3 ratio as a natural MORB (open circle with error bars). It is evident that an increase of water content would shift the pressure to a lower value. In these two examples the pressure change related to the presence of H2O ( 0.3 wt %) in MORB melts is within 50–60 MPa. However, it should be noted that small systematic inconsistencies when comparing ‘dry’ barometric models with our two approaches may be due to the presence of water in the natural systems. For example, the generally lower pressures (with very low or even negative pressure values) observed for segment A3 using the model of Yang et al. (1996; Figs 10a and 11) may be the result of these calculations assuming anhydrous conditions.
|
H2O·hH2O)/In conclusion, the estimation of the pressure of partial crystallization in MORBs evolving along the Ol + Pl + Cpx cotectic does not require knowledge of the melt water content. Because the effect of water (in the range 0–1 wt% H2O) on the pressure at which multiple saturation is reached is comparable with the pressure uncertainty reported for most experimental apparatus, the precision of the barometers cannot be improved.
Depths of partial crystallization along the MAR near Ascension (7–11°S)
The calculations of partial crystallization pressures provide a snapshot of magma storage and distribution along the spreading axis. Crystallization pressures obtained by different models are converted to depths and plotted vs latitude of sampling in Fig. 11. Crystal fractionation is calculated to have last occurred at upper mantle depths and in the oceanic crust beneath N-MORB segments A1, A2 and A4. In contrast, the water-bearing ferro-basaltic magmas collected along segment A3 last underwent crystallization only within the oceanic crust.
The high crystallization pressures obtained for most N-MORB, particularly from segments A1 and A2, and to a lesser extent A4, are consistent with data for typical slow-spreading ridges reported by Dmitriev (1998), Michael & Cornell (1998) and Herzberg (2004). Those researchers applied the pressure-estimation techniques to a global MORB glass database and noted a correlation between crystallization pressure and spreading rate, extent of melting, and ridge segmentation. Crystallization at high pressure beneath slow-spreading ridges is the result of a relatively cold mantle environment, low melt fraction in the mantle, low magma supply and little heat transfer to the surface (Michael & Cornell, 1998; Herzberg, 2004). The upper bound of crystallization pressures for samples from segment A1 is consistently subcrustal, becoming deeper towards the north (Fig. 10), suggesting the influence of a colder mantle north of the ridge–transform intersection (Shen & Forsyth, 1995).
The crustal depths of magma evolution recorded in the chemistry of segment A3 MORB are more typical for fast-spreading systems (Michael & Cornell, 1998; Herzberg, 2004) than for ridges with slow spreading rates. This is not exceptional along the MAR and similar observations have been made for the slow-spreading Reykjanes Ridge (57°N, MAR). Here, crystallization at low pressures was suggested to be related to the influence of the hot Icelandic mantle plume (e.g. Herzberg, 2004), which, through lateral channeling into the ridge, heats the surrounding mantle and hampers extensive partial crystallization at elevated depths. In the study area there is no indication of the presence of a hot mantle plume. The existence of a thicker crust of 11 km beneath segment A3 [in contrast to 5 km beneath A1 as found by Minshull et al. (1998) and Bruguier et al. (2003)] is, however, a clear indication for a high magma supply in this region—it may be this that leads to warming of the mantle through repeated melt transit and hinders crystallization at mantle depths.
|
CONCLUSION: TOWARDS AN ACCURATE PREDICTIVE MODEL FOR PRE-ERUPTIVE CONDITIONS OF MORB |
|---|
|
TOPABSTRACTINTRODUCTIONGEOLOGICAL BACKGROUND AND SAMPLE...METHODS AND DATA TREATMENTSRESULTSRESULTS FROM FORWARD AND...DISCUSSIONCONCLUSION: TOWARDS AN ACCURATE...SUPPLEMENTARY DATAREFERENCES |
|---|
The methods applied in this study not only are barometers but also have the potential to constrain temperature and relative water content from the composition of quenched MORB glasses, if the genetic relationships with a parental MORB can be well defined. The efficiency of our approach to constrain pressure of crystallization is comparable with that of other models (Danyushevsky et al., 1996
; Yang et al., 1996; Herzberg, 2004; Villiger et al., 2007) but the advantage of our method is that it takes the effect of water on LLDs into account. For example, the forward approach used in this study is helpful in constraining pressure and simultaneously discriminating between nearly anhydrous magmas (e.g. segment A1) and magmas in which water affects the LLDs (e.g. segment A3). If the water content of the melts is known from an independent method, the inverse approach is also a powerful way to constrain pre-eruptive temperatures. Although MORB magmas are considered to be almost anhydrous or strongly water-undersaturated, knowledge of the effects of small amounts of water on the liquidus temperatures of olivine, plagioclase and clinopyroxene is crucial for the successful modeling of phase relations in MORB. However, the quantitative effect of low water concentrations on the liquidus temperatures of the main silicate phases in basaltic systems is poorly known, as very few experiments in which H2O activities were determined have been carried out on relevant systems. Such information is, however, crucial for all models that aim to quantify correctly the point of multiple saturation (olivine + plagioclase + clinopyroxene) as a function of pressure, temperature and water content.
|
SUPPLEMENTARY DATA |
|---|
Supplementary data for this paper are available at Journal of Petrology online.
|
ACKNOWLEDGEMENTS |
|---|
This is publication 14 of the priority program SPP1144 ‘From Mantle to Ocean: Energy-, Material- and Life-cycles at Spreading Axes’ of the German Science Foundation (DFG) (Project Ho1337/10). We thank Alexei Ariskin and Jun-Ichi Kimura for helpful reviews of an earlier draft. The authors wish to thank Othmar Müntener and three anonymous reviewers for their suggestions, which allowed us to improve the clarity and quality of the paper considerably. The authors acknowledge Wendy Bohrson for her editorial comments and efforts.
*Corresponding author. Telephone: + 49 511 762-2443. Fax: + 49 511 762-3045. E-mail: r.almeev{at}mineralogie.uni-hannover.de
|
REFERENCES |
|---|
Almeev RR, Ariskin AA. Mineral–melt equilibria in a hydrous basaltic system: computer modeling. Geochemistry International (1996) 34:563–573.
Almeev RR, Ariskin A, Pletchov P Yu. Calculations of mineral-melt equilibria in tholeiitic system: MELTS versus COMAGMAT. Lithos (2004) 73:1–2. S1.[Web of Science]
Ariskin AA. Phase equilibria modeling in igneous petrology: use of COMAGMAT model for simulating fractionation of ferro-basaltic magmas and the genesis of high-alumina basalt. Journal of Volcanology and Geothermal Research (1999) 90:115–162.[CrossRef][Web of Science]
Ariskin AA, Barmina GS. COMAGMAT: development of a magma crystallization model and its petrological applications. Geochemistry International (2004) 42:S1–S157.[Web of Science]
Ariskin AA, Tsekhonya TI, Frenkel MY. Computerized barometry and genetic interpretation of Central Atlantic basalt glasses. Geochemistry International (1992) 29:117–125.
Ariskin AA, Frenkel MY, Barmina GS, Nielsen RL. COMAGMAT: a Fortran program to model magma differentiation processes. Computers and Geosciences (1993) 19:1155–1170.[CrossRef]
Asimow PD, Dixon JE, Langmuir CH. A hydrous melting and fractionation model for mid-ocean ridge basalts: Application to the Mid-Atlantic Ridge near the Azores. Geochemistry, Geophysics, Geosystems (2004) 5. Q01E16, doi:10.1029/2003GC000568.
Bender JF, Hodges FN, Bence AE. Petrogenesis of basalts from the project Famous Area: experimental study from 0 to 15 kbars. Earth and Planetary Science Letters (1978) 41:277–302.[CrossRef][Web of Science]
Berndt J, Koepke J, Holtz F. An experimental investigation of the influence of water and oxygen fugacity on differentiation of MORB at 200 MPa. Journal of Petrology (2005) 46:135–167.
Bezos A, Humler E. The Fe3+/Fe ratios of MORB glasses and their implications for mantle melting. Geochimica et Cosmochimica Acta (2005) 69:711–725.[CrossRef][Web of Science]
Bourdon E, Hemond C. Looking for the ‘missing endmember’ in South Atlantic Ocean mantle around Ascension Island. Mineralogy and Petrology (2001) 71:127–138.[CrossRef][Web of Science]
Bruguier NJ, Minshull TA, Brozena JM. Morphology and tectonics of the Mid-Atlantic Ridge, 7°–12°S. Journal of Geophysical Research (2003) 108. doi:10.1029/2001JB001172.
Danyushevsky LV. The effect of small amounts of H2O on crystallisation of mid-ocean ridge and backarc basin magmas. Journal of Volcanology and Geothermal Research (2001) 110:265–280.[CrossRef][Web of Science]
Danyushevsky LV, Sobolev AV, Dmitriev LV. Estimation of the pressure of crystallization and H2O content of MORB and BABB glasses: calibration of an empirical technique. Mineralogy and Petrology (1996) 57:185–204.[CrossRef][Web of Science]
Dmitriev LV. Chemical variability of mid-ocean ridge basalts as a function of the geodynamic setting of their formation. Petrology (1998) 6:314–334.[Web of Science]
Fisk MR, Bence AE. Experimental crystallization of chrome spinel in FAMOUS basalt 527-1-1. Earth and Planetary Science Letters (1980) 48:111–123.[CrossRef][Web of Science]
Fisk MR, Schilling JG, Sigurdsson H. An experimental investigation of Iceland and Reykjanes ridge tholeiites: I. Phase relations. Contributions to Mineralogy and Petrology (1980) 74:361–374.[Web of Science]
Fontignie D, Schilling JG. Mantle heterogeneities beneath the South Atlantic: a Nd–Sr–Pb isotope study along the Mid-Atlantic Ridge (3°S–46°S). Earth and Planetary Science Letters (1996) 142:209–221.[CrossRef][Web of Science]
Freise M. Differenzierung von Basalten einer ‘Large Igneous Province’ am Beispiel des Kerguelen Plateaus: eine experimentelle Studie. (2004) 105. PhD thesis, Leibniz University, Hannover.
Fujii T, Bougault H. Melting relations of a magnesian abyssal tholeiite and the origin of MORBs. Earth and Planetary Science Letters (1983) 62:283–295.[CrossRef][Web of Science]
Fujii T, Scarfe CM. Composition of liquids coexisting with spinel lherzolite at 10 kbar and the genesis of MORBs. Contributions to Mineralogy and Petrology (1985) 90:18–28.[CrossRef][Web of Science]
Gaetani GA, Grove TL, Bryan W. Experimental phase relations of basaltic andesite from hole 839B under hydrous and anhydrous conditions. In: Proceedings of the Ocean Drilling Program, Scientific Results, 135—Hawkins J, Parson L, Allan J, eds. (1994) College Station, TX: Ocean Drilling Program. 557–565.
Graham DW, Jenkins WJ, Schilling J.-G, Thompson G, Kurz MD, Humphris SE. Helium isotope geochemistry of mid-ocean ridge basalts from the South Atlantic. Earth and Planetary Science Letters (1992) 110:133–147.[CrossRef][Web of Science]
Grove TL, Bryan WB. Fractionation of pyroxene-phyric MORB at low pressure: an experimental study. Contributions to Mineralogy and Petrology (1983) 84:293–309.[CrossRef][Web of Science]
Grove TL, Kinzler RJ, & Bryan W. Natural and experimental phase relations of lavas from Serocki Volcano. In: Proceedings of the Ocean Drilling Program, Scientific Results, 106/109—Detrick R, Honnorez J, et al, eds. (1990) College Station, TX: Ocean Drilling Program. 9–17.
Grove TL, Kinzler RJ, Bryan WB. Fractionation of mid-ocean ridge basalt (MORB). In: Mantle Flow and Melt Generation at Mid-Ocean Ridges. Geophysical Monograph, American Geophysical Union—Morgan J, Blackman D, Sinton J, eds. (1992) 71:281–310.
Hanan BB, Kingsley RH, Schilling JG. Pb isotope evidence in the South Atlantic for migrating ridge–hotspot interactions. Nature (1986) 322:137–144.[CrossRef]
Herzberg C. Partial crystallization of mid-ocean ridge basalts in the crust and mantle. Journal of Petrology (2004) 45:2389–2405.
Hirschmann MM, Baker MB, Stolper E. The effect of alkalis on the silica content of mantle-derived melts. Geochimica et Cosmochimica Acta (1998) 62:883–902.[CrossRef][Web of Science]
Holtz F, Becker A, Freise M, Johannes W. The waterundersaturated and dry Qz–Ab–Or system revisited. Experimental results at very low water activities and geological implications. Contributions to Mineralogy and Petrology (2001) 141:347–357.[Web of Science]
Juster TC, Grove TL, Perfit MR. Experimental constraints on the generation of Fe–Ti basalts, andesites, and rhyodacites at the Galapagos spreading center, 85°W, and 95°W. Journal of Geophysical Research (1989) 94:9251–9274.
Kagi R, Muntener O, Ulmer P, Ottolini L. Piston-cylinder experiments on H2O undersaturated Fe-bearing systems: An experimental setup approaching fO2 conditions of natural calc-alkaline magmas. American Mineralogist (2005) 90:708–717.
Kinzler RJ, Grove TL. Primary magmas of mid-ocean ridge basalts. 1. Experiments and methods. Journal of Geophysical Research (1992) 97:6885–6906.
Langmuir CH, Klein E, Plank T. Petrological systematics of mid-ocean ridge basalts: constraints on melt generation beneath ocean ridges. In: Mantle Flow and Melt Generation at Mid-Ocean Ridges. Geophysical Monograph, American Geophysical Union—Morgan J, Blackman D, Sinton J, eds. (1992) 71:183–277.
Michael P. Regionally distinctive sources of depleted MORB: Evidence from trace elements and H2O. Earth and Planetary Science Letters (1995) 131:301–320.[CrossRef][Web of Science]
Michael PJ, Chase RL. The influence of primary magma composition, H2O and pressure on mid-ocean ridge basalt differentiation. Contributions to Mineralogy and Petrology (1987) 96:245–263.[CrossRef][Web of Science]
Michael PJ, Cornell WC. Influence of spreading rate and magma supply on crystallization and assimilation beneath mid-ocean ridges: Evidence from chlorine and major element chemistry of mid-ocean ridge basalts. Journal of Geophysical Research (1998) 103:18325–18356.[CrossRef]
Minshull TA, Bruguier NJ, Brozena JM. Ridge–plume interactions or mantle heterogeneity near Ascension Island? Geology (1998) 26:115–118.
Minshull TA, Bruguier NJ, Brozena JM. Seismic structure of the Mid-Atlantic Ridge, 8–9°S. Journal of Geophysical Research (2003) 108(B11)). 2513, doi:10.1029/2002JB002360.
Moeller H. Magma genesis and mantle sources at the Mid-Atlantic Ridge east of Ascension Island. (2002) 149. PhD thesis, Christian-Albrechts Universität, Kiel.
Myers JD, Johnston AD. Phase equilibria constraints on models of subduction zone magmatism (overview). In: Subduction. Top to Bottom. Geophysical Monograph, American Geophysical Union—Bébout GE, ed. (1996) 96:229–251.
Schilling JG, Thompson G, Kingsley R, Humphris S. Hotspot-migrating ridge interaction in the South Atlantic. Nature (1985) 313:187–191.[CrossRef]
Shen Y, Forsyth DW. Geochemical constraints on initial and final depths of melting beneath midocean ridges. Journal of Geophysical Research—Solid Earth (1995) 100:2211–2237.[CrossRef]
Sisson TW, Grove TL. Experimental investigations of the role of H2O in calc-alkaline differentiation and subduction zone magmatism. Contributions to Mineralogy and Petrology (1993) 113:143–166.[CrossRef][Web of Science]
Stolper EM. A phase diagram for mid-ocean ridge basalts: preliminary results and implications for petrogenesis. Contributions to Mineralogy and Petrology (1980) 74:13–27.[CrossRef][Web of Science]
Stolper E. Water in silicate glasses: an infrared spectroscopic study. Contributions to Mineralogy and Petrology (1982) 81:1–17.[CrossRef][Web of Science]
Toplis MJ, Carroll MR. An experimental study of the influence of oxygen fugacity on Fe–Ti oxide stability, phase relations, and mineral–melt equilibria in ferro-basaltic systems. Journal of Petrology (1995) 36:1137–1170.
Thy P, Lofgren GE. Experimental constraints on the low-pressure evolution of transitional and mildly alkalic basalts: the effect of Fe–Ti oxide minerals and the origin of basaltic andesites. Contributions to Mineralogy and Petrology (1994) 116:340–351.[CrossRef][Web of Science]
Thy P, Lesher CE, Fram M. Low-pressure experimental constraints on the evolution of basaltic lavas from site 917, southeast Greenland continental margin. In: Proceedings of the Ocean Drilling Program, Scientific Results, 152—Saunders AD, Larsen HC, Wise SW Jr, eds. (1998) College Station, TX: Ocean Drilling Program. 359–372.
Thy P, Lesher CE, & Mayfield J. Low-pressure melting studies of basalt and basaltic andesite from the southeast Greenland continental margin and the origin of dacites at site 917. In: Proceedings of the Ocean Drilling Program, Scientific Results, 163—Larsen HC, Duncan RA, Allan JF, Brooks K, et al, eds. (1999) College Station, TX: Ocean Drilling Program. 95–112.
Tormey DR, Grove TL, Bryan WB. Experimental petrology of normal MORB near the Kane Fracture Zone: 22°–25°N, mid-Atlantic ridge. Contributions to Mineralogy and Petrology (1987) 96:121–139.[CrossRef][Web of Science]
Villiger S, Ulmer P, Muntener O. Crystallization of MORB in the mantle: implications from major element variations of glasses from equilibrium and fractional crystallization experiments. Journal of Geophysical Research, Solid Earth (2007) 112. B01202, doi:10.1029/2006JB004342.
Walker D, Shibata T, DeLong SE. Abyssal tholeiites from the Oceanographer Fracture Zone II. Phase equilibria and mixing. Contributions to Mineralogy and Petrology (1979) 70:111–125.[CrossRef][Web of Science]
Yang H-J, Kinzler RJ, Grove TL. Experiments and models of anhydrous, basaltic olivine–plagioclase–augite saturated melts from 0·001 to 10 kbar. Contributions to Mineralogy and Petrology (1996) 124:1963–1973.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||