Earth and Planetary Science Letters 252 (2006) 423–436 www.elsevier.com/locate/epsl Origin of high-Al N-MORB by fractional crystallization in the upper mantle beneath the Galápagos Spreading Center ⁎ Deborah Eason , John Sinton Department of Geology and Geophysics, School of Ocean and Earth Science and Technology, University of Hawaii, Honolulu, HI, USA Received 24 May 2006; received in revised form 30 September 2006; accepted 30 September 2006 Available online 15 November 2006 Editor: R.W. Carlson Abstract The Galápagos Spreading Center (GSC) includes lavas with chemical compositions ranging from N-MORB to transitional and more enriched MORB. N-MORB dominate the region west of 95.5°W, far from the influence of the Galápagos hotspot. However, some N-MORB glass samples from the GSC have very high Al contents (N16.0 wt.% Al2O3 at N8.5 wt.% MgO), similar to high-Al N- MORB from other slow and intermediate spreading ridges and close to fracture zones elsewhere. GSC high-Al N-MORB are dominated mineralogically by uniform plagioclase compositions (An 80–82) with only 1–2% olivine (Fo 85–87), and have glass compositions with higher Al2O3 and lower SiO2 than is predicted by normal MORB fractionation trends. Forward modeling using the MELTS and pMELTS algorithms constrained by crustal thickness measurements indicates that high-Al, low-Si MORB can be produced by high-pressure crystallization in the upper mantle using the same source as normal (low-Al) GSC N-MORB. Although high-Al glasses can be obtained by very low extents of partial melting of this mantle source, such melting models result in significant misfits in other major element oxides, especially SiO2. Models involving significant evolution with up to 20% olivine and clinopyroxene crystallization at pressures of 0.3–0.4 GPa can account for the complete major and selected trace element compositions of these unusual MORB samples. We suggest that high-pressure fractionation is enhanced by conductive cooling of the upper mantle in this area of the GSC, consistent with other recent models correlating mantle crystallization with slow spreading mid-ocean ridges and fracture zones. © 2006 Elsevier B.V. All rights reserved. Keywords: MORB; mid-ocean ridge; Galápagos rift; basalt fractionation; pMELTS; mantle 1. Introduction Experiment (G-PRIME), which included detailed bathy- metric mapping, and the collection of seismic refraction, The nearly east–west trending GSC separates the reflection, and magnetics data and rock samples along Cocos and Nazca plates in the eastern equatorial Pacific an ∼800 km-long portion of the ridge extending from (Fig. 1). Samples from this study were collected during 90.5°W, just north of the Galápagos Archipelago and the Galápagos Plume Ridge Interaction Multidisciplinary the inferred location of the Galápagos mantle plume, to 98°W, the region of the GSC considered to be a “normal” mid-ocean ridge [1]. Full spreading rates along this ⁎ Corresponding author. Tel.: +1 808 956 9544. portion of the ridge vary from 46 mm/yr at 98°W to E-mail address: [email protected] (D. Eason). 56 mm/yr at 91°W near the Galápagos hotspot [2]. 0012-821X/$ - see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.epsl.2006.09.048 424 D. Eason, J. Sinton / Earth and Planetary Science Letters 252 (2006) 423–436 Fig. 1. (a) Sample locations along part of the western Galápagos Spreading Center. Station numbers are those for the G-PRIME expedition [1,3]. ALVIN (ALV) samples from Hey et al. [63]. High-Al sample locations are shown as triangles. Thick black line denotes the GSC axis from Sinton et al. [63]. (b) Global distribution of samples in PetDB [64] with Al2O3 N16.0 wt.%, MgON8.5 wt.%. Open squares denote fracture zone samples. The G-PRIME sample suite includes chemical com- 1.68 wt.% TiO2, 0.05–0.12 wt.% P2O5), with K/Ti ratios positions varying from normal (N-) to transitional (T-) b0.09. GSC N-MORB also have low water contents and more enriched (E-) mid-ocean ridge basalts (MORB) (no greater than 0.22 wt.% H2O, with an average of [1,3]. N-MORB dominate the area west of the propa- 0.16 wt.% H2O). Cushman et al. [3] argued that av- gating rift at 95.5°W, which appears to be the limit of erage Galápagos N-MORB could be produced from a geochemical influence of the Galápagos hotspot. Sam- mean fraction of partial melting (F¯) ∼0.06 of a source ples west of the propagating rift show little or no evi- with ∼34±1 ppm K, 133±3 ppm H2O, 2250±50 ppm dence of plume enrichment, while samples to the east Na2O, and 1050±25 ppm Ti. show increasing incompatible element and isotopic en- Some of the N-MORB glass samples from the GSC richment approaching the plume (∼91.7°W) [3–7]. have unusually high Al2O3 contents (N16.0 wt.%) at Galápagos N-MORB are characterized by moderate MgON8.5 wt.%. These samples have higher aluminum to high MgO values (ranging from 6.9 wt.% MgO to and lower silica than predicted by normal MORB frac- nearly 10 wt.% MgO) and low concentrations of ele- tionation trends while maintaining very low potassium ments that are incompatible during mantle melting and sodium values. According to MELTS models of (≤0.08 wt.% K2O, 1.67–2.56 wt.% Na2O, 0.77– their glass compositions [8], they are saturated with D. Eason, J. Sinton / Earth and Planetary Science Letters 252 (2006) 423–436 425 plagioclase at low pressures even at MgON9.0 wt.%, tube. Whole rocks were crushed in an alumina swing consistent with experimental data reported for other mill, and powders were analyzed for major elements on high-Al MORB (e.g., [9,10]). High-MgO basalts with fused disks following methods similar to those of unusually high aluminum content also have been Norrish and Hutton [18]. Trace elements were analyzed reported for various mid-ocean ridges (see Fig. 1), on pressed powder pellets. Peak intensities for the trace although their origin is not fully understood (e.g., [11– elements were corrected for backgrounds, line interfer- 13]). These high-Al MORB are distinct from the high-Al ences and matrix absorption using methods similar to basalts and andesites often associated with arcs, which Chappell [19]. Corrected intensities were calibrated tend to be more evolved with higher K and H2O contents. against a wide range of natural rock standards. Accuracy Such high-Al, low-Si basalts appear to be restricted to and precision data for this system are reported in Sinton ridges with slow spreading rates or close to fracture et al. [20]. zones and ridge terminations. Along the GSC, high-Al Glass and mineral compositions were collected using samples are located close to ridge segment ends or the University of Hawaii Cameca SX-50 five-spectrom- associated with failing rifts (Fig. 1). eter electron microprobe. Major and minor element The extensive dataset available from G-PRIME makes analyses for GSC glasses are reported in Cushman et al. the Galápagos an ideal location to examine the melting [3]. Mineral analyses reported here are averages of three and fractionation processes that give rise to high-Al spots collected from individual crystals in selected high- MORB compositions. Glass data for the sample suite Al MORB. Plagioclase was analyzed for Si, Al, Fe, Mg, were reported in Cushman et al. [3], and whole-rock XRF Ca, Na and K using an accelerating voltage of 15 kV, data are reported for Galápagos N-MORB in Table 1. 20 nA beam current, and 10 μm beam diameter. Peak Estimates of GSC crustal thickness from multichannel counting times were 30 s for Si, Mg, Al, and Na and 60 s seismic reflection data [1,14] put a limit on melt for Fe, Ca, and K. Background counting times were 30 s productivity, giving us an important modeling constraint. for Fe, Ca, and K; 15 s for Si, Mg, and Al; and 20 s for To supplement the glass and whole rock analyses, we Na. Na was analyzed first in each acquisition to minimize collected mineral compositions, zoning profiles, and modes loss due to volatilization. Samples were calibrated in five high-Al GSC samples. Using this combination of against mineral standards Lake County plagioclase (Si petrographic and chemical data, we are able to model the and Al), San Carlos olivine (Fe and Mg), Amelia albite feasibility of various melting and fractionation paths (Na), anorthite (Ca) and orthoclase (K). A PAP–ZAF using the MELTS and pMELTS algorithms [8,15–17] by matrix correction was applied to all analyses. comparing the phase compositions and abundances re- Olivine was analyzed for Si, Mg, Fe, Ca, Mn and Ni sulting from each evolution path with the chemistry of using an accelerating voltage of 20 kV, 30 nA beam the rocks themselves. We first establish a working model current, and 10 μm beam diameter. Peak counting times that produces the array of normal GSC N-MORB under were 60 s for Mg and Si; 80 s for Ca and Ni; 50 s for Fe; reasonable ridge melting and crystallization conditions, and 40 s for Mn. Background counting times were 30 s then use the determined source composition to model the for Mg and Si; 40 s for Ca and Ni; 25 s for Fe; and 20 s for evolution of the high-Al samples. By varying the pres- Mn. Samples were calibrated against San Carlos olivine sure and extent of partial melting of the GSC source as (Mg), Springwater olivine (Fe), Verma garnet (Mn), well as the pressure and extent of subsequent fractional diopside (Si and Ca) and Ni–metal (Ni) standards.