At mid-ocean ridges, a variety of crustal processes overprint mantle-derived melts and can obscure original mantle compositions. To address the nature of this crustal filter, we report 87Sr/86Sr ratios from plagioclase phenocrysts and host glasses in mid-oceanic-ridge basalts from the Juan de Fuca Ridge, Blanco Transform Zone, and the Southwest Indian Ridge. Microdrilled isotopic analyses reveal significant inter- and intracrystalline disequilibria within individual samples. These disequilibria suggest that a range of isotopically distinct melt components contribute to individual plagioclase crystals and to the magmas that transport them to the surface. Low Cl/K values both in the host glass and in plagioclase-hosted melt inclusions largely rule out incorporation of seawater-derived material as an explanation for differences in 87Sr/86Sr. Instead, the observed heterogeneity implies derivation of magmas from isotopically diverse mantle sources. Importantly, the range of Sr isotope values preserved in a single sample is similar to the range of compositions seen at the ridge segment scale. Unlike analyses of host glass compositions, which are the result of extensive crustal processing, isotopic analyses of phenocryst phases record fine-scale aggregation of these distinct mantle-derived melts and are thus an important and underutilized tool in interpreting the nature of the mid-oceanic-ridge basalts.


The role of heterogeneity in the upper mantle in controlling the chemical and isotopic compositions of mid-oceanic-ridge basalts (MORBs) is a key issue for constraining the nature of Earth’s depleted upper mantle reservoir. MORBs are noted for their high degree of homogeneity relative to ocean island basalts and arc volcanics (Hofmann, 1997), and although this has been argued to reflect melting of a broadly homogeneous mantle (Hofmann, 1997) or larger degrees of melting (Rubin and Sinton, 2007), there is also widespread consensus that upper mantle heterogeneity contributes to variations in MORB chemistry (Batiza, 1984; Holness and Richter, 1989; Rubin et al., 2009; Salters and Dick, 2002). What remain less clear are the scales at which such heterogeneities are manifest within upper mantle material, and the degree to which variable mantle sources contribute to individual MORB magmas. One important limitation for many existing studies is that they are based on the composition of quenched MORB glasses, which represent the liquid fraction present at the time of eruption. However, these liquids are also the product of a complex magma supply system, and thus mantle-derived variations are obscured by extensive crystal differentiation, melt-rock reaction, melt aggregation, and mixing (Rubin et al., 2009; Sinton and Detrick, 1992). Variable melt inclusion compositions from an individual MORB suggest that magmas aggregate from diverse melts, but these data are difficult to interpret in terms of mantle heterogeneity alone (Sobolev and Shimizu, 1993; Van Orman et al., 2002; Danyushevsky et al., 2002; Cottrell et al., 2002). Here we present microdrilled Sr isotope compositions of plagioclase crystals from MORBs from slow and intermediate spreading centers. These data provide a robust means to gauge contributions from heterogeneous mantle sources, and document a range of complex relationships between plagioclase phenocrysts and their host glass. Importantly, the ranges of compositions observed within crystals and between crystals and host glass are comparable to that evident at the scale of ridge segments. The Sr isotope variations we observe reflect contributions from highly heterogeneous mantle sources at temporal or spatial scales sufficiently fine to be sampled by individual crystals during crystal growth.

Mid-ocean ridges are the most productive magmatic environments on Earth, producing some 20 km3/yr of magma (Hofmann, 1997). Correlations between spreading rates, extent of magma differentiation, and chemical and isotopic variance (Rubin et al., 2009) highlight the importance of crustal processes in modifying primitive MORB compositions, and suggest that studies of erupted glass alone are insufficient in resolving the range of mantle contributions and the degree to which these contributions are overprinted within the crust. Studies of olivine- and plagioclase-hosted melt inclusions demonstrate that large compositional diversity exists in the melts that contribute to individual MORBs (Sobolev and Shimizu, 1993; Adams et al., 2011; Laubier et al., 2012). However, there is no consensus as to the causes of these variations. Suggested mechanisms include mixing of melts derived from a range of mantle source components (Laubier et al., 2012; Maclennan, 2008), near-fractional melting of a homogeneous mantle source (Sobolev and Shimizu, 1993), and melt-rock reaction or diffusional fractionation during melt transport (Van Orman et al., 2002; Danyushevsky et al., 2002; Spiegelman and Kelemen, 2003). Further, questions remain about the effects of inclusion formation on melt composition (Cottrell et al., 2002). Although radiogenic isotope compositions have been used to place important constraints on melt inclusions (Maclennan, 2008; Saal et al., 2005), technical limitations have largely prevented precise measurements of radiogenic isotope signatures in trace element–depleted MORB melt inclusions. However, a largely unexploited source of information exists in the early-formed phenocryst phases themselves. The resistance of crystalline solids to mechanical mixing and re-equilibration allow phenocrysts to potentially preserve a record of the geochemical conditions at early phases of magmatic differentiation. Plagioclase phenocrysts are particularly useful because of the prevalence of feldspar in MORBs (Bryan, 1983), relatively slow diffusion rates of many elements (Grove et al., 1984; Costa et al., 2010), and the relatively high content of a number of useful trace elements (e.g., Sr, Ti, Ba, light rare earth elements [LREEs]) not present in olivine (Adams et al., 2011). The utility of isotopic profiles within plagioclase has been demonstrated dramatically by studies of arc rocks (Davidson and Tepley, 1997), revealing the complexity of processes in the generation of arc magmas. Measuring the isotopic composition of phenocryst phases thus provides an independent and robust measure of the nature of the parental melts that contribute to an individual magma.


Plagioclase and groundmass samples were drilled using a Merchantek MicroMill at Oregon State University (United States). Core and rim samples were drilled from multiple phenocrysts in one-inch epoxy mounts. Additional samples were drilled from the groundmass directly adjacent to the phenocryst to test for variation between the phenocryst and the host liquid, using the method detailed in Charlier et al. (2006). Groundmass was mostly glassy; however, multiple samples of groundmass were drilled and measured to ensure homogeneity. Sample weight was estimated at 20–80 μg yielding 10–20 ng of Sr. Sr was separated during column chemistry using cation exchange resin with blanks averaging 30 pg. Further details of methodology are found in Ramos and Tepley (2008) and references therein. The high-precision 87Sr/86Sr measurements of unspiked aliquots were conducted by thermal ionization mass spectrometry at New Mexico State University using a dynamic peak jumping routine. All ratios were normalized to 86Sr/88Sr = 0.1194 to account for mass fractionation. 84Sr/86Sr was also monitored during analyses, and these values and the NBS 987 standard values are reported in Table DR1 in the GSA Data Repository1. Reproducibility was estimated from multiple runs of the E32 glass and long-term machine reproducibility of NBS 987 at 0.00002. Major element analyses of glass and melt inclusions were performed on the Cameca SX-100 at Oregon State University. Only glassy melt inclusions were chosen for analysis.


We have examined Sr isotope systematics of plagioclase-phyric MORB samples from three spreading ridges in order to evaluate the range of Sr isotope compositions preserved in phenocrysts and to compare these to the composition of the erupted liquid fraction, as represented by quenched glass or groundmass. Micromilling techniques allow sampling of multiple zones within phenocrysts, which are then purified by low-blank column chemistry before analysis by thermal ionization mass spectrometry (Davidson and Tepley, 1997; Charlier et al., 2006). Low concentrations of Sr in these plagioclase (100–280 ppm) necessitated sampling of relatively large crystals (>500 μm). This study utilizes on-axis MORBs with abundant large plagioclase phenocrysts, which appear to be restricted to ridges with ultraslow to intermediate spreading rates (Flower, 1980). Highly plagioclase-phyric lavas (>15 mod%) have been considered to represent anomalous MORBs (Bryan, 1983; Allan et al., 1989); however, a comparison of host glass compositions between highly plagioclase-phyric MORBs and more typical sparsely phyric to aphyric samples from the same ridge segments shows that highly plagioclase-phyric samples have glass compositions that overlap those of aphyric MORBs, and do not appear anomalous in any respect other than modal plagioclase abundance (Fig. 1). As a result, we conclude that these lavas are a valid proxy for studying MORB petrogenesis in general. In addition, the similarity in compositional range between phyric and aphyric lavas suggests that the high modal proportions of plagioclase are likely the result of physical processes and not controlled by the chemical properties of the host lava.

Our primary observation (Fig. 2) is that most plagioclase phenocrysts studied exhibit a range of Sr isotope compositions, and in most cases the range of Sr isotopes observed in a small number of plagioclase crystals rivals the range observed in all lavas from that particular ridge segment (Fig. 3). Sample D22 from Juan de Fuca Ridge contains host glass with an average 87Sr/86Sr ratio of 0.702454 (n = 3) and plagioclase ranging from 0.702349 ± 0.000011 to 0.702479 ± 0.000013. Sample 15-1-2 from the Southwest Indian Ridge (SWIR) contains host glass with an average 87Sr/86Sr of 0.704307 (n = 2) and plagioclase ranging from 0.703380 ± 0.000013 to 0.703808 ± 0.00001. Sample A91 1R2 from the Blanco Transform Zone contains host glass with an average 87Sr/86Sr of 0.702355 (n = 3) and plagioclase ranging from 0.702408 ± 0.000011 to 0.702481 ± 0.000014. At the Blanco Transform Zone, the more radiogenic values we observe in plagioclase are consistent with plagioclase-hosted melt inclusion compositions, which suggest that the plagioclase crystallized from a more enriched liquid than the host glass (Adams et al., 2011). In contrast, SWIR 15-1-2 includes a highly radiogenic glass (0.7043) with significantly less radiogenic phenocrysts (∼0.7036). The high 87Sr/86Sr in the glass is consistent with contribution from a continental-like end member that has been identified within the mantle source beneath the SWIR (Meyzen et al., 2005). Another sample studied from the same ridge segment (17-1-3; see the Data Repository) shows less radiogenic glass (average 87Sr/86Sr of 0.703006, n = 3) with a higher degree of homogeneity between the glass and phenocrysts, suggesting that the magma storage conditions are highly variable even within a single segment.

In addition to disequilibria between crystals and the host glass, there are also resolvable internal variations within some plagioclase phenocrysts (Fig. 2C). Intracrystalline Sr isotopic disequilibria, where present, provide evidence that a single phenocryst has interacted with multiple generations of melts of varying Sr isotope composition during growth. Due to low concentration of Sr and the relatively large amount of plagioclase needed for high-precision analysis, our sampling was coarse and only allowed for Sr isotopic analyses of crystal cores and rims, meaning that the observed variation should be considered a minimum. Even with this limited spatial resolution, our data demonstrate that some plagioclase phenocrysts in MORBs preserve significant isotopic variations, and that the Sr isotope compositions of these crystals often differ from the melts that ultimately transport the plagioclase to the site of eruption. In order to preserve these isotopic disequilibria, crystals cannot be kept at magmatic temperatures for an extended period, as this would result in diffusive re-equilibration (Davidson et al., 2001). This time was estimated in crystals preserving disequilibria using an anorthite-corrected diffusion coefficient for Sr (Zellmer et al., 1999) and the distance between drill locations. In order to erase the disequilibria across 250 μm, the shortest length scale we measured, the crystal would have to be held at 1200 °C for over 450 years, suggesting that these crystals were not held at magmatic temperatures for significant periods. The intracrystalline disequilibria were likely formed during episodic crystal growth through interaction with isotopically distinct melts.


One potential way to generate Sr isotope diversity is through assimilation of seawater-altered crustal rocks, and this is a noted process affecting MORBs, although it is typically more common in fast spreading environments (Michael and Cornell, 1998). As seawater has higher 87Sr/86Sr in comparison to most MORBs (∼0.709 and ∼0.703, respectively; Bach et al., 2003), assimilation of altered basalt or seawater-derived brines will produce more radiogenic Sr isotope magma compositions. We can assess the role of seawater assimilation through Cl and Cl/K systematics. Because seawater is highly elevated in Cl and Cl/K with respect to mantle-derived basalts, assimilation of seawater, seawater-derived brines, or seawater-altered ocean crust is evident in Cl/K values above the interpreted mantle range of 0.05–0.08 (Michael and Cornell, 1998). The range of Cl/K observed in groundmass glass and plagioclase-hosted melt inclusions from our samples extends from values within the typical mantle range to values that are slightly higher, up to ∼0.17, which is consistent with addition of relatively small amounts of seawater-derived Cl (up to 0.2%; Fig. 4). Recent studies have expanded the range of mantle Cl estimates to include all of the values we measured (Kendrick et al., 2012). Mixing calculations (Fig. 4) show that addition of the small amount of assimilant required to elevate Cl and Cl/K would have almost no effect on 87Sr/86Sr. The potential variation of Sr isotope composition due to this assimilation is much less than the differences we observe within plagioclase crystals and between crystals and host glass. Thus addition of seawater-derived components is unlikely to explain the observed variation in 87Sr/86Sr in our samples. Given that the isotopic compositions of MORB sources are known to be variable over a range of length scales (Batiza, 1984; Holness and Richter, 1989), including variation at the subsegment scale (Bergmanis et al., 2007), we conclude that the most likely source of the Sr isotope variations we see are contributions from different mantle sources.

Our data have important ramifications for the nature of the mantle sources that contribute to MORBs, and show that the compositional heterogeneity in individual MORBs evident in melt inclusion populations (Sobolev and Shimizu, 1993; Adams et al., 2011) is also matched by variations in isotopic compositions. However, as the chemical compositions of melt inclusions may be modified during melting (Sobolev and Shimizu, 1993), melt transport (Van Orman et al., 2002), crustal processing, and inclusion formation (Cottrell et al., 2002), they are difficult to interpret solely in terms of mantle source heterogeneity. In contrast, the Sr isotope composition remains largely unmodified by melt transport processes, and thus more closely reflects the Sr isotope compositions derived from mantle sources. The variations in 87Sr/86Sr we observe within phenocrysts, and between phenocrysts and groundmass, demonstrate the variability of mantle sources that contribute to individual mid-ocean ridge systems, and show that these source contributions vary at spatial or temporal scales fine enough to be sampled by individual crystals. This suggests that the apparent homogeneity of MORBs can result from extensive aggregation of isotopically diverse melts during crustal magma processing rather than reflecting homogeneity of melt sources (Lundstrom et al., 2000; Rubin et al., 2009). The apparent homogeneity of MORBs is further accentuated by the emphasis on quenched glass compositions in MORB data sets. Our data show that using only glass analyses dramatically limits our abilities to characterize the nature of primitive MORB liquids and the crustal processing that overprints them. Trace element and isotopic data from MORB phenocrysts are a previously underexploited source of information on crustal construction processes. Use of such phenocryst information, together with glass data, represents a fundamental change in how MORB lavas are analyzed and in the potential sophistication of our models for heat and mass transport in the oceanic crust.

This work was funded by National Science Foundation grant OCE-0927773. Thanks to F.C. Ramos and his students S. Scott and C. Dimond for help with TIMS analysis at New Mexico State University. Thanks to D. Burns for assistance with EMP analyses. Thanks to Ken Rubin and two anonymous reviewers for their helpful comments, which greatly improved this manuscript. Thanks to P.M. Gregg, D.W. Graham, and A.W. Burleigh for discussions and reading of the manuscript. Thanks to all who contributed plagioclase-phyric samples for this project, without which our work would have been impossible.

1GSA Data Repository item 2013069, methods, Tables DR1–DR3, and Figure DR1, is available online at www.geosociety.org/pubs/ft2013.htm, or on request from editing@geosociety.org or Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301, USA.