Oxygen isotope ratios of olivine have become a widely used tool for the study of magmatic systems, especially in the interpretation of source heterogeneities in mantle plume–derived ocean island basalts. The underlying assumption is that fresh minerals provide a better guide to magma δ18O than bulk rock analyses and that olivine is also likely to be a major phenocryst phase in primitive magmas. However, distinctions between source compositions and the effects of subsequent magma evolution have not always been thoroughly scrutinized. For the Azores samples investigated here, we can demonstrate that the δ18O variation (+4.84‰ to +5.25‰ Vienna standard mean ocean water) observed in the olivine phenocryst population is closely linked to evolution in the host magmas during ascent to the surface. We observe a linear, positive correlation between forsterite (Fo) content and δ18O in all of the individual island lava suites. This forces us to conclude that the low oxygen isotope ratios result from combined assimilation and fractional crystallization processes, the assimilant being hydrothermally (temperature > 250 °C) altered, lower oceanic crust. Linear regression of the measured δ18O olivine values to Fo89 suggests a homogeneous mantle source with δ18O = +5.2‰ ± 0.1‰.


Oxygen isotopes have been widely used as a petrogenetic tool to decipher the source of igneous rocks (e.g., Magaritz et al., 1978; Harris et al., 2000; Garcia et al., 2007; Day et al., 2009; Gurenko et al., 2011). Early studies used the oxygen isotopic fractionation between coexisting minerals to develop thermometers for igneous systems (e.g., Anderson et al., 1971; Kyser et al., 1981), whereas more recent studies tend to focus on the isotopic composition of melts and their sources. Experimentally and thermodynamically derived fractionation factors between coexisting minerals, and between minerals and host magma, provide a means of estimating the bulk oxygen isotopic composition of melts and their derivatives (e.g., Eiler, 2001; Zhao and Zheng, 2002).

Over the past two decades, advances in analytical techniques from conventional to laser fluorination and more recently ion microprobe analysis have increased the amount of data from natural and synthetic samples (Eiler, 2001; Bindeman, 2008; Eiler et al., 2011). This has afforded precise determination of oxygen isotopic variability in basaltic rocks, which in turn may provide tighter constraints on the nature of their mantle source regions (e.g., Magaritz et al., 1978; Thirlwall et al., 2006). Equilibrium fractionation between coexisting phases at mantle temperatures is an underlying assumption. In the case of olivine, it is widely established that this mineral represents an early liquidus phase in the crystallization of basaltic melts and therefore may provide the best proxy for the oxygen isotopic composition of the primary mantle melt and its source region. The fact that forsteritic olivines are rarely zoned in δ18O is evidence in support of this (Eiler et al., 2011). However, as with any crystallizing system, there are complexities that need to be considered. For example, wall-rock assimilation processes can significantly change the composition of a rising magma, such that earlier formed minerals that crystallized deeper and when magmas were hottest and most able to promote fusion of their host rocks may be out of chemical equilibrium with their host magma.

Our objective here was to constrain the primary oxygen isotope composition of Azores lavas, including, for the first time, the islands west of the Mid-Atlantic Ridge. We show that, despite the fact that most olivines have forsterite and δ18O compositions very close to mantle values, reaction with hydrothermally altered crust was nearly ubiquitous. We conclude that the range in mantle oxygen isotope composition beneath the Azores region is significantly smaller than previously inferred and that in order to better constrain the oxygen isotope composition of the mantle, the effects of assimilation need to be appraised more closely.


Lavas from the Azores Islands

The samples investigated in this study are lavas selected from Corvo, Faial, Pico, São Jorge, and São Miguel Islands. These islands are thought to reflect the surface expressions of a mantle plume and so have been studied intensively for their chemical and isotopic composition (e.g., White et al., 1976; Turner et al., 1997; Beier et al., 2008). Incompatible trace element ratios and Sr-Nd-Pb isotope ratios of the Azores lavas are consistent with a mantle source that is heterogeneous on a small scale (Fig. 1), with significant trace element and isotope variability occurring even within a single island (Turner et al., 1997; Elliott et al., 2007; Beier et al., 2007). Oxygen isotope values, however, are relatively scarce for these islands and most studies so far focused on the islands located east of the Mid-Atlantic Ridge (Widom and Farquhar, 2003; Turner et al., 2007). This will permit future comparison with data from the Mid-Atlantic Ridge (Cooper et al., 2004).


Olivines from selected lavas (MgO > 5 wt%) that are known to have distinct and variable radiogenic isotopic compositions (Fig. 1) were hand-picked to avoid contamination, alteration, and, where visible, inclusions. Single grains were coarsely crushed in a steel mortar to obtain splits that were treated separately to obtain mineral chemical compositions via electron microprobe analysis and the oxygen isotope composition via laser fluorination from the same crystal. All analyses were performed at the GeoZentrum Nordbayern (University of Erlangen-Nürnberg). A description of the method is given in the GSA Data Repository1.


δ18O Values of Azores Olivines

The new oxygen isotope ratios from single olivines are presented in Table DR1 in the Data Repository. The islands in the center of the Azores archipelago, and therefore closest to the plume center (Terceira; Bourdon et al., 2005), have low δ18O values, extending to values lower than normally considered to be representative of pristine mantle olivine with δ18O = +5.2‰ ± 0.2‰ (Mattey et al., 1994). However, low δ18O olivines are also observed on the islands furthest from the plume center at the margin of the plateau (e.g., São Miguel Island in the east, and Corvo Island in the west). The overall range for the olivines is δ18O = +4.84‰–5.25‰ (Vienna standard mean ocean water); thus, the olivine population of the Azores lavas exceeds the range in δ18O of the pristine mantle.

Mineral Chemistry

The composition of the olivines investigated in this study is evident from plots of major and minor element oxides versus forsterite (Fo) contents (Fig. 2A), which vary from Fo77 to Fo90. Calcium and Ni display similar behavior during magmatic differentiation as their concentrations decrease continuously (CaO 0.41–0.17 wt%; NiO 0.31–0.08 wt%) with decreasing Fo. However, for a given Fo content, the NiO concentrations (see Table DR1) are slightly, but systematically, lower in olivines from Corvo relative to olivines from islands east of the Mid-Atlantic Ridge.


The oxygen isotope fractionation factor between olivine and basaltic melt at mantle temperatures is well constrained (Eiler, 2001). However, there is increasing evidence, from detailed studies of individual lava flows (Wang and Eiler, 2008) and of individual volcanoes (Harris et al., 2000), that oxygen isotopes can provide clues to assimilation processes and there is enhanced utility in studies that combine phenocryst compositions with oxygen isotope determinations from the same crystal. Assuming that the partition coefficient (KD) of Fe-Mg between olivine and melt is ∼0.3 (Roeder and Emslie, 1970), the oxygen isotope composition of primary basalts will be most closely approximated by the oxygen isotope composition of olivines with Fo ∼89 (host magma Mg# ∼71). For the Azores, magmatic differentiation histories are similar across the islands (cf. Beier et al., 2008; Genske et al., 2012) and dominated by olivine and clinopyroxene fractionation. These minerals typically have δ18O values lower than the melt, thus their removal will lead to an increase in the δ18O value of the differentiated melt (e.g., Muehlenbachs and Byerly, 1982); this is in stark contrast with our observations. The range of forsterite content in our data set is very large, but correlates positively with δ18O. This correlation (r2 ∼0.81; Fig. 2B) allows us to extrapolate to δ18O values for putative mantle olivines (Fo89–93; results are given in Table DR1). The relationship implies a major role for assimilation fractional crystallization (AFC) processes, as we discuss in detail further in the following, where we aim to constrain the primary oxygen isotope composition of Azores mantle source.

Primary Oxygen Isotopic Compositions of the Azores Mantle

The most striking observation of this study is the tight correlation between δ18O and Fo. This can be used to constrain the primary oxygen isotope signals of the individual islands; i.e., the δ18O of Fo(>89). It is intriguing that there is no resolvable difference in the δ18O-Fo arrays between the islands (Fig. 2B), even though each of these has a distinct mantle source in terms of Sr-Nd-(Pb-Hf) isotopes (cf. Fig. 1; Table DR1). That is, linear regression of the measured δ18O(ol) data indicates that the mantle sources beneath the Azores islands are extremely homogeneous, at least in terms of their oxygen isotopic composition. This implies that the variability observed in radiogenic isotopes must reflect a high-temperature origin mostly due to intramantle processes, with only a limited contribution from recycled low-temperature components (e.g., subducted sediments and basalt, both of which are typified by a large range in δ18O).

The calculated δ18O(Fo89) values for each island have an average of +5.2‰ ± 0.1‰ (2 standard deviations; Table DR1), in excellent agreement with estimates for pristine upper mantle olivine (Mattey et al., 1994). If their source were characterized by Fe depletion (e.g., Schaefer et al., 2002; Turner et al., 2007), and conceivably contained Fo ∼93 olivines, it would have a slightly higher projected oxygen isotope ratio [δ18O(ol) ∼+5.29‰]. From this, we conclude that the oxygen isotope ratios of Azores mantle olivine [i.e., Fo(>89)] are remarkably homogeneous. Whether our observation may apply more widely remains a matter to be fully tested by other studies. Nevertheless, published olivine data from off-ridge and ridge-centered ocean island basalts (OIB) reveal very similar δ18O compositions: Hawaii, Mauna Loa, δ18O(Fo90) = +5.24‰ ± 0.09‰; Canary Islands, La Palma, δ18O(Fo89) = +5.20‰ ± 0.12‰; Tristan da Cunha, Gough Island, δ18O(Fo > 85) = +5.19‰ ± 0.14‰; Azores islands, δ18O(Fo89) = +5.20‰ ± 0.10‰ (Wang and Eiler, 2008; Day et al., 2009; Harris et al., 2000; this study; Fig. 2B). Thus, there is growing evidence that the primary δ18O composition of plume sources may be relatively uniform.

In order to further evaluate the departure of the lower Fo Azores olivines to lower δ18O values, we evaluate AFC models similar to those developed for Hawaii (Wang and Eiler, 2008) and Gough Island (Harris et al., 2000).


Because our measured data set is limited to five islands, we applied the equation derived in Figure 2B to a larger sample set. This sample set includes primitive lavas (MgO = 5–12 wt%) from the western Azores (Flores and Corvo), the central Azores (Faial, Pico, Graciosa, Terceira, and São Jorge), and the eastern island of São Miguel (Turner et al., 1997; Beier et al., 2007, 2008; Genske et al., 2012). In detail, we first estimated Fo values in equilibrium with host lava compositions (Roeder and Emslie, 1970) and subsequently calculated their δ18O based on the equation of the correlation. To account for olivine-melt δ18O fractionation, we added 0.4‰ to the δ18O(Fo) value to estimate δ18O(melt) values (Eiler, 2001). We also applied olivine-liquid thermometers following the model of Putirka (2008), where we used equilibrium-liquid compositions from the larger Azores data set. The temperature results (given in Table DR1) are in good agreement with previously published temperature estimates for Azores lavas (Beier et al., 2012; Genske et al., 2012) and reinforce their potential to promote significant fusion and assimilation of the overlying oceanic crust (cf. DePaolo, 1981).

The oxygen isotopic composition of hydrothermally altered oceanic crust (AOC) is significantly different from that of pristine mantle. Analyses of AOC confirm that high-temperature, altered gabbro layers may extend down to δ18O values of +2.5‰ (e.g., Alt and Honnorez, 1984; Gao et al., 2006), although globally the average ratio is not significantly below +4‰ (Thirlwall et al., 2006). As illustrated in Figure 3B, we performed AFC modeling based on a broad correlation of Sr/Nd with δ18O and using the equations of DePaolo (1981). The MgO content of the lavas was used to monitor the degree of magmatic differentiation. Because the range in δ18O(melt) at a given trace element concentration (e.g., Sr and Nd) is relatively restricted and similar on each island, we develop a holistic model for δ18O behavior during magma differentiation that can apply to all of the islands. Changing the proportion of assimilation relative to fractionation (r) has large effects on the modeled δ18O values, but relatively minor effects on Sr/Nd or 87Sr/86Sr ratios. The proportion of the assimilant with δ18O ∼ +4‰ that best simulates the data in Figure 3B corresponds to r ∼ 0.25.

The range of 87Sr/86Sr in the lavas exceeds that of the model across a range of different proportions of assimilation (i.e., r = 0.1–0.5). Furthermore, the correlation coefficient for the olivine population (Fig. 2B) suggests that δ18O evolution reflects very similar assimilation processes across the islands and by implication that the assimilant beneath each island is similar. For example, if significant amounts of altered conduit material had been assimilated on the different islands, the slopes in δ18O versus forsterite would be more variable than observed (Fig. 2B). Thus, we would anticipate different trends for different localities (i.e., volcanoes) or stratigraphic units (i.e., post-shield versus shield stages), as observed in the Hawaiian volcanic chain (Wang and Eiler, 2008).

We suggest that the δ18O of the assimilant (hydrothermally altered gabbroic layers of the oceanic crust) is characterized by values <+4.5‰. This value has previously been suggested for AOC in OIB settings (e.g., Eiler, 2001; Wang and Eiler, 2008). Moreover, it would appear that assimilation of hydrothermally AOC is ubiquitous across the Azores archipelago. Whereas mid-oceanic ridge basalts generally show constant or higher δ18O with increasing magma evolution (e.g., Wanless et al., 2011), OIB magmas typically have lower oxygen isotope ratios that result from assimilation of ocean crust that underwent variable degrees of hydrothermal alteration. In the following we discuss the significance of our findings for the interpretation of mantle sources in OIB settings, with a focus on the origin of the Azores plume source.

Implications for OIB genesis

Deviations of the olivine oxygen isotopic ratio from the accepted mantle range of Mattey et al. (1994) led numerous authors to speculate whether this can be used to constrain the nature of source components within mantle plumes. The fact that oxygen isotopes are only efficiently fractionated in near surface environments at low temperatures has been taken as evidence that anomalously high or low olivine oxygen isotope ratios (i.e., <+5.0‰ or >+5.4‰) reflect subducted and recycled oceanic crust with or without sediment (e.g., Turner et al., 2007). However, assimilation of AOC may not always be evident from the major and trace element ratios in the host lavas. Conversely, the much greater variability of Sr isotope composition in the mantle (e.g., White et al., 1976; Hart et al., 1992) means that small volumes of assimilated AOC may not be detectable using 87Sr/86Sr isotopes alone. The homogeneous δ18O of +5.2‰ ± 0.1‰ in Azores olivines leads to the hypothesis that this plume may contain less evidence for recycled components than previously suggested (e.g., Beier et al., 2007; Turner et al., 2007).


Olivines from the Azores islands display a strong (r2 ∼0.81), nearly linear covariation between mineral major element chemistry and δ18O. A trend of decreasing δ18O with decreasing Fo leads us to conclude that AFC processes, involving the gabbroic layers of the underlying hydrothermally AOC as the dominant assimilant (δ18O ∼ +4.0‰), dictated the δ18O of the magmatic systems beneath all of these islands. The homogeneity of primary oxygen isotope signatures inferred for the Azores mantle is in stark contrast with their highly variable Sr-Nd-(Hf-Pb) isotope systems. Our results do not demand the presence of recycled material in the Azores mantle source; rather, they suggest that the time-integrated radiogenic isotopes variations were produced by intramantle processes.

This project was funded by projects BE4459/1-1 and BE4459/4-1 of the Deutsche Forschungsgemeinschaft. We acknowledge the help of Victor Hugo Forjaz and the Observatório Vulcanológico e Geotérmico dos Açores during various stays in the Azores. We thank M. Joachimski and D. Lutz for providing the mass spectrometers for oxygen measurements in Erlangen. Reviews by C. Harris and M. Thirlwall considerably improved the manuscript. This is contribution 241 from the ARC Centre of Excellence for Core to Crust Fluid Systems (http://www.ccfs.mq.edu.au) and 866 in the GEMOC Key Centre (http://www.gemoc.mq.edu.au).

1GSA Data Repository item 2013128, detailed method description and results from laser fluorination and electron microprobe analysis, including international certified standards measured during the course of this study, 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.