because of the limited geographic distribution of the melt inclusion and submarine glass dataset; because there are no samples in common between the two distributions; and because the melt inclusions may reflect magma compositions that precede magnetite and ilmenite saturation. Thus, to first order, our global result is not inconsistent with the results of Waters and Lange (2016) and Crabtree and Lange (2011) who found congruence when they compared magnetite‐ilmenite oxybarometry to wet‐chemical titration on the same suite of very fresh aphyric lavas.
A more direct comparison can be made between the olivine‐hosted melt inclusions and submarine glasses erupted along the active Mariana Arc, and back arc basin (BAB) glasses erupted at depth along the associated back arc spreading center: the Mariana Trough. Both datasets apply the same method (XANES) to arrive at fO2 estimates, and both sample suites comprise basaltic to basaltic andesite glasses representing similar stages of differentiation in thin crust (similar MgO). The Mariana arc front samples record fO2s that are on average 0.73 log units higher than the Mariana trough samples (Mariana arc: QFM+0.95±0.36 for n = 107 vs Mariana trough: QFM+0.22±0.30 for n=37, respectively) (Brounce et al., 2014; Kelley & Cottrell, 2009).
Another direct comparison can be made between submarine basalts in ridge (MORB) and BAB tectonic settings. Here we find that BAB from the Mariana trough (n=37) record significantly higher fO2s than MORB globally (n=160) by 0.4 log units (tstatistic = 7.8, tcritical = 2.0, degrees of freedom [df] = 41, p‐value << 0.001). This comparison is particularly germane for inferring the effect of subduction on mantle fO2 because submarine back‐arc ridges and mid‐ocean ridges are tectonically similar and differences in their melt chemistry can be largely attributed to the influence of subduction (Stolper & Newman, 1994). As indices of subduction influence in Mariana trough lavas go from negligible to significant (e.g., as H2O contents and the ratios of fluid mobile to fluid immobile incompatible trace elements increase), Fe3+/∑Fe ratios (and fO2s) also increase (Brounce et al., 2014; Kelley & Cottrell, 2009, 2012). In the Marianas, volcanics erupted over the course of the arc’s maturation also record increasing fO2 with increasing subduction influence. Modern arc tholeiites record similar fO2s to the boninites that erupted during the early stages of slab influence on the mantle wedge; and both lithologies are more oxidized than the forearc basalts that tapped the mantle prior to slab influence (Brounce et al., 2021; Brounce et al., 2015). The volatile and trace element signals of subduction appear intimately tied to elevated fO2s in space and in time.
Mantle lithologies recovered from arc settings comprise primarily forearc and arc peridotites. Forearc peridotites are exposed on trench walls and may sample ancient lithospheric mantle (Parkinson & Pearce, 1998), mantle wedge metamorphosed by the subducting slab (Fryer et al., 1985), or processes associated with subduction initiation (Birner et al., 2017). In comparison, arc peridotites rapidly ascend to the surface as xenoliths encased within their basaltic hosts at arc front volcanoes. The mean fO2 recorded by forearc peridotites is statistically indistinguishable from the mean fO2 recorded by ridge peridotites (tstatistic = 0.73, tcritical = 2, df = 131, p‐value = 0.47) (Fig. 3.3). As discussed by Birner et al. (2017), this result contrasts with Parkinson and Pearce (1998)’s study of forearc peridotites from the Izu‐Bonin subduction zone, primarily because we apply the spinel activity model of Sack and Ghiorso (1991) instead of Nell and Wood (1991). Yet, consistent with Parkinson and Pearce (1998), Birner et al. (2017) show that peridotites that have interacted with slab‐influenced melts do yield elevated fO2. This influence is additionally evident in the distribution of fO2 recorded by arc xenoliths from five studies (Table 3.1), which lies significantly higher, by 0.65 log units, than ridge peridotites (tstat = 4.4, tcrit = 2.0, df = 90, p‐value << 0.001) or forearc peridotites. Another unique characteristic of sub‐arc peridotites is the extended range of melt extraction they record. Spinel Cr#, commonly taken as a proxy for melt extraction, extends to much higher values (> 60) in sub‐arc peridotites than in ridge peridotites. This extended range of melt extraction may provide an opportunity to investigate the relationship between extent of melting and fO2. For example, Benard et al. (2018b) found a weak positive correlation (p‐value > 0.06) between fO2 and modal orthopyroxene, which they interpreted as evidence of fO2 falling with melt extraction; however, the positive correlation between spinel Cr# and fO2 in these same samples suggests the relationship between fO2 and melt extraction may be more complicated. No correlation exists between fO2 and orthopyroxene mode or spinel Cr# in the Tonga peridotites of Birner et al. (2017). More work is needed to better constrain the effects on fO2 of extracting melt from the mantle.
Plumes.
Mantle plumes are thermal upwellings that impinge on the lithosphere (French & Romanowicz, 2015; Montelli et al., 2006; Sleep, 1992). Capable of generating low degree mantle melts that more ably sample mantle heterogeneity, mantle plumes can produce ocean island basalts (OIB) (Dasgupta et al.; McKenzie & Onions, 1983; Stracke et al., 2005) and xenoliths wrenched from the lithospheric mantle (Frey & Roden, 1987). Both melts and xenoliths at ocean islands offer opportunities for oxybarometry; however, it remains challenging to interpret the relationship between lithospheric mantle peridotites and OIB. We begin with the volcanics.
Melt inclusions and submarine basalts erupting at and around the mantle plumes of Hawaii, Erebus, Iceland, and the Canary Islands record, on average, QFM +0.1 (7 XANES spectroscopic studies with n= 334 samples; Table 3.1, Fig. 3.2e). The Fe‐ XANES‐based fO2 in this case is higher than the record of magnetite–ilmenite pairs (n=47) by 0.35 log units (tstatistic = 3.8, tcritical = 2, df = 69, p‐value = 0.003); however, we emphasize that these datasets have no samples in common. There is no meaningful difference between the fO2s recorded by OIBs and MORBs. In the case of OIB volcanics, most XANES studies have either interrogated a geographic gradient in fO2 (e.g., Shorttle et al., 2015) or the effect of differentiation (crystallization and degassing) on fO2 (e.g. Brounce et al., 2017; Helz et al., 2017; Moussallam et al., 2016; Moussallam et al., 2014). We therefore observe a bimodal distribution of fO2 recorded by plume‐affected glasses (Fig. 3.2e). Glasses erupted along mid‐ocean ridges that approach plumes, and glasses affected by degassing, record lower fO2, while primitive and relatively undegassed melt inclusions record higher fO2 (Fig. 3.2e, and see discussion). The authors of the detailed studies that have interrogated the fO2 of plumes and plume‐affected ridges have inferred plume mantle source fO2s anywhere from 0.4 to 2 log units higher than the average fO2 recorded by the volcanics (filled circles in Figure 3.2e). In all cases, the authors have suggested that the mantle sources of these OIBs are more oxidized than those of MORB, and that this may be due to incorporation of recycled components (e.g. Brounce et al., 2017; Moussallam et al., 2019; Helz et al., 2017; Moussallam et al., 2016; Moussallam et al., 2014; Shorttle et al., 2015). Investigators have drawn such inferences by projecting the glass Fe3+/∑Fe ratios along compositional trends (e.g., to more primitive, less degassed, and more enriched compositions) or geographic trends (e.g., along a ridge toward a plume). We discuss some of these projections in greater detail in Section 3.4.1.1.
We have additionally compiled data from 13 studies to calculate the fO2 of 143 ocean island xenoliths. Unlike all of the ridge peridotites and
