Oxidized Mantle Sources of HIMU and EM-type Ocean Island Basalts

Oxygen fugacity (fO2) is a fundamental variable in igneous petrology with utility as a potential tracer of recycled surficial materials in the sources of mantle-derived lavas. It has been postulated that ocean island basalts (OIB) have elevated fO2 relative to mid-ocean ridge basalts (MORB) owing to more oxidized source regions. To clarify this issue, trace-element systematics of olivine grains are reported from OIB lavas with HIMU (high-; Mangaia, Canary Islands), enriched mantle (EM; Samoa; São Miguel, Azores Islands) and depleted MORB mantle (DMM; Pico, Azores) Sr-Nd-Pb-Os isotopic signatures, to constrain the fO2 of each magmatic system. Despite sampling distinct mantle reservoirs based on radiogenic isotope systematics, these OIB suites show similar fO2, ranging from +1.5 to +2.9 FMQ, with an average of 2.0 ± 0.7 FMQ, significantly higher than MORB at +0.6 ± 0.2 FMQ using the same oxybarometer. OIBs show no correlation between fO2 and bulk rock isotopic ratios or parental magma compositions. The lack of correlations with isotopic signatures likely results from radiogenic isotope signatures being hosted in volumetrically minor trace element enriched mantle lithologies, while fO2 reflects the volumetrically dominant mantle component. Higher fO2 in OIB relative to MORB implies a uniformly oxidizing plume source mantle that may be the result of either a common oxidized oceanic crust-rich reservoir parental to all modern plume lavas, or preservation of un-degassed and oxidized mantle domains formed early in Earth history.


20
Oxygen fugacity (fO2) is an intensive variable in igneous petrology that controls the 21 geochemical behavior of redox-sensitive elements such as Fe, V, Cr, S, C and H. It is defined as 22 the chemical potential of molecular oxygen (O2) in equilibrium with an igneous system and, like 23 all equilibria, oxygen fugacity depends on temperature and pressure. It is therefore normally 24 discussed in igneous petrology relative to mineral redox buffers, with the most common being 25 the fayalite-magnetite-quartz, or FMQ buffer (Lindsley, 1991). Oxygen fugacity varies significantly 26 in natural Earth systems by ~nine orders of magnitude, from the reduced metallic core to an 27 atmosphere that contains ~20% molecular O2. Igneous systems also show large variations in fO2, 28 with arc basalts and alkaline continental basalts showing systematically higher fO2 relative to 29 plume and ridge basalts (Carmichael, 1991;Brounce et al. 2014). Oxygen fugacity in arc basalts is 30 elevated by ~+1 to +5 log units FMQ above ambient mantle, although it is currently debated 31 whether the high value of arc basalts results from subduction-related metasomatism of their 32 mantle source, or from differentiation and degassing processes (Lee et al. 2005;Kelley and 33 Cottrell, 2009;Brounce et al. 2014;Tang et al. 2018).

34
Oxygen fugacity is traditionally constrained in volcanic rocks using the Fe +3 /Fe ratio of 35 volcanic glasses coupled with the experimental formulation of Kress and Carmichael (1991). The 36 Fe +3 /Fe ratio of volcanic glasses is determined in several ways, including wet chemistry and 37 Mössbauer spectroscopy. X-ray Absorbance Near-Edge Spectroscopy (XANES) has recently 38 allowed for high spatial resolution coupled with relatively fast sample throughput (Cottrell et al. 39 2011;2013;Moussallam et al. 2014;2019;Brounce et al. 2014). However, XANES 40 qualitative analyses of hydrous glasses and melt inclusions can be compromised by beam damage 41 . Additionally, it has been shown that Fe +3 /Fe ratio of melts can change on 42 the order of minutes by interaction with atmospheric oxygen ). Due to these 43 challenges and issues, as well as the low preservation potential of pristine volcanic glasses, alternative methods of determining magmatic fO2, such as V/Sc and Zn/Fe ratios in bulk rocks 45 (Lee et al. 2005;2010), as well as the partitioning of V into olivine (Canil, 1997;Mallmann and 46 O'Neill, 2009;2013;Nicklas et al. 2018;2019;2021), have been developed. The oxidation state 47 of vanadium in magmas varies from V +3 to V +5 ; the former being much more compatible in olivine, 48 regardless of temperature, pressure and melt composition (Canil, 1997;Wang et al. 2019). The 49 most important advantage of Vanadium-in-olivine oxybarometry over XANES Fe oxybarometry, 50 is the possibility to obtain fO2 values of the melt at the first crystallization of primitive olivine, in 51 many cases prior to any magmatic degassing. In contrast, XANES measures the fO2 as glass 52 quenches, after the melt might have been modified by degassing and or assimilation processes.

53
Additionally, V-in-olivine oxybarometry values are relatively hard to reset, as V diffusion in olivine 54 is fairly slow with a diffusion coefficient on the order of 10 -14 m 2 /s (Chakraborty, 2010). This 55 method can give erroneous results however, if the measured olivines are xenocrysts and did not 56 crystallize from the rock in which they are found. Vanadium-in-olivine oxybarometry can be 57 readily applied to primitive olivine-phyric lavas from a variety of settings.

58
In contrast to arc basalts, the fO2 of ocean island basalts (OIB) has only recently received 59 significant attention. It has been postulated, based on XANES measurements in glassy melt 60 inclusions, that Hawaiian basalts are oxidized relative to mid-ocean ridge basalts (MORB), 61 although degassing of sulfur has led to substantial modification of their observed fO2 62 (Moussallam et al. 2016;Brounce et al. 2017). A similar argument for degassing of sulfide and 63 oxidation has also been demonstrated for intraplate alkaline lavas from Mt Erebus in Antarctica 64 (Moussallam et al. 2014). The high values of fO2 (relative to MORB) measured in spinel grains 65 hosted in residual mantle xenoliths from Cape Verde also suggest an oxidized mantle source for 66 Cape Verde magmas (Ryabchikov et al. 1995). Basaltic glasses from the Reykjanes Ridge adjacent 67 to Iceland show a positive correlation between oxygen fugacity and proxies for geochemical 68 enrichment, suggesting that significant amounts of oxidized surficial material are present within 69 the Iceland plume (Shorttle et al. 2015;Novella et al. 2020). The idea that OIB source mantle is 70 uniformly oxidized was extended by Moussallam et al. (2019) (Woodhead, 1996), 89 while Samoan lavas represent the EM2 (Enriched Mantle) isotopic endmember (Jackson et al. 90 2007a). The Canary Island lavas are 'HIMU-type' (Day et al., 2010), and samples from the same 91 Canary Island volcanos have also been measured by Moussallam et al. (2019)   values. Information on the modeling of clinopyroxene accumulation can be found in Table 3.

133
Calculated fO2 of the OIB samples are listed in Table 4 with the relatively high S +6 /S (0.17 ± 0.11) of Samoan lavas (Labidi et al. 2015). The lack of 165 correlation of fO2 with isotopic signatures in the OIB dataset contrasts to prior observations made 166 using MORB datasets. For example, the XANES study of Cottrell and Kelley (2013) showed that 167 isotopically enriched MORB have resolvable lower fO2, which they attributed to reduced carbon 168 amount in their source regions. This correlation could also possibly instead reflect isotopically 169 enriched MORB being generally more volatile rich, and thus having degassed more S. Our new 170 average OIB fO2 value overlaps with the range of global arc basalts (i.e., Carmichael, 1991;Kelley 171 and Cottrell, 2009) and also coincides with that of high MgO Siberian meimichites (~+2.5 FMQ) 172 estimated using V-in-olivine oxybarometry (Mungall et al. 2005).
Although the data of Moussallam et al. (2019) showed that OIB are indeed oxidized, melt 174 inclusions XANES is perhaps not the best way of constraining OIB source region fO2 for two main 175 reasons. Firstly, degassing has affected the measured fO2 of the samples, and thus only a 176 minimum fO2 can be calculated. It has been argued that some modern Canary lavas are the most 177 S-rich contemporary lavas on Earth and are thus especially susceptible to substantial modification 178 by degassing (Taracsak et al. 2019). Secondly, several of the samples from that study showed high 179 volatile concentrations (up to ~3% H2O) which can cause significant analytical problems during 180 XANES analysis ) and those authors also did not seek to correlate fO2 with 181 isotopic evidence for recycled material in their sample set. The concordance of our average OIB 182 fO2 with the plume fO2 inferred by Moussallam et al. (2019) is remarkable, and more V-in-olivine 183 oxybarometry and XANES measurements of the same samples will serve to clarify the utility of 184 these two oxybarometry methods for OIB studies. terrigenous sediments) even ~30% continental crustal assimilation is unlikely to have significantly 211 affected the fO2 of a mantle-derived melt (Grocke et al 2005). The studied OIB suites are all 212 situated on oceanic crust which is relatively Fe-rich, but not grossly more so than the OIB parental 213 lavas themselves, meaning that significant (>20%) quantities of assimilation would be necessary 214 to modify their fO2. Large amounts of assimilation of oxidized altered oceanic crust from the Jackson et al., 2007;Day et al., 2010;Waters et al., 2020), leading to correlations with fO2, which 217 are not observed (Fig. 3). In particular, while the available data are limited, no correlation is seen 218 between O isotopes (Day et al., 2009;2010) and fO2, further demonstrating that assimilation was 219 a minor process. Finally, although crustal assimilation has been documented in selected Azores 220 lavas using B and Li isotopes (Genske et al. 2014 it is also disputed whether the isotopically enriched-endmember in the Canary plume derives 252 from pyroxenite or peridotite (Day et al. 2009(Day et al. , 2012Gurenko et al. 2009). Pyroxenite is 253 conventionally thought to be the result of hybridization of recycled AOC and peridotite (Sobolev 254 et al. 2005), and therefore would likely be more Fe +3 -rich (i.e., high fO2) than ambient peridotite.

255
Either pyroxenite in the sources of the studied OIB is not more oxidized than peridotite, or 256 pyroxenite-derived melt represents a minor portion of the parental melts contributing to the 257 studied lavas so as to be undetectable by the method use here. In either scenario, more data are 258 necessary on pyroxenite and peridotite-derived OIB to clarify the issue. It is worth noting that previous studies have concluded that pyroxenite is lacking in the source of Azores lavas (Sobolev 260 et al. 2007).

271
Whether the AOC is present as a separate, oxidized pyroxenite lithology or the isotopic signatures 272 of AOC have been imparted onto normal peridotite without a separate lithology being present is 273 unsettled (Herzberg et al. 2014).

274
For the purpose of our modeling, we assume that AOC isotopic signatures in the Mangaia 275 source are accompanied by oxidized AOC Fe. AOC has a relatively high Fe +3 /Fe of 0.22 ±0.08 but 276 can locally reach values as high as 0.36 (Evans, 2012). Even Archean AOC is likely highly oxidized, 277 as oxidation of AOC is largely the result of serpentinization reactions in the presence of water 278 and has little to do with atmospheric O2 content (i.e., Kasting, 2014). Assuming that ambient 279 MORB mantle has 8.05 wt.% total FeO (McDonough and Sun, 1995) with a Fe +3 /Fe = 0.05 (Cottrell and Kelley, 2011) and recycled AOC has 10.43 wt.% total FeO (Gale et al. 2013) with a Fe +3 /Fe of 0.06 to 0.07. Assuming temperature is constant and mantle spinel does not change 283 in composition by addition of AOC except to increase in Fe +3 content, addition of 5% AOC is 284 calculated using the method of Ballhaus et al. (1991) to raise the fO2 of the HIMU source by 0.34-285 0.58 log units FMQ. These differences are possibly resolvable using our method, which has 286 uncertainties varying from 0.13 to 0.59 log units FMQ. (e.g., Hart et al. 1992;Farley et al. 1992;Stracke et al. 2005), which has been termed either   (Day & Hilton, 2011), and Mangaia 332 lavas have lower 3 He/ 4 He, in some cases even lower than MORB lavas (Parai et al., 2009). Azores 333 lavas are largely MORB-like with regard to He isotopes, ranging from 7.2 to 11.1 Ra (Moriera et 334 al. 1999;2012;Madureira et al. 2014). It is notable that regardless of the wide variation in He 335 isotopic signatures in the studied OIB, there seems to be no variation in their fO2.

336
If the "common" component sampled by OIB is more oxidized relative to the MORB 337 source mantle, then a model explaining its high fO2 is necessary. If the common component is 338 simply a constant amount of relatively young, recycled AOC, as suggested by Stracke et al. (2005), 339 this could be a potential source of the MORB-OIB fO2 dichotomy, but this model does not explain the high-3 He/ 4 He signature seen in Samoan OIB, as recycled crust is predicted to have low-primordial reservoir (Class and Goldstein, 2005), its high fO2 may reflect a more oxidized  (Stracke et al., 2005) or an early-formed geochemically depleted source 361 (Hart et al., 1992;Mundl-Petermeier et al., 2020), more OIB fO2 data from more plumes, including 362 those with extreme isotopic signatures, are necessary to clarify the issue. Of particular interest is 363 the measurement of fO2 in samples with 182 W/ 184 W anomalies, which are sensitive indicators of the presence of early-formed components in mantle plume sources. The current dataset only 365 includes one such sample, OFU-4-14, that shows a strongly negative  182 W anomaly of -17.3 ± 366 4.5 (Mundl-Petermeier et al. 2020) and 3 He/ 4 He ratio of 25 Ra (Jackson et al. 2007) but close to 367 average OIB fO2 of +1.78 +0.25 -0.22 FMQ.

368
Our new fO2 data support the paradigm that recycled surficial materials such as AOC and 369 sediments are present in the sources of OIB (Hofmann, 1997), possibly leading to their high fO2 370 value. There is a strong theoretical framework to support this idea, as mass-balance calculations 371 in subduction zones (Evans, 2012;Brounce et al. 2019) show that the majority of contemporary 372 subducted oxidants (Fe +3 , S +8 , C +4 ) are not emitted by arc volcanoes, indicating that they are 373 brought into the deep mantle. Indeed, the global study of Evans (2012) suggested that only ~10% 374 of the total subducted oxidants, and almost none of the subducted Fe +3 , is emitted by global arc 375 volcanoes. There is substantial isotopic variation found in OIB (Zindler and Hart, 1984;Hofmann, 376 1997) likely because of varying proportions of recycled AOC, as well as terrigenous and pelagic 377 sediments. These different lithologies are likely to show substantially different oxygen fugacity, 378 with AOC and terrigenous sediment being oxidized relative to the mantle, and pelagic organic-379 rich sediments being reduced. This study shows that the quantity and type of recycled material 380 does not appear to impact the fO2 of the lava. Prior to this study, no work has attempted to 381 correlate isotopic signatures of OIB with oxygen fugacity. More data will show whether the OIB 382 source mantle is indeed uniformly more oxidized than MORB and whether any variation is shown 383 with isotopic signatures.     Table 4. MORB isotopic data are from Workman and Hart (2005) and MORB fO2 data are from Nicklas et al. (2018). Symbols as in Fig. 1.  Table 1: Average concentrations of elements (in ppm) in the cores of primitive olivine crystals from each OIB sample. Data were reduced assuming uniform Si concentrations, which are taken from the literature, and vary the least of all major elements in stochiometric olivine. Nnumber of olivine grains analyzed for each sample, SiO2 -assumed silica content of the olivine in wt.%, 2s -two standard deviations of the average concentration. For full dataset including concentrations in each spot analyzed, see online supplemental datasets.   Table 3: Average clinopyroxene V concentrations as determined by a LA-ICP-MS procedure identical to the one for olivine. Bulk rock V concentrations used to determine fO2 is also listed, along with a bulk V concentration from which 5%, 10%, 20%, 40% and 50% of the measured clinopyroxene removed. Finally, the calculated fO2 for each sample is listed, along with an fO2 calculated using the bulk rock compositions with varying amounts of clinopyroxene removed. As demonstrated by the modeling shown here, clinopyroxene accumulation has a negligible effect on the calculated fO2 unless extremely large amount of clinopyroxene was accumulated in the bulk rock samples. CPX -clinopyroxene, 2s -two standard deviations, N -number of clinopyroxenes analyzed.