log units below the buffer (Frost & McCammon, 2008; Stagno, 2019). This process explains the production of kimberlites and some carbonatites, the rarity of which in the Archean has been ascribed to more reducing mantle conditions by 0.5 to 0.7 log units (Foley, 2011). Indeed, using the V/Sc redox proxy, a well‐characterized mantle eclogite suite from Lace in the Kaapvaal craton, with demonstrated spreading ridge‐derived origin, has yielded an fO2 (FMQ) of –1.7±1.1 for the ambient mantle at ca. 3 Ga (Aulbach & Viljoen, 2015). Based on this value, redox melting occurs at substantially shallower depths (~100 to 80 km; Aulbach & Stagno, 2016) with the formation of carbonate–silicate magmas, precluding the formation of carbonatite below typically thicker continental lithosphere, and possibly restricting the formation of kimberlites to unusually oxidized and/or warm mantle regions. Using additional mantle and orogenic eclogite suites sampling Proterozoic to Archean spreading ridges, the ambient mantle was shown to have evolved across the Archean–Proterozoic boundary, from ∆logƒO2 of ~FMQ‐1.19±0.33 to FMQ‐0.26±0.44 (Aulbach & Stagno, 2016). This evolution is exactly mirrored by that recorded in komatiites, which show an increase in ~1.3 log units between 3.48 and 1.87 Ga (Nicklas et al., 2018, 2019; Fig. 2.3a). This increase and the absolute fO2 values recorded across the Archean–Proterozoic are well within the range of, and therefore not in conflict with, proposed values after magma ocean solidification (~FMQ‐3 to FMQ) obtained from thermodynamic and climate modelling (Pahlevan et al., 2019). On the other hand, the compositions of inclusions in sublithospheric diamonds testify to the strong chemical and redox gradients, ranging from carbonates to solidified iron–nickel–carbon–sulfur melts (Brenker et al., 2007; Kaminsky et al., 2015; Smith et al., 2016). Indeed, diamond formation is favoured by contrasting redox states, which are typically achieved when deeply subducted slabs containing oxidized components are juxtaposed with the predominantly reducing, metal‐saturated mantle at depths > 250–300 km (Foley, 2011; Rohrbach & Schmidt, 2011). Thus, we take the relatively high fO2 retrieved from the ferric iron content of majorite inclusions in sublithospheric diamonds (Kiseeva et al., 2018) to indicate local mantle conditions in the vicinity of recycled slabs that are not representative of the mantle as a whole.
2.4. THE MANTLE GREAT OXIDATION EVENT: FACT OR ARTEFACT?
The redox evolution of Earth’s interior, as inferred from V/Sc (Fig. 2.3a), is central to the debate on the role of the mantle in the transition from an anoxic to an oxygenated atmosphere during the so‐called Great Oxidation Event (GOE) around 2.4–2.2 Ga, which was preceded by “whiffs of oxygen” (Lyons et al., 2014) as shown in Fig. 2.3b. The mantle–asthenosphere connection was hypothesized early (Kasting et al., 1993; Holland, 2002), but subsequently dismissed by various workers finding only small or no difference in the redox state of Archean and post‐Archean magmas (e.g., Canil, 1997; Delano, 2001; Li & Lee, 2004; Trail et al., 2011; Rollinson et al., 2017). In contrast, the data shown in Fig. 3.3c document a redox evolution based on samples derived from both ambient mantle and deeper “plume” mantle, which crosses the fO2 region where volcanic gases become sources as opposed to sinks of O2 when first whiffs of oxygen appeared (highlighted by the bar Fin = Fout after Holland, 2002) at ca. 3.5–3.2 Ga, reaffirming a role for the mantle in atmospheric oxygenation. Although the relationship between the redox state of a magma and the gases released to the atmosphere at any point in time during magmatic activity is controlled by multiple parameters (e.g., Gaillard et al., 2011; Moussallam et al., 2019), it is inescapable that the oxidation state of the sum of the gases emitted over the course of volcanism must reflect that of the magma with which they were once in equilibrium. The cause of this apparent mantle redox evolution possibly goes back to Earth’s origins. Once sufficient mass had been accreted, FeO disproportionated into Fe metal at high pressure, about 10% of which was extracted into the core, and Fe2O3, which was left behind in the magma ocean (Armstrong et al., 2019) and, thus, inherited by the rocky mantle where it partitioned into bridgmanite (Frost et al., 2004), corresponding to an increase in mantle O2 (Wade & Wood 2005). This was followed by gradual exhumation of O2‐enriched lowermost material by mantle convection processes. This was put forward by both Aulbach and Stagno (2016) and Nicklas et al. (2019) as one plausible explanation for the fO2 trends in their data. Early subduction of oxidized surficial material to raise mantle fO2 was offered as an alternative process to explain the fO2 trend (Nicklas et al., 2019). Interestingly, Archean placer diamonds formed between 3.5 and 3.1 Ga show geochemical markers of an origin by reduction of oxidized fluids accompanied by input of Archean sediments (Smart et al., 2016). However, mantle oxidation by subduction implies that O2 would have to be taken from the ocean–atmosphere system and cause a corresponding reduction in order to produce the oxidizing material carried down. Similarly, subduction of reduced material (Duncan & Dasgupta, 2017) may cause a transient increase in atmospheric O2, but, if not permanently sequestered into a “hidden” reservoir, would have a reducing effect on the mantle and ultimately cause it to release more reducing gases upon melting.
Figure 2.3 (a) V/Sc and (B) calculated LogfO2 (normalized to the FMQ buffer) corrected to 1 GPa of mid‐ocean ridge‐derived (meta)basalts sampling the ambient convecting mantle (modified from Aulbach and Stagno, 2016), where error bars on V/Sc represent 1σ of the mean per sample suite, those on ΔFMQ are propagated 1σ errors of V/Sc, and those on the age represent age ranges or 1σ errors for isochron ages reported in the literature. In two suites, samples with the lowest V/Sc were excluded for a conservative estimate. Range of V/Sc for continental Archean basalts in (a) from Li and Lee (2004). Shown for comparison in (b) are fO2 estimated by Nicklas et al. (2018; 2019) for picritic and komatiitic systems, based on the redox‐dependent distribution of V between liquidus olivine and melt. Note that because of an unresolved inter‐calibration issue, their MORB fO2 estimate is offset to higher values (ΔFMQ + 0.60±0.15) from those derived in other works (~–0.4±0.4; Li and Lee, 2004; Frost and McCammon, 2008; +0.1; Berry et al., 2018), and all results were shifted down by one log unit for comparability to the V/Sc‐derived MORB fO2 of –0.4. Range of fO2 in the magma ocean at ca. 4.5 Ga permitted based on chondritic D/H in the water ocean, which places an upper limit on H2 escape from the atmosphere, from Pahlevan et al. (2019) is shown for comparison (vertical blue-grey arrow). Timing of the Great Oxidation Event (GOE) around 2.4 Ga and “whiffs of oxygen” from Lyons et al. (2014) (c) shows the speciation of continental and oceanic volcanic gases, with mixtures thereof lying between these two curves, as a function of fO2 based on modern volcanic volatile fluxes (6.00 x 1012 for C, 1.15 x 1014 for H, and 2.13 x 1012 for S), illustrating the transition from such gases acting as sinks (reducing) and sources (oxidizing) of O2 compared to the estimated average modern O2 input (from burial of reduced components) and its upper and lower limit
(modified from Li & Lee, 2004).
Regardless of the extent and timing of mantle oxidation, both the eclogite and the komatiite datasets contain outliers testifying to redox heterogeneity in the convecting mantle, possibly inherited from the aforementioned accretion and magma ocean processes and preserved in mantle regions that had not been remixed at the time of melt generation (Nicklas et al., 2019). This raises the question of the timescales needed to mix deep oxidized mantle regions into the upper convecting mantle. Gu et al. (2016) showed that an oxidized lower mantle assemblage left over after core formation would be less dense than its reduced equivalent, facilitating its ascent, which they modelled to be completed by 3.6 Ga ago, although the exact magnitude of the density contrast is debated (Liu et al., 2018). In contrast, the data shown in Fig. 3a–c