Redox variations in Mauna Kea lavas, the oxygen fugacity of the Hawaiian plume, and the role of volcanic gases in Earth’s oxygenation

Significance

Volcanic degassing at Mauna Kea volcanoes decreases the oxidation state of both Fe and S in the magmas, consistent with recent results from Kilauea volcano. The least degassed magmas from Mauna Kea are more oxidized than midocean ridge basalt magmas, possibly due to plate tectonic cycling of material from the oxidized surface to the deep mantle, which is then returned to the surface as a component of buoyant plumes. If the degassing trend observed for Mauna Kea applies to basaltic systems more broadly, decompression does not lead to significant changes in the reducing capacity of the exsolved gases and thus cannot explain the rise of oxygen at the Archean/Proterozoic boundary.

Abstract

The behavior of C, H, and S in the solid Earth depends on their oxidation states, which are related to oxygen fugacity (fO₂). Volcanic degassing is a source of these elements to Earth’s surface; therefore, variations in mantle fO₂ may influence the fO₂ at Earth’s surface. However, degassing can impact magmatic fO₂ before or during eruption, potentially obscuring relationships between the fO₂ of the solid Earth and of emitted gases and their impact on surface fO₂. We show that low-pressure degassing resulted in reduction of the fO₂ of Mauna Kea magmas by more than an order of magnitude. The least degassed magmas from Mauna Kea are more oxidized than midocean ridge basalt (MORB) magmas, suggesting that the upper mantle sources of Hawaiian magmas have higher fO₂ than MORB sources. One explanation for this difference is recycling of material from the oxidized surface to the deep mantle, which is then returned to the surface as a component of buoyant plumes. It has been proposed that a decreasing pressure of volcanic eruptions led to the oxygenation of the atmosphere. Extension of our findings via modeling of degassing trends suggests that a decrease in eruption pressure would not produce this effect. If degassing of basalts were responsible for the rise in oxygen, it requires that Archean magmas had at least two orders of magnitude lower fO₂ than modern magmas. Estimates of fO₂ of Archean magmas are not this low, arguing for alternative explanations for the oxygenation of the atmosphere.

Modern arc and back-arc basin magmas have higher ratios of Fe³⁺ to total Fe (i.e., Fe³⁺/[Fe³⁺+Fe²⁺], or Fe³⁺/ΣFe), and S⁶⁺ to total S (i.e., S⁶⁺/[S⁶⁺+S^2-], or S⁶⁺/ΣS), than midocean ridge basalt (MORB) magmas (1–7). Based on the known relationships between Fe³⁺/ΣFe, S⁶⁺/ΣS, oxygen fugacity (fO₂), and melt composition (8, 9) and on recent observations that Fe³⁺/ΣFe of subduction-related magmas are only weakly dependent on factors such as crystal−liquid fractionation and degassing (7, 10, 11), the higher Fe³⁺/ΣFe ratios of arc and back-arc basalts relative to MORBs have been attributed by most authors to elevated fO₂ in their mantle sources resulting from addition of oxidized hydrous fluids or silicate melts from subducted slabs (7, 11, 12). Ocean island basalts (OIBs) are more variable in fO₂, with Fe³⁺/ΣFe and S⁶⁺/ΣS ratios that range from lower than to higher than MORBs (4, 13–17), but it is unclear whether these differences relate to variable amounts of ancient, more oxidized subducted material in the mantle sources of OIB (e.g., ref. 18) or to other factors related to differences in the petrogenesis of OIB and MORB magmas [e.g., temperature and pressure of melting (15), degassing of certain volatile elements (16, 17)].

There are several processes by which Fe³⁺/ΣFe and S⁶⁺/ΣS ratios could change during magmatic differentiation, and these could contribute to the observed variability in fO₂ among basalts erupted at Earth’s surface. One possibility is crystal−liquid fractionation: In a system closed to oxygen, olivine and pyroxene crystallization would result in an increase in the Fe³⁺/ΣFe of residual liquids; alternatively, spinel crystallization could decrease this ratio in residual liquids. Degassing is another possibility: Degassing and loss of H₂ can lead to oxidation of residual liquids (19). Likewise, although CO₃²⁻ species are thought to be the dominant forms of dissolved C in terrestrial magmas, CO is a nonnegligible component of exsolved gases and degassing, and loss of CO-bearing vapor could lead to oxidation of residual liquids (20). Also, both oxidized and reduced forms of S are dissolved in basaltic melts and coexisting gas. Progressive degassing could lead to either oxidation or reduction of the residual melt phase, depending on the speciation of S in the melt and gas. Because the direction of this latter effect has been thought to be pressure-dependent, it has been proposed that the rise of atmospheric oxygen near the Archean−Proterozoic boundary reflected the generation of subaerial crust and a subsequent, progressive decrease in the pressure at which magmas erupted, resulting in an increase in average oxidation state of degassed volcanic volatiles (21, 22).

To isolate and quantify the effects of degassing on the fO₂ of magmas and to place quantitative constraints on the ability of the solid Earth’s fO₂ to impact pO₂ at Earth’s surface via magmatic degassing, we measured both Fe³⁺/ΣFe and S⁶⁺/ΣS ratios in submarine glasses from the Mauna Kea volcano in Hawaii that record shallow degassing of an S+H₂O+CO₂-bearing gas phase. These samples are from a well-characterized sample suite recovered by the Hawaii Scientific Drilling Project (HSDP) (23) that records the transition from submarine to subaerial eruption during the shield-building phase of Mauna Kea’s growth. In this closely related suite of samples, the concentrations of volatiles are known to vary primarily due to shallow-level degassing, with slight variations in the major element composition that suggest only a small amount of concomitant crystal−liquid fractionation (24). For this suite of OIB glasses, it is reasonable to attribute observed correlations of volatile contents of glasses with their Fe³⁺/ΣFe and S⁶⁺/ΣS ratios nearly entirely to the effects of shallow-level degassing on the valence states of dissolved Fe and S in the liquid, and thereby to the effects of degassing on fO₂.

Materials and Methods

The HSDP recovered >3,000 m of core, most of which samples 150,000- to >500,000-y-old subaerial and submarine basalts from Mauna Kea volcano (25, 26). The glasses range from undegassed to degassed with respect to H₂O and S, although even the so-called undegassed samples likely also lost significant CO₂ during ascent (24). The low-SiO₂ glasses have, on average, higher H₂O contents than high-SiO₂ glasses, but the total S contents of the two suites are similar (Fig. 1 A−D and ref. 24).

Fig. 1.

Observations of the redox states of Fe and S in Hawaiian submarine volcanic glasses from XANES compared with H₂O and S abundances. (A) Fe³⁺/ΣFe ratios v. S concentrations, (B) Fe³⁺/ΣFe ratios v. H₂O concentrations, (C) S⁶⁺/ΣS ratios v. S concentrations, (D) S⁶⁺/ΣS ratios v. H₂O concentrations, and (E) S⁶⁺/ΣS ratios v. Fe³⁺/ΣFe ratios in the glasses. Representative error bars for each axis are shown in each panel (see SI Appendix). Sulfur and H₂O abundance data are from ref. 24. The vertical dashed lines in A and C are the approximate S content at sulfide saturation for HSDP2 glasses with the highest FeO* content (∼12.5 wt%), taken from a line of best fit through S versus FeO* contents of MORB that are sulfide saturated (20, 24). Measurements of S⁶⁺/ΣS = 0 in A, C, and E reflect analyses that lack spectral evidence for S⁶⁺, i.e., S⁶⁺ is below the analytical detection limit. The black dashed curve in E shows the calculated trend of Fe³⁺/ΣFe vs. S⁶⁺/ΣS ratios for a representative HSDP2 glass as fO₂ is varied. The initial fO₂ is based on the Fe³⁺/ΣFe ratio and major element glass composition of glass sample SR0915-0.4 (26), at 1,200 °C and 1 atm using the calibration of ref. 8. The corresponding S⁶⁺/ΣS ratio was then calculated from fO₂ using the calibration of ref. 9. Tick marks show the calculated fO₂ relative to the QFM buffer at 1,200 °C and 1 atm. Data from ref. 17.

We focused on glassy samples from two intervals of the HSDP2 core drilled in 1999 (23). The first suite consists of high-SiO₂ pillow-rim glasses recovered from 2,703 to 2,841 m below sea level (mbsl), and the second suite consists of low-SiO₂ pillow-rim glasses recovered from 2,438 to 2,470 mbsl. The difference in SiO₂ contents between the two suites is thought to arise as the result of differences in the pressure of melt segregation in the mantle source of Mauna Kea magmas [low-SiO₂ lavas segregate at higher pressures (27)] and/or differences in the extent of melt−rock interaction before or during melt segregation [low-SiO₂ lavas reflect lower extents of melt−rock interaction (26)]. The H₂O and S contents are positively correlated in both suites, and the samples have variable but generally low CO₂ contents (0.16 wt% to 0.81 wt% H₂O; 350 ppm to 1,440 ppm S; 30 ppm to 189 ppm CO₂; see SI Appendix). The FeO* (total Fe, expressed as FeO) concentrations of both suites span a narrow range (10.2 wt% to 12.4 wt%; SI Appendix, Fig. S6) and do not correlate with S contents, but instead plot below the empirical trend in Fe vs. S contents of basaltic magmas of comparable major element compositions at sulfide saturation (SI Appendix, Fig. S6 and ref. 24). Based on these observations, the volatile contents of these submarine glasses have been inferred to reflect variable extents of degassing of S+H₂O+CO₂-bearing vapor from magmas that were shallowly stored below the seafloor (5 bars to 50 bars total pressure; SI Appendix, Fig. S6 and ref. 24) and then erupted on the seafloor.

To determine the effect of degassing on magmatic fO₂, we measured Fe³⁺/ΣFe and S⁶⁺/ΣS ratios of these pillow glasses by micro-X-ray absorption near-edge structure (µ-XANES) spectroscopy (see SI Appendix for a description of analytical methods, spectral fitting procedures, and error assessments for both Fe and S; see Dataset S1 for results).

Results and Discussion

The high- and low-SiO₂ suites each have Fe³⁺/ΣFe and S⁶⁺/ΣS ratios that decrease with decreasing H₂O and S contents (Fig. 1 A−D), and the glasses within each suite that have the highest S and H₂O concentrations also have the highest proportions of oxidized Fe (Fe³⁺) and S (S⁶⁺). The low-SiO₂ glasses have a narrower range of Fe³⁺/ΣFe and S⁶⁺/ΣS ratios and have Fe³⁺/ΣFe ratios 0.01 to 0.02 units higher than the high-SiO₂ glasses at similar S concentrations (Fig. 1A). Despite the subtle differences between the high- and low-SiO₂ glasses, the two suites together define a single overall trend in Fe³⁺/ΣFe and S⁶⁺/ΣS space (Fig. 1E). Our results demonstrate that degassing of a S+H₂O+CO₂ vapor is correlated with reduction of Fe³⁺ to Fe²⁺ and of S⁶⁺ to S^2- (Fig. 1). This result is consistent with recent observations of variable Fe³⁺/ΣFe ratios in olivine-hosted melt inclusions from Erebus, Kilauea (OIB; shown for comparison with our Mauna Kea data in Fig. 1 A and B), and Agrigan (arc) volcanoes, although the major element compositions, the S, H₂O, and CO₂ concentrations, and Fe³⁺/ΣFe ratios in the case of Erebus and Agrigan differ significantly from those studied here, and the Agrigan trend differs from what is observed for other arc glasses (see SI Appendix) (10, 16, 17, 28). Our work extends the observation to include direct measurements of the simultaneous reduction of S⁶⁺/ΣS ratios.

Magmatic fO₂ values relative to the quartz−fayalite−magnetite (QFM) oxygen buffer for each of our samples at 1 atmosphere and 1,200 °C were calculated based on their Fe³⁺/ΣFe ratios and major element compositions (see SI Appendix) (Figs. 2 and 3 A and B and refs. 8 and 29). Magmatic fO₂ decreases from ∼QFM+1.0 in the least degassed samples (i.e., those with highest S and H₂O contents) to ∼QFM−0.4 in the most degassed samples (Fig. 3). This ∼1.4 orders of magnitude range in fO₂ from the least degassed to the most degassed samples is consistent with inferred and measured fO₂ values for other Hawaiian lavas, based on Fe−Ti oxide oxybarometry (30), gas-mixing experiments (31), gas equilibria (32–34), and XANES measurements of olivine-hosted melt inclusions and matrix glasses from Kilauea (Fig. 2 and refs. 17, 28). A key point is that whole-rock measurements (vertical gray bars, Fig. 2), which represent the final, degassed lava and can be altered rapidly and pervasively at Earth’s surface, record highly variable Fe³⁺/ΣFe ratios that span more than four orders of magnitude fO₂ (Fig. 2), making it difficult to infer from such measurements what the preeruptive fO₂ may have been for Hawaiian magmas. The Fe−Ti oxide record (light gray horizontal bar, Fig. 2), the constraints from gas-mixing experiments (black horizontal bar, Fig. 2), and the Fe³⁺/ΣFe ratios measured in less degassed melt inclusions (blue vertical bars, Fig. 2) all indicate that the fO₂ of the crystallizing magmas is near QFM+0.2 to QFM+0.8, broadly consistent with the fO₂ of the least degassed magmas from this study.

Fig. 2.

Histogram comparing Fe³⁺/ΣFe ratios in 29 HSDP2 glasses (this study, red and pink bars) and 44 glasses from Kilauea volcano (17) (blue bars), all measured using XANES, compared with those in whole rocks measured by wet chemistry (gray bars). Also shown are ranges in fO₂ values at the temperatures relevant for each dataset, relative to QFM (shown as shaded horizontal bars) for Hawaiian volcanism based on analyses of volcanic gases (33, 38) analyses of coexisting Fe−Ti oxides (30), or gas mixing experiments (31). The heavy red vertical dashed line is the average of three undegassed HSDP2 samples. The fO₂ scale on the upper horizontal axis was calculated as in Fig. 1, for a representative MORB composition (42), and should be taken as approximate fO₂ for OIB samples.

Fig. 3.

(A) The fO₂ values (relative to QFM) for each of the HSDP glasses (calculated as in Fig. 1) vs. S concentration of the glass (24). Also shown are the results of fractional (red curve) and batch (blue curve) degassing calculations based on the D-Compress software from ref. 35. Tick marks represent increments of decompression from 120 bars to 1 bar total pressure. Symbols are as in Fig. 1. (B) Same as A, except fO₂ is shown vs. H₂O content of the glass.

We modeled the change in magmatic fO₂ with progressive degassing of a C−O−H−S vapor species using the software package D-Compress, which executes the gas−melt equilibrium model of ref. 35. This thermodynamically based model assumes that, at equilibrium, the chemical potentials of volatile species dissolved in a fluid (i.e., gas)-saturated silicate melt must be equal in the silicate melt and the fluid. Given some constraints on the system (i.e., initial H₂O and CO₂ contents, magmatic fO₂), D-Compress computes the concentration and speciation of C, H, O, and S in coexisting fluid and silicate melt as functions of pressure and temperature, based on experimental calibrations of the solubilities of H₂, H₂O, CO, CO₂, SO₂, H₂S, and S₂ species in silicate melts, and calculations of homogeneous equilibria in the gas phase for the same species. With D-Compress, we calculated redox conditions and coexisting vapor and silicate liquid compositions by progressively decreasing total pressure (at a constant temperature of 1,190 °C; the effect of variable temperature on fO₂ relative to QFM is negligible; see item i below) from the assumed starting pressure of 120 bars down to 1 bar, both for fractional (red curves, Fig. 3) and batch (blue curves, Fig. 3) degassing (see SI Appendix, Modeling the Impact of Degassing on Melt fO2).

Both the fractional and batch models predict a decrease in magmatic fO₂ during the degassing that would accompany depressurization during magma ascent. The fractional model results in a decrease in magmatic fO₂ from QFM+1.1 to QFM−0.2, or about 90% of the observed change in fO₂ observed in the Mauna Kea glasses (Fig. 3). Although the trends are similar for fractional and batch calculations, the models decrease in fO₂ only to QFM (batch) and QFM−0.2 (fractional) when they reach 1 bar (i.e., not quite as much as the fractional degassing model).

The decrease in fO₂ associated with degassing observed in the Mauna Kea glasses is dominated by changes in the speciation of S in the vapor phase. We demonstrate this first by using the same model to simulate degassing of S-free melts in which H₂O or CO₂ was the only dissolved volatile component (H₂O-only case, SI Appendix, Fig. S7A; CO₂-only case, SI Appendix, Fig. S8). We chose melts with much higher volatile contents than observed in the HSDP2 samples of this study to enhance the fO₂ response to degassing. In the H₂O-only and CO₂-only cases (see SI Appendix, Fig. S8 for CO₂-only case), the batch-degassing modeling predicts that the fully degassed basaltic melts (i.e., melts decompressed to 1 bar) are, respectively, 0.6 and 0.1 log units higher in fO₂ than the undegassed basaltic magmas that start with 4.5 wt% H₂O or 1,250 ppm CO₂ [note, however, that neither of these extremes results in a change in fO₂ large enough to explain the oxidized nature of arc and back-arc basalts relative to MORB (SI Appendix, Fig. S7A for H₂O-only case; SI Appendix, Fig. S8 for CO₂-only case)]. In the fractional degassing modeling of a H₂O-only vapor produces a basaltic melt at 1 bar total pressure that is 0.7 log units higher in fO₂ than the undegassed basaltic magma, slightly higher than the batch calculation.

In both the batch and fractional degassing scenarios when H₂O is the only volatile component, the increase in melt fO₂ results from the fact that dissolved H₂ is negligible in terrestrial magmas at low pressure (36), but the amount of H₂ in vapor coexisting with hydrous melt is not. Consequently, when dissolved H₂O is transferred from the melt into the vapor, H₂ is generated in the vapor by disproportionation of a fraction of the H₂O. This disproportionation must likewise generate an equal amount of O₂, but, since, at these fO₂ levels, fH₂/fO₂ in the vapor is much greater than 1, the excess oxygen generated by disproportion of H₂O is partitioned into the coexisting melt, resulting in increases in fO₂, and Fe³⁺/ΣFe and S⁶⁺/ΣS ratios in the melt. In C-bearing systems, the assumed negligible amount of CO dissolved in the melt but the nonnegligible CO content of the vapor leads to the same effect when CO₂ is transferred from the melt in the vapor (see SI Appendix, Fig. S8).

We also simulated the degassing of basaltic melt in which S-bearing species are the dominant volatile components dissolved in the melt. The model basaltic melt contains 6,000 ppm S, no C, and 128 ppm total H (the amount of H necessary to make H₂S and 0.1 wt% H₂O). In this S-dominant case, the model predicts that, for batch degassing, on decompression to 1 bar, the melt is 1.3 log units lower in fO₂ than the undegassed melt (SI Appendix, Fig. S7B); the fO₂ decreases by more than 1.5 log units for fractional degassing. In systems in which S is the dominant volatile element, the change in fO₂ during degassing is more complex than for CO₂- and H₂O-dominant systems because S is present both as sulfide and sulfate in the melt, and in multiple oxidation states (i.e., S₂, H₂S, and SO₂) in the vapor as well. We nevertheless are able to infer that the decrease in fO₂ upon degassing for the modeled S-dominant basaltic melt (SI Appendix, Fig. S7B) is dominated by the conversion of sulfide dissolved in the melt to SO₂ in the vapor. That is, as suggested by many previous authors (e.g., refs. 6, 10, 16, 17, 28, 30, and 37), degassing reactions such as 3Fe₂O₃(melt) + FeS(melt) = 7FeO(melt) + SO₂(vapor) lead to a decrease in the Fe³⁺/ΣFe ratios, to a decrease in S⁶⁺/ΣS ratios of the residual S dissolved in the melt (i.e., Fig. 1 A, C, and E), and thus to a decrease in the fO₂ of the system.

In a system containing H₂O, S, ± CO₂ at the fO₂ levels appropriate to Hawaiian magmas, the tendency of H₂O and CO₂ degassing to increase the fO₂ is overwhelmed by the tendency to decrease the fO₂ due to degassing of S-bearing species. For Hawaiian magmas, it is known that CO₂ is the dominant degassing volatile component until pressures of ∼100 bars (e.g., ref. 38), at which point H₂O and S come out nearly simultaneously. The degassing of dominantly CO₂ at pressures higher than that at which H₂O and S begin to transfer from melt to vapor in significant quantities increases fO₂ slightly, but further decreases in pressure that result in simultaneous degassing of H₂O and S-bearing species at low pressure lead to significant decrease in fO₂ (SI Appendix, Fig. S8). We emphasize, however, that the effects of degassing on fO₂ will depend on the sequence of loss of the volatile species with decreasing pressure, which in turn depends on both the solubilities of volatile species and on the initial abundances of volatile components in the undegassed magma.

The least degassed Mauna Kea glasses record an fO₂ that is approximately one order of magnitude higher than MORB glasses at similar MgO contents; i.e., corresponding to Fe³⁺/ΣFe of ∼0.22 for the HSDP2 samples, compared with Fe³⁺/ΣFe of ∼0.16 for MORB (SI Appendix, Fig. S10). This result is broadly similar to results from XANES measurements of matrix glasses and melt incusions from Kilauea volcano (17, 28), although the highest fO₂ values from the Mauna Kea glasses (QFM+1.0; this study) are higher than those from Kilauea (QFM+0.7; refs. 17 and 28). However, the Fe³⁺/ΣFe ratios and inferred fO₂ values of degassed Mauna Kea glasses overlap with those of average undegassed MORB, and, if the reducing effects of low-pressure degassing were not taken into account, it might be concluded that parental magmas and mantle sources at Mauna Kea had fO₂ values similar to those of MORB parental magmas and their mantle sources. However, the appropriate comparison is between the least-degassed samples from Mauna Kea and the MORB glasses. Although this comparison indicates that undegassed Mauna Kea magmas were, on eruption, more oxidized than MORB, caution must be exercised in generalizing this statement to their parental magmas and to their mantle source regions.

The mantle sources of Hawaiian magmas are at higher pressures and temperatures than MORB magmas (e.g., ref. 39 and references therein), and corrections must be made to the fO₂ values based on measurements of fractionated, lower-pressure and -temperature glasses to determine the fO₂ values that prevailed in their mantle sources.

• i)The relative change in fO₂ of silicate liquids and the fO₂ of the QFM oxygen buffer at constant pressure and increasing temperature is small [<0.005 log unit decrease from 1,200 °C to 1,400 °C at 1 GPa, and <0.02 log unit decrease in the same temperature range at 3 GPa (8, 29)], so the effect of the temperature differences between MORBs [1,400 °C to 1,525 °C (39)] and undegassed Hawaiian magmas [∼1,725 °C (39)] on inferences about their source fO₂ relative to the QFM oxygen buffer can be neglected.
• ii)In contrast, the relative change in fO₂ of silicate liquids and the fO₂ of the QFM oxygen buffer at constant temperature and increasing pressure is nonnegligible: The fO₂ of a typical basaltic silicate liquid increases by 0.39 log units relative to QFM when pressure increases from 1 bar to 3 GPa at 1,300 °C and by 0.36 log units relative to QFM over the same pressure interval at 1,450 °C (8, 29). Thus, assuming a typical average pressure of ∼1.0 GPa for MORB melt generation (e.g., ref. 40) and ∼3 GPa for Hawaiian melt generation (e.g., ref. 41 and references therein), the approximately one order of magnitude higher fO₂ (relative to the QFM buffer at 1 atmosphere) of the undegassed HSDP2 glasses relative to MORB glasses increases to ∼1.14 orders of magnitude (relative to the QFM buffer at the pressure of melt generation) difference in their sources after correcting for pressure alone.
• iii)Considerable (and different amounts of) olivine fractionation occurred between the generation of primary melts in their mantle source regions and the eruption of the melts from which MORB and Hawaiian glasses were quenched. Assuming that crystallization occurs at a constant total oxygen content for the system, the effect of olivine crystallization is to increase Fe³⁺/ΣFe and increasing fO₂ with progressive fractionation, potentially overprinting differences that may have existed between the unfractionated primary melts equilibrated with their residual mantle sources. Mauna Kea primary liquids are thought to have as much as 18 wt% to 20 wt% MgO, assuming they equilibrated with Fo₉₁ olivines in their sources (41), whereas MORB primary magmas are thought to have only 12 wt% to 14 wt% MgO, assuming they equilibrated with Fo₉₀ olivines in their sources (e.g., ref. 42; see SI Appendix). If this is correct, then primary Mauna Kea liquids fractionated a larger amount of olivine than MORB magmas (44% olivine removed for Mauna Kea, compared with 23% for MORB) to reach 7 wt% MgO. We calculated the Fe³⁺/ΣFe ratios and fO₂ of primary mantle melts for MORB and Mauna Kea by taking an undegassed composition from each setting and adding the equilibrium composition olivine back to the melt in 0.1% increments, treating Fe²⁺ and Fe³⁺ as conservative elements (see SI Appendix). This can account for 0.5 ± 0.1 log units (the uncertainty estimate accounts for the range of primary melt MgO contents described above) of the observed difference in fO₂ between undegassed Mauna Kea and MORB melts. When combined with the temperature and pressure effect on silicate melts in items i and ii, this results in a calculated difference in fO₂ between Mauna Kea and MORB sources of 0.66 ± 0.1 log units, compared with the ∼1.16 log unit difference between the fO₂ values of the erupted, highly fractionated melts preserved as glasses.
• iv)It has been suggested that lower temperatures of melting can generate melts with higher abundances of Fe³⁺ than melts produced during normal MORB melting and that it might be difficult to distinguish unambiguously melts of an intrinsically oxidized source from those of a lower-temperature source because both would have relatively high Fe³⁺/ΣFe ratios (43). Although we have not tried to quantify this effect, if the reverse temperature effect were true (i.e., that higher temperatures lead to melts with lower Fe³⁺ abundances; e.g., ref. 15), the fact that the Hawaiian plume is generally thought to have an elevated mantle potential temperature of ∼1,725 °C compared with 1,400 °C to 1,525 °C for MORB sources (39) suggests that unmelted Hawaiian sources have even higher fO₂ relative to MORB sources than suggested by direct comparison of differences in Fe³⁺/ΣFe ratios between the melts of these sources. To the degree that this temperature effect is important, it would suggest that the residual sources of the Hawaiian melts studied here are more than ∼0.66 log units more oxidized relative to QFM than MORB primary magmas; i.e., the ∼0.66 log unit higher fO₂ of Hawaiian primary magmas relative to MORB magmas would be a lower limit to the differences in the fO₂ values of their source regions before melting.

After considering these petrogenetic factors, we conclude that the difference in fO₂ between the sources of Mauna Kea magmas and MORBs, while less than that suggested by the “raw” Fe³⁺/ΣFe ratios of undegassed Mauna Kea and MORB glasses, is still significant at $\stackrel{\sim }{>}$0.6 ± 0.1 log units relative to QFM. The higher fO₂ values of undegassed Hawaiian magmas relative to MORBs in turn reflect differences in the fO₂ values of the Hawaiian plume relative to MORB sources. This suggests that differences in fO₂ in the modern upper mantle are not restricted to the ∼1.3 orders of magnitude difference between the subarc mantle wedge (generally attributed to recycling of oxidized sediments and crust by subduction) and the mantle beneath ridges (likely representative of most of the upper mantle) summarized in the Introduction. If, as is widely thought (18, 25, 44), the sources of OIBs, including Hawaii, contain near-surface material recycled by plate tectonics into the deep mantle that is then returned to the surface in the form of ocean island volcanism, the higher fO₂ value relative to the QFM buffer inferred for the mantle sources of the Mauna Kea samples relative to the sources of MORBs could simply reflect that the recycled plume component(s) are more oxidized than the average mantle. This hypothesis is also consistent with observations that Mauna Kea submarine glasses from the studied intervals in the HSDP2 core have relatively radiogenic Sr (0.7035 to 0.7037) and nonradiogenic Nd isotopic compositions [0.51216 to 0.51219 (25)], two characteristics consistent with the presence of recycled components in the Hawaiian plume: either recycled sediment and/or oceanic crust (e.g., refs. 25, 45, and 46). If we approximate the primary melt of Mauna Kea magmas as a simple mixture between a primary melt of garnet-bearing lherzolite, with 5.6 wt% FeO* (3 GPa melt composition from ref. 47) and 1.0 wt% Fe₂O₃ [the Fe₂O₃ content corresponding to a melt at 3GPa, 1,450 °C, and QFM (8)], and average altered oceanic crust, with 5 wt% FeO* and 1.97 wt% Fe₂O₃ (13), it suggests that ∼20% of the sources of Mauna Kea primary melts (with Fe³⁺/ΣFe ∼0.142; see SI Appendix) is derived from altered oceanic crust in the Hawaiian plume. This is similar to the high end of the range of other estimates for the amount of recycled material present in the Hawaiian plume [4 to 20% (45, 48, 49)] based on the major and trace element and isotopic compositions of Hawaiian lavas.

Using our results as a starting point, we now consider the influence of gas evolved during degassing of basaltic magmas on the oxidation state of the atmosphere. Volcanic gases are an important input to the inventory of H-, S-, and C-bearing molecules in the atmosphere, oceans, and crust (e.g., ref. 50), and it has been hypothesized that a change in the oxidation state of volcanic gases played a role in the “Great Oxidation Event” at ∼2.3 Ga (e.g., refs. 21 and 51–55). The general idea is that, before the rise of atmospheric O₂, volcanic gas compositions were, overall, more reducing compared with their Proterozoic and Phanerozoic counterparts, leading to increased sinks for atmospheric O₂ that would have limited the accumulation of O₂ produced by oxygenic photosynthesis (i.e., volcanic gases had a value of f > 1 according to Holland’s criterion, Eq. 1 below). Then, a change in magmatic fO₂ (22) or volcanic degassing conditions (21, 53) led to a change in the composition of volcanic gases added to the atmosphere that resulted in their being more oxidizing, such that volcanic gases no longer contained enough reducing power to titrate the photosynthetic production of O₂, which thereafter began to accumulate in the atmosphere [i.e., volcanic gases had a value of f < 1 (54)]. This capacity for volcanic gases to react with and remove atmospheric O₂ has been referred to as the “reducing power,” and is characterized using Holland’s criterion based on the parameter, f (54) (m_i is the mole fraction of species i in a volcanic gas), defined as

$$f=\frac{m_{H2}+0.6\hspace{0.17em}{m}_{CO}-0.4\hspace{0.17em}{m}_{CO2}+3\hspace{0.17em}{m}_{H2S}}{3.5\hspace{0.17em}(m_{SO2}+m_{H2S})}+\frac{1}{3.5}.$$ [1]

We calculated the f value for the incremental gas composition in equilibrium with the decompressing magma that is provided from the modeled degassing trajectory of Mauna Kea lavas shown in Fig. 3. The value of f at all pressure increments is ∼0.5, does not vary significantly upon decompression from 120 bars to 1 bar, and is consistent with an atmosphere that allows for the accumulation of photosynthetically produced O₂. Thus, to the degree that degassing of Hawaiian magmas can represent degassing of Archean magmas, they would not have been sufficiently reducing to preclude the accumulation of O₂ in the ancient atmosphere if oxygenic photosynthesis was prevalent at that time. We note that the relative constancy from our modeling of f during decompression is inconsistent with suggestions that a systematic decrease in the pressure of degassing from 100 bars to 1 bar (i.e., a result of the production of subaerial continental crust, thus enabling magmas to degas directly to the atmosphere) could have led to atmospheric oxygenation (i.e., by reducing f from a value of >1 to <1). These suggestions were based on the increase in SO₂/H₂S ratio of volcanic gases with decreasing pressure, and the assumption that, at high pressure, emissions of SO₂ are negligible. However, the SO₂/H₂S ratio of volcanic gases at these pressures can be ∼0.5 to 1.0, indicating that SO₂ emissions at these pressures cannot be assumed to be negligible. Moreover, there are significant changes in C abundance and speciation in the gas phase over this pressure range that must be considered simultaneously with changes in sulfur speciation. Model degassing trajectories for modern MORB-like and arc-like magmas (SI Appendix, Fig. 11) demonstrate that the f value of these gases can be low at high pressures when CO₂ dominates the gas composition (at 1,000 bar, f = −16 for MORB and f = −10 for arc magmas) and increase dramatically as the total C content of the gas decreases and the proportion of CO₂ to CO decreases (at 10 bar, f = −5 for MORB and f = −3 for arc magmas). Nevertheless, the degassing scenarios examined here have f < 1 for the entire degassing trajectory, and they do not suggest that any reasonable changes in volatile contents or pressures of degassing would lead to the production of volcanic gases with f > 1. Thus, given their fO₂ values, degassing of any of the most abundant igneous rocks on Earth today would, regardless of the pressure of degassing or their volatile contents, support oxygen accumulation in the atmosphere. Only if we decrease the fO₂ of the degassing magma in our calculations to QFM−2.3 do the released incremental volcanic gases have f > ∼1 (solid gray line, SI Appendix, Fig. S11C). Consequently, we take this to be the upper limit of fO₂ for degassing magmas that would have been capable of diminishing atmospheric O₂ concentrations. These results are broadly similar to calculations presented by ref. 55, except that they place the upper limit of fO₂ for degassing magmas capable of diminishing atmospheric O₂ concentrations somewhat higher, near QFM−1.8. In that work, average volcanic gas-phase chemistry was estimated from present-day measurements of volcanic gas emissions, which may contribute to the difference in upper limits between our study and that of ref. 55.

Our degassing calculations suggest that changes in the average degassing pressure of the most abundant modern basaltic magma types (i.e., the calculations on Mauna Kea lavas as representative of OIBs and on representative MORB and arc compositions summarized in SI Appendix, Fig. S11) could not have strongly influenced the oxygen-poor conditions at Earth’s surface before ∼2.3 Ga. For volcanic processes to have been responsible for inhibiting the rise of oxygen in the Archean, magmas would have to have been, on average, more reduced than modern MORB magmas by more than two orders of magnitude (SI Appendix, Fig. S11). Vanadium contents in olivines in Archean komatiites suggest that magmatic fO₂ in Archean komatiites were between QFM−1 and QFM+2 (56), which is outside of the range of fO₂ levels required to prevent oxygenation of the atmosphere, based on our calculations, as well as those presented by ref. 55. If these trace metal constraints accurately describe the redox state of average Archean magmas, then volcanic gases from Archean magmas would also have been unable to limit oxygenation of the atmosphere if photosynthetic microbes were present at that time. We conclude that, if volcanic processes are required to play a major role in inhibiting the rise of oxygen in the Archean, either photosynthetic microbes were not present or Archean basalts were significantly more reducing than modern magmas or than current evidence suggests for Archean basalts.