ARTICLE

https://doi.org/10.1038/s41467-020-16493-1 OPEN

Mantle data imply a decline of oxidizable volcanic

gases could have triggered the Great Oxidation
Shintaro Kadoya 1 ✉, David C. Catling 1, Robert W. Nicklas2, Igor S. Puchtel3 & Ariel D. Anbar4

Aerobic lifeforms, including humans, thrive because of abundant atmospheric O2, but for
much of Earth history O2 levels were low. Even after evidence for oxygenic photosynthesis
1234567890():,;

appeared, the atmosphere remained anoxic for hundreds of millions of years until the ~2.4 Ga
Great Oxidation Event. The delay of atmospheric oxygenation and its timing remain poorly
understood. Two recent studies reveal that the mantle gradually oxidized from the Archean
onwards, leading to speculation that such oxidation enabled atmospheric oxygenation. But
whether this mechanism works has not been quantitatively examined. Here, we show that
these data imply that reducing Archean volcanic gases could have prevented atmospheric O2
from accumulating until ~2.5 Ga with ≥95% probability. For two decades, mantle oxidation
has been dismissed as a key driver of the evolution of O2 and aerobic life. Our findings
warrant a reconsideration for Earth and Earth-like exoplanets.

1 Department of Earth and Space Sciences/Cross-Campus Astrobiology Program, University of Washington, Box 351310, Seattle, WA 98195-1310, USA.
2 Geoscience Research Division, Scripps Institution of Oceanography, La Jolla, CA 92093, USA. 3 Department of Geology, University of Maryland, College
Park, MD 20742, USA. 4 School of Earth and Space Exploration and School of Molecular Sciences, Arizona State University, Tempe, AZ 85287, USA.
✉email: shintaro.kadoya@gmail.com

NATURE COMMUNICATIONS | (2020)11:2774 | https://doi.org/10.1038/s41467-020-16493-1 | www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-16493-1

T
he geological record of mass-independent sulfur isotope Because forg changed little with a conventional carbon cycle
fractionation shows that O2 first inundated the atmosphere model, it has been proposed that a secular decline of efficient O2
during the Great Oxidation Event (GOE) between 2.4 and sinks, rather than an increase in forg, caused the tipping point that
2.1 Ga1. However, redox-sensitive iron and molybdenum isotope initiated the GOE; this point would have been reached when the
data suggest the presence of O2 in the 3.2–3.0-Ga marine photic efficient O2 sink flux fell below the organic burial flux of
zone2,3, which implies that O2-producing cyanobacteria existed O219,22,32,33. Models show that atmospheric O2 would then
long before the GOE. Indeed, models show that, under a globally rapidly accumulate, until oxidative weathering of the continents
anoxic atmosphere, cyanobacteria-derived O2 produces photic became a significant sink, causing O2 concentrations to level off34.
zone oxygen oases4,5. Geological data also indicate the presence of Proxies suggest that the O2 level, however, was still far below
methanotrophy and oxidative nitrogen cycling in Neoarchean today’s atmospheric O2 concentrations [e.g., ref. 35].
oceans and lakes, which suggest that O2 was oxidizing sulfides Qualitatively, the pre-GOE O2 sink flux could have declined if
and ammonium to make sulfate and nitrate, respectively6–8. Such the mantle’s oxidation state increased over time32,36,37. The
inferences are consistent with phylogenetic interpretations that proportion of reducing gases in volcanic emissions depends
oxygenic photosynthesis evolved by the mid-Archean [e.g., ref. 9]. inversely on the oxygen fugacity (fO2 ) in their magma source
In addition, paleoredox proxies suggest O2 transients at 2.5 region, the upper mantle. Thus, if the Archean mantle’s fO2 was
Ga6,7,10–16, and point to the possibility of pervasive oxygenation low, the H2/H2O and CO/CO2 ratios in Archean volcanic gases
of the oceans over continental margins17,18. would have been high, suppressing atmospheric O2 levels.
The reasons for the delay of oxygenation of the atmosphere for Until recently, studies suggested that the mantle’s fO2 had been
hundreds of millions of years after the advent of oxygenic pho- similar to the modern value since at least the early Archean38–42.
tosynthesis remain unknown. Numerous hypotheses rely on the The mantle fO2 estimated by many of these studies has uncer-
notion that the atmospheric O2 level is determined by a kinetic tainties that vary between ~1 log10 unit38,43 and ~2 log10
balance between O2 supply and consumption19–23. If the rapid units41,42. Changes in the oxygen fugacity of the mantle as small
and efficient O2 sinks are larger than the O2 supply, there are as ~0.5 log10 units have been suggested to have a significant effect
excess reductants, and the atmosphere remains anoxic even in the on atmospheric redox evolution32; therefore, prior results do not
presence of global oxygenic photosynthesis. Rapid sinks include preclude involvement in the GOE.
net reactions of O2 with reducing gases emanating from the solid A notable study of mantle redox evolution is Li and Lee40, who
Earth, e.g., H2, CO, H2S, SO2, and CH4. used the V/Sc ratios of a large database of primitive basalts to
One possibility is that an increase in O2 fluxes overwhelmed report that mantle oxidation state had not changed by more than
the efficient O2 sinks, causing the GOE. However, the available 0.3 log10 units since the Archean. The use of the V/Sc oxyba-
data are not particularly persuasive. Photosynthetic O2 produc- rometer on large amounts of published basalt data relies upon the
tion is accompanied by organic matter (CH2O) in the net reaction assumption that all of the studied basalts are the result of similar
CO2 + H2O = CH2O + O2. Because respiration or oxidative degrees of partial melting, and that they sample a spinel peridotite
decay reverse this reaction (presently in ~50 years), the long-term source with a primitive mantle V/Sc of ~5. None of these are
net flux of O2 occurs when organic matter is segregated from O2 necessarily safe assumptions. Average degree of partial melting
and buried. was likely higher in the Archean due to higher mantle potential
Long-term changes in the burial rate of organic matter—and temperature. In addition, not all of the studied basalts were likely
hence in the net supply of O2—can be inferred from C isotopes in generated from melting of spinel peridotite, and residual garnet
sedimentary rocks. Organic matter concentrates 12C relative to can have a potentially strong effect on V/Sc44. Finally, the mantle
13C, leaving marine inorganic carbonate 12C depleted; as such, sources of basalts vary greatly in the degree of previous melt
carbon isotopes in marine carbonates and kerogens track forg, the depletion, and remelting of a previously depleted source can
fraction of total carbon buried as organic matter. Isotopic mass generate a low V/Sc melt and an erroneously reduced fO2 . For all
balance shows little difference between the average organic burial these reasons, the conclusions and quoted uncertainties of Li and
fraction for 3.6–2.5 Ga of forg = 0.15 ± 0.02 [2 standard deviations Lee40 must come into question.
(SD)] and that for 1.8–1.0 Ga of forg = 0.18 ± 0.02 [2 SD]24, sug- Two new studies reveal an fO2 trend. Aulbach and Stagno45
gesting no significant increase in O2 flux. carefully filtered V/Sc data to only include Archean basalts
Of course, the inferred organic burial fraction and associated formed in a mid-ocean ridge (MOR)-like environment, while
O2 flux depend on assumptions about the carbon cycle. Chal- Nicklas et al.46 calculated the fO2 of ultramafic lavas directly
lenges to a conventional carbon mass balance model include 12C- without making assumptions about their source compositions.
enriched carbon sequestered into seafloor carbonates25 or into These new data, shown in Fig. 1a, indicate that mantle fO2
authigenic carbonates26, or isotopic weathering inputs that scaled increased by ~1.3 log10 units from the early Archean to Proter-
with the amount of atmospheric O224,27,28. The first two ozoic, and likely represent the current best estimate for mantle
hypotheses lack supporting evidence: Archean seafloor carbonate redox evolution.
is not isotopically lighter than typical marine sedimentary car- Both datasets in Fig. 1a show a similar fO2 trend, but each
bonates29, and the abundance of authigenic carbonates is low dataset was determined using a different oxybarometer. The fO2
before the GOE compared with afterward30, and is relatively derived from different oxybarometers shows a systematic offset,
small today31. The third idea is plausible: the lack of oxidative the reason for which is currently unclear, as oxybarometry per-
weathering in the Archean could modulate forg27. However, the formed using different methods on modern rocks gives values
average organic content of Archean sedimentary rocks (3.59 wt%) that vary outside of analytical uncertainty47. One possible
is indistinguishable from that of Proterozoic (3.56 wt%)24, and explanation involves the degassing of volatile species, such as SO2,
the cumulative distribution of total organic content in organic- which have the potential to strongly reduce a lava shortly prior to
rich Archean sedimentary rocks is identical to that of Neogene eruption48. Degassing may lead to an offset between these two
rocks1. datasets because the V-partitioning oxybarometer of Nicklas
Here we are interested in examining the potential implications et al.46 measures the fO2 of komatiites and picrites, while the V/Sc
of the trend in mantle redox state on the GOE. A lower Archean ratio oxybarometer of Aulbach and Stagno45 infers the fO2 of
O2 flux may be a possible factor, but that is not the foxus basalts and picrites. Komatiites are high-temperature, high-degree
of this paper.

2 NATURE COMMUNICATIONS | (2020)11:2774 | https://doi.org/10.1038/s41467-020-16493-1 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-16493-1 ARTICLE

a +2 In this study, we show that the new data imply that reducing
Archean volcanic gases would prevent atmospheric O2 from
accumulating, and then the GOE would occur by or after ~2.5 Ga
ΔFMQ [log10 bar]
+1 with ≥95% probability. Thus, we conclude that secular oxidation
of the mantle have indeed triggered oxygenation of the
atmosphere.
0
Results
–1 Oxygenation parameter. We evaluate whether the atmosphere
was prone to oxygenation at any given time using an oxygenation
Nicklas et al. (2019) parameter, Koxy22,33,34. This parameter is the ratio of O2 source
Aulbach and Stagno (2016)
–2 fluxes to kinetically efficient O2 sink fluxes, which we consider
0 1 2 3 4 here to be dominated by oxidizable volcanic gases
Time [Ga]
O2 source fluxes
b +1 K oxy
: ð1Þ
non-weathering O2 sink fluxes ðmay incl: excess reductantsÞ

Fluxes are quantified in units of T mol O2 yr−1. When Koxy < 1,

ΔFMQ [log10 bar]

0 gaseous volcanic O2 sinks exceed O2 sources, and excess H2 builds

up until balanced by escape to space. When Koxy > 1, O2 sources
exceed efficient O2 sinks, and O2 builds up until balanced by
oxidative weathering. Box modeling coupled to photochemistry
–1 shows that Koxy = 1 defines the point when the atmosphere
becomes oxic (see Fig. 7b of Claire et al.34). For detailed infor-
Nicklas et al. (2019) mation, see the “Methods” section and the Supplementary
–2
Aulbach and Stagno (2016) information.
0 1 2 3 4
Evolution of the oxygenation parameter. As an initial, illus-
Time [Ga]
trative trial, we took the organic burial fraction (forg) to be con-
Fig. 1 Evolution of the oxygen fugacity of mantle. The oxygen fugacity (fO2) stant at 20%, which is a canonical value adopted by previous
is in log10 units relative to the Fayalite–Magnetite–Quartz (FMQ) buffer. In researchers [e.g., refs. 21,32], to isolate the effect of the secular
a, we show original data of Aulbach and Stagno45 and Nicklas et al.46. change in mantle fO2 shown in Fig. 1b. As shown in Fig. 2a, the
Dotted and dashed lines in a represent a linear fit for the data of Aulbach results are that the oxygenation parameter (Koxy) monotonically
and Stagno45 and Nicklas et al.46, respectively, showing similar trends. Both increases with time.
datasets should converge on an average modern fO2 value inferred from In Fig. 2a, Koxy larger than unity occurs when the production of
mid-ocean ridge basalt (MORB). So, in b, we anchor the datasets to the O2 via organic burial and deposition of pyrite (i.e., the numerator
MORB-inferred fO2 of the modern mantle, 0.250. Thus, 0.2 is added to of Eq. (20)) exceeds the consumption of O2 from oxidizable
Aulbach and Stagno data45 and −1.00 to Nicklas et al. data46. The black volcanic gases (i.e., the denominator of Eq. (20)). Then
solid line and gray shaded region in b represent the median value and 95% atmospheric O2 accumulates, and oxidative weathering kicks in
confidence interval of oxygen fugacity, respectively. The error bar and balances the excess O2 production22,33,34. In contrast, when
represents uncertainty of 1σ. The gray shaded region corresponds to a Koxy is smaller than unity, the production of O2 via organic burial
variation of the slope of the linear fit, which is propagated from the and deposition of pyrite (i.e., the numerator of Eq. (20)) is smaller
variations of the samples (i.e., the error bars). Note that in b, the variation than the consumption of O2 by reducing volcanic gases (i.e.,
of the fO2 of the modern mantle, which is discussed later, is neglected. the denominator of Eq. (20)). Under such a condition, atmo-
spheric O2 cannot accumulate, and the buildup of excess of
reducing gases, such as CH4 and H2, is limited by their
partial melts that are undersaturated in sulfur, while basalts are decomposition in the upper atmosphere and the escape of
lower-degree, sulfur-saturated melts. Hence, sulfur degasses more hydrogen to space22,33,34.
from basaltic lavas than komatiitic lavas, i.e., decreases the fO2 of The 5% probability quantile of Koxy (the lower end of gray
basalts more than that of komatiite46. shaded region) crosses unity (a dotted gray line) at 2.62 Ga in the
As each mantle oxidation trend only compares samples ana- purely illustrative case of Fig. 2a. This means that at 2.62 Ga and
lyzed by the same oxybarometry method, and each study also later, the probability that Koxy is larger than unity, allowing
analyzed modern mid-ocean ridge basalt (MORB) with its atmospheric oxygenation, is 95% or more. Hereafter, we designate
respective method, we can anchor each trend to the current this 95% threshold as the oxic transition time.
canonical modern MORB value of +0.2 ± 0.3 log10 units above In Eq. (20), Koxy also depends on forg, which has fluctuated over
the fayalite–magnetite–quartz (FMQ) buffer as determined by X- time. In Fig. 2b, which we subsequently call our standard case, we
ray absorption near-edge spectroscopy49,50. Upon anchoring, the included temporal changes in forg from Krissansen-Totton et al.24
two mantle oxidation trends overlap within their respective derived from the carbon isotope record (Supplementary Fig. 1).
uncertainties, lending credence to the idea of mantle oxidation Unlike in Fig. 2a, Koxy fluctuates in Fig. 2b because of fluctuations
(Fig. 1b). in forg derived from the carbon isotope record. However, Koxy still
Although it has been speculated that the secular trends in increases with time. Here, the oxic transition time of this standard
the mantle fO2 in Fig. 1a and the timing of the GOE may be case is 2.48 Ga, which is delayed because of relatively low values
related45,46, this hypothesis has never been quantitatively tested in of forg before 2.5 Ga (Supplementary Fig. 1). Hence, even when we
a redox model for the surface environment using data-derived include temporal changes of forg, the atmosphere still becomes
trends of forg and mantle fO2 . oxic at 2.48 Ga (or afterward) with 95% (or more) probability.

NATURE COMMUNICATIONS | (2020)11:2774 | https://doi.org/10.1038/s41467-020-16493-1 | www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-16493-1

a For comparison, we also did calculations for three hypothetical

10 scenarios where the mantle fO2 was fixed at FMQ-2, FMQ-1, and

Oxygenation parameter (K oxy)

95% prob. quantile
Median FMQ (Fig. 2c). We also included temporal fluctuations in forg, as
5% prob. quantile in Fig. 2b, which are reflected in the fluctuations of Koxy. In these

Oxic
scenarios, an oxic transition for Koxy does not occur, although
Koxy occasionally becomes greater than 1 in the FMQ-1 case
1 within the gray shaded region of uncertainty in Fig. 2c. Thus, the
transitions from Koxy < 1 to Koxy > 1 in Fig. 2a, b are mainly due to
the temporal increase in mantle fO2 taken from the fit in Fig. 1.

Anoxic
Discussion
0.1 As shown above, an empirically-inferred increase in mantle fO2
0 1 2 3 4 causes an increase in Koxy, resulting in the shift of the atmosphere
Time [Ga] from reducing to oxic (Fig. 2). However, this result does not
exclude a role for other processes in the oxidation of the atmo-
b sphere as explained below.
10 The result of the standard case shows that uncertainties in the
Oxygenation parameter (K oxy)

95% prob. quantile

Median fO2 time series and the oxygen flux estimates from the carbon
5% prob. quantile Oxic isotope record propagate through the calculations, so that the
uncertainty envelope allows an oxic regime even before ~2.5 Ga
(Fig. 2b). On the other hand, geological evidence, such as mass-
1 independent sulfur isotope fractionation1, indicates an anoxic
atmosphere for the Archean Earth before 2.4–2.3 Ga though a
recent study suggests that the GOE occured somewhat ear-
Anoxic

lier than 2.4 Ga51. This discrepancy in the timing of the GOE
might also be caused by not considering other processes that may
0.1 delay the oxic transition.
0 1 2 3 4 To investigate the effect of processes proposed by previous
Time [Ga] studies20–22,27,52,53, we did some sensitivity studies and obtained
the same trends as previous studies. As explained in Supple-
c mentary Note 1, Koxy decreases if the degassing pressure is high,
10 if the carbon and/or sulfur degassing fluxes are low, and/or if the
Oxygenation parameter (K oxy)

FMQ rate of magnetite deposition via serpentinization is high, which

Oxic

generates oxidizable hydrogen that is a sink for O2.

If Koxy was lower, the atmosphere in the early Archean would
1 be more reducing, and the oxic transition time would be delayed,
because of the following processes: a secular decrease in the
FMQ-1 degassing pressure due to a transition from submarine to sub-
Anoxic

aerial volcanism [e.g., refs. 20,52], a secular increase in the carbon

0.1 and/or sulfur degassing due to an increase in their continent and/
FMQ-2 or ocean floor reservoir21, and/or a secular decrease in the
magnetite deposition flux via serpentinization, which might result
from a decrease in the degree of partial melting of the mantle
0 1 2 3 4
caused by secular cooling22. The magnitude of the Archean ser-
Time [Ga]
pentinization flux of H2 is debated, since it is only significant
Fig. 2 Evolution of the dimensionless oxygenation parameter, Koxy. Solid today from slow-spreading centers where ultramafic rocks are
lines represent median values, and the shaded region bounds 5% to 95% exposed53.
probability quantiles. These are obtained by 10,000 times Monte-Carlo There is another possible process that contributed to the delay
simulations. Gray dotted lines highlight Koxy = 1, above which the in th oxic transition along with slow mantle oxidation. Because of
atmosphere is oxic; otherwise, it is anoxic. a Organic burial fraction is the lack of oxidative weathering in the Archean, the carbon iso-
constant at 20%, and the secular increase in the oxygen fugacity of mantle tope input into the atmosphere-ocean could have been relatively
in Fig. 1b is considered. b Both a change in organic burial fraction from heavy compared with mantle values, and so less organic burial
Krissansen-Totton et al.24 and secular increase in the oxygen fugacity in was needed for the mass balance27.
Fig. 1b are considered. c Oxygen fugacity is assumed to be constant at three We also investigated the uncertainty of the anchoring value of
different levels (blue: FMQ-2, black: FMQ-1, orange: FMQ). The fluctuations the mantle fO2evolution (Supplementary Note 1). For the standard
come from imposed changes in organic burial fraction. In b, the parameter case (Fig. 2), we modeled the evolution of the mantle fO2
Koxy exceeds unity by ~2.5 Ga with >= 95% probability. anchoring the trend to the average fO2 of modern MORB, i.e.,
ΔFMQ0 = +0.2 [see also Eq. (4)]50. However, this reference value
On the other hand, the probability of Koxy > 1 is less than 50% has an uncertainty of 0.3 log10 units50. Anchoring the fO2 trend to
before 3.38 Ga, for example, 30% at 3.6 Ga (Fig. 2b, see also a lower ΔFMQ0 value causes lower mantle fO2in the past, resulting
Supplementary Fig. 2). This indicates that the atmosphere was in a more reducing atmosphere in the Archean and delaying the
likely reducing in the early Archean despite the possible pre- oxic transition time (Supplementary Fig. 8).
sence of oxygenic photosynthesis. Thus, the increase in fO2 of the Thus, all of the processes discussed in the preceding three
mantle would cause the atmosphere to shift from reducing to paragraphs might affect the GOEʼs timing and so are potentially
oxic, and the shift seemed to occur before ~2.5 Ga. complicationg factors.

4 NATURE COMMUNICATIONS | (2020)11:2774 | https://doi.org/10.1038/s41467-020-16493-1 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-16493-1 ARTICLE

Observations show that erupted volcanic gas is more oxidized, It is also noteworthy that the similar trend of the man-
i.e., has a larger ΔFMQ, than its source melt54,55. The relative tle fO2 with time is observed by the measurement of fO2 of different
oxidative state of the volcanic gas results from reactions within a rocks46,59. This similarity indicates that the increase in fO2 of melts
closed gas mixture due to cooling54,55. Consequently, it has been follows the same trend of the decreasing mantle fO2 with
proposed that secular cooling of the mantle could have facilitated time, despite the difference in the degree of partial melting, which
the GOE55. Also, observations seem to indicate that it is also supports our assumption that the fO2 of the volcanic gas
not possible to calculate the oxidation effect, i.e., Koxy, of the increased with time.
volcanic gas using the ΔFMQ of its source, although we did this The assumption of the anchoring value of the fO2 bears con-
in this study. However, the assumptions of these previous sideration. In this study, we anchored the fO2 of the volcanic gas
studies54,55 need to be reconsidered, as follows. and the melt at the value of the modern MORB, implicitly
Recent work56 discusses how cooling affects the oxidation state assuming that the fO2 of the melt does not change after the melt
and the Koxy of a volcanic gas mixture considering two stages of decouples from the gas. However, if the fO2 of the melt chan-
degassing, i.e., the stage where volcanic gas is buffered by its
ges after the decoupling from the gas, it also changes the timing of
source melt (melt-buffered stage), and the stage where the vol-
the GOE as indicated by the parameter study of the anchoring
canic gas is a closed system (closed stage). For the closed stage,
valur of the fO2 (Supplementary Fig. 8).
the same conclusion was reached as in the previous studies54,55,
Of course, a major question is what drove the increase in
i.e., cooling increases ΔFMQ of a closed volcanic gas compared
mantle fO2 and, hence, could have driven the GOE. One possi-
with that of its source melt. However, in a closed gas mixture,
bility is that convection-driven homogenization of an initially
oxidation of a gas should be accompanied by a reduction of
redox-stratified primordial mantle was responsible for this
another gas, so any reaction in the closed gas mixture does not
change45,46,62. The basic idea is that in the early deep mantle,
change the overall O2 sink in the gas mixture56. Hence, to eval-
Fe2+ disproportionated to Fe3+ and Fe metal due to high pres-
uate the Koxy of volcanic gas, we can neglect the effect of reactions
sure. The latter was lost to the core, leaving a more oxidized lower
after the volcanic gas separated from its source, i.e., the observed
mantle below a relatively reduced upper mantle. According to
difference in the oxidation state between erupted volcanic gas and
Andrault et al.62, the primitive mantle contained excess of Fe3+
its source melt.
corresponding to ~60% of an ocean’s worth of oxygen.
For the melt-buffered stage, cooling results in reduction of a gas
However, another possible driver of upper mantle oxidation
mixture and a decrease in Koxy if the ΔFMQ of the source melt
has recently gained evidence in its favor. Ancient air dissolved in
is buffered and constant56. This is because cooling decreases the
inclusions of seawater in Archean quartz shows that the nine
absolute value of fO2 of the FMQ buffer (Supplementary Fig. 4). A
isotopes of Xe become increasingly isotopically heavy throughout
trend of smaller Koxy with lower mantle temperature is shown in
the Archean and early Proterozoic until the GOE63. The most
Supplementary Fig. 3a. However, the oxic transition time is
plausible explanation is a very rapid escape of hydrogen to space
insensitive to temperature if the temperature is higher than the
that dragged along ionized Xe atoms, which would have fractio-
solidus temperature of dry peridotite (~1390 K)57, as shown in
nated Xe isotopes because of mass-dependent escape64.
Supplementary Fig. 3b. The mantle temperature in the Archean
Substantial loss of a strong reducing agent, i.e., hydrogen,
would be between 1600 and 1900 K [e.g., refs. 57–59]. Hence, the
would have oxidized the Earth, with the oxidation affecting the
secular cooling of the mantle would affect less the GOE.
reservoir from which the hydrogen originated. Thus, the mantle
Note that Holland’s f number [e.g., ref. 32] was used in
would become gradually more oxidized because the hydrogen
Moussallam et al.48, while Koxy was used in Kadoya et al.56.
comes from the decomposition of water in volcanic melts, sche-
However, the temperature dependence of Holland’s f number is
matically represented as 3FeO + H2O → Fe3O4 + H2. The upper
essentially the same as that of Koxy. This point is explained in
mantle (down to ~660-km depth) contains the equivalent of
Supplementary Note 2.
~20% of an ocean’s worth of oxygen as Fe3+ [ref. 65, p. 207]. The
A decrease in mantle temperature might delay the oxic tran-
Xe isotope data require Archean hydrogen loss from the
sition time if the temperature was lower than the solidus tem-
equivalent of ~10 s of percent of an ocean64. Thus, if the fO2 trend
perature of dry peridotite (Supplementary Fig. 3b). Degassing
in Fig. 1b explains the GOE, as shown in Fig. 2b, the fO2 trend, in
would occur under such a low temperature in arc volcanism
because the hydrous phase of the subducted crust can lower the turn, may have been driven by hydrogen escaping to space from
melting temperature [e.g., ref. 60]. However, the island arc basalts the Earth’s pre-GOE anoxic atmosphere [e.g., refs. 19,66].
are often more oxidized than the mid-ocean ridge basalts or Hydrogen escape as a mechanism for mantle oxidation36 has
oceanic island basalts61. Hence, we cannot conclude that arc been proposed previously, but has been rejected over the last two
volcanism, whose degassing temperature would be low, delayed decades because of reports of seemingly constant mantle fO2
the oxic transition time. through time39,67,68. Our results suggest that this mechanism may
A question might arise about our assumption that the fO2 of the need to be reconsidered.
volcanic gas was equal to that of the upper mantle. Degassing can In summary, we examined whether new data for increasing
reduce the source melt, e.g., SO2 degassing48. However, during mantle oxygen fugacity (fO2 ) since the Archean could explain the
the melt-buffer stage, i.e., when the volcanic gas mixture interacts GOE, when O2 first accumulated in the Earth’s atmosphere. The
with the ambient melt, the fO2of the gas will be equal to that of the onset of the GOE can only be properly quantified by considering
melt. In addition, during the closed stage, i.e., after the gas sources and sinks of oxygen in a global redox balance of the
mixture decouples from the ambient melt, any reaction within the surface environment. The oxygenation parameter Koxy used for
gas mixture does not change the oxygenation effect of the gas56. this purpose is defined as the ratio of O2 sources to kinetically
Thus, the oxygenation effect of the gas should be examined using rapid sinks. For an anoxic atmosphere, Koxy < 1, while for an oxic
the fO2 of the gas when the gas decouples from the melt. Also, at atmosphere, Koxy > 1; by evaluating when Koxy = 1, we deter-
that time, the fO2 of the gas would equal to that of the ambient mined how mantle fO2 trends affected the onset of the GOE.
A more reducing mantle with low fO2 produces a greater
melt, which has already experienced degassing. Therefore, we can
proportion of reducing volcanic gases. So, we found that the data-
calculate Koxy using the fO2 from the Archean that is anchored
derived trend of mantle fO2 likely prevented O2 building up in the
today to modern MORB fO2.
atmosphere with relatively high probability (~70% at 3.6 Ga) and

NATURE COMMUNICATIONS | (2020)11:2774 | https://doi.org/10.1038/s41467-020-16493-1 | www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-16493-1

caused an oxic transition from Koxy < 1 to Koxy > 1 with >95% We take H2O and CO2 to be redox-neutral, so hydrogen (H2), carbon monoxide
probability after 2.5 Ga. (CO), and methane (CH4) are sinks of oxygen, as follows:
Our calculated timing of the GOE is relatively insensitive to the 1
H2 þ O2 ! H2 O; ð6Þ
mantle potential temperature, but depends on the assumed 2
degassing pressure, and total outgassing fluxes of carbon and 1
sulfur relative to total hydrogen. An additional oxidative sink of CO þ O2 ! CO2 ; ð7Þ
2
ferric iron in magnetite deposition has a minor effect, unless this
flux in the Archean exceeded ten times the modern flux. CH4 þ 2O2 ! CO2 þ 2H2 O: ð8Þ
If a trend in mantle fO2 controlled the timing of the GOE, then Hence, one mole of H2 degassing corresponds to 0.5 mole of O2 consumption.
the cause of mantle oxidation is ultimately important for setting Similarly, one mole of CO and CH4 degassing corresponds to 0.5 and 2 moles of O2
the tempo of biological evolution because macroscopic, energy- consumption, respectively.
On the other hand, the burial of organic matter (CH2O) is a source of oxygen,
intensive aerobic life was impossible when O2 levels were negli- as follows:
gible. Possible drivers of mantle redox evolution are the mixing of
CO2 þ H2 O ! CH2 O þ O2 : ð9Þ
a redox heterogeneous mantle or the time-integrated oxidative
effect of the breakdown of mantle water in volcanism, and the Hence, one mole of the burial of organic matter corresponds to one mole of O2
escape of hydrogen to space. Such processes could also apply to production.
We also assume that SO2 is redox-neutral, so H2S is a sink of oxygen, as follows:
other Earth-like planets elsewhere, and would thus determine
whether such planets could be habitats for complex aerobic life 3
H2 S þ O2 ! SO2 þ H2 O: ð10Þ
with high O2 demand69. 2
Also, reduction of SO2 must work as a source of oxygen, as follows:

Methods 1 1 5
SO2 þ FeO ! FeS2 þ O2 : ð11Þ
The redox tipping point of the atmosphere can only be quantified by considering 2 2 4
the global redox flux balance of the early atmosphere and ocean, which is as Hence, one mole of SO2 degassing corresponds to 1.25 mole of O2 production. On
fundamental as mass or energy conservation22,65, p. 221–223]. the other hand, one mole of H2S degassing corresponds to 0.25 mole of O2
To evaluate the redox tipping point of the atmosphere, we use the oxygenation consumption.
parameter, Koxy, which is the ratio of the O2 source fluxes (Foxi) to non-weathering As explained above, degassing of H2, CO, CH4, and H2S is a sink of oxygen.
O2 sink fluxes (Fred)22,34,69: Accordingly, Fred is expressed as follows:
1 1 1
K oxy

F oxi
: F red ¼ F H2 þ F CO þ 2F CH4 þ F H2 S : ð12Þ
ð2Þ 2 2 4
F red
On the other hand, burial of organic matter and SO2 degassing are sources of
The fluxes, Foxi and Fred, are calculated using fluxes of volcanic gas, such as CO and oxygen. Accordingly, Foxi is expressed as follows:
SO2, organic burial flux, and pyrite burial flux. In the subsequent sections, we will 5
describe a model of each flux, and then explain models of Foxi and Fred. F oxi ¼ F org þ F SO2 : ð13Þ
4

Fluxes related to source and sink of oxygen. In this section, we explain the Derivation of oxygenation parameter. Substituting Eqs. (12) and (13) into Eq.
components, which are used to calculate the O2 source (Foxi) and a kinetically rapid (2), Koxy is rewritten as follows:
sink of O2 (Fred). For the calculation of Foxi and Fred, see the next section.
Hydrogen is degassed to the ocean–atmosphere system as hydrogen molecules 4F org þ 5F SO2
K oxy ¼ : ð14Þ
(H2), water vapor (H2O), methane (CH4), and hydrogen sulfide (H2S). 2F H2 þ 2F CO þ 8F CH4 þ F H2 S
Accordingly, a total flux of hydrogen (Fhydrogen) can be expressed as follows:
We now define χc and χs as follows:
F hydrogen ¼ F H2 þ F H2 O þ 2F CH4 þ F H2 S ; ð3Þ F carbon F
χc
; χ
sulfur : ð15Þ
F hydrogen s F hydrogen
where Fx is a flux of x. Methane contains the equivalent of two H2 molecules, so
F CH4 is weighted by a factor of 2. In addition, we define r H2 as the ratio of a flux of hydrogen molecule to a total flux
Carbon is degassed to the ocean–atmosphere system as carbon dioxide (CO2), of hydrogen:
carbon monoxide (CO), and methane (CH4), and deposited as organic matter (org) F H2
and carbonate (carb). Since we assume that carbon is in a steady state, the total flux r H2
: ð16Þ
F hydrogen
of carbon (Fcarbon) is equal to input and output fluxes, which can be expressed as
follows: Similarly, we defined the following parameters:
F CO F CH4 F org
F carbon
F CO2 þ F CO þ F CH4 ¼ F org þ F carb ; ð4Þ r CO
;r
;f
; ð17Þ
F carbon CH4 F carbon org F carbon
where Fx is a flux of x as in Eq. (3).
F SO2 F H2 S
The ratio of the organic burial to the total carbon flux (i.e., forg = Fcarb/Fcarbon) r SO2
;r
: ð18Þ
can be evaluated using the geological record of carbon isotopes in organic and F sulfur H2 S F sulfur
inorganic carbon [e.g., ref. 24]. Supplementary Fig. 1 shows the temporal change in If we substitute from the above for the various gas fluxes and flux of organic burial,
the forg, which is derived by Krissansen-Totton et al.24. Eq. (14) can be rewritten as
Similarly, sulfur (S) is degassed to the ocean–atmosphere system as sulfur
dioxide (SO2) and hydrogen sulfide (H2S), and deposited as pyrite (FeS2) and 4f org F carbon þ 5r SO2 F sulfur
K oxy ¼
 : ð19Þ
sulfate (e.g., CaSO4). However, we neglect the deposition of sulfate because we
2r H2 F hydrogen þ 2r CO þ 8r CH4 F carbon þ r H2 S F sulfur
focus on the Archean eon, the surface environment was anoxic, and there was very
little sulfate (~2.5 μM, i.e., 0.01% of modern level) in the Archean ocean70,71. Since If we now divide the above equation by Fhydrogen, we arrive at the form
we assume that sulfur is in steady state, a total flux of sulfur (Fsulfur) is equal to
input and output fluxes, which can be expressed as follows: 4f χ þ 5r SO2 χ s
K oxy ¼
org c  : ð20Þ
2r H2 þ 2r CO þ 8r CH4 χ c þ r H2 S χ s
1
F sulfur
F SO2 þ F H2 S ¼ F FeS2 ; ð5Þ
2 Kasting22 argues that an additional O2 flux, in the form of F Fe3 O4 , could be
included in the denominator of Eq. (2). If we take FeO as the redox reference state
where Fx is a flux of x as in Eq. (3). of iron for the Archean surface environment, then the formation of ferric iron in
magnetite (Fe3þ 2þ
2 Fe O4 ) in iron formations or during serpentinization and its
Sources and sinks of oxygen. In this section, we explain models of the O2 source burial removes oxidizing power from the surface environment or, equivalently, is
(Foxi) and a kinetically rapid sink of O2 (Fred), which are used for the calculation of an input flux of reducing power. We initially neglect the deposition of magnetite,
Koxy (Eq. (2)). and then examine its specific influence later. When the magnetite deposition is

6 NATURE COMMUNICATIONS | (2020)11:2774 | https://doi.org/10.1038/s41467-020-16493-1 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-16493-1 ARTICLE

taken into account, Koxy can be rewritten as

4f org χ c þ 5r SO2 χ s Cr 2H2 O Dr H2 O
K oxy ¼
 ; ð21Þ ð1 þ AÞr H2 O þ 2χ c þ χs ¼ 1: ð31Þ
2r H2
F
þ 2r CO þ 8r CH4 χ c þ r H2 S χ s þ F Fe3 O4 ð1 þ BÞr 2tot þ Cr 2H2 O r tot þ Dr H2 O
hydrogen

where F Fe3 O4 is an O2 consumption via magnetite deposition in the unit of T mol

O2 yr−1. For example, the modern F Fe3 O4 is 0.05 ∼ 0.2 T mol O2 yr−1 72,73. Oxygen fugacity of mantle. As explained above, we need the oxygen fugacity, fO2 ,
of volcanic gas to calculate the volcanic gas speciation. To evaluate the Koxy of the
gas, we can use the fO2 of the source of the gas, i.e., the mantle melt56. In addition,
Equilibrium state of volcanic gases. In this section, we will explain how to both experimental75 and empirical data from natural samples76 show that the
calculate the flux ratios, i.e., rx for each volatile species x in Eq. (20). mantle melt records the same fO2 as its mantle residue, and therefore its mantle
To calculate fluxes of volcanic gases, we first assume that equilibrium states are
source. Hence, we evaluate the Koxy using the fO2 of the mantle.
achieved for volcanic volatiles in the silicate melt:
8 According to Aulbach and Stagno45 and Nicklas et al.46, the oxygen fugacity of
> H2 O ¼ H2 þ 12 O2 the mantle has increased by +0.9 ± 0.2 [2 SD] in log10 units during the past 3 Gyr.
>
>
< CO2 ¼ CO þ 12 O2 Since two independent datasets show the same trend, secular oxidation of the
: ð22Þ mantle is corroborated. In addition, the oxygen fugacity of the modern mantle is
>
> CO2 þ 2H2 O ¼ CH4 þ 2O2
>
: +0.2 ± 0.6 [2 SD] in log10 units above the FMQ buffer50. Here, FMQ is the
SO2 þ H2 O ¼ H2 S þ 2 O2
3
fayalite–magnetite–quartz synthetic buffer, which defines fO2 at a particular
These equations have equilibrium constants that are as follows, in terms of temperature and pressure, e.g., fO2 = 10−8.5 bar at 1200 °C (1473.15 K and 0.5
fugacities (fx) for each volatile species x: MPa). Hence, we modeled the evolution of the mantle fO2 in log10 units above the
8 f f 0:5
FMQ buffer as follows:
>
> K 1 ¼ Hf 2 O2
>
> ΔFMQ ¼ ΔFMQ0 þ at: ð32Þ
>
>
H2 O
>
> 0:5
>
< K 2 ¼ CO
f fO 2
where t is time in units of Ga, and the slope, a, is −0.29 ± 0.05 [2 SD] in units of per
f CO2
: ð23Þ Ga. For the standard case, we set ΔFMQ0 at 0.2 log10 units50, and later discuss the
>
> f f2
>
> K 3 ¼ f CH4f 2O2 effect of the variation of this parameter.
>
>
>
CO 2 H 2 O
>
> 1:5 Figure 1b shows the evolutionary range of oxygen fugacity for the standard
: K ¼ f H2 S f O2 case with its uncertainties. Further details for the comparison of the datasets of
4 f f SO2 H2 O
Fig. 1b are given in Supplementary Note 3.
We assume that gas fluxes will be in proportion to their fugacities, e.g.,
f H2 ¼ Ptot ´ F H2 =F tot . Here, Ptot is a total pressure under which degassing occurs. Other input parameters. As indicated in Eqs. (30) and (31), the flux ratios, rx in
Ftot is a total degassing flux that can be written as Eq. (20), depend on the total flux ratios of carbon and sulfur relative to hydrogen,
χc and χs. In addition, the flux ratios also depend on the temperature and pressure
F tot  F H2 O þ F H2 þ F carbon þ F sulfur ; ð24Þ
of the system.

 Initially, we assume modern values for the total flux ratios of carbon and sulfur
¼ F hydrogen r H2 O þ r H2 þ χ c þ χ s : ð25Þ relative to hydrogen. The modern total degassing fluxes of hydrogen, carbon, and
sulfur (in all their forms) are 97 ± 20, 9 ± 2, and 2.2 ± 0.7 T mol yr−1, respectively
Then, we obtain the following equations: [ref. 65, p. 203 and p. 221]. Hence, present-day χc = 0.1 ± 0.03 and χs = 0.023 ±
8 K1 rH2 0.0086. Later, we examine the effect of these parameters on the results of our
>
> 0:5 ¼ r
>
>
fO H2 O modeling.
>
>
2

>
> K2
¼ r CO Similarly, we use a modern mantle potential temperature of 1623.15 K (1350 °C)77
< 0:5 rCO2
fO 2 and a pressure of 0.5 MPa78 first. We examine the sensitivity to these parameters later.

 :
rCH4 r H2 O þrH2 þχ c þχ s 2
ð26Þ
>
> K3 2
> f 2O2 Ptot ¼ rCO2
>
> rH2 O
>
> K Data availability
>
: 1:54 Ptot ¼ rH2 S ´ rH2 O þrH2 þχc þχ s
f O2
r r
SO2 H2 O The datasets generated during the current study are available in the Zenodo repository
[https://doi.org/10.5281/zenodo.3668382].
where fO2 is a oxygen fugacity, which we will explain later. We calculate
equilibrium constants of the above reactions using data of NIST74. According to
Eqs. (3)–(5), we also obtain the following equations: Code availability
8 The source code used in this study is also available in the Zenodo repository [https://doi.
< r H2 þ r H2O þ 2r CH4 χ c þ r H2 S χ s ¼ 1
> org/10.5281/zenodo.3668382].
r CO2 þ r CO þ r CH4 ¼ 1 : ð27Þ
>
: r þr ¼1
SO2 H2 S Received: 15 October 2019; Accepted: 6 May 2020;
Hence, given mantle temperature, a total pressure, and oxygen fugacity in the
mantle, and solving Eqs. (26) and (27), we obtain fluxes of each molecular species.
Hereafter, we explain how to solve Eqs. (26) and (27). First of all, we defined
variables as follows:
K1 K K K4
A
; B
0:52 ; C
3 P2tot ; D
P ; References
f 0:5
O2 f O2
f O 2 f O 1:5 tot
2 2
1. Lyons, T. W., Reinhard, C. T. & Planavsky, N. J. The rise of oxygen in Earth’s
early ocean and atmosphere. Nature 506, 307–315 (2014).
r tot
ð1 þ AÞr H2 O þ χ c þ χ s : 2. Planavsky, N. J. et al. Evidence for oxygenic photosynthesis half a billion years
Then, Eq. (27) can be rewritten as follows: before the Great Oxidation Event. Nat. Geosci. 7, 283–286 (2014).
8
r 2
r  3. Satkoski, A. M., Beukes, N. J., Li, W. Q., Beard, B. L. & Johnson, C. M. A
>
> ð1 þ AÞr H2 O þ 2χ c C rH2 O r CO2 þ χ s D rH2 O r SO2 ¼ 1 redox-stratified ocean 3.2 billion years ago. Earth Planet. Sci. Lett. 430, 43–53
>
>
>
<
tot tot

r 2 (2015).
r þ Br þ C H2 O
r ¼ 1 : ð28Þ 4. Kasting, J.F., Schopf, J.W. & Klein, C. Models relating to Proterozoic
>
>
CO 2 CO 2 rtot CO 2
>
>
r  atmospheric and oceanic chemistry. in The Proterozoic Biosphere: A
>
: r SO2 þ D r r SO2 ¼ 1
H 2 O
Multidisciplinary Study (eds. Schopf, J. W. & Klein, C.) 1185–1187
tot
(Cambridge University Press, Cambridge, 1992).
Hence, 5. Olson, S. L., Kump, L. R. & Kasting, J. F. Quantifying the areal extent and
r 2tot dissolved oxygen concentrations of Archean oxygen oases. Chem. Geol. 362,
r CO2 ¼ ; ð29Þ 35–43 (2013).
ð1 þ BÞr 2tot þ Cr 2H2 O
6. Godfrey, L. V. & Falkowski, P. G. The cycling and redox state of nitrogen in
r tot the Archaean ocean. Nat. Geosci. 2, 725–729 (2009).
r SO2 ¼ : ð30Þ 7. Garvin, J., Buick, R., Anbar, A. D., Arnold, G. L. & Kaufman, A. J. Isotopic
r tot þ Dr H2 O
evidence for an aerobic nitrogen cycle in the latest Archean. Science 323,
The unknown variable, f H2 O , is obtained by solving the following equation: 1045–1048 (2009).

NATURE COMMUNICATIONS | (2020)11:2774 | https://doi.org/10.1038/s41467-020-16493-1 | www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-16493-1

8. Stüeken, E. E. & Buick, R. Environmental control on microbial diversification 40. Li, Z.-X. A. & Lee, C.-T. A. The constancy of upper mantle fO2 through time
and methane production in the Mesoarchean. Precambrian Res. 304, 64–72 inferred from V/Sc ratios in basalts. Earth Planet. Sci. Lett. 228, 483–493
(2018). (2004).
9. Schirrmeister, B. E., Gugger, M. & Donoghue, P. C. J. Cyanobacteria and the 41. Trail, D., Watson, E. B. & Tailby, N. D. The oxidation state of Hadean
Great Oxidation Event: evidence from genes and fossils. Palaeontology 58, magmas and implications for early Earthas atmosphere. Nature 480, 79–82
769–785 (2015). (2011).
10. Kaufman, A. J. et al. Late Archean biospheric oxygenation and atmospheric 42. Rollinson, H., Adetunji, J., Lenaz, D. & Szilas, K. Archaean chromitites show
evolution. Science 317, 1900–1903 (2007). constant Fe3+/Fe in Earth’s asthenospheric mantle since 3.8 Ga. Lithos 282,
11. Anbar, A. D. et al. A whiff of oxygen before the Great Oxidation Event? 316–325 (2017).
Science 317, 1903–1906 (2007). 43. Canil, D. & Fedortchouk, Y. Olivine-liquid partitioning of vanadium and
12. Reinhard, C. T., Raiswell, R., Scott, C., Anbar, A. D. & Lyons, T. W. A late other trace elements, with applications to modern and ancient picrites. Can.
Archean sulfidic sea stimulated by early oxidative weathering of the Mineral. 39, 319–330 (2001).
continents. Science 326, 713–716 (2009). 44. Wang, J. et al. Oxidation state of arc mantle revealed by partitioning of V, Sc,
13. Duan, Y. et al. Molybdenum isotope evidence for mild environmental and Ti between mantle minerals and basaltic melts. J. Geophys. Res. 124,
oxygenation before the Great Oxidation Event. Geochim. Cosmochim. Acta 74, 4617–4638 (2019).
6655–6668 (2010). 45. Aulbach, S. & Stagno, V. Evidence for a reducing Archean ambient mantle
14. Kendall, B., Brennecka, G. A., Weyer, S. & Anbar, A. D. Uranium isotope and its effects on the carbon cycle. Geology 44, 751–754 (2016).
fractionation suggests oxidative uranium mobilization at 2.50 Ga. Chem. Geol. 46. Nicklas, R. W. et al. Secular mantle oxidation across the Archean-Proterozoic
362, 105–114 (2013). boundary: Evidence from V partitioning in komatiites and picrites. Geochim.
15. Kendall, B., Creaser, R. A., Reinhard, C. T., Lyons, T. W. & Anbar, A. D. Cosmochim. Acta 250, 49–75 (2019).
Transient episodes of mild environmental oxygenation and oxidative 47. Mallmann, G. & O’Neill, H. S. Calibration of an empirical thermometer and
continental weathering during the late Archean. Sci. Adv. 1, e1500777 (2015). oxybarometer based on the partitioning of Sc, Y and V between olivine and
16. Stüeken, E. E., Buick, R. & Anbar, A. D. Selenium isotopes support free O-2 in silicate melt. J. Petrol. 54, 933–949 (2013).
the latest Archean. Geology 43, 259–262 (2015). 48. Moussallam, Y. et al. The impact of degassing on the oxidation state of basaltic
17. Kendall, B. et al. Pervasive oxygenation along late Archaean ocean margins. magmas: a case study of Kilauea volcano. Earth Planet Sci. Lett. 450, 317–325
Nat. Geosci. 3, 647–652 (2010). (2016).
18. Ostrander, C. M. et al. Fully oxygenated water columns over continental 49. Berry, A. J., Stewart, G. A., O’Neill, H. S. C., Mallmann, G. & Mosselmans, J. F.
shelves before the Great Oxidation Event. Nat. Geosci. 12, 186–191 (2019). W. A re-assessment of the oxidation state of iron in MORB glasses. Earth
19. Catling, D. C., Zahnle, K. J. & McKay, C. P. Biogenic methane, hydrogen Planet. Sci. Lett. 483, 114–123 (2018).
escape, and the irreversible oxidation of early Earth. Science 293, 839–843 50. O’Neill, H. S. C., Berry, A. J. & Mallmann, G. The oxidation state of iron in
(2001). mid-ocean ridge basaltic (MORB) glasses: implications for their petrogenesis
20. Kump, L. R. & Barley, M. E. Increased subaerial volcanism and the rise and oxygen fugacities. Earth Planet. Sci. Lett. 504, 152–162 (2018).
of atmospheric oxygen 2.5 billion years ago. Nature 448, 1033–1036 51. Philippot, P. et al. Globally asynchronous sulphur isotope signals require re-
(2007). definition of the Great Oxidation Event. Nat. Commun. 9, 1–10 (2018).
21. Holland, H. D. Why the atmosphere became oxygenated: a proposal. Geochim. 52. Gaillard, F., Scaillet, B. & Arndt, N. T. Atmospheric oxygenation caused by a
Cosmochim. Acta 73, 5241–5255 (2009). change in volcanic degassing pressure. Nature 478, 229–233 (2011).
22. Kasting, J. F. What caused the rise of atmospheric O2? Chem. Geol. 362, 13–25 53. Keir, R. S. A note on the fluxes of abiogenic methane and hydrogen from mid-
(2013). ocean ridges. Geophy. Res. Lett. 37, L24609 (2010).
23. Lee, C. T. A. et al. Two-step rise of atmospheric oxygen linked to the growth of 54. Oppenheimer, C. et al. Influence of eruptive style on volcanic gas emission
continents. Nat. Geosci. 9, 417–424 (2016). chemistry and temperature. Nat. Geosci. 11, 678–681 (2018).
24. Krissansen-Totton, J., Buick, R. & Catling, D. C. A statistical analysis of the 55. Moussallam, Y., Oppenheimer, C. & Scaillet, B. On the relationship between
carbon isotope record from the Archean to Phanerozoic and implications for oxidation state and temperature of volcanic gas emissions. Earth Planet. Sci.
the rise of oxygen. Am. J. Sci. 315, 275–316 (2015). Lett. 520, 260–267 (2019).
25. Bjerrum, C. J. & Canfield, D. E. New insights into the burial history of organic 56. Kadoya, S., Catling, D. C., Nicklas, R. W., Puchtel, I. S. & Anbar, A. D. Mantle
carbon on the early Earth. Geochem. Geophys. Geosyst. 5, Q08001 (2004). cooling causes more reducing volcanic gases and gradual reduction of the
26. Schrag, D. P. Authigenic carbonate and the history of the global carbon cycle. atmosphere. Geochem. Perspect. Lett. 13, 25–29 (2020).
Science 339, 1383–1383 (2013). 57. Hirschmann, M. M. Mantle solidus: experimental constraints and the effects of
27. Derry, L.A. Organic carbon cycling and the lithosphere. in Treatise on peridotite composition. Geochem. Geophys. Geosyst. 1, 2000GC000070 (2000).
Geochemistry (Second Edition) (eds. Holland, H. D. & Turekian, K. K.) 58. Nisbet, E. G., Cheadle, M. J., Arndt, N. T. & Bickle, M. Constraining the
239–249 (Elsevier, Oxford, 2014). potential temperature of the Archean mantle—a review of the evidence from
28. Daines., S. J., Mills, B. J. W. & Lenton, T. M. Atmospheric oxygen regulation at komatiites. Lithos 30, 291–307 (1993).
low proterozoic levels by incomplete oxidative weathering of sedimentary 59. Aulbach, S. & Arndt, N. T. Eclogites as palaeodynamic archives: evidence for
organic carbon. Nat. Commun. 8, 14379 (2017). warm (not hot) and depleted (but heterogeneous) Archaean ambient mantle.
29. Nakamura, K. & Kato, Y. Carbonatization of oceanic crust by the seafloor Earth Planet. Sci. Lett. 505, 162–172 (2019).
hydrothermal activity and its significance as a CO2 sink in the early archean. 60. Till, C. B., Grove, T. L. & Withers, A. C. The beginnings of hydrous mantle
Geochim. Cosmochim. Acta 68, 4595–4618 (2004). wedge melting. Contrib. Mineral. Petrol. 163, 669–688 (2012).
30. Fallick, A.E., Melezhik, V.A. & Simonson, B. M. The ancient anoxic biosphere 61. Carmichael, I. S. E. The redox states of basic and silicic magmas: a reflection of
was not as we know it. in Biosphere Origin and Evolution (eds. Dobretsov, N., their source regions? Contrib. Mineral. Petrol. 106, 129–141 (1991).
Kolchanov, N., Rozanov, A., & Zavarzin, G.) 169–188 (Springer US, Boston, 62. Andrault, D. et al. Large oxygen excess in the primitive mantle could be the
MA, 2008). source of the great oxygenation event. Geochem. Perspect. Lett. 6, 5–10 (2018).
31. Sun, X. L. & Turchyn, A. V. Significant contribution of authigenic carbonate 63. Avice, G. et al. Evolution of atmospheric xenon and other noble gases inferred
to marine carbon burial. Nat. Geosci. 7, 201–204 (2014). from Archean to Paleoproterozoic rocks. Geochim. Cosmochim. Acta 232,
32. Holland, H. D. Volcanic gases, black smokers, and the Great Oxidation Event. 82–100 (2018).
Geochim. Cosmochim. Acta 66, 3811–3826 (2002). 64. Zahnle, K. J., Gacesa, M. & Catling, D. C. Strange messenger: a new history of
33. Catling, D. C. & Claire, M. W. How Earthas atmosphere evolved to an oxic hydrogen on Earth, as told by xenon. Geochim. Cosmochim. Acta 244, 56–85
state: a status report. Earth Planet. Sci. Lett. 237, 1–20 (2005). (2019).
34. Claire, M. W., Catling, D. C. & Zahnle, K. J. Biogeochemical modelling of the 65. Catling, D.C. & Kasting, J.F. Atmospheric Evolution on Inhabited and Lifeless
rise in atmospheric oxygen. Geobiology 4, 239–269 (2006). Worlds. (Cambridge University Press, New York, 2017).
35. Planavsky, N. J. et al. A case for low atmospheric oxygen levels during Earth’s 66. Zahnle, K.J. & Catling, D.C. Waiting for oxygen. in Special Paper 504: Early
middle history. Emerg. Top. Life Sci. 2, 149–159 (2018). Earth’s Atmosphere and Surface Environment (eds. Shaw, G. H.) 37–48
36. Kasting, J. F., Eggler, D. H. & Raeburn, S. P. Mantle redox evolution and the (Geological Society of America, Boulder, 2014).
oxidation state of the Archean atmosphere. J. Geol. 101, 245–257 (1993). 67. Canil, D. Vanadium in peridotites, mantle redox and tectonic environments:
37. Kump, L. R., Kasting, J. F. & Barley, M. E. Rise of atmospheric oxygen and the Archean to present. Earth Planet. Sci. Lett. 195, 75–90 (2002).
"upside-down” Archean mantle. Geochem. Geophys. Geosyst. 2, 68. Lee, C. T. A., Leeman, W. P., Canil, D. & Li, Z. X. A. Similar V/Sc systematics
2000GC000114 (2001). in morb and arc basalts: implications for the oxygen fugacities of their mantle
38. Canil, D. Vanadium partitioning and the oxidation state of archaean komatiite source regions. J. Petrol. 46, 2313–2336 (2005).
magmas. Nature 389, 842–845 (1997). 69. Catling, D. C., Glein, C. R., Zahnle, K. J. & Mckay, C. P. Why O-2 is required
39. Delano, J. W. Redox history of the Earth’s interior: implications for the origin by complex life on habitable planets and the concept of planetary
of life. Orig. Life Evol. Biosph. 31, 311–341 (2001). “oxygenation time”. Astrobiology 5, 415–438 (2005).

8 NATURE COMMUNICATIONS | (2020)11:2774 | https://doi.org/10.1038/s41467-020-16493-1 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-16493-1 ARTICLE

70. Canfield, D. E., Habicht, K. S. & Thamdrup, B. The Archean sulfur cycle and Competing interests
the early history of atmospheric oxygen. Science 288, 658–661 (2000). The authors declare no competing interests.
71. Crowe, S. A. et al. Sulfate was a trace constituent of Archean seawater. Science
346, 735–739 (2014). Additional information
72. Sleep, N.H. Dioxygen over geologic time. in Metal Ions in Biological Systems, Supplementary information is available for this paper at https://doi.org/10.1038/s41467-
Vol. 43, Biogeochemical Cycles of Elements (eds. Sigel, H. & Sigel, R.) 49–73 020-16493-1.
(Taylor & Francis, Boca Raton, FL, 2005).
73. Canfield, D. E., Rosing, M. T. & Bjerrum, C. Early anaerobic metabolisms. Correspondence and requests for materials should be addressed to S.K.
Philos. Trans. R. Soc. B 361, 1819–1834 (2006).
74. Chase, M.W.NIST-JANAF thermochemical tables, Fourth edition (American Peer review information Nature Communications thanks Bruno Scaillet and the other,
Institute of Physics, 1998). anonymous, reviewer(s) for their contribution to the peer review of this work. Peer
75. Davis, F. A. & Cottrell, E. Experimental investigation of basalt and peridotite reviewer reports are available.
oxybarometers: Implications for spinel thermodynamic models and Fe3+
compatibility during generation of upper mantle melts. Am. Mineral. 103, Reprints and permission information is available at http://www.nature.com/reprints
1056–1067 (2018).
76. Birner, S. K., Cottrell, E., Warren, J. M., Kelley, K. A. & Davis, F. A. Peridotites Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
and basalts reveal broad congruence between two independent records of published maps and institutional affiliations.
mantle fo2 despite local redox heterogeneity. Earth Planet. Sci. Lett. 494,
172–189 (2018).
77. Herzberg, C., Condie, K. & Korenaga, J. Thermal history of the Earth and its Open Access This article is licensed under a Creative Commons
petrological expression. Earth Planet. Sci. Lett. 292, 79–88 (2010). Attribution 4.0 International License, which permits use, sharing,
78. Holland, H.D. The Chemical Evolution of the Atmosphere and Oceans. adaptation, distribution and reproduction in any medium or format, as long as you give
(Princeton University Press, Princeton, 1984). appropriate credit to the original author(s) and the source, provide a link to the Creative
Commons license, and indicate if changes were made. The images or other third party
material in this article are included in the article’s Creative Commons license, unless
Acknowledgements indicated otherwise in a credit line to the material. If material is not included in the
Funding support came from NSF Frontiers in Earth System Dynamics award No. article’s Creative Commons license and your intended use is not permitted by statutory
1338810. regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this license, visit http://creativecommons.org/
licenses/by/4.0/.
Author contributions
D.C.C. designed the study. S.K. performed the calculation and data analysis. S.K. with D.
C.C. wrote the paper with input from R.W.N., I.S.P., and A.D.A. All authors discussed © The Author(s) 2020
the results and commented on the paper.

NATURE COMMUNICATIONS | (2020)11:2774 | https://doi.org/10.1038/s41467-020-16493-1 | www.nature.com/naturecommunications 9