You are on page 1of 9

nature astronomy

Article https://doi.org/10.1038/s41550-023-02102-w

Venus’s atmospheric nitrogen explained by


ancient plate tectonics

Received: 22 December 2022 Matthew B. Weller 1,5 , Alexander J. Evans 1


, Daniel E. Ibarra 1,2,3
&
Alexandria V. Johnson 4
Accepted: 12 September 2023

Published online: xx xx xxxx


Venus is the least understood of the terrestrial planets. Despite broad
Check for updates similarities to the Earth in mass and size, Venus has no evidence of plate
tectonics recorded on its young surface, and Venus’s atmosphere is
strikingly different. Numerical experiments of long-term planetary
evolution have sought to understand Venus’s thermal–tectonic history with
indeterminate results. However, Venus’s atmosphere is linked to interior
evolution and can be used as a diagnostic to constrain planetary evolution.
Here we compare the present-day Venusian atmosphere to atmospheres
generated by long-term thermal–chemical–tectonic evolution models.
We find that a continuous single-plate stagnant lid regime operating since
antiquity (magma ocean solidification) explains neither the present-day
observed atmospheric abundances of N2 and CO2, nor the surface pressure.
Instead, the Venusian atmosphere requires volcanic outgassing in an early
phase of plate-tectonic-like activity. Our findings indicate that Venus’s
atmosphere results from a great climatic–tectonic transition, from an early
phase of active lid tectonics that lasted for at least 1 Gyr, followed by the
current stagnant lid-like mode of reduced outgassing rates.

Venus is shrouded by a thick 92-bar atmosphere composed of ~96.5% is rare in the Solar System, with the Earth being the only commonly
CO2, ~3% N2 and trace gases, resulting in an extreme greenhouse climate recognized case11,12. In a plate-tectonic-like state (hereafter referenced
with surface temperatures exceeding 700 K (ref. 1). The surface of Venus as active lid), the outermost rigid layer of a planet, its lithosphere, is
is covered by vast volcanic plains estimated to have been emplaced broken into tectonic plates, bounded by spreading centres and sub-
within the last 0.3–1.0 Gyr (refs. 2,3). Due to this voluminous volcan- duction zones, that move in response to mantle convective forces13.
ism, the surface record of Venus’s evolution before ~1 billion years ago Active lids have abundant volcanism focused predominantly along
(Ga) is notably absent from detection, leading to a substantial debate mid-ocean ridge spreading centres. Melt ascending from the inte-
regarding Venus’s climatic, tectonic and atmospheric history1,4–8. Yet, rior, emplaced at the surface by volcanism, allows for the transport
in contrast to its surface, Venus’s thick atmosphere likely preserves and release of volatiles to the atmosphere through outgassing14. In a
a long-term record of the planet’s surface and interior evolution as stagnant lid state, similar to Mars today15, the lithosphere is sufficiently
outgassed species and isotopic ratios. viscous to impede any long-term movement induced by the underlying
At present, Venus’s surface is tectonically and volcanically quies- mantle. A planet in a stagnant lid state lacks the hallmark signatures
cent, showing no clear evidence of Earth-like plate tectonic activity9. of active lids, namely global networks of volcanoes (for example,
Yet, determining whether Venus once had active tectonics, as previ- along mid-ocean ridges and subduction zones). Hence, compared to
ously suggested10, is critical to untangling Venus’s tectonic and geo- active lids, stagnant lids have significantly decreased volcanic rates
logic history. Active lid tectonics, of which plate tectonics is a subset, and outgassing16,17.

Department of Earth, Environmental and Planetary Sciences, Brown University, Providence, RI, USA. 2Institute at Brown for Environment and Society,
1

Brown University, Providence, RI, USA. 3Department of Earth and Planetary Sciences, University of California, Berkeley, CA, USA. 4Department of Earth,
Atmospheric, and Planetary Sciences, Purdue University, West Lafayette, IN, USA. 5Present address: The Lunar and Planetary Institute/USRA, Houston,
TX, USA. e-mail: mweller@lpi.usra.edu

Nature Astronomy
Article https://doi.org/10.1038/s41550-023-02102-w

a 20
atmospheric N2 to within an order of magnitude (Table 1). In contrast,
10
Venusian atmospheric N2 (present) stagnant lids can generate only 10–40% of the observed atmospheric
N2 at present, for the lowest and highest initial mantle and surface tem-
Earth atmospheric
N2 (present)
perature cases, respectively (Fig. 1b and Table 1).
The minimum initial mantle temperature required to generate the
0K
observed atmospheric N2 mass, assuming an upper limit with no atmos-
N2 (kg)

,75
l id =2
ti ve T mi
2,750
K

Present
T mi =
Ac
pheric loss, is shown in Table 2 for both active and stagnant lid states
10
15
Active lid:
(from linear regression of data in Table 1; r2 ≥ 0.976 for all regressions).
Ts = 100–700 K Table 2 highlights that for increasing surface temperatures, lower initial
Tmi = 2,550–2,950 K
Stagnant lid: mantle temperatures are required to match the observed atmospheric
d Ts = 100–700 K
gna
nt li Tmi = 2,550–2,950 K N2. These results suggest that for an early active lid state on Venus,
Sta
relatively small changes in the initial mantle temperature allow the
b potential of more Earth-like surface temperatures (Ts ≤ ~300 K) that
Venusian atmospheric CO2 (present)
can still match current N2 observations and yet permit the possibility
of an early habitable Venus4,8,24. To produce the observed atmospheric
20 Earth surficial
10 N2 mass, stagnant lids would require unreasonably high initial mantle
CO2 reservoir
(present)
temperatures (Tmi > 4,300 K) that would exceed Venus’s mantle solidus.
0K
CO2 (kg)

ve
l id T mi
=2
,75
We further consider outgassing of CO2. For simplicity, we do
ti
Ac
not consider any intermediate stage of carbon silicate weathering25

Present
T mi =
2,75
0 K
and instead assume outgassed CO2 feeds directly into the atmosphere
(Fig. 1b). Our results show the observed atmospheric CO2 mass for
10
15 Venus can be matched and/or exceeded for active lid cases but can-
Earth atmospheric CO2 (present)

gna
nt li
d not be attained by any of the stagnant lid cases. For active lid cases,
Sta
observed atmospheric CO2 masses are reached within ~90–580 Myr
–1 0 1 2 3 4
10 10 10 10 10 10
(Fig. 1b). However, all active lids cases overpredict the observed atmos-
Time (Myr) pheric CO2 by 30–380% (Table 1). In contrast, stagnant lid cases are
Fig. 1 | Modelled cumulative atmospheric mass of outgassed N2 and CO2 over unable to reach observed atmospheric CO2 values, underpredicting
time for an active and stagnant lid Venus. a, Outgassing of N2 to the Venusian observations by 40–90% (Fig. 1b and Table 1).
atmosphere over time in active and stagnant lid states. b, Similarly, outgassing The accumulated Venusian atmospheric surface pressure as a
of CO2 to the Venusian atmosphere. In a and b, shaded regions indicate the range function of time is shown in Fig. 2. Both active and stagnant lid cases can
of modelled atmospheric mass of N2 and CO2, respectively, over initial mantle reach the observed value for Venus; however, in simulations extended
temperatures Tmi; lighter shades indicate effects of surface temperature Ts alone; to the present day, almost all active lid cases exceed the observed sur-
solid outlines indicate minimum and maximum outgassing (at Ts = 100 K and
face pressure, but only a narrow range of stagnant lid cases can achieve
Ts = 700 K, respectively); dashed and dotted lines provide reference modelling
a close match. For active lid states, current Venusian surface pressures
for the case Tmi = 2,750 K and Ts = 300 K; orange and blue lines indicate present-
are attained in ~27–155 Myr (Fig. 2), whereas stagnant lid states barely
day values1,35,78 for Venus and Earth, respectively. In b, the dashed blue lines show
achieve observed surface pressures in relatively recent times.
extrapolation (in lieu of detail) back in time for the atmospheric and total surficial
reservoir of CO2 for the Earth35. Our results, summarized in Fig. 3, show that the outgassed atmos-
phere of a purely stagnant lid, even with generous assumptions (for
example, no atmospheric loss or sequestering mechanisms), fails to
reproduce key observational data, the present-day N2 and CO2 abun-
Thermal, tectonic and outgassing evolution dances and surface pressure. Only active lid conditions can develop
A planet’s atmosphere is established through cumulative volcanic Venus-like N2 and CO2 abundances and surface pressures. However,
outgassing of species such as N2 (refs. 18,19). N2 is largely chemically an active lid state over the lifetime of Venus would overproduce key
inert and is stable in the atmosphere over long timescales20. Hence, atmospheric observables, particularly for high internal and surface
atmospheric N2 can be used as a critical component in deciphering the temperature cases and, critically, would not be consistent with the
outgassing and tectonic histories of planets, particularly when cou- observations of a stagnant lid-like state for Venus today26–28. Without
pled with CO2 inventories and atmospheric surface pressure. Through initially considering significant loss of an atmospheric N2 reservoir
our models of thermal–chemical–tectonic evolution (Methods; Sup- (Supplementary Text 5) the only active lid case we find that matches
plementary Text 1 and 2), we examine the two standard endmembers N2 observations are those of a relatively cool initial starting state for
of planetary volcanic outgassing (active and stagnant lid states) Venus (Table 2 and the right edge of window of atmosphere emplace-
across a range of plausible initial solid-state mantle temperatures ment in Fig. 3). However, even this case results in an overabundance
(Tmi = 2,550–2,950 K; Supplementary Text 3), surface temperatures of CO2 and surface pressure (Table 1) and suggest the importance of
(Ts = 100–700 K) and initial conditions for Venus8,21–23 (Methods; Supple- other mechanisms such as Venus-specific weathering reactions29,30 and
mentary Text 4). The observed abundance constraints for the Venusian atmospheric loss31 that are likely to have occurred over the co-evolution
atmosphere should be considered the minimum targets to be reached of the atmosphere and surface. For our results, approximately
in our analysis. As illustrated in Fig. 1a, our models of outgassed N2 high- 130–160% of a CO2-equivalent atmosphere (204–231 bar) would need
light that the present-day atmospheric N2 of Venus can only be achieved to be sequestered, lost, or otherwise removed (Table 1). Yet, given that
with an early active lid state, whereas all fully stagnant lid cases cannot present-day Venus is in a stagnant lid-like regime, if a tectonic transition
match the present-day atmospheric N2 mass. Most moderate to higher from an active lid to a stagnant lid tectonic state occurred sufficiently
initial mantle temperature active lid cases (Tmi ≥ 2,650 K) rapidly reach early in Venus’s history5,32–34, this would offer a compelling pathway to
the currently observed atmospheric N2 values by ~300 Myr with the prevent substantial overproduction of the atmosphere.
nominal mantle and surface temperature case (Tmi = 2,750 K; Ts = 300 K)
attaining the currently observed N2 mass by ~240 Myr. Importantly, Implications for Venusian evolution
for active lids that last for more than ~600 Myr (Fig. 1a), the highest Although both tectonic endmember states generate a thick CO2 rich
initial mantle and surface temperatures overpredict current Venusian atmosphere, the current mass of both CO2 and N2 cannot be readily

Nature Astronomy
Article https://doi.org/10.1038/s41550-023-02102-w

Table 1 | Ratio of model-derived to observed N2, CO2 and Table 2 | Predicted initial mantle temperature required to
surface pressure at present for Venus match N2 observations for Venus

Tectonics Tmi (K) Ts (K) Model:Obs Tectonics Ts (K) Predicted Tmi (K) r2 Excess T (K)
N2 CO2 Ps 100 2,585 0.978 –
100 0.72 1.32 1.88 200 2,578 0.978 –
300 0.81 1.48 2.06 300 2,572 0.978 –
2,550
500 0.93 1.68 2.29 Active 400 2,563 0.976 –
700 1.09 1.96 2.60 500 2,555 0.976 –
100 1.83 3.04 3.80 600 2,549 0.977 –
300 1.99 3.25 4.04 700 2,544 0.977 –
2,650
500 2.22 3.53 4.35 100 4,599 0.999 2,233
700 2.52 3.88 4.74 200 4,537 0.998 2,171
Active
100 2.15 3.44 4.26 300 4,478 0.996 2,112
300 2.37 3.70 4.55 Stagnant 400 4,458 0.999 2,092
2,750
500 2.65 4.03 4.92 500 4,436 0.999 2,070
700 3.06 4.46 5.39 600 4,366 0.998 2,000
100 4.09 4.83 5.86 700 4,300 0.997 1,934
300 4.69 4.84 5.91 Regressions take N2 results from Table 1 for a fixed surface temperature across different
2,950 initial mantle temperatures for both active and stagnant lids. r2 indicates robustness of fit.
500 5.40 4.84 5.96 For stagnant lids, the excess temperature reflects the mantle temperature versus the solidus
temperature (Supplementary Fig. 3) required to generate Venusian N2 observables. Required
700 6.24 4.84 6.02
stagnant lid temperatures all are substantially above the solidus. Active lids lie substantially
100 0.11 0.11 0.35 below the solidus temperature and are not reported.

300 0.14 0.15 0.43


2,550
500 0.18 0.19 0.53 2.2 times the total surficial carbon inventory for the Earth35. Despite
suggestions of a moderately outgassed Venus (Earth-relative) from
700 0.22 0.26 0.64
present-day atmospheric 40Ar data36,37, our N2 results, in addition
100 0.15 0.15 0.45 to the similar carbon inventories for both the Earth and Venus and
300 0.18 0.20 0.53 non-radiogenic 36Ar concentrations nearly two orders of magnitude
2,650 greater than the Earth’s38, are consistent with Venus being highly out-
500 0.22 0.26 0.64
gassed39 relative to Earth.
700 0.26 0.33 0.75 While we find at least a short period of active lid convection is
Stagnant
100 0.19 0.21 0.56 required to match Venus’s atmospheric observables, an important
300 0.22 0.27 0.65 observation is that Venus currently lacks plate tectonics9, requiring
2,750 that in the past Venus underwent a global tectonic state shift5,32,33. From
500 0.26 0.33 0.75
our results and observations of Venus’s atmosphere, we can bracket the
700 0.30 0.42 0.86 plausible window for the tectonic transition to be between ~100 Myr
100 0.28 0.37 0.79 (consistent with atmospheric emplacement window) (Fig. 3) and up
300 0.32 0.44 0.89
to ~2.4 Gyr.
2,950 Our results indicate that a shift in tectonic states can match current
500 0.35 0.52 0.99
N2 observables without substantially overproducing CO2 or requiring
700 0.40 0.62 1.11 substantial atmospheric loss or sequestration. We consider a smooth
Initial model mantle temperatures were taken at 200 K steps in surface temperature Ts for and continuous shift in tectonic states (active to stagnant) analogous
both active and stagnant lid states. Table entries near unity indicate the Venusian observable to previous approaches10,40 at 100 Myr, 1 Gyr and 3.7 Gyr (100 Myr and
is met by model output, whereas values less than and greater than unity indicate the model
1 Gyr are shown in Fig. 4a,b). For an early transition in tectonics, only
underpredicts and overpredicts the observable, respectively. Ps, surface pressure.
high initial mantle temperature cases (Tmi > 2,800 K; all Ts) can roughly
account for outgassed N2 values seen at present (4–8% below the cur-
explained by either an active or stagnant lid state alone. The present-day rent values). The present numerically modelled atmospheric mass
atmosphere of Venus may better be explained by outgassing of a of CO2 ranges from ~140–161% of current values. Allowing for a later
plate-tectonic-like active lid regime before a later transition to a stag- transition in tectonics at 1 Gyr (Fig. 4b), allows all temperature cases
nant lid state. In aggregate, the observables of N2, CO2 and surface (Tmi = 2,550–2,950 K) to match or exceed atmospheric N2, except for
pressure allow us to determine a plausible window for this transition the lowest surface temperature cases and the lowest initial mantle
(Fig. 3). Our results require Venus to have had an early active lid state for temperature (Tmi = 2,550 K; Ts < 500 K). However, only the cases of low
its first few 100 Myr up to ~2.4 Gyr (Fig. 3 and Table 1). In this scenario, initial mantle and high surface temperatures (Tmi < 2,600 K; Ts ≥ 500 K)
the current Venusian atmosphere need not to have been exclusively allow for N2 to be within 15% of current values. CO2 produced at pre-
set primordially from a magma ocean state as suggested6,7 and instead sent from these cases (Tmi < 2,600 K; Ts ≥ 500 K) range from 151–168%
may have been substantially modified through outgassing in different of current atmospheric values. The discrepancy between observed
tectonics states. CO2 concentrations in our simulations and the atmosphere is likely
Our simulations indicate that early purely active lid states result a product of chemical reactions with the surface and some form of
in substantially thicker atmospheres than are observed presently for early atmospheric loss29,31,41. Increasing the transition time to 3.7 Gyr
Venus. The current atmospheric reservoir of CO2 for Venus is about (corresponding to the estimates for the last resurfacing age2,3) do not

Nature Astronomy
Article https://doi.org/10.1038/s41550-023-02102-w

24 Underpredicts Overpredicts
Venusian Ps (present) 10
10
2
Ps (bar)
Matches observations
Surface pressure (bar)

23
10
1
10

Melt mass (kg)


lid 0K
ti ve =2
,75 22 CO2 (kg)
Ac T mi 10
0
Earth Ps (present)
10 Ts > 100 K Ts > 500 K
21 Tmi > 2,600 K Tmi < 2,600 K
Present 10

–1 =2
,75
0K N2 (kg)
10 T mi Active lid: 10
20 Active lid
Ts = 100–700 K

Present
an t lid Tmi = 2,550–2,950 K Stagnant lid
Window of atmospheric
–2 St agn 19 emplacement
10 Stagnant lid: 10
Ts = 100–700 K –1 0 1 2 3 4
Tmi = 2,550–2,950 K 10 10 10 10 10 10
18
10
–3 10
10
–1 0 1
10
2
10
3
10
4 Time (Myr)
10 10
Time (Myr) Fig. 3 | Atmospheric observables compared to simulation of active and
Fig. 2 | Surface pressures (Ps) and cumulative extrusive melt produced as a stagnant lid tectonic states. Shaded boxes on the timeline (red arrows)
function of time for active and stagnant lid tectonic states. Description of indicate active (green) and stagnant (grey) lid tectonic state simulation results
results follows from Fig. 1. that match present observations of Venus1,35,78. Each box indicates the timeframe
in which some combination of simulation mantle and surface temperatures
can produce results that are present on Venus. The window of atmospheric
emplacement (blue shaded region) depends on N2 and CO2 production; it is
result in substantially different predictions due to the lower outgassing determined when N2 and CO2 match or fail to match observations, the leading
rates of active lids after ~1 Gyr, with the exception of allowing lower (early) edge occurs at Tmi > 2,600 K and Ts > 100 K, and the trailing (late) edge
surface temperatures to outgas sufficient N2 by the present (Tmi < 2,600; at Tmi < 2,600 K and Ts > 500 K. The active lid tectonic state can match present
300 K ≤ Ts ≤ 500 K; consistent with predictions from Table 2). Results conditions in Ps, CO2, and N2 if atmospheric loss, and/or other sequestration
above Tmi = 2,600 K produce an overshoot of N2 from ~120–400% of processes are considered; however, the stagnant lid tectonic state can only
atmospheric levels. These results also produce 142–160% of current match surface pressure Ps starting a few hundred Ma—with CO2 levels only
matched at ~10 Gyr, and N2 production is never sufficient to match observed
atmospheric CO2 values. In aggregate, with a transition in tectonic
atmospheric levels. The present-day atmosphere of Venus cannot have been
states, 0.7–10.5 bar of N2 and 125–149 bar of CO2 are available to be lost
produced under a stagnant tectonic lid state, whereas active lid tectonics are
to space, sequestered, or recycled back into the interior of the planet.
able to match key atmoshperic observations.
Early active lid states naturally lead to thick atmospheres and cur-
rent day Venusian levels of N2. These developing atmospheres would
lead to substantial climate change due to outgassing25, which can force
a transition in the global tectonic state from early active lid to stagnant 1 Gyr of early active lid convection on Venus and further extending
lid convection5,7,32,42,43. For the scenario of large scale resurfacing of the the younger boundary for the atmospheric window (Fig. 3) towards
planet within the last 300 Myr to 1 Gyr, followed by periods of reduced, the present, in agreement with estimates for the last resurfacing age
but nonzero, volcanic rates and tectonic activity to present5,44,45, Venus of ~3.7 Gyr (refs. 2,3).
either would have started cooler (mantle/surface) with a simple transi- Our results can be extended to early phases of active lid activity on
tion in tectonic state, or the transition is dynamic, potentially lasting similarly sized bodies (for example, Venus and Earth). Active lid planets
several Gyr to the present5. are predicted to have high outgassing rates and thick early atmospheres
In this work, we have not considered the effects of a preexisting (Fig. 2). In this state, Earth-like observables for atmospheric N2 and total
atmosphere6, or of atmospheric loss. If we consider the effects of an surficial inventory of CO2 can be matched by 100 Myr. These results
atmosphere that retains some amount of CO2 from a prior magma suggest early active lids have a substantially reduced window of time
ocean phase (Supplementary Text 4), CO2 concentrations may be (tens of megayears; Fig. 1b) in which to circumvent the accumulation of
matched in stagnant lid cases where ~34–79 bars of CO2 are already a thick CO2 rich atmosphere that precludes the existence and stability
present at the start of the experiments. Initial concentrations of CO2 of liquid water at the surface of the planet. In this scenario in the first
would need to be significantly greater for stagnant lid cases to match 100 Myr of solid-state evolution, an active lid planet to potentially
the observations if atmospheric loss is considered. However, N2 would maintain liquid water at the surface must lose to space or otherwise
still insufficiently outgas to match observations. In the active lid and sequester from the atmosphere ~40 bars of CO2.
transitioning cases, CO2 concentrations would be substantially greater In contrast, due to the lower outgassing rates of stagnant lids, it is
than currently observed atmospheric levels, indicating the loss of predicted these planets would produce substantially thinner atmos-
potentially hundreds of bars of CO2. pheres, particularly early in their evolution (Fig. 2). Stagnant lid con-
Atmospheric loss due to high EUV flux of the early Sun is likely vection requires at least 1 Gyr to match Earth-like values for N2 (Fig. 1a)
nonnegligible for the first 300–1,000 Myr (ref. 31). Ref. 31 found that and more than 2 Gyr for total surficial CO2 inventories (Fig. 1b). These
thick CO2 atmospheres may be required to prevent atmospheric results suggest that early stagnant lid planet have a substantially larger
escape, which would suggest that the thinner atmospheres of stag- window of time (gigayears) in which CO2 dominated atmospheres are
nant lids (without a large preexisting atmosphere) may be easier developing, potentially allowing for the stability of liquid water at the
to strip, potentially including a substantial fraction of N2 (ref. 46), surface of the planet.
whereas active and transitional lid cases with thicker CO2 rich atmos- Our results indicate Venus-type atmospheres may be a more
pheres (Fig. 1b) may better retain their N2. Although the details of loss likely consequence of early active lids when compared to stagnant
are complex (Supplementary Text 5), for either tectonic state some lids. The present state on Venus may result from an early runaway-
fraction of CO2 and N2 is expected to be lost in the first billion years plate-tectonic-like state in which any putative recycling, loss, or seques-
of the Solar System, moderating the N2 overshoot and limiting CO2 tration failed to prevent the rapid accumulation of CO2 in the atmos-
overproduction inherent with an early active lid. After ~0.1–1.0 Gyr, phere, leading to an extreme greenhouse state and Venus today.
loss effects would diminish. Qualitatively considering atmospheric If Venus was in an early active lid state, there are open questions:
loss indicates that a transition in tectonic regimes may become even did Venus ever harbour life, and can vestiges from this early life-bearing
more important to match atmospheric N2 abundances and not over- epoch be detectable from the atmosphere? The existence of a thick
produce CO2. This further suggests a minimum timeframe of at least CO2-rich atmosphere may imply that life never existed, that it never

Nature Astronomy
Article https://doi.org/10.1038/s41550-023-02102-w

simulations, heat generated within the interior is required to balance


a
10
20
heat loss from the surface as follows:
Venusian atmospheric N2 (present)

̇ ) − ρm fpm Lpm = As qs − Ac qc
Vm (H − ρm Cm Tm (1)
Earth atmospheric
N2 (present)

where Vm is the mantle volume, Cm is the mantle’s heat capacity, ρm is


N2 (kg)

the density of the mantle, Tm ̇ is the time averaged mantle temperature,


H is a radiogenic heat production term, Lpm is the latent heat of melt,

Transition 100 Myr


15
10
fpm is the melt production, As and Ac are the surface and core areas, and

Transition 1 Gyr
qs and qc are the heat losses from the surface and core. A simplified

Present
version of equation (1) is used to track the evolution of a single layer
core, Vm (ρc Cc Tċ ) = −Ac qc . Relevant values are given in Supplementary
Table 1.
b For the mantle, the general heat flux relationship is given by the
Nusselt-Rayleigh scaling form23,53–55:
Venusian atmospheric CO2 (present)
β
20
10 Earth surficial qm Ra
Nu = =( ) . (2)
CO2 reservoir ΔT Racr
(present) km
d
CO2 (kg)

Present Here the nondimensional heat flow is given as Nu (from Nusselt’s


Transition 100 Myr

Transition 1 Gyr

name); the Rayleigh number is Ra (from Rayleigh’s name); the critical


15
Rayleigh number is Racr, qm is the heat flux out of the mantle, km is the
10
Earth atmospheric CO2 (present) mantle thermal conductivity, ΔT is the temperature difference across
the mantle, d is the convective depth scale (taken as the convective
–1 0 1 2 3 4
10 10 10 10 10 10 mantle thickness) and β is the scaling parameter. For plate-tectonic-like
Time (Myr) active lid convection, the heat loss from the mantle can be treated as
Fig. 4 | Outgassed N2 and CO2 abundances as driven by tectonic regime the heat loss from the surface. The Rayleigh number indicates the
change at 100 Myr and 1 Gyr. a, Outgassing of N2 to the Venusian atmosphere vigour of convection:
over time with transition from active to stagnant lid states. b, Similarly,
outgassing of CO2 to the Venusian atmosphere. Transitioning at either 100 Myr gρm αΔTd3
Ra = (3)
or 1 Gyr to a stagnant lid tectonic state creates better accord with present-day κηa
observations. In a and b, the results are the same as the yellow-coloured results
in Fig. 1, up to the point of transition, after which the outgassing atmosphere where g is surface gravity, α is the mantle thermal expansivity, κ is
production is reduced. The earlier transition shows the strongest deviation thermal diffusivity and ηa is the internal viscosity, which is taken at the
from pure active or stagnant lid states. The 100 Myr transition results are shown
midpoint of the mantle (for example, (Ts − Tc ) / 2 )). Relating the time
in green, the 1 Gyr results are shown in pink. The other markings are the same
scale for boundary layer breakaway following ref. 13 results in an expres-
as in Fig. 1. The transition at 1 Gyr shows effectively no deviation (of up to a few
sion for the convective velocity u as a function of the Rayleigh
percent) from earlier transitioning cases. A transition in tectonic states from
an early active lid to a later stagnant lid can achieve present-day atmospheric
number:
observations of both outgassed species. 2β
κ Ra
u = c0 ( ) ( ) (4)
d Racr

established a sufficient rate of carbon sequestration or that any carbon where c0 is a constant13.
was recycled into the interior. If life existed, and carbon was released The previous formulations suffice to account for plate-tectonic-like
back into the atmosphere, it may be possible to detect organic car- active lid convection; however, modifications are required to account
bon isotopes. DAVINCI and other future atmospheric missions can for convection underneath a thick immobile surface, or a stagnant
well constrain C and N isotopic ratios, with 15N/14N determining how lid. Schematically, the differences between active and stagnant lid
much atmosphere was lost to space. This determination will place bet- convection are illustrated in Supplementary Fig. 1. Within a stagnant
ter constraints on the cumulative outgassing and escape rates of the lid, most temperature variation occurs within the conductive lid and
Venusian atmosphere, both of which can better constrain the timing is consequently not available to drive convection. Instead, convection
of a potential transition in tectonic states for Venus and explore the is driven by the temperature contrast across the rheologic sublayer
potential of early Venusian habitability. between the base of the stagnant lid and the nearly isothermal mantle.
Solving for the temperature at the base of the stagnant lid Tsl (ref. 47),
Methods it follows that
Parametrized thermal evolution models
Tsl = Tm − ΔTrh (5)
We simulate the long-term thermal–chemical evolution of Venus
with parametrized one-dimensional models of interior evolution that
couples the evolution of the planet from the interior to the atmos- with
phere13,23,47–51. Parametrized models have two important strengths:
the first is that they can isolate processes and behaviours that more T2m
ΔTrh = 2.23R (6)
complex models cannot individually address, which otherwise leads E
to results in which the physics describing a behaviour is difficult to where R is the universal gas constant and E is the activation energy.
pinpoint; and second, these classes of models can reach parameter In contrast to the active lid convective regime where the thickness of
ranges that more complex models are unable to achieve52. For these the lithosphere can be directly determined from mantle heat loss, the

Nature Astronomy
Article https://doi.org/10.1038/s41550-023-02102-w

growth of the stagnant lid follows from an energy balance at the base where a– f are constants from published experiments. The relation-
of the lithosphere and rheologic sublayer, or the heat flux through the ships in equations (10a) and (10b) are experimentally anhydrous and
conductive lid (dsl)53: constrained to depths corresponding to ~10 Gpa or less. For simplic-
dry dry
ity, we assume that both Tsol and Tliq are of a linear form with higher
ρm Cm (Tm − Tsl ) dsl̇ = qsl − qm (7) pressures and/or greater depths.
H2O in our simulations is an incompatible species (melt favoured)
where qsl is the heat flux at the top of the conductive stagnant lid, dsl̇ allowing for changes in the water concentrations for the bulk solid
is the lid thickness over time; qm can be solved by substituting ΔTrh mantle over time. Water partitioning in the melt is estimated following:
for ΔT in equation (2). The heat flux out of the stagnant lid qsl is
CH2 O
determined through Fourier’s law, using the temperature contrast Xmelt = ( ) (11)
across the stagnant lid and the stagnant lid thickness. Inspection of DH2 O + φ (1 − DH2 O )
equations (5)–(7) implicitly suggests that the depth scales of convec-
tion between plate-tectonic-like active lids and stagnant lids are where Xmelt is the concentration of water in the melt (kg H2O per kg of
starkly different. Inherent in the assumption of active lid convection melt), CH2 O is the weight fraction concentration of water in the bulk
is that the surface participates in mantle overturn, thus suggesting solid mantle, DH2 O is the bulk distribution coefficient of water and φ
that the convective depth is the entire mantle depth d. Conversely, is the melt fraction. The melt fraction is averaged in our melt layer
in the stagnant lid, the convective mantle occurs beneath the thick (calculated from onset depth of melting to cessation depth of melt-
stagnant lid. Thus, a modification to equations (3) and (4), substitut- ing). Clinopyroxene phases are not depleted, and consequently this
ing mantle depth d with the new stagnant lid convective thickness approach is similar to equilibrium batch melting. These assumptions
of d − dsl̇ , with dsl̇ (as constrained by equation (7)) to potentially be a ensure the solidus and liquidus functions are continuous over the
substantive portion of the entire mantle thickness d. The power law melt domain.
exponent β in equations (2) and (4) is taken to be 0.33 for endmember To account for the effects of dissolved water components on
active lid and 0.2 for endmember stagnant lid convection23, per Sup- the anhydrous solidus and liquidus equations (10a) and (10b) are
plementary Table 1. modified62:
dry
Viscosity parametrization Tsol = Tsol − ΔTH2 O (12a)
Our viscosity formulation depends on temperature, pressure, water
content48,49 and melt fraction (‘Melting parametrization’): dry
Tliq = Tliq − ΔTH2 O (12b)
E + pV
η (Tm , p, H2 O, φ) = A0 fH2 O exp ( − γφ) (8)
RTm
where Tsol and Tliq are the composite hydrous solidus and liquidus and
dry dry
Tsol and Tliq are the anhydrous solidus and liquidus temperatures. ΔTH2 O
A0 and γ are empirical constants49,56–58, p is pressure and V is the is the shift in temperatures from dry to melting hydrous curves, which
activation volume. We assume viscosity is controlled by diffusion scales as the water concentration in the melt:
creep and therefore behaves as a Newtonian fluid. Our formulation γ
relates an empirical relationship that defines fugacity in terms of water ΔTH2 O = KXmelt (13)
concentrations (OH)59:
−r
K and γ are constants as determined by ref. 62. The melt fraction
2 3
fH2 O = exp (c0 + c1 lnCOH + c2 ln COH +c3 ln COH ) (9) φ then can be parametrized with the following power law
relationship:
β
dry
where fH2 O is the water fugacity, c0 − c3 are experimental constants, r T − (Tsol − ∆TH2 O (Xmelt ))
is the fugacity constant and COH is the water concentration, which is φ=( ) (14)
dry dry
Tliq − Tsol
expressed in units of atomic H/106 Si. We assume an olivine mineral
ratio of Fe2SiO4 to Mg2SiO4 of 1:9. Viscosity is therefore modified by the
effects of oxygen fugacity fH2 O and melt fraction φ (refs. 49,59), as where T is the time dependent temperature of the mantle at a
shown in equations (8) and (9), with both quantities evolving together given depth.
with time and temperature. Calibration of the numerical simulations We account for the effects of depletion of melt and residuum51,64.
is shown in Supplementary Text 1 and Supplementary Fig. 2. The difference in solidus is estimated at ~150 K between undepleted
and depleted (that is, harzburgite) peridotite64. The solidus of the
Melting parametrization residuum is increased linearly by degree of melt (maximum 40%) up to
Melting serves as a form of direct communication between the inte- a maximum value of 150 K, similar in principle to the approach of ref. 51.
rior and surface of a planet, directly coupling to the thermal evolu- For simplicity and to account for melt buoyancy65,66, melt generated
tion (equation (1)) and viscosity (equation (8)) models. We use the above 12 GPa is assumed to be transported upward towards the surface
well-studied model for the generation of partial melt from adiabatic and below 12 GPa to be extracted downward deeper into the mantle.
decompression60,61 and models of hydrous mantle melting62. Calculation of the melt zone includes a determination of the upper
To parametrize the solidus and liquidus for our experiments, we and lower boundary of melt production. These are determined where
adopt a standard form of a simple polynomial expression between the solidus crosses the mantle adiabat (Supplementary Fig. 3). We
temperature and pressure63: calculate the thermal profile (the venotherm) in two parts (Supple-
mentary Fig. 1): a conductive lithosphere and the convective adiabat.
dry
Tsol = aP2 + bP + c (10a) The difference between the upper and lower crossing is taken to be the
average partial melt zone thickness. The partial melt zone thickness is
dry a function of the thermal evolution, the mantle water content and the
Tliq = dP2 + eP + f (10b)
amount of previous melt and depletion that has occurred.

Nature Astronomy
Article https://doi.org/10.1038/s41550-023-02102-w

Melt extraction and outgassing (ref. 70) (the fayalite–magnetite–quartz buffer). On differentiated bod-
For a planet operating in a plate-tectonic-like regime, the vast major- ies in our Solar System, this value represents the average fO2 of basalts72.
ity of volcanism is accommodated through decompression melt at The concentration of water in the Venusian mantle is unknown, but
spreading centres (on Earth, mid-ocean ridges). Melt production in estimates range from a nominally dry ~5–50 ppm to a relatively water
this case simply follows49 rich ~800 ppm (refs. 36,73–75). We assume the bulk solid-state Venusian
mantle initially contains 600 ppm of both H2O and CO2 (refs. 20,76)
fpm = φ∗ Dmelt Spr (15)
and 40 ppm of N2 (ref. 77) at the beginning of our simulations. The
solubility of N in silicate melt can be substantially greater, ranging
where φ* is the melt fraction that reaches the surface and Dmelt is from ~200 ppm to a few weight percent (with the assumption that the
the depth of the melting layer to the surface, which changes over time. partition coefficient is 0 and the mantle is oxidizing, which is likely
The spreading rate of the mid-ocean ridges follows the function appropriate for melt sourced in Earth-like upper mantles). Melt there-
fore can be highly enriched when compared to the background mantle.
Spr = 2MORL uc (16)
We allow for a maximum potential concentration of 400 ppm of N2
that may go into the melt, provided there is sufficient N2 available in
where MORL is the ridge length and uc is the convective velocity the surrounding mantle, allowing for this enrichment. We examine
in a plate-tectonic-like active lid13,49. MORL is assumed to be both con- the results of differing N2 concentrations in Supplementary Text 6 and
stant with time and to scale with the surface area difference between Supplementary Figs. 7–9. Concentrations of species in the mantle are
Earth and Venus. updated as melting progresses, reducing the total availability of spe-
Convection within a stagnant lid planet is dominated by regions of cies that may enter the melt. We allow for total depletion of volatiles
passive upwellings and downwellings, which stands in stark contrast to from the mantle; however, no model explored reaches this condition.
the much more active stirring and vigorous convection of a Volatiles are extracted from the melt that reaches the surface and are
plate-tectonic-like system. Passive upwelling velocities are approxi- released to the atmosphere as constrained by the pressure-sensitive gas
mately proportional to the velocity of the interior ui, such that ui < uc. speciation of ref. 70 (‘Melt extraction and outgassing’). For simplicity,
Holding all things equal between active and stagnant lids, adiabatic and given the lack of constraints, we assume the initial atmosphere in
decompression melt is suppressed due to the thicker stagnant lid when our numerical experiments to be negligible of order 1 × 10−3 bars of
compared to active lid convection (illustrated in Supplementary Figs. 3 pressure. This is done to focus the experiment on the process of outgas-
and 4). Increasing boundary layer thicknesses require an increase in sing after solid-state convection initiates. The tenuous atmosphere is
the mantle temperatures to generate similar levels of melt. With lower assumed to be the product of volcanic outgassing, with the chemical
convective velocities fluxing new material into the melt zone and also makeup of the initial atmosphere reflecting the initial outgassed spe-
suppressed adiabatic melting, melt production then is generally more cies from the speciation model at the first timestep.
inefficient when compared to active lid systems67 and is expressed as

Data availability
4ui φ D2melt Data used to generate Figs. 1–4 are available at https://doi.org/10.5281/
fpm = As (17)
d2 zenodo.7570178.

here, the melt channel thickness Dmelt is measured from the base of the Code availability
lithosphere to the base of the melt zone. The depth of the convecting MATLAB is a commercial code. The C parametrized thermal evolution
mantle is a proxy for convective cell size and conveniently normalizes code used here and described in the Methods sections, in addition to
melt production across the entire surface area As of the planet. the MATLAB analysis code(s), are available from the authors upon
The general expression for the volatile flow rate into the atmos- reasonable request.
phere is as follows:
References
Ωv = ρm fpm χd (18)
1. Bullock, M. A. & Grinspoon, D. H. The recent evolution of climate
on Venus. Icarus 150, 19–37 (2001).
where χd is the outgassing efficiency factor following from refs. 49, 2. Schaber, G. G. et al. Geology and distribution of impact craters on
68, 69 (see Supplementary Text 2 for further discussion). Volatiles are Venus: what are they telling us? J. Geophys. Res. 97, 13257–13301
extracted from the interior as melting progresses, allowing for feed- (1992).
back in mantle viscosity and the long-term thermal evolution. Regas- 3. Strom, R. G., Schaber, G. G. & Dawson, D. D. The global resurfacing
sing and release of volatile species from metamorphic reactions are not of Venus. J. Geophys. Res. Planets 99, 10899–10926 (1994).
considered. A comparison of convective velocity, boundary layer thick- 4. Way, M. J. et al. Was Venus the first habitable world of our solar
ness and melt production for our models is shown in Supplementary system? Geophys. Res. Lett. 43, 8376–8383 (2016).
Fig. 4. The general overall model setup is illustrated in Supplementary 5. Weller, M. B. & Kiefer, W. S. The physics of changing tectonic
Fig. 5. To link the melt evolution (from the convection experiments) to regimes: Implications for the temporal evolution of mantle
outgassing (atmosphere evolution), we employ the gas speciation of convection and the thermal history of Venus. J. Geophys. Res.
model of ref. 70 (see Supplementary Text 2 and Supplementary Fig. 6 Planets 125, e2019JE005960 (2020).
for additional discussion). 6. Gillmann, C. et al. Dry late accretion inferred from Venus’s
coupled atmosphere and internal evolution. Nat. Geosci. 13,
Initial composition and atmosphere 265 (2020).
Given the current uncertainty of Venus’s internal composition, we 7. Gillmann, C. & Tackley, P. Atmosphere/mantle coupling
restrict our consideration to Earth-like compositions. Our models and feedbacks on Venus. J. Geophys. Res. Planets 119,
follow from estimates of the Bulk Silicate Earth composition71. This 1189–1217 (2014).
assumption offers two critical controls: (1) the Bulk Silicate Earth 8. Way, M. J. & Del Genio, A. D. Venusian habitable climate
assumption constrains the outgassing species and (2) also constrains scenarios: modeling Venus through time and applications to
models of the interior structure to be consistent with a MgSiO3 man- slowly rotating Venus-like exoplanets. J. Geophys. Res. 125,
tle. The redox conditions of the pre-melt mantle are fixed at FMQ-1.5 e2019JE006276 (2020).

Nature Astronomy
Article https://doi.org/10.1038/s41550-023-02102-w

9. Solomon, S. C. et al. Venus tectonics - An overview of Magellan 33. Weller, M. B. & Lenardic, A. On the evolution of terrestrial planets:
observations. J. Geophys. Res. Planets 97, 13199–13255 (1992). bi-stability, stochastic effects, and the non-uniqueness of tectonic
10. Phillips, R. J., Bullock, M. A. & Hauck, S. A. Climate and states. Geosci. Front 9, 91–102 (2018).
interior coupled evolution on Venus. Geophys. Res. Lett. 28, 34. O’Neill, C. et al. A window for plate tectonics in terrestrial planet
1779–1782 (2001). evolution? Phys. Earth Planet. Inter. 255, 80–92 (2016).
11. McGill, G. E. Venus tectonics: another Earth or another Mars? 35. Lécuyer, C., Simon, L. & Guyot, F. Comparison of carbon, nitrogen
Geophys. Res. Lett. 6, 739–741 (1979). and water budgets on Venus and the Earth. Earth Planet. Sci. Lett.
12. Kaula, W. M. & Phillips, R. J. Quantitative tests for plate tectonics 181, 33 (2000).
on Venus. Geophys. Res. Lett. 8, 1187–1190 (1981). 36. Namiki, N. & Solomon, S. C. Volcanic degassing of argon
13. Schubert, G., Turcotte, D. L. & Olson, P. Mantle Convection in the and helium and the history of crustal production on Venus.
Earth and Planets (Cambridge Univ. Press, 2001). J. Geophys. Res. 103, 3655 (1998).
14. Cann, J. R. et al. A review of melt migration processes in the 37. Kaula, W. M. Constraints on Venus evolution from radiogenic
adiabatically upwelling mantle beneath oceanic spreading argon. Icarus 139, 32 (1999).
ridges. Philos. Trans. R. Soc. A. 355, 283–318 (1997). 38. Hoffman, J. H., Hodges, R. R., Donahue, T. M. & McElroy, M. B.
15. Nimmo, F. & Stevenson, D. J. Influence of early plate tectonics on Composition of the Venus lower atmosphere from the Pioneer
the thermal evolution and magnetic field of Mars. J. Geophys. Res. Venus Mass Spectrometer. J. Geophys. Res. Space Phys. 85,
Planets 105, 11969–11979 (2000). 7882–7890 (1980).
16. Guimond, C. M., Noack, L., Ortenzi, G. & Sohl, F. Low volcanic 39. Halliday, A. N. The origins of volatiles in the terrestrial planets.
outgassing rates for a stagnant lid archean Earth with Geochim. Cosmochim. Acta 105, 146–171 (2013).
graphite-saturated magmas. Phys. Earth Planet. Inter. 320, 40. Krissansen-Totton, J., Fortney, J. J. & Nimmo, F. Was Venus ever
106788 (2021). habitable? Constraints from a coupled interior–atmosphere–
17. Noack, L., Rivoldini, A. & Van Hoolst, T. Volcanism and outgassing redox evolution model. Planet. Sci. J. 2, 216 (2021).
of stagnant-lid planets: Implications for the habitable zone. 41. Bierson, C. J. & Zhang, X. Chemical cycling in the Venusian
Phys. Earth Planet. Inter. 269, 40–57 (2017). atmosphere: a full photochemical model from the surface to
18. Som, S. M. et al. Earth’s air pressure 2.7 billion years ago 110 km. J. Geophys. Res. 125, e2019JE006159 (2020).
constrained to less than half of modern levels. Nat. Geosci. 9, 42. Lenardic, A., Jellinek, A. M. & Moresi, L. N. A climate induced
448–451 (2016). transition in the tectonic style of a terrestrial planet. Earth Planet.
19. Stüeken, E. E. et al. Mission to planet Earth: the first two billion Sci. Lett. 271, 34–42 (2008).
years. Space Sci. Rev. 216, 31 (2020). 43. Foley, B. J., Bercovici, D. & Landuyt, W. The conditions for
20. Marty, B. The origins and concentrations of water, carbon, plate tectonics on super-Earths: inferences from convection
nitrogen and noble gases on Earth. Earth Planet. Sci. Lett. 313, models with damage. Earth Planet. Sci. Lett. 331–332,
56–66 (2012). 281–290 (2012).
21. Boujibar, A., Driscoll, P. & Fei, Y. Super-earth internal structures 44. Byrne, P. K. et al. A globally fragmented and mobile lithosphere
and initial thermal states. J. Geophys. Res. 125, e2019JE006124 on Venus. Proc. Natl Acad. Sci. 118, e2025919118 (2021).
(2020). 45. Filiberto, J., Trang, D., Treiman, A. H. & Gilmore, M. S. Present-day
22. Taylor, F. & Grinspoon, D. Climate evolution of Venus. J. Geophys. volcanism on Venus as evidenced from weathering rates of
Res. https://doi.org/10.1029/2008JE003316 (2009). olivine. Sci. Adv. 6, eaax7445 (2020).
23. Breuer, D. & Moore, W. B. in Treatise on Geophysics 46. Tian, F., Kasting, J. F., Liu, H.-L. & Roble, R. G. Hydrodynamic
(ed. Gerald Schubert) 299–348 (Elsevier, 2007). planetary thermosphere model: 1. Response of the Earth’s
24. Warren, A. O. & Kite, E. S. Narrow range of early habitable Venus thermosphere to extreme solar EUV conditions and the
scenarios permitted by modeling of oxygen loss and radiogenic significance of adiabatic cooling. J. Geophys. Res. https://doi.
argon degassing. Proc. Natl Acad. Sci. 120, e2209751120 (2023). org/10.1029/2007JE002946 (2008).
25. Walker, J. C. G., Hays, P. B. & Kasting, J. F. A negative feedback 47. Grasset, O. & Parmentier, E. M. Thermal convection in a
mechanism for the long-term stabilization of the Earth’s surface volumetrically heated, infinite Prandtl number fluid with
temperature. J. Geophys. Res. 86, 9776 (1981). strongly temperature-dependent viscosity: implications for
26. Rolf, T. et al. Dynamics and evolution of Venus’ mantle through planetary thermal evolution. J. Geophys. Res. Solid Earth 103,
time. Space Sci. Rev. 218, 70 (2022). 18171–18181 (1998).
27. Solomatov, V. S. & Moresi, L. N. Stagnant lid convection on Venus. 48. Sandu, C. & Kiefer, W. S. Degassing history of Mars and the
J. Geophys. Res. 101, 4737–4753 (1996). lifespan of its magnetic dynamo. Geophys. Res. Lett. 39,
28. Bjonnes, E., Johnson, B. C. & Evans, A. J. Estimating Venusian L03201 (2012).
thermal conditions using multiring basin morphology. Nat. Astron 49. Sandu, C., Lenardic, A. & McGovern, P. The effects of deep water
5, 498–502 (2021). cycling on planetary thermal evolution. J. Geophys. Res. Solid
29. Zolotov, M. Y. Gas–solid interactions on venus and other solar Earth 116, B12404 (2011).
system bodies. Rev. Mineral. Geochem. 84, 351–392 (2018). 50. Grott, M., Morschhauser, A., Breuer, D. & Hauber, E. Volcanic
30. Taylor, F. W., Svedhem, H. & Head, J. W. Venus: the atmosphere, outgassing of CO2 and H2O on Mars. Earth Planet. Sci. Lett. 308,
climate, surface, interior and near-space environment of an 391–400 (2011).
Earth-like planet. Space Sci. Rev. https://doi.org/10.1007/s11214- 51. Morschhauser, A., Grott, M. & Breuer, D. Crustal recycling, mantle
018-0467-8 (2018). dehydration, and the thermal evolution of Mars. Icarus 212,
31. Johnstone, C. P., Lammer, H., Kislyakova, K. G., Scherf, M. & 541–558 (2011).
Güdel, M. The young sun’s XUV-activity as a constraint for lower 52. Weller, M. B., Lenardic, A. & Moore, W. B. Scaling relationships
CO2-limits in the Earth’s archean atmosphere. Earth Planet. Sci. and physics for mixed heating convection in planetary interiors:
Lett. 576, 117197 (2021). isoviscous spherical shells. J. Geophys. Res. Solid Earth 121,
32. Weller, M. B., Lenardic, A. & O’Neill, C. The effects of internal 7598–7617 (2016).
heating and large scale climate variations on tectonic bi-stability 53. Schubert, G. Subsolidus convection in the mantles of terrestrial
in terrestrial planets. Earth Planet. Sci. Lett. 420, 85–94 (2015). planets. Annu. Rev. Earth Planet. Sci. 7, 289–342 (1979).

Nature Astronomy
Article https://doi.org/10.1038/s41550-023-02102-w

54. Schubert, G., Stevenson, D. & Cassen, P. Whole planet cooling 74. Marty, B. & Yokochi, R. Water in the early Earth. Rev. Mineral.
and the radiogenic heat source contents of the Earth and moon. Geochem. 62, 421–450 (2006).
J. Geophys. Res. Solid Earth 85, 2531–2538 (1980). 75. Peslier, A. H., Schönbächler, M., Busemann, H. & Karato, S.-I.
55. Stevenson, D. J., Spohn, T. & Schubert, G. Magnetism and thermal Water in the Earth’s interior: distribution and origin. Space Sci.
evolution of the terrestrial planets. Icarus 54, 466–489 (1983). Rev. 212, 743–810 (2017).
56. Hirth, G. & Kohlstedt, D. in Inside the Subduction Factory, Vol. 138 76. Saal, A. E., Hauri, E. H., Langmuir, C. H. & Perfit, M. R. Vapour
(ed. Eiler, J.) 83–105 (AGU Geophysical Monograph, 2003). undersaturation in primitive mid-ocean-ridge basalt and
57. Liebske, C. et al. Viscosity of peridotite liquid up to 13 GPa: the volatile content of Earth’s upper mantle. Nature 419,
implications for magma ocean viscosities. Earth Planet. Sci. Lett. 451–455 (2002).
240, 589–604 (2005). 77. Grewal, D. S., Dasgupta, R., Hough, T. & Farnell, A. Rates of
58. Karato, S. & Wu, P. Rheology of the upper mantle: a synthesis. protoplanetary accretion and differentiation set nitrogen budget
Science 260, 771–778 (1993). of rocky planets. Nat. Geosci. 14, 369–376 (2021).
59. Li, Z.-X. A., Lee, C.-T. A., Peslier, A. H., Lenardic, A. & Mackwell, S. J. 78. Baines, K. H. et al. in Comparative Climatology of Terrestrial Planets
Water contents in mantle xenoliths from the Colorado (eds Bullock, M. A., Mackwell, S. J., Simon-Miller, A. A. &
Plateau and vicinity: implications for the mantle rheology and Harder, J. W.) 137–160 (Univ. of Arizona Press, 2013).
hydration-induced thinning of continental lithosphere.
J. Geophys. Res. Solid Earth 113 https://doi.org/10.1029/ Acknowledgements
2007JB005540 (2008). This work was supported by funds provided by A.J.E., Brown University
60. McKenzie, D. The generation and compaction of partially molten and NASA’s Solar System Workings programme (grant number
rock. J. Petrol. 25, 713–765 (1984). 80NSSC23K0167), which partially funded M.B.W. Additional support
61. McKenzie, D. & Bickle, M. J. The volume and composition was provided through the USRA/LPI Urey fellowship for M.B.W.
of melt generated by extension of the lithosphere. J. Petrol. 29,
625–679 (1988). Author contributions
62. Katz, R. F., Spiegelman, M. & Langmuir, C. H. A new M.B.W., A.J.E. and A.V.J. conceptualized the project. A.J.E., M.B.W. and
parameterization of hydrous mantle melting. Geochemistry, D.E.I. devised the methodology and M.B.W. and A.J.E. performed the
Geophys. Geosystems 4, 1073 (2003). investigation. Visualization was done by M.B.W. Funding acquisition
63. Hirschmann, M. M. The mantle solidus: experimental constraints was handled by M.B.W. and A.J.E. All authors contributed to the writing
and the effect of peridotite composition. Geochemistry Geophys. and editing of the manuscript.
Geosystems https://doi.org/10.1029/2000GC000070 (2000).
64. Maaløe, S. The solidus of harzburgite to 3 GPa pressure: the Competing interests
compositions of primary abyssal tholeiite. Mineral. Petrol. 81, The authors declare no competing interests.
1–17 (2004).
65. Ohtani, E., Nagata, Y., Suzuki, A. & Kato, T. Melting relations of Additional information
peridotite and the density crossover in planetary mantles. Chem. Supplementary information The online version
Geol. 120, 207–221 (1995). contains supplementary material available at
66. Ohtani, E., Suzuki, A. & Kato, T. in Properties of Earth and Planetary https://doi.org/10.1038/s41550-023-02102-w.
Materials at High Pressure and Temperature (eds Manghnani, M. H. &
Yagi, T.) 227–239 (American Geophysical Union, 1998). Correspondence and requests for materials should be addressed
67. Hauck, S. A. & Phillips, R. J. Thermal and crustal evolution of Mars. to Matthew B. Weller.
J. Geophys. Res. Planets https://doi.org/10.1029/2001je001801
(2002). Peer review information Nature Astronomy thanks Helmut Lammer,
68. Grott, M., Breuer, D. & Laneuville, M. Thermo-chemical evolution Cedric Gillmann and the other, anonymous, reviewer(s) for their
and global contraction of Mercury. Earth Planet. Sci. Lett. 307, contribution to the peer review of this work.
135–146 (2011).
69. O’Neill, C., Lenardic, A., Jellinek, A. M. & Kiefer, W. S. Melt Reprints and permissions information is available at
propagation and volcanism in mantle convection simulations, www.nature.com/reprints.
with applications for Martian volcanic and atmospheric evolution.
J. Geophys. Res. https://doi.org/10.1029/2006JE002799 (2007). Publisher’s note Springer Nature remains neutral with regard to
70. Gaillard, F. & Scaillet, B. A theoretical framework for volcanic jurisdictional claims in published maps and institutional affiliations.
degassing chemistry in a comparative planetology perspective
and implications for planetary atmospheres. Earth Planet. Sci. Springer Nature or its licensor (e.g. a society or other partner) holds
Lett. 403, 307–316 (2014). exclusive rights to this article under a publishing agreement with
71. McDonough, W. F. & Sun, S. S. The composition of the Earth. the author(s) or other rightsholder(s); author self-archiving of the
Chem. Geol. 120, 223–253 (1995). accepted manuscript version of this article is solely governed by the
72. Herd, C. D. K. Basalts as probes of planetary interior redox state. terms of such publishing agreement and applicable law.
Rev. Mineral. Geochem. 68, 527–553 (2008).
73. Grinspoon, D. H. Implications of the high D/H ratio for the sources © The Author(s), under exclusive licence to Springer Nature Limited
of water in Venus’ atmosphere. Nature 363, 428–431 (1993). 2023

Nature Astronomy

You might also like