WARMING EARLY MARS WITH A DENSE ATMOSPHERE OF CO2 AND
H2: A REVIEW OF RECENT NUMERICAL AND EXPERIMENTAL
ADVANCES
M. Turbet, F. Forget, J.M. Hartmann, H. Tran, W. Fakhardji, Laboratoire de Météorologie Dynamique,
IPSL, CNRS, Sorbonne Université, École Normale Supérieure, PSL Research University, École Polytechnique,
75005 Paris, France (martin.turbet@lmd.ipsl.fr), D. Mondelain, Univ. Grenoble Alpes, CNRS, LIPhy, 38000
Grenoble, France, C. Boulet, Université Paris-Saclay, CNRS, Institut des Sciences Moléculaires d'Orsay (ISMO),
91405, Orsay, France, O. Pirali, AILES Beamline SOLEIL Synchrotron, L'Orme des Merisiers, Saint-Aubin
91190, France ; Institut des Sciences Moléculaires d'Orsay, CNRS, Université Paris-Saclay, Orsay 91405,
France.

Introduction: To this day, the enigma of Mars'
ancient climate remains unexplained (Wordsworth et
al. 2016, Haberle et al. 2017, Kite 2019). What was
the atmosphere of Mars made of at the time when –
near the Late Noachian epoch – lakes and valley networks (see e.g., Fassett & Head 2008a,b, Hynek et al.
2010, Goudge et al. 2015) were formed? Although
there is a broad consensus to date that liquid water is
the main culprit, the nature of the atmosphere at the
time (i.e., the atmospheric pressure and composition)
remains essentially unknown (Wordsworth et al.
2016, Haberle et al. 2017, Kite 2019).
Many hypotheses have been proposed and tested
so far to warm the surface of Mars enough to keep
surface water reservoirs liquid. However, none of
them provide satisfactory results:
- Accumulating several bars of CO2 – to values
even higher than what is allowed by the few existing
observations (see Kite 2019, Fig. 9) – into the atmosphere does not work (Forget et al. 2013, Wordsworth
et al. 2013). This is because CO2 condenses (possibly
on the surface) and because the CO2 greenhouse effect
saturates above 2-3bar (for a low-gravity planet like
Mars) due to the cooling by Rayleigh scattering
(Wordsworth et al. 2010, Forget et al. 2013, Kopparapu et al. 2014, Ramirez et al. 2014, Turbet & Tran
2017). The greenhouse effect of the CO2 ice clouds
(Forget et al. 2013, Kitzmann 2016) and water ice
clouds (Wordsworth et al. 2013, Ramirez & Kasting
2017, Turbet et al. 2020a) do not allow Mars to be
heated sufficiently either.
- Accumulating volcanic gases (SO2, H2S) does
not work, because the volcanic aerosols produce overall a net cooling effect (Tian et al. 2010, Halevy &
Head 2014, Kerber et al. 2015).
- Meteoritic impacts do not work, because their
cumulated impact on the hydrological cycle is too
small to carve valley networks (Turbet 2018, Steakley
et al. 2019, Turbet et al. 2020a).
Given the apparent failure of these hypotheses, a
new hypothesis has recently emerged, according to
which the warm climate of early Mars would have
been sustained by a dense CO2+H2 atmosphere
(Ramirez et al. 2014, Wordsworth et al. 2017, Turbet

et al. 2020b). From a climatic point of view, a dense
CO2+H2 atmosphere is the most favored scenario to
date, as it is the only composition known to explain
the apparent persistence of surface liquid water, and
thus the presence of Martian valley networks and
lakes, despite the reduced solar luminosity of the time
(about 75% of today's). This scenario, which requires
that the atmosphere and presumably the mantle of
Mars have been chemically reduced in the past, has
attracted considerable attention in recent years
(Ramirez et al. 2014, Batalha et al. 2015, Wordsworth
et al. 2017, Turbet et al. 2019, Ramirez 2019, Turbet
et al. 2020b, Kamada et al. 2020, Hayworth et al.
2020, Ramirez et al. 2020, Godin et al. 2020, Turbet
& Forget 2021, Wordsworth et al. 2021, Guzewich et
al. 2021, Kamada et al. 2021). This has also revived
interest in studying the climatic effects of large impacts that may have generated large amounts of H2
(Haberle et al. 2019); although we recall the readers
that these large impacts are not contemporaneous with
the periods of lake and valley network formation (Fassett & Head 2008, 2011, Turbet et al. 2020a).
In this presentation, we will review our recent advances to test the CO2+H2 hypothesis. Firstly on the
experimental side, to better constrain the spectral
opacity of a CO2+H2 rich atmosphere, which determines the amount of H2 needed to heat Mars, and thus
the plausibility of this scenario. Secondly on the numerical side, to better understand the impact of a
CO2+H2 rich atmosphere on the hydrological cycle,
which can then be compared with the geologic records
(e.g., maps of lakes and valley networks).
Results from laboratory experiments: We conducted a series of laboratory experiments, using the
Fourier-Transform Spectroscopy (FTS) technique
(Turbet et al. 2019, 2020b, Fakhardji et al. 2022) and
the Cavity Ring Down Spectroscopy (CRDS) technique (Mondelain et al. 2021). Most of these experiments are part of the COMPLEAT (COntinuum
Measurements for PLanEtary ATmospheres) French
ANR project, which aims at better evaluating the continuum absorptions (CIA, dimer, far wings) of hybrid
gas mixtures. Fig. 1 summarizes all the measurements
we have made so far of the collision-induced absorption (CIA) produced by the pair CO2+H2. The

measurements are in good agreement (see Fig. 1) with
our theoretical calculations in black (Turbet et al.
2020b, Mondelain et al. 2021), at least in the 20-50µm
CO2 spectral window, where the CO2+H2 CIA matters
the most for the greenhouse effect calculations. The
agreement with measurements performed in the 2.12.55µm spectral region (Mondelain et al. 2021,
Fakhardji et al. 2022) provide additional confidence
that the calculations perform quite well.
Our calculations can be extended to a wide range
of temperatures (typically from 100 to 600K), as
shown in Fig. 2. The resulting look-up tables are made
available to the community (Turbet et al. 2020b,
Mondelain et al. 2021, Fakardhji et al. 2022), and can
be directly implemented as new opacity sources in radiative transfer models.
Results from 1-D and 3-D climate model simulations: We implemented these new CO2+H2 opacity
sources in our early Mars 1-D and 3D climate models
(using the LMD 3-D Generic Global Climate Model
– GCM). We first evaluated – following Ramirez et
al. 2014, Wordsworth et al. 2017, Turbet et al. 2020b
– how the mean surface temperature of early Mars
evolves as a function of CO2 and H2 atmospheric reservoirs, as illustrated in Fig. 3 (taken from Turbet &
Forget 2021). With our 1-D radiative-convective climate model calculations (Turbet et al. 2020b), we find
that the amount of H2 required to warm early Mars is
about 7% (for 2 bar of CO2), i.e. between the previous
results of Ramirez et al. 2014 (15 %) and Wordsworth
et al. 2017 (2.5 %), as illustrated in Fig. 3 (black, pink
and green data points). We further constrain these
numbers using 3-D Global Climate Model simulations (see Fig. 3; red, blue and orange data points) assuming different surface water reservoirs (Turbet &
Forget et al. 2021).
More importantly, we explored the impact of a
dense CO2+H2 atmosphere on the hydrological cycle,
for different CO2 surface pressures (1 and 2 bar), H2
mixing ratios (from 0.5 to 30%), and water reservoir
distributions (see Fig. 3, maps on the left), assuming
a 40° obliquity. We find that the adiabatic cooling
mechanism (Wordsworth et al., 2013) that leads to the
preferential accumulation of ice deposits in the southern highlands in cold climates (the so-called ’icy highland’ scenario) also works in warm climates, with impact crater lakes acting as the main water reservoirs.
This produces rainfall localized in the southern highlands of Mars (see an example in Fig. 4), whether
oceans are present or not. Fig. 4 illustrates that the observed distribution of valley networks can match –
with a few exceptions detailed in Turbet & Forget
2021 – the observed distribution of valley networks
and impact crater lakes.
References:
Batalha et al. 2015, Icarus vol. 258
Bouley et al. 2016, Nature vol. 531
Fakhardji et al. 2022, JQSRT vol. 283

Fassett & Head 2008a, Icarus vol. 198
Fassett & Head 2008b, Icarus vol. 195
Fassett & Head 2011, Icarus vol. 211
Forget et al. 2013, Icarus vol. 222
Godin et al. 2020, JGR Planets vol. 125
Goudge et al. 2015, Icarus vol. 260
Guzewich et al. 2021, JGR Planets vol. 126
Haberle et al. 2017, Cambridge Univ. Press
Haberle et al. 2019, GRL vol. 46
Halevy & Head 2014, Nature Geoscience vol. 7
Hayworth et al. 2020, Icarus vol. 345
Hynek et al. 2010, JGR Planets vol. 115
Kamada et al. 2020, Icarus vol. 338
Kamada et al. 2021, Icarus vol. 368
Kerber et al. 2015, Icarus vol. 261
Kite 2019, Space Science Reviews vol. 215
Kitzmann 2016, ApJ letters vol. 817
Kopparapu et al. 2014, ApJ letters vol. 787
Mondelain et al. 2021, JQSRT vol. 260
Ramirez et al. 2014, Nature Geoscience vol. 7
Ramirez & Kasting 2017, Icarus vol. 281
Ramirez 2019, RNAAS vol. 3
Ramirez et al. 2020, JGR Planets vol. 125
Steakley et al. 2019, Icarus vol. 330
Tian et al. 2010, EPSL vol. 295
Turbet & Tran 2017, JGR Planets vol. 122
Turbet 2018, PhD manuscript, Sorbonne University
Turbet et al. 2019, Icarus vol. 321
Turbet et al. 2020a, Icarus vol. 335
Turbet et al. 2020b, Icarus vol. 346
Turbet & Forget 2021, eprint arXiv:2103.10301
Wordsworth et al. 2010, Icarus vol. 210
Wordsworth et al. 2013, Icarus vol. 222
Wordsworth et al. 2016, AREPS vol. 44
Wordsworth et al. 2017, GRL vol. 44
Wordsworth et al. 2021, Nature Geoscience vol. 14

Figure 1 – CO2-H2 collision-induced absorption (in units of 10-5 cm-1 amagat-2) as a function of wavelength (in
micron units), and at 296K. The upper panel gives a global representation, while the lower panel offers a zoom
in the 2.1-2.55 µm spectral region. The black line corresponds to a semi-empirical calculation (Turbet et al.
2020b, Mondelain et al. 2021). Laboratory measurements are taken from Turbet et al. 2020b (in red), Mondelain
et al. 2021 (in magenta) and Fakhardji et al. 2022 (in brown). For reference, we added the N2-H2 CIA from
HITRAN (used in Ramirez et al. 2014) and the CO2-H2 CIA calculated in Wordsworth et al. 2017. We also added
the rough locations of the main CO2 infrared windows. We did not add the measurements of Godin et al. 2020
which exhibit much larger error bars.

Figure 2 - CO2-H2 collision-induced absorption (in units of 10-5 cm-1 amagat-2) as a function of wavelength (in micron units), and calculated for different temperatures (100, 300 and 500K). For reference,
we added the rough locations of the main CO2 infrared windows.

Figure 3 - Annual global mean surface temperatures (K) as a function of H2 mixing ratio, for two different CO2
partial pressures (solid lines = 2 bar; dashed lines =1 bar), and four different types of simulations. These include 1D simulations (black) based on Turbet et al. (2020b), 3-D simulations from Turbet & Forget 2021 assuming a dry
(i.e., low water content) planet (brown), a planet with a small 100 m GEL ocean (orange), and a planet with a large
550 m ocean (blue). For reference, we also added the 1-D results from Ramirez et al. 2014 (green) and Wordsworth
et al. 2017 (pink).

Figure 4 – On left, map of the cumulated annual liquid water runoff (Turbet & Forget 2021) calculated for a dense
CO2+H2 atmosphere (2bar of CO2, 15% of H2). In this simulation, almost the entire water reservoir (a few meters
global equivalent layer) is trapped in liquid form in the impact crater lakes, due to the (warm) adiabatic cooling effect
(Turbet & Forget 2021). On right, map of the observed distribution of valley networks (Bouley et. al. 2016). Both maps
were computed using the pre-True Polar Wander (TPW) topography of Bouley et al. 2016 (present-day topography,
but without Tharsis and all the younger volcanic features; and with a 20° TPW rotation).