Evaluating Atmospheric and Surface Drivers for O₂ Variations at Gale Crater as Observed by MSL SAM

Atmospheric photochemistry comprises the set of chemical reactions involving atmospheric species that is primarily driven by solar radiation. Generally, atmospheric photochemistry at Mars tends not to be spatially localized. The photochemistry that occurs at some local volume of the atmosphere would be determined by the composition and temperature of the volume, as well as the incident solar radiation flux. All these quantities vary gradually over the Mars atmosphere, mediated by the relatively short transport and mixing timescales. Simulations using global circulation models (GCMs) found both the vertical and horizontal mixing timescales to be less than 100 sols (e.g., Barnes et al. 1996), except over the polar regions at altitudes 20–40 km, where the timescales can be higher than 200 sols (Waugh et al. 2019). Given Curiosity's location in Gale crater at −5° N, these air masses, restricted by the polar vortex, would not have a significant influence on the ambient atmosphere around Curiosity and thus are not relevant to this study. Although the relief of ∼6 km at Gale crater has some effects on the local atmospheric circulation (e.g., Rafkin et al. 2016), the high eddy diffusivities up to 10⁷ cm² s⁻¹ (e.g., Taylor et al. 2007; Pathak et al. 2009) and the growth of the planetary boundary layer to several kilometers in the daytime (Guzewich et al. 2017; Fonseca et al. 2018) would have facilitated mixing of the atmosphere between the inside and outside of the crater. Modeling of the circulation within the crater also found fast mixing timescales of 1 sol (Pla-García et al. 2019). With the transport and mixing timescales being shorter than the 100 sol timescale of the O₂ variations, the O₂ variations observed by Curiosity at Gale crater and any photochemical process driving them are likely not localized to Gale crater, but rather part of a larger global phenomenon.

With this assumption, we use a 1D globally averaged coupled ion–neutral photochemical model to investigate oxygen photochemistry in the Martian atmosphere. The details for this model have previously been described in Lo et al. (2021). Spanning the surface to 240 km altitude at a resolution of 1 km, this model calculates the number densities of the various species and the rates of the various reactions at steady state using an exhaustive reaction list with the most updated reaction rates and cross sections, making it well suited for a comprehensive investigation of the effects of photochemistry on the surface O₂ VMR. While the use of a 1D photochemical model offers efficiency in our search for any potentially important but previously ignored reaction in its exhaustive reaction list and our investigation into the effect of different boundary conditions (Section 3.2), many phenomena that are relevant and important to our problem cannot be modeled by a 1D photochemical model. To mitigate these limitations, we use average temperatures and number densities of CO₂, N₂, Ar, and H₂O from the Mars Climate Database (MCD, version 5.3) at different seasons and solar activity levels for our model inputs. The MCD is a collection of results from the Mars Planetary Climate Model (PCM), developed at the Laboratoire de météorologie dynamique (LMD) of the French Centre national de la recherche scientifique (CNRS)  (Forget et al. 1999; Millour et al. 2018). Results from the MCD have been extensively validated using available observational data and span the altitude range of the photochemical model, allowing for ease of its incorporation as inputs. Separate temperature and number density profiles are constructed for different seasons and solar activity levels. For seasons, a set of aphelion profiles based on MCD results from L_S = 60° to 90° (aphelion occurs at L_S = 71°) and a set of perihelion profiles based on MCD results from L_S = 180° to 330° are constructed to account for the dependence of the photochemical processes on the highly varying solar radiation flux from the eccentricity of the Martian orbit. The different solar activity levels are set through the MCD "solar maximum" and "solar minimum" climatology scenarios. In addition to being able to describe 3D circulation, the PCM models a wide variety of phenomena not in the photochemical model, such as dust and ice aerosols, CO₂ ice formation and sublimation on the ground and in the atmosphere, and the water cycle with modeling of cloud microphysics. These phenomena are key to an accurate description of the temperatures and H₂O abundance, important variables for controlling oxygen photochemistry.

Figure 2 shows the main reactions that make up oxygen photochemistry at Mars, as elucidated from the photochemical model. Although we have included a maximal set of oxygen reactions available in the literature in our photochemical model, many of these reactions have turned out to be insignificant, and the overall picture we obtained is consistent with previous results from other photochemical models (e.g., Atreya & Gu 1995; Krasnopolsky 2010; Viúdez-Moreiras et al. 2020b). Ultimately, O₂ is sourced from the photodissociation of CO₂ and H₂O by solar ultraviolet (UV) radiation. Based on the model, the highest rates for O₂ production come from seven reactions:

$$\begin{eqnarray}&&2{\rm{O}}+{\rm{M}}\to {{\rm{O}}}_{2}+{\rm{M}}\end{eqnarray} \tag{ P1 }$$

$$\begin{eqnarray}&&\mathrm{OH}+{\rm{O}}\to {{\rm{O}}}_{2}+{\rm{H}}\end{eqnarray} \tag{ P2 }$$

$$\begin{eqnarray}&&{\mathrm{HO}}_{2}+{\rm{O}}\to {{\rm{O}}}_{2}+\mathrm{OH}\end{eqnarray} \tag{ P3 }$$

$$\begin{eqnarray}&&2{\mathrm{HO}}_{2}\to {{\rm{H}}}_{2}{{\rm{O}}}_{2}+{{\rm{O}}}_{2}\end{eqnarray} \tag{ P4 }$$

$$\begin{eqnarray}&&{\mathrm{NO}}_{2}+{\rm{O}}\to {{\rm{O}}}_{2}+\mathrm{NO}\end{eqnarray} \tag{ P5 }$$

$$\begin{eqnarray}&&{{\rm{O}}}_{3}+h\nu \to {{\rm{O}}}_{2}+{\rm{O}}\end{eqnarray} \tag{ P6 }$$

$$\begin{eqnarray}&&{{\rm{O}}}_{3}+{\rm{H}}\to {{\rm{O}}}_{2}+\mathrm{OH}.\end{eqnarray} \tag{ P7 }$$

Zoom In Zoom Out Reset image size

Here M represents a third body, typically CO₂ at Mars, required for the conservation of momentum in the reaction. The majority of O₂ loss occurs via three reactions:

$$\begin{eqnarray}&&{{\rm{O}}}_{2}+h\nu \to 2{\rm{O}}\end{eqnarray} \tag{ L1 }$$

$$\begin{eqnarray}&&{{\rm{O}}}_{2}+{\rm{H}}+{\rm{M}}\to {\mathrm{HO}}_{2}+{\rm{M}}\end{eqnarray} \tag{ L2 }$$

$$\begin{eqnarray}&&{{\rm{O}}}_{2}+{\rm{O}}+{\rm{M}}\to {{\rm{O}}}_{3}+{\rm{M}}.\end{eqnarray} \tag{ L3 }$$

Rate coefficients for the above reactions are provided in Table B1 in Appendix B. For ease of discussion, we classify the seven production reactions and three loss reactions into four groups. The first group, which we shall refer to as "CO₂ photodissociation," comprises reactions (P1) and (L1). The second group of "HO_x chemistry" comprises (P2), (P3), (P4), and (L2), and the third group of "NO_x chemistry" comprises (P5). The fourth group of "O₃ photochemistry" comprises (P6), (P7), and (L3).

The first group of "CO₂ photodissociation" is so named because it comprises the channel for the formation of O₂ from CO₂. O is primarily produced from the photodissociation of CO₂, either directly or through a two-step process involving CO. This O is then rapidly converted into O₂ through (P1), with the reverse reaction of (L1) maintaining a small amount of O and preventing O₂ from increasing indefinitely. These two reactions of (P1) and (L1), together with the CO₂ photodissociation reaction, are among the oldest known reactions in the Martian atmosphere, and they set the stage for the scientific problem of the "stability of the Martian atmosphere." The fast (P1) versus the slow spin-forbidden CO + O + M → CO₂ + M (reaction constants k at 200 K of ∼10⁻³³ cm⁶ molecule⁻² s⁻¹ and ∼10⁻³⁶ cm⁶ molecule⁻² s⁻¹, respectively; Tsang & Hampson 1986) would result in the accumulation of substantial CO and O₂ in the atmosphere, and at a ratio of 2:1.

The observed quantities of CO and O₂ at Mars are 2 orders of magnitude smaller than predicted by the "CO₂ photodissociation" reactions alone, and in a 1:2 ratio rather than the predicted 2:1. The "HO_x chemistry" reactions were proposed as a solution. Sourced from the photodissociation of H₂O, the HO_x species (H, OH, and HO₂) form two catalytic cycles regenerating CO₂ from CO, O, and O₂. The first cycle, proposed by McElroy & Donahue (1972), comprises

$$\begin{eqnarray*}&&\mathrm{CO}+\mathrm{OH}\to {\mathrm{CO}}_{2}+{\rm{H}}\end{eqnarray*}$$

$$\begin{eqnarray}&&{{\rm{O}}}_{2}+{\rm{H}}+{\rm{M}}\to {\mathrm{HO}}_{2}+{\rm{M}}\end{eqnarray} \tag{ L2 }$$

$$\begin{eqnarray}&&{\mathrm{HO}}_{2}+{\rm{O}}\to {{\rm{O}}}_{2}+\mathrm{OH},\end{eqnarray} \tag{ P3 }$$

with the net effect of converting CO and O into CO₂. The second cycle, proposed by Parkinson & Hunten (1972), comprises

$$\begin{eqnarray*}&&\mathrm{CO}+\mathrm{OH}\to {\mathrm{CO}}_{2}+{\rm{H}}\end{eqnarray*}$$

$$\begin{eqnarray}&&{{\rm{O}}}_{2}+{\rm{H}}+{\rm{M}}\to {\mathrm{HO}}_{2}+{\rm{M}}\end{eqnarray} \tag{ L2 }$$

$$\begin{eqnarray}&&2{\mathrm{HO}}_{2}\to {{\rm{H}}}_{2}{{\rm{O}}}_{2}+{{\rm{O}}}_{2}\end{eqnarray} \tag{ P4 }$$

$$\begin{eqnarray*}&&{{\rm{H}}}_{2}{{\rm{O}}}_{2}+h\nu \to 2\mathrm{OH},\end{eqnarray*}$$

with the net effect of converting CO and O₂ into CO₂.

From the perspective of O₂, HO_x chemistry introduces H₂O as a second ultimate source of O₂ through (P2), with a column-integrated rate ∼85% that of (P1) at aphelion and a much larger 7–8 times that of (P1) at perihelion when the H₂O abundance is significantly higher. The first cycle, with 10 times the rate of the second cycle (e.g., Nair et al. 1994; Krasnopolsky 2010), is net neutral with respect to O₂. Net O₂ loss with the breaking of the O–O bond can occur through the second cycle with H₂O₂ photodissociation, or earlier with HO₂ photodissociation or reaction with H after its primary production reaction of (L2). Collectively, these three reactions make up ∼30% of the breaking of the O–O bond, while (L1) makes up ∼50% (Lefèvre & Krasnopolsky 2017). The dominance of (P3) over (P4) also results in the preferential consumption of O over O₂, tipping the equilibrium CO:O₂ ratio toward higher O₂ .

NO_x chemistry comprises a cycle converting O into O₂ through reaction with NO and NO₂. NO, produced from the reaction of CO₂ with N (produced from N₂ photodissociation at ∼120 km), first reacts with O to form NO₂, and further reaction of NO₂ with O produces O₂ and regenerates NO through (P5). The column-integrated rate for (P5) is about 30% that of (P2) during the perihelion season, and about 10% during aphelion. NO₂ can also be produced from reaction of NO with HO₂ (this constitutes the remaining ∼20% of O–O bond breaking; Lefèvre & Krasnopolsky 2017), but with the subsequent production of O₂ through (P5) there is no net O₂ loss through this channel.

Finally, O₃ photochemistry comprises the rapid interconversion between O₂ + O and O₃ via (P6) and (L3) and conversion of O₃ back into O₂ with OH through (P7). All these reactions preserve the O–O bond and are together net neutral with regard to O₂. As a result of this and the significantly higher rate of (P6) and (L3) over the other O₂ formation and loss reactions, we can effectively treat each O₃ molecule as the combination of an O₂ molecule and an O atom for the purposes of bookkeeping the O₂ in the Martian atmosphere.

Understanding how the various O-containing species are linked to O₂ through photochemical reactions allows us to examine the photochemical drivers behind the observed O₂ variations. From the VMRs of the various O-containing species provided in Table 1, we find that only CO₂, CO, and H₂O have sufficiently high VMRs to function as potential oxygen reservoirs for driving the O₂ variations. An important point to note is that although the atmospheric H₂O reservoir has a maximum VMR of only 200 ppmv, there is significant exchange with the surface H₂O reservoirs, and thus the amount of available oxygen can actually be higher. Indeed, sublimation and condensation of frost at the higher latitudes (particularly the north polar cap) are a significant driver for variations of atmospheric H₂O on the seasonal timescale (e.g., Montmessin et al. 2017). This is handled in the photochemical model through fixing the H₂O number density profile, such that the atmospheric H₂O reservoir is never depleted and the steady-state number densities and reaction rates are calculated based on this fixed density profile. For O, even though it has a somewhat high VMR at 135 km altitude, it is important to remember that the overall density of the atmosphere decreases with altitude. Taking into account the difference in pressures between 135 km and the surface, the actual amount of O from 135 km will correspond to a much smaller (and insufficient) VMR if brought down to the surface. This is also reflected in the lower column-averaged VMR, which is strongly weighted toward the surface in its calculation. While the abundances of O, OH, HO₂, H₂O₂, NO, and NO₂ are too low for them to function as oxygen reservoirs, these trace species can act as bridges connecting O₂ to the CO₂, CO, and H₂O reservoirs.

Even if the CO₂, CO, and H₂O reservoirs contain more oxygen than the O₂ variations, not all the oxygen in these reservoirs will be liberated to eventually form O₂. From Figure 2, we can see that photodissociation is the key gatekeeper step in the conversion of CO₂, CO, and H₂O into O₂, with O and OH functioning as intermediaries. Since the observed O₂ variations are deviations from an equilibrium VMR of ∼1700 ppmv, it would be the variations in the photodissociation rates over time, rather than the total rates, that would provide the additional release of oxygen to drive O₂ increases above its equilibrium value.

Photodissociation of CO₂ can produce O in the ground (³ P) or excited (¹ D) state, with the combined rate peaking at ∼30 km altitude in our dust-free photochemical model. The preponderance of CO₂ in the atmosphere means that CO₂ photodissociation rates depend mostly on the solar UV flux, and thus the 40% increase in solar UV flux from aphelion to perihelion translates into a comparable increase of 50% in the column-integrated rate (Table 2). High solar activity also increases rates by about 10% compared to low solar activity. If we assume the CO₂ photodissociation rate to vary sinusoidally over the Martian year between the maximum and minimum values for each solar activity scenario and all liberated O from this variation to be immediately converted into O₂, we will find O₂ variations with a column-integrated amplitude of ∼4 × 10¹⁸ O₂ molecules cm⁻² for both solar activity scenarios. If we further assume that these variations are well mixed down from the photodissociation peak at 30 km to the surface with a 10 km scale height, then we would get an increase of ∼4 × 10¹² molecules cm⁻³ in the surface O₂ number density. Even with the two preceding assumptions that would maximize the variations in surface O₂, we get a value that is more than an order of magnitude smaller than our requirement of 10¹⁴ molecules cm⁻³.

Photodissociation of CO occurs high in the atmosphere, with rates peaking at ∼140 km altitude and dropping to zero below 90 km. Overall rates and their variation with season are much lower compared to CO₂ photodissociation (Table 2). Applying the same treatment as in CO₂ photodissociation, we find that increased CO photodissociation rates can only give rise to a very small increase of 10⁸ molecules cm⁻³ in O₂ surface number density, making it unlikely that CO photodissociation would play a significant role in driving the observed O₂ variations.

With a peak at ∼50 km altitude, H₂O photodissociation rates vary significantly with the seasons (Table 2), driven by the highly varying H₂O abundance (Table 1) arising from exchange with the surface reservoirs. Nonetheless, these seasonal variations are not sufficient, providing an increase of only 10¹¹ molecules cm⁻³ in the surface O₂ number density under the two maximalist assumptions.

Just as variations in the photodissociation rates of CO₂, CO, and H₂O could drive O₂ variations by controlling the release of oxygen, variations in the rates of the various O₂ loss processes can do the same by taking O₂ out of the atmosphere. Ultimate loss of O₂ occurs through O₂ photodissociation into O in (L1) and the loss of HO₂ after (L2). While most HO₂ is regenerated into O₂ through (P3), a fraction is lost through photodissociation, reaction with itself ((P4), followed by photodissociation of the H₂O₂ product), and reaction with H.

O₂ photodissociation occurs primarily at the surface. Seasonal variations in its rate are too small, translating to a variation of only 10¹¹ molecules cm⁻³ under the two assumptions.

The HO₂ loss reactions of photodissociation and reaction with H both peak at 20–30 km altitude, driven primarily by the HO₂ density peak there. We find that the maximum O₂ variation from HO₂ photodissociation is 10¹⁰ molecules cm⁻³, while those from reaction with H to form OH and to form H₂O and O are 10¹² and 10¹⁰ molecules cm⁻³, respectively. HO₂ can also react with itself to form H₂O₂, which then photodissociates with a peak rate also at 20–30 km altitude. Maximum seasonal variation from this channel corresponds to 10⁸ molecules cm⁻³.

Altogether, the seasonal variability of the production and loss channels can only account for O₂ variations that are an order of magnitude smaller than observed. In addition to the maximalist assumptions that all liberated O is converted into O₂ and the O₂ variations are confined to the bottom 10 km of the atmosphere, the photodissociation rates above are calculated assuming an aerosol-free atmosphere. In reality, aerosols from dust and condensates can increase atmospheric UV opacities, reducing photodissociation rates. This effect is particularly significant below ∼40 km altitude and is stronger during the perihelion season (e.g., Montmessin et al. 2006; Määttänen et al. 2013), thus countering the increase in the solar UV flux then. The UVISMART radiative transfer model, which includes these aerosol effects and has also been validated with in situ data from Curiosity, found that UV irradiance at Gale crater generally varies throughout the year within a band of ±15% (Viúdez-Moreiras 2021), smaller than the 40% used in the photochemical model. With their peak rates occurring within the aerosol haze layer, photodissociation of CO₂, O₂, HO₂, and H₂O₂ would be significantly affected by the inclusion of aerosol effects. Their rates would vary less with the seasons, making them even less viable drivers for the observed amplitudes in the O₂ variations.

Thus, even though there are oxygen reservoirs in the Martian atmosphere with sizes larger than the magnitude of the O₂ variations, the processes linking these reservoirs to O₂ do not vary enough to produce sufficiently large variations about the equilibrium O₂ VMR in the required time frame. The O₂ variations cannot be driven by atmospheric photochemistry alone, and we need to look toward surface processes for an answer.

Before we do so, however, a caveat has to be noted. The photochemical stability of the Martian atmosphere is still not well understood today, and this poor understanding shows up as very low modeled values for long-term equilibrium CO and O₂ VMRs. A common "solution" for the low O₂ VMRs in 1D photochemical models is to introduce a surface flux through the lower boundary condition, either directly with an appropriate magnitude to match the observations or indirectly by setting the abundance (e.g., Atreya & Gu 1995). This O₂ surface flux seems plausible, but justification has generally been poor in the instances where it has been invoked. We will examine the processes that can contribute to this surface flux later in Section 3.2. The same is not plausible for CO, however, giving rise to the "CO problem," where modeled CO VMRs are as much as 7 times lower than the observations. Lefèvre & Krasnopolsky (2017) provides an excellent exposition of the problem. GCMs with photochemistry, such as PCM (Forget et al. 1999) and GEM-Mars (Neary & Daerden 2018), generally avoid the O₂ and CO problems by picking an initial atmosphere consistent with the observations and then evolving the atmosphere while subjecting O₂ and CO to only transport effects, or taking advantage of the long photochemical lifetimes of the two species and stopping the models before the simulation time becomes too long and instabilities set in. For the results we have reported earlier from our photochemical modeling, we have set the O₂ surface boundary condition to zero flux. Particularly for an investigation of O₂ VMR variations effectively at the surface, setting a nonzero surface flux would overwhelm any effects from the atmosphere's response to the different solar fluxes and H₂O content during different seasons, and thus would be contrary to our goal of investigating photochemical drivers. Our calculated O₂ VMRs are, as a result, more than an order of magnitude smaller than the observed values over almost all altitudes. These low modeled O₂ abundances could be due to the actual O₂ production being faster (perhaps from larger photodissociation cross sections and/or missing reactions) and/or actual O₂ loss being slower (from smaller O₂ photodissociation cross sections). Our conclusion here that atmospheric photochemistry cannot be the sole driver of the surface O₂ VMR variations observed by SAM has to be accepted with the important caveat that we may still be missing key pieces in our understanding of Martian atmospheric photochemistry.

Before we examine what specific processes at the surface could drive the atmospheric O₂ VMR variations that SAM has observed just above the surface, we shall first investigate the effects of a general oxygen flux out of the surface while staying agnostic on the exact mechanism of this release. The oxygen release is not necessarily in the form of O₂; rather, some other O-containing species can be photochemically converted to O₂ after its release into the near-surface atmosphere. We can see from Figure 2 that this could be the case for O, OH, HO₂, and H₂O₂. We investigate the significance of such releases using our photochemical model, this time by setting fluxes of these species individually at various magnitudes as the lower boundary condition, rather than setting them to be zero as in our earlier investigation. We found the effect of these surface fluxes to be minimal, strongly suppressed by photochemistry. When compared against the zero surface flux atmospheres, only the particular species with the surface source flux exhibits a significant increase in abundance (except in the case of HO₂, where H₂O₂ also increased), and this is restricted to the bottom few altitude bins of the model. O₂ abundances remain relatively unchanged.

Setting the O₂ surface flux gives a different picture, however. With a flux of 5 × 10¹⁰ cm⁻² s⁻¹, we are able to obtain an O₂ VMR of 1900 ppmv at the surface under the perihelion low solar activity scenario. While the average O₂ level in the Martian atmosphere can be maintained by this baseline surface flux, a significantly higher flux above this baseline value would be required to supply the amount of O₂ in the observed O₂ increases. We recall the minimum requirement of 10⁷ O₂ molecules cm⁻³ s⁻¹ that we have calculated earlier. Assuming that all released O₂ is confined to and well mixed within the bottom 10 km of the atmosphere, this will translate into the requirement of a surface flux of 10¹³ cm⁻² s⁻¹. This inclusion of a surface flux does not have a significant impact on the O₂ photochemical lifetime, however, and it remains long at 5 × 10⁸ s. Thus, while the addition of a surface source would solve the problem of low modeled O₂ abundances in the atmosphere, additional loss processes are also needed to reduce the O₂ lifetime to the 10⁷ s timescale of the variations.

With this new understanding, we can narrow our investigation to only releases of O₂ and not of other O-containing species. We apply a similar strategy to the one we used earlier investigating the photochemical drivers: identify sufficiently large oxygen reservoirs at the surface or in the shallow subsurface, as well as processes that enable the exchange of oxygen between these reservoirs and the atmosphere. Oxygen can be chemically or physically sequestered in these reservoirs and released into the atmosphere through chemical decomposition or breakdown of the physical structure trapping the O₂.

3.2.1. Perchlorates and Chlorates

3.2.2. Brines

3.2.3. H₂O₂

3.2.4. Airborne Dust

In the original version of Trainer et al. (2019), which provided the first introduction and analysis of the data set for this study, there was an unfortunate coding error in the calculation of the correlation between the atmospheric O₂/Ar ratio and the daily maximum relative humidity as measured by REMS. Although the correction only slightly improved the ${\chi }_{\nu }^{2}$ value of a linear fit between the two variables from 3.34 to 3.06, the corrected Figure S8(h) (reproduced in Figure D1(a)) now appears to indicate some relationship between the normalized O₂ and H₂O that is perhaps more complex than a simple linear function. Here we will be exploring any potential relationship between O₂ and H₂O further. Figure 3 shows the traces of a variety of H₂O-related quantities from REMS plotted over the normalized O₂ VMR from Figure 1.

Zoom In Zoom Out Reset image size

REMS makes extensive measurements of the relative humidity of the immediate atmosphere around Curiosity, at ∼1.6 m height from the ground arising from the position of REMS on the rover's mast (Gómez-Elvira et al. 2012). This relative humidity can then be used in conjunction with REMS air temperature measurements to determine the water vapor content (or H₂O VMR). Typically, relative humidity increases through the Martian night, reaching a maximum at around 0400–0600 local mean solar time (LMST; Martínez et al. 2017). However, these high nighttime relative humidities actually correspond to the lowest absolute H₂O VMRs over a Martian sol. In the day, the H₂O VMR is much higher, but air temperatures are also higher, resulting in very low daytime relative humidities with uncertainties often higher than 100% (Savijärvi et al. 2015; Martínez et al. 2016). Because of this, we will be using the more reliable nighttime relative humidity values for this study. We do not believe that this would introduce a significant bias into our investigation of a relationship between O₂ and H₂O. The O₂ changes over a seasonal timescale are unlikely to be driven by processes with a diurnal timescale, and strong convective mixing in the day would also have ensured efficient exchange from the near-surface atmosphere to as high as 4 km altitude on the diurnal timescale (e.g., Savijärvi et al. 2015). Furthermore, the majority of the SAM measurements were taken at night (Table A1), and thus the nighttime REMS measurements would also be more relevant for catching any unexpected relevant processes.

Generally, the seasonal variability of the daily maximum relative humidity is controlled mostly by near-surface temperatures (Martínez et al. 2017; Savijärvi et al. 2019a). From Figure 3, we find the relative humidity to increase as temperatures start to fall at L_S ∼ 215°, reaching a peak at L_S ∼ 90° when temperatures are the lowest. Although the normalized O₂ VMR also increased over a similar period, the two quantities do not follow each other closely (we calculated a correlation coefficient of 0.47). At the beginning of MY 32, the O₂ VMR was sustained to at least L_S ∼ 250° (and possibly even the end of the year) despite the decrease in relative humidity after L_S ∼ 90°. In MY 33, the O₂ measurements mostly match the relative humidity levels, except for the highest measurement on sol 1319, which continued the increasing O₂ trend despite significantly lower relative humidity. If relative humidity is to be a driver for the O₂ variations, it would do so by controlling the physical processes exchanging oxygen between the atmosphere and the surface or subsurface reservoirs, such as the evaporation and deliquescence of brines. Generally, such a physical process that releases O₂ into the atmosphere will also release (rather than consume) H₂O, and thus we should expect that a higher relative humidity will inhibit the release of both H₂O and O₂. This is contrary to the observed increase of O₂ with relative humidity. Furthermore, the association between only increases in O₂ and increases in relative humidity, but not between the decreases of either variable, suggests that relative humidity per se can at most provide a trigger for the process of increasing atmospheric O₂ abundance through its own increase. These two considerations make it difficult for relative humidity to be a direct driver for the process behind the O₂ variations. Neither is the near-surface temperature, which drives the relative humidity variations, a likely contributor to the O₂ variations—Trainer et al. (2019) did not find any apparent relationship between the O₂/Ar ratio and various temperature variables.

In both MY 33 and 34, the (daily minimum) H₂O VMR started to increase at L_S ∼ 60°, reaching a maximum just before L_S ∼ 180°. This reflects the transport of H₂O sublimating from the north polar region to the equatorial region, where Curiosity is (e.g., Knutsen et al. 2022). The high VMR is then maintained until L_S ∼ 330°, possibly from isolation of the local circulation within Gale crater over the cooler months and exchange via a diurnal cycle of adsorption and desorption with the ground (Savijärvi et al. 2019b). Tracing through Figure 3, we find that the O₂ VMR does not follow this H₂O VMR (the correlation coefficient is even lower at 0.08). The low H₂O VMR over the first half of MY 32 spans both high and low O₂. For MY 33, low H₂O in the first half is accompanied by high O₂, while high H₂O in the second half is accompanied by lower O₂. H₂O VMR would control the rate of chemical processes, such as the previously discussed formation of oxychlorines with O₃.

We have also plotted in Figure 3 the rate of change of H₂O VMR, which may suggest O₂ production or loss mechanisms that involve a change in the H₂O VMR. These changes in the H₂O VMR appear to be more strongly correlated with O₂ (correlation coefficient of 0.58), particularly apparent over sols 250–700 and 1100–1500. It is highly unlikely that the rate of change of H₂O VMR per se would be a control for a process driving O₂ variations, and the observed correlation is more likely to point instead to an indirect causal relationship between the two species. For example, the production of O₂ from the decomposition of H₂O₂ released from a subsurface reservoir would also produce H₂O, resulting in a correlated increase in the VMR of both species. With this perspective, it may perhaps be more physically meaningful to compare the rate of change of H₂O VMR with the rate of change of O₂ VMR (which would point to a process that simultaneously liberates H₂O and O₂), but the sparseness of the SAM O₂ data set makes it difficult to calculate the gradient reliably.

Trainer et al. (2019) noted a potential link through their Figure 13(a) between the O₂ VMR and atmospheric dust opacities as measured by MSL Mastcam (Smith et al. 2016). Specifically, the O₂ variations in MY 33 appear to inversely follow the variations in dust opacity closely. However, this is not the case in MY 32, when the significant increase in opacity at L_S ∼ 150° and L_S ∼ 250° did not translate into any significant change in the O₂ from the measurements spanning sols 638–830. Thus, we do not find sufficient evidence for a strong link between the normalized O₂ and dust opacities from the data.

Over sols 45–1869 spanned by our data set, the Curiosity rover traveled from its landing location at Bradbury Landing to Vera Rubin Ridge (VRR), covering a total distance of 17.6 km. Elevation ranged between −4521 and −4172 m relative to the aeroid, generally increasing with time as Curiosity ascended Aeolis Mons (also informally referred to as Mt. Sharp). Here we will trace the SAM measurements along Curiosity's traverse over time in Gale crater to examine any relationships between the O₂ variations and the local geologic context where the measurements were taken. Figure C1 shows a map of the traverse, and we will also provide rover odometry readings corresponding to when the SAM atmospheric VMR measurements were made to provide some intuition for the spatial separation of the measurements. In short, we do not see strong indications of any relationship between the normalized O₂ VMRs and the changing local geology and mineralogy around Curiosity over its journey.

Geology exposed on the surface from Bradbury Landing to VRR is predominantly sedimentary, believed to be deposited mostly under subaqueous conditions (e.g., Grotzinger et al. 2015), but also to a smaller extent under subaerial conditions (e.g., the Stimson formation; Banham et al. 2018). Stratigraphically, the traverse can be divided into two, with the Bradbury group exposed on flat plains of Aeolis Palus over the first ∼9 km and the Mt. Sharp group as Curiosity started up the slopes of Aeolis Mons.

The Bradbury group comprises fine-grained sandstones, pebbly sandstones, and conglomerates (Vasavada et al. 2014). Cross-stratification in the sandstones and the low to moderate level of rounding of the pebbles point to deposition in a fluvial and fluvio-deltaic environment with a short transport distance (Williams et al. 2013; Grotzinger et al. 2015). Low chemical index of alteration values suggest minimal open-system chemical weathering, and compositional variability in the Bradbury group primarily arises from different mixtures of primary igneous minerals (Siebach et al. 2017), probably supplied by erosion of the crater wall (Grotzinger et al. 2015). Beds in the Bradbury group are approximately horizontal; thus, elevation can serve as a proxy for stratigraphic position. This also suggests that surface units are underlain by the same down-section units encountered at a lower elevation elsewhere along the traverse. Certainly, this assumption works best at shallow depths, but as discussed earlier, only the top few meters would be able to exchange effectively with the atmosphere from the thermal and diffusional perspectives and be relevant to our study.

The first measurement in our data set on sol 45 (odometer reading 294 m) was made halfway between Bradbury landing and the first major scientific stop at Yellowknife Bay, where Curiosity arrived on sol 125 at odometer reading 612 m, about 320 m later. Yellowknife Bay is a topographic depression (∼20 m below Bradbury landing), with exposed bright stratified rocks (Grotzinger et al. 2014). Thermal inertias measured at Yellowknife Bay are relatively high at 350–550 J m⁻² K⁻¹ s^−1/2 (Figure D3; Vasavada et al. 2017) and have been attributed to a greater extent of cementation of the bedrock (Grotzinger et al. 2014). In contrast with the majority of the Bradbury group, the Sheepbed mudstone at Yellowknife Bay is hypothesized to be deposited in a lacustrine environment (Grotzinger et al. 2014). SAM EGA experiments on the John Klein (JK) and Cumberland (CB) samples from the Sheepbed member measured drastically different oxycholorine contents (JK: 0.09 wt%; CB: 1.05 wt%; plotted in Figure D4) (Sutter et al. 2017). The measurement on sol 77, taken at odometer reading 490 m also on the way to Yellowknife Bay, did not show a significant change in the normalized O₂ VMR from the first measurement. Curiosity subsequently left Yellowknife Bay on sol 299 (odometer reading 762 m), heading southwest toward the Darwin and Cooperstown waypoints. SAM atmospheric measurements over this period from sol 278 to sol 434 (odometer readings of 727–3978 m) did not exhibit any significant excursions in the normalized O₂ VMR from its departure.

A significant increase in the O₂ VMR was observed between the measurements from sols 434 and 538, corresponding to a sub–1 km traverse between Cooperstown and Dingo Gap (odometer readings 3978–4966 m) and an elevation change from −4495 to −4487 m. Surface bedrock exposures and thermal inertias over this traverse between Cooperstown and Dingo Gap were not notably different from what had been previously encountered since departure from Yellowknife Bay.

Subsequently, normalized SAM O₂ VMRs remained relatively constant up to sol 830. Over the period spanned by these four measurements, the Curiosity rover did an extensive investigation at the Kimberley over sols 571–634 (Rice et al. 2017) and arrived at Pahrump Hills on sol 753 (odometer reading 9458 m). Rocks at the Kimberley are characterized by high potassium content (Siebach et al. 2017), but this is unlikely to have significant implications for atmospheric O₂. The SAM EGA experiment on the Windjana (WJ) sample measured intermediate oxycholorine content at 0.23 wt% (Figure D4; Sutter et al. 2017). Pahrump Hills marks the transition from the Bradbury group to the Mt. Sharp group, and the atmospheric measurement on sol 753 was made immediately on Curiosity's arrival. The Murray formation, at the base of the Mt. Sharp group and exposed at Pahrump Hills, is composed of predominantly mudstone, with some sandstone and conglomerate (Grotzinger et al. 2015). The Murray formation is believed to be deposited in a lacustrine environment at near-neutral pH (Grotzinger et al. 2015; Rampe et al. 2017), with significantly higher hematite content and a higher degree of open-system chemical weathering compared to the Bradbury group (Rampe et al. 2020a). SAM EGA analysis on the three samples from the Pahrump Hills outcrop (Confidence Hills (CJ), Mojave (MJ), and Telegraph Peak (TP)) found a low 0.05–0.10 wt% oxychlorine content (Figure D4; Sutter et al. 2017).

The increase in O₂ VMR over sols 830–1319 (odometer readings from 9722 to 12,948 m, elevations from −4457 to −4430 m) spans a long traverse covering Pahrump Hills, Emerson plateau (odometer reading 10,749 m), Namib and High Dunes in the Bagnold dune field (odometer reading 11,719 m), and Naukluft plateau (odometer reading 12,654 m). The Emerson and Naukluft plateaus are composed of the aeolian Stimson sandstone formation, which lies unconformably above the Murray formation. The sandstone contains highly spherical bimodally sorted grains in meter-thick crossbeds, pointing to an origin from the lithification of an ancient dune field (Banham et al. 2018). The Stimson formation also has a somewhat higher thermal inertia (410–440 J m⁻² K⁻¹ s^−1/2; Fraeman et al. 2016). While its bulk geochemical composition is similar to that of the Bradbury group, suggesting a common source region (Bedford et al. 2020), the degree of chemical alteration is limited to alteration halos around fractures and acidic leaching (Yen et al. 2017), lacking the phyllosilicates that can be found in the fluvial/lacustrine/deltaic Bradbury group (Bedford et al. 2020).

While transiting between the two Stimson exposures at the Emerson and Naukluft plateaus, Curiosity took a detour to approach and study the active Bagnold dune field over sols 1162–1254. The dune sands were found to be depleted in H₂O at ∼1% (Ehlmann et al. 2017; Sutter et al. 2017), and evolution temperatures suggest that the water resides in hydroxylated phases rather than as water adsorbed onto sediment grains. Nonetheless, oxychlorine content is moderate at 0.21 wt% (the Gobabeb (GB) samples; Figure D4). The minimal cementation of the dune sands also returned some of the lowest thermal inertia measurements (Figure D3; Vasavada et al. 2017).

The subsequent decrease in O₂ VMR over sols 1319–1457 (odometer readings 12,948–14,376 m, elevation −4430 to −4380 m) occurred as Curiosity left Naukluft plateau and traveled south toward its second encounter with the Bagnold dune field (Nathan Bridges Dune and Mount Desert Island). Over this period, Curiosity drove over the Murray formation, but the Stimson formation remains exposed as a capping unit for buttes and mesas in the vicinity (Fraeman et al. 2016). While the O₂ variations over sols 830–1457 match those of Curiosity's sojourn around the Stimson formation during sols 924–1469, it is unclear from a geologic perspective how it would have contributed to the elevated O₂ levels. Furthermore, Curiosity's detour to the Bagnold dune field over sols 1162–1254 brought it as far as ∼400 m from the nearest Stimson exposure. If proximity to the Stimson formation was a contributing factor to the elevated O₂ levels, then the sol-1252 measurement not showing a decrease (but in fact an increase from sol 1145) would indicate that this proximity effect would apply in places at least 400 m from the Stimson formation. If we now apply a 400 m radius to the beginning and end of Curiosity's encounter with the Stimson formation, we should find elevated O₂ for the entire period from sol ∼690 to sol ∼1510. There was, however, no elevation in the normalized O₂ levels for the observations on sols 753 and 830. SAM EGA on the three surface samples over this period (Oudam (OU), Marimba (MB), and Quela (QL)) did not find any oxychlorine content from lack of an O₂ peak (Figure D4; Clark et al. 2021).

After this final decrease in O₂ VMR, there are no further significant O₂ variations within our data set. Over the remaining duration of our data set, Curiosity arrived at Nathan Bridges Dune on sol 1602 (odometer reading 15,649 m), left the Bagnold dune field on sol 1662 (odometer reading 16,163 m), and arrived at Vera Rubin Ridge (VRR) on sol 1795 (odometer reading 17,409 m). The mineralogical and geochemical composition at this downwind location of the dune field is very similar to that at the upwind location at the Namib and High Dunes (Rampe et al. 2020a). A later derivatization experiment on the Ogunquit Beach sample from the Mount Desert Island ripple field detected N-containing organics and unidentified high molecular weight compounds (Millan et al. 2022). The VRR is composed of the Pettegrove and Jura members of the Murray formation (Edgar et al. 2020), characterized by strong absorption features at 530 and 860 nm suggestive of hematite (Fraeman et al. 2016). However, the mineralogical and geochemical composition of the VRR is very similar to that of the lower members of the Murray formation encountered earlier, and the strong hematite signal is due to grain size differences and/or the hematite occurring in phases with stronger absorptions (Rampe et al. 2020b; Frydenvang et al. 2020; Horgan et al. 2020; Jacob et al. 2020; Thompson et al. 2020).

Over Curiosity's traverse on the floor of Gale crater, DAN also made regular measurements of the hydrogen content in the top ∼1 m of the surface, where the rover is located (Litvak et al. 2008). As pointed out earlier, this also corresponds to the annual thermal skin depth, and thus the part of the shallow subsurface that we can expect to be able to exchange with the atmosphere to drive our O₂ variations. Variability in the measured WEH reflects surface compositional changes under Curiosity over its traverse, such as the low WEH measurements over sandy aeolian ripples (Tate et al. 2018, 2019). Figure D2 shows the WEH measured by DAN over the period spanned by our data set. We find that the relationship between DAN WEH and the normalized SAM O₂ VMR is weak in general. There is an increase in WEH between Cooperstown and Dingo Gap (sol ∼500), but the WEH has fallen back down by the time of the observation on sol 538 to a level similar to that around the observation on sol 434. Furthermore, there is a drop in WEH before the observation on sol 830, while the observation returned a level normalized O₂ VMR. In MY 33, although the measured WEH is somewhat lower than in MY 32, there is no clear trend in time that indicates a correlation or anticorrelation with the monotonic O₂ increase.

In the original publication of the SAM mixing ratio data in Trainer et al. (2019), it was noted that after sol 1000 the retrieval of the CO VMR was suspect:

Although we have measured the CO in each atmospheric experiment through MSL sol 1,711, measurements after sol 1,000 show significantly elevated signal at m/z 12 and therefore very high CO mixing ratios. The VMR for CO obtained in MY 32, near L_S 250° (MSL sol 830, TID 25232, Tables 1 and S1) shows the onset of this trend, in which the CO begins to diverge from the repeated seasonal trend in Ar (Figure 8). In addition to the derived CO mixing ratio, more than doubling from MY 32 to MY 33, the elevated measurements have not decreased or shown any seasonal modulation in MY 33 and MY 34, in contrast to the O2 measurement. This behavior is suspect, and at this time, a possible contamination or instrument effect cannot be ruled out. Especially because we know the m/z 12 signal to be highly sensitive to such effects (Franz et al. 2015), we are cautiously omitting the questionable observations from this paper. Those CO measurements thus require further investigation and will be reported at a later time.

It is important to note that the decrease in the confidence of the CO VMR derivation does not impact the O₂ measurement that was reported, and that is the cornerstone of the work presented herein. One of the most important factors to consider is that the detection of CO is marginal, given its low relative abundance compared to CO₂ and N₂, both of which produce mass fragments that interfere with the CO signal in the mass spectrometer.

The primary peak for O₂ is its molecular ion at m/z = 32, which is not impacted by the same factors as CO because (1) m/z = 32 is not a prevalent HC fragment (as compared to m/z = 12–14) and (2) it is the highest signal fragment for O₂ and there is only a minor contribution from CO₂.

Zoom In Zoom Out Reset image size

It can be seen in Figure E1 that the background signal for O₂ has been very stable throughout the mission, and there are no signs to indicate degradation of the instrument or changes in the performance at this mass channel. In contrast, the background for m/z = 12 shows more variability.

Figure E2 shows the full set of atmospheric measurements that have been conducted by the SAM QMS to date that are available on PDS. The time of year/season of the measurements is indicated with the asterisk. Measurements have continued from MY 34, although not at the same cadence as the prior campaign that led to the Trainer et al. (2019) paper. Other mission measurement priorities, specifically the drilling and organic analysis measurements by the SAM instrument suite, have reduced the cadence of the atmospheric measurements. In particular, there was a large gap from mid–MY 34 to mid–MY 35 to allow the instrument resources to be dedicated to the solid sample analysis, a top priority for mission science. Currently, the mission is in the middle of a campaign to take measurements through the end of MY 37.

Zoom In Zoom Out Reset image size

As noted in Trainer et al. (2019), high signal levels in the wet chemistry experiment on sol 1909 caused a shift in the sensitivity of the instrument detectors, requiring a change in the QMS electron multiplier gain setting. Mixing ratio determinations for atmospheric runs following that shift have indicated a change in the calibration ratios needed to properly derive the relative composition of the primary gases. In situ calibrations similar to that reported in Franz et al. (2017) have been performed on Mars, and the analysis of these data to derive new calibration constants is ongoing. At this time the team is not able to publish high-fidelity VMRs for this later data set that are of high enough confidence to support meaningful scientific interpretation.