Origin of Molecular Oxygen in Comets: Current Knowledge and Perspectives

Abstract

The Rosetta Orbiter Spectrometer for Ion and Neutral Analysis (ROSINA) instrument onboard the Rosetta spacecraft has measured molecular oxygen (O₂) in the coma of comet 67P/Churyumov-Gerasimenko (67P/C-G) in surprisingly high abundances. These measurements mark the first unequivocal detection of O₂ in a cometary environment. The large relative abundance of O₂ in 67P/C-G despite its high reactivity and low interstellar abundance poses a puzzle for its origin in comet 67P/C-G, and potentially other comets. Since its detection, there have been a number of hypotheses put forward to explain the production and origin of O₂ in the comet. These hypotheses cover a wide range of possibilities from various in situ production mechanisms to protosolar nebula and primordial origins. Here, we review the O₂ formation mechanisms from the literature, and provide a comprehensive summary of the current state of knowledge of the sources and origin of cometary O₂.

1 Introduction

While atomic oxygen is the third most abundant element in the Universe, its molecular form, O₂, has been particularly elusive to observers in various astrophysical environments. O₂ is expected to be present in the interstellar medium (ISM) in both its solid (ice) and gas phases. However, O₂ is a highly chemically reactive molecule with no electric dipole moment, which makes it hard to detect. Indeed, solid O₂ is yet to be detected in interstellar grains or in protoplanetary disks, and gas-phase O₂ has only been observed in two interstellar clouds with very low concentrations relative to H₂ (Orion: $\mbox{O}_{2}/\mbox{H}_{2} \approx 0.3\mbox{--}7.3 \times 10^{-6}$, and $\rho$ Oph A: $\mbox{O}_{2}/\mbox{H}_{2} \approx 5 \times 10^{-8}$; see e.g. Goldsmith et al. 2011; Chen et al. 2014; Larsson et al. 2007; Liseau et al. 2012). In the Solar System, O₂ has been detected on the icy moons of Jupiter and Saturn, where it is produced via photolysis and radiolysis followed by sputtering (e.g. Hall et al. 1995; Johnson 1996; Johnson and Quickenden 1997; Johnson and Jesser 1997; Cruikshank et al. 1998; Johnson et al. 2006; Bahr et al. 2001; Teolis et al. 2010), and from energetic particles accelerated in their magnetospheres (e.g. Spencer et al. 1995; Noll et al. 1997; Sieger et al. 1998). O₂ is also present in trace amounts in the atmosphere of Mars, and results from the complex photochemistry taking place in these environments (Barker 1972). Meanwhile, O₂ is the second most abundant atmospheric component at about 21% by volume in Earth’s atmosphere, but it is the only known planetary environment in which O₂ is biologically produced.

Before the Rosetta mission, comets in general seemed to be oxygen poor, which was expected based on the generally low levels of O₂ observed in interstellar and protoplanetary environments, and its high reactivity. However, with the unprecedented capabilities of the European Space Agency’s Rosetta mission came unprecedented discoveries. One of Rosetta’s most perplexing findings was the persistent detection of O₂ in the coma of comet 67P/Churyumov-Gerasimenko (hereafter: 67P/C-G) in surprisingly high abundances. These measurements were made by the Double Focusing Mass Spectrometer of the Rosetta Orbiter Spectrometer for Ion and Neutral Analysis (ROSINA/DFMS; Balsiger et al. 2007), and marked the first ever unequivocal detection of O₂ in a cometary environment. ROSINA/DFMS measured an O₂ concentration of 1–10% relative to water in the coma of 67P/C-G, with a mean value of $3.80 \pm 0.85\%$ (Bieler et al. 2015). The presence of O₂ in the coma of 67P/C-G was later independently confirmed by the Alice imaging spectrograph onboard Rosetta, with a somewhat larger mean relative column abundance of 25% (Keeney et al. 2017a). In addition, the recent reevaluation of the Neutral Mass Spectrometer (NMS) measurements from the Giotto flyby yielded data consistent with O₂ in the coma of comet 1P/Halley, with concentrations similar to those detected in the coma of 67P/C-G (Rubin et al. 2015b). Thus, the presence of cometary O₂ may not be unique to 67P/C-G, but rather it may be a common attribute of the otherwise dynamically different Jupiter-family (e.g. 67P/C-G) and Oort Cloud (e.g. 1P/Halley) comets.

The first ROSINA/DFMS measurements of O₂ in the coma of 67P/C-G were made over a six-month period between September 2014 and March 2015 at 10–30 km from the nucleus (Bieler et al. 2015). The measured local O₂ abundance relative to H₂O was persistently high over this period with a mean value of $3.80 \pm 0.85\%$, making O₂ the fourth most abundant species in the coma after H₂O, CO₂, and CO. Over this six-month time interval, O₂ was found to strongly correlate with H₂O, which led to the conclusion that the origin and/or release mechanism of O₂ and H₂O are linked. In addition, the analysis of Bieler et al. (2015) suggested that the O₂/H₂O ratio in the coma of 67P/C-G remained fairly constant, with a scatter of less than about 10%. Fougere et al. (2016) showed that the strong correlation with H₂O extended through February, 2016. Their modeling results also provided an O₂/H₂O production rate of 1–2%. While ROSINA/DFMS did not detect O₃ (ozone), it did measure the chemically related species H₂O₂ (hydrogen peroxide) and HO₂ (hydroperoxyl) with relative abundances of $\mbox{H}_{2}\mbox{O}_{2}/\mbox{O} _{2} \approx \mbox{HO}_{2}/\mbox{O}_{2} \approx 6 \times 10^{-4}$ (Bieler et al. 2015).

The remarkably high concentration of gaseous cometary O₂ contrasts with the remarkably low O₂ concentrations across the Universe. Because comets are leftover planetesimals whose accretion and growth took place in the protosolar nebula (PSN), they contain some of the best-preserved materials from before and during planet formation. Cometary nuclei likely agglomerated either from minimally-altered icy grains formed in the ISM, or from more processed grains directly condensed in the PSN. Hence, comets are often thought of as “cosmic freezers”. Whether O₂ is present in all or a multitude of comets, in just a few, or in 67P/C-G only, the production mechanism and origin of cometary O₂ would provide crucial information about the formation of our Solar System, and for interstellar-, disk-chemistry, and astrochemical models. Since its first detection in 67P/C-G, several hypotheses have emerged for the production and origin of cometary O₂. These hypotheses cover a wide range of possibilities from in situ production mechanisms to various primordial or protosolar nebula origins. The proposed formation mechanisms of cometary O₂ existing in the current literature are shown in Fig. 1.

Fig. 1

Proposed explanations for the origin (red and orange), formation mechanisms (purple), and trapping mechanisms (blue) of O₂ from the literature. Implications of the different scenarios to the ROSINA data, i.e. key measurements required to validate the different approaches, are listed in the black boxes. Formation mechanisms in the gray boxes are highly unlikely because of the affected layers would have been removed by erosion since 67P/C-G entered the inner Solar System. References: M16: Mousis et al. (2016b), T16: Taquet et al. (2016), B15: Bieler et al. (2015), M18: Mousis et al. (2018), YG17: Yao and Giapis (2017), D17: Dulieu et al. (2017), AB: Bouquet et al. (2018)

One possibility for the presence of molecular oxygen in the coma of 67P/C-G is that it was present before comet formation. In that case, the nucleus would have been agglomerated from O₂-rich icy grains formed in the ISM that did not undergo further chemical modification before and after being incorporated into the nucleus (primordial origin). If cometary O₂ is primordial, then the comet would have agglomerated in the coldest regions of the disk, where chemical reactions may be slow enough to preserve O₂ in cometary nuclei. Possible production mechanisms for primordial O₂ in interstellar ices include the radiolysis of H₂O ice by galactic cosmic rays (GCRs) in a low-density environment, such as a molecular cloud (Mousis et al. 2016b). Radiolysis of icy grains requires very large GCR fluxes, but the irradiation itself would stabilize the O₂ formed by the development of cavities in the H₂O ice matrix, where the O₂ molecules get trapped. Alternatively, oxygen atom recombination on the surfaces of interstellar grains in warm ($\sim20\mbox{--}30~\mbox{K}$) and dense ($\geq 10 ^{5}~\mbox{cm}^{-3}$) dark clouds could produce O₂ in the observed abundances (Taquet et al. 2016). The latter scenario requires warmer and denser conditions than typically found in dense clouds, but is in good agreement with the physical conditions in $\rho$ Oph A, as noted in Taquet et al. (2016). Both of these primordial scenarios require an intimate relationship between O₂ and H₂O; thus explaining the strong observed correlation between O₂ and H₂O. These observations must be taken in conjunction with other measurements pertaining to the origins of cometary H₂O. For instance, the elevated D/H ratio observed at 67P/C-G pre-inbound equinox (Altwegg et al. 2015) would be another constraint on the (primordial) formation scenario of 67P/C-G’s ice.

Gaseous O₂ could have also been formed and incorporated into icy grains in the PSN before being agglomerated into the nucleus. Luminosity outbursts (Taquet et al. 2016) or accretion bursts (Bieler et al. 2015) could have increased the temperature to exceed 100 K, aiding the formation of gas-phase O₂. Subsequent rapid cooling would then be needed to trap the O₂ in amorphous icy grains. In this case though, CO and N₂ (and other volatiles) should also be trapped in amorphous ice and correlate with H₂O, which was not the case for observations over a six-month period by Bieler et al. (2015). O₂ also readily traps in clathrates, but in that case it would be expected to strongly correlate with CO and even CO₂ instead of H₂O (Mousis et al. 2016b). In addition, O₂ may form crystalline ice in the PSN, but that would make the N₂ and Ar depletion problematic in 67P/C-G compared to the protosolar values (Mousis et al. 2016b). However, O₂ embedded in interstellar icy grains may remain stable even during a phase transition upon heating in the disk. While CO, Ar, and N₂ would get trapped in clathrates from the gas phase as the disk cools, O₂ would remain trapped from its initial incorporation into the H₂O matrix, maintaining its close association with H₂O (Mousis et al. 2016b). O₂ with an adsorption energy of 1660 K (8.6 kJ/mol) (Laufer et al. 2018; Smith et al. 2015) would be more strongly attached to the ice than N₂ (adsorption energy of 1155 K or 7.2 kJ/mol; Laufer et al. 2018; Smith et al. 2015) and CO (adsorption energy of 1420 K or 7.6 kJ/mol; Laufer et al. 2018; Smith et al. 2015). Ar, with an adsorption energy of 866 K or 7.5 kJ/mol (Laufer et al. 2018; Smith et al. 2015), is even more volatile. PSN formation scenarios do not necessarily require a strong correlation with H₂O, but require a correlation with CO, CO₂, and/or N₂ (Fig. 1). Most recently, Mousis et al. (2018) proposed that an alternative mechanism for O₂ formation is radiolysis of icy grains being cycled between the mid-plane and the upper layers of the disk multiple times. While this turbulent mixing would have allowed the icy grains to undergo stronger GCR irradiation while in the upper layers of the disk, and would explain a strong correlation between O₂ and H₂O, as they point out the timescale needed to make the O₂ in the abundances observed by ROSINA/DFMS exceeds the formation timescale of the comet.

O₂ may have also formed after nucleus accretion, either via radiolysis, or in situ. Past radiolysis in the Kuiper-belt seems unlikely to be the main source of O₂ on the comet because the affected layers would be removed by erosion since 67P/C-G entered the inner solar system, and current radiolysis would be indicated by variable O₂/H₂O driven by illumination (Bieler et al. 2015). However, endogenic radiolysis may contribute to the observed O₂ in the coma of 67P. In this scenario, radiation chemistry from an endogenic source, such as decaying radionuclides would produce O₂ homogeneously in the ice of the nucleus. While endogenic radiolysis is unlikely to have produced all the observed O₂ today, it may have significantly contributed to it (Bouquet et al. 2018). In situ O₂ formation is possible via water-group ions accelerated by the solar wind and energetic electrons hitting various oxidized surfaces and triggering Eley-Rideal (ER) reactions (Yao and Giapis 2017). Thus, O₂ may be formed on the surface of the nucleus, spacecraft, or the ROSINA/DFMS instrument itself (hyperthermal surface ionization). On the other hand, O₂ formation via the dismutation/disproportionation of H₂O₂ co-sublimating with H₂O was proposed based on laboratory experiments (Dulieu et al. 2017). The ER and the H₂O₂ dismutation/disproportionation reactions, as well as endogenic radiolysis require a strong correlation with H₂O at all times.

This paper reviews the possible O₂ formation mechanisms, as we currently understand them, and evaluates their strengths, weaknesses, and limitations based on the currently available, published Rosetta measurements. Sections 2 and 3 provide detailed reviews of pre-agglomeration and post-agglomeration O₂ formation mechanisms from the literature. A comprehensive summary of the current state of knowledge, as well as conclusions are given in Sect. 4. Finally, we suggest future key measurements and data analysis for a deeper understanding of the origin of cometary O₂.

2 Pre-agglomeration Formation and Trapping Scenarios

This section discusses the possibility of O₂ formation in the icy grains precursors of comets. Different scenarios of O₂ formation, proposed both in the ISM and in the PSN, are reviewed.

2.1 O₂ Formation in the ISM

As detailed below, the formation of O₂ can take place in the ISM in dark clouds either through surface reactions on interstellar grains, or via radiolysis and/or photolysis of icy grains.

2.1.1 Exogenic Radiolysis and O₂ Trapping in Icy Grains

Among the possibilities considered, the radiolysis of cometary water ice has attracted significant attention. A first scenario linking the production of O₂ to the ice of the comet nucleus was proposed (Bieler et al. 2015). These authors estimated, considering 67P/C-G’s thermal history, that any quantity of O₂ formed prior to approach at closer distances to the Sun should have been lost due to outgassing and/or erosion rapidly during the first one or two orbits around the Sun. It was therefore suggested that O₂ must have already been incorporated into 67P/C-G at the time of its formation in the PSN (Mousis et al. 2016b). In this context, radiolysis by Cosmic Ray Flux (CRF) is able to deposit large enough energy doses that create latent tracks in the bulk ice in which O₂ can be formed from the fragments H, O, OH issued from the radiolysis.

The possibility of radiolytic production of O₂ in micron to millimeter-sized icy grains present either in the outer PSN or in the ISM has been investigated by Mousis et al. (2016b). In both cases, the energy available for radiolysis was assumed to be provided by the galactic CRF impacting icy grains. The authors used the energy range and CRF distribution from Yeghikyan (2011) and Cooper et al. (2003) for their calculations, respectively. They also assumed that O₂ is produced by radiolysis of water ice through the chemical reaction

$$ 2\mbox{H}_{2}\mbox{O} \longrightarrow 2\mbox{H}_{2}+\mbox{O}_{2}, $$

(1)

with an amount of energy needed to alter one H₂O molecule $W_{r}=235~\mbox{eV}$ (Johnson 1991). Furthermore, the authors assumed that all the energy absorbed by water ice is used to form O₂. Under those circumstances, the time $\tau$ needed to reach a given O₂/H₂O ratio in icy grains has been established as follows:

$$ \tau = \frac{W_{r} \cdot N_{A}}{E_{CR} \cdot M_{\text{H}_{2}\text{O}}} \times f_{\text{H}_{2}\text{O}} $$

(2)

where $N_{A}$ ($\mbox{mol}^{-1}$) is the Avogadro constant, $M_{\text{H}_{2}\text{O}}$ ($\mbox{kg}\,\mbox{mol}^{-1}$) is the molar mass of water, $E_{CR}$ ($\mbox{eV}\,\mbox{kg}^{-1}\,\mbox{yr}^{-1}$) is the CRF energy dose received by water ice, which depends on the gas number density, and $f_{\text{H}_{2}\text{O}}$ is the fraction of altered H₂O molecules, corresponding to two times the fraction of O₂ produced. Figure 2 summarizes the results of the calculations presented by Mousis et al. (2016b) and shows that the timescales needed to produce an O₂ fraction in the 1–10% range in icy grains lying at 30 AU in the PSN midplane are much longer than the lifetime of the PSN ($\sim 0.25\mbox{--}2.5~\mbox{Gyr}$ vs. a few Myr). The CRF energy dose absorbed by icy grains located at this distance from the Sun is within the $\sim(5\mbox{--}60)\times 10^{16}~\text{eV}\,\text{kg}^{-1}\,\text{yr}^{-1}$ range, depending on the disk’s surface density (between 10 and $10^{3}~\mbox{g}\,\mbox{cm}^{-2}$; Hersant et al. 2001). We note that the distance of 30 AU was intended as a figure of merit representing the outer part of the disk in Mousis et al. (2016b). It does not quite correspond with where the outermost planets formed because our solar system was more compact, and then relaxed to the current configuration. However, the disk surface density is constant to a good approximation, and so the production rate from GCRs should not vary much inward. Any variation would not be enough to change the conclusion that the production timescale of O₂ in the observed relative abundances is longer than the age of the disk.

Fig. 2

Time evolution of the fraction of O₂ molecules produced by cosmic rays in icy grains located in various environments (derived from Mousis et al. 2016b). The PSN case (blue curve) considers the irradiation of an icy grain placed at 30 AU in the PSN midplane. The ISM case (red curve) considers the irradiation of an icy grain located in the interstellar medium

Mousis et al. (2016b) also considered the case in which the CRF has supposedly undergone a significant enhancement by a factor of $\sim 100$ because of a close supernova explosion. Because these events are too brief (∼ a few kyr), such enhancements cannot significantly decrease the time needed to form O₂ in the PSN midplane. Meanwhile, if icy grains have grown to sizes larger than tens of meters in the PSN, then the deepest layers would be more difficult to alter, requiring longer timescales for efficient O₂ formation. Mousis et al. (2016b) also considered the possibility of an icy grain receiving the maximum CRF energy dose estimated by Yeghikyan (2011) ($\sim1.20\times 10^{20}~\mbox{eV}\,\mbox{kg}^{-1}\,\mbox{yr}^{-1}$), a value corresponding to a surface density of $10^{-3}~\mbox{g}\,\mbox{cm}^{-2}$, which is typical of those found in molecular clouds. The authors determined that only $\sim 1.3\mbox{--}13~\mbox{Myr}$ of radiolysis was needed to generate an O₂ fraction in the 1–10% range in icy grains located in environments with such low column densities (see Fig. 2).

Mousis et al. (2016b) considered the radiolysis to be effective because of the different diffusion between the products of the radiolysis reaction. At temperatures encountered in the ISM or the PSN, the diffusion of O₂ in water ice is greatly reduced, which is not the case for H₂ (Johnson 1991). This difference in diffusion disfavors further reactions between O₂ and H₂, and helps to preserve O₂ in the ice. Given this assumption about the effectiveness of the radiolysis, the first consequence of the CRF irradiation of the ice mantles is the creation of latent tracks, i.e. channels void of H₂O molecules but filled with their fragments. These fragments participate in the reconstruction of the ice bulk by forming H₂O molecules (Eq. (3)) which close the open tracks. By doing so, they create cavities playing the role of confinement volumes in which O₂ could be formed by the coupling of two oxygen atoms. Reconstruction can obviously come from the successive addition of two atomic hydrogen to the oxygen atom. However, radiolysis of H₂O ice may create molecules other than H₂ and O₂. The OH radical is produced via the reaction $\mbox{H}_{2}\mbox{O} \rightarrow \mbox{H} + \mbox{OH}$ (Johnson 1991), which was not considered by Mousis et al. (2016b). The presence of the OH radical opens new chemical routes:

$$\begin{aligned} &\mbox{OH} + \mbox{OH} \rightarrow \mbox{H}_{2}\mbox{O} + \mbox{O}, \end{aligned}$$

(3)

$$\begin{aligned} &\mbox{OH} + \mbox{OH} \rightarrow \mbox{H}_{2}\mbox{O}_{2}, \end{aligned}$$

(4)

$$\begin{aligned} &\mbox{OH} + \mbox{O} \rightarrow \mbox{HO}_{2} \end{aligned}$$

(5)

that can be followed by the formation of O₂ according to

$$\begin{aligned} &\mbox{H}_{2}\mbox{O}_{2} + \mbox{OH} \rightarrow \mbox{HO}_{2} + \mbox{H}_{2}\mbox{O} \end{aligned}$$

(6)

$$\begin{aligned} &\mbox{H}_{2}\mbox{O}_{2} + \mbox{O} \rightarrow \mbox{HO}_{2} + \mbox{OH} \end{aligned}$$

(7)

$$\begin{aligned} &\mbox{HO}_{2} + \mbox{OH} \rightarrow \mbox{O}_{2} + \mbox{H}_{2}\mbox{O} \end{aligned}$$

(8)

The confinement in the cavities makes all these processes possible. The question is how can O₂ survive in these voids? To address this issue, Mousis et al. (2016b) investigated the question of the stability of the radiolytically produced O₂ molecules in the water matrix of ice grains. To do so, a sampling of the representative structures of O₂ in solid water ice has been obtained using a strategy based on first principle periodic density functional theory quantum calculations (Lattelais et al. 2011, 2015; Ellinger et al. 2015). These calculations, performed in the case of hexagonal ice Ih using the Vienna ab initio simulation package (VASP) (Kresse and Hafner 1993, 1994), show that when 2, 3, or 4 adjacent H₂O molecules were removed from the hexagonal lattice, well defined cavities form with shapes depending on the positions of the dismissed entities. The stabilization energies of the embedded O₂ were found to be in the order of 0.2–0.5 eV, namely the energy between water dimer stabilization and ice cohesion energies. This implies that the presence of O₂ does not perturb the ice structure until its further ejection due to sublimation or abrasion. A typical example of O₂ trapping is illustrated in the case of 4 removed H₂O molecules in Fig. 3. Similar results have been obtained in the case of O₂ stabilization in amorphous ice.

Fig. 3

Optimized crystalline structure for O₂ trapped inside the cavity ($n = 4$). The dashed lines show the unit cell considered in the VASP simulation. A vacuum of 15 Å high is used over the topmost H₂O bi-layer to avoid spurious interactions between adjacent slabs in the periodic treatment. The initial positions of destroyed H₂O are illustrated by the dotted circles. The arrow points to the embedded O₂ molecule

Based on all these considerations, Mousis et al. (2016b) concluded that, even with a strong CRF enhancement due to the presence of a nearby supernova, the radiolysis of icy grains is not fast enough in the PSN to create amounts of O₂ comparable with those observed in 67P/C-G. Instead, they showed that icy grains must be placed in low-density environments such as molecular clouds to enable radiolysis to work efficiently. They also found that the irradiation process favors the stabilization of O₂ molecules in the icy matrix via the development of cavities and is compatible with both amorphous and crystalline ice structures. From this set of constraints, two possible formation scenarios of 67P/C-G in the PSN have been derived by the authors: (1) agglomeration from pristine amorphous icy grains/particles formed in the ISM and (2) agglomeration from multiple guest clathrates including O₂ that formed during the cooling of the disk subsequent to the vaporization of the amorphous icy grains entering the PSN (see Sect. 2.3.2). With the incorporation of O₂ in the cavities created by CRF in the icy matrix, scenario 1 naturally provides an explanation for the strong correlation found between the O₂ and H₂O production rates observed in 67P/C-G’s coma (Bieler et al. 2015). However this scenario makes implausible the accretion of 67P/C-G from clathrates and crystalline ices originating from the PSN, as proposed by Mousis et al. (2016a) and Lunine et al. (1991a). A way to reconcile scenario 2 with the strong O₂-H₂O correlation would be to assume that the icy grains (i) initially formed as in scenario 1 and (ii) subsequently experienced an amorphous-to-crystalline phase transition in the 130–150 K temperature range when entering the disk (Kouchi et al. 1994; Maldoni et al. 2003; Ciesla 2014). In this context, all volatiles initially adsorbed by ISM amorphous ice would be released in the PSN gas phase during phase transition. With the cooling of the disk, these volatiles would have been later trapped in the clathrates formed with the crystallized icy grains. O₂ is unique because, in spite of the ice phase transition, this molecule would remain stable within the icy matrix thanks to the strength of the interaction between O₂ and the surrounding H₂O molecules that would remain stable, or eventually increase, upon crystallization.

The exogenous radiolysis scenario makes it possible to rationalize also why H₂O₂ and HO₂ are detected only in very low abundances in 67P/C-G, and O₃ not detected at all. These chemically-related species are formed during radiolysis and are effectively present into the cavities of the cometary ice. Our calculations show that all the oxygenated actors, i.e. H₂O₂, the corresponding radical HO₂, atomic oxygen O, and the OH radical are stabilized within the ice matrix with comparable energies, close to the stabilization energy of water. Due to the confinement situation, the cavity appears as a chemical reactor in which H₂O₂ and HO₂ are converted into O₂ (see reaction network above). It is the formation of O₂ which is at the origin of the failure to observe the reactants leading to its formation. As to O₃, it is so reactive that it will be destroyed by any radical formed in the radiolysis.

2.1.2 Recombination of Oxygen Atoms on the Surfaces of Interstellar Icy Grains

Gas-phase chemistry is not sufficient to account for the observed abundance of interstellar molecules because of the inefficiencies of gas phase routes. Grain-surface chemistry is thus required to explain the observed abundances of water and other molecules, like H₂ or O₂ in molecular clouds (Tielens and Hagen 1982; D’Hendecourt et al. 1985; Hasegawa et al. 1992). Surface O₂ is involved in the chemistry reaction network of water ice (Tielens and Hagen 1982; Cuppen and Herbst 2007), and forms through the recombination of atomic oxygen on icy grain surfaces.

The formation of O₂ in the ISM in dark clouds can take place either through surface reactions on interstellar grains, or via radiolysis and/or photolysis of ice grains. In the case of surface reactions, O₂ is formed through the recombination of atomic oxygen on icy grain surfaces. O₂ efficiently reacts with O to form O₃, and H to form HO₂. The hydrogenation of O₃ is the main destruction process of O₃ and leads to the formation of O₂ along with HO (Mokrane et al. 2009). Formation of O₂ through radiolysis or photolysis takes place when energetic particles and UV photons break bonds triggering chemistry within the icy mantle of the grains, which could possibly lead to the formation of O₂. However, this process seems unlikely for several reasons. First, although UV photolysis produces O atoms that may recombine to form O₂, it also produces H atoms which react with O₂ to reform H₂O. Second, laboratory experiments show that if radiolysis does not occur when water is condensing onto the grain surface, then water radiolysis overproduces H₂O₂ while producing O₂ (Teolis et al. 2006).

Taquet et al. (2016) conducted a parameter study in their astrochemical model to evaluate the formation of O₂ in dark clouds. They varied several chemical and physical parameters with the aim of reproducing the relative abundances of related species (H₂O, H₂O₂, HO₂ and O₃) with respect to O₂ observed in 67P/C-G. The authors were able to reproduce the O₃/O₂ and HO₂/O₂ abundance ratios by introducing an activation barrier of $\sim300~\mbox{K}$ for the $\mbox{O}+\mbox{O}_{2}$ and $\mbox{H}+\mbox{O}_{2}$ reactions. However, under those conditions, H₂O₂ is overproduced by one order of magnitude compared to the observations, suggesting that other chemical processes are at play. In their simulations, Taquet et al. (2016) were only able to reproduce high O₂ abundances, comparable to those measured for 67P/C-G and when the density of H nuclei in the gas phase was very high ($\ge10^{5}~\mbox{cm}^{-3}$). Simulations using intermediate temperatures, $\sim20~\mbox{K}$, and low cosmic ray flux also favored the formation and survival of O₂, with only a weak impact from visual extinctions. The temperature of $\sim20~\mbox{K}$ enhances the mobility of oxygen atoms on the surface of the grain, which favors oxygen recombination forming O₂. At the same time, it allows efficient sublimation of atomic hydrogen, limiting the impact of hydrogenation reactions that destroy O₂. The results presented by Taquet et al. (2016) suggest that the physical conditions of $\rho$ Oph A, where Hershel observations revealed the presence of O₂, HO₂ and H₂O₂ gases in proportions consistent with the observations at 67P/C-G, is a favorable environment for the formation and survival of O₂ because of its high density ($\sim10^{6}~\mbox{cm}^{-3}$) and warm temperature for a starless core (dust temperature of $\sim20~\mbox{K}$). Furthermore, Taquet et al. (2016) showed that O₂ is mostly formed in the innermost layers of the icy mantles and that it is well preserved during the formation of the protoplanetary disk. Contrary to O₂, CO and N₂ mostly form in the outermost layers of the icy mantle. Taquet et al. (2017) detail the impact of layering of the icy mantle during the transition from dark cloud to protoplanetary disk and its consequences to the volatile content of cometary ices. The outer layers of the icy grains are more likely to evaporate in the outer layer of the protoplanetary disk releasing N₂ and CO in the gas phase, while the inner ice layers where most of the O₂ would be trapped remains frozen. N₂ and CO can subsequently freeze out onto ices again once they reach the colder disk mid-plane. As CO is slightly less volatile than N₂, it freezes out more efficiently than N₂, which leads to a low N₂/CO abundance, in agreement with observations made at 67P/C-G (Rubin et al. 2015a). Furthermore, as refreezing is independent of the amount of H₂O in the icy mantle, this would also explain the low correlation of N₂ and CO with H₂O in cometary ices. On the other hand, the inner layer, where most of the O₂ is thought to be trapped in H₂O ice, remains unprocessed, which explains the good correlation between O₂ and H₂O reported by Bieler et al. (2015). The preservation of the inner layers of the icy grain mantles from the dark cloud stage until their incorporation into cometary nuclei is supported by the observation of a distributed source of halogens in the coma of 67P/C-G (Dhooghe et al. 2017). Thus, the scenario of O₂ production and the subsequent evolution of ice during the transition from dark cloud to protoplanetary disk discussed in Taquet et al. (2016) and Taquet et al. (2017) is consistent with observations of other volatiles (such as N₂ and CO) made at 67P/C-G.

2.2 O₂ Formation in the PSN

2.2.1 During Protosolar Collapse/Protoplanetary Disk Formation

Formation of O₂ during protostellar collapse and protoplanetary disk formation was evaluated by Taquet et al. (2016) using an astrochemical model for gas grain chemistry (Furuya et al. 2016). A fluid approach is used to describe the physical evolution of the cloud during its collapse and during the formation of an accretion disk described by the $\alpha$-viscosity prescription. The fluid parcels are traced with the 2D axisymmetric and semi-analytical model initially developed by Visser et al. (2009). The initial composition of the gas and solid phase of the parent cloud was obtained from a molecular cloud model (Furuya et al. 2015) in which the amount of O₂ ice is negligible at the onset of the collapse. The authors found that some gaseous O₂ can form (up to $10^{-6}\mbox{H}_{2}$), but that O₂ trapped in H₂O ice does not form efficiently. Because most of the O-bearing species are in the ices at the onset of the collapse, gaseous O₂ can only form through photodissociation or desorption of H₂O by stellar UV followed by subsequent gas phase reactions. In their simulations, this formation process failed to reproduce the observations of O₂ made in 67P/C-G. They found that the O₂/H₂O ratio in the ice was several orders of magnitude lower, except for a few parcels of the upper layers of the disk where it is higher than $10^{-2}$. They also found that O₂ was associated with CO₂ rather than with H₂O because CO₂ ice reformation was more favorable than H₂O reformation.

Taquet et al. (2016) ran another simulation where they modified the initial conditions in order to have a O₂/H₂O ratio of 5% in the ice, comparable to the ratio in 67P/C-G, at the beginning of the collapse. This simulation shows that this ratio is largely preserved, indicating that O₂ produced in an earlier stage, for instance at the molecular cloud stage, would be preserved during the formation of the protoplanetary disk.

2.2.2 During the Vertical Transport of Icy Grains

The O₂ formation in icy grains irradiated during their transport towards the upper layers of the PSN has been investigated by Mousis et al. (2018). In this scenario, the icy grains present in the midplane were lifted toward the upper layers of the disk and dragged down over a large number of cycles, due to turbulent mixing (see Fig. 4). Consequently, these grains spent a non-negligible fraction of their lifetime in the disk’s upper regions, where CRF irradiation was stronger than in the PSN midplane.

Fig. 4

Illustration of the vertical transport of small icy grains toward disk regions where irradiation is strong and favors the formation of O₂ (figure adapted from Mousis et al. 2018). Grains remain concentrated in the midplane of the disk because of gravitational settling and gas drag

The computations performed by Mousis et al. (2018) suggest that, even if a significant fraction of the icy particles have followed a back and forth cycle towards the upper layers of the disk over 10 million of years in a static PSN, the amount of O₂ created via radiolysis is at least ∼ two orders of magnitude lower than the Rosetta observations, making the proposed mechanism completely negligible if at play. This timespan exceeds by far the formation timescale of 67P/C-G, which has been estimated to range between 2.2 and 7.7 Myr after the formation of Ca-Al-rich inclusions in the PSN (Mousis et al. 2017). They also found that an increase of the CRF up to a factor of ∼100 due to a close supernova explosion would not substantially change the O₂/H₂O ratio in icy grains because the timespan of such an event (a few kyr) is too short (Mousis et al. 2016b).

2.3 Different O₂ Trapping Scenarios

2.3.1 Trapping in Amorphous Ice in the PSN Prior to Comet Formation

If O₂ were produced in the molecular cloud prior to the formation of the protosolar nebula itself, might it have been trapped in grains and delivered to the protosolar nebula as the latter formed? Trapping and transport of the pre-nebular molecular oxygen in grains that reach the comet-forming zone in the protosolar nebula is a more complex problem, given the range of environments and the time dependence of processes delivering grains to the disk, along with possible modification of grains as they enter the disk (e.g., Lunine et al. 1991b).

Hence, Taquet et al. (2016) invoke the variable luminosity of low mass stars to construct a scenario in which luminosity outbursts sublimate ice from grains, allowing O₂ to be produced by gas phase cosmic ray energized chemistry, and then frozen out into grains of amorphous ice formed as the stellar luminosity tails off quickly. A series of such luminosity excursions could lead to the production and trapping of O₂ in ice grains directly within the planet-forming disk. However, they discard this disk-origin model for O₂ because, they argue, other species such as N₂ would become trapped with the O₂, which is not consistent with the observed depletion and lack of correlation with H₂O in 67P/C-G, in contrast with O₂’s strong correlation (N₂ should be more poorly trapped in amorphous ice than O₂, since the measured binding energy for adsorption of N₂ is about 2/3 that of O₂ (see Introduction), though the key here is the difference in correlation with H₂O). Another concern, not raised by Taquet et al. (2016), is whether the luminosity fall-off and cooling are fast enough to form the amorphous ice required to trap the molecular oxygen.

Taquet et al. (2016) favor a dark cloud site, outside of and perhaps prior to the protosolar nebula, for the chemical production of O₂ from H₂O. Although consistent with the cometary data, the consequent complexities of delivery to the disk including the potential for removal of the O₂ from the ice by even modest heating, must be considered in contemplating a molecular cloud source.

2.3.2 Trapping in Clathrate Hydrate

Trapping of volatile species in nebular clathrate hydrate has been treated by numerous groups over a number of years, but with less attention paid to molecular oxygen until very recently. Tse et al. (1986) prepared a clathrate hydrate with O₂ as the sole “guest” species (the molecule occupying the void or cage site in the water ice H-bonded structure) and found it to induce the so-called structure II clathrate, with 16 small and 8 large cages per unit cell. The dominant clathrate-forming species in the solar nebula, CO and CH₄, both induce structure I clathrate formation (Lunine and Stevenson 1985), but O₂ will incorporate as a secondary guest species in the structure I. Mousis et al. (2016b) numerically computed O₂ incorporation in clathrate in the nebula by considering a two-component gas of O₂ and CO, employing a statistical mechanical formalism first used comprehensively for solar system studies by Lunine and Stevenson (1985), but with molecular interaction (Kihara) parameters from the most recent data sets available. Mousis et al. (2016b) found that the ratio of O₂ to CO in the clathrate hydrate was up to twice that in the coexisting gas phase of the nebula. Given the lack of evidence for direct condensation of the class of volatile species O₂, N₂, and argon as molecular ices, clathrate hydrate may have been an important carrier of molecular oxygen in the protoplanetary disk. This scenario has been ruled out by Mousis et al. (2016b) because it predicts a strong O₂-CO correlation instead of the one observed for O₂-H₂O, regardless the seasons or heliocentric distance of the comet.

2.3.3 O₂ Condensation in the Protosolar Nebula

As mentioned above, direct condensation of O₂ (and other volatiles) in the colder regions of the PSN also needs to be considered in the context of O₂ in comet 67P/C-G. O₂ and other volatiles condensed from the gas phase in the PSN may have incorporated into the forming nucleus if the nucleus was formed in a colder environment than needed for clathration. Pure Ar, N₂, CO, and O₂ condense at similar temperatures ($\sim20\mbox{--}25~\mbox{K}$) in the PSN (Fray and Schmitt 2009). Therefore, their similar condensation temperature implies that if direct condensation is responsible for the observed O₂ in the coma of 67P/C-G, then both Ar and N₂ should be present in the nucleus in near-solar proportions.

ROSINA/DFMS measured the abundance of O₂ and other volatiles relative to H₂O in the coma. While it is important to note that measurements of volatile abundances in the coma do not necessarily reflect abundances in the nucleus, they currently are our best estimate of the possible composition of the nucleus. As measured by ROSINA/DFMS, Ar/CO (Balsiger et al. 2015) and N₂/CO (Rubin et al. 2015a) in the coma of 67P/C-G have been found to be significantly depleted compared to the solar abundances, by a factor of $\sim90$ and 10, respectively (Mousis et al. 2016b). In addition, N₂ was found to be strongly depleted with respect to CO in three Oort cloud comets (Cochran et al. 2000; Cochran 2002). This depletion implies that Ar and N₂ did not likely form substantial amounts of pure condensates in the PSN where comet 67P/C-G was formed. However, a mixture of pure condensates and clathrates in the nucleus of 67P/C-G is also possible.

O₂ was observed to correlate strongly with H₂O during the six-month approach-phase interval reported by Bieler et al. (2015). If such strong correlation is representative of the outgassing behavior through the entire orbit, then it is likely that either O₂ and H₂O desorb together, or that O₂ is produced from H₂O (see Sect. 3 for details).

3 Post-agglomeration Formation Scenarios

3.1 In Situ O₂ Formation Processes

3.1.1 Eley-Rideal Reactions

An alternative hypothesis concerning a class of rarely-considered Eley-Rideal mechanisms (ER) was recently proposed by Yao and Giapis (2017) to explain the presence of O₂ in the coma of 67P/C-G. In these ER reactions, energetic water-derived ions (e.g. $\mbox{H}_{2}\mbox{O}^{+}$, $\mbox{H}_{3}\mbox{O}^{+}$, $\mbox{OH}^{+}$, $\mbox{O}^{+}$) are accelerated by the solar wind and are pulled back toward the nucleus (or other surfaces). Based on laboratory experiments, when the ionized water-group species collide with oxidized Si/Fe or other mineral surfaces, they abstract an oxygen atom and form an excited precursor state. The dissociation of this precursor state generates energetic $\mbox{O}_{2}^{-}$ anions, which are then converted to O₂ by photo-detachment. $\mbox{O}_{2}^{-}$ and energetic O₂ produced this way was proposed to account for the O₂ detected in the coma of 67P/C-G (Yao and Giapis 2017). Since O₂ is produced from H₂O on multiple possible surfaces in the proposed ER reactions, the two species must be closely associated at all times, and the $\mbox{O}_{2}^{-}$ production rate must be proportional to the accelerated cometary H₂O ion flux. In addition to energetic water-derived ions, O₂ could also be produced in collisions of (1) energetic hydronium ions, (2) energetic neutral water molecules, and (3) excited atomic oxygen (thermal or hyperthermal) on the surfaces of the nucleus, the spacecraft, or the DFMS instrument itself.

While the proposed ER reactions do explain the strong correlation between H₂O and O₂ as reported by Bieler et al. (2015), the energetic water ion flux (Nilsson et al. 2017; Fuselier et al. 2015) seems to be too low to account for the observed high O₂ abundance in the coma of 67P/C-G. Recently, Heritier et al. (2018) concluded that both the integrated energetic cometary water-group ion flux and the maximum fraction of O₂ produced via ER reactions (assuming 100% yield) were insufficient for ER reactions to be the dominant O₂ source. In fact, they estimated that the O₂ production rate by ER reactions is at least 4 orders of magnitude lower than the observed O₂ production rate, even when using the conservative low O₂ concentration of 1%. In addition, Heritier et al. (2018) argue that the energetic water ions that drive ER reactions were anti-correlated with the O₂ density measured by ROSINA/DFMS, at least for March 5–23, 2016. At the same time, Yao and Giapis (2018) argued that “cold” water ions, whose flux is about 100 times that of the “accelerated” water ions (Nilsson et al. 2015) bombarding the spacecraft could produce O₂ in situ. Whether this O₂ would be detected by DFMS is uncertain as the DFMS response under exposure to a low-pressure water plasma containing O₂ is unknown (Yao and Giapis 2018). It is also important to note that Alice has detected O₂ independently of ROSINA (Keeney et al. 2017a), and post-perihelion Alice results indicate that the relative O₂/H₂O decreases with increasing impact parameter above the nucleus (Keeney et al. 2017b). Although the Alice measurements may arguably be affected by spacecraft-produced O₂, the observed O₂/H₂O variation with impact parameter would be unlikely within this framework. Finally, ER reactions require a direct collision with the appropriate atom, which is a low probability process.

3.1.2 O₂ Formation via H₂O₂ Dismutation/Disproportionation

Dulieu et al. (2017) proposed dismutation of cometary H₂O₂ as a source of O₂ observed in the comae of 67P/C-G by ROSINA/DFMS (Bieler et al. 2015) and Halley by the Giotto mass spectrometer (Rubin et al. 2015b). The peroxide is suggested to have a primordial origin, co-produced with water ice on surfaces of interstellar dust grains via hydrogenation reaction of O-bearing species, and transported into the comet-forming regions of the protoplanetary disks (Taquet et al. 2016). The comet’s volatile constituents, including water ice and peroxide, undergo temperature-dependent sublimation during perihelion approach. The peroxide either desorbs as intact molecules, or dismutates as the global reaction

$$ 2 \mbox{H}_{2}\mbox{O}_{2} \rightarrow 2 \mbox{H}_{2}\mbox{O} + \mbox{O}_{2}. $$

(9)

Based on results from their laboratory experiments summarized below, the authors (Dulieu et al. 2017) assert this peroxide-derived O₂ as the source of molecular oxygen detected in the coma of 67P/C-G at abundances ranging from $\sim1\mbox{--}10\%$ relative to H₂O.

In the first set of experiments, the authors co-deposited molecular O₂ with an atomic beam of D onto an amorphous silicate substrate held at 10 K. These species condensed and reacted to form a thin film ($< 1~\mbox{ML}$ (monolayer)) of O₂, D₂O, and D₂O₂, which are thermally desorbed and detected with a mass spectrometer when the substrate was heated to $> 200~\mbox{K}$ at $10~\mbox{K}\,\mbox{min}^{-1}$. The key result was the presence of O₂ and D₂O desorption peaks at 175 K, accompanying the sublimation of the less volatile D₂O₂. Dulieu and coworkers interpreted the higher temperature O₂ and D₂O peaks as the signature of D₂O₂ dismutation. While a fraction of the D₂O₂ sublimes intact, some are converted to D₂O and O₂ on the surface and released in the gas phase. Above 180 K, the ratio O₂/D₂O₂ is about 0.4; nearly half that of D₂O/D₂O₂ $\sim1$, and consistent with the stoichiometry of the dismutation reaction.

The second set of experiments was a repeat of the first set, except performed on a 25 ML-thick compact H₂O ice pre-deposited on top of the silicate substrate. A distinct O₂ peak was observed at 155 K, immediately following the main H₂O/HDO desorption. Solid-state isotope exchange led to the conversion of D₂O and D₂O₂ formed via $\mbox{O}_{2}+\mbox{D}$ reactions into HDO and HDO₂, respectively. However, it is not clear what fraction of the deuterated species was substituted with H even to form H₂O₂, which is not discussed.

Several differences were noted compared to the first set of experiments: the O₂ peak was observed at 155 K, 20 K lower compared to the first experiments. Further, no concurrent HDO₂ desorption peak was observed with O₂. Instead, the HDO₂ sublimation followed the O₂ peak at 160 K. The authors claim a dismutation origin for the O₂, since they report a HDO desorption rate twice that of O₂ at 155 K. However, this is not consistent with the data in Fig. 3 of their manuscript (reproduced here as Fig. 5) where the HDO rate is about ten times larger compared to O₂. Furthermore, to explain the absence of HDO₂, the authors inferred a highly efficient peroxide dismutation (close to unity) attributed to the high concentration of peroxide that accumulates on the silicate substrate following H₂O sublimation, and the catalytic effect of the substrate favoring peroxide dismutation. Above 160 K, the three species continue to effuse out with O₂/HDO₂ and D₂O/D₂O₂ ratios more consistent with dismutation.

Fig. 5

Desorption rates of O₂, HDO, HDO₂, and H₂O from heating a thin film produced from the reaction of D atoms with O₂ condensed onto the surface of a 25 ML-thick compact amorphous ice. The H atoms are introduced in the deuterated species via isotope exchange. Reproduced from Dulieu et al. (2017)

The dismutation process requires the incorporation of primordial peroxide into the cometary nucleus. Though astrochemical models (Tielens and Hagen 1982; van Dishoeck et al. 2013) predict the occurrence of peroxide on interstellar grains, and laboratory experiments support the formation of H₂O₂ via hydrogenation reactions with O-bearing species (Dulieu et al. 2010; Ioppolo et al. 2008; Jing et al. 2011; Oba et al. 2009), an unambiguous detection of solid peroxide in the molecular clouds has not been reported to date. H₂O₂ has been reported in the gas phase, but at a small abundance of $10^{-10}$ relative to H₂ (Bergman et al. 2011). Smith et al. (2011) analyzed the infrared spectra of several molecular clouds, and placed upper limits on H₂O₂ abundance ($9 \pm 4\%$ relative to water ice) by assessing the possible contribution of the 3.5 μm peroxide feature measured in the laboratory to the 3.47 μm interstellar feature. However, other H₂O₂-specific absorption features expected at 7 and 11.3 μm were not identified, probably due to masking by underlying H₂O and silicate features (Smith et al. 2011). Further, it is not clear if the primordial H₂O₂ can survive the evolution of the dark clouds to protoplanetary disks and be transported unaltered to the comet-forming regions.

It is also challenging to reconcile the highly efficient (50–100%) peroxide dismutation with other laboratory experiments where stable films of pure crystalline H₂O₂ ($\sim98\%$ purity) is obtained from the isothermal distillation of H₂O-H₂O₂ films at 167 K (Loeffler et al. 2006b). This temperature exceeds the dismutation temperature of 155 K. Using a quartz crystal microbalance, Loeffler and colleagues reported no mass loss ($<0.1$ H₂O₂ ML, the detection limit of the microbalance) from a distilled crystalline peroxide film (Loeffler et al. 2006b). Any H₂O₂ dismutation into O₂ and H₂O, which are thermally unstable at 167 K, would have resulted in a mass loss. While we note the use of different substrates (gold vs. silicate), the substrate-promoted dismutation is likely not important for multi-layered peroxide films obtained by Loeffler et al. (2006b).

Finally, H₂O₂ can fragment into O₂ from electron impact in the ionization region of the mass spectrometer (Loeffler et al. 2006a). Also, collision with the chamber walls can dissociate H₂O₂ to O₂. It is unclear if the desorption flux reported by Dulieu et al. (2017) were corrected for fragmentation.

3.2 Post-accretion Radiolysis

One can wonder how radiolysis affected cometary ices over their orbital history during their 4.56 billion-year evolution in the Solar System. Bieler et al. (2015) noted that comets are subject to radiolysis on 3 different timescales: (i) billions of years during their residence in the Kuiper belt, (ii) over a period of a few years when they enter the inner Solar System, and (iii) on very short timescales when they approach perihelion. The authors estimated that the skin depth for producing O₂ is in the range of meters during their residence in the Kuiper belt. However, once a comet migrates towards the inner Solar System, it easily loses surface material over depths reaching several meters during each orbit. This material loss implies that, according to Bieler et al. (2015), no O₂ produced during the Kuiper belt phase is likely to remain in the amount exceeding the percentage level. Radiolysis and photolysis by solar wind and UV flux in the inner Solar System may also produce O₂ at the top micrometer levels of cometary surfaces. Meanwhile, based on measurements of continuous mass loss through outgassing, Bieler et al. (2015) estimated that 67P/C-G’s active surface areas lost a depth of several centimeters during their sampling period between September 2014 and March 2015. Because their data show a constant O₂/H₂O ratio during the studied time interval instead of a continuous decrease, the authors ruled out the hypothesis that a recent production by radiolysis or photolysis may be the source of the measured O₂. In addition, Bieler et al. (2015) excluded the possibility of instant creation of O₂ by radiolysis or photolysis because these processes would lead to variable O₂/H₂O ratios as a result of different heliocentric distances and illumination conditions.

3.2.1 Endogenic Radiolysis

A possible source of molecular oxygen in 67P/C-G is radiation chemistry of ice induced by an endogenic source—decaying radionuclides present in the chondritic dust that represents 4/5th of the mass of the comet (Rotundi et al. 2015). Short-lived (²⁶Al, ⁶⁰Fe) and long-lived (⁴⁰K, ²³⁵U, ²³⁸U, ²³²Th) radionuclides emit energetic alpha particles (helium nuclei), beta particles (electrons), and gamma rays that interact with the surrounding materials, including ice. Assuming an even repartition of the radionuclide-bearing dust, this process would produce O₂ homogeneously within the ice, in conformity with the observation of an O₂ signal correlated with water (Bieler et al. 2015).

Using a production model originally developed for Earth water-rock mixtures (Hofman 1992; Lin et al. 2005a,b; Bouquet et al. 2017), Bouquet et al. (2018) estimated a maximum value of 1% O₂/H₂O produced within 67P/C-G over the age of the Solar System, which is below the 3.8% value derived from the ROSINA/DFMS observations. However, this estimate relies on O₂ production yields from ice irradiation experiments, which are motivated by the need to understand the evolution of icy surfaces exposed to planetary radiation belts (e.g. Europa). These experiments are therefore performed at projectile energies lower than the values relevant to radionuclide decay emissions, and gamma ray irradiation is usually not considered. Additionally, the results of Teolis et al. (2017) indicate that O₂ production yields in ice are higher for low projectile energies and high ice temperatures ($>80~\mbox{K}$). Conditions in comet 67P/C-G are therefore unfavorable for O₂ production, as projectile energies are high and the temperature has stayed below 80 K for most of the comet’s history (Mousis et al. 2017). For these reasons, the 1% value reported in Bouquet et al. (2018), which uses the highest yields from the literature, is an upper limit.

It should be noted that H₂O₂ is also produced by ice radiation chemistry in a more consistent manner than O₂ (Teolis et al. 2017) regarding projectile energy. The findings of Bouquet et al. (2018) point to the potential for endogenic radiolysis to produce amounts of H₂O₂ on the order of a tenth of a percent compared to water. Such quantities are incompatible with the very low abundance of H₂O₂ detected by ROSINA/DFMS (Bieler et al. 2015), indicating that either production was prevented, or that H₂O₂ was destroyed by a chemical process, possibly producing O₂ (Dulieu et al. 2017).

Endogenic radiolysis cannot account for the abundance of O₂ measured by ROSINA, but it may have produced a non-negligible fraction of it, either directly, or through production and destruction of H₂O₂. This homogenous distribution of O₂ in the ice would fit the observed correlation with water. More laboratory data is required to better constrain this fraction.

4 Conclusions

The presence of large amounts of O₂ in the coma of 67P/C-G continues to be perplexing. While several possible scenarios have been postulated for the formation of cometary O₂, the current state of analysis of the published Rosetta data preclude the determination of its dominant source and origin. All O₂ formation mechanisms currently proposed in the literature rely heavily and solely on the only available published O₂ data, which only covers a six-month time interval from Rosetta’s approach phase (Bieler et al. 2015). This limitation raises the issue of how representative the time interval is for the entire orbit of 67P/C-G, which would be key for deciphering the origin of O₂. Is the O₂-H₂O correlation strong throughout the orbit, or does it change with heliocentric distance? Are there any hemispherical differences between the correlations? Do the large-scale changes or the lack thereof more or less agree with the small-scale changes, and if not, then why? Does the O₂/H₂O ratio remain fairly constant at different orbital segments, or does it change systematically/randomly/hemispherically/with illumination/with temperature? Is the bulk of cometary O₂ produced by one dominant mechanism, or do multiple production mechanisms contribute to the signal? Does the dominant production mechanism change at different parts of the orbit? These are just a handful of questions that need to be answered in order to assess the viability of each scenario reviewed in this paper. Some of the answers most likely lie in the systematic analysis of the Rosetta data, and some issues will likely require future missions to address.

If O₂ is primordial, then the O₂/H₂O should remain nearly constant with heliocentric distance, regardless of the hemisphere (see Fig. 1). That is because O₂ and H₂O ices in the nucleus would be physically related. That means that the production rate of O₂ must be related to the production rate of H₂O at all times. When more H₂O is released from the nucleus due to higher sublimation rates, more O₂ is also released, resulting in a nearly constant O₂/H₂O with season, hemisphere, and heliocentric distance. On the other hand, if O₂ has a present-day origin and is produced in situ, then the O₂/H₂O is expected to change with the seasons and heliocentric distance. If O₂ is produced in situ from water-group ions accelerated by the solar wind hitting various surfaces and triggering ER reactions, then O₂ production would have to depend on: (i) the H₂O production rate, (ii) on the intensity of the solar UV/photoelectrons and the solar wind/energetic electrons, and (iii) dynamic processes in the solar wind including its interaction with the coma (shocks, discontinuities, etc.). UV photons and photoelectrons ionize the cometary H₂O molecules, while the solar wind and energetic electrons drag these water-derived ions back onto the surface, initiating the ER reactions. The H₂O production rate scales with heliocentric distance ($R_{h}$) as a function of roughly $R_{h}^{-c2}$, where $c2 = 5$ and $c2 = 7$ pre- and post-perihelion, respectively (Hansen et al. 2016). At the same time, the solar wind and solar UV scale with $R_{h}$ as $\sim1/R_{h}^{2}$. Therefore, the production rate of O₂ should scale with $\sim R_{h}^{-(c2-2)}$, while the production rate of H₂O is $R_{h}^{-c2}$, meaning that the O₂/H₂O ratio should follow a $\sim 1/R_{h}^{2}$ dependence. Thus, if ER reactions produce the majority of the observed O₂, then the O₂/H₂O ratio should be several factors (by a factor of $\sim5$) higher at perihelion than at 3 AU. Therefore, analysis of the O₂/H₂O over the extent of the mission would be needed to conclude whether or not the ratio changes with cometocentric distance. In addition, ER reactions are independent of the illumination conditions and the nucleus temperature. At the same time, H₂O₂ dismutation/disproportionation reactions are temperature-dependent, and the O₂/H₂O₂ ratio should vary below about 180 K (Dulieu et al. 2017). At $R_{h} \geq 3~\mbox{AU}$ the conditions were ideal for identifying differences in the O₂/H₂O₂ between the two hemispheres due to the different hemispheric temperatures; hence, looking at time periods prior to the inbound equinox and following the outbound equinox may be a good start. We also note that a possible lack of correlation between O₂ and H₂O may occur as a result of H₂O re-condensation onto the nucleus after the more volatile O₂ has already sublimated away. This re-sublimated H₂O would not contain O₂ anymore; hence, would weaken the correlation between O₂ and H₂O.

A potential strong correlation between O₂ and CO, and/or O₂ and CO₂ (hemispheric and/or over a given period along the orbit) would make it more likely that O₂ was formed in the PSN instead of the ISM, and that it is not produced in situ. In that case, determining the O₂/CO ratio over time and comparing it to calculations from the clathrate model of Mousis et al. (2016b) could provide more definitive answers as to what phase O₂ is stored in the nucleus. Furthermore, the time scales calculated by Mousis et al. (2016b) may yield a different conclusion about the formation location of icy grains if the GCR flux (H and Fe) were estimated using a different approach, i.e. that of Webber and Yushak (1983). While the Fe flux is lower than the proton flux based on the latter approach, the electronic stopping in H₂O ice is higher, which could potentially make Fe contribute more heavily to O₂ production in icy grains. In future parameter studies, both the issue of GCR flux and the depth-dependency of the O₂ yield (Teolis et al. 2017) will need to be evaluated.

The conundrums created by the discovery of O₂ in 67P/C-G would be readily dismissed as perhaps a peculiarity were it not for the conclusion that 1P/Halley data also show evidence of this molecule in similar amounts (see Sect. 1 Introduction). The inference that 1P/Halley may have been an Oort cloud comet, rather than a Jupiter family object as 67P/C-G, might argue for a widespread occurrence of O₂ in the protoplanetary disk, any other explanation (such as both comets being anomalous) seeming to beg the question. It is thus incumbent upon modelers to focus additional effort on mechanisms for producing this molecule either in the protoplanetary disk, precursor clouds, or in the comet itself. The prospect that a sample of a comet nucleus might be returned to Earth, thanks to the selection for additional development of the CAESAR mission in NASA’s New Frontiers program, opens up the possibility of determining how this molecule is incorporated in the ice. Until then, the difficult task of remote sensing of this symmetric molecule or its atomic fragments will have to suffice for additional data.