Main

Measurements of the coma of 67P/Churyumov–Gerasimenko (hereafter 67P) were made between September 2014 and March 2015 with the ROSINA-DFMS mass spectrometer5 on board the Rosetta spacecraft. For the present study, we analysed 3,193 mass spectra taken in this time period. Because of the high mass resolving power and sensitivity of ROSINA-DFMS, it was possible to differentiate unambiguously between the three main species present in the narrow mass range centred at mass-to-charge (m/z) ratio 32 Da/e, namely, molecular oxygen (O2), atomic sulfur (S) and methanol (CH3OH); such differentiation has not been achieved by previous in situ or remote sensing measurements at comets. Figure 1 shows several DFMS measurements centred at the O2 peak. The green and orange lines show data taken before the close encounter with 67P. Only minor signatures from the tenuous neutral gas atmosphere of the Rosetta spacecraft can be identified, and even after long thruster firing manoeuvres, which use N2O4 as an oxidizer, the contamination of the O2 signal remains small (see the green line in Fig. 1). Measurements taken while orbiting 67P, shown as the light blue, dark blue and purple lines in Fig. 1, show a clear increase of the O2 signal, indicative of cometary O2. These three measurements were taken at decreasing distances (r) from the comet nucleus, and follow the predicted 1/r2 signal dependence that is expected for conserved cometary species, giving further confidence in our detection.

Figure 1: DFMS mass spectra around 32 Da/e normalized to the spectrum with the largest signal.
figure 1

The black labels indicate the three major species found in the coma of 67P at 32 Da/e. The green labels and green line identify contamination peaks from thruster firings, showing that their contributions to the O2 peak are very low. The light blue, dark blue and purple lines represent measurements taken at different distances from the comet nucleus.

PowerPoint slide

Full size image

As previously reported, the local number densities in the coma vary spatially and temporally6,7 for different compounds. Figure 2 shows correlation plots of H2O with O2, CO and N2. Red symbols represent data from 17 to 23 October 2014, for which previously published data on N2 are available. During this time the spacecraft was in closed orbits at 10 km from the nucleus centre. Black symbols are data from 1 September 2014 to 31 March 2015. For both periods, O2 clearly shows the strongest correlation with H2O. While CO shows a high correlation with H2O from 17 to 23 October 2014, the correlation for the whole data set is fairly low. N2 shows the weakest correlation with H2O of all three species. The strong correlation between H2O and O2, with a Pearson correlation coefficient of 0.88 (and even 0.97 for the October data), indicates that they are of similar origin in the nucleus and that their release mechanisms are linked, in contrast to CO and N2 which have a similar volatility but do not show a strong correlation with H2O (see Fig. 2 for correlation coefficient values). The O2/H2O ratio decreases for high H2O abundances, which might be caused by surface water ice produced by a cyclic sublimation–condensation process8, although the total amount of surface ice is limited9.

Figure 2: Correlation between H2O and O2, CO and N2.
figure 2

a, H2O and O2; b, H2O and CO; c, H2O and N2. All three panels share a common y axis. Numbers on x and y axes are proportional to number density but in arbitrary units. Red crosses mark a subset of data for which N2 data are also available. Panel a shows the strong correlation between H2O and O2, which is observed for all data. In contrast, the correlation of CO with H2O (b) varies over time, which leads to a low overall correlation between those two species. N2 has the lowest correlation with H2O of the compared species for the October data (c).

PowerPoint slide

Full size image

A plausible mechanism for the strong O2/H2O correlation would be the production of O2 by radiolysis or photolysis of water ice. Here we follow the convention that photolysis refers to ultraviolet photons breaking bonds, whereas radiolysis refers to more energetic photons or fast electrons and ions depositing energy into the ice and ionizing molecules. Creation of sputtered O2 by radiolysis has been demonstrated in laboratory experiments10 and is observed for the icy moons of Jupiter—Europa, Ganymede and Callisto11,12,13—as well as for the rings of Saturn3. Comets are subject to radiolysis over various timescales: (a) over billions of years, while they reside in the Kuiper belt; (b) over the period of a few years once they enter the inner Solar System; and (c) on very short timescales, as for the present radiolysis. In the Kuiper belt, the skin depth for producing O2 is in the range of metres, although the produced O2 may diffuse deeper into the porous nucleus. Once a comet begins its residence in the inner Solar System, it loses its surface material to a depth of several metres during each orbit around the Sun, therefore we can safely assume that no O2 from radiolysis in the Kuiper belt phase remains in 67P at the percentage level. Radiolysis and photolysis by solar wind and ultraviolet radiation in the inner Solar System only affect the top few micrometres of the cometary surface. Taking account of 67P’s continuous mass loss through outgassing, we estimate the actively outgassing surface areas to be lost to a depth of several centimetres over the time from August 2014 to March 2015. If recent production by radiolysis or photolysis (only affecting the top few micrometres) were the source of the measured O2, our data would show a continuous decrease of the O2/H2O ratio over the examined time period as the active surface continues to be shed over that time. Apart from the variations related to H2O abundance, Fig. 3 shows that we do not observe a systematic change in the O2/H2O ratio over several months. Instantaneous creation of the measured O2 by radiolysis or photolysis seems, overall, unlikely, and would lead to variable O2 ratios due to different illumination conditions. Given that radiolysis and photolysis, on any of the discussed timescales, do not seem to be plausible production mechanisms, the preferred explanation of our observations is the incorporation of primordial O2 into the cometary nucleus.

Figure 3: O2/H2O ratio over several months.
figure 3

There seems to be no systematic increase or decrease of the O2/H2O ratio. The variances happen on very short timescales and can be explained by the decrease of the O2 ratio for high H2O abundances. It is not fully understood if the higher variability of the O2 ratio from October to the end of December 2014 can be attributed to orbital changes of the spacecraft or to physical changes of the cometary nucleus.

PowerPoint slide

Full size image

Despite great efforts by remote sensing campaigns, information on primordial O2 is still limited. Solid O2 has not yet been detected in interstellar ices, and upper limits for the O2/H2O ice ratios of <0.5 and for O2/CO ratios of <1 are in agreement with our findings, but such high upper limits do not provide useful constraints14,15. Gaseous O2 has only been detected in two interstellar clouds so far4,16,17, and is generally known to have surprisingly low abundances4. Reports of very low upper limits for O2 in a protostellar envelope suggest the material infalling to the accretion disk is very poor in molecular oxygen18. This has been ascribed to the high volatility and reactivity of O2, in particular the rapid transformation of O and O2 to H2O ice on cold grains19.

However, in one of the two interstellar clouds where O2 has been detected (the ρ Ophiuchi dense core), the chemically related species HO2 and H2O2 have been measured20,21 with gaseous abundance ratios of HO2/O2 ≈ H2O2/O2 ≈ 0.6 × 10−3. Interestingly, the abundance ratios determined by DFMS for the coma of 67P are very close to these interstellar values: HO2/O2 = (1.9 ± 0.3) × 10−3 and H2O2/O2 = (0.6 ± 0.07) × 10−3 (see Fig. 4). If these gas-phase abundance ratios reflect those in the cometary ice, this would support the existence of primordial O2. The ρ Ophiuchi A core has been suggested to have experienced slightly higher temperatures of around 20–30 K over its lifetime (which is also typical of estimates for the comet-forming conditions in the outer early solar nebula), compared to ∼10 K for most other dense interstellar clouds21,22. If higher temperatures are indeed needed to produce significant amounts of O2, this would indicate that our Solar System was formed from an unusually warm molecular cloud, consistent with the low abundance of N2 in 67P (ref. 23).

Figure 4: DFMS spectra for some of the common products of radiolysis of water ice.
figure 4

a–c, Products are O2H (seen in a), 16O18O and O2H2 (seen in b) and O3 (not seen in c). These data were recorded on 20 October 2014 at around 01:00 utc. With the exception of O3 (c), all the previously mentioned species are measured and can clearly be identified in the mass spectra of DFMS. The grey area consists of the statistical error and a 10% uncertainty from the individual pixel gains on the detector (here N is the number of molecules on the detector).

PowerPoint slide

Full size image

One aspect of the present results that remains unexplained is the high value (a few per cent) of the O2/H2O ratio in 67P. Models of gas-grain chemistry in molecular clouds predict O2/H2O ratios at least an order of magnitude lower19. They also over-predict ozone (O3): we found no evidence for the presence of ozone (see Fig. 4), with an upper limit of 1 × 10−6 relative to water.

An alternative explanation for the presence of O2 is the incorporation of gaseous O2 into water ice in the protosolar nebula, and chemical models of protoplanetary disks do in fact show high abundances of gaseous O2 in the comet forming zone24. Rapid cooling from >100 K to less than 30 K would then be needed to form amorphous water ice with trapped O2 on dust grains. This could happen when young disks experience increased heating due to accretion bursts onto the star, followed by a rapid drop in temperature as soon as the burst is over. These O2 rich grains then need to be accreted into larger bodies before further chemical modification occurs.

Finally, we discuss radiolysis of icy grains before accretion. When produced by radiolysis in water ice, O2 can remain trapped in voids, while hydrogen can diffuse out10. This prevents the hydrogenation of O2, which is otherwise a dominant reaction for the destruction of molecular oxygen, and could lead to increased and stable levels of O2 in the solid ice25. Incorporation of such icy grains into the comet nucleus would explain the observed strong correlation with H2O, in contrast to N2 which is trapped from the gas phase and shows a lower correlation with water. However, O3 (resulting from O2 radiolysis) has been reported to be trapped in Ganymede’s surface25, and at such a concentration would just be detectable by DFMS for O2 levels at a few per cent of H2O, but no O3 could be detected. A further consequence would be that these icy grains have been incorporated into the comet mostly unaltered, a process that is much debated, but which has recently been proposed again26 and that would also be in accordance with the measured high D/H ratio in 67P (ref. 27).

We note that our findings do not significantly affect our understanding of the global distribution of elemental O in the interstellar medium, as O2 in ice with an abundance of a few per cent relative to H2O accounts only for a small fraction of the total oxygen inventory.

Methods

Data reduction

The integration time for all evaluated spectra is 20 s. A single mass spectrum consists of the signals of 512 individual pixels, which are arranged along the dispersive axis of the mass spectrometer. We take the gain degradation of each individual pixel over time into account by determining its gain with a calibration sequence dedicated to this task every few weeks.

For every recorded mass spectrum, a third-degree polynomial is fitted to the baseline of the mass spectrum and subsequently subtracted from the signal. To avoid contamination in the baseline fit from the measured peaks, the centre of the mass spectrum (with the peaks) is ignored for those fitting purposes. Peaks are fitted with a Gaussian-shape curve. This introduces a systematic underestimation of the absolute magnitude of the signal, as at the 1% level and below the peaks are broader than Gaussian, but cancels out for the reported ratios of O2/H2O as both species are underestimated similarly. In a post-processing procedure, outliers are removed. An outlier occurs if one of the following criteria is met: (a) the signal amplitude changes by more than two orders of magnitudes between two measurements; (b) the centre of the fitted peak is not within a given window around the predicted position; (c) the width of the fitted peak curve is wider than a set limit.

For the CO measurements, the fraction contributed to the signal by CO fragments from CO2 is subtracted. As DFMS measures one mass line at a time in high resolution mode, there are no simultaneous measurements of two species that are separated by more than 1 Da/e. To calculate ratios, the two measurements have to occur within 20 min; if no pair of measurements can be found within 20 min, the data point is ignored. In our error analysis we include uncertainties for the individual gain levels of the detector, the species branching ratios and their relative sensitivities as systematic errors. Individual pixel gains are treated as statistical error, as the peak position on the detector is not constant. A careful analysis confirmed that the O2/H2O ratio is independent of the peak centre positions.

Sample size

No statistical methods were used to predetermine sample size.

Spacecraft outgassing background

To clearly identify the measured O2 as cometary in origin, all non-cometary sources of O2 must be considered and excluded. The Rosetta spacecraft produces a neutral gas cloud of its own, mainly due to diffusion of volatiles out of spacecraft material and desorption of re-deposited volatiles from the spacecraft. For example, by changing the spacecraft attitude, different spacecraft elements are illuminated by the Sun, which then warm up and release condensed gas. The orange line in Fig. 1 shows the low-level signals from this spacecraft contamination for O2, S and CH3OH, referred to as “background”. Those measurements were taken several days before the encounter with 67P. It is not possible to distinguish this background signal from any potential cometary signature with DFMS, but it has been well characterized before the arrival at 67P and is usually orders of magnitude lower than the measured O2 signals28. To keep the background influence as low as possible, we only considered mass spectra where both the O2 and H2O abundances are at least 5 times larger than the corresponding spacecraft contamination. Another potential source of O2 is the oxidizer, N2O4, used by the Rosetta spacecraft during thruster firings. Measurements taken shortly after a large thruster firing manoeuvre from June 2014 (still before arrival at the comet) show minor contaminations around 32 Da/e, but not directly affecting the O2 peak (see green curve in Fig. 1). Although contamination from thruster firings is small, DFMS measurements are usually performed hours after thruster firings, in order to minimize influence thereof. Finally we can exclude the production of O2 inside the instrument through a careful review of all oxygen-bearing molecules up to 150 Da/e, which could potentially fragment into O2 in the DFMS electron impact ion source. Many minor species contain O2 but these are too low in abundance to account for the large amount of O2 detected. The remaining possibility is CO2, which is very abundant in the coma of 67P (ref. 6). However, owing to its molecular structure it only fragments into CO and O, not into O2 (ref. 29). Finally, we exclude the production of O2 from H2O in the instrument. For 81 mass spectra taken from May to the end of June 2014 (before the encounter with 67P) we determine an O2 abundance of (0.18 ± 0.07)% relative to H2O, which is a factor of 20 lower than the cometary values.

Coma chemistry

The production of O2 from atomic oxygen in the neutral gas coma of 67P is possible through the reaction

Owing to the tenuous nature of the coma, only very few collisions are expected within the first 100 km above the nucleus surface30. With increasing activity and therefore denser coma one would expect a higher O2/H2O ratio as both reactants (atomic oxygen and OH) are related to H2O as well as the collision frequency. This is clearly not observed. Production of a significant amount of O2 from coma chemistry hence seems unfeasible.

Correlation with H2O

The measured O2 signal shows a very strong dependence on radial distance (r) from the comet. It increases by roughly one order of magnitude when the radial distance from the comet decreases from 30 km to 10 km. This is in agreement with a predicted 1/r2 dependence of the number density profile of a non-reactive species. Examining the data further, we observe a strong correlation between H2O and O2 (see Fig. 2, Pearson correlation coefficient R = 0.88) for data from September 2014 to March 2015. This correlation indicates that O2 and H2O are both of a similar cometary origin. In contrast, there is no correlation between O2 and H2O for measurements taken before the arrival at the comet (R = −0.01). The observed temporal variations in the O2/H2O ratio are largely due to a nonlinear correlation between H2O and O2 for high water densities, where the O2 ratio drops with increasing H2O abundance. The correlation similarly supports the ruling out of CO2 as a source of the O2, as previous studies6,23 and Fig. 2 show a lack of correlation between the abundance of H2O and other species like CO, CO2 and N2.

Radiolysis

The production of O2 from water ice by radiolysis is the result of several reactions, where initially H, O and OH are produced, followed by subsequent rearrangement to form H2, HO2 and H2O2 and ultimately O2 (ref. 2). In a systematic study of the irradiation of pure cubic crystalline water ice at 10 K, it was estimated that comets in the Oort cloud would reach an equilibrium concentration of about 0.6% of H2O2 after ∼105 years, and that the concentration of O2 would be significantly lower31. As stated by the authors, the relative proportion of O2 and H2O2 in actual astrophysical ices can be different as the presence of H2O2 ices may enhance the production of O2 and vice versa.

Cosmic rays consist of energetic particles (mostly H+) that can penetrate inside the cometary nucleus. Their penetration depth depends on their energy. The more energetic a cosmic ray, the deeper it deposits its energy into the cometary nucleus. Only the most energetic components, the Galactic cosmic rays (GCRs), can penetrate deeply into the cometary body. Today, the GCR flux peaks at energies around a few hundreds of MeV. At energies higher than ∼1 GeV, the GCR flux drops exponentially, the flux at 10 GeV being approximately 100 times lower than the flux at 1 GeV. GCRs with energies near the GCR maximum of flux typically deposit their energy in the nucleus at depths of the order of metres to tens of metres32. There is some uncertainty on the evolution of the cosmic ray flux over the history of the Solar System related to the occurrence of supernovae which can drastically increase the cosmic ray flux at high energies (up to ∼PeV) during relatively short time periods (of the order of 105 years) or to the orbital motion of the Sun and of the Galaxy. There are also uncertainties on the cross-sections, the relative proportion of high-Z cosmic rays, and the role of porosity and defects in the cometary ice33.

During 67P’s stay in the Kuiper belt (∼4.5 billion years) GCRs may have produced a significant amount of O2. This O2 is produced where the GCR energy is deposited, that is, in the first tens of metres below the comet surface. Once cometary objects enter the inner Solar System they lose (owing to their activity and depending on their perihelion distance) material from the surface in the range of several metres per orbit around the Sun. Therefore, as most products built up during the stay in the Kuiper belt reside in the outer few metres, O2 produced by radiolysis should be released quickly on the first solar passages. 67P’s perihelion distance has been within the orbit of Jupiter for the past 250 years, possibly even for more than 5,000 years, and since a close encounter with Jupiter in 1959, the perihelion distance of 67P has been about 1.3 au with an orbital period of 6.4 years (ref. 34). Accumulated over the last perihelion passages we can assume that 67P has lost hundreds of metres from its surface.

However, O2 molecules may have diffused into the comet and penetrated into deeper regions of the cometary nucleus. According to the estimated erosion of the comet, this requires significant diffusion of the O2 molecules. Such efficient diffusion would dilute the produced O2, resulting in a lower O2/H2O ratio hardly compatible with the elevated amount of O2 observed by DFMS.

Recent radiolysis

The flux of the low energy component such as solar energetic particles (SEP) is several orders of magnitude higher than the flux of GCRs. The SEP flux peaks at energies around a few tens of MeV and only penetrates the first centimetres of the comet32. Considering the comet erosion rate, the production of O2 by low energy cosmic rays would have to be very fast to account for the observed amount of O2, requiring unrealistically high O2 production rates. Consequently, it appears unlikely that most of the O2 in 67P has been built up over the past few years.

O2 production through surface sputtering

It has been shown that sputtering of refractory materials from the cometary surface due to the solar wind is occurring at 67P (ref. 35) and a clear difference between the southern and northern latitudes in the measured abundances of the sputtered species was demonstrated. This apparent spatial difference is explained via the asymmetry of the neutral coma, with higher number densities in the northern hemisphere that is preferentially exposed to the Sun for the time span under study. The solar wind, which is responsible for the sputtering, is therefore attenuated more efficiently by these denser parts of the coma and thus has limited or no access to the surface. We find that the O2/H2O ratio is independent of latitude and relatively constant over a period of 7 months. Furthermore, the major part of the top surface of 67P accessible by the solar wind does not contain any water ice9. This suggests that sputtering cannot be the main source of the detected molecular oxygen. Moreover, the sputter yields by solar wind ions are orders of magnitude too low to explain the observed amount of O2.

Code availability

The data reduction software was written in the Julia and Python programming languages and is available upon request from A.B.