Stellar Occultation by Comet 67P/Churyumov-Gerasimenko Observed with Rosetta's Alice Far-Ultraviolet Spectrograph

TL;DR

This study reports on a serendipitous stellar occultation observed by Rosetta's Alice spectrograph, revealing detailed measurements of water and oxygen absorption near Comet 67P, and analyzing their spatial distribution and correlation.

Contribution

First detection of O2 absorption in a stellar occultation by Comet 67P, providing high-resolution data on H2O and O2 distribution close to the nucleus.

Findings

O2 and H2O column densities decrease with impact parameter.

O2 decreases faster than H2O with distance from nucleus.

O2/H2O ratio decreases as H2O column density increases.

Abstract

Following our previous detection of ubiquitous H2O and O2 absorption against the far-UV continuum of stars located near the nucleus of Comet 67P/Churyumov-Gerasimenko, we present a serendipitously observed stellar occultation that occurred on 2015 September 13, approximately one month after the comet's perihelion passage. The occultation appears in two consecutive 10-minute spectral images obtained by Alice, Rosetta's ultraviolet (700-2100 A) spectrograph, both of which show H2O absorption with column density $>10^{17.5} \mathrm{cm}^{-2}$ and significant O2 absorption ($\mathrm{O2/H2O} \approx 5$-10%). Because the projected distance from the star to the nucleus changes between exposures, our ability to study the H2O column density profile near the nucleus (impact parameters $<1$ km) is unmatched by our previous observations. We find that the H2O and O2 column densities decrease with increasing impact parameter, in accordance with expectations, but the O2 column decreases $\sim3$ times more quickly than H2O. When combined with previously published results from stellar appulses, we conclude that the O2 and H2O column densities are highly correlated, and O2/H2O decreases with increasing H2O column.

Figures

Equations

$$n=\frac{fQ}{4\pi r^{2}v}$$

$$N$$

$$=\frac{fQ}{4\pi v}{\int }_0^{\infty }\frac{ds}{s^2+r_{\mathrm{sc}}^2-2s{r}_{\mathrm{sc}}\cos \theta }$$

$$Q(R_h)$$

$$v(R_h)$$

$$\frac{Q(R_h)}{v(R_h)}=\frac{4\pi N/f}{{\int }_0^{\infty }(s^2+r_{\mathrm{sc}}^2-2s{r}_{\mathrm{sc}}\cos \theta {)}^{-1}ds}$$

Full text

comets: individual (67P) – ultraviolet: planetary systems

††journal: AJ

\turnoffedits

1 Introduction

The double-focusing mass spectrometer (DFMS) of the Rosetta Orbiter Spectrometer for Ion and Neutral Analysis (ROSINA; Balsiger et al., 2007) has found that \ceO2 is the fourth most abundant parent species in the coma of Comet 67P/Churyumov-Gerasimenko (67P/C-G), behind only \ceH2O, \ceCO2, and \ceCO (Le Roy et al., 2015; Fougere et al., 2016). The ubiquitous, abundant presence of \ceO2 was surprising since it had never been detected in a comet before (Bieler et al., 2015).

Feldman et al. (2016) confirmed the presence of substantial \ceO2 in the coma of 67P/C-G during gaseous outbursts with Alice, Rosetta’s ultraviolet spectrograph (Stern et al., 2007). Later, Keeney et al. (2017) directly detected \ceO2 absorption in Alice data using stellar sight lines temporarily projected near the nucleus (“stellar appulses”). These sight lines were observed over a wide range of heliocentric distances (1.2-2.3 AU) and impact parameters (4-20 km), yielding $\log{N_{\ce{H2O}}}=15.2$-17.1 (all column densities, $N$, herein are quoted in units of $\mathrm{cm}^{-2}$) and a median value of $N_{\ce{O2}}/N_{\ce{H2O}}=25$%.

Several Rosetta instruments can directly detect \ceH2O in the coma of 67P/C-G: Alice, ROSINA, the Visible and Infrared Thermal Imaging Spectrometer (VIRTIS; Coradini et al., 2007), and the Microwave Instrument for the Rosetta Orbiter (MIRO; Gulkis et al., 2007). All but ROSINA are remote-sensing instruments that measure or infer column densities along a line of sight, and $N_{\ce{H2O}}$ measured by Alice (Keeney et al., 2017) agrees with VIRTIS values near perihelion (Bockelée-Morvan et al., 2016). However, only Alice and ROSINA can directly detect \ceO2, and the relative abundance of $\ce{O2}/\ce{H2O}$ in the Alice data was nearly an order of magnitude larger than the average ROSINA value ($n_{\ce{O2}}/n_{\ce{H2O}}=3.85\pm 0.85$%; Bieler et al., 2015, where $n$ is number density measured at the spacecraft position). The consistency in $N_{\ce{H2O}}$ between Alice and VIRTIS suggests that the methodology of Keeney et al. (2017) is trustworthy, but the discrepancy in $\ce{O2}/\ce{H2O}$ remains puzzling.

Here we present a stellar occultation by 67P/C-G observed with Alice. We describe our observations in Section 2, and our analysis procedure in Section 3. Section 4 compares the \ceH2O and \ceO2 column densities for our stellar occultation with those of Keeney et al. (2017) and ROSINA measurements (Bieler et al., 2015; Hansen et al., 2016), and Section 5 summarizes our main findings.

2 Observations

The A0 IV star HD 4150 was occulted by 67P/C-G on 2015 September 13, approximately one month after the comet’s perihelion passage. The occultation was serendipitously observed by Alice during the course of normal operations. During this phase of the mission, Alice was integrating nearly continuously to catalog emissions from the near-nucleus coma (Pineau et al., 2019).

Figure 1 shows a $\sim 5\times 7.5~{}\mathrm{km}^{2}$ subsection of a Rosetta navigation camera (NAVCAM) image taken at 13:36:02 UT, while HD 4150 was occulted by 67P/C-G and $\sim 20$ minutes before it entered the Alice slit. The Alice slit is $5\fdg 5$ long and has a dog-bone shape that is twice as wide at the top and bottom as in the center (Stern et al., 2007); its position with respect to the nucleus is shown in red in Figure 1, and the transition between the wide-bottom and narrow-center regions of the slit is evident. Figure 1 also shows the approximate path of HD 4150 as it emerges from behind the small lobe of 67P/C-G and crosses the Alice slit from left to right at an angle of $\sim 35\arcdeg$ over the “neck” of the nucleus.

HD 4150 appears in two consecutive 10-minute spectral images with start times of 13:52:04 and 14:02:49 UT, respectively. For the first five minutes of the first exposure, the star is occulted by the comet nucleus. The star remains in the Alice slit for the remainder of the first exposure and the first five minutes of the second exposure. During these exposures, 67P/C-G was 1.30 AU from the Sun, Rosetta was orbiting 313 km from the comet center, and the solar phase angle was $108\arcdeg$.

The spectra of HD 4150 extracted from these exposures are shown in green (first exposure) and brown (second exposure) in Figure 2. The observed stellar fluxes are corrected for the reduced amount of time the star was in the slit during these exposures. Nevertheless, differences between the two exposures are evident; most notably, the first (green) spectrum has less flux from 1350-1600 Å.

HD 4150 was re-observed (“revisited”) on 2016 June 6 when it was far from the nucleus, at an off-nadir angle of $79\arcdeg$. The purpose of this 15-minute integration, which started at 06:02:24 UT, was to characterize the intrinsic stellar spectrum without the presence of foreground coma absorption (see Keeney et al., 2017 for details). When HD 4150 was revisited, the comet was 3.15 AU from the Sun and the solar phase angle was $67\arcdeg$. The intrinsic stellar spectrum is shown in gray in Figure 2, and has considerably more flux from 1250-1800 Å than the spectra obtained immediately after the occultation.

The analysis below, and that of Keeney et al. (2017), assumes that the stellar spectrum obtained when the star was revisited while far from the nucleus is equivalent to the intrinsic stellar spectrum (i.e., that there is no foreground coma absorption at that time). Although it is true that \ceO2 is extremely volatile, with a sublimation temperature in vacuum of $\sim 30$ K (Fray & Schmitt, 2009), it has been found to be strongly correlated with \ceH2O in the coma of 67P/C-G (Bieler et al., 2015; Fougere et al., 2016), and the production rate of \ceH2O was down by 2-3 orders of magnitude at 3.15 AU compared to 1.30 AU (Hansen et al., 2016; Gasc et al., 2017). However, even if our assumptions are incorrect and a small amount of foreground coma absorption is present in the revisit spectrum, then this would cause us to underestimate the amount of foreground coma absorption at the time of the occultation111The same can be said for the stellar appulse analysis of Keeney et al. (2017)..

3 Results and Analysis

The spectra of HD 4150 taken immediately after the occultation by 67P/C-G were analyzed analagously to the stellar appulse spectra of Keeney et al. (2017), except in one regard. The UV-bright stars that Keeney et al. (2017) used to study the near-nucleus coma all had measurable far-UV flux down to $\sim 900$ Å (see Figure 1 of Keeney et al., 2017), but HD 4150 is an A0 star with almost no flux below $\sim 1250$ Å (see Figure 2). This lack of flux blueward of 1250 Å has important consequences for our analysis.

Although \ceH2O and \ceO2 have relatively large cross sections from 1250-1800 Å (Chung et al., 2001; Yoshino et al., 2005), other abundant coma species do not (i.e., \ceCO and \ceCO2; Cairns & Samson, 1965; Yoshino et al., 1996), which means that they cannot be directly constrained by our data. Fortunately, at the time of the occultation VIRTIS was acquiring data while pointed $\sim 3.6$ km above the sunward limb of the nucleus. Thus, we adopt their contemporaneous column density ratio of $N_{\ce{CO2}}/N_{\ce{H2O}}=0.310\pm 0.034$ (Bockelée-Morvan et al., 2016) for our analysis. The position of the VIRTIS-H aperture with respect to the Alice slit and the path of HD 4150 are shown in Figure 1.

To isolate the coma signature from the intrinsic stellar flux and interstellar absorption, we divided the stellar spectra taken immediately post-occultation (i.e., the green and brown spectra in Figure 2) by the spectrum of the star taken much later (i.e., the gray spectrum in Figure 2), after first scaling them to have the same median flux from 1850-1950 Å. This procedure reduces our sensitivity to the uncertainty in the amount of time the star was in the slit during our long exposures (Keeney et al., 2017). The resulting normalized spectrum quantifies the amount of foreground coma absorption in the post-occultation exposures.

Keeney et al. (2017) modeled far-UV absorption from ten molecular species (see their Table 3 and Figure 2 for adopted cross sections) in Alice spectra normalized as above. We use the same procedure here, except that we only fit wavelengths in the range 1250-1950 Å because HD 4150 has insufficient flux blueward of 1250 Å (see Figure 2). Consequently, we remove \ceCO from our fits because it has no appreciable cross section redward of 1050 Å (Cairns & Samson, 1965).

The column density of \ceH2O is fit directly and allowed to vary in the range $\log{N_{\ce{H2O}}}=14$-18. The abundances of all other species are fit relative to $N_{\ce{H2O}}$; \ceO2 is allowed to vary in the range $N_{\ce{O2}}/N_{\ce{H2O}}=0$-1, \ceCO2 is fixed at the ratio measured by VIRTIS ($N_{\ce{CO2}}/N_{\ce{H2O}}=0.310\pm 0.034$; Bockelée-Morvan et al., 2016), and all other species (\ceCH4, \ceC2H2, \ceC2H6, \ceC2H4, \ceC4H2, and \ceH2CO) are constrained to have $N/N_{\ce{H2O}}<0.01$ (Le Roy et al., 2015). All species except \ceCO and \ceCO2 are treated the same as they were in Keeney et al. (2017).

Our fits to the spectra of HD 4150 taken immediately after its occultation by 67P/C-G are shown in Figure 3-4. The top panels show the normalized spectra in black, and our best-fit absorption profiles from \ceH2O (blue), \ceO2 (green), \ceCO2 (brown), and all other species (purple). The ensemble fit from all species is shown in pink, and the shaded regions represent 95% ($2\sigma$) confidence bands. Recall that $N_{\ce{CO2}}/N_{\ce{H2O}}$ is held fixed at the value measured contemporaneously by VIRTIS (Bockelée-Morvan et al., 2016). The bottom panel shows the residual of the ensemble fit as a function of wavelength, with $1\sigma$ flux uncertainty overlaid in orange. Masked regions that are not used to constrain the fits are shown in lighter hues in both panels.

We searched for systematic offsets in the best-fit \ceH2O and \ceO2 column densities using Monte Carlo simulations to compare the values retrieved from forward-modeled data with Poissonian noise and known input values. However, unlike in Keeney et al. (2017), we found no evidence for systematic offsets. Reassuringly, the large optical depth of the \ceH2O absorption makes the fitting results more robust. Thus, we adopt the \ceH2O and \ceO2 column densities returned by our fits as final.

4 Discussion

The decrease in the \ceH2O column density from the first exposure of HD 4150 to the second (see Figure 3-4) agrees with the naïve expectation that $N\propto\rho^{-1}$ (Haser, 1957), where $\rho$ is the impact parameter with respect to the nucleus. $N_{\ce{H2O}}$ decreases by a factor of $\sim 1.4$ between the two exposures when $\rho$ increases by a factor of $\sim 1.3$ (we estimate $\rho\approx 0.3$ km and 0.4 km for the first and second exposures, respectively; see Figure 1). However, the contemporaneous VIRTIS measurement ($\log{N_{\ce{H2O}}}=17.00\pm 0.04$ at $\rho\approx 2.7$ km; Bockelée-Morvan et al., 2016)222Bockelée-Morvan et al. (2016) list $\rho=3.6$ km for the contemporaneous exposure, which is measured from the center of the nucleus. The value we list is measured with respect to the limb of the nucleus for consistency with Figure 1. is not as consistent with $N\propto\rho^{-1}$, finding a factor of $\sim 6$ decrease in $N_{\ce{H2O}}$ compared to the first post-occultation exposure when $\rho$ increases by a factor of $\sim 9$.

Of course, the Haser (1957) model, which assumes spherically symmetric outflow from a point source, is far too simplistic to be directly applied to data from Rosetta 67P/C-G. From Rosetta’s vantage point embedded in the coma of 67P/C-G, the nucleus is clearly resolved and far from spherical (Figure 1), and observed gas (e.g., Migliorini et al., 2016) and dust (e.g., Gerig et al., 2018) outbursts are not axisymmetric. Thus, it is questionable whether the analytic prediction of Haser (1957, namely, $N\propto\rho^{-1}$) holds for Alice data.

The best-fit values of the \ceO2 column density are $\log{N_{\ce{O2}}}=16.82\pm 0.04$ for the first post-occultation exposure (Figure 3) and $\log{N_{\ce{O2}}}=16.20\pm 0.12$ for the second exposure (Figure 4), corresponding to a factor of $\sim 4$ decrease. It is unclear why $N_{\ce{O2}}$ is decreasing almost three times more quickly than $N_{\ce{H2O}}$. Keeney et al. (2017) found no clear trend in $N_{\ce{O2}}$ as a function of impact parameter when $\rho\approx 5$-15 km (see their Figure 9). However, the exposures of HD 4150 probe the coma much closer to the nucleus than any of our previous sight lines, so we cannot confidently extrapolate the Keeney et al. (2017) measurements to $\rho<1$ km.

The most puzzling result of Keeney et al. (2017) was that the relative abundance of \ceO2 with respect to \ceH2O inferred from Alice data (median $N_{\ce{O2}}/N_{\ce{H2O}}=25$%) was nearly an order of magnitude larger than the average value of $n_{\ce{O2}}/n_{\ce{H2O}}=3.85\pm 0.85$% found by ROSINA (Bieler et al., 2015). In fact, the datasets almost seemed to be mutually exclusive; Keeney et al. (2017) did not find a single high-quality example with $N_{\ce{O2}}/N_{\ce{H2O}}<10$% (although there were several upper limits), and ROSINA almost never measured $n_{\ce{O2}}/n_{\ce{H2O}}>10$% (Bieler et al., 2015; Fougere et al., 2016). Keeney et al. (2017) speculated that the discrepancy could be attributed to comparing column density along a line of sight near the nucleus with number density measured in situ at the spacecraft position, but no firm conclusions could be drawn. It is therefore reassuring that we measure a relative abundance of $\ce{O2}/\ce{H2O}$ that is nearly identical to the average ROSINA value in one of the post-occultation exposures (see Figure 4).

However, whereas Keeney et al. (2017) did not find any instances of $\ce{O2}/\ce{H2O}$ consistent with ROSINA measurements in absorption, the emission-line technique of Feldman et al. (2016) frequently infers $\ce{O2}/\ce{H2O}$ values near the sunward limb of the nucleus in Alice data that are consistent with ROSINA values. For example, an estimate of $\ce{O2}/\ce{H2O}$ can be derived from exposures taken just before and just after the occultation using the ratio of the semi-forbidden O I] 1356 Å line to H I Ly$\beta$. If we assume that all of the O I] 1356 Å emission near the limb comes from electron impact on \ceH2O (Makarov et al., 2004), \ceO2 (Kanik et al., 2003), and \ceCO2 (Mumma et al., 1972) at an energy of 100 eV, then the brightness of the C I 1657 Å line suggests that 25-30% of the 1356 Å brightness comes from electron impact dissociation of \ceCO2. Similarly, large off-nadir steps along the slit suggest that 25-30% of the H I Ly$\beta$ brightness comes from resonant scattering in the coma. After these corrections, we estimate that $\ce{O2}/\ce{H2O}\approx 4$% in the Alice exposures surrounding the occultation. Although this estimate is predicated on many assumptions, it is nevertheless reassuring that it is identical to the value we derived for our second exposure (see Figure 4).

The dust outburst peaked at approximately 13:30:00 UT (Rinaldi et al., 2018), $\sim 20$ minutes before our first post-occultation exposure began. At the beginning of the first Alice exposure, the radiance of the dust emission had decayed to $\sim 20$% of its peak compared to the quiescent level, and by the beginning of the second Alice exposure it had returned to the quiescent level altogether (see Figure 4 of Rinaldi et al., 2018). Thus, one plausible explanation for the differing $\ce{O2}/\ce{H2O}$ levels in the two Alice exposures is a non-constant production rate of \ceO2 (i.e., the amount of \ceO2 in the first exposure is affected by the dust outburst).

Figure 5 shows $N_{\ce{O2}}$ and $N_{\ce{O2}}/N_{\ce{H2O}}$ as a function of $N_{\ce{H2O}}$ for the two post-occultation exposures of HD 4150 and the stellar appulses of Keeney et al. (2017). Owing to the smaller impact parameters probed, our occultation data find larger $N_{\ce{H2O}}$ values and smaller $N_{\ce{O2}}/N_{\ce{H2O}}$ than the appulses, and overall there is a clear trend of decreasing $\ce{O2}/\ce{H2O}$ with increasing $N_{\ce{H2O}}$ in Alice data (e.g., when $N_{\ce{H2O}}<10^{16}~{}\mathrm{cm}^{-2}$ the median $N_{\ce{O2}}/N_{\ce{H2O}}=41$%, whereas the median $N_{\ce{O2}}/N_{\ce{H2O}}=16$% for larger $N_{\ce{H2O}}$).

This trend, , suggests that we are able to detect an $\ce{O2}/\ce{H2O}$ abundance consistent with the ROSINA measurements in Figure 4 simply because the \ceH2O column density is sufficiently large. A $\mathrm{S/N}$-dependent detection threshold for $N_{\ce{O2}}$ in Alice data is consistent with this trend, but we see no clear evidence that our measurements are strongly affected by such a selection bias (Figure 5). Further, this hypothesis cannot explain all of the Alice measurements since Keeney et al. (2017) found two examples where $N_{\ce{O2}}/N_{\ce{H2O}}>40$% when $N_{\ce{H2O}}>10^{16}~{}\mathrm{cm}^{-2}$. Thus, a full explanation for the discrepancy in $\ce{O2}/\ce{H2O}$ abundance between Alice and ROSINA measurements remains elusive.

[TABLE]

[TABLE]

[TABLE]

[TABLE]

5 Summary

We have presented far-UV spectra of the A0 IV star HD 4150, which was serendipitously observed by Rosetta’s Alice imaging spectrometer on 2015 September 13. HD 4150 was observed in two 10-minute integrations immediately after being occulted by 67P/C-G, and revisited approximately 9 months later when its line of sight was far from the nucleus. By comparing the two epochs of stellar spectra, we were able to quantify the amount of \ceH2O and \ceO2 within 1 km of the nucleus.

We find that $N_{\ce{H2O}}\propto\rho^{-1}$ in our consecutive exposures of HD 4150, but $N_{\ce{O2}}$ decreases $\sim 3$ times faster than $N_{\ce{H2O}}$. We have also measured a value of $N_{\ce{O2}}/N_{\ce{H2O}}$ that is consistent with ROSINA measurements of $n_{\ce{O2}}/n_{\ce{H2O}}$ (Bieler et al., 2015; Fougere et al., 2016). , we confirm the strong correlation between \ceO2 and \ceH2O first reported by ROSINA (Bieler et al., 2015; Fougere et al., 2016), find a general decrease in $\ce{O2}/\ce{H2O}$ with increasing $N_{\ce{H2O}}$ in Alice data . This trend of decreasing $\ce{O2}/\ce{H2O}$ with increasing $N_{\ce{H2O}}$ partially explains the initial discrepancy in $\ce{O2}/\ce{H2O}$ between Alice and ROSINA.

Several Rosetta instruments (e.g., Alice, ROSINA, VIRTIS, MIRO) can detect \ceH2O in the coma of 67P/C-G. However, only Alice and ROSINA can directly detect \ceO2, so the differing Alice and ROSINA measurements of $\ce{O2}/\ce{H2O}$ in the coma of 67P/C-G remain mysterious,

Rosetta is an ESA mission with contributions from its member states and NASA. We thank the members of the Rosetta Science Ground System and Mission Operations Center teams, in particular Richard Moissl and Michael Küppers, for their expert and dedicated help in planning and executing the Alice observations. The Alice team acknowledges continuing support from NASA via Jet Propulsion Laboratory contract 1336850 to the Southwest Research Institute. This research has made use of the SIMBAD database, operated at CDS, Strasbourg, France. Rosetta (Alice)