the origin of water vapor and carbon dioxide in jupiter's stratosphere

Icarus 159, 112–131 (2002)doi:10.1006/icar.2002.6929

The Origin of Water Vapor and Carbon Dioxide in Jupiter’s Stratosphere

E. Lellouch and B. Bezard

Observatoire de Paris, F-92195 Meudon, FranceE-mail: [email protected]

J. I. Moses

Lunar and Planetary Institute, Houston, Texas 77058-1113

G. R. Davis

University of Saskatchewan, Saskatoon, S7N 5E2, Canada

P. Drossart

Observatoire de Paris, F-92195 Meudon, France

H. Feuchtgruber

Max-Planck Institute fuer Extraterrestrische Physik, 85740 Garching, Germany

E. A. Bergin

Harvard–Smithsonian Center for Astrophysics, Cambridge, Massachusetts 02138

R. Moreno

Institut de RadioAstronomie Millimetrique, 38400 St-Martin d’Heres, France

and

T. Encrenaz

Observatoire de Paris, F-92195 Meudon, France

Received December 28, 2001; revised May 17, 2002

Observations of H2O rotational lines from the Infrared Space Ob-servatory (ISO) and the Submillimeter Wave Astronomy Satellite(SWAS) and of the CO2 ν2 band by ISO are analyzed jointly to de-termine the origin of water vapor and carbon dioxide in Jupiter’sstratosphere. Simultaneous modelling of ISO/LWS and ISO/SWSobservations acquired in 1997 indicates that most of the strato-spheric jovian water is restricted to pressures less than 0.5 ±0.2 mbar, with a disk-averaged column density of (2.0 ± 0.5) ×1015 cm−2. Disk-resolved observations of CO2 by ISO/SWS reveallatitudinal variations of the CO2 abundance, with a decrease of atleast a factor of 7 from mid-southern to mid-northern latitudes,and a disk-center column density of (3.4 ± 0.7) × 1014 cm−2. Theseresults strongly suggest that the observed H2O and CO2 primarilyresult from the production, at midsouthern latitudes, of oxygenatedmaterial in the form of CO and H2O by the Shoemaker–Levy 9 (SL9)impacts in July 1994 and subsequent chemical and transport evolu-tion, rather than from a permanent interplanetary dust particle or

satellite source. This conclusion is supported by quantitative evo-lution model calculations. Given the characteristic vertical mixingtimes in Jupiter’s stratosphere, material deposited at ∼0.1 mbar bythe SL9 impacts is indeed expected to diffuse down to the ∼0.5 mbarlevel after 3 years. A coupled chemical-horizontal transport modelindicates that the stability of water vapor against photolysis andconversion to CO2 is maintained over typically ∼50 years by thedecrease of the local CO mixing ratio associated with horizontalspreading. A model with an initial (i.e., SL9-produced) H2O/COmass mixing ratio of 0.07, not inconsistent with immediate post-impact observations, matches the observed H2O abundance andCO2 horizontal distribution 3 years after the impacts. In contrast,the production of CO2 from SL9-produced CO and a water com-ponent deriving from an interplanetary dust component is insuffi-cient to account for the observed CO2 amounts. The observationscan further be used to place a stringent upper limit (8 × 104 cm−2

s−1) on the permanent water influx into Jupiter. This may indicatethat the much larger flux observed at Saturn derives dominantly

112

0019-1035/02 $35.00c© 2002 Elsevier Science (USA)

All rights reserved.

WATER AND CO2 IN JUPI

from a ring source, or that the ablation of micrometeoroids leadsdominantly to different species at Saturn (H2O) and Jupiter (CO).Finally, the SWAS H2O spectra do not appear fully consistent withthe ISO data and should be confirmed by future ODIN and Herschelobservations. c© 2002 Elsevier Science (USA)

Key Words: Jupiter’s atmosphere; spectroscopy.

1. INTRODUCTION

The detection of water vapor in the upper atmospheres of thefour Giant Planets and Titan and of carbon dioxide in Jupiter,Saturn, and Neptune by the Infrared Space Observatory (ISO)has provided definite evidence for an external source of oxygenin these objects (Feuchtgruber et al. 1997, 1999, Lellouch et al.1997a, 1998, Coustenis et al. 1998, Moses et al. 2000a). The in-volved amounts are on the order of (1–30) × 1014 moleculescm−2 for H2O and (4–10) × 1014 molecules cm−2 for CO2.Oxygen is delivered at the top of the atmosphere in the formof H2O and/or other oxygen-bearing species (e.g., CO, CO2,CH3OH . . .),which undergo both downward transport andchem-ical conversion through reaction with the photolysis products ofmethane. From an elementary model, in which the sole lossmechanism of water was assumed to be transport to its conden-sation sink in the lower stratosphere, Feuchtgruber et al. (1997,1999) and Lellouch et al. (1997a, 1998) estimated the water in-put fluxes into the Giant Planets to be in the range 105–107 cm−2

s−1, with large uncertainties due to the uncertain time scale forvertical transport, described by the use of a vertical eddy dif-fusion (Kv) coefficient profile. A much more comprehensivemodel, including detailed chemical schemes and constraints onthe eddy Kv profile from the distribution of hydrocarbons, wasdeveloped for Saturn by Moses et al. (2000a). This model re-sulted in an improved determination of the total oxygen flux atSaturn, found to be (4 ± 2) × 106 cm−2 s−1.

The origin of this external flux was discussed by Feuchtgruberet al. (1997) and Moses et al. (2000a). Based on its ubiquitouscharacter and on the order of magnitude and similarity of theinput fluxes into the four Giant Planets, the most likely sourcewas concluded to be interplanetary dust particles (IDP). Moseset al. (2000a) further discussed the origin of this interplanetarydust and suggested a dominant production by the short-period,Halley-type comets. The contribution from planetary rings as thesource of vapor or solid debris into the planetary atmosphereswas also evaluated. The ring vapor source was estimated to be ∼2orders of magnitude less important than the direct IDP source.Ionization and diffusion of dust produced by ring erosion wasfound to be potentially more significant, at least in Saturn, pos-sibly producing localized enhancements in the H2O abundanceat specific latitudes magnetically connected to the rings. Such aring signature for Saturn’s water has been claimed from HST ob-servations (Prange et al. 1998), but Moses et al. (2000a) foundthat such a localized distribution of water did not provide as
good a fit to the disk-averaged ISO observations as a globaldistribution did.
TER’S STRATOSPHERE 113

Carbon monoxide, detected in Jupiter (Beer 1975), Saturn(Noll et al. 1986), and Neptune (Marten et al. 1993), is a cen-tral species in oxygen photochemistry, as it can be synthesizedfrom the photolysis of water and lead to the production of CO2.It is, in fact, the first claim by Beer and Taylor (1978) that COwas concentrated in Jupiter’s stratosphere that prompted the-oretical studies on the possible role of micrometeoritic bom-bardment and sputtered oxygen from the Galilean icy satellitesas sources of external oxygen in Jupiter (Prather et al. 1978,Strobel and Yung 1979). Unfortunately, as subsequent observa-tions (see reviews in Noll and Knacke 1998 and Bezard et al.2002) resulted in contradictory results as to the CO presenceand abundance in Jupiter’s stratosphere, reliable estimates ofthe oxygen influx were not possible. The situation is ambiguousas well in Saturn, where the vertical distribution of CO remainsunknown (Noll and Larson 1990, Moses et al. 2000a). Recently,from high-resolution 5-µm observations of Jupiter, Bezard et al.(2002) made a convincing case of an enhanced CO mixing ratioin Jupiter’s stratosphere compared to its tropospheric (∼6 bar)abundance, with implications on the CO stratospheric produc-tion rate.

The situation for Jupiter is especially complicated becauseof the chemical modifications of its stratosphere that resultedfrom the collision of Comet Shoemaker–Levy 9 (hereafter SL9)fragments in July 1994. Indeed, one of the many effects of thisevent was the delivery of massive amounts of oxygenated mate-rial at submillibar pressures (for reviews of the impact-inducedchemistry, see Lellouch 1996, Moses 1996, Zahnle 1996). Mostof this material was seen in the form of CO, with CO columndensities as high as 1.5 × 1018 cm−2 for the large impact sites,corresponding to CO masses of about 1.5 × 1014 g (Lellouchet al. 1997b). CO is chemically stable in Jupiter’s stratosphere(Moses 1996) and is indeed observed to persist since the SL9 im-pacts (Moreno 1998, Marten et al. 1998). Nontrivial amounts ofH2O were also measured, typically 1% of CO or more (Bjorakeret al. 1996a,b, Carlson et al. 1995a,b, Sprague et al. 1996,Encrenaz et al. 1997). Since CO2 can be built progressivelyfrom the photolysis of H2O (Moses et al. 1995, Moses 1996),it is conceivable that in addition to CO, at least a fraction ofthe oxygen seen in Jupiter’s stratospheric H2O and CO2 is SL9-related.

The goal of this paper is to try and distinguish between thevarious possible sources of external oxygen for Jupiter, usingobservations from the Infrared Space Observatory (ISO) andthe Submillimeter Wave Astronomical Satellite (SWAS), andphotochemical models. Such a task is obviously needed beforesafe interpretations can be made on topics such as the origin ofoxygen and the activity of small bodies in the outer solar system.Observations are presented in Section 2. Models are described inSection 3, with special emphasis on a simplified evolution modelof SL9-derived oxygen species. The case for a dominantly SL9origin for both H2O and CO2 is discussed in Section 4, and
implications for the magnitude and nature of other sources aredrawn.

H
114 LELLOUC
2. OBSERVATIONS

2.1. H2O Observations

Pure rotational lines of water have been observed in Jupiterby three different experiments (Figs. 1 and 2). The first detec-tion was achieved by the Short-Wavelength Spectrometer in-strument (SWS; de Graauw et al. 1996) of ISO (Kessler et al.1996) in Fabry–Perot (FP) mode (AOT SWS-07), at a meanspectral resolution of 31,000. The instrumental aperture in thismode was 17′′ × 40′′, with the long axis roughly aligned withthe planet central meridian. The initial detection was obtained onApril 22, 1997, at 39.38 µm and 43.89 µm (Feuchtgruber et al.1997, Lellouch et al. 1998). The 40.33-µm line was later de-tected on June 4, 1997. Improved observations were performed
on November 22, 1997 (revolution 738 of ISO), showing the de-tection of five lines at 39.38 µm (actually a resolved doublet of resolved by adjusting the slopes of the nearby continua to match
FIG. 1. The various H2O lines (histograms) observed by ISO/SWS (top four panels), ISO/LWS (bottom left two panels), and SWAS (Sept. 1999, bottomright panel) are compared with two background models with different H2O input fluxes (see text). Solid line: H2O flux = 1.65 × 106 cm−2 s−1, adjusted to
fit the SWAS spectrum. Dashed line: H2O flux = 7.3 × 105 cm−2 s−1, adjustelines.
ET AL.

lines at 39.375 µm and 39.379 µm), 40.33 µm, 43.89 µm, and44.19 µm. These five lines, whose preliminary analysis has beenpresented elsewhere (Lellouch et al. 1997a, 1998, Feuchtgruberet al. 1997, 1999), constitute our first dataset.

At longer wavelengths, two other rotational lines of water at66.44 and 99.49 µm were clearly detected by the ISO Long-Wavelength Spectrometer (LWS; Clegg et al. 1996). These ob-servations were made on 12 and 26 November 1997, respec-tively, following initial detections on 23 May 1997. The LWSwas employed in Fabry–Perot mode (AOT LWS03; Davis et al.1995) at a mean spectral resolution of about 9000. The LWSinstrument aperture (100′′) encompassed the entire jovian disk.The data were processed using standard algorithms described bySwinyard et al. (1996, 1998), including two corrections. First,an instrumental uncertainty in the LWS grating position was

d to fit the ISO/SWS lines. Both models severely overpredict the ISO/LWS

WATER AND CO2 IN JUPITER’S STRATOSPHERE 115

FIG. 2. The various H2O lines (histograms) observed by ISO/SWS (first four panels), and ISO/LWS (last two panels) are compared with two “Shoemaker–−8
Levy 9” models. Solid line: H2O mixing ratio = 4 × 10 , confined to p0 < 0.5 mbar levels, and spatially uniform. Dashed line:“physical” model, including the
vertical and horizontal distribution of H2O (see Figs. 6 and 8), calculated at t + 3

continuum models (such models are discussed in Section 3); andsecond, the independently measured stray light contribution toeach LWS detector was explicitly subtracted from the data. Theflux calibration, as for all LWS data, was based on a model ofUranus. It is shown below that the combination of these observa-tions, which have not been presented before, with the ISO/SWSdata, is of key importance in distinguishing the origin of jovianwater.

Finally, on September 1–6 and 19–23, 1999, theSubmillimeter–Wave Astronomical Satellite (SWAS; Melnicket al. 2000) detected emission from the 556.936-GHz(538.3-µm) H2O line from heterodyne observations with resolv-ing power ∼3 × 105 (Bergin et al. 2000). Like with ISO/LWS,the instrumental beam (3.3′ × 4.5′) is much larger than Jupiter.This observation was repeated on January 24–29, 2001, and theline was detected again with similar characteristics, although
with a somewhat lower signal-to-noise ratio. As already notedin Bergin et al. (2000) and as is frequent in heterodyne spec-
years. The models satisfactorily match all ISO/SWS and ISO/LWS lines.

troscopy of strong continuum sources, the SWAS spectra ofJupiter exhibit broad wings (±100 km s−1 from line center)which cannot be reproduced in models and are most likely instru-mental. The estimated mean continuum brightness temperatureover the bandwidth (350 MHz) is 137 ± 4.5 K for the September1999 data and 128.5 ± 7 K for January 2001.

An overview of the observational parameters for the threeH2O datasets is given in Table I. All of the ISO observationsand the SWAS 1999 spectrum are shown in Figs. 1 and 2. Theadopted units follow the practice of the individual instrumentalteams. The x-axis is expressed in wavelength (micrometers) forISO/SWS and ISO/LWS and velocity (km/s) for SWAS. The y-axis is expressed in flux units (Jy for ISO/SWS, W cm−2 µm−1

for ISO/LWS) and Rayleigh–Jeans temperature (K) for SWAS.Each of the observations possesses its own nominal calibration,and the data are presented as such. However, the calibration of
the Fabry–Perot data is too uncertain (about 25–30% precisionfor ISO/LWS and even worse for ISO/SWS) to usefully constrain

116

quality. All Ident formal e

LELLOUCH ET AL.

TABLE IParameters for the H2O and CO2 Observations

Date Instrument Mode Wavelength (µm) Resolution Aperture

1997/11/22 SWS FP 39.38, 40.33, 43.89, 44.19 31,000 17′′ × 40′′1997/11/12 LWS FP 66.44 8,000 100′′1997/11/26 LWS FP 99.49 9,500 100′′1997/05/25–26 SWS Grating 13.6–16.2 1,200 14′′ × 27′′1999/9 SWAS Heterodyne 538.3 300,000 ∼200′′ × 270′′
2001/1 SWAS Heterodyne ′′ ′′
continuum models, and in what follows, these observations areanalyzed in terms of line-to-continuum ratios.

2.2. CO2 Observations

Observations of CO2 were performed by SWS in gratingmode, at a mean resolution of 1200. The instrumental aper-ture in this mode is 14′′ × 27′′, offering a modest spatial reso-lution on the ∼40′′ jovian disk. Preliminary observations wereobtained on April 12, 1996, in AOT SWS01 mode, covering theentire 2.36–45.2 µm range. They showed the first detection ofCO2 in Jupiter through emission in its ν2 vibrational mode near14.98 µm (Lellouch et al. 1998). However, these data are af-fected by significant instrumental fringing, making their quanti-tative interpretation uncertain. The CO2 detection was confirmedon May 23, 1997, again with the SWS01 mode. Here, we willfocus on improved observations that were obtained on May 25–26, 1997. They were taken in the AOT SWS06 mode, coveringthe 13.6–16.2 µm range. This mode provides better resolutionand signal-to-noise and enables an improved removal of instru-mental fringing (Salama et al. 1997). In addition, besides thenominal pointing on Jupiter’s center, observations were obtainedwith the aperture center targeted on the north and south poles ofJupiter. In all cases, the long side of the SWS slit was aligned(to within 2◦) along the polar axis. These observations weretaken as part of a program aimed at searching for enhancementsand longitudinal variations of hydrocarbons in auroral zones(Bezard et al. 2001). Spectra were acquired at System III longi-tudes of 40, 175, and 280◦ in the north region, and 56, 188, and316◦ in the south region. These observations (Fig. 3) revealeda marked latitudinal variation of the CO2 emission, decreasingfrom the south pole position to the equator and becoming un-detectable at the north pole (Feuchtgruber et al. 1999). Thisvariation already suggests that the presence of CO2 in Jupiter’sstratosphere is related to the SL9 impacts, which occured at∼44◦ S. In contrast, the CO2 emission does not show any obviouslongitudinal variation, and we use here the longitude-averagedspectra.

The SWS06 observations in this spectral range have a nominalcalibration uncertainty of 15% (Salama et al. 1997). An addi-tional important aspect for the CO2 observations is the pointing

SO/SWS individual observations have an indepen-rror of ±1.5′′. In addition, the different bands of the

538.3 300,000 ∼200 × 270

SWS instrument have relative pointing offsets. In the directionparallel to the long side of the slit, band 3A (which includes the15-µm band) is shifted by 1.8′′ from band 1A (2.38–2.60 µm),which was used for the nominal pointing. Therefore, the CO2

observations are affected by a combination of systematic andrandom pointing errors, at a total level of 2–3′′. A “best guess”pointing was estimated from the observed continuum level, asdescribed below.

FIG. 3. The CO2 ν2 14.98 µm observed by ISO/SWS in May 1997 forthe north, center, and south regions. Histograms: data. Thin solid lines: CO2

is uniformly mixed at pressures lower than p0 = 0.5 mbar, with mixing ratiosof 13 × 10−9, 7 × 10−9, and 1.5 × 10−9 in the south, center, and north region,respectively. Thick dashed line: “physical” model, based on the CO2 horizontal
distribution of Fig. 8 and the vertical distribution (thick dashed line) of Fig. 6.Features at 14.86, 14.92, and 15.02 µm belong to the C2H2 ν5 band.

I
WATER AND CO2 IN JUP
2.3. CO Observations

In this paper, we also make use of CO millimeter-wave ob-servations to constrain our models. These observations are partof a long-term (still ongoing) monitoring of some of the chem-ical species (CO, HCN, and CS) that have been delivered toJupiter by the SL9 impacts. A detailed presentation of the ob-servations from August 1994 to July 1997 is given in Moreno(1998). The observations we use were acquired at the Pico VeletaIRAM 30-m radiotelescope in July 1997, i.e., during the sameperiod as the ISO H2O and CO2 data. Essentially they consistof heterodyne measurements of rotational CO, CS, and HCNlines. The spatial resolution is typically 10′′, and the lines areobserved at three different latitudes on the jovian disk: −44, 0,and +44◦. The observations constrain the residence level of thevarious species, as characterized by a maximum pressure levelp0, and their beam-averaged mixing ratios above this level atthe three target latitudes. The analysis is described in detail inMoreno (1998), with some highlights in Marten et al. (1998).The results most relevant to our study are (i) the p0 pressure,best determined for CS, is equal to 0.1–0.3 mbar, (ii) the COmixing ratio exhibits a factor-of-3 contrast between −44◦ and+44◦, and (iii) the CO column density at the equator is foundto be (1.4 ± 0.6) × 1016 cm−2. Although we do not refit the datathemselves here, we use these results to constrain our photo-chemical and transport models, as detailed in Section 3.3.

3. MODELS

Modelling of the ISO and SWAS data was performed us-ing a standard radiative nonscattering transfer code, includingthe collision-induced absorption of H2–He–CH4 and molecularline opacities by NH3, PH3, H2O, and CO2. Ammonia and phos-phine affect the apparent continuum level mostly in the vicinityof the water 557-GHz (J = 1–0 line of NH3 at 572.5 GHz andJ = 2–1 doublet of PH3 at 533.8 GHz) and 99.5-µm (J = 5–4multiplet of NH3) lines. Information on the temperature, am-monia, and phosphine profiles was extracted from ISO/SWSgrating spectra covering the 7–16 µm range. Specifically, thetemperature and NH3 profiles were taken from Lellouch et al.(2001) and the PH3 profile from Fouchet et al. (2000a). Fit-ting the entire ISO spectrum requires cloud opacity. FollowingLellouch et al. (2001), we introduced a cloud with a base at900 mbar, a scale height equal to 0.15 times the gas scale height,and a total cloud opacity τcl = 1.0. This opacity was appliedto model all ISO lines. To account for the likely decrease ofthe cloud opacity toward longer wavelengths, we used τcl = 0.2when modelling the SWAS data. The resulting submillimeter(300–700 GHz) continuum is in agreement with the Griffinet al. (1986) measurements. This model produces a 135-K bright-ness temperature at 557 GHz (insensitive to the precise valueof τcl), in agreement with the SWAS continuum. Similarly, themodel-calculated continuum fluxes at 66 and 99 µm are 5%
lower and 7% higher, respectively, than measured, well withinISO/LWS calibration uncertainties.

Molecular lines were assumed to be Voigt-shaped. H2-broadened H2O linewidths, depending on the J and K quantumnumbers, were taken from the results of Brown and Plymate(1996) on the ν2 band of H2O.

Synthetic spectra were calculated monochromatically, con-volved with the instrumental functions, and integrated overviewing angles at the planet. A complication with these high-resolution H2O observations is that, with an equatorial veloc-ity of 12.7 km/s, rotational smearing contributes an additionalbroadening of ∼20 km/s in the observations, equivalent to an ef-fective resolving power of 15,000. As in Lellouch et al. (2001),this observational effect was included in our model spectra bypartitioning the jovian disk into a square grid of regular spac-ing. Synthetic spectra were calculated at each point of the grid,shifted in frequency to account for the local projected diurnalvelocity, and finally summed with appropriate weights. Becauseof the uncertainty in absolute calibration, as discussed above,models in the vicinity of each spectral line were multiplied by aconstant factor to match the observed continuum, i.e., the H2Oemissions were analyzed in terms of line-to-continuum ratios.In the case of the SWAS data, because of the apparent broadlines that are most likely instrumental, there is considerable un-certainty in the actual continuum level and therefore in the line-to-continuum ratio. We force models to match observations at±20 km/s from line center and evaluate the fits over a restricted±20 km/s range, although this choice is obviously arbitrary tosome extent.

In the case of CO2, the continuum levels in the south andnorth points spectra were used to attempt the retrieval of pointinginformation. For the “nominal pointing” (i.e., with the center ofthe aperture exactly located at Jupiter’s south and north pole,respectively), the model overpredicts the continuum flux in thesouth by 13% and underpredicts it in the north by 18%. Perfectmatching of the continua can be achieved by “displacing” theaperture by 1.8′′ parallel to its long axis towards the south. Thisis consistent with the order of magnitude of the pointing errorsdiscussed above. Although the validity of this approach is limitedby the 15% absolute calibration uncertainty (in particular, at diskcenter, the model overpredicts the flux by 13% regardless ofpointing errors), we adopt this 1.8′′ shift to the south in definingour best-guess pointing. With this shift, the latitudes sampled bythe center, south, and north apertures extend from 48◦S to 34◦N,25◦S to 90◦S, and 14◦N to 90◦N, respectively.

The above models were run in combination with a suite ofmodels for the vertical and/or horizontal distribution of H2Oand CO2.

3.1. Uniform Models

The first models we considered have no physical basis, butthey are simple and provide a first insight into the order of mag-nitudes involved. We here assumed that H2O and CO2 haveuniform mixing ratios in the stratosphere above some level. For
H2O, it is natural to take this level as the condensation level,and the corresponding distribution is assumed to follow the

H
118 LELLOUC
saturation curve below this level. In doing so, we neglect anycontribution due to injection of water cirrus particles into thestratosphere and subsequent evaporation. Although there is ev-idence for moist convection in Jupiter’s atmosphere (Gieraschet al. 2000), we consider it unlikely that it could transport ma-terial from the ∼300-mbar cloud tops all the way up to themillibar region. In this framework, the ISO/SWS data indicatea best fit mixing ratio of (1.5 ± 0.4) × 10−9 (Lellouch et al.1998, Feuchtgruber et al. 1999), a column density of (2.9 ±0.7) × 1015 cm−2, and a condensation level near 15 mbar. How-ever, the ISO/LWS data would suggest an inconsistent (0.5 ±0.15) × 10−9 mixing ratio. In addition, the best reproductionof the SWAS data is achieved by a mixing ratio of 2 × 10−9

above the saturation level, but the calculated line profile is thenunacceptably broad (Bergin et al. 2000). The SWAS data, inthemselves, indicate that the bulk of the water resides at pres-sures less than 5 mbar. From a theoretical point of view, H2O,which has a high-altitude source (both in the case of a steadyexternal source and in the case of SL9-derived material) anda low-altitude condensation sink, is expected to show a de-creasing mixing ratio with increasing pressure. There is there-fore every reason to discard the vertically uniform distributionfor H2O.

CO2 does not condense at Jupiter’s tropopause. Thus a pos-sible cutoff in the CO2 distribution must be specified arbitrar-ily. Assuming that CO2 is present only at pressures less than10 mbar—a zeroth-order approximation of the distribution ex-pected from a steady-state photochemical model (see below)—the CO2 emission at disk center indicates an average 4 × 10−10

mixing ratio, and a 4.1 × 1014 cm−2 column. For a cutoff at30 mbar (respectively 3 mbar), the mixing ratio becomes 2.1 ×10−10 (respectively 1.1 × 10−9), and the column density is 6.2 ×1014 cm−2 (respectively 3.2 × 1014 cm−2). For a given cutoffpressure, the average mixing ratio and column density in thesouth region are typically 1.8 times larger than at disk center; inthe north region, they are at least four times smaller than at diskcenter.

A possible source for oxygen compounds in Jupiter’s observ-able atmosphere is convective transport from deep hot layers, ashas been proposed for CO (e.g., Fegley and Lodders 1994). In thedeep atmosphere, conversion of oxidized carbon to reduced car-bon proceeds mainly through H + H2CO + M → CH3O + M(Yung et al. 1988), and the CO abundance at observable levels(1 ± 0.2 ppb, Bezard et al. 2002) is determined by quenching ofthis reaction. Using the formalism developed by Smith (1998)to determine this quench level, Bezard et al. (2002) showed thatvertical transport can provide the observed CO abundance for aneddy mixing characteristic of free convection. The observableCO mole fraction depends linearly on the O/H ratio and, in amore complex fashion, on the convective mixing coefficient Kv .A solar value of the O/H ratio would require Kv ∼ 5 × 108 cm2

s−1, while a 3 times solar ratio implies Kv ∼ 108 cm2 s−1. AnO/H ratio as high as 10 � would require an eddy mixing coeffi-
cient of approximately 107 cm2 s−1, lower than expected fromfree convection.
ET AL.

Exchange between various forms of oxidized carbon (notablyCO, CO2, and H2CO) is slightly more rapid, and quenching be-tween them takes place at lower temperature levels. Conversionof CO2 into CO likely occurs through CO2 + H → CO + OH.We calculated the amount of CO2 that can be upwelled to theupper atmosphere by quenching of this reaction. Chemical timeconstants (tchem) were calculated using Baulch et al.’s (1992) ki-netic data. The mixing time constant, given by tmix ≈ L2

eff/Kv ,where Leff is a mixing length scale, was estimated for the rangeof Kv needed to reproduce the observed 1 ppb of CO (Bezardet al. 2002). Adopting Smith’s (1998) formulation, we foundthat for the above reaction, Leff ≈ 0.13 H , where H is the atmo-spheric scale height. Assuming an O/H ratio in the range 1–10 �,we accordingly varied Kv between 107 and 5 × 108 cm2 s−1 andderived quench levels for the above reaction between 910 and990 K. The corresponding CO2/CO ratio is proportional to theO/H elemental ratio and varies between 0.003 (for O/H = 1 �)and 0.04 (for O/H = 10 �). With the CO mole fraction frozen to1 ppb, the CO2 abundance upwelled to observable levels is thusin the range 3 × 10−12–4 × 10−11. We calculated that a uniformCO2 profile with a 4 × 10−11 mixing ratio produced a 15-µmemission close to the upper limit indicated by the northern spec-trum. Thus, the nondetection of CO2 in the northern region isconsistent with our expectations for a small internal source ofCO2, and we did not consider the internal source any further.

3.2. “Background” Models

In a second class of models, we assumed that H2O is spatiallyinvariable and that it results from a steady micrometeoritic inter-planetary source. We generated such H2O profiles by adaptingto Jupiter the 1-D photochemical model developed for Saturn byMoses et al. (2000a,b). This model includes a comprehensivedescription of hydrocarbon and oxygen chemistry (102 species,559 reactions). Turbulent mixing is parameterized by an ad-justable vertical eddy diffusion coefficient profile (Kv), chosento match the distribution of CH4 and other major hydrocarbons asobserved by ISO (Fouchet et al. 2000b). A physical descriptionof condensation processes and of the ablation and depositionof external material is also included. The model is generallysuccessful at reproducing the abundances and vertical distribu-tion of the major C2 hydrocarbons (Moses et al. 2001). Regard-ing oxygen compounds, we followed the approach of Moseset al. (2000a) and assumed that the incoming material is a com-bination of H2O, CO, CO2, and CH3OH, similar to cometarymaterial.

An oxygen influx similar in composition and magnitude to thenominal Saturn model, i.e., fluxes of 1.5 × 106 H2O cm−2 s−1,8.0 × 105 CO cm−2 s−1, 7.5 × 104 CO2 cm−2 s−1, and 7.0 × 104

CH3OH cm−2 s−1, gives a satisfactory match of the SWAS H2O1999 observation (optimum fitting is achieved with a H2O flux of1.65 × 106 cm−2 s−1). However, this model considerably over-estimates all of the ISO lines. A good overall match of the four
ISO/SWS H2O lines is achieved by reducing the H2O flux to7.3 × 105 cm−2 s−1, keeping the other fluxes unchanged (Fig. 1).

I

v


FIG. 4. The H2O (solid line) and CO2 (short-dashed line) vertical distribu-tion in the nominal “background model.” Input fluxes are 7.3 × 105 H2O cm−2

s−1, 8.0 × 105 CO cm−2 s−1, 7.5 × 104 CO2 cm−2 s−1, and 7.0 × 104 CH3OHcm−2 s−1. The vertical eddy diffusion (Kv) coefficient used in this model is alsoshown (long-dashed line). The dashed–dotted line represents the H2O “hybrid”model combining a background source and a SL9 source.

The associated H2O and CO2 mixing profiles are shown in Fig. 4,in which the Kv profile is also presented. The H2O column den-sity is 1.65 × 1015 cm−2. However, this model overestimates theISO/LWS data by a factor of 3, and underestimates the SWASline by a factor of 2 (Fig. 1). Thus, the different water lines givediscrepant values for the H2O column, by as much as a totalfactor of 6–7.

The CO2 profile of Fig. 4 has a column density of 9 ×1013 cm−2. This is about 4.5 times less than the average (diskcenter) column density estimated above. The associated emis-sion underestimates the observed CO2 band contrast by a factorof about 3 at disk center and about 5 in the south and is reason-ably consistent with the upper limit in the north.

As we discuss below, the recent CO observations of Bezardet al. (2002) indicate a much higher CO flux than postulatedhere. However, it appears that the H2O and CO2 abundances arerelatively insensitive to the CO flux. Boosting the latter from8 × 105 cm−2 s−1 to 4 × 106 cm−2 s−1 leads to an increaseof the stratospheric columns of H2O and CO2 by only ∼20%.In contrast, increasing the CO2 flux readily increases the CO2

column density. A model with input H2O and CO2 fluxes of3.4 × 105 cm−2 s−1 and 4.0 × 105 cm−2 s−1, respectively, doesfit the average CO2 emission but, being by essence spatially uni-form, it does not explain the CO2 hemispheric asymetry, whichin itself implies that CO2 is SL9-related.

3.3. “Shoemaker–Levy 9” Model

3.3.1. Elementary Analysis

The 66.44 and 99.49 µm H2O lines observed by ISO/LWSare intrinsically stronger than those seen by ISO/SWS, and
they are observed at a spectral resolution ∼3 times lower. As

they are already saturated for the background model (e.g., the66.44-µm line core opacity is equal to 10.7), a way to reducethem preferentially is to decrease their equivalent width by re-stricting the H2O column to low pressure levels. Doing so, theLorentz linewings of the 66.44 and 99.49 µm lines are sup-pressed, while the Doppler core is virtually unchanged, and theresulting contrast after instrumental convolution is decreased.This does not happen for the lines observed by ISO/SWS whichremain largely unsaturated. In general, the relative 66.44-µm/

39.375-µm/39.379-µm line contrast is a good indicator of thevalidity of a model, as a model matching these three lines simul-taneously will allow a good fit of all seven ISO H2O lines. Wefind that a profile in which water is confined to pressures lessthan p0 = 0.5 mbar (with a uniform mixing ratio of 4 × 10−8

there) matches this requirement (Figs. 2 and 5). For this profile,which corresponds to a column density of 2.0 × 1015 cm−2, the

FIG. 5. The H2O 39.38-µm doublet and 66.44-µm line for several mod-els. Thick lines represent simple SL9 models in which the maximum pressurelevel where H2O is located, p0, is varied. Thick solid line: p0 = 0.5 mbar,mixing ratio above this level q = 4 × 10−8. Thick dashed line: p0 = 0.2 mbar,q = 1 × 10−7. Thick dashed–dotted line: p0 = 0.8 mbar, q = 2.2 × 10−8. Thinlines represent model calculations based on vertical distributions calculated byour “vertical transport model” (see text) at t + 3 years. Thin solid line (indis-tinguishable from thick solid line): nominal eddy Kv profile (the correspondingdistribution is the thick short-dashed line of Fig. 6 rescaled by 1.8 × 10−7). Thindashed line: same initial mixing ratio, but the eddy K profile is divided by 5.
Thin dashed–dotted line: same initial mixing ratio, but the eddy Kv profile ismultiplied by 2.

120 LELLOUC

66.44-µm, 39.375-µm, and 39.379-µm lines have core opticaldepth of 23, 2.75, and 0.92, respectively, to be compared with10.7, 1.5, and 0.5 for the background model of Fig. 4. Fig. 5demonstrates the sensitivity of the 39.38-µm doublet-66.44-µmline comparison in defining the typical pressure level at whichwater is located. Given the noise level and the imperfect matchof the weak 39.379-µm line, we estimate that p0 is defined to liebetween 0.3 and 0.7 mbar and that the water column density is(2.0 ± 0.5) × 1015 cm−2. Examination of weighting functionsindicates that at infinite spectral resolution, peak sensitivity oc-curs near 0.35 mbar at 39.379 µm, 0.1 mbar at 39.375 µm,and 0.01 mbar at 66.44 µm. This model still underestimates theSWAS 1999 observation. We come back to this issue later.

A H2O distribution confined to submillibar levels is stronglyreminiscent of post SL9-impacts distributions. From millimeter-wave observations, Lellouch et al. (1997b) found that 1 dayafter the G impact, the CO generated by this impact was re-stricted to pressures less than p0 = 0.1 mbar, within a factor of2. From similar observations in July 1997, Moreno (1998) in-ferred a maximum residence pressure for CO, p0 = 0.2 mbar,also within a factor of 2. It seems therefore plausible to observep0 = 0.3–0.7 mbar for water 3.3 years after impacts.

The total mass of H2O in Jupiter’s stratosphere implied bythis (spatially uniform) model is (3.8 ± 1) × 1013 g. This is about0.05 times the total mass of CO delivered by the SL9 impacts, asestimated by Lellouch et al. (1997b), i.e., assuming that a largefragment holds 1/5 of the total comet mass. This is also typically5 times more than the amount of post-SL9 H2O reported by mostobservers, but in agreement with the mass found by Bjorakeret al. (1996b) (see further discussion below). Thus, at this point,the SL9 origin for H2O does not appear grossly inconsistent withearlier findings.

The CO2 observations are nearly concommittant with the H2Oobservations (34 and 40 months, respectively, after the impacts).For a SL9 source, it is a reasonable assumption to adopt a simi-lar vertical profile for CO2 and H2O. Assuming that CO2 is uni-formly mixed at pressures less than p0 = 0.5 mbar, the ISO/SWSobservations indicate mean mixing ratios of (13 ± 3) × 10−9,(7 ± 1.5) × 10−9, and <1.5 × 10−9, for the south, center, andnorth regions, respectively. The associated total column densitiesare (6.3 ± 1.5) × 1014, (3.4 ± 0.7) × 1014, and <7 × 1013 cm−2.These columns are essentially independent of the precise loca-tion of CO2, as the ν2 band is optically thin (maximum lineoptical depth = 0.15) and does not provide information on thevertical profile of CO2. Thus, variations by factors of ∼2 fromsouth to center and ∼5 from center to north are seen. These arecomparable to, though somewhat larger than, the spatial con-trasts seen in post-SL9 CO 3 years after the impacts (Moreno1998).

3.3.2. A Physical Model

The above analysis suggests that a “SL9-type” vertical profile
of H2O and a “SL9-type” horizontal distribution of CO2 providethe best overall fit of the entire ISO data set. This result, and in
H ET AL.

particular the persisting presence of H2O 3 years after the im-pacts, is at first glance surprising, and the question of whether theinvolved amounts are plausible, and, if so, what the implicationson vertical and horizontal transport are, must be investigated.

Chemistry. Material deposited and/or created by the comet-ary impacts undergoes both chemical evolution and vertical andhorizontal (longitudinal and meridional) transport. We assumethat the only oxygen compounds deposited/created by the SL9impacts are those detected just after the impacts, CO and H2O.We neglect OCS, which was reported to be present at the 1%of CO level 4 days after the K impact, but had disappeared1 month later (see Lellouch 1996). We also note that CO2 wasnot detected, with an upper limit of (2–8)% of H2O for a largeimpact (Encrenaz et al. 1997).

The main chemical reactions controlling the long-term evolu-tion of water and CO2 are the following (see Moses et al. 1995,2000a, Moses 1996 for the complete oxygen chemistry):

H2O → OH + H (1)

OH + H2 → H2O + H (2)

OH + CO → CO2 + H (3)

CO2 → CO + O (4)

CO2 → CO + O(1D) (5)

O(1D) + H2 → OH + H. (6)

The photolytic rates at Jupiter’s distance for H2O and CO2 (re-actions 1, 4, and 5) are typically J1 = 8.25 × 10−8 s−1 and J4 +J5 = 2.05 × 10−10 + 8.0 × 10−10 = 1.0 × 10−9 s−1 at middlestratospheric pressures (see Fig. 7 below). At 170 K, reactionrates for reactions 2, 3, and 6 are k2 = 3.3 × 10−17 cm3 s−1,k3 = 1.5 × 10−13 cm3 s−1, and k6 = 1.1 × 10−10 cm3 s−1. Otherreactions, such as those involving OH with the products ofmethane photolysis, appear to be of secondary importance. Asour goal is primarily to study the stability of H2O and formationof CO2 over the first few years, we do not model the subsequentfate of O radicals.

Considering only reactions (1) to (3), the evolution of H2Othus mostly depends on the competition between reaction (3),which converts it to CO2, and reaction (2), which recycles itfrom OH. A higher CO mixing ratio increases the conversionof H2O to CO2. (This non-steady-state situation is thus in sharpcontrast with the background model described in the previoussection, in which an increased CO flux leads to an increase ofthe steady-state abundances of both H2O and CO2). With theshort-lived OH being in photochemical equilibrium, the localproduction rate of CO2, equal to the loss rate of CO, is alsoequal to the net loss rate of H2O,

P(CO2) = L(CO) = L(H2O) − P(H2O)/(k2

)
= J1[H2O] 1 +
k3qco,

I
where [H2O] is the concentration of water and qco is the COmixing ratio (=[CO]/[H2]).

The effective H2O chemical lifetime is then given by τH2O =J−1

1 (1 + k2k3qco

). With J1 = 8.25 × 10−8 s−1, and high values ofqco characteristic of conditions just after impact (qco ∼ 1.5 ×10−4), conversion of OH to CO2 occurs quickly and τH2O ∼350 days. This is confirmed by the detailed calculations of Moseset al. (1995) and Moses (1996), showing a significant decreaseof water and building of CO2 after 1 year. However, these cal-culations do not account for the horizontal spreading of CO,which by causing a decrease of the local qco will increase theH2O lifetime. For example, for CO mixing ratios of 3 × 10−6, asobserved in May 1995 at the impact latitude (Moreno 1998), theH2O chemical lifetime is already equal to 28 years. Once CO2

has built up significantly, the lifetime of H2O is further increasedby the partial conversion of CO2 to H2O (reactions 5, 6, and 2).

Vertical diffusion occurs on a timescale τv = H 2/Kv , whereH and Kv are the atmospheric scale height and the vertical eddydiffusion coefficient in the region of interest (∼0.2 mbar). UsingH = 40 km and Kv = 1 × 105 cm2 s−1 (Gladstone et al. 1996,Moses et al. 2001) gives τv = 5 years. Similarly, the timescalefor horizontal diffusion can be written as τh = L2/Kh , where Lis a characteristic horizontal length and Kh an effective horizon-tal eddy diffusion coefficient. We take L to be the equator-to-poledistance (110,000 km). We use Kh = 2.5 × 1011 cm−2 s−1, as in-ferred by Moreno (1998) from the evolution of the CO latitudinaldistribution over 1994–1997 (we will derive hereafter a similarvalue). This value is over an order of magnitude higher than theKyy suggested by Friedson et al. (1999) from the meridionalspreading of the dust; however, the value derived from the dis-tribution of a gas is more appropriate here, as the dust probablysettles to deeper levels due to sedimentation. This gives τh = 16years. This is actually valid for the latitudinal transport. Longitu-dinal mixing certainly occurs more quickly, due to the existenceof zonal wind shears (see below). The similarity of all transporttimescales with the H2O chemical lifetime illustrates the com-plexity of the problem, which for an accurate resolution, wouldrequire a full 3-D and time-dependent photochemical model.

Our goals are to investigate if, 3 years after impacts, (i) thevertical profile we determine for water is consistent with a re-alistic Kv profile and (ii) the total amounts of H2O and CO2

and the horizontal distribution of CO2 are consistent with plau-sible values for Kh and the initial amount of water. We thereforedecouple the vertical and horizontal evolutions. We first solvethe vertical diffusion equation for H2O, assuming no horizontaland chemical evolution. We then, independently, solve the hori-zontal diffusion equation for all oxygen species, including theirchemical sources and sinks but neglecting any vertical diffusion.

Vertical profile of H2O. The usual 1-D vertical diffusionequation

dq 1 d(

dq)

dt=

n dzKvn

dz


FIG. 6. Evolution of the vertical distribution of a chemically inert gasrelative to an initial distribution restricted to pressure levels lower than p0 =0.1 mbar. The vertical eddy diffusion coefficient is taken from Fig. 4. Solidline: Initial distribution. Short-dashed line: Distribution at t + 3 years. Long-dashed line: Distribution at t + 5 years. The short-dashed profile, multiplied by1.8 × 10−7, allows a good fit of all H2O ISO lines. The thick lines are for thecase of initial uniform mixing at p < p0 and the thin lines are for the case of amixing ratio initially varying as (p/p0)−0.5.

is solved with a Crank–Nicholson semiexplicit scheme. For ini-tial conditions, water is assumed to be present at p < 0.1 mbaronly, as observed for CO (Lellouch et al. 1997b), and with auniform mixing ratio there. We nominally use the eddy Kv pro-file of Fig. 4, and we test scaling factors on this profile. Theinitial H2O mixing ratio at p < 0.1 mbar, qi , is a free parameteradjusted to give optimum fits to the data.

After 3 years, the eddy Kv profile of Fig. 4 leads to a wa-ter profile (shown as the thick short-dashed line in Fig. 6) infull agreement with the 66.44 µm/39.375 µm ratio (Fig. 5). Aninitial qi mixing ratio of 1.8 × 10−7 allows a simultaneous fitof all H2O lines observed by ISO. The associated profiles (ini-tial, and at t + 3 years and t + 5 years) are shown in Fig. 6.Similar to the discussion on the allowed range of p0, we con-sider that the range of scaling factors on the Kv profile allowedby the ISO data is 0.6 to 4. This indicates Kv = 4 × 104–3 ×105 cm−2 s−1 at ∼0.2 mbar. This reasonable result, in itself,is interesting, as independent constraints on eddy diffusion inthe midstratosphere are lacking (Fouchet et al. 2000b, Moseset al. 2001). Although there is some uncertainty on the initialdeposition/formation level of H2O and its vertical profile justafter the impacts, this affects only marginally the range of ac-ceptable Kv values. For example, the mixing profile of waterjust after the impacts may actually be increasing with altitude atp < 0.1 mbar (Lellouch et al. 1997b). As proved in Fig. 6, thisaffects very little the resulting profiles at t + 3 years and t + 5years.

The CO2 data can also be reproduced with such a verticalprofile. The ratio of the aperture-averaged CO abundance to
2
the planet-averaged H2O abundance is 0.35 in the south, 0.18

122 LELLOUC

in the center, and <0.03 in the north. Column densities are verysimilar to those found for the case of uniform mixing at p <

0.5 mbar.

Horizontal model and CO2 production. We here take thecomplementary point of view. We neglect vertical diffusion,i.e., hold the vertical profile of all compounds fixed at its ini-tial postimpact value (uniform mixing at p < 0.1 mbar), andwe study the horizontal and chemical evolution of the oxygencompounds.

Postimpact observations of aerosol debris (e.g., West 1996,Simon and Beebe 1996, Sanchez–Lavega et al. 1998, Friedsonet al. 1999) indicate different timescales and different mecha-nisms for the longitudinal and meridional spreading of the SL9dust. The formation of a “SL9 belt” at the impact latitude is ob-served as early as September 1994, while it took about 3 yearsfor the debris to reach the −20◦ latitude. The longitudinal dis-persion results from advection by zonal winds in the presenceof meridional or vertical shears, with an effective velocity ofup to 10–20 m/s. In contrast, meridional motions as slow as6 cm/s (poleward) and 42 cm/s (equatorward) were measuredby Sanchez–Lavega et al. (1998). Friedson et al. (1999) find thatadvection by the meridional circulation calculated by West et al.(1992) does not explain the equatorward spreading, which theyinterpret as transport from large-scale quasi-geostrophic eddies.This suggests different spreading mechanisms in the longitu-dinal and meridional directions, at least at the altitude coveredby the aerosol debris (1 to 400 mbar). Whether this is also truefor gases in the submillibar region is unknown. In particular, itmay be expected that the organized zonal circulation decreaseswith altitude, while the effect of eddy mixing increases. Obser-vational data are not conclusive in this respect. Moreno (1998)reports that 10 months after the impacts, the CS line intensityshowed longitudinal variations at 44◦S. However, these vari-ations do not appear correlated with the position of the largeimpacts, so their interpretation remains ambiguous. We here as-sume that meridional transport occurs as an effect of transienteddies that can be parameterized with the use of a single hor-izontal eddy diffusion coefficient Kh . We further assume thatzonal mixing is initially very fast, so that our initial conditioncorresponds to uniform mixing ratios in a latitude band. Mixingratios at all subsequent times are then independent of longitude,and the evolution equation can be written as

∂ni

∂t= 1

r2 cos θKh

∂(cos θ ∂ni

∂θ

)∂θ

+ (Pi − Li )(θ ),

where ni is the number density of species i , θ is latitude, andP and L are chemical production and loss terms, respectively,as described by the six reactions above. A similar but slightlydifferent approach was used by Moreno (1998) who calculatedthe evolution of CO (neglecting any chemical terms) by solvingthe diffusion equation in full spherical geometry ∂n

∂t = Kh�n,
i.e., assuming that zonal mixing and meridional mixing proceedat the same speed.
H ET AL.

FIG. 7. Photolysis rate profiles as a function of latitude. Solid lines: 0◦.Short-dashed lines: 30◦. Long-dashed lines: 60◦. Dashed–dotted lines: 88◦. Notethat the increase in the photolysis rates at lower stratospheric pressures is dueto multiple Rayleigh scattering. At all latitudes, the photolysis rates eventuallydecrease rapidly with decreasing depth as photons are lost due to molecularabsorption and scattering.

The above equation was solved using a Crank–Nicholsonscheme, recalculating the local production and loss terms ateach time step. The photolysis rates were calculated using di-urnally averaged transmission functions. As it turned out thatthe J values in the relevant pressure region (0.1 mbar) are es-sentially latitude-independent (Fig. 7), we simply used con-stant values of J1 = 8.25 × 10−8 s−1, J4 = 2.05 × 10−10 s−1

and J5 = 8.0 × 10−10 s−1. Note that for the range of H2O thatwe find (column <3 × 1016 cm−2 at any time step), the waterphotolysis, which occurs primarily longward of 170 nm, remainsoptically thin.

For initial conditions, we assumed a uniform CO mixing ratioof 2.5 × 10−5 in a band extending from 37◦S to 52◦S latitude.This corresponds to the dilution of an initial 1.5 × 10−4 COmixing ratio in four large sites (8500 km radius, such as G or K).Based on the classification of Hammel (1996), and the results ofBezard et al. (1997) for HCN, Lellouch et al. (1997b) assumedthat the total CO mass corresponds to that delivered in five large(i.e., class 1) sites, but using only four sites gives the best match
to the CO abundances observed in 1997. Other free parametersare the initial H2O/CO ratio in the band, and the value of Kh .


FIG. 8. The horizontal distribution of CO, H2O, and CO2 3 years afterimpacts calculated from our “horizontal-chemical” transport (i.e., neglectingvertical transport). The horizontal eddy diffusion coefficient is 2 × 1011 cm2

s−1. Initial conditions are a CO mixing ratio of 2.5 × 10−5 in a 15◦ wide latitudeband around 44◦S and a H2O/CO volume mixing ratio of 0.11. The calculatedmixing ratios after 3 years agree very well with the CO observations of Moreno(1998) and the H2O and CO2 observations presented here (see also Figs. 2and 3).

Best fits are obtained for a volume H2O/CO ratio of 0.11 andKh = 2 × 1011 cm2 s−1. Figure 8 shows the latitudinal distribu-tion of CO, H2O, and CO2 after 3 years. For comparison with thedata, the abundances are expressed in terms of column densities.The good agreement with the measurements can be further seenin Figs. 2 and 3, where the actual data are compared with modelsbased on the horizontal distributions of Fig. 8 and the verticaldistribution of Fig. 6 (thick short-dashed). The combination ofthese latitudinal and vertical distributions is shown in Fig. 9.

FIG. 9. The H2O distribution 3 years after impacts, computed from ourbest fit SL9 model (see Fig. 8 caption for parameters). This figure is obtained
by combining the latitudinal distribution of H2O from Fig. 8 with the verticalprofile at t + 3 years from Fig. 6 (thick short-dashed line).
ITER’S STRATOSPHERE 123

This figure hence represents our estimate of the 2-D (latitude,pressure) distribution of H2O 3 years after the impacts. It is onlyapproximate, due to our approach of decoupling the vertical andhorizontal evolutions.

We now need to come back to the SWAS data. In Fig. 10,we separately show the Sept. 1999 and Jan. 2001 data, com-pared to a suite of models. A model based on the calculatedvertical distribution of water at t + 3 years (Fig. 6), rescaled bya (horizontally constant) 1.8 × 10−7 factor, as was found satis-factory for the ISO data, matches the width of the line observedby SWAS but underestimates its contrast, both for the 1999 and2001 data. However, these data were taken after the ISO ob-servations, and the progressive downward diffusion of watercontributes to desaturating the SWAS line, thereby increasingits contrast. This effect is not quite sufficient, as the same (stillhorizontally constant) model calculated at t + 5 years and t +6.5 years continues to underestimate the SWAS data. Using the

FIG. 10. The SWAS 1999 and 2001 observations, compared with various“SL9 models.” Solid lines: H2O profile at t + 3 years (thick short-dashed line ofFig. 6), with initial 1.8 × 10−7 mixing ratio. A very similar fit can be obtainedwith a 4 × 10−8 mixing ratio restricted to p < 0.5 mbar. Long-dashed lines:Same initial condition, but the H2O profile is now calculated at the relevanttime (t + 5 years and t + 6.5 years, respectively). Short-dashed line: profilescalculated from the physical model, i.e., with initial H2O/CO = 0.11 and the
vertical and horizontal distributions at the appropriate time. None of the modelsadequately matches the SWAS data.

124 LELLOUC

horizontal distribution of H2O, as calculated for the relevantepochs, does not improve the situation much. We are thereforeunable to completely reconcile the SWAS and the ISO data. Con-sideration of uncertainties in the SWAS baseline could help bringthe ISO and SWAS data into agreement; nonetheless, furtherheterodyne observations of H2O in the 556.935-GHz (or other)line by ODIN (currently in orbit) or Herschel (to be launchedin 2007) are needed. In addition, as the H2O 557-GHz line islocated in the wings of the strong NH3 572-GHz line, its contrastis sensitive to the details of the continuum model. For example,increasing the NH3 abundance by a factor of 2 at all levels leadsto a decrease of the 557-GHz continuum brightness temperatureby 2.5 K (Fig. 11), and the water line contrast increases by 10%,bringing our preferred model almost into agreement with thedata. In this case, however, the overall continuum model tendsto underestimate the photometric measurements of Griffin et al.(1986) (see Fig. 11). When fully calibrated, the Cassini/CIRSJupiter spectra, obtained in Dec. 2000 and covering Jupiter’sfar-infrared spectrum shortward of 1 mm, will be very useful inthis respect.

Our approach, which for the sake of simplicity separates thevertical, longitudinal, and latitudinal transport, gives only an ap-proximate description of the evolution of the SL9-derived oxy-gen material. Therefore, the values of the free parameters (initialCO and H2O mixing ratios, latitudinal eddy diffusion coeffi-cient) should be taken with caution. In particular, we identifytwo limitations of our model, which fortunately act in oppositedirections as regards the efficiency of conversion of water toCO2. First, since we treat the longitudinal transport as instanta-neous, our initial conditions imply a dilution of the site mixingratios in a latitude band. This causes an underestimate of the CO2

build-up during the phase of longitudinal diffusion. Second, thevertical diffusion of CO and H2O that accompanies the latitu-

FIG. 11. Our models of Jupiter’s continuum at 250–750 GHz. Solid line:nominal model. Dashed line: model in which the NH3 abundance is multiplied
by 2 at all levels. Open squares are the broadband measurements of Griffin et al.(1986). Filled triangles are the SWAS continuum measurements.
H ET AL.

FIG. 12. Long-term evolution of the volume partitioning of SL9-derivedoxygen. Note that the O build-up after ∼30 years is artificial because reactionsinvolving atomic oxygen are not tracked in the model.

dinal spreading causes the vertically integrated production rateof CO2 to slowly decrease with time, an effect that our modelcannot handle. To what extent these two inaccuracies compen-sate for one another can be investigated only with the help ofa three-dimensional, time-dependent, chemical model. Yet, ourcalculations illustrate that the SL9 source for water is a realisticcase, explaining most features of our observations.

In the framework of this conclusion, and using our best fitparameters, Fig. 12 shows the temporal evolution of the parti-tioning of oxygen between the four species considered in ourmodel over 270 years. CO is the dominant species at all times.Water remains the second dominant species for about 150 years,after which its photolytical destruction becomes important. CO2,which has slowly built up from the H2O photolysis, reaches itsmaximum after about 50 years, and then starts itself to be de-pleted by photolysis, leading to a build-up of O. The increaseof O in our model is artificial because we do not track any reac-tions involving O. A variety of reactions with methane photolysisproducts like CH3, C2H2, C2H4 . . . , actually occur, leading tospecies such as H2CO, HCO, HCCO, CH3CO . . . , whose ulti-mate fate is dominantly to recycle CO. O can also react with H2

to recycle H2O, especially at very high and very low altitudes.At any rate, O does not build up importantly before ∼30 years,so that the oxygen partition until then should be accurate. Theimportance of horizontal transport in increasing the effectiveH2O lifetime can be seen by comparing our Fig. 12 with Fig. 10of Moses (1996).

While a SL9 origin for CO2 is obvious from the latitudinalvariations of the data themselves, we at this point have twoarguments that the observed H2O also primarily results fromSL9. The first one is the inferred vertical distribution of H2Ofrom the combined ISO/SWS and ISO/LWS data. The secondone is the fact that a SL9 origin for water explains at the same
time the absolute abundance of H2O and CO2 and the horizontal


FIG. 13. The CO2 latitudinal distribution 3 years after impacts calculatedfrom our horizontal-chemical transport under other assumptions. Thin solid line:water is assumed to derive from a steady background source, with the verticalprofile of Fig. 4 (solid line). In this case, a spatially and temporally constant1 × 10−8 H2O mixing ratio is used in the calculations. The resulting CO2 mixingratios are ∼50 times too small. Thick solid line: same, but a 4 × 10−7 H2Omixing ratio is used. The average CO2 amount now matches the observations, butthe distribution does not show enough latitudinal variation. Dashed line: “hybridmodel.” The initial H2O/CO volume mixing ratio is taken as 0.066, and no initialCO2 is introduced. Dashed–dotted line: same, but an initial CO2/CO = 0.005is assumed.

distribution of CO2. The strength of this argument must be fur-ther assessed by examining the situation in which CO2 is builtfrom SL9-derived CO and from background water. For this, wereran our horizontal model by fixing the H2O profile to the time-independent background profile of Fig. 4, which provided thebest compromise to the ISO/SWS data. The CO initial mixingratio at p < 0.1 mbar and the horizontal eddy Kh were kept un-changed to match the observations of Moreno (1998). In thiscase, the H2O mixing ratio at 0.1 mbar is only ∼10−8, and thewater column at p < 0.1 mbar is only 1.1 × 1014 cm−2, 7%of the total stratospheric column. The resulting CO2 horizontaldistribution is shown in Fig. 13 (thin solid line). The calcu-lated average abundance is ∼50 times smaller than observed.Matching the observed CO2 with such a model would requirea spatially uniform H2O mixing ratio at p < 0.1 mbar of about4 × 10−7, much larger than indicated by the background mod-els. In addition, the resulting CO2 distribution does not showenough latitudinal variation compared to the data (Fig. 13, thicksolid line).

4. DISCUSSION

4.1. Initial Abundance of SL9-Derived H2O

We find that we can explain the entire ISO data set if theH2O/CO ratio in the material created by the SL9 impacts was∼0.11 in volume, i.e., 0.07 in mass. This would indicate a H O
2
mass of about 1.0 × 1013 g for the large impacts. This is several


times larger than inferred from most postSL9 H2O observations.Indeed, for large impactors (G, K; class 1 of Hammel 1996), ob-servations at 2.7 and 6–8 µm of the plume/splashback phaseindicated water masses of only (0.6–2.8) × 1012 g, i.e., typi-cally 1% of CO (Bjoraker et al. 1996a, Carlson et al. 1995a,b,Sprague et al. 1996, Encrenaz et al. 1997). Water masses typi-cally 10 times lower were found for midsize (class 2 of Hammel1996) impactors R and W (Encrenaz et al. 1997, Sprague et al.1996). However, except for the Carlson et al. (1995a,b) andEncrenaz et al. (1997) NIMS–Galileo low-resolution observa-tions of the ν3 band of H2O, these determinations were basedon high-energy water lines, sensitive only to the hottest watercomponents, so they probably represent lower limits. This isconfirmed by the fact that much larger masses were measuredfrom the analysis of 22.6 and 23.9-µm water lines, 40 min afterthe G impact. From these data, Bjoraker et al. (1996b) reporteda mass of 3 × 1012 g above 3 µbar and 1.0 × 1013 g above10 µbar. These numbers must still be viewed as a lower limitas more H2O could be hidden at lower temperatures and deeperlevels, and are thus consistent with our finding. Still, the Bjorakeret al. (1996b) data do not provide an unambiguous determi-nation of the H2O/CO ratio in the impact-generated material.Near-infrared (2.0–2.4 µm) spectra acquired by Meadows andCrisp (1995, 1999) would be more suitable for this purpose,as they contain CO and H2O features simultaneously, but theiranalysis is fraught with complexity. We finally note that fromthe detection of a 22-GHz MASER emission on class 2 site E,Cosmovici et al. (1996) estimated a water mass of 5 × 1011 g.Once scaled up by a factor of 10 to account for the class differ-ence between impacts E and G, this is consistent with the resultsof Bjoraker et al. (1996b) and our inference.

A H2O/CO mass ratio of 0.07 is also fully consistent with theimpact chemistry calculations by Zahnle (1996). For a reactingparcel assumed to be an equal mix of jovian air and cometarymaterial, Zahnle (1996) finds that the H2O/CO ratio is hugelysensitive to the parcel O/C ratio, varying from essentially 0 forO/C < 1 to 0.5 for O/C = 2. According to these calculations,H2O/CO = 0.07 in mass is reached for O/C ∼ 1.3. As noted byZahnle (1996), however, no single value of the O/C ratio canexplain the ensemble of species observed after the impacts; inparticular, the presence of abundant HCN and CS2 implies thatchemistry occured as well in dry shocked jovian air.

4.2. A Hybrid Model

The biggest surprise is not, therefore, to see H2O and CO2

deriving from the SL9 impacts, but rather that the backgroundcomponent, attributed to be responsible for the water seen inSaturn, Uranus, Neptune, and Titan, does not show any obvioussignature in the Jupiter data. This must be quantified by puttingan upper limit on the steady influx rate. Doing so, however, is notstraightforward. A first method is to find the maximum H2O col-umn and associated background flux that does not overpredictany of the observed lines. In the framework of the full photo-
chemical background model, the line that indicates the lowest

126 LELLOUC

amount of water is the 66.44-µm line. Accounting for the noiseon this feature, it can be fit with at most a H2O column densityof 4.5 × 1014 cm−2 s−1, and the H2O flux is 8 × 104 cm−2 s−1.Note that the water column is 3.5 times less than the one indi-cated by the SWS lines, but the associated flux is 9 times lower,indicating a highly nonlinear character of the chemistry. For aCO flux of 4 × 106 cm−2 s−1, as typically estimated by Bezardet al. (2002), the CO2 column density in this background modelis 8.3 × 1013 cm−2, not inconsistent with the CO2 observations.Thus, a 8 × 104 cm−2 s−1 H2O flux is certainly a robust upperlimit; however, it is not fully satisfactory because it does not leadto a coherent explanation of all the ISO observations. A morecomplete approach is to construct a two-component model. Inwhat follows, we simplify the problem by describing the SL9component as spatially uniform and vertically uniformly mixeddown to some limiting pressure p0.

As noted above, in the case of a purely SL9 model, p0 =
0.3–0.7 mbar, similar but nominally slightly deeper than the ratio is a priori unknown. We believe that the associated water level reached by CO in 1997, 0.1–0.3 mbar according to the flux, 4 × 104 cm−2 s−1, may represent a good estimate of the
FIG. 14. Fit of the ISO H2O lines with the hybrid model. Short-dashed liLong-dashed lines: model for the background component (4 × 104 cm−2 s−1 flux

H ET AL.

millimeter–wave observations of Moreno (1998). Restrictingthe SL9 H2O to p < 0.2 mbar allows us to introduce a deeper,background component and still retain an acceptable overall fitof the ISO lines. Specifically, the combination of SL9-derivedH2O with q = 6 × 10−8 at p < 0.2 mbar (column density 1.2 ×1015 cm−2) and of a background component with 3.5 ×1014 cm−2 column density is tolerable, though it tends to un-derestimate the SWS lines and to overestimate the LWS lines.(The SWAS data are still underpredicted.) The associated waterprofile is shown in Fig. 4. Its disk-averaged column density is1.5 × 1015 cm−2. As illustrated in Fig. 14, the SL9 componentproduces most of the contrast of the ISO/SWS lines, while thebackground component is slightly dominant for the ISO/LWSlines. The nonlinearity of the LWS lines is clearly visible, thecombined model giving much less contrast that the sum of twomodels with the individual components. It is not possible to as-sess the uniqueness of this solution, as the SL9-derived mixing

nes: model for the SL9 component of this model (6 × 10−8 at p < 0.2 mbar).). Solid lines: model for the total hybrid distribution.


steady input flux into Jupiter; however, we conservatively adopt8 × 104 cm−2 s−1 as an upper limit.

The CO2 observations must also be explained in this context.For reasons explained above, the background water componenthas a negligible contribution to the formation of CO2. For theSL9 component model, the initial H2O/CO volume mixing ratiois taken as 0.11 × 1.2 × 1015/2 × 1015 = 0.066. This producesapproximately two-thirds of the observed CO2. While this isprobably a tolerable mismatch, a way to reach complete agree-ment is to assume that some CO2 was actually formed by impactthermochemistry. In this case, a CO2/CO initial ratio of about0.005 is required (Fig. 13). This is probably not unreasonable inview of Zahnle’s (1996) calculations of parcel composition as afunction of O/C ratio. Indeed, from figure 16 of Zahnle (1996),for a O/C ratio of 1.25, consistent with a H2O/CO volume ratioof 0.066, the CO2/CO ratio is ∼3.5 × 10−3. We also note thatthis model has an initial CO2/H2O volume mixing ratio of 7.5%,which remains marginally consistent with the 2–8% upper limitinferred by Encrenaz et al. (1997).

4.3. NonSL9 Oxygen in Jupiter’s Stratosphere

Although the associated fits are not quite as good as in thecase of a “pure SL9” model, the hybrid model may be favoredbecause (i) it allows a full reconciliation of the vertical distri-bution of SL9-derived H2O with that of CO and (ii) it acco-modates some background H2O, whose presence is expected.The H2O flux is, however, quite low, with a preferred valueof 4 × 104 cm−2 s−1 and an upper limit of 8 × 104 cm−2 s−1.This contrasts with the H2O fluxes into the other outer plan-ets (Feuchtgruber et al. 1997, Moses et al. 2000a). The fol-lowing discussion focuses on the comparison with Saturn, forwhich the most detailed modeling is available. Feuchtgruberet al. (1997) initially suggested a 3 × 105 − 5 × 106 cm−2 s−1

H2O flux at Saturn. Moses et al. (2000a) refined this estimateto 5 × 105 − 2 × 106 cm−2 s−1. They further found that, inorder to explain the ISO H2O and CO2 observations, the in-put of oxygen into Saturn must occur in the form of at leasttwo different species—H2O (or any species that would rapidlyform H2O in the atmosphere, such as OH, O(1D). . .) plus aspecies with a C–O bond (CO, CO2 . . .). The total oxygen fluxinto Saturn is (4 ± 2) × 106 cm−2 s−1, ∼10–90% of which iswater. We note, however, that most of the models consideredby Moses et al. (2000a) have a water flux of 1.5 × 106 cm−2

s−1; the only model (their model D) with a significantly lowerwater flux (5 × 105 cm−2 s−1) includes a 2 × 106 cm−2 s−1

flux of atomic oxygen. At high temperatures characteristic ofSaturn’s (and Jupiter’s) thermospheres, a large fraction of atomicO is converted to H2O through the reaction O + H2 → H + OH(followed by OH + H2 → H2O + H), which dominates overO + CH3 → H2CO + O above the methane homopause. Thus,to first order, a large O flux has the same effect as a large H2O
flux, unless O is introduced entirely below the methane ho-mopause and above ∼70 mbar, where O + CH3 dominates the

loss of atomic O. Taking 1.5 × 106 cm−2 s−1 as a typical valueof the “effective” (i.e., including O) H2O flux into Saturn, weconclude that the H2O flux into Jupiter is at least 20–40 timesless than in Saturn.

From the analysis of high-quality 5-µm Jupiter observations,Bezard et al. (2002) found evidence for a non-SL9 stratosphericcomponent of CO, with a (2–7) × 1016 cm−2 column density, in-dicative of a non-zero production or deposition of CO in Jupiter’sstratosphere. The associated flux depends primarily on the eddyKv at the tropopause. Our eddy Kv profile, having Kv = 700 cm2

s−1 at the tropopause, leads to a (2–7) × 106 cm−2 s−1 CO fluxinto Jupiter. Thus, the CO/H2O flux ratio in Jupiter is at leastequal to 30, with a preferred value of 100. In Saturn, the COexternal flux cannot be determined unambiguously, as current5-µm observations are inconclusive as to the presence of COin Saturn’s stratosphere (Noll and Larson 1990, Moses et al.2000a). However, as discussed above, the various models ofMoses et al. (2000a) indicate that water makes at least 10%(and 30–40% if atomic O is included) of the incoming oxygen.Therefore, the CO/H2O flux at Saturn is certainly below 10, andprobably below ∼3.

Thus, taken all together, the data indicate very similar to-tal oxygen fluxes into Jupiter and Saturn, but very differentCO/H2O flux (or production rate) ratios. Several explanationsmay be considered. A first possibility would be that microme-teoritic impacts predominantly lead to CO at Jupiter and H2Oat Saturn. A true compositional difference of the grains hittingJupiter and Saturn is unlikely to produce this effect, since theH2O/CO ratio in icy grains is expected to increase with decreas-ing heliocentric distance (since CO is more volatile than H2O,it is lost at greater distances than H2O). Alternatively, the ab-lation process could favor the formation of CO at Jupiter, as aresult of a difference in impact velocities (∼60 km s−1 at Jupiterand ∼35 km s−1 at Saturn). For instance, if a complex organiccomponent within the dust grain ablates along with an oxygen-bearing silicate or metal–oxide component, then CO may be pro-duced from the resulting gas–phase chemistry. However, basedon ablation models by Moses et al. (2001), we estimate that theresulting ablation temperatures on Jupiter are only ∼100–200 Khigher than on Saturn. Whether this temperature change canmake a difference (e.g., in ablation of kerogens) is not known.Note that subsequent stratospheric photochemistry cannot ex-plain the CO/H2O difference between the two planets. Two ef-fects combine to make CO production actually more effectiveon Saturn than on Jupiter: (1) icy grains ablate at lower pres-sures on Jupiter, at altitudes predominantly above the methanehomopause so that the ablated oxygen compounds tend to re-act with hydrogen to form hydrogen–oxygen bonds rather thancarbon–oxygen bonds, and (2) unsaturated hydrocarbons likeC2H2 and C2H4 are more abundant on Saturn than on Jupiter, fa-cilitating the photochemical conversion of H2O to CO on Saturn(see Moses et al. 2000a).

These considerations therefore put into question the hypoth-esis that interplanetary dust particles are the main source of

128 LELLOUC

external oxygen in Jupiter (and in the Giant Planets in general).Bezard et al. (2002) discuss the case where most of the oxy-gen supply to Jupiter is in the form of atomic O originatingfrom the Galilean satellites or the Io plasma torus. This mecha-nism was promoted by Strobel and Yung (1979) to explain thepresence of CO in Jupiter’s stratosphere. However, atomic Ois readily ionized in the magnetosphere and its precipitation isrestricted to auroral zones. Bezard et al. (2002) show that giventhe fluxes estimated to power the auroral emissions, this mech-anism is, on a planetary-averaged basis, a negligible source ofCO. In addition, as discussed above, this mechanism actuallyalso produces vast amounts of H2O because of the importanceof the O + H2, O+ + H2, and O+ + H reactions in the upperatmosphere. Specifically, we calculated that a model with a pureO flux of 4 × 106 cm−2 s−1 matches the CO column (4.2 ×1016 cm−2 above 300 mbar), but the H2O column is much toolarge (7.8 × 1015 cm−2).

It is remarkable that the background CO/H2O flux ratio wederive for Jupiter is reminiscent of (but even higher than) the rel-ative CO/H2O volume ratio we infer for the SL9-derived mate-rial, potentially suggesting a similar origin. Bezard et al. (2002)make the intriguing proposition that the stratospheric CO theysee reflects, at least in part, material that was deposited in therelatively recent past in one or several SL9-like event(s) and thathas diffused downward. With our eddy Kv profile, the e-foldingtimescale for CO diffusion across the 300-mbar bottleneck (seeFig. 4) is 320 years. Bezard et al. (2002) calculate the time-averaged CO production rate from recent events following thesize distribution of Levison et al. (2000), rescaled by using the“S factor” of Bottke et al. (2002), with a truncature at 320 years.They find it to be somewhat too low to fully account for the ob-served CO. Bringing it into agreement with data requires eitherincreasing the event frequency by a significant but reasonablefactor (∼5–10) or increasing the CO lifetime against verticaltransport by reducing the minimum eddy Kv to 100–300 cm2

s−1. These events also deliver H2O, with a H2O/CO volume ratioof 7–11% according to our models. However, in contrast to CO,H2O is subject to photolytic loss (Fig. 12) as well as condensationloss. In Table II, we estimate the fraction of initial CO and H2Othat persists above 300 mbar as a function of time after a SL9type impact. In Table II, the fifth column quantifies the differen-tial loss of H2O with respect to CO. It assumes that the two lossprocesses for H2O (condensation and photolysis) are uncoupled.This is true because, as mentioned above, the photolysis of H2Oalways occurs in an optically thin regime. The differential lossof H2O is not very effective until about 50 years. As can be seenfrom Table II, the calculated CO/H2O mixing ratio agrees withthe inferred value at times longer than ∼80 years. This wouldimply that the idea of “old SL9” impacts causing the observedCO is consistent with our H2O inferences if most of the CO wasdelivered in one or more relatively large impacts at least ∼100years ago rather than in many smaller and more recent impacts.

While this may provide a reasonable explanation, the low mi-crometeoritic flux at Jupiter must still be explained in view of the

H ET AL.

TABLE IILoss of SL9-produced CO and H2O

Time (years) f (CO)a fc(H2O)b f p(H2O)c fc (H2O) f p (H2O)f (CO)

dCO/H2Oe

10 1.00 1.00 0.76 0.76 1350 1.00 0.93 0.53 0.49 20

100 0.97 0.56 0.43 0.25 40200 0.70 0.17 0.31 0.075 130300 0.41 0.055 0.23 0.031 325

a Fraction of initial CO column persisting above 300 mbar.b Fraction of initial H2O column persisting in the stratosphere, assuming

condensation loss only.c Fraction of initial H2O column persisting in the stratosphere, assuming pho-

tolytic loss only.d See text.e Assuming an initial value of 10.

presence of external water in the other three Giant Planets andTitan. The orbital properties of the IDPs are not well known (seediscussion in Moses et al. 2000a), but in any case, the focusingfactor—and therefore the IDP flux—is higher for Jupiter than itis for all other Giant Planets. An upper limit of 8 × 104 cm−2 s−1

is marginally consistent with the estimated H2O flux at Uranus((6–16) × 104 cm−2 s−1; Feuchtgruber et al. 1997). In contrast,the inferred fluxes at Saturn ((4 ± 2) × 106 cm−2 s−1; Moseset al. 2000a), Neptune ((1.2–150) × 105 cm−2 s−1; Feuchtgruberet al. 1997) and Titan ((8–28) × 105 cm−2 s−1; Coustenis et al.1998) are much higher. While the rings may provide the dom-inant source at Saturn, as proposed by Prange et al. (1998),alternative sources for the other bodies are more difficult toidentify. Kuiper belt dust might be expected to be more im-portant at Neptune than at the other Outer Planets because ofmigration losses due to shattering by interstellar grains or scat-tering by each of the Giant Planets in succession. An alternateexplanation for the high Neptune flux would be to invoke arecent cometary event as well. Levison et al.’s (2000) esti-mates of cometary impact rates imply an equivalent flux perunit area ∼4 times larger at Neptune than at Jupiter. The pres-ence of HCN in Neptune’s stratosphere could be a signature ofsuch an event, although alternate explanations have been pro-posed (Marten et al. 1993, Lellouch et al. 1994). The simi-larity of the fluxes at Saturn and Titan is equally difficult toexplain, since Saturn’s rings do not provide oxygen to Titan(Moses et al. 2000a). A possibility is that Hyperion is a sig-nificant source of dust (and hence oxygen) to Titan, as advo-cated by Banaszkiewicz and Krivov (1997). We finally note thatin the case of Titan, photochemical models (Lara et al. 1996,Wong et al. 2002) require a large CO flux ((8–11) × 105 cm−2

s−1) in order to maintain the conservation of O atoms, adopt-ing a ∼5 × 10−5 CO mixing ratio. Evaporation from a surfaceocean and steady meteoritic influx have been invoked to sus-tain this flux. It is also not inconceivable that the CO abundance
in Titan is not in equilibrium and was affected by an impactevent.


5. SUMMARY AND PROSPECTS

The main results of our study can be summarized as follows:

(1) Rotational lines of H2O were detected in Jupiter by boththe Short-Wavelength Spectrometer and the Long-WavelengthSpectrometer of the Infrared Space Observatory in 1997, andby the Submillimeter Wave Astronomical Satellite in 1999 and2001. All the lines are detected in emission and probe watervapor in Jupiter’s stratosphere. Additionally, carbon dioxide15-µm emission was detected in disk-resolved ISO observa-tions and exhibits latitudinal variations, with a strong decreasefrom southern midlatitudes to northern midlatitudes.

(2) Although each individual H2O line may be fit by a modelin which water derives from a steady micrometeoritic interplan-etary source, the different lines then indicate discrepant valuesfor the H2O column density (from 4.5 × 1014 cm−2 to 2.8 ×1015 cm−2) and input flux (8 × 104 to 1.5 × 106 cm−2 s−1).

(3) The ISO SWS and LWS lines can be reconciled if mostof the stratospheric jovian water is restricted to pressures lessthan 0.5 ± 0.2 mbar, with a disk-averaged column density of(2.0 ± 0.5) × 1015 cm−2.

(4) The CO2 emission indicates a decrease of the CO2 columndensity by a factor of 2 from 45◦S ((6.3 ± 1.5) × 1014 cm−2) tothe equator ((3.4 ± 0.7) × 1014), and at least a factor 7 from45◦S to 45◦N (<7 × 1013 cm−2).

(5) The low pressure levels inferred for H2O and the markedlatitudinal variation of CO2 strongly suggest that both species re-sult from the delivery of oxygenated material by the Shoemaker–Levy 9 impacts in July 1994 and subsequent chemical and dy-namical evolution.

(6) A vertical transport model indicates that given typical val-ues of the vertical eddy diffusion coefficient in Jupiter’s strato-sphere (Kv ∼ 1 × 105 cm2 s−1 at∼0.2 mbar), material depositedat ∼0.1 mbar by impact chemistry—as was observed for CO—indeed diffuses down to the ∼0.5 mbar level after 3 years.

(7) Chemical evolution of a CO–H2O mix in proportionscharacteristic of SL9 impact chemistry indicates the progres-sive photolysis of H2O and conversion to CO2 (and ultimatelythe photolytic conversion of CO2 to CO). However, a coupledchemical–horizontal transport model indicates that the stabilityof water vapor is maintained over typically ∼50 years by thedecrease of the local CO mixing ratio associated with horizontalspreading.

(8) A model with an initial (i.e., SL9-produced) H2O/COmass mixing ratio of 0.07, not inconsistent with immediatepostimpact observations, matches the observed H2O abundanceand CO2 horizontal distribution 3 years after the impacts. In con-trast, the production of CO2 from SL9-derived CO and a watercomponent coming from an interplanetary dust component atthe flux level indicated in (2) is insufficient to account for theobserved CO2 amounts.

(9) The ISO observations can further be used to place a strin-gent upper limit (at most 8 × 104 cm−2 s−1, with 4 × 104 cm−2


s−1 as a more likely value) on the permanent water influx intoJupiter. This may indicate that the much larger flux observed atSaturn derives dominantly from a ring source, or that ablationof icy micrometeoroids leads primarily to the production of COat Jupiter and H2O at Saturn.

(10) Even if allowance is made for the temporal evolution ofthe SL9-derived material, the SWAS H2O spectra do not appearfully consistent with the ISO data and should be confirmed byfuture ODIN and Herschel observations.

Our essential conclusion, which may seem unexpected, thatjovian stratospheric H2O derives from the SL9 impacts, is basedon a combination of observational (LWS/SWS comparison) andtheoretical (consistency of H2O and CO2 amounts with modelexpectations) arguments, but the ultimate direct proof is stilllacking. It is of course unfortunate that ISO and SWAS were notoperational before the SL9 impacts. For now, the best demon-stration would obviously be evidence for latitudinal variations injovian H2O. Observational prospects in this respect are, however,not very good. H2O in Jupiter can in principle be observed fromthe ground with spatial resolution, either from large ground-based radiotelescopes in the 183.3-GHz line, which under ex-ceptionally dry conditions is not saturated in the Earth’s atmo-sphere, or in weak 10-µm lines. In both cases, however, chancesof detection are marginal at best. The prospect is somewhat bet-ter for 10-µm SOFIA airborne observations in 2005. In the moredistant future, observations of H2O at 1716 and 2640 GHz withthe submillimeter space telescope Herschel and its instrumentsHIFI (heterodyne) and PACS (bolometer array) will result ineasy detections of water with a spatial resolution of about 10′′.However, this facility will not become operational until 2007,at which time spatial variations of H2O will be largely sub-dued (only a 10% contrast is expected between −60◦ and +60◦

latitudes), considerably weakening the crucial aspect of the ob-servation.

ACKNOWLEDGMENTS

This study is based on observations with ISO, an ESA project with instru-ments funded by ESA Member States (especially the principal investigators’countries: France, Germany, The Netherlands, and the UK), and with partici-pation of ISAS and NASA. We are indebted to K. Zahnle, G. Bjoraker, andK. Lodders for useful discussions. GRD is grateful for the assistance of T. R.Fulton, S. D. Sidher and B. M. Swinyard in the analysis of the LWS data. GRD’sparticipation in this research was funded by the Natural Sciences and Engineer-ing Research Council of Canada. E. L. thanks T. Fouchet for his teaching inpreparing Fig. 9.

REFERENCES

Banaszkiewicz, M., and A. V. Krivov 1997. Hyperion as a dust source in thesaturnian system. Icarus 129, 289–303.

Baulch, D. L., C. J. Cobos, R. A. Cox, C. Esser, P. Frank, T. Just, J. A. Kerr,M. J. Pilling, J. Troe, R. W. Walker, and J. Warnatz 1992. Evaluated kineticdata for combustion modeling. J. Phys. Chem. Ref. Data 21, 411–734.

Beer, R. 1975. Detection of carbon monoxide in Jupiter. Astrophys. J. 200,L167–169.

130 LELLOUC

Beer, R., and F. W. Taylor 1978. The abundance of carbon monoxide in Jupiter.Astrophys. J. 221, 1100–1109.

Bergin, E. A., and 21 colleagues 2000. Submillimeter wave astronomy satelliteobservations of Jupiter and Saturn: Detection of 557-GHz water emissionfrom the upper atmosphere. Astrophys. J. 539, L147–150.

Bezard, B., C. A. Griffith, D. M. Kelly, J. H. Lacy, T. Greathouse, and G. Orton1997. Thermal infrared imaging spectroscopy of Shoemaker–Levy 9 impactsites: Temperature and HCN retrievals. Icarus 125, 94–120.

Bezard, B., P. Drossart, T. Encrenaz, and H. Feuchtgruber 2001. Benzene on thegiant planets. Icarus 154, 492–500.

Bezard, B., E. Lellouch, D. Strobel, J.-P. Maillard, and P. Drossart 2002. Carbonmonoxide on Jupiter: Evidence for both internal and external sources. Icarus158, 95–111.

Bjoraker, G. L., S. R. Stolovy, T. L. Herter, G. E. Gull, and B. E. Pirger 1996a. De-tection of water after the collision of fragments G and K of Comet Shoemaker–Levy 9 with Jupiter. Icarus 121, 411–421.

Bjoraker, G. L., S. R. Stolovy, T. L. Herter, G. E. Gull, and B. E. Pirger 1996b.KAO observations of water and methane in the G impact site. In InternationalConference on the SL9-Jupiter Collision, 3–5 July 1996, Meudon, France,p. I–3.

Bottke, W. F., Jr., A. Morbidelli, R. Jedicke, J.-M. Petit, H. F. Levison, P. Michel,and T. S. Metcalfe 2002. Debiased orbital and absolute magnitude distributionof the near-Earth objects. Icarus 156, 399–433.

Brown, L. R., and C. Plymate 1996. H2-broadened H216O lines in four infrared

bands between 55 and 4045 cm−1. J. Quant. Spectrosc. Radiat. Transfer 56,263–283.

Carlson, R. W., P. R. Weissman, J. Hui, M. Segura, W. D. Smythe, K. H. Baines,T. V. Johnson, P. Drossart, T. Encrenaz, F. Leader, and R. Mehlman 1995a.Some timing and spectral aspects of the G and R collision events as ob-served by the Galileo near infrared mapping spectrometer. In Proceedings ofEuropean SL–9/Jupiter workshop, (R. West and H. Bahnhardt, Eds.), pp. 69–74.

Carlson, R. W., P. R. Weissman, J. Hui, M. Segura, W. D. Smythe, K. H. Baines,T. V. Johnson, P. Drossart, T. Encrenaz, F. Leader, and R. Mehlman 1995b.Galileo infrared observations of the Shoemaker–Levy 9 G and R fireballsand flash. In Abstracts for IAU colloquim 156: The collision of comet p/

Shoemaker–Levy 9 and Jupiter, p. 15.

Clegg, P. E., and 61 colleagues 1996. The ISO long-wavelength spectrometer.Astron. Astrophys. Lett. 315, 38–42.

Cosmovici, C. B., S. Montebugnoli, A. Orfei, S. Pogrebenko, and P. Colom 1996.First evidence of planetary water maser emission induced by the comet/Jupitercatastrophic impact. Planet. Space Sci. 44, 735–739.

Coustenis, A., A. Salama, E. Lellouch, T. Encrenaz, G. L. Bjoraker, R. E.Samuelson, T. de Graauw, H. Feuchtgruber, and M. F. Kessler 1998. Evi-dence for water vapor in Titan’s atmosphere from ISO/SWS data. Astron.Astrophys. Lett. 336, 85–89.

Davis, G. R., I. Furniss, W. A. Towlson, P. A. R. Ade, R. J. Emery, W. M. Glen-cross, D. A. Naylor, T. J. Patrick, R. C. Sidey, and B. M. Swinyard 1995.Design and performance of cryogenic, scanning Fabry–Perot interferome-ters for the long-wavelength spectrometer on the Infrared Space Observatory.Appl. Opt. 34, 92–107.

de Graauw, T., and 61 colleagues 1996. Observing with the short-wavelengthspectrometer. Astron. Astrophys. Lett. 315, 49–54.

Encrenaz, T., P. Drossart, R. W. Carlson, and G. Bjoraker 1997. Detection ofH2O in the splash phase of G- and R-impacts from NIMS–Galileo. Planet.Space Sci. 45, 1189–1196.

Fegley, B., Jr., and K. Lodders 1994. Chemical models of the deep atmospheresof Jupiter and Saturn. Icarus 110, 117–154.

Feuchtgruber, H., E. Lellouch, T. de Graauw, B. Bezard, T. Encrenaz, and
M. Griffin 1997. External supply of oxygen in the giant planets. Nature 389,159–161.
H ET AL.

Feuchtgruber, H., E. Lellouch, T. Encrenaz, B. Bezard, A. Coustenis, P. Drossart,A. Salama, T. de Graauw, and G. R. Davis 1999. Oxygen in the strato-spheres of the giant planets and Titan. In The Universe as Seen by ISO,(P. Cox and M. F. Kessler, Eds.) ESA SP-427, pp. 133–136, Noordwijk,The Netherlands.

Fouchet, T., E. Lellouch, B. Bezard, T. Encrenaz, P. Drossart, H. Feuchtgruber,and T. de Graauw 2000a. ISO–SWS observations of Jupiter: Measurementof the ammonia tropespheric profile and of the 15N/14N isotopic ratio. Icarus143, 223–243.

Fouchet, T., E. Lellouch, B. Bezard, H. Feuchtgruber, P. Drossart, andT. Encrenaz 2000b. Jupiter’s hydrocarbons observed with ISO–SWS. Astron.Astrophys. Lett. 355, 13–17.

Friedson, A. J., R. A. West, A. K. Hronek, N. A. Larsen, and N. Dalal 1999.Transport and mixing in Jupiter’s stratosphere inferred from Comet S-L9 dustmigration. Icarus 138, 141–156.

Gierasch, P. J., A. P. Ingersoll, D. Banfield, S. P. Ewald, P. Helfenstein,A. Simon–Miller, A. Vasavada, H. H. Breneman, D. A. Senske, and the GalileoImaging Team 2000. Observation of moist convection in Jupiter’s atmosphere.Nature 403, 628–630.

Gladstone, G. R., M. Allen, and Y. L. Yung 1996. Hydrocarbon photochemistryin the upper atmosphere of Jupiter. Icarus 119, 1–52.

Griffin, M. J., P. A. E. Ade, G. S. Orton, E. I. Robson, W. K. Gear, I. G. Nolt, andJ. V. Radistitz 1986. Submillimeter and millimeter observations of Jupiter.Icarus 65, 244–256.

Hammel, H. 1996. HST imaging of Jupiter shortly after each impact: Plumesand fresh sites. In The Collision of Comet Shoemaker–Levy 9 and Jupiter(K. Noll, H. A. Weaver, and P. D. Feldman, Eds.), pp. 111–120. CambridgeUniv. Press, Cambridge, UK.

Kessler, M. F., A. J. Steinz, M. E. Anderegg, J. Clavel, G. Drechsel, P. Estaria,J. Faelker, J. R. Riedinger, A. Robson, B. G. Taylor, and S. Ximenez de Ferran1996. The Infrared Space Observatory (ISO) mission. Astron. Astrophys. Lett.315, 27–31.

Lara, L.-M., E. Lellouch, J. J. Lopez-Moreno, and R. Rodrigo 1996. Verticaldistribution of Titan’s atmospheric neutral constituents. J. Geophys. Res. 101,23,261–23,383; erratum: 1998, 103, 25,775.

Lellouch, E. 1996. Chemistry induced by the impacts: Observations. In TheCollision of Comet Shoemaker–Levy 9 and Jupiter (K. Noll, H. A. Weaver,and P. D. Feldman, Eds.), pp. 213–242. Cambridge Univ. Press, Cambridge,UK.

Lellouch, E., P. N. Romani, and J. Rosenqvist 1994. The origin and verticaldistribution of HCN in Neptune’s stratosphere. Icarus 108, 112–136.

Lellouch, E., H. Feuchtgruber, T. de Graauw, T. Encrenaz, B. Bezard, andM. Griffin 1997a. Deuterium and oxygen in the giant planets. In First ISOWorkshop on Analytical Spectroscopy, (A. M. Heras, K. Leech, N. R. Trams,and M. Perry, Eds.) ESA SP-419, 131–135, Noordwijk, The Netherlands.

Lellouch, E., B. Bezard, R. Moreno, D. Bockelee-Morvan, P. Colom, J. Crovisier,M. Festou, D. Gautier, A. Marten, and G. Paubert 1997b. Carbon monoxidein Jupiter after the impact of Comet Shoemaker–Levy 9. Planet. Space Sci.45, 1203–1212.

Lellouch, E., H. Feuchtgruber, T. Encrenaz, B. Bezard, P. Drossart, M. Griffin,and G. R. Davis 1998. D/H ratio and oxygen source: A Jupiter-Saturn com-parison. In The Jovian System After Galileo. The Saturnian System be-fore Cassini–Huygens. International Symposium, Nantes, 11–15 May 1998,pp. 21–22.

Lellouch, E., B. Bezard, T. Fouchet, H. Feuchtgruber, T. Encrenaz, and T. deGraauw 2001. The deuterium abundance in Jupiter and Saturn from ISO–SWSobservations. Astron. Astrophys. 370, 610–622.

Levison, H. F., M. J. Duncan, K. Zahnle, M. Holman, and L. Dones 2000.Planetary impact rates from ecliptic comets. Icarus 143, 415–420.

Marten, A., D. Gautier, T. Owen, D. B. Sanders, H. E. Matthews, S. K. Atreya,R. P. J. Tilanus, and J. R. Deane 1993. First observations of CO and HCN

of Comet Shoemaker–Levy 9 and Jupiter (K. Noll, H. A. Weaver, and P. D.

WATER AND CO2 IN JUPI

on Neptune and Uranus at millimeter wavelengths and the implications foratmospheric chemistry. Astrophys. J. 406, 285–297.

Marten, A., R. Moreno, H. E. Matthews, and T. Owen 1998. Monitoring ofJupiter’s stratosphere within the four years following the collision of CometShoemaker–Levy 9. Bull. Am. Astron. Soc. 30, 1069.

Meadows, V. S., and D. Crisp 1995. Impact plume composition from near-infrared spectroscopy. In Proceedings of European SL–9/Jupiter Workshop(R. West and H. Boehnhardt, Eds.), pp. 239–244.

Meadows, V. S., and D. Crisp 1999. The composition of the SL9 plumesfrom time-resolved near-infrared spectra. Bull. Am. Astron. Soc. 31,1191.

Melnick, G. J., and 20 colleagues 2000. The Submillimeter Wave AstronomySatellite: Science objectives and instrument description. Astrophys. J. 539,L77–85.

Moreno, R. 1998. Observations Millimetriques et Submillimetriques desPlanetes Geantes. Etude de Jupiter Apres la Chute de la Comete SL9. Ph.D.thesis, Universite Paris-6.

Moses, J. I. 1996. SL9 impact chemistry: Long-term photochemical evolution.In The Collision of Comet Shoemaker–Levy 9 and Jupiter (K. Noll, H. A.Weaver, and P. D. Feldman, Eds.), pp. 243–268. Cambridge Univ. Press,Cambridge, UK.

Moses, J. I., M. Allen, and G. R. Gladstone 1995. Nitrogen and oxygen photo-chemistry following SL9. Geophys. Res. Lett. 22, 1601–1604.

Moses, J. I., E. Lellouch, B. Bezard, G. R. Gladstone, H. Feuchtgruber, andM. Allen 2000a. Photochemistry of Saturn’s atmosphere. II. Effects of aninflux of external oxygen. Icarus 145, 166–202.

Moses, J. I., B. Bezard, E. Lellouch, G. R. Gladstone, H. Feuchtgruber, andM. Allen 2000b. Photochemistry of Saturn’s atmosphere. I. Hydrocarbonphotochemistry and comparisons with ISO observations. Icarus 143, 244–298.

Moses, J. I., T. Fouchet, B. Bezard, E. Lellouch, G. R. Gladstone,H. Feuchtgruber, and M. Allen 2001. Comparative planetology: Lessons fromphotochemical modeling of the upper atmospheres of Jupiter and Saturn. Bull.Am. Astron. Soc. 33, 1126.

Noll, K. S., and R. F. Knacke 1998. Response to comment on “Carbon monoxidein Jupiter after Comet Shoemaker–Levy 9.” Icarus 133, 322–324.

Noll, K. S., and H. P. Larson 1990. The spectrum of Saturn from 1990 to2320 cm−1: Abundances of AsH3, CH3D, CO, GeH4, NH3, and PH3. Icarus89, 168–189.

Noll, K. S., R. F. Knacke, T. R. Geballe, and A. T. Tokunaga 1986. The detectionof carbon monoxide in Saturn. Astrophys. J. 309, L91–94.

Prange, R., T. Fouchet, and R. Courtin 1998. FUV remote sensing of the atmo-sphere of Saturn with HST: Polar haze and latitudinal dependence of
water abundance. In The Jovian System After Galileo. The Saturnian System

before Cassini–Huygens. International Symposium, Nantes, 11–15 May 1998,pp. 25–26.

Prather, M. J., J. A. Logan, and M. B. McElroy 1978. Carbon monoxide inJupiter’s upper atmosphere: An extraplanetary source. Astrophys. J. 223,1072–1081.

Salama, A., H. Feuchtgruber, A. Heras, F. Lahuis, K. J. Leech, R. Lorente,P. Morris, B. Vandenbussche, and E. Wieprecht 1997. Topics in SWS dataanalysis and calibrations. In First ISO Workshop on Analytical Spectroscopy,(A. M. Heras, K. Leech, N. R. Trams, and M. Perry, Eds.) ESA SP-419, 17–21,ESA Publications Division ESTEC, Noordwijk, The Netherlands.

Sanchez-Lavega, A., J. M. Gomez, J. F. Rojas, J. R. Acaretta, J. Lecacheux,F. Colas, R. Hueso, and J. Arregui 1998. Long-term evolution of Comet SL-9impact features: July 1994–September 1996. Icarus 131, 341–357.

Simon, A., and R. Beebe 1996. Jovian tropospheric features: Wind field, mor-phology, and motion of long-lived systems. Icarus 121, 319–330.

Smith, M. D. 1998. Estimation of a length scale to use with the quenchlevel approximation for obtaining chemical abundances. Icarus 132,176–184.

Sprague, A. L., G. L. Bjoraker, D. M. Hunten, F. C. Witteborn, R. W. H.Kozlowski, and D. H. Wooden 1996. Water brought into Jupiter’s atmosphereby fragments R and W of Comet SL-9. Icarus 121, 30–37.

Strobel, D. F., and Y. L. Yung 1979. The Galilean satellites as a source of CO inthe jovian upper atmosphere. Icarus 37, 256–263.

Swinyard, B. M., and 54 colleagues 1996. Calibration and performanceof the ISO Long-Wavelength Spectrometer. Astron. Astrophys. Lett. 315,43–48.

Swinyard, B. M., M. J. Burgdorf, P. E. Clegg, G. R. Davis, M. J. Griffin, C. Gry,S. J. Leeks, T. Lim, S. Pezzuto, and E. Tommasi 1998. In-orbit performanceof the ISO Long-Wavelength Spectrometer. In Proceedings Infrared Astro-nomical Instrumentation (SPIE 3354), Kona, March 1998, pp. 888–899.

West, R. 1996. Particles in Jupiter’s atmosphere from the impacts of CometP/Shoemaker–Levy 9. In The Collision of Comet Shoemaker–Levy 9 andJupiter (K. Noll, H. A. Weaver, and P. D. Feldman, Eds.), pp. 269–292.Cambridge Univ. Press, Cambridge, UK.

West, R. A., A. J. Friedson, and J. F. Appleby 1992. Jovian large-scale strato-spheric circulation. Icarus 100, 245–259.

Wong, A.-S., C. G. Morgan, and Y. L. Yung 2002. Evolution of CO in Titan.Icarus 155, 382–392.

Yung, Y. L., W. A. Drew, J. P. Pinto, and R. R. Friedl 1988. Estimation of thereaction rate for the formation of CH3O from H + H2CO: Implications forchemistry in the Solar System. Icarus 73, 516–526.

Zahnle, K. 1996. Dynamics and chemistry of SL9 plumes. In The Collision

Feldman, Eds.), pp. 183–212. Cambridge Univ. Press, Cambridge, UK.

the origin of water vapor and carbon dioxide in jupiter's stratosphere

Documents