RADIATION SYNTHESIS OF CARBON DIOXIDE IN ICE-COATED CARBON: IMPLICATIONS FOR INTERSTELLAR GRAINS AND ICY MOONS

There is ample evidence for the common occurrence of solid CO₂ on the icy satellites of the outer planets. Reflectance spectra of Galilean satellites (Europa, Callisto, and Ganymede) acquired by Galileo's Near-Infrared Mapping Spectrometer showed infrared features indicative of CO₂ (Carlson et al. 1996; Hansen & McCord 2008; Hibbitts et al. 2000, 2003). More recently, Cassini's Visible and Infrared Mapping Spectrometer has revealed the presence of surficial CO₂ in the icy regolith of the Saturnian moons Encealdus, Dione, Phoebe, Iapetus, and Hyperion (Brown et al. 2006; Buratti et al. 2005; Cruikshank et al. 2010). Additionally, Cassini's Ion Neutral Mass Spectrometer recently detected a CO₂ exosphere at Rhea, which is thought to generated from sputtering of condensed CO₂ (Teolis et al. 2010). Ground-based observations have also identified condensed CO₂ via overtone absorptions at ∼2 μm on Ariel, Umbriel, and Titania, the moons of Uranus (Grundy et al. 2006) and on Neptune's Triton (Cruikshank et al. 1993). Radar measurements by the Mars Reconnaissance Orbiter (Phillips et al. 2011) have recently shown that solid CO₂ levels in the Martian southern pole are 30 times higher than a previous estimate (Leighton & Murray 1966). CO₂ has also been detected in the dust tails of various cometary nuclei (Feaga et al. 2007; Feldman et al. 2009; Ootsubo et al. 2010).

Solid CO₂ has also been detected in the interstellar medium (ISM) by the Spitzer Space Telescope and the Infrared Space Observatory (Boogert et al. 2004; Gerakines et al. 1999; Gibb et al. 2004). Solid CO₂ is a major constituent of the icy mantles coating the micron-sized dust grains in these clouds with abundance between 9% and 37% relative to water ice, the dominant component (Whittet et al. 2007). Furthermore, CO₂ is detected in quiescent clouds such as Elias 16 (Whittet et al. 1998), or in clouds with an embedded protostar such as W33 A (Gibb et al. 2000), which collapse to form protoplanetary disks.

Understanding the origin of molecules, like CO₂, in diverse environments such as the satellites of the outer planets or dense clouds in the ISM is a central quest in astrophysics, one that is advanced by complementing observations with modeling and laboratory experiments.

Several competing mechanisms have been proposed to explain the origin of CO₂ on the icy moons. One hypothesis suggests that CO₂ is endogenic, i.e., it was either incorporated in the regolith from direct accretion during its formation or formed via aqueous alteration of carbon-bearing materials present in the satellites' interior (Carlson et al. 2009). Alternatively, CO₂ could also be exogenic, since carbonaceous materials required for CO₂ synthesis could be delivered to these satellites by meteorites. This idea is supported by the enhancement in infrared absorption due to CO₂ in the vicinity of fresh crater sites, for instance on Europa and Callisto (Hibbitts et al. 2000). Ionizing radiation (magnetospheric ions, solar photons, electrons) impinging on the surface of these satellites could produce CO₂ by initiating chemical reactions between water molecules and carbonaceous material. In addition, CO₂ may be produced by implantation of magnetospheric carbon ions into the icy regolith (Strazzulla 2011).

In the cold ISM, oxidation of the abundant solid CO could lead to the formation of observed CO₂. The oxygen atoms needed for this process may have been accreted in the grain, especially in the dense clouds where radiation is sparse, or they may have resulted from the dissociation of neighboring molecules in the ice, such as O₂ or H₂O, by radiation such as Lyα, cosmic rays or stellar winds. In our recent work (Raut & Baragiola 2011), we found that the reaction of CO and cold O atoms is relatively inefficient at making CO₂, mostly because oxygen reacts preferentially with O atoms to form O₂ and O₃. Furthermore, other laboratory studies have demonstrated that irradiation of solid CO (Loeffler et al. 2005) or mixtures of CO with water ice or solid O₂ (Gerakines et al. 2001; Satorre et al. 2000) can produce CO₂, but the maximum CO₂: CO ratios obtained are much smaller than those observed in the ISM. On the other hand, grains on which CO₂ is produced from CO + O reaction and CO radiolysis can, if heated, end up with high CO₂: CO ratio due to the much higher volatility of CO as compared to CO₂ (Raut & Baragiola 2011). Such heating may result from the energy deposited by fast heavy cosmic rays on the dust grains and induce selective desorption (Bringa & Johnson 2003; Leger et al. 1985).

Although little is known about the composition of the interstellar dust grains, spectroscopic evidence suggests a carbonaceous component. Specifically, the 2175 Å bump in the stellar extinction curves is thought to be caused by the UV absorption shown by many carbon-bearing species. Some possible candidates are small graphitic grains, ultraviolet light-processed hydrogenated amorphous carbon, hyper-fullerene carbon particles, and even poly-aromatic hydrocarbons (Draine 2003). These carbon-containing grains are coated with an ice mantle that is composed primarily of water ice, but also contain other molecules such as solid CO, O₂, and CH₃OH. Radiation processing of the oxygen-bearing ices could yield radicals such as O or OH that react with the carbonaceous dust to form interstellar CO and CO₂. Such reactions cause mass loss or etching of carbon; this is a well-known effect that dates back to 1950s, when the degradation of carbonaceous material within frozen biological samples was observed under the electron microscope (Heide 1982; Talmon et al. 1979). Etching was attributed to oxidation; energetic electrons dissociate water molecules to form radicals (O, OH, HO₂) that oxidized carbon into CO, CO₂, and also formed various hydrocarbons and H₂, which were seen to desorb from the samples. Recently, several groups have investigated etching for different carbon allotropes (graphite, amorphous carbon, carbon nanotubes) with focused electron and ion beams in the presence of ambient water vapor or other gases such as O₂ (Krasheninnikov & Banhart 2007; Lee et al. 1999; Spinney et al. 2009; Yuzvinsky et al. 2005). These studies have explored etching as a tool to making nanostructures, which are important in integrated circuits and other electronic devices; less attention has been given to the reaction products of oxidative etching. Most of the experiments have been performed at room temperature, where CO and CO₂ desorbed readily from the underlying carbon and were ultimately removed via vacuum pumps.

Seeking to understand CO₂ abundance in extraterrestrial environments, several groups have investigated the formation of CO and CO₂ upon irradiation of water-ice coatings on carbonaceous materials at low temperatures. In contrast to the carbon-etching experiments performed at room temperature, the ice coating can help trap the oxidation products such as CO and CO₂. Mennella et al. reported CO and CO₂ synthesis from irradiation with Lyα photons (Mennella et al. 2006) and energetic ions (Mennella et al. 2004) of thin ice films condensed on hydrogenated carbon grains. Similarly, Gomis & Strazzulla (2005) observed CO₂ formation from irradiation of ice films deposited on top of different carbon-bearing substrates such as asphaltite and irradiated benzene. The results of these ion-impact studies differ in the temperature dependence and in the total production of CO and CO₂, likely due to the use of different substrates (carbon grain size, porosity, degree of hydrogenation, etc.), and, at least in part, due to the neglect of interference effects in the infrared absorption measurements (Teolis et al. 2007).

To better understand the mechanisms for radiation-induced carbon oxidation, we undertook experiments with well-defined surfaces of amorphous ¹³C substrates cleaned by removing atmospheric contamination by Ar ion sputtering in ultrahigh vacuum. The irradiations were made with mass-analyzed 100 keV protons at 20 and 120 K, temperatures typical of interstellar ice grains and icy moons in the outer solar system, and characterization of the radiation effects was performed with reflection infrared absorption spectroscopy taking due account of interference effects (Teolis et al. 2007), together with microgravimetry.

The experiments were performed in an ultrahigh vacuum chamber cryopumped to a base pressure of ∼10⁻¹⁰ Torr. It houses a quartz-crystal microbalance (QCM) cooled by a closed-cycle refrigerator. The QCM is coated with a ¹³C foil (∼50 nm thick, >98% isotopic purity) made by Arizona Carbon Foils using electron beam evaporation of carbon and condensation onto glass slides previously coated with a parting layer such as BaCl₂. The foil separates from the glass slide and floats when immersed into HPLC-grade high-purity water. A small uniform section of the foil is layered on top of the polished gold electrode of the QCM. Placing the foil-coated quartz crystal under a heat lamp for several hours, followed by baking the QCM in a high-vacuum oven (∼10⁻⁶ Torr), removes liquid water trapped between the foil and the gold surface. Care was taken to ensure that the foil coverage over the active area of the QCM was uniform and smooth, since creases and tears in the foil cause non-uniform mass loading, affecting the performance of the QCM and decreasing the specular reflectance of the carbon-coated gold. The ¹³C-coated QCM is irradiated with 100 keV Ar⁺ ions to fluences of ∼10¹⁴ ions cm⁻² to sputter-remove adventitious hydrocarbons that coat the foil during transfer to the vacuum system. The amount of material removed is <1.5% of the foil mass.

The use of isotopic ¹³C foil ensured two advantages. The solid¹³CO₂ has an infrared absorption band at 2275 cm⁻¹ (4.40 μm), shifted from the 2345 cm⁻¹ (4.26 μm) band of solid ¹²CO₂ that can accumulate in the film from spurious sources, such as gas from the accelerator beam line and gas desorbed from the walls of the vacuum chamber by thermal or particle-induced desorption. Furthermore, the absorption bands of solid¹³CO₂ are shifted from those of residual gaseous ¹²CO₂ present in the external path of the infrared beam, a perturbation that is present only when the quality of the dry air purge is not optimum.

Thin ice films were vapor-deposited at 20 and 120 K on top of the carbon foil. The column densities of the ice films, measured by the QCM, were 276 × 10¹⁵ H₂O cm⁻², corresponding to a thickness of ∼100 nm. We irradiated the films at their growth temperature at near normal incidence (9°) with 100 keV protons from an ion accelerator with a magnetic mass filter. At 100 keV, the 1.3 μm ion range for protons is much larger than the ∼0.1 μm ice thickness (Ziegler & Biersack 2006). In fact, the ions penetrate the 50 nm carbon layer and are implanted into the gold electrode of the QCM.

Infrared reflectance spectra R were collected under specular conditions at a 35° angle of incidence using a Thermo-Nicolet Nexus 670 spectrometer at 2 cm⁻¹ resolution. The spectra were divided by the reflectance spectrum of the bare carbon–gold substrate R_o and converted to optical depth, −ln(R/R_o). Subtraction of straight baselines from the optical depth spectra, followed by integration in the absorption region, yields band areas. Dividing the band areas by the effective band strength A* gives the amount of CO₂, η, produced in the sample. The qualifier effective implies that the value of band strength is not a unique property of the molecule in the solid phase, but rather depends on other factors such as geometry (reflectance versus transmission), refractive index, and thickness of the films (Teolis et al. 2007). Therefore, using band strengths published elsewhere can introduce error in estimates of η. Instead, we performed experiments where known quantities of ¹³CO₂ in the range of 1–10 ML, as measured by the microbalance, were deposited on top of the carbon foil and then covered with ∼100 nm thick ice films. These experiments gave A* = (6.1 ± 0.6) × 10⁻¹⁷ cm mol⁻¹, a value that corresponds to our experimental conditions and should not be used for films of a different thickness or for other reflectance geometries.

Mass loss due to sputtering of water ice by the proton bombardment is measured with the QCM. It can decrease the film thickness significantly at high fluences. To maintain the original film thickness of ∼100 nm, we deposited water ice to compensate for sputtering loss following irradiation steps above a fluence of ∼5× 10¹⁵ H⁺ cm⁻². The importance of maintaining a constant film thickness is to avoid errors in quantifying CO₂ levels using infrared spectroscopy, since, as mentioned earlier, A* depends on thickness due to interference effects.

In another experiment, we deposited 25 ML of ¹³CH₄ on top of the ¹³C substrate at 20 K. We then covered the film with 267 ML of water ice and irradiated the film with 100 keV protons, while monitoring the consumption of methane and the production of ¹³CO₂. We performed this experiment to compare CO₂ production between ice-coated carbon grains and ice-coated solid CH₄.

Figure 1 shows the evolution of the ν₃ absorption feature of ¹³CO₂ at 2275 cm⁻¹ versus irradiation fluence for 100 nm thick H₂O films deposited over the carbon substrate at 20 K and 120 K. The films were irradiated at the temperature of film growth. In some instances, the spectra curve at higher wavenumbers (∼2320 cm⁻¹) due to the fluctuating presence of ¹²CO₂ gas in the path of the infrared beam. The ¹³CO₂ absorption is sufficiently redshifted from the ¹²CO₂ absorption, such that changes in the continuum caused by fluctuations in the atmospheric CO₂ levels do not interfere with the ¹³CO₂ measurements. For comparison, we also show in Figure 1 the ¹³CO₂ absorption in the infrared spectra of molecular clouds in the direction of protostars S140: IRS 1 and W 33A. The differences between interstellar CO₂ spectra and laboratory spectra are discussed later (see Section 5).

Zoom In Zoom Out Reset image size

The fluence dependences of ¹³CO₂ production at 20 and 120 K are shown in Figure 2. We observe that η has an initial linear growth with fluence and approaches steady-state values at ∼30 × 10¹⁵ H⁺ cm⁻² at both temperatures. The onset of saturation is apparent in the data at 120 K and is suggested at 20 K. The saturation value η_s at 120 K is about 2.5 times larger than that at 20 K. Even at saturation fluences, we found no evidence of CO above our sensitivity of 1.4 × 10¹⁵ molecules cm⁻², implying a CO/CO₂ ratio less than 1.4 and 0.52 at 20 and 120 K, respectively.

Zoom In Zoom Out Reset image size

Figure 3 shows normalized ν₃ absorption spectra of ¹³CO₂. Spectra in the top panel are from 1 ML of ¹³CO₂ produced from 100 keV irradiation of an ice-coated carbon foil at 20 K. Heating the film to 140 K did not produce significant changes, except for a slight sharpening of the band. The CO₂ produced from radiolysis remains trapped in the ice coat beyond 80 K, which is the desorption temperature for pure CO₂ films. Spectra in the middle panel belong to 2 ML ¹³CO₂ deposited on top of the carbon foil and coated with 267 ML of the ice layer at 20 K. Comparing the spectrum of the deposited film at 20 K with the spectra of an ion-irradiated sample shown in the top panel highlights additional structures at 2306 cm⁻¹ and ∼2260 cm⁻¹. These two features disappear above 100 K during ramping the film to desorption whereas the main CO₂ is not observed above 150 K, consistent with previous desorption data (Malyk et al. 2007).

Zoom In Zoom Out Reset image size

The bottom panel shows synthetic spectra generated with Fresnel's equations using optical constants of ¹³CO₂, H₂O, and amorphous carbon (aC; Teolis et al. 2007). One spectrum belongs to a 100 nm thick film with 1% CO₂ mixed in water ice and is obtained using optical constants calculated with the effective medium theory (Born et al. 1999). The other spectrum is for a 1 nm layer of CO₂, capped by 100 nm H₂O. The synthetic spectra are generated using unpolarized light at 35° incidence and gold substrate. Comparing the experimental and the synthetic spectra yields information about the physical state of the radiolytic-generated CO₂, which we will discuss below.

In Figure 4, we compare the fluence dependence of CO₂ synthesis obtained from this work with data from several published experiments (Gomis & Strazzulla 2005; Mennella et al. 2004), where a variety of carbonaceous substrates are coated with thin ice films and irradiated with light ions with different energies. The CO₂ is normalized to the maximum yield obtained in each experiment. Note the large fluence shift between data for carbon and hydrogenated carbon.

Zoom In Zoom Out Reset image size

4.1. Physical State of Radiolytic CO₂

4.2. CO₂ Synthesis and Destruction Mechanisms

The atomic constituents of the CO₂ molecules are separate prior to irradiation. Therefore, CO₂ formation requires oxidation of the carbon atoms by oxygen-bearing species produced by radiolysis of water near the ice/carbon interface. The energy deposited by the fast ions in water goes into making transient reactive species such as H, O, OH in their neutral, excited, and ionic forms, which react further via multiple pathways to form stable end-products such as H₂, O₂, HO₂, HO₃, and H₂O₂ (Spinks & Woods 1990). In principle, any oxygen-containing species can lead to oxidation of the carbon substrate. However, O₂ does not react with carbon at low temperatures and the same is true for H₂O₂, as judged by the lack of CO₂ formation when high-purity hydrogen peroxide was deposited on carbon grains at 16 K in a different, but related, experiment. We begin by examining possible reaction pathways that produce O-bearing radicals from radiolysis of water capable of oxidizing carbon at the substrate.

An impinging ion X can directly break a water molecule

$$\begin{equation} X + {\rm H}_2 {\rm O } \to {\rm H} + {\rm OH} \end{equation} \tag{ 1 }$$

or cause ionization resulting in electron emission

$$\begin{equation} X + {\rm H}_2 {\rm O} \to {\rm H}_2 {\rm O}^ + + {e}^{ -}. \end{equation} \tag{ 2 }$$

The electron and ionic H₂O in (2) can further react with other water molecules

$$\begin{equation} {\rm H}_2 {\rm O}^ + + {\rm H}_2 {\rm O} \to {\rm H}_3 {\rm O}^ + + {\rm OH}^ - \end{equation} \tag{ 3 }$$

$$\begin{equation} e^ - + {\rm H}_2 {\rm O} \to {\rm H} + {\rm OH}^ - . \end{equation} \tag{ 4 }$$

Additionally, the ions can also cause direct excitation of the water molecule

$$\begin{eqnarray} X + {\rm H}_2 {\rm O} &\to& {\rm H}_2 {\rm O}^* \to {\rm H} + {\rm OH}\\ \ms &\to& {\rm H}_2 {\rm O}^* \to {\rm O} + 2{\rm H}\\ \ms &\to& {\rm H}_2 {\rm O}^* \to {\rm O} + {\rm H}_{\rm 2}. \end{eqnarray} \tag{ 5 }$$

Reactions (6) and (7) leading to production of O atoms are ∼10% compared to reaction (5) (Slanger & Black 1982). This leads to a radiolyzed ice film more abundant in OH and H, compared to O atoms. The fate of the H and OH radicals is strongly influenced by diffusion, which is temperature dependent. H atoms can diffuse, even at low temperatures such as 20 K, and combine with OH to reform water or associate with another H to form H₂ and escape the ice. In contrast, diffusion of OH is minimal at 20 K (Johnson 1997). OH radicals not consumed in reaction with H to reform water react to form hydrogen peroxide:

$$\begin{equation} {\rm OH} + {\rm OH} \to {\rm H}_2 {\rm O}_2. \end{equation} \tag{ 8 }$$

Though H₂O₂ is synthesized in our films via channel (8), its steady-state abundance, ∼0.3 × 10¹⁵ H₂O₂ cm⁻² (derived from Loeffler et al. 2006), is below the detection limit. At higher temperatures, the OH diffuses away from the ion track, inhibiting H₂O₂ synthesis (Loeffler et al. 2006).

Near the carbon–water interface, OH radicals can also react with the carbon to form CO₂ in steps via creation of an intermediate CO, since it is unlikely that two OH radicals are available to oxidize a single carbon atom at the same instant:

$$\begin{eqnarray} {\rm OH} &+& {\rm C}({\rm s}) \to {\rm CO} + {\rm H}\nonumber\\ {\rm OH} &+& {\rm CO}\; \to {\rm CO}_2 + {\rm H}. \end{eqnarray} \tag{ 9 }$$

The H atoms in Equation (9) can recombine to form H₂.

Additionally, CO₂ synthesis can also occur via dissociative electron attachment (DEA), where secondary electrons emitted from the carbon during irradiation (Baragiola et al. 1979; Ritzau & Baragiola 1998) are temporarily captured by water, making short-lived anions that dissociate to make OH radicals:

$$\begin{equation} e^ - + {\rm H}_2 {\rm O} \to {\rm H}_2 {\rm O}^{ - *} \to {\rm H} + {\rm OH}^ - . \end{equation} \tag{ 10 }$$

The backward secondary electron yield from carbon foils due to 100 keV protons is ∼3 (Pauly et al. 2005; Ritzau & Baragiola 1998). Formation of DO₂ and D₂O₂ in ice by DEA has been demonstrated in deuterated water ice irradiated with electrons of energy > 5 eV (Pan et al. 2004; Simpson 1997).

Once formed, a CO₂ molecule can dissociate back to CO from a direct impact from ion X:

$$\begin{equation} X + {\rm CO}_2 \to {\rm CO} + {\rm O}. \end{equation} \tag{ 11 }$$

In addition, since the CO₂ is embedded in a water-ice matrix, reactions between H₂O and CO₂ molecules initiated by ions can also lead to CO₂ destruction. Carbonic acid is the dominant product observed in experiments where mixtures of CO₂ and H₂O are irradiated with keV ions, along with trace amounts of HCO, HCOOH, H₂CO, CO₃, and O₃ (Hudson & Moore 2001; Strazzulla et al. 2005a, 2005b). However, we did not observe H₂CO₃ or other molecules in our experiment, indicating, as in the case of CO, that their concentrations were below the detection limits.

4.3. Rate Equation: Creation Yield, Destruction Cross Sections and Steady-state Values

The rate equation below describes the CO₂ creation and destruction during ion irradiation:

$$\begin{equation} \frac{{d\eta }}{{dF}} = y_c - \sigma _d \eta, \end{equation} \tag{ 12 }$$

where y_c is the creation yield (CO₂/proton) and σ_d is the net destruction cross section for CO₂. The CO₂ column density approaches a near-constant, steady-state value when y_c ≈ σ_d × η. Integration of Equation (12) gives

$$\begin{equation} \eta (F) = \frac{{y_c }}{{\sigma _d }}[ {1 - \exp ({ - \sigma _d \;F})}]. \end{equation} \tag{ 13 }$$

The solid lines in Figure 2 are the fits of Equation (13) to data. The fit yields a steady-state CO₂ value η_s = (2.8 ± 0.1) × 10¹⁵ CO₂ cm⁻² and an effective σ_d = (4.1 ± 0.3) × 10⁻¹⁷ cm² at 120 K. The onset of saturation is not clear from the data at 20 K. Extrapolation of the fit suggests η_s ∼ 1.6 × 10¹⁵ CO₂ cm⁻² and σ_d ∼ 2.5 × 10⁻¹⁷ cm². The initial creation yield y_c (CO₂/ion) = σ_d × η_s is 0.05 ± 0.01 and 0.1 ± 0.01 at 20 and 120 K, respectively. In the low-F limit, Equation (13) reduces to η (F) = y_c F. The higher values of y_c and η_s at 120 K are likely due to increased availability and mobility of the OH radicals. Previous studies (Loeffler et al. 2006) have shown that H₂O₂ production from ion irradiation in pure water-ice film decreases with increasing temperature. This is attributed to out-diffusion of OH radicals from the ion tracks. Since fewer OH radicals are consumed in H₂O₂ synthesis, more are available for C oxidation at 120 K via the reaction channels shown in Equation (9).

The production of CO, below our detectability limit, is seen in experiments with hydrogenated carbon grains having large surface areas and is roughly proportional to CO₂ (Mennella et al. 2004). This means that CO and CO₂ go through a conversion cycle without net loss. Satorre et al. (2000) have shown that, for a wide range of concentrations of CO₂ in water and without a carbon substrate (i.e., y_c = 0), the ratio of CO to CO₂ reaches a constant value k at high fluences. Then, Equation (12) can be rewritten as

$$\begin{equation} \frac{{d\eta }}{{dF}} = y_c + \left[ {\sigma _{12} k\eta - \sigma _{21} \eta } \right] - \sigma _{2c} \eta, \end{equation} \tag{ 14 }$$

where σ₁₂, σ₂₁, and σ_2c are the cross sections for converting CO into CO₂, CO₂ into CO, and CO₂ into carbonic acid and other hydrocarbons (Hudson & Moore 2001; Strazzulla, et al. 2005a), respectively, where we have neglected back-conversion to CO, CO₂. This implies that the effective destruction cross section for CO₂

$$\begin{equation*} \sigma _d = - \sigma _{12} k + \sigma _{21} + \sigma _{2c} \end{equation*}$$

is small, 100 times smaller than σ₂₁ in initially pure CO film irradiated with 200 keV protons (Loeffler et al. 2005), where carbonic acid cannot be produced. The increase in σ_d with temperature may be due to the easier escape of dissociation fragments of CO₂ from the surrounding molecular environment (cage effect; Spinks & Woods 1990).

4.4. Absence of CO

4.5. Comparison with Previous Studies

These experiments show that carbon dioxide can be formed by radiolysis at the interface between water ice and carbon. Similar situations exist in the ISM, such as in the case of clouds with embedded protostars, and in the cold edges of protoplanetary disks, where carbonaceous dust grains coated with water ice are continually impacted by GeV cosmic rays or stellar winds. The maximum column density of CO₂ that can be formed by this mechanism at one interface is of the order of a monolayer (as seen in our experiments at 20 K). Therefore, this mechanism is insufficient to explain the high amount of CO₂ (∼10%–40% relative to water ice) observed toward different molecular clouds (Whittet et al. 1998), but should contribute to the total interstellar CO₂ abundance. More CO₂ would be formed if the incident radiation passes through multiple interfaces; such would be the case of nanoscale carbon particles dispersed in ice or of porous carbon grains containing embedded water. Additionally, since H₂ is the most dominant gas in the ISM (Tielens 2005), it is likely that the carbonaceous grains are hydrogenated. Other studies suggest that more CO₂ can form on hydrogenated grains at lower fluences.

The position of the ν₃ absorption band observed in this study (2275.2 ± 0.5 cm⁻¹) is in agreement with other measurements in ultra-high vacuum for ¹³CO₂ diluted in water ice (Malyk et al. 2007). An upward shift to 2281 cm⁻¹ holds for ¹³CO₂ diluted in ¹³CO ice and produced by radiolysis at 16 K (Loeffler et al. 2005). The astrophysically important case is a CO matrix with natural isotopic abundance. Previous isotopic studies have shown that the shift due to a matrix of lighter C isotopes is small, ∼3 cm⁻¹ (Falk 1987). A common interstellar environment is expected to be pure CO₂ ice, where ¹³CO₂ is naturally diluted in ¹²CO₂, with ν₃ peaking at 2281–2283 cm⁻¹ (Falk 1987), depending on the phase, amorphous or crystalline. These three environments (pure CO₂, CO₂ diluted in CO, or water) are sufficient to explain the observations of absorption band positions of condensed ¹³CO₂ in different molecular clouds (2275–2283 cm⁻¹), without the need to invoke thermal processing or the presence of methanol or other species (Boogert et al. 2000). As an example, in Figure 1 we compare the ν₃ absorption observed in the star-forming regions S 140: IRS1 and W 33A with that produced from radiolysis of carbon–water ice in our experiments. The peak position of the ¹³CO₂ feature toward S 140:IRS1 is shifted compared to our laboratory spectrum, indicating that this molecule is not in water but possibly pure or in CO. In contrast, the peak of the ¹³CO₂ feature toward W 33A is at 2277 cm⁻¹, close to the laboratory value of CO₂ diluted in water ice. We do not discuss line widths since they result from a convolution of intrinsic widths with a distribution of peak positions due to a population of environments along each line of sight.

We now consider the implications of our results on the observations of carbon dioxide on the icy moons of the outer solar system. We do not compare CO₂: H₂O ratios since they depend on the thickness of the water coverage and on the inhomogeneous composition of satellite surfaces. Since there are no reports of ¹³CO₂ in icy satellites, we limit our discussions to ¹²CO₂. The observed CO₂ band position ranges from 2347 cm⁻¹ (Phoebe and Iapetus; Buratti et al. 2005; Clark et al. 2005) to 2354.6 cm⁻¹ (Hyperion; Cruikshank et al. 2010). The CO₂ on Galilean satellites such as Callisto and Ganymede peaks at 2348.5–2349 cm⁻¹ (Hibbitts et al. 2000, 2003). These frequencies are substantially higher than the range seen in interstellar ices, 2342–2346.2 cm⁻¹ (Gerakines et al. 1999) and to the laboratory values for pure CO₂, 2343.3 cm⁻¹ (Baratta & Palumbo 1998; Falk 1987) as well as for CO₂ mixed in water, 2340 cm⁻¹ (Sandford & Allamandola 1990). Other studies show that the frequency of the ν₃ band of CO₂ adsorbed on metal oxide surfaces shifts upward. For instance, the peak position is at 2367 cm⁻¹ for CO₂ adsorbed on Al₂O₃ (Ramis et al. 1991) and 2351–2363 cm⁻¹ on porous inorganic zeolites (Montanari & Busca 2008). The CO₂ observed on the icy moons could be produced by the mechanism presented in this work and be readsorbed on mineral surfaces after diffusion or desorption followed by gravitational return. This absorbed state would have the observed band position.

The radiolytic interface process discussed here might also apply to synthesis of other molecules in extraterrestrial environments via combination of atoms of various solids embedded in ice.

We acknowledge financial support from NASA Cosmochemistry.