Secondary Ion–Ice Chemistry in UV-Irradiated Oxygen Ice: New Astrochemical Models Including Bulk Ionic Species Authors: Kristen Darnell, Daniel Lopez-Sanders, Deaton Warren, River-Allen Carroll, Emma Stanley, Liton Majumdar, Paola Caselli, Christopher R. Arumainayagam, Christopher N. Shingledecker Published: 2026 ABSTRACT Although energetic and nonenergetic processing of interstellar dust grain surfaces and ice is thought to be one of the dominant mechanisms for the extraterrestrial synthesis of prebiotic molecules, the role that “ion–ice” chemistry (bulk ion–neutral and ion–ion reactions within the dust-grain ice mantle) plays is still a largely unexplored topic. In this work, we present the results of our study to investigate the astrochemical importance of ions produced within dust grain ice mantles via irradiation by VUV photons, though the methods are applicable to other kinds of ionizing radiation. To this end, we present new models of the chemical evolution of a pure oxygen ice irradiated by UV photons, using a chemical network that explicitly includes bulk reactions involving secondary ions. A comparison of our calculations with previous data suggests that ion–neutral and ion–ion reactions may play a critical role in understanding the chemistry of interstellar ice. Introduction It has now been more than half a century since the publication of the pioneering work of Herbst and Klemperer, which arguably marked the genesis of the field of astrochemical modeling. A number of key predictions from that work have since been verified. Among these was the claim that gas-phase ion-neutral reactions, beginning with those involving H3 +, were essential to the formation of polyatomic molecules in interstellar environments. These reactions are often both barrierless and exothermic, making them efficient even in dense molecular clouds, which are typically around 10 K (see the review by Ceccarelli et al. and references therein for more details). In addition, especially for cases where the neutral reactant is polar, the long-range forces involved mean that such reactions can have a negative temperature dependence and rate coefficients significantly larger than the ∼10–9 cm3 s–1 predicted by the Langevin equation. Just as ions are critical for the formation of polyatomic molecules in the gas phase, so too are interstellar dust grains. In interstellar environments, dust acts as a quasi-catalyst for chemical reactions. In diffuse and translucent molecular clouds, the surfaces of the bare grains act as important reaction sites, especially for the formation of H2. − So too are the ice mantles which form on dust grains in dense molecular clouds. Our understanding of the importance of dust-grain ice mantles in explaining the observed chemical complexity of interstellar environments has grown over the history of astrochemistry. Grain-related processes are now thought to be critical in explaining the presence of terrestrial-like saturated or semisaturated organic species, including “prebiotic” molecules which may be implicated in the origins of life on Earth and possibly elsewhere. Nevertheless, the addition of surface and bulk ice-mantle grain chemistry, though doubtless a more accurate representation of chemistry in the real interstellar medium, remains an active area of research. The unfortunate combination of the importance of grain-related processes in the interstellar medium (ISM), combined with a comparative lack of knowledge and data regarding relevant processes has led one astrochemical modeler to make the now well-known jest that interstellar grain chemistry represents the “last refuge of the scoundrel.” Here, we will highlight just a few of the current areas of uncertainty that pose challenges for ice modeling, generally. First, there exist substantial uncertainties regarding the physisorption binding energies of most of the hundreds of species usually included in grain-chemical networks. − These binding energies, which have a range of values depending on many factors, one example being the local surface morphology, are typically approximated with a single value, with some very recent exceptions. , Moreover, though it is now well established both via experiments and theory that grain and ice-mantle surface chemistry is dominated by diffusive Langmuir–Hinshelwood type reactions, − whether and how chemistry within the ice mantle occurs at low temperatures has been the subject of considerable recent work. There is a growing consensus that nondiffusive processes almost certainly play a key role. ,− Finally, and most pertinent for this work, the chemical networks used for simulating interstellar grain chemistry have historically included only neutral species. That ion-ice processes can occur efficiently under interstellar conditions was suggested in the pioneering theoretical study of Woon, who found that cations such as HCO+ and CH3 + were able to react barrierlessly with water molecules in ice via pathways that could lead to formic acid and methanol, two astrochemically relevant species (see also Woon). The reaction between low-energy CH3 + cations and water ice was investigated in detail by Nakai et al., who confirmed experimentally that the reaction suggested by Woon does indeed occur under interstellar conditions. In that work, astrochemical models using the code of Furuya et al. found that the CH3 +-ice reaction was competitive at the edges of the simulated molecular cloud where the visual extinction was low, and calculated enhanced abundances of gas-phase CH3 +. The adsorption of gas-phase cations undoubtedly also occurs in real interstellar environments. The interaction between gas-phase cations and ice surfaces was the subject of a recent modeling study by Cui and Herbst, who found that such reactions could be efficient in forming complex species. However, there exist other mechanisms for driving ion-ice chemistry. While externally produced energetic photons are quickly extinguished in molecular clouds, galactic cosmic rays reach even the centers of most dense molecular clouds, and are an important source of continuous ice processing in the ISM. , The essential role of cosmic rays to astrochemistry has been recognized from the outset of astrochemical modeling in the first model of Herbst and Klemperer. They noted that cosmic rays drive the formation of H3 + via collisions with H2, which also generate an internal flux of UV photons. These photons can further ionize gaseous species in molecular clouds. Direct cosmic ray collisions with dust grains and their ice mantles similarly result in the formation of charged species within a roughly cylindrical region known as the “track” which forms around the path of the cosmic ray through the solid. Bombardment by internal UV photons represents yet another means of ionizing, albeit with a lower efficiency than with cosmic rays. Although some fraction of the charged species produced within ices undoubtedly undergo rapid neutralization by recombination, , the remainder may participate in an internal ion-ice chemistry whose importance may be comparable to that known to occur in the gas-phase. The ionization of an ice-mantle species creates, in addition to the atomic or molecular ion, a secondary electron, which can subsequently multiply the chemical effects. When such secondary electrons lose sufficient energy through collisions with bulk species, they can become “low-energy” electrons (<20 eV), which can have significant effects of their own. ,− In this work, we present the first computational study of secondary ion-ice chemistry driven by the UV photon (<11.3 eV) bombardment of a cosmic ice analogue of pure molecular oxygen, following the experimental work of Gerakines et al. Though galactic cosmic rays are indeed the ultimate drivers of ion chemistry (both gas and solid phase) in dense molecular clouds, they generate an internal flux of VUV photons which trigger broadly similar reactions as the cosmic rays themselves. It is these VUV photons which we simulate in this work, following the laboratory conditions of Gerakines et al. The key distinction between the present work and all our previous implementations of this modeling approach , is our relaxation of the requirement that ionization is always immediately followed by charge recombination, such that in previous studies, ionic species were never present in the simulated ice. Here, for the first time, we allow ionization to persist, yielding long-lived ionic species that participate in a network of ion–neutral, ion–ion, and dissociative recombination reactions within the bulk ice. Thus, the three primary innovations of this work are (i) the first astrochemical kinetic model to include cations, anions, and free electrons as persistent species within the bulk ice mantle, allowing them to undergo nondiffusive ion–neutral, ion–ion, and dissociative recombination reactions in the solid phase. (ii) two new kinetic parameters to define the rate coefficients of the additional ion-neutral and ion–ion reactions, extending the nondiffusive rate formalism of Shingledecker et al. to charged species; and (iii) the first computational study of secondary ion chemistry in a UV-irradiated cosmic ice analogue. We note, regarding point (i), that this is distinct from recent models treating Eley–Rideal reactions of gas-phase ions with ice surfaces, which do not include internally generated bulk ionic species. Our paper is laid out as follows: In Section we describe our theoretical and computational approach, in Section we present the results of our trials and describe their astrochemical significance, and finally, we summarize our findings in Section . Theory Formation of Ions Within Cosmic Ice Here, we are focused on the chemical contribution of ions produced within cosmic ice. In dense molecular clouds, as noted in the introduction, there exist two main mechanisms which could drive the ionization of dust-grain ice mantle species; namely, galactic cosmic rays , and the UV photons they generate through the excitation of H2. The basis of the theoretical approach used in our models is that of Shingledecker and Herbst, which begins with the assumption of a set of four possible outcomes when some ice constituent, A, encounters some particle of ionizing radiation R1 aA⇝aA++e− R2 aA⇝aA++e−→aA*→bB*+cC* R3 aA⇝aA*→bB+cC R4 aA⇝aA* where the lower case letters represent the relevant stoichiometric coefficients. Processes () and () correspond to the ionization of A, while () and () represent the effects of excitation to a higher vibrational or electronic state. Shingledecker and Herbst furthermore derive expressions for the rate coefficients ()–(), albeit for cases where A encounters a particle such as an energetic proton. This formalism was later expanded by Shingledecker et al. to treat irradiation by photons. In large part because grain-chemical networks have only included neutral species, previous modeling works using one of the above approaches have made the assumption that in every case, the ionization of A proceeds via () rather than (), i.e., that ionization is followed by rapid charge recombination to yield some excited suprathermal products. In contrast, here we assume as a first-order approximation that the two ionization processes, () and (), occur with equal probability. Briefly, to obtain rate coefficients (k photo) for ()–(), we begin with the standard formula for photoprocesses 1 kphoto=∫σ(λ)I(λ)dλ where the resulting value is a function of the wavelength-dependent absorption cross sections (σ­(λ)) and intensities (I(λ)) integrated over all wavelengths. To obtain the ultimately utilized expression for the rate coefficients, we express eq as 2 kphoto=∫σ(λ)I(λ)dλ∫I(λ)dλ∫I(λ)dλ=σ̅Φ such that the rate coefficient can be written as the product of an average cross section for a given species and the photon flux integrated over all wavelengths. As described in Mullikin et al., the individual k photo values for ()–() are then multiplied by some branching fractions, f br. For a given species, we divide the branching fractions such that all ionization processes for that species add to unity (Type R1 and R2 in Table ), and all excitation processes (R3 and R4) for a given species likewise add to unity. 1Photoprocesses Included in Our Modelnumbertypeprocess f br σ̅ (cm2)P1R1 O⇝O++e− 0.500.00P2R1 O2⇝O++O− 0.253.86 × 10–20 P3R1 O2⇝O2++e− 0.253.86 × 10–20 P4R1 O3⇝O++O2− 0.1670.00P5R1 O3⇝O2++O− 0.1670.00P6R1 O3⇝O3++e− 0.1670.00P7R2 O⇝O* 0.500.00P8R2 O2⇝O*+O* 0.503.86 × 10P9R2 O3⇝O2*+O* 0.500.00P10R3 O2⇝O+O 0.502.13 × 10–18 P11R3 O3⇝O2+O 0.505.60 × 10–18 P12R4 O⇝O* 1.000.00P13R4 O2⇝O2* 0.502.13 × 10–18 P14R4 O3⇝O3* 0.505.60 × 10–18 Photoprocesses Included in Our Model Mullikin et al. also introduces the factor δ, which accounts for any discrepancy due to unavailable data, such as the difference between the actual solid phase cross sections compared with the gas-phase values used here. Thus, the rate coefficients for photodriven processes in our model are given by 3 kphoto=fbrσ̅Φδ Shown in Table are the cross sections and branching fractions for the various processes used here, taken from Mullikin et al. Each separate process in that table is given a label prefixed with “P” while the generic type of process using the four pathways given above and described in Shingledecker et al. are denoted R1 to R4. For ionization processes, the total branching fraction is split between pathway types R1 and R2, where applicable. Likewise for excitation, branching fractions are split between R3 and R4. Though gas-phase molecular oxygen has an ionization energy of ∼12 eV, slightly higher than the UV photon energies used by Gerakines et al., these values are typically lowered by a few eV in solids, leading to a nonzero cross section for processes P2 and P3. We note that processes P7 and P12 share the same net stoichiometry (O⇝O*) but represent fundamentally distinct physical mechanisms with different rate expressions. P7 is an R2-type process (ionization of O followed immediately by charge recombination), whose rate is proportional to the ionization cross section of O; since the ionization potential of atomic oxygen (13.6 eV) exceeds the maximum photon energy employed here, this cross section is effectively zero. P12 is an R4-type process (direct electronic excitation of O without ionization), which in principle carries a nonzero excitation cross section. The identical net reaction therefore reflects two distinct physical channels, not a duplication. The Chemistry of Solid-Phase Ions Thus formed within the ice as described in Section , the ions in our simulations are allowed to react chemically. Whether, or by which mechanism, reactions occur in low-temperature cosmic ice is still a matter of ongoing investigation. Here we assume, as in Shingledecker et al., that rapid reactions can and do occur between neighbors even at temperatures as low as 5 K, in other words, nondiffusively. This approach was motivated by a now large body of experimental studies showing persuasively that even complex organic molecules (COMs) can be formed via fast solid-phase nondiffusive reactions ,,− and, when incorporated into simulations of astrophysical environments, yield predicted abundances of ice-mantle species that have proven to be in general agreement with recent JWST observations. − As described in Shingledecker et al., the fast nondiffusive bulk rate coefficient for some reaction A + B is given by 4 kfast=fbr[ν0A+ν0Bnbulk]exp(−EactABTice) where, ν0 A and ν0 B are the vibrational frequencies of A and B, n bulk is the number of species in the ice mantle, exclusive of the surface layers, E act is the activation energy in units of Kelvin, if present, and T ice is the ice temperature. If one of the vibrational frequencies is much larger than the other, for example where ν0 A ≫ ν0 B, then eq can be simplified to 5 kfast≈fbr[νn−nnbulk]exp(−EactABTice) where νn–n = ν0 A. Hereafter, we will refer to νn–n as the neutral–neutral trial frequency. As originally described in Shingledecker et al., the species undergoing fast solid-phase reactions were all neutrals. For the purposes of this work, we have modified eq slightly to also treat the solid-phase reaction of ions. First, for reactions between ions and neutrals of the forms A+ + B and A + B–, we replace νn–n with another frequency νi–n, the ion-neutral trial frequency. Finally, for reactions between two charged species of the form A+ + B–, we invoke a third frequency, νi–i, the ion–ion trial frequency. Here, the anions do not form directly from any of the processes in Table , but rather, result from, e.g., electron attachment reactions. The three frequencies in eq represent the different long-range forces experienced between the reactants and account for the potentially higher overall reaction rate between a cation and an anion relative to a reaction involving two neutral species, even in an environment where diffusion is severely inhibited. As we describe in Section , these three frequencies were varied in our models to ascertain their influence on the resulting predicted abundances. As noted by Theulé et al., the order of magnitude of the pre-exponential frequencies can vary widely, depending on the underlying mechanism and phase where reactions are occurring, e.g., gas or liquid. The same work further notes an enormous possible range of values for first-order processes between 1013 s–1 in the gas phase to as low as ∼10–3 s–1 in other cases, with those in condensed media being lower. The underlying causes for such variation are not, as far as we know, currently understood but could be related to effects such as the entropy of activation. Physically, such variation could reflect the strong configurational constraints imposed by the condensed-phase environment. In the language of transition-state-theory, this could manifest as an effective entropy-of-activation effect, though we do not regard that explanation as unique here. Model & Chemical Network Reaction Network The reaction network in this work was limited to oxygen species only, since it is arguably the simplest radiation-chemical system, and for which there exist relevant experimental data in the work of Gerakines et al. We note that a number of additional experimental studies of photolysis and radiolysis of pure oxygen ice are relevant to the broader context of this network. In particular, Zhen and Linnartz report UV-photodesorption and photochemistry cross sections for O2 ice, and Sivaraman et al. examine the temperature-dependent formation of ozone under 5 keV electron irradiation; the ozone yields reported in both works are qualitatively consistent with the photochemical pathways included in our model, and both works are in broad agreement with our findings, described below, regarding the dominant chemical pathways to ozone formation. The starting point for our network was the oxygen network in Mullikin et al., shown in Table , which in turn was based on previous work by Shingledecker et al. and Shingledecker et al. Our base network, which included only reactions between neutral species, was expanded with the reactions shown in Table to include charge recombination reactions of the form 6 A++B−→C+D using the reaction networks in Pastina and LaVerne, Anicich, and the NIST Chemical WebBook. Beyond extending ion-neutral reactions beyond the gas-phase and into our grain-chemical network, we have further added two more classes of reactions involving two species with unlike charges; namely, ion–ion recombination of the form 7 A++B−→C*+D* and dissociative recombination between cations and electrons of the form 8 A++e−→C*+D* 2Base Chemical Network Used to Model Solid-Phase O2 Photo-Processing and Subsequent Chemistry, Taken from Mullikin et al. and Shingledecker et al. non-photon-induced reactions O3*+O3→O2+O2+O2 O2*+O→O3* O*+O3*→O2+O2 O+O→O2* O2*+O→O2+O O2*+O2*→O2+O2 O*+O→O+O O2*+O2→O2+O2 O2*+O3*→O2+O2+O O*+O→O2* O2*+O3→O2+O2+O O3*+O3*→O2+O2+O2 O*+O2→O3* O*+O*→O+O O+O3*→O2+O2 O*+O2→O+O2 O*+O2*→O3* O2+O3*→O2+O2+O O+O3→O2+O2 O*+O2*→O+O2 O+O2→O3* O*+O3→O2+O2 Base Chemical Network Used to Model Solid-Phase O2 Photo-Processing and Subsequent Chemistry, Taken from Mullikin et al. and Shingledecker et al. 3New Reactions Involving Charged Species Added for This Worknumberreactionbranching fractionN1 O+e−→O− 1.00N2 O+O−→O2− 1.00N3 O+O2−→O3− 1.00N4 O++e−→O* 1.00N5 O++O−→O2* 0.50N6 O++O−→O*+O* 0.50N7 O++O2→O2++O 1.00N8 O++O2−→O3* 0.50N9 O++O2−→O*+O2* 0.50N10 O++O3−→O*+O3* 0.50N11 O++O3−→O2*+O2* 0.50N12 O2+e−→O2− 1.00N13 O2+O−→O3− 1.00N14 O2++e−→O2* 0.50N15 O2++e−→O*+O* 0.50N16 O2++O−→O3* 0.50N17 O2++O−→O*+O2* 0.50N18 O2++O2→O3++O 1.00N19 O2++O2−→O*+O3* 0.50N20 O2++O2−→O2*+O2* 0.50N21 O2++O3−→O2*+O3* 0.33N22 O2++O3−→O2*+O2*+O* 0.33N23 O2++O3−→O3*+O*+O* 0.33N24 O3+e−→O3− 1.00N25 O3+O−→O2−+O2 1.00N26 O3+O2−→O3−+O2 1.00N27 O3+O3+→O2+O2+O2+ 1.00N28 O3−+O−→O2−+O2− 1.00N29 O3++e−→O3* 0.50N30 O3++e−→O2*+O* 0.50N31 O3++O−→O*+O3* 0.50N32 O3++O−→O2*+O2* 0.50N33 O3++O2→O2++O3 0.415N34 O3++O2→O3++O2 0.085N35 O3++O2→O2+O2+O+ 0.50N36 O3++O2−→O2*+O3* 0.33N37 O3++O2−→O2*+O2*+O* 0.33N38 O3++O2−→O3*+O*+O* 0.33N39 O3++O3−→O3*+O3* 0.33N40 O3++O3−→O3*+O2*+O* 0.33N41 O3++O3−→O2*+O2*+O2* 0.33 New Reactions Involving Charged Species Added for This Work In both of the two cases above, we assume that the products are created in an excited suprathermal state, given the high exothermicity of the process. All new reactions involving ions were assumed to proceed barrierlessly. These ion-induced reactions enter the standard rate-equation formalism as additional formation and destruction terms, with rate coefficients given by eq using the ionic trial frequencies ν i–n or ν i – i as appropriate, and with the assumption of E act = 0 for all charged-species reactions. UV photolysis is included explicitly via the photoprocess rate coefficients in Table , which compete with all chemical reactions at each time step. Thermal diffusion is deliberately excluded: our nondiffusive framework (Section ) is motivated by experimental evidence that bulk ice reactions can occur even at temperatures where thermal diffusion is negligible. Grain surface reactions are included, but should be viewed with somewhat more skepticism, since our comparison here is with data relevant for bulk chemistry. We also note that the present model assumes spatially uniform (homogeneous) chemical processing throughout the ice mantle, which is a fundamental limitation of standard rate-equation astrochemical models, though the region over which abundances are averaged can be subdivided as in the multilayer model of Taquet et al. and Vasyunin et al., yielding greater physical realism in terms of ice composition. In the context of UV-photon irradiation, where energy is deposited relatively uniformly over the penetration depth of thin ice films, this assumption is better justified than it would be for MeV heavy-ion bombardment, where spatial inhomogeneity along ion tracks can be significant. Sputtering, compaction, and structural modifications of the ice are not modeled in this work, though those processes are natural future extensions of this work. Model Parameters In order to examine the role of condensed-phase ion–molecule reactions under well-constrained conditions, we have simulated previous laboratory experiments directly, rather than an interstellar region. Specifically, we have sought to reproduce the experiments reported in Gerakines et al., which studied the effects of irradiating a pure molecular oxygen ice deposited on a MgF2 substrate at a temperature of 10 K with UV photons (<11.3 eV) from a microwave-discharge hydrogen lamp (MDHL) with a photon flux of 2.2 × 1014 photons cm–2 s–1 as reported in Gerakines et al. Though this is a direct simulation of an experiment, our method could be easily extended to real interstellar sources via, e.g., adjusting the flux to an appropriate ISM value. Here, a suite of simulations were carried out to simulate this ice irradiation using a variety of kinetic parameters, with the goal of understanding the dependence of the underlying chemistry on these values. A full list of the models run, and the values of the varied parameters, is given in Table . 4Model Sets Described Here, and Their Associated Parametersmodelνn–n (s–1)νi–n (s–1)νi–i (s–1)15.00 × 108 3.01 × 107 6.93 × 1011 21.00 × 1013 1.00 × 1013 1.00 × 1013 31.00 × 1013 1.00 × 1013 1.00 × 1010 41.00 × 1013 1.00 × 1013 1.00 × 108 51.00 × 1013 1.00 × 1010 1.00 × 1013 61.00 × 1013 1.00 × 1010 1.00 × 1010 71.00 × 1013 1.00 × 1010 1.00 × 108 81.00 × 1013 1.00 × 108 1.00 × 1013 91.00 × 1013 1.00 × 108 1.00 × 1010 101.00 × 1013 1.00 × 108 1.00 × 108 111.00 × 1010 1.00 × 1013 1.00 × 1013 121.00 × 1010 1.00 × 1013 1.00 × 1010 131.00 × 1010 1.00 × 1013 1.00 × 108 141.00 × 1010 1.00 × 1010 1.00 × 1013 151.00 × 1010 1.00 × 1010 1.00 × 1010 161.00 × 1010 1.00 × 1010 1.00 × 108 171.00 × 1010 1.00 × 108 1.00 × 1013 181.00 × 1010 1.00 × 108 1.00 × 1010 191.00 × 1010 1.00 × 108 1.00 × 108 201.00 × 108 1.00 × 1013 1.00 × 1013 211.00 × 108 1.00 × 1013 1.00 × 1010 221.00 × 108 1.00 × 1013 1.00 × 108 231.00 × 108 1.00 × 1010 1.00 × 1013 241.00 × 108 1.00 × 1010 1.00 × 1010 251.00 × 108 1.00 × 1010 1.00 × 108 261.00 × 108 1.00 × 108 1.00 × 1013 271.00 × 108 1.00 × 108 1.00 × 1010 281.00 × 108 1.00 × 108 1.00 × 108 Model Sets Described Here, and Their Associated Parameters For Model 1, we used a method for achieving optimal agreement between calculated and experimental data based on the approach used in Mullikin et al. The three aforementioned trial frequencies, νn–n, νi–n, and νi–i, as well as the δ factors for select reactions described below were allowed to vary. The overall agreement between a given model and the experimental data was characterized with the value, χ, given by 9 χ=∑i=1n(mi−ei)2n where e i and m i are the ith experimental and model data points out of a total of n. Regarding the variation of δ values, the results of a sensitivity analysis carried out by us found that calculated O3 abundances were sensitive to processes with a nonzero cross section, σ̅, excepting P10. Results and Discussion Best-Fit Model In Table , we show values for the δ parameters for each photo process that yielded the best fit to experimental data. Shown in Figure are the calculated abundances for O3 vs fluence for Model 1. Here, the fluence is defined as the product of the flux and the time. In Model 1, which resulted in χ = 0.484, the value of νi–i is set to a value of ∼7 × 1011 s–1, several orders of magnitude higher than the other frequency values of νn–n = 5 × 108 s–1 and νi–n = 3 × 107 s–1. This comparatively high value of ν i–i , coupled with a low value for νi–n, tends toward the limiting case where ions produced in the bulk recombine very quickly, corresponding to the assumption made in our prior work where solid-phase ions were not present in the bulk. ,, Previously, we assumed that ionization proceeded solely via process (). A comparison between the abundances of Model 1 here and the results described in Mullikin et al. shows broadly good agreement, as expected, though the exact parameters obtained differ somewhat due to the greatly expanded network. We note that Gerakines et al. reported a peak O3/O2 ratio of ∼36%, whereas the maximum ratio reached in Model 1 is ∼24%. However, as in Mullikin et al., we interpret this value as a percentage of the total initial oxygen atoms. Multiplying our final abundances by a factor of 1.5, to account for this, brings our results in exact agreement with the experimental value. 5Delta Value for Photoprocesses in Model 1numbertypeprocessδP1R1 O⇝O++e− 1.00P2R1 O2⇝O++O− 0.437P3R1 O2⇝O2++e− 1.54P4R1 O3⇝O++O2− 1.00P5R1 O3⇝O2++O− 1.00P6R1 O3⇝O3++e− 1.00P7R2 O⇝O* 1.00P8R2 O2⇝O*+O* 1.90P9R2 O3⇝O2*+O* 1.00P10R3 O2⇝O+O 1.00P11R3 O3⇝O2+O 22.1P12R4 O⇝O* 1.00P13R4 O2⇝O2* 99.0P14R4 O3⇝O3* 99.0 Delta Value for Photoprocesses in Model 1 Model 1 calculated abundances of O3 in best-fit model compared with experimental data. Other Models As shown in Table , in addition to Model 1, our best-fit model, we also ran a number of other simulations to investigate the effects of key parameters on the resulting agreement with experimental data. The resulting ozone abundances are shown in Figure . In Models 2–28, all δ values were set to unity, while νn–n, νi–n, and νi–i were tested using all possible combinations of three different frequency values: low (1.00 × 108 s–1), medium (1.00 × 1010 s–1), and high (1.00 × 1013 s–1). These models highlight the importance of the δ values in obtaining a good fit to the experimental data, and all of these test simulations resulted in values of χ > 12 denoting poor fit. In general, we found that models with higher relative values for νi–n gave the worst agreement, in line with the findings of our best-fit model. We also consistently found that, with our expanded chemical network, varying the values of νi–n and νi–i affected the calculated abundance of ozone much more than variations of νn–n. This finding suggests that, though the dominant formation pathways of many of the species described below may be ultimately governed by neutral–neutral reactions, the included ionic chemistry had a significant effect on the abundances of those neutral species in our models. Models 2–28 (solid lines) compared with experimental data (points). In the following sections, the chemistry of neutrals and ions is analyzed. We focus on the results of the best-fit Model 1. Neutral Species In this section, we detail the chemistry of the neutral species in our chemical network (O, O2, and O3), as well as of their suprathermal counterparts. Each of their major formation and destruction reactions is shown in Figure . We note that Table describes the meaning of the different line styles in the reaction contribution figures presented here. Relative importance of formation and destruction reactions for neutral species and their suprathermal counterparts in Model 1. Here, panel (a) shows O3, panel (b) shows O3*, panel (c) shows O, panel (d) shows O*, panel (e) shows O2, and panel (f) shows O2*. See Table for an explanation of the different linestyles. 6Line Styles in Reaction Contribution Plots Line Styles in Reaction Contribution Plots O3 and O3 * As shown in Figure a, the chemistry of ozone largely follows what was found previously in Mullikin et al. Here, as there, the dominant formation route for ozone is the two-step process 10 O+O2→O3*⇝O3 where suprathermal O3* is first formed from the reaction of ground state atomic and molecular oxygen, or between suprathermal oxygen and ground state molecular oxygen. In both cases, suprathermal ozone is destroyed via quenching by the bulk ice (see Figure b). Conversely, at intermediate fluences, the underlying chemistry of O3 in the model changes dramatically, highlighting the potentially critical role of ions for understanding the chemistry of irradiated low-temperature solids, with the dominant formation pathway (∼70%) becoming N33 O3++O2→O2++O3 In Model 1, ozone is destroyed mainly by either photolysis or reaction with O3 +, again highlighting the potential importance of ions in bulk chemistry. O and O* As can be seen in Figure c, the chemistry of atomic oxygen largely follows a similar pattern as O3. At early fluences, neutral chemistry dominates, with the main formation route being P10 O2⇝2O and the major destruction pathways being the reformation of molecular oxygen 11 O+O→O2 or reaction with O2 to form suprathermal ozone. However, as with O3, ion-neutral reactions become more important at later fluences, as depicted in Figure c. After a fluence of 1011 photons cm–2 is reached, the reactions N18 O2++O2→O3++O and N7 O++O2→O2++O become more significant, representing roughly 40% and 20% of the O formation, respectively. For the case of suprathermal oxygen, depicted in Figure d, the situation is quite simple. At all fluences, it is mainly formed via P8 O2⇝2O* Similarly, the major destruction pathway throughout the model is quenching by the ice. O2 and O2* For O2 reactions, represented in Figure e, the dominant formation route at early fluences (before ∼ 1012 cm–2) comes simply from the quenching of suprathermal O2. After that time, the bimolecular reaction between ground-state oxygen atoms becomes more important. Destruction is initially dominated by electronic excitation leading to suprathermal molecular oxygen. After ∼ 1012 cm–2, ion-neutral destruction routes become more important, including the reaction of molecular oxygen with O2 + (N18), as well as the reactions O3++O2→O2++O3(N33) and N35 O3++O2→2O2+O+ For suprathermal molecular oxygen, shown in Figure f, at all model times the dominant formation route is via the excitation process P13. Similar to the other suprathermal species, the dominant destruction pathway is via quenching. Ions Figure shows the calculated abundances of the cations and anions in our simulation as a function of fluence. There, one can see that the combined peak abundance of all charged species remains very low, with a value of ∼4.6 × 10–7% at its peak. Here, we describe the main formation and destruction pathways of key ionic species. Total abundance of solid-phase ions vs fluence. Before a fluence of ∼1011 cm–2, the abundance of ions is quite low. Afterward, however, there is a noticeable increase in their abundance, with O2 – being initially the most abundant. As depicted in Figure a, this anion is formed at all model times mainly via the associative attachment of an electron N12 O2+e−→O2− and is destroyed via a number of reactions, with the most important after ∼1013 cm–2 being the charge transfer reaction N26 O3+O2−→O2+O3− Relative importance of formation and destruction reactions for (a) O2 – and (b) O3 – in Model 1. See Table for an explanation of the different linestyles. After a fluence of ∼1014 cm–2, O3 – becomes the dominant charged species in Model 1. As shown in Figure b, it is initially formed mainly via N13 O2+O−→O3− with the reaction O3+O2−→O3−+O2(N26) becoming important at later fluences. The destruction of O3 – occurs entirely through ion-recombination reactions. We note that the plots of the eight ion-recombination reactions obscure each other in Figure b. Electrons Finally, the addition of electrons in our network is another novel feature and a byproduct of the ionizing radiation that real astrophysical ices are continuously exposed to. As shown in Figure , electron chemistry is dominated by the ionization and subsequent recombination with the dominant ice constituents. Specifically, they are produced overwhelmingly from the photoionization of O2 and destroyed via recombination with O2 and O3. We note that in these preliminary simulations, we have not included processes which are critical for very low energy (<5 eV) electrons, such as dissociative electron attachment (DEA), though such processes undoubtedly occur in real irradiated ices. Future simulations will examine the effects of low-energy electrons in more detail. Major reactions with electrons. The Impact of Ions in our Models To directly assess the role of ionic channels in Model 1, we also carried out a control model in which only the photoprocess channels yielding charged products were suppressed, while all neutral channels, trial frequencies, and other Model 1 parameters were held fixed. In this control, agreement with the Gerakines et al. O3 data deteriorates markedly: the unweighted root-mean-square deviation χ increases from 0.484 (Model 1) to 13.93 in this control case, with the peak modeled O3 abundance falling from 24% to 3% of the initial O2. Calculated abundances for Model 1 and this control simulation are compared directly in Figure . Thus, while it remains the case that neutral–neutral reactions are the dominant formation routes for ozone in our new network, the overall impact of adding ionic reactions here is nontrivial on the overall chemistry, at least for the ion–optimized parameter set of Model 1. Calculated abundances for Model 1 (blue curve) and a control model using Model 1 input parameters, but with ion–forming photoprocesses disabled (red–dashed curve). Conclusions In this work, we describe novel astrochemical simulations which seek to elucidate the effects of solid-phase ions on the chemistry of low-temperature solids, specifically, within interstellar dust grain ice mantles. Expanding on our previous models described in Shingledecker et al. and Mullikin et al., we have used the method of Shingledecker and Herbst to model the production of ionic species formed from the photoionization and photoexcitation of atoms and molecules within the ice. To examine the effects of such ions on solid-phase chemistry under well-constrained conditions, we carried out simulations of pure O2 ice at 10 K irradiated by UV photons, following the work of Gerakines et al., who tracked the production of ozone under those conditions. The models described here show that inclusion of ionic species and related reactions is possible in three-phase, slightly modified astrochemical codes, with the caveat that such additions do substantially increase the size, complexity, and uncertainty of the chemical networks. Our results further indicate that ionic pathways can be kinetically important for reproducing the observed O3 growth in Model 1. Though neutral channels generally dominate O3-forming steps, particularly at early and late fluence, ion-neutral and ion–ion processes become more important during midfluence chemistry. The stronger sensitivity of the calculated abundances to ν i–n and ν i–i than to ν n–n further suggests that, in this network, ionic reactions had a nontrivial role in regulating the neutral species abundances. Future studies using this method could fruitfully be applied to studying systems where solid-phase ion-neutral and ion–ion reactions are likely to play a larger role, such as those involving the formation of salts or the kinds of charged species which have been detected in cometary material and real interstellar dust-grain ice mantles. , Overall, though, the results of our calculations are in line with other recent studies, notably by Woon and Woon who studied bulk-ice chemistry, and both Nakai et al. and Cui and Herbst, who studied surface processes, in showing that, as is the case with gas-phase processes, the reactions of charged species on interstellar dust grains likely represent an important class of chemical processes that will need to be examined in more detail in the future. Of particular utility for future modeling studies would be experimental data on the fluence-dependent abundance of multiple components of an ice-analogue, as well as information about the total concentration and makeup of ions in situ in the ice.