In this paper, we explain the HNO3 observations made with the MIPAS/Envisat instrument in the northern polar region at the time of the October-November 2003 solar proton event. Increases of 0.5–5.5 ppbv are seen at altitudes 35–58 km after the onset of the event. Results from the Sodankylä Ion and Neutral Chemistry model are in good agreement with the MIPAS observations, except at around 58 km where the modeled mixing ratios are about a factor of two larger. According to the model results, HNO3 production at altitudes above 35 km is almost entirely due to ion-ion recombination between NO3− and H+ cluster ions. At 35 km and below, there is contribution also from N2O5 reacting with ions as well as from NO2 reacting with OH.
 In the middle atmosphere, large solar proton events (SPEs) produce significant amounts of odd hydrogen HOx (H + OH + HO2) and odd nitrogen NOx (N + NO + NO2). These species induce both short and long-term changes of ozone [e.g., Jackman et al., 2008, and references therein].
 HNO3 is a reservoir species for NOx and an important constituent in the middle atmosphere because of its role in polar stratospheric cloud formation, denitrification, and ozone depletion. Satellite observations have shown substantial amounts of HNO3 in the upper stratosphere of the polar regions during winter time [López-Puertas et al., 2005; Stiller et al., 2005; Orsolini et al., 2005, 2008]. These enhancements have been explained by particle precipitation taking place during polar night conditions when HNO3 loss by photodissociation is not effective. Possible mechanisms that could produce HNO3 during and after particle forcing include N2O5 conversion in reactions with positive ion clusters, recombination between positive and negative ions, and reactions involving either NO2 and OH or NO and HO2 [Böhringer et al., 1983; Kawa et al., 1995; Aikin, 1997; Butkovskaya et al., 2005].
 The October-November 2003 solar proton event, a.k.a. the Halloween event, is one of the largest SPEs of the last 50 years. The MIPAS/Envisat instrument observed two phases in the change of HNO3 [López-Puertas et al., 2005]. The first was a rapid increase during the proton forcing, followed by a recovery to initial values within a few days. Second, a slower, continuous build-up occurred after the event, lasting until the stratospheric warming event in late December 2003. The slow after-event increase in the upper stratosphere can be explained by NOx conversion to N2O5, which then reacts with cluster ions to form HNO3 [Stiller et al., 2005]. However, the large HNO3 amounts produced in late October during the SPE are significantly underestimated in 3-D atmospheric models [e.g., Jackman et al., 2008].
 In this paper, we use the Sodankylä Ion and Neutral Chemistry model to study the production of HNO3 during the Halloween event. The model results are compared to the observations from the MIPAS/Envisat instrument. We discuss the different production mechanisms of HNO3 and their importance, pointing out the connection to HOx production and ozone depletion.
2. MIPAS/Envisat Observations
 MIPAS (Michelson Interferometer for Passive Atmospheric Sounding), on board the European Space Agency's Envisat satellite, is a high-resolution mid-IR limb sounder that allows measurement of the kinetic temperature and a large number of atmospheric species, including HNO3, with good global coverage [Fischer et al., 2008]. The original MIPAS standard measurement mode covers tangent altitudes from 6 to 68 km. Due to the Sun-synchronous orbit of Envisat, observations are made at around 10 a.m. and 10 p.m. local time.
 The retrieval of HNO3 data analyzed here was performed with the IMK-IAA data processor [von Clarmann et al., 2003, and references therein]. IMK data version V3O_HNO3_9 was used here, which is based on ESA level 1b spectra. Altitude resolution of the data is ≈4 km in the stratosphere and ∼8 km in the mesosphere. The precision (all random errors) is between 4 and 8% between 25 and 45 km, increasing to 12% at 60 km, while the estimated accuracy, with spectroscopic uncertainty as main additional error source, is between 5 and 15%.
 The MIPAS HNO3 data retrieved by IMK/IAA have been extensively validated against MIPAS-balloon and MIPAS ESA-retrieved HNO3, and satellite measurements of Odin/SMR, ILAS-II and ACE-FTS [Wang et al., 2007]. In general they show a good consistency with the other data sets with mean differences of about ±0.5 ppbv and standard deviations of the differences varying from 0.5 to 1.5 ppbv.
 In the beginning of the Halloween event, MIPAS was shut down to protect its electronics from high-energy particles. This resulted in a data gap on Oct 27 and 28. However, observations were continued on Oct 28 before the strongest peak of the proton forcing occurred. Measurements made on Oct 26 provide a reference to which those observed after the onset of the SPE can be contrasted.
3. Ion-Neutral Chemistry Modeling
 The Sodankylä Ion and Neutral Chemistry model, a.k.a. SIC, is a 1-D tool for ionosphere–atmosphere interaction studies. The first version was developed in the late 1980s for ionospheric data interpretation [Turunen et al., 1996]. The latest version 6.10.0 solves the concentrations of 65 ions, of which 36 are positive and 29 negative, as well as 16 minor neutral species. Altitude range is from 20 to 150 km. A recent description of SIC is given by Verronen et al. . Below we briefly summarize some details of the modeling work.
 Ionization rates due to proton precipitation are calculated using GOES-11 satellite proton flux data, available from, e.g., the NOAA National Geophysical Data Center (www.ngdc.noaa.gov/stp/stp.html). For details of the rate calculation, see Verronen et al. [2005, and references therein].
 In the model, there are three potential HNO3 production mechanisms that are intensified by SPEs because of increases in the concentrations of the reactants, both ions and neutrals. 1) N2O5 conversion to HNO3 due to ionic reactions is included according to Böhringer et al. , but taking the rate coefficient of the reaction with all X(H2O)n ions to be 10−11 cm3 s−1. X is considered to be either H+ or NO3−, and n is the order of hydration with values ranging between 0 and 8. 2) Recombination between positive and negative ions is modeled using two rate coefficients for all reactions, 6.0 × 10−8 × (300/T)0.5 cm3 s−1 for two-body reactions and 1.25 × 10−25 × (300/T)4 cm6 s−1 for three-body reactions [Arijs et al., 1987]. HNO3 production is due to recombination between H+(H2O)n and NO3−(Y)n ions, where Y is either HNO3 or H2O. 3) The three-body reaction NO2 + OH + M has a rate coefficient of 2.0 × 10−30 × (300/T)3.2 cm6 s−1 [Sander et al., 2003]. Currently, our model does not include the reaction NO + HO2, which may contribute in the order of 0.5 ppbv additional HNO3 in the middle to upper stratosphere [Brühl et al., 2007].
 Two locations were chosen for the modeling: 70°N/0°E and 85°N/90°W. This allowed us to contrast the effects of different illumination conditions on the results. Before modeling the SPE effects, SIC was initialized to conditions of late October. MIPAS total number density, water vapour, and temperature, averaged over the SPE period, were used in all modeling. The SPE modeling was executed between Oct 26, 00:00 UT, and Nov 07, 00:00 UT. In the middle and lower stratosphere the model clearly underestimated the initial concentrations of HNO3, N2O5, and NO2, so the starting values for the SPE modeling, on Oct 26, were set using MIPAS data. After SPE modeling, a MIPAS averaging kernel appropriate for the polar region was applied to the results in order to make them comparable to the coarser altitude resolution of the observations.
Figure 1 shows calculated ionization rates for the Halloween event at 35, 46, and 58 km altitude. The proton forcing is strongest between Oct 28 12:00 UT and Oct 29 15:00 UT. Therefore, it can be expected that most of the rapid production of HNO3 occurs during that time. The secondary peaks of ionization around Oct 30 00:00 UT and Nov 03 00:00 UT are relatively more pronounced at 35 km. The ionization before Oct 28 is not visible with the rate scale used (not shown).
Figure 2 presents model results and observations of HNO3 at 35, 46, and 58 km. MIPAS shows increases between about 0.5 and 5.5 ppbv, depending on the latitude and altitude. The largest increase due to the SPE is seen at 46 km/85°N, the smallest at 35 km/70°N. Most of the enhancements occur on Oct 28–29 during the intense proton forcing period. Model predictions of HNO3 agree with the observations in many cases, in magnitude of enhancement as well as in temporal behavior. However, at 35 km/85°N and 58 km/85°N the model overestimates HNO3 production by ≈30% and ≈100%, respectively. At 58 km/85°N, 46 km/85°N, and 35 km/70°N a significant decrease is observed after about Nov 01 but not captured by the model. A similar decrease occurs also at 35 km/85°N a few days later. These features are discussed later in Section 5. Nevertheless, it is clear that the model chemistry can produce the high mixing ratios of HNO3 observed by MIPAS.
 At 46 km, the observed behavior of HNO3 at 85°N is quite different from that at 70°N. At 85°N the produced HNO3 generally persists with no significant loss for several days while at 70°N a decrease is observed on each day after Oct 29. A similar latitude-dependent difference in behavior is also seen at 58 km. The model results provide a valuable aid in understanding the data by filling in between the data sample points at better temporal resolution. At 70°N, there is a diurnal cycle in HNO3 mixing ratio at these altitudes. Production caused by proton precipitation at night is interrupted by fast photodissociation in daytime. At 85°N, there is very little solar illumination present at any time, no photodissociation is going on, and the daytime decrease is not observed. At 35 km, there is no photodissociation at any of the latitudes because the radiation between 190–330 nm, dissociating HNO3, is absorbed at higher altitudes. It seems that the local times of MIPAS observations are not always optimal to observe the highest values of the diurnal cycle which, according to the model, occur just before sunrise. This is especially clear at 58 km/70°N.
Figure 3 presents the modeled HNO3 production rates at 70°N on Oct 28 near midnight, i.e., around the time when the largest enhancements take place. The situation is similar throughout that night (not shown). In the upper stratosphere and mesosphere the main process producing HNO3 during solar proton forcing is the recombination between negative and positive ions. It accounts for about 95%, 75%, and 35% of the total production at 40 km, 35 km, and 30 km, respectively. At 30 km and below, N2O5 conversion and OH+NO2+M become equally important.
5. Discussion and Conclusions
 The significant decrease of HNO3 after Nov 01 which is seen in some of the observations, e.g., at 46 km/85°N, is difficult to explain by chemistry. It is likely that these changes are related to mixing with air from lower latitudes where SPE forcing is weaker and/or part of the produced HNO3 has been photodissociated. MIPAS observations of CO and CH4 (not shown) support this argument. They indicate weakening of the vortex after around Nov 01 which is manifested by increased CH4 and decreased CO inside the vortex region.
 At 35 km/85°N, the model overestimates the HNO3 production because long chemical lifetime and possible dynamical changes make 1-D modeling of N2O5 challenging. Unlike in the other cases, the modeled N2O5 mixing ratios differ from those of MIPAS on Oct 28–29, being about a factor of 3 higher (not shown). Thus, the HNO3 production by N2O5 conversion is overestimated. We studied the model sensitivity to N2O5 by making a shorter model run, started on Oct 28 00:00 UT. Again we used MIPAS data from Oct 26 as starting point but multiplying N2O5 by 0.65 and decreasing the rate of N2O5 conversion from 10−11 to 0.5 × 10−11 cm3 s−1 to better match the observed values on Oct 28–29. The results, shown in Figure 2, indicate a significantly better agreement with MIPAS HNO3 than those from the original run, the model now showing about 10% higher values. Therefore, most of the HNO3 overestimation can in this case be explained by the overestimation of N2O5 in the model.
 The model overestimates HNO3 also at 58 km/85°N. Sensitivity studies indicate (not shown) that H2O background has relatively little impact on the amounts of H+(H2O)n and HNO3, a 30–50% decrease leading to only about 5% decrease in HNO3. NOx overestimation could affect the balance between CO3− and NO3− cluster ions, favoring the latter which participate in HNO3 production. However, we find that the model NO2 is in relatively good agreement with MIPAS (not shown). It is possible that the uncertainties of reaction rate coefficients of the ion chemical reactions could be used to explain at least part of the difference. A detailed study of these is beyond the scope of this work.
 Our results shown in Figure 3 agree for the most part with the conclusions of Aikin , who pointed out the ion-ion recombination as the most probable source of HNO3 in the upper stratosphere and mesosphere in a case of electron precipitation, with possible contribution from N2O5 conversion below 36 km. On the other hand, Aikin found the NO2 + OH + M source to be negligible, while our results indicate a contribution comparable to the N2O5 conversion. It should be noted that at altitudes below ≈35 km the relative importance of the three processes shown in Figure 3 is quite sensitive to the amount of N2O5 and NO2, which can vary significantly with, e.g., latitude and solar illumination.
 Most of the SPE-produced HNO3 is eventually photodissociated to release HOx, which is the most important catalyst causing ozone depletion in the mesosphere. Verronen et al.  showed that because HNO3 accumulates at night and it is then rapidly converted to HOx at sunrise when ozone depletion is efficient, the current parameterizations of SPE-related HOx production, which ignore the role of HNO3, can lead to a significant underestimation of ozone depletion. At higher latitudes, where HNO3 photodissociation is absent throughout the day (e.g., at 46 km/85°N in Figure 2), the time delay between HNO3 production and HOx release can vary from days to weeks and dynamical processes may play an important role in how ozone responds to particle forcing. Therefore, it seems very important to accurately model HNO3 when studying atmospheric effects of particle precipitation.
 PTV and CFE were supported by the Academy of Finland projects THERMES (123275, Thermosphere and Mesosphere Affecting the Stratosphere) and SERCHAM (109054, Solar Energetic Radiation and Chemical Aeronomy of the Mesosphere), respectively. The IAA team was supported by the Spanish project ESP2004-01556 and EC FEDER funds. The IMK team was supported by BmBF under contract 50EE0512 and by DFG under the emphasis program CAWSES, project MANOXUVA. PTV's visit to Instituto de Astrofísica de Andalucia in April 2008 was supported by the Väisälä foundation.