We present simultaneous in situ observations of OH, HO2, ClONO2, HCl, and particle surface area inside a polar stratospheric cloud undergoing rapid heterogeneous processing. A steady-state analysis constrained by in situ observations is used to show that concentrations of OH calculated during a processing event are extremely sensitive to the assumptions regarding aerosol composition and reactivity. This analysis shows that large perturbations in the abundance of OH are consistent with the heterogeneous production of HOCl via ClONO2 + H2O → HOCl + HNO3 and removal via HOCl + HCl → Cl2 + H2O in a polar stratospheric cloud. If the cloud is composed of supercooled ternary solution (STS) aerosols and solid nitric acid trihydrate (NAT) particles, comparison with observations of OH show that modifications to surface reactivity to account for high HNO3 content in STS aerosols and low HCl coverage on NAT particles are appropriate. These results indicate that with the low HCl levels in this encounter and in a processed polar vortex in general, reactions on STS aerosols dominate the total heterogeneous processing rate. As a consequence, the formation of NAT does not lead to significantly faster reprocessing rates when HCl concentrations are low and STS aerosols are present. Model calculations that include these modifications to uptake coefficients for STS and NAT will lead to significantly slower reprocessing and faster recovery rates of chlorine in the springtime Arctic polar vortex.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The heterogeneous conversion of inorganic chlorine species from inactive to active forms fuels the catalytic cycles that destroy ozone in the wintertime Arctic polar vortex [Anderson and Toon, 1993; Rodriguez, 1993]. The most important reactions,
convert HCl and ClONO2 to Cl2 or HOCl on liquid and solid aerosols composed of HNO3/H2SO4/H2O mixtures. It is known that a large number of small liquid aerosols and solid particles coexist in polar stratospheric clouds (PSCs) [Toon et al., 2000]. It is also known that the stoichiometry of small liquid aerosols measured in PSCs formed in lee-wave events in the Arctic is consistent with supercooled ternary solutions (STS) and that of larger solid particles is consistent with nitric acid trihydrate (NAT) [Schreiner et al., 1999; Voigt et al., 2000].
 Because the rates of reactions (1)–(3) depend strongly on the microphysical properties of the particle, the distinction between liquid and solid phases is crucial to the rate of heterogeneous processing of chlorine [Ravishankara and Hanson, 1996; Carslaw et al., 1997]. While it is generally accepted that some combination of reactions and aerosols are responsible for the conversion of chlorine from inactive to active forms, the detailed mechanism, including the dependence on the concentrations of the reactants and products and the phase and composition of aerosols, remains largely untested by direct observations.
 HOx (OH and HO2) is strongly coupled to reactions (2) and (3) through the gas phase production and loss of HOCl:
In the conditions of the winter polar vortex where HOCl concentrations are relatively high and HOx levels are low, measurements of HOx are a highly sensitive probe of recent heterogeneous processing [Jaeglé et al., 1997]. In fact, while modeled ClO concentrations are insensitive to the composition and phase of aerosols assumed in trajectory models [Kawa et al., 1997], modeled HOx concentrations can be used to deduce the character of the heterogeneous processing event [Jaeglé et al., 1997].
 We present measurements of HOx and the species needed to constrain the rates of reactions (1)–(5) obtained aboard the NASA ER-2 high altitude research aircraft during a processing event on 5 March 2000 in the Arctic polar vortex. The fast photochemistry of HOx and HOCl enables a quantitative evaluation of reactions (2) and (3) using an instantaneous steady-state analysis based on simultaneous in situ observations. This observation-based approach is distinct from trajectory and box models that are hampered by uncertainties introduced by long integration times. The comparison between the steady-state analysis and observations will be used to test our understanding of reactions (2) and (3) directly, and of reaction (1) by inference. The results of these comparisons will be used to clarify the mechanism that reprocesses chlorine in the recovery period of the Arctic polar vortex.
 The measurements were obtained using instruments flown aboard the NASA ER-2 during the SOLVE/THESEO campaign deployed out of Kiruna, Sweden (68°N, 20°E). Measurements include simultaneous in situ observations of OH, HO2, ClONO2, ClO, ClOOCl, O3, HCl, CH4, H2O, NO, NO2, HNO3, NOy, temperature, and particle number density and volume. Most of the instrument capabilities are summarized by Newman et al. . New instruments that detect HNO3 and ClOOCl are described elsewhere (K. A. McKinney et al., in preparation; R. M. Stimpfle et al., submitted to Journal of Geophysical Research). The instruments that measure HNO3 and NOy each have the capability to distinguish between the gas and aerosol phases.
Figure 1 shows the concentrations of OH, ClONO2 and HCl, temperature, and particle surface area obtained during the ER-2 flight on 5 March 2000. In (a), the observed OH is compared to a parameterization of OH as a function of solar zenith angle (SZA). The parameterization represents the concentration of OH expected at a given SZA based on a fit to OH in the absence of heterogeneous reactions [Hanisco et al., 2001]. This fit is adjusted by 30% to match the higher OH resulting from additional HOx sources (e.g., Cl + CH4) inside the vortex. The tight correlation in the second half of the flight is typical; deviations by more than 30% are extremely rare and immediately identify the presence of a perturbation to normal HOx photochemistry.
 The large enhancements and smaller suppressions relative to the fit indicate the presence of both atypical sources and sinks of HOx during this flight. The concentrations of OH correlate with the concentrations of ClONO2 (Figure 1b) that show large variations resulting from partial recovery of the air parcels in the 10 days prior to the ER-2 intercept. Backtrajectories indicate that the parcels with large values of ClONO2 were not exposed to temperatures below 195 K in the 9 days prior to the day of the ER-2 intercept. Many of the trajectories of these parcels traveled just inside the edge of the vortex where temperatures are highest. The low values of HCl (∼275 ppt) in these partially recovered parcels are typical of the processed vortex, due to the much slower recovery time for HCl compared to ClONO2.
 Large particle surface areas and low temperatures are coincident with the enhanced levels of OH. The temperature measured by the MMS instrument on the ER-2 is shown in Figure 1c. Also shown are the threshold temperatures for NAT and H2O ice formation [Hanson and Mauersberger, 1988; Marti and Mauersberger, 1993]. The temperatures in the chemically perturbed region of the flight (3.5 × 104 < UT < 4.5 × 104 s) are below the NAT formation temperature and, according to air parcel trajectories, had been for at least 12 hours prior to the intercept. The temperatures in this region of the flight are the lowest temperatures experienced by these air parcels in the 10 days prior to intercept. The temperature determined with the NASA Goddard trajectory model 24 hours prior to the measurements is shown for reference. In the chemically perturbed region the temperature drops by 3–6 K over the preceding 24 hr period, with an average gradient of 0.16 K hr−1. The sharp increase in temperature centered at UT ∼ 40000 s occurs during a dive to 16 km.
 At the coldest temperatures there are large enhancements of the particle surface area (Figure 1d) derived from the size distributions measured by the MASP instrument [Baumgardner et al., 1996]. Inside the chemically perturbed region of the flight the particle surface areas are at least 5 μm2 cm−3, indicating the presence of a PSC. The surface area is dominated by two size modes inside the PSC, centered at particle diameters (D) of 0.5 μm (0.3 < D < 0.7 μm) and 1.0 μm (0.7 ≤ D < 1.25 μm). The particle surface areas in these two modes are roughly equal in most of the PSC. At the lowest temperatures, where total surface area reaches 80 μm2 cm−3 the surface area in the large particle mode dominates the total area. There is not a corresponding increase in any of the larger size bins recorded by the MASP instrument that would indicate the presence of large ice particles or the large NAT particles (∼20 μm) observed frequently earlier in the winter [Fahey et al., 2001].
 The composition of the two aerosol modes is not measured directly. However, indirect evidence suggests that the 0.5 μm mode is predominantly STS and the 1 μm mode is predominantly NAT. Remote measurements of the NASA DC-8 aerosol lidar on the same day indicate that large clouds of small spheres (liquid) and larger nonspheres (solids) were present in the region of the vortex encountered during this flight (E. V. Browell, personal communication). Because of the large aerosol volume, the PSC is likely to contain large amounts of nitric acid. Liquid nitric acid aerosols are likely to be STS and solid particles are likely to be NAT or NAD (nitric acid dihydrate) [Peter, 1997 and references therein].
Figure 1e shows the weight percent of HNO3 in the liquid aerosols determined from a parameterization for liquid HNO3/H2SO4/H2O aerosol composition [Carslaw et al., 1995]. The aerosol volume in the 0.5 μm mode inferred from the MASP measurement is consistent with the volume indicated by the STS parameterization. Simultaneous observations of enhanced HNO3 in the condensed phase by the CIMS and NOy instruments indicate the presence of particles containing HNO3 in the D = 1–2 μm size range (S. Dhaniyala et al., in preparation). Analyses of the vapor pressure relationships of this mode are consistent with NAT particles (S. D. Brooks et al., in preparation; K. A. McKinney et al., in preparation).
 Although there is likely to be some mixture in the phase of particles in each mode, the analysis will assume that each mode is made up entirely of either liquid or solid particles. The uncertainty of this assumption is at least as large as the uncertainty of the aerosol surface area derived from the MASP measurements, ±32% with liquid (spherical) aerosols and ±45% with solid (nonspheres). The possibility that all aerosols are liquid or solid will be explored below.
3.1. Production and Loss Terms
 The most significant production and loss terms of HOx and HOCl inside the Arctic vortex are shown in Figure 2. It is conceptually useful to include HOCl with HOx in order to emphasize the distinction between the rate determining steps of the primary production and loss terms and the reactions that simply cycle OH, HO2, and HOCl. This distinction is important because reactions (4) and (5) comprise a null sequence with respect to both HOx and HOCl. These reactions become important when coupled with the heterogeneous reactions (2) and (3) that lead to net production and loss of HOx and HOCl. The heterogeneous production sequence of HOx is given by
The sequence that describes the heterogeneous removal of HOx is
In both these sequences the heterogeneous production reaction (2) and loss of HOCl reaction (3) are the rate determining steps.
 In this analysis HOx and HOCl are assumed to be in steady state. This assumption is reasonable given the short lifetime of HOx (τ ∼ 1 minute) and HOCl (τphotolysis ∼ 1 hour). For the conditions of this flight, the predicted gas phase instantaneous steady-state concentration of HOCl differs from the diurnally averaged gas phase steady state by less than 10%. Furthermore, the measurements occurred in the late afternoon local time. Air parcels in the perturbed region had been exposed to sunlight (SZA < 90°) for a minimum of 7 hrs and for an average of 9.6 hours prior to the ER-2 intercept. Thus, it is unlikely that the enhancements result from the tail end of a prior nighttime processing event.
 The differential equation for HOx can be expressed in terms of the rate determining steps of these reactions and the gas phase processes that produce and remove HOx:
where Pgas and Lgas refer to the gas phase production and loss rates [Hanisco et al., 2001]. The heterogeneous terms include the first order rate, , where γ is the reaction probability, A is the surface area, and is the mean molecular collision frequency. Equation (6) will be solved for OH rather than HO2 because concentrations of OH are directly controlled by the production and loss terms shown in Figure 2. HO2 concentrations are influenced by these terms and, in addition, partitioning reactions that interconvert OH and HO2.
 The in situ observations are used directly in the calculation of production and loss rates when possible. Only 105 observations of ClONO2 are reported for this flight. All observations within ±30 s of the ClONO2 measurements are averaged into bins matching these times. The HNO3 observations are scaled by a constant offset to match the NOy observations in regions of low aerosol volume where these measurements are expected to be identical. Concentrations of O(1D), Cl, and HOCl are determined from steady-state relations using measurements and calculated photolysis rates. Photolysis rates along the flight track are calculated using a photolysis code constrained by ozone column from TOMS satellite observations and profiles from ozone sondes and ER-2 observations (R. J. Salawitch, personal communication). Rate constants are taken from the JPL-97 and JPL-00 recommendations [DeMore et al., 1997; Sander et al., 2000]. The rate constant of reaction (4) is determined from the mean of two recent laboratory measurements [Knight et al., 2000; Nickolaisen et al., 2000]. Parameterizations of heterogeneous reaction rate constants are taken from JPL-00, with modifications discussed below.
3.2. Parameterizations of γ
 Heterogeneous reaction rates for reactions (1)–(3) on binary H2SO4/H2O solution aerosols are given in the JPL-00 recommendation. This parameterization includes terms that account for low concentrations of HCl [Shi et al., 2001]. However, there are no recommendations that account for the presence of HNO3 in liquid aerosols. Measurements of reactions (1) and (2) in HNO3/H2SO4/H2O solutions with large amounts of HNO3 by Hanson  show decreased reaction probabilities compared with pure binary sulfuric acid solutions. The smaller reaction probabilities were reproduced by reducing the bulk uptake coefficient (Γb) and surface uptake coefficient (Γs) in the parameterization of the pure H2SO4/H2O solution by factors of up to two and ten, respectively [Hanson, 1998]. The divisors used by Hanson  are approximately given by the relations that depend on the weight percent of HNO3 (wt%):
These relations yield Db = 2 and Ds = 10 at wt% > 25%, the same as used by Hanson  for the highest concentration of HNO3 in that study.
 The reaction probabilities for reactions (1) and (2) using Γb/Db and Γs/Ds in the JPL-00 parameterization are shown in Figure 3a for conditions representative of the PSC. The coefficient for reaction (3) is determined from the JPL-00 parameterization without modification. The modification reduces γ1 by a factor of ∼2–6, and γ2 by ∼2, depending on temperature. It should be noted that other laboratory measurements show smaller effects of HNO3 on the reaction probabilities of γ1 and γ2 in HNO3/H2SO4/H2O solutions [Elrod et al., 1995; Zhang et al., 1994]. Based on these studies, treating the reactivity on STS the same as a pure H2SO4/H2O solution is also reasonable. The unmodified coefficients for reactions on pure H2SO4/H2O aerosols are shown in Figure 3c. The significance of these differences will be discussed below.
 The reaction probabilities determined from the parameterizations by Carslaw and Peter  and Carslaw et al.  are listed in Table 1 and shown in Figure 3b. The parameterizations for reactions (2) and (3) include the modification based on the partial pressure of HCl during the ER-2 encounter, pHCl ∼ 1.6 × 10−6 Pa, which is approximately 1000 times smaller than the experimental conditions by Abbatt and Molina [1992a, 1992b] and 10 times smaller than the conditions by Hanson and Ravishankara [1992, 1993]. The parameterizations without the modification for low HCl, shown in Figure 3d, are 10–100 times greater. These uncorrected reaction probabilities are approximately equal to the JPL-00 recommendations for NAT at temperatures below 192 K. The differences between the alternative schemes will be discussed below.
 The average gas phase production and loss rates of HOx inside the PSC are summarized in Table 2. Production is dominated by the photolysis of HNO3, followed by Cl initiated methane oxidation, and O(1D) reaction with H2O. The primary loss mechanism for HOx is the reaction of OH with HNO3, with small contributions from reaction of OH with ClO and HO2. The average discrepancy between the gas phase production and loss rates in the chemically perturbed region is 3.1 × 103 molecules cm−3 s−1 with a maximum of 2.5 × 104 molecules cm−3 s−1. Enhanced production of HOx occurs through reaction (2) followed by reaction (5). For [ClONO2] = 1 ppb, γ2 = 0.01 and surface area of 4 μm2 cm−3, the instantaneous production rate of HOCl via reaction (2) is 5.4 × 103 molecules cm−3 s−1, which is comparable to the gas phase rates of HOx production listed in Table 2. Thus, it is entirely reasonable that instantaneous production of HOx from heterogeneous reactions causes the enhancements in OH.
Table 2. Average Production and Loss Rates of HOx in 103 Molecules cm−3 s−1 Inside the PSC (3.5 × 104 s < UT < 4.5 × 104 s)a
 The comparisons of the concentration of OH calculated from equation (6) with observations of OH are shown in Figure 4. In panel (a), the predicted OH using only gas phase chemistry is compared with observations. The agreement outside of the processed region (UT > 45000 s) is within the uncertainty in the gas phase production and loss rates (±50%). In this region, the ratio of calculated to measured OH is 0.87 ± 0.14, typical of this steady-state calculation using in situ observations from the ER-2 payload outside the polar vortex [Hanisco et al., 2001]. In the processed region the gas phase model underpredicts OH by almost a factor of 10 where aerosol surface area is highest. Though less noticeable, there are regions early in the flight where the model overpredicts OH by 60%. These differences indicate that the heterogeneous processing can be either a net source or sink of HOx, depending on the relative rates of reactions (2) and (3). Inside the PSC the underpredictions of the model correspond to regions where concentrations of ClONO2 are elevated ([ClONO2] > 500 ppt).
 The calculation that includes heterogeneous reactions, shown in Figure 4b, closely matches the observations. This calculation uses the ternary gammas shown in Figure 3a for particles in the 0.5 μm mode and the gammas for NAT shown in Figure 3b for particles in the 1 μm mode. With these gammas, the heterogeneous rates account for over 40% of the total production and over 20% of the loss rate of HOx inside the PSC on average (see Table 2). The calculation reproduces both the large enhancements and the smaller depressions in OH relative to the gas phase model. Overall, the agreement is good, the ratio of calculated to measured OH is 1.02 ± 0.25, but there are several points where the model underpredicts the measurements by as much as 50%. These discrepancies fall within the uncertainties of the terms in equation 6 that are constrained by observations (e.g., Pgas, Lgas, [ClONO2], [HCl], [HOCl], total surface area, etc.). The uncertainty of these terms determined from adding each weighted uncertainty in quadrature is ±0.85 (1 − σ). However, since the aerosol composition and uptake coefficients are not known, the overall uncertainty is much larger than this number.
 The assumptions about the gammas that represent the heterogeneous rates used to calculate OH in Figure 4b are reasonable, but alternative assumptions should also be considered. Figure 4c shows the concentration of OH calculated using the unmodified JPL-00 H2SO4/H2O coefficients shown in Figure 3c for the 0.5 μm mode and the modified NAT coefficients for the 1 μm mode. The unmodified uptake coefficient for reaction (2) is larger than that shown in Figure 3a by a factor of ∼2. With this gamma, the heterogeneous production of HOCl is larger by a factor of two, leading to an average overestimate of OH by 40% in the chemically perturbed region, with some observations overpredicted by a factor of 2.5. This is particularly noticeable in the region dominated by heterogeneous production of HOx (40000 s < UT < 45000 s).
Figure 4d shows the OH calculated without the modifications for surface coverage of HCl on NAT shown in Figure 3d. The modification to the gammas on NAT surfaces reduces the reaction probabilities of γ3 by orders of magnitude, but reduces γ2 only slightly. With the unmodified NAT coefficients the total heterogeneous removal rate of HOCl is larger by a factor of 2–3, resulting in a large underprediction of OH in the region dominated by 1 μm particles. Given the uncertainty of the measured and derived quantities in equation (6) (±0.85), the discrepancies in Figures 4c and 4d are not likely due to errors in these terms, but to errors the treatment of the surface reactivities and/or the aerosol composition.
 The comparisons shown in Figure 4 are consistent with heterogeneous processing of ClONO2 and HCl by reactions (2) and (3) during the ER-2 encounter. The observed OH is well reproduced by the model that accounts for the two distinct particle compositions and includes modifications to uptake coefficients that account for HNO3 content in liquid STS aerosols and low HCl coverage on NAT surfaces. However, there are a large number of permutations that can be assumed for particle composition and reactivity that are not shown in Figure 4. For example, all aerosols can be assumed to be STS; all could be NAT; nitric acid dihydrate (NAD), sulfuric acid tetrahydrate (SAT) and/or H2O ice could be present in some amount; etc. The concentration of OH calculated assuming that the aerosols are either all liquid or all solid is shown in Figure 5. These calculations use the unmodified reaction coefficients shown in Figures 3c and 3d. If all the aerosols are assumed liquid the faster heterogeneous rates lead to a much greater overprediction of OH than shown in Figure 4c. If all aerosols are assumed to be NAT the slower rates lead to a more severe underprediction than that shown in Figure 4d.
 The calculations in Figure 5 illustrate a basic feature of the liquid and solid heterogeneous rates. The reactions on liquid aerosols lead to rapid production of HOCl and the reactions on solids lead to rapid removal of HOCl. The large enhancement of OH compared to the gas phase prediction is consistent with a significant contribution from liquid aerosols. At the same time, the rates for the liquid-only case are too fast, indicating that a reduction in γ2 and/or the liquid surface area is required.
 The largest overprediction of OH occurs at the lowest temperatures and highest surface areas where NAT or some other solid particles are likely to be significant. Thus, it is reasonable that the smaller rate of ClONO2 + H2O compared to the liquid only case is due in part to the presence of NAT. The difference between γ2 and γ3 on NAT without the correction for HCl coverage is at least a factor of 100. As a result, the calculation is extremely sensitive to the amount of NAT assumed in this calculation. If a significant amount of NAT is included in the calculation, the smaller gamma for HOCl + HCl is required to match the observations. Though the exact nature of the heterogeneous reactions remains uncertain, the observed OH can be matched only with some combination of slower liquid (D = 0.5 μm) and slower solid (D = 1 μm) rates.
 With a combination of STS and NAT included in the model, there are ambiguities in how to treat the surface area in each mode. The underpredictions in Figure 4b correspond to data where the surface area in the 1 μm mode is greater than 20 μm cm−3. It is possible that the selection criteria for liquid and solid phases lead to an undercount of the surface area in the liquid mode. If only 10% of the solid phase were assumed to be liquid, the calculations would match the observations throughout the PSC. Also, the possibility that small amounts of SAT, NAD, or ice are present in the PSC cannot be excluded. The reactivities of reactions (2) and (3) are thought to be the same on NAD as on NAT; the reactivity of reaction (2) is only slightly higher on SAT than on NAT; and the reactivity of reaction (3) is thought to be the same on SAT as on NAT. Thus, it is not possible to use the model comparison with observed OH to comment on the possible presence of SAT or NAD. The rates of reactions (2) and (3) are significantly faster on ice than on NAT (Figure 3b). It is possible that a small amount of ice or ice coverage on NAT could explain the discrepancy in Figure 3b at UT = 42000 s where the temperature crosses the frost point. If only 1% of the NAT surface area were ice or ice coating on NAT, the calculated OH would match the measured OH where T ∼ Tice.
 The discrepancies in Figure 4c and d are larger than the uncertainties of the observed and derived quantities used in equation (6), but it is possible that the discrepancies in Figures 4b and 4c are due to errors in the composition assumed for the 0.5 μm and 1 μm modes. For example, the overprediction of OH in Figure 4c could be a result of an overestimate of the liquid STS surface area. A reduction by a factor of 2 in the STS surface area would result in a similar result as Figure 4b. Likewise, the underprediction in Figure 4d could result from a lack of ice particle surface area in the model. If roughly 10% of the 1 μm surface area in the region that is underpredicted (40000 s < UT < 42500 s) is treated as ice, the calculation in this region would match the observations. Though these examples are not likely, the possibility that they might occur underscores the uncertainty of these results and highlights the need for simultaneous in situ measurements of particle composition and additional laboratory measurements of these heterogeneous reactions.
 If the modifications that account for HNO3 content in liquid aerosols and HCl coverage on NAT are required for reactions (2) and (3), they are also likely to be required for reaction (1). With these modifications, the rates of reactions (1)–(3) on NAT surfaces are significantly slower than on STS aerosols. As a result, the heterogeneous processing is dominated by reactions on liquid aerosols. Figure 6a shows the fractional contribution of reactions on STS aerosols to the total processing rate. Also shown is the fraction of the surface area in the 0.5 μm mode. Only at the lowest temperatures and highest NAT surface areas are reactions on NAT significant, but even under these conditions reactions on STS aerosols dominate the total conversion rate of ClONO2 and HCl to Cl2. This result implies that the formation of NAT under low HCl conditions has little impact on the total reprocessing rate of HCl and ClONO2.
 Together, these modifications have a large effect on the rate of the reprocessing of chlorine in the recovery period of the vortex. Figure 6b shows the instantaneous removal rate of ClONO2 by reactions (1) and (2) calculated using the reactivities that include modifications for HNO3 content in liquid aerosols and HCl coverage on NAT (Figures 3a and 3b) compared to the rate using the unmodified gammas (Figures 3c and 3d). (The heterogeneous removal rate of HCl is similar because reaction (1) is involved in both removal processes and the rates of reactions (2) and (3) are within a factor of two.) The modifications result in a net decrease in the rate of ClONO2 removal by a factor of ∼5 at the coldest temperatures. At warmer temperatures the rate of ClONO2 removal is largely unchanged and it is nearly identical to that determined from liquid H2SO4/H2O rates alone. This is consistent with analyses of the observed onset of chlorine activation at ∼195 K [Toohey et al., 1993; Kawa et al., 1997].
 During the recovery period of the Arctic vortex the magnitude of the removal rate of ClONO2 is particularly important. Unlike the initial processing that occurs in December and January when gas phase processes are dormant, the reprocessing must compete with gas phase recovery rates. During this flight any NO2 produced quickly reacts with ClO to form ClONO2, so that the production rate of ClONO2 in the PSC is equivalent to that of NO2 (i.e., the rates of HNO3 + hv and OH + HNO3 taken from Table 2). With the modified reactivities the heterogeneous removal is slower than the gas phase production above T ∼ 192.5 K, about 1.5 K lower than without these modifications. Thus, the factor of ∼5 reduction in the ClONO2 removal rate is roughly equivalent to a 1.5 K change in temperature. These changes are particularly significant because of the steep gradient in temperatures inside the Arctic vortex. Back trajectories indicate that, on average, parcels intercepted during this flight experienced temperatures below 192.5 K only 2% of the time during the 10 days prior to intercept compared to 23% of the time below 194 K. As shown by Carslaw et al. , the modifications made to the NAT reactivity alone lead to substantially slower reprocessing rates. The additional modification to the reactivity of STS aerosols would further reduce the calculated processing rates in conditions similar to those encountered here.
 The steady-state analysis of the HOx system inside a PSC provides a unique mechanism to compare in situ observations directly with heterogeneous reaction rates. The significant differences between the model results in this analysis underscore the importance of the correct representation of aerosol composition and reactivity. The assumptions regarding composition and reactivity are critically important in accurately predicting processing rates of chlorine in model calculations and accurately reproducing observations of HOx. This is in sharp contrast to model calculations of ClO that are hampered by the ambiguities associated with long integration times.
 Correlations with ClONO2, low temperatures, and large surface areas suggest that enhanced OH results from heterogeneous production and suppressed levels from the heterogeneous removal of HOCl. The steady-state analysis shows that the observations are consistent with the following:
Heterogeneous reactions of ClONO2 + H2O and HOCl + HCl both occur.
These reactions are fast and compete with gas phase production and loss rates of HOx.
The observed OH is reproduced by treating 0.5 μm aerosols as liquid STS and 1.0 μm particles as NAT.
The best agreement is obtained by reducing the liquid rate for ClONO2 + H2O on 0.5 μm aerosols (STS) and reducing the rate of HOCl + HCl on 1 μm aerosols (NAT).
 Given that these reactions occur, confirmation of the effects of HNO3 in STS and of HCl on NAT by further laboratory studies is crucial. These results indicate that with the low HCl levels in this encounter and in a processed Arctic polar vortex in general, reactions on STS aerosols are likely to dominate the total heterogeneous processing rate. As a consequence, NAT formation does not lead to significantly faster reprocessing rates when HCl concentrations are low. Model calculations that include these modifications to uptake coefficients for STS and NAT will determine significantly slower reprocessing and faster recovery rates of chlorine in the Arctic polar vortex.
 We thank the pilots and crew of the NASA ER-2, the logistical support of NASA AMES, and the engineering staff of the Harvard group for making the measurements possible. The NO and NOy data were provided by the Fahey group and the O3 data by E. C. Richard at the NOAA Aeronomy laboratory. Photolysis rates were provided by R. J. Salawitch at the NASA Jet Propulsion Laboratory. This work was supported by the NASA Upper Atmospheric Research Program.