Heterogeneous Kinetics of H 2 O , HNO 3 and HCl on HNO 3 1 hydrates ( α-NAT , β-NAT , NAD ) in the range 175-200 K 2

10 Experiments on the title compounds have been perfor med using a multidiagnostic stirred-flow 11 reactor (SFR) in which the gasas well as the cond ensed phase has been simultaneously 12 investigated under stratospheric temperatures in th e range 175-200 K. Wall interactions of the 13 title compounds have been taken into account using Langmuir adsorption isotherms in order 14 to close the mass balance between deposited and des orbed (recovered) compounds. Thin solid 15 films at 1 μm typical thickness have been used as a proxy for a tmospheric ice particles and 16 have been deposited on a Si window of the cryostat with the optical element being the only 17 cold point in the deposition chamber. FTIR absorpti n spectroscopy in transmission as well as 18 partial and total pressure measurement using residu al gas MS and sensitive pressure gauges 19 have been employed in order to monitor growth and e vaporation processes as a function of 20 temperature using both pulsed and continuous gas ad mission and monitoring under SFR 21 conditions. Thin solid H2O ice films were used as the starting point through t, with the 22 initial spontaneous formation of α-NAT followed by the gradual transformation of αβ23 NAT at T > 185 K. NAD was spontaneously formed at s omewhat larger partial pressures of 24 HNO3 deposited on pure H 2O ice. In contrast to published reports the formati on of α-NAT 25 proceeded without prior formation of an amorphous H NO3/H2O layer and always resulted in 26 β-NAT. For αand β-NAT the temperature dependent accommodation coeffi ci nt α(H2O) 27 and α(HNO3), the evaporation flux J ev(H2O) and Jev(HNO3) and the resulting saturation vapor 28 pressure P eq(H2O) and Peq(HNO3) were measured and compared to binary phase diagra ms of 29 HNO3/H2O in order to afford thermochemical check of the ki netic parameters. The resulting 30

investigated under stratospheric temperatures in the range 175-200 K. Wall interactions of the 13 title compounds have been taken into account using Langmuir adsorption isotherms in order 14 to close the mass balance between deposited and desorbed (recovered) compounds. Thin solid 15 films at 1 µm typical thickness have been used as a proxy for atmospheric ice particles and 16 have been deposited on a Si window of the cryostat with the optical element being the only 17 cold point in the deposition chamber. FTIR absorption spectroscopy in transmission as well as 18 partial and total pressure measurement using residual gas MS and sensitive pressure gauges 19 have been employed in order to monitor growth and evaporation processes as a function of 20 temperature using both pulsed and continuous gas admission and monitoring under SFR 21 conditions. Thin solid H 2 O ice films were used as the starting point throughout, with the 22 initial spontaneous formation of α-NAT followed by the gradual transformation of α-β-23 NAT at T > 185 K. NAD was spontaneously formed at somewhat larger partial pressures of 24 HNO 3 deposited on pure H 2 O ice. In contrast to published reports the formation of α-NAT 25 proceeded without prior formation of an amorphous HNO 3 /H 2 O layer and always resulted in 26 β-NAT. For αand β-NAT the temperature dependent accommodation coefficient α(H 2 O) 27 and α(HNO 3 ), the evaporation flux J ev (H 2 O) and J ev (HNO 3 ) and the resulting saturation vapor 28 pressure P eq (H 2 O) and P eq (HNO 3 ) were measured and compared to binary phase diagrams of 29 HNO 3 /H 2 O in order to afford thermochemical check of the kinetic parameters. The resulting 30 H 2 SO 4 /HNO 3 /H 2 O supercooled solutions (type Ib) or pure H 2 O ice (type II) (Zondlo et al. 48 2000) and are formed during the polar winter season when temperatures are sufficiently low 49 in order to allow H 2 O supersaturation that ultimately leads to cloud formation in the dry 50 stratosphere subsequent to ice nucleation (Peter, 1997). 51 Ozone is depleted during the Arctic and Antarctic spring season after unreactive chlorine 52 reservoir compounds, ClONO 2 and HCl, are converted into molecular chlorine and rapidly 53 photolyze into active atomic chlorine during the spring season (Solomon, 1990  and prevents formation of reservoir species with longer atmospheric residence times. 71 The study of HNO 3 interaction with ice in the temperature and pressure ranges typical of the 72 UT/LS is crucial in order to understand the de-nitrification process initiated by reaction (R1) 73 and its effectiveness in the overall ozone destruction mechanism. To this purpose, many 74 research groups (Voigt et  The existence of several crystalline hydrates of nitric acid has been confirmed for several 86 years. Hanson and Mauersberger (1988) have identified two stable hydrates, namely, nitric 87 acid monohydrate (NAM, HNO 3 H 2 O) and nitric acid trihydrate (NAT, HNO 3 3H 2 O) the 88 latter of which is thought to be of atmospheric importance. Several distinct crystalline 89 hydrates of HNO 3 have been found by Ritzhaupt and Devlin (1991)  Hanson (1992) also measured the uptake coefficient of HNO 3 on NAT using a cold coated-130 wall flow tube with HNO 3 deposited on ice condensed on the cold flow tube walls and 131 reported γ NAT (HNO 3 ) > 0.3. A rapid uptake was observed which decreased as the surface 132 coverage or dose of HNO 3 increased. Furthermore, the observed steady state partial pressure 133 of HNO 3 over the ice substrate is about a factor of 5 higher than the HNO 3 vapor pressure 134 over NAT and thus indicates that no hydrate was actually formed during the experiments. 135 Therefore, the observed uptake has most likely to be attributed to uptake on other cold 136 surfaces in the flow reactor. 137 Reinhardt et al. (2003) reported γ NAT (HNO 3 ) = 0.165 in the temperature range 160 to 170 K. 138 They used a slow flow reaction cell coupled with DRIFTS (Diffuse Reflectance Infrared 139 Fourier Transform Spectroscopy) for the detection of adsorbed species and downstream FTIR 140 for the detection of gas phase HNO 3 . 141 In the investigation of the properties of binary chemical systems the behavior of the simple 142 single-component systems is an important stepping stone. Hynes et al. (2002) observed 143 continuous uptake of HNO 3 on water-ice films below 215 K and time dependent uptake above 144 215 K, with the maximum uptake γ ice (HNO 3 ) decreasing from 0.03 at 215 K down to 0.006 at 145 235 K. They also observed that the uptake of HCl at 218 K on ice surfaces previously dosed 146 with HNO 3 is reversible. Furthermore, the adsorption of HNO 3 on ice surfaces which 147 contained previously adsorbed HCl indicates that HCl is displaced from surface sites by 148 In this work, the results for the kinetics of H 2 O and HNO 3 gas interacting with 150 spectroscopically characterized HNO 3 hydrates will be presented. The independent 151 measurement of the absolute rate of evaporation R ev [molec s -1 cm -3 ] and the accommodation 152 coefficient α of H 2 O and HNO 3 on αand β-NAT substrates is performed using a 153 combination of steady state and real-time pulsed valve experiments. Results on the kinetics of 154 the ternary system HCl on HNO 3 hydrates will also be presented. All experiments reported in 155 this work have been performed using a multidiagnostic stirred flow reactor (SFR) in the 156 molecular flow regime, which has been described in detail before (Chiesa and Rossi, 2013;157 Iannarelli and Rossi, 2014). In addition, all experiments have been performed under strict 158 mass balance control by considering how many molecules of HNO 3 , HCl and H 2 O were 159 present in the gas vs. the condensed phase (including the vessel walls) at any given time. 160 These experiments have been described by Iannarelli and Rossi (2015). Most importantly, the 161 consistency of the accommodation and evaporation kinetics has been checked using the 162 method of thermochemical kinetics (Benson, 1976) Table 1 reports its characteristic parameters. Briefly, it consists of a low-pressure 172 stainless steel reactor, which may be used under static (all valves closed) or stirred flow (gate 173 valve closed, leak valves open) conditions. We use absolute total pressure measurement and 174 calibrated residual gas mass spectrometry (MS) to monitor the gas phase and FTIR 175 spectroscopy in transmission for the condensed phase. Thin solid films of up to 2 µm 176 thickness are grown on a temperature controlled Si substrate and an average of 8 scans are 177 recorded at 4 cm -1 resolution in the spectral range 700-4000 cm -1 at typical total scan time of 178 45-60 s. 179 The 1" Si window is the only cold spot in the reactor exposed to admitted gases and therefore 180 the only place where gas condensation occurs. This allows the establishment of a 1:1 181 correspondence between the thin film composition and the changes in the gas partial pressures 182 in the reactor. Experimental proof of mass balance has previously been reported for this setup 183 (Delval et al., 2003;Chiesa and Rossi, 2013;Iannarelli and Rossi, 2014;2015). 184 The introduction of HNO 3 in the system forced us to slightly modify the inlet system used 185 previously (Iannarelli and Rossi, 2014) in order to take into account the fact that HNO 3 is an 186 extremely "sticky" molecule that interacts with the internal surfaces of the reservoir vessel of 187 the inlet system as well as with the reactor walls of the SFR (Iannarelli and Rossi, 2015). We 188 therefore minimized the volume of the admission system and only retained the absolutely 189 necessary total pressure gauge for measuring the absolute inlet flow rate (molecule s -1 ). 190 Similarly to the case of HCl and H 2 O (Iannarelli and Rossi, 2014) we have described the 191 HNO 3 interaction with the reactor walls using a Langmuir adsorption isotherm and 192 determined the concentration of HNO 3 in the ice sample after calibration of HNO 3 following 193 the methodology described in Iannarelli and Rossi (2015). Table 2  The protocol for the growth of α-NAT, β-NAT and NAD thin films has also been described 199 in Iannarelli and Rossi (2015 The growth protocols for α-NAT and NAD are similar and start after the deposition of a pure 208 ice film: the temperature of the Si substrate is held in the range 180 to 185 K for α-NAT and 209 at 168 K for NAD. The sample is exposed for approximately 10 min at SFR conditions to 210 HNO 3 vapor at flow rates in the range 3 to 7×10 14 molecule s -1 for α-NAT and 9×10 14 211 molecule s -1 for NAD. The typical total dose of HNO 3 admitted into the reactor is 2 to 3×10 17 212 molecules and 4×10 17 molecules for α-NAT and NAD, respectively, with almost all of it 213 adsorbed onto the ice film. In both cases, we observe the formation of a new phase after 214 approximately 5 min of exposure as shown in the change of the FTIR absorption spectrum. 215 The present experimental conditions seem to show that no nucleation barrier is present for α-216 NAT and NAD growth, in agreement with previous works (Hanson, 1992;Middlebrook et al., 217 1992;Biermann et al., 1998). In contrast, Zondlo et al. (2000) have shown that crystalline  218   growth occurs via an intermediate stage of supercooled H 2 O/HNO 3 liquid forming over ice.  219 After exposure the temperature of the substrate is set to the desired value for the kinetic 220 experiments on α-NAT or NAD as a substrate. 221 The protocol for the growth of β-NAT is different compared to NAD and α-NAT hydrates as 222 it only starts after the growth of an α-NAT film. After the HNO 3 flow has been halted, the α-223 NAT/ice system is set to static conditions and the temperature increased to 195 K. During the 224 temperature increase the α-NAT film converts to β-NAT as shown by means of FTIR 225 spectroscopy Iannarelli and Rossi, 2015), and once the conversion is 226 completed the temperature is set to the desired value to start the kinetic experiments using β-227 NAT as substrate. Typical growth protocols under mass balance control showing both the 228 FTIR transmission as well as the corresponding MS signals of HNO 3 as a function of 229 deposition time have been published previously (Iannarelli and Rossi, 2015). 230 In all samples used for this work, we never have a pure HNO 3 hydrate because we always 231 operate under conditions of excess or comparable amounts of pure ice. Excess ice has been 232 shown to have a stabilizing effect on both α-NAT and β-NAT (Weiss et al., 2016) and in all 233 our experiments the presence of excess ice has been confirmed by FTIR spectra (Iannarelli 234 and Rossi, 2015). 235

Experimental Methodology 236
The experimental methodology used in this work is an extension of the methodology reported 237 (1) 244 All terms in Equation (1) are flow rates in molec s -1 with the terms from left to right 245 corresponding to molecules admitted into the reactor (F in ), molecules desorbing from the 246 F in (X) + F des (X) + F ev (X) = F SS (X) + F ads,w (X) + F ads,ice (X) reactor walls (F des ), molecules evaporating from the ice surface (F ev ), molecules effusing 247 through the leak valve into the MS chamber (F SS ), molecules adsorbing onto the reactor walls 248 (F ads,w ) and molecules adsorbing onto the ice film (F ads,ice ). 249 Under the assumption that the adsorption onto the walls may be described as a Langmuir-type 250 adsorption, Eq. (1) may be expressed as follows for a gas X: 251 where V is the reactor volume in cm 3 , R in (X) the rate of molecules X admitted into the 253 chamber in molec⋅s -1 ⋅cm -3 , N TOT the total number of molecules X adsorbed on the reactor 254 walls, k des,w (X) the desorption rate constant from the reactor walls in s -1 , θ the fractional 255 surface coverage in terms of a molecular monolayer, R ev (X) the rate of evaporation of X from 256 the ice in molec⋅s -1 ⋅cm -3 , R SS (X) the rate of effusion through the leak valve in molec⋅s -1 ⋅cm -3 , 257 S w and S film the surfaces of the reactor walls and the thin film in cm 2 , α w (X) and α film (X) the  Table 2. 266 In the case of H 2 O, once the selected substrate has been grown according to the protocol 267 briefly described above, the film is set to a chosen temperature. After steady state conditions where ω(H 2 O) is the calculated gas-surface collision frequency in s -1 and is reported in (4) 280 where k esc (X) is the effusion rate constant of gas X out of the reactor in s -1 (see Table 1). 281 Finally, [X] SS is used to calculate R ev (X) using Eq. (2). 282 Subsequently, the film is set to a higher temperature, In this case the advantage of the PV technique as a real-time method of observation is lost. 290 Therefore, in order to measure the kinetics of HNO 3 gas in the presence of α-NAT, β-NAT 291 and NAD ice films we have used the two-orifice method first described by Pratte et al. (2006). 292 It has been modified to take into account the interaction of HNO 3 with the internal walls of 293 the SFR. The two-orifice method has also been used to measure the kinetics of H 2 O on HNO 3 294 hydrates in order to compare these results with the results of PV experiments for H 2 O. 295 The two-orifice (TO) method allows the separation of the rate of evaporation R ev (X) and the 296 condensation rate constant k c (X) of a gas X by choosing two different escape orifices and 297 measuring the corresponding value of concentration [X] SS at steady state of gas X inside the 298 reactor. By alternatively opening the small orifice (S) and both orifices (M) (see Figure 1), 299 two steady state equations hold for a probe gas X which are reported in Eqs. (5) and (6)  The kinetic parameters R ev (X) and k c (X) are calculated from Eqs. (7) and (8) as follows: 307 This method leads to larger uncertainties for both R ev (X) and k c (X) compared to the combined 310 We also used the combination of real-time PV and steady state experiments using HCl as a 317 probe gas and applied the experimental method described previously in order to measure the 318 kinetics of HCl, R ev (HCl) and α (HCl), in the presence of α-NAT and β-NAT ice films. 319 Once the kinetics R ev (X) and k c (X) have been measured using the combination of PV and 320 steady state experiments (H 2 O, HCl) or the two-orifice method (HNO 3 , H 2 O), we may 321 calculate the equilibrium vapor pressure P eq (X) for each gas according to Eq. (9): 322 where R is the molar gas constant in cm 3 Torr K -1 mol -1 , T the temperature of the thin film in 324 K and N A Avogadro's constant in molec mol -1 . The largest uncertainty in our experiment is that of the flow rate introduced into the reactor, 420 which is assigned a relative error of 25%. The flow rate measurement affects the calibration 421 of the MS and therefore the measurement of all the concentrations in the reactor (Eq. 4). 422 Therefore, we estimate a global relative error of 30% for PV experiments and double this 423 uncertainty for TO experiments because Equations (7) and (8) imply a difference of two large 424 numbers in many cases, as discussed above. We therefore assign a global 60% relative error 425 to results obtained in TO experiments. 426 components which leads to a single degree of freedom for the system. At constant 477 temperature different HNO 3 /H 2 O mixing ratios will lead to different values of P eq (H 2 O) if we 478 stay on an isotherm. This corresponds to a vertical cut in the binary phase diagram for β-NAT 479 in Figure 5. It shows that we expect P eq (H 2 O) values between a factor of ten or so for the 480 experimental points that "fill" the NAT phase diagram more or less homogeneously within 481 the used T range. 482 In addition, Figure 5 shows that the majority of points are in the rectangular shaded area Turning to the upper panel of Figure 6 we display a series of FTIR transmission spectra from 507 700 to 4000 cm -1 at specific times during the repetitive pulsing experiment which are 508 indicated in the lower panel by a series of color-coded "sp1" and continuing going from red to 509 purple. The principal peak positions have been collected in Table 3 and will be discussed 510 below in terms of changes in the "pure" α-NAT/ice absorption spectra owing to the presence 511 of increasing adsorbed HCl. The enlarged IR-spectral range in the upper panel of Figure 6 512 displays the effect of the HCl adsorption particularly well by showing a non-monotonic 513 sequence of IR absorption peaks not present in the "pure" reference spectra from Iannarelli 514 and Rossi (2015). The raw MS data from the lower panel of Figure 6 have been used to 515 calculate the kinetic and thermodynamic data displayed in Figure 8. 516 Figure 7 displays raw data from repetitive pulsed dosing of HCl onto a β-NAT/ice substrate in 517 analogy to Figure 6. The eleven individual pulses corresponded to (6-7)×10 16 molecule per 518 pulse resulting in a total HCl dose of approximately 4 x 10 17 molecules which amounts to 519 1300 molecular monolayers or so. Like in Figure 6 the upper panel displays a series of color-520 coded FTIR absorption spectra in transmission with the principal peak positions collected in 521 Table 3. As for Figure 6 the MS steady-state levels at the different temperatures will be used 522 to derive the kinetic and thermodynamic data of Figure 9 as a function of temperature. 523 In addition, Figure S6 presents an enlarged graph for the non-exponential decay of a HCl 524 pulse interacting with both αand β-NAT on a 30 s time scale consisting of a fast and a 525 slowly-decaying portion. The evaluation of such pulsed admission MS signals has been 526 presented in the past (Iannarelli and Rossi, 2014, Supplemental Information (SI)) and the 527 present analysis and fitting of the HCl MS signals follows the same scheme. 528 A look at Table 3 should provide an answer as to whether or not there is an identifiable 529 spectral fingerprint of HCl adsorbed on α-or β-NAT in the FTIR absorption spectrum of the 530 combined αor β-NAT/HCl system displayed in Figure 6 and Figure 7. 531 The first column of Table 3  The series of FTIR absorption spectra displayed in Figure 6 shows the non-monotonous 540 change of intensity at this transition (1328 cm -1 ): sp1 (red), sp2 (yellow) and sp3 (green) 541 display the growth of a shoulder to the red of the 1375 cm -1 peak, sp4 (turquoise), sp5 (blue) 542 and sp6 (purple) show the separate peak in its decline (1328 cm -1 ) owing to evaporation of 543 HCl together with NAT. 544 For β-NAT the analogous situation is displayed in the second and fourth column of Table 3  545 and Figure 7. Here the presence of HCl is more discrete within the FTIR absorption spectrum 546 of β-NAT as Table 3 suggests the well-separated peak to the blue of the 3227 cm -1 ice peak at 547 3360 cm -1 to be a HCl tracer as it looks very similar to the HCl/H 2 O system (Iannarelli and 548 Rossi, 2014; Chiesa and Rossi, 2013). The peaks identified to appear in the FTIR spectrum 549 upon HCl adsorption may be found in the fifth column of Table 3 which displays the principal 550 IR peaks in the reference HCl/H 2 O system, except the 1200 cm -1 vibration found in column 1 551 and 2 whose origin remains unclear. 552 In order to restrain the number of independent measurements on this ternary system to a 553 practical level we had to make some assumptions and/or simplifications in order to measure 554 the unknown parameters of Eq. (2) for each gas used. Specifically, we made the following 555 doped α-NAT is lower by a factor of approximately 10 compared to P eq (HCl) of crystalline 597 hexahydrate in the overlapping temperature range (177.5-193.5 K). 598 P eq (HCl) of amorphous HCl/H 2 O mixtures is higher by a factor of 20 compared to P eq (HCl) of 599 HCl-doped α-NAT at low temperatures (177.5 K) with the difference being constant or 600 slightly decreasing at high temperatures (199.5 K) where P eq (HCl) of the amorphous mixture 601 is only a factor of 4 higher than P eq (HCl) of α-NAT. 602 P eq (HNO 3 ) on HCl-doped α-NAT films is equal within experimental error to P eq (HNO 3 ) of α-603 NAT films free of adsorbed HCl. It is lower by a factor of 1000 compared to P eq (H 2 O) on 604 pure ice in the measured temperature range 177.5-199.5 K. 605 indicates that adsorbed HCl molecules seem to have no effect on the rate of evaporation of 617 HNO 3 from β-NAT films in the presence of HCl as well, at least in the given T range. 618  Table 1) on α-639 NAT shows two distinct regimes of temperature dependence, as well. Figure 10 Table 4 reports a synopsis of the kinetic (J ev ) as well as the thermodynamic (P eq ) parameters 650 calculated for all experiments of the present work. 651 The considerable scatter in the kinetic data, reflected in the significant uncertainties of Eqs. 652 high temperature in common hexagonal ice that finally turned out to be a perturbed hexagonal 681 ice structure (Kuhs et al., 2012). 682 In the case of β-NAT we have good agreement between PV (dotted line) and TO (solid line) 683 experiments of P eq (H 2 O) as shown in the van 't Hoff representation displayed in Figure 11. 684 As already mentioned, the ice surface is exposed to a series of pulses of H 2 O during PV 685 experiments. The free sites may be saturated before the introduction of each consecutive pulse 686 resulting in the discrepancy between PV and TO experiments. We therefore believe that the 687 results from PV experiments are more precise but less accurate owing to partial surface 688 saturation whereas the TO experiments are less precise but more accurate. We chose the latter interface. This is the first time such a large discrepancy between two kinetic measurements 733 techniques has been observed. As expected, thermodynamic results are not affected for 734 reasons of microscopic reversibility because both forward (α(H 2 O)) and reverse reactions 735 (J ev (H 2 O)) are affected to the same extent which cancels out for the calculation of the values 736 of thermodynamic parameters. 737 Figure S3 of Supplement C shows the results of PV experiments using H 2 O as a probe gas on 738 α-NAT and β-NAT substrates. Red and black circles represent the decay of series of two 739 pulses on αand β-NAT, respectively, with the first and second pulse labeled accordingly. In 740 the case of α-NAT films (red circles), the decay of the second pulses is equal to within 20-741 30% of the decay of the initial pulses, and only in a few cases at temperatures higher than 180 742 K is the decay of the second pulse significantly slower than the initial pulse. In the case of β-743 NAT films, the decay of second pulses is consistently slower than the decay of first pulses in 744 most cases. This indicates that the surface of β-NAT films exposed to a transient 745 supersaturation of H 2 O vapor is more prone to saturation compared to α-NAT. because no sharp pulses could be generated with pure HNO 3 , presumably owing to the 794 tendency of nitric acid to stick to the inner surfaces, mainly on stainless (austenitic) steel. This 795 has been verified by measuring the Langmuir adsorption on that same surface ( Figure S1, 796 Table 2). The following equations define the corresponding straight lines based on the present 797 measurements. For α-NAT (Eqs. (17) and (18)) and β-NAT (Eqs. (19) and (20) For αand β-NAT we obtain Σ∆H ev 0,α and Σ∆H ev 0,β equal to 318.8 and 324.8 kJ/mol, 822 respectively, when we use the average of the TO and PV experiment for H 2 O and the TO 823 value listed above for HNO 3 evaporation. Specifically, we have used (63.4 ± 9.6) and (128.6 824 ± 42.2) for H 2 O-and (76.1 ± 14.4) and (96.5 ± 12.0) for HNO 3 -evaporation for αand β-825 NAT, respectively, as displayed above. Finally, we arrive at the difference Σ∆H ev 0,α -826 Σ∆H ev 0,β = -6.0 ± 20.0 kJ/mol which shows that β-NAT is marginally more stable than α-827 NAT. This is true despite the fact that the standard heat of evaporation for HNO 3 in α-NAT 828 (ΔH ! HNO [ ) is significantly larger than for β-NAT by 32.1 kJ/mol which may be expressed 829 by the fact that the calculated "affinity" of HNO 3 towards ice in the α-NAT is larger than for  Despite the considerable scatter of the data displayed in Figure 14 it may be pointed out that 888 the enthalpy of HCl evaporation is identical for αand β-NAT within the stated experimental 889 uncertainty: We compare ∆H 0 ev (HCl) of 78.4 ± 11.4 and 69.6 ± 5.8 kJ/mol for αand β-NAT 890 (equations (23) and (25)). On the other hand, we have equality, perhaps fortuitously, between 891 E ev (HCl) and ∆H 0 ev (HCl) for α-NAT following equations (22) Table 4) whereas the saturation effect seems not to affect the kinetic data for α-NAT. 902 The anomalously large experimental uncertainty for HNO 3 TO experiments on α-NAT 903 displayed in Table 4 certainly has to do with the restricted temperature interval over which we 904 were able to monitor α-NAT before it converted to β-NAT. This may be seen in the synoptic 905 overview of the van't Hoff plots for HNO 3 interacting with NAT displayed in Figure S4 of 906 Supplement D. This restricted T range is also visible in Figure 13A for J ev (HNO 3 ) from α-907 NAT.. 908

Atmospheric implications and conclusion 909
In this study we have confirmed that exposure of ice films to HNO 3 vapor pressures at 910 temperatures found in the stratosphere leads to formation of NAT hydrates. 911 Of the two known forms of NAT, namely α-NAT and β-NAT, the latter is the 912 thermodynamically stable one whereas metastable α-NAT is likely to be of lesser importance 913 in the heterogeneous processes at UT/LS atmospherically relevant conditions. 914 HCl kinetic measurements on α-NAT and β-NAT indicate that HCl does not displace a 936 significant number of HNO 3 molecules from the ice surface upon deposition, but rather that 937 HCl and HNO 3 do not strongly interact with each other in the condensed phase and that HCl 938 evaporates faster. This observation is also supported by the slower rates of evaporation and 939 the correspondingly higher values of the HNO 3 evaporation activation energy on α-NAT and 940 β-NAT, E ev (HNO 3 ) = (178.0 ± 27.4) and (102.0 ± 8.6) kJ mol -1 (see Table 4), respectively, 941 compared to the activation energy for HCl evaporation on HCl 6H 2 O, E ev (HCl) = (87.0 ± 17) 942 kJ mol -1 . This also is consistent with a larger calculated interaction energy of HNO 3 with H 2 O 943 ("affinity") in α-NAT compared to β-NAT (Weiss et al., 2016) despite the fact that ∆H f 0 (α-944 NAT) is less stable by 6.0 ± 20 kJ/mol compared to β-NAT. 945 A look at Table 5  with r, a, rh and N ML being the radius of the ice particle, shell thickness, relative humidity in 952 % and the number of molecules cm -2 corresponding to one monolayer. J ev rh and J ev max are the 953 evaporation fluxes of H 2 O at rh and rh = 0, the latter corresponding to the maximum value of 954 J ev which we calculate following Equation (2) or (8). The salient feature of this simple 955 evaporation model is the linear rate of change of the radius or diameter of the particle, a well-956 and widely known fact in aerosol physics in which the shrinking or growing size (diameter) of 957 an aerosol particle is linear with time if the rate of evaporation is zero order, that is 958 independent of a concentration term. 959 Table 5 lists the evaporation life times which are not defined in terms of an e-folding time 960 when dealing with first-order processes. In this example the lifetime is the time span between 961 the cradle and death of the particle, this means from a given diameter 2r and "death" at 2r = 0. 962 The chosen atmospheric conditions correspond to 190   (a) M in kg; A Si in m 2 ; V in m 3 ; R = 8.314 J K -1 mol -1 . "One side" corresponds to front or rear side of Si-window.

1200
In order to calculate the accommodation coefficient α using equation (3) we have used 2ω as the total collision 1201 frequency for both sides of the Si-window.
1202 1203 interaction with the internal stainless steel (SS304) surfaces of the SFR.   Table 1 and Table 2 The principal peak positions are listed in Table 3  The principal peak positions are listed in Table 3