The maintenance of elevated active chlorine levels in the Antarctic lower stratosphere through HCl null-cycles

The Antarctic ozone hole arises from ozone destruction driven by elevated levels of ozone destroying (“active”) chlorine in Antarctic spring. These elevated levels of active chlorine have to be formed first and then maintained throughout the period of ozone destruction. It is a matter of debate, how this maintenance of active chlorine is brought about in Antarctic spring, when the rate of formation of HCl (considered to be the main chlorine deactivation mechanism in Antarctica) is extremely high. 5 Here we show that in the heart of the ozone hole (16-18 km or 100-70 hPa, in the core of the vortex), high levels of active chlorine are maintained by effective chemical cycles (referred to as HCl null-cycles hereafter). In these cycles, the formation of HCl is balanced by immediate reactivation, i.e. by immediate reformation of active chlorine. Under these conditions, polar stratospheric clouds sequester HNO3 and thereby cause NO2 concentrations to be low. These HCl null-cycles allow active chlorine levels to be maintained in the Antarctic lower stratosphere and thus rapid ozone destruction to occur. For the observed 10 almost complete activation of stratospheric chlorine in the lower stratosphere, the heterogeneous reaction HCl+HOCl, the production of HOCl via HO2 +ClO, with the HO2 resulting from CH2O photolysis, is essential. These results are important for assessing the impact of changes of the future stratospheric composition on the recovery of the ozone hole. Our simulations indicate that, in the lower stratosphere, future increased methane concentrations will not lead to enhanced chlorine deactivation (through the reaction CH4 +Cl−→ HCl+CH3) and that extreme ozone destruction to levels below≈ 0.1 ppm will occur until 15 mid-century.

The period mid-September to early October is the period during which most of the Antarctic ozone loss occurs. This period of high ClO x ends abruptly with the rapid formation of HCl leading to deactivation of chlorine (Grooß et al., 1997;Grooß et al., 2011).
We present box-model calculations of Antarctic chlorine chemistry and ozone depletion which allow the chemical mecha-25 nisms and the impact of particular reactions to be studied in detail. In particular, we apply a detailed analysis of the temporal development of the rates of the key chemical processes based on a unique algorithm for the determination of chemical pathways (Lehmann, 2004). We demonstrate that for the efficacy of the HCl null-cycles it is essential that the heterogeneous reaction (Prather, 1992;Crutzen et al., 1992) HOCl + HCl → Cl 2 + H 2 O (R2) 30 as well as the gas-phase reaction (Crutzen et al., 1992) CH 3 O 2 + ClO → CH 3 O + ClOO (R3) 2014) had occurred by this time. This assumption likely constitutes a slightly too early onset of denitrification, but the impact of this assumption is minimal during polar night. The impact of denitrification has been explored in sensitivity studies (see appendix).
To integrate the system of stiff ordinary differential equations describing the chemistry we employ the solver SVODE (Brown 25 et al., 1989) that does not use the family approximation. The chemical kinetic data are taken from Sander et al. (2011). The photolysis rates are calculated in spherical geometry (Becker et al., 2000) for every hour using a climatological ozone profile for ozone hole conditions from HALOE measurements (Grooß and Russell, 2005).
Of particular importance for the simulations discussed here is the representation of the photolysis of CH 2 O. We have employed the recommended setup for both cross sections and quantum yields for 223 K (Sander et al., 2011). Using the photolysis 30 quantum yields suggested by Röth and Ehhalt (2015) yields very similar results to those presented here (Zafar, 2016). Branching ratios of 0 and 100% cannot not occur in reality (Röth and Ehhalt, 2015) but were used here as lower and upper limits. A possible temperature dependence of both the cross sections the quantum yields of the photolysis of CH 2 O could be important, as it could potentially lead to greater production of HO 2 in the photolysis of CH 2 O for the temperature range relevant here.
However, an accurate temperature dependence of these kinetic data for temperatures in the polar lower stratosphere could not be considered here due to lack of laboratory information (Smith et al., 2006;Röth and Ehhalt, 2015).
Heterogeneous chemistry is calculated on ice, nitric acid trihydrate (NAT), liquid ternary particles (H 2 O/H 2 SO 4 /HNO 3 ) and cold liquid binary aerosols. Temperature dependent uptake coefficients of heterogeneous reactions on liquid ternary and binary 5 aerosols are taken from parametrisations by Shi et al. (2001) as recommended (Sander et al., 2011). For uptake coefficients for reactions on NAT the parametrisation of Carslaw and Peter (1997) is used, based on laboratory measurements of Hanson and Ravishankara (1993). NAT particles are assumed to form from supercooled ternary solutions (STS) droplets. A HNO 3 supersaturation of three (corresponding to about 10 K supercooling) is required for NAT formation. The NAT particle density is assumed to be 3 · 10 −3 cm −3 . This NAT particle density is lower than assumed by Grooß et al. (2011) but is in better 10 agreement with observations.

Pathway analysis
A pathway in a chemical reaction system is a set of reactions converting some reactants of interest into some products of interest through some intermediate species, for which no net production or destruction occurs. An integer factor ("multiplicity") may be assigned to each reaction. The algorithm used here for the automatic determination of all significant pathways in a chemical 15 reaction system was developed by Lehmann (2004). As input it requires a set of chemical reaction equations and reaction rates, which are usually provided by a chemical model. Starting from the individual reactions (and their rates) as initial pathways, longer pathways are constructed step by step by connecting shorter ones. If a newly formed pathway contains sub-pathways, it is split into these. A rate for each pathway is calculated. Pathways with rates below a pre-described threshold are deleted already during the construction process, in order to avoid an intractably large number of pathways ("combinatorial explosion"). 20 3 Results

Maintenance of chlorine activation
We conducted box-model simulations for which the impact of mixing is neglected; however mixing across the Antarctic vortex edge is frequently overestimated substantially in current chemistry climate models (Hoppe et al., 2014). For the box-model, the development of temperature, potential temperature and solar zenith angle along a typical air parcel trajectory in Antarctic 25 spring is shown in Figure 1 (panels a-c). The initial titration between HCl and ClONO 2 via Reaction R1 occurred before June (as usually in the Antarctic, e.g. Santee et al., 2008) so that the model simulation starts with about one ppb of ClO x and with near zero values of ClONO 2 . Little further chemical change occurs ("sleeping chemistry") as long as solar zenith angles are large (until end of July), but with decreasing solar zenith angle, HCl further decreases, leading to increasing ClO x and, subsequently, chemical ozone destruction (e.g., Salawitch et al., 1988;Santee et al., 2005;Santee et al., 2008;     With decreasing ozone, the Cl/ClO ratio shifts increasingly towards Cl (Douglass et al., 1995;Grooß et al., 1997), so that HCl production via the reactions of Cl with CH 4 and CH 2 O (reactions R4 and R11) increases substantially by more than an order of magnitude between end of August and end of September ( Figure 1, panel g). Indeed, HCl formation through reaction 5 R4 is commonly considered as the main chlorine deactivation mechanism in Antarctic spring.
The formation rate of HCl via reactions R4 and R11 at the end of September amounts to more than 0.5 ppb per day. If these reactions proceeded unbalanced, chlorine activation and thus ozone depletion would be stopped within days. However, using an algorithm for the determination of chemical pathways (Lehmann, 2004) cycles C1 and C2 are identified for this period: Cycle C1 was first formulated by Crutzen et al. (1992) demonstrating the importance of reaction R3 for ozone hole chemistry.
Reaction R3 is essential for the production of a HO 2 radical from the CH 3 formed in reaction R4.
In cycles C1 and C2, HOCl is produced at the same rate as HCl because in both reactions R4 and R11, for each HCl molecule produced, also an HO 2 radical is generated (from CH 3 via R5, R3, and R6 or from CHO via R12). This HO 2 radical reacts further to form HOCl, which then reacts heterogeneously with HCl (in reaction R2) so that there is no net production of HCl.
We therefore refer to cycles C1 and C2 in the following as "HCl null-cycles". For the HCl null-cycle argument it is essential that Reaction 3 dominates the loss of CH 3 O 2 and indeed alternative reactions for CH 3 O 2 (e.g. the CH 3 O 2 selfreaction) were discussed (Müller and Crutzen, 1994). However, all relevant alternative reactions are included in the chemical scheme used here and the competition of the different reaction pathways of CH 3 O 2 is addressed in the pathway analysis that was employed (Lehmann, 2004).
These HCl null-cycles are effective in ensuring that there is no net production of HCl even though the speed of reactions R4 5 and R11, and thus the production of HCl, increases by about two orders of magnitude during September ( Figure 1, panel g).
We find that during the first 10 days of September, HCl is recycled at a slow rate of 88 ppt in 10 days through cycle C1, at a rate of 26 ppt in 10 days through cycle C2. However, for the last days of September (21-30), the rate of recycling of HCl in cycles C1 and C2 is much more rapid, the rates are 3.7 ppb in 10 days and 2.9 ppb in 10 days respectively.
Thus cycles C1 and C2 constitute the chemical mechanism responsible for the maintenance of high levels of active chlorine 10 (and thus of continued ozone destruction) under conditions of increasingly rapid HCl formation in reactions R4 and R11.

The path to full activation of HCl
However cycles C1 and C2 cannot explain the decrease of HCl from values of ≈ 1 ppb in early July to very low values (about 0.2 ppb) during late August and September and thus the complete activation of chlorine as observed (Santee et al., 2008).
Using the pathway analysis (Lehmann, 2004), we identify two chemical cycles, C3 and C4 which are responsible for the decline of HCl during August and September: Reaction R19 may also proceed in two steps: First, both cycles C3 and C4 require sufficiently fast heterogeneous reactions to be present. Second, in both cycles, it is important that HO 2 radicals are produced without simultaneous HCl formation (in contrast to cycles C1 and C2). The HO 2 radicals lead to formation of HOCl so that there is net HCl loss through reaction R2. The formation of HO 2 radicals in C3 and C4 is thus the key process responsible for the decline of HCl in August and September.
The dominant source of HO x radicals under the conditions of the polar lower stratosphere in late winter and early spring is 5 not the production of O( 1 D) radicals through ozone photolysis with subsequent reaction with H 2 O, but rather the the radical channel of the photolysis of CH 2 O (reaction R14) (Crutzen et al., 1992;Müller and Crutzen, 1994;Crowley et al., 1994). The photolysis of CH 2 O possesses two product channels; reaction R14 (radical channel) leading to the formation of two HO 2 radicals and the molecular channel which does not lead to production of HO x . The branching ratio between reactions R14 and R23 is uncertain; it is about 30% for the radical channel R14 for the conditions in question here (Röth and Ehhalt, 2015). Furthermore, both the photolysis cross sections and the branching ratio for the photolysis of CH 2 O are likely temperature dependent (with the intensities of the maxima of each absorption band increasing with lower temperature), but this dependence is not well characterised in laboratory measurements for temperatures below 200 K (Smith et al., 2006;Röth and Ehhalt, 2015) which are relevant here.
To demonstrate the importance of cycle C3 and in particular the HO 2 production through reaction R14, we conducted two sensitivity runs, one assuming 100% efficiency for radical channel R14 and zero for the molecular channel R23 (red line in Figure 2) and vice versa (blue line in Figure 2). These assumptions constitute upper and lower limits of HO 2 production in 5 reaction R14 that will not be reached in reality (see methods).
Without HO 2 production in the CH 2 O photolysis there is little reduction in HCl between late August and late September and, consequently, lower ClO x prevails leading to slower ozone loss than in the reference case. Assuming 100% efficiency for the HO 2 producing channel in the CH 2 O photolysis (red line in Figure  During the period of near zero values of HCl during late August to mid September, no substantial decrease of ClO x and thus no deactivation through an increase in HCl occurs (red line in Figure 2). This is the case because cycles C1 and C2 (net reactions R10 and R13) are still active and prevent the net formation of HCl. Solely the ongoing production of HO 2 radicals 15 causes somewhat enhanced levels of HOCl (0.1 to 0.2 ppb) during this period. Photolysis of HOCl inhibits an accumulation of larger amounts of HOCl.

The role of ClONO 2 and denitrification
For the conditions in the heart of the ozone layer, in the lower stratosphere, which are considered here, reaction ClO + NO 2 + M → ClONO 2 + M (R24) 20 and the related chemistry involving ClONO 2 only play a minor role. This is the case mainly because polar stratospheric clouds (PSCs) consisting of NAT or STS (Supercooled ternary solutions) exist almost continuously throughout the simulation sequestering most HNO 3 from the gas-phase. This leads to low NO 2 concentrations even in the presence of sunlight in spring.
Consequently, throughout Antarctic winter and spring, and in particular during the main ozone loss period from late August to late September, the rate of the heterogeneous reaction between HCl and HOCl (R2) is substantially larger than the rate of HCl 25 + ClONO 2 (R1), which is caused by the slow rate of formation of ClONO 2 . The ClO radical preferentially reacts with HO 2 (forming HOCl) rather than with NO 2 (forming ClONO 2 ).
For the efficacy of cycles C1 and C2, low gas-phase concentrations of NO 2 (and thus of HNO 3 ) are necessary. It is however not important whether the removal of HNO 3 from the gas-phase occurs temporarily (uptake in particles) or permanently (denitrification). In our box-model simulations the impact of denitrification is taken into account by assuming that denitrification has 30 occurred by the start of the simulation (4.5 ppb initial HNO 3 at 1 June). Overall, for the conditions in the core of the Antarctic vortex considered here, there is very little impact of denitrification on ozone depletion (see appendix).

Sensitivity of ozone loss on stratospheric methane and chlorine levels
In the coming decades, the composition of the Antarctic lower stratosphere will change considerably, in particular the stratospheric halogen loading will continue to decrease (e.g., WMO, 2014;Chipperfield et al., 2015). Repeating the reference run with Cl y halved (blue line in Figure 3) (typical conditions for ≈2050) results in lower active chlorine ClO x and thus slower ozone loss rates, as expected. However, in this case the ozone depletion period in the lower stratosphere is longer, so that 5 extremely low ozone values (below 0.2 ppb) are reached also for the initial amount of Cl y halved, albeit about three weeks later than in the reference case. The fact that in our simulation the impact of a reduced amount of Cl y is mainly observed in September is consistent with the recent conclusion that signs of healing in the Antarctic ozone layer have emerged for September (Solomon et al., 2016). Note that in the reference case ozone depletion stops because ozone is reaching extremely low values although temperatures are still low enough for PSCs and heterogeneous processing; the same mechanism works for halved Cl y , 10 solely the low values are reached later in the season due to the slower ozone loss rate.
Doubling methane in the simulation leads to a speed up of Reaction R4, thereby also to enhanced production of CH 2 O and thus also to a speed up of Reaction R11. But doubling methane does not lead to an enhanced deactivation, because cycles C1 and C2 are active and inhibit the deactivation effect of Reactions R4 and R11. However, surprisingly at first, doubling of methane means that more CH 2 O is produced in cycle C1 leading to somewhat faster HCl activation in late August and early 15 September through cycle C3 and thus to higher ClO x and a somewhat faster ozone depletion during this period.
A simulation combining doubled methane and halved Cl y shows results very close to those for halved Cl y , but again the doubled methane does not lead to a substantially faster deactivation and practically the same very low ozone values are reached in late October. Thus we suggest that the very low ozone values observed today in the core of the polar vortex (Solomon et al., 2005) will continue to be measured for decades to come and that the recovery of the stratospheric chlorine loading should lead 20 to a shift by some weeks of the low values to later in the season, consistent with recent observations (Solomon et al., 2016).

Multi-trajectory simulations
We have repeated the reference run for a set of realistic trajectories in the vicinity of the South pole including diabatic descent and latitude variations (Grooß et al., 2011). The initial values (for 1 August) of HCl and ClO x were chosen consistently with the reference run. The results of these runs (Fig. 4) show a significant variability in the decline of HCl and ozone loss rate, but 25 all show a rather similar behaviour as the reference run. Namely, very little chemical change in polar night until early August ("sleeping chemistry") and a relatively slow additional chlorine activation until early September. Next, a maintenance of high ClO x values during mid-September to early October (due to cycles C1 and C2) accompanied by rapid ozone loss. This period of high ClO x ends abruptly with the rapid formation of HCl (and thus chlorine deactivation), which occurs in a similar way for the individual trajectories in the time frame late September to early October.  For the heart of the ozone hole in the Antarctic lowermost stratosphere, formation of HCl through reactions R4 and R11 is very rapid for enhanced chlorine levels with the rate of HCl formation increasing by more than a factor of ten during September. We have shown that high levels of active chlorine are maintained nonetheless because formation of HCl is balanced in very effective HCl null-cycles, allowing rapid chemical destruction of ozone to proceed. Further, for the depletion of HCl to very low values, 5 formation of HO 2 is essential, with photolysis of CH 2 O being the major net source of HO 2 . Owing to the uptake of HNO 3 in NAT and STS particles, NO x chemistry and the formation of ClONO 2 are of minor importance. Chlorine is finally deactivated if an imbalance in the null-cycles occurs leading to a rapid and almost complete conversion of the activated chlorine into HCl and thereby putting a halt to ozone depletion (Douglass et al., 1995;Grooß et al., 1997;Grooß et al., 2011).   c Figure A1. Simulations for different initial ozone mixing ratios. Simulation as in the reference run but with different initial ozone mixing ratios (2.0 to 2.6 ppm).
We have conducted six additional box-model simulations, identical to the reference run (initial ozone 2.2 ppm), but with 5 different initial ozone mixing ratios ranging from 2.0 to 2.6 ppm (Fig. A1) Similarly, the timing of the rapid increase in HCl and chlorine deactivation is different for the different initial ozone mixing ratios (Fig. A1).
Importantly however, the general pattern of chlorine activation and decrease of HCl is very similar until about 20 September indicating that the efficacy of the HCl null-cycles is not affected by the initial ozone mixing ratio.  Figure A2. Simulation for the impact of denitrification. The impact of denitrification is tested by initialising the model run with HNO 3 doubled and halved.
The formation of polar stratospheric clouds impacts polar chemistry by taking up HNO 3 and thereby reducing HNO 3 in the gas-phase (Crutzen and Arnold, 1986;Toon et al., 1986;Salawitch et al., 1988 (STS) particles the reduction of gas-phase HNO 3 lasts only as long as temperatures are low enough for the particles to exist, because the particles are too small to sediment significantly. Nitric acid trihydrate particles, however, can grow to large sizes allowing substantial sedimentation speeds, so that HNO 3 is removed from stratospheric air masses permanently (a process referred to as "denitrification", Fahey et al., 2001;Molleker et al., 2014). Denitrification is observed regularly in the lower stratosphere in Antarctica and in cold winters in the Arctic (e.g., Davies et al., 2006;Santee et al., 2008).

5
For the case considered here, PSCs that sequester HNO 3 from the gas-phase exist almost continuously until mid September ( Fig. 1). Therefore, the gas-phase concentration of HNO 3 during the period of chlorine activation, maintenance of active chlorine, and chemical ozone depletion is controlled by PSCs and not by denitrification.
Nonetheless, we conducted a sensitivity simulation initialised with half (2.25 ppb) or twice (9 ppb) the amount of initial HNO 3 (Fig. A2). These assumptions result in practically the same ozone depletion in late September as in the reference case 10 (Fig. A2). For mid September, the simulation with 9 ppb initial HNO 3 shows about 10% more active chlorine than the reference run and thus a somewhat more rapid ozone loss (ozone in mid September ≈ 0.1 ppm less than in the reference run). For half the initial HNO 3 (2.25 ppb) the situation is reversed, but differences are even smaller.
Thus, as to be expected, the impact of the initial HNO 3 mixing ratio on the simulated ozone loss and chlorine activation is only moderate. Solely in the short periods when HNO 3 is released to the gas-phase and some of this gas-phase HNO 3 is 15 converted to NO x , the resulting production of ClONO 2 is enhanced in the case of a greater initial HNO 3 mixing ratio. Therefore, a greater initial HNO 3 mixing ratio (assumption of no denitrification) leads to a somewhat faster HCl depletion (through R1), a stronger chlorine activation and thus a little bit more chemical ozone loss (Fig. A2).
Overall, like for the sensitivity on the initial ozone mixing ratio, the general pattern of chlorine activation and decrease of HCl is very similar for the results of the reference run and the results under the (rather extreme) assumptions of reduced and 20 enhanced denitrification (Fig. A2).