Bromine atom production and chain propagation during springtime Arctic ozone depletion 1 events in Barrow , Alaska 2 3

Department of Chemistry, Purdue University, West Lafayette, IN, USA 7 8 Department of Earth and Atmospheric Sciences and Purdue Climate Change Research Center, 9 Purdue University, West Lafayette, IN, USA 10 11 School of Earth and Atmospheric Sciences, Georgia Institute of Technology, Atlanta, GA, USA 12 13 National Center for Atmospheric Research, Boulder, CO, USA 14 15 now at: Cooperative Institute for Research in Environmental Sciences, University of Colorado 16 Boulder, Boulder, CO, USA 17 18 now at: Chemical Sciences Division, NOAA Earth System Research Laboratory, Boulder, CO, 19 USA 20 21 now at: Department of Atmospheric and Ocean Sciences, University of Colorado Boulder, 22 Boulder, CO, USA 23 24


Introduction 47
The springtime depletion of boundary layer ozone in the Arctic has been the subject of 48 intense research for several decades. Early observations revealed a strong correlation between 49 ozone depletion events (ODEs) and enhancements in filterable bromine . 50 This discovery led researchers to propose a mechanism for the bromine-catalyzed destruction of 51 ozone. 52 natural tundra snow collected near Barrow, AK lend further evidence to this mechanism, and 62 also suggest Br 2 production from OH produced photochemically within the snowpack. 63 Once present in the gas-phase, bromine atoms can be regenerated through radical-radical 64 reactions of BrO with XO (where X = Br, Cl, or I), NO, OH, or CH 3 OO to propagate the chain 65 reaction and continue the destruction cycle of ozone. If BrO photolyzes or reacts with NO, O 3 is 66 regenerated, and there is a null cycle with respect to O 3 ; however, these two pathways represent 67 efficient routes for Br atom propagation. Thus R3 serves to make R2 effective in destruction of 68 O 3 . At the same time, Br atoms could be recycled through heterogeneous reactions of HOBr 69 with bromide in the condensed phase to release Br 2 to the gas-phase via the now well-known 70 "bromine explosion" mechanism (Vogt et al., 1996; Tang and McConnell, 1996;Fan and Jacob, 71 1992 Evidence for reaction sequence R5 -R8 has been provided through laboratory studies, which 77 found that Br 2 was produced when frozen bromide solutions were exposed to gas-phase HOBr 78 (Huff and Abbatt, 2002;Adams et al., 2002). 79 To efficiently sustain the ozone destruction cycle to the point of near complete loss of 80 boundary layer ozone ([O 3 ] < 2 ppb), bromine atoms must be continually recycled through some 81 combination of the above mechanisms. The gas-phase reaction cycle described by Reactions R1 82 -R3 has generally been considered to be the dominant pathway for Br reformation during ODEs. 83 Thus, it has been assumed that the rate of ozone destruction can be estimated as Equation 1 84 (Hausmann and Platt, 1994;Zeng et al., 2006), or as Equation 2 if chlorine chemistry is 85 considered through Reaction R9 (Le Bras and Platt, 1995;Platt and Janssen, 1995). 86 (1) 87 -d O3 dt = 2(k 3 · BrO 2 + k 9 · BrO · ClO ) (2) 88 BrO + ClO à Br + OClO (R9) 89 However, these approximations assume that the ozone destruction rate is dominated by the BrO 90 + XO reaction, which in turn necessitates efficient gas-phase recycling of Br; therefore, a 91 relatively long bromine chain length would be required to account for observed rates of ozone 92 destruction. It is, however, possible that Br atoms are generated mostly by Br 2 photolysis, 93 followed by BrO termination, e.g. by R5, in which case a short gas-phase bromine radical chain 94 length would be implied. The chain length for any process depends on the rates of the 95 propagation relative to the production and destruction reactions (Kuo, 1986). In the stratosphere, 96 the Br/BrO catalytic cycle can have a chain length ranging from 10 2 to 10 4 (Lary, 1996). In the 97 troposphere, there is significantly less solar radiation and many more available sinks; thus, 98 radical chain lengths can be much shorter. For example, the chain length of the tropospheric 99 HO x cycle has been estimated to be ~ 4 -5 (Ehhalt, 1999;Monks, 2005), increasing to 10 -20 100 near the tropopause (Wennberg et al., 1998). The halogen radical chain lengths in the Arctic 101 troposphere have so far not been determined, thus, it is difficult to evaluate whether Equations I 102 and II are appropriate for estimating ozone depletion rates. 103 Modeling studies using typical Arctic springtime conditions to simulate ODEs have 104 concluded that ozone depletion cannot be sustained without considering the heterogeneous 105 recycling of reactive bromine on snow or aerosol surfaces (Michalowski et al., 2000;Piot and 106 Von Glasow, 2008). Michalowski et al. (2000) determined that the rate of ozone depletion in 107 their model was limited by the mass transfer rate of HOBr to the snowpack (i.e., the rate at which 108 Br is recycled through the heterogeneous mechanism) and that the depletion of ozone is nearly 109 completely shut down when snowpack interactions are removed. Piot  While it is evident that the reactions occurring on snow and aerosol surfaces are likely the 120 initial source of halogen species to the polar boundary layer and that heterogeneous bromine 121 recycling on these surfaces must be considered for HOBr and HBr, the relative importance of 122 gas-phase versus heterogeneous recycling of Br is uncertain. The goal of this work was to 123 investigate gas-phase Br atom propagation in terms of the bromine chain length in comparison to 124 the production of Br atoms through heterogeneous recycling and surface emissions of Br 2 and 125 BrCl. Here, we present results from our study using a multi-phase, zero-dimensional model 126 constrained with time-varying measurements of molecular halogens, O 3 , CO, NO, NO 2 , and 127 VOCs from the 2009 Ocean-Atmosphere-Sea Ice-Snowpack (OASIS) campaign in Barrow,128 Alaska. This work builds on the analysis presented in Thompson et al. (2015). By constraining 129 our model with observations, we were able to conduct an in-depth study of the halogen atom 130 recycling occurring under varying conditions that were observed during the campaign. 131 132 2 Experimental 133

Measurements and Site Description 134
The analysis presented herein utilizes observations conducted during the OASIS field 135 campaign that occurred during the months of February through April of 2009 in Barrow, AK. 136 The goal of the OASIS study was to investigate the chemical and physical processes occurring 137 within the surface boundary layer during ozone and mercury depletion events in polar spring. 138 This study resulted in the largest suite of simultaneous and co-located atmospheric measurements 139 conducted in the Arctic near-surface atmosphere to date, and represents a unique opportunity for 140 in-depth examination of a multitude of chemical interactions in this environment. 141 Atmospheric measurements were conducted from instrument trailers located near the 142  Thompson et al. (2015). We will describe the model only briefly here. High Iodine scenario incorporates a flux that is adjusted such that I 2 averages 0.5 pptv across the 180 simulated period. This results in IO mole ratios ranging from 5 to 10 pptv. 181 All gas-phase rate constants used in this model were calculated for a temperature of 248 182 K, consistent with ambient conditions in Barrow for the period simulated. Table 1  and CO at ten-minute time steps. All other gas-phase species are allowed to freely evolve. 193 Surface fluxes (represented as volumetric fluxes) are used for HONO and I 2 and are scaled to 194 J(NO 2 ) as a proxy for radiation as both of these species are likely to be produced 195 photochemically. Further discussion regarding HONO can be found in Thompson et al. (2015). also included, as well as mass transfer and dry deposition for certain species using the method 207 and mechanism of Michalowksi et al. (2000). It was intended to utilize strictly multiphase 208 chemistry to produce halogen radical precursors using this mechanism, however, the 209 heterogeneous production mechanisms could not reproduce observed Br 2 or Cl 2 from OASIS. 210 This likely reflects the complex but not fully understood condensed phase chemistry and physics 211 that leads to production of Br 2 (and Cl 2 ) (Pratt et al., 2013). Thus, Br 2 and Cl 2 concentrations 212 were fixed at the observed levels (see Thompson 2000) and assume a deposition velocity that is 10 times greater than for O 3 , leading to a k t of 225 1.67x10 -5 s -1 . We assume an equivalent k t for the oxidized nitrogen compounds (HNO 3 , HO 2 NO 2 , 226 HONO, and N 2 O 5 ). The rate of transfer out of the snowpack of emitted species, assumed here to 227 be limited only by vertical mixing, is estimated as 1.67x10 -5 s -1 , or equivalent to the rate of 228 deposition of the halogen acids (Michalowski et al., 2000). This is chosen as a best estimate of 229 vertical mixing because the deposition of halogen acids is likely to be only limited by the rate of 230 vertical mixing. The mass transfer coefficient of atmospheric species to and from the particle 231 phase is calculated as a first-order process as described in Jacob (2000). suggests that Br 2 should indeed be present and above the instrument detection limit during the 261 daytime. 262 Overall, the model captures the temporal cycle of HOBr well, but often overpredicts 263 daytime maximum mole ratios ( Figure 2C). A case in point is the much higher than observed 264 HOBr levels on 31 March, which corresponds to a similar overprediction of BrO. Observed Br 2 265 is relatively high on this day, and given that the model is forced to these observations, if the high 266 Br 2 is due to instrument artifact of HOBr conversion on the inlet, this could account for the 267 model discrepancy here. 268 HO 2 is essential for the heterogeneous recycling of bromine (via Reactions R5 -R7). 269 Therefore, it is important that our model accurately simulates HO 2 for this analysis. In Figure  270 2D we show a comparison of simulated HO 2 (black trace) and observed HO 2 from OASIS for 271 this period (red data), measured using a CIMS developed for peroxy radicals (Edwards et

Chain length 278
The ozone destruction cycle as described in Reactions R1 -R3 is a chain reaction 279 mechanism catalyzed by BrO x . The effectiveness of a catalytic cycle can be can be quantified by 280 considering the chain length, that is, the number of free radical propagation cycles per 281 termination or per initiation. We have not, until the OASIS2009 campaign, had the high quality 282 measurements available to enable a reliable estimation of the bromine radical chain length in the 283

Arctic. 284
The length of the chain in a radical propagation cycle is limited by termination steps that 285 destroy the chain carriers and result in relatively stable atmospheric species. Thus, the chain 286 length can be defined as the rate of propagation divided by the rate of termination. Alternatively, 287 the chain length can also be calculated using the rate of initiation. If the total bromine radical 288 population is at steady-state, the rate of initiation is equal to the rate of termination; thus, for 289 short-lived radical species, the two methods for calculating chain length should be approximately We used our model to calculate the chain length for bromine radical propagation across 294 the 7-days of the simulated period using both Method 1 and 2 as shown in Equations 5 and 6. 295 Because bromine radicals are generated photolytically, the chain length is calculated for daytime 296 only, defined here as approximately 7:00 to 20:00 Alaska Standard Time (AKST it should be noted that if we omit these reactions and consider only those that result in a net O 3 324 loss, it would be expected that the chain length would be shorter. Indeed, model simulations were 325 performed without these two terms and the determined chain lengths were on average 80% lower 326 than those presented here. BrO reaction with CH 3 OO is included in both the numerator and 327 denominator in Equation 5 because this reaction has two channels, one that propagates the Br 328 chain and one that terminates it. 329 In Figure 3, we present the results of these calculations for the Base Model, which show 330 that the two methods for calculating bromine chain length are in reasonably good agreement, 331 although there are small differences between the two methods throughout the time-series. This 332 agreement is a test of our basic understanding of the radical chemistry. The inset graph in Figure  333 3 shows a linear regression of the two methods for the chain length calculation. The coefficient 334 of determination (r 2 ) of 0.79 confirms the good temporal agreement between the two methods; 335 however, the slope of 0.72 indicates that Method 1 is generally higher than Method 2 throughout 336 (with some periods of exception). This offset reveals that either Method 1 is slightly 337 overestimating the chain length, or that Method 2 is underestimating it. The numerator is 338 identical in Equations 5 and 6, therefore, the denominator must be driving this discrepancy, with 339 either the denominator term in Method 1 too low or the denominator term in Method 2 too high 340 (or some combination thereof). If it's the case that the Method 1 denominator is too low, then it 341 must be concluded that there are important BrO x sink terms that are missing from the calculation. 342 If, however, the denominator of Method 2 is too high, this would imply that our measurements of 343 these BrO x precursors are too high, which, as discussed above, is a known likelihood at least for 344 the Br 2 measurements. 345 In Equation 6, we also do not include photolysis of organobromine compounds because 346 the rate of Br atom production from this pathway is small (e.g., ~ 100 molecules·cm -3 ·s -1 for 347 bromoform at mid-day) compared to Br atom production from Br 2 photolysis (~1.3x10 7 348 molecules·cm -3 ·s -1 at mid-day assuming 5 pptv of Br 2 ). Photolysis of bromine nitrate (BrONO 2 ) 349 is included, however, the prevalence of and production of this compound in the Arctic is highly days. This implies that, for these days, ozone depletion is strongly dependent upon initiation 377 processes, and most BrO radicals produced terminate the chain via reactions R5 and R10 in less 378 than two cycles. Reaction R12 will also efficiently terminate the chain, however, the relative 379 importance of R10 and R12 depend upon the relative abundances of BrO  recycling through the "bromine explosion", which emits Br 2 and BrCl from surface reactions, 384 must be of critical importance for ODEs occurring at the surface, as was previously concluded by 385 Piot and von Glasow (2008) and Michalowski et al. (2000). 386 A question to address regarding the relatively small chain length calculated for Br is to 387 what extent the chain length is dependent on NO 2 . As discussed in Thompson et al. (2015) and 388 further investigated in Custard et al. (2015), NO 2 at Barrow can be greater and more variable 389 than at very remote sites due to its proximity to anthropogenic emissions sources. We find that 390 the chain length calculation is relatively insensitive to NO 2 concentrations and so it is robust for 391 the range of conditions encountered at Barrow. As discussed in Custard et al. (2015), while NO 2 392 can inhibit the bromine chain through reactions R10 and R12 (i.e., decreasing the chain length), 393 enhanced NO 2 will also reduce available HO 2 , thereby decreasing the HO 2 available to terminate 394 the chain (i.e., increasing the chain length  No other adjustments were made to the model for these sensitivity runs. 410 Table 2  bromine sinks, such as aldehydes (e.g., propanal and butanal, which were free to evolve in our 420 model; HCHO and CH 3 CHO are fixed to observations) and HO 2 (see Thompson et al., 2015). 421 Iodine has a larger effect on the Br chain length. When Low Iodine is added to the "Br Only" 422 simulation, the chain increases from 1.17 -1.51 to 1.21 -1.67, primarily due to the very fast 423 cross-reaction between IO and BrO. Interestingly, there is no significant difference in the 424 calculated chain length between the "Br and Low Iodine" and the "Br and High Iodine" 425 simulations, potentially due to the increased competition for NO by I atoms. The addition of Cl 426 to the "Br and I" simulation imparts a slight decrease to the Br chain length. This may be 427 explained by the competition between BrO and ClO for reaction with NO and/or IO, as well as 428 the additional Br sinks in the presence of Cl chemistry. Regardless, overall there is more Br 429 available for reaction with O 3 when Cl is present due to the interhalogen reactions, thereby 430 increasing the rate of ozone depletion (see Thompson et al., 2015 for further discussion on ozone 431 depletion rates). 432 There are several conclusions that can be drawn from Figure 3 and Table 2: 1) there is a 433 distinct difference in bromine chain length between O 3 -depleted and non-depleted days with a 434 significantly larger chain length when ozone is present, and 2) for all simulations, the average 435 bromine chain is much shorter than expected (given that gas-phase recycling has, to date, been 436 assumed to be highly efficient). The chain length is greatest when ozone is present because 437 many of the species that propagate the Br chain (e.g., BrO, ClO, IO, and to a lesser extent OH At these levels, the BrO + IO reaction is more important than even BrO + BrO, accounting for 482 8% on average and a maximum of 39%. In the Base + Low Iodine scenario (not shown), the BrO 483 + IO reaction contributes 4%, which is at times comparable to BrO + BrO and greater than BrO 484 + ClO, even at the low IO concentrations in this simulation (~1 pptv). 485 The short gas-phase chain length calculated for bromine propagation indicates that there 486 are large reactive bromine (BrO x ) sinks terminating the chain reaction. Figure 7 presents the 487 rates of the most important BrO x termination reactions, with the y-axis expressed as the 488 cumulative rate of reaction. Here it can be seen that reaction of BrO with NO 2 is the dominant 489 sink for BrO x on non-ODE days for the conditions encountered at Barrow, while Br reaction with 490 CH 3 CHO is most important when O 3 is depleted. That HO 2 is a significant sink, and would be 491 more so in less anthropogenically-impacted Polar Regions, points toward the importance of 492 heterogeneous recycling through the bromine explosion mechanism. During ozone depletion, 493 such as the major event from days 26 -28 March ([O 3 ] < 5ppbv) when BrO is mostly absent, 494 CH 3 CHO becomes the primary sink term for Br, and HCHO is relatively more important. The 495 strength of the CH 3 CHO sink is much greater than is HCHO, as noted previously by Shepson et 496 al. (1996). Of note are the relatively similar magnitudes of the total rate of reaction of the 497 propagation and termination reactions shown in Figures 6 and 7, respectively, which of course 498 must be the case for a chain length near 1. This accounts for the short bromine chain length 499 determined here. This also implies then that to sustain elevated bromine radical concentrations 500 necessary to deplete O 3 requires an equally large Br 2 source (initiation) term, likely in the form 501 of a significant surface Br 2 flux. 502 503

Ozone loss rate 504
Since the chain length calculations seem to suggest a larger than expected contribution of 505 heterogeneous bromine recycling to Br atom production, to examine this further, we calculated 506 the rate of net ozone loss by Br and Cl in the Base Model using Equation 7 and compared this 507 rate to that estimated by Equation 2 (Platt and Janssen, 1995; Le Bras and Platt, 1995). 508 Additionally, the total simulated chemical ozone loss in the Base Model was calculated from 509 The method in Equation 2 assumes that the rate of ozone loss is equivalent to the rate at which 517 Br is regenerated through BrO reaction with itself and ClO (thus assuming efficient gas-phase 518 propagation and a long chain length), whereas Equation 7 accounts for all net ozone destruction 519 by Br and Cl, by correcting for those reactions that release a triplet oxygen atom and reform O 3 . 520 In other words, this method accounts for the fact that some BrO radicals react to terminate the 521 chain (and at steady state, an equivalent BrO x production rate is necessary). Figure 8A Figure 8B. Here it can be seen from the pink 530 data that halogen chemistry accounts for 99% of the total chemical O 3 loss under the conditions 531 simulated here. Importantly, the O 3 loss rate estimation presented in Equation 2 accounts for 532 only 44% of the total chemical O 3 loss rate (shown as the green data in Figure 8B). This 533 quantitatively expresses the conclusion that the gas-phase recycling of bromine is not as efficient 534 as previously considered and that it is often the case, for Barrow, that BrO x terminations must 535 often, through reactions R5 or R10, be followed by heterogeneous production of Br 2 through 536 condensed-phase reactions of HOBr and/or BrONO 2 . Indeed, the two methods for estimating 537 ozone loss rate agree the most when BrO, and thus, the gas-phase chain length are the greatest, or 538 in other words, when the 2k[BrO] 2 term, present in both Equations 2 and 5, is most important. A 539 very significant conclusion from this analysis is that the chemical O 3 loss rate is largely 540

Bromine atom production 547
If it is the case that heterogeneous recycling is of such importance, it may be that 548 Reaction R5 (BrO + HO 2 ) competes favorably with Reaction R3 (BrO + BrO). Panel A of 549 Figure 9 shows the rates of reactions R5 and R3. This plot demonstrates that the rate of reaction 550 of BrO with HO 2 is often of a comparable or greater magnitude than the BrO self-reaction, and 551 remains significant throughout the simulated period. Because the BrO + HO 2 reaction is of 552 primary importance for the bromine explosion mechanism, this result supports the hypothesis 553 that heterogeneous recycling may be equally or even more important than gas-phase recycling of 554 reactive bromine. 555 Given that the chain length is small, it must be that initiation is an important source of Br 556 atoms. To further examine the question of surface emissions versus gas-phase recycling, we 557 determined the rate of production of Br atoms via photolysis of Br 2 and BrCl (Equation 9), as 558 both are emitted from the surface as products of the bromine explosion, compared to the rate of 559 Panel B of Figure 9 compares the results of Equations 9 and 10, showing the total rate of Br atom 567 production separated into primary Br production (purple) and gas-phase Br regeneration 568 (orange); Panel C plots the fraction of total Br atom production that is due to primary production 569 from Br 2 and BrCl emissions. The majority of the time during this 7-day period Br atom 570 production from Br 2 and BrCl emissions (Equation 9) accounts for 40% or greater of the total, 571 and at times reaches over 90%. This explains both how ozone depletion can be rapid despite the 572 short calculated bromine radical chain length, as well as the difference found between the two 573 methods of estimating O 3 loss rate in Figure 7. It can be concluded from this analysis, then, that 574 the heterogeneous recycling of bromine can be of equal or greater importance to the evolution of 575 ODEs than gas-phase Br regeneration through radical recycling reactions. 576 577

Conclusions 578
The analysis presented here suggests that the gas-phase recycling of bromine species may 579 be less important than commonly believed, and we conclude that heterogeneous recycling, 580 primarily through the snowpack, is critical for the evolution of ODEs/AMDEs, consistent with 581 results by Piot and von Glasow (2008) and Michalowski et al. (2000). Indeed, the gas-phase 582 bromine propagation chain length is much shorter than expected, suggesting that much of the Br 583 present in the gas-phase is primary Br from surface emissions. Again note that our calculation of 584 chain length includes photolysis of BrO and BrO + NO, which do not result in net O 3 loss. Had 585 we omitted these two reactions, which we have found are in fact dominating the radical 586 propagation, the chain length would be, on average, 80% shorter. We find that between 40 -95% 587 of Br atoms are produced from surface emissions of Br 2 and BrCl. It is possible that iodine may 588 also play a potential role in facilitating heterogeneous bromine production through surface 589 emissions of IBr, though observations of this compound have not yet been achieved. 590 The production of Br 2 is quite complex and is dependent on many factors, including the 591 relative concentrations of bromide and chloride (among others), the availability of atmospheric 592 oxidants, such as ozone (e.g., Oum et al., 1998;Pratt et al., 2013)