Formation of Gas-Phase Bromine from Interaction of Ozone with

ARTICLE pubs.acs.org/JPCA

Formation of Gas-Phase Bromine from Interaction of Ozone with Frozen and Liquid NaCl/NaBr Solutions: Quantitative Separation of Surficial Chemistry from Bulk-Phase Reaction N. W. Oldridge and J. P. D. Abbatt* Department of Chemistry, University of Toronto, 80 St. George Street, Toronto, Ontario, Canada M5S 3H6 ABSTRACT: The formation kinetics of gas-phase bromine (Br2) from interaction of gas-phase ozone (O3) with frozen and liquid solutions of NaCl (0.55 M) and NaBr (largely from 1.7 to 8.5 mM) have been studied from -40 to 0 °C in a coated-wall flow tube coupled to a chemical ionization mass spectrometer. The reactive uptake coefficient for O3 is deduced from the product formation rate and then studied as a function of experimental conditions. In particular, for both the liquid and frozen solutions, we find that the uptake coefficient is inversely dependent on the gas-phase O3 concentration in a manner that is quantitatively consistent with both surface- and bulk-phase kinetics. The reaction is fastest on acidic media (pH of the starting solution down to 2) but also proceeds at an appreciable rate on neutral substrates. Above 253 K, the uptake coefficient increases with increasing temperature on frozen solutions, consistent with an increasing brine content. The similarity of the absolute magnitude and form of the kinetics on the frozen and liquid substrates suggests that the reaction on the frozen solution is occurring with the associated brine, and not with the ice bulk or a quasi-liquid layer existing on the ice. The implications of these results to bromine activation in the tropospheric boundary layer are made.

’ INTRODUCTION The reaction between aqueous O3 and aqueous bromide has been well-studied for many decades.1-3 In its simplest form, the mechanism is thought to be O3 þ Br- f OBr- þ O2

ðR1Þ

OBr- þ Hþ T HOBr

ðR2Þ

HOBr þ Br- f Br2 þ OH-

ðR3Þ

where the reaction proceeds faster in the presence of a general acid.4 The net reaction is commonly written as 2Br- þ O3 þ 2Hþ f Br2 þ O2 þ H2 O

ðR4Þ

While the reaction in solution is well-characterized, recent studies have focused on analogous surface chemistry that might occur with these reactants. In particular, Hunt et al. used aerosol chamber experiments, chemical kinetics modeling, and molecular dynamics simulations to show that the amount of bromine produced from O3 þ Br- with deliquesced sodium bromide aerosol in a chamber was an order of magnitude greater than predicted from the known aqueous- and gas-phase chemistry.5 These workers proposed that the majority of bromine r 2011 American Chemical Society

production was from a reaction between ozone and bromide ions at the air-aqueous interface. This is consistent with earlier results from Anastasio and Mozurkewich who had measured bromine production from the interaction of ozone with deliquesced NaBr aerosols.6 They had observed 100-1000 times the amount of gas-phase bromine that is expected when one considers only reactions in the bulk aerosol. This discrepancy in production rate was attributed to unknown reactions on the walls of their glass reaction flask, but heterogeneous reactions on the aerosol surfaces may also have been partly responsible. Clifford and Donaldson were the first to show direct evidence that a heterogeneous reaction occurred between gas-phase ozone and aqueous bromide, by measuring a pH increase at the airaqueous interface with the two reactants present.7 They showed that the rate of reaction was dependent on the concentration of ozone at the interface and that the kinetics exhibited behavior related to adsorption at the surface. It is also known that bromide ions in ices are reactive to ozone. In particular, a study by Oum et al. showed that bromine gas is released when frozen seawater is exposed to ambient levels of Received: January 4, 2011 Revised: February 7, 2011 Published: March 09, 2011 2590

dx.doi.org/10.1021/jp200074u | J. Phys. Chem. A 2011, 115, 2590–2598

The Journal of Physical Chemistry A

ARTICLE

gas-phase ozone at close to freezing temperatures.8 This reaction is of special interest to atmospheric bromine activation mechanisms because it does not require light to proceed. In particular, it allows for the release of bromine to the atmosphere during the Arctic winter, prior to sunrise. Thus, it may represent a mechanism for the formation of “seed” bromine required to initiate the springtime Arctic bromine explosion that ultimately leads to boundary layer ozone depletion.9 Using spectroscopic techniques sensitive to the surface of frozen sodium bromide solutions, Wren et al. have shown that the reaction proceeds with modification of the surface chemical structure.10 The kinetics of the surface-phase reaction are very much enhanced for the frozen solutions when compared to liquid solutions. In this paper, we describe results from a detailed kinetics study of the oxidation of bromide in frozen salt solutions by gas-phase ozone. The motivation is both to better understand Arctic bromide activation mechanisms but also to examine the rate of the surficial reaction in depth. In particular, although surfacephase chemistry has been demonstrated in this reaction system, no experimental study has as yet simultaneously measured the kinetics of the bulk- and surface-phase chemistry, thus directly determining the relative importance of each. Also, examination of the kinetics of this reaction under varying experimental conditions of temperature, substrate pH, and reactant concentrations has not been performed. With this information, we hope to be better able to determine whether the reactitn occurs on the ice surface, with the brine associated with the ice, or in the bulk. This is a challenging, yet general, goal for the larger ice chemistry community.

’ METHODS i. Experimental Section. Experiments were performed in a 90 torr, coated-wall flow tube coupled to a chemical ionization mass spectrometer (CIMS) for monitoring gas-phase composition.11 In particular, ozone was added in excess to the flow tube, by flowing it through a movable injector. The ozone was generated by passing a flow of oxygen and nitrogen over a Hg pen-ray lamp and then through a home-built ozone absorption monitor consisting of a second mercury lamp, a 254 nm interference filter, a 10 cm long absorption cell, and a photodiode. The flow rates of oxygen and nitrogen were each varied so that the total flow through the ozone generator and injector was 230 sccm. Nitrogen was also added directly to the flow tube at its upstream end. In this manner, 630-700 sccm of gas were routed through the flow tube. When the injector was pulled back, ozone flowing from the injector diffused through the carrier gas to the flow tube walls, allowing heterogeneous reactions to take place. The gas-phase products of such reactions were swept away by the carrier flow to the ion-molecule flow reactor of the CIMS. The reagent ion used was iodide, formed by passing a dilute flow of methyl iodide (CH3I) in nitrogen through a static charge eliminator containing Po-210 (NRD Inc.). The relevant ion-molecule chemistry at the pressure of the ion-molecule region (90 torr) occurs via clustering reactions

I- þ O3 T I- •O3

ðR5Þ

I- þ H2 O T I- •H2 O

ðR6Þ

I- þ Br2 T I- •Br2 -

ðR7Þ

or via analogous reactions proceeding via the I-•H2O reagent ion. These ions passed from the ion-molecule flow reactor to a quadrupole mass spectrometer that was operated in the negative ion mode for detection of iodide-molecule clusters.12 Methyl iodide was added to the carrier gas using a homemade permeation tube consisting of a Teflon tube filled with methyl iodide that was plugged at both ends by a Teflon plug held in place by a crimped metal tube. Although roughly 105 counts/s of signal at m/z = 127 (i.e., I-) represented the most abundant ion observed with a dry carrier gas, the most abundant ion (with roughly the same intensity) becomes m/z = 145 (i.e., I-•H2O) when the CIMS is connected to a flow tube containing either ice or liquid solutions. Also, when ozone is added to the flow tube under the highest gas-phase concentrations used, a large peak at m/z = 175 appears with intensity just below104 counts/s (i.e., I-•O3). Br2 was the sole product observed in this experiment, as three peaks at m/z = 285, 287, 289 corresponding to the isotopes of Br2 clustering with I-. A Br2 calibration was performed immediately after most experiments, with the reaction substrate still in the flow tube. With ozone still flowing but with the injector inserted so that the ozone was not exposed to the reactive surface, a small flow of Br2 was released to the flow tube from a fixed-volume manifold consisting largely of glass and Teflon. The pressure drop in the manifold over time was converted to the equivalent concentration of Br2 in the flow tube in molecules/cm3. During the course of experiments, the sensitivity varied from 2.5 10-9 to 5 1010 counts/s per molecule/cm3, and detection limits (S/N = 1, 1 s) of between 109 and 1010 molecules/cm3. It was important to calibrate under the conditions of the experiment because the intensity of the reagent ions (I-, I-•H2O) varied somewhat from run to run. Solutions of NaCl (99.0%, ACP Chemicals) and KBr (99.0%, ACP Chemicals) were prepared daily with typical concentrations of 0.55 M NaCl and 1.7 mM or 8.5 mM KBr. This concentration of NaCl is the same as that in seawater, and for bromide these values are 2 and 10 times, respectively. Solutions that were more concentrated in bromide were also prepared at times (see below). For those, the amount of NaCl added was decreased to keep the ionic strength the same. If applicable, aqueous HCl or NaOH was added to increase or decrease acidity. Solutions without HCl or NaOH added were found to have pH = 5.6 prior to freezing in the flow tube. To form an ice film, 4-7 mL of these solutions was used to wet the inner walls of a 1.5 cm i.d. glass tube insert, which was manually turned while inside the cool flow tube. The solution froze over a time period of 10-15 s. The flow tube was then left stationary in the cooling jacket, the injector was aligned in the center of the flow tube, and the system was closed. It was assumed that the flow of nitrogen became saturated with respect to the vapor pressure above ice in the first few centimeters of the flow tube—thus, ice loss further downstream (where reactions took place) was minimized. Indeed, at the end of each experiment, the flow tube was removed and ice was observed to still completely cover the inside of the flow tube. The ice films were fully transparent, with no substantial evidence for frostiness or cracks. The reproducibility of the update coefficients in this work (roughly (30%) is a measure of the maximum degree to which variability in the ice films from run to run affects the results. Also, experiments were repeated a second time on the same ice surface, and the yields did not systematically vary from run to run, even at the lowest temperatures where the surface will be most solidlike. Thus, we are confident that what is 2591

dx.doi.org/10.1021/jp200074u |J. Phys. Chem. A 2011, 115, 2590–2598


ARTICLE

Integrating, this gives kobs ¼

½Br2 ½O3 t

ðE3Þ

where reaction time has been calculated from the plug flow approximation using the reaction distance in the flow tube and carrier gas flow velocity (45-57 cm/s, based on flow conditions). In principle, the observed rate constant (i.e., rate coefficient), kobs, should be corrected for ozone concentration gradients that may exist within the flow tube.14,15 However, the kinetics in this reaction are sufficiently slow that this correction never amounted to more than 0.1% and so was neglected. This rate constant, kobs, can be related to the reactive uptake coefficient, γobs14 γobs ¼ Figure 1. Br2 signal as a function of injector position when exposed to a bromide-containing ice film.

measured here represents kinetics in/on ice media, not on desiccated salts. Note that this experimental procedure has been shown to form ice films that are smooth at the molecular level.13 Once the ion signals had stabilized subsequent to injection of ozone, the injector was pulled back a measured distance. The injector was left in this position until signals had stabilized again, after which the injector was again withdrawn. A slight time delay in appearance of Br2 signal sometimes occurred, probably due to conditioning of surfaces downstream of the injector tip. These steps were repeated for the entire length of the flow tube save the last 5-7 cm, since this is the area that is used to saturate the N2 flow with H2O vapor. The injector was always returned to its starting position, to confirm that the background had not changed during the course of an experiment. Typical signal intensity of Br2 as a function of injector distance is shown in Figure 1. Note that a background signal arises from small deposits of salt present on the tubes that connect the flow tube to the CIMS and from a short distance of ice that may be exposed to injector flow at the start of each experiment. However, we are only interested in the slope of these plots over the region where the ozone is exposed to the ice-halide surface. The linearity of this plot confirms that ozone is in excess in this experiment, i.e., it is not being significantly depleted by the heterogeneous reaction, and that the films are chemically uniform in an axial dimension. ii. Data Analysis. A goal of these experiments was to determine the reactive uptake coefficient for ozone that gives rise to Br2 production via reaction R4, i.e. γ¼

number of Br2ðgÞ molecules formed ðE1Þ number of collisions between O3ðgÞ and surface

To extract the kinetics, it is necessary to know the concentration of ozone in the gas phase and the number of molecules of Br2 molecules formed per ozone collision with the wall. In particular, assuming first-order kinetics for an ozone molecule that collides with the surface d½Br2 d½O3 ¼ ¼ kobs ½O3 dt dt

ðE2Þ

2rkobs ω

ðE4Þ

where r is the inner radius of the glass insert in the flow tube and ω is the mean molecular speed of ozone in the gas phase. When all uncertainties are taken into consideration, the uptake coefficients are accurate to better than (50%. Heterogeneous reactions with ozone on both solid and liquid substrates are frequently observed to obey Langmuir-Hinshelwood-like kinetics, via a mechanism between a surface-bound species and an adsorbed species that is in equilibrium with the gas phase. (See a compilation of these reactions in McCabe and Abbatt.16) Assuming that mass accommodation does not limit the kinetics, Ammann et al. describe the corresponding surficial uptake coefficient as17 1 þ K½O3 g 1 ¼ ωσ γs 4Kks I

ðE5Þ

where σ is the surface area taken up by one adsorbed ozone molecule, K is the Langmuir adsorption equilibrium constant of O3, [O3]g is the concentration of ozone in the gas phase, and ksI is the first-order surface-phase rate constant for the O3 þ Brreaction. ksI = ksII [Br-]s, where [Br-]s is the concentration of bromide at the surface and ksII is the second-order surface-phase rate constant. By contrast, the uptake coefficient describing loss of a gasphase molecule within the liquid phase can be described by the following expression18 pffiffiffiffiffiffiffiffiffi 4RTH Dk1 I ðE6Þ γl ¼ ω where R is the gas constant, T is the temperature, H is the Henry’s law constant for the gas-phase molecule in the liquid phase, D is its diffusion coefficient in the liquid phase, and klI is the liquidphase rate constant for the first-order reaction. Br2 production is treated as a pseudo-first-order reaction, with klI = klII[Br-]l. Using emulsions, Koop et al. showed that ice made from solutions containing NaCl and NaBr will have a liquid component to temperatures below -50 °C.19 It is then possible that the ices, made from NaCl/KBr solutions, will contain a liquid component down to the coldest temperature at which experiments were performed (-40 °C) with, albeit, increasingly less liquid available for reaction as the temperature decreases. Of course, the presence of the flow tube substrate may alter the phase of the films relative to what was observed by Koop et al. 2592



Figure 2. γ vs temperature for sea ice with [Br-] = 8.4 mM, pH = 5.6, [O3] = 1.7 1014 molecules/cm3. [NaCl] = 0.55 M for this figure and all subsequent figures.

For reactions that may occur both on a surface and in the bulk of a medium, the observed reaction probability is the sum of the individual reaction probabilities for the bulk and surface processes. For example, see Crowley et al.20 and references therein. That is, the system may have two different media available for reaction: (i) a solid, liquid, or quasi-liquid-like surface and (ii) bulk liquid existing as a brine in pockets or veins within the ice matrix. We exclude the possibility that reaction occurs in the bulk solid phase, due to low diffusion coefficients and the low solubility of ions in ice.

’ RESULTS AND DISCUSSION i. Dependence of Kinetics on Temperature. Uptake coefficients were measured on frozen acidic surfaces to determine whether the variation in substrate structure would affect the reaction kinetics. For the 8.5 mM bromide solution (Figure 2), there is a steep increase in the uptake coefficient above -30 °C, and the uptake coefficient remains invariant with temperature under colder conditions. Results for 1.7 mM Br- frozen solutions (not shown) are similar with the increase occurring above -20 °C. Although there will be liquid content in the frozen solution to very low temperatures, note that at temperatures below about -22 °C, i.e., the point at which sodium chloride begins to crystallize out of the aqueous portion as NaCl 3 2H2O, there will be a sudden drop in the liquid content of the solution. Thus, at high temperatures where γobs rises with temperature it is possible that the data are consistent with liquid-phase chemistry dominating the Br2 production. At lower temperatures, where γobs remains constant with temperature, the data are more consistent with chemistry at the surface of a solid phase. For the plateau region at low temperatures, the uptake coefficients for two different bromide concentrations, 8.4 and 1.7 mM, are very similar, 2 10-9 and 3 10-9, respectively. It can be concluded that for this range of concentrations, the concentration of the original solution does not significantly affect the number of surface sites available for reaction; i.e., if the ice surface is saturated with bromide when a 1.7 mM solution is frozen, it will certainly be saturated when an 8.4 mM solution is frozen, and no concentrations effect will be observed.

ARTICLE

Figure 3. γ vs [O3] for frozen solution with [Br-] = 8.4 mM, pH = 1.97, T = -20 °C.

ii. Effect of Ozone Concentration. The dependence of γobs on the gas-phase concentration of ozone is shown in Figure 3. A strong inverse dependence in the uptake coefficient is observed that is qualitatively consistent with a Langmuir-Hinshelwoodlike mechanism.17 In particular, as the gas-phase ozone concentration increases, more and more of the surface sites become occupied. Ozone molecules that collide with the surface will then have a lower probability of adsorbing and reacting. Thus, the uptake coefficient decreases as ozone increases to an asymptotic value of zero in this regime. However, in Figure 3, it is clear that the uptake coefficient in the high ozone concentration limit is above zero. To quantify the asymptote, note that at high [O3]g, the numerator in eq E5 1 þ K[O3]g ≈ K[O3]g. Thus

ωσK½O3 g 1 ωσ ¼ ½O3 g II II γs 4Kks ½Br s 4ks ½Br- s

ðE7Þ

which gives the following equation for γs, linear with [O3]g-1: γs ¼

4ks II ½Br- s 1 ½O3 g ωσ

ðE8Þ

A plot of γobs vs 1/[O3], shown in Figure 4, has a positive intercept on the vertical axis, corresponding to γobs = 1.1 10-8. For reasons to be developed more fully below, we believe that this uptake coefficient arises from bulk-phase chemistry, whereas surface-phase chemistry becomes of greater relative importance as the ozone concentration decreases. To test whether the lower [O3]g results conform to Langmuir-Hinshelwood kinetics, the asymptotic, bulk-phase uptake coefficient is subtracted from each data point in Figure 3, and the data are then plotted in the following form ωσK½O3 g 1 ωσ ¼ þ II γs 4Kks ½Br s 4Kks II ½Br- s

ðE9Þ

which is equivalent to 1 ¼ A þ AK½O3 g γs 2593

ðE10Þ



ARTICLE

Figure 4. γ vs 1/[O3] for frozen solution with [Br-] = 8.4 mM, pH = 1.97, T = -20 °C. The linear fit of the data is γ = (4.2 ( 0.4) 106/[O3] þ (1.1 ( 0.4) 10-8. Note that we have only used the high [O3] points in this plot.

Figure 6. γ vs pH for frozen solution with [Br-] = 8.4 mM, T = -20 °C, [O3] = 1.5 1014 molecules/cm3.

and the lower temperature. However, the nature of the adsorption mechanism of these species to ice may be quite different than with ozone. By contrast, somewhat smaller values of about 10-16 cm3/molecule have been measured for ozone adsorption to ice surfaces containing phenanthrene between -10 and -30 °C.22 Note, however, that there are salts in the frozen substrates which may affect surface properties making them more liquidlike. iii. Effect of Acidity. The kinetics are dependent on the acidity of the solution from which the ice film was formed (Figure 6). This is in line with bulk-phase studies that have shown that the reaction progresses faster under acidic conditions.2,4 In particular, Liu et al. proposed a mechanism for HOBr production that is expanded from eq R1 to eq R24 O3 þ Br- T BrOOO-

ðR8Þ

BrOOO- þ Hþ f HOBr þ O2

ðR9Þ

BrOOO- þ H2 O f HOBr þ O2 þ OH-

Figure 5. 1/ γ vs [O3] for frozen solution with [Br-] = 8.4 mM, pH = 1.97, T = -20 °C. The linear fit of the data is 1/γ = (2.3 ( 0.3) 10-7 [O3] þ (0.16 ( 0.41) 107.

where A¼

ωσ 4Kks II ½Br- s

ðE11Þ

A plot of γobs-1 versus ozone concentration is linear (Figure 5), confirming the Langmuir-Inshelwood functional relationship, with a slope = AK = 2.3 10-7 cm3/molecule and an intercept A = 0.16 107. From these, a Langmuir adsorption constant K ∼ 10-13 cm3/molecule can be calculated at -20 °C. By comparison, note that K is on the order of 10-10-10-11 cm3/molecule for HNO3 and HCl adsorption to ice surfaces between -34 and -59 °C.21 The larger values for these molecules are consistent with these strong acids adsorbing more strongly than O3 to ice

ðR10Þ

where eq R9 proceeds faster in acidic environments. In the data, the acidity dependence is relatively weak with γobs changing by less than a factor of 3 as the acidity of the starting solution increased from pH = 8.4 to 2.7. This is consistent with eq R8 being largely rate limiting, and a small effect arising from acidity effects in eq R9. Note that Hunt et al. propose a mechanism where charge transfer between bromide and ozone occurs whereby the kinetics would again not be driven by acidity.5 The experiments presented here do not allow us to differentiate these pathways; however, it is not necessary to invoke this mechanism to explain the data. Although experiments at low ozone concentrations were challenging to perform, we were nevertheless interested in determining whether the same acidity dependence is apparent in the kinetics under these conditions versus those at [O3]g = 1.5 1014 molecules/cm3 that were employed for the data in Figure 6. Indeed, for experiments conducted at 2.0 1013 molecules/cm3 where the surface-phase reaction is dominant, the uptake coefficient for a pH 5.6 frozen solution was also just slightly lower (1.8 10-7) than for a pH 1.9 solution (2.7 10-7), for solutions of 86 mM bromide concentration. 2594



ARTICLE

Figure 7. log(γ) vs log([Br-]) on sea ice at -20 °C, [O3] = 1.7 1014 molecules/cm3, pH = 1.97. The linear fit of the data is log γ = (0.72 ( 0.04)(log [Br-]) - (6.08 ( 0.05).

Figure 8. γ vs [O3] for aqueous solution with [Br-]l = 8.6 mM, pH = 1.95, T = 0 °C.

iv. Effect of Bromide Concentration. If the reaction of ozone with bromide is occurring in a bulk liquid phase, then the reactive uptake coefficient would be expected to scale with the square root of the bromide concentration, according to eq E6. A surface-phase process that is dominated by reaction with surficial bromide in a Langmuir-Hinshelwood-like manner would vary linearly if the surface is not saturated with bromide; if it is saturated, then there would be no dependence. When the log of the observed uptake coefficient for one set of gas-phase ozone and pH conditions is plotted versus the log of the bromide concentration, the slope of the line of best fit is 0.72 (see Figure 7) suggestive of a process dominated by a bulk-phase reaction, but with a surface component as well. The ozone partial pressure used for these measurements is relatively high, about 1.7 1014 molecules/cm3, in the regime where Figure 3 illustrates the reaction is dominated by the bulk-phase reaction but still has some surficial character as well. As in section iii above, we were interested in the relationship between the uptake coefficient and the bromide concentration under low gas-phase ozone concentrations, where the surface reaction dominates. To do this, pH 1.9 frozen solutions were used, and it was observed for O3 concentrations of 2.0 1013 molecules/cm3 that the uptake coefficient increased from 1.4 10-7 to 2.7 10-7 as the bromide concentration of the solution increased from 8.5 to 85 mM. This weak dependence on bromide concentration in the surface-reaction-dominated regime is consistent with the observations made below -30 °C (see section i) where we inferred that the surface is saturated with bromide ions for frozen solutions from 1.7 to 8.5 mM composition. These measurements at -20 °C at low ozone concentrations confirm this weak relationship and indicate that the surface is also approaching saturation with bromide ions for 8.5-85 mM frozen solutions. Note that Wren et al. (2010) also observed a saturation in the surface-phase kinetics on frozen sodium bromide surfaces at quite low concentrations of bromide.10 v. Effect of Iodide. Enami et al. found that micromolar amounts of iodide catalyzed the oxidation of bromide in aqueous nanodroplets exposed to ozone.23 They proposed that HOI is formed in sufficiently acidic environments and that HOI goes on

I- þ O3 þ Hþ T HOI þ O2

ðR11Þ

HOI þ Hþ þ 2Br- f H2 O þ IBr2 -

ðR12Þ

IBr2 - T I- þ Br2

ðR13Þ

to oxidize bromide:

These workers note that the reaction proceeds in this manner due to the relative inertness of O3 to bromide (klII(O3 þ Br-) = 205-248 M-1 s-1, at 25 °C) compared to its reactivity with iodide (klII(O3 þ I-) = 1.2 109 M-1 s-1 at 25 °C).3,4 To test this mechanism, experiments were performed with a solution containing [NaCl] = 0.55 M, [Br-] = 8.4 mM, and [I-] = 54 μM at T = -20 °C, with pH = 5.6 and [O3] = 2 1014 molecules/cm3. The uptake coefficient measured for this frozen solution was γobs = 1.5 10-8, which is at most 50% higher than the mean γobs value for the equivalent noniodized solutions. The “iodide-free” solutions contain no more than [I-] = 3.9 μM, based on the stated impurity levels in the commercial NaCl. Within our experimental uncertainties this is not a significant increase. Based on this one result, we believe that the kinetics in the experiments are driven by bromide ions alone and that iodide does not significantly accelerate the chemistry. vi. Comparison with Aqueous-Phase Chemistry. Through the results presented so far, it is clear that chemistry with an aqueous component of the frozen solutions may be responsible for much of the reactivity observed. To validate this, the frozen solution results were compared to those obtained using a liquid solution substrate held at 0 °C. The same experimental setup was used as in the ice experiments except that the ice coating on the inside of the glass tube insert was replaced with a 20 cm long glass cylinder cut in half lengthwise and with a semicircular cap at each end, i.e., a “boat” that could contain liquid. As with the ice experiments, pulling the injector back repeated times over the course of the experiment gave the same product signal intensity each time indicating that the composition of the liquid does not change significantly during the experiments. In the data analysis the amount of surface area presented by the liquid surface was taken into account. To test this experimental approach, two experiments were performed at [Br-] = 8.4 mM, pH = 5.57, T = -20 °C 2595



Figure 9. γ vs 1/[O3] for aqueous solution with [Br-]l = 8.4 mM, pH = 1.95, T = 0 °C. The linear fit of the data is γ = (0.78 ( 0.14 107)/[O3] þ (7.66 ( 1.8) 10-8. Note that we have only used the high [O3] points in this plot.

Figure 10. 1/γ vs [O3] for aqueous solution with [Br-]l = 8.6 mM, pH = 1.95, T = 0 °C. The linear fit shown is 1/γ = (1.4 ( 0.2 10-7)[O3] þ 0.06 ( 0.22 107.

with [O3] = 2 1014 molecules/cm3. The flow tube coated with ice gave a value of γ = 1.1 ( 0.3 10-8, and two experiments with an identical solution frozen in the boat gave γ = 1.9 ( 1.1 10-8; i.e., within experimental uncertainty uptake coefficients in the boat are the same as in the flow tube. The experimental results with the liquid solutions are shown in Figures 8-12. Figure 8 illustrates that the kinetics exhibit the same functional form as those on frozen solutions, implying that Langmuir-Hinshelwood kinetics prevail for the O3 þ Brreaction on liquid surfaces in confirmation of the work by Clifford and Donaldson.7 The value of the uptake coefficient asymptote at high ozone concentrations in Figure 8 is evaluated in Figure 9, where the same data analysis process employed above was used. We believe this asymptote (7.7 10-8) represents reactivity in the bulk liquid. To test this assumption, the expected uptake coefficient in this regime can be calculated using eq E6 and literature values for the

ARTICLE

Figure 11. γ vs pH on aqueous solution with [Br-]l = 8.4 mM, [O3] = 1.7 1014 molecules/cm3, T = 0 °C.

Figure 12. log γ vs log([Br-]l) for aqueous solutions at 0 °C, [O3] = 1.7 1014 molecules/cm3, pH = 1.97. The fit shown is log γ = (0.56 ( 0.05) log([Br-]) - 7.84 ( 0.06.

liquid-phase rate constant for ozone reacting with bromide, the aqueous O3 diffusion coefficient, and the Henry’s law constant of ozone in solution. In particular, we use the data from the JPL evaluation to estimate the Henry’s law constant to be 0.020 M/ atm at 0 °C, where the Schumpe parameters have been employed to evaluate the effect of salts on the solubility in aqueous solution.24 For the rate constant in solution, the room temperature rate constant and an activation energy of 37 kJ/mol were used to estimate the kinetics at 0 °C.3 Finally, the liquid-phase diffusion coefficient is taken to be 8.9 10-6 cm2/s, as calculated by using the Einstein relation to adjust the measured room temperature diffusion coefficient of ozone in water.25 Using this approach, the uptake coefficient at 273 K of a 8.5 mM bromide solution is calculated to be 1.1 10-7. Given the uncertainties involved, this value is in remarkable agreement with the measured value of 7.7 10-8. This confirms that the asymptote in the plot of the uptake coefficient versus ozone concentration is indeed a direct measurement of the bulk reactivity. Figure 10 is linear, validating the Langmuir-Hinshelwoodlike relationship at low ozone on liquid solutions. From eqs 2596


The Journal of Physical Chemistry A E9-E11, the slope = AK = 1.4 10-7 cm3/molecule and the intercept = A = 0.06 107. Thus, the Langmuir adsorption equilibrium constant K ∼ 10-13 cm3/molecule for this saltwater surface, equal to the value obtained on the frozen solution. This suggests that ozone is also adsorbing to a liquid surface on the frozen solution. For reference, K = 5 10-16 cm3/molecule has been measured for ozone adsorption to an aqueous surface containing anthracene.26 This value may be smaller than ours given that the measurements were done at room temperature. Also, our substrates are salt solutions and not pure water. As with the frozen solutions, acidity was found to increase γobs on the liquid substrate as it did on the ice films (Figure 11), with the pH dependence stronger in aqueous solution at 0 °C than with ice at -20 °C. Here, γobs increases by a factor of 7 as acidity increased from pH 5.5 to 2.5. γobs for the ice showed an increase of only 50% over this range. It is possible that Br2 production is more proton-limited in aqueous solution than in the highly concentrated liquid portion of ice. Finally, the relationship between γ and [Br-] (Figure 12) is fully consistent with a liquid-phase reaction. The slope of this log-log plot is 0.56, and the predicted value (eq E6) is 0.5.

’ CONCLUSIONS i. Reaction Mechanism. The experimental results allow us to draw conclusions about both the nature of the bromide oxidation reaction—i.e., whether it is occurring in the surface versus the bulk—and also on whether the reaction with the frozen solutions occurs with a brine or with an ice surface that has a reactivity different from a brine. To summarize the results, note that the reaction kinetics are consistent with both a bulk-phase reaction and a surface-phase reaction operating simultaneously. In particular, the dependence of the kinetics on ozone concentration indicate that bulk-phase kinetics dominate at high ozone concentrations, whereas surfacephase kinetics is of greater relative importance at low ozone. Also, we find that the rate of the bulk-phase chemistry at high ozone on liquid solutions is in quantitative agreement with the predicted rate using literature parameters for the solubility and diffusivity of ozone in solution and of its reactivity with bromide. At low ozone concentrations, the reactive uptake coefficient for ozone increases with decreasing ozone concentrations because the surface is unsaturated with adsorbing ozone molecules in this regime. At high ozone, the surface kinetics become of less importance relative to the bulk-phase kinetics because the surface becomes saturated with ozone. Interestingly, with the reaction between iodide and ozone being so fast, the conclusions of Rouviere et al. are that the bulk chemistry dominates surface chemistry when aqueous iodide particles are exposed to ozone.27 These conclusions are in agreement with previous work in the literature that has suggested that a surface-phase reaction is more important than the bulk under ambient conditions5 and with reports that the surface-phase chemistry is operative.7 However, this is the first time that an experiment has simultaneously quantified the importance of the two mechanisms as a function of the reaction conditions. Indeed, in our opinion it is the clearest example of the quantitative separation of bulk- and surface-phase kinetics demonstrated for any multiphase reaction studied to date. The fact that ozone participates in a surface-phase reaction is not surprising given the very large number of ozone reactions that are well-described by Langmuir-Hinshelwood kinetics.

ARTICLE

However, this is the first demonstration that ozone also adsorbs to the surfaces of ionic liquids, despite it having a large mass accommodation coefficient into such surfaces.27 In a recent compilation of these kinetics, our group demonstrated that there is a remarkable similarity in the kinetics that extends for such ozone reactions across a wide range of reactants and surfaces.16 In particular, we illustrate that the pseudo-first-order rate constant for loss of surface species (bromide in this case) via reaction with gas-phase ozone is between 10-3 and 10-1 s-1 at room temperature for the following: PAH loss on pure organic substrates and on aqueous surfaces, loss of alkene functional groups within self-assembled monolayers bound to a surface, oxidation of soot and mineral dust surfaces, and oxidation of frozen hexadecene and oleate coated particles. Similarly, the value of ksI for the 0 °C data can be estimated to be ∼10-3 s-1 by using the experimental value for A in eq E11, where the crosssectional area of the absorbed ozone molecule is taken to be 5 10-15 cm2. As concluded earlier,16 the similarity of this value to the equivalent rate constant for such a diverse set of reactants and surfaces implies that the rate of the surface chemistry is in large part determined by the transformation of the surface adsorbed ozone molecule into a more reactive form; i.e., the kinetics are tied more to the transformation of the ozone molecule than to its subsequent reaction with a surface species. A major goal of this work was to determine the phase in which the oxidation chemistry occurs with the frozen solutions. Within the ice chemistry community, this has proven to be highly challenging because chemistry can occur in bulk ice, on the surface of ice which may or may not have liquidlike properties, in the brine associated with the frozen solution, or on the surface of the brine. Given the very strong similarity between the form of the kinetics measured on true aqueous solutions and on frozen solutions (i.e., qualitatively similar dependencies of the kinetics on pH, bromide concentration, ozone concentration), we believe that the chemistry—whether at the surface or in the bulk—occurs with a brine formed when freezing a solution. Of course this does not rule out the possibility that the chemistry is occurring with the ice surface that has a liquidlike structure and contains solute ions (a “quasibrine”), but the kinetics on such a surface would have to be coincidentally the same as those with the true brine that will be present in equilibrium with the ice. Thus, we do not feel the need to invoke the concept of such a new phase to explain our data. Although this is the first experimental verification that brine chemistry dominates, this is not a surprising conclusion. It is wellknown that solutions exclude solutes into an increasingly concentrated brine as freezing occurs. Given that the frozen solutions appear to react as a brine, then it is possible that this excluded brine layer surrounds the ice crystals that form and is thus exposed to ozone in the gas phase. Also, the temperature at which most of these studies were conducted, -20 °C, is not so low that the brine will have fully frozen. Indeed, the studies conducted at considerably lower temperatures (see Figure 2) suggest that the chemistry will be less aqueous-like under these conditions. At a quantitative level, it is important to assess whether the kinetics of the reaction are different between those on aqueous and frozen substrates. In particular, it is known that solutes are excluded when solutions freeze, and so it is possible that the rates of the reaction would increase with an ice substrate, if the surfaces are not already saturated with bromide. The results indicate that under low ozone conditions, where the surface reaction dominates on ozone-unsaturated surfaces, the uptake coefficients are roughly the same (to within a factor of 3) on ice than on aqueous 2597


The Journal of Physical Chemistry A solutions. Their similarity is another indication that the reaction on the frozen solutions is indeed occurring on a brine coexisting with ice that is similar in reactivity to the surface of an aqueous solution. In the limit of high ozone conditions where bulk-phase kinetics prevail, the kinetics are faster on the liquid substrate by a factor of 7. Part of the temperature dependence for the bulkphase reaction will arise from the activation energy in the rate constant between dissolved bromide and ozone. However, it is also not clear whether the full surface of the frozen solution is reactive, i.e., covered with brine. When comparing frozen- to aqueous-phase reactivity, note that Wren and Donaldson report a factor of 60 increase in the kinetics when moving from solution to ice, where the experiments were done with high ozone and with a technique sensitive to the uppermost layers of the substrate.10 Although these results appear at odds with those presented here, comparison between the two data sets is complicated by the fact that our measurements at high ozone concentrations are not sensitive to the surface reaction, being dominated by the bulk reaction instead. Also, our work was done with mixed chloride and bromide solutions, whereas that of Wren and Donaldson used only bromide. ii. Atmospheric Implications. There are implications for bromine activation chemistry on both ice substrates and marine aerosol. On ice, the study of Oum et al. illustrated that this reaction may provide a source of gas-phase bromine to the atmosphere that can arise in the absence of sunlight.8 That is, this dark reaction may provide some of the gas-phase bromine that is the seed for the autocatalytic release of bromine that arises through HOBr uptake to bromide-containing ice surfaces such as frozen seawater or snow to which sea-salt aerosol has been deposited. Thus, it is important to know the kinetics of this reaction on ice. In particular, for such a slow reaction, the ratedetermining step will not be mass transfer from the atmosphere to the ice-covered surface as it is for much faster reactions (see, e. g., Huff and Abbatt28). Quantitatively, a rough comparison to the results of Oum et al., who conducted their study at low ozone conditions where the surface reaction dominates and with solutions of seawater composition, can be done. From the Br2 production rates of Oum et al. and the ozone concentration, we can infer an ozone uptake coefficient of 2 10-9 for their study using eq E2. In our case, we start with the uptake coefficient corresponding to the A value measured on ice at -20 °C for pH 1.9 frozen solutions with 8.5 mM bromide composition, i.e., γ = 6 10-7 (with a roughly a factor of 3 uncertainty). This value may be larger than that of Oum et al. for a number of reasons, including the higher acidity and the higher bromide concentrations. Also, the solution studied by Oum et al. was close to seawater composition whereas our work was with chloride/bromide mixtures. Comparison to the results of Wren and Donaldson10 on pure bromide solutions is somewhat more involved given that these workers are only sensitive to the reaction in the top surface layers and were operating at high ozone concentration, where our experiments are only sensitive to the bulk reaction. For liquid aerosol in the atmosphere, the implications for the kinetics of this reaction under ambient conditions is that use of bulk-phase kinetic parameters will significantly underestimate the rate of the reaction. Thus, we are in agreement with the conclusions of Hunt et al. who inferred that a surface reaction dominates reactivity on aqueous bromide-containing particles.5 We confirm that the surface-phase kinetics are significantly faster

ARTICLE

than the bulk-phase kinetics for low, atmospheric ozone concentrations on aerosol similar in composition to marine particles.

’ AUTHOR INFORMATION Corresponding Author

*E-mail [email protected].

’ ACKNOWLEDGMENT This work was supported by NSERC. ’ REFERENCES (1) Taube, H. J. Am. Chem. Soc. 1942, 64, 2468. (2) Haruta, K.; Takeyama, T. J. Phys. Chem. 1981, 85, 2383. (3) Haag, W. R.; Hoigne, J. Environ. Sci. Technol. 1983, 17, 261. (4) Liu, Q.; Schurter, L. M.; Muller, C. E.; Aloisio, S.; Francisco, J. S.; Margerum, D. W. Inorg. Chem. 2001, 40, 4436. (5) Hunt, S. W.; Roeselova, M.; Wang, W.; Wingen, L. M.; Knipping, E. M.; Tobias, D. J.; Dabdub, D.; Finlayson-Pitts, B. J. J. Phys. Chem. A 2004, 108, 11559. (6) Anastasio, C.; Mozurkewich, M. J. Atmos. Chem. 2002, 41, 135. (7) Clifford, D.; Donaldson, D. J. J. Phys. Chem. A 2007, 111, 9809. (8) Oum, K. W.; Lakin, M. J.; Finlayson-Pitts, B. J. Geophys. Res. Lett. 1998, 25, 3923. (9) Vogt, R.; Crutzen, P. J.; Sander, R. Nature 1996, 383, 327. (10) Wren, S. N.; Kahan, T. F.; Jumaa, K. B.; Donaldson, D. J. J. Geophys. Res., [Atmos.] 2010, 115, 8. (11) Frinak, E. K.; Abbatt, J. P. D. J. Phys. Chem. A 2006, 110, 10456. (12) Thornton, J. A.; Braban, C. F.; Abbatt, J. P. D. Phys. Chem. Chem. Phys. 2003, 5, 4593. (13) Abbatt, J. P. D.; Bartels-Rausch, T.; Ullerstam, M.; Ye, T. J. Environ. Res. Lett. 2008, 3, 5. (14) Howard, C. J. J. Phys. Chem. 1979, 83, 3. (15) Poschl, U.; Canagaratna, M.; Jayne, J. T.; Molina, L. T.; Worsnop, D. R.; Kolb, C. E.; Molina, M. J. J. Phys. Chem. A 1998, 102, 10082. (16) McCabe, J.; Abbatt, J. P. D. J. Phys. Chem. C 2009, 113, 2120. (17) Ammann, M.; Poschl, U.; Rudich, Y. Phys. Chem. Chem. Phys. 2003, 5, 351. (18) Gas-Liquid Reactions; Dankwerts, P. V., Ed.; McGraw-Hill: New York, 1970. (19) Koop, T.; Kapilashrami, A.; Molina, L. T.; Molina, M. J. J. Geophys. Res., [Atmos.] 2000, 105, 26393. (20) Crowley, J. N.; Ammann, M.; Cox, R. A.; Hynes, R. G.; Jenkin, M. E.; Mellouki, A.; Rossi, M. J.; Troe, J.; Wallington, T. J. Atmos. Chem. Phys. 2010, 10, 9059. (21) Cox, R. A.; Fernandez, M. A.; Symington, A.; Ullerstam, M.; Abbatt, J. P. D. Phys. Chem. Chem. Phys. 2005, 7, 3434. (22) Kahan, T. F.; Donaldson, D. J. Environ. Res. Lett. 2008, 3, 6. (23) Enami, S.; Vecitis, C. D.; Cheng, J.; Hoffmann, M. R.; Colussi, A. J. J. Phys. Chem. A 2007, 111, 8749. (24) Chemical Kinetics and Photochemical Data for Use in Atmospheric Studies; Sander, S. P. e. a., Ed.; JPL Publication 06-2, 2006; Vol. Evaluation 15. (25) Johnson, P. N.; Davis, R. A. J. Chem. Eng. Data 1996, 41, 1485. (26) Mmereki, B. T.; Donaldson, D. J. J. Phys. Chem. A 2003, 107, 11038. (27) Rouviere, A.; Sosedova, Y.; Ammann, M. J. Phys. Chem. A 2010, 114, 7085. (28) Huff, A. K.; Abbatt, J. P. D. J. Phys. Chem. A 2000, 104, 7284.

2598


Formation of Gas-Phase Bromine from Interaction of Ozone with

Recommend Documents