The effect of cations on NO2 production from the photolysis of aqueous thin water films of nitrate salts.

The photochemistry of nitrate ions in bulk aqueous solution is well known, yet recent evidence suggests that the photolysis of nitrate may be more eﬃcient at the air–water interface. Whether and how this surface enhancement is altered by the presence of different cations is not known. In the present studies, thin aqueous films of nitrate salts with different cations were deposited on the walls of a Teflon chamber and irradiated with 311 nm light at 298 K. The films were generated by nebulizing aqueous 0.5 M solutions of the nitrate salts and the generation of gas-phase NO 2 was monitored with time. The nitrate salts fall into three groups based on their observed rate of NO 2 formation ( R NO 2 ): (1) RbNO 3 and KNO 3 , which readily produce NO 2 ( R NO 2 4 3 ppb min (cid:2) 1 ), (2) Ca(NO 3 ) 2 , which produces NO 2 more slowly ( R NO 2 o 1 ppb min (cid:2) 1 ), and (3) Mg(NO 3 ) 2 and NaNO 3 , which lie between the other two groups. Neither differences in the UV-visible spectra of the nitrate salt solutions nor the results of bulk-phase photolysis studies could explain the differences in the rates of NO 2 production between these three groups. These experimental results, combined with some insights from previous molecular dynamic simulations and vibrational sum frequency generation studies, show that cations may impact the concentration of nitrate ions in the interface region, thereby directly impacting the effective quantum yields for nitrate ions. surface active relative to the NO 3 (cid:2) ion and that the surface propensity of NO 3 (cid:2) is dependent on the magnitude of the electric field between the cation and NO 3 (cid:2) . 88,95 The present paper is directed to elucidating the impact cations have on nitrate ion photolysis in thin films of aqueous aerosols deposited on a Teflon substrate, by monitoring NO 2 production.

The presence of ions at the interface has significant impacts on the chemistry in atmospheric aerosols, and on environmental surfaces, where interfacial ions can be oxidized at the surface without the need for gaseous oxidants to diffuse into the bulk. 15,16 There is also evidence that the photolysis of species at the interface has higher quantum yields than those in the bulk. [34][35][36][37][38][39][40][41][42][43][44][45][46][47][48] In short, interfacial kinetics and mechanisms may be significantly different compared to those in the bulk phase.
The role of cations in the composition of the interface of aqueous solutions has not been studied directly but there has been substantial computational work on the effects of these ions on the organization of water molecules. Monte Carlo simulations by Hribar et al. 49 determined that large cations with low charge densities, e.g., K + and Rb + , result in the surrounding water molecules being hydrogen bonded, whereas smaller ions, e.g., Na + and Li + , with accompanying higher charge densities can break the hydrogen bonded network. This is consistent with ab initio calculations that show that smaller cations more strongly interact with water molecules, 50 and was confirmed by molecular dynamic (MD) simulations that show coordination numbers of water molecules increase around cations with increasing size. 51 Though computational studies have clearly shown that cations can alter surrounding water molecules, it is not well established whether cations can also affect chemical processes through effects on the concentration of anions at the air-water interface.
Nitrate ions are a common constituent of sea salt aerosols (100-400 mM) 52 due to oxidation of surface halide ions by oxides of nitrogen (N 2 O 5 , NO 2 , NO 3 and ClONO 2 ),  as well as uptake of nitric acid from the gas phase. 81 Nitrate ions photolyze with actinic radiation, producing NO 2 and OH via (1a) 82 and (2), 83 and NO 2 À and O( 3 P) through (1b). 82 Molecular dynamics simulations, 45,[84][85][86] surface sensitive spectroscopy studies, 21,33,[87][88][89] and surface tension measurements 6,85 have been used to examine nitrate ions in solution, and most have concluded that in bulk aqueous solutions of NaNO 3 (Z 300 water molecules), nitrate is not in surface excess compared to the bulk. In bulk nitrate ion photolysis, O À and NO 2 formed via reaction (1a) have a high recombination rate due to solvent cage effects, thus regenerating nitrate ion. 82 Only photolysis products that escape the solvent cage are able to either escape into the gas-phase or diffuse into the bulk of solution. At room temperature the bulk-phase quantum yields are B0.01 for OH production and an order of magnitude lower (B0.001) for O( 3 P) formation (1b) at 305 nm. 82,[90][91][92][93] In smaller water clusters (32-300 water molecules), 86 NO 3 À ions are predicted to be present at the air-water interface. In the interface region, nitrate ions are less solvated than those in the bulk, 86 resulting in increased NO 2 production yields. 33,[46][47][48]94 Computational modeling and X-ray photoelectron spectroscopy show that cations (Na + and Rb + ) have a strong impact on Br À surface enhancement, affecting the ion's partitioning of halide ions at the interface. 32 96 The coated chambers were then filled with synthetic air (Scott-Marrin Inc., NO x o 0.001 ppm, SO 2 o 0.001 ppm, Riverside, CA) that flowed through a water bubbler and mixed with dry air to adjust the final relative humidity (RH). For each MNO 3 experiment (except for KNO 3 ), the RH in the chamber was B5% above the deliquescence relative humidity (DRH) of that salt ( Table 1), so that the salts were aqueous solutions. In the KNO 3 system it was difficult to obtain reproducible and stable water vapor concentrations above 90% RH, so the experiments with KNO 3 were performed at 85%, below the DRH of 92.2 AE 0.4%.
To the best of our knowledge, there are no reported literature values for the DRH of RbNO 3 . The DRH of RbNO 3 was therefore measured experimentally by placing a saturated RbNO 3 solution in a humidity calibrator (Vaisala, HMK15) and sampling with a relative humidity-temperature probe (Vaisala, HMP 338) every hour until three consistent measurements were obtained to ensure equilibration and reproducibility.
Irradiation of the thin aqueous salt film on the walls of the Teflon chamber was carried out using eight externally mounted narrowband UVB lamps (l max B 311 nm) which overlap with the nitrate np* absorption band. The reaction chamber is actively cooled using an external fan to prevent dehydration of the film. All experiments were conducted at 298 AE 2 K. The NO 2 concentrations were monitored as a function of time by periodically sampling with a chemiluminescence nitrogen oxides analyzer (ThermoElectron Corp., Model 42C). Calibration of the nitrogen oxides analyzer was performed using known mixtures of NO 2 (4.57 ppm; Scott Marrin Inc.) or NO in N 2 at levels similar to those detected during photolysis.

UV/visible spectra
The UV absorption spectrum was taken for each MNO 3 at 0.5 nm resolution using a Cary 50 UV-visible spectrophotometer for nitrate ion concentrations of 0.5 M and 2 M using a 1 cm and 0.1 cm UV-visible grade quartz cell, respectively.

Bulk photolysis studies
The formation rates of OH from NO 3 À photolysis were measured in bulk MNO 3 solutions by adding benzene as a chemical probe and measuring the stable product phenol. Benzene was chosen as the chemical probe because it has been previously proven useful for trapping OH, 97 does not have a counter cation and will not affect solution pH. Air-saturated solutions of 5.0 mM MNO 3 and 10.0 mM benzene were placed in an airtight 2 cm, far-UV, quartz cuvette (Spectrocell), placed in a temperaturecontrolled illumination chamber at 298 K with continuous stirring and illuminated with 313 nm light from a monochromatic illumination system with a 1000 W Hg/Xe lamp. For each illumination experiment, a dark control (with corresponding rate of phenol formation R dark ) was also performed, consisting of a quartz cell wrapped in aluminum foil and treated identically (sample composition and temperature) as the illuminated sample.
On each day of experiments, two measurements were also performed: (1) a blank control containing benzene only (with corresponding phenol rate R blank ) and (2) measurement of the photon flux in the chamber using 10 mM 2-nitrobenzaldehyde (2NB) as a chemical actinometer to test for possible lamp intensity fluctuations. 98 The rate of photolysis of 2NB did not differ significantly from day to day ( J 2NB = (7. Aliquots of sample were removed from the quartz cells at measured time intervals during illumination and the phenol concentration was measured using high performance liquid chromatography (HPLC). The HPLC system consisted of a Shimadzu SPD-10A UV-Vis detector, LC-10AT pump, and a 250 Â 33 mm, 5 mM bead, BetaBasic-18 column (Thermo Hypersil-Keystone). The HPLC eluent was 30 : 70 acetonitrile-water mixture at a flow rate of 0.5 ml min À1 . For each experimental day, the HPLC was calibrated with aqueous phenol standards to span the range of values detected. The rate of phenol formation, R Phen , during illumination was determined from plots of phenol concentrations versus illumination time using a linear regression. Rates of phenol formation in dark or blank samples were r1% of the sample value. The measured R Phen was converted to the rate of OH formation, R OH , by dividing by the yield of phenol formed from the reaction of OH with benzene (Y Phen = 0.69). 97 For the majority of experiments, the benzene concentration was 10.0 mM which was sufficiently high to trap all of OH generated from the 5.0 mM MNO 3 solutions. To determine the benzene concentration needed to trap all of the generated OH, separate experiments at either constant concentrations of NaNO 3 or Mg(NO 3 ) 2 with varied benzene concentrations were conducted (Fig. S1, ESI †). For both of these sets of experiments, the rate of phenol formation increased with increasing benzene concentration until the concentration ratio [benzene] : [NO 3 À ] Z 1, at which point the rate remained constant, indicating that all of the OH was scavenged. Solutions were typically at a pH of 5.1 AE 0.2 (unadjusted) with a less than AE0.3 change in pH by the end of illumination.

Error analysis
All errors are reported as 2s, where s is the sample standard deviation, defined as where N, the number of samples, was 3-5, depending on the measurement. 99

Chemicals
The salts NaNO 3 Fig. 1 shows the formation of NO 2 as a function of illumination time (Fig. 1a) and the corresponding rates of NO 2 production (R NO 2 , Fig. 1b) during the illumination of thin films containing NO 3 À with different cations: Na + , K + , Rb + , Ca 2+ and Mg 2+ . All experiments with the exception of KNO 3 were conducted at 5% above the deliquescence relative humidity (DRH, Table 1) so that the salts were aqueous solutions, which was confirmed by visual inspection.

Results and discussion
As mentioned in the experimental section, the experiments for KNO 3 were performed just under the DRH, at 85%. To gain added insight into how this might affect the rate of NO 2 production, photolysis of NaNO 3 was studied as a function of relative humidity above and below the deliquescence point (Fig. 2). Above the DRH (73.8% for NaNO 3 ), there is no more than a 15% difference in the R NO 2 . This is consistent with water activity measurements by Tang and Munkelwitz 100 in which the concentration of NO 3 À was shown to change by only B18% for NaNO 3 aerosols with RH ranging from 75-87%. Since the photolysis rate is independent of nitrate concentration but dependent on the moles of NO 3 À (see ESI †), the increased water (and film volume) at higher relative humidity has only a minor effect on the rate of nitrate photolysis. In the range of 50-75% RH, below the DRH of NaNO 3 , NO 2 continues to be generated but at a smaller rate. While the salt is solid under these conditions, it holds significant amounts of surface-adsorbed water 101-104 which increases ion mobility at the interface and provides some liquid-like character to the surface. These results for NaNO 3 suggest that R NO 2 from KNO 3 would be higher if photolysis experiments were conducted 5% above the DRH rather than 7% below, and hence the data in Fig. 1 for KNO 3 represent lower limits. Values of the R NO 2 in Fig. 1 fall into three groups: (1) RbNO 3 and KNO 3 , which rapidly photolyze to form NO 2 4 3 ppb min À1 , (2) Ca(NO 3 ) 2 , which photolyzes more slowly (R NO 2 r 1 ppb min À1 ), and (3) Mg(NO 3 ) 2 and NaNO 3 , which have intermediate rates. The difference in the R NO 2 between these groups is unexpected because the only difference is the counter cation. Cations have been generally thought to be spectator ions, and believed to play a minor role in both chemical and photochemical processes.
One potential source of difference between the cation-nitrate salts could be that they possess different UV absorption spectra, resulting in more rapid photolysis of RbNO 3 and KNO 3 compared to Mg(NO 3 ) 2 , NaNO 3 and Ca(NO 3 ) 2 . To probe this possibility, the UV-visible absorption spectra of aqueous MNO 3 solutions (0.5 M and 2 M) were measured (Fig. S2, ESI †). The molar absorptivies of the MNO 3 salts for both 0.5 M and 2 M are all centered around 302 AE 1 nm, and do not vary significantly between the different salts, suggesting that changes in the absorption spectra cannot be responsible for the increase in the R NO 2 for the different cationnitrate salts. However, studies by Hudson et al. 105 showed that as Ca(NO 3 ) 2 , Mg(NO 3 ) 2 , and NaNO 3 become more concentrated, the lowest electronic absorption (np*) shifts to shorter wavelengths and decreases in intensity. From their results, Ca(NO 3 ) 2 showed the largest change in light absorption, with a 10% decrease from 0.5 M to 5.8 M. These minor changes suggest that differences in the absorption spectra cannot account for the large differences in the R NO 2 between MNO 3 . Even though concentrations in the thin films do change depending on RH, this should not affect the photolysis rate since there was no significant light attenuation in the thin films (see calculation in ESI †).
Another possibility to explain the photolysis differences between the different salts ( Fig. 1) is that there could be a bulkphase photochemical process in which cations disrupt the solvent cage around nitrate ions, allowing photolysis products to more easily escape. However, one photolysis study of bulk nitrate solutions 106 reported that the quantum yield of NO 2 À formation via reaction (1b) was not sensitive to the nature of the counter ion. As illustrated in reaction (1a) and (1b), NO 2 and NO 2 À are both products of NO 3 À photolysis and therefore we would expect that if cations were affecting the solvent cage in the bulk, then the NO 2 À and OH yields would be affected as well. To verify that a bulkphase process does not affect nitrate photolysis, bulk-phase OH solution measurements were performed by following phenol production with time. Fig. 3 shows that the R OH is not statistically different between the five MNO 3 salts, with an average value of 1.9 AE 0.08 mM min À1 . The corresponding average value of F(OH) is (1.1 AE 0.4) Â 10 À2 (see Chu and Anastasio 97 for details of the calculation), which agrees well with past studies reporting a value of B0.01 at room temperature. 82,[90][91][92][93]97 These OH experiments thus suggest that the nature of the cations is not influencing the production of nitrate photo-products in the bulk. However, the bulk-phase measurements were conducted at concentrations that were a factor of 600 (for RbNO 3 ) to 2040 (for NaNO 3 ) times lower than the thin film experiments due to the low solubility of benzene in aqueous solution (23 mM). Thus there is a possibility that a higher R OH would result in the bulk studies if experiments could be conducted at higher nitrate ion concentrations. To test this, the influence of nitrate ion concentration on the F(OH) in the bulk-phase was determined using NaNO 3 , which has the largest concentration difference between the bulk and thin film photolysis experiments.  In this case, sodium benzoate was used as the chemical probe due to its much larger solubility in water compared to benzene (4240 mM), allowing for nitrate ion concentrations closer to those in the thin-film experiments to be used. The F(OH) was found to be independent of nitrate ion concentrations between 0.05 mM to 0.5 M, with an average value of (1.0 AE 0.5) Â 10 À2 (Fig. S3, ESI †). The absence of a bulk-phase explanation suggests that cations may be affecting the partitioning of NO 3 À near the surface, where an incomplete solvent cage enhances photo-product formation. The observed R NO 2 can be expressed as a combination of contributions from the interface region and the bulk: In eqn (4), I l is the photon flux from the narrow-band UVB lamps, s l is the absorption cross-section of NO 3 À , f int and f b are the quantum yields for NO 2 formation from NO 3 À photolysis in the interface region and the bulk, respectively, and N int and N b are the number of available nitrate ions in each region. Given that the same number of moles of NO 3 À was added into each reaction chamber, the amount of nitrate in the bulk will be the same for each MNO 3 . Similarly, the bulk quantum yields for the different salts are the same (Fig. 3). Therefore, this suggests that the first term (f int Â N int ) is responsible for the observed differences in the rates of production of NO 2 .
To assess if interfacial chemistry is responsible for the observed change in R NO 2 , a combination of our experimental data and previous MD simulations was used. Assuming that bulk nitrate photochemistry dominates NO 2 release from NaNO 3 , the ratio of the R NO 2 production for KNO 3 (R K NO 2 ) to NaNO 3 (R Na NO 2 ) can be expressed as a combination of contributions from the interface region and the bulk (eqn (5)): It is assumed that the interface region for NaNO 3 does not contribute significantly to the production of NO 2 because of the relatively small concentrations of nitrate ions near the interface. Eqn (5) can be rewritten in the form of eqn (6): The second term in eqn (6) is taken to be B1 since (a) the bulkphase studies (Fig. 3) show that f b K = f b Na and (b) equal amounts of nitrate were added to the chamber for both NaNO 3 and KNO 3 , so the number of ions in the bulk (which forms most of the 800 nm thick film) is essentially the same; even in the case of enhancement of nitrate ions in the interface region for the potassium salt (see below), the interface layer (taken as 1 nm) is too small to affect the bulk-phase number significantly. Thus The measured ratio of the rates of gas-phase NO 2 production for the potassium salt compared to the sodium salt is 1.73 AE 0.34 (Fig. 1). Thus eqn (6) becomes: Rearranging this to solve for f int K gives If nitrate ions were equally distributed over the interface and bulk regions, i.e., there was no enhancement of NO 3 À at the KNO 3 interface compared to that in the NaNO 3 bulk, then N int . Applying this value and the bulk phase f b Na = 0.011 in eqn (8) gives f int K = 6.4. Since the quantum yield for NO 2 production must be r1.0, this is not reasonable and suggests that some combination of enhancement of nitrate ions at the interface for the KNO 3 salt (relative to NaNO 3 ) and an enhancement in the quantum yield at the interface over that for the bulk must be occurring.
The results of previous MD simulations suggest that the nitrate ion concentrations in the interface region, although smaller than in the bulk, are a factor of 10 higher for KNO 3 compared to NaNO 3 . 45,47 This factor arises from the fraction of nitrate ions at the interface determined from the integrated density profile area of NO 3 À within AE0. distribution throughout the films), but this is not constrained by measurements or modeling. Despite the uncertainties, the calculated interfacial quantum yield for KNO 3 is much larger than the bulk quantum yield of 0.011, emphasizing that cations play a strong role in the ion organization and photochemistry in thin films. In this case, interfacial nitrate accounts for roughly 40% of NO 2 production from KNO 3 photolysis, compared to 7% for the NaNO 3 films (see calculation in ESI †). MD simulations also show that 2 M Mg(NO 3 ) 2 85 has a ratio of 0.2 in the number of nitrate ions in the interface region compared to the bulk, resulting in a factor increase of 7 more NO 3 À in the interface region compared to NaNO 3 . It would therefore be expected that the rate of NO 2 production would also be higher for the magnesium salt compared to the sodium salt and that the order for R NO 2 would be KNO 3 4 Mg(NO 3 ) 2 4 NaNO 3 , consistent with data in Fig. 1. The measured value for R NO 2 from Mg(NO 3 ) 2 is about 10% higher than NaNO 3 (ratio is 1.1 AE 0.2), but using eqn (6) and assuming an interfacial quantum yield of 0. 6 95 For example, Ca 2+ nitrate solutions showed considerably more contact-ion pairing (with no solvent molecules between ions) compared to Mg 2+ . This resulted in Mg(NO 3 ) 2 having 50% more free nitrate ions at the air-water interface compared to Ca(NO 3 ) 2 . As shown in Fig. 1, the R NO 2 for Mg(NO 3 ) 2 is approximately 2.9 times faster than Ca(NO 3 ) 2 . Therefore, one possibility is that free nitrate at the interface may photolyze more efficiently than cation-nitrate ion pairs at the interface.

Conclusions
This study shows that the rate of production of gas-phase NO 2 from thin films of nitrate salts is enhanced for RbNO 3 and KNO 3 , somewhat less for Mg(NO 3 ) 2 and NaNO 3 and even less for Ca(NO 3 ) 2 . Neither differences in the UV-visible spectra of the aqueous solutions of the salts, nor changes in bulk-phase photolysis for the different salt solutions, are consistent with this enhancement. Previous MD simulations and VSFG studies suggest that higher concentrations of nitrate ions at the air-water interface, especially those that do not have strong columbic attractions with their counter-cation, may explain this enhancement. Thus cations can play a role in photochemical processes at interfaces, possibly by changing the concentration of nitrate ions in the interface region.