Research Article

Experiments and Simulations of Ion-Enhanced Interfacial Chemistry on Aqueous NaCl Aerosols

See allHide authors and affiliations

Science  14 Apr 2000:
Vol. 288, Issue 5464, pp. 301-306
DOI: 10.1126/science.288.5464.301


A combination of experimental, molecular dynamics, and kinetics modeling studies is applied to a system of concentrated aqueous sodium chloride particles suspended in air at room temperature with ozone, irradiated at 254 nanometers to generate hydroxyl radicals. Measurements of the observed gaseous molecular chlorine product are explainable only if reactions at the air-water interface are dominant. Molecular dynamics simulations show the availability of substantial amounts of chloride ions for reaction at the interface, and quantum chemical calculations predict that in the gas phase chloride ions will strongly attract hydroxl radicals. Model extrapolation to the marine boundary layer yields daytime chlorine atom concentrations that are in good agreement with estimates based on field measurements of the decay of selected organics over the Southern Ocean and the North Atlantic. Thus, ion-enhanced interactions with gases at aqueous interfaces may play a more generalized and important role in the chemistry of concentrated inorganic salt solutions than was previously recognized.

Processes at the air-water interface may play a key role in the uptake and reactions of gases with liquid particles in the troposphere. For example, SO2 has been shown to form a unique bound complex at the interface that participates in its oxidation to sulfate (1–3). Similarly, reactions between gases such as Cl2 and O3 and inorganic ions such as Br and I in solution have been shown to occur not only in the bulk liquid but also at the interface, leading to enhanced uptake and/or reactions of the gases (4–6).

Zetzsch and co-workers (7) report that an unidentified chlorine atom precursor is generated upon irradiation of seawater aerosol with simulated sunlight in the presence of ozone. Subsequently, our laboratory has shown that gaseous Cl2(8) is generated from deliquesced sea salt particles in the presence of O3 and 254-nm radiation. Photolysis of O3 generates OH radicals in the gas phase, which can be taken up into the particlesEmbedded Image(1) Embedded Image(2) Embedded Image(3) Embedded Image(4)where hυ is light of wavelength 254 nm, O(1D) is an electronically excited oxygen atom, g indicates gaseous, and aq indicates aqueous. OH is known to react in solution with Cl, ultimately generating Cl2 Embedded Image(5) Embedded Image(6) Embedded Image(7) Embedded Image(8) Embedded Image(9)These reactions are dominant for Cl2 production under the chamber conditions. However, a variety of additional reactions can also contribute, such as HOCl + Cl + H+ → Cl2 + H2O, which dominates under atmospheric conditions. Such bulk aqueous-phase chemistry is included in current models of sea salt reactions in the marine boundary layer (9, 10).

Because of the complex composition of sea salt (11), constituents such as various trace metals have the potential for catalysis. Here we studied reactions of NaCl alone in order to simplify the potential chemistry. A comprehensive computational chemical kinetics model, which incorporates the known gas- and aqueous-phase chemistry in the particles as well as the mass transfer processes, is used to explore quantitatively the chemistry of this simplified system. We show that experimental results and model predictions can only be reconciled if reactions of gases with ions at the interface control the chemistry. Molecular dynamics simulations of NaCl dissolved in water clusters and quantum chemical calculations support this mechanism. Moreover, they suggest that strong interactions of gases with ions at the interface of concentrated solutions are probably a general property of atmospheric aerosol particles and should be taken into account in understanding and modeling atmospheric processes.

Aerosol chamber studies of the O3–NaCl photochemical system. Dry NaCl particles in air were placed in a chamber described in detail elsewhere (12) and the relative humidity (RH) was increased above the deliquescence point (75% RH) to form concentrated salt solutions. At a RH of 82% and temperature of 297 K, typical of these experiments, the particles are concentrated salt solutions with a concentration of 20 weight percent (wt. %) (13). Ozone was added, but no production of Cl2was observed until the mixture was irradiated at 254 nm, generating OH via reactions 1 through 4. After the addition of ozone, differential optical absorption spectroscopy was used to follow O3, other gas-phase species such as CO2 were measured with Fourier-transform infrared spectroscopy (FTIR), and Cl2 was measured by atmospheric pressure ionization–mass spectrometry (API-MS).

Figure 1 shows Cl2 production in three experiments at different initial ozone concentrations. The Cl2 production is similar to that observed from sea salt particles (8), showing that the minor components are not important in this reaction. Although the initial ozone concentration decreases by a factor of about 6, the rate of initial Cl2production decreases by a factor of less than 2 and the peak Cl2 concentrations by a factor of less than 4. Indeed, the initial rate of Cl2 production appears to follow primarily the particle surface area and volume (which tend to change to a similar extent simultaneously in these experiments). This is expected if a key step in the Cl2 production is the reaction of OH with Cl, because modeled OH concentrations for the gas-phase reactions in this system are relatively insensitive to the initial O3 concentration.

Figure 1

Typical decay of O3 and formation of Cl2 as a function of time during the photolysis at 254 nm of a mixture of O3 and deliquesced NaCl particles in air at 82% RH and 297 K. Conditions in the three experiments were as follows: (▪) O3 = 3.4 × 1014 molecule cm–3, 1.9 × 105particle cm–3 with a count median diameter of 224 nm, geometric standard deviation σ = 1.93, surface area 6.1 × 1010 nm2 cm–3, and volume 5.1 × 1012 nm3 cm−3; (▴) O3 = 2.4 × 1014 molecule cm−3, 1.4 × 105 particle cm−3 with a count median diameter of 227 nm, σ = 1.97, surface area 4.7 × 1010 nm2cm–3, and volume 3.9 × 1012nm3 cm–3; (•) O3 = 0.63 × 1014 molecule cm–3, 1.1 × 105 particle cm–3 with a count median diameter of 209 nm, σ = 2.1, surface area 3.6 × 1010nm2 cm−3, and volume 3.1 × 1012 nm3 cm−3. The formation of Cl2 is shown in each experiment by the symbols □, ▵, and ○, respectively.

Computational kinetic model studies. To evaluate mechanisms of production of Cl2 in this system with known gas- and bulk aqueous-phase chemistry, we used a computer kinetic model (14). The model, MAGIC (Model ofAerosol, Gas and InterfacialChemistry), includes 17 gas-phase species (Table 1) undergoing 52 reactions, combined with 32 aqueous-phase species undergoing 99 reactions, including reactions 1 through 9. Mass transfer of species between the gas and aqueous phase (15) is treated with the method of Schwartz (16). Namely, the temporal variations of the concentrations of the gas-phase species, Cg, and the aqueous phase species, Caq, are described by a system of coupled differential equations of the formEmbedded Image(10)Embedded Image(11)where kmt is the mass transfer coefficient, wL is the dimensionless volumetric liquid water mixing ratio, HA is the Henry's Law coefficient, R is the universal gas constant,T is the chamber temperature, 〈Raq〉 is the spatially averaged bulk aqueous-phase reaction rate, and Rg is the gas-phase reaction rate. Because of the high ionic strengths encountered in the droplets, the reactivity of the species in the aqueous phase is described by their activities. Activity coefficients are calculated explicitly for the species H+, Na+, Cl, ClO3, OH, CO2, and O2 with the Pitzer ion interaction approach (17). For other species, the Guntelberg approximation of the Debye-Hückel limiting law is used (18). An additional term, Rint, is added to Eqs. 10 and 11 to account for interfacial reactions.

Table 1

Gas- and aqueous-phase species included in the computational model.

View this table:

Model predictions with the use of conventional physical and chemical processes for the experiment with an initial O3concentration of 2.4 × 1014 molecule cm−3 are shown in Fig. 2. The predicted Cl2 concentrations are three orders of magnitude smaller than observed because of the initially neutral pH of the particles. As a result, the generation of chlorine atoms from the HOCl intermediate via reaction 6 with H+ is slow relative to the rapid reverse of reaction 5 of HOClback to Cl and OH, and the subsequent production of Cl2 via reactions 7 through 9 is small. To confirm the effects of acid on the predicted Cl2 formation, we performed model runs in which a constant pH of either 3.5 or 4.0 was assumed. As seen in Fig. 2, the predicted initial rate of formation of Cl2 at a pH of 4.0 is smaller than observed because of the second-order kinetics for the self-reaction 8 of Cl2 in the bulk, and the predicted peak concentration was too small by a factor of 2. Although Cl2 is produced more rapidly at a pH of 3.5, the peak concentrations are predicted at later times and are much larger than those measured experimentally. In addition, there is no known source of acid in the experimental system, which would provide a pH of 4.0 or less.

Figure 2

Experimental data and model-predicted O3 decay and Cl2 formation for the experiment in Fig. 1 in which O3 = 2.4 × 1014molecule cm–3. For this experiment, conventional gas-particle transport and chemical processes in the gas and bulk aqueous phases (base case) were used without fixing the pH or by assuming a constant pH (= −log a H+) of either 4.0 (□) or 3.5 (▿). Experimental data are shown by •, O3 and ○, Cl2.

In short, conventional chemical and physical processes involving transport of gases to the particle surface, mass accommodation, diffusion, and known reactions in the bulk aqueous-phase of the particles, do not explain the experimental observations.

Molecular dynamics simulations. The solvation of NaCl and the nature of ion pairs in clusters and in concentrated bulk solutions have been explored theoretically (19–21). There is evidence that in aqueous solutions of inorganic salts, the large and polarizable anions are more readily available at the surface than the small nonpolarizable cations. This effect is present at the surfaces of both bulk solutions and finite-size molecular clusters. For example, a negative surface potential has been measured for NaCl solutions (22) and attributed in part to enhanced chloride ion concentrations near the surface. Theoretical studies report that in clusters of hundreds of water molecules, Cl can be moved within several molecular layers of the surface before the free energy starts to increase significantly, whereas the free energy for Na+ increases continuously from the center of the cluster to the surface (23). Molecular dynamics simulations of Na+(H2O)n and Cl(H2O)n(n = 4 to 14) clusters show that full ion solvation occurs in the former case, whereas surface solvation dominates for the chloride anion (24).

To understand the nature of the interface of microbrine droplets with which OH and other reactant gases in this system interact, we performed a series of molecular dynamics simulations of microbrines of different sizes, consistently keeping the NaCl concentration at the bulk saturation value. We investigated clusters ranging from NaCl(H2O)9 to Na32Cl32(H2O)288, and with the use of two-dimensional (2D) periodic boundary conditions, an “infinite” open surface (actually a double surface) with the unit cell formed by Na96Cl96(H2O)864. Two potential models, one nonpolarizable (25) and one explicitly including the polarizabilities of water and ions (26), were compared. Standard Lennard-Jones parameters for the ions were used in both cases (27). All interactions have been included for cluster studies, whereas for the slab with an open surface, an interaction cutoff of 12 Å was applied and long-range interactions were accounted for with the particle-mesh Ewald method (28). Simulations were run for 500 ps after a sufficiently long equilibration period of 250 ps. The slab with a flat surface mimics the atmospheric microbrines with typical diameters of 200 nm better than the small clusters with hundreds of atoms do. We therefore report our results primarily on the basis of slab simulations.

Figure 3 depicts a snapshot from the molecular dynamics run showing a typical arrangement of the open surface of the slab. The corrugated surface contains water molecules (red and white balls) and a large number of chloride ions (yellow balls). Sodium ions (green balls) are almost missing from the surface. This happens because the small sodium cations fit well into the hydrogen-bonded water structure and therefore are fully solvated, whereas the chloride anions are too large and, as a consequence, are to a great extent pushed toward the surface of the system. This effect is also present consistently in finite size clusters of different sizes and seems to be pertinent to concentrated solutions of ions of different relative sizes in a polar medium.

Figure 3

Snapshot of molecular dynamics predictions of typical open surface of a slab consisting of 96 NaCl molecules and 864 water molecules. The large yellow balls are Cl ions, the smaller green balls Na+, and the red and white balls are water molecules.

The degree of surface exposure of the ions will strongly influence their surface reactivity toward gas phase reactants. To quantify this, we have applied a standard procedure of evaluating the accessible surface area of the two ions by rolling a particle of the size of the OH radical (radius of 1.7Å) over the surface of the slab at each time step of the dynamical run. The relative accessible areas (with respect to the total surface) of the Na+ and Cl ions from both the polarizable and nonpolarizable potential model are shown in Fig. 4. Both models qualitatively predict that although the Na+ ions are fully solvated and almost absent at the surface, the Cl ions occupy a significant part of the surface of the slab. Quantitatively, the nonpolarizable and polarizable models predict that 3.3 and 11.9% of the surface is covered by chloride ions, respectively, whereas both models predict that the sodium cations occupy less than 0.2% of the surface.

Figure 4

(left). Predicted relative surface-accessible areas for Na+ and Cl with the use of the polarizable (solid lines) and nonpolarizable (broken lines) potential models, respectively.

The dramatic difference between the solvation of the two ions is not only due to their different sizes but also to a much larger polarizability of Cl, which also explains the quantitative discrepancy between the two potential models. Comparison of previous molecular dynamics simulations of clusters with only one type of ion to experimental measurements (24) as well as our preliminary first-principle molecular dynamics with a density functional interaction model (29) for Cl(H2O)6 indicates that the polarizable model is the more accurate one (24). Assuming that about 12% of the surface is covered by Cland that the ratio between the number of NaCl and water species is 1:9, the chloride ion actually has more surface exposed than an average water molecule.

Our simulations confirm that chloride anions, unlike the sodium cations, occupy a significant fraction of the microbrine surface. This is a robust effect that qualitatively depends neither on the size of the microbrines nor, within relatively broad margins, on temperature. Two different potential models coherently predict this behavior, and the more accurate model also predicts more surface chloride anions.

Finally, in agreement with experimental findings (4,5), we found that chloride ions strongly attract surrounding reactive particles. Using practically converged quantum chemical calculations [coupled clusters with single and double excitations and perturbative triples with an augmented correlation corrected Dunning polarization double-zeta basis set and with correction for the basis set superposition error (30)], we have evaluated affinities of Cl toward the OH radical and compared them to those of the water molecule. The complexation energy rises dramatically from 4.9 to 16.9 kcal mol−1 from H2O to Cl, suggesting that chloride anions can actually enhance the scavenging of reactive species from the atmosphere with which they then undergo chemical reactions.

Integration of experiments and simulations. The combination of experimental observations, computer kinetic modeling with the use of only gas- and bulk aqueous-phase chemistry, molecular dynamics simulations, and quantum chemical calculations strongly suggests that chemistry at the interface is responsible for the observed production of Cl2. We propose that the key step in the generation of Cl2 in the chamber is the reaction of OH with chloride at the interfaceEmbedded Image(12)which then reacts further, ultimately forming Cl2. The structure of the bound complex between OH and Cl is likely to involve interaction between the hydrogen of the OH and the chloride ion. Such an intermediate has been observed in the gas phase upon electron impact on a mixture of HCl and N2O (31) and has been explored experimentally and theoretically in electron spin resistance studies of the low-temperature reaction of OH with chloride ions (32). We assume here that the OH...Cl intermediate undergoes a self-reaction in a manner similar to that of another chlorine-containing radical anion, Cl2(reaction 8)Embedded ImageEmbedded Image(13)Other potential mechanisms that could generate Cl2 were also considered (33); this one is presented because it provides good overall fits to the experimental data and is thermodynamically and kinetically feasible.

Figure 5 shows schematically the overall chemistry and mass transfer processes considered in this model. In contrast to the mechanism represented by reactions 5 through 9, the interfacial reaction proposed here does not require an acid for Cl2 production. Instead, OH is produced. Cl2 is known to react rapidly with OH so that if the droplets become highly alkaline, Cl2 is taken up into the aqueous phase and hydrolyzed, which results in much smaller predicted gas-phase concentrations of Cl2. However, the particles in the experiment are buffered by small amounts of gaseous CO2 [∼13 parts per million (ppm)] that are present from the CaCO3 drying agent in the particle diffusion dryer. The computer kinetic model includes the uptake of CO2 and the aqueous-phase chemistry of carbonate and bicarbonate. As a result of these buffering reactions, the predicted pH in the droplets reaches the range of 8 to 9.

Figure 5

(right). Schematic diagram of kinetic model of gas particle system.

Production of Cl2 is modeled under the assumption that a three-step process occurs at the interface, in addition to conventional transport into and reaction in the bulk aqueous phase of the particles. The rate of reaction 12 is set equal to the number of OH-particle surface collisions per second, multiplied by the ratio of the chloride ion to water concentrations, and scaled by a factor of 1.6 for the relative areas of a chloride ion and water, consistent with the polarizable molecular dynamics results that predict 12% of the surface is covered by chloride ions. A time-dependent scaling factor is also included to account for the fact that a chloride ion already existing as a bound (OH...Cl) complex would not react with an incoming gas phase OH. This approach is applicable when the collisions are random with all surface sites without a strong propensity for hitting a chloride ion. This model is conservative because OH collisions may result in sticking to the surface with longer residence times, which will give a higher reactivity than the estimate based simply on collision rates. The rate constant for reaction 13 is assumed to be the same as that for the self-reaction of Cl2 , 1.8 × 109 l mol−1 s−1 (34). The (OH...Cl) intermediate was assumed to exist in a volume corresponding to a 1-nm shell on the outside of the particle. In addition to its self-reaction, it is also assumed to decompose back to reactants with a rate constant of 1 × 104s−1, which was chosen to provide the best fit to the experimental observations. However, similar fits can also be obtained with smaller decomposition rate constants for OH...Clalong with smaller values of the probability for reaction 12 forming OH...Cl.

Figure 6 shows the model-predicted concentrations of O3 and Cl2 when the surface reactions are included, along with the full suite of gas- and liquid-phase chemistry and mass transfer processes for the bulk phases described above for the three different initial ozone concentrations (Fig. 1). The model predictions are much more consistent with both the time-dependent and the peak levels of Cl2 observed experimentally than the base case, which uses only known gas and bulk aqueous-phase chemistry, or the predictions that assume a constant pH of 3.5 to 4.0 (Fig. 2). Indeed, the predicted Cl2concentrations are always within 50% of those observed experimentally, and the time profiles are in qualitative agreement despite the uncertainties in the fates of (OH....Cl) at the interface and the simple reaction scheme used.

Figure 6

Model-predicted Cl2formation (○) and O3 loss (▴) when the interface reactions are included in computational kinetic model. Experimental data, dotted lines; model predictions, solid lines. The model takes into account both dilution, which occurs during the experiment while sampling into the API-MS to measure Cl2, and small losses of Cl2 at the chamber walls and from photolysis, which were determined independently with authentic samples of Cl2. (A through C) correspond to three experiments with different initial O3 concentrations as in Fig. 1.

Application to atmospheric marine boundary layer. Computational kinetic studies were also performed for conditions more representative of the marine boundary layer (35,36) using the mechanism that includes the surface reactions. Salt particles were assumed to be present with a number concentration of 10 cm−3 and a diameter of 2 μm. Figure 7 shows the model predictions for Cl atoms and OH radicals for a 24-hour diurnal cycle for remote marine boundary layer conditions. The reported diurnal averages for OH and Cl (35) over 24 days are also shown, which include light and dark periods as well as times when sea salt particles had been washed out by storms. Given the different time scales and conditions involved in the model and measurements and that the model does not include transport or emissions, the model predictions and the field results are in excellent agreement. The pH of the particles was predicted to have dropped to ∼4 after 24 hours, so acid-catalyzed bulk aqueous phase chemistry also contributes to the production of chlorine atoms (∼60% of the total). Model calculations were also carried out for a Lagrangian field experiment over the North Atlantic Ocean (36). In this case, the model predicts a chlorine atom concentration of 3.7 × 104 atoms cm−3 at noon, compared to a measured value of 6.5 × 104 atoms cm–3 for a 2-hour period leading up to local noon, again in good agreement given the uncertainties in both the model and the measurements. Under these more polluted conditions, the pH of the particles is predicted to fall to ∼3.2 after 2 hours, so that bulk aqueous phase chemistry predominates in the chlorine atom production (80% of the total). Under these atmospheric conditions, the major source of Cl2 from the bulk, aqueous phase chemistry is the acid-catalyzed reaction of HOCl with Cl. In the absence of the interface reaction, chlorine atoms are generated primarily from OH + HCl, ultimately forming gas phase HOCl that is taken up into the aqueous phase. Clearly, the surface reactions will be most important at pH values above about 4.

Figure 7

Model-predicted Cl atom and OH radical concentrations over a 24-hour period under remote boundary layer conditions. Initially, concentrations of ethane, propane, dimethyl sulfide, NO and O3 are 370 parts per trillion (ppt), 11 ppt, 200 ppt, 3 ppt, and 25 parts per billion, respectively, consistent with measured values in the Southern Ocean (35, 40). Monodisperse salt particles are present with a number concentration of 10 cm−3 and a diameter of 2 μm.

Given the ubiquitous occurrence of sea salt particles in the marine boundary layer and the potential importance of their reactions in mid-latitudes and in the Arctic, it is critical that such chemistry at interfaces be taken into account in field, laboratory, and tropospheric modeling studies. Furthermore, there is an urgent need for the development and application of experimental methods of detection of such radical surface species. Molecular dynamics simulations suggest that the availability of anions at the interface is not unique to chloride. If this availability is determined largely by the size and polarizability of the anion, then bromide should be even more evident at the surface of concentrated sea salt particles, enhancing its chemistry relative to chloride as has been observed in a variety of laboratory (37) and field (38) studies. Another factor in this enhancement may be that Br, unlike Cl, is oxidized by O3 at room temperature (39).Furthermore, quantum chemical calculations have shown that the complexation energy of O3 and Cl is larger than O3 and H2O (5.5 versus 1.6 kcal) and might be expected to be larger for O3 and Br as well. Finally, the possible role of anions such as sulfate and nitrate that are considered unreactive in the atmosphere in increasing the scavenging of gases and, hence, altering their heterogeneous chemistry, also needs to be considered in assessing the contribution of highly concentrated inorganic salt particles to chemistry in the atmosphere.

  • * To whom correspondence should be addressed. E-mail: bjfinlay{at}


View Abstract

Navigate This Article