on CaCO3 surfaces under dry and wet conditions

Abstract. With increasing NO2 concentration in the troposphere, the importance of NO2 reaction with mineral dust in the atmosphere needs to be evaluated. Until now, little is known about the reaction of NO2 with CaCO3. In this study, the heterogeneous reaction of NO2 on the surface of CaCO3 particles was investigated at 296 K and NO2 concentrations between 4.58×1015 molecules cm−3 to 1.68×1016 molecules cm−3, using diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) combined with X-ray photoelectron spectroscopy (XPS) and scanning electron microscopy (SEM), under wet and dry conditions. Nitrate formation was observed under both conditions, while nitrite was observed under wet conditions, indicating the reaction of NO2 on the CaCO3 surface produced nitrate and probably nitrous acid (HONO). Relative humidity (RH) influences both the initial uptake coefficient and the reaction mechanism. At low RH, surface −OH is formed through dissociation of the surface adsorbed water via oxygen vacancy, thus determining the reaction order. As RH increases, water starts to condense on the surface and the gas-liquid reaction of NO2 with the condensed water begins. With high enough RH (>52% in our experiment), the gas-liquid reaction of NO2 with condensed water becomes dominant, forming HNO3 and HONO. The initial uptake coefficient γ0 was determined to be (4.25±1.18)×10−9 under dry conditions and up to (6.56±0.34)×10−8 under wet conditions. These results suggest that the reaction of NO2 on CaCO3 particle is unable to compete with that of HNO3 in the atmosphere. Further studies at lower NO2 concentrations and with a more accurate assessment of the surface area for calculating the uptake coefficient of the reaction of NO2 on CaCO3 particle and to examine its importance as a source of HONO in the atmosphere are needed.


Introduction
It has been estimated that 1000-3000 Tg of mineral aerosols are emitted into the atmosphere annually (Jonas et al., 1995).The main components of mineral dust include quartz, feldspar, carbonate (e.g.calcite, dolomite) and clay.Each of these components provides surfaces and reactants for heterogeneous reactions in the atmosphere.Many gas phase species in the atmosphere could also condense or adsorb onto mineral dust during long-range transport to impact atmospheric chemistry and climate change (Chen, 1985;Quan, 1993;Carmichael et al., 1996;Zhang et al., 2000).Modeling studies suggested that approximately 40% of nitrate formation is associated with mineral aerosols (Dentener et al., 1996).Aerosol samples taken in East Asia showed a good correlation between nitrate and calcium (Zhuang et al., 1999;Song and Carmichael, 2001;Sullivan et al., 2007).
In fresh Asian mineral aerosols, calcium is found primarily in the form of calcium carbonate (CaCO 3 ) (Song et al., 2005;Okada et al., 2005).The observed association of nitrate with calcium in mineral dust samples after long range transport suggests that after exposure to nitrogen oxides in polluted region, calcium carbonate in mineral dust had been converted Published by Copernicus Publications on behalf of the European Geosciences Union.
T. Zhu et al.: Kinetics and mechanisms of heterogeneous reaction of NO 2 on CaCO 3 surfaces to calcium nitrate, which in turn can change the composition and morphology of calcium-containing aerosols and enhance their hygroscopicity and the rates of heterogeneous reactions, thus influence atmospheric chemistry.Calcium nitrate can also alter the optical properties of aerosols to impact radiation, cloud formation, and global climate change.
Laboratory studies have demonstrated that nitric acid (HNO 3 ) can react with CaCO 3 to form calcium nitrate (Fenter et al., 1995;Goodman et al., 2000;Hanisch and Crowley, 2001;Krueger et al., 2003a, b).The measured uptake coefficient accounting for the BET area of the samples was determined to be (2.5±1)×10−4 for HNO 3 on CaCO 3 under dry conditions (Goodman et al., 2000).Johnson et al. (2005) determined the initial uptake coefficient of HNO 3 on CaCO 3 to be (2±0.4)×10−3 .The net reaction probability of HNO 3 on CaCO 3 particles was found to increase with increasing relative humidity, from 0.003 at RH=10% to 0.21 at 80% (Liu et al., 2008a).Vlasenko et al. (2006) reported an uptake coefficient of HNO 3 on CaCO 3 to be 0.11 at 33% relative humidity and a humidity dependence on Arizona Test Dust.
Compared to nitric acid, little is known about the reaction of NO 2 on CaCO 3 .The uptake coefficients of NO 2 on mineral oxide particles, such as Al 2 O 3 , Fe 2 O 3 , TiO 2 , have been determined to be on the order of 10 −7 (Underwood et al., 1999), which is consistent with the results of the reaction of NO 2 on mineral dust samples from the Cape Verde Islands (Ullerstam et al., 2003).Boerensen et al. (2000) reported the reactive uptake coefficient of NO 2 on Al 2 O 3 at levels of 10 −9 .
No literature report about the uptake coefficient of NO 2 on CaCO 3 has been found.Considering that the concentration of NO 2 in the atmosphere is at least an order of magnitude higher than that of HNO 3 , the reaction of CaCO 3 particles with NO 2 may comprise an important atmospheric reaction that lead to the formation of calcium nitrate during the long range transport of mineral dust.This reaction could be even more important in East Asia, where industrial regions with high NO 2 emissions are located in the path of long range transport of Asian Dust.
To understand the importance of the heterogeneous reaction of NO 2 on mineral dust, it is important to measure the uptake coefficient of NO 2 on CaCO 3 particles.Similar to the reaction of HNO 3 on CaCO 3 , the reaction of NO 2 on CaCO 3 may also change under the influence of surface water.Therefore, it is interesting to study the reaction mechanisms and to measure the uptake coefficient of the reaction of NO 2 on CaCO 3 particles under dry and wet conditions.
In this study, the reactions of NO 2 on CaCO 3 surfaces was investigated in situ using diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) in combination with Xray photoelectron spectroscopy (XPS), ion chromatography (IC), and scanning electron microscopy (SEM).Initial uptake coefficients were measured and the reaction mechanisms were studied under different relative humidity (RH).

Experimental
The infrared (IR) spectrum of the particle surface was investigated on line with DRIFTS during the reaction.The reactor of DRIFTS is a vacuum reaction chamber (Model HVC-DR2) surrounding a Harrick Scientific diffuse reflectance accessory (Model DRA-2CS) located in the sampling compartment of a Nicolet Nexus FTIR Spectrometer equipped with a mercury cadmium telluride (MCT) detector.Data from 128 scans, 4 cm −1 resolution, taken over about 80 s were averaged for one spectrum.
Besides online infrared spectroscopy measurements, for a number of selected experiments, the reaction products on the particle surface and extractable soluble components were also measured off line using X-ray photoelectron spectroscopy (XPS) and Ion Chromatography (IC), respectively.The morphological changes were observed using SEM.Further details of the instruments have been described previously (Li et al., 2006).
CaCO 3 (>99.999%,Alfa Aesar) powder was prepared by grinding and the size of the ground particle was about 1-10 µm with a mean value of 5.6 µm, as measured with a laser particle sizer (MasterSizer 2000, Malvern).The size and morphology of the ground CaCO 3 particles was also measured with SEM (KYKY1000B, KYKY Technology Development LTD., Beijing, China) (Fig. S1: http://www.atmos-chem-phys.net/10/463/2010/acp-10-463-2010-supplement.pdf).The ground CaCO 3 particles have various shapes; to simplify the estimation of surface area, we assumed the CaCO 3 particles have cubic shape.Using the size distribution measured with the laser particle sizer and using cubic shape for all CaCO 3 particles, the specific geometric surface area of the CaCO 3 particles (Ap) was calculated to be 0.19 m 2 g −1 .The volumetric BET (Brunauer, Emmett and Teller model) surface area of particles was measured with an ASAP2010 BET apparatus (Micromeritics Co., USA).The BET surface area, A BET , was determined to be 4.91 m 2 g −1 , which is 25.8 times the specific geometric surface area Ap of CaCO 3 particles.
For experiments with DRIFTS, about 20 mg CaCO 3 particles was placed in the sample holder, pressed flat using a glass slide, and stored in a desiccator as previously described (Li et al., 2006).Water vapor was produced and gases were mixed using the apparatus displayed in Fig. S2: http://www.atmos-chem-phys.net/10/463/2010/acp-10-463-2010-supplement.pdf.All tubes and surfaces were made of inert glass or Teflon with the exception of the stainless steel in situ reactor.Concentrations of NO 2 entering the reactor were adjusted by mixing a NO 2 standard gas (710 ppm in N 2 , Center of National Standard Material Research) with pure N 2 (>99.999%)using two mass flow controllers (FC-260, Tylan, Germany).The flow lifetime in the DRIFTS cell is 2 s.In the reactor, the NO 2 diffused to the particle surface and reacted with CaCO 3 .The formation of the reaction products was monitored using DRIFTS.When we use SiO 2 particles to react with NO 2 as a control experiment under dry conditions, we did not detect NO − 3 formation.Humidity was regulated by mixing dry nitrogen with water vapor by bubbling through two glass grits in pure water (Millipore Corporation, Billerica, MA, USA).Humidity was monitored using a temperature sensor (PT100; Vaisala, Vantaa, Finland) and a humidity sensor (HM1500; Vaisala).
After the reactions, the absolute numbers of nitrate and nitrite ions formed during the reaction were determined with IC.The reacted CaCO 3 particles were sonicated in 10 mL of pure water (Millipore Corporation), and then the filtered solution was analyzed using an IC (Dionex DX-500 system), which was equipped with an Ionpac AS-11HC-4 mm analytical column and a conductivity detector.Details of analyzing conditions can be found in a previous paper (Li et al., 2006).
The XPS spectra were taken on an AXIS-Ultra instrument from Kratos Analytical using monochromator Al K α radiation (225 W, 15 mA, 15 kV) and low-energy electron flooding for charge compensation.To compensate for surface charges effects, binding energies were calibrated using C1s hydrocarbon peak at 284.80 eV.The data were converted into VAMAS file format and imported into CasaXPS software package for data processing and curve-fitting.

Reaction under dry conditions
Experiments under dry conditions were carried out by first heating the CaCO 3 at 623 K for 2 h, thus eliminating nearly all of the adsorbed water.A mixture of NO 2 and N 2 with no water vapor was admitted at 296 K.
Figure 1 shows typical IR spectra of CaCO 3 when reacted with NO 2 under dry conditions for 0 to 100 min.Nitrate was clearly observed on the CaCO 3 surface.The peak at 816 cm −1 was assigned to the out-of-plane bending of nitrate ν 2 .The peak at 1043 cm −1 was assigned to the symmetric stretching of nitrate ν 1 and the peak at 748 cm −1 was assigned to the in-plane bending of nitrate ν 4 .Peaks at 1295, 1330, and 1350 cm −1 were assigned to the asymmetric stretching of nitrate ν 3 .The ν 3 mode of nitrate on the CaCO 3 surface has many peaks, which is different from that of the nitrate ions in Ca(NO 3 ) 2 .This can be attributable to the degenerate ν 3 mode of adsorbed nitrate of different coordination with surface (monodentate, bidentate, and bridging) (Underwood et al., 1999).
The asymmetric stretching of adsorbed nitric acid has been reported to be at 1710/1680 cm −1 (Nakamoto, 1997;Finlayson-Pitts et al., 2003).In our study, to confirm the vibration frequency of the adsorbed HNO 3 , calcium sulfate (CaSO 4 ) particles were exposed to gas phase HNO 3 .The IR absorption peaks of adsorbed HNO 3 were observed at 1670, and 1720 cm −1 ; therefore, we can assign the peaks at 1670 and 1703 cm −1 to adsorbed HNO 3 , which suggests that HNO 3 was formed when NO 2 reacted with CaCO 3 .The peaks at 1630 and 3520 cm −1 were assigned to the bending and asymmetric stretching of crystal hydrate water.Compared with the vibration frequencies of the crystallized water in Ca(NO 3 ) 2 •4H 2 O at 1640 cm −1 and 3490 cm −1 , a shift of both peaks occurred, due to the decreased hydrogen bonding.The asymmetric stretching shifted to the higher wave number, and the bending shifted to the lower wave number.The peaks at 3140 and 3330 cm −1 were assigned to surface adsorbed water, both peaks decreased as the reaction proceeded until the peaks for crystal hydrate water at 1630 cm −1 and 3520 cm −1 appeared.
Figure 2 shows the formation of nitrate on CaCO 3 through the reaction of NO 2 as a function of time.The amount of nitrate formed is represented by the integrated absorbance (I A ) of the IR peak area between 1077 and 1013 cm −1 .As the reaction proceeded, the nitrate concentration on the calcite surface increased with time and the rate of nitrate formation increased with higher NO 2 concentrations.The nitrate formation process could be divided into three stages: at stage I nitrate concentration on the surface increased linearly with time.This stage lasted about 10 to 20 min, depending on NO 2 concentration.Stage II was the transition stage when the increase of I A was slowing down, and stage III was www.atmos-chem-phys.net/10/463/2010/Atmos.Chem.Phys., 10, 463-474, 2010  the growth stage when I A was increasing at a constant rate, smaller than that of stage I.
The amount of nitrate ions formed on the particles {NO − 3 } was determined with a linear relationship between I A in the range of 1073-1013 cm −1 and the amount of nitrate determined by ion chromatography {NO − 3 }: The conversion factor f was found to be independent of reaction time and NO 2 concentration as long as the experiment was completed at a stage when the absorption of the nitrate band was still growing.For our study, f was calculated to be 1.89×10 18 molecules/I A .
Using f , the formation rate of nitrate d{NO − 3 }/dt was calculated, and the reaction order of NO 2 on CaCO 3 particles were obtained from the slope of the double-log plots of d{NO − 3 }/dt versus NO 2 concentrations (Fig. 3).Using the slope of the initial linear portion (0-10 min) to determine ln{dNO3−}/dt}, the reaction order of NO 2 in stage I was determined as 1.69±0.19and in stage II 0.49±0.36 in the range of NO 2 concentration in this study.

Reaction under wet conditions
Experiments of CaCO 3 reaction with NO 2 under wet conditions were carried out using non-heat-treated CaCO 3 particles at 296 • K with RH=60-71%.Figure 4a shows typical IR spectra of CaCO 3 when NO 2 entered the reactor under wet conditions, at reaction times from 0 to 60 min.
The peak at 1048 cm −1 was assigned to the symmetric stretching ν 1 of nitrate, the peak at 1344 cm −1 to the  asymmetric stretching ν 3 of nitrate.The peak at 1251 cm −1 was assigned to the asymmetric stretching ν 3 of nitrite.The peak at 1688 cm −1 was assigned to the adsorption peak of HNO 3 and the peak at 1636 cm −1 to the bending of the surface adsorbed water; the peak at 1472 cm −1 appeared in the range of peaks for the asymmetric stretching of carbonate and was assigned to the carbonate vibration peak.Compared to the corresponding peak observed under dry conditions, these peaks had all shifted more or less.We attributed these shifts to the changed environment caused by increased surface adsorbed water under wet conditions.
The peaks at 1525 and 1540 cm −1 were in the range of asymmetric stretching of bicarbonate and were assigned to the bicarbonate vibration peaks (Al-Hosney and Grassian, 2004;Al-Abadleh et al., 2005).Negative peaks at 1472, 1525, and 1540 cm −1 were similar to the negative peaks from the reaction of HNO 3 with CaCO 3 (Fig. S4: http://www.atmos-chem-phys.net/10/463/2010/acp-10-463-2010-supplement.pdf).This indicated that the intermediate products HCO − 3 and H 2 CO 3 were formed during the reaction of NO 2 with CaCO 3 .The result is consistent with the findings of (Al-Hosney and Grassian, 2004) that H 2 CO 3 was the intermediate product when HNO 3 , SO 2 , HCOOH and CH 3 COOH reacted with CaCO 3 .
The formation of nitrate and nitrite was further confirmed by the XPS spectrum of CaCO 3 particles reacting with 1.06×10 15 molecules cm −3 of NO 2 for 30 min at RH=80±2%, which demonstrated distinctly the existence of N of +5 valence and +3 valence as shown in Fig. 5a, and the clear formation of nitrate and nitrite shown in IC chromatogram of Fig. 5b.
Figure 4b shows the time curves of integrated absorbance of 1048 cm −1 for nitrate, 1248 cm −1 for nitrite, 1472+1525+1540 cm −1 for CaCO 3 , and 3500 cm −1 for H 2 O.During the first 10 min it shows clearly the fast increases in the concentrations of H 2 O and nitrites, and the fast decrease in CaCO 3 , while the absorbance of nitrate increase gradually.After 10 min, the nitrite concentration started to decrease, while the increase in the concentrations of H 2 O and nitrate, and the decrease in the concentrations of CaCO 3 maintained at relative slow and steady rates.It was inferred that the increase of adsorbed nitric acid caused the release of nitrite in the form of gaseous HONO which led to the decrease of the concentration of nitrite ions.
Figure 6 shows the nitrate formation as a function of time at different NO 2 concentrations during the reaction of CaCO 3 with NO 2 under wet conditions (RH=60-71%).The integrated absorbance was in the range of 1077 and 1013 cm −1 .Unlike the reaction under dry conditions, under wet conditions, the nitrate on the CaCO 3 surface increased linearly with time but did not become saturated during the experimental period, and the nitrate formation rate increased with increasing NO 2 concentration.
Using the conversion factor f =1.89×10 18 molecules/I A , the integrated absorbance of the nitrate peak in the range of 1077 and 1013 cm −1 was converted to the number of nitrate ions formed on CaCO plot of the nitrate formation rate d{NO − 3 }/dt versus the NO 2 concentration (Fig. 7).Slope of the curve was 0.94±0.10,this means that under wet conditions, the NO 2 reaction on CaCO 3 particles is first order with respect to NO 2 in the NO 2 concentration range of this study.
Here we assumed that the conversion factor f is the same under both wet and dry conditions, this assumption may cause bias in NO − 3 concentrations, because the spectra under dry and wet conditions could be different, which may influence the conversion factor f .We tried our best to reduce this possible bias: (1) the peak we used for quantitative analysis was the symmetric stretching of nitrate ν 1 (with an integration range of 1073-1013 cm −1 ), where there is no interference from other peaks.When changed from dry to wet conditions, the peak maintained the same shape and shifted slightly, from 1043 cm −1 to 1048 cm −1 ; (2) the conversion factor is also related to the sample property such as sample stack density and particle size; in our experiment the same sample preparation procedure were used for both the dry and wet experiment this should minimize the difference caused by sample property.In a future study, the relationship of the conversion factor and RH should be verified.

The influence of water vapor
Calcite is the main form of CaCO 3 in the environment and its (1014) crystal plane has the lowest energy state (de Leeuw and Parker, 1998).Both laboratory and computer modeling studies have demonstrated that at certain relative humidity, a hydrated layer on CaCO 3 could be dissociated by  oxygen vacancy to form stable −OH that remained surfacebound even in a high vacuum (Stipp et al., 1996;Mc-Coy and LaFemina, 1997;de Leeuw and Parker, 1998).
The results of Kuriyavar et al. (2000) and Stipp (1999) demonstrated that the dissociation of water could produce surface hydroxyl on CaCO 3 surfaces.Species including H + , OH + , HCO − 3 , Ca(OH) − , and Ca(HCO 3 ) − can exist on the surfaces of calcite and alter its surface properties (Thompson and Pownall, 1989).XPS analysis of our CaCO 3 particle sample showed two peaks of O1s at 531.3 eV and 533.1 eV (Fig. S5: http://www.atmos-chem-phys.net/10/463/2010/acp-10-463-2010-supplement.pdf),corresponding to the oxygen atoms of CaCO 3 and −Ca(OH) (Stipp, 1999).The oxygen atoms of −Ca(OH) accounted for 3.4-3.6% of the total number of atoms, and the oxygen atom ratio of Ca(OH):CaCO 3 was 7.3:100.The presence of −Ca(OH) implied that water can have a major effect on the CaCO 3 surface properties.Therefore, understanding water adsorption on surfaces of CaCO 3 particles is very important.
From the IR spectra of the CaCO 3 particle samples, we integrated the area of the stretching vibration of −OH from 3644 to 2983 cm −1 and plotted it as a function of RH at 296 K (Fig. S6, left axis: http://www.atmos-chem-phys.net/10/463/2010/acp-10-463-2010-supplement.pdf), it demonstrated that the proportion of surface bound water increased with RH.The NO − 3 formation rate on the CaCO 3 particle samples can be expressed as: where {} denotes the concentration of surface species (e.g.nitrate ions and active site on CaCO 3 ), [] denotes the concentration of species in gas phase, and m, n, and q are the reaction order for CaCO 3 , NO 2 and H 2 O, respectively.At the Atmos.Chem.Phys., 10, 463-474, 2010 www.atmos-chem-phys.net/10/463/2010/initial stage of the reaction, the number of NO − 3 ions formed on the surface of CaCO 3 is small compared to the surface active site, {CaCO 3 } can be considered constant.
The reaction rate of NO 2 on CaCO 3 particle was investigated as a function of RH (Fig. 8a).The nitrate formation rate decreases initially with RH and then increases slightly with the increasing RH.
The reaction order of H 2 O was obtained from slope double-log curve of the NO − 3 formation rate versus the water concentration at a constant NO 2 concentration (Fig. 8b).The reaction order of H 2 O(g) was −0.44 with RH<50%, and 0.20 with RH>50%.
The change of nitrate formation rate with RH indicates the influence of water vapor on the reaction mechanism of NO 2 on CaCO 3 particles.At low RH, the surface −OH determines the reaction rate and water competes with NO 2 for the active sites (surface −OH), and the uptake coefficient decreases with the increase in RH.
As RH increases, water starts to condense on the surface to form micro-puddles of water (Krueger et al., 2005), and gas-liquid reaction of NO 2 with condensed water may become feasible.With RH high enough (>52% in our experiment), the gas-liquid reaction of NO 2 with condensed water may become rate limiting, and the nitrate formation rate increases with RH.This is consistent with the observation that at low RH, the formation of the active site on the surface of the CaCO 3 , e.g.−OH, involved the adsorbed water, which can alter the reactivity of the particle surface (Elam et al., 1998;Hass et al., 1998;Eng et al., 2000;Hass et al., 2000).Water could easily adsorb onto the surface −OH, and compete with the adsorption of NO 2 (Stumm, 1992).At high RH, as more condensed water forms on the surface of CaCO 3 , the reaction on the calcite surface can change from gas-solid to gas-liquid reaction.Thus, water plays an important role in determining the mechanism of the heterogeneous reactions of NO 2 with CaCO 3 .Details of the mechanism will be discussed below.

Uptake coefficient
The reactive uptake coefficient or reaction probability, γ , is defined as the ratio of the reactive gas-surface collision rate to the total gas-surface collision rate.The reactive uptake coefficient of NO 2 by the surface of the CaCO 3 particles is where N(NO 2 ) is the number of reactive NO 2 collisions with the surface, Z is the total gas-surface collision rate of NO 2 on the surface of CaCO 3 , A is the surface area of CaCO 3 particles (BET surface area was used here), ω is the mean speed of the gas molecule, and n is the concentration of gas phase species.The rate of reactive collisions can be obtained from the formation rate of nitrate d{NO3−}/dt, the reactive uptake coefficient then becomes: Previous studies show that the reaction of NO 2 with particles was through N 2 O 4 (Finlayson-Pitts et al., 2003), then the reactive NO 2 loss rate in the gas phase is twice of the nitrate formation rate.The nitrate formation rate d{NO − 3 }/dt could be calculated from the curve of nitrate formation as a function of time, and the nitrate formation rate in the initial stage of the reaction www.atmos-chem-phys.net/10/463/2010/was used for calculating the initial uptake coefficient γ 0 .Table 1 lists the initial uptake coefficient of NO 2 on CaCO 3 particles under both dry (RH<10%) and wet (RH=60-71%) conditions.The initial uptake coefficient under dry conditions was (4.25±1.18)×10−9 , while under wet conditions, it is slightly lower, with a value of (2.54±0.13)×10−9 .These values are at the same level as those reported by Boerensen et al. (2000) about the reactive uptake coefficient of NO 2 on Al 2 O 3 .

Discussion
We conclude from the above results that water vapor has an important influence on the heterogeneous reaction mechanism of NO 2 with CaCO 3 .The reaction mechanism is different under dry and wet conditions.Under dry conditions, −OH is produced on the calcite surface via the dissociation of water by oxygen vacancy and this seems to be the rate determining factor for the reaction.As RH increases and more water condenses on CaCO 3 surface, surface condensed water participates in the reaction.We discuss the reaction mechanisms for both of these two conditions below.

Mechanism under dry conditions
Under dry conditions, the reaction mechanism proceeded as shown in Fig. 9a.At low RH, there was still some water vapor remaining and the surface adsorbed water on the CaCO 3 surface equilibrated with water vapor, as shown in Reaction (R1): Surface-adsorbed water existed on the CaCO 3 surface despite heating at 623 • K for 2 h.The adsorbed water on the CaCO 3 surface was dissociated by an oxygen vacancy site(O) to form the active site S•(OH)H that was stable even at high temperature and high vacuum: Finlayson-Pitts et al. ( 2003) found NO 2 was likely to form nitrogen dioxide dimers (nitrogen tetroxide, N 2 O 4 ) at high concentrations, NO 2 was in equilibrium with N 2 O 4 in the gas phase, then N 2 O 4 was adsorbed to the particle surface: N 2 O 4 reacted with the surface active site forming adsorbed nitric acid HNO 3 and nitrous acid HONO, as illustrated in Reaction (R5): Atmos.Chem.Phys., 10, 463-474, 2010 www.atmos-chem-phys.net/10/463/2010/Compared to "sticky" HNO 3 , HONO was unlikely to adsorb onto surface, but rather being released into the gas phase (Goodman et al., 1999;Borensen et al., 2000).Although we could not detect gas phase HONO, we did detect adsorbed HNO 3 with DRIFTS.
No adsorbed N 2 O 4 was observed on the particle surface, suggesting N 2 O 4 could be treated as an intermediate in the reaction: Assuming the HNO 3 formed was quickly converted to NO − 3 on the surface of CaCO 3 .Then Combining Eqs. ( 6), ( 7), ( 8), and ( 9), one can get Equation (10) shows that, at the initial stage of the reaction, the amount of surface active sites was relatively large and formation of HNO 3 on the CaCO 3 surface was a second-order reaction with respect to NO 2 .This is consistent with the measured reaction order of 1.69±0.19as illustrated in Fig. 3, even though the NO 2 concentration range is small due to the limits from experiment conditions.Underwood et al. (1999Underwood et al. ( , 2001) ) suggested that under dry conditions NO 2 might first react to form a surface nitrite, which was then oxidized by NO 2 .However, in our experiment, we did not detected surface nitrite under dry conditions, the reason for this is likely the high NO 2 concentration used in our experiment compared to the studies Underwood et al. (1999Underwood et al. ( , 2001)), and this confirms the mechanism involving the reaction of N 2 O 4 at the high NO 2 pressure of the present study.
Based on the mechanism discussed above, under dry conditions, surface nitrate should increase exponentially and reach saturation due to the depletion of surface reactive sites; which is consistent with the experimental observation at the stage I and II (Fig. 2).However, at the third stage, the nitrate was found grow continuously instead of leveling off gradually.This indicates the rate law may not work at the stage III.Other reaction process, such as the reaction of NO 2 with water in newly formed Ca(NO 3 ) 2 may contribute to the continuous formation of surface nitrate at this stage.Further study is needed to understand the mechanism at stage III.

Mechanism under wet conditions
Here, the wet conditions refer to the conditions when RH>52%.The reaction mechanism under wet conditions was illustrated by the general scheme in Fig. 9b.NO 2 was adsorbed on CaCO 3 to form adsorbed NO 2 (a): Previous studies have demonstrated that the disproportionation reaction of NO 2 with surface adsorbed water was a first-order reaction, with the adsorption of NO 2 being the rate-limiting step (Svensson et al., 1987;Jenkin et al., 1988).
The nitrite and nitrate formed on the particle surface was attributed to the reaction of surface-adsorbed water with NO 2 ; i.e., adsorbed NO 2 (a) reacted with condensed water H 2 O(a) to form adsorbed HNO 3 (a) and gaseous HONO(g).
At the initial stage, HONO dissolved in adsorbed water to form nitrite.As the concentration of HNO 3 (a) increased, the surface pH decreased and nitrite escaped from the surface as gaseous HONO.Meanwhile, HNO 3 (a) reacted with CaCO 3 to form H 2 CO 3 and Ca(NO 3 ) 2 (Al-Hosney and Grassian, 2004).
2HNO 3 (a)+CaCO 3 (s) Afterwards, H 2 CO 3 decomposed in adsorbed water to form H 2 O and CO 2 , resulting in the decrease of HCO − 3 and CO 2− 3 on the surface.
Ca(NO 3 ) 2 has strong hygroscopic property and the Ca(NO 3 ) 2 particles were in the form of solution droplets at RH>10% (Liu et al., 2008b), this caused the broadened peak for adsorbed water in the IR spectrum.
Since the adsorption of NO 2 was the rate-limiting step and HNO 3 (a) was the intermediate product, the rate of nitrate formation in the presence of excess surface water could be expressed as the following.
Equation ( 11) shows the surface nitrate formation with respect to NO 2 as a first order reaction; this is consistent with the experimental result of 0.94±0.10.Under wet conditions, N 2 O 4 could also be involved in the reaction of NO 2 with the condensed phase water, and the reaction order for loss of NO 2 was reported as 1.6±0.2(Finlayson-Pitts et al., 2003).Lee and Schwartz (1981) reported that for the NO 2 reaction with liquid water, the apparent reaction order in NO 2 could change from 1 to 2, depending on the reaction regime, such as phase mixed, convective mass transport, and diffusive mass transport.The presence of N 2 O 4 in solution is also a factor determining the apparent reaction order in NO 2 .Under convective mass transport conditions, the NO 2 reaction order should be 1.Our experiments found that the NO 2 reaction order is 0.94±0.10.However, the surface/volume ratio in our experimental was about 10 6 m −1 , this is much higher than the 100-800 m −1 interfacial area per unit liquid volume in the study of Lee and Schwartz (1981).Thus the reaction under our experimental conditions is unlikely limited by the convective mass transport; further analysis is needed to explain the reason for reaction order of 0.94±0.10 in NO 2 .
In our experiments, the NO 2 concentration range was narrow for reaction order determination: 4.81×10 15 to 1.22×10 16 molecule m −3 under dry conditions, and 4.58×10 15 to 1.14×10 16 molecules m −3 under wet conditions.This is unfortunately limited by the current experimental setup.With DRIFTS system the gas concentration can neither be too low due to the detection limit of the instrument nor too high because of the fast nitrate formation in the initial stage of the reaction, which may hinder obtaining exact reaction rate from the kinetic curve.However, this narrow range should not prevent us from discussing the reaction order, as long as we do not extrapolate the results outside the range of NO 2 concentration.Moreover, the correlation coefficient is high in Fig. 3 (R=0.98)which implies the reaction order is reliable within the NO 2 concentration range.If future experimental conditions allowed, one should expand the range of NO 2 concentrations for rate order determination.

Implications for the troposphere
Under dry conditions (RH<10%), formation of N 2 O 4 is the first reaction step.In the troposphere, the NO 2 concentrations are orders of magnitude smaller than that used in our experiments, the N 2 O 4 concentrations in the troposphere should be even lower than our experiment conditions, with a γ 0 of (4.25±1.18)×10−9 , the reaction of NO 2 on CaCO 3 under dry conditions may not have important implication in the atmosphere.
Table 1 shows that under dry and wet conditions, γ 0 were varied within a range of 2.5-4.3×10−9 , which is narrower than we anticipated.At RH>52%, with the increase of water condensed on the surface of CaCO 3 particles, the γ 0 of NO 2 on CaCO 3 particles calculated using BET surface area could have been underestimated.As significant amount of condensed water could make the surface of CaCO 3 particles smoother and less pores, the available surface area for the reaction should be close to the geometric surface area of a cubic shape particle.
When calculating the uptake coefficient, to use which type of surface area depends on the surface properties of the particles.For our experiment, at wet conditions and with significant amount of condensed water on the surface, we should use geometric surface area to calculate γ 0 .
Using 0.19 m 2 g −1 as the specific geometric surface area of CaCO 3 particle samples, from Table 1 we calculated γ 0 at RH=60-71% to be (6.56±0.34)×10 −8 ; this is more than twenty times higher than the value calculated with BET surface area.
The uptake coefficient of HNO 3 on CaCO 3 has been measured by many researchers using different methods, it was found to increase with RH (Vlasenko et al., 2006;Liu et al., 2008) and varies from 10 −5 ∼10 −4 (Goodman et al., 2000;Underwood et al., 2000;Johnson et al., 2005) at 0% RH to 0.01-0.1 at >20% RH.The large variation of the uptake coefficient of HNO 3 on CaCO 3 may due to different methods as well as the different surface areas used in calculating uptake coefficient (Fenter et al., 1995;Hanisch and Crowley, 2001).Even considering that NO 2 concentrations in the atmosphere is orders of magnitude higher than that of HNO 3 , our experimental results suggest that the reaction of NO 2 on CaCO 3 particle is unable to compete with that of HNO 3 in the atmosphere.
However, under wet conditions or RH>52%, in order to obtain reliable uptake coefficients, one needs to carefully choose the type the surface area and determine it accurately.Moreover, under wet conditions, the IR, XPS, and IC evidences for the formation of nitrite and its subsequent decrease on CaCO 3 particle suggests the reaction of NO 2 with CaCO 3 could be a source of HONO in the atmosphere.

Conclusions
The heterogeneous reaction of NO 2 with CaCO 3 under dry and wet conditions produced nitrate.Under wet conditions, nitrite formation was observed with IR, XPS, and IC, indicating the reaction of NO 2 on the CaCO 3 surface probably produced nitrous acid (HONO).Relative humidity (RH) influenced both the initial uptake coefficient and the reaction mechanism.The initial uptake coefficient of NO 2 on CaCO 3 particles γ 0 was found to change with increasing relative humidity.Using BET surface area, γ 0 under dry conditions (RH∼0%) was determined to be (4.25±1.18)×10−9 , while under "wet" conditions (RH=60-71%), it was slightly lower, with a value of (2.54±0.13)×10−9 .Our results show that relative humidity had a major influence on the surface properties of CaCO 3 particles by not only influencing the initial uptake coefficient but also by changing the reaction mechanism.At low RH, surface −OH produced via the dissociation of surface-adsorbed water by an oxygen vacancy on the CaCO 3 surface was the rate determining factor in the reaction.The reaction was second order in NO 2 .As RH increases, water starts to condense on the surface and the gas-liquid reaction of NO 2 with the condensed water begins.With high enough RH (>52% in our experiment), the gasliquid reaction of NO 2 with condensed water becomes dominant, producing HNO 3 and HONO.At this stage, the reaction appeared to be first order in NO 2 .
At high RH, condensed water could make the surface smoother with less pores, other than the BET surface area, the geometric surface area should be used to calculate γ 0 .Using 0.19 m 2 g −1 as the specific geometric surface area of CaCO 3 particle samples, we calculated γ 0 at RH=60-71% to be (6.56±0.34)×10−8 ; this is more than twenty times higher than that under dry conditions.Moreover, CaCO 3 is alkaline and can react with NO 2 to form calcium nitrate which has strong hygroscopicity and could deliquesce at RH>10%.This will help exposing CaCO 3 to NO 2 , until CaCO 3 is fully reacted.
The uptake coefficient of HNO 3 on CaCO 3 ranged from 10 −4 to 0.1, depending on RH.Our results show that the heterogeneous reaction of NO 2 on CaCO 3 is unable to compete with that of HNO 3 in producing calcium nitrate on mineral aerosol in the atmosphere.However, the NO 2 reaction with water adsorbed on CaCO 3 could be a source of HONO in the atmosphere.Further efforts to detect HONO formed in the reaction of NO 2 on CaCO 3 particles are necessary.

Figure 2 .
Figure 2. Nitrate formation as a function of time during the reaction of CaCO3 with NO2.The nitrate concentration is represented by the integrated absorbance of the peak area between 1077 and 1013 cm -1 .The NO2 concentration was changed from 4.81×10 15 to 1.22 ×10 16 molecule m -3 .The data points and the error bars are the average value and the standard deviation of three duplicate experiments.

Fig. 2 .
Fig. 2. Nitrate formation as a function of time during the reaction of CaCO 3 with NO 2 .The nitrate concentration is represented by the integrated absorbance of the peak area between 1077 and 1013 cm −1 .The NO 2 concentration was changed from 4.81×10 15 to 1.22×10 16 molecule m −3 .The data points and the error bars are the average value and the standard deviation of three duplicate experiments.

27Figure 3 .
Figure 3. Double-log curves of nitrate-producing rate versus NO2 concentration.The data points and error bars are the average values and standard deviations of three replicate experiments.The reaction order was of NO2 1.69 ± 0.19 in initial stage I (0-10 min), and 0.49 ± 0.36 in transition stage II.

Fig. 3 .
Fig. 3. Double-log curves of nitrate-producing rate versus NO 2 concentration.The data points and error bars are the average values and standard deviations of three replicate experiments.The reaction order was of NO 2 1.69±0.19 in initial stage I (0-10 min), and 0.49±0.36 in transition stage II.

Figure 6 .Fig. 6 .
Figure 6.Nitrate formation as a function of time during the reaction of CaCO 3 with NO 2 at 60-71% RH.The nitrate concentration is represented by the integrated absorbance of the peak area between 1077 and 1013 cm -1 .The NO 2 concentration was changed from 4.58×10 15 to 1.14 ×10 16 molecule m -3 .The data points and the error bars are the average value and the standard deviation of three duplicate experiments.Fig.6. Nitrate formation as a function of time during the reaction CaCO 3 with NO 2 at 60-71% RH.The nitrate concentration is represented by the integrated absorbance of the peak area between 1077 and 1013 cm −1 .The NO 2 concentration was changed from 4.58×10 15 to 1.14×10 16 molecule m −3 .The data points and the error bars are the average value and the standard deviation of three duplicate experiments.

Figure 7 .
Figure 7. Double-log plot of the nitrate formation rate versus the NO 2 concentration.The reaction was at 296°K and RH = 66 ± 2%.The data points and error bars are the average values and standard deviations of two or three replicate experiments.

Fig. 7 .
Fig. 7. Double-log plot of the nitrate formation rate versus the NO 2 concentration.The reaction was at 296 • K and RH=66±2%.The data points and error bars are the average values and standard deviations of two or three replicate experiments.

Figure 8 .Fig. 8 .
Figure 8. Nitrate formation rate during the reaction of CaCO 3 with NO 2 (a) and double-log curve of NO 3 -formation rate versus [H 2 O] (b) at different RH and a constant NO 2 concentration of 6.88 × 10 15 molecules cm -3 .Fig. 8. Nitrate formation rate during the reaction of CaCO 3 with NO 2 (a) and double-log curve of NO − 3 formation rate versus [H 2 O] (b) at different RH and a constant NO 2 concentration of 6.88×10 15 molecules cm −3 .

Figure 9 .Fig. 9 .
Figure 9. Schematics of the mechanism of NO2 reaction on the surface of CaCO3 particles under (a) dry and (b) wet conditions.

Table 1 .
Initial uptake coefficient of NO 2 on CaCO 3 particles under dry and wet conditions.The RH was measured at 296 K.