Heterogeneous reaction of N 2 O 5 with illite and Arizona Test Dust particles

Introduction Conclusions References


Introduction
Mineral dust particles, lifted into the atmosphere from arid and semi-arid regions with a global annual flux of ∼ 2000 Tg (Textor et al., 2006), can impact direct radiative forcing by scattering and absorbing solar radiation (Balkanski et al., 2007) and also modify indirect radiative forcing by serving as cloud condensation nuclei (Twohy et al., 2009) and ice nuclei (DeMott et al., 2003;Klein et al., 2010).After being mobilized, dust particles with a mass mean diameter of < 10 µm can stay in the troposphere for a few days and be transported over thousands of kilometres (Prospero, 1999;Fairlie et al., 2010).The heterogeneous reactions of mineral dust particles during transport can directly and/or indirectly impact the levels of many important trace gases, including NO x , O 3 , and HO x radicals (Dentener et al., 1996;de Reus et al., 2005;Wang et al., 2012;Zhu et al., 2010).In addition, the chemical aging of dust particles (e.g.formation of particulate nitrate and/or sulphate) (Laskin et al., 2005;Matsuki et al., 2005;Mori et al., 2003;Sullivan et al., 2007) can modify their hygroscopicity and ability to serve as cloud condensation nuclei (Krueger et al., 2003;Shi et al., 2008;Sullivan et al., 2009;Tobo et al., 2010).Finally, heterogeneous processing can influence the ice nucleation properties of mineral dust particles (Cziczo et al., 2009;Kanji et al., 2013;Niedermeier et al., 2010;Sullivan et al., 2010).
N 2 O 5 is formed in the reaction of NO 2 with NO 3 radicals, the latter formed by the oxidation of NO 2 by O 3 (Reaction R1) (Wayne et al., 1991).N 2 O 5 thermally decomposes back to NO 2 and NO 3 radicals, leading to a dynamic equilibrium between NO 2 , NO 3 , and N 2 O 5 (Reaction R2) which is usually achieved within a few minutes under most conditions in the lower atmosphere (Crowley et al., 2010b;Osthoff et al., 2007).
N 2 O 5 plays a significant role in tropospheric chemistry by contributing to the removal of NO x and the formation of particulate nitrate (Dentener and Crutzen, 1993;Evans and Jacob, 2005;Brown et al., 2006) as well as heterogeneous chlorine activation through the formation of ClNO 2 (e.g.Osthoff et al., 2008;Thornton et al., 2010;Phillips et al., 2012;Finlayson-Pitts et al., 1989).The atmospheric NO x and O 3 burdens are sensitive to the variation of γ (N 2 O 5 ) in the range of 0.001-0.02(Macintyre and Evans, 2010).In general N 2 O 5 is only important during the nighttime because NO 3 radicals (precursor and equilibrium partner) are rapidly photolysed and react with NO during the day (Wayne et al., 1991).
The uptake of N 2 O 5 onto mineral dust particles has been investigated using bulk dust samples in a Knudsen reactor (Seisel et al., 2005;Karagulian et al., 2006;Wagner et al., 2008), airborne particles in an aerosol chamber (Mogili et al., 2006a) and in an aerosol flow tube (Wagner et al., 2008(Wagner et al., , 2009)).Recently, aerosol flow tubes with detection of N 2 O 5 by cavity ring-down spectroscopy were deployed to study the reaction of N 2 O 5 with Saharan dust aerosol (Tang et al., 2012).The same apparatus has been used in this study to investigate the heterogeneous uptake of N 2 O 5 onto airborne Arizona test dust (ATD) and illite particles.
In order to assess the atmospheric importance of the uptake of N 2 O 5 to mineral dust accurately, it is necessary to understand how strongly this parameter is correlated with the composition (mineralogy) of the dust particles.One might, for instance expect that, being a di-acid anhydride, N 2 O 5 uptake will be favoured on particles which are alkaline and/or which have a high affinity to water, so that the efficiency of uptake would be enhanced at high relative humidity (RH).
To date, the database of reliable measurements of N 2 O 5 uptake at atmospherically relevant RH is very small and the conclusions appear counter intuitive, with both positive and negative impacts on the uptake coefficient reported for increases in RH.For example, the uptake coefficient of N 2 O 5 , γ (N 2 O 5 ), onto quartz is reported to be enhanced by a factor of 4 when increasing RH from 0 % to 43 % RH (Mogili et al., 2006a).Similarly, for CaCO 3 , γ (N 2 O 5 ) increased from (4.8 ± 0.7) × 10 −3 at 0 % RH to (19.4 ± 2.2) × 10 −3 at 71 % RH (Wagner et al., 2009).In contrast, γ (N 2 O 5 ) on Saharan dust particles showed no dependence (Tang et al., 2012) or slightly negative dependence (Wagner et al., 2008) on RH.Previous aerosol flow tube studies of the uptake of N 2 O 5 onto quartz and ATD were only carried out at two different relative humidities (0 % and 29 %) (Wagner et al., 2009) and no definite conclusions regarding the effect of RH could be made.
We extend this database by investigating the effects of relative humidity on the uptake of N 2 O 5 to illite, one of the most abundant clay minerals in dust particles (Chester, 1990;Claquin et al., 1999;Nickovic et al., 2012) and one of the most efficient ice nuclei in the troposphere (Eastwood et al., 2008;Zimmermann et al., 2008).Illite, with the general formula M x [Si 6.8 Al 1.12 ]Al 3 Fe 0.25 Mg 0.75 O 20 (OH) 4 (where M is a monovalent interlamellar cation), is a non-expansive clay mineral characterized by aluminosilicate layers containing one octahedral alumina sheet sandwiched by two tetrahedral silica sheets (Hatch et al., 2012).At room temperature one monolayer of surface-adsorbed water is formed at ∼ 15 % RH (Hatch et al., 2012), and the amount of adsorbed water increases to 0.15-0.2g water per gram illite (corresponding to ∼ 60-80 formal monolayers of water) at ∼ 70 % RH (Schuttlefield et al., 2007;Hatch et al., 2012).For comparison, we have also investigated the uptake of N 2 O 5 to ATD, which is essentially ground sand from the Arizona desert and which, although possessing a mineralogy that does not correspond closely to that of globally important sources of atmospheric mineral dust aerosols (e.g.Saharan or Asian dust) has often been used as a laboratory surrogate for investigations of heterogeneous reactivity (Crowley et al., 2010a) and cloud nucleation efficiency of mineral dust particles (Sullivan et al., 2010;Vlasenko et al., 2005).Though the uptake of N 2 O 5 onto mineral dust particles has been confirmed to lead to the formation of particulate nitrate (Seisel et al., 2005;Tang et al., 2012), it is still not clear why different mineral dust components show variable heterogeneous reactivity towards N 2 O 5 (Crowley et al., 2010a).Investigation of the uptake of N 2 O 5 onto different dust components can shed light on the reaction mechanisms, and, for example, indicate which factors control the rate of heterogeneous reaction of N 2 O 5 .

Experimental
The heterogeneous reaction of N 2 O 5 with airborne ATD and illite particles was investigated using an aerosol flow tube (AFT) operated at room temperature and atmospheric pressure of N 2 .

Aerosol flow tube
A schematic diagram of the experimental set-up is given in Fig. 1.The flow tube is a vertically mounted Pyrex tube with a length of 120 cm and an inner diameter of 4.1 cm.A flow (F A + F B + F D ) of 2800 cm 3 (STP) min −1 (sccm) containing dispersed illite or ATD was introduced into the top of the flow tube via the side arm.Gaseous N 2 O 5 was eluted from a crystalline sample held at 223-233 K with a small N 2 flow (F K , 10-40 sccm) and diluted by F C to a total flow of 200 sccm.This flow was then transported through a 1/8 PFA tube (inner diameter: ∼ 1.5 mm) into the lower 10 cm of the stainless steel injector (inner diameter: 5 mm) and then into the centre of the AFT.The inner wall of the lower 10 cm of the injector was coated with Teflon (FEP) to reduce the loss of N 2 O 5 .Another small flow (F E , 10 sccm) was used to purge the annular space between the injector and the 1/8 Teflon tube.The position of the injector could be adjusted to vary the interaction time between N 2 O 5 and dust aerosols.The total flow through the reaction volume was typically 3010 sccm, resulting in a linear flow velocity of ∼ 4.2 cm s −1 and a Reynolds number of 112, indicating that the flow is laminar with an entrance length of ∼ 26 cm required to fully develop the laminar flow.The mixing length was calculated to be ∼ 42 cm (i.e. a mixing time of ∼ 10 s) (Keyser, 1984), using a diffusion coefficient of 0.085 cm 2 s −1 for the diffusion of N 2 O 5 in N 2 at atmospheric pressure (Wagner et al., 2008).
The wall of the flow tube was kept dusty and therefore highly reactive towards N 2 O 5 .In this case the loss of N 2 O 5 onto the wall was close to being gas phase diffusion limited and was thus largely independent of fluctuations in the wall loss rate constant caused, for example, by variations of the dust particle concentration in the AFT.Measurement of the N 2 O 5 wall loss rates before and after the uptake experiments confirmed that the loss of N 2 O 5 onto the wall of the flow tube was limited by gas phase diffusion.

Dust aerosol generation and characterization
The illite sample was obtained from the Source Clay Minerals Repository, University of Missouri, Columbia, USA.Arizona test dust particles (nominal 0-10 µm) were purchased from Powder Technology Inc., Burnsville, MN, USA.Dispersed illite samples were generated using a commercially available rotating brush generator (RBG), and then entrained into an 800 sccm flow (F D ).ATD samples were dispersed using a self-built aerosol generator as described in Wagner et al. (2009).The aerosol flow was diluted by the carrier gas (F A + F B , 2000 sccm) to a total flow of 2800 sccm, transported through a 1/4 aluminium tube, and delivered into the reaction volume via the side arm.The ratio of F A to F B could be varied in order to adjust the relative humidity up to 67 %.Before being diluted by the carrier gas, the aerosol flow (800 sccm) was delivered into a 5 L glass vessel (with a residence time of 6-7 min) to smooth out any spikes in the dust aerosol concentration.
At the bottom of the flow tube, particle-free air was added to increase the total flow to 5 L min −1 prior to further dilution by a factor of 20 using a TSI 3302A aerosol diluter, and measurement by a TSI 3321 Aerodynamic Particle Sizer (APS).The APS provided both the size distribution and an analogue output proportional to the aerosol number concentration and which was synchronized to the N 2 O 5 signals.The APS measures the time of flight of a particle over a fixed distance to derive the equivalent aerodynamic diameter, D a , which is the diameter of the spherical particle of unity density with the identical aerodynamic properties of the dust particle under investigation (Hinds, 1996).If the density of the particle (ρ) is known, the equivalent Stokes diameter can be derived: In this study, the density for both ATD and illite particles was assumed to be 2.7 g cm −3 .The same density was used for ATD in a previous study (Wagner et al., 2009).In order to take into account the non-sphericity of dust particles, a shape factor of 1.36 has been proposed (Hinds, 1996).In a previous study, Wagner et al. (2008) compared the time-integrated particle mass with size-resolved particle number concentration measurement using an APS and derived a correction factor of 1.6.These considerations lead us to conclude that the surface area of the dust particles might be overestimated (and thus the uptake coefficient underestimated) by a factor of up to ∼ 2. The aerodynamic size distributions of illite and ATD are displayed in Fig. 2. The average surface area, calculated using the Stokes diameters, is 21.1 µm 2 per particle for ATD and 3.56 µm 2 per particle for illite.The average surface area was calculated by dividing the total surface area (per volume) of all the particles with the number (per volume) of the particles.No significant change in particle size distribution occurred over the course of an experiment (∼ 1 h) for either ATD or illite.

N 2 O 5 generation and detection
Crystalline N 2 O 5 was synthesized by reacting NO 2 with excess O 3 in a glass reactor and trapping the product in a cold finger kept at −78 • C using a dry-ice-ethanol bath (Fahey et al., 1985).A large excess of O 3 ensured that all the NO 2 was oxidised.O 3 was generated by electrical discharge of O 2 , which had been passed through silica gel to remove any residual water vapour.We have not measured the HNO 3 impurity in this study but do not expect it to be larger than the < 5 % reported previously from this lab (Wagner et al., 2008).
N 2 O 5 was detected using a highly sensitive thermaldissociation cavity ring-down spectrometer (TD-CRD) as described previously (Schuster et al., 2009;Crowley et al., 2010b).The limit of detection was usually less than about 5 ppt (5 s sampling time).A counterflow based gas-particle separation method was deployed in order to minimise entry of particles into the TD-CRD without the use of filters.As shown in Fig. 1, ∼ 200 sccm flow (6 SLM -F H − F I ) was sampled from the flow tube through a 1/8 Teflon tube, diluted by carrier gas (F H + F I ) to a total flow of 6 SLM (standard litre per minute), and then pumped through the TD-CRD.A 200 sccm counter flow (F G ) was fed into the annular space between the 1/8 Teflon tube and a 1/4 steel tube to prevent particles being sampled.This set-up enabled particle-free air (< 1 particle cm −3 , measured by a TSI 3010 condensation particle counter) to be sampled.Efficient gasparticle separation was very important because the TD-CRD is very sensitive to aerosol light scattering, and because deposition of dust particles onto the inner wall of the sampling tubing should be avoided to minimize the loss of N 2 O 5 during transport to the TD-CRD.
The N 2 O 5 concentration in the flow tube is much greater than in the TD-CRD.This arises mainly through the dilution effect of the counter flow but is also caused by adding a large carrier gas flow (F H + F I ) to rapidly transport N 2 O 5 from the flow tube to the optical cavity of the TD-CRD.The overall dilution factor (645) was experimentally determined by introducing a known amount of NO 2 into the flow tube through the injector and measuring its post-dilution concentration at 662 nm by the TD-CRD.NO 2 , instead of N 2 O 5 , was used to determine the dilution effect because its losses through the flow tube are negligible.As NO 2 and N 2 O 5 have different diffusion coefficients, they will be differently diluted in the counterflow, which adds uncertainty (estimated as not more than ∼ 20 %) to the dilution factor.However, as the uptake kinetics is determined by the relative change of N 2 O 5 concentrations, and we show that the uptake coefficients are in any case not dependent on the initial N 2 O 5 concentration, this is not significant.

Results
Two typical data sets showing the response of the N 2 O 5 mixing ratio (in parts per trillion, pptv, where 1 pptv ∼ 2.5 × 10 7 molecule cm −3 at STP) to the introduction of illite and ATD aerosols into the flow tube are displayed in Fig. 3.The obvious anticorrelation between the N 2 O 5 mixing ratio (measured by the CRD) and the aerosol number concentration (measured by the APS) indicates substantial interaction between N 2 O 5 and the illite/ATD particles.A cursory inspection of the data shows that even short spikes in the dust concentration are accompanied by reductions in N 2 O 5 of similar duration.This indicates that, under the operating conditions used, the flow tube is in sufficiently rapid steady state to deliver accurate uptake coefficients.Figure 3 also shows that when the dust aerosol number concentration returned to 0 particles cm −3 after the dust was switched out of the reactor (after ∼ 500-550 s in Fig. 3), the measured N 2 O 5 level recovered to the initial value.This indicates that the gas-particle separation was efficient and there was no additional loss of N 2 O 5 caused by dust particles being progressively deposited onto the inner wall of the sampling tubing during the experiment.
When the number of reactive sites on the dust surface does not change significantly during the reaction time, the loss of N 2 O 5 in the flow tube can be described by (2) where [N 2 O 5 ] t and [N 2 O 5 ] 0 are the measured N 2 O 5 mixing ratios at the interaction time (of N 2 O 5 with dust aerosols) of 0 and t(s), respectively, k w is the pseudo-first-order loss rate constant of N 2 O 5 onto the wall of the flow tube (s −1 ), k d is the pseudo-first-order loss rate constant of N 2 O 5 onto dust particle surface (s −1 ), γ exp is the effective (or experimentally measured) uptake coefficient of N 2 O 5 onto dust particles, c is the average molecular speed of N 2 O 5 (24 096 cm s −1 at 296 K), N d is the dust aerosol number concentration (particle cm −3 ), and A d is the average surface area of dust particles (cm 2 per particle).The wall loss constant, k w was about ∼ 0.04 s −1 which can be compared to a first-order loss rate constant due to uptake to illite of k d ∼ 0.1 when N d = 5000 cm −3 .Note that it is the stability (rather than size) of k w during an experiment that limits the experimental accuracy.Uptake experiments were conducted by introducing bursts of dust aerosol (usually around 5-10 min in duration) into the flow tube at 5-6 different injector positions.Equations ( 2) and (3) suggest that, at each fixed contact time (t), that is, at each fixed injector position, the measured N 2 O 5 concentration, [N 2 O 5 ] t , should show an exponential dependence on the aerosol number concentration, N d , if the wall loss rate (k w ) does not change during the experiment.While the N 2 O 5 decay does not span a sufficient range to prove that these equations are appropriate, the experimental data set displayed in Fig. 4, plotting the measured N 2 O 5 concentrations versus the dust aerosol number concentration for both illite and ATD particles at three different injector positions, is consistent with this.According to Eqs. ( 2) and ( 3), the slopes of such plots, 0.25γ exp cA d t, depends linearly on the contact time, t.Typical results for ATD and illite particles, confirming the expected linear relation, are displayed in Fig. 5. Here, the slope is equal to 0.25γ exp cA d which can be used to derive the effective uptake coefficient (γ exp ) when combined with A d from the aerosol size distribution measured by the APS. Figure 5 shows that at 0 % RH the value of 0.25γ exp cA d for illite is similar to that for ATD; however, the average surface area of ATD particles is much larger than that of illite particles, suggesting that the uptake of N 2 O 5 onto illite particles is more efficient.The non-zero intercept (∼ 5 s) in Fig. 5, is the result of non-instantaneous mixing of the main-flow and the injector flow in the AFT (see above).
The rate of uptake of a trace gas onto aerosol particles is reduced by the concentration gradient close to the particle surface, resulting in an underestimation of the true uptake coefficient, γ .This effect can be corrected by using the following expression (Fuchs and Sutugin, 1970): where Kn is the Knudsen number.For mono-dispersed particles, Kn is given by where r is the radius of the particle (cm) and D(N 2 O 5 ) is the gas phase diffusion coefficient of N 2 O 5 (0.085 cm 2 s −1 at 296 K).The aerosol particles used in this study are not mono-dispersed and Kn was calculated by The ratio of the mass of absorbed water to that of dry illite (lower panel, red curve, right y axis) is also plotted as a function of RH (Hatch et al., 2012).
where N I and Kn(i) are the aerosol number concentration and the Knudsen number in the ith size bin with the radius of r I , respectively.The heterogeneous reaction of N 2 O 5 with illite aerosol particles was investigated at five different relative humidities with initial N 2 O 5 concentrations in the range of (11-21) × 10 12 molecule cm −3 .Uptake experiments were repeated at least three times at each different RH for both illite and ATD.This is the first time that the heterogeneous interaction of N 2 O 5 with illite has been investigated.As shown in the lower panel of Fig. 6, γ (N 2 O 5 ) onto illite particles was determined to be (9.1 ± 3.9) × 10 −2 , (9.3 ± 0.8) × 10 −2 , (7.2 ± 2.1) × 10 −2 , (4.9 ± 0.6) × 10 −2 , and (3.9 ± 1.2) × 10 −2 , when the RH was 0 %, 27 %, 33 %, 50 %, and 67 %, respectively.The diffusion correction factor, defined as (γ -γ exp )/γ exp , is ∼ 20 % at high RH and as large as ∼ 40 % at low RH owing to the larger uptake coefficient.
The uptake of N 2 O 5 onto airborne ATD particles was examined at five different relative humidities (RH) with initial N 2 O 5 concentrations in the range of (8-24) × 10 12 molecule cm −3 .No dependence of γ (N 2 O 5 ) on the initial N 2 O 5 concentration was found, consistent with the previous aerosol flow tube study of the uptake of N 2 O 5 onto ATD particles (Wagner et al., 2009).The result is summarized in Table 1 and shown in Fig. 6 (upper panel).
Our study suggests that γ (N 2 O 5 ) onto ATD particles is at most only weakly dependent on RH (0-67 %), and the data set may be described with an average, RH independent value of (6.3 ± 1.6) × 10 −3 (1σ ).The diffusion correction factor is only ∼ 5 % due to the relatively inefficient uptake of N 2 O 5 onto ATD particles.

Discussion
Very large uptake coefficients were observed for illite, which decreased by a factor of 2-3 as RH was increased from 0 to 67 %. Figure 6 also shows the water adsorption isotherm of illite (lower panel, red curve, right y axis) reported by (Hatch et al., 2012).One possible explanation for the decrease in (N 2 O 5 ) with increasing RH is an increasing rate of deactivation of reactive surface groups (e.g.OH) by physisorbed H 2 O.For γ to decrease with increasing RH requires that the deactivation effect must outrun the rate of increase of surface hydrolysis of N 2 O 5 as RH increases.However, as hydrolysis (solvation/ionization of N 2 O 5 to NO + 2 and NO − 3 ) will require more than one adjacent, surface-adsorbed H 2 O molecule, the latter will likely only take place under conditions of multi-layer adsorption, with an unknown RH threshold.In this context, we note that illite, with the general formula: M x [Si 6.8 Al 1.12 ]Al 3 Fe 0.25 Mg 0.75 O 20 (OH) 4 (Hatch et al., 2012), has four OH groups in each structure unit.Previous experimental and theoretical work has shown that N 2 O 5 reaction on mineral surfaces is partially controlled by the availability of surface OH groups.Seisel et al. (2005) showed that, for Saharan dust, the infrared absorption of surface OH groups at 3756 cm −1 and 3725 cm −1 decreased with exposure to N 2 O 5 and concluded that the reaction proceeds via two parallel processes: reaction with the OH groups on the mineral dust surface and the heterogeneous hydrolysis of N 2 O 5 by surface adsorbed water.
Using density functional theory to investigate the reactivity of N 2 O 5 with (Si(OH) 4 ) 2 (a simplified model of a silica surface), Messaoudi et al. (2013) concluded that surface reaction of N 2 O 5 with OH groups on the silica surface is more favorable than its hydrolysis.If similar mechanisms also operate for illite, an increase of RH will lead to "deactivation" of surface OH groups by adsorbed surface water and consequently a decrease of the overall surface reactivity towards N 2 O 5 .
In this respect it is interesting to note that at 67 % RH, the value of the uptake coefficient (∼ 0.04) is similar to γ (N 2 O 5 ) onto liquid water surface within the experimental uncertainties (Ammann et al., 2013).This may indicate that at higher RH the heterogeneous surface hydrolysis contributes significantly to the uptake of N 2 O 5 whilst at lower RH the more rapid reaction with surface OH groups dominates.The loss of surface reactivity of mineral dust at increasing RH has been previously observed for other trace gases (e.g.H 2 O 2 ) (Pradhan et al., 2010) and O 3 (Mogili et al., 2006b;Nicolas et al., 2009).
The results for ATD reveal a rather different picture, with lower uptake coefficients (factor ∼ 10 lower than illite at 0 % RH) and (at most) a weak dependence on RH.The lower uptake coefficients may be related to the mineral composition of ATD which consists mainly of feldspar and quartz (Broadley et al., 2012), which may have less (and/or less reactive) surface OH groups.The weak dependence on RH is probably related to the fact that the hygroscopic growth of ATD particles is very small (Gustafsson et al., 2005;Vlasenko et al., 2005) and therefore even at high RH, the amount of adsorbed water on the surface does not contribute significantly to N 2 O 5 solvation/ionization but may still result in deactivation of surface OH groups as shown by Goodman et al. (2001).Surface sensitive methods would be required to confirm this postulate.
Results from previous studies of the heterogeneous reaction of N 2 O 5 with airborne mineral dust particles are compiled in Fig. 7. Values of γ (N 2 O 5 ) reported by Knudsen-cell studies (Karagulian et al., 2006;Wagner et al., 2008) are significantly larger than that measured in this work, presumably the result of using bulk samples and the geometric area of the sample holders to calculate the uptake coefficients, which are then upper limits.The drawbacks of using bulk samples in investigation of heterogeneous reactions have been documented previously (Crowley et al., 2010a) and results from bulks samples are not considered further here.
The two sets of uptake coefficients for Saharan dust reported by this group (Wagner et al., 2008;Tang et al., 2012) differ by about a factor of two and display a different dependence on relative humidity.Reasons for this, related to the inferior detection and sampling scheme used in the earlier study, are discussed by Tang et al. (2012).
Figure 7 shows that the uptake coefficients measured for illite in this study are the largest to date followed by Saharan dust (SDCV).The difference between uptake coefficients for  illite and Saharan dust is especially significant at low humidity and may be related to the presence of reactive, surface-OH groups on illite as discussed above.The lowest uptake coefficients are observed for dry samples of ATD and calcite.For ATD our results agree well with those reported by Wagner et al. (2009), in which an aerosol flow tube with similar flow conditions as this study was used.The present study extended the RH range investigated to 67 % and confirms the weak trend in γ (N 2 O 5 ) (slightly negative) with increasing RH.
Figure 7 highlights the importance of studying heterogeneous reactions of N 2 O 5 on mineral dust at atmospherically relevant relative humidities as different minerals or mineral dust components display different behavior.Indeed, while CaCO 3 is the least reactive at low humidity, the uptake coefficient of N 2 O 5 at RH close to 70 % is similar for both CaCO 3 and Saharan dust, and only slightly lower than for illite (which displays the opposite trend with RH as described above).The significant increase of γ (N 2 O 5 ) onto CaCO 3 at high RH bears resemblance to its water adsorption isotherm (Gustafsson et al., 2005) and is likely to be related to the formation of more reactive Ca(OH)(CO 3 H) on the surface at higher RH (Al-Hosney et al., 2005).
The lack of RH dependence of γ (N 2 O 5 ) onto Saharan dust particles is potentially related to the fact that Saharan dust is a complex mixture of different minerals which have different heterogeneous reactivity towards N 2 O 5 , which may also vary both in positive or negative sense with increasing RH.We note that the reactivity of N 2 O 5 on illite and Saharan dust particles tends to a common value at high RH, suggesting that, under these conditions, heterogeneous hydrolysis drives the N 2 O 5 uptake onto clay minerals.Indeed, the uptake coefficients at high RH are similar in magnitude to that of N 2 O 5 onto aqueous solutions (Ammann et al., 2013).
Further kinetics and mechanistic investigations of the effect of relative humidity on the uptake of N 2 O 5 onto other important minerals present in dust particles (e.g.kaolinite and feldspars) will help to improve our understanding of the heterogeneous reaction of N 2 O 5 with mineral dust particles and allow us to transfer laboratory derived uptake coefficients to the atmosphere with more confidence.The present database (Crowley et al., 2010a) on the uptake of N 2 O 5 (and indeed most trace gases) to mineral aerosol is too incomplete to enable accurate, mineralogically distinct (or dust source region dependent) global modelling of the effects of N 2 O 5 uptake to dust particles.As long as this situation persists, average uptake coefficients, derived from experiments on Saharan dust samples, may represent the best alternative.Arizona dust is unlikely to be a useful alternative in this regard.

Conclusions
The heterogeneous interaction of N 2 O 5 with illite and Arizona test dust (ATD) particles was investigated using an aerosol flow tube at room temperature and different relative humidities (up to 67 %).For illite, the uptake coefficient was found to be very large and to decrease from (9.1 ± 3.9) × 10 −2 at 0 % RH to (3.9 ± 1.2) × 10 −2 at 67 % RH.Much lower uptake coefficients, (6.3 ± 1.6) × 10 −3 (1σ ), and a significantly weaker dependence on RH was found for ATD.Whilst the uptake coefficients for interaction of N 2 O 5 with illite calcite and Saharan dust tend to a common value of ∼ 0.03 at high relative humidities, the much weaker interaction of N 2 O 5 with ATD (γ ≈ 0.004 at similar RH) may indicate that its common usage as a laboratory surrogate for atmospheric mineral dust may not necessarily lead to conclusions that can be applied with confidence to the mineralogically distinct "global mineral dust" which is dominated by emissions from the Saharan and Asian desserts.

Fig. 2 .
Fig. 2. Typical size distribution of ATD and illite particles measured at the downstream end of the aerosol flow tube.

Fig. 3 .
Fig. 3. Response of the N 2 O 5 mixing ratio (left axis) to the introduction of mineral dust aerosols (right y axis) into the flow tube at RH = 0 % when the injector was at 70 cm.The time between acquisition of neighbouring data points is ∼ 3 s.

Fig. 4 .
Fig. 4. Exponential dependence of the measured N 2 O 5 mixing ratio on the dust aerosol number concentration at three different injector positions at RH = 0 %.

Fig. 5 .
Fig. 5. Plot of 0.25γ exp cA d t, versus the contact time (t) for N 2 O 5 with dust aerosols in the AFT.The lines are least-squares fits to the illite (solid line) and ATD (dashed line) data sets.The error bars are statistical (2σ ) only.

Fig. 6 .
Fig.6.Uptake coefficients of N 2 O 5 at different relative humidities to illite and ATD.The ratio of the mass of absorbed water to that of dry illite (lower panel, red curve, right y axis) is also plotted as a function of RH(Hatch et al., 2012).

Figure 7 .
Figure 7.Comparison of relative humidity dependence of N coefficients to airborne mineral dust particles.

Fig. 7 .
Fig. 7. Comparison of relative humidity dependence of N 2 O 5 uptake coefficients to airborne mineral dust particles.

Table 1 .
Uptake coefficients for N 2 O 5 on illite and ATD.