Self-limited uptake of α-pinene-oxide to acidic aerosol : the e ff ects of liquid-liquid phase separation and implications for the formation of secondary organic aerosol and organosulfates from epoxides

Introduction Conclusions References


Introduction
The reactive uptake of volatile organic compounds (VOCs) by tropospheric aqueous aerosols has recently gained attention as a potential source of secondary organic aerosol (SOA) and organosulfate species (Ervens et al., 2011;Lim et al., 2010;McNeill et al., 2012;Volkamer et al., 2009;Sareen et al., 2010;Kroll et al., 2005;Noziére et al., 2010;Galloway et al., 2009;Liggio et al., 2005).Volatile compounds can react in the particle phase (e.g., hydrolyzing or oligomerizing) after uptake to form lowvolatility products.Recently, interest has grown in reactive uptake for aerosols with significant water content (McNeill et al., 2012).Aqueous uptake and processing of organic matter may be important in explaining the extreme levels of oxidation (O : C ≥ 1) observed in secondary organic aerosol (Lee et al., 2011;McNeill et al., 2012).The focus of this study is the reactive uptake of epoxides to acidic sulfate aerosol and understanding the effects of particle composition on acid-mediated reactive uptake.
Epoxides have been identified as potential SOA precursors in both laboratory studies and thermodynamic calculations, in particular through their ability to form organosulfates (OSs) through acid-catalyzed ring opening (Iinuma et al., 2009;Lal et al., 2012;Minerath and Elrod, 2009; Paulot et G. T. Drozd et al.: Self-limited uptake of α-pinene oxide to acidic aerosol al., 2009).Recent observations of OSs in ambient samples have led to laboratory studies aimed at determining their formation mechanisms and organic precursors (Surratt et al., 2006(Surratt et al., , 2007(Surratt et al., , 2008;;Lin et al., 2011;Hatch et al., 2011;Lal et al., 2012;Minerath et al., 2008Minerath et al., , 2009;;Minerath and Elrod, 2009;Darer et al., 2011;Hu et al., 2011;Perri et al., 2010).OS yields are known to depend on particle acidity and total aerosol volume (Surratt et al., 2007;Iinuma et al., 2009;Lal et al., 2012;Hu et al., 2011).In addition, while initial OS formation may drive uptake to aerosol, less-substituted OSs or those with nearby electron-withdrawing groups may readily hydrolyze to diol compounds (Hu et al., 2011).A recent study employing low αPO concentrations reacting in bulk sulfuric acid solutions did not show significant formation of organosulfates; rather unsaturated compounds formed almost exclusively following opening of the αPO 4-member ring (Bleier and Elrod, 2013).Laboratory experiments have observed monoterpenederived epoxide uptake to extremely acidic aerosol (pH ∼ 0), but results at acidities above 0 and less than 7 have not been reported.We conducted experiments over a wide range of aerosol pH and αPO concentrations to explore the importance of epoxide uptake under ambient conditions.
Reactive uptake can be strongly affected by particle morphology and phase separation of particle organic and inorganic components.After liquid-liquid phase separation, a core-shell morphology with the organic phase coating the outer surface of the particle has been observed (Bertram et al., 2011;Smith et al., 2012;You et al., 2012;Song et al., 2012a, b;Ciobanu et al., 2009).Such a morphology change is expected to impact aerosol heterogeneous chemistry by changing the surface composition from aqueous to organic (You et al., 2012).Zuend and Seinfeld (2012) also showed via calculations that liquid-liquid phase separation can dramatically impact gas-particle partitioning of semivolatile species (Zuend et al., 2010;Zuend and Seinfeld, 2012).Particles with significant water content may exist in several morphologies.These could include a fully mixed aqueous/organic particle, a phase separated aqueous core with an organic shell, and an aqueous phase partially engulfed by an organic phase (Smith et al., 2012;Ciobanu et al., 2009;Kwamena et al., 2010;Reid et al., 2011;Song et al., 2012a).The organic-rich phase of phase-separated acidic particles may be proton-depleted, with the aqueous phase retaining the initial acidity (Pavia et al., 1999).In addition, uptake studies with bulk phase mimics of acidic sulfate aerosol particles (e.g., sulfuric acid solutions) do not replicate this complicated phase behavior and may only represent initial uptake rates to systems with low levels of organics.By studying uptake to both bulk solutions and particles over a range of organic content and acidities, we begin to elucidate the effects of particle morphology on uptake to acidic aerosol.

Aerosol reaction chamber setup and operation
All chamber experiments were conducted in a ∼ 3.5 m 3 Teflon chamber as shown in Fig. 1.The chamber is run in steady-state operation with a constant gas flow of 13 Lpm for a chamber residence time of about 4 h.Previous static chamber studies of epoxides with similar aerosol acidity, particle concentration, and αPO concentration indicate that reactive uptake reaches steady state after about 2 h (Lin et al., 2011;Iinuma et al., 2009).While uptake timescales vary with particle composition and αPO concentration, the previously observed timescales for uptake suggest our chamber residence time is long enough to accommodate complete uptake.We confirmed this experimentally, as described below.
Prior to each experiment, the bag was rinsed with deionized water and flushed with dry nitrogen to remove any material present on the chamber walls.All chamber experiments were conducted at approximately 50 % RH (relative humidity) and 25 • C, except for experiments with a particularly high particle acidity of pH = −1, which were run at 25 % RH.A hygrometer (Vaisala) was used to monitor the humidity and temperature of the chamber.The conditions for each experiment are listed in Table 1; also shown are the results of each experiment to be discussed below.
The bag was filled with a combination of three flows: humidified nitrogen, ammonium sulfate/sulfuric acid aerosol in nitrogen, and αPO in nitrogen.The final humidity was adjusted by combining a nitrogen flow that passed through a water bubbler filled with de-ionized water and a second flow of dry nitrogen.The total humid-nitrogen flow was 11 Lpm.An atomizer (TSI-3076) produced seed particles by atomizing solutions of H 2 SO 4 with (NH 4 ) 2 SO 4 (total solute concentration 0.2 M) with a nitrogen flow rate of ∼ 2 Lpm.Particle acidity was altered by adjusting the ratio of H 2 SO 4 : (NH 4 ) 2 SO 4 in the atomizing solution.Particle pH values were calculated at the designated chamber RH using E-AIM (Wexler and Clegg, 2002;Clegg et al., 1992;Carslaw et al., 1995;Clegg and Brimblecombe, 2005;Massucci et al., 1999), assuming particles have the same inorganic composition as the atomizing solutions.pH was calculated using E-AIM mole fraction-based activity data: pH = − log 10 (a H+ ).In order to achieve precise growth measurements, the atomizer output was size-selected for 150 nm particle diameter using a DMA (TSI-3080) operating at a 8 : 0.8 sheath to sample flow ( Lpm) ratio.Particle concentrations in the bag were in the range of 1000-3000 cm −3 .αPO (Sigma-Aldrich, > 97 %) vapor was injected at variable concentrations by passing nitrogen over liquid αPO held at varying temperature, which was controlled using a cold finger setup.To run below ambient temperatures, the cold finger was immersed in a dewar filled with either ice (0 • C) or an NaCl/ice bath (−20 • C).Data from typical experiments with a range of acidities and αPO concentrations are shown in Fig. 2. To show all the experiments on the same axes, the measured diameters (D t ) are normalized to the final diameter (D f for each experiment).After a stable initial diameter measurement is achieved for the seed particles (∼ 150 nm), the αPO flow was initiated.After injecting αPO for between 4 and 8 h, the particles in the chamber attain a stable output diameter.Previous studies using this chamber under similar operating conditions (Sareen et al., 2013) have shown that additional mixing did not perturb the outlet flow gas or aerosol concentrations, suggesting that at these flow rates and with the existing chamber configuration, the reactor is wellmixed.

Gas phase uptake to bulk surfaces
The phase behavior and diffusivity of αPO reactive uptake products were observed in additional experiments in which gas-phase αPO was taken up by bulk aqueous sulfuric acid samples.Four vials with 10 mL of sulfuric acid of varying concentration in water (10, 3, 1, and 0.1 M) and one vial with 3 mL of pure αPO were placed beneath a large inverted beaker.These concentrations of the bulk solutions were picked to match roughly the pH values predicted by E-AIM for the particles in the chamber experiments (−1, −0.5, 0, 1).This created a sustained exposure of the acid surface to room temperature αPO vapor (0.82 Torr, 25 • C) (Lal et al., 2012).Vials after 18 h of αPO exposure are shown in Fig. 3a.In the 10 M acid solution, a light-red   The data are aligned so that t = 0 corresponds to the beginning of particle growth.Chamber residence time was ∼ 4 h.Final growth values are attained within 4-8 h.
layer was formed at the solution surface within several hours and continued to thicken with longer αPO exposure.The 3 M solution became slightly cloudy, and none of the other solutions formed visible products from αPO exposure.The top and bottom layer of the acid solutions were extracted with a pipette, and the UV-Vis absorbance spectra of these fractions were measured.Digital photographs of the 10 M reaction vials were used to monitor the time-dependent penetration of the colored reaction products into the bulk solution and estimate the aqueous-phase diffusion coefficient of αPO.Control experiments in which the acid solutions were exposed to ambient laboratory air under the trapped beaker in the absence of αPO resulted in no color change.

Slow addition of liquid αPO to bulk acid solution
To deliver a large volume fraction of αPO, 3 mL of liquid αPO was slowly added at 750 µL h −1 to 3 mL of sulfuric acid solutions with a syringe pump to achieve a 50 % volume fraction of αPO after 3 h.In contrast to the gas-phase uptake case (Fig. 3a), for both the 10 M and 3 M acid concentrations, visible phase separation occurred; the vial with 10 M acid is shown in Fig. 3b.

Uptake of αPO to particles: effect of αPO concentration and particle acidity
The volume-growth factor, defined as the ratio between the final and initial volumes (V f /V i ), increased with the gas-phase αPO concentration.These results are displayed  in Fig. 4 and Table 1 for experiments for several particle acidities.High particle acidity and gas-phase αPO concentration resulted in very high growth factors and particle organic content.For particles with pH ∼ −0.5 and 5 ppm αPO, the growth factor is greater than 2 and the organic fraction of the particle, (V f / V i − 1) / V f , reaches nearly 50 %.
The clear trend of increased uptake with particle acidity indicates that αPO only forms SOA under conditions of reactive uptake.This is consistent with the relatively low aqueous solubility of αPO (219 mg L −1 or roughly 0.02 % by mass).Previous measurements have also shown αPO uptake to be strongly dependent on particle acidity (Surratt et al., 2006(Surratt et al., , 2007;;Lin et al., 2011;Iinuma et al., 2009;Lal et al., 2012).Iinuma et al. (2009) ran experiments with acidic (pH = 0) and neutral aerosol, but only observed uptake at pH = 0. Surratt and co-workers also observed a strong pH dependence for the reactive uptake of isoprene-derived epoxides, with greater uptake at low pH, consistent with acid-catalyzed reactions driving uptake to aerosol (Lin et al., 2011).

Partitioning coefficients
To quantify the partitioning of a gas-phase component to the aqueous aerosol, we use an effective partitioning coefficient, K p,eff (m 3 µg −1 ): where C p,tot is the increase in total particle mass concentration from gas uptake, C g the mass concentration of organic precursor in the gas phase, and C p,tot the total particle mass concentration, all expressed in µg m −3 .Given the chamber operating parameters, and assuming that the chamber behaves as a classical continuous-flow stirredtank reactor (CSTR), the final C g value in the chamber will typically be within 10 % and never less than ∼ 70 % of the input concentration of αPO (Fogler, 2009).The resulting uncertainty in K p,eff is small compared to the variation in the measured K p,eff values.The partitioning coefficient is shown as a function of particle growth factor in Fig. 5.A value of 2.8 × 10 −4 m 3 µg −1 was measured by Iinuma et al. (2009) under conditions of 50 ppb αPO, 4 × 10 −6 cm 3 m −3 seed concentration, and particles with pH = 0 (Iinuma et al., 2009).For that study this corresponded to a volume growth factor of ∼ 1.1.We measured K p,eff under similar conditions with 200 ppb αPO to be in the range of 0.4-0.6 × 10 −4 m 3 µg −1 .This is good agreement given measurement uncertainty and our observation that the partitioning coefficient decreases with increasing αPO concentration (Fig. 5).
As shown in Fig. 5, the partitioning coefficient decreases with increasing growth factor.In other words, as the organic fraction of the particle becomes greater, the affinity of αPO for the particle decreases.The trend of uptake with growth factor suggests that changes in particle composition and/or morphology upon αPO uptake play a major role in determining αPO partitioning to acidic particles.

Uptake of αPO to bulk solutions
The uptake of αPO to bulk sulfuric acid solutions was strongly pH-dependent.The reactive uptake of αPO by the 10 M H 2 SO 4 solution was made evident by the formation of a visible red layer at the top of the solution within 3 h.When left to sit over 48 h, this layer darkened and grew thicker.The mass loading of αPO in the sulfuric acid solution is low enough that phase separation is not evident.Comparing the relevant timescales in the system (uptake, reaction, and diffusion), our observation of a quickly forming yet slowly growing colored layer suggests that diffusion of organics through the liquid phase is the ratelimiting step.Due to structural similarities, we expect the diffusivities of αPO and its reaction products to be similar.We note that, from the results of Bleier and Elrod (2013), the timescale for reaction of αPO in 10 M sulfuric acid will be ∼ 5 min, suggesting a short reacto-diffusive length of ∼ 0.05 cm.Therefore, we conclude that αPO will react near the interface, and analysis of the depth of the colored layer can give an estimate of the diffusivity of the colored products.UV-Vis spectrophotometry confirms the formation of strongly light-absorbing products at high solution acidities and that this chemistry is reversible upon dilution with water (see Supplement).The change in thickness of the red layer over time allowed an estimate of the diffusivity for the reactive uptake products of αPO.After 45 h the products formed a layer roughly 1.5 cm thick.Fick's law (Eq.2) gives a relationship between diffusivity, D, the distance traveled in a given direction, x, and the elapsed time, t: (2) For our system, x is the depth of the colored layer, or the distance from the solution surface to the bottom of the colored region, and t is the elapsed time after beginning exposure to αPO.The diffusivity of the reaction products   −1 (magenta diamonds), −0.5 (red circles), 0 (blue triangles).
Partitioning coefficients increase with increasing particle acidity and decrease with increasing growth factor.
is calculated to be 9 × 10 −6 cm 2 s −1 .This value is similar to the room temperature, low-concentration limit of glucose in water (Gladden and Dole, 1953), and does not indicate particularly low diffusivity.Liquid-liquid phase separation was observed at higher mass loadings of αPO.Using a syringe pump, liquid αPO was slowly injected to the surface of sulfuric acid solutions.The flow rate was set to give a similar volume-addition rate of αPO to the acid solution as was observed in the particle uptake experiments, doubling the solution volume over 3 h.Phase separation occurred for both the 10 and 3 M sulfuric acid solutions, but the organic phase in the 3 M solutions was not colored, consistent with our other bulk experiments.The αPO concentration in a solution or particle will affect the product distribution, because a higher organic concentration will favor oligomerization.Lal et al. (2012) observed a change in the product distribution in bulk studies of the mixed liquids with different αPO : acid mass ratios, suggesting that the concentration of organics at the bulk surface or in a particle affects the product distribution.Standard organic synthesis has shown that, at least in bulk sulfuric acid solutions, the hydrolysis products involve opening of the 4-member ring in αPO (Coelho et al., 2012).This reaction mechanism was also evident in the recent work of Bleier and Elrod (2013).This chemistry is not likely to be reversible and retains a double bond that will allow for further reaction, such as oligomerization at high αPO concentrations.The estimation of relatively fast diffusion rates in these systems and the decreased K p values in high-mass loading experiments suggest that products of αPO uptake will saturate the particle volume and then phase separate with increased loading.G. T. Drozd et al.: Self-limited uptake of α-pinene oxide to acidic aerosol

Loss of water upon drying
Particles were dried after organic uptake in order to get more information about their phase/morphology.The results from drying are shown in Table 1 and also Fig. 6 as diameter growth vs. the fraction of water lost from the particle.The fraction of water lost is calculated as where V wet and V dry are the particle volumes before and after the drier, and V (H 2 O) RH,wet and V (H 2 O) RH,dry are the volumes of water at the RH conditions before and after the drier.The volumes of water were calculated using E-AIM, models I and II (Wexler and Clegg, 2002;Clegg et al., 1992;Carslaw et al., 1995;Clegg and Brimblecombe, 2005;Massucci et al., 1999).We observe a clear trend in the fraction of the total particle water lost with the increase in particle diameter.Greater particle diameter growth (i.e., a thicker organic coating) was correlated with decreased water loss from the particles.The residence time in the drier was ∼ 7 s.It may be the case that on longer timescales (from ∼ 1 min to several minutes) particles may reach equilibrium.Nonetheless, our results suggest that the organic component of the particle inhibited water evaporation, as we discuss in the following paragraphs.
Varying particle size and changes in water activity may partially explain the trend in particle organic content.Given the relation between particle diameter (d p ), diffusion coefficient (D) and equilibration time (τ D ) (Shiraiwa et al., 2011), the variation in the relative amounts of drying is greater than would be explained by the variation in particle diameter alone.As seen in Fig. 7, the variation in final particle diameter is ∼ 152-182 nm, which would only give rise to ∼ 30 % variation in equilibration timescales.The diffusivity of the particles is indeed likely to decrease with increased organic material, regardless of particle morphology, but particle size alone does not explain the variability in approach to equilibrium.This suggests that at least the effective water diffusivity may decrease significantly with higher particle fractions of αPO SOA.Another possible cause for the variation in the fraction of water loss is a change in the water activity of the particles after organic deposition.Water activity can be calculated as (Bilde and Svenningsson, 2004) where γ w is the activity coefficient for water, n w the moles of water, i s the van't Hoff factor for solute species "s", Table 1.Summary of experimental conditions.a pH values calculated using atomizing solution composition in E-AIM calculations.b Ratio of final to initial particle volume: Vf/Vi.and n s the moles of solutes "s".The variation in water activity for our systems was calculated using this equation with the molalities of the inorganic species (including water) predicted from E-AIM and assuming the αPO SOA organic material to have a density of 1 g cm −3 , van't Hoff factor of 1, and a molecular weight of 150 g mol −1 .The Raoult's law (mole fraction) part of the water activity expression only explains about 7 % of the variation in water evaporation, much less than the factor of 2 exhibited.Thus the variation in fraction of water lost would be parameterized, empirically, in varying the water activity coefficient.
The trend in water loss due to drying will be partially due to differences in equilibration timescales and the mole fractions of water present.Empirically this could be represented as a change in the water activity coefficient for particles with different amounts of αPO SOA.The large variation in activity coefficients is not inconsistent with formation of an organic coating.
Separation of the particles into water-rich and organic-rich liquid phases is suggested by our bulk-phase studies.This is consistent with recent studies by Bertram et al. (2011), which showed phase separation for particles of ammonium sulfate and organic compounds that have atomic O : C ratios of less than 0.7, and product studies of αPO + H 2 SO 4 suggest material with O : C of 0.2-0.5 is formed (Lal et al., 2012;Iinuma et al., 2009;Coelho et al., 2012;Song et al., 2012b).In their studies, Bertram et al. (2011) observed a core-shell morphology for phase-separated particles, with the organic phase on the outside.Inhibition of water loss from particles with high organic content is consistent with slow diffusion of water out of the particle through the organic-rich phase of the particle.The observed trend in water loss highlights the potential importance of phase separation in predicting water uptake/loss.

Uptake coefficients
The dynamic uptake of gases to the particle surface is determined by the uptake coefficient.It characterizes the collision efficiency for uptake of an organic molecule to a surface.Assuming a set timescale to grow to the final diameter, we can estimate this parameter, and then compare this value to previous measurements and values for similar cases.Our experiments do not directly yield the timescale to reach the final growth factor, because the experimental timescale is determined by the time to reach steady state in the continuous-flow chamber.We can use the estimate of ∼ 1.5 h from the batch reactor experiments of Iinuma et al. (2009) as a guide .Assuming a constant particle density, there is a simple relation between the mass flux and molecular collision with a particle that yields the effective uptake coefficient (Seinfeld and Pandis, 2006): where D p is the change in particle diameter, γ eff the uptake coefficient, ω the molecular speed, C gas the condensing species' gas-phase concentration, MW gas the molecular weight of the condensing species, ρ the particle density, and t the duration of condensation.As a first approximation, we can use this relation to calculate an average uptake coefficient during the course of an experiment.Using the timescale from Iinuma et al. (2009), we calculate reactive uptake coefficients between 1 × 10 −6 and 110 × 10 −6 .These values, shown in Fig. 7, are particularly low for reactive uptake of organics.Lal et al. (2012) have measured uptake coefficients for αPO to bulk sulfuric acid surfaces to be 4.6×10 −2 , and uptake coefficients for other reactive organics are in the range of 1 × 10 −3 (Lal et al., 2012;Liggio et al., 2005).The absorbing phase in the previous studies of αPO to bulk solution does not reach high volume fractions of organics, so these measurements might reflect only initial uptake of αPO to particles, before accumulation (or without the presence) of organic material.Our low values for the effective uptake coefficients show that the uptake coefficient depends on the volume fraction of organics and decreases as the particle accumulates organic material.Since αPO uptake is driven by reaction with acid in the aqueous phase, this is again consistent with the formation of an organic-rich phase at the gas-aerosol interface.

Conclusions
We have demonstrated via bulk and aerosol chamber measurements that the reactive uptake of αPO to acidic aerosol is self-limiting.Liquid-liquid phase separation at high organic loadings, supported by experiments with bulk mixtures, is a likely cause of this phenomenon.Both the effective partition coefficients and uptake coefficients decreased for particles with higher volume fractions of organics, and inhibited water loss was observed at high organic loadings.In experiments using bulk solutions, phase separation was observed for solutions with high volume fractions of αPO and its reaction products.Similar effects are possible for aqueous aerosol SOA formation in other systems with O : C ratios < 0.7 (and therefore liquid-liquid phase separation is predicted) (Bertram et al., 2011;Song et al., 2012b).In scenarios where isoprene epoxydiols or glyoxal dominates uptake to the aerosol aqueous phase, the O : C ratio is expected to be > 0.7, so this effect may not limit SOA formation via those pathways (McNeill et al., 2012).However, more experimental evidence is needed to confirm this prediction.
To date, studies have only shown uptake of monoterpenederived epoxides at an aerosol pH of zero.Our results support this fact and show that, even at an aerosol pH of 1.0, no observable uptake occurred for αPO.This suggests that, under typical ambient conditions, significant formation of monoterpene-derived SOA or organosulfate compounds is not likely to occur.Both our observations of complex phase behavior and pH-dependent product formation suggest that flow tube measurements of epoxide uptake to bulk solutions may only apply to particles with low organic content and equivalent acidity, and that particle organic content strongly affects product yields/identities.Supplementary material related to this article is available online at: http://www.atmos-chem-phys.net/13/8255/2013/acp-13-8255-2013-supplement.pdf.

Fig. 3 .
Fig. 3. (a) Photograph of vials with sulfuric acid solutions: 10, 10, 3, 1, and 0.1 M (left to right).At the far left is a control vial of 10 M acid exposed to room air for 19 h; the others were exposed to the room temperature vapor pressure of αPO for 19 h.(b) Photograph of a reaction vial with slow addition of liquid αPO (750 µL h −1 ) to 10 M acid solution.Two distinct layers are present: a dark-red bottom layer and a lighter-colored yellowish top layer.The appearance of a dark layer on top is an optical effect. Figur

Fig. 5 .
Fig. 5. Effective partitioning coefficients as a function of particle growth factor.Data are shown for three particle pH levels: Figure 5.

Fig. 6 .
Fig.6.Fraction of water removed during drying from particles after uptake of αPO vs. diameter growth.

Figure 1 .
Figure 1.Schematic of the continuous flow experimental chamber setup.

Fig. 7 .
Fig. 7. Uptake coefficient dependence on growth factor for particles in the pH range of −1 to 0. Calculated pH values for the seed aerosol are −1 (magenta diamonds), −0.5 (red circles), and 0 (blue triangles).

Table 1 .
Summary of experimental conditions.
a pH values calculated using atomizing solution composition in E-AIM calculations.b Ratio of final to initial particle volume: V f / V i .