Interactive comment on “ Heterogeneous reaction of peroxyacetic acid and hydrogen peroxide on ambient aerosol particles under dry and humid conditions : kinetics , mechanism and implications

Abstract. Hydrogen peroxide (H2O2) and organic peroxides play important roles in the cycle of oxidants and the formation of secondary aerosols in the atmosphere. Recent field observations have suggested that the budget of peroxyacetic acid (PAA, CH3C(O)OOH) is potentially related to the aerosol phase processes, especially to secondary aerosol formation. Here, we present the first laboratory measurements of the uptake coefficient of gaseous PAA and H2O2 onto ambient fine particulate matter (PM2.5) as a function of relative humidity (RH) at 298 K. The results show that the PM2.5, which was collected in an urban area, can take up PAA and H2O2 at the uptake coefficient (γ) of 10−4, and both γPAA and γH2O2 increase with increasing RH. The value of γPAA at 90 % RH is 5.4 ± 1.9 times that at 3 % RH, whereas γH2O2 at 90 % RH is 2.4 ± 0.5 times that at 3 % RH, which suggests that PAA is more sensitive to the RH variation than H2O2 is. Considering the larger Henry's law constant of H2O2 than that of PAA, the smaller RH sensitivity of the H2O2 uptake coefficient suggests that the enhanced uptake of peroxide compounds on PM2.5 under humid conditions is dominated by chemical processes rather than dissolution. Considering that mineral dust is one of the main components of PM2.5 in Beijing, we also determined the uptake coefficients of gaseous PAA and H2O2 on authentic Asian Dust storm (ADS) and Arizona Test Dust (ATD) particles. Compared to ambient PM2.5, ADS shows a similar γ value and RH dependence in its uptake coefficient for PAA and H2O2, while ATD gives a negative dependence on RH. The present study indicates that, in addition to the mineral dust in PM2.5, other components (e.g., soluble inorganic salts) are also important to the uptake of peroxide compounds. When the heterogeneous reaction of PAA on PM2.5 is considered, its atmospheric lifetime is estimated to be 3.0 h on haze days and 7.1 h on non-haze days, values that are in good agreement with the field observations.


Introduction
Peroxide compounds, including hydrogen peroxide (H 2 O 2 ) and organic peroxides, play an important role in the chemistry of the atmosphere, because they serve as oxidants for the conversion of S(IV) to S(VI) in the atmospheric aqueous phase, resulting in the formation of sulfate aerosol (Calvert et al., 1985;Lind et al., 1987;Stein and Saylor, 2012).Peroxide species also serve as a reservoir for HO x (OH and HO 2 ) radicals (Wallington and Japar, 1990;Vaghjiani and Ravishankara, 1990;Atkinson et al., 1992;Ravetta et al., 2001) and RO x (RO and RO 2 ) radicals (Lightfoot et al., 1991;Reeves and Penkett, 2003).Moreover, recent laboratory studies have indicated that peroxide compounds, especially organic peroxides, significantly contribute to the formation and aging of secondary organic aerosols (SOA) (Claeys et al., 2004;Docherty et al., 2005;Surratt et al., 2006;Paulot et al., 2009;Huang et al., 2013;Xu et al., 2014;Zhao et al., 2015).
Published by Copernicus Publications on behalf of the European Geosciences Union.

Q. Q. Wu et al.: Heterogeneous reaction of peroxyacetic acid and hydrogen peroxide
The peroxide compounds are mainly produced by the bimolecular reaction of HO 2 and RO 2 radicals (e.g., reactions R1 and R2), and their minor sources include the ozonolysis of alkenes and biomass burning (Lee et al., 2000).
Their traditional removal pathways include reacting with OH radicals, photolysis and deposition (Lee et al., 2000).Recent studies have combined field and model data to ascertain the importance of heterogeneous loss.For example, de Reus et al. (2005) have demonstrated that, on the subtropical island, the concentration of gaseous H 2 O 2 was largely overestimated by a standard gas phase chemical mechanism.When the heterogeneous uptake of H 2 O 2 and/or HO 2 radicals on the surface of aerosols was accounted for in the model, the observed and modeled values were in better agreement.In addition, a series of laboratory studies have addressed the importance of the heterogeneous reaction of H 2 O 2 on model or authentic mineral dust particles (Pradhan et al., 2010a, b;Wang et al., 2011;Zhao et al., 2011aZhao et al., , b, 2013;;Romanias et al., 2012Romanias et al., , 2013;;Zhou et al., 2012;El Zein et al., 2014).For example, Pradhan et al. (2010a) have indicated that the heterogeneous reaction of H 2 O 2 on dust aerosols could compete with its photolysis and significantly affect the HO x radical budget.Romanias et al. (2012Romanias et al. ( , 2013) ) have confirmed that the heterogeneous reaction of H 2 O 2 on mineral dust had an important effect on the fate of HO x radicals.El Zein et al. (2014) also suggested that the lifetime of H 2 O 2 removed by heterogeneous reaction was comparable with its photolysis on severe dust storm periods.Our recent study has indicated that H 2 O 2 could enhance the uptake of oxygenated volatile organic compounds (OVOCs) onto the surface of mineral dust particles (Zhao et al., 2014).
To the best of our knowledge, to date, there has been no laboratory experimental evidence for the importance of the heterogeneous reactions of organic peroxides in the atmosphere.As an important organic peroxide, peroxyacetic acid (PAA, CH 3 C(O)OOH) has been frequently detected over both rural and urban areas (Lee et al., 1995;Hua et al., 2008;He et al., 2010;Zhang et al., 2010;Liang et al., 2013;Phillips et al., 2013).The typical concentration of PAA is comparable to that of H 2 O 2 , i.e., several tens to hundreds of pptv in summer, and the maximum concentration surpasses 1 ppbv over Mazhuang, a rural site in Shandong Province, China (Zhang et al., 2010), and the boreal forest (Phillips et al., 2013).Our field observation results have suggested that heterogeneous reactions on aerosol particles might be an important removal pathway for PAA in the atmosphere (Zhang et al., 2010;Liang et al., 2013).Therefore, we use PAA as a representative organic peroxide to investigate the kinetics and mechanisms of its heterogeneous reactions on ambient PM 2.5 as well as mineral dust particles over a wide range of relative humidities (3-90 %).We also estimate the contri-bution of heterogeneous reactions to the PAA budget in the atmosphere.As a comparison, we investigate the kinetics of H 2 O 2 uptake on PM 2.5 .

Generation of gaseous PAA and H 2 O 2
PAA aqueous solution was synthesized by mixing H 2 O 2 aqueous solution with acetic acid aqueous solution, using H 2 SO 4 as a catalyst (Dul'neva and Moskvin, 2005;Zhao et al., 2007).The mixing aqueous solution was kept in the dark for 24 h at room temperature to make sure PAA reached its maximum balanced concentration.The PAA concentration in this primary solution (S1) was 1.3 M. The solution was stored at 277 K in the dark before use.At the beginning of every experiment, a PAA solution (S2) (4 × 10 −5 M) was prepared by diluting S1 with ultrapure water and then 100 mL S2 was added into a 1 L bubbler.The S2 in the bubbler, kept in a 277 K water bath in the dark, was bubbled by a N 2 flow at a rate of 70 standard mL min −1 to generate gaseous PAA.The PAA-containing N 2 flow was then mixed with an O 2 flow and water vapor flow generated by another bubbler.The resulting gas mixture (reactant gas) was introduced into the filterbased flow reactor as described in Sect.2.2.3.A H 3 PO 4 solution (5 × 10 −3 M) was used to scrub gaseous peroxide in a glass scrubbing coil.The collection efficiency was 85 % for PAA and 100 % for H 2 O 2 at 277 K (Hua et al., 2008;Liang et al., 2013).The peroxide-containing scrubbing solution was analyzed immediately by an online high performance liquid chromatography (HPLC, Agilent 1200).The method is described in detailed in Sect.2.3.The concentration of PAA was 300 ± 30 pptv in the gas mixture.To ensure a constant concentration of gaseous PAA, the bubbling solution of PAA was renewed every day.The generation of gaseous H 2 O 2 was similar to that of PAA.The concentration of the bubbling solution of H 2 O 2 was 1.14 × 10 −3 M. The resulting H 2 O 2 concentration was 510 ± 40 pptv in the gas mixture.

Preparation of particle-loaded filters
The PM 2.5 samples were collected on the roof of a six-story teaching building (26 m above the ground) at the campus of Peking University (PKU), Beijing, China.PKU is located in the northwestern downtown area, with two major traffic arteries passing by.An ambient particulate sampler (TH-16A, Wuhan Tianhong Instruments Co., Ltd) was used to collect the PM 2.5 particles on the PTFE filters with four parallel channels operating simultaneously, and the sampling flow of each channel was 16.7 standard L min −1 .The PM 2.5 samples were collected for 6 days, from 31 July to 06 August 2014.Sampling was conducted twice a day for 11.5 h each time (daytime, 09:00-20:30 LT; nighttime, 21:00-08:30 LT).During the sampling period, 31 July to 03 August were haze days and 03 to 05 August were non-haze days.Haze is caused by a large number of fine particles (e.g., dust, smoke, salt) with RH less than 90 % and makes the visibility less than 10 km (Li, 2010).Here, we differentiated non-haze days from haze days based on two criteria.One was the visibility of a mountain (by eye) that is about 10 km away from the sampling site.The other was the national ambient air quality standard grade II in China, i.e., an average PM 2.5 mass concentration of lower than 75 µg m −3 .The PM 2.5 particle-loaded filters were sealed and kept at 255 K before use.ADS and ATD particles were separately used to prepare the mineral dust particle-loaded filters.Mineral dust particles were resuspended using a custom-built resuspension apparatus and then collected on the PTFE filters.The resuspension apparatus consists of three parts, i.e., a glass inlet, a stainless filter holder and a vacuum pump.First, we put a known number of mineral dust particles into the glass inlet and then turned off the inlet.Secondly, we turned on the vacuum pump and a negative pressure was then formed in this resuspension system.Finally, we turned on the inlet, and the particles were resuspended with the help of airflow and collected onto the filter.To compare the experimental results for PM 2.5 sampled on non-haze and haze days, the mass of ADS or ATD on the filter was carefully controlled at 0.3 and 1.3 mg for the lower and higher particle mass, respectively.

Uptake experiments
A filter-based flow reactor was used to measure the uptake coefficients of gaseous peroxides on aerosol particles.The schematic of this experimental apparatus is shown in Fig. 1.The reactor is composed of two perfluoroalkoxy resin filter holders (Savillex Corporation) connected in parallel.One reactor contains a blank PTFE filter, and the other contains a particle-loaded filter.The tubing system is made of Teflon tubes.The peroxide-containing gas mixture (20 % O 2 + 80 % N 2 ) was used at a flow rate of 2.7 standard L min −1 and was introduced into the blank reactor or the particle-loaded reactor via two unreactive stainless steel valves.After exiting the reactor, the peroxide-containing gas was directed into a glass scrubbing coil in a 277 ± 0.1 K water bath, in which a H 3 PO 4 solution (5 × 10 −3 M) was used as the eluent to scrub the peroxide at a rate of 0.2 mL min −1 .The same particleloaded filter was used to measure the uptake coefficient at a continuously increasing RH ranging from 3 to 90 % and then the measurement was repeated in reverse, at a decreasing RH from 90 to 3 %.We have compared the uptake coefficients of PAA on the exposed PM 2.5 filter that has been used in the PAA uptake experiments and the unexposed PM 2.5 filter that has not been used for any experiments at 60 % RH, and no obvious difference was observed between the two uptake coefficients (Table 1).Therefore, we think the reuse of the filter for experiments at different RH has no significant effect on the results.
The uptake experiment at a certain RH took 2 h for PAA and 1 h for H 2 O 2 , including the time for the balance of peroxide on a blank filter and a particle-loaded filter.The balance concentrations of PAA/H 2 O 2 have been detected for at least three times.Then the RH was directly changed to another RH without any treatment for the filter samples.All the experiments were conducted at 298 ± 2 K, ambient pressure and in the dark.
The uptake coefficients of gaseous peroxide can be calculated using the following equations (Molina et al., 1996;Zhao et al., 2010): where {C} is the total uptake of gaseous peroxide by particle surfaces, molecules; Z is the collision frequency, molecules s −1 ; [C] is the number concentration of gaseous peroxide, molecules m −3 ; ω is the mean molecular speed, m s −1 ; R is the universal gas constant, kg m 2 s −2 mol −1 K −1 ; T is the temperature, K; A es is the effective surface area of particles, m 2 ; M x is the molecular weight, kg mol −1 .The uptake onto the particles is equal to the loss of the gaseous re- actant and this gas phase loss can be calculated by the difference between the reactant concentrations at the inlet and outlet of the reactor.Here, we define the fractional loss of the reactant (L f ) as Eq. ( 4): where [C] in and [C] out are the concentrations of the reactant at the inlet and outlet of the reactor, molecules m −3 , respectively.Since no obvious uptake of peroxide on the blank filter occurred, the reactant concentration at the outlet of the blank reactor can be treated as the initial concentration at the inlet of the reactor for the uptake on aerosols.Therefore, Eq. ( 1) can be expressed as Eq. ( 5): where V g is the flow rate of the reactant-containing gas, m 3 s −1 .The values of γ on PM 2.5 , ADS and ATD particles in the next test are calculated by the A es estimated in Sect.2.4.

Analysis of peroxides, soluble species and elements
Peroxide compounds were measured by HPLC coupled with a post-column derivatization module.The length of the column is 150 mm (Alltima AQ 5µ).The details of this method have been reported in our previous study (Hua et al., 2008).Briefly, this method is based on the determination of the fluorescent dimer produced by the reaction of POPHA and peroxides with the catalysis of hemin.The HPLC mobile phase was a H 3 PO 4 solution (pH = 3.5) at a flow rate of 0.5 mL min −1 .The formed fluorescent dimer was analyzed by a fluorescence detector.The time of collecting a chromatogram was 10 min for PAA and 5.0 min for H 2 O 2 .The retention times of PAA and H 2 O 2 were 8.9 and 4.0 min, respectively.05 Aug a 2.30 × 10 −4  2.40 × 10 −4  05 Aug b 2.45 × 10 −4 2.33 × 10 −4 Note: a daytime; b nighttime; exposed PM 2.5 particles, which have been used in the PAA uptake experiments; unexposed PM 2.5 filter, which has not been used for any experiments.
We used the ultrasonic method to extract the soluble compounds in particle samples.Each sample was exposed to ultrasonic treatment in ice water with 10 mL Milli-Q water for 30 min.The extracted soluble compounds were measured by ion chromatography (IC, Dionex ICS2000 and ICS2500).The analytical columns for cations and anions were Dionex CS 12A and Dionex AS 11, respectively.Here, the measured compounds include eight inorganic ions (i.e., K + , Ca 2+ , Na + , Mg 2+ , NH + 4 , Cl − , NO − 3 and SO 2− 4 ) and four organic acids (i.e., formic acid, acetic acid, pyruvic acid and oxalic acid).
We used acid digestion to extract elements in particles through a microwave digestion system (CEM MARS, USA).Elements in the extractions were measured by inductively coupled plasma mass spectroscopy (ICP-MS, Thermo X series).The measured elements include Mg, Al, P, Ca, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, As, Se, Mo, Cd, Ba, Tl, Pb, Th and U.

Estimation of effective surface area
The effective surface area (A es ) is a key factor in the uptake of a specific compound from the gas phase onto aerosol particles.The uptake coefficient (γ ) estimated by the geometric filter surface area (A gs ) is several orders of magnitude higher than that by the Brunauer-Emmett-Teller (BET) surface area (Shen et al., 2013).To date, accurate estimation of A es of the particle sample has been a challenge for the determination of γ .Bedjanian et al. (2013) have measured the uptake of HO 2 radicals on ATD particles and showed a pseudo-logarithmic relationship between the uptake and the particle mass.In the present study, ambient particles were loaded on the filter in an agglomerated state, extremely different from their status in the atmosphere, where they are highly dispersed.Obviously, neither the geometric surface area nor the BET surface area can represent the A es of the ambient particle samples on the filter.Here, we estimated A es by investigating the relationship between the uptake and loaded particle mass.Eq. ( 5) shows that the fractional loss (L f ) of a specific gaseous reactant due to the uptake of the filter-loaded particles was directly proportional to A es .The value of A es should depend on the loaded particle mass.Therefore, we used the relationship between L f and particle mass (M a ) to estimate the value of A es .Figure 2 shows the relationship between the L f of gaseous PAA versus the loaded particle mass.Although L f appeared to have a linear relationship with particle mass in the low particle mass region, it generally fit with the logarithmic function of particle mass, with a correlation coefficient r = 0.88.This empirical logarithmic relationship is given in Eqs. ( 6) to ( 8): For ADSL f = 0.099 × ln(M a ) + 0.26, ( 7) where M a is the mass of the particles, mg.The mass used for measuring the L f of ADS is 0.18, 0.37, 0.81, 1.05, 1.16, 1.63, 1.86 and 2.46 mg, respectively.The mass for measuring the L f of ATD is 0.27, 0.48, 0.83, 1.07, 1.36, 1.58, 1.76, 2.02, 2.57 and 3.00 mg, respectively.In the low particle mass region, the particles were highly dispersed on the filter and A es increased rapidly with increasing particle mass; in the high particle mass region, particles highly overlapped and agglomerated with each other on the filter, and A es was closer to A gs (12.43 cm 2 ).Here, we assume that there exists a critical particle mass (M a,c ) for which A es is equal to A gs .When the particle mass is greater than M a,c , A es tends to be constant, i.e., the A gs .For M a,c , the corresponding fractional loss of PAA is L f c .We used an iterative method to determine M a,c .The termination criterion of this iterative method was the relative error (R el ) of L f c towards the average of all the L f values (L f ) that were larger than L f c and the calculation method was expressed in Eq. ( 9).Here, we set R el as 5 % to terminate the iteration.The procedure of the iteration was as follows: (i) start M a,c with 0.10 mg; (ii) calculate series values of L f by inputting a range of M a (0.01 to 2.00 mg) into Eq.( 6); (iii) calculate R el by the calculated values of L f and Eq. ( 9); (iv) if R el is larger than 5 %, reset M a,c with an added mass of 0.01 mg (i.e., 0.11 mg); (v) repeat steps (ii-iv) until R el is less than 5 %, and then obtain the expected M a,c and L f c .The calculated L f c was 4.89 × 10 −1 , which was similar to the experimental result in Fig. 2., i.e., 4.90 × 10 −1 .Based on the directly proportional relationship between A es and L f , A es can be expressed in Eq. ( 10).
The estimation of A es for filter-loaded PM 2.5 , ADS and ATD particles can be expressed as the respective logarithmic functions in Eqs. ( 11) to (13): For PM 2.5 A es = 3.75 × ln(M a ) + 12.0, For ADSA es = 3.66 × ln(M a ) + 9.59, For ATDA es = 3.01 × ln (M a ) + 10.3, where M a represents the filter-loaded particle mass, mg; A es represents the effective surface area of particles, cm 2 .The mass of the filter-loaded PM 2.5 and the estimated A es values are listed in Table 2.A es for PM 2.5 changes with the particle mass, ranging from 3.2 to 13.8 cm 2 ; A es for ADS is 6.1 and 10.9 cm 2 , respectively; A es for ATD is 6.4 and 11.2 cm 2 , respectively.The uptake coefficients on PM 2.5 particles, ADS and ATD particles below are all calculated with these A es values.

Results and discussion
3.1 Uptake of PAA and H 2 O 2 on PM 2.5 The uptake coefficient of PAA (γ PAA ) on PM 2.5 particles was measured over a wide range of RH (3-90 %). Figure 3 shows the γ PAA profile on PM 2.5 with respect to increasing/decreasing RH. γ PAA increases with increasing RH on ; ADS h and ATD h , the mass of mineral dust about 1.3 mg; ADS l and ATD l , the mass of mineral dust, about 0.3 mg; a uptake coefficient calculated by total surface area of the particles using size distribution, representing the lower limit; b uptake coefficient calculated by BET area, representing the lower limit; the errors represent the relative standard deviation between γ on particles of ascending and descending RH.
both daytime and nighttime PM 2.5 samples.The values of γ PAA on nighttime PM 2.5 samples are similar to those on daytime PM 2.5 samples.Additionally, although the mass of PM 2.5 collected on a haze day is significantly different from that on a non-haze day, the γ PAA values are similar under these two different weather conditions (Table 3).In general, γ PAA rises from (0.89 ± 0.26) × 10 −4 at 3 % RH to (4.41 ± 0.92) × 10 −4 at 90 % RH.Table 3 also lists the lower limit of γ PAA on PM 2.5 , which are calculated using the total surface area of the particles using size distribution (see the details in Sect.4, and Eqs.21 and 22).The lower limit is on the order of 10 −6 -10 −5 .The empirical equation of γ PAA plotted against water activity (a H 2 O ; here, a H 2 O = RH/100) can be expressed as Eq. ( 14) and the measured and modeled γ PAA on PM 2.5 are shown in Fig. 4.
We also determined the uptake coefficients of H 2 O 2 on PM 2.5 over the RH range of 3 to 90 %.Before this experiment, we compared the measured uptake coefficients of H 2 O 2 on two PM 2.5 samples; one had been used to measure the uptake coefficient of PAA and the other had not been used for any measurements.The results show that the relative error between the above two experiments was 1.0-7.4% among different RH (3-90 %).Therefore, there is no obvious difference between the uptake coefficients of H 2 O 2 on used and unused PM 2.5 samples.Figure 5 shows the γ H 2 O 2 on PM 2.5 that had been used to measure γ PAA , over 3-90 % RH.The empirical equation of γ H 2 O 2 as a function of a H 2 O can be expressed as Eq. ( 15) and the measured and modeled γ H 2 O 2 on PM 2.5 is shown in Fig. 4.
The value of γ H 2 O 2 , similar to γ PAA , shows a positive correlation with RH.The average value of γ H 2 O 2 changes from (1.12 ± 0.20) × 10 −4 at 3 % RH to (2.70 ± 0.37) × 10 −4 at 90 % RH.The positive RH dependence of γ H 2 O 2 has been reported by Pradhan et al. (2010b).They have measured γ H 2 O 2 on authentic mineral dust particles (i.e., Gobi dust particles and Saharan dust particles).Figure 6 shows the ratios of γ PAA,90 % RH to γ PAA,3 % RH (R γ PAA ) and γ H 2 O 2 ,90 % RH to γ H 2 O 2 ,3 % RH (R γ H 2 O 2 ).Although the R γ PAA values are more variable on haze days than those on non-haze days, the average value of R γ PAA shows no obvious difference at different times and under different weather conditions, varying over the narrow range of 4.4 ± 0.6 to 6.3 ± 2.7.On average, R γ PAA is 5.4 ± 1.9.It is interesting to note that R γ H 2 O 2 is 2.4 ± 0.5 (see Fig. 6b), which is much lower than R γ PAA .Although γ H 2 O 2 has a positive RH dependence on PM 2.5 as well, H 2 O 2 is less sensitive to RH variance compared to PAA.For peroxide compounds, if a physical process, especially the dissolution, dominates their uptake on PM 2.5 , the R γ H 2 O 2 should be larger than R γ PAA , because the Henry's law constant of H 2 O 2 is 100 times larger than that of PAA (298 K) (8.47 × 10 2 M atm −1 for PAA and 8.43 × 10 4 M atm −1 for H 2 O 2 ) (O' Sullivan et al., 1996).This expectation, however, is at odds with our experimental results.Hence, we speculate that the physical pro-  cess is not the main pathway for the uptake of peroxide compounds on PM 2.5 .In addition, the values of γ PAA and γ H 2 O 2 on PM 2.5 were measured with increasing RH from 3 to 90 % and then the measurements were repeated by using the same sample with decreasing RH from 90 to 3 %.Interestingly, we find that the γ PAA and γ H 2 O 2 can be well repeated in these two cases (see Figs. 3 and 5).The independence of γ PAA and γ H 2 O 2 from reaction time also indicates that PM 2.5 has a sustained reactivity for the uptake of peroxide compounds at different RH, which falls into the category of reactive uptake as suggested by Crowley et al. (2010).The detailed mechanism is described in Sect.3.3.The present study is the first investigation on the kinetics of the heterogeneous reactions of PAA and H 2 O 2 on PM 2.5 particles.Recent studies have already indicated the importance of mineral dust for H 2 O 2 uptake (Pradhan et al., 2010a, b;Wang et al., 2011;Zhao et al., 2011aZhao et al., , b, 2013;;Romanias et al., 2012Romanias et al., , 2013;;Zhou et al., 2012;El Zein et al., 2014).For PAA, however, no data regarding its kinetics on mineral dust have been available in the literature.Therefore, we investigated the heterogeneous reaction of PAA on mineral dust as a comparison of that on PM 2.5 .

Uptake of PAA and H 2 O 2 on mineral dust
Mineral dust is an important component of atmospheric aerosols in Beijing; it comprises 6.0 and 6.2 % of PM 2.5 on haze days and non-haze days, which is similar to the reported values (7.1-12.9%) (Sun et al., 2004;Yang et al., 2011;Zhang et al., 2013).To determine whether the mineral dust dominates the uptake of PAA on PM 2.5 , we measured the γ PAA on two kinds of mineral dust particles, i.e., ADS and ATD particles.The measured γ PAA values are listed in Table 3. γ PAA on low mass ADS (ADS l ) increases from (0.84 ± 0.01) × 10 −4 at 3 % RH to (3.21 ± 0.08) × 10 −4 at 90 % RH and γ PAA on high mass ADS (ADS h ) increases from (1.37 ± 0.02) × 10 −4 at 3 % RH to (2.62 ± 0.01) × 10 −4 at 90 % RH.On the surface of ATD, however, γ PAA shows a negative RH dependence, from (2.42 ± 0.02) × 10 −4 at 3 % RH to (1.17 ± 0.03) × 10 −4 at 90 % RH on low mass ATD (ATD l ) and decreasing from (1.86 ± 0.01) × 10 −4 at 3 % RH to (0.91 ± 0.04) × 10 −4 at 90 % RH on high mass ATD (ATD h ).Table 3 also lists the lower limit of γ PAA on ADS and ATD, which are calculated by the BET surface area of the particles.The lower limits of γ PAA on ADS and ATD are on the order of 10 −6 -10 −5 .The positive correlations between RH and γ PAA on ADS are similar to that on PM 2.5 .Similar positive RH dependence has also been observed for the uptake of Saharan dust (Pradhan et al., 2010b) and aged CaCO 3 particles (Zhao et al., 2013).This negative RH dependence on ATD is similar to the previously reported γ H 2 O 2 on ATD and mineral oxides (e.g., α-Al 2 O 3 , Fe 2 O 3, TiO 2 , SiO 2 ) (Pradhan et al., 2010a;Zhao et al., 2011a;Romanias et al., 2012Romanias et al., , 2013;;El Zein et al., 2014).The reasons for the discrepancies in the RH dependence of γ P AA are discussed in Sect.3.3.The empirical equation of γ PAA against a H 2 O on ADS and ATD can be expressed as Eqs.( 16) and ( 17), respectively: We also determined the uptake coefficient of H 2 O 2 on ADS and ATD over the RH range of 3 to 90 %.The measured γ H 2 O 2 on ADS and ATD is shown in Fig. 7.The  A es for ADS l and ADS h is 6.1 and 10.9 cm 2 , respectively; A es for ATD l and ATD h is 6.4 and 11.2 cm 2 , respectively.
The empirical equation of γ PAA against a H 2 O on ADS and ATD can be expressed as Eqs.( 18) and ( 19), respectively: It is noted that, although the γ values of H 2 O 2 and PAA on mineral dust particles obtained with the low mass loading are not the same with those with high mass loading, they have the same RH dependence.The differences among ADS l , ADS h , ATD l and ATD h are mainly caused by two reasons: the un-  certainty of the A es estimation method and the experimental error.

Reaction mechanisms
In general, the uptake of a gas onto particles can be attributed to physical processes (e.g., physisorption and dissolution) and/or chemical processes (e.g., catalytic reaction, acid-base reaction, redox reaction and thermal decomposition).In Sect.3.1, we have provided evidence that the chemical processes dominate the uptake of peroxide compounds on PM 2.5 .Here, we discuss the potential chemical pathways.The composition of PM 2.5 determines the relative importance of physical and chemical processes.In general, PM 2.5 is mainly composed of mineral dust, sulfate, nitrate, ammonium compounds, soot and organic matter (Eldred et al., 1997;He et al., 2001;Hueglin et al., 2005;Sun et al., 2006;Huang et al., 2014).In this study, we have measured the con- centrations of elements and soluble ions in PM 2.5 samples.
The results are shown in Table 5.The concentration of mineral dust was estimated by multiplying 14.3 by the concentration of the Al element; the ratio was suggested by Zhang et al. (2013) for PM 2.5 in urban Beijing.The estimated mineral dust accounts for 6.0 ± 4.3 and 6.2 ± 3.1 % of PM 2.5 mass concentration on haze days and non-haze days, respectively.The concentration of SO 2− 4 is 42.26 ± 7.88 µg m −3 on haze days, which is about 7 times that on non-haze days.The concentrations of NO − 3 and Cl − on haze days are also about 6.9-7.3 times those on non-haze days.
There have been several studies of the mechanism of H 2 O 2 uptake on mineral dust particles.Zhao et al. (2011a) have found that the uptake of H 2 O 2 on both SiO 2 and α-Al 2 O 3 particles decreased with increasing RH.On SiO 2 particles, the contribution of physisorption to H 2 O 2 uptake increased from 59 % at 12 % RH to 80 % at 76 % RH; on α-Al 2 O 3 particles, the catalytic decomposition dominated H 2 O 2 uptake even at high RH, probably due to its high surface reactivity.Although the γ H 2 O 2 on both SiO 2 and α-Al 2 O 3 particles decreased with increasing RH, the reduction was more pronounced on the physical process dominated SiO 2 particles.El Zein et al. (2014) observed a negative correlation between RH and γ H 2 O 2 on ATD particles and suggested that the uptake of H 2 O 2 on ATD particles was a catalytic process and that it was not limited by site-filling.Thus, the catalytic reaction of mineral dust might be important to the uptake of peroxide compounds on PM 2.5 .But this reaction alone cannot explain the positive RH dependence for the γ on PM 2.5 .Therefore, some other pathways may also be important to the uptake of peroxide compounds onto PM 2.5 .Based on the characteristics of peroxide compounds, in addition to catalytic reaction, acid-base reaction, redox reaction, thermal decomposition, and aqueous reaction are considered as the potential pathways.
With respect to acid-base reactions, we must consider that H 2 O 2 and PAA are both weak acids (pK a = 11.6 for H 2 O 2 , Marinoni et al., 2011; pK a = 8.2 for PAA, Evans and Upton, 1985) and can react with alkaline substances.A number of studies have demonstrated that the heterogeneous reaction of an acidic vapor on alkaline materials is enhanced with increasing RH (Santschi and Rossi, 2006;Preszler et al., 2007;Sullivan et al., 2009).However, PM 2.5 in Beijing is acidic (e.g., pH = 5.57, Wang et al., 2005).The concentrations of ions of strong acids such as SO 2− 4 and NO − 3 make up 60.9 % of PM 2.5 mass on haze days, and 41.3 % on non-haze days (see Table 5).Even though there are some basic components (such as NH + 4 and CaCO 3 ), we believe they are already neutralized or acidified.Therefore, acid-base reactions on PM 2.5 may not be important for the uptake of H 2 O 2 and PAA.
Both PAA and H 2 O 2 have strong oxidative capacity and can react with the reducing substances on aerosol particles, especially in the presence of water.Zhao et al. (2013) found that γ H 2 O 2 on sulfite-coated calcium carbonate particles is 3-10 times higher than that on the pristine calcium carbonate particles.This enhancement increased with increasing RH.1.12 ± 0.41 0.14 ± 0.10 Cd b 1.09 ± 0.32 0.22 ± 0.17 In addition, transition metals make up 0.9 % of PM 2.5 mass on haze days and 1.2 % on non-haze days.Both PAA and H 2 O 2 can undergo catalytic reactions with transition metals, leading to the formation of highly reactive species, such as OH, RO and RO 2 radicals (Koubek and Edwards, 1963;Lin and Gurol, 1998;Zhang et al., 1998;Hiroki and LaVerne, 2005).Nawrot et al. (2009) have studied PM 2.5 samples in 20 European locations and suggested that H 2 O 2 would decompose and form OH radicals in the presence of transition metals (e.g., Cu, Fe, Mn, V and Ti).Petigara et al. (2002) have reported that the decomposition rate of H 2 O 2 is enhanced by the presence of organic matter and manganese.Therefore, the redox reactions may be important to the uptake of peroxide compounds on PM 2.5 .
It is noted that PAA, which has a hydroperoxyl group (-OOH) and a carbonyl group (C = O), is less stable than H 2 O 2 (Kunigk et al., 2012) and can more readily undergo thermal decomposition.The O-O bond dissociation enthalpies at 298 K of PAA and H 2 O 2 are 48 and 50 kcal mol −1 , respectively (Bach et al., 1996).In addition, PAA is prone to hydrolysis in the presence of water (reactions 3 and 4) (Yuan et al., 1997).This is consistent with our experimental result that R γ PAA is larger than R γ H 2 O 2 .
In considering the role of aqueous reactions, water-soluble inorganic salts including sulfate and nitrate make up a substantial fraction (35-58 %) of PM 2.5 (Sun et al., 2004;Wang et al., 2005).As shown in Table 5, the concentration of Cl − , NO − 3 and SO 2− 4 accounts for 61.9 and 42.0 % of PM 2.5 mass on haze days and non-haze days, respectively.These salts can greatly increase the water content of the particles under humid conditions.When RH exceeds the deliquescence relative humidity (DRH) of these inorganic salts, PM 2.5 may be covered with an aqueous film on the particle surface or exist in a liquid phase state.The DRH is suggested to be 79 % for (NH 4 ) 2 SO 4 , 39 % for NH 4 HSO 4 and 62 % for NH 4 NO 3 at 298 K (Cziczo et al., 1997;Lightstone et al., 2000), and the DRH of PM 2.5 is even lower than that of the individual salt particles (Seinfeld and Pandis, 2006).Under humid conditions, the deliquesced particles and/or the aqueous film on the particle surface becomes a medium for aqueous reaction.In this aqueous phase, soluble salts will release anions.The anions can potentially enhance the dissolution of Fe minerals (Rubasinghege et al., 2010), resulting in a larger uptake of peroxide compounds by Fe catalysis (Chevallier et al., 2004;Pignatello et al., 2006).Furthermore, Zhao et al. (2013) have provided experimental evidence for the effect of a soluble salt on γ H 2 O 2 .They found that nitrate coated on calcium carbonate particles decreased the γ H 2 O 2 by 30-85 % at 3 % RH, but increases γ H 2 O 2 by a factor of 1-8 with increasing RH from 20 to 75 %, as compared to the γ H 2 O 2 on the uncoated particles.Mineral dust can undergo atmospheric aging from its emission, which modifies its surface with coating sulfates and nitrates (Sullivan et al., 2007).The aged authentic mineral dust particles (e.g., ADS dust, Gobi dust and Saharan dust) are coated with salts, while the mineral oxide (e.g., SiO 2 , TiO 2 and α-Al 2 O 3 ) and ATD particles have no or few soluble salt coatings.For example, in this study, the measured concentration of SO 2− 4 in ADS and ATD particles was 20.3 and 0.2 µg mg −1 , respectively.The coatings on the particles can lead to the formation of a surface aqueous film, in which the aqueous reactions may occur.This observation helps explain the differences in RH dependence of the uptake of peroxides on aged authentic particles and unaged mineral oxide and ATD particles.In short, the aqueous reactions that occur in the aqueous film or liquid particles formed by the deliquescence of soluble salts may play important roles in the uptake of peroxide compounds on PM 2.5 and aged mineral dust particles.
In summary, chemical processes rather than physical processes dominate the heterogeneous reaction of peroxide compounds on PM 2.5 and aged mineral dust particles.The soluble inorganic components in authentic particles play an important role in the uptake of peroxide compounds.The uptake of peroxide compounds on PM 2.5 is probably affected by the combined effects of catalytic reactions, redox reactions, thermal decomposition, and aqueous reactions.

Conclusions and implications
The present study is the first to measure the uptake coefficient of gaseous PAA and H 2 O 2 on ambient PM 2.5 and on mineral dust over a wide range of RH values (3-90 %).Both γ PAA and γ H 2 O 2 on PM 2.5 have a positive correlation with RH.In general, both γ PAA and γ H 2 O 2 are on the order of 10 −4 .The γ PAA values show no obvious differences between haze days and non-haze days.Both γ PAA and γ H 2 O 2 on Asian Dust storm (ADS) particles show a similar RH dependence compared to PM 2.5 , but on Arizona Test Dust (ATD), both γ PAA and γ H 2 O 2 show a negative RH dependence.This observation provides evidence that, in addition to the mineral dust, other components in PM 2.5 , such as soluble inorganic salts and organic compounds, may greatly contribute to the uptake of peroxide compounds.The ratio of γ PAA,90 % RH to γ PAA,3 % RH (R γ PAA ) is larger than the ratio of γ H 2 O 2 ,90 % RH to γ H 2 O 2 ,3 % RH (R γ H 2 O 2 ), while the Henry's law constant of H 2 O 2 is 100 times that of PAA; besides, authentic particles show a sustained surface reactivity for the uptake of peroxide compounds.These two experimental results suggest that chemical processes dominate the uptake of peroxide compounds onto PM 2.5 and aged mineral dust.The potential chemical processes include catalytic reactions, redox reactions, thermal decomposition and aqueous reactions.The heterogeneous processes of H 2 O 2 have already been taken into account as an important removal pathway (de Reus et al., 2005;Liang et al., 2013).To the best of our knowledge, there has been almost no consideration of the heterogeneous removal pathways for organic peroxides.
Field observations have shown that the atmospheric lifetime of PAA is 4.1-5.8h in summer in Beijing (Zhang et al., 2010;Liang et al., 2013).To explain this result, we at first considered the traditional removal mechanism for PAA, including the gas phase chemical reactions (OH radical reaction and photolysis) and deposition (Jackson and Hewitt, 1999).The concentration of OH radicals has a positive correlation with solar ultraviolet irradiation and changes in different seasons.The mean concentration of OH radicals on a non-haze summer day was estimated as 3.4 × 10 6 molecule cm −3 in the 35-45  and Khalil, 2004), where Beijing is located.In addition, the concentration of OH radicals on a haze day is onefourth of that on a non-haze day (Liang et al., 2013).
The reaction rate constant of OH radicals with PAA is 3.7 × 10 −12 cm 3 molecule −1 s −1 (Jenkin et al., 1997;Saunders et al., 2003).Hence, the lifetime of PAA with respect to the OH radical reaction is 88.3 h on a haze day and 22.1 h on a non-haze day.Using the reported cross sections of PAA by Orlando and Tyndall (2003), the lifetime of PAA against photolysis is about 28 days on haze days and 21 days on non-haze days.In these studies, we assume that the planetary boundary layer is 1000 m and the dry deposition of PAA is 0.27 cm s −1 (Wesely, 1989;Hall et al., 1999), both on haze and non-haze days.The lifetime of PAA against dry deposition is 4.3 days.The estimated overall lifetime of PAA is 44.2 h on a haze day and 17.6 h on a non-haze day.Obviously, this lifetime is much longer than the field observation results, especially on haze days, indicating that the heterogeneous reaction of PAA on ambient particles would be a removal pathway for gaseous PAA.
In order to estimate the PAA lifetime with respect to the heterogeneous reactions, we assume that all PM 2.5 particles are spheres and the heterogeneous reaction of PAA on PM 2.5 is a pseudo-first-order reaction.The lifetime of PAA can be calculated by Eq. ( 20) (Ravishankara, 1997): where A v is the surface area per unit volume of PM 2.5 , m 2 m −3 .Assuming each mode of aerosol fine particles is a log-normal distribution, the particles number can be expressed as Eq. ( 21) (Seinfeld and Pandis, 2006): where i = 1, 2, 3 correspond to the nucleation mode (3-20 nm), Aiken mode (20-100 nm), and accumulation mode (100-1000 nm), respectively; N i is the number concentration; D p i is the geometric mean diameter, m; σ i is the geometric standard deviation of the ith mode.The recommended values of N t,i , D p i and σ i are suggested by Yue et al. (2009).
The value of A v can be calculated by Eq. ( 22): where M a is the mass of the PM 2.5 particles, kg; ρ is the density of the PM 2.5 particles, 1.42 × 10 3 kg m −3 for a haze period and 1.96 × 10 3 kg m −3 for a non-haze period (Hu et al., 2012); D p is the mean diameter of the total particles, m; and V is the volume of sampling air, m 3 .The number percentage of coarse mode particles (1000-2500 nm) is less than 0.02 % of the fine particle number (3-1000 nm) (Wu et al., 2008) and the corresponding surface area of the coarse mode is about 0.4 % of the total surface area.Therefore, the surface area of the coarse mode particles (1000-2500 nm) could be negligible and D p is 114.6 nm for haze-day PM 2.5 particles and 62.4 nm for non-haze PM 2.5 particles.The mean mass concentration is 123 µg m −3 on a haze day and 23 µg m −3 on a non-haze day and the corresponding A v is 4.5 × 10 3 µm 2 cm −3 on a haze day and 1.2 × 10 3 µm 2 cm −3 on a non-haze day, which is similar to the literature results (Wehner et al., 2008;He et al., 2010).Here, we use the mean uptake coefficient of PAA on PM 2.5 at 60 % RH, i.e., γ = 2.70 × 10 −4 , to estimate the lifetime of PAA.The calculated lifetime of PAA against heterogeneous reaction is 3.2 h on a haze day and 11.9 h on a non-haze day, which are more important than photolysis and decomposition and can compete with OH reaction on haze days.Considering heterogeneous reaction, gas phase reaction and deposition, the estimated lifetime of PAA is 3.0 h on a haze day and 7.1 h on a non-haze day, which is similar to the field measurement results.Thus, the heterogeneous reaction on PM 2.5 is likely to be an important removal pathway for PAA.
The fate of peroxide compounds on aerosols will greatly impact the budget of peroxide compounds themselves as well as the cycle of radicals in the atmosphere.The formation of PAA and H 2 O 2 is related to the self-reaction of HO 2 radical and the reaction of HO 2 radicals with RO 2 radicals, while the photolysis of PAA and H 2 O 2 releases HO x radicals and RO x radicals.Therefore, peroxide compounds can be treated as a temporary reservoir of HO x radicals and RO x radicals.Besides, PAA has a close relation with peroxyacetyl nitrate (PAN).In high NO x (NO + NO 2 ) areas, such as urban areas, NO 2 will combine with acetyl peroxy (CH 3 C(O)OO) radicals to form PAN by competing with HO 2 radicals that will donate H to the CH 3 C(O)OO radical to form PAA. The uptake of PAA onto the particle surface will result in a sink for the CH 3 C(O)OO radical, hence reducing PAN, which is an important carrier of NO x and regionally transports NO x from urban areas to rural and remote areas, affecting oxidant (e.g., O 3 and OH radical) distribution there (Fischer et al., 2014).Moreover, through the heterogeneous uptake, the peroxide compounds are introduced onto the surface of particles, which might enhance the atmospheric aerosol oxidative capacity and then change the composition of the aerosols.For example, Zhao et al. (2014) have suggested that the coexistence of H 2 O 2 could enhance heterogeneous oxidation of OVOCs and the yield of organic acids, such as formic acid and acetic acid.Moreover, peroxide compounds have the potential to enhance the heterogeneous reaction of SO 2 and promote sulfate formation.Hence, the heterogeneous reaction of peroxide compounds on aerosols may help to explain the high concentration of sulfates during haze episodes when other oxidants (e.g., OH radicals) are limited.Therefore, we suggest that the current atmospheric models should take into account the heterogeneous reactions of peroxide compounds on aerosols.

Figure 1 .
Figure 1.Schematic diagram of the experimental apparatus.MFC, mass flow controller; PF, particle-loaded filter; BF, blank filter; PO g , gaseous peroxide compound; PO aq , aqueous peroxide compound; BP, bubbler for peroxide vapor; BW, bubbler for water vapor; HPLC, high-performance liquid chromatography.The scrubbing coil, BP and BW were kept in 277 and 298 K water baths, respectively.

Figure 2 .
Figure 2. The trend line of fractional loss (L f ) of PAA against PM 2.5 mass (M a ) (60 % RH).Error bars are 1 standard deviation.Solid line (a), the logarithmic trend line of L f against M a among all mass values; dotted line (b), the linear correlation of L f against M a in the low mass region; dotted line (c), the nearly constant L f against M a in the high mass region.

Figure 3 .
Figure3.Profiles of uptake coefficients of gaseous PAA on PM 2.5 over a range of RH (3-90 %); N_ase, γ PAA was measured with ascending RH on nighttime PM 2.5 particles; N_des, γ PAA was measured with descending RH on nighttime PM 2.5 particles; D_ase, γ PAA was measured with ascending RH on daytime PM 2.5 particles; D_des, γ PAA was measured with descending RH on daytime PM 2.5 particles.

Figure 4 .
Figure 4.The uptake coefficients of PAA and H 2 O 2 on PM 2.5 particles.The red line and the blue line in (a) and (b) represent the empirical fit of γ PAA and γ H 2 O 2 , respectively.
Figure 5. Profiles of uptake coefficient of gaseous H 2 O 2 on PM 2.5 over a range of RH (3-90 %); N_ase, γ H 2 O 2 was measured with ascending RH on nighttime PM 2.5 particles; N_des, γ H 2 O 2 was measured with descending RH on nighttime PM 2.5 particles; D_ase, γ H 2 O 2 was measured with ascending RH on daytime PM 2.5 particles; D_des, γ H 2 O 2 was measured with descending RH on daytime PM 2.5 particles.

Figure 6 .
Figure 6.The ratio of γ at 90 % RH to γ at 3 % RH (R γ PAA and R γ H 2 O 2 ) on PM 2.5 .Dh, daytime of haze day; Nh, nighttime of haze day; Dn, daytime of non-haze day; Nn, nighttime of non-haze day.

Figure 7 .
Figure 7. Uptake coefficient of H 2 O 2 on ADS and ATD particles.The red line and the blue line in (a) and (b) represent the empirical fit of γ H 2 O 2 on ADS and ATD particles, respectively.
the unit is µg m −3 ; b the unit is ng m −3 .

Table 2 .
Summary of the collected mass and effective surface area of PM 2.5 on the filter, and its ambient average mass concentrations on haze and non-haze days.
Note: a daytime; b nighttime; Aes, effective surface area; Ma , mass of PM 2.5 .

Table 3 .
The uptake coefficients γ (× 10 −4 ) of PAA on PM 2.5 , ADS, and ATD under different relative humidity conditions.The values in the parentheses are the lower limit of γ (× 10 −5 ).Note: PM 2.5h , haze day PM 2.5 ; PM 2.5n , non-haze day PM 2.5 Table 4 summarizes the literature result of γ H 2 O 2 and its RH dependence on different types of mineral dust.Apart from γ H 2 O 2 on authentic Gobi dust, authentic Saharan dust and aged particles, all γ H 2 O 2 values show a negative RH dependence.

Table 4 .
Summary of the uptake coefficients of H 2 O 2 on mineral dust particles in literature data.
× 10 −4 at 90 % RH.Although the values of γ H 2 O 2 at low and high mass loading are not identical, all γ H 2 O 2 values on ADS show a positive correlation with RH, and all γ H 2 O 2 values on ATD show a negative correlation with RH.

Table 5 .
The average concentration of ions, organic acids and elements of PM 2.5 on haze and non-haze days.The errors represent the relative standard deviation.