Environ. Sci. Technol. 2004, 38, 254-259
Sampling Artifacts of Acidity and Ionic Species in PM2.5 RAVI KANT PATHAK,† X I A O H O N G Y A O , ‡ A N D C H A K K . C H A N * ,† Department of Chemical Engineering, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong, and Institute for Environment and Sustainable Development, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong
Although sampling artifacts of acidity, ammonium, nitrate, and chloride in airborne particulate pollutants can be reduced by the use of denuders to absorb interfering gases, artifacts due to interparticle interactions still remain. In this study, the contribution of individual artifact reactions to particle evaporation and the effects of aerosol composition on the extents of sampling artifacts in PM2.5 were investigated. Samples were collected using a Harvard honeycomb denuder/filter-pack system at an urban site and a rural site in Hong Kong. The results show that the formation of artifacts can be categorized into two regimes: ammonium rich (AR) samples with a molar ratio [NH4+]/ [SO42-] greater than 1.5 and ammonium poor (AP) samples with a molar ratio [NH4+]/[SO42-] less than or equal to 1.5. The urban samples were all AR samples, and they were characterized by high nitrate and low in situ free acid concentrations. In contrast, the rural samples were all AP samples and they were characterized by low nitrate and high in situ free acid concentrations. We have developed a methodology to estimate the contribution of each artifact reaction to the sampling loss of nitrate, chloride, ammonium, and acidity. In the AR samples, the evaporation of HNO3 and HCl and concomitant evaporation of NH3 were the principal reactions in determining the extent of the sampling loss of nitrate and chloride. In the AP samples, the evaporation of HNO3 and HCl alone was the principal reaction instead, especially at high sampling loss. The in situ free acid concentration, a function of aerosol composition and ambient conditions, is a more useful parameter than strong acidity in understanding the sampling loss of acidity, nitrate, and chloride from the collected particles.
Introduction Airborne acidic gaseous and particulate pollutants have been associated with a variety of adverse health effects in humans, including impairment of pulmonary functions, an increase in the rate of asthma occurrence and other respiratory ailments, and possibly even excess mortality (1-3). Most PM2.5 sampling periods are 24 h, during which sampling artifacts of acidity, ammonium, nitrate, chloride, and sulfate can occur and thus change the composition of the collected aerosols (4-8). * Corresponding author phone: (852) 2358-7124; fax: (852) 23580054; e-mail:
[email protected]. † Department of Chemical Engineering. ‡ Institute for Environment and Sustainable Development. 254
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 1, 2004
In general, gas-particle interactions, interparticle interactions, and dissociation of semivolatile species can occur during sampling (4-6, 9, 10). The use of denuders to remove all the interfering gases before particle collection can minimize such gas-particle interactions. Nevertheless, interparticle interactions between collected particles and evaporation of semivolatile species can occur (6). Interparticle interactions, that is, the mixing of the collected particles with different chemical natures during sampling, especially at high relative humidity (RH) when the particles are deliquesced aqueous droplets, can lead to significant sampling loss of semivolatile species due to the following reactions (6):
NH4+ + Cl- S NH3(g) + HCl(g)
(rxn1)
NH4+ + NO3- S NH3(g) + HNO3(g)
(rxn2)
H+ + Cl- S HCl(g)
(rxn3)
H+ + NO3- S HNO3(g)
(rxn4)
A notable example of an interparticle interaction is the mixing of ammonium sulfate and sodium chloride particles, which can account for some of the chloride depletion in sea-salt aerosols in field observations (4). Prior to sampling, (NH4)2SO4 and NaCl particles are externally mixed. Upon collection at high ambient relative humidity, rxn 1 (NH4+ + Cl- S NH3(g) + HCl(g)) takes place, leading to the evaporation of HCl and hence to an increase in the observed chloride depletion percentage in sea-salt aerosols. Under conditions when the particles are dry, interparticle interactions are not expected to be significant. Nevertheless, negative sampling artifacts can result from the dissociation reactions of NH4NO3 S NH3(g) + HNO3(g) and NH4Cl S NH3(g) + HCl(g), if NH4NO3 and NH4Cl are present. The dissociation is favored because of the removal of the associated gases by the denuders to shift the gas-particle equilibrium to the gas phase. The term strong acidity is often used to describe the acidity of atmospheric aerosols. Strong acidity is the total H+ from the strong acids in an aqueous extract of atmospheric samples. H+ must exist in the form of free acid to participate in the sampling loss reactions of HNO3 and HCl. In ambient aerosols, some H+ is associated with bisulfate. The dissociation of bisulfate to release the free acid depends on the aerosol composition and the ambient conditions. Hence, the in situ acidity (or the free acid concentration) of atmospheric aerosols is a more relevant parameter than strong acidity in understanding sampling loss. Despite the importance of sampling artifacts, there has been little research carried out on the role of aerosol composition in artifact formation. It has been reported that the percentage of acidity loss decreases as the aerosol acidity increases (6). The goal of this study is to (1) evaluate the contributions of the individual reactions (rxns 1-4) to the sampling losses in measuring semivolatile species in PM2.5 and (2) to investigate the role of in situ acidity (free acid) and aerosol composition in the sampling losses using a denuder/filterpack system.
Method Daily PM2.5 samples were collected from sites at Tsim Sha Tsui (TST) East and at the Hong Kong University of Science 10.1021/es0342244 CCC: $27.50
2004 American Chemical Society Published on Web 11/22/2003
TABLE 1. Acidity and the Concentration of Ionic Species in PM2.5 free acid concentration in PM2.5 using AIM2, nmol/m3 sample day
concentrations on Teflon AR filter, nmol/m3 or ionic APb H+ c NH4+ SO42- NO3- Cl- Na+ ratiod
concentrations on backup filters, nmol/m3 H+
NH4+
NO3-
Cl-
[A]/[S]a
[H+]free (% of H+]SA,amb)
[HSO4-] (% of [H+]SA,amb)
24 h average meteorological data DRH* e %
RH %
temp °C
5 Dec 00 AR 6 Dec 00 7 Dec 00 8 Dec 00 9 Dec 00 10 Dec 00 11 Dec 00 14 Dec 00 15 Dec 00 16 Dec 00 average
39 40 65 38 40 31 22 22 23 38 36
228 235 256 141 141 196 216 94 106 201 182
141 142 152 93 83 121 116 57 72 116 109
11 20 27 6 6 7 13 6 7 39 14
5 9 10 3 5 8 9 4 4 7 7
11 20 15 10 10 12 25 4 6 8 12
0.92 0.86 0.99 0.97 1.07 0.93 1.03 0.99 0.83 0.88 0.95
13 NAf 9 15 15 11 5 14 NA 15 11
21 25 32 24 8 12 14 22 21 14 20
22 20 28 23 22 16 8 27 NA 21 21
12 NA 13 16 11 7 9 9 7 8 10
1.73 1.83 1.91 1.76 1.92 1.70 1.98 2.00 1.69 1.81 1.84
13 (25) NA 20 (27) 15 (28) 15 (27) 11 (26) 5 (19) 14 (39) NA 15 (28) 14 (28)
39 (75) NA 54 (73) 38 (72) 40 (73) 31 (74) 22 (81) 22 (61) NA 38 (72) 36 (72)
72 72 70 67 65 74 77 64 71 70
74 80 80 77 76 74 84 74 71 77 77
20 20 20.5 19 20.5 20 18 17 17 19.5 20
12 Mar 02 AP 13 Mar 02 14 Mar 02 15 Mar 02 16 Mar 02 17 Mar 02 average
160 184 92 47 82 106 112
181 163 134 115 171 129 149
166 165 104 90 139 95 122
6 2 3 2 9 4 4
5 3 5 4 4 4 5
6 4 8 3 5 1 5
1.01 1.04 1.07 0.90 0.89 1.18 1.02
17 23 27 5 4 2 13
8 9 9 11 8 16 10
13 15 22 7 5 12 12
12 17 14 8 7 6 11
1.14 1.05 1.38 1.40 1.29 1.50 1.30
80 (45) 103 (50) 52 (44) 20 (38) 26 (30) 36 (33) 53(42)
97 (55) 104 (50) 66 (56) 32 (62) 60 (70) 73 (67) 72 (58)
45 43 50 65 62 53
87 88 84 90 85 72 84
21 21 23 23 20 20 21
a Ambient ammonium to sulfate molar ratio in PM b AP and AR stand for the ammonium rich and ammonium poor samples, 2.5 aerosols. respectively. All the AP and AR samples were collected at TST and HKUST, respectively. c H+ stands for the measured strong acidity. d Ionic ratio ) ([H+]SA + [NH4+] + [Na+])/(2[SO42-] + [NO3-] + [Cl-]), using the ambient concentrations of the species in PM2.5. e Predicted DRH* of ambient aerosols at 25 °C by AIM2 (16). f Not available.
and Technology (HKUST) in Kowloon, Hong Kong (HK), during December 5-16, 2000, and March 12-17, 2002, respectively. TST is an urban site with heavy traffic, and HKUST is a rural site. More information about the sites is available in refs 11 and 12. PM2.5 samples were collected using a channel of a speciation sampler equipped with a cartridge containing a Harvard honeycomb denuder and a filter-pack system at the flow rate of 10 L/m with a cut point of 2.5 µm (13). The meteorological data such as temperature and relative humidity were obtained from the CCAR-IESD meteorological station at HKUST (14). The Harvard honeycomb denuder and the filter-pack system use two denuders in series to remove acidic and basic gases prior to the collection of particles on the filter. The first denuder was coated with 1% (w/v) Na2CO3/1% (w/v) glycerol in a 50% H2O/50% methanol mixture (v/v) to absorb SO2, HNO3, and HCl gases. The second denuder was coated with a citric acid solution (4% (w/v) citric acid and 2% (w/v) glycerol in methanol) to absorb NH3. The filter pack is comprised of a 47 mm diameter Teflon filter with 2 µm pores (Teflo R2PJ047, Gelman Science, USA) for particle collection, a nylon backup filter for absorbing evaporated HCl and HNO3, and an additional Teflon back filter coated with the citric acid solution for absorbing evaporated NH3. The USEPA recommended procedures for sampling, filter handling, extraction, and the strong acidity measurements were followed (15). The ionic concentrations of the aqueous extract were determined using ion chromatography (IC). The detailed procedures for IC analyses are described elsewhere (12). The overall detection limits for strong acidity, NH4+, Na+, SO42-, NO3-, and Cl- were 1-2 nmol/m3. Two injections from every sample were taken to ensure the consistency of the IC analyses. Data with deviations larger than 10% in the concentrations between the two injections were rejected. A similar rejection scheme was applied to the strong acidity measurements.
Results and Discussion Acidity Loss, Strong Acidity, and in Situ Acidity. The amount of acidity loss ([H+]loss) due to the evaporation of HCl and
HNO3 from the particles collected on the Teflon filters can be estimated from the amounts of ammonium, nitrate, and chloride measured on the backup filters (6):
[H+]loss ) [Cl-]nylon + [NO3-]nylon - [NH4+]CAT
(1)
where the subscripts loss, nylon, and CAT stand for the amount of loss of a species from the Teflon filter, the amount collected on the nylon filter, and the amount collected on the citric acid coated Teflon filter, respectively. The ambient strong acidity of PM2.5 ([H+]SA,amb) was estimated as the sum of the apparent strong acidity measured on the Teflon filter ([H+]SA,Teflon) and the acidity loss ([H+]loss) (6):
[H+]SA,amb ) [H+]SA,Teflon + [H+]loss
(2)
where the subscripts SA, amb, and Teflon stand for strong acidity, ambient concentration, and the concentrations on the Teflon filter, respectively. The ambient concentrations of NH4+, NO3-, and Cl- in PM2.5 can be estimated from the sum of the concentrations on the Teflon and the corresponding backup filters. SO42- is not volatile. Its ambient concentration was therefore taken from that on the Teflon filters only. The losses of ionic species due to evaporation are listed in Table 1. In our discussion below, the extent of the negative sampling artifacts will be expressed in terms of the loss percentages. For instance, the percentage of strong acidity loss is
loss % ) {([H+]SA,amb - [H+]SA,Teflon) × 100%}/[H+]SA,amb (3) As discussed earlier, in situ acidity (or the free acid, [H+]free) is a useful parameter in understanding the sampling loss of nitrate and chloride. Model III of AIM2 (Aerosol Inorganic Model) was used to obtain [H+]free and the relative humidity at which particles are completely deliquesced (DRH*) at 25 °C (16). The DRH* is used to determine the state (dry/wet) and the possibility of the mixing of collected particles. AIM2 was used in this study for its comprehensive aerosol chemistry and thermodynamics and its ability to predict accurately the VOL. 38, NO. 1, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
255
FIGURE 1. The dependence of [H+]free/[HSO4-] on (RHamb - DRH*) in ambient aerosols. Filled symbols indicate the AP samples.
FIGURE 2. The dependence of [H+]free/[HSO4-] on the normalized water content ([H2O]/[SO42-]amb) in ambient aerosols. acidity of acidic aerosols (17). The 24 h average ambient relative humidity and the average concentration of strong acidity, NH4+, Na+, SO42-, NO3-, and Cl- in ambient aerosols were used as input. [H+]free depends on the extent of dissociation of bisulfate ions, which can be enhanced by a higher water content in the aerosols (18). The amount of free acid is related to the presence of liquid water in the aerosols. As shown in Table 1, the 24 h average ambient RH was higher than the estimated DRH* in each sample. Hence, it is reasonable to expect that the collected particles contained liquid water most of the time during the sampling periods. Free acid was thus available, and the mixing of the collected particles, which existed as droplets, was possible. Any solid NH4NO3 and NH4Cl particles would have dissolved to form aqueous solutions. Figure 1 shows that the ratio of free acid to bisulfate is linearly correlated to the difference between the 24 h average ambient RH and the DRH* of each sample. In all the figures of this paper, the filled and unfilled symbols denote the AP (ammonium poor) and AR (ammonium rich) samples, respectively. The rationale of this classification will be described in detail later. The linear relationship shown in Figure 1 reflects the importance of the hygroscopic properties to the acidity of PM2.5. Figure 2 also shows a positive correlation between the [H+]free/[HSO4-] ratio and the water content of the aerosol normalized with respect to the sulfate concentration ([H2O]/[SO42-]amb). Overall, only 28% (19-39%) and 42% (30-50%) of the measured strong acidity ([H+]SA,amb) existed as free acid ([H+]free) in the ambient aerosols in the AR and AP samples, respectively (see Table 1). The major fraction of the measured strong acidity exists in the form of bisulfate (HSO4-) in the ambient aerosols. Concentrations and Characteristics of PM2.5. Table 1 summarizes the strong acidity and the concentrations of the ionic species of PM2.5 measured on the Teflon and backup filters. The cation (including the strong acidity [H+]SA)-toanion ratio of the ambient concentrations of the ionic species 256
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 1, 2004
FIGURE 3. The nitrate-to-sulfate, chloride-to-sulfate, and free-acidto-ammonium concentration ratios in the ammonium poor and ammonium rich samples. in PM2.5 was close to unity. The samples can be clearly classified into two categories on the basis of their ammonium ([A]) to sulfate ([S]) molar ratio in ambient concentrations as ammonium rich (AR for [A]/[S] > 1.5) and ammonium poor (AP for [A]/[S] e 1.5). All the samples collected from the HKUST site (rural) in March 2002 were AP and more acidic, and those from the TST site (urban) in December 2000 were AR and less acidic. Sulfate in general is uniformly distributed in HK (19). Although there were daily variations in the sulfate concentration at TST and HKUST, the average sulfate concentrations at these sites were similar. The difference in the [A]/[S] ratios is mainly attributed to the ambient ammonia concentration at HKUST (25-164 nmol/m3) being lower than that at TST (145-358 nmol/m3). The AP and AR samples had different compositions and characteristics of their sampling losses. Figure 3 shows that the free-acid-to-ammonium ([H+]free/ [NH4+]amb) ratio sharply decreases from 0.6 to 0.15 as [A]/[S] increases from 1 to 1.5 (AP samples). At [A]/[S] e 1.5, both bisulfate and free acid exist. The addition of ammonia neutralizes the acid to form ammonium until [(NH4)3H(SO4)2] is formed at [A]/[S] ) 1.5. At [A]/[S] > 1.5 (AR samples), the [H+]free/[NH4+]amb ratio becomes very small and almost constant at 0.08. At 1.5 < [A]/[S] < 2, additional ammonia is available to stabilize nitrate and, to a lesser extent, chloride in the system. Figure 3 also shows the normalized nitrate concentration (mole of nitrate per mole of sulfate) and the normalized chloride concentration as a function of the [A]/[S] ratio. The normalized nitrate concentration is relatively constant at [A]/[S] below 1.5 but increases dramatically as the ratio increases from 1.5. In fact, the excess ammonium, defined as the amount of ammonium in excess of that required for [A]/[S] ) 1.5 (excess [NH4+] ) ([A]/[S] - 1.5)[SO42-]amb), was almost identical to the nitrate concentration in both the measured concentrations on the Teflon filter and the estimated ambient concentrations in the AR samples, as shown in Figure 4. The exceptions were the two days with the lowest (8 nmol/m3, Dec 11) and highest (27 nmol/m3, Dec 16) HNO3 gas-phase concentrations. This suggests that the nitrate concentration increases with a similar increase in the excess ammonium in the AR samples forming nitrate or nitrate-sulfate salts of ammonium. The presence of ammonium nitrate in the ammonium-sulfate-nitrate system was consistent with theoretical predictions. We found that the product of the ammonium and nitrate concentrations on the Teflon filter (1.74(0.35-4.68) ppb2) was much larger than that required for ammonium nitrate to remain in the particulate phase (0.51(0.3-1.5) ppb2), that is, [NH4+]Teflon × [NO3-]Teflon > Kp, where Kp is the dissociation constant of ammonium nitrate in the ammonium-sulfate-nitrate system (20). Figure 4 represents a very unique characteristic of
TABLE 2. Summary of the Sampling Loss percentage of loss (%)a sample day
[H+]
free
[H+]
SA
([H+]
b SA,nitrate)
+]
[NH4
([NH4+]nitrate)c
[NO3-] ([NO3-]ammonium)d
[Cl-] ([Cl-]ammonium)e
5 Dec 00 6 Dec 00 7 Dec 00 8 Dec 00 9 Dec 00 10 Dec 00 11 Dec 00 14 Dec 00 15 Dec 00 16 Dec 00 average
100 NA 45 100 100 100 100 100 NA 100 93
25 (65) NA 12 (68) 28 (59) 27 (67) 26 (70) 15 (47) 39 (75) NA 28 (72) 25 (65)
Ammonium Rich Samples 9 (65) 10 (NA) 12 (68) 16 (59) 12 (67) 6 (70) 6 (47) 22 (75) 19 (NA) 7 (72) 11 (65)
67 (62) 50 (NA) 51 (78) 79 (62) 79 (55) 70 (52) 38 (74) 82 (61) NA 35 (48) 61 (60)
71 (62) NA 57 (78) 84 (62) 69 (55) 47 (52) 53 (74) 69 (61) 63 (NA) 53 (48) 63 (60)
12 Mar 02 13 Mar 02 14 Mar 02 15 Mar 02 16 Mar 02 17 Mar 02 average
21 22 52 25 15 6 25
10 (52) 11 (46) 23 (62) 9 (47) 4 (44) 2 (66) 10 (53)
Ammonium Poor Samples 4 (52) 5 (46) 7 (62) 9 (47) 4 (44) 11 (66) 7 (53)
70 (32) 89 (29) 90 (26) 82 (71) 35 (67) 77 (88) 74 (50)
69 (32) 83 (29) 75 (26) 69 (71) 65 (67) 58 (88) 70 (50)
a For example: percentage of [H+] + + + b % SA loss ) {[H ]SA,amb - [H ]SA,Teflon} × 100/[H ]SA,amb. The % loss of other species was similarly calculated. loss ([H+]SA,nitrate) denotes the % of strong acidity loss due to NO3- loss (rxn 4). It is the same as the % of free acidity loss due to NO3- loss (rxn 4); % loss ([H+]SA,chloride) (rxn 3) ) 100% - % loss ([H+]SA,nitrate). c % loss ([NH4+]chloride) (rxn 1) ) 100% - % loss ([NH4+]nitrate) (rxn 2). d % loss ([NO3-]H+) (rxn 4) ) 100% - % loss ([NO3-]ammonium) (rxn 2). e % loss ([Cl-]H+) (rxn 3) ) 100% - % loss ([Cl-]ammonium) (rxn 1).
FIGURE 4. The equality of the excess ammonium and the nitrate concentration in the AR samples. Excess ammonium ) ([A]/[S] -1.5)[SO42-]amb. the AR samples and justifies the use of [A]/[S] ) 1.5 as the demarcation criterion of the AP and AR samples. The normalized chloride concentration shows a weaker dependence on the [A]/[S] ratio, primarily because ammonium chloride is more volatile than ammonium nitrate. Chloride in ambient PM2.5 in HK is mostly from sea-salt particles and forms an external mixture with ammonium sulfate and nitrate particles prior to particle collection (12, 21). In view of Figures 3 and 4, it can be concluded that the neutralization of the free acid by ammonia and the formation of nitrate and/or nitrate-sulfate salt of ammonium were the principal processes in the AP and AR samples, respectively. These processes in turn affect the evaporation loss of collected particles, as will be discussed later. Sampling Loss. As determined earlier, the ionic species existed in the form of aqueous solutions in all samples most of the time. The interparticle interactions probably took place under such conditions. In fact, the AP samples were collected with the ambient RH always higher than the average DRH*, suggesting that the samples were most likely deliquesced all the time during each sampling period. In the AR samples, there were fewer than 6 h during each sampling period when the ambient RH was lower than the DRH*. Evaporation of solid NH4NO3 and NH4Cl could take place, if they were
present. However, our AIM2 simulation results reveal that neither NH4NO3 nor NH4Cl was present when the DRH* of the particles was lower than the ambient RH for a few hours in some samples. Hence, we attribute all the sampling loss to rxns 1-4. Table 2 summarizes the sampling losses of free acidity, strong acidity, ammonium, nitrate, and chloride. The percentage of free acid loss ([H+]loss × 100/[H+]free) should not be confused with the percentage of strong acidity loss ([H+]loss × 100/[H+]SA,amb). On average, the AR samples had a much higher (93%) percentage of free acid loss than the AP samples had (25%) while the percentage of strong acidity loss was low for both. The percentage of ammonium loss from the Teflon filter was low, 11% (6-19%) and 7% (4-11%) in the AP and AR samples, respectively. In contrast, the percentage of nitrate loss and the percentage of chloride loss from the Teflon filter in both the AR and AP samples were both significant. In the AP samples, the percentage of free acid loss was low (∼20%), due to the low nitrate and high free acid concentrations (Figure 3). The percentage of nitrate loss was high (∼80%) because the large amount of free acid resulted in the almost complete evaporation of nitrate. In the AR samples, the free acid concentration was lower than the total of the nitrate and chloride concentrations. Thus, almost all the free acid was lost and the percentage of nitrate loss was lower than that in the AP samples. Thermodynamically, more ammonium was available to stabilize the nitrate and hence the percentage of nitrate loss was lower in the AR samples (see Figure 4) than in the AP samples. Contribution of Rxns 1-4 to Sampling Loss. The amount of HCl released due to rxn 1 (NH4+ + Cl- S NH3(g) + HCl(g)) is associated with the same amount of NH3 released. The same is true for HNO3 in rxn 2 (NH4+ + NO3- S NH3(g) + HNO3(g)). The material balance equations for the species involved in rxns 1-4 are
[NH3]rxn1 + [NH3]rxn2 ) [NH4+]CAT
(4)
[HCl]rxn1 + [HCl]rxn3 ) [Cl-]nylon
(5)
-
[HNO3]rxn2 + [HNO3]rxn4 ) [NO3 ]nylon VOL. 38, NO. 1, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
(6) 9
257
FIGURE 5. The dependence of the nitrate loss and the chloride loss on the acidity loss. Since H+ needs to be available to form HCl or HNO3, rxn 1 and rxn 2 can be written as
NH4+(aq) S NH3(aq) + H+(aq)
H+(aq) + Cl-(aq) S HCl(g) (rxn1)
NH4+(aq) S NH3(aq) + H+(aq)
H+(aq) + NO3-(aq) S HNO3(g) (rxn2)
Although the source of H+(aq) is different in rxns 1-4, the same H+(aq) is equally available for all the reactions. Hence, the ratio of evaporated HCl to HNO3 via rxns 1 and 2 and that via rxns 3 and 4 are the same, that is,
[HCl]rxn1/[HNO3]rxn2 ) [HCl]rxn3/[HNO3]rxn4 ) RCN
(7)
where RCN is the chloride-to-nitrate concentration ratio on the nylon filter,
RCN )
[HCl]nylon [HNO3]nylon
)
([HCl]rxn1 + [HCl]rxn3) ([HNO3]rxn2 + [HNO3]rxn4)
(8)
The contributions of individual reactions to the loss of acidity, ammonium, nitrate, and chloride were determined by solving eqs 4-7. They are shown in terms of percentage contributions in brackets next to the total percentage loss of the semivolatile species in Table 2. On average, 65% and 53% of the strong acidity loss was associated with nitrate loss (rxn 4, H+ + NO3- S HNO3(g)) in the AR and AP samples, respectively, with the rest attributable to chloride loss (rxn 3, H+ + Cl- S HCl(g)). Similarly, 65% and 53% of the ammonium loss was associated with nitrate loss (rxn 2) in the AR and AP aerosols, respectively, with the rest attributable to chloride loss (rxn 1). Furthermore, an average of 60% and 50% of the loss of nitrate (and of the loss of chloride) was associated with the loss of ammonium in the AR and AP samples, respectively. Sampling Loss Correlations. As shown in Table 2, nitrate (chloride) can evaporate as HNO3 (HCl) alone or with NH3. Figures 5 and 6 illustrate the contributions of the evaporative reactions with and without NH3 to the sampling loss of nitrate and chloride. Shown in Figure 5 are the nitrate loss and the chloride loss as a function of the acidity loss. The acidity loss ([H+]loss) was positively correlated with the nitrate loss ([NO3-]loss ) 0.5[H+]loss + 4.5) and with the chloride loss ([Cl-]loss ) 0.5[H+]loss + 4.5) in the AP samples. The intercepts can be interpreted as the nitrate and chloride loss due to ammonium loss. The identical slopes and intercepts of the fitting equations of the nitrate loss and the chloride loss suggest that nitrate and chloride contributed equally to the acidity loss (same slope) and ammonium loss (same intercept) in the AP samples. Thus, on average, RCN ) 1 in the AP 258
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 1, 2004
FIGURE 6. The dependence of the nitrate loss and the chloride loss on the ammonium loss. samples, which is close to the results of 53% contribution by nitrate to the acidity loss based on solving eqs 4-7 and shown in Table 2. Since chloride and nitrate loss contributed equally to a total acidity loss of ∼13 nmol/m3 and a total ammonium loss of ∼10 nmol/m3 in the AP samples, it is estimated that, on average, about 55% of the chloride loss and nitrate loss was associated the acidity loss. This is consistent with the solution to eqs 4-7 in which the acidity loss contributed to 50% of the nitrate loss and the chloride loss. Therefore, it can be concluded that, on average, rxn 3 (H+ + Cl- S HCl(g)) and rxn 4 (H+ + NO3- S HNO3(g)) were the principal reactions in the AP samples, especially with high sampling loss, although rxn 1 and rxn 2 played a minor role as indicated by the small positive intercepts. In contrast, the acidity loss was not correlated with either the nitrate loss or the chloride loss in the AR samples ([H+]loss ∼ 13 nmol/m3). Figure 6 shows that the ammonium loss was positively correlated with the nitrate loss ([NO3-]loss ) 0.68[NH4+]loss + 8) and the chloride loss ([Cl-]loss ) 0.32[NH4+]loss + 4) in the AR samples. The intercepts indicate the loss of nitrate and chloride due to rxn 3 and rxn 4 alone. However, the ammonium loss was constant at about 9 nmol/m3, independent of the nitrate and chloride loss in the AP samples. The slope and intercept of the correlation between the nitrate loss and the ammonium loss were double those between the chloride loss and the ammonium loss in the AR samples. This implies that the nitrate loss contributed roughly 66% to the acidity loss (the intercepts) and to the ammonium loss (the slopes) in the AR samples, that is, RCN ) 0.5. This is also consistent with the solution to eqs 4-7 in which 65% of acidity loss and ammonium loss was due to nitrate loss, as shown in Table 2. At zero ammonium loss, there was 8 nmol/m3 of nitrate loss and 4 nmol/m3 of chloride loss, which were associated with a total of 12 nmol/m3 of acidity loss in the AR samples. This is close to the constant acidity loss of 13 nmol/m3 in the AR samples shown in Figure 5. Similarly, there was a 4.5 nmol/m3 loss of nitrate and a 4.5 nmol/m3 loss of chloride at zero acidity loss in the AP samples as shown in Figure 5. These losses at zero acidity loss were due to rxns 1 and 2, consistent with the result of a total of 9 nmol/m3 of ammonium loss in the AP samples shown in Figure 6. These results imply that, on average, rxn 1 (NH4+ + Cl- S NH3(g) + HCl(g)) and rxn 2 (NH4+ + NO3- S NH3(g) + HNO3(g)) were the principal reactions to the nitrate and chloride loss in the AR samples. Based on experimental measurements using a Harvard honeycomb denuder/back filter system, this paper demonstrates that the ambient [A]/[S] ratio of 1.5 is a critical ratio in determining the sampling artifacts of nitrate and chloride in PM2.5. We have developed a methodology to estimate the contribution of each reaction in the sampling artifacts of nitrate, chloride, ammonium, and acidity. Samples with [A]/[S] ratios less than 1.5 are characterized by low nitrate
concentrations and high in situ acidity (free acid). Rxn 3 and rxn 4 were the principal reactions of the sampling loss of nitrate and chloride. Samples with ratios larger than 1.5 are characterized by high nitrate concentrations and low in situ acidity; rxn 1 and rxn 2 were the principal reactions instead. Although we did not collect samples under dry conditions for direct comparisons, our measurements suggest that interparticle interactions can be very severe under high humidity conditions. The influence of humidity on sampling artifacts warrants more experimental verification.
Acknowledgments The financial support from the Hong Kong Research Grants Council via the Earmarked Grant HKUST6071/99P is gratefully acknowledged.
Literature Cited (1) Thurston, G. D.; Ito, K.; Lippmann, M. Environ. Health Perspect. 1988, 79, 73-82. (2) Dockery, D. W.; Pope, C. A., III; Xu, X.; Spengler, J. P.; Ware, J. H.; Fay, M. E.; Ferris, B. G., Jr.; Fpeizer, F. E. N. Engl. J. Med. 1993, 29, 1753-1759. (3) Spengler, J. D.; Brauer, M.; Koutrakis, P. Environ. Sci. Technol. 1990, 24 (7), 946-955. (4) Yao, X. H.; Fang, M.; Chan, C. K. Environ. Sci. Technol. 2001, 35, 600-605. (5) Tsai, C. J.; Perng, S. N. Atmos. Environ. 1998, 32, 1605-1613. (6) Koutrakis, P.; Thompson, K. M.; Wolfson, J. M.; Spengler, J. D.; Keeler, G. J.; Slater, J. L. Atmos. Environ. 1992, 26A, 987-995. (7) Dasch, J. M.; Cadle, S. H.; Kennedy, K. G.; Mulawa, P. A. Atmos. Environ. 1989, 23, 2775-2782. (8) Harrison, R. M.; Sturges, W. T.; Kitto, A. M. N.; Li, Y. Atmos. Environ. 1990, 24A, 1883-1888.
(9) Zhang, X. Q.; McMurry, P. H. Atmos. Environ. 1992, 26A (18), 3305-3312. (10) Cheng, Y. H.; Tsai, C. J. J. Aerosol Sci. 1997, 28 (8), 1553-1567. (11) Zhuang, H.; Chan, C. K.; Fang, M.; Wexler, A. S. Atmos. Environ. 1999, 33, 4223-4233. (12) Yao, X. H.; Fang, M.; Chan, C. K. Atmos. Environ. 2003, 37, 743751. (13) Partisol 2300; R&P Co. Operating manual: partisol 2300 speciation sampler; R&P Part No. 42-006439; R&P Co. Inc.: New York, 2000. (14) CCAR-IESD meteorological station at HKUST, Hong Kong; http://ccar.ust.hk/dataview/aws/current/ust.py. (15) USEPA Determination of strong acidity of atmospheric fineparticles (
