Subscriber access provided by UNIV OF CONNECTICUT
Article
Measuring and Modeling the Salting-out Effect in Ammonium Sulfate Solutions Chen Wang, Ying Duan Lei, Satoshi Endo, and Frank Wania Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/es5035602 • Publication Date (Web): 16 Oct 2014 Downloaded from http://pubs.acs.org on October 29, 2014
Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Environmental Science & Technology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.
Page 1 of 20
Environmental Science & Technology
1
Measuring and Modeling the Salting-out Effect in Ammonium Sulfate
2
Solutions
3
Chen Wang,1 Ying Duan Lei,1 Satoshi Endo,2 Frank Wania1,*
4
1
5
of Toronto Scarborough, 1265 Military Trail, Toronto, Ontario, Canada M1C 1A4
6
2
7
Environmental Research, Permoserstrasse 15, D-04318 Leipzig, Germany
8
* To whom correspondence should be addressed:
[email protected], +1-416-287-7225
9
Abstract. The presence of inorganic salts significantly influences the partitioning behavior of
10
organic compounds between environmentally relevant aqueous phases, such as sea water or
11
aqueous aerosol, and other, non-aqueous phases (gas phase, organic phase, etc.). In this study,
12
salting-out coefficients (or Setschenow constants) (KS [M-1]) for 38 diverse neutral compounds in
13
ammonium sulfate ((NH4)2SO4) solutions were measured using a shared headspace passive
14
dosing method and a negligible depletion solid phase microextraction technique. The measured
15
KS were all positive, varied from 0.216 to 0.729, and had standard errors in the range of 0.006-
16
0.060. Compared to KS for sodium chloride (NaCl) in the literature, KS values for (NH4)2SO4 are
17
always higher for the same compound, suggesting a higher salting-out effect of (NH4)2SO4. A
18
polyparameter linear free energy relationship (pp-LFER) for predicting KS in (NH4)2SO4
19
solutions was generated using the experimental data for calibration. pp-LFER predicted KS
20
agreed well with measured KS reported in the literature. KS for (NH4)2SO4 was also predicted
21
using the quantum-chemical COSMOtherm software and the thermodynamic model AIOMFAC.
22
While COSMOtherm generally overpredicted the experimental KS, predicted and experimental
23
values were correlated. Therefore, a fitting factor needs to be applied when using the current
24
version of COSMOtherm to predict KS. AIOMFAC tends to underpredict the measured
25
KS((NH4)2SO4), but always overpredicts KS(NaCl). The prediction error is generally larger for
26
KS(NaCl) than for KS((NH4)2SO4). AIOMFAC also predicted a dependence of KS on the salt
27
concentrations, which is not observed in the experimental data. In order to demonstrate that the
28
models developed and calibrated in this study can be applied to estimate Setschenow coefficients
29
for atmospherically relevant compounds involved in secondary organic aerosol formation based
Department of Chemistry and Department of Physical and Environmental Sciences, University
Department of Analytical Environmental Chemistry, UFZ - Helmholtz Centre for
ACS Paragon Plus Environment
1
Environmental Science & Technology
Page 2 of 20
30
on chemical structure alone, we predicted and compared KS for selected α-pinene oxidation
31
products.
32
Introduction
33
Atmospheric aerosol particles, cloud and fog droplets often contain both organic species and
34
inorganic salts.1,2 The inorganic salts’ impact on the activity or solubility of neutral organic
35
solutes in aqueous solutions can be described by the following empirical equation:
36
log(γ/γ0) = log(S0/S) = KS [salt]
37
where γ0 and γ are activity coefficients of an organic solute in pure water and salt solutions, S0
38
and S are the solubilities of a solute in pure water and salt solutions, KS (M-1) is the empirical
39
Setschenow or salting-out coefficient and [salt] (mol/L) is the salt solution concentration. The
40
salting-out effect can also be described in an alternative form as follows:3
41
log(K1/salt water/K1/water)= KS [salt]
42
where K1/salt water and K1/water are equilibrium partitioning coefficients of an organic solute between
43
the aqueous phase and another non-aqueous phase (as described in equation 1 in ref.3). When KS
44
is positive, there is a shift of the partitioning equilibrium to the non-aqueous phase due to the
45
salting-out effect, and vice versa. This effect has important implications regarding the reactivity,
46
transport and fate of organic compounds in atmospheric waters, where the concentration of
47
dissolved salts often varies widely with decreasing water content of a hydrometeor, i.e. from
48
cloud droplets to fog droplets to aerosol particles.2,4 The influence can be particularly significant
49
for aerosol particles because of the relatively large amount of salts present. These inorganic salts
50
not only contribute to the total mass of aerosol,1 but also play an important role in the phase
51
partitioning equilibria of organic species involved in secondary organic aerosol formation.5 In
52
addition, the inorganic salts together with the cycle of relative humidity in the atmosphere
53
influence the phase transition of aerosol particles.6 Liquid-liquid phase separation was observed
54
to occur in both real-world and laboratory-generated particle samples containing ammonium
55
sulfate and secondary organic materials, forming an inorganic rich phase and an organic rich
56
phase.7 In the electrolytes-organics mixture, ions have a high affinity for water and a low affinity
(1)
(2)
ACS Paragon Plus Environment
2
Page 3 of 20
Environmental Science & Technology
57
for organic compounds, especially the non-polar compounds, and two phases will form in order
58
to maintain a minimum system free energy.5,6
59
Among the various inorganic salts present in atmospheric aerosol, cloud, and fog in continental
60
regions, ammonium (NH4+), sulfate (SO42-), and nitrate (NO3-) are often the most abundant.
61
Marine aerosols additionally include sodium (Na+) and chloride (Cl-).1 When comparing the
62
salting-out effect of SO42-, NO3- and Cl-, several studies8,9 found NO3- to have the smallest and
63
SO42- the highest effect.
64
Most of the KS values reported so far were for NaCl or seawater, which is relevant for the marine
65
environment and to some extent for atmospheric water. However, Setschenow constants for
66
organic compounds in (NH4)2SO4 solution (subsequently referred to as KS((NH4)2SO4)) are
67
available only for a very limited number of chemicals. When Xie et al.10 reviewed the reported
68
data for Setschenow constants in various salt solutions in 1997, KS((NH4)2SO4) was available for
69
only six aromatic solutes. Görgényi et al.9 measured salting-out effects of 27 different salts,
70
including (NH4)2SO4, but only for four organic solutes. Furthermore, most of the previous
71
measurements were performed under relatively low salt concentrations (0.5 M, i.e., similar to the
72
salinity of sea water), whereas the salting-out effect in highly concentrated salt solutions is more
73
relevant in the atmosphere: concentrations of inorganic salts in aqueous aerosols can be very
74
high (i.e. reach saturation or even supersaturated concentrations). In addition, no data are
75
available to systematically compare the salting-out effect for the two major salts in the
76
environment, NaCl and (NH4)2SO4. As a consequence, more KS((NH4)2SO4) are required for
77
accurate atmospheric phase distribution assessments of organic compounds, including those
78
implicated in secondary organic aerosol (SOA) formation. Particularly, measurements at higher
79
salt concentrations are required to test the dependence of KS on salt concentrations.
80
One reason for this lack of empirical data might be that the determination of Setschenow
81
coefficients requires highly precise measurements of aqueous solubility or partitioning properties
82
(i.e., air-water, organic solvent-water) at different ionic strength. Until recently, few methods
83
allowed for a relatively simple, reliable and inexpensive determination of the salting-out effect
84
for diverse compounds. Jonker et al.11 proposed a solid phase microextraction (SPME) technique,
85
where KS in NaCl solutions (subsequently referred to as KS(NaCl)) is determined by measuring
ACS Paragon Plus Environment
3
Environmental Science & Technology
Page 4 of 20
86
SPME fiber-water partitioning coefficients at different salt concentrations. Endo et al.3 modified
87
this SPME method, developed another shared headspace passive dosing (SHPD) approach to
88
measure KS, and applied both methods to determine KS(NaCl) for a large number of organic
89
solutes with diverse structures. Their large and reliable data set also provided a basis for the
90
development of a predictive model.3
91
Due to the lack of data on the salting-out effect of (NH4)2SO4 and the significant role of
92
(NH4)2SO4 in the atmosphere, the objective of this work was to measure Setschenow coefficients
93
for a large and diverse group of organic compounds in (NH4)2SO4 solutions at a wide range of
94
salt concentrations using recently developed experimental approaches described in the literature.
95
While it may seem more sensible to measure the salting-out effect for atmospherically relevant
96
substances directly, such compounds are not always commercially available and may often pose
97
experimental challenges. In order to be able to compare KS((NH4)2SO4) with KS(NaCl), the
98
organic compounds tested here were those used in previous work for NaCl.3 These compounds
99
were not meant to represent compounds found in aerosols, but were selected to include
100
compounds with diverse functional groups that interact by a variety of intermolecular forces.
101
This was necessary, because we aimed to use the measured data to develop and test predictive
102
models for estimating KS((NH4)2SO4) from chemical structures. Models calibrated with such a
103
diverse dataset can be applied to predict KS for other compounds, including those found in
104
aerosol.
105
Materials and Methods
106
Setschenow coefficients for 38 neutral organic compounds in (NH4)2SO4 solutions were
107
measured in this study (names are listed in the first column of Table 1).
108
Materials. The 38 organic chemicals were purchased from Sigma Aldrich. Solvents ethyl acetate,
109
hexane, methanol and acetone (Omnisolv) were obtained from EMD Millipore Chemicals.
110
Ammonium sulfate and sodium chloride in analytical grade were from VWR and Fisher
111
Scientific, respectively. Olive oil was purchased from a grocery store in Toronto. Water was
112
treated with a Barnstead Nanopure water system. Polydimethylsiloxane fibers (PDMS, 30 µm
113
coating thickness, 13.2 µL/m coating volume) and polyacrylate fibers (PA, 36 µm coating
114
thickness, 16.5 µL/m coating volume) were from Polymicro Technologies (Phoenix, AZ).
ACS Paragon Plus Environment
4
Page 5 of 20
Environmental Science & Technology
115
Experimental Methods. The methods for the determination of KS include the SHPD method and
116
negligible depletion SPME technique as detailed in Endo et al.3 The SHPD method and SPME
117
method were compared in Endo et al.3 using a series of 2-alkanones. The two methods showed
118
good agreement, although the SPME method has slightly lower standard errors. One of the two
119
methods was chosen for a compound based on its air-water partitioning behavior. The SHPD
120
approach is suitable for relatively volatile compounds (e.g., alkylbenzenes) and for compounds
121
that tend to adsorb to interfaces (e.g., highly fluorinated compounds). The fiber types in the
122
SPME method were chosen to avoid significant solute depletion and also to reduce the
123
equilibration time for the compounds between fiber and the solutions.3
124
In brief, 27 compounds were measured with the SPME method, where SPME fibers (PA or
125
PDMS, 2 cm length) were deployed in the same volume (8 mL) of pure water or salt solutions
126
spiked with the same amount of test solutes (Table S1 in Supporting Information). Due to the
127
small volume of the fibers, only negligible amounts of test solutes will be absorbed and the
128
aqueous concentrations will not significantly decrease. At equilibrium (more than 24 h
129
deployment), the relative concentrations in the fibers placed in pure water and in salt solutions at
130
different concentrations are a measure of the degree of salting-out. Four replicates were made for
131
each salt concentration. The fibers were extracted in 500 µL ethyl acetate and analyzed with gas
132
chromatography mass spectrometry (GC/MS). Measurements were made in solutions with up to
133
six different concentrations of (NH4)2SO4, which were 0%, 5%, 10%, 20%, 30% and 40% (w/v),
134
corresponding to molar concentrations of 0, 0.38, 0.76, 1.52, 2.27 and 3.03 M.
135
The SHPD method was applied in measuring KS for 11 relatively volatile compounds in
136
(NH4)2SO4 solutions (Table S1 in Supporting Information). Briefly, in a closed container, test
137
solutes transfer from the spiked non-volatile solvent (100 mL olive oil) through the shared
138
headspace into pure water and salt solutions (25 mL) in small petri dishes as described in Figure
139
1 of ref.3. At equilibrium, 800 µL (×4 replicates) of aqueous solutions were sampled through a
140
septum using a gas tight syringe and then liquid-liquid extracted with 600 µL of n-hexane.
141
Extracts were analyzed using GC/MS. The differences in concentrations in the different aqueous
142
solutions reflect the degree of salting-out. In this method, three or four salt concentrations (0%,
143
5%, 10%, and 20 % (w/v)) were used depending on the solubilities of the compounds.
ACS Paragon Plus Environment
5
Environmental Science & Technology
Page 6 of 20
144
With pH values of the (NH4)2SO4 solutions in this study varying from 5.28 to 5.5 without
145
buffering, all organic solutes were predominantly in the neutral form during the experiment. In
146
both methods, KS can be determined by linear regression analysis based on eqs. 2 and 3 in ref.3
147
More detail on these two methods is provided elsewhere3 and can be found in the Supporting
148
Information of this article (Table S1). To validate the measured data, a few measurements were
149
made with a third method, namely a headspace gas chromatographic (HS-GC) method. Details
150
for this method are given in the Supporting Information. Because previous studies found KS to be
151
independent of the solute concentrations11,12 and only weakly dependent on temperature,13,14 only
152
one organic solute concentration and one temperature (room temperature) were used in the
153
experiments.
154
GC/MS Analysis. The compounds were analyzed on a GC/MS system, consisting of a 6890 GC
155
with a 5973 MS detector and a 7683 autosampler (Agilent Technologies). Helium was used as
156
carrier gas (1.2 mL/min). The columns used were DB-5 (60 m × 0.25 mm i.d. × 0.25 µm, Agilent
157
J&W, USA) and DB-WAX (30 m × 0.25 mm i.d. × 0.25 µm, Agilent J&W, USA). The GC oven
158
temperature program was different for different analytes. The MS was operated in electronic
159
ionization and selective ion monitoring mode with a GC/MS interface temperature of 250 °C.
160
The samples (1 µL) were injected in splitless mode and the injector temperature was held at
161
225 °C. Internal standards were used for quantification of analytes. More details on columns and
162
internal standards can be found in Supporting Information (Table S1).
163
Modeling. Three different modeling approaches were used to predict an organic chemical’s
164
KS((NH4)2SO4). While the quantum-chemical COSMOtherm software and the group contribution
165
method AIMOFAC15 only require information on the solute’s molecular structure, the poly-
166
parameter linear free energy relationship (pp-LFER) approach depends on the availability of
167
empirical data for calibration.
168
Multiple linear regression analysis and the experimentally determined KS values for 38 solutes
169
were used to calibrate a pp-LFER for KS((NH4)2SO4). Specifically, experimental KS values can
170
be correlated with compound specific solute descriptors using:3
171
KS = c + aA + bB + sS + eE + vV
(2)
ACS Paragon Plus Environment
6
Page 7 of 20
Environmental Science & Technology
172
where A, B, S, E and V are solute descriptors representing a solute's hydrogen-bond acidity,
173
hydrogen-bond basicity, dipolarity/polarizability, excess molar refraction and McGowan volume
174
(cm3/mol) divided by 100, respectively. The regression coefficients are denoted by a, b, s, e and
175
v; c is the regression constant. Such a pp-LFER describes, and can be used to predict, the salting-
176
out effect based on molecular interaction properties.3 Previously, pp-LFERs have been shown to
177
reliably predict KS(NaCl).3 Solute descriptors were taken from the UFZ-LSER database16 or
178
estimated with Absolv from ACD/Labs (Advanced Chemistry Development, Inc., Toronto,
179
Canada) and are given in the Supporting Information.
180
KS((NH4)2SO4)
181
BP_TZVPD_FINE_C30_1401 parameterization, COSMOlogic GmbH & Co. KG, Leverkusen,
182
Germany 2014), which relies on a quantum mechanical calculation and a statistical
183
thermodynamics approach to predict physical-chemical properties.17,18 In this approach, structure
184
files (called COSMO files) of organic solutes and salt ions, which had been optimized with
185
TURBOMOLE, are fed into a program that calculates partition coefficients between pure water
186
and (NH4)2SO4 solutions (Kwater/salt water) at a temperature of 298 K. The term K1/salt water/K1/water in
187
the Setschenow relationship (eq. 2) is equivalent to Kwater/salt water, so KS [M-1] can be calculated
188
from the salt concentration and log Kwater/salt water.
189
AIOMFAC is a thermodynamic model that predicts activity coefficients of different chemical
190
species in inorganic-organic mixtures based on a group-contribution approach.15 In the present
191
study, AIMOFAC was used to calculate activity coefficients of compounds in pure water and salt
192
solutions at 298 K. Setschenow constants can then be calculated from eq. 1. AIOMFAC requires
193
the input of temperature, and the mole fractions of organic solute and inorganic salt in the
194
mixture. Since a very small amount of the organic solute was used (0.05), it
273
was excluded from the final regression model, indicating that the excess molar refraction does
274
not influence the salting-out process significantly. The standard errors for all the regression
275
coefficients are small, from 0.025 to 0.045. As in the pp-LFER equation for KS(NaCl),3 the
276
regression coefficients for A, B and S are negative, while that for V is positive. This indicates that
277
the salting-out effect of (NH4)2SO4 is negatively related to the ability of the target compounds to
278
engage in polar interactions (hydrogen-bond interaction and dipole interaction), while it is
279
positively related to the molecular size of the compounds, which agrees with previous findings
280
from other studies.3,10,12 The main difference for the two salts is that the values of a, b and s were
281
much closer to 0 for NaCl (from -0.060 to -0.042)3 than those for (NH4)2SO4 (from -0.194 to -
(3)
ACS Paragon Plus Environment
11
Environmental Science & Technology
Page 12 of 20
282
0.085). As explained in Endo et al.,3 the presence of NaCl increases the polar interaction energies
283
of aqueous solutions toward neutral organic solutes. For (NH4)2SO4, this increase of the polar
284
interactions is much larger as indicated by the larger influences of a, b and s. This means the
285
systematically higher KS((NH4)2SO4) compared to KS(NaCl) is mostly due to the coefficient v for
286
molar volume (0.287 vs. 0.171) and the constant term c (0.300 vs. 0.112).
287
Measured Setschenow constants in (NH4)2SO4 solutions have been rarely reported and only 10
288
data, mostly for aromatic compounds, were found in the literature. Comparisons were made
289
between these experimental data from the literature9,10 and estimations with the four-descriptor
290
pp-LFER model (Figure 2). Setschenow constants obtained using the pp-LFERs model are in
291
excellent agreement with experimental data, despite the large uncertainties and the different
292
methods used for measuring KS in the literature. This further supports that the pp-LFER model
293
correctly accounts for the intermolecular interactions in the ion-organic solute-water system. 1.0 data from this study data from literature
measured KS((NH4)2SO4)
0.8
1:1 line 0.6
0.4
0.2
0.0 0.0
0.2
0.4
0.6
0.8
1.0
pp-LFER estimated KS((NH4)2SO4)
294 295 296
Figure 2. Measured and pp-LFER calculated KS values for (NH4)2SO4. Squares and circles represent data from this study and the literature, respectively. The line indicates 1:1 agreement.
297
COSMOtherm Calculation. Partition coefficients between pure water and (NH4)2SO4 solution
298
(Kwater/salt
299
compounds in this study and 10 compounds from the literature.9,10 Values of log Kwater/salt water for
300
all compounds are positive, indicating a salting-out effect of (NH4)2SO4. COSMOtherm
water)
were calculated using COSMOtherm at a temperature of 298 K for the 38
ACS Paragon Plus Environment
12
Page 13 of 20
Environmental Science & Technology
301
predictions were made at different (NH4)2SO4 concentrations, 5%, 10%, 20%, 30%, and 40%
302
(w/v), which cover the experimental concentration range. Setschenow constants calculated at
303
different salt concentrations are almost identical (SD from 0.01 to 0.05 for all the 38 compounds),
304
so only KS calculated at 5% (w/v) are reported in the following discussion.
305
COSMOtherm predictions are higher than the experimental KS((NH4)2SO4) values for all the
306
compounds. The ratios between COSMOtherm-predicted and experimental KS values range from
307
2.1 to 4.5, with an average of 2.7. Other studies using COSMOtherm to predict KS also noted
308
overestimation problems.3,22,23,24 Although KS((NH4)2SO4) values were higher than the
309
experimental ones, there was a correlation between COSMOtherm predictions and measurements
310
(R2=0.58, SD=0.078) (Figure 3, green line). Tri-n-butyl phosphate and bisphenol A are two
311
outliers in this linear regression. While the discrepancy for tri-n-butyl phosphate is likely mainly
312
due to the measurement errors discussed before, no obvious reasons could be identified for
313
bisphenol A. Uncertainties in both measurement and COSMOtherm calculations could have
314
contributed to the relatively poor correlation. By removing these two compounds, the R2 of the
315
correlation increased to 0.74 (SD=0.062) (Figure 3, purple dashed line). A relationship between
316
predicted and measured KS (KS = 0.334 KS,COSMOtherm + 0.0604 ) is suggested to be used in the
317
future for correction of the salting-out effect of (NH4)2SO4 predicted by COSMOtherm. This
318
relationship between COSMOtherm prediction and measured KS((NH4)2SO4) values is almost
319
identical to a similar relationship derived for KS(NaCl).3
320
AIOMFAC Calculation. Activity coefficients at 298 K of different organic compounds in water
321
at six different concentrations of either (NH4)2SO4 (0, 5, 10, 20, 30, 40 % (w/v)) or NaCl (0, 6,
322
10, 14, 20, 36 % (w/v)) were predicted using AIOMFAC and KS values were then calculated
323
using eq. 1. As a group contribution method, AIOMFAC critically depends on the availability of
324
the appropriate group interaction parameters. Because AIOMFAC was developed primarily for
325
atmospheric applications, such parameters are available for only 17 of the compounds with
326
experimental KS for (NH4)2SO4 and NaCl reported here and by Endo et al.3 (Table S2). Moreover,
327
as the error bars in Figure 4 show, the AIOMFAC-predicted KS varied substantially at different
328
salt concentrations. For both (NH4)2SO4 and NaCl, variations are large for compounds with a
329
higher KS. The AIOMFAC-estimated KS values for (NH4)2SO4 vary more with the salt
330
concentration than those for NaCl. A large variation of KS at different salt concentrations
331
(especially for (NH4)2SO4) contradicts our experimental observations.
ACS Paragon Plus Environment
13
Environmental Science & Technology
Page 14 of 20
1.0 data from this study data from literature
measured KS((NH4)2SO4)
0.8 y = 0.334x + 0.0604 R² = 0.74
0.6
tri-n-butyl phosphate 0.4 bisphenol A 0.2
y = 0.261x + 0.144 R² = 0.58 1:3 line
0.0 0.0
0.5
1.0
1.5
2.0
2.5
3.0
COSMOtherm prediced KS((NH4)2SO4)
332 333 334 335 336
Figure 3. Measured and COSMOtherm calculated KS values for (NH4)2SO4. The black line indicates the 1:3 relation between measured and COSMOtherm estimated values. The green line and equation indicate the linear regression for all plotted data. The purple dashed line and equation indicate the regression without tri-n-butyl phosphate and bisphenol A.
337
The AIOMFAC predicted-KS, averaged over the six salt concentrations, is compared with the
338
measured KS for (NH4)2SO4 and NaCl in Figure 4. AIOMFAC tends to underpredict the
339
measured KS((NH4)2SO4), but always overpredicts KS(NaCl). The prediction error is generally
340
larger for KS(NaCl) than for KS((NH4)2SO4).
341
To illustrate the difference between AIOMFAC and pp-LFER predictions more clearly, the
342
AIOMFAC-predicted KS is compared directly with the pp-LFER-predicted KS for 102
343
compounds, including the 17 compounds for which experimental KS are available here and in
344
Endo et al.3 (Figure 4). The compounds used and their solute descriptors are given in Table S3 in
345
the Supporting Information. The compounds used for this comparison all fall within the pp-
346
LFER model’s applicability domain according to a leverage calculation. AIOMFAC and pp-
347
LFER-predicted KS are correlated for both (NH4)2SO4 (R2=0.533, SD=0.156) and NaCl
348
(R2=0.803, SD=0.102). However, the relationship differs from 1:1 agreement and is different
349
between the two salts: AIOMFAC-predicted KS(NaCl) are mostly higher than pp-LFER
350
predictions, whereas AIOMFAC-predicted KS((NH4)2SO4) are more scattered.
ACS Paragon Plus Environment
14
Page 15 of 20
Environmental Science & Technology
1.4
(NH4)2SO4 y = 1.206x - 0.217 R² = 0.53
1.2 1.0
AIOMFAC predicted KS(NaCl)
AIOMFAC predicted KS((NH4)2SO4)
1.4
1:1 line
0.8 0.6 0.4 Ks predicted by pp-LFER
0.2
y = 2.797x - 0.209 R² = 0.80
1.2 1.0
1:1 line
0.8 0.6 0.4 Ks predicted by pp-LFER
0.2
Ks measured in this study
0.0
Ks measured by Endo et al.
0.0 0.0
351
NaCl
0.2
0.4
0.6
0.8
1.0
1.2
measured or pp-LFER predicted KS((NH4)2SO4)
1.4
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
measured or pp-LFER predicted KS(NaCl)
352 353 354 355 356 357
Figure 4. AIOMFAC and measured or pp-LFER calculated KS values for (NH4)2SO4 (left) and NaCl (right). The AIOMFAC predictions are the average of the KS calculated at six different salt concentrations and the error bar shows the standard deviation. In both figures, the orange triangles show compounds with experimental KS values in this study and in Endo et al.3 The solid line indicates 1:1 agreement and the dashed line designates the linear regression between two prediction methods.
358
Relative Merits of the Prediction Methods. Each of the three different approaches to predict
359
the salting-out effect has merits and limitations. Both pp-LFER and AIOMFAC methods are easy
360
to apply because of their simplicity and general availability of model parameters. However, they
361
also have limitations. As with any other empirical and semi-empirical prediction, these models
362
are limited by their calibration data. For instance, even though the 38 calibration compounds in
363
this study were selected to represent compounds with solute descriptors covering a wide range,
364
the pp-LFER approach might still not work well for compounds that are outside of its
365
applicability domain. A statistical analysis such as the leverage calculation is recommended
366
before applying the pp-LFER to a compound to test whether the compound falls within the
367
model’s applicability domain. The group contribution method AIOMFAC, as discussed
368
previously, is limited by the availability of interaction parameters for the functional groups
369
present in a compound and will overlook intramolecular interactions that are not captured during
370
the calibration. COSMOtherm, on the other hand, is based on quantum mechanical calculations
371
and requires no experimental data. In theory, it should be able to calculate KS for any compound
372
given its molecular structure. It is also the only one of the three methods that considers different
ACS Paragon Plus Environment
15
Environmental Science & Technology
Page 16 of 20
373
conformations of a molecule, which should give more realistic results. However, based on the
374
results of the present and previous studies, the current version of COSMOtherm always tends to
375
overpredict the salting-out effect3,22,23,24 and predictions for some compounds can be outliers,
376
such as bisphenol A for (NH4)2SO4 or caffeine for NaCl.3 While more investigations are needed
377
to improve this approach, based on this work on (NH4)2SO4 and early work on NaCl,3
378
COSMOtherm in combination with a calibrated regression is still a promising tool for predicting
379
the salting-out effect of diverse compounds. Because laboratory measurements can hardly be
380
performed under supersaturated conditions, prediction models are of great significance for such
381
situations. Among all the three models, only AIOMFAC calculates activity coefficients of
382
electrolytes, water, and organics at highly concentrated and even supersaturated conditions. The
383
experimental data for the pp-LFER and the regression with COSMOtherm predictions were
384
obtained in (NH4)2SO4 solution up to 3 M. The applicability of these two methods to predict the
385
salting-out effect in supersaturated solutions needs further testing.
386
Implications for Chemical Partitioning of SOA Compounds. COSMOtherm, pp-LFER
387
and AIOMFAC were applied to study the influence of the salting-out effect on the partitioning of
388
fourteen α-pinene oxidation products between gas and aqueous phase. They were chosen because
389
of their known contribution to secondary organic aerosol formation and the availability of
390
structural information in the literature.5 Structures and Absolv-predicted solute descriptors for
391
these compounds are given in Table S4 in the Supporting Information. Setschenow constants for
392
(NH4)2SO4 predicted with these three approaches are shown in Figure 5. The linear regression in
393
Figure 3 was used to correct the COSMOtherm-predicted KS. Predictions from pp-LFER and
394
COSMOtherm are very close, while AIOMFAC predicts a higher salting-out effect for the
395
fourteen compounds. COSMOtherm-predicted Henry’s law constants or air-water partitioning
396
coefficients (Kair/water, L water/L air) of these compounds range from 2.2×10-16 to 4.2×10-7 (L
397
water/L air) at 25 °C, covering 9 orders of magnitude. According to the KS predicted with pp-
398
LFER and COSMOtherm, Kair/water for the fourteen α-pinene oxidation products increases by
399
approximately a factor of 10 to 100 due to the presence of 4 M (NH4)2SO4 (saturation
400
concentration at 25°C). Concentrations of (NH4)2SO4 in aqueous aerosols often exceed saturation
401
and also other types of salts may be present, which means there will be even more salting-out if
402
additivity of the salt effects in mixed electrolytes is assumed.25 The increase in Kair/water when
403
using the AIOMFAC-predicted KS is higher, ranging from 100 to 10,000 times for the fourteen
ACS Paragon Plus Environment
16
Page 17 of 20
Environmental Science & Technology
404
compounds. As previously discussed, AIOMFAC-predicted KS could possibly have a larger error
405
compared to the other two methods.
406
The ions present in aerosol water are predominantly NH4+, Na+, H+, SO42-, NO3-, and Cl-. Very
407
few research has focused on the salt effect of individual ions.9 Even though additivity is always
408
assumed when salt effects for electrolytes mixtures are considered (e.g. AIOMFAC), this
409
assumption should be tested experimentally and methods should be developed to separately
410
quantify the salt effect of individual ions. This would help in the development of models for
411
predicting the salting-out effect of any combinations of ions for any electrolyte mixtures, such as
412
complex aerosol water systems. 1.4 1.2
KS ((NH4)2SO4)
1.0
pp-LFER COSMOtherm, average AIOMFAC, average
0.8 0.6 0.4 0.2 0.0
413 414 415 416 417 418
Figure 5. KS((NH4)2SO4) for 14 α-pinene oxidation products predicted with COSMOtherm, ppLFER and AIOMFAC. Structures of the 14 compounds5 are given in Table S4 in the Supporting Information. COSMOtherm and AIOMFAC predictions are the average of KS at six different salt concentrations and the error bar shows the standard deviation. Values for COSMOtherm are corrected with the regression equation in Figure 3.
419
Finally, we should note that the methods introduced here only predict the extent to which
420
the presence of ammonium sulfate affects the solvation of organic compounds in water.
421
They cannot quantify how the presence of salts might influence the reactive fate of the
422
solvated organic compounds in water. An example is the effect of ammonium sulfate on the
423
hydration reaction of glyoxal or the reaction of the salt with glyoxal.26-28
424
Acknowledgement
ACS Paragon Plus Environment
17
Environmental Science & Technology
Page 18 of 20
425
We acknowledge funding from the Natural Sciences and Engineering Research Council of
426
Canada.
427
Supporting information Available
428
A table with information on the methods used for each compound, tables with solute descriptors
429
and a description of the headspace gas chromatographic method. This material is available free
430
of charge via the Internet at http://pubs.acs.org.
431
Literature Cited
432 433 434 435 436 437 438 439
1. Zhang, Q.; Jimenez, J. L.; Canagaratna, M. R.; Allan, J. D.; Coe, H.; Ulbrich, I.; Alfarra, M. R.; Takami, A.; Middlebrook, A. M.; Sun, Y. L.; Dzepina, K.; Dunlea, E.; Docherty, K.; DeCarlo, P. F.; Salcedo, D.; Onasch, T.; Jayne, J. T.; Miyoshi, T.; Shimono, A.; Hatakeyama, S.; Takegawa, N.; Kondo, Y.; Schneider, J.; Drewnick, F.; Borrmann, S.; Weimer, S.; Demerjian, K.; Williams, P.; Bower, K.; Bahreini, R.; Cottrell, L.; Griffin, R. J.; Rautiainen, J.; Sun, J. Y.; Zhang, Y. M.; Worsnop, D. R., Ubiquity and dominance of oxygenated species in organic aerosols in anthropogenically-influenced Northern Hemisphere midlatitudes. Geophys. Res. Lett. 2007, 34, L13801, doi:10.1029/2007GL029979.
440 441 442
2. Lim, Y. B.; Tan, Y.; Perri, M. J.; Seitzinger, S. P.; Turpin, B. J., Aqueous chemistry and its role in secondary organic aerosol (SOA) formation. Atmos. Chem. Phys. 2010, 10, 1052110539.
443 444 445
3. Endo, S.; Pfennigsdorff, A.; Goss, K. U., Salting-out effect in aqueous NaCl solutions: Trends with size and polarity of solute molecules. Environ. Sci. Technol. 2012, 46, 14961503.
446 447 448
4. Ervens, B.; Turpin, B. J.; Weber, R. J., Secondary organic aerosol formation in cloud droplets and aqueous particles (aqSOA): a review of laboratory, field and model studies. Atmos. Chem. Phys. 2011, 11, 11069-11102.
449 450 451
5. Zuend, A.; Seinfeld, J. H., Modeling the gas-particle partitioning of secondary organic aerosol: the importance of liquid-liquid phase separation. Atmos. Chem. Phys. 2012, 12, 3857-3882.
452 453 454
6. Shiraiwa, M.; Zuend, A.; Bertram, A. K.; Seinfeld, J. H., Gas-particle partitioning of atmospheric aerosols: interplay of physical state, non-ideal mixing and morphology. Phys. Chem. Chem. Phys. 2013, 15, 11441-11453.
455 456 457 458
7. You, Y.; Renbaum-Wolff, L.; Carreras-Sospedra, M.; Hanna, S. J.; Hiranuma, N.; Kamal, S.; Smith, M. L.; Zhang, X.; Weber, R. J.; Shilling, J. E.; Dabdub, D.; Martin, S. T.; Bertram, A. K., Images reveal that atmospheric particles can undergo liquid-liquid phase separations. Proc. Nat. Acad. Sci. U.S. 2012, 109, 13188-13193.
459 460
8. Grover, P. K.; Ryall, R. L., Critical appraisal of salting-out and its implications for chemical and biological sciences. Chem. Rev. 2005, 105, 1-10.
ACS Paragon Plus Environment
18
Page 19 of 20
Environmental Science & Technology
461 462
9. Gorgenyi, M.; Dewulf, J.; Van Langenhove, H.; Heberger, K., Aqueous salting-out effect of inorganic cations and anions on non-electrolytes. Chemosphere 2006, 65, 802-810.
463 464
10. Xie, W. H.; Shiu, W. Y.; Mackay, D., A review of the effect of salts on the solubility of organic compounds in seawater. Mar. Environ. Res. 1997, 44, 429-444.
465 466
11. Jonker, M. T. O.; Muijs, B., Using solid phase micro extraction to determine salting-out (Setschenow) constants for hydrophobic organic chemicals. Chemosphere 2010, 80, 223-227.
467 468
12. Schwarzenbach, R. P.; Gschwend, P. M.; Imboden, D. M., Environmental Organic Chemistry. 2nd ed.; John Wiley & Sons: Hoboken, NJ, USA, 2003.
469 470
13. May, W. E.; Wasik, S. P.; Freeman, D. H., Determination of the solubility behavior of some polycyclic aromatic hydrocarbons in water. Anal. Chem. 1978, 50, 997-1000.
471 472
14. Paul, M. A., The solubilities of naphthalene and biphenyl in aqueous solutions of electrolytes. J. Am. Chem. Soc. 1952, 74, 5274-5277.
473 474
15. Zuend, A.; Marcolli, C.; Luo, B. P.; Peter, T., A thermodynamic model of mixed organicinorganic aerosols to predict activity coefficients. Atmos. Chem. Phys. 2008, 8, 4559-4593.
475 476 477 478
16. Endo, S., Watanabe, N., Ulrich, N., Bronner, G., Goss, K.-U., UFZ-LSER database v 2.1 [Internet], Leipzig, Germany, Helmholtz Centre for Environmental Research-UFZ. 2014 [accessed on 15.05.2014]. Available from: http://www.ufz.de/index.php?en=31698&contentonly=1&lserd_data[mvc]=Public/start.
479 480
17. Klamt, A., From Quantum Chemistry to Fluid Phase Thermodynamics and Drug Design. Elsevier: Amsterdam, The Netherlands, 2005.
481 482
18. Eckert, F.; Klamt, A., Fast solvent screening via quantum chemistry: COSMO-RS approach. AlChE Journal 2002, 48, 369.
483 484 485 486
19. Ali, N.; Dirtu, A. C.; Van den Eede, N.; Goosey, E.; Harrad, S.; Neels, H.; t Mannetje, A.; Coakley, J.; Douwes, J.; Covaci, A., Occurrence of alternative flame retardants in indoor dust from New Zealand: Indoor sources and human exposure assessment. Chemosphere 2012, 88, 1276-1282.
487 488
20. Long, F. A.; McDevit, W. F., Activity coefficients of nonelectrolyte solutes in aqueous salt solutions. Chem. Rev. 1952, 51, 119-169.
489 490 491
21. Oleszek-Kudlak, S.; Shibata, E.; Nakamura, T., The effects of temperature and inorganic salts on the aqueous solubility of selected chlorobenzenes. J. Chem. Eng. Data 2004, 49, 570-575.
492 493 494 495
22. Oleszek-Kudlak, S.; Grabda, M.; Shibata, E.; Eckert, F.; Nakamura, T., Application of the conductor-like screening model for real solvents for prediction of the aqueous solubility of chlorobenzenes depending on temperature and salinity. Environ. Toxicol. Chem. 2005, 24, 1368-1375.
496 497 498
23. Kholod, Y. A.; Gryn'ova, G.; Gorb, L.; Hill, F. C.; Leszczynski, J., Evaluation of the dependence of aqueous solubility of nitro compounds on temperature and salinity: A COSMO-RS simulation. Chemosphere 2011, 83, 287-294.
499 500
24. Schroder, B.; Freire, M. G.; Varanda, F. R.; Marrucho, I. M.; Santos, L.; Coutinho, J. A. P., Aqueous solubility, effects of salts on aqueous solubility, and partitioning behavior of
ACS Paragon Plus Environment
19
Environmental Science & Technology
Page 20 of 20
501 502
hexafluorobenzene: Experimental results and COSMO-RS predictions. Chemosphere 2011, 84, 415-422.
503 504 505
25. Gordon, J. E.; Thorne, R. L., Salt effects on non-electrolyte activity coefficients in mixed aqueous electrolyte solutions - 2. Artificial and natural sea waters. Geochim. Cosmochim. Act. 1967, 31, 2433-2443.
506 507
26. Ip, H. S. S.; Huang, X. H. H.; Yu, J. Z. Effective Henry's law constants of glyoxal, glyoxylic acid, and glycolic acid. Geophys. Res. Lett., 2009, 36, L01802, doi:10.1029/2008GL036212.
508 509 510
27. Yu, G.; Bayer, A. R.; Galloway, M. M.; Korshavn, K. J.; Fry, C. G.; Keutsch, F. N. Glyoxal in aqueous ammonium sulfate solutions: Products, kinetics and hydration effects. Environ. Sci. Technol. 2011, 45, 6336-6342.
511 512 513 514
28. Kampf, C. J.; Waxman, E. M.; Slowik, J. G.; Dommen, J.; Pfaffenberger, L.; Praplan, A. P.; Prevot, A. S. H.; Baltensperger, U.; Hoffmann, T.; Volkamer, R. Effective Henry's law partitioning and the salting constant of glyoxal in aerosols containing sulfate. Environ. Sci. Technol. 2013, 47, 4236-4244.
515
516
TOC Art
517
ACS Paragon Plus Environment
20