1-octanol-water partitioning as a classifier of water soluble organic matters: Implication for solubility distribution

ABSTRACT

Water-soluble organic matters (WSOMs) play an important role in determining magnitudes of climatic and environmental impacts of organic aerosol particles because of their contributions to hygroscopic growth and cloud formation. These processes are dependent on water solubility as well as distribution of this property in a particle, yet no method has been available to quantify such characteristics. In this study, we developed a theoretical framework to classify WSOM by 1-octanol-water partitioning that has a strong correlation with water solubility. 1-octanol-water partitioning coefficient also has a strong correlation with a traditional solid phase extraction method, facilitating interpretation of data from the technique. The theoretical analysis demonstrated that the distributions of WSOM classified by 1-octanol-water partitioning depend on (1) the volume ratio of 1-octanol and aqueous phases, and (2) extraction steps. The method was tested by using organic aerosol particles generated by smoldering of a mosquito coil, which serves as a surrogate for biomass burning particles. The WSOM extracted from the mosquito coil burning particles was classified by 1-octanol-water partitioning at different volume ratios. These solutions, including both the 1-octanol and aqueous phases, were nebulized to generate particles for measurements using an online aerosol mass spectrometer. The mass spectra indicated that highly oxygenated species tend to be highly soluble, while high molecular weight compounds are less soluble. Linear combinations of these mass spectra allowed the estimation of the mass fractions of WSOM partitioned to 1-octanol and aqueous phases, thereby facilitating the evaluation of the mass fractions of cloud condensation nuclei (CCN) active materials.

© 2017 American Association for Aerosol Research

EDITOR:

• Spyros Pandis

1. Introduction

Aerosol particles in the atmosphere affect both the global climate and local environments (Carslaw et al. 2010; Fiore et al. 2012). For instance, aerosol particles serve as cloud condensation nuclei (CCN) when water vapor is supersaturated, leading to the formation of cloud droplets (Farmer et al. 2015). Aerosol particles also influence visibility as well as the radiation budget of the Earth by scattering solar radiation (Ramanathan et al. 2001). This effect is particularly pronounced when these particles uptake water vapor, as hygroscopic growth increases diameter of particles (Yu et al. 2006; Wang et al. 2008). In addition, aerosol particles induce adverse health effects, including respiratory problems (Laumbach and Kipen 2012). The deposition pattern of aerosol particles in the human body depends on their hygroscopic growth properties because the diameter change in particles affects their dynamics (Londahl et al. 2008). Understanding the hygroscopic properties of aerosol particles is important for investigating their climatic and environmental impacts.

Hygroscopic properties as well as CCN activities of pure inorganic and organic compounds have been investigated by both theoretical and experimental approaches, demonstrating that water solubility (S) is a key parameter to control water uptake properties (Shulman et al. 1996; Petters and Kreidenweis 2008; Sullivan et al. 2009; Riipinen et al. 2015). These studies demonstrated that organic compounds can be classified to three categories, based on their water solubility (Sullivan et al. 2009; Kuwata et al. 2013). For instance, highly soluble compounds, including levoglucosan and malonic acid, are highly CCN active and hygroscopic (Prenni et al. 2001; Mochida and Kawamura 2004; Rosenorn et al. 2006). On the other hand, slightly soluble species such as adipic and pinonic acids are only slightly CCN active (Kuwata et al. 2013). The theoretical analysis of CCN activation mechanisms by the Köhler theory provides an explanation for the experimental results (Petters and Kreidenweis 2008; Kuwata et al. 2013; Riipinen et al. 2015). Highly soluble species deliquesce at subsaturated humidity (i.e., relative humidity [RH < 100%]) to form aqueous particles. On the other hand, slightly soluble species deliquesce only at the supersaturated regime (RH > 100%). Insoluble species (S < ∼10⁻⁴ [g cm⁻³]) exhibit no water uptake, while highly soluble species (S > ∼10⁻¹ [g cm⁻³]) significantly contribute to water uptake. Compounds with marginal solubility (∼10⁻⁴ [g cm⁻³] < S < ∼10⁻¹ [g cm⁻³]) are considered as slightly soluble. It is noted that these values separating different types of organic compounds have some variability, depending on particle size and mixing state (Petters and Kreidenweis 2008; Sullivan et al. 2009; Kuwata et al. 2013). A recent theoretical analysis demonstrated that the distribution of water solubility is important for CCN activity when multiple organic compounds exist in a single particle (Riipinen et al. 2015).

A significant fraction of atmospheric aerosol particles comprises organic compounds, which include tens of thousands of species (Hamilton et al. 2004; Jimenez et al. 2009). Classification of these organic compounds is crucial, as the complexity of the chemical composition inhibits identification at the molecular level (Hallquist et al. 2009). One of the most popular classifications is the separation of water soluble (WSOM) and insoluble organic matters (Psichoudaki and Pandis 2013). Further separation of WSOM is required to understand water uptake properties of organic aerosol particles, as it contains a broad spectrum of compounds (Kuwata et al. 2013; Psichoudaki and Pandis 2013). For this reason, separation of WSOM has been conducted using solid-phase extraction as well as high performance liquid chromatography techniques (Decesari et al. 2000; Sullivan and Weber 2006; Asa-Awuku et al. 2008; Polidori et al. 2008). WSOM extraction by multiple solvents was also found as a useful technique to separate them (Mihara and Mochida 2011). Although these techniques are demonstrated as useful in classifying WSOM, relevance between these classification methods and water solubility has been unclear, preventing the detailed theoretical analysis of water uptake properties. In addition, WSOM classifications by these techniques are not predictable, requiring a significant effort for characterizing methods (Sullivan and Weber 2006). Development of a method to infer solubility distribution of actual aerosol particles will be advantageous in promoting our understanding on solubility dependent water uptake properties of organic materials.

In this study, we explored the possibility to use 1-octanol-water partitioning as a method to classify WSOM by their water solubility (S). The purpose of the study is to provide a theoretical background to classify WSOM by 1-octanol-water partitioning, and qualitatively test the method using a complex mixture of organic compounds. Water solubility and 1-octanol-water partitioning coefficient (K_OW) are tightly correlated, allowing separation of complex mixtures of organic compounds by water solubility. Unlike solid phase extraction methods, K_OW can be reliably predicted using functional group contribution methods, allowing detailed interpretation of data without extensive characterizations (Meylan and Howard 1995; Meylan et al. 1996). A theoretical analysis is provided. The method is tested for organic material produced by smoldering of a mosquito coil, which has similar chemical characteristics as ambient biomass burning particles (Ng et al. 2011; Li et al. 2012). The result from the experiment using a mosquito coil qualitatively supports the theory, suggesting the further quantitative verification of the method, which includes experimental tests by standard chemicals, could be promising.

2. Theory

In this section, a theoretical background for the classification of organic compounds by 1-octanol-water partitioning is provided, after introducing relationships between K_OW and S as well as notation for solutions after liquid–liquid extractions.

2.1. K_OW–S relationship

The value of K_OW of compound i (K_{OW_i}) is defined by the following equation (Sangster 1989):[1] where and are the molar concentrations of compound i in the 1-octanol and aqueous phases, respectively. The correlation between K_OW and S has been established both for molar and mass concentrations (Kenaga and Goring 1980; Chiou et al. 1982). In this study, we focus on K_OW–S (mass concentration) relationship, as the unit is more useful in describing water uptake properties of aerosol particles. Several K_OW–S relationships have been proposed using various experimental datasets, as compared in Figure 1 and Table 1 (Kenaga and Goring 1980; Valvani et al. 1981; Coates et al. 1985; Isnard and Lambert 1989).

Figure 1. K_OW–S relationships of organic compounds. Empirical relationships for these quantities are also shown (Table 1).

Table 1. The relationships between K_OW and S proposed in literature. Unit conversion factors were applied to original equations so that S can be represented in g cm⁻³.

Reference	Equation
Kenaga and Goring (1980)	log KOW = −0.642 − 0.8 log S
Valvani et al. (1981)	log KOW = −0.63 − 1.05 log S
Coates et al. (1985)	log KOW = −1.79 − 0.94 log S
Isnard and Lambert (1989)	log KOW = 0.3 − 0.72 log S

Due to the nature of empirical correlations, the proposed K_OW–S relationships have some variability, especially for highly soluble (low K_OW) compounds. To find the best K_OW–S relationship for atmospheric application, the empirical correlations proposed by these studies were tested for atmospherically relevant compounds. The list of water-soluble atmospheric relevant compounds, which include both primary and secondary species, is provided in Table S1 (Yu et al. 1999; Graham et al. 2002). Experimental values for K_OW and S were used when these values are available in literature (Yalkowsky and Dannenfelser 1992; Hansch et al. 1995; Yalkowsky and He 2003; Hartz et al. 2006). Functional group contribution methods such as KOWWIN v 1.68 and WSKOW v. 1.41, which are implemented in Estimations Programs Interface for Windows (EPI Suite) from the United States Environmental Protection Agency (US EPA), were employed when experimental data are not available (Meylan and Howard 1995; Meylan et al. 1996). Predicted values of K_OW and S by these programs agreed well with these experimental data for atmospheric relevant chemical species (Figure S1). Figure 1 shows that Kenaga and Goring (1980) provide the best estimates for K_OW–S relationship for atmospheric relevant species. The parameterization by Kenaga and Goring (1980) successfully reproduced S of the atmospheric relevant compounds (Table S1) from K_OW within the error of Δlog S = 0.39 (mean squared error), providing an estimate of the uncertainty.

2.2. Notation for solutions extracted by 1-octanol-water partitioning

Numerous 1-octanol and aqueous solutions are produced, following liquid–liquid extractions by 1-octanol and water (Figure 2).

Figure 2. Notations for different solutions.

Providing appropriate notations to these solutions will facilitate discussing characteristics of chemical species in them. An aqueous solution of WSOM is obtained by dissolving sampled aerosol particles in water using an offline (e.g., filter sampling) or online (e.g., Particle-into-Liquid-Sampler: PILS) technique (Weber et al. 2001). This solution is called as A₀. 1-octanol is added to an aliquot of A₀, classifying organic compounds by 1-octanol-water partitioning. The 1-octanol phase is named as O₁A₀, whereas the aqueous phase is called as A₁A₀. Organic compounds in O₁A₀ are further separated by adding water, while 1-octanol could be added to A₁A₀. The procedure can be repeated until the desired classification of organic compounds is achieved. In general, the notation of a solution following N extractions (X_N: X is A or O) is given as follows:[2] where N is a positive integer. This notation rule is employed throughout the study.

2.3. Single extraction from aqueous solution (N = 1)

By employing the definition of molar concentration, K_{OW_i} for the single extraction case is also represented as[3]

In the case of dilute solutions, the total volumes of the 1-octanol () and aqueous () phases are predominantly determined by the volumes of the solvents. The total number of moles of compound i in the original WSOM solution () is equivalent to the sum of the number of moles in the 1-octanol () and aqueous () phases.[4]

Combining Equations (3[3] ) and (4[4] ), the mole fractions of compound i in the 1-octanol () and aqueous () phases are represented as[5] and[6] respectively. In these equations, the volumetric ratio of 1-octanol and aqueous phases is defined as . Both and are only dependent on K_{OW_i} and . The chemical characteristics of compound i determines K_{OW_i}. On the other hand, is a controllable variable, meaning that the classification of organic compounds can be regulated by this parameter.

As discussed in Section 2.1, K_OW of atmospheric relevant compounds could be converted to S using the empirical relationship developed by Kenaga and Goring (1980). Using the relationship (Table 1), Equations (5[5] ) and (6[6] ) can be converted as follows to provide and as a function of S_i (water solubility of compound i):[7] [8]

These equations quantitatively relate f_i and water solubility.

Figure 3 shows some examples of and , which are plotted as functions of K_OW and S. Three different values of (0.01, 0.1, and 1) are employed for the calculation. The calculated distributions of and have sigmoidal shapes when plotted in logarithmic scales. Compounds with lower values of K_OW tend to partition to the aqueous phase, especially when the values of are smaller (Figure 3a).

Figure 3. (a) Fractions of organic materials partitioning to 1-octanol and aqueous phases following single extraction, plotted against K_OW. (b) Same as (a) but plotted against S.

This trend is translated to the capability of 1-octanol-water partitioning as a solubility classifier (Figure 3b). When is equal to 1, the aqueous phase is enriched by organic compounds with high solubility (S > 10⁻¹ [g cm⁻³]). On the other hand, a significant fraction of much less soluble species (S ∼ 10⁻³ [g cm⁻³]) could also partition to the aqueous phase for a lower value of (). This theoretical analysis demonstrates that 1-octanol-water partitioning is a useful tool to obtain solubility distribution of WSOM. The estimated distribution functions shown in Figure 3 are relatively broad due to the nature of a single extraction. The classification resolution would be improved by repeating extractions (multiple extractions) at the same values of, although the method could be time and labor intensive.

2.4. Multiple extractions (N ≥ 2)

The classification resolution of organic compounds by 1-octanol-water partitioning is further improved by repeating the liquid–liquid extraction multiple times. The fractions of compound i in the 1-octanol and aqueous phases following the Nth extraction () are denoted as and , respectively. In the case of multiple extractions, these values can be related with K_{OW_i} by the following equations:[9] [10]

In Equations (9[9] ) and (10[10] ), r^(N-1) is volume fraction of X_N-1 solution, which is used for the Nth extraction. The value of r^(N-1) is equal to unity when the whole part of X_N−1 is used for the Nth extraction, whereas it is less than one if only a fraction of X_N−1 is employed for the Nth extraction. Combinations of Equations (9[9] ), (10[10] ), and parameterization by Kenaga and Goring (1980) yield and as functions of solubility.[11] [12]

Figure 4 shows examples of and . The values of are assumed as (. The distributions of , and shown in Figure 4 are similar to , , and shown in Figure 3, respectively. On the other hand, and (Figure 4) are significantly different from , and shown in Figure 3. The distributions of and are approximated well by lognormal distributions, and they do not resemble sigmoidal distributions any more. An important implication from this theoretical analysis is that the classification resolution by 1-octanol-water partitioning can be improved by repeating the liquid–liquid extraction multiple times.

Figure 4. (a) Fractions of organic materials partitioning to 1-octanol and aqueous phases following multiple extractions, plotted against K_OW. (b) Same as (a) but plotted against S.

3. Experiment

A mosquito coil, which does not contain any artificially synthesized compounds (Tennen Jochuugiku, Kicho-no-uzumaki, natural pyrethrum spiral, Kincho, Osaka, Japan), was burned in an enclosed 100-L stainless steel container. The container was equipped with an air inlet and outlet. Particle free air was continuously supplied from the air inlet. The chemical composition of aerosol particles emitted from the mosquito coil burning was measured using the Time-of-Flight Aerosol Chemical Speciation Monitor (ToF-ACSM; Aerodyne Research, Inc., Billerica, MA, USA) (Froehlich et al. 2013). Simultaneously, aerosol particles were sampled using two sets of 47-mm Teflon filters (P0450, Sigma-Aldrich, St. Louis, MO, USA) and filter holders (BGI, Inc., Butler, NJ, USA) at flowrates of 5 L min⁻¹. The collected samples were stored in a freezer (−27°C) until analysis. A schematic diagram of the experimental setup is available in the online supplementary information (SI; Figure S2).

The filter samples were extracted using approximately 10 mL of purified water (Type I, 18.2 MΩ·cm) in an ultrasonic bath for 30 min. The resulting aqueous solution was filtered using a PTFE syringe filter (514-0070, VWR International, Radnor, PA, USA) to remove debris. The pore size of the filter is 0.2 µm. A fraction of the solution was nebulized for the ToF-ACSM analysis to obtain the mass spectrum of WSOM (A₀). An aliquot of the aqueous solution was used for liquid–liquid extraction by 1-octanol (Wako first grade, Wako Pure Chemical Industries, Osaka, Japan). The volumes of the aqueous solution and 1-octanol were measured using syringes with metal needles. These liquids were mixed in a separation funnel. Only single extraction experiments were conducted. The values of were set as 0.01, 0.1, and 1. The separation funnel was vigorously shaken to form an emulsion. Subsequently, the funnel was placed at a funnel stand at least for an overnight period, separating 1-octanol and aqueous phases clearly. The 1-octanol and aqueous phases were separately sampled, and transferred to glass vials.

Subsequently, the 1-octanol solutions were diluted more than 10 times by methanol (ACS grade, 99.8%, VWR International). The aqueous phase solutions were used for analysis without any dilution. These solutions were nebulized using glass nebulizers. A mass flow controller (MC-20 SLPM-D, Alicat Scientific, Inc., Tucson, AZ, USA) was used to regulate the particle free air supplied to the nebulizer. Following nebulization, the solvents were removed using a diffusion dryer (Model 42000, Brechtel Manufacturing, Inc., Haywood, CA, USA) filled with activated charcoal (granular 6–10 mm, Wako Pure Chemical Industries Osaka, Japan) or silica gel (Chameleon 83000.360, VWR International). Charcoal was used for the 1-octanol solution diluted by methanol, whereas silica gel was employed for aqueous solutions. As methanol is a highly volatile solvent, it does not affect quantification by the ACSM. A trace amount of residual 1-octanol signal, which is negligibly small compared with that from mosquito coil burning particles, were observed after removal of organic vapor by charcoal (Figure S3). The chemical composition of organic aerosol particles was measured using the ToF-ACSM. The mass spectra measured by the ToF-ACSM were analyzed using a fragmentation table (Allan et al. 2004).

4. Results and discussion

4.1. Visual inspection of the 1-octanol and aqueous phases

Figure 5 shows a photo of O₁A₀ and A₁A₀ of mosquito coil burning particles, classified at = 0.01, 0.1, and 1. In general, light-absorbing species tend to partition to the 1-octanol phase. On the other hand, A₁A₀ (= 1) is almost colorless, whereas A₁A₀ (= 0.01) is slightly yellow. This result indicates that light-absorbing organic compounds are relatively less soluble because A₁A₀ (= 1) should only contain highly soluble species (Figure 3).

Figure 5. A photograph of 1-octanol (O₁A₀) and aqueous (A₁A₀) phases separated at = 0.01, 0.1, and 1. The samples were prepared by dissolving organic aerosol particles that were generated by mosquito coil burning in water. See the text for details.

Light absorbing species in biomass burning particles is known to be dominantly composed of humic like substances (HULIS; Hoffer et al. 2006; Wu et al. 2016), which is frequently separated from aerosol samples using solid phase extraction techniques (Graber and Rudich 2006; Fan et al. 2013). One of the most commonly used materials for solid phase extraction of HULIS is XAD-8. The separation characteristics of organic compounds by XAD-8 are compared with K_OW in Figure A1. In general, organic compounds can easily penetrate XAD-8 column if the values of their K_OW is smaller than unity (log K_OW < 0). On the other hand, organic compounds with higher values of K_OW (K_OW > 1 or log K_OW > 0) tend to stay on the surface of XAD-8 column. The experiment by Sullivan and Weber (2006) demonstrated that fluvic and humic acids from Suwannee River, which serve as surrogate of HULIS, do not penetrate XAD-8, although K_OW of these compounds are unknown, The implication of this experiment is that HULIS likely have relatively large values of K_OW (K_OW > 1). Thus, we hypothesize that HULIS in mosquito coil burning particles dominantly partitioned to 1-octanol phase, producing light-absorbing solutions.

4.2. ACSM mass spectra

Figure 6 shows the ACSM mass spectra of mosquito coil burning particles, which are (a) directly measured from the burning chamber (online) and (b) WSOM (A₀). Ordinate of the figure is fractional contribution of ions at each m/z to the whole mass spectra. The online measurement data have a minor contribution of m/z44 signal and intense peaks of hydrocarbons, such as m/z55 (C₄H₇⁺) and 57 (C₄H₉⁺). Intensities of these ions indicate that mosquito coil burning particles are not highly oxygenated, as m/z44 is considered as a marker for degree of oxygenation (Ng et al. 2011). The marker ion for levoglucosan-like species (m/z60), which is known as an indicator of biomass burning particles, is clearly identified (Cubison et al. 2011). These results are qualitatively consistent with those of a previous study on mosquito coil burning particles (Li et al. 2012). However, some high m/z ions reported in the previous work (m/z121, 137, and 167) are not distinct in Figure 6a. The dissimilarity could be due to the differences in burning condition and/or types of mosquito coils. The ACSM mass spectrum of WSOM extracted from mosquito coil (A₀) has a higher fraction of m/z44 than that obtained in the online measurement. In addition, the intensities of the hydrocarbon peaks are much less than those in the online data. These results show that the WSOM fraction is more oxygenated. The fraction of high-molecular weight ions (m/z > 100) for A₀ is less than that for the online measurement, implying that most of high molecular weight species are insoluble in water.

Figure 6. Normalized mass spectra of organic materials quantified by the ToF-ACSM. (a) Online measurement and (b) WSOM.

Figure 7 shows the mass spectra of A₁A₀ and O₁A₀ separated at different values of. The most significant signal for A₁A₀ (= 1) is m/z44, indicating that the organic material in this fraction is highly oxygenated (Ng et al. 2011). The mass spectrum also shows a strong signal at m/z60, implying the significant abundance of levoglucosan-like species (Cubison et al. 2011). On the other hand, the m/z44 signal for O₁A₀ (= 1) is much less pronounced than that for A₁A₀ (= 1). The intensity of the marker ion for levoglucosan-like species (m/z60) is also almost negligible for O₁A₀ (= 1). Furthermore, the signals for hydrocarbons (e.g., m/z55 and 57) and high molecular weight ions (m/z >100) for O₁A₀ (= 1) are much more intense than that for A₁A₀ (= 1). Taking the theoretical consideration (Section 2.3) and K_OW–S relationship (Table 1) into account, these results experimentally demonstrate the following: (1[1] ) both highly oxygenated and levoglucosan-like species are highly soluble in water, and (2[2] ) hydrocarbon-like and high molecular weight compounds are less soluble, although they could be a part of WSOM. These findings are also supported by comparing A₁A₀ and O₁A₀ at different values of (= 0.01 and 0.1). The intensities of m/z44 for A₁A₀ decrease for lower values of, indicating that highly oxygenated compounds predominantly exist in highly soluble fraction. In fact, m/z44 signals for the less soluble fractions (O₁A₀ [= 0.01 and 0.1]) are as small as “hydrocarbon-like organic aerosol (HOA)” fractions retrieved by the Positive Matrix Factorization (PMF) analysis of atmospheric organic aerosol particles (Ng et al. 2011). Similarly, fractions of high molecular weight ions (m/z > 100) are consistently larger for less soluble fractions.

Figure 7. Mass spectra of WSOM measured by the ToF-ACSM following classification by 1-octanol-water partitioning (single extraction). (a) Aqueous phase (A₁A₀, = 1), (b) 1-octanol phase (O₁A₀, = 1), (c) aqueous phase (A₁A₀, = 0.1), (d) 1-octanol phase (O₁A₀, = 0.1), (e) aqueous phase (A₁A₀, = 0.01), and (f) 1-octanol phase (O₁A₀, = 0.01).

4.3. Fractions of WSOM partitioned to 1-octanol and aqueous phases

Mass fractions of WSOM partitioned to the 1-octanol and aqueous phases were estimated, combining the mass spectra of A₀, A₁A₀, and O₁A₀. The estimation was performed by numerically minimizing the following metric:[13] where denotes the normalized signal intensity (i.e., fraction) of ions at m/z = i observed for X_n solution and a is the parameter to be optimized (0 ≤ a ≤ 1) (Figure S4). The mass fraction of organic materials partitioned to the 1-octanol phase corresponds to a, while that for the aqueous phase is equivalent as (1 − a). The numerical calculation successfully reconstructed the original mass spectrum of A₀, confirming the validity of the approach (Figure S5). A merit of this experimental approach is that absolute concentrations of WSOM in the aqueous and 1-octanol phases do not affect reconstruction of the mass spectra as long as mass spectra of these species are accurately quantified.

Figure 8 shows the estimated fractions of organic materials partitioned to the 1-octanol and aqueous phases. The optimized values of a are 0.16 ( = 0.01), 0.33 ( = 0.1), and 0.55 ( = 1). In general, the fractions of organic materials partitioned to the 1-octanol phase increase with , which is in good agreement with the theoretical prediction (Figure 3). The obtained value of a ( = 0. 01) indicates that solubility of approximately 84% of extracted WSOM could be higher than 10⁻⁴ (g cm⁻³), potentially contributing to CCN activation (Figure 3; Petters and Kreidenweis 2008; Kuwata et al. 2013). It should be noted that particle size also needs to be considered for CCN activation in the atmosphere (Dusek et al. 2006). Furthermore, the fraction of WSOM in the highly soluble range for CCN activation (S > 10⁻¹ [g cm⁻³]) is estimated to be around 45% on the basis of the values of a ( = 1) and the theoretical evaluation for solubility classification (Figure 3; Petters and Kreidenweis 2008; Kuwata et al. 2013).

Figure 8. Mass fractions of WSOM partitioned to 1-octanol and aqueous phases at = 0.01, 0.1, and 1. These values are estimated from the ACSM mass spectra. See the text for details.

The majority of the uncertainty in estimating the solubility distribution using this experimental approach is likely associated with the empirical correlation between K_OW and S. Considering the very good agreement (slope = 1.007, r² = 0.999) between the observed and reconstructed WSOM mass spectra (Figure S5), which occurs only at a certain value of a, (Figure S4), there would not be a significant error in estimating fractions of compounds those partitioned to aqueous and 1-octanol phases. The good correlation also supports that there was not a significant artifact associated with separation and measurement processes. On the other hand, the empirical relationship between K_OW and S has an uncertainty of Δlog (S) = ±0.39. The uncertainty associated with the nature of the empirical correlation should be considered for quantitative interpretation of the data.

The promising result from the mosquito coil burning experiment encourages further verification and development of the method. For instance, quantitative experiments using atmospheric relevant pure species will help validating the method in more detail. A systematic test of multiple extraction approaches, which was indicated to be successful for our preliminary tests (not shown), will enhance our capability to classify WSOM by the 1-octanol-water partitioning method. Development of an appropriate data inversion algorithm, which employs the theoretical curves as kernel functions (Figures 3 and 4), will be helpful to estimate solubility distribution from experimental data. Experimentally determined solubility distribution by such a method will be eventually useful for predicting the CCN activity of aerosol particles (Riipinen et al. 2015).

5. Conclusions

A classification method for water-soluble organic matters (WSOMs), which employs 1-octanol-water partitioning, was developed. The partitioning coefficients of organic compounds between 1-octanol and water are known to correlate with their water solubility. A theoretical analysis of the method demonstrated that the solubility of organic compounds partitioning to the aqueous phase tends to be higher when a volume ratio of 1-octanol to water () is large. In addition, the theoretical analysis showed that classification resolution by the partitioning method is improved by multiple extraction approaches.

The developed concept was evaluated using organic aerosol particles generated by smoldering of a mosquito coil. Following sample collection, WSOM in the collected samples were extracted. A mass spectrum of WSOM was recorded using the Time-of-Flight Aerosol Chemical Speciation Monitor (ToF-ACSM). In addition, WSOM was classified by the 1-octanol-water partitioning method at = 0.01, 0.1, and 1. The mass spectra of WSOM partitioned to both the 1-octanol and aqueous phases were measured using the ToF-ACSM. WSOM partitioned to aqueous phases were highly oxygenated, whereas those for 1-octanol phases contained significant fractions of hydrocarbon peaks. In fact, the fraction of m/z44 for WSOM partitioned to a 1-octanol phase (= 0.01) was as insignificant as that of hydrocarbon-like organic aerosol (HOA).

The reconstruction of the WSOM mass spectrum allowed the estimation of the fractions of materials partitioned to them. The analysis indicated that the major fraction of WSOM was sufficiently soluble for contributing to cloud condensation nuclei (CCN) activation. The promising result from the mosquito coil burning experiment encourages conducting detailed verification experiment using standard chemicals in the future. In addition, future development of an appropriate data inversion algorithm will be helpful in establishing an accurate estimation method of solubility distribution of WSOM, which is important for predicting the CCN activity of aerosol particles.

Acknowledgments

The authors acknowledge Jing Chen for useful discussions and assistance in the experiment.

Funding

This research was supported by the National Research Foundation Singapore under its Singapore NRF Fellowship scheme (National Research Fellow Award, NRF2012NRF-NRFF001-031), the Earth Observatory of Singapore (EOS), and Nanyang Technological University.

Appendix:

Comparison with solid phase extraction

Classification of WSOM in atmospheric aerosol particles has been conducted using solid phase extraction technique (Duarte and Duarte 2005; Sullivan and Weber 2006). Comparison of the existing technique with 1-octanol-water extraction method, which is newly introduced in the present study, is beneficial in comparing outputs from different techniques. Sullivan and Weber (2006) tested penetration efficiency of XAD-8 column using 37 compounds. Among these species, values of K_OW and penetration efficiency of XAD-8 column were compared for 33 compounds. Experimentally determined values of K_OW were employed when these data are available. Values of K_OW estimated using KOWWIN v1.68 were also used for this comparison. All the data for the comparison are summarized in Table S2.

Figure A1 compares K_OW with penetration ratios through XAD-8 column. All the compound with high penetration ratios (∼100%) have relatively low values of K_OW (log K_OW ≤ 0.59). On the other hand, the values of K_OW for non-penetrating species are higher ((log K_OW ≥ −0.48). The clear separation of penetration experiment data for XAD-8 by K_OW demonstrates that classification of WSOM by 1-octanol-water partitioning is comparable to traditional solid phase extraction. An advantage of 1-octanol-water partitioning is that the flexibility of the method allows designing shapes and thresholds of WSOM classification.

Figure A1. Comparison of penetration ratio of organic compounds through XAD-8 column measured by Sullivan and Weber (2006) with (a) experimentally determined K_OW, and with (b) estimated values of K_OW.