Liquid-liquid phase separation in particles containing secondary organic 1 material free of inorganic salts

14 Particles containing secondary organic material (SOM) are ubiquitous in the atmosphere and play 15 a role in climate and air quality. Recently, research has shown that liquid-liquid phase separation 16 (LLPS) occurs at high relative humidities (RH) (greater than 95 %) in α-pinene-derived SOM 17 particles free of inorganic salts while LLPS does not occur in isoprene-derived SOM particles free 18 of inorganic salts. We expand on these findings by investigating LLPS in SOM particles free of 19 inorganic salts produced from ozonolysis of β-caryophyllene, ozonolysis of limonene, and photo20 oxidation of toluene. LLPS was observed at greater than ~95 % RH in the biogenic SOM particles 21 derived from β-caryophyllene and limonene while LLPS was not observed in the anthropogenic 22 SOM particles derived from toluene at 290 ± 1 K. This work combined with the earlier work on 23 LLPS in SOM particles free of inorganic salts suggests that the occurrence of LLPS in SOM 24 particles free of inorganic salts is related to the average oxygen-to-carbon elemental ratio (O:C) of 25 the organic material. When the average O:C is between 0.25 and 0.60, LLPS was observed, but 26 when the average O:C was between 0.52 and 1.3, LLPS was not observed. These results help 27 Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2017-408, 2017 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 11 May 2017 c © Author(s) 2017. CC-BY 3.0 License.


Introduction
Secondary organic material (SOM) is produced in the atmosphere by the oxidation of volatile organic compounds (VOCs) such as α-pinene and isoprene from trees and toluene from anthropogenic sources.Once formed, the low-volatility and semivolatile oxidation products can partition to the particle phase to form SOM-containing particles (Hallquist et al., 2009;Ervens et al., 2011).SOM accounts for approximately 20-80 % of the submicrometer particle mass in the atmosphere (Zhang et al., 2007;Jimenez et al., 2009).Although the exact chemical composition of SOM in atmospheric particles remains an active area of research, laboratory and field studies have shown that the average oxygen-to-carbon elemental ratio (O : C) of these particles ranges from 0.2 to 1.0 (Chen et al., 2009;Jimenez et al., 2009;Heald et al., 2010;Takahama et al., 2011).
As the relative humidity (RH) varies in the atmosphere, SOM-containing particles can undergo several different phase transitions with implications for the cloud condensation nuclei (CCN) properties, optical properties, reactivity, and growth of these particles (Martin et al., 2000;Raymond and Pandis, 2002;Bilde and Svenningsson, 2004;Zuend et al., 2010;Kuwata and Martin, 2012;Brunamonti et al., 2015).One possible phase transition that SOM particles may undergo as RH varies in the atmosphere is liquidliquid phase separation (LLPS) (Pankow, 2003;Marcolli et al., 2006;Ciobanu et al., 2009: Zuend andSeinfeld, 2012;Veghte et al., 2013;O'Brien et al., 2015).LLPS in particles containing both SOM and inorganic salts has been the focus of many recent studies.These studies have shown that SOM particles mixed with inorganic salts can undergo LLPS in the atmosphere when the O : C of the organic material is less than 0.56, but LLPS may not occur when the O : C of the organic material is greater than 0.80 (Bertram et al., 2011;Krieger et Table 1.Experimental conditions for production and collection of SOM particles by ozonolysis.Included is the measured separation relative humidity (SRH) upon moistening and mixing relative humidity (MRH) upon drying of the collected particles.Particles were collected on hydrophobic substrates using an electrostatic precipitator or single-stage impactor.The uncertainty in the MRH and SRH values al., 2012;Smith et al., 2012;Song et al., 2012aSong et al., , 2013;;Schill and Tolbert, 2013;You et al., 2013You et al., , 2014)).
Recently, researchers have also focused on LLPS in SOM particles free of inorganic salts.Petters et al. (2006) suggested that a miscibility gap in particles containing organic polymers at high RH may lead to a non-classical pathway for CCN activation.Renbaum-Wolff et al. (2016) showed that α-pinene-derived SOM free of inorganic salts can undergo LLPS at high RH values (∼ 95 to 100 %), which could result in altered CCN properties.In addition, they showed that LLPS in SOM particles will lead to a different hygroscopic parameter, κ, at subsaturated conditions compared to supersaturated conditions.The implication is that the CCN activity of SOM particles, if they undergo LLPS, is higher than predicted from subsaturated hygroscopicity measurements.Related, Hodas et al. (2016), using a combination of measurements and modeling of surrogates of oligomers, showed that the prevalence of LLPS at high RH can contribute to differences in hygroscopicity above and below water saturation.Most recently, Rastak et al. (2017) observed that isoprenederived SOM particles do not undergo LLPS even at high RH.Rastak et al. (2017) used these results together with thermodynamic calculations to explain the hygroscopic properties of biogenic organic aerosol particles in the laboratory and the field.

Methods
2.1 Production of secondary organic materials SOM particles were generated via β-caryophyllene ozonolysis and limonene ozonolysis in a flow tube reactor and an oxidation flow reactor (OFR) (Table 1), and via toluene photooxidation in the OFR (Table 2).The method of particle generation in the flow tube reactor was described in Shrestha et al. (2013) and the methods of particle generation in the OFR (Kang et al., 2007) were given in Liu et al. (2015).The flow tube reactor was operated at a flow rate of 3.5 L min −1 (with a residence time of 38 s) and < 5 % RH.The OFR was operated at flow rates of 7.0 L min −1 (with a residence time of 110 s) and 13 ± 3 % RH.Both reactors were operated at a temperature of 293 ± 2 K.
Table 1 lists the experimental conditions for the production of SOM via ozonolysis.For the particle generation via ozonolysis, ozone was produced by irradiating pure air (Aadco 737 Pure Air Generator) with ultraviolet emission from a mercury lamp (λ = 185 nm).Ozone concentrations used for ozonolysis ranged from 12 to 30 ppm for βcaryophyllene and 13 to 30 ppm for limonene (Table 1).β-Caryophyllene and limonene (Sigma-Aldrich, ≥ 99 %) were dissolved in 2-butanol (Sigma-Aldrich, ≥ 99.5 %).These organic solutions were injected into a glass round-bottom flask held at 310 K, where the organic liquids vaporized at the tip of a syringe.The organic vapor was then swept into the reactor where ozonolysis took place to form SOM and particles.
The injected precursor concentrations were 0.03-0.7 ppm for β-caryophyllene and 0.07-2.0ppm for limonene in the main flow of the reactor.In the ozonolysis experiments, butanol served as an OH radical scavenger.
Table 2 presents the experimental conditions for the production of SOM via photo-oxidation.For the particle generation via photo-oxidation, hydroxyl radicals were produced in the OFR by the photochemical reactions: Ozone was again produced by irradiating pure air (Aadco 737 Pure Air Generator) with ultraviolet emission from a mercury lamp (λ = 185 nm).Ozone concentrations used in the photo-oxidation studies were 30 ppm (Table 2).Toluene (Sigma-Aldrich, 99 %) was injected and vaporized in a flask, and the vapors were swept into the OFR by purified air.The injected toluene concentrations were 0.2-1.0ppm.The mass concentration of SOM particles during the generation process was determined from measurements of the number-diameter distribution of SOM particles in the flow tube reactor or OFR and assuming a material density of 1200 kg m −3 (Liu et al., 2013).The number-diameter distributions were measured with a scanning mobility particle sizer (SMPS; TSI Inc.).The O : C ratio of the toluene-derived SOM studied here was determined using a high-resolution aerosol mass spectrometer (HR-ToF-AMS; Aerodyne Research Inc.).Data analysis was based on the approach described by Chen et al. (2011).

Production of supermicron SOM particles on hydrophobic substrates
At the outlet of the flow tube reactor and OFR, the submicrometer SOM particles were collected on hydrophobic surfaces.The limonene-derived SOM and toluene-derived SOM particles were collected onto glass slides coated with trichloro(1H,1H,2H,2H-perfluorooctyl)silane (Sigma-Aldrich, 97 %).The coating procedure is described in Knopf (2003).The β-caryophyllene-derived SOM particles were collected onto Teflon substrates.Two different methods were used to collect submicron particles on hydrophobic substrates (see Tables 1 and 2).The first method used was an electrostatic precipitator (TSI 3089, USA).In this case, the resulting SOM particles on the hydrophobic substrates were smaller than ∼ 10 µm.From experience in our laboratory, detection of LLPS with our microscope setup is the clearest when the size of the particles are roughly 20-80 µm.As a result, the following method was used to coagulate the sub-10 µm particles into 20-80 µm particles: first the substrate containing the SOM particles was placed in a RH-controlled flow cell (Parsons et al., 2004;Pant et al., 2006;Song et al., 2012b).The RH in the flow cell was then set to over 100 % for 30-60 min to grow and coagulate the SOM particles.The RH in the flow cell was then decreased to ∼ 80-90 % RH to evaporate the water.During the experiments, the particles were observed using a reflectance microscope (Zeiss Axiotech, 50×).These growth and coagulation processes resulted in SOM particles consisting of 20-80 µm in diameter (Song et al., 2015;Renbaum-Wolff et al., 2016).
In the second method used to collect SOM particles collected on a hydrophobic substrate, a single-stage impactor was used (Prenni et al., 2009;Pöschl et al., 2010;Hosny et al., 2016).In this case, the SOM particles after collection were as big as 100 µm due to coagulation during the collection process.Since the particles were already large enough for the LLPS experiments, they were used directly without  1) and (b) limonene-derived SOM for the mass concentrations of 7000 µg m −3 (limonene 3, Table 1).Note that the light gray circles at the center of the particles are an optical effect due to the hemispherical nature of the particles.Illustrations are shown below the images for clarity.Green: SOM-rich phase.Blue: water-rich phase.The scale bar is 20 µm.
the need for the growth and coagulation experiments described above.Both methods used to collect SOM particles collected both the water-soluble and water-insoluble components of the SOM particles.

Optical microscopy of supermicron SOM particles
For the LLPS experiments, the hydrophobic substrate containing SOM particles with sizes in the range of 20 to 80 µm in diameters was mounted in a temperature and RH controlled flow cell coupled to an optical reflectance microscope (Zeiss Axiotech, 50× objective) (Parsons et al., 2004;Pant et al., 2006;Song et al., 2012b).The temperature of the cell was 290 ± 1 K in all experiments.RH in the cell was regulated by varying the ratio of a dry and humidified N 2 flow.The total flow rate was ∼ 1200 sccm.The RH was determined from measurements of the temperature with a thermocouple and measurements of the dew point/frost point with a chilled mirror sensor (General Eastern, Canada).The RH was calibrated using the deliquescence RH for pure ammonium sulfate particles (80 % RH at 293 K; Martin, 2000).After calibration, the uncertainty of the RH was ±2.0 % based on the reproducibility of multiple deliquesce measurements.At the beginning of LLPS experiments the SOM particles were equilibrated at ∼ 100 % RH for 15 min.Then the RH was reduced from ∼ 100 to ∼ 0 % RH at a rate of 0.1 to 0.5 % RH min −1 , and subsequently increased to ∼ 100 % RH at a rate of 0.1 to 0.5 % RH min −1 .We did not observe a dependence of LLPS on the RH ramp rate, although only a narrow range of rates were used.During the humidity cycle, optical images of the SOM particles were recorded every 5-10 s using a CCD camera.

β-Caryophyllene-derived and limonene-derived SOM particles
Humidity cycles at 290 ± 1 K were performed for βcaryophyllene-derived SOM particles generated with mass concentrations of 15-4000 µg m −3 and limonenederived SOM generated with mass concentrations of 80-7000 µg m −3 (Table 1).In all cases, LLPS was observed at high RH.Table 1 summarizes the results during humidity cycles.Shown in Fig. 1a and Movie S1 (Supplement) are examples of optical images of a β-caryophyllene-derived SOM particle as a function of increasing RH for the particle mass concentrations of 2000-4000 µg m −3 .Shown in Fig. 1b and Movie S2 (Supplement) are examples of optical images of a limonene-derived SOM particle as a function of increasing RH for the particle mass concentrations of 7000 µg m −3 .For both types of SOM particles, only one phase was observed for RH values from 0 to ∼ 90 %.Note the light-colored circle in the center of the particles at 90.5 % RH for β-caryophyllene-derived SOM and at 95.0 % RH for limonene-derived SOM is an optical effect due to the light scattering from a hemispherical particle (Bertram et al., 2011).In Fig. 1, LLPS is observed at 91.5 % RH for the β-caryophyllene-derived SOM particle and at 95.3 % RH for the limonene-derived SOM particle.LLPS began  with the formation of many small inclusions of a second phase, and in both cases the phase transition occurred over a narrow range of RH.The small inclusions coagulated to larger droplets in the β-caryophyllene-derived SOM at 92.5 % RH and in the limonene-derived SOM at 96.1 % RH (Fig. 1a and b).At the highest RH investigated, a core-shell morphology is observed.Such a core-shell morphology on a hydrophobic substrate has been observed previously by Song et al. (2012b) in particles containing organic, ammonium sulfate and water, although a different morphology can result in the absence of the hydrophobic substrate (Reid et al., 2011).After formation of the core-shell morphology consisting of an inner and outer phase, the two liquid phases co-existed as high as ∼ 100 % RH.We assume that the inner phase is a water-rich phase, while the other phase is an organic-rich phase, since the size of the inner phase decreases as the RH decreases (Renbaum-Wolff et al., 2016).
The surface tension of water and the surface tension of more-oxidized and less-oxidized organics is consistent with this assumption (Jasper, 1972).Upon drying, the two liquid phases merge into one liquid phase.This mixing process occurred at 90.9 % RH for β-caryophyllene-derived SOM and 95.6 % RH for limonene-derived SOM.Movies of the mixing process are shown in the Supplement (Movies S3 and S4).Shilling et al. (2009) showed that the O : C of SOM can depend on the particle mass concentration used to generate the SOM.To determine whether the occurrence of LLPS depends on the SOM particle mass concentrations used when generating the SOM, particle mass concentrations ranging from 15 to 7000 µg m −3 were investigated (Table 1).Illustrated in Fig. 2a and b is the RH at which two phases were observed during humidity cycles as a function of the mass concentrations of the β-caryophyllene-derived and limonenederived SOM samples.Triangles represent mixing relative humidity (MRH) of two liquid phases upon drying and circles represents separation relative humidity (SRH) upon moistening.
LLPS was observed at 93.6 ± 1.5 % RH in the βcaryophyllene-derived SOM particles for the particle mass concentrations of 15-4000 µg m −3 (Fig. 2a).In the limonenederived SOM particles, LLPS occurred at 96.1 ± 2.1 % RH for the particle mass concentrations of 80-7000 µg m −3 (Fig. 2b).LLPS occurred at 96.0 ± 0.7 % RH in α-pinenederived SOM particles for the mass concentrations of 75-11000 µg m −3 (Renbaum-Wolff et al., 2016) (Fig. 2c).As shown in Fig. 2, the SRH and MRH of the β-caryophyllene-Table 3. Summary of the LLPS results as well as the average oxygen-to-carbon atomic ratios (O : C) of the studied SOM particles.The standard deviation (σ ) of the separation relative humidity (SRH) and mixing relative humidity (MRH) is derived from several cycles of RH for different SOM mass concentrations.SRH = 0 and MRH = 0 indicate phase separation was not observed.The average O : C values for SOM from toluene photo-oxidation are based on the current study (Sect.2.1).The average O : C values for the other SOM are taken from the literature.Since the average O : C can depend on oxidant time and oxidation conditions, we chose literature data that were closest to the experimental conditions used in the LLPS studies (Supplement, Sect.S1 and Table S1).derived SOM, limonene-derived SOM, and α-pinene-derived SOM particles do not depend strongly on the SOM particle mass concentrations used to generate the SOM.

Toluene-derived SOM
Humidity cycles were also performed for SOM particles generated from photo-oxidation of toluene using particle mass concentrations of 80-1000 µg m −3 in the reactor (Table 2).None of the toluene-derived SOM particles underwent LLPS during RH cycling even at high RH (Table 2).Shown in Fig. 3 and Movie S5 (Supplement) are optical images of a toluene-derived SOM particle for mass concentrations of 80-100 µg m −3 .Images in Fig. 3 and Movie S5 were recorded as the RH was increased.No LLPS was observed in the SOM particle during RH cycling between 0 and 100 %.Rastak et al. (2017) 4. The experimental conditions for the studies reported in (a) were not necessarily similar to the experimental conditions for the studies reported in (b), even if the same precursor volatile organic compound was used.conditions used when studying LLPS (Supplement, Sect.S1 and Table S1).
Based on the data shown in Table 3, there appears to be a relationship between the occurrence of LLPS and the average O : C of the organic material: when the average O : C was between 0.34 and 0.44, LLPS was observed, but when the average O : C was between 0.52 and 1.30, LLPS was not observed.This trend is also apparent in Fig. 4a, where SRH data in Table 3 are plotted versus the O : C data in Table 3.
SOM have an average O : C and a spread (or distribution) of O : C values.Similar to SOM, systems containing two organics and water also have a spread in O : C and an average O : C. Hence, as a starting point to understanding LLPS in SOM, we considered previous studies that explored the miscibility gap in bulk solutions containing two organics and water (see Table 1 in Ganbavale et al., 2015).When the average O : C of the organic material was low in a system containing two organic components with water, LLPS was observed.For example, LLPS was observed in a mixture of 1-butanol (O : C = 0.25), 1-propanol (O : C = 0.20), and water (Gomis-Yagües et al., 1998) and in a mixture of 1-pentanol (O : C = 0.20), acetone (O : C = 0.33), and water (Tiryaki et al., 1994).On the other hand, when the average O : C of the organic material was high in a system containing two organics and water, LLPS was not observed.For example, LLPS was not observed in a mixture of acetic acid (O : C = 1.00), ethanol (O : C = 0.50), and water (Pickering, 1893).We conclude that the relationship between average O : C and LLPS in SOM observed here is not inconsistent with previous LLPS studies in systems containing two organics and water.In addition to the average O : C, the spread in O : C in organic mixtures will also be important for LLPS (Renbaum-Wolff et al., 2016).

Implications
As mentioned in the introduction, Petters et al. (2006), Hodas et al. (2016), Renbaum-Wolff et al. (2016), and Rastak et al. (2017) showed using thermodynamic calculations that SOM particles that undergo LLPS at high RH values have modified CCN properties.Hence, LLPS should be considered when predicting the CCN properties of SOM particles derived from α-pinene ozonolysis, β-caryophyllene ozonolysis, and limonene ozonolysis.A caveat is that the mass concentrations used when generating the SOM particles in our experiments was larger than normally found in the atmosphere (Zhang et al., 2007;Jimenez et al., 2009;Spracklen et al., 2011;Li et al., 2015).Additional studies are needed to confirm LLPS in SOM particles generated using more atmospherically relevant SOM mass concentrations.
Figure 4b suggests that κ is related to the average O : C of the organic material.Figure 4a and b combined suggests that when the average O : C is small, LLPS occurs and the difference between κ HGF and κ CCN is large.On the other hand, when the average O : C is large, LLPS does not occur and the difference between κ HGF and κ CCN is small.Figure 4 provides additional support for the suggestion that the LLPS is related to the discrepancies between κ HGF and κ CCN .

Figure 1 .
Figure 1.Optical images of SOM particles with increasing RH: (a) β-caryophyllene-derived SOM for the mass concentrations of 2000-4000 µg m −3 (β-caryophyllene 3, Table1) and (b) limonene-derived SOM for the mass concentrations of 7000 µg m −3 (limonene 3, Table1).Note that the light gray circles at the center of the particles are an optical effect due to the hemispherical nature of the particles.Illustrations are shown below the images for clarity.Green: SOM-rich phase.Blue: water-rich phase.The scale bar is 20 µm.

Figure 2 .
Figure 2. RH at which two phases were observed during humidity cycles of individual particles of (a) β-caryophyllene-derived SOM and (b) limonene-derived SOM from this study, and (c) α-pinene-derived SOM from Renbaum-Wolff et al. (2016) as a function of the SOM mass concentrations.For all panels, circles represent onset of phase separation upon moistening (i.e., separation relative humidity, SRH) and triangles represent mixing of two liquid phases upon drying (i.e., mixing relative humidity, MRH).Yellow shaded region indicates two phases present and green shaded region indicates one phase prevalent in SOM.SOM mass concentrations were derived from measured number-diameter distributions and assuming a material density of 1200 kg m −3 (Liu et al., 2013).

Figure 3 .
Figure 3. Optical images of toluene-derived SOM for the particle mass concentrations of 80-100 µg m −3 (toluene 2, Table 2) with increasing RH.Illustrations of the images are shown for clarity.Green: SOM-rich phase.Size bar is 20 µm.
did not observe LLPS in isoprene-derived SOM particles for the mass concentrations of 60-1000 µg m −3 3.3 Relation between LLPS and O : C Summarized in Table 3 are the average SRH and MRH values determined in our work and by Renbaum-Wolff et al. (2016) and Rastak et al. (2017).The average SRH and MRH values are based on several cycles of RH for different SOM mass concentrations.Also included in Table 3 are the average O : C values for the studied SOM particles.The average O : C values for SOM from toluene photo-oxidation are based on the current study (Sect.2.1).The average O : C values for the other SOM are taken from the literature.Since the average O : C can depend on oxidant time and oxidation conditions, we chose literature data that were closest to the experimental

Figure 4 .
Figure 4. (a) Separation relative humidity (SRH) as a function of the average O : C of the organic material.Shown are the results from Table 3. β-Caryophyllene-derived SOM (red), limonene-derived SOM (light green), and toluene-derived SOM (cyan) from this study, isoprene-derived SOM (orange) from Rastak et al. (2017) and α-pinene-derived SOM (blue) from Renbaum-Wolff et al. (2016) as a function of O : C. RH value of 0 % indicates that LLPS did not occur.The O : C values of the SOM particles were taken from Table 3.(b) The difference between κ HGF and κ CCN , denoted as κ, as a function of the average O : C of the SOM.Data taken from Table4.The experimental conditions for the studies reported in (a) were not necessarily similar to the experimental conditions for the studies reported in (b), even if the same precursor volatile organic compound was used.
is ±2.0 % RH, due to the uncertainty in the RH measurements.
*Values derived from number-diameter distribution measured by an SMPS and analyzed using a material density of 1200 kg m −3 .

Table 2 .
Experimental conditions for production and collection of SOM produced by photo-oxidation.Included are the measured separation relative humidity (SRH) upon moistening and mixing relative humidity (MRH) upon drying of the collected particles.Particles were collected on hydrophobic substrates using an electrostatic precipitator or single-stage impactor.SRH = 0 and MRH = 0 indicate LLPS was not observed during humidity cycles.Values derived from number-diameter distribution measured by an SMPS and analyzed using a material density of 1200 kg m −3 .

Table 4 .
Literature data of measured average O : C, κ HGF , κ CCN , and the difference between κ HGF and κ CCN , denoted as κ, of SOMs.The average O : C values are based on measurements from the individual studies referenced here.The experimental conditions for the studies reported in Table4were not necessarily similar to the experimental conditions for the studies reported in Table3, even if the same precursor volatile organic compound was used.