Liquid–liquid phase separation and morphologies in organic particles consisting of α-pinene and β-caryophyllene ozonolysis products and mixtures with commercially-available organic compounds

Liquid–liquid phase separation (LLPS) in organic aerosol particles can impact several properties of atmospheric particulate matter, such as cloud condensation nuclei (CCN) properties, optical properties, and gas-to-particle partitioning. Yet, our understanding of LLPS in organic aerosols is far from complete. Here, we report on LLPS of one-component and two-component organic particles consisting of α-pinene-20 and β-caryophyllene-derived ozonolysis products and commercially-available organic compounds of relevance to atmospheric organic particles. In the experiments involving single-component organic particles, LLPS was observed in 8 out of 11 particle types studied. LLPS almost always occurred when the oxygen-to-carbon elemental ratio (O:C) was ≤ 0.44, but did not occur when O:C was > 0.44. The phase separation occurred by spinodal decomposition and nucleation and growth mechanism, and when 25 LLPS occurred, two liquid phases co-existed up to ~100% relative humidity (RH). In the experiments

Many researchers have focused on LLPS in particles containing organic material mixed with inorganic salts. They found that LLPS can occur when the O : C of the organic material is smaller than 0.8 (Bertram et al., 2011;Krieger et al., 2012;Song et al., 2012aSong et al., , 2012bSchill and Tolbert, 2013;You et al., 2013You et al., , 2014. More recently, studies on LLPS in organic aerosol particles free of inorganic salts have shown that LLPS occurs in SOA generated in environmental chambers when the average O : C of the organic material is smaller than roughly 0.5 across the RH range of ∼ 95 % to ∼ 100 % (Renbaum-Wolff et al., 2016;Rastak et al., 2017;Song et al., 2017;Ham et al., 2019) with implications for the CCN properties of the SOA Hodas et al., 2016;Renbaum-Wolff et al., 2016;Ovadnevaite et al., 2017;Rastak et al., 2017;Liu et al., 2018;Ham et al., 2019). In addition, Song et al. (2018) showed that LLPS occurs in organic particles containing one commercially available organic compound when the O : C is smaller than 0.44, while LLPS occurs in organic particles containing two commercially available organic species when the average the O : C is smaller than ≤ 0.58.
In the following, we investigate LLPS in particles containing one and two organic species generated from ozonolysis products of α-pinene and β-caryophyllene, which are atmospherically relevant, and commercially available organic compounds. α-Pinene and β-caryophyllene are the most abundant types of monoterpene (C 10 H 16 ) and sesquiterpenes (C 15 H 24 ) in the atmosphere, respectively (Guenther, 1995;Sakulyanontvittaya et al., 2008;Henrot et al., 2017). However, studies of LLPS and morphologies for α-pinene and β-caryophyllene oxidation products are still rare. Our results can provide additional insight into the O : C range required for LLPS in organic particles free of inorganic salts. Moreover, our results can provide the chemical complexity of organic particle effects on LLPS. These observations should improve our understanding of LLPS behavior and provide more accurate constrained values of the O : C ratio for LLPS.
The results from these studies should also improve the understanding and modeling of CCN activity of SOA free of inorganic salts.  Table 1 presents the physical properties of the organic compounds investigated. In this study, 11 organic species were used, including seven products from the ozonolysis of αpinene and β-caryophyllene and four commercially available organic compounds. These species covered an O : C range of 0.13-1.00 (Table 1). All species were liquid at room temperature.
Seven of the products from the ozonolysis of α-pinene and β-caryophyllene were synthesized. The detailed synthesis methods for these species are described in Bé et al. (2017). Using 1 H NMR, 13 C NMR, and IR spectroscopy, the ozonolysis products were characterized to confirm their identity and purity. The purity of all synthesized compounds was > 95 %. All products contained a carboxylic acid, ketone, and/or aldehyde, which are abundant organic functional groups in the atmosphere (Hallquist et al., 2009;Nozière et al., 2015). The O : C range of the ozonolysis products was between 0.13 and 0.44 (Table 1). To achieve O : C ratios up to 1.00, we used commercially available organic compounds (Sigma-Aldrich, purities ≥ 97 %) ( Table 1).

Preparation of particles consisting of one and two organic species
Particles consisting of either one or two organic compounds were prepared at room temperature without the addition of a solvent. Particles consisting of the commercially available organic compounds were nebulized directly on siliconized hydrophobic glass slides (Hampton Research, Canada). Particles consisting of ozonolysis products were slightly viscous. To form particles on a substrate, these ozonolysis products were picked up with the tip of a pipette, and the pipette was then flicked towards a hydrophobic glass slide. Particles consisting of two organic compounds were prepared using mixtures (1 : 1 mass ratio) of pure organic species without addition of a solvent. To prepare the mixtures with a 1 : 1 mass ratio, each organic species was weighed in a vial and then combined. After mixing, the solutions were homogenous based on visual inspection. Particles were generated from these mixtures and deposited on hydrophobic slides either by nebulization (for the mixtures involving commercially available organic compounds) or by the flicking method via the tip of a pipette as described above (for the ozonolysis products). This method of producing two-component organic particles did not work for α-pinene ozonolysis products and β-caryophyllinic acid due to the stickiness of these materials. Hence, these materials were not included in the systems used to generate two-component organic particles.

Optical microscopy for observation of liquid-liquid phase separation
The organic particles on hydrophobic glass slides were placed into an RH-and temperature-controlled flow cell coupled to an optical microscope (Olympus BX43, 40× objective, Japan) (Parsons et al., 2004;Pant et al., 2006;Bertram et al., 2011;Song et al., 2012aSong et al., , 2018Ham et al., 2019). During all experiments, the temperature inside the flow cell was maintained at 291 ± 1 K. The RH was controlled by a continuous flow of a wet and dry N 2 mixture with a total flow rate of 500 sccm. The temperature and RH were monitored by a humidity and temperature sensor (Sensirion, SHT 71, Switzerland). RH inside the flow cell was calibrated by measuring the deliquescence RH of four different pure inorganic salts (potassium carbonate, sodium chloride, ammonium sul-fate, and potassium nitrate) (Winston and Bates, 1960). The RH uncertainty from the calibration was ±1.5 %. At the beginning of LLPS experiments, organic particles inside the flow cell were equilibrated at ∼ 100 % RH for 15-20 min. If LLPS was observed, the RH was decreased from ∼ 100 % to ∼ 5 %-10 % lower than the RH at which the two liquid phases merged into one phase, followed by an increase in RH to ∼ 100 %. If LLPS was not observed, the RH was decreased from ∼ 100 % to ∼ 0 %, followed by an increase to ∼ 100 %. During all experiments, the RH was adjusted at a rate of 0.1 %-0.2 % RH min −1 . The optical images during experiments were recorded every 5 s using a CMOS (complementary metal-oxide-semiconductor) detector (Di-giRetina 16, Tucsen, China). Organic particles were selected in the diameter range of 30-100 µm, which was required for LLPS experiments. Each organic species was measured four to five times within this size range.

Results and discussion
3.1 Liquid-liquid phase separation in particles containing one organic species A total of 11 different types of particles containing one organic species were investigated for LLPS at 291 ± 1 K. Out of the 11 different types of one-component organic particles studied, eight underwent LLPS during humidity cycles (Table S1 in the Supplement). LLPS occurred in all one-component organic particles containing α-pinene and βcaryophyllene ozonolysis products. Shown in Fig. 1 and Movies S1-S7 in the Supplement are optical images recorded while the RH was decreased for all the cases in which LLPS was observed in one-component organic particles. For these cases, two liquid phases were always observed at ∼ 100 % RH. As the RH was decreased, the two liquid phases merged into one liquid phase at ∼ 95 % RH, except for particles of β-caryophyllinic acid ( Fig. 1e and Movie S5). For β-caryophyllinic acid particles, the two liquid phases merged into one liquid phase at 82.1 % RH ( Fig. 1e and Movie S5).
Shown in Fig. 2 and Movies S8-S14 are optical images of the same seven particles shown in Fig. 1 and Movies S1-S7, except the images were recorded while the RH was increased rather than decreased. At low RH values, the particles contained one phase. As the RH increased, LLPS occurred at ∼ 95 % RH for all cases except for β-caryophyllinic acid particles, which underwent LLPS at 82.9 % RH ( Fig. 2e and Movie S12). At the onset of LLPS, many small inclusions formed in the particles. As the RH was further increased, the small inclusions coagulated and coalesced, and the particles continued to grow ( Fig. 2 and Movies S8-S14). At ∼ 100 % RH, all particles contained two liquid phases. These results for LLPS occurrence in the organic particles were consistent within the studied size ranges (∼ 30-100 µm in diameter).
LLPS in the single-component organic particles occurred by nucleation and growth, as well as spinodal decomposition, with RH increasing. Particles of β-nocaryophyllonic acid and suberic acid monomethyl ester had two liquid phases with inclusions by the nucleation and growth mechanism ( Fig. 2d and Movies S11), while the other single-component organic particles had a core-shell morphology by spinodal decomposition (Fig. 2a, b, c, e, f, g and Movies S8, 9, 10, 12, 13, 14). The nucleation and growth mechanism is a phase transition that has to overcome an energy barrier to form stable nuclei of the second phase within a liquid (Shelby, 1995;Papon et al., 1999;Ciobanu et al., 2009;Song et al., 2012a). Spinodal decomposition is a phase transition that occurs within a liquid without an energy barrier (Shelby, 1995;Papon et al., 1999;Ciobanu et al., 2009;Song et al., 2012a). Previous studies also observed LLPS by spinodal decomposition in α-pinene-derived SOA, β-caryophyllenederived SOA, and limonene-derived SOA (Renbaum-Wolff et al., 2016;Song et al., 2017;Ham et al., 2019). We ex- pect that the inner phase consisted mainly of water, while the outer phase consisted mainly of organic molecules because the amount in the inner phase was reduced in size as the RH was decreased (Renbaum-Wolff et al., 2016;. This assumption has also been reported in several other studies (Renbaum-Wolff et al., 2016;Song et al. 2017Song et al. , 2018Ham et al. 2019). The surface tension of water and the surface tensions of organics are consistent with this assumption (Jasper, 1972). 0.44 (Fig. 3a, gray symbols). Our results are also consistent with LLPS lower and LLPS upper determined for SOA produced from α-pinene and β-caryophyllene (Renbaum-Wolff et al., 2016;Song et al., 2017;Ham et al., 2019). In all cases, LLPS upper was ∼ 100 % RH.
The values of LLPS lower and LLPS upper determined in the experiments using a decreasing RH were within the uncertainty of LLPS lower and LLPS upper values determined in the experiments using an increasing RH (Tables S1 and S2). In addition, no dependence on particle size was observed for LLPS lower and LLPS upper within the size range investigated (30-100 µm). . Relative humidity (RH) for LLPS as a function of the O : C of the organic particle consisting of (a) one organic species and (b) two organic species for increasing RH. Open blue circles are the LLPS upper boundary (LLPS upper ) with increasing RH, and closed blue circles are the LLPS lower boundary (LLPS lower ) with increasing RH. Each data point includes four to five particles within particle size ranges from 30 to 100 µm. The gray diamonds represent the result from Song et al. (2018). Error bars represent 2σ of multiple measurements and the uncertainty from the RH calibration. The solid and dashed lines are Sigmoid-Boltzmann fits to all the data of LLPS lower and LLPS upper .

Liquid-liquid phase separation in particles containing two organic species
To better mimic the complexity of real aerosol compositions, we also studied LLPS in particles containing two organic species. Table S2 lists the 25 different mixtures investigated using combinations of β-caryophyllene ozonolysis products and commercially available organic compounds. In total, 23 out of the 25 two-component organic particle types investigated underwent LLPS ( Fig. 3b and Table S2). Shown in Fig. 4 and Movies S15-S19 are examples of images of two- component organic particles that underwent LLPS during a decrease in RH. Shown in Fig. 5 and Movies S20-S24 are the same five particles, but the images were recorded as the RH was increased. Out of the 23 particle types that underwent LLPS, 22 of the particle types formed a core-shell morphology with decreasing RH. Only one particle type (β-caryophyllonic acidβ-nocaryophyllonic acid) was observed to have both coreshell and partially engulfed morphology with increasing RH (Fig. S1). As discussed in Sect. 3.1, the inner phase is expected to be mainly water, while the outer phase is expected to be mainly organic material (Renbaum-Wolff et al., 2016;Song et al., 2017Song et al., , 2018. As RH was decreased, the two liquid phases merged into one phase. For example, particles of β-caryophyllene aldehyde-PEG-400 merged into one phase at 39.9 % RH ( Fig. 4e and Movie S19).
Interestingly, particles showed different morphologies of core-shell, partially engulfed, and inclusions after LLPS occurred for RH increasing. These different morphologies have also been observed previously (Kwamena et al., 2010;Reid et al., 2011;Song et al., 2012aSong et al., , 2013. The different morphologies could result from the phase separation mechanisms, volume ratios, and different functional groups (dicarboxylic acid vs carboxylic acid and ketone), which can result in different interfacial energies and spreading coefficients (Kwamena et al., 2010;Reid et al., 2011;Song et al., 2013;Stewart et al., 2015;Gorkowski et al., 2019). In the experiments with two-component organic particles and increasing RH, in most cases (19 out of the 23 particle types that underwent LLPS), phase separation began with the abrupt formation of many small inclusions (e.g., Fig. 5a, b, e and Movies S20, 21, 24). This behavior suggests spinodal decomposition as the mechanism for LLPS. In contrast, the mechanism for LLPS in the particles containing ozonolysis products mixed with pyruvic acid was likely the nucleation and growth mechanism based on the appearance of nucleation from the interior of the particles as the RH increased (Figs. 5c,d and Movies S22,23). Illustrated in Fig. 3b (blue symbols) is the lower RH boundary for LLPS (LLPS lower ) and upper RH boundary for LLPS (LLPS upper ) determined in the experiments with two-component organic particles. LLPS was observed in all cases when the average O : C ≤ 0.67. When LLPS was observed, LLPS upper was ∼ 100 % RH. These results are similar to previous results from Song et al. (2018) (gray symbols in Fig. 3b), even though they studied different types of two-component organic particles. Figure 3b also presents Sigmoid-Boltzmann fits of all data points from

Atmospheric implications
The O : C of organic materials has been used to interpret and parameterize hygroscopicity , oxidation (Heald et al., 2010;Kroll et al., 2011), and mixing thermodynamics of organic aerosol particles Hodas et al., 2016). Previous studies have shown that LLPS in mixed organic and inorganic aerosol particles often occurs for O : C < 0.8 (Bertram et al., 2011;Krieger et al., 2012;Song et al., 2012aSong et al., , 2012bSchill and Tolbert, 2013;You et al., 2013You et al., , 2014. Even in the absence of inorganic salts, the occurrence of LLPS was dependent on the O : C of organic materials (Renbaum-Wolff et al., 2016;Song et al., 2017Song et al., , 2018Ham et al., 2019). Our results show that as compositional complexity increased from one organic species to two organic species, LLPS occurred over a wider range of average O : C values of organic materials (increasing from 0.44 to 0.67) (Figs. 3a and b). Considering the chemical complexity and the O : C ratio of organic particles in the troposphere (0.20 < O : C < 1.00) (Zhang et al., 2007;Hallquist et al., 2009;Jimenez et al., 2009;Heald et al., 2010;Ng et al., 2010), our result provided additional evidence that LLPS is likely a common feature of organic aerosols free of inorganic salts. A caveat is that the mixing ratio of 1 : 1 for two organic species and the chemical complexity used in our experiments are rather simple compared to the chemical complexity found in the atmosphere (Zhang et al., 2007;Hallquist et al., 2009;Jimenez et al., 2009). Further studies are needed to confirm LLPS in organic aerosols comprising more complex mixtures with different mixing ratios. In addition to the O : C ratio, the types of organic functional groups present in the molecules are also likely important for LLPS (Song et al., 2012b) because different functional groups lead to different strengths of intermolecular interactions with water. Further studies are needed to elucidate the effect of functional groups on the occurrence of LLPS in organic particles. LLPS in aerosol particles has an impact on heterogeneous reactions (Folkers et al., 2003;Anttila et al., 2006;McNeill et al., 2008;Lam et al., 2019) and CCN Hodas et al., 2016;Renbaum-Wolff et al., 2016;Ovadnevaite et al., 2017;Rastak et al., 2017;Liu et al., 2018). The reactive uptake of gas-phase oxidants can differ depending on the number of phases present in the particles (Folkers et al., 2003;Anttila et al., 2006;McNeill et al., 2008;Lam et al., 2019). For example, the effective OH uptake coefficient of a 3methylglutaric acid-ammonium sulfate particle decreased by a factor of ∼ 2.4 compared to a 3-methylglutaric acid particle (Lam et al., 2019). The occurrence of LLPS in organic aerosol particles at high RH, as observed in the current study, is important since LLPS at high RH can lower the barrier to CCN activation by decreasing the surface tension of the particles Rastak et al., 2017;Liu et al., 2018). A decrease in surface tension and lowering of the bar-rier to CCN can lead to an increase in cloud droplet numbers in the atmosphere, with implications for modeling the indirect effect of aerosols on climate Rastak et al., 2017).
Data availability. Underlying material and related items for this paper are located in the Supplement.
Author contributions. MS, AKB, and RJT designed the study. Y-CS and MS conducted LLPS experiments and analyzed the data. AGB, FMG, and RJT produced ozonolysis products. Y-CS and MS prepared the paper with contributions from AGB, STM, FMG, AKB, and RJT.
Competing interests. The authors declare that they have no conflict of interest.