Six mother solutions were used in this study to generate aerosol droplets which were grouped into two types, OA/NaCl and 1,2,6-hexanetriol (HEX)/ ( N H 4 ) 2 S O 4 . The solutions were prepared by double-distilled water (18.2 MΩ cm; Milli-Q; MilliporeSigma, Burlington, MA, United States). Taking account of that sulfates are one of the major inorganics in ambient aerosols, we adjusted the pH of the solutions by adding different amounts of sulfate acid (guaranteed reagent, Xilong Scientific Co., Ltd., Guangdong, China). Solute NaCl and ( N H 4 ) 2 S O 4 were purchased from Tong Guang Fine Chemical Co. (Beijing, China). Solute OA (analytical reagent) was purchased from Xilong Scientific Co., Ltd. (Guangdong, China); HEX was purchased from Shanghai Macklin Biochemical Co.,Ltd. (Shanghai, China). Relevant information of the mother solutions is listed in Table 1. The pH of mother solutions was measured with a pH meter (Mettler Toledo Instruments Co., Ltd., Shanghai, China). All volume fractions of organics in the mother solutions used in this work were 10%. The organic-to-inorganic mass ratios (OIR) of Solution HEX-I ∼IV are all 2/1.

A schematic of the laser tweezers Raman spectroscopy (LTRS) system is shown in Figure 1. Details of the single particle technique used here can be seen in our previous work [30,31]. In brief, we generated the aerosol droplets of desired chemical composition by a medical nebulizer (Mint PN100) and used a self-developed laser tweezers to trap a single levitated droplet (4 ∼10 μm in radius). The droplet was trapped in a tailored chamber where the relative humidity (RH) could be regulated at a maximum rate of 5% RH/min by mixing dry and humidified nitrogen gas flow; the value of RH was monitored with a RH probe (HC2A-S, ROTRONIC). A laser beam with a wavelength of 532 nm (Excelsior-532-200, Spectra Physics) was used to both trap the aerosol droplet and excite its Raman signal, which was recorded by a spectrograph (SpectaPro 2300i, Acton) equipped with a liquid nitrogen cooled CCD (Spec-10, Princeton Instruments). We observed the LLPS of the single levitated droplet, which was generated from different mother solutions, during a dehumidifying process and determined the SRH according to the readouts of the RH probe.

The disturbance of the RH-controlled nitrogen flow was discussed in the Supplemental Material (Section S1) by calculating relevant dimensionless numbers including the Weber number and the Ohnesorge number. The results indicate that compared with the gravity, inertia and viscosity of the droplet, the surface tension is the dominant factor that impacts the morphology of the droplet. Thus, under the effect of self-surface tension, the single levitated droplet can be considered to be a stationary sphere after achieving a balance with the ambient atmosphere (see an image of the trapped droplet shown in Supplementary Figure S3 in the Supplemental Material).

The phase separation in substrate-deposited droplets has been directly observed by bright-field imaging or scanning electron microscopy [17–19,32]. However, for levitated droplets, the defocus of the trapped droplets blurs the direct imaging. Instead, the time-resolved Raman spectra of the trapped droplets can be used to detect the LLPS efficiently [13–15,33–36]. The Raman spectra demos from different kinds of microdroplets are shown in Figure 2. The trapped droplet works as an enhancing cavity and will overlap stimulated sharp peaks at wavelengths commensurate with whispering gallery modes (WGMs) on the spontaneous Raman spectra (see Figure 2A). The WGM peaks are extremely sensitive with the morphology of the droplet, indicating that if the droplet undergoes phase separation and becomes nonspherical (e.g., partially engulfed), WGM peaks in the spectra will quench (see Figure 2C). Thus, the moment when WGM peaks start to quench can be determined as the onset of phase separation, particularly for partially engulfed phase separation.

For the other mode of phase separation, the core-shell morphology, the sphericity of the phase separated droplet remains and its Raman spectra still retain WGM peaks (see Figure 2B). Nonetheless, the radial homogeneity is destroyed because of the separation of the hydrophilic core and hydrophobic shell. Thus, fitting the Raman spectra with the Mie scattering model for a homogenous droplet, the fitting errors between the measured and simulated WGM peaks are supposed to increase drastically (i.e. model failure). Therefore, the moment when the homogenous fitting errors start to soar can be determined as the onset of core-shell phase separation. The homogenous Mie scattering fitting model used in this work was developed by Preston et al. [37]. All the Raman spectra used here is normalized by area.

The number of organic species in a single atmospheric particle is on the order of thousands, among which water insoluble organic compounds (WIOCs) take a quite important part. Here, we used oleic acid as a surrogate of WIOC and detected the phase separation of inherently separated aerosols.

Figure 3A shows the time-resolved Raman spectra of the aerosol droplet generated from Solution OA-I, which vividly depicts the progression of phase separation. The brand range of 3100 ∼3600 cm⁻¹ corresponds to the bending and stretching modes of O-H of water, while the range of 2,850 ∼ 2,950 cm⁻¹ corresponds to the C-H mode of organics (OA here). In the beginning, the RH was set at a constant value of 85% for around 650 s to let the freshly trapped droplet reach a stable state. It can be seen that in the beginning (before ∼230 s, marked with a green dashed line in Figure 3), there are no pronounced bright dots in the spectra which indicates that no WGMs have emerged in the freshly trapped droplet. It is because that the droplet was generated by nebulizing the mixture of OA/NaCl/Water and a number of emulsions of OA would certainly be contained in the nascent droplet. The random motion of the inclusions destroys the radial homogeneity and gives rise to the absence of the WGMs (see the schematic diagram in Figure 3A). After this, no more than one bright dot presents in each spectral snapshot which indicates that at most one WGM peak has been detected in the spectra of the droplet. It means that the droplet was heterogenous, in other words, phase-separation with a partially engulfed structure took place. The rather complex but very weak resonant structure in the spectra stems from that the volume of the aqueous inorganic phase may be far greater than that of the hydrophobic organic phase because of the prior stochastic mixing of OA emulsions and water, and an approximately spherical cavity occurs for the aqueous volume (see the schematic diagram in Figure 4A). The imperfect spherical volume is supposed to lead to complexity in the WGM fingerprint and a low enhancement due to the low quality of the WGMs [14].

After adding sulfate acid, the phase separation of OA-II is shown in Supplementary Figure S1 (see the Supplemental Material). It can be seen that after the transient evolution of the nascent droplet, there are still several weak but non-eliminable WGM peaks in the spectral snapshots (100 ∼400 s). It may be explained by that the added sulfate acid worked as a surfactant and reduced the surface tension between air and organic phase and the interface tension between organic and inorganic phase, which then changed the morphology of the phase separated droplet from partially engulfed structure to core-shell structure. However, such core-shell structure was quite fragile and lasted only for a short while. As the RH continued to decrease, the concentration of the inorganics as well as the surface tension between air and inorganic phase increased, which turned the core-shell structure back to partially engulfed structure. This change is shown in the spectral snapshots that as RH decreased, after ∼430 s, the droplet was dehydrated and similar low-quality WGMs emerged. Besides, the interim process from partially-engulfed to core-shell is also illustrated in the time-resolved Raman spectra, where no WGMs were recorded during 400 ∼430 s.

The formation of ambient secondary organic aerosols (SOAs) is inextricable from the generation of organosulfates and nitrooxy-organosulfates [3,38,39], which then contributes to the emergence of various water soluble organic compounds (WSOCs). Most of the WSOCs are humic-like substances, including dicarboxylic acids, polyols, esters, and ethers. We used 1,2,6-hexanetriol as the surrogate of WSOC and detected the phase separation of internally miscible aerosols.

Figure 5 illustrates the phase separations of aerosol droplets generated from aqueous HEX/ ( N H 4 ) 2 S O 4 at different pH levels. For HEX-I (pH = 6.9), from Figure 5A1, we can see that there are numerous WGMs from the beginning of the time-resolved Raman spectra. After the droplet reaches a stable state with the ambient environment, there are still three WGMs in the spectral snapshots, one of which is quite prominent. It means that the droplet could not be partially-engulfed, in other words, the droplet was expected to be either core-shell or homogenous. In Figure 5B1, the fitting errors of the measured WGMs calculated by the homogenous Mie scattering model are on the order of 10^–1 ∼ 10⁰, which are far greater than those from homogenous droplets (compared with Figure 5B2), time = 0–500 s). It also indicates that although the droplet was generated from a homogenous internally miscible solution, it underwent phase separation just after entering the optical well. Thus, the SRH of HEX-I is presumably higher than the initial RH (88%) in the trapping chamber.

For HEX-III (pH = 2.8), Figure 5A2) shows that before 690 s, the droplet was homogenous with five obvious WGMs in its spectra, and the fitting errors in Figure 5B2 are pretty minor. However, as the RH keeps decreasing, the fitting errors surge promptly (note the logarithmic scale), indicating that the phase state of the droplet changed during 690–700 s (marked with a dashed line in the figure). The corresponding RH shown in Figure 5C2 is in the range of 73%–75%, at which the SRH of HEX-III may locate. Previous works found that the mechanism of phase separation in aerosol microdroplet was mainly determined by the OIR of the droplet. Empirical summaries indicate that when 1/1.5 < OIR < 2, the LLPS occurs dominantly via spinodal decomposition [35]. Considering the OIR of HEX-III, the droplet may undergo phase separation by spinodal decomposition and cluster growth and aggregation (see Figure 4B).

For HEX-IV (pH = 1.1), from Figures 5A3–C3, we can see that the SRH of HEX-IV falls within the range of 66%–68%. In the period of 0–2,600 s, the reduction in RH leads to an increase in the Raman intensity from the C-H vibrations and a decrease in the intensity from O-H (this phenomenon is more obvious in Figure 5A1). After that, we elevated the ambient RH; the droplet started to absorb the moisture along with a decrease of C-H mode and an increase of O-H mode. Besides, the convoluted WGMs in the spectra regress to around six main peaks at around 3200 s (marked with a grey dashed line in the figure). It means that the droplet subsequently returned to be homogenous, albeit a hysteresis took place compared with the increase of RH. Moreover, the mixing relative humidity (MRH, the RH level at which phase mixing occurs) can be determined according to the slump of the fitting errors, which is in the range of 72%–75% and mildly higher than the SRH. It is inline with the results of Losey et al. [28]. It also shows that water molecules can go through the hydrophobic shell and be absorbed by the hydrophilic core during the process of water condensation. It might stem from that the molecules of organics are much bigger than the molecules of water so that the latter can transfer through the pores and channels formed in the former.

The SRHs of aerosol droplets generated from Solution HEX-I ∼IV are illustrated in Figure 6. It can be seen that the SRH of HEX decreases as pH decreases, which means that the acid enhances the miscibility of organics/inorganics. Considering the acidity found in ambient aerosol particles, it is conceivable that aerosols would undergo phase separation at lower RH than the values reported in [17,18]. Moreover, the pH-dependent SRHs reported by Losey et al. [28] are also drawn in the same graph. Our results agree well with that of Losey et al. [28] in the pH range of 3 ∼5. The discrepancy occurs when the pH is below 3 and increases in the range of 1 ∼2. It may result from the different ambient conditions of the droplets which are laser-levitation (leading to a morphology of sphere) and substrate-deposition (leading to a morphology of spherical crown) respectively. Besides, the researched droplets in our work were also obviously smaller than that in Losey et al. [28]. Considering the realistic size and levitating state, the droplets used in our work better mimicked the natural airborne aerosols Kucinski et al. [40]. found that the size of micro-particles was influential in the phase separation. The smaller size and the spherical morphology enhance the Kelvin effect, making the droplets in our work easier to undergo phase separation.