The hygroscopic properties of the particles were investigated using a “see-through” inertia impactor apparatus equipped with an optical microscope. The experimental setup is described in detail elsewhere (Ahn et al., 2010). Briefly, the apparatus was composed of three parts: (A) see-through impactor, (B) optical microscope, and (C) humidity controlling system. The Parafilm-M substrate onto which the aerosol particles were deposited was mounted on the impaction plate in the see-through impactor. The RH inside the impactor was controlled by mixing dry and wet (saturated with water vapor) N2 gases. The wet N2 gas was obtained by bubbling through two de-ionized water reservoirs. The flow rates of the dry and wet N2 gases were controlled by mass flow controllers to obtain the desired RH in the range of ∼ 3–90 %, which was monitored by a digital hygrometer (Testo 645). The digital hygrometer was calibrated using a dew-point hygrometer (M2 Plus-RH, GE), providing RH readings of ±0.5 % reproducibility. The particles on the impaction plate were observed through a nozzle throat using an optical microscope (Olympus, BX51M) equipped with 50× / 0.5 numerical aperture objective. The images of particles were recorded by a digital camera (Canon EOS 5D, full frame, Canon EF f/3.5 L macro USM lens) during the humidification (by increasing RH from ∼ 3 to 90 %) and dehydration (by decreasing RH from ∼ 90 to 3 %) experiments. The image size was 5616 × 3744 pixels and the image recording condition was set according to the ISO500. The exposure time was 0.4 s and the depth of focus (DOF) was F/3.5. All the hygroscopic experiments were carried out at room temperature (T= 23 ± 1 ∘C).

The changes in particle size with the variation of RH were monitored by measuring the particle areas in the optical images (Ahn et al., 2010). The particle images were processed using image analysis software (Matrox, Inspector v9.0). The size of the imaging pixel was calibrated using 10 µm Olympus scale bars. Although particles larger than 0.5 µm in 2-D diameter (Dp) could be analyzed using this system, image fields with particles of 1–10 µm were selected because high quality hygroscopic OM data can be obtained from them. Each humidity condition was sustained for at least 2 min. A steady state (or quasi-equilibrium) is reached in much shorter time than 2 min, which can be clearly confirmed by the sudden morphological and/or size change of particles/droplets responding to the RH change and no further morphological and/or size change thereafter. Just to make sure the equilibrium state is attained after the RH change, the optical image of an “ensemble of particles” was recorded after 2 min from each step of RH change.

In situ RMS measurements were carried out under controlled RHs to observe the structural changes in the hydrated water in MgCl2 crystalline solids. The apparatus is composed of three parts: (A) see-through impactor, (B) Raman microscope/spectrometer, and (C) humidity controlling system. The experimental setup is similar to that used for optical microscopy, with the Raman spectrometer replacing the optical microscope; i.e., the particle impaction and RH controlling system are the same as for the stand alone optical microscopy system. The RH was monitored using a digital hygrometer (Testo 635). The digital hygrometer was calibrated using a dew-point hygrometer (M2 Plus-RH, GE), providing RH readings with ±1.5 % reproducibility. The in situ Raman spectra of the particles deposited on Parafilm-M or Al substrates were recorded using a confocal Raman micro-spectrometer (XploRA, Horiba Jobin Yvon) equipped with a 50× / 0.5 numerical aperture objective (Olympus). The optical image size was 908 × 680 pixels. An excitation laser with a wavelength of 532 nm and 3 mW power was used and the scattered Raman signals were detected using an air cooled multichannel charge-coupled device (CCD) detector. The data acquisition time was 10 s for each measurement. The spectral resolution was 1.8 cm-1. The spectra and images were acquired and processed using Labspec6 software.

OM images have a higher spatial resolution (5616 × 3744 pixels) than the RMS optical images (908 × 680 pixels), so that they can be processed more accurately for monitoring the 2-D size changes and the two measurements were performed separately. Unlike the closely monitored OM images, which were recorded in RH steps of ∼ 0.3 % to generate the 2-D area growth curves (Ahn et al., 2010), the Raman spectra of the individual particles were recorded in RH steps of ∼ 3.0 % (∼ 10 min for an ensemble of particles at each RH) to detect the phase changes and complement the OM data.

During the humidification process, NaCl–MgCl2 particles with XNaCl≥ 0.3 exhibit two-stage deliquescence transitions, as is generally expected for binary electrolytic mixture particles. Figure 3 shows the 2-D area ratio plot for the humidification and dehydration processes as a function of the RH for a NaCl-rich particle of XNaCl= 0.9, which is similar to seawater (XNaCl= 0.89) (Lide, 2002), together with the optical images shown at the transition points. Initially at RH = 3.1 %, the optical image of the particle clearly shows its angular crystalline nature. During the humidification process, the particle size remains constant until RH = ∼ 14 %, where a slight decrease in size is observed due to water adsorption at the lattice imperfections of the solid salts in the particle and/or structural rearrangement inside the crystal lattice (Mikhailov et al., 2004; Ahn et al., 2010). A first deliquescence transition was observed from RH = 15.1 to 15.8 %, where the particle morphology changed somewhat at RH = 15.1 %, and the size increased noticeably at RH = 15.8 %. With further increases in RH, the partially aqueous particle gradually grew until RH = ∼ 70.0 %, after which a sharp size increase was noted. The optical image at RH = 73.9 % revealed a solid inclusion, which completely dissolved to form a homogeneous droplet, with the second transition completing at RH = 74.1 %. Subsequently, with further increases in RH, the aqueous droplet undergoes continuous hygroscopic growth, as shown for the larger droplet at RH = 84.5 %. The first phase transition at RH = 15.8 %, i.e., MDRH of the NaCl–MgCl2 system, is assigned to the deliquescence of the MgCl2-dominant eutonic solids in the particle. The observed MDRH deviates from the calculated value of RH = 36.1 % according to the AIOMFAC model, whereas it is comparable to the observed DRH (i.e., 15.9 (±0.3) %) of crystalline MgCl2 ⚫ 4H2O for nebulized pure MgCl2. At the MDRH, the particle consists of a mixed phase of liquid droplets (eutonic solution) and a NaCl solid inclusion, which is clearly visible in the optical image of the droplet at RH = 73.9 %, just before the complete dissolution of pure NaCl at the second DRH of 74.1 %. The second DRH is consistent with the DRH of 74.1 % observed for a nascent SSA particle levitated in an electro-dynamic balance (EDB) by Tang et al. (1997).

The other NaCl-rich particles with compositions of XNaCl≥ 0.3 (e.g., XNaCl= 0.5 in Fig. 4a) also exhibited two-stage phase transitions during the humidification process: the first transition at MDRH (RH = 15.9 (±0.2) %) due to deliquescence of the eutonic component and the second transition due to complete deliquescence of the particles. In the equimolar mixture particles (XNaCl= 0.5, Fig. 4a), the observed second DRH of 67.6 % (on average, 67.3 (±0.4) %, in terms of reproducibility when the humidification and dehydration processes were repeated) is consistent with the theoretical DRH of 67.5 % calculated using the AIOMFAC model. For an equimolar mixing ratio, Chan et al. (2000) estimated a MDRH of 32.0 % and a second DRH of 70 % from the phase diagram of the bulk NaCl–MgCl2–H2O system. For the NaCl-rich particles with compositions of XNaCl≥ 0.3, the MDRH is independent of the compositions. On the other hand, the second DRHs were dependent on the compositions and shift toward a pure NaCl limit (DRH = 75.1 (±0.5) %) with increasing NaCl mole fraction, as observed for NaCl–KCl and NaCl–NaNO3 mixture particles (Li et al., 2014; Gupta et al., 2015).

During the dehydration process, the representative NaCl-rich particle of XNaCl= 0.9 (Fig. 3) showed a two-stage phase transition. The liquid droplet decreased gradually in size with decreasing RH and became supersaturated with respect to NaCl below RH of 74.1 % (DRH for XNaCl= 0.9), where the droplet still appeared to be homogeneous (as shown in the optical image at RH = 71.2 %). With the further decreases in RH, the droplet size decreased abruptly at RH = 45.9–45.6 % due to the crystallization of NaCl in the droplet. At RH = 45.6 %, the first ERH for XNaCl= 0.9, the particle was composed of a mixed phase of the eutonic solution and NaCl solids, as observed clearly from the crystal segments in the optical image. With further decreases in RH, the mixed phase droplet shrank gradually until RH = 5.3–5.1 %, with the MgCl2-dominant eutonic component in the particle crystallized. At RH = 5.1 % (MERH), a completely effloresced solid particle was formed, which is clearly seen by the reappearance of an overall angular shape and a bright crystalline segment in the optical image (similar to that at RH = 3.1 %, at the start of the humidification process). The measured first ERH and MERH for the particles with a composition of XNaCl= 0.9 vary among the particles, 1–10 µm size, in the same optical image field and are in the range of RH = 48.2–45.4 % and RH = 5.6–5.0 %, respectively.

During the dehydration process, the other NaCl-rich particles with compositions of XNaCl≥ 0.3 also exhibited two-stage phase transitions. For example, for a particle with XNaCl= 0.5 (Fig. 4a), the first transition due to the efflorescence of pure NaCl occurs at RH = 35.5–35.0 % and the second due to the efflorescence of the eutonic component at RH = 5.7–5.5 %. Chan et al. (2000) reported the first ERH of 38 % (within the range of the first ERHs of 40.5–35.0 % observed in this study for 38 particles), but no MERH (in this work, it was observed in the RH range of 6.9–5.0 %), when they performed the dehydration experiments on a single levitated aqueous droplet of NaCl–MgCl2 with an equimolar mixing ratio (XNaCl= 0.5) in an EDB. The first ERHs for the NaCl-rich mixture particles with compositions of XNaCl≥ 0.3 are dependent on the compositions and shift toward the pure NaCl limit (ERH = 47.6–45.7 %) with increasing NaCl mole fraction, as observed for NaCl–KCl and NaCl–NaNO3 mixture particles (Li et al., 2014; Gupta et al., 2015). On the other hand, the MERH, like their MDRH, is almost independent of the particle compositions. The MERHs, however, were observed over a wide RH range (RH = 10.4–2.9 %) due to the stochastic nature of nucleation leading to efflorescence.

In the case of NaCl-rich particles with compositions of XNaCl= 0.1 and 0.2, three types of particles showing 2- or 3-stage deliquescence transitions were observed. Figure 5a–c show the hygroscopic behavior of representative “type A, B, and C” particles, where their particle/droplet optical images are also shown at each transition point. During the humidification process, a type A particle of XNaCl= 0.2 (Fig. 5a) showed an initial small decrease in size due to structural rearrangement before the first deliquescence transition at RH = 15.5–15.8 %. At its MDRH of 15.8 %, bright solids in the particle disappeared due to dissolution of the eutonic component. With further increases in RH, it gradually grew until RH = 56.7 %, after which an abrupt increase in size was observed with the dissolution of pure NaCl, i.e., the second deliquescence transition completing at RH of 57.0 % (resulting in a homogeneous droplet in the optical image).

For type B particles, three deliquescence transitions and one intermediate efflorescence transition were observed during the humidification process (e.g., Fig. 5b). Figure 5b shows that the particle size remained constant until RH = ∼ 15.5 %, and then underwent an abrupt first mutual deliquescence transition at RH of 15.5–15.8 % (called MDRH1). The particle grew gradually until the RH reached 21.9 %. Subsequently, an efflorescence-like transition with a decrease in size and a bright, crystallized solid particle were observed at RHs of 21.9–24.9 % and 24.9 %, respectively. Thereafter, the size of the type B particle remained almost constant until RH = 31.0 %, which was followed by a small decrease in size (structural rearrangement) at RH = 32.0 % and then by a second mutual deliquescence transition at RH = 32.3–33.0 % (MDRH2). The observed MDRH2 of 33.0 % for the type B particle is attributed to the complete dissolution of the crystalline MgCl2 ⚫ 6H2O-dominant eutonic part. With further increases in RH, the size of the type B particle increased gradually until the third deliquescence transition with the dissolution of pure NaCl (NaCl solid inclusion in the optical image at RH = 56.7 % disappears in the optical image at RH = 57.0 %), which occurs at RH of 56.7–57.0 %, similar to the type A particle (Fig. 5a). The type C particle (Fig. 5c) did not show the first deliquescence transition at RH of ∼ 15.8 % (seen in Fig. 5a and b) or recrystallization (seen in Fig. 5b), but its size remained unchanged until RH = 31.0 %, followed by a small decrease in size due to the structural rearrangement at RH = 32.0 %, and a sharp increase in size from RH = 32.3–33.0 %, which is the first deliquescence transition or MDRH. The second deliquescence transition was observed as usual at RH = 56.7–57.0 %.

During the dehydration process, the NaCl-rich particles of XNaCl= 0.2 of type A–C particles showed two-stage phase transitions (Fig. 5a–c). The liquid droplets decreased gradually in size with decreasing RH and became supersaturated with NaCl below RH = 57.0 % (DRH for XNaCl= 0.2). With further decreases in RH, the droplet sizes decreased noticeably at RH = 24.1–23.9 % for type A particles (Fig. 5a), RH = 25.1–24.9 % for type B particles (Fig. 5b), and RH = 24.9–23.9 % for type C particles (Fig. 5c), due to the crystallization of NaCl. At the first ERHs of 23.9, 24.9, and 23.9 %, the partially aqueous type A–C particles, respectively, are composed of a mixed phase of a eutonic solution and NaCl solid. With further decreases in RH, type A and B droplets gradually shrank until the final perceptible decreases in sizes were observed at RH = 6.1–5.2 % (Fig. 5a) and at RH = 5.4–5.2 % (Fig. 5b), where completely effloresced type A and B solid particles, respectively, were formed. Their low MERH of 5.2 % suggests that the MgCl2 ⚫ 4H2O-dominant eutonic component was crystallized for the type A and B particles. For the type C particle, complete efflorescence occurred at RH = 23.9–23.7 % with a MERH of 23.7 % (Fig. 5c). The high MERH suggests that the MgCl2-dominant eutonic part crystallized heterogeneously as the MgCl2 ⚫ 6H2O moiety on crystalline NaCl seeds, leading to the observed MDRH of ∼ 33.0 % for the type C particle.

To clearly explain the above observations, particles with a composition of XNaCl= 0.2 sitting on a Parafilm-M substrate were examined using in situ RMS. Figure 6 shows the Raman spectra and corresponding optical images, which were recorded during the in situ RMS measurements of particles of XNaCl= 0.2. As shown in Fig. 6a, at RH = 4.5 %, the type A particle appears irregular and shows a typical OH-stretching signature for crystalline MgCl2 ⚫ 4H2O. At RH = 16.6 %, the type A particle becomes round-shaped and the broad OH-stretching peak for free water indicates that it is deliquesced. During the dehydration process, as shown in Fig. 6b, the type A round droplet at RH = 10.7 % becomes irregular-shaped, i.e., effloresces at RH = 7.6 %, forming a crystalline MgCl2 ⚫ 4H2O-dominant eutonic phase. For type A particles, the NaCl and MgCl2 ⚫ 4H2O-dominant eutonic solids are dissolved and formed during the humidification and dehydration processes.

As shown in Fig. 6c, during the humidification process, the type B particle appears irregular at low RH (i.e., 4.6 %) and shows a OH-stretching signature for crystalline MgCl2 ⚫ 4H2O. At RH = 16.4 %, the particle becomes round and the broad OH-stretching signal for free water indicates that it is deliquesced. With a small increase in RH (to ∼ 16.6 %), however, the type B particle changes its morphology and shows the typical OH vibration signatures for crystalline MgCl2 ⚫ 6H2O, indicating the occurrence of efflorescence. At RH = 33.9 %, the type B particle appears round and shows a free water OH peak again, indicating that further deliquescence occurs below RH = 33.9 %. During the dehydration process, as shown in Fig. 6d, the type B round droplet at RH = 15.3 % becomes irregular-shaped, i.e., effloresces at RH = 10.1 %, forming a crystalline MgCl2 ⚫ 4H2O-dominant eutonic phase. For type B particles, NaCl and MgCl2 ⚫ 4H2O-dominant eutonic solids are formed during the dehydration process, and despite the MgCl2 ⚫ 4H2O-dominant eutonic solids being dissolved, the MgCl2 ⚫ 6H2O-dominant eutonic solids are formed through efflorescence and are dissolved, and finally, the NaCl solids deliquesce during the humidification process. Efflorescence of laboratory-generated particles during humidification has not been reported previously.

As shown in Fig. 6f, at RH = 4.5 %, a type C particle appears irregular and crystalline and shows the typical OH-stretching signature for crystalline MgCl2 ⚫ 6H2O. At RH = 34.1 %, the type C particle appears round and shows a free water OH peak, indicating that deliquescence occurs below RH = 34.1 %. As shown in Fig. 6e, the type C particle effloresced at a higher ERH of 16.8 % and shows a typical OH-stretching signature for crystalline MgCl2.6H2O. For the type C particles, NaCl and MgCl2 ⚫ 6H2O-containing crystals are formed and dissolved during the dehydration and humidification processes, as clearly confirmed by the OM and in situ RMS measurements. On the other hand, this type C particle does not form MgCl2 ⚫ 6H2O again at the next dehydration process, as shown in Fig. 6g, indicating that this particle is no longer type C and the different types of particles are formed somewhat randomly because efflorescence is a kinetic or rate-driven process depending on many factors.

Table 1 shows the encountering frequencies of type A–C particles at various NaCl–MgCl2 mixing ratios in their respective OM experiments. For all mixing ratios except XNaCl= 0.1 and 0.2, type A particles, which form a crystalline MgCl2 ⚫ 4H2O-dominant eutonic phase at MERH, are dominant. Owing to the decreasing availability of condensed water during dehydration, large kinetic barrier and/or diffusional resistance for hydrate-ion nucleation make the formation of the crystalline MgCl2 ⚫ 6H2O structure difficult, which is already discussed for the nebulized pure MgCl2 aerosols.

For particles of XNaCl= 0.1 and 0.2, the encountering frequencies of the type B particles are 29 and 45 %, respectively, and type C are 47 and 18 %, respectively, indicating that the formation of MgCl2 ⚫ 6H2O-containing crystals is easier at these mixing ratios. Aqueous moieties in particles were reported to effloresce more easily by heterogeneous nucleation in the presence of seeds (Schlenker and Martin, 2005; Li et al., 2014; Gupta et al., 2015). For type C particles, the crystallization of MgCl2 ⚫ 6H2O takes place at an ERH range of 23.7–11.9 % during dehydration (e.g., see Fig. 5c), suggesting that the NaCl crystals of an optimal size (XNaCl= 0.1 and 0.2) can act as seeds for the heterogeneous nucleation of MgCl2 ⚫ 6H2O. Heterogeneous nucleation of (NH4)2SO4 in presence of optimally sized kaolinite seeds was also reported by Pant et al (2006). On the other hand, the heterogeneous efflorescence of MgCl2 ⚫ 6H2O is also observed during the humidification process for type B particles (Fig. 5b). Efflorescence during humidification was once reported for Amazonian rain forest aerosols (Pöhlker et al., 2014), where it was claimed that the impacted ambient organic–inorganic mixed aerosols initially had amorphous or poly-crystalline structures and underwent restructuring through kinetic water and ion mobilization, resulting in the crystallization of inorganic salts during hydration. During the humidification process, the MgCl2 ⚫ 4H2O-dominant eutonic solids in the effloresced type B particles of XNaCl= 0.1 and 0.2 dissolve at MDRH1 (i.e., 15.9 (±0.3) %) and some moisture is also adsorbed on the surface of the crystalline NaCl moiety. The [Mg(6H2O)]2+ hydrate-ions appear to be mobilized in the presence of sufficient condensed water at the observed ERH range of ∼ 16.1–24.9 % during hydration (∼ 23.7–11.9 % during dehydration) and the kinetic barrier is overcome by heterogeneous nucleation on the optimally sized NaCl seeds, leading to structural growth and crystallization of the MgCl2 ⚫ 6H2O moiety. Therefore, particles of XNaCl = 0.1 and 0.2 frequently show a characteristic MDRH of ∼ 33.0 (±0.5) % for the dissolution of the MgCl2 ⚫ 6H2O-dominant eutonic solids. On the other hand, the crystalline NaCl seeds for the particles of XNaCl≥ 0.3, being larger than the optimal size, appear to inhibit the smooth diffusion of water kinetically, which is a primary requirement for the nucleation of [Mg(6H2O)]2+. Therefore, for particles of XNaCl≥ 0.3, crystalline MgCl2 ⚫ 6H2O is not formed in the timescale of the measurements.

In the case of particles of XNaCl= 0.05, two-stage deliquescence transitions for type A particles were mainly observed: mutual deliquescence occurring at MDRH of 16.0 (±0.3) % and the second DRH for the remnant NaCl at RH of 40.5 (±1.5) %. For XNaCl= 0.05, the frequency with which particles of types B and C were encountered was much lower (Table 1), most likely because the NaCl seeds were smaller than the optimal size.

For the NaCl-rich particles of XNaCl= 0.05, 0.1, and 0.2, the second DRHs were dependent on the compositions and shifted toward the pure NaCl limit (DRH = 75.1 (±0.5) %) with increasing NaCl mole fraction. Figure 7 plots the measured DRHs for the NaCl-rich particles with various compositions as a function of the NaCl mole fraction, showing that the experimental second DRH values are in good agreement with the values calculated from the AIOMFAC model. On the other hand, the observed MDRH (i.e., 15.9 (±0.3) %) for type A particles deviates from that predicted by the thermodynamic considerations in the model. In contrast, the observed MDRH2 for the type B particles and MDRH for the type C particle of 33.0 (±0.5) % agree with the predicted DRH for MgCl2 ⚫ 6H2O.

In the case of XNaCl= 0.2, the measured first ERH and MERH vary among type A and B particles, 1–10 µm in size, on the same optical image field and are in the range of RH = 27.1–23.9 % (average first ERH = 25.0 %) and RH = 10.2–5.2 % (average MERH = 7.1 %), respectively. In contrast, the type C particles exhibit MERH in the range of 23.7–15.3 % (average MERH = 18.1 %). For type A and B particles of XNaCl= 0.1, the measured first ERH and MERH vary in the range of RH = 17.4–12.4 % (average first ERH = 15.8 %) and RH = 9.4–6.5 % (average MERH = 7.2 %), respectively, whereas the type C particles exhibit MERH in the range of 15.2–11.9 % (average MERH = 13.3 %). In the case of XNaCl= 0.05, the measured first ERH and MERH vary in the range of RH = 15.5–13.0 % (average first ERH = 13.2 %) and RH = 9.3–6.5 % (average MERH = 6.9 %), respectively. Figure 8 shows the measured ERHs for the NaCl-rich particles with various compositions as a function of the NaCl mole fraction.

Figure 4b and c present the 2-D area ratio plots for the MgCl2-dominant particles with the eutonic composition (XNaCl= 0.026) and of XNaCl= 0.01 as a function of RH. During the humidification process, the particles of XNaCl= 0.026 and 0.01 show single-stage deliquescence phase transitions from RH = 15.7 to 16.1 % and from RH = 15.6 to 16.1 %, respectively. With further increases in RH, the particles showed continuous and gradual hygroscopic growth. During the dehydration process, the particles of XNaCl= 0.026 and 0.01 decreased gradually in size until they exhibited hysteresis with single-stage efflorescence transitions from RH = 5.3 to 4.5 % and from RH = 4.8 to 4.6 %, respectively. Therefore, the hygroscopic behavior of these particles with MgCl2-dominant compositions is similar to that of the nebulized pure MgCl2 particles (Fig. 1b).

The probability of forming a crystalline MgCl2 ⚫ 6H2O structure for particles of XNaCl= 0.026 and 0.01 is quite low (Table 1), due to the low probability of heterogeneous nucleation in the absence of optimally sized seeds, as explained above for the case of XNaCl= 0.05. NaCl, which is mostly in the aqueous phase, cannot act as a crystalline seeds for heterogeneous nucleation at low RHs, or the sizes of those NaCl seeds are too small for heterogeneous nucleation leading to the formation of a hexahydrate (MgCl2 ⚫ 6H2O) structure.