Water Adsorption vs Phase Transition of Aerosols Monitored by a Quartz Crystal Microbalance.

A quartz crystal microbalance (QCM) with a high sensitivity of 0.1 ng was applied to monitor the oscillation frequency variation (Δf) of standard single species, two-component systems with typical ambient aerosol compositions, and ambient aerosol filter samples as a function of relative humidity (RH) and determine their deliquescence RH (DRH) and phase transition. Δf is associated with the adsorption or desorption process of water molecules for solid samples and physical properties of the sample film during solid-to-aqueous phase transition (deliquescence). During the pre-deliquescence stage, the water adsorption process led to the increased mass with decreasing Δf, especially for the hydrates such as MgCl2 and Ca(NO3)2, which have more than 20% and 40% increased mass, respectively. The water adsorption process might cause a mass deviation of ambient particulate matter measurement using similar instrument principles. During the deliquescence stage, the observed rapid increasing Δf with RH was caused by a significant change in the physical properties (such as density and viscosity) of the sample film. The determined DRH for a given single-component system is consistent with the results estimated from the thermodynamic models. For a complex system, the QCM can determine the DRH1st well as the eutonic point and track the possible variation of the physical properties of inorganic or with organic acid mixture systems. During the post-deliquescence stage, the gradual increasing trend of Δf with RH for Ca(NO3)2 and an external mixture of NaCl-Ca(NO3)2 was mainly contributed by a stronger RH dependent of physical properties for Ca(NO3)2(aq). Overall, this study provides the possible physical properties variation of common aerosol composition as a function of RH, which was consistent with the results calculated from the thermodynamic models. The stronger water adsorption for MgCl2 and Ca(NO3)2 with solid-like viscosity at RH < DRH might lead to different chemical reactivities in the atmospheric chemistry in addition to the radiative forcing of aerosols caused by the hysteresis.

Introduction

Aerosols play an important role in the climate system via scattering or absorbing radiation directly or acting as cloud condensation nuclei (CCN) to modify the cloud albedo or lifetime and to affect the radiation balance indirectly.[1−4] Thus, the efficiency of aerosols affecting the Earth’s radiative forcing is controlled by the composition and the ability to uptake water. The physical phase of aerosols can affect the impact of aerosols on the atmosphere radiation budget and chemical processes. The hysteresis process during solid-to-aqueous aerosol phase transition can directly affect the uncertainty of the aerosol physical phase as the ambient relative humidity (RH) is in the range from efflorescence RH (ERH) to deliquescence RH (DRH). Model simulation has shown that the hysteresis process of solid-aqueous ammonium sulfate aerosol transition can result in 20% uncertainty in the sulfate direct climate forcing.[5] The Intergovernmental Panel on Climate Change (IPCC) reported that a significant uncertainty of radiative forcing is due to the contribution of aerosol direct effect (ADE).[6] Xing et al. estimated the health impact of ADE in PM2.5-related mortality in East Asia, North America, Europe, resulted from the enhancement of PM2.5 concentration in near-surface through atmospheric dynamics, is 3–6 times higher than reduced mortality from decreased from temperature due to ADE.[7] Therefore, it is crucial to investigate the uncertainty from the physical state (i.e., solid, liquid, or aqueous) of aerosol particles and hysteresis effect from solid-to-aqueous aerosol transition.

The physical state of aerosols can lead to different chemical reactivities for reactions occurring either on the surface or inside of aerosols to affect the formation of secondary aerosol or aerosol number concentration.[8] The physical phases of aerosols are highly related to its composition and the exposed RH history (hysteresis effect). Generally, the soluble components of aerosols include organic acids and soluble inorganic species such as ammonium sulfate, ammonium nitrate, sodium chloride, and sodium nitrates.[9] Several laboratory studies have used a variety of aerosol hygroscopicity measurement techniques (e.g., Fourier transform infrared spectroscopy (FTIR), quartz crystal microbalance (QCM), fluorescence spectroscopy, Raman spectroscopy, electrodynamics balance (EDB), optical microscopy (OM), and hygroscopicity tandem differential mobility analyzer (HTDMA)) to examine the phase transitions of various aerosol compositions.[10−17] Choi and Chan investigated the organic species absorb water and alter the hygroscopicity of inorganic aerosols using an EDB.[11] Fong et al. applied FTIR to monitor the liquid water content as a function of RH based on the absorption of the water stretching band at 3400–3600 cm–1 to determine the phase transition.[12] The DRH could be determined when the absorption peak intensity varies significantly with the phase transition. Optical microscopy can be applied to estimate the ratio of the cross-sectional areas between dry particle and wet particle and then determine the DRH where the ratio and the shape change rapidly.[13] This technique can also go with other detection methods such as Raman or electron microscopy (e.g., scanning electron microscopy (SEM) or transmission electron microscopy (TEM)) to provide the spectral information or detail morphology and elemental distribution.[18] In thermodynamics, the DRH of a well-known composition can be estimated as the Gibbs free energy (GFE) favors the aqueous phase if the physical properties of the ionic species are available. Some thermodynamic models in atmospheric chemistry, such as Aerosol Inorganic–Organic Mixture Functional groups Activity Coefficients (AIOMFAC),[19] the Extended Aerosol Inorganics Model (E-AIM),[20] and ISORROPIA,[21] are usually applied to determine the portion of aqueous and solid phases for a given common atmospheric aerosol composition based on GFE for different phases with water activities. Despite the good agreement between the DRH estimated from thermodynamic models and FTIR or QCM measurement from previous studies, some discrepancies between the morphological features determined the DRH using an electron microscope with Raman spectra technique and a previous study using attenuated total reflection (ATR)-FTIR remain questionable, in particular for Ca(NO3)2. Liu et al. determined that the deliquescence of amorphous Ca(NO3)2 began at ∼7% RH and transformed to solution droplets above 10% RH using micro-Raman spectrometry,[22] and Laskin et al. observed the morphological changes at 8% RH using ESEM.[23] These discrepancies could be due to the unique property of applied samples or the limited sensitivity and low signal-to-noise of water envelope in Raman spectra under low RH conditions.

A quartz crystal microbalance (QCM) is highly sensitive to mass and physical properties variation of adhered species.[15,24,25] For a solid-phase sample, the frequency would have a negative response of oscillation frequency with increasing mass. As the phase transition of sample changes from solid to aqueous phase, the QCM would show a positive frequency response for this transition.[15] Previous QCM measurements have shown that the QCM can measure the DRH of organic acids and inorganic salts and mixtures of organic acids with ammonium sulfate[14] and can determine the mass labilities of films of secondary organic material from QCM measurements of the evaporation rate and vapor mass concentrations.[26] Schuttlefield et al. reported the quantification of monolayer of water uptake by atmospherically relevant particles including organic films, SiO2, and clay mineral samples using coupling QCM with ATR-FTIR.[16,17] The oscillation frequency variation (Δf) of QCM response is associated with the adsorption or desorption process of water molecules for solid samples and physical properties of a sample film during the solid-to-aqueous phase transition stage. For the liquid sample, Δf is related to the liquid density and viscosity on the liquid film of the crystal surface (Kanazawa equation, eq ):[25]where f0 is the fundamental oscillation frequency of the bare crystal, ηl and ρl are the absolute viscosity and density of the liquid, respectively, and μq and ρq are the elastic modulus and density of the quartz. Δf is increased as the viscosity and density of liquid are lowered.

To investigate the interaction between water vapor and ambient aerosols with complex composition, this study applied the QCM to monitor the responses of single species, two-component mixtures, including ammonia sulfate, sodium chloride, calcium nitrate, and succinic acid, and ambient aerosol filter samples collected using a Micro-Orifice Uniform Deposit Impactor (MOUDI) sampler in Kinmen, Taiwan. The DRHs determined from the QCM response variation as a function of RH were compared with the model simulation estimated from thermodynamic models. Furthermore, the QCM response of filter samples was monitored to investigate the phase transition of ambient aerosols and their hygroscopic processes with possible uncertainty discussion.

Results and Discussion

Phase Transition of a Single Component

Figure shows the frequency difference, Δf, and RH variation with time for the quartz crystal with and without ammonium sulfate during the whole hygroscopic process. Without ammonium sulfate, the Δf of bare crystal gradually decreased as the RH increases, likely due to the adsorption of water on the surface of the quartz crystal. The major influence of the water adsorption process tends to increase the mass on the crystal, and that is different from the possible contribution of liquid physical properties such as viscosity and density, which tends to increase the frequency. The maximal frequency variation was −12 Hz at RH of ∼90%, corresponding to 16 ng of water adsorption. For the ammonium sulfate system, Δf as a function of RH had three stages of variation for deliquescence: (1) pre-deliquescence, (2) deliquescence, and (3) post-deliquescence, while (4–6) are for the efflorescence process. During the pre-deliquescence stage, Δf gradually decreased as RH increased, similar to the trend observed for the bare crystal but higher decreasing rate, likely due to the stronger water adsorption on ammonium sulfate than that on the bare crystal. The maximal decrease of Δf was −20 Hz before reaching the phase transition point, corresponding to 27 ng of adsorption, and that showed a more significant variation as compared with that of the bare quartz crystal. For the monolayer of water vapor adsorbed on the gold plated electrode surface, the mass change was estimated to be 4.8 ng of water with a diameter of 4 × 10–10 m for a water molecule. The overall water adsorption during the pre-deliquescence stage was more than monolayer and varied with the applied chemical species due to the interaction of water vapor with the surface. During the deliquescence stage, Δf showed a rapidly increasing trend as RH was approaching the deliquescence point (∼80%), which leads to a significant change in the physical properties of the sample film to increase the oscillation frequency instead of decreasing frequency with the addition of water mass.[15,25] During the post-deliquescence stage, Δf showed an increasing trend with RH and might reach a stable value under the applied condition.

Figure 1

Measured RH (blue line), Δf data of bare crystal (green line), and (NH4)2SO4 (red line) as a function of time. The gray color was applied to separate different stages as described in Section with 1–3 for the deliquescence process and 4–6 for the efflorescence process.

In the RH decreasing cycle where the efflorescence occurred, Δf showed a gradually decreasing trend first, then a sudden drop phenomenon, and a slightly increasing trend afterward. The Δf decreasing trend in the first two stages (4 and 5 in Figure ) of the efflorescence process was likely due to the physical properties variation. As the RH decreased, the water evaporation from the aqueous phase (NH4)2SO4 can lead to a higher density and induce a sudden viscosity enhancement over the aqueous (NH4)2SO4 to reduce Δf as the phase transition of liquid-to-solid phase (efflorescence) occurs. The decreasing Δf trend as RH is approaching 0% (6 in Figure ) was mainly determined by the evaporation of adsorbed water to reduce the overall mass.

To provide a semi-quantitative analysis, the measured ΔfR at a given RH condition was normalized to the ΔfS value caused by the deposited dry sample (RH < 1%) to obtain a normalized frequency, ΔfN, with a starting value at −100% relative to a bare crystal using the following equation:

The corresponding ΔfN was calculated for the time frame and plotted as a function of RH as shown in Figure A for (NH4)2SO4, NaCl, Ca(NO3)2, MgCl2, and succinic acid during the increasing RH cycle. The maximum decreased ΔfN prior to deliquescence was ∼30% for MgCl2, 45% for Ca(NO3)2, 10% for (NH4)2SO4, <5% for NaCl, and 5% for succinic acid. Before reaching the DRH, Ca(NO3)2 and MgCl2 had significantly higher water adsorption than NaCl, (NH4)2SO4, and succinic acid likely due to the hydrate formation in calcium nitrate and magnesium chloride (tetrahydrate and hexahydrate, respectively) or stronger water adsorption on the surface of the hydrates. The more significant water adsorption for (NH4)2SO4 in this study as compared with other study (10% vs ∼0%)[14] might be due to the much smaller quantity of chemicals applied (∼100 ng vs 10 μg). Nevertheless, the hygroscopic inorganic salts deliquesced at RH < 90% while non-hygroscopic succinic acid showed a limited variation with RH in this study likely due to the fact that the DRH is higher than the upper limit of RH, 90%. An additional test was applied to the LiCl sample and showed a fast Δf increasing at RH ∼11% (Figure S1) to confirm that the phase transition can be determined at a low RH.

Figure 2

(A) ΔfN as a function of RH for (NH4)2SO4, NaCl, Ca(NO3)2, MgCl2, and succinic acid. (B) Derivative ΔfN to RH as a function of RH.

As d(ΔfN)/|dRH| > 0, the response of ΔfN was associated with the phase transition and physical property variation. Using (NH4)2SO4 as an example, the DRH for a given sample was determined at d(ΔfN)/|dRH| ≥ 1 as criteria for the phase transition. The DRH based on this criteria was determined at 80 ± 1%, 75 ± 2%, 52 ± 3%, and 37 ± 1% for (NH4)2SO4, NaCl, Ca(NO3)2, and MgCl2, respectively, in good agreement with the reported values from other experimental results and the thermodynamics models from AIOMFAC and E-AIM (Table ).[19,20] The determined DRH of Ca(NO3)2 was significantly higher than that reported by different techniques, <10%,[22,23] but agreed with the 54 ± 2% DRH using ATR-FTIR reported by Schuttlefield et al.[17] However, the absorbed water molecules might affect the morphology or the spectral variation, but the physical properties, such as viscosity and density, for the salts with adsorbed water behave more like a solid phase.

Table 1

Comparison of DRH for a Single Component

sample	(NH4)2SO4	NaCl	Ca(NO3)2	MgCl2
DRH (AIOMFAC)	79.9%	75.6%	51.8%	37.6%
DRH (E-AIM III)	79.9%	75.3%
DRH (this study)	80 ± 1%	75 ± 2%	52 ± 3%	37 ± 1%
DRH from the literature	79.9 ± 1% (Martin, 2000[3] Gysel et al., 2002[27])	75.0 ± 0.7% (Martin, 2000[3] Gysel et al., 2002[27])	<10% (Liu et al. 2008[22] Laskin et al.[23])	32.8±0.2% (Greenspan, 1977[28])
			54 ± 2% (Schuttlefield et al. 2007[17])

To discuss the adsorption of water vapor, the Brunauer–Emmett–Teller (BET) isotherm was applied to describe the dynamic equilibrium between the vapor and the surface of the uppermost molecules in adsorbed stacks at the saturation ratio of adsorbate less than 0.3.[29,30] However, there were some species such as MgCl2 and Ca(NO3)2, having significantly higher water adsorption (more than five layers at RH = 30%), which might be caused by the stronger interaction of water molecules other than van der Waals force with the surface.[30−32] The adsorbed film has properties similar to the liquid to lead to a strong interaction with water vapor and usually occurs at a high saturation ratio, higher than 0.4.[30] Thus, the results suggest that the surface of MgCl2 and Ca(NO3)2 has a liquid-like interaction with water vapor at a low saturation ratio, while the physical phase of the film is more toward solid-like without physical property change. The adsorption of water molecules on the surface mainly increased the overall mass and caused a negligible influence on the physical properties of the chemical species.

Phase Transition of a Mixture

The mixtures applied in this study include internal mixtures of NaCl-MgCl2, NaCl-Ca(NO3)2, and (NH4)2SO4-succinic acid and an external mixture of NaCl/Ca(NO3)2 in a mole ratio of 1:1. Figure shows ΔfN and d(ΔfN)/|dRH| of the NaCl-MgCl2 mixture as a function of RH as compared with the relative mass of mixture estimated from AIOMFAC. The decreased ΔfN before reaching the DRH1st at 36 ± 1% for the NaCl-MgCl2 mixture is likely due to significant water adsorption of MgCl2. The decreased ΔfN was less than that of pure MgCl2 to reflect the water adsorption process by part of the mixture, mainly by MgCl2. DRH1st was slightly less than the 37 ± 1% DRH of MgCl2 and was contributed by the eutonic composition of the NaCl-MgCl2 mixture with residual NaCl crystal. The estimated DRH1st is also consistent with the DRH1st estimated from AIOMFAC at 36.7%. As the RH kept increasing, part of the residual NaCl crystal dissolved into the aqueous phase to maintain the thermodynamic equilibrium. The response of ΔfN before reaching DRH2nd showed a gradually increasing trend, which might be contributed by the decreasing viscosity and the decreased quantity of NaCl(s). The DRH2nd based on the ΔfN variation was estimated at 71 ± 2%, slightly higher than that estimated from AIOMFAC at 67.8%. The ΔfN during the decreasing RH process also clearly showed the efflorescence process likely associated with the hysteresis effect, which confirms the observed two DRH transitions. A summary of DRH of the internal mixture salts from experimental studies and thermodynamic models is summarized in Table .

Figure 3

(A) Measured ΔfN and the relative mass derived from the thermodynamic model as a function of RH for the NaCl-MgCl2 mixture (NaCl mole fraction = 0.5). (B) Derivative ΔfN to RH as a function of RH.

Table 2

Comparison of DRH for Mixtures

sample (mole ratio)	NaCl+MgCl2 (1:1)	NaCl+Ca(NO3)2 (1:1)	(NH4)2SO4+succinic acid (1:1)
DRH (AIOMFAC)	DRH1st: 36.7%	DRH1st: 47.0%	74.1%
	DRH2nd: 67.8%	DRH2nd: 59.3%
DRH (this study)	DRH1st: 36 ± 1%	52 ± 2%	70 ± 3%
	DRH2nd: 71 ± 2%
DRH from the literature	DRH1st: 15.9 ± 0.3%		77.5–79% (measured by EBD) [Choi and Chan, 2002][11]
	DRH2nd: 56.7–57.0% (measured by OM) [Gupta et al., 2015][18]		>40% (measured by FTIR) [Miñambres et al., 2013][33]

For the internal mixture of NaCl-Ca(NO3)2, the ΔfN profile as a function of RH was mainly dominated by the influence of Ca(NO3)2 with strong water adsorption at RH < 50% followed by a rapid transition up to RH ∼75% as shown in Figure . Similar to the NaCl-MgCl2, the decreased ΔfN was less than that of pure Ca(NO3)2 to reflect the adsorption water associated with the fraction of a material having stronger ability of water adsorption. The ΔfN variation could only reveal one DRH at 52 ± 2%, while the AIOMFAC model estimated two DRHs at 47% and 59% for DRH1st and DRH2nd, respectively. Although the two DRHs were not distinguishable by the QCM, the predicted DRH2nd by AIOMFAC was near the RH corresponding to the maximum d(ΔfN)/|dRH| at 62% (Figure B). ΔfN variation in the RH decreasing process can also show the hysteresis effect to support the observed DRH. For the external mixture of NaCl and Ca(NO3)2 as shown in Figure , ΔfN variation was able to differentiate the contribution of NaCl and Ca(NO3)2, respectively, to show the influence of the water adsorption ability and the individual DRH consistent with the variation of individual species. At RH decreasing loop, the efflorescence process also has a noticeable hysteresis effect. As the RH returned to 5%, ΔfN was shifted to −85% different from the initial −100%. This ΔfN difference at RH < 5% before and after the RH cycle sometimes occurred for NaCl samples and might be caused by the crystallization process of NaCl forming grain size or contact surface different from that at the previous cycle. This frequency shift did not affect the determination of DRH. With this comparison of the internal and external mixture of NaCl and Ca(NO3)2, the ΔfN response of the internal mixture at RH > DRH1st was contributed by the gradual dissolution of NaCl with possible viscosity variation of the mixture solution, different from that of the pure individual species.

Figure 4

(A) ΔfN as a function of RH for the NaCl-Ca(NO3)2 internal mixture (NaCl mole fraction = 0.5). (B) Derivative ΔfN to RH and the relative mass derived from the thermodynamic model as a function of RH for the NaCl-Ca(NO3)2 internal mixture.

Figure 5

(A) ΔfN as a function of RH for the NaCl-Ca(NO3)2 external mixture (NaCl mole fraction = 0.5). (B) Derivative ΔfN to RH as a function of RH for the NaCl-Ca(NO3)2 external mixture.

For the internal mixture of (NH4)2SO4-succinic acid, ΔfN had a similar decreasing level as that for succinic acid before the RH reached ∼70% as shown in Figure S2, suggesting that the surface properties of the mixture are dominated by succinic acid to control the adsorption of water vapor. As RH kept increasing, the ΔfN started to show an increasing trend at RH = 70 ± 3%, lower than the 80 ± 1% DRH of (NH4)2SO4. The AIOMFAC model estimated the DRH of the (NH4)2SO4-succinic acid mixture at 74%, which is close to the upper end of the observed DRH but different from other studies, 77.5–79% measured by EBD[11] and higher than 40% measured by FTIR,[33] as summarized in Table . The observed DRH might be controlled by the eutonic point for a mixture of (NH4)2SO4 and (NH4)3H(SO4)2 due to the possible proton contribution from succinic acid. Using the E-AIM model, the substitution of a small quantity of NH4+ with H+ for (NH4)2SO4, even at 0.1%, can have DRH1st at 69% followed by the gradual dissolution of residual (NH4)2SO4(s) and a complete dissolution at RH = 79%. The previously reported DRH for such (NH4)2SO4-succinic acid mixture has a higher range from >40% to ∼79% likely due to the response and sensitivity of different detection methods. Furthermore, it is suspected that the acidity contributed from succinic acid can cause the particles to partially deliquesce at a lower RH to affect the physical properties and reactivity of the particles. In the comparison of ΔfN between the stable post-deliquescence and RH < 5%, ΔfN increment for the (NH4)2SO4-succinic acid mixture is ∼50% of that for pure (NH4)2SO4. The lower ΔfN increment for the (NH4)2SO4-succinic acid mixture might be caused by the influence of remaining solid succinic acid in the aqueous phase on the overall frequency response. Therefore, the ΔfN increment is associated with the fraction of soluble species, which is also reported by Arenas et al.[14]

Application to Ambient Aerosol Filter Samples

Figure shows the ΔfN and d(ΔfN)/|dRH| as a function of RH for two ambient aerosol filter samples. The ΔfN of sample_A with a larger particle size range of 10.0–17.7 μm had a more significant water adsorption trend than that of sample_B with a smaller particle size range of 0.32–0.56 μm. The ΔfN variation suggests the DRH at 52 ± 2% and 68 ± 2% for sample_A and sample_B, respectively. The IC analysis showed that the major water-soluble ions were Na+, Cl–, Ca2+, and NO3– for sample_A and NH4+, NO3–, and SO42– for sample_B as shown in Figure S3. With similar electric equivalents of (Na+ vs Cl–) and (Ca2+ vs NO3–), the main constituent salts for sample_A might be the NaCl-Ca(NO3)2 mixture at a molar ratio of ∼1:0.7. The decreased ΔfN before deliquescence was contributed by Ca(NO3)2, similar to the mixture salt as discussed in Section . For sample_B, it is likely to be the (NH4)2SO4-NH4NO3 mixture with a molar ratio of 3:2. From thermodynamic models, the deliquescence was estimated as DRH1st at 47.0% and DRH2nd at 63.0% for sample_A, while DRH1st at 65.5% and DRH2nd at 76.0% for sample _B as shown in Figure . The QCM could only resolve one DRH, which was located between DRH1st and DRH2nd. The overall trend suggests that the water adsorption process was significant for the larger mode particles with one apparent DRH for the ambient aerosol samples. Similar to the internal mixture sample discussed in Section , the ΔfN response of ambient samples at DRH1st < RH < DRH2nd was contributed by the gradual dissolution of the second species combining a possible viscosity variation of the mixture solution. For the mixture composition, the viscosity was relatively RH-dependent as compared with a pure species. The physical properties variation might affect the uptake of gas species and the efficiency of aqueous chemical reactions.[9,34]

Figure 6

(A) ΔfN as a function of RH for Kinmen ambient aerosol filter samples. (B) Derivative ΔfN to RH as a function of RH for Kinmen ambient aerosol filter samples.

Figure 7

Relative mass derived from the thermodynamic model as a function of RH for NaCl-Ca(NO3)2 at a molar ratio of 1:0.7 (blue line) and (NH4)2SO4-NH4NO3 at a molar ratio of 3:2 (red line).

Due to the significant water adsorption observed on ambient aerosol samples, it is vital to examine the possible issue on the current ambient particular matter measurement instrument, such as tapered element oscillating microbalance (TEOM), which has a similar technique principle as QCM. If the ambient aerosols are composed of significant hydrates such as Ca(NO3)2 or MgCl2, the strong water adsorption might cause a possible overestimation of particular matter mass using a TEOM instrument. Our result showed that 15% of decreased ΔfN for sample_A at RH = 35–40% was due to the increase in water adsorption mass, which would be likely a 15% overestimation of PM mass assuming a TEOM instrument was applied. The significant water adsorption for Ca(NO3)2 or MgCl2 below the DRH might also cause possible discrepancy of DRH determination if the technique is solely based on the increased mass or varied morphology.

Conclusions and Atmospheric Implications

In this study, a high sensitivity QCM was applied to investigate water adsorption and phase transition as a function of RH variation for a single component, mixtures, and ambient aerosol filter samples. During the pre-deliquescence stage, the water adsorption process led to the increased mass with decreasing oscillation frequency. Even though there was significant water uptake on the sample film, especially for MgCl2 and Ca(NO3)2 hydrates, the chemical species remained as a solid phase without a change in physical properties (density or viscosity), which disagrees with that previously reported as more liquid-like morphology of aerosols.[22] During the deliquescence stage, the density and viscosity variation of the sample film from solid-to-aqueous phase transition increased the oscillation frequency. The determined DRH for single-component systems is consistent with the results estimated from the thermodynamic models, AIOMFAC and E-AIM. During the post-deliquescence stage, ΔfN approached a quasi-steady-state condition with the completion of increasing water mass and decreasing the viscosity coefficient as the RH increased. The gradual increasing trend of ΔfN for Ca(NO3)2 and the external mixture of NaCl-Ca(NO3)2 suggests a stronger RH-dependent physical properties for Ca(NO3)2(aq). The physical properties can then affect the aqueous phase reactions by modifying the diffusion efficiency of reactant molecules in the aqueous phase. For the ambient samples, there was strong water adsorption before reaching deliquescence if the composition was composed of hydrates species (Ca(NO3)2 for the studied case) consistent with the mixture results. This water adsorption might affect the mass, spectral absorption, or morphology, but the viscosity of material remained as solid-like. Overall, this QCM study provides the physical phase variation as the RH varied for inorganic salts, mixture compounds, and ambient aerosol, which was consistent with the results calculated from the thermodynamic models. The stronger water adsorption for MgCl2 and Ca(NO3)2 at RH < DRH with solid-like viscosity might lead to different chemical reactivities in the atmospheric chemistry in addition to the uncertainty on radiative forcing of aerosols due to hysteresis.

Experiment and Simulation Setup

Figure shows a schematic of the experimental setup composed of two digital mass flow controllers (MFCs), a time-resolved quartz crystal microbalance (QCM, CHI421, CH Instrument, Inc.), and a thermo-hygrometer (HH311, Omega Engineering Inc.). The relative humidity (RH) was controlled by adjusting the dry-to-wet flow ratio using two MFCs with a total flow rate of 50 mL min–1. Dry air was provided directly from the air supply, while wet air was prepared with flow passing through a bottle partially filled with water. The RH was recorded in the range of 5 < RH < 90%. The flow mixture was then passed through an acrylic flow cell (∼ 0.2 cm3) attachment with a quartz crystal coated with a gold plated electrode (area of 0.196 cm2). The exiting flow was then went through a chamber (∼ 8 cm3) having a thermo-hygrometer to monitor the RH and temperature. The temperature variation within the experiment was less than 2 °C and had a negligible influence (less than 2 Hz fluctuation) on the DRH determination. Both oscillation frequency and RH were recorded every second. The RH value was averaged when the RH reached a plateau for each stepwise RH.

Figure 8

Schematic diagram illustrating the experimental setup for the phase transition study of the substrate using a QCM with two mass flow controllers.

QCM Responses

To calculate the oscillation frequency of QCM responses, the Sauerbrey equation was applied:[24]where Δf is the normalized frequency change (Hz), Δm is the mass change (g), A is the crystal area, ρq is the density of quartz (ρq = 2.648 g/cm3), and μq is the shear modulus of quartz for the AT-cut crystal (μq = 2.947 × 1011 g/cm·s–2). The QCM controller in this study has a fundamental frequency (f0) of 7.995 MHz, with the sensitivity factor at a net change of −1 Hz corresponding to 1.34 ng of mass adsorbed based on eq .

Sample Preparation

The chemical species applied in this study included ammonium sulfate (≥99%, Riedel-de Haën), sodium chloride (99.7%, J. T. Baker), magnesium chloride (magnesium chloride 6-hydrate, 100.6%, J. T. Baker), calcium nitrate (calcium nitrate tetrahydrate, 99.0–103.0%, Alfa Aesar), succinic acid (≥ 99.0%, Sigma-Aldrich), and lithium chloride (99.7%, J. T. Baker) with a purity greater than 99% and were used directly without further treatment. The samples were prepared at a concentration of 0.050 g L–1 by dissolving the chemicals into MilliQ water (18.2 MΩ·cm). A 2 μL solution of the selected solution was pipetted on the gold plated electrode surface of the quartz crystal to form a droplet of ∼1 mm radius and dried using a dry air flow (RH < 1%) at 50 mL min–1 for ∼30 min to reach a constant Δf, the difference from that of the bare quartz crystal, which reflects the mass of dry samples on the surface as ΔfS in eq . The two-component internal mixture salt solution was prepared at a molar ratio of 1:1 solution. For the external mixture salt (NaCl/Ca(NO3)2 case), two droplets containing the desired single species, NaCl and Ca(NO3)2, respectively, were separately deposited on the electrode surface of quartz without mixing.

Ambient Aerosol Filter Samples

The ambient aerosol filter samples in this study were collected in Kinmen (24.41°N, 118.29°E), Taiwan, on April 12, 2015, in cooperation with a project for aerosol observation. The Micro-Orifice Uniform Deposit Impactor (MOUDI, 110R, MSP Corporation, Shoreview, Minnesota, USA) with 47 mm Teflon filters (P/N R2PL047, PALL Corporation, Ann Arbor, Michigan) as an impaction substrate was used in this study for classifying the particle size of aerosols according to aerodynamic size by inertial impaction in several multi-nozzle stages stacked in series. The MOUDI consists of 10 impaction stages with cut-point diameters (in ambient conditions) of 17.7, 10.0, 5.6, 3.2, 1.8, 1.0, 0.56, 0.32, 0.18, 0.1, and 0.056 μm, respectively. Two filter samples with a particle size range of 10.0–17.7 μm and 0.32–0.56 μm were selected to differentiate the QCM response of the coarse mode and accumulation mode of particles, respectively. The ion chromatography (IC) analysis for the filter samples was performed in the Research Center for Environmental Changes of Academia Sinica to provide the soluble composition information, which was also adopted in the thermodynamic models to evaluate the phase transition.

RH Control

The flow cell with and without sample was exposed to flow at RH < 5% and RH of ∼90% cycles three times to determine the response of phase transition before performing a finer adjustment to determine the DRH of the applied composition. The RH was controlled by adjusting the flow ratio between dry and moist flows with the following concept. The RH was set up with four stages with an increasing rate of ∼5% every 3 min as the RH was increased from <5% to ∼60% as shown in Figure . The increasing rate was then reduced to ∼2% every 3 min from ∼60 to ∼90% to provide a finer resolution. After the system reached the maximum RH, the dry-to-wet flow ratio was increased to control the RH at approximately −2% every 3 min from ∼90 to ∼40% and approximately −5% every 3 min from ∼40 to <5%. The RH was changed stepwise to allow a few minutes for the equilibrium. The oscillation frequency variation was averaged as the RH reached a plateau for each stepwise RH. Due to the hysteresis effect, the DRH can be determined from the increasing RH loop, while the efflorescence relative humidity (ERH) was from the decreasing RH loop. The ERH was influenced partially by the heterogeneous crystallization process with a higher value than that received purely from the homogeneous process and not discussed in this study.

Thermodynamic Models for Phase Transition

Two thermodynamic models, Aerosol Inorganic–Organic Mixture Functional groups Activity Coefficients (AIOMFAC)[19] and the Extended Aerosol Inorganics Model (E-AIM),[20,35] were applied to estimate the phase diagram of single or mixture composition. The mass as a function of RH is usually represented as “relative mass”, normalized to the mass in dry conditions. The DRH of a chemical species is derived from the solubility limit caused by the thermodynamic equilibrium between the solid and aqueous solution phases. Therefore, the DRH of a single-component species is the water activity of the aqueous solution when the system reaches the thermodynamic equilibrium. The ion activity product (IAP) calculated by the activity coefficients of the ions from the chemical species would also stay as a constant even with the addition of a second species to the system. There are usually two stages of the deliquescence process in a two-component system. The first stage deliquescence process is caused by dissolving both two species in a fixed ratio, so the DRH1st was a constant and called the eutonic point. The second deliquescence stage is the complete dissolution of the rest of the solid-phase species, and the DRH2nd would vary with the mixing ratio of the chemical species. Therefore, the DRH of two-component species in different mole fractions can be estimated using the AIOMFAC or E-AIM model. In this study, the E-AIM model provides only the DRH of ammonium sulfate and sodium chloride, while other compositions were determined using AIOMFAC.