When exposed to water vapor, particles will absorb/absorb water vapor, leading to depletion of water vapor in the system. The amount of water absorbed/adsorbed by particles can be determined from the measured change in water vapor pressure (if the volume remains constant), and the RH can be calculated from the final water vapor pressure when the equilibrium is reached. The amount of water associated with particles can be determined as a function of RH by varying the initial water vapor pressure.

Commercial instruments, usually designed to measure the Brunauer–Emmett–Teller (BET) surface areas using nitrogen or helium (Torrent et al., 1990), have been utilized to investigate hygroscopic properties of atmospherically relevant particles (Ma et al., 2010b, 2012b; Hung et al., 2015). For example, Ma et al. (2010b) integrated an AUTOSORB-1-C instrument (Quantachrome, US) with a water vapor generator and employed this apparatus to investigate hygroscopic properties of NaCl, NH4NO3 and (NH4)2SO4. The measured DRH values and mass hygroscopic factors were found to agree very well with those reported in the literature (Ma et al., 2010b). This method has proven to be very sensitive; as shown in Fig. 2, change in adsorbed water as small as the < 0.5 monolayer can be reliably quantified (Q. Ma et al., 2013). In addition to CaSO4 and gypsum, this instrument was also employed to investigate hygroscopic properties of fresh and aged Al2O3, MgO and CaCO3 particles (Ma et al., 2012a).

A similar instrument (Micromeritics ASAP 2020) was employed by Hung et al. (2015) to examine the hygroscopicity of black carbon, kaolinite and montmorillonite particles at 301 K, and a sensitivity of sub-monolayers of adsorbed water could be achieved. Assuming a dry particle diameter of 200 nm, the single hygroscopicity parameters, κ, were determined to be ∼0.002 for montmorillonite and < 0.001 for both black carbon and kaolinite (Hung et al., 2015).

This technique is able to quantify particle water content for unsaturated samples and is sensitive enough to measure adsorbed water; however, it cannot be (at least has not been) used to examine supersaturated samples. This technique, which is independent of particle size and morphology, can also be used to investigate hygroscopic properties of ambient aerosol particles in an offline manner. For example, a physisorption analyzer was used to study hygroscopic properties of ambient aerosol particles collected in Beijing during an Asian dust storm, and one monolayer of adsorbed water was formed on these particles at 46 % RH (Ma et al., 2012b).

The katharometer, also known as the thermal conductivity detector, can be used to measure water vapor concentration. Lee and co-workers employed a katharometer to investigate liquid water content of aerosol particles collected on filters (Lee and Hsu, 1998, 2000; Lee and Chang, 2002). In this setup (Lee and Chang, 2002), aerosol particles were collected on a Teflon filter and then equilibrated with a helium flow at a given RH; after the equilibrium was reached, the particle-loaded filter was purged with a dry helium flow, which was subsequently directed to a katharometer to measure the water vapor concentration. As a result, the liquid water content associated with particles at a given RH could be quantified. The performance of this new method was systematically examined (Lee and Hsu, 1998, 2000; Lee and Chang, 2002), and the measured water-to-solute ratios at different RH during both humidification and dehumidification processes were found to agree well with those reported in the literature for several compounds, including NaCl, NH4Cl, Na2SO4, (NH4)2SO4 and NH4NO3.

Mikhailov et al. (2011, 2013) also developed a katharometer-based method to investigate aerosol hygroscopicity. The instrument, called a filter-based differential hygroscopicity analyzer (FDHA), is described elsewhere (Mikhailov et al., 2011), and a brief introduction is provided here. In this apparatus, a humidified helium flow was split into two identical flows which were then passed through a pair of differential measurement cells: the reference cell contained a blank filter, and the sample cell contained a filter laden with particles (typically less than 0.1 mg). The difference in water vapor concentrations in these two cells, caused by absorption/adsorption of water by particles loaded on the filter, was measured using a differential katharometer, and the amount of water taken up by particles could be quantified by integration of the katharometer signals over time. This instrument could measure hygroscopic growth at very high RH (up to 99 %).

Hygroscopic properties of (NH4)2SO4, NaCl, levoglucosan, malonic acid, and mixed (NH4)2SO4–malonic acid particles were examined using FDHA at different RH during humidification and dehumidification (Mikhailov et al., 2013), and the measured mass growth factors agreed well with those reported in the literature. This instrument was further employed to investigate hygroscopic properties of particles collected from a pristine tropical rainforest (near Manaus, Brazil) (Mikhailov et al., 2013), a suburban boreal forest site (near the city of St. Petersburg, Russia) (Mikhailov et al., 2013) and a remote boreal site (the Zotino Tall Tower Observatory, ZOTTO) in Siberia (Mikhailov et al., 2015). Figure 3 displays the measured hygroscopic properties of aerosol particles collected at the ZOTTO site. As shown in Fig. 3, both supermicrometer and submicrometer particles started to uptake a substantial amount of water at ∼70 % RH; nevertheless, efflorescence took place at different RH, with ERH being ∼35 % RH for submicrometer particles and ∼50 % RH for supermicrometer particles (Mikhailov et al., 2015). It was suggested that the observed difference in ERH could be explained by the difference in organic contents in submicrometer and supermicrometer particles (Mikhailov et al., 2015): submicrometer particles contained larger fractions of organic materials, consequently leading to the reduction of ERH.

The katharometer-based technique can be used to determine particle water content for unsaturated and supersaturated samples, independent of particle size and morphology (Lee and Chang, 2002; Mikhailov et al., 2013). It has also been successfully used as an offline method to investigate hygroscopic properties of ambient aerosol particles (Mikhailov et al., 2013, 2015). It remains to be tested whether this technique is sensitive enough to investigate water adsorption of a few monolayers or less.

Knudsen cell reactors are low-pressure reactors widely used to investigate heterogeneous uptake of trace gases (Al-Abadleh and Grassian, 2000; Karagulian and Rossi, 2005; Karagulian et al., 2006; Wagner et al., 2008; Liu et al., 2009; Zhou et al., 2012). This technique was also employed in several studies to explore water adsorption by particles with atmospheric relevance (Rogaski et al., 1997; Seisel et al., 2004, 2005). For example, the initial uptake coefficient was reported to be 0.042±0.007 for uptake of water vapor by Saharan dust at 298 K (Seisel et al., 2004). Another study (Rogaski et al., 1997) found that pretreatment with SO2, HNO3 and H2SO4 could significantly increase water uptake by amorphous carbon. Knudsen cell reactors are normally operated in the molecular flow regime, and thus water vapor pressure used in these experiments is extremely low. As a result, although these measurements can provide mechanistic insights into the interaction of water vapor with particles at the molecular level, limited information on aerosol hygroscopicity under atmospheric conditions can be provided.

In a simple manner, the change in particle mass due to water uptake can be measured using an analytical balance under well-controlled conditions (Hänel, 1976; McInnes et al., 1996; Hitzenberger et al., 1997; Diehl et al., 2001). For example, Diehl et al. (2001) investigated hygroscopic properties of 10 pollen species at room temperature, using an analytical balance housed in a humidification chamber. The masses of pollen samples were measured at 0, (73±4) and (95±2) % RH. The average ratios of the mass of adsorbed water to dry mass increased from around 0.1 at 73 % RH to ∼1 at 95 % RH (Diehl et al., 2001), suggesting that pollen samples can adsorb a substantial amount of water at elevated RH.

Analytical balance was also employed to investigate hygroscopic properties of ambient aerosol particles. McInnes et al. (1996) employed an analytical balance to explore the hygroscopic properties of submicrometer marine aerosol particles collected on filters and found that liquid water accounted for up to 9 % of the dry particle mass at 35 % RH and up to 29 % of the dry particle mass at 47 % RH. In another study (Hitzenberger et al., 1997), size-segregated aerosol particles were collected on aluminum foils using a nine-stage cascade impactor in downtown Vienna, and their hygroscopic properties were examined using an analytic balance. Aerosol hygroscopicity was found to be strongly size dependent (Hitzenberger et al., 1997), and the mass ratios of particles at 90 % RH to those at dry conditions were found to be 2.35–2.6 for particles in the accumulation mode and 1.16–1.33 for those in the coarse mode.

Similarly to humidity-controlled analytical balance, thermogravimetric analyzers (TGAs) can directly measure the mass change in particle samples at different temperatures to investigate aerosol hygroscopicity. Commercial TGA instruments are typically integrated with automated systems for humidity generation and control. They can control temperature and RH very precisely and are very sensitive in mass measurement (typically down to 1 µg or even better).

Thermogravimetric analyzers, sometimes also called vapor sorption analyzers (VSAs), have been employed by several groups to investigate hygroscopic properties of atmospherically relevant particles. For example, water uptake by CaCO3 and Arizona test dust was measured at room temperature using a Mettler-Toledo TGA with an accuracy of 1 µg in mass measurement (Gustafsson et al., 2005), and about four monolayers of adsorbed water were formed at 80 % RH for both mineral dust samples. A similar instrument was utilized to determine the DRH of dicarboxylic acids and their sodium salts at different temperatures (Beyer et al., 2014; Schroeder and Beyer, 2016), and the DRH was found to decrease with temperature for malonic acid, from 80.2 % at 277 K to 69.5 % at 303 K (Beyer et al., 2014). This method was also used to probe water adsorption by different soot particles (Popovitcheva et al., 2001, 2008a, b), although no details of the instrument used were provided. It is worth noting that TGA and/or VSA have been widely used to investigate hygroscopic properties of pharmaceutical materials. For example, at room temperature anhydrous theophylline was observed to transform to hydrate at 62 % RH, and its DRH was determined to be 99 % (Chen et al., 2010).

Very recently, Tang and co-workers systematically evaluated the performance of a vapor sorption analyzer to investigate hygroscopic properties of particles of atmospheric relevance (Gu et al., 2017b). The instrument, with its schematic diagram shown in Fig. 4, has two sample crucibles housed in a temperature- and humidity-regulated chamber, and one crucible is empty so that the background is simultaneously measured and subtracted. DRH values of six compounds, including (NH4)2SO4 and NaCl, were determined at different temperatures (5–30 ∘C) and found to agree well with literature values. In addition, the mass change as a function of RH (up to 90 %), relative to that at 0 % RH, was also found to agree well with those calculated using the E-AIM model (Clegg et al., 1998) for (NH4)2SO4 and NaCl at 5 and 25 ∘C. Therefore, it can be concluded that the vapor sorption analyzer is a reliable technique to study hygroscopic properties of atmospherically relevant particles.

The vapor sorption analyzer was used to examine hygroscopicity of CaSO4⋅2H2O at 25 ∘C (Gu et al., 2017b), and the results are displayed in Fig. 5. The hygroscopicity of CaSO4⋅2H2O was found to be very low, and the sample mass was only increased by < 0.5 % when RH was increased from 0 % to 95 %. This instrument was very sensitive to the change in sample mass due to water uptake; for example, as shown in Fig. 5b, a relative mass change of < 0.025 % within 6 h could be accurately determined. This instrument was further employed to investigate hygroscopic properties of perchlorates (Gu et al., 2017a; Jia et al., 2018), Ca- and Mg-containing salts (Guo et al., 2019), and primary biological particles (Tang et al., 2019), which play significant roles in the environments of the Earth and Mars. To our knowledge, the VSA technique has not yet been used to explore hygroscopic properties of ambient aerosol particles.

It was proposed in 1959 (Sauerbrey, 1959) that a film attached to the electrodes of a piezoelectric quartz resonator would cause a decrease in the resonance frequency, given by Eq. (1): 1Δf=-Cf⋅Δm, where Δf is the change in resonance frequency, Δm is the mass of the film, and Cf is a constant specific to the quartz resonator that can be experimentally calibrated. Equation (1), known as the Sauerbrey equation, forms the basis for using the piezoelectric quartz resonator as a microbalance, which is usually called quartz crystal microbalance (QCM). QCM is a highly sensitive technique for particle mass measurement and could be extended to investigate aerosol hygroscopicity. In a typical experiment, a particle film is first coupled to the quartz crystal, and RH is then varied, with the resonance frequency being simultaneously recorded. According to Eq. (1), change in the mass of the particle film, due to change in RH, is proportional to the change in resonance frequency. Hygroscopicity measurements only need the information of relative mass change (relative to that under dry conditions), and as a result, knowledge of Cf is not required. QCM has a very high sensitivity in mass measurement, and it has been reported that the change in mass on the order of a few percent of a monolayer can be reliably determined (Tsionsky and Gileadi, 1994).

A QCM was used to measure the DRH of a number of inorganic and organic salts, including NaCl, (NH4)2SO4, CH3COONa and CH3COOK (Arenas et al., 2012), and the measured values agreed very well with those reported in previous work. Several studies (Thomas et al., 1999; Demou et al., 2003; Asad et al., 2004; Liu et al., 2016) have utilized QCM to explore hygroscopic properties of organic compounds of atmospheric interest. For example, Demou et al. (2003) quantitatively determined the amount of water taken up by dodecane, 1-octanol, octanoic acid, 1,5-pentanediol, 1,8-octanediol and malonic acid at room temperature. The DRH was measured to be ∼72 % for malonic acid and ∼95 % for 1,8-octanediol, and in general compounds with higher oxidation state showed higher hygroscopicity (Demou et al., 2003). Another study (Asad et al., 2004) found that exposure to O3 would substantially increase the hygroscopicity of oleic acid. Using a QCM, Zuberi et al. (2005) explored the effect of heterogeneous reactions on hygroscopic properties of soot particles. As shown in Fig. 6, while water adsorption was very limited for fresh soot particles, hygroscopicity of soot particles was significantly increased after heterogeneous reactions with OH/O3 and HNO3 (Zuberi et al., 2005).

QCM has also been applied to study hygroscopic properties of mineral dust particles, including oxides (Schuttlefield et al., 2007a), clay minerals (Schuttlefield et al., 2007b; Yeşilbaş and Boily, 2016) and authentic dust samples (Navea et al., 2010; Yeşilbaş and Boily, 2016). For example, Yeşilbaş and Boily (2016) measured the amount of water taken up by 21 different types of mineral particles up to 70 % RH at 25 ∘C and found that particle size played a critical role in water adsorption by these minerals. At 70 % RH, submicrometer-sized particles could adsorb up to ∼5 monolayers of water, while the amount of water adsorbed by micrometer-sized particles can reach several thousand monolayers (Yeşilbaş and Boily, 2016). Another study (Hatch et al., 2008) suggested that ∼3 monolayers of adsorbed water were formed on CaCO3 particles at 78 % RH, and internal mixing with humic and fulvic acids could substantially increase the hygroscopicity of CaCO3.

It should be pointed out (as this is often not fully considered) that a few assumptions are required for the Sauerbrey equation to be valid (Rodahl and Kasemo, 1996), including that (i) the film deposited on the quartz crystal is rigid, i.e., internal friction is negligible; and that (ii) the film is perfectly coupled to the quartz crystal, i.e., there is no slip between the film and the crystal. The Sauerbrey equation may not hold if these conditions are not fulfilled, and the stiffness of the particle film would significantly affect the quartz resonator response (Dybwad, 1985; Pomorska et al., 2010; Vittorias et al., 2010; Arenas et al., 2012). Rodahl and Kasemo (1996) suggested that the Sauerbrey equation can offer reliable mass change measurement only if the film is thin enough and does not slide on the QCM electrode. In addition, as supersaturated films formed on the quartz crystal are unstable, QCM may not be able to explore hygroscopic properties of supersaturated samples.

Piezoelectric bulk wave resonators, which work in a way similar to the QCM, have been used for monitoring aerosol mass concentrations (Thomas et al., 2016; Wasisto et al., 2016). When particles are deposited onto the resonator surface, the resonance frequency will be linearly reduced with the particle mass. Very recently, a new method based on piezoelectric bulk wave resonators was developed to investigate aerosol hygroscopicity (Zielinski et al., 2018). Aerosol particles were first collected on the resonator surface and then exposed to changing RH. Measured DRH and ERH values were found to agree with the literature for NaCl and (NH4)2SO4; in addition, good consistency between experimentally measured and E-AIM predicted hygroscopic growth curves was found for NaCl, (NH4)2SO4 and NaCl/malonic acid mixture (Zielinski et al., 2018). Therefore, this technique appears to be a very promising method for aerosol hygroscopicity measurements.

In addition to the gravimetric method, the beta gauge method is widely used to measure aerosol mass concentrations in a semi-continuous way (Courtney et al., 1982; Chow, 1995; McMurry, 2000; Solomon and Sioutas, 2008; Kulkarni et al., 2011). A beta gauge measures the attenuation of beta particles emitted from a radioactive source through a particle-loaded filter, and if properly calibrated, attenuation of beta particles through the filter can be used to quantify the mass of particles loaded on the filter (McMurry, 2000). The mass of aerosol particles, after being collected on a filter, was measured at different RH in a closed chamber using a beta gauge to determine the aerosol liquid water content (Speer et al., 1997). Laboratory evaluation showed that the liquid water content of (NH4)2SO4 determined using this method agreed well with those measured gravimetrically (Speer et al., 1997), and when compared to humidification, a hysteresis was found during dehumidification for (NH4)2SO4. The ability to observe hysteresis is related to the use of hydrophobic substrate (for example, Teflon is usually a good option) in particle sampling. In addition, the beta gauge method was preliminarily employed to explore hygroscopic properties of submicrometer ambient aerosol particles (Speer et al., 1997). Further tests with other compounds, in addition to (NH4)2SO4, are required to validate the robustness and reliability of this method.

Another widely employed semi-continuous technique for aerosol mass measurement is tapered-element oscillating microbalance (TEOM) (Patashnick and Rupprecht, 1991; Chow et al., 2008; Solomon and Sioutas, 2008; Kulkarni et al., 2011). In a typical TEOM instrument, the wide end of a tapered hollow tube is mounted on a base plate, and its narrow end is coupled to a filter used to collect aerosol particles (Kulkarni et al., 2011). The oscillation frequency of the tapered hollow tube depends on the mass of particles collected on the filter and can be used to measure particle mass if properly calibrated (Kulkarni et al., 2011). Rogers et al. (1998) explored the possibility of using TEOM to measure aerosol liquid water content. Increase in particle mass was observed when a humid particle-free air flow was passed through a particle-loaded filter in the TEOM, and the particle mass started to decrease after a dry particle-free air was introduced (Rogers et al., 1998). This suggested that TEOM had the potential to examine hygroscopic properties of aerosol particles, though further experimental evaluation is needed to assess its performance.

All the techniques discussed in Sect. 3.2 determine particle water content through direct measurement of sample mass or properties that are related to the sample mass, and hence there is no requirement on particle shape. Some of these techniques, such as thermogravimetric analysis (Gustafsson et al., 2005) and quartz crystal microbalance (Schuttlefield et al., 2007a; Yeşilbaş and Boily, 2016), are sensitive enough to investigate water adsorption down to one or a few monolayers, while other techniques, such as the analytic balance, may not be sensitive enough for this application. If particles are supported on proper substrates (such as hydrophobic films), these techniques can be used to investigate hygroscopic properties of supersaturated samples, as demonstrated for the beta gauge method (Speer et al., 1997) and the piezoelectric bulk wave resonators (Zielinski et al., 2018). Nevertheless, supersaturated solutions formed in the majority of these applications may not be stable enough for hygroscopic growth measurements, and as a result measurements have been rarely reported for supersaturated samples. In principle these techniques can all be used offline to investigate ambient aerosol particles if samples with enough mass can be collected. Analytical balance (McInnes et al., 1996; Hitzenberger et al., 1997) and the beta gauge method (Speer et al., 1997) have been used to explore hygroscopic properties of ambient aerosols; to our knowledge, application of thermogravimetric analysis, quartz crystal microbalance, TOEM and piezoelectric bulk wave resonators to ambient samples is yet to be demonstrated.

Optical microscopy was employed to investigate phase transition of atmospheric particles as early as in the 1950s (Twomey, 1953, 1954). In these two studies (Twomey, 1953, 1954), a large number of aerosol particles collected in Sydney were found to deliquesce at 71 % RH–75 % RH, implying that they consisted mainly of sea salt. Since then, optical microscopy has been widely used to study hygroscopic properties of atmospherically relevant particles, and herein we only introduce representative studies conducted in the last 2 decades.

Bertram and co-workers (Parsons et al., 2004a, b, 2006) developed a flow cell-optical microscope apparatus to investigate phase transitions of individual particles deposited on glass slides coated with hydrophobic films. As shown in Fig. 7, the glass slide was placed in a flow cell mounted on a cooling stage for temperature regulation. A dry nitrogen flow was mixed with a humidified nitrogen flow and then delivered into the flow cell through the inlet, and the two flows were regulated using two mass flow controllers to adjust water vapor pressure (and thus RH) in the flow cell. Phase transitions of particles deposited on the glass slide were monitored using a microscope, and particle images were recorded using a CCD camera.

The performance of this apparatus was evaluated by measuring the DRH of (NH4)2SO4 particles from ∼260 to 300 K (Parsons et al., 2004b), and the measured DRH agreed well with those reported in the literature. This setup was then used to investigate the deliquescence of malonic, succinic, glutaric and adipic acid particles from 243 to 293 K (Parsons et al., 2004b) and deliquescence and crystallization of (NH4)2SO4 and NaCl particles internally mixed with organic compounds (Pant et al., 2004; Parsons et al., 2004a). It was found that if (NH4)2SO4 or NaCl particles contained substantial amounts of organic materials, their DRH would be significantly reduced, and these particles were more likely to be aqueous in the troposphere (Pant et al., 2004). A similar instrument was employed to investigate deliquescence and efflorescence of HIO3 and I2O5 particles (Kumar et al., 2010), and the DRH at 293 K was reported to be 81 % for HIO3 and 85 % for I2O5. Li and co-workers employed an optical microscope to investigate hygroscopic properties of individual particles emitted from residential coal burning (Zhang et al., 2018), collected over the Arctic (Chi et al., 2015) and collected during haze events at an urban site in northern China (W. J. Li et al., 2014; Sun et al., 2018). It was found that during hydration urban haze particles typically had core-shell structure at 60 % RH–80 % RH and fully deliquesced at > 80 % RH, while during dehydration most of these particles remained aqueous at > 50 % RH (Sun et al., 2018).

As illustrated by Fig. 8a, besides deliquescence and efflorescence, atmospheric aerosols can also undergo liquid–liquid phase separation (LLPS), leading to coexistence of two liquid phases (Bertram et al., 2011; You et al., 2012, 2014; Freedman, 2017). LLPS can impact the direct and indirect radiative forcing of atmospheric aerosol particles as well as their heterogeneous reactivity, and therefore has received increasing attention in the last several years (You et al., 2012; Freedman, 2017). Optical microscopy has played an important role in understanding LLPS of atmospherically relevant particles (Bertram et al., 2011; You et al., 2012, 2014). Figure 8b shows optical microscopic images of an internally mixed particle during an experiment in which RH was decreased while temperature was kept at ∼291 K (Bertram et al., 2011), and the particle contained (NH4)2SO4 and 1,2,6-trihydroxyhexane with a mass ratio of 1:2.1. As shown in Fig. 8b, at high RH the particle existed as an aqueous droplet, and LLPS happened when RH was decreased, leading to the formation of two liquid phases; efflorescence took place with further decrease in RH, leading to the formation of a solid (NH4)2SO4 core coated with an organic liquid layer.

In addition to identification of phase transitions, analysis of optical microscopic images recorded can also be used to determine particle size change and, as a result, hygroscopic growth factors (Ahn et al., 2010; Eom et al., 2014; Gupta et al., 2015). For instance, Ahn et al. (2010) employed an optical microscope to investigate hygroscopic properties of NaCl, KCl, (NH4)2SO4 and Na2SO4 particles collected on TEM grids and found that their measured hygroscopic growth factors agreed well with those reported in the literature for all four types of particles examined. A following study (Eom et al., 2014) compared the influence of six types of supporting substrates (including TEM grid, Parafilm-M, aluminum foil, Ag foil, silicon wafer and cover glass) on hygroscopicity measurements using optical microscopy and concluded that TEM grids were the most suitable substrate for this application. Optical microscopy was also used to study hygroscopic properties of MgCl2 and NaCl-MgCl2 mixed particles (Gupta et al., 2015), and hygroscopic properties (including DRH and growth factors) of these particles were found to differ significantly from NaCl. Since MgCl2 is an important component in sea salt aerosol, this work can have significant implications for hygroscopicity and thus climatic impacts of sea salt aerosol (Zieger et al., 2017).

Optical microscopy can be (and has been widely) coupled to suitable spectroscopic techniques such as FTIR (Liu et al., 2008b), Raman spectroscopy (Liu et al., 2008c) and fluorescence (Montgomery et al., 2015), and if so chemical information can be simultaneously provided.

Electron microscopy has been widely used in laboratory and field studies to examine composition, mixing state and morphology of atmospheric particles, as summarized by a few excellent review articles (Prather et al., 2008; Posfai and Buseck, 2010; Li et al., 2015; Ault and Axson, 2017). Herein we discuss exemplary studies to illustrate how electron microscopy can help improve our knowledge of aerosol hygroscopicity. This section is further divided into two parts, i.e., scanning electron microscopy (SEM) and transmission electron microscopy (TEM).

Ebert et al. (2002) developed an environmental scanning electron microscope (ESEM) technique to explore hygroscopic properties of individual particles, and the instrument they used had a spatial resolution of 8–15 nm. Changes in particle morphology could be used to identify phase transitions (deliquescence and efflorescence), and growth factors could be derived from observed change in particle size at different RH. Their measured DRH and hygroscopic growth factors (Ebert et al., 2002) were in good agreement with results reported by the previous literature for NaCl, (NH4)2SO4, Na2SO4 and NH4NO3. However, ERH could not be accurately determined due to the influence of the substrate onto which particles under investigation were deposited (Ebert et al., 2002).

ESEM, coupled to energy disperse X-ray analysis (EDX), was employed to investigate hygroscopic properties of a wide range of atmospheric particles, including (NH4)2SO4 (Matsumura and Hayashi, 2007), sea spray (Hoffman et al., 2004), aerosol particles collected in nickel refineries (Inerle-Hof et al., 2007), agricultural aerosol (Hiranuma et al., 2008), pollen (Pope, 2010; Griffiths et al., 2012) and protein (Gomery et al., 2013). For example, Hoffman et al. (2004) found that both NaNO3 and NaNO3/NaCl particles existed as amorphous solids even at very low RH and exhibited continuous hygroscopic growth, instead of having clear DRH; furthermore, EDX analysis showed that Cl was enriched in the core of dried NaCl/NaNO3 particles (Hoffman et al., 2004), implying that during dehumidification NaCl started to crystallize first because of its lower solubility. This finding may have important implications for chemical and radiative properties of marine aerosol particles (Quinn et al., 2015). In another study (Pope, 2010), ESEM observations revealed that birch pollen gains swelled internally but did not take up water on the surface significantly, even at 93 % RH; however, liquid water could be observed on the particle surface when RH was > 95 %. Hiranuma et al. (2008) found that most aerosol particles collected at a cattle feedlot in Texas did not take up a significant amount of water at 96 % RH, though a small fraction of coarse particles became deliquesced at ∼75 % RH and their sizes were doubled at 96 % RH compared to their original sizes.

SEM/EDX was utilized by Krueger et al. (2003) to monitor changes in phase, morphology and composition of individual mineral dust particles after heterogeneous reaction with gaseous HNO3. For the first time, laboratory work showed that solid mineral dust particles could be transformed to aqueous droplets due to heterogeneous reactions (Krueger et al., 2003). As displayed in Fig. 9, solid CaCO3 particles were converted to spherical droplets as heterogeneous reaction with gaseous HNO3 proceeded (Krueger et al., 2003), and this was caused by the formation of Ca(NO3)2 which had very low DRH (Al-Abadleh et al., 2003; Kelly and Wexler, 2005). A following study (Krueger et al., 2004) examined heterogeneous reactions of HNO3 with mineral dust samples collected from four different regions, using SEM/EDX. It was suggested that calcite and dolomite particles exhibited large reactivity towards HNO3 and could be transformed to aqueous droplets, while no morphological change was observed for gypsum, aluminum silicate clay and quartz particles after exposure to HNO3 (Krueger et al., 2004).

The new laboratory discovery by Krueger et al. (2003) has been supported by a number of field measurements (Li et al., 2015; Tang et al., 2016a), in some of which SEM was also utilized. For example, Laskin et al. (2005) provided the first evidence demonstrating that in the ambient air solid nonspherical CaCO3 particles could be transformed to aqueous droplets which contained Ca(NO3)2 formed in heterogeneous reaction with nitrogen oxides. ESEM was also applied to examine mineral dust particles collected in Beijing (Matsuki et al., 2005) and southwestern Japan (Shi et al., 2008), and both studies found that some Ca-containing particles existed in aqueous state even at RH as low as 15 % because heterogeneous reactions with nitrogen oxides converted CaCO3 to Ca(NO3)2. Similarly, it was shown by SEM/EDX measurements (Tobo et al., 2010, 2012) that Ca-containing mineral dust particles in remote marine troposphere were transformed to aqueous droplets, because CaCl2 was formed in heterogeneous reaction of CaCO3 with HCl.

Compared to SEM, transmission electron microscopy (TEM) has better spatial resolution and can resolve features down to 1 nm or even smaller. TEM and AFM (atomic force microscopy) were employed by Buseck and colleagues (Posfai et al., 1998) to examine ambient particles collected on TEM grids under vacuum and ambient conditions. It was found that particle volumes were up to 4 times larger under ambient conditions compared to vacuum conditions. Several years later Buseck and co-workers (Wise et al., 2005) developed an environmental transmission electron microscope (ETEM) which enabled individual particles to be characterized under environmental conditions. The performance of this instrument was validated by measuring DRH and ERH of NaBr, CsCl, NaCl, (NH4)2SO4 and KBr particles in the size range of 0.1–1 µm, and good agreement was found between their measured values and those reported by previous work for all of the five compounds investigated (Wise et al., 2005).

The ETEM technique was further employed to investigate hygroscopic properties of a wide range of atmospheric particles, including NaCl-containing particles (Semeniuk et al., 2007b; Wise et al., 2007), biomass-burning particles (Semeniuk et al., 2007a) and potassium salts (Freney et al., 2009). The DRH of NaCl particles internally mixed with insoluble materials was determined to be ∼76 % (equal to that for pure NaCl), while internal mixing with other soluble compounds (e.g., NaNO3) would reduce the DRH (Wise et al., 2007). DRH and ERH were reported to be 85 % and 56 % for KCl and 96 and 60 % for K2SO4, while KNO3 displayed continuous hygroscopic growth (Freney et al., 2009); in addition, deliquescence and efflorescence of internally mixed KCl/KNO3 and KCl/K2SO4 were also examined (Freney et al., 2009). In another study (Adachi et al., 2011), aerosol particles, mainly sulfate internally mixed with weakly hygroscopic organic materials, were collected at Mexico City, and their hygroscopic properties were investigated using ETEM. It was found that only the sulfate part was deliquesced at elevated RH, while all the particles containing deliquesced sulfate did not necessarily became spherical. It was further suggested that the actual light scattering ability was 50 % larger than that estimated by Mie theory, which assumes particle sphericity (Adachi et al., 2011).

Recently cryogenic TEM has been deployed to explore morphology, hygroscopic properties and chemical composition of atmospheric particles (Veghte et al., 2014; Patterson et al., 2016). For example, it was observed that most nascent sea spray aerosol particles were homogeneous aqueous droplets, and upon exposure to low RH they would be quickly reorganized and undergo phase separation (Patterson et al., 2016).

Atomic force microscopy (AFM) is a widely used technique in surface chemistry and surface science. Compared to other microscopic techniques (e.g., optical microscopy, FTIR microscopy, TEM and SEM), AFM has several unique advantages. It does not require a vacuum condition and thus can be operated under environmental conditions; in addition, it has a high spatial resolution down to the nanometer level and offers 3-D imaging (Morris et al., 2016).

In the past 2 decades, AFM has been gradually utilized in atmospheric chemistry to observe 3-D morphology of aerosol particles, and its application in atmospheric chemistry started with observation of surfaces of single crystals with atmospheric relevance. For example, AFM was employed to study the (100) cleavage surface of NaCl during exposure to water vapor (Dai et al., 1997). A uniform layer of water was formed on the surface, and surface steps started to evolve slowly at ∼35 % RH; when RH increased to ∼73 % (approximately the DRH of NaCl), the step structure disappeared abruptly due to deliquescence of the surface (Dai et al., 1997). This pioneering work demonstrated that AFM had the potential to be used to determine the DRH of hygroscopic salts, in addition to providing rich information on surface structure change during exposure to water vapor. AFM was later used to observe the MgO(100) and CaCO3(1014) surface during exposure to water vapor and gaseous nitric acid (Krueger et al., 2005). Instabilities of oscillations in AFM images were observed, indicating that deliquescence of nitrate salts, which were formed in heterogeneous reaction with nitric acid, occurred at elevated RH (Krueger et al., 2005).

To our knowledge, AFM was successfully used in 1995 to characterize aerosol particles collected using a low-pressure impactor (Friedbacher et al., 1995). Three years later, Posfai et al. (1998) used AFM to examine individual particles collected above the North Atlantic Ocean at different RH. The particle volume was observed to be 4 times larger under ambient conditions (measured by AFM) compared to that in the vacuum (measured by TEM) (Posfai et al., 1998). Another study (Wittmaack and Strigl, 2005) used AFM to measure height-to-diameter ratios of ambient particles and concluded that some particles may exist in the supersaturated metastable state at around 50 % RH. Non-contact environmental AFM was used to examine uptake of water vapor by NaCl nanoparticles at RH below DRH (Bruzewicz et al., 2011). NaCl nanoparticles started to adsorb water at RH well below its DRH (75 %), and a liquid-like surface layer with a thickness of 2–5 nm was formed at 70 % RH, suggesting that deliquescence of NaCl nanoparticles was much more complicated than an abrupt first-order phase transition.

Very recently, Tivanski and co-workers (Ghorai et al., 2014; Laskina et al., 2015b; Morris et al., 2015, 2016) developed an AFM-based method to investigate hygroscopicity of particles deposited on substrates and systematically evaluated its performance by measuring hygroscopic growth factors of NaCl, malonic acid and a binary mixture of NaCl with malonic or nonanoic acid. It was found that hygroscopic growth factors derived from 3-D volume-equivalent diameters always agreed well with H-TDMA results; however, hygroscopic growth factors derived from 2-D area-equivalent diameters showed significant deviation from H-TDMA results for some types of particles (Morris et al., 2016). An example is displayed in Fig. 10, suggesting that at 80 % RH, the hygroscopic growth factor of NaCl particles derived from the volume-equivalent diameter was equal to that determined using H-TDMA, significantly larger than that derived from the area-equivalent diameter. Such deviation was caused by anisotropic growth of particles (Morris et al., 2016), and the extent of the deviation depended on the particle composition and their hydrate state at the time when they were collected on the substrate.

In addition to hygroscopicity measurement, AFM was used in several studies to characterize morphology, structure and other physicochemical properties of atmospheric particles (Lehmpuhl et al., 1999; Freedman et al., 2010; Laskina et al., 2015a). For example, AFM measurements found that organic and soot particles would shrink after interactions with O3, while inorganic particles remained unchanged (Lehmpuhl et al., 1999). Freedman et al. (2010) employed AFM coupled to Raman microscopy to characterize atmospheric particles under ambient conditions and observed core-shell structure for some organic particles. A recent study (Laskina et al., 2015a) characterized particles collected on substrates using AFM, Raman microscopy and SEM and suggested that microscopy techniques operated under ambient conditions would offer the most relevant and robust information on particle size and morphology. Conventional AFM offers no chemical information; however, it can be (and has already been) coupled to spectroscopic techniques (such as FTIR) (Dazzi et al., 2012; Ault and Axson, 2017; Dazzi and Prater, 2017), enabling detailed physical and chemical properties to be provided with high spatial resolution. Very recently, the peak force infrared microscopy, a type of scanning probe microscopy, was developed to investigate IR absorption and mechanical properties of ambient aerosol particles (Wang et al., 2017b), and a spatial resolution of 10 nm could be achieved.

Scanning transmission X-ray microscopy (STXM) is a novel technique which can provide spatial distribution of physical, chemical and morphological information of individual particles (de Smit et al., 2008) and has been recently employed to investigate atmospheric particles (Ault and Axson, 2017). For example, Ghorai and Tivanski (2010) developed a STXM-based method to study hygroscopic growth of individual submicrometer particles and proposed a method to quantify the mass of water associated with individual particles at a given RH. DRH and ERH values of NaCl, NaBr, and NaNO3, determined using STXM (Ghorai and Tivanski, 2010), agreed very well with previous results, and mass hygroscopic growth factors were also reported for these particles. In a following study (Ghorai et al., 2011), STXM was used to investigate hygroscopic growth of individual malonic acid; in addition to measured mass hygroscopic growth factors, near-edge X-ray absorption fine structure spectroscopy (NEXAFS) acquired using STXM suggested that keto-enol tautomerism occurred for deliquesced malonic acid particles (Ghorai et al., 2011). The keto-enol equilibrium constants were found to vary with RH, with enol formation favored at high RH (Ghorai et al., 2011).

Hygroscopic growth of submicrometer (NH4)2SO4, measured using STXM/NEXAFS (Zelenay et al., 2011a), agreed well with previous studies; furthermore, analysis of STXM images and NEXAFS spectra suggested that phase separation occurred for internally mixed (NH4)2SO4–adipic acid particles, and adipic acid was partially enclosed by (NH4)2SO4 at high RH (Zelenay et al., 2011a). An environmental chamber was constructed to be directly coupled to a STXM instrument (Kelly et al., 2013), and this setup was utilized to explore hygroscopic properties of NaCl, NaBr, KCl, (NH4)2SO4, levoglucosan and fructose (Piens et al., 2016). Measured mass hygroscopic growth factors were compared with those predicted by a thermodynamic model (AIOMFAC) (Zuend et al., 2011), and good agreement between measurement and prediction was found for all the compounds investigated (Piens et al., 2016). In another study, Zelenay et al. (2011b) utilized STXM/NEXAFS to investigate hygroscopic properties of submicrometer tannic acid and Suwannee River Fulvic acid used as proxies for humic-like substances found in atmospheric aerosol. Both compounds exhibited continuous water uptake, and at 90 % RH around one water molecule was associated with each oxygen atom contained by tannic acid, while approximately two water molecules were associated with each oxygen atom contained by Suwannee River Fulvic acid (Zelenay et al., 2011b).

STXM/NEXAFS has already been applied to explore hygroscopicity of ambient particles. For example, Pöhlker et al. (2014) collected aerosol particles from the Amazonian forest during periods with anthropogenic impacts and then analyzed these particles using STXM-NEXAFS at different RH. Substantial changes in particle microstructure were observed upon dehydration, very likely caused by efflorescence and crystallization of sulfate salts (Pöhlker et al., 2014). Piens et al. (2016) employed STXM-NEXAFS to examine hygroscopicity of atmospheric particles collected from the Department of Energy's Atmospheric Radiation Monitoring site in the Southern Great Plains. As shown in Fig. 11, compared to particles with medium and low hygroscopicity, particles with high hygroscopicity always contained larger fractions of Na and Cl (Piens et al., 2016).

Hygroscopicity measurements using microscopic techniques typically rely on changes in particle diameter measured microscopically. Therefore, it would be non-trivial for these techniques to quantify hygroscopic growth factors for non-spherical particles. In addition, these techniques may not be sensitive enough to investigate water adsorption. Since single particles deposited on supporting substances are usually examined, these techniques can be employed to investigate supersaturated samples if proper supporting substances are used. They have also been widely used to explore hygroscopic properties of ambient aerosol particles which were collected on proper substances. As discussed in Sect. 3.4, microscopic techniques can be and have widely been coupled to spectroscopic tools, and if so chemical information could be simultaneously provided.

Fourier transform infrared spectroscopy (FTIR), a vibrational absorption spectroscopy, has been widely employed in laboratory (Goodman et al., 2000; Eliason et al., 2003; Asad et al., 2004; Hung et al., 2005; Najera et al., 2009; Li et al., 2010; Tan et al., 2016; Tang et al., 2016b) and field work (Maria et al., 2002; Russell et al., 2011; Takahama et al., 2013, 2016, 2019; Kuzmiakova et al., 2016) to characterize the chemical composition of aerosol particles. It can also be used in aerosol hygroscopicity studies. When water is adsorbed or absorbed by particles, change in IR absorption of particles under investigation due to water uptake can be recorded as a function of RH, and therefore hygroscopic properties of these particles can be characterized. One advantage of FTIR is that it can be coupled with a range of accessories to form different experimental configurations, including transmission FTIR (Cziczo et al., 1997; Braban et al., 2001; Goodman et al., 2001; Zhao et al., 2006; Song and Boily, 2013; Leng et al., 2015; Zawadowicz et al., 2015), attenuated total reflection-FTIR (ATR-FTIR) (Schuttlefield et al., 2007a; Navea et al., 2010; Hatch et al., 2011; Zeng et al., 2014; Q. N. Zhang et al., 2014; Yeşilbaş and Boily, 2016; Navea et al., 2017; Gao et al., 2018), diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) (Gustafsson et al., 2005; Ma et al., 2010a; Joshi et al., 2017; Ibrahim et al., 2018) and micro-FTIR, for which FTIR is coupled with a microscope (Liu et al., 2008a; Liu and Laskin, 2009). Particles under investigation are typically deposited on proper substrates, though aerosol particles can also be studied using transmission FTIR (Cziczo et al., 1997; Cziczo and Abbatt, 2000; Zhao et al., 2006; Zawadowicz et al., 2015). FTIR has been used in a large number of studies to investigate hygroscopic properties of atmospherically relevant particles, and herein we only introduce and highlight a few representative examples.

Micro-FTIR was employed to investigate hygroscopic properties of CH3SO3Na particles (Liu and Laskin, 2009) and NH4NO3 (Wu et al., 2007). Figure 12a shows IR spectra of CH3SO3Na particles during humidification, and no significant change in IR spectra was observed when RH was increased from 0 % to 70 %; however, when RH was increased to 71 %, IR absorption attributed to the v(H2O) band (at ∼3400 cm-1) became very evident and its intensity increased with further increase in RH, indicating that the deliquescence of CH3SO3Na particles occurred at 71 % RH. In addition, at < 71 % RH two groups of narrow and structured bands, typically observed for crystalline samples, were observed for CH3SO3Na particles. The first one, centered at ∼1197 and 1209 cm-1, was attributed to asymmetrical stretching of ν8(-SO3-), and the other one, centered at 1062 cm-1, was attributed to symmetrical stretching of ν3(-SO3-). When RH was increased to 71 %, both bands were significantly broadened and shifted to lower wavelengths, further confirming that the DRH of CH3SO3Na particles was ∼71 %. IR spectra of CH3SO3Na particles during dehumidification are displayed in Fig. 12b. Complete disappearance of IR absorption at ∼3400 cm-1 and significant change in the shape and position of IR peaks of ν8(-SO3-) and ν3(-SO3-) were observed when RH was decreased from 49 % to 48 %, suggesting that the ERH of CH3SO3Na was around 48 %.

FTIR spectra can also be used to investigate hygroscopic growth quantitatively if IR absorbance can be calibrated. In the work by Liu and Laskin (2009), the absorbance ratio of v(H2O) (at ∼3400 cm-1) to ν8(-SO3-) (at ∼1192 cm-1) was calibrated and then used to calculate water-to-solute ratios (WSR, defined as mole ratios of H2O to CH3SO3-) of aqueous CH3SO3Na particles. As shown in Fig. 12c, WSR values determined using FTIR (Liu and Laskin, 2009) agreed well with those reported in a previous study (Peng and Chan, 2001b) using the electrodynamic balance (EDB). In another study (Liu et al., 2008a), DRH, ERH and WSR measured using micro-FTIR were found to agree well with those reported in the literature for NaCl, NaNO3 and (NH4)2SO4 particles. ATR-FTIR can be used in a similar way to micro-FTIR to investigate phase transitions and WSR of atmospherically relevant particles and has been applied to a number of compounds, including NaCl (Schuttlefield et al., 2007a; Zeng et al., 2014), NaNO3 (Tong et al., 2010b; Q. N. Zhang et al., 2014), Na2SO4 (Tong et al., 2010b), NH4NO3 (Schuttlefield et al., 2007a), (NH4)2SO4 (Schuttlefield et al., 2007a), CH3SO3Na (Zeng et al., 2014), sodium formate (Gao et al., 2018), and sodium acetate (Gao et al., 2018).

In addition, ATR-FTIR (Schuttlefield et al., 2007a, b; Hatch et al., 2011; Navea et al., 2017), DRIFTS (Ma et al., 2010a; Joshi et al., 2017; Ibrahim et al., 2018) and transmission FTIR (Goodman et al., 2001) have been employed to investigate water adsorption by insoluble particles, such as mineral dust. Figure 13 displays IR spectra of adsorbed water on SiO2 at different RH, as measured using DRIFTS at 30 ∘C. As shown in Fig. 13, two intensive peaks appeared in IR spectra at elevated RH (Ma et al., 2010a), one at 2600–3800 cm-1 attributed to the O–H stretching mode and the other one at ∼1630–1650 cm-1 attributed to the bending mode of H–O–H. Both peaks can be used to quantify the amount of adsorbed water, though surface OH groups may also contribute to the IR absorbance at ∼3400 cm-1 (Goodman et al., 2001; Tang et al., 2016a). The intensity of the third peak at 2100–2200 cm-1, attributed to the association mode of H–O–H, was much smaller (Ma et al., 2010a). It is possible but non-trivial to convert IR absorbance to the amount of adsorbed water, and the procedure used can be found elsewhere (Goodman et al., 2001; Ma et al., 2010a; Joshi et al., 2017; Ibrahim et al., 2018). It was found that the three-parameter BET equation (Joyner et al., 1945) could well describe water adsorption as a function of RH on mineral oxides (such as SiO2, TiO2, Al2O3, and MgO) (Goodman et al., 2001; Ma et al., 2010a; Joshi et al., 2017), authentic mineral dust from different sources (Joshi et al., 2017; Ibrahim et al., 2018) and Icelandic volcanic ash (Joshi et al., 2017). Another study (Hatch et al., 2011) suggested that, compared to the two-parameter BET equation, the Freundlich adsorption isotherm could better approximate the amount of water adsorbed by kaolinite, illite, and montmorillonite at different RH.

Raman spectroscopy is complementary to infrared spectroscopy. Bands which are weak in infrared spectroscopy can be strong in Raman spectroscopy, and vice versa. Compared to infrared spectroscopy, Raman spectroscopy is much less sensitive to H2O, despite the symmetric stretching vibration of H2O being Raman active, and this characteristic limits application in Raman spectroscopy in exploring particles with low hygroscopicity. Meanwhile, Raman spectroscopy is very sensitive to crystalline structures, making it very useful for investigating particle-phase transition. For example, Raman spectroscopy was employed to probe phase transformation of levitated (NH4)2SO4, Na2SO4, LiClO4, Sr(NO3)2, KHSO4, RbHSO4 and NH4HSO4 microparticles (Tang et al., 1995), and the occurrence of metastable solid states was observed under ambient conditions for Na2SO4, LiClO4, Sr(NO3)2 and bisulfates. Raman spectroscopy was also used to investigate hygroscopic properties of supersaturated droplets (Zhang and Chan, 2000; Zhang and Chan, 2002b), such as (NH4)2SO4 and MgSO4.

For regular spherical droplets, their Raman spectra may overlap with strong morphology-dependent resonances (Zhang and Chan, 2002b). Nevertheless, if individual droplets were deposited on proper substrates, Raman spectra with high quality (i.e., high signal-to-noise ratios) could be obtained using confocal micro-Raman spectroscopy (Wang et al., 2005; Li et al., 2006). For example, micro-Raman spectrometry was successfully used to investigate hygroscopic properties of (NH4)2SO4, Ca(NO3)2 and NO2-aged Ca(NO3)2 particles deposited on fluorinated ethylene propylene slides (Liu et al., 2008c; Zhao, 2010). Herein we use (NH4)2SO4 as an example to illustrate how Raman spectroscopy can be used to determine hygroscopic properties of atmospherically relevant particles. Figure 14 shows Raman spectra and microscopic images of an (NH4)2SO4 particle at different RH during humidification and dehumidification processes (Liu, 2008). When RH was increased to 80 % during humidification, the Raman peak centered at ∼3450 cm-1, attributed to the stretching vibration of H2O, started to become evident, whereas during dehumidification this peak disappeared when RH was decreased to 37 %. This suggested that deliquescence and efflorescence of (NH4)2SO4 took place at 80 and 37 % RH, respectively.

As discussed in previous work (Ling and Chan, 2007; Liu et al., 2008c; Zhao, 2010), the occurrence of deliquescence and efflorescence of (NH4)2SO4 could also be identified from the change in position and full width at half maxima (FWHM) of the Raman peak at 970–980 cm-1 (due to symmetrical stretching of sulfate, ν1-SO42-). As shown in Fig. 14, during humidification ν1-SO42- was shifted from 975 to 980 cm-1 when RH was increased to 80 %, and meanwhile its FWHM increased from 6 to 9 cm-1, implying the occurrence of deliquescence. For comparison, during dehumidification when RH was decreased to 37 %, ν1-SO42- was shifted from 978 to 975 cm-1 and the corresponding FWHM decreased from ∼10 to 6 cm-1, suggesting that efflorescence took place at ∼37 % RH. Phase transitions could be further inferred from microscopic images (Liu et al., 2008c; Zhao, 2010). Figure 14 shows that the particle under investigation became spherical when it was deliquesced (at 80 % RH) and became irregular when efflorescence occurred (at ∼37 % RH).

The peak intensity ratio of the stretching vibration of H2O to symmetrical stretching of sulfate is proportional to the molar ratio of H2O to sulfate in the solution and could be used to quantify the water-to-solute ratios (WSR) in aqueous (NH4)2SO4 droplets if properly calibrated (Liu et al., 2008c). WSR values determined using Raman spectroscopy (Liu et al., 2008c) were found to agree well with those reported in the literature as a function of RH for (NH4)2SO4 and Ca(NO3)2 during humidification and dehumidification processes (Stokes and Robinson, 1948; Tang and Munkelwitz, 1994; Clegg et al., 1998; Kelly and Wexler, 2005). In addition, Liu et al. (2008c) employed micro-Raman spectroscopy to study heterogeneous reaction of CaCO3 with NO2 and revealed that solid CaCO3 particles were converted to aqueous droplets after heterogeneous reaction with NO2, due to the formation of Ca(NO3)2.

Raman spectroscopy has been employed in a number of studies to investigate hygroscopic properties of organic aerosols and mixed particles (Ling and Chan, 2007, 2008; Yeung et al., 2009, 2010; Yeung and Chan, 2010; Ma and He, 2012; Q. Ma et al., 2013; Q. X. Ma et al., 2013). During humidification–dehumidification processes, oxalic acid was converted to oxalate when mixed with NaCl (Q. X. Ma et al., 2013) or Ca(NO3)2 (Ma and He, 2012), and such conversion would lead to significant change in the hygroscopic properties of mixed particles. When a hygroscopic sulfate, such as (NH4)2SO4 or Na2SO4), was mixed with a hygroscopic calcium salt, such as Ca(NO3)2 or CaCl2, gypsum, the hygroscopicity of which was very limited, would be formed by humidification. Raman spectroscopy was also used to explore hygroscopic properties of NH4NO3/(NH4)2SO4 mixed particles (Ling and Chan, 2007), and the formation of double salts, including 3(NH4NO3)⋅(NH4)2SO4 and 2(NH4NO3)⋅(NH4)2SO4, was observed for the first time during crystallization. The effects of malonic, glutaric and succinic acids on the hygroscopic properties of (NH4)2SO4 particles were explored using Raman spectroscopy (Ling and Chan, 2008). Partial crystallization of (NH4)2SO4/malonic acid droplets took place at 16 % RH, while (NH4)2SO4/glutaric acid and (NH4)2SO4–succinic acid particles became completely effloresced at ∼30 % RH. In addition, partial deliquescence with solid inclusions was observed at 10 % RH–79 % RH for (NH4)2SO4/malonic acid, 70 %-80 % for (NH4)2SO4/glutaric acid, and 80 % RH–90 % RH for (NH4)2SO4/succinic acid particles.

Water molecules in aqueous solutions can exist in two states, i.e., solvated water which interacts directly with ions and free water which interacts with other water molecules. Chan and co-workers (Choi et al., 2004; Choi and Chan, 2005) developed a method to explore the state of water molecules in single droplets levitated in an EDB. Pyranine, a water-soluble dye, was added into the droplets. When excited by radiation at ∼345 nm, Pyranine would emit fluorescence, and the spectra peaked at ∼440 nm (attributed to the presence of solvated water) and ∼510 nm (attributed to the presence of free water). The amounts of solvated and free water can be derived by combining mass hygroscopic growth factors (determined using the EDB) and the ratio of fluorescence intensity at 440 nm to that at 510 nm (Choi et al., 2004). It was found that for NaCl, Na2SO4 and (NH4)2SO4, efflorescence of supersaturated droplets occurred when the amount of solvated water was equal to that of free water (Choi et al., 2004; Choi and Chan, 2005). Imaging analysis further revealed that solvated water and free water were homogeneously distributed in the droplets for some types of droplets, e.g., MgSO4, but heterogeneously distributed for other types of droplets, such as NaCl and Na2SO4 (Choi and Chan, 2005).

In another study (Montgomery et al., 2015), fluorescence microscopy was used to monitor structural change in particle aggregates with RH. In this work NaCl particle aggregates were collected on wire meshes and then coated with Rhodamine which would generate fluorescence. Particle aggregates collapsed and became more compact when RH was increased from 0 % to 52 % (Montgomery et al., 2015), lower than the DRH of NaCl (∼75 %). Hosny et al. (2013) developed fluorescence lifetime imaging microscopy (FLIM) to determine the viscosity of individual particles by measuring the viscosity-dependent fluorescence lifetime of fluorescent molecular rotors. The viscosity of a particle is of interest because it is closely related to the phase state of the particle and largely determines diffusion in the particle (Koop et al., 2011; Reid et al., 2018). FLIM was used to investigate the viscosity of ozonated oleic acid particles and secondary organic particles formed by myrcene ozonolysis, and their viscosity was observed to increase largely with decreasing RH and increasing extent in oxidative aging (Hosny et al., 2016).

In addition to the spectroscopic and microscopic methods discussed in Sect. 3.3 and 3.4, there are a number of other surface characterization techniques which can be used to explore water adsorption on surfaces, e.g., sum frequency generation spectroscopy (Ma et al., 2004; Liu et al., 2005; Jubb et al., 2012; Ault et al., 2013), atmospheric pressure X-ray photoelectron spectroscopy (Ketteler et al., 2007; Salmeron and Schlogl, 2008; Yamamoto et al., 2010a), or scanning tunneling microscopy (Wendt et al., 2006; He et al., 2009). These techniques, which are able to provide fundamental and mechanistic insights into water–surface interactions, have mainly been applied to surfaces of single crystals, and their usefulness for particles with direct atmospheric relevance is yet to be demonstrated. As a result, these techniques are not further discussed here, and readers are referred to the aforementioned literature and references therein for more details.

Infrared and Raman spectroscopy can be used to quantify particle water content for unsaturated and supersaturated samples, with no restriction imposed by particle shape or morphology. Infrared spectroscopy is very sensitive to adsorbed water and has been widely used to investigate water adsorption (Tang et al., 2016a), as discussed in Sect. 3.3.1. In contrast, Raman spectroscopy is not sensitive enough to detect adsorbed water; nevertheless, recent work (Gen and Chan, 2017) showed that electrospray surface-enhanced Raman spectroscopy was able to detect surface-adsorbed water. One important advantage for infrared and Raman spectroscopy is that simultaneous measurement of chemical composition can be provided; therefore, they have been coupled to other techniques (such as optical microscope or electrodynamic balance) to further understand hygroscopic properties of atmospherically relevant particles, as discussed in Sects. 3.3, 3.4, 4.1 and 4.2. Infrared and Raman spectroscopy have been widely employed to characterize ambient aerosol particles collected on proper substrates, and therefore they can be used to explore hygroscopic properties of ambient particles in an offline manner.

The tandem differential mobility analyzer (TDMA) was pioneered in 1978 and called the aerosol mobility chromatograph at that time (Liu et al., 1978). The terminology “TDMA” was first introduced in 1986 in a study (Rader and McMurry, 1986) which showed that size change as small as 1 % could be readily measured. In addition to size change due to humidification (humidity-TDMA), TDMAs can also be used to measure particle size change due to other processing such as heating (Bilde et al., 2015a). H-TDMA is probably the most widely used technique for aerosol hygroscopicity measurement in both laboratory (Gibson et al., 2006; Herich et al., 2009; Koehler et al., 2009; Wex et al., 2009; Good et al., 2010b; Wu et al., 2011; Hu et al., 2014; Lei et al., 2014; Gomez-Hernandez et al., 2016; Jing et al., 2016; Zieger et al., 2017) and field studies (McMurry and Stolzenburg, 1989; Swietlicki et al., 2008; Ye et al., 2011, 2013; X. Wang et al., 2014; Yeung et al., 2014b; Atkinson et al., 2015; Cheung et al., 2015; Wu et al., 2016; Sorooshian et al., 2017). There are a number of H-TDMAs developed and used by individual research groups, and all the instruments follow the same principle. Recently these instruments have also become commercially available, e.g., from Brechtel Manufacturing Inc. (Lopez-Yglesias et al., 2014) and MSP Corporation (Sarangi et al., 2019). Swietlicki et al. (2008) provided a good description of the operation principle, and discussed potential error sources for H-TDMA measurements; Duplissy et al. (2009) analyzed the result from an intercomparison of six different H-TDMAs and recommended guidelines for design, calibration, operation and data analysis for H-TDMAs. Below we describe in brief how a typical H-TDMA works.

As illustrated in Fig. 19, polydisperse ambient or laboratory-generated aerosol particles were sampled through an aerosol dryer to reduce the RH to < 15 %, and the dry aerosol flow was passed through a neutralizer and then the first DMA (DMA1) to generate quasi-monodisperse aerosol particles. After that, the aerosol flow was delivered through a humidification section to be humidified to a given RH, and aerosol particles exiting the humidification section were monitored using the second DMA (DMA2) coupled with a condensation particle counter (CPC) to provide number size distributions. The growth factor (GF) is defined as the ratio of aerosol mobility diameter at a given RH to that at dry condition. The raw H-TDMA data should be inverted to retrieve the actual growth factor probability density function (Rader and McMurry, 1986; Gysel et al., 2009; Good et al., 2010a), and currently the inversion algorithm developed by Gysel et al. (2009) is widely used. One major uncertainty for H-TDMA measurements stems from the accuracy of RH in the second DMA, and considerable efforts are needed to minimize the RH and temperature fluctuation (Swietlicki et al., 2008; Duplissy et al., 2009; Massling et al., 2011; Lopez-Yglesias et al., 2014). The residence time in the humidification section should exceed 10 s for aerosol particles to reach the equilibrium under a given RH, while it should not be more than 40 s due to potential evaporation of semi-volatile species (Chan and Chan, 2005; Duplissy et al., 2009). In addition, it is important to check the H-TDMA performance via comparing the measured GF with theoretical values for reference aerosol particles, such as (NH4)2SO4 and NaCl (Swietlicki et al., 2008; Duplissy et al., 2009).

In typical laboratory studies (Herich et al., 2009; Koehler et al., 2009; Jing et al., 2016; Zieger et al., 2017), aerosol size is measured at different RH using the H-TDMA to get the RH-dependent GF. Humidograms, in which GF are plotted as a function of RH, are shown in Fig. 20 for NaCl and synthetic sea salt aerosol particles, suggesting that at a given RH, GF of sea salt aerosol is 8 %–15 % smaller than NaCl aerosol (Zieger et al., 2017). Since both NaCl and synthetic sea salt aerosol particles are non-spherical under dry conditions, growth factors were reported after shape factor correction. The difference in GF between NaCl and synthetic sea salt aerosols was attributed to the presence of hydrates (such as the hydrates of MgCl2 and CaCl2) with lower hygroscopicity (when compared to NaCl) in synthetic sea salt (Zieger et al., 2017).

H-TDMA has been widely used to investigate hygroscopic growth of secondary organic aerosol (Prenni et al., 2007; Duplissy et al., 2008; Wex et al., 2009; Good et al., 2010b; Massoli et al., 2010; Duplissy et al., 2011; Alfarra et al., 2013; Zhao et al., 2016), which significantly contributed to submicrometer aerosol particles over the globe (Zhang et al., 2007). Using an aerosol flow tube, Massoli et al. (2010) generated secondary organic aerosols (SOA) via OH oxidation of α-pinene, 1,3,5-trimethylbenzenen (TMB), m-xylene and a 50:50 mixture of α-pinene and m-xylene and measured their hygroscopic growth at 90 % RH using a H-TDMA. As shown in Fig. 21, measured GF at 90 % RH ranged from 1.05 (non-hygroscopic) to 1.35 (moderately hygroscopic) for SOA systems examined, increasing linearly with O:C ratios determined using an Aerodyne High Resolution Time-of-Flight Mass Spectrometer (Massoli et al., 2010). In addition, for most SOA systems studied, single hygroscopicity parameters (κ) derived from H-TDMA measurements were smaller than these derived from CCN activity measurements (Massoli et al., 2010). Gaps between hygroscopic growth and cloud activation have also been observed in a number of other studies for SOA (Prenni et al., 2007; Juranyi et al., 2009; Petters et al., 2009; Wex et al., 2009; Good et al., 2010b; Whitehead et al., 2014; Zhao et al., 2016). One major reason for such gaps is that SOA usually contain substantial amount of slightly soluble materials, which only undergo partial dissolution under water-subsaturated conditions but can be dissolved to a significantly larger extent under water supersaturated conditions (when more water is available). Further discussion on reconciliation between hygroscopic growth and cloud activation can be found elsewhere (Petters et al., 2009; Wex et al., 2009). In another study (Y. J. Li et al., 2014), H-TDMA was used to explore hygroscopic properties of SOA formed via OH oxidation and direct photolysis of methoxylphenol (a model compound for biomass-burning aerosol) in the aqueous phase. For SOA generated from aqueous-phase OH oxidation, GF at 90 % RH was observed to increase linearly with the O:C ratio, but the slope was around 3 times smaller than that reported by Massoli et al. (2010).

Since RH scanning is time-consuming, in most ambient applications H-TDMA measurements are usually carried out at a fixed RH (mostly 90 % and 85 % to a less extent) for one or a few dry particle diameters (Swietlicki et al., 2008; Kreidenweis and Asa-Awuku, 2014; Cheung et al., 2015). Usually at least one diameter in the center of Aitken mode (∼50 nm) and one size in the center of the accumulation mode (∼150 nm) are selected (Swietlicki et al., 2008). The second DMA is typically scanned over a diameter range to cover a corresponding GF range between 0.9 and 2.0 (sometimes up to 2.5) at 90 % RH (Swietlicki et al., 2008). However, there have been a few studies which measured GF of size-selected ambient aerosols as a function of RH (Santarpia et al., 2004; Cheung et al., 2015). For example, Cheung et al. (2015) measured the GF of ambient aerosol particles (100 and 200 nm) as a function of RH (10 %–93 %) in Hong Kong using a H-TDMA and found that the derived κ values at (or above) 90 % RH were significantly larger than those derived at 40 % RH. Each set of such measurements took ∼3 h, limiting its application to periods with large fluctuation in aerosol composition (Cheung et al., 2015).

To further understand hygroscopic properties of ambient aerosol particles, aerosol hygroscopicity closure studies have been widely carried out (Swietlicki et al., 1999; Dick et al., 2000; Gysel et al., 2007; Cerully et al., 2011; Wu et al., 2013, 2016; Schurman et al., 2017; Hong et al., 2018). In such studies, hygroscopic growth measurements using H-TDMA are concurrently performed with aerosol chemical composition measurements, and measured growth factors can then be compared to these calculated based measured chemical composition. Aerosol chemical compositions were usually measured offline in the early stage (Swietlicki et al., 1999; Dick et al., 2000) and have been increasingly determined online with high time resolution using aerosol mass spectrometry (Gysel et al., 2007; Wu et al., 2013) and single-particle mass spectrometry (X. M. Wang et al., 2014; K. Li et al., 2018). For example, Wu et al. (2013) used a H-TDMA to measure aerosol hygroscopic growth at 90 % RH and an Aerodyne High Resolution Time-of-Flight Mass Spectrometer (HR-ToF-AMS) to measure size-resolved aerosol chemical composition at a middle-level mountain area in central Germany. Single hygroscopicity parameters, κhtdma, derived from growth factors measured using H-TDMA, were compared to those derived from aerosol composition (κchem), assuming ideal mixing. If the average compositions of submicron particles were used to calculate κchem, reasonably good agreement between κhtdma and κchem was found for 250 nm particles while no correlation was observed for 100 nm particles (Wu et al., 2013). If size-resolved aerosol compositions were used to calculate κchem, as shown in Fig. 22, good closure between κchem and κhtdma were found for all four particle sizes. Figure 22 also reveals that κchem were significantly larger than κhtdma, indicating that ideal mixing assumption may overestimate aerosol hygroscopic growth (Wu et al., 2013). Simultaneous H-TDMA and HR-ToF-AMS measurements were also carried out at a coastal suburban site in Hong Kong (Yeung et al., 2014a). Approximations for growth factors of organic aerosols, using the fraction of m/z 44, the oxygen-to-carbon ratio and PMF-resolved organic factors from HR-ToF-AMS measurements, did not yield better closure results, likely because of the overall dominance of sulfate during the whole measurement period.

H-TDMA measurements in Shanghai at wintertime showed that aerosol particles (250 nm in dry diameter) could be classified into two modes according to their hygroscopicity (X. Wang et al., 2014). The first mode had growth factors of ∼1.05 at 85 % RH, mainly containing fresh elemental carbon and minerals, as revealed by measurements using a single-particle mass spectrometer (aerosol time-of-flight mass spectrometer). In contrast, the second mode had growth factors of ∼1.46 at 85 % RH and were enriched with elemental carbon and organic carbon particles internally mixed with secondary inorganic materials.

While most H-TDMAs only work at around room temperature, Weingartner et al. (2002) designed a H-TDMA which could measure hygroscopic growth of aerosol particle below 0 ∘C (temperature: -20 to 30 ∘C; RH: 10 %–90 %). Measured hygroscopic growth factors showed good agreement with theoretical calculations for (NH4)2SO4, NaCl and NaNO3 at both 20 and -10 ∘C (Gysel et al., 2002). This instrument was subsequently deployed at a high-alpine site (3580 m above sea level) to investigate hygroscopic properties of ambient aerosol particles at -10 ∘C (Weingartner et al., 2002), and the average GF at 85 % RH were measured to be 1.44, 1.49 and 1.53 for aerosol particles with dry diameters of 50, 100 and 250 nm.

RH in the troposphere frequently exceeds 90 %, and it is desirable to investigate hygroscopic growth of aerosol particles at > 90 % RH. Hennig et al. (2005) developed a high-humidity TDMA which could be operated at 98 % RH, and the absolute accuracy of RH at 98 % was ±1.2 %. It was found that within the uncertainties, the measured GF in the RH range of 84 %–98 % agreed well with theoretical values (Hennig et al., 2005). The Leipzig Aerosol Cloud Interaction Simulator (LACIS), a laminar flow tube designed to study cloud formation and growth, could be operated at stable RH ranging from almost 0 % up to 99.1 % (Stratmann et al., 2004), and aerosol particles and/or droplets exiting the flow tube were detected using an optical particle sizer especially developed for this instrument. LACIS was employed to study hygroscopic growth of (NH4)2SO4 and NaCl aerosol particles at 85.8 % RH–99.1 % RH (Wex et al., 2005). At 99 % RH, measured GF values agreed well with these predicted assuming solution ideality for NaCl, whereas for (NH4)2SO4, solution ideality assumption would overestimate GF values by up to 20 % (Wex et al., 2005). In a following study (Niedermeier et al., 2008), LACIS was used to investigate hygroscopic growth of sea salt aerosol up to 99.1 % RH.

Long duration is required by the second DMA to measure size distributions of humidified aerosol particles, and therefore the H-TDMA technique is usually quite slow. It typically takes ∼30 min for a traditional H-TDMA to determine GF values at a given RH for five different dry diameters (Cerully et al., 2011; Pinterich et al., 2017b). Instruments with fast duty cycles are of great interest and have been developed and deployed (Sorooshian et al., 2008; Pinterich et al., 2017a, b). For example, after replacing the second DMA (used in the traditional H-TDMA) with a water-based fast integrated mobility spectrometer which could provide 1 Hz size distribution measurements (Pinterich et al., 2017a), the improved instrument, called the humidity-controlled fast integrated mobility spectrometer (HFIMS), only took ∼3 min to measure GF of particles with five different dry diameters at a given RH (Pinterich et al., 2017b).

Since the upper size limit is < 1000 nm for a typical DMA and GF values at 90 % RH can be > 2 for atmospheric particles, most H-TDMAs can only be used for particles with dry diameters smaller than 500 nm (McFiggans et al., 2006; Swietlicki et al., 2008). Several instruments, which could measure hygroscopic growth of aerosol particles larger than 500 nm in dry diameter, have been developed (Kreisberg et al., 2001; Hegg et al., 2007; Massling et al., 2007; Snider and Petters, 2008; Kaaden et al., 2009; Kim et al., 2014). One obvious approach to overcome the DMA sizing limit is to use optical particle counters for particle sizing, as adopted by some previous studies (Kreisberg et al., 2001; Hegg et al., 2007; Snider and Petters, 2008). Another approach is to use aerodynamic particle sizers (APS) for particle sizing (Massling et al., 2007; Kaaden et al., 2009; Schladitz et al., 2011; Kim and Park, 2012). For example, a H-DMA-APS was developed to explore hygroscopic growth of large aerosol particles (Massling et al., 2007; Kaaden et al., 2009). As shown in Fig. 23, the dry aerosol flow was first delivered through a custom-built high-aerosol flow-DMA (HAF-DMA) which could select particles with dry mobility diameters over 1000 nm, and the dry aerosol flow exiting the DMA was split into two identical flows; the first flow was directly sampled by the first APS to measure the aerodynamic size distribution under dry conditions, and the second flow was first delivered through a humidifier to be humidified to a given RH (e.g., 90 %) and then sampled into the second APS so that the aerodynamic size distribution of the humidified aerosol was measured.

The utilization of H-TDMAs to measure aerosol hygroscopic growth factors assumes particle sphericity. Some particles in the atmosphere, such as mineral dust and soot, are known to be non-spherical, and therefore GF measured using H-TDMA may not correctly reflect the amount of aerosol liquid water (Weingartner et al., 1997; Rissler et al., 2005; Vlasenko et al., 2005; Koehler et al., 2009; Tritscher et al., 2011). Very recently, an instrument, called differential mobility analyzer-humidified centrifugal particles mass analyzer (DMA-HCPMA), was developed to measure mass change in submicron aerosol particles at different RH (10 %–95 %) (Vlasenko et al., 2017). In this setup, a dry aerosol flow was delivered through a DMA to produce quasi-monodisperse particles and then through an aerosol humidifier to be humidified to a give RH; after that, the aerosol flow was delivered through a centrifugal particle mass analyzer (which would classify aerosol particles according to their mass-to-charge ratios) (Olfert and Collings, 2005; Rissler et al., 2014; Kuwata, 2015) and then a CPC so that aerosol particle mass could be determined as a function of RH (Vlasenko et al., 2017). The measured mass growth factors were found to agree well with theoretical values for (NH4)2SO4 and NaCl, and this newly developed DMA-HCPMA setup was successfully deployed to explore hygroscopic properties of ambient aerosol particles (Vlasenko et al., 2017). It can be expected that DMA-HCPMA would significantly improve our knowledge of hygroscopicity of non-spherical aerosol particles.

Cavity ring-down spectroscopy (CRDS), a highly sensitive method for optical extinction measurement, has been extensively employed for gas and aerosol detection (Brown, 2003; Baynard et al., 2006, 2007; Langridge et al., 2011; Sobanski et al., 2016; Peng et al., 2018). For a typical CRDS setup, a laser beam pulse is coupled into a high-finesse optical cavity (which has one high-reflectivity mirror on each end) from one end of the cavity, and the decay of the intensity of the light transmitted from the other end is monitored. The change in decay lifetimes of transmitted light intensity can be related to the extinction coefficient, αext, using Eq. (3) (Baynard et al., 2007; Langridge et al., 2011): 3αext=RLc1τ-1τ0, where RL is the ratio of the distance between the two mirrors to the length of the cavity filled with aerosol particles, c is the speed of light (m s-1), and τ and τ0 are the measured decay lifetimes of light intensity with and without aerosol particles present in the cavity. If aerosol particles delivered into the cavity are monodisperse, the extinction coefficient of each individual particle, σext, can be calculated using Eq. (4) (Freedman et al., 2009): 4σext=σextNp, where Np is the aerosol number concentration (cm-3).

A CRD spectrometer was employed by Tolbert and co-workers to investigate the effects of RH on aerosol optical extinction at 532 nm, and its schematic diagram is depicted in Fig. 24 (Beaver et al., 2008). The aerosol flow generated using an atomizer was delivered through diffusion dryers to reduce its RH to < 10 % and passed through a DMA to produce quasi-monodisperse aerosol particles. The aerosol flow was then delivered into the first cavity to measure the aerosol optical extinction at 532 nm under dry conditions; after that, the aerosol flow entered a humidifier to be humidified to a given RH and was then delivered into the second cavity to measure the aerosol optical extinction under the humidified condition. In the final, the aerosol flow was sampled by a CPC to measure the number concentration. For (NH4)2SO4 aerosol particles in the size range of 200–700 nm, the measured optical growth factors at 80 % RH, defined as the ratio of the extinction coefficient at 80 % RH to that under dry conditions, were found to be in good agreement with those calculated from diameter-based growth factors using Mie theory (Garland et al., 2007).

CRDS was used to examine the effect of RH on aerosol optical extinction for phthalic acid, pyromellitic acid and 4-hydroxybenzoic acid aerosol particles in the size range of 150–500 nm (Beaver et al., 2008). The optical growth factors were found to be smaller for the three organic compounds examined, compared to (NH4)2SO4. For example, for aerosol particles with a dry diameter of 335 nm, optical growth factors at 80 % RH were measured to be 1.3 and 1.1 for phthalic and pyromellitic acid (Beaver et al., 2008), compared to 3.0 for (NH4)2SO4. Optical extinction coefficients of 4-hydroxybenzoic acid particles at 80 % RH were smaller than those under dry conditions (Beaver et al., 2008), implying that morphological and structural change may occur for these particles during humidification. Similarly, optical growth factors of illite and kaolinite aerosol particles were found to be < 1 at 50 and 68 % RH (Attwood and Greenslade, 2011), due to structural rearrangement of clay mineral particles after water uptake. Optical growth factors of internally mixed aerosol particles, which contained (NH4)2SO4 and organic materials, were also studied (Garland et al., 2007; Robinson et al., 2013, 2014). Another study (Michel Flores et al., 2012) measured optical growth factors (at wavelengths of 355 and 532 nm) at 80 and 90 % RH for aerosol particles with different extents of optical absorption ranging from purely scattering (e.g., (NH4)2SO4) to highly absorbing (e.g., nigrosine) and found good agreement between measured optical growth factors and those calculated using Mie theory.

CRDS has also been widely deployed to investigate optical extinction of ambient aerosol particles at different RH (X. L. Zhang et al., 2014; Atkinson et al., 2015; Brock et al., 2016a). For example, an eight-channel CRD spectrometer was developed by NOAA Earth System Research Laboratory (Langridge et al., 2011). This instrument could measure aerosol optical growth factors at three wavelengths (405, 532 and 662 nm) simultaneously and has been successfully deployed for aircraft measurements (Langridge et al., 2011).

In addition to CRDS, broadband cavity enhanced spectroscopy (BBCEAS), also called cavity enhanced differential optical absorption spectroscopy (CE-DOAS), is an alternative high-finesse cavity-based technique with high sensitivity in optical extinction measurements (Platt et al., 2009; Washenfelder et al., 2013, 2016). Compared to CRDS, one major advantage of BBCEAS is that optical extinction can be measured as a function of wavelength. BBCEAS, as described in detail elsewhere (Platt et al., 2009; Varma et al., 2013; Washenfelder et al., 2013, 2016; Zhao et al., 2014; Wang et al., 2017a; Z. Y. Li et al., 2018), has also been widely used in gas and aerosol measurements. Zhao et al. (2014) utilized BBCEAS to measure aerosol optical extinction at 641 nm as a function of RH, and for 200 nm (NH4)2SO4, the measured optical growth factors agreed well with those calculated using Mie theory. The instrument was further deployed to simultaneously measure optical extinction of ambient submicrometer aerosol at < 20 % and 85 % RH at Hefei Radiation Observatory. The result is displayed in Fig. 25, suggesting that the optical growth factors at 85 % RH varied from ∼1 to > 2.5 during the campaign (Zhao et al., 2017).

Humidified nephelometry, which was first developed as early as in the 1960s (Pilat and Charlson, 1966; Covert et al., 1972), has been widely used to measure aerosol light scattering coefficients at different RH (Rood et al., 1985; Carrico et al., 1998; Li-Jones et al., 1998; Day et al., 2000; Malm et al., 2000a, b; Koloutsou-Vakakis et al., 2001; Fierz-Schmidhauser et al., 2010a; Zieger et al., 2010, 2013; Kreidenweis and Asa-Awuku, 2014; Zhang et al., 2015; Titos et al., 2016). Due to its high time resolution, this technique is very suitable for online measurement of ambient aerosols. A very recent review paper (Titos et al., 2016) summarized and discussed theories, history, measurement uncertainties and ambient applications of this technique in a comprehensive manner. As a result, herein we only introduce in brief its basic principle, representative instrumental configurations and exemplary applications.

The scattering enhancement factor, f(RH), defined as the ratio of the aerosol scattering coefficient at a given RH to that at dry conditions, is typically reported by humidified nephelometry measurements (Kreidenweis and Asa-Awuku, 2014; Titos et al., 2016). Figure 26 shows the schematic diagram of a humidified three-wavelength integrating nephelometer (TSI 3563) at 450, 550 and 700 nm (Fierz-Schmidhauser et al., 2010c). The aerosol flow was first delivered through an aerosol humidifier which could increase the RH to 95 % and then through an aerosol dryer to reduce the RH to below 40 %. After that, the aerosol flow was sampled into the nephelometer to measure aerosol scattering coefficients at three different wavelengths. The flow exiting the nephelometer was pulled through a mass flow controller (to control the sample flow rate) by a pump. The performance of the aerosol dryer could be adjusted to vary the RH of the flow entering the nephelometer, and thus scattering coefficients could be measured as a function of RH (40 %–90 %); in addition, using such a configuration, light scattering properties of supersaturated aerosol particles, i.e., the hysteresis effect, could be examined (Fierz-Schmidhauser et al., 2010c). The humidified nephelometer was used to measure light scattering properties of monodisperse (NH4)2SO4 and NaCl aerosol particles with dry diameters of 100, 150, 240 and 300 nm, and the measured f(RH) values agreed with these predicted using Mie theory (Fierz-Schmidhauser et al., 2010c). Some instruments could measure aerosol light scattering at different RH in a simultaneous manner, via using two or more nephelometers in parallel (Carrico et al., 1998).

A number of previous studies have carried out field measurements of f(RH) at various locations over the globe (Zieger et al., 2013; Kreidenweis and Asa-Awuku, 2014; Titos et al., 2016). As summarized by Titos et al. (2016), f(RH) values (for 80 % RH–85 % RH) were larger for marine sites (ranging from 1.5 to 3.5), when compared with most continental sites; furthermore, f(RH) values were found to be in the range of 1.1–2.1 for dust particles, and larger f(RH) values observed for dust may be caused by the co-presence of sea salt aerosol. A field study (Li-Jones et al., 1998) carried out on Barbados (West Indies) found that f(RH) values (for RH in the range of 67 %–83 %) were very small (1.0–1.1) for mineral dust transported from North Africa, indicating that large variation in ambient RH may not lead to significant change in optical properties of mineral dust aerosol.

Since aerosol light scattering coefficients depend on particle size and refractive index in a complex manner even for spherical particles, it is not straightforward to link f(RH) with the aerosol liquid water content (Kreidenweis and Asa-Awuku, 2014). A number of studies (Malm and Day, 2001; Fierz-Schmidhauser et al., 2010b; Zieger et al., 2010; Chen et al., 2014; Kreidenweis and Asa-Awuku, 2014; Kuang et al., 2017, 2018) have discussed how measured f(RH) values could be used to derive single hygroscopicity parameters (κ) (Petters and Kreidenweis, 2007) and aerosol liquid water contents. In addition, it should be emphasized that humidity-dependent aerosol scattering coefficients (as well as aerosol extinction and absorption coefficients) themselves are important parameters to assess the impacts of aerosols on visibility and direct radiative forcing.

Photoacoustic spectroscopy has been developed and deployed to measure aerosol optical absorption in a direct manner (Arnott et al., 2003; Lack et al., 2009; Lewis et al., 2009; Moosmuller et al., 2009; Gyawali et al., 2012; Langridge et al., 2013; Lack et al., 2014). In brief, the aerosol flow is continuously sampled into a cell which serves as an acoustic resonator section and illuminated by a modulated laser beam. The laser radiation absorbed by aerosol particles is transferred to the surrounding air as heat, leading to the generation of an acoustic wave which is amplified in the resonator and detected using a microphone (Moosmuller et al., 2009; Gyawali et al., 2012). The signal intensity measured by the microphone is proportional to optical absorption and can be used to derive aerosol optical absorption coefficients after proper calibration (Moosmuller et al., 2009; Gyawali et al., 2012). In principle, hygroscopic growth of aerosol particles at elevated RH would lead to increase in particle size and thus enhancement in aerosol optical absorption due to the lensing effect (Lewis et al., 2009). Nevertheless, several studies suggested that photoacoustic spectroscopy measurements at high RH are likely to significantly underestimate the actual aerosol optical absorption (Arnott et al., 2003; Lewis et al., 2009; Langridge et al., 2013). For example, Langridge et al. (2013) used photoacoustic spectroscopy at 532 nm to measure optical absorption of several types of aerosol particles with various hygroscopicity, morphology and refractive indices and found that the measured absorption exhibited strong low biases at high RH. The underestimation of optical absorption is due to that acoustic signals are affected by evaporation of aerosol liquid water when aerosol particles absorb radiation and get heated. As a result, Langridge et al. (2013) concluded that photoacoustic spectroscopy was not a suitable technique to measure aerosol optical absorption at elevated RH. Similarly, other techniques used for direct measurement of aerosol optical absorption, such as the filter-based method and photothermal interferometry, did not perform well at elevated RH either (Schmid et al., 2006; Sedlacek and Lee, 2007).

An indirect method has been developed (Khalizov et al., 2009; Xue et al., 2009; Brem et al., 2012; Chen et al., 2015) to explore the effect of RH on aerosol optical absorption, which was calculated as the difference between aerosol light extinction and scattering. In the setup developed by Brem et al. (2012), aerosol light extinction and scattering at three wavelengths (467, 530 and 660 nm) were measured at different RH using an optical extinction cell and a nephelometer. As RH was increased from 38 to 95 %, light absorption of nigrosine aerosol was enhanced by a factor of ∼1.24 for all three wavelengths (Brem et al., 2012). In some other work (Khalizov et al., 2009; Xue et al., 2009; Chen et al., 2015), CRDS, instead of the optical extinction cell, was used to measure the aerosol optical extinction.