Simultaneous measurement of the relative humidity dependent aerosol light extinction, scattering, absorption and single-scattering albedo with a humidified cavity-enhanced albedometer

Hygroscopic aerosols take up water and grow with increasing relative humidity (RH), giving rise to large changes 15 in light extinction (bext), scattering (bscat), absorption (babs), and single scattering albedo (SSA, ). The optical hygroscopic growth factors for each parameter (f(RH)ext,scat,abs,) are thus important for assessing aerosol effects on regional air quality, atmospheric visibility, and radiative forcing. The RH dependence of aerosol scattering and extinction has been studied in many laboratory and field studies. However, owing partly to the absence of suitable instrumentation, there are few reports of the RH dependence of aerosol absorption and . In this work, we report the development of a humidified cavity-enhanced 20 albedometer (H-CEA) for simultaneous measurements of f(RH)ext,scat,abs, at  = 532 nm from 10% to 90% RH. The instrument’s performance was evaluated with laboratory-generated ammonium sulphate, sodium chloride and nigrosin aerosols. Measured hygroscopic growth factors for different parameters were in good agreement with model calculations and literature reported values, demonstrating the accuracy of the H-CEA for measuring RH-dependent optical properties.

Multiple techniques have been developed to characterize aerosol hygroscopic behaviour over the last few decades and have been described in several reviews (Kreidenweis and Asa-Awuku, 2014;Titos et al., 2016;Zhao et al., 2019;Tang et al. 5 2019). The instrument commonly used to characterise particle size growth in the laboratory and in field applications is the hygroscopic tandem differential mobility analyser (H-TDMA). The growth factor (GF) is obtained by measuring the ratio of particles under humid and dry conditions (Swietlicki et al. 2008;Tang et al., 2019).
Similarly, the enhancement factor (f(RH)) for optical properties is defined as: 10 ,, , where bext,scat,abs is either the extinction, scattering, or absorption coefficient and ω is the single scattering albedo (SSA, ω = bscat / bext). bext,scat,abs and ω are functions of wavelength  and relative humidity, typically > 80% or 30-40% (Dry). By combining H-TDMA with an optical instrument, size dependent aerosol optical hygroscopicity can be measured (Michel Flores et al., 2012). 15 Advances in optical methods have allowed significant progress to be made in studying f(RH). In early work, Baynard et al. (2006) developed a dual-channel CRDS for extinction measurements ( = 532 nm) at 80% RH and under dry conditions (RH < 10%) to measure f(RH)ext. The instrument was used for measuring the RH dependence of the extinction of inorganic aerosols and a mixture of organic and ammonium sulphate aerosols (Baynard et al., 2006;Garland et al., 2007). In 2009, Massoli et al. (2009a) developed a six-channel CRDS for measuring aerosol extinction at  = 355, 532 and 1064 nm at 25%, 20 60% and 85% RH. In 2011, an eight-channel CRDS for aircraft observations was developed by Langridge et al. (2011).
Three channels ( = 532 nm) were used for RH dependent extinction measurement at < 10%, 70% and 95% RH. Recently, Zhao et al. (2017) reported the development of a humidified broad-band cavity-enhanced aerosol extinction spectrometer (H-BBCES) for measuring the effect of hygroscopicity on extinction at  = 461 nm. These high-finesse cavity based spectroscopy methods provide valuable, sensitive and accurate methods for in-situ measurement of bext and f(RH) ext 25 (Baynard et al., 2006;Langridge et al., 2011;Michel Flores et al., 2012;Zhao et al., 2017).
For measuring f(RH)scat, the humidified nephelometer is widely used (Covert et al., 1972). Commercial nephelometers (Titos et al., 2016) can measure the scattering enhancement factor at a fixed RH or scan over the hygroscopic growth curve between 40 and 90% RH (Yan et al., 2009;Fierz-Schmidhauser et al., 2010;Chen et al., 2014;Zhang et al., 2015). This approach has been used to investigate how hygroscopic growth affects aerosol scattering properties and to establish the 30 relationship between aerosol chemical composition, diameter hygroscopic growth and hygroscopicity parameter (Zieger et al., 2010;Chen et al., 2014;Zhang et al., 2015;Kuang et al., 2017). For moderately hygroscopic aerosols, the uncertainty of f(RH)scat was about 20-40% and mainly arose from uncertainties in RH, dry reference state and nephelometer measurement https://doi.org/10.5194/amt-2020-6 Preprint. Discussion started: 27 January 2020 c Author(s) 2020. CC BY 4.0 License. uncertainties (Titos et al., 2016). The truncation error is unavoidable in current commercial nephelometers (Massoli et al., 2009b) and modifications such as reducing the lamp power or replacing the original RH sensor inside the nephelometer chamber are sometimes needed to improve the accuracy of f(RH)scat measurements (Zhao et al., 2019).
Suitable instruments for measuring f(RH)abs and f(RH)ω are lacking, however. Filter-based instruments such as aethalometers and particle soot absorption photometers suffer from a change in the nature of the suspended state of the 5 sample and are not suitable for high RH operation (Arnott et al., 2003). Photoacoustic spectroscopy (PAS) is a direct and insitu method for measuring babs (Lack et al., 2006;Moosmüller et al., 2009), but sample RH is commonly controlled in low RH conditions (10-30% RH) owing to an evaporation-induced bias of about 7-30% (Lack et al., 2009;Langridge et al., 2013). In contrast, recently developed cavity-based albedometers allow simultaneous in situ measurements of bext, bscat, babs and ω in the same sample volume. These instruments include the CRDS-albedometer (Thompson et al., 2008;Ma et al., 10 2012), the BBCES-albedometer (Zhao et al., 2014;Xu et al., 2018a) and the cavity attenuated phase shift spectroscopy (CAPS)-albedometer (Onasch et al., 2015), which combine CRDS/BBCES/CAPS with integrating spheres (IS). These albedometers are suitable for operating under high RH conditions and have advantage of same sample conditions over independent measurements of different parameters with different instruments (Wei et al., 2013;Zhao et al., 2014;Xu et al., 2018a). 15 In this work, we report the first demonstration of a humidified cavity-enhanced albedometer (H-CEA) that combines a BBCES-albedometer with a humidigraph system for simultaneous and accurate measurement of f(RH)ext,scat,abs,ω at  = 532 nm. The instrument performance was tested with laboratory-generated ammonium sulphate, sodium chloride and nigrosine aerosols. The H-CEA provides a new method for direct and in-situ measurement of multiple optical hygroscopic parameters with a single instrument. 20 2 Experimental setup and methodology 2.1 Humidified cavity-enhanced albedometer Figure 1 shows a schematic diagram of the H-CEA. The instrument consisted of a BBCES-albedometer and a RH controlled system. The albedometer combined BBCES with an integrating sphere (IS) for simultaneous in situ measurements of bext and bscat, babs and  in the same sample volume (Zhao et al., 2014;Xu et al., 2016;Xu et al., 2018aXu et al., , 2018b. bext was measured 25 over the range from 518 to 554 nm with a spectral resolution of 0.11 nm. bscat was measured as an integrated scattering coefficient over the spectral region from 528 to 537 nm (centered at 532 nm). Measurement of bext and bscat allow calculation of babs and  at a centre wavelength of 532 nm. Truncation reduction tubes narrowed the maximum truncation angle of the integrating sphere to 1.22º . Truncation losses are negligible for particle diameters below 1 μm. High-reflectivity mirrors (R = 99.971% at λ = 532 nm) were mounted at each end of the truncation reduction tubes. The flow rates of the sample (Fs) and 30 purified air near the front mirror (Fi, sample inlet part) and the end mirror (Fo, sample outlet part) were 1.45, 0.05 and 0.1 L https://doi.org/10.5194/amt-2020-6 Preprint. Discussion started: 27 January 2020 c Author(s) 2020. CC BY 4.0 License. min -1 , respectively. Some modifications of previous instruments have been made for RH dependent measurements, including (1) a temperature and RH sensor (T/RH sensor-2) inserted close to the inlet of the cavity cell to accurate measure the RH of the humidified sample, and (2) thermal insulation of the optical cavity, sampling tubes, and the Nafion humidifier to reduce the effect of changes in ambient temperature.
Data retrieval and calibration of the extinction and scattering measurements have been described elsewhere (Zhao et al., 5 2014;Xu et al., 2018a). Detection limits of optical parameters were determined from a Gaussian fit to the frequency distribution of a time series measurement of zero air (Xu et al., 2018a). With a 6 s acquisition time (an average of 150 individual spectra, each with 40 ms exposure time), the 3 detection limits for bext, bscat and babs were better than 0.81, 0.25 and 0.77 Mm -1 , respectively. The accuracy of the albedometer was evaluated with laboratory-generated polystyrene latex (PSL) spheres and ammonium sulphate aerosols. The total uncertainties in bext, bscat, babs and  measurements were estimated 10 to be 3.8%, 3.5%, 5.2% and 3.3%, respectively. Uncertainties in bext mainly arose from the calibrations of the mirror reflectivity (1 -R, ~2.5%) and effective sample length (RL, ~ 0.6%), as well as from particle losses. Uncertainty contributions associated with the scattering measurement include the scattering factor determination (< 2%), scattering truncation losses (< 1%, for submicron particles) and particle losses. Since measurements of bext and bscat were over mostly the same sample volume, the effect of particle losses on ω measurement was considered negligible. 15

Humidigraph system
The humidigraph system (as shown in Fig.1) consisted of a gas humidity-adjusting system that generated humid gas, and a second Nafion humidifier (MD-700-24S-3, Perma Pure) to humidify the aerosol sample. Dry zero air was divided into two paths separately controlled by two mass flow controllers (MFC): one was humidified with a Nafion humidifier (FC-125, Perma Pure, the water was supported by an automatic temperature-controlled water bath), and the other was used as dry 20 bypass air to adjust the RH of the air. A program controlled the mixing ratios of the humid air and bypass air, allowing the RH of the zero air to be varied from 2 -98% RH (monitored with T/RH sensor-1, Rotronic HC2-C05, with accuracies of ± 0.2 ºC and ± 1.5 % RH at 23 ºC ). The zero air was then used as the sheath air to humidify the sampling aerosol flow. The achievable RH of the aerosol sample ranged from 10-90% (as shown in Fig. 2) and was monitored with the T/RH sensor-2.
A cycle from 10% to 90% RH and back again with the humidigraph system took 15-20 min depending on experimental 25 conditions.

Laboratory test and model evaluation f(RH) measurement
The performance evaluation of H-CEA was carried out using laboratory-generated monodispersed particles of ammonium sulphate, sodium chloride and nigrosin. The aerosol generation system is shown in Fig. 1. Polydisperse aerosol particles were generated with a constant output atomizer (TSI 3076), dried in a diffusion dryer (TSI 3062), and then charged with an 30 aerosol neutralizer (TSI 3077) (Zhao et al., 2013;2014). A quasi-monodisperse size distribution of particles was selected using a differential mobility analyzer (DMA, TSI 3080L), diluted in dry zero air, and transferred to the H-CEA and a https://doi.org/10.5194/amt-2020-6 Preprint. Discussion started: 27 January 2020 c Author(s) 2020. CC BY 4.0 License. condensation particle counter (CPC, TSI 3076) for measuring optical properties and particle number concentration, respectively. A scanning mobility particle sizer (SMPS) in front of the H-CEA measured the size distribution of size-selected particles.
The extinction, scattering and absorption coefficients of size-selected particles can be calculated using the following equation (Pettersson et al., 2004): where N(Dp) is the number concentration of particles in size bin dDp with mean diameter Dp. m = n + i k is the CRI of the particles, where n and k are the real and imaginary parts of the CRI, respectively. x = πDp/ is the size parameter. are the extinction, scattering or absorption cross sections (ext,scat,abs) for a particle of given size and can be obtained with the measured bext,scat,abs and N(Dp) (ext,scat,abs = bext,scat,abs / N(Dp)). For chemically homogeneous 10 spherical particles, ext,scat can be calculated using Mie theory.
The model calculation of f(RH)ext,scat was similar to the method described in Zhao et al. (2017). The method was as follows: (a) the measured size distribution of size-selected particles was used to calculate a normalized size distribution; (b) the hygroscopic growth factors for a particular diameter Dp (GF(RH, Dp)) were calculated with the Extended Aerosol  (Querry, 1987) for water, ammonium sulphate and sodium chloride at λ = 532 nm, respectively.

Particle wall loss
Particle losses in the H-CEA were evaluated using laboratory-generated ammonium sulphate particles by measuring the number concentrations of size-selected particles (as shown in Fig. 3). Losses considered included particle loss in the tube, in https://doi.org/10.5194/amt-2020-6 Preprint. Discussion started: 27 January 2020 c Author(s) 2020. CC BY 4.0 License.
the Nafion humidifier and in the cavity cell. For particle diameters above 40 nm, the literature reported losses in the Nafion humidifier were estimated to be less than 1% (Bohensky et al., 2014). For particles larger than 100 nm, the measured total particle loss in the H-CEA was less than 7%, with a < 1% loss in the sample tube and Nafion humidifier, and < 6% loss in the cavity cell. The larger particle losses for small particles were mainly caused by the Brownian diffusion to the walls of the tube and cavity cell (von der Weiden, Drewnick, and Borrmann 2009). For ambient measurements, particle losses 5 contributed less than 3% uncertainty to the extinction and scattering measurements.

Effect of water vapour on optical signal
To confirm that the influence of water vapour on the measurements of bext, bscat and babs was negligible, the relative changes in light intensity (ICCD), scattering intensity (IPMT), and bext,scat,abs for humidified air were measured during a full RH cycle. As shown in Fig. 4a, ICCD did not show obvious changes with RH. IPMT increased slightly at high RH, but the overall increment 10 in IPMT was less than 2.5% for RH ranging from 10% to 88%. The calculated values of bext, bscat and babs did not show obvious changes with RH (Fig. 4b). The respective standard deviations were 0.15, 0.15 and 0.33 Mm -1 and were within the precision of the albedometer. These results demonstrated that water vapour content had no statistically significant effect on the measurement of bext, bscat and babs.

Uncertainty analysis for f (RH)ext,scat,abs, measurements 15
A power-law dependence of bext and bscat on RH is commonly used to describe hygroscopic behaviour (f(RH)ext,scat = (1 -RH/100) - ) (Quinn et al., 2005;Massoli et al., 2009a). Using this parameterization, Titos et al. (2016) comprehensively analysed the uncertainty in f(RH)scat measurements with Monte Carlo simulations (Saltelli et al., 2000). We used a similar method in our study to evaluate the uncertainty in f(RH)ext,scat measurements with H-CEA. The uncertainty in f(RH) mainly arose from the uncertainties in the extinction and scattering measurements and the measurement of RH. The relative 20 uncertainties of these properties are denoted δf(RH), δbext,scat, and δRH. These properties can be described with the following expressions: In the simulation, δbext and δbscat were the relative uncertainties of bext and bscat with values of 3.8% and 3.5% respectively.
bext and bscat were randomly selected from 10 to 1000 Mm -1 by taking into account upper and lower limits of the uncertainties (ranging from 1+δbext,scat to 1-δbext,scat). Values of bext,scat of the aerosol sample in wet and dry conditions (Dry = 20% RH) were calculated with the power-law functions (Eqs. 5 and 6). The relative uncertainty of RH measurement (δRH ~ 1.5%) contributed to the total uncertainties of bext,scat, which will be enlarged with the increment of γext and γscat. hygroscopic parameter  and RH with a reference RH of 20%. In the simulation, RH was varied from 20% to 95% with a step of 1%, and the hygroscopic parameter  was varied from 0 to 1 with a step of 0.01. The average errors of f(RH)ext and f(RH)scat were estimated to be 11% and 10% for RH in the range from 20% to 95%. For moderately hygroscopic aerosols ( ~ 0.5), the uncertainties of f(RH)ext and f(RH)scat ranged from 7% to 25%. The uncertainties reported of our H-CEA 5 instrument were about half those of the humidified nephelometer (20-40% for f(RH)scat) reported by Titos et al. (2016). A power-law dependence of babs on RH has not been reported in previous studies and is not expected. That is because condensation of water or other nonabsorbing species onto a particle does not increase the total amount of absorbing molecules on that particle, unlike the contribution of condensing species to particle scattering. For this reason, the Monte Carlo simulation was not used to model the uncertainty in f(RH)abs. If any power-law relationship exists, the expected values 10 of γabs and γω were much smaller than γext,scat, the uncertainty of f(RH)abs and f(RH) should be less than f(RH)ext and f(RH)scat.

Laboratory results
A series of laboratory experiments were conducted to evaluate the performance of the H-CEA. For measurement of f(RH)ext,scat, size-selected nonabsorbing inorganic particles (ammonium sulphate and sodium chloride) were used. The results were compared with the model calculations and values reported in the literature. Measurement of f(RH)abs, was assessed 15 using size-selected strongly absorbing organic particles (nigrosin). Results were compared to previous findings in the literature. Figure 6 shows the RH dependent extinction and scattering for the size-selected ammonium sulphate particles of 200, 250, 300 and 350 nm (dry diameter, with RH of ~ 20%, the same as below for sodium chloride and nigrosin). The black circles 20 and red dots represent the measured extinction and scattering hygroscopic enhancement (f(RH)ext,scat,m). The blue lines represent the model calculation results with Mie theory using the theoretical GF(RH) from the E-AIM model (Fierz-Schemidhauser et al., 2010;Michel Flores et al., 2012). There were no significant differences in the values of f(RH)scat and f(RH)ext for particles of four different diameters, and their values in the range of ~10% to 77% RH were close to 1, consistent with model calculations. The sudden increase in f(RH)ext,scat was due to the deliquescence of ammonium sulphate particles. 25

Ammonium sulphate and Sodium Chloride
The measured deliquescence RH (DRH) of ammonium sulphate was 77-78 %, close to the literature reported value (79.9 % RH at 298 K) (Tang and Munkelwitz, 1993). The small difference may arise from slight temperature and RH differences at different parts of the system as well as the accuracy of the T/RH sensor (±1.5%). Table 1 Table 1). Figure 7 shows the f(RH)ext,scat values for size-selected 200, 250, 300 and 350 nm sodium chloride particles. The measured DRH of sodium chloride was ~73 %, which was close to the value reported in the literature (75.3% RH at 298 K) 5 (Tang and Munkelwitz, 1993). Table 1 lists the measured f(RH)ext/scat at 80% and 85% RH. The f(RH)scat values for 300 nm particles were larger than the values reported by Fierz-Schemidhauser et al. (2010) using a humidified nephelometer instrument. This difference likely arises from the accuracy of our RH measurement of our sample. Our deliquescence points are about 2% lower RH than literature values, suggesting that the actual RH in our system was marginally higher than 80% or 85%. As f(RH)ext/scat shows a strong positive dependence on RH, the actual values are likely to be lower in our instrument. 10 This effect is most marked for sodium chloride particles where the enhancement factor (0.3 at 85% RH, 300 nm particles) is about double that of the ammonium sulphate particles (0.1 at 85% RH, 300 nm particles). In addition, differences in the particle size distribution between our study and that of Fierz-Schemidhauser et al. (2010) may also contribute to the observed difference in f(RH)scat. The f(RH)ext,scat values of sodium chloride particles were in good agreement with the model calculations, which were obviously larger than those of ammonium sulphate. 15

Nigrosin
The performance evaluation of the H-CEA for f(RH)abs and f(RH) measurements was demonstrated with laboratory generated nigrosin aerosol, which absorbs strongly at 532 nm (Xu et al., 2018a). Figure 8 shows the measured f(RH)ext,scat,abs, for size-selected 200 and 300 nm nigrosin particles. For 200 nm nigrosin particles, the measured f(RH)ext, f(RH)scat, f(RH)abs and f(RH) at 80% RH were 1.22±0.00, 1.34±0.00, 1.12±0.01 and 1.10±0.00, respectively. These values 20 were 1.26±0.01, 1.34±0.02, 1.18±0.01 and 1.06±0.01, respectively, for 300 nm nigrosin particles. The measured f(RH)scat of nigrosin particles at 200 and 300 nm were roughly the same, however, the measured f(RH)ext,abs, were slightly different for these two particle sizes, which showed a hygroscopic increase in absorption. In this study, the measured f(RH)ext values at 80% RH were comparable with Michel Flores et al.'s results (1.18±0.06 for 200 nm particles, and 1.19±0.04 for 300 nm particles) using a humidified CRDS instrument (Michel Flores et al., 2012). The measured f(RH)abs value in this study was also 25 consistent with the value (~ 1.14) reported by Brem et al. (2012). f(RH) shown a monotonic increase with RH, which may be due to a decrease in the imaginary part of CRI.

Conclusions
In this paper, we report the development and characterization of a humidified cavity-enhanced albedometer (H-CEA) for the simultaneous measurements of multiple optical hygroscopic parameters (f(RH)ext,scat,abs,ω) of the same sample. Monte Carlo 30 simulations were used to evaluate the uncertainties in f(RH)ext,scat measurements. For moderately hygroscopic aerosols, the https://doi.org/10.5194/amt-2020-6 Preprint. Discussion started: 27 January 2020 c Author(s) 2020. CC BY 4.0 License.