Measuring Light Absorption by Organic Aerosols: Correction Factors for Solvent

11 Recent studies have shown that organic aerosol (OA) could have a non-trivial role in 12 atmospheric light absorption at shorter visible wavelengths. Good estimates of OA absorption 13 are therefore necessary to accurately calculate radiative forcing due to these aerosols in climate 14 models. One of the common techniques used to measure OA light absorption is the solvent 15 extraction technique from filter samples which involves the use of a spectrophotometer to 16 measure bulk absorbance of the solvent-soluble organic fraction of particulate matter. Measured 17 bulk absorbance is subsequently converted to particle-phase absorption coefficient using 18 correction factors. The appropriate correction factors to use for performing this conversion under 19 varying scenarios of organic carbon (OC) to total carbon (TC) mass ratios has been an 20 unexplored area of research. The conventional view is to apply a correction factor of 2 for water21 extracted OA based on Mie calculations. 22 Here, we performed a comprehensive laboratory study involving three solvents (water, methanol, 23 and acetone) to investigate the corrections factors for converting from bulk-to-particle phase 24 Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2018-1200 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 21 November 2018 c © Author(s) 2018. CC BY 4.0 License.

. The atmospheric 48 mass of OC can be 3-12 times larger than that of BC (Husain et al., 2007;Zhang et al., 2008) 49 which warrants its inclusion as an atmospheric light absorber. Only recently have a few global 50 modeling studies incorporated radiative forcing by organic aerosol (OA) absorption (Chung et 51 al., 2012;Feng et al., 2013;Lin et al., 2014). Thus, having accurate estimates for OA absorption 52 is necessary to help improve climate models.

53
A convenient and prevalent methodology of measuring OA absorption is based on collecting 54 aerosol particles on a filter substrate followed by extracting the organic compounds into a 55 solvent. This is a good analytical method used in many studies as it excludes any interference 56 from EC and provides only the OC absorption spectra (Mo et al., 2017;Chen and Bond, 2010;57 Liu et al., 2013). The absorbance of the organic chromophores in the solvent extract is measured 58 using an ultraviolet-visible (UV-Vis) spectrophotometer and the imaginary part of the complex 59 refractive index of the solvent is calculated using the measured absorbance and extract 60 concentrations. The absorbance values can be converted to corresponding bulk phase absorption 61 coefficients (b abs,bulk ), and the imaginary index along with an assumed number size distribution 62 can be used as inputs to Mie theory for calculating the absorption coefficient for the dissolved 63 OC particles. Past studies have suggested that b abs,bulk is not representative of the corresponding 64 particulate-phase organic aerosol absorption coefficient (b abs,OA ) (Liu et al., 2013;Moosmüller et 2013;Liu et al., 2016;Liu et al., 2014;Washenfelder et al., 2015). These correction factors have 71 been deemed relatively independent of particle size as long as the diameters do not get much 72 smaller than the wavelength of light, in which case their values fall to a range of 0.69 -0.75 (Sun 73 et al., 2007). Further, these factors were calculated assuming homogeneous composition and 74 external mixing states of the OC aerosol in addition to fixed values of the real part of the 75 refractive index and effective density. No attempts have been made to quantify the variation in 76 these correction factors with varying aerosol intrinsic properties, such as the single scattering 77 albedo, and EC/OC ratios, even though these properties influence the types and fractions of 78 organics extracted by a given solvent (Zhang et al., 2013;Saleh at al., 2014) thereby affecting 79 b abs,bulk .

80
In-situ measurement of particulate-phase absorption coefficient is commonly and accurately 81 accomplished using a photoacoustic spectrometer (PAS) (Lack et al., 2006;Arnott et al., 2005;82 Arnott et al., 2003). However, a single-wavelength PAS cannot distinguish between absorption 83 by OC and BC aerosol and it typically measures the total particle-phase absorption coefficient 84 (b abs,tot ) of the aerosol population in the cell (Moosmüller et al., 2009). One must make use of a 85 multi-wavelength PAS using which b abs,OA could be separated out from that of BC, based on the 86 Here, we burnt a range of different biomass fuels under different combustion conditions and the 93 resulting aerosol emissions were passed through various in-situ instruments while simultaneously 94 being collected on quartz-fiber filters. The particle phase absorption coefficient was obtained 95 using integrated photoacoustic-nephelometer spectrometers (IPNs) at wavelengths 375, 405 and 96 1047 nm. Organics collected on quartz-fiber filters were extracted in water, acetone, and 97 methanol, and corresponding b abs,bulk values were calculated. These values were compared with 98 corresponding b abs,OA , and the change in b abs,OA /b abs,bulk with varying single scattering albedo 99 (SSA) values and OC/TC ratios was established. SSA was parametrized with the OC/TC ratios 100 with trends similar to those observed by Pokhrel et al., (2016). AÅE from spectroscopic data for 101 bulk and particle phase measurements were compared, and the Mie Theory based correction 102 factor was also verified for a few samples. stainless-steel combustion chamber housing a fan for mixing and recirculation (Sumlin et al.,107 2018b). Aerosol samples were generated by burning several types of biomass including pine, fir, 108 grass, sage, and cattle dung (sources are given in the Supplementary Information). During a ranged from 0.55-1, and SSA values ranged from 0.56-0.98 for wavelengths of 375, 405, and 115 1047 nm.

116
For one set of experiments, the particles were directly sampled from the chamber; in another set, 117 the sampling was done from a hood placed over the burning biomass. A diffusion dryer removed 118 excess water from the sample stream, and the gas-phase organics were removed by a pair of 119 activated parallel-plate semi-volatile organic carbon (SVOC) denuders. The gas-phase organics 120 were stripped to reduce artifacts produced by the adsorption of organic vapors on the quartz 121 filters. The aerosols were finally sent to a 208-liter stainless-steel barrel, from which they were 122 continuously sampled by the three IPNs and a scanning mobility particle sizer (SMPS, TSI, Inc.).

123
The SMPS was not used in all the experiments, however, it gave an estimate of the range over Quartz filters (Pallflex Tissuquartz) collected during sampling were split into four quarters, and 138 each quarter was extracted using either deionized water, acetone, hexane, or methanol. The 139 absorption by hexane extracts were low and prone to errors, so data for its extracts were not 140 analyzed. The filters were placed in a covered beaker along with 3-5 ml of the solvent for 24 141 hours. The solvent volumes were measured both before and after the extraction and the 142 differences between the two measurements were within 8%. The extracts were then passed 143 through syringe filters with 0.22 μm pores to remove any impurities introduced by the extraction 144 process.

145
The light absorbance of the extracts was measured using a UV-Vis spectrophotometer (Varian

151
where V l is the volume of solvent the filter was extracted into, V a is the volume of air that passed 152 over the given filter area, and l is the optical path length that the beam traveled through the 153 cuvette (1 cm). The absorbance at a given wavelength is normalized to the absorbance at 700 nm 154 to account for any signal drift within the instrument. The resulting absorption coefficient (m -1 ) 155 was multiplied by ln(10) to convert from base 10 (provided by the UV-Vis spectrophotometer) to 156 natural logarithms.  BC light absorption coefficient at shorter wavelengths (b abs,BC (λ 1 )) was calculated by:

165
where λ 1 is the wavelength at which the absorption will be calculated and AÅE is defined for a  OC/TC ratios were assumed to be constant for a given steady state IPN reading, which allowed 178 us to relate the absorption data to the OC/TC data. solution. The calculated α/ρ was then used to determine k for the WSOC by: algorithm was used to extract the real part of the refractive index (n) using data from the SMPS 198 and IPN (Sumlin et al., 2018a). A sensitivity analysis was performed by varying the n value from 199 1.4 to 2, and the change in Mie calculated absorption was within 18%. The size distribution for 200 the WSOC was estimated assuming the same geometric mean and standard deviation as that of 201 the original aerosol, but with number concentrations calculated based on the extracted mass.

202
Calculations for the number concentration are provided in the SI.

203
After the size distribution and complex refractive index were determined, they were used to 204 calculate the absorption coefficient based on Mie Theory, which was then compared to b abs,bulk to 205 determine Mie based correction factors for converting from bulk to particle phase absorption.       We also fit an exponential curve to b abs,OA (λ)/b abs,bulk (λ) as a function of EC/OC values (y = k 0 + 272 k 1* (1-exp(-(EC/OC)/k 2 ))). The fits had comparable RMSE values, but because strong 273 relationships between the SSA and OC/TC ratios were previously established, the plots described

287
The AÅE values, for organics extracted in different solvents and those obtained from b abs,OA are 288 compared in for OC in both the particle-and bulk-phase decreased at lower OC/TC ratios indicating a drastic 298 decrease in the spectral dependence of OC co-emitted with BC. Table 4 shows that AÅE values 299 for OA were greater than or close to their bulk counterparts at OC/TC ratios close to 1. These 300 bulk measurements of AÅE suggest that they deviate significantly from the spectral dependence 301 of OC in the particle phase, and future studies and models should not use AÅE data from bulk 302 measurements to be representative of the particle phase.

304
The absorption coefficient determined from the bulk absorbance using Eq. (1) was compared to