Light absorption properties of laboratory-generated tar ball particles

Tar balls (TBs) are a specific particle type that is abundant in the global troposphere, in particular in biomass smoke plumes. These particles belong to the family of atmospheric brown carbon (BrC), which can absorb light in the visible range of the solar spectrum. Albeit TBs are typically present as individual particles in biomass smoke plumes, their absorption properties have been only indirectly inferred from field observations or calculations based on their electron energy-loss spectra. This is because in biomass smoke TBs coexist with various other particle types (e.g., organic particles with inorganic inclusions and soot, the latter emitted mainly during flaming conditions) from which they cannot be physically separated; thus, a direct experimental determination of their absorption properties is not feasible. Very recently we have demonstrated that TBs can be generated in the laboratory from droplets of wood tar that resemble atmospheric TBs in all of their observed properties. As a follow-up study, we have installed on-line instruments to our laboratory set-up, which generate pure TB particles to measure the absorption and scattering, as well as the size distribution of the particles. In addition, samples were collected for transmission electron microscopy (TEM) and total carbon (TC) analysis. The effects of experimental parameters were also studied. The mass absorption coefficients of the laboratory-generated TBs were found to be in the range of 0.8–3.0 m g at 550 nm, with absorption Ångström exponents (AAE) between 2.7 and 3.4 (average 2.9) in the wavelength range 467–652 nm. The refractive index of TBs as derived from Mie calculations was about 1.84− 0.21i at 550 nm. In the brown carbon continuum, these values fall closer to those of soot than to other light-absorbing species such as humic-like substances (HULIS). Considering the abundance of TBs in biomass smoke and the global magnitude of biomass burning emissions, these findings may have substantial influence on the understanding of global radiative energy fluxes.

of their absorption properties is not feasible. Very recently we have demonstrated that TBs can be generated in the laboratory from droplets of wood tar that resemble atmospheric TBs in all of their observed properties. As a follow-up study we have installed on-line instruments to our laboratory set-up generating pure TB particles to measure the absorption and scattering, as well as size distribution of the particles. In addition, 15 samples were collected for transmission electron microscopy (TEM) and total carbon (TC) analysis. The effects of experimental parameters were also studied. The mass absorption coefficients of the laboratory generated TBs were found to be in the range of 0.8-3.0 m 2 g −1 at 550 nm, with absorption Ångström exponents (AAE) between 2.7 and 3.4 (average 2.9) in the wavelength range 467-652 nm. The refractive index of 20 TBs as derived from Mie calculations was about 1.84-0.21i at 550 nm. In the brown carbon continuum these values fall closer to those of soot than to other light-absorbing species such as humic-like substances (HULIS). Considering the abundance of TBs in biomass smoke and the global magnitude of biomass burning emissions, these findings may have substantial influence on the understanding of global radiative energy 25 fluxes. 16216

Introduction
Tar balls (TBs) are abundant and represent a peculiar particle type emitted from biomass burning. They can be readily identified by transmission electron microscopy (TEM) by their morphology, chemical composition, and amorphous structure. TBs are homogeneous and spherical particles that can withstand the high-energy electron 5 beam of the TEM. They are most often present in external mixture, i.e. as individual standalone particles. Their sizes range from 30 to 500 nm in optical diameter as determined by TEM (Pósfai et al., 2004;Cong et al., 2009;Adachi and Buseck, 2010;Fu et al., 2012;China et al., 2013). Very recently we have demonstrated that TBs can be generated in the laboratory from droplets of wood tar that resemble atmospheric 10 TBs in all of their observed properties (Tóth et al., 2014). These particles belong to the family of atmospheric brown carbon (BrC) which can absorb light in the visible range of the solar spectrum . Chung et al. (2012) have estimated that the global contribution of BrC to light absorption may be as high as 20 % at 550 nm. Given that the estimated contribution of humic-like substances (HULIS) to 15 solar absorption can be only few per cent at 500 nm (Hoffer et al., 2006), a substantial fraction of BrC absorption may be attributed to TBs. So far a direct experimental determination of absorption properties of TBs has not been feasible because in biomass smoke TBs coexist with various other particle types from which they cannot be separated. Thus, their absorption properties have been so far only indirectly inferred from 20 field observations (Hand et al., 2004;Chakrabarty et al., 2010) or calculations based on their dielectric functions obtained from electron energy-loss spectrometry (Alexander et al., 2008). Hand et al. (2004) were the first to estimate the optical properties of TB by measuring the optical properties of ambient particles emitted from biomass burning during 25 the YACS (Yosemite Aerosol Characterization Study) conducted from July to September 2002 in the western United States. The derived (estimated from OC/EC and scattering data) ensemble complex index of refraction of TBs was found to be 1.56-0. at 632 nm, indicating that the TBs do absorb light. The difference between the measured absorption between 370 and 880 nm was the highest in periods when TBs were the predominant particle type, suggesting that the absorption Ångström exponent of TB was different from 1. Back-trajectory analyses showed that the particles measured were affected by long-range transport, thus the residence time of the particles allowed 5 photochemical and ageing processes to take effect. These effects can be observed on the distribution of the elements in individual TB particles. Whereas carbon and nitrogen were homogenously distributed over the entire particle volume, the abundance of oxygen was strongly enhanced in the ∼ 30 nm outmost shell of the particles. Since these particles were affected by atmospheric processing, some of their properties might be 10 different from those of the freshly emitted TB particles. Alexander et al. (2008) investigated individual particles ("carbon spheres") from ambient aerosols collected above the Yellow Sea during the Asian Pacific Regional Aerosol Characterization Experiment (ACE-Asia). The morphological properties (size, structure, and mixing state) of the carbon spheres observed by TEM were similar to 15 those characteristic of TB particles. The refractive indices of individual carbon spheres were derived from theoretical calculations based on electron energy-loss spectra and were found to be centred around 1.67-0.27i at 550 nm. The authors also calculated the wavelength dependence of the absorption and found AAE of 1.5 which is not much different from that reported for BC (Schnaiter et al., 2003;Moosmüller et al., 2009). 20 The derived mass absorption coefficients of the carbon spheres were in the range of 3.6-4.1 m 2 g −1 , almost as high as those of BC (4.3-4.8 m 2 g −1 ) (Alexander et al., 2008). Chakrabarty et al. (2010) measured the optical properties of tar balls from smoldering combustion of Ponderosa pine and Alaskan Pine duffs in the laboratory. They found the index of refraction of TB particles similar to those of humic-like substances 25 (Hoffer et al., 2006). The wavelength dependent absorption Ångström exponents were split into 2.3-2.8 and 4.2-6.4 in the spectral range of 532-780 nm and 405-532 nm, respectively. The TB particles were almost spherical, having a carbon-to-oxygen ratio of about 6, as determined by SEM-EDX. Introduction The absorption properties of BrC including TBs are very important in regional and global modelling of the radiative budget, as well as in interpreting satellite-based radiation measurements. In spite of being an abundant particle type among BrC particles, TBs have so far eluded direct measurements of their optical properties, since they always coexist with other particle types and UV-absorbing gaseous species in biomass 5 smoke. By generating pure TB particles in the laboratory without concurrent emission of other combustion products, we have directly measured the optical properties of TBs for the first time in aerosol science. In this paper we report the fundamental optical properties of laboratory generated TBs generated under different conditions.

10
For particle generation liquid tar was produced by dry distillation of wood, as described in our previous paper (Tóth et al., 2014). Dry European turkey oak wood (Quercus cerris) chops (25 mm × 10 mm × 10 mm) were placed in a Kjeldahl flask (100 mL) fixed above a Bunsen burner in a slightly down-tilted position. Liquid condensate was collected in a 40 mL vial in which it separated into "oily" and "aqueous" phases (Maschio et al., 1992). Both phases were aged further on a ∼ 300 • C plate to concentrate the solutions. The concentrates were taken up with high purity methanol (J. T. Baker, HPLC Gradient), the concentration of solutions used for particle generation were 1-3 g L −1 .
A modified experimental setup similar to that used in previous experiments (Tóth et al., 2014) was applied for particle generation. In order to maintain the concentration 20 of the generated particles constant for a longer time that is necessary for measuring the size distribution and optical properties of the particles, particles were generated with an ultrasonic atomizer used in a fashion described previously by Okuyama and Lenggoro (2003). The production of tar droplets from their solution in methanol was performed in a plastic flask placed above the ultrasonic nebulizer (1.6 MHz, Exo Terra 25 Fogger, PT2080, Rolf C. Hagen Corp.), held in a water bath at room temperature. The nebulizer flask was continuously rinsed with purified nitrogen (Messer, purity 99.5 %) ACPD 15,[16215][16216][16217][16218][16219][16220][16221][16222][16223][16224][16225][16226][16227][16228][16229][16230][16231][16232][16233][16234]2015 Light absorption properties of laboratory generated tar ball particles  (Anderson et al., 1996). The data were recorded with a time resolution of 5 s, the raw light absorption and scattering data were corrected 15 according to Bond et al. (1999) and Anderson and Ogren (1998), respectively. All data were also corrected for standard temperature and pressure. The absorption Ångström exponents of the particles were calculated from the measured and corrected absorption coefficient for the wavelength range between 467 and 652 nm with the equation (Moosmüller et al., 2011): where A 467 and A 652 are the absorbances measured at the two different wavelengths. The size distribution was measured in the range of 7-800 nm with a Differential Mobility Particle Sizer (DMPS), constructed at the University of Helsinki.
The generated particles were collected on Whatman QMA quartz filters (pre-baked 25 at 680 • C for 6 h). The elemental composition (CHNS) of the particles on filters was measured by elemental analyser (EuroVector EA3000 Pella Inc., USA) fixed on 13.1 mm spots of quartz filters placed in the filter holder that were used for sampling for elemental analysis as well.
The morphologies of the particles were studied in bright-field TEM images obtained using a Philips CM20 TEM operated at 200 kV accelerating voltage. The possible presence of an internal structure was checked in high-resolution electron micrographs and 5 in selected-area electron diffraction patterns. The electron microscope was equipped with an ultra-thin-window Bruker Quantax X-ray detector that allowed the energydispersive X-ray analysis (EDS) of the elemental compositions of individual particles. Spectra were acquired for 60 s, with the diameter of electron beam adjusted to include the entire individual TB particles.

Morphology, elemental composition and structure of the generated particles
Two samples were collected for TEM analysis to investigate the morphology and elemental composition of the generated particles, one representing the particles gen-15 erated from the aqueous phase of the tar, whereas the other was collected from the oily phase. In both cases the oven temperature was set to 650 • C, the flows and other experimental parameters were similar to those applied for samples collected for TC analysis.
As it can be observed in Fig. 1, the particles generated from the aqueous phase 20 were spherical. From the oily phase more irregularly shaped particles with oval twodimensional outlines were produced, indicating that in the latter case the particles were not perfectly solid at the time of collection. It was observed during the TEM analysis that all of the generated particles can withstand the high-energy electron beam of the instrument: they did not evaporate or shrink while exposed to the electron beam. Introduction The observed sizes of the particles vary widely (up to ∼ 360 nm in diameter), the number size distribution peaks at ∼ 100 nm as determined from the TEM images. Bimodal number size distribution was obtained from the DMPS measurements for the particles produced from both the aqueous and oily phase of the tar (Fig. 2). For the particles aged at 650

Conclusions References
• C the two modes are centred around 20-40 and 100-140 nm.

5
The number size distributions of nigrosin and the blank (pure methanol) are unimodal peaking at 117 and 41 nm, respectively. Nevertheless, in cases when the ageing temperature was higher than 500 • C the mass and volume of the particles are dominated by the larger particles: at least 86 and 70 % of the total mass is represented by the larger particle mode in the case of the aqueous and oily samples, respectively. Considering 10 that both the absorption and scattering efficiencies are very small for small particles, the optical properties are also determined by the particles of the larger mode. (Here we note that the mass absorption coefficient was calculated only for size distributions in which the relative contribution of the second mode to total volume was larger than 93 %.) The sizes of the particles of the second mode were similar to those determined 15 for ambient TB particles observed in samples from K-puszta and Southern Africa (Pósfai et al., 2004). The EDS spectra of the particles generated from both the aqueous and oily phase indicated that the particles consist predominantly of carbon and oxygen. In the case of the particles formed from the aqueous phase the average carbon to oxygen molar 20 ratio was 10 : 1, with 90 mol % C (RSD = 10 %), 9 mol % O (RSD = 16 %) and N, Na, Si, S, K only in trace amounts. The limitations of determining molar ratios by this method are described in detail elsewhere (Pósfai et al., 2003). It should be noted that the spectra were practically indistinguishable from those obtained from atmospheric TBs. Both HRTEM images and electron diffraction confirm that the particles in both samples are perfectly amorphous, lacking even the short-range order that is characteristic of ns-soot (Buseck et al., 2014

Measurement uncertainties
In order to estimate the measurement uncertainties, nigrosin dye (Sigma-Aldrich, Acid black 2, water soluble) was measured with the same setup that was used for the measurements of TBs. The nigrosin was also dissolved in methanol and particles were generated with the process used for the TB samples. Oven temperature was set to 5 65 • C in order to evaporate methanol from the droplets without inducing compositional changes of nigrosin. According to Massoli et al. (2009), the scattering coefficient of absorbing particles with single scattering albedo (SSA) = 0.4 (at 532 nm) is overestimated by 25 % using the Anderson and Ogren correction (Anderson and Ogren, 1998) for the raw data measured by a TSI nephelometer. Since in our case the SSA of the generated nigrosin was ∼ 0.4 at 550 nm, the scattering coefficient might be also overestimated by ∼ 25 %. The uncertainty of the measurements of Particle Soot Absorption Photometer (PSAP) whose measurement principle is very similar to that of the CLAP is 20-30 % (Bond, 1999). It was demonstrated that the presence of organic compounds (secondary organic aerosol, SOA) causes positive bias and enhances the uncertainty 15 of the PSAP (Cappa et al., 2008;Lack et al., 2008). This effect has to be considered in the case of particles generated from tar which contain condensable organic compounds as well. Based on the above, if we consider that the CLAP overestimated the absorption of nigrosin by 25 % and the scattering is also overestimated by 25 %, we obtain a refractive index of 1.65-0.29i and 1.77-0.27i for nigrosin at wavelengths of 550 20 and 652 nm, respectively. The index of refraction of nigrosin at 633 nm was reported to be 1.67-0.26i (Pinnick et al., 1973). By assuming that the absorption is similar at both 633 and 652 nm, Mie calculations using the refractive index of nigrosin (1.67-0.26i ) and the measured size distribution yield scattering and absorption coefficients at 652 nm higher by ∼ 17 % and lower by ∼ 2 %, respectively, as compared to those Introduction  Table 1 summarizes the measured optical properties of the particles produced from the aqueous phase. At 650 • C the measured mass absorption coefficients of the TBs generated from the aqueous phase of the wood tar varied between 2.4 and 3.2 m 2 g −1 C, the average being 2.7 m 2 g −1 C at 550 nm. Taking into account the potential positive bias in 5 absorption measurements (see discussions in Sect. 2) and related uncertainties (e.g. uncertainty of total carbon measurements) and the fact that the mass-to-carbon ratio of TBs is about 1.2 this range translates into mass absorption coefficients of about 0.8-3.0 m 2 g −1 (see Table 1). These values are similar to that characteristic for BC (∼ 7 m 2 g −1 , Schnaiter et al., 2003;Clarke et al., 2004;Taha et al., 2007). The mass 10 absorption coefficient of HULIS is a factor of 25-100 lower (∼ 0.032 m 2 g −1 , Hoffer et al., 2006). The range of the measured mass absorption coefficients for the particles generated from the oily phase of wood tar was found to be largely similar to that obtained for the particles from the aqueous phase. However, the former is not evaluated since the 15 particles generated from the oily phase morphologically differ from atmospheric TB particles.

Ångström exponent of generated tar balls
The absorption Ångström exponents of particles generated varied between 2.7 and 3.7 (2.7-3.4 and 3.1-3.7 for the aqueous and oily phase, respectively) in the spectral  . The lack of highly ordered structures in laboratory-generated TB particles as observed by HRTEM, and the carbon-to-oxygen ratios measured by EDS reasonably explain the measured Ångström exponents. These values fall between those of BC and humic-like substances, the less polar fraction of the water soluble fraction of the aerosol (Hoffer et al., 2006).

5
The optical properties of the generated TB particles were measured continuously while the oven was gradually cooled from the temperature of ∼ 650 • C. In the case of particles generated from the aqueous phase of the tar the Ångström exponent did not change significantly down to about 550 • C, below which it drastically increased.
The same phenomenon was observed for particles generated from the oily phase, 10 the Ångström exponent (and also the SSA, not shown in the figure) changed rapidly only below a certain oven temperature (< 580 • C). This finding implies that the optical properties of tar balls are markedly different from those of the bulk tar material, and suggests that the chemical transformations induced by heat shock or atmospheric ageing that produce rigid and refractory spherical particles also significantly alter the 15 absorption properties of the resulting TB particles.

Index of refraction of tar ball particles
The indices of refraction of particles generated from the aqueous phase and aged at 650 • C were calculated based on the method of Guyon et al. (2003), also used in Hoffer et al. (2006). Since the SSA of TB samples varied between 0.4 and 0.5 (at 550 nm) the 20 measured absorption and scattering coefficients were corrected as described for the nigrosin particles. The obtained index of refraction was 1.94-0.21i (at 550 nm). Based on the nigrosin measurements, if we assume that the measured scattering coefficient is overestimated by a further 17 %, the real part of the obtained index of refraction can be considered as an upper limit, as it is overestimated by about 5 %, the average value is 1.84-0.21i. This is comparable to the complex refractive index of individual carbon spheres -in particular in its imaginary part -calculated from TEM electron energy loss Introduction  (Alexander et al., 2008). The real part of the index of refraction as measured in our experiment is higher by about 10 % than the one calculated for the carbon spheres.

Conclusions
Tar balls have been shown to be abundant in the troposphere impacted by biomass smoke which is now the main global source of anthropogenic aerosol particles. The 5 contribution of TBs to the number concentration of particles could be as high as 80 % in the vicinity of biomass burning sources (Pósfai et al., 2003), while it was in the range of 6-14 % away from the sources (Adachi et al., 2011), as observed using TEM. At a site that represents regional background conditions (K-puszta) the abundance of TBs varied from 0 to 40 % depending on the season and time of sampling (Pósfai et al., 10 2004). Even over the Himalaya TB particles accounted for 3 % of all observed particles (Cong et al., 2009). Near the Arctic, in Hyytiälä during a pollution episode 1-4 % of the particles were identified as TBs (Niemi et al., 2006). Given the abundance of TBs in the global troposphere and their relatively high absorption efficiency over the entire solar spectrum, their contribution to column absorption can be clearly significant. This is particularly true for immense geographical regions impacted by Atmospheric Brown Clouds (ABCs) where TBs may make a contribution to solar absorption comparable to that of BC. The last question that remains is where TBs are positioned in the blackto-brown carbon continuum of atmospheric aerosols Sun et al., 2007). Somewhat surprisingly, their optical properties suggest that they are 20 not very far from BC or amorphous carbon, despite their markedly different formation mechanism and chemical composition. On the other hand, it is clear that TBs are very much different from faintly coloured species such as HULIS or SOA in their absorption properties. We suggest that TBs are on the dark side of brownness of aerosol carbon, but clearly out of the BC regime both in terms of their key absorption parameters (e.g. Introduction