Morphology and mixing state of aged soot particles at a remote marine free troposphere site: Implications for optical properties

2.1 Sampling Site and Measurements

The PMO is located in the summit caldera of the Pico Volcano (at 2225 m above sea level, 38.47°N, 28.40°W) in the Azores, Portugal. The mountaintop observatory (Figure S1 in the supporting information) is typically above the marine boundary layer [Honrath et al., 2004; Kleissl et al., 2007; Rémillard et al., 2012]. PMO often receives air masses from North America and sometimes from Africa or Europe, and it is an ideal site to study free tropospheric aerosol transported over long distances [Dzepina et al., 2014].

We performed retroplume analysis using the Lagrangian particle dispersion model Flexible Particle (FLEXPART) [Stohl et al., 2005; Owen and Honrath, 2009] to study the origin of air masses and trajectories to PMO. We selected two specific events to study soot particles with different ages and transport patterns representing: (a) recirculation over the ocean from the southwest Atlantic (event 1, 6 and 7 July 2012), and (b) transport from North America (event 2, 20 and 21 July 2012) (Figure 1). FLEXPART also provides an estimate of the plume age that was higher for event 1 (~15.7 days) than for event 2 (~9.5 days).

Particles were collected on nuclepore membranes [China et al., 2013], and individual particle morphology and mixing state were studied using a field emission SEM (Hitachi S-4700) coupled with an energy dispersive X-ray spectrometer (EDS). Additional experimental details are provided in the supporting information.

2.2 Particle Classification and Soot Mixing State

First, soot particles were identified using their unique aggregate nature, since soot consists of small carbonaceous spherical monomers (Figure S2 in the supporting information). In this classification, “soot” includes soot particles that are mixed with or coated by other material. We note that SEM operates in vacuum conditions and provides only surface information (see the supporting information for details); these caveats may skew the number of soot particles toward larger fractions than actually present in the atmosphere.

Second, soot particles were visually examined to determine their mixing state. They were categorized into four groups: (1) embedded [Adachi and Buseck, 2008], when soot was fully engulfed in coating material; (2) partly coated, when the coating material filled the soot internal voids; (3) thinly coated, when coating on soot was minimal; and (4) partially encapsulated, when only a part of the soot was inside a host particle (Figure 2f). The details of soot classification and limitations are discussed elsewhere [China et al., 2013].

2.3 Parameterization of Soot Morphology

Soot particles are considered to be fractal-like aggregates, because they exhibit self-similar structures over several length scales [Oh and Sorensen, 1997]. Thus, they can be described using a mass fractal dimension (D_f) [Mandelbrot, 1982]. For particles with D_f < 2, D_f can be calculated from 2-D images using the ensemble method [Oh and Sorensen, 1997; Wentzel et al., 2003] (see the supporting information). The soot particles collected at the PMO were typically highly compacted (with expected D_f > 2); therefore, we did not use the ensemble method to estimate the scaling law. Instead, we used the measurable projected area of the aggregate (A_a) and its maximum length (L_max) [Lee and Kramer, 2004] given by

(1)

where we assumed the logarithm of A_a to be proportional to the logarithm of the particle mass. We underline that the power law exponent, D_2f represents a 2-D estimate of the fractal dimension that should be proportional but not equal to D_f. We applied the power relation (equation 1) using ensembles of soot particles to provide a measure of the fractal morphology of compact soot, inclusive of coating material filling voids between monomers.

The size of each particle was estimated from 2-D SEM images and is expressed in terms of the projected area equivalent diameter, D_Aeq. We used roundness and convexity to further quantify the compactness of soot (Figure S3 in the supporting information). Details on the image processing and morphological parameters are described in China et al. [2014].

2.4 Modeling of Soot Optical Properties With Different Compaction

We modeled climate-relevant optical properties of soot using synthetic soot particles that closely mimic the characteristics of aerosols sampled at the PMO. We generated the synthetic aggregates using a random walk algorithm [Richard et al., 2011] with an average monomer diameter of 34 nm as measured from the PMO samples. We used three different values for the numbers of monomers (N) based on the analysis of the PMO samples: 66, an intermediate value of 150 considering the range of observed monomer overlaps in thinly coated particles, and a high value of 247 considering a constant packing density of 0.36 [Radney et al., 2014; Zangmeister et al., 2014] (see supporting information). Furthermore, we generated synthetic soot aggregates with different levels of compaction: lacy (1.83 < D_f < 1.87), semicompact (2.00 < D_f < 2.40), and highly compact (D_f > 2.67) (Table S1 and Figure S6 in the supporting information).

The optical properties of the observationally constrained synthetic particles were calculated at 10 wavelengths, ranging from 370 nm to 950 nm, using the discrete dipole approximation (DDA-DDSCAT7.3 code) [Draine and Flatau, 1994, 2013]. The wavelengths were selected according to channels used by several remote sensing and in situ instruments. We used a wavelength-dependent refractive index for soot [Chang and Charalampopoulos, 1990]. The optical properties were averaged over 1000 random synthetic particle orientations sufficient to obtain DDA solution convergence for soot aggregates [Scarnato et al., 2013]. A detailed description of the methods for the aggregate generation and DDA simulations is provided elsewhere [Scarnato et al., 2013].

We numerically investigated the sensitivity of the spectral optical properties to compactness, neglecting the possible effect of coating. For comparison, soot optical properties were also modeled using Mie theory assuming spherical particles with the same mass-equivalent diameters [Mätzler, 2002].

3.1 Mixing States of Soot Particles

We analyzed a total of 1317 particles for event 1 and 806 for event 2. The relative abundance of soot particles in event 2 was almost double (~54%) than that in event 1 (27%). Soot particles at PMO were typically highly compact (Figures 2a–2c), while lacy chain-like soot aggregates were rarely observed.

Freshly emitted soot is typically bare or very thinly coated lacy structure [Wentzel et al., 2003; Adachi et al., 2007; China et al., 2014]. During atmospheric processing, soot mixes with other compounds yielding different mixing configurations and sometimes soot restructuring. The fractions of soot mixing categories are shown in Figure 2g. We found a higher fraction of embedded and partly coated soot (87%) in event 2 than in event 1 (57%) and a higher fraction of thinly coated soot in event 1 (37%) than in event 2 (9%), suggesting different atmospheric processing. Condensation of coating material may have been prominent for the less aged soot during event 2; however, it is also possible that the mixing material of the two events had different volatilities. The fraction of partially encapsulated soot was small for both events (<7%), suggesting that coagulation processes were not significant. In addition, semiquantitative EDS analysis showed that embedded particles contained C and O, suggesting that the embedding material was likely organic with a small, but detectable fraction of S. However, for both events the relative abundance of S was higher for the partially encapsulated particles than for the embedded particles (Table S2 in the supporting information). Similarly, in clean air above the remote Southern Ocean near Tasmania, Pósfai et al. [1999] found that 10–45% of the sulfate particles contained soot.

3.2 Morphology of Soot Particles

The fractal dimension of soot reflects its aging and is affected by several factors, including emission sources, emission conditions, and atmospheric processing pathways [Adachi et al., 2007; Chakrabarty et al., 2014]. Compact soot particles have greater D_f than lacy aggregates [Liu et al., 2008]. For example, soot aggregates in Figures 2b and 2c presumably had a 3-D D_f close to 3 or a 2-D fractal dimension D_2f close to 2. On average, D_2f estimated using equation 1 was greater for event 2 (1.89 ± 0.02) than for event 1 (1.82 ± 0.02) (Figure S5 in the supporting information). Greater D_2f reflects more coating or more compactness. In Table 1, we report the D_2f values for the three classes of soot.

Embedded soot for event 1 had the largest average D_Aeq, followed by partly coated and thinly coated soot, suggesting that the size of the particle increased with coating (Table 1). In contrast, the difference in the sizes of embedded and thinly coated soot in event 2 was small, suggesting that the soot cores might have been smaller in size or that the coating was thinner than for the particles in event 1. The details of the size distributions for the different soot types are discussed in Figure S4 in the supporting information. Note that thinly coated soot particles observed at the PMO often exhibited compact round aggregated shape, characterized by little void space between the monomers. The distribution of convexity (roundness; Figure S4 in the supporting information) for thinly, partly coated and embedded soot is shown in Figures 2h and 2i. Thinly coated soot particles were more compact in event 1 (convexity = 0.85) than in event 2 (convexity = 0.79), possibly due to the fact that particles during event 1 were likely more aged and processed. However, for both events, the thinly coated soot was highly compact compared to freshly emitted soot collected, for example, near freeway on-ramps (convexity = 0.69) [China et al., 2014].

A recent laboratory study showed that oxidized (OH initiated) soot adsorbed water upon humidification and turned into efficient CCN, leading to additional soot compaction [Khalizov et al., 2013]. Therefore, cloud or water processing may have been responsible for the compact shape of the thinly coated soot abundantly observed at PMO. Several previous laboratory studies reported the collapse of soot under the influence of water [Weingartner et al., 1995; Mikhailov et al., 1999; Lewis et al., 2009; Radney et al., 2014; Zangmeister et al., 2014]. It has been hypothesized that the capillary forces induced during condensation or filling of soot cavities with water may be responsible for the soot restructuring [Tritscher et al., 2011]. However, others argue that capillary forces during the evaporation of water, instead of the condensation, drive the restructuring of soot [Ebert et al., 2002; Ma et al., 2013]. The collapse of soot leads to a reduction in its surface area compared to freshly emitted soot.

3.3 Optical Properties of Soot Particles

We assessed the sensitivity of the soot optical properties to compaction by performing DDA simulations. Figure 3 (top) contains the C_abs and C_sca cross sections, single scattering albedo (SSA), and the asymmetry parameter (g) normalized by the corresponding values obtained for lacy soot. Simulations are shown for three different numbers of monomers (N = 66, 150, and 247). We found a wavelength dependence in the C_abs ratio (C_{abs-highly compact}/C_abs-lacy) as a function of compactness. Highly compact soot aggregates have slightly greater C_abs compared to the lacy soot for wavelengths larger than 500 nm, while the opposite is true for shorter wavelengths. Using T matrix, Liu et al. [2008] also found a similar pattern and suggested two possible reasons: (1) as particles become more compact, less of the absorbing material is directly exposed to the incident light, resulting in lower C_abs [Scarnato et al., 2013], and (2) C_abs increases for compacted particles due to an increased electromagnetic interaction between monomers [Liu et al., 2008], and this second effect seems to dominate at longer wavelengths.

The C_sca ratio (C_{sca-highly compact}/C_sca-lacy) is >1.5 for N = 66, and >2.0 for N = 150 and N = 247 in the visible wavelength range, but the ratio decreases (although always >1) in the near-infrared. The C_sca ratio is the highest (2.66) for N = 247 at 660 nm for highly compact soot. As the effect of compaction is substantially greater for scattering than for absorption, the SSA increases as the lacy soot collapses [Abel et al., 2003]. Increased SSA and extinction have also been recently demonstrated in laboratory-generated soot compacted by water processing [Radney et al., 2014]. The SSA ratio (SSA_{highly compact}/SSA_lacy) decreases as the wavelength increases. The g factor is a function of both particle size and shape [Liu et al., 2008; Kahnert et al., 2012]. We found that lacy soot has a higher g compared to the semicompact and highly compact soot with N = 60. However, g exhibits a different behavior for N = 150, with semicompact soot having higher g, followed by lacy and highly compact soot. Similarly, Liu et al. [2008] found that g reaches the highest value for a D_f~2 and starts decreasing as the particle becomes more compact. Furthermore, we computed the optical properties of soot with the same mass-equivalent diameter using Mie calculations to investigate how closely they approach the DDA values. SSA estimated using Mie is within ~19% ([SSA_DDA-SSA_Mie]/SSA_DDA) of the DDA value for highly compact soot, whereas it differs considerably for lacy soot (~149%) for all three cases of N at 525 nm (Figure S7 in the supporting information).

We estimated the ratio of the top of the atmosphere direct radiative forcing (TOA-DRF) for semicompact and highly compact soot with respect to lacy soot. We performed the calculations following the conceptual formulation by Chylek and Wong [1995]. Since the surface albedo (a) has an important role in determining the TOA-DRF, we performed the calculations for two extreme values: a = 0.06 and 0.8 (see supporting information for details) that represent a lower limit for the ocean albedo and an upper limit for the cloud albedo, respectively. The ratio of the TOA-DRF is close to 1 or ≪ 1, depending on compactness, number of monomers, and surface albedo (Figure 3, bottom graph). Compaction due to aging typically results in a reduction of the soot TOA-DRF for both values of a. For high surface albedo (a = 0.8) the TOA-DRF estimated from DDA differs substantially from Mie for lacy and semicompact soot having a normalized ratio [DRF_DDA-DRF_Mie]/DRF_DDA ≤28%; however, it differs by <12% for highly compact soot over the studied range of wavelengths and N (Figure S8 in the supporting information). Note that the effects of soot coating were not considered here. Coating would increase the absorption but would also result in much higher scattering, and the SSA would increase with coating. For example, SSA can increase by a factor of ≤3 at 670 nm depending on the coating material and mixing scenarios [Cheng et al., 2014]. Therefore, our calculations that neglect the coating effect represent a lower limit for the SSA increase.

As mentioned earlier, soot particles observed at PMO were likely cloud processed. Cloud processing is likely the reason for the highly compact shape with minimal coating. Soot particles may have been incorporated into water droplets or ice crystals during long-range transport. A recent T-matrix study [Mishchenko et al., 2014] on soot-water mixtures shows that orientation-averaged C_abs strongly depends on soot morphology and the position inside a water droplet. Orientation-averaged C_abs of a soot-water mixture is larger for lacy soot than for compact soot, and C_abs tends to decrease as the soot particle moves from the center of the droplet to the boundary. This effect implies that the soot morphology may have an important role in the semidirect effect as well, with soot compaction resulting in a lower in-cloud absorption and therefore lower positive forcing.