Morphology of diesel soot residuals from supercooled water droplets and ice crystals: implications for optical properties

The experiments were conducted in the environmental chamber facility at PNNL as part of SAAS in November of 2013 and January of 2014. Supplementary figure S1 shows the schematic of the experimental setup. Soot particles were generated using a diesel generator (Pramac P6000S-4810) under a 5000 W load using a load bank (Simplex, Swift-E). The soot particles were then extracted from the generator exhaust and diluted with pure air by a factor of 10–11 at 130 °C using a heated venturi pump before cooling to room temperate. Co-emitted species, such as volatile organic carbon components and NO_x, were partially removed using a charcoal denuder. Soot particles were size selected (120 nm diameter) (supplementary figure S2) according to their electrical mobility with a differential mobility analyzer (DMA, TSI, 3081). Mobility size selected soot particles (figure S2) were injected (8600 cm⁻³) into the environmental chamber.

A single particle mass spectrometer, SPLAT II (Zelenyuk et al 2009), was used for detailed real-time characterization of soot particles generated by the diesel engine. Particle characterization included measurements of mass, vacuum aerodynamic diameter (supplementary figure S2), and composition of individual particles sampled directly from the exhaust after mass or mobility classification. These data were used to determine particle effective density, material density, mass–mobility exponent, dynamic shape factors in free molecular and transition flow regimes, average diameter of monomers, number of monomers, and void fraction of soot agglomerates as described in details in a separate publication (Zelenyuk et al 2014).

Ice nucleation measurements of nascent soot particles were performed using a compact ice chamber (CIC) (Friedman et al 2011). Briefly, the CIC consists of two vertical independent temperature controlled parallel plates that are coated with an ice layer ∼0.5 mm thick to produce ice-supersaturated humidity conditions inside the ice chamber. The evaporation section of the CIC, again coated with an ice layer ∼0.5 mm thick, was maintained at a target temperature (−20 °C or −40 °C). The sheath and sample flow rates were 10 and 1 LPM, respectively, which led to an aerosol residence time of ∼12 s within the chamber. The temperature gradient between the vertical plates was adjusted such that using Murphy and Koop (2005) vapor pressure formulations, the desired conditions of relative humidity with respect to water (RH_w) were obtained at a given fixed temperature. We performed laboratory experiments at two temperatures: −20 and −40 °C and at each temperature the RH_w in the CIC was set to ∼108 ± 3%. These humidity conditions generated supercooled droplets. At −20 °C we did not observe ice crystals, while ice crystals were formed at −40 °C due to homogeneous freezing. We confirmed the absence of ice crystals and the presence of supercooled droplets at −20 °C by observing the size distribution of particles (supercooled droplets are in the range of 2.5–4.2 μm, while ice crystals are in the range of 6–9 μm in diameter) exiting the CIC using an optical particle counter (OPC; CLiMET, model CI-3100) (Kulkarni et al 2012, Kulkarni et al 2015). These results are in agreement with previous work where particles generated by miniCAST soot generator were used (Friedman et al 2011). Ice crystals and supercooled droplets were seperated from the interstitial particles using a pumped counterflow virtual impactor (PCVI) (Kulkarni et al 2011) and passed through a diffusion dryer to collect the dry residuals. The 50% cut-off size of PCVI was set to transmit droplets and ice particles larger than ∼4 μm. The residuals of the transmitted cloud hydrometeors were impacted on substrates for electron microscopy analysis using a four-stage cascade impactor (MPS-4G1). The samples analyzed here were collected on the fourth stage with a 50% cut-off aerodynamic diameters of 50 nm. Soot particles were collected on 300 mesh transmission electron microscopy (TEM) copper lacey formvar grids (Ted Pella, Inc.). As a reference, nascent soot particles were also collected directly from the environmental chamber, before entering into the CIC. Nascent soot particles were also collected on nucleopre polycarbonate membranes using a custom built aspirated sampler for scanning electron microscopy (SEM) analysis.

Individual nascent soot particles, and residuals from ice crystals and supercooled droplets were investigated using a field-emission SEM (Hitachi S-4700) and a TEM (JEOL JEM-2010). Freshly emitted soot particles are often represented as fractal-like aggregates due to their self-similar structures over several length scales (Oh and Sorensen 1997) and can be described by the following scaling law.

$$\begin{eqnarray}&&N={k}_{g}{\left(\displaystyle \frac{2{R}_{g}}{{d}_{p}}\right)}^{{D}_{f}},\end{eqnarray} \tag{ 1 }$$

where N is the number of monomers per aggregate, D_f is the mass fractal dimension, R_g is the radius of gyration, d_p is the monomer diameter, k_g is the fractal proportionality constant or fractal prefactor. We used the ensemble method to calculate the 3D fractal dimension of nascent soot particles from 2D images (Brasil et al 1999, Chakrabarty et al 2006, China et al 2014). However, soot particles from supercooled and ice crystal residuals were very compact (see section 3.1), presumably resulting in a D_f > 2. In this limit, N would be underestimated due to the significant overlap of the monomers in the 2D image. Therefore, instead of using equation (1), we calculated a 2D fractal dimension (D_2f such that for a sphere D_2f would be equal to 2) using the directly measurable area of each aggregate (A_a) and its maximum length (L_max) (Lee and Kramer 2004) using the following scaling law (China et al 2015):

$$\begin{eqnarray}&&{A}_{a}={k}_{2g}{\left({L}_{{\rm{max}}}\right)}^{{D}_{2f}}\end{eqnarray} \tag{ 2 }$$

For direct comparison purpose, we calculated D_2f for nascent soot particles as well (supplementary figure S3). We also used several other 2D morphological descriptors such as projected area equivalent diameter (D_Aeq), aspect ratio, roundness and convexity to characterize nascent soot, and supercooled droplet and ice crystal residuals. The D_Aeq is defined as the diameter of a spherical particle of the same projected area. The aspect ratio represents the level of elongation of the particle defined as the ratio of the longest dimension (L_max) to the orthogonal width (W_max). Roundness is the ratio of the projected area of the particle (A_p) to the area of a circle with L_max as the diameter. Higher roundness indicates particles that are more compact. The convexity is defined as the ratio of A_p and the area of the convex hull polygon (smallest convex polygon in which the particle is inscribed). In this study, we assessed the morphological characteristics of 200–240 individual particles for each samples. A detailed description of the image processing and the analysis of the morphological parameters and their limitations is provided elsewhere (China et al 2013, China et al 2014).

We investigated the effect of compaction on the optical properties of soot particles as a function of wavelength, using the discrete dipole approximation (DDA-DDSCAT7.3) code (Draine and Flatau 1994, 2013). In particular in this study, we discuss parameters relevant to radiative forcing calculations, including the absorption (C_abs) and scattering cross-sections (C_sca), the single scattering albedo (SSA = C_sca/(C_abs + C_sca)), and the asymmetry parameter (g). We used a random walk aggregation method to generate model soot particles for the DDA simulations (Richard et al 2011). The detailed method for the generation of the aggregates and for the DDA simulations are described elsewhere (Scarnato et al 2013). We modeled soot aggregates with 100 monomers with a diameter of 23 nm, in agreement with our observed mean values for nascent soot from electron microscopy and the measurements by SPLAT II, ice crystal and supercooled droplet residuals (see sections 3.1 and 3.2). We report the morphological parameters of the modeled soot aggregates representing nascent soot, supercooled droplet residuals, and ice crystal residuals in the supplementary table S1. We computed values of scattering and absorption cross sections averaged over 1000 random particle orientations (see Scarnato et al (2013) for discussion on the solution convergence due to the average of number of orientations). We performed the calculations at four different wavelengths of 450, 532, 781 and 550 nm; the first three corresponding to wavelengths used by several laser-based instruments for aerosol optical characterization and the last one to allow for a direct comparison with data available in literature that are often reported at 550 nm. We used the soot refractive index provided by Chang and Charalampopoulos (1990) for all the calculations (Scarnato et al 2015).

The composition analysis by SPLAT II revealed that soot particles generated by this diesel engine were composed of 80% elemental carbon, 5% oxygenated organics, and 15% PAHs, yielding an estimate for soot material density of 1.9 g cm⁻³. Furthermore, the measurements on the mass- and mobility-selected particles indicate, for example, that nascent soot particles with mobility diameter of 120 nm had mass of 0.56 fg, volume equivalent diameter of 83 nm, void fraction of 0.67, and were comprised on average of 67 monomers with diameter of 22 nm.

We begin by comparing the morphologies of soot residuals from supercooled water droplets and ice crystals with the morphology of nascent soot particles. In figure 1 we show examples of TEM images of individual nascent soot particles, and soot residuals from supercooled droplets and ice crystals. The open fractal-like structure of nascent soot particles exhibited a D_f of 1.53 ± 0.02 and a k_g of 3.44 ± 0.06. For diesel soot, previous studies showed D_f values ranging between 1.20 and 1.82 depending on engine conditions and combustion properties (e.g., Luo et al 2005, Li et al 2011). The center and right panels in figure 1 show the morphology of supercooled droplet residuals and ice crystal residuals, respectively. The residuals appear to be more compact than nascent soot.

Zoom In Zoom Out Reset image size

Figure 2 compares the distribution of aspect ratio (left panel), roundness (middle panel) and convexity (right panel) for nascent soot, soot from supercooled droplet residuals and soot from ice crystal residuals, respectively. Soot particles from ice crystal residuals are the most compact (higher roundness and convexity, and lower aspect ratio) followed by supercooled droplet residuals and nascent soot particles. On average, soot from ice crystal residuals is more compact than supercooled droplet residuals as indicated by the respective ∼34% and ∼10% increases in roundness compared to nascent soot), suggesting that the freezing process results in more extensive restructuring of soot. A consistent pattern is seen also from the 2D fractal dimension (table 1); D_2f is highest for soot from ice crystal residuals, followed by soot from supercooled droplet residuals, and finally, by nascent soot. Higher D_2f also represents more compact structures.

Zoom In Zoom Out Reset image size

These observarions of morphological changes of soot particles are in accord with the measurements conducted by SPLAT II on soot particles coated with secondary organic aerosol (SOA). In a separate set of the SAAS experiments it was shown that fractal-like nascent soot particles become compacted after coating and subsequent thermal removal of SOA coating. As a result, the effective density of soot particles increases from 0.58 g cm⁻³ for uncoated fractal nascent soot particles to 0.96 g cm⁻³ for the compact soot.

Table 1 summarizes the mean and standard deviation values for selected morphological descriptors (aspect ratio, roundness and convexity) including 2D fractal dimension and prefactor. Mean values of D_Aeq in table 1 show that soot particles from the ice crystal residuals (201 nm) exhibited the largest sizes followed by soot from supercooled residuals (179 nm) and nascent soot particles (153 nm). This is also consistent with their size distribution (supplementary figure S4). Coagulation of soot before entering the CIC can be responsible for the overall larger size of soot from supercooled droplet and ice crystal residuals. In addition, we suspect that some interstitial particles along with the desired size supercooled water droplets and ice crystals may have been transmitted through the PCVI (Pekour and Cziczo 2011).

Some degree of compaction (roundness ∼0.45) of soot residuals from the supercooled droplets can occur either during droplet condensation on soot particles due to surface tension (Kutz and Schmidt-Ott 1992) and capillary forces (Tritscher et al 2011) or during evaporation (Ma et al 2013). We observed that ice crystal residuals were more compact (roundness ∼0.55) than supercooled droplet residuals, suggesting that the atmospheric freeze-drying process perhaps playing a role in further restructuring of the soot particles. Expansion of water during freezing leads to presure developement inside the droplet and stresses in the shell of the frozen droplet, resulting in the growth of spikes or bulges (deformation of the shell) on the frozen droplet surface and even shattering of the shell if sufficient elastic energy is stored (Hobbs and Alkezweeny 1968, Johnson and Hallett 1968, King and Fletcher 1973, López and Ávila 2012). The pressure that develops inside the droplet and can be in the order of ∼7600 kPa for 7 mm droplets (Visagie 1969) and ∼8900 kPa for 10 mm droplets (King and Fletcher 1973). The magnitude of the pressure depends on the parameters such as droplet size, temperature, total freezing time and visco-elastic properties of ice (see discussion in supplementary material figures S5 and S6). The pressure inside the droplet increases with freezing time and it is higher for colder freezing temperatures (King and Fletcher 1973). The mechanical stress during freezing can alter the shape of the soot particle and the magnitude of the soot–water contact angle and filling angle (Butt and Kappl 2009), resulting in changes of the capilary force (Rahman 2001) and finally leading to the compaction of the soot structure. Previous studies suggested that the developement of surface stresses and compression forces during freezing process results in deformed shapes of biological material (Zhang et al 2006), wooden fibers (Macfarlane et al 2012) and materials used in pharmaceutical science (Zijlstra et al 2004). We suggest that similar stresses and compression are reponsible for the compaction of soot particles that we observed during the atmospheric freeze-drying process. The drying process (that occurs within the flow line between PCVI and cascade impactor; see section 2.2) can also contribute to the compaction of soot particles. We believe that both freezing and drying processes lead to higher compaction of the soot particles.

The optical properties of single nascent soot particles, as well as soot particles from supercooled droplet residuals and ice crystal residuals were simulated using the DDA model described earlier (supplementary table S2). Figure 3 shows the spectral dependence of several optical properties, including absorption (C_abs) and scattering (C_sca) cross-sections, SSA and asymmetry parameter (g) of soot residuals from supercooled droplets and ice crystals, normalized by the correspondent property of nascent soot. DDA predicts higher optical cross sections and SSA for compact aggregates. Soot residuals from ice crystals and supercooled droplets have a slightly higher C_abs (C_{abs-IC residuals}/C_abs-nascent is 1.05 and C_{abs-SCD residuals}/C_abs-nascent is 1.03 at 550 nm) than nascent soot. However, soot residuals from ice crystals have substantially higher C_sca (C_{sca-IC residuals}/C_sca-nascent is 1.37 at 550 nm) than nascent soot. These results are consistent with an increase in modeled C_sca for highly compact soot collected in the free troposphere in the North Atlantic (China et al 2015). Similarly, a previous T-matrix simulation study showed an increase in C_abs and C_sca at 870 nm for compact soot compared to less compact soot (Liu et al 2008). Less material is exposed to the light when soot particles are more compact, decreasing the absorption. At the same time, monomer-monomer interactions increase, resulting in an overall small enhancement (<10%) of absorption. Similarly, increase in monomer–monomer interactions results in higher scattering for compact soot (China et al 2015). Overall, our results show higher SSA for compact soot (IC residuals) than for lacey soot (nascent) and the SSA ratios (SSA_{IC residuals}/SSA_nascent) are 1.40, 1.30, 1.28 and 1.15 at 450, 532, 550 and 781 nm, respectively. These results are consistent with direct measurement of optical properties using photoacoustic and cavity ring-down spectroscopy by Radney et al (2014). They found that soot compaction increases SSA by a factor of ∼1.2 at 405 nm for relatively smaller monomer diameter (17 nm) and smaller soot size (geometric mean mobility diameter of soot: 116.7 nm) compared to our study. However, future studies should focus on in situ measurements of the optical properties of supercoooled droplet residuals and ice crystal residuals simultaneously with the morphological characterization of the soot particles. Our DDA simulation results are also consistent with a previous simulation study (Kahnert and Devasthale 2011, Liu et al 2008) that used T-matrix. They also found an increase in SSA due to soot compaction and the magnitude of the difference in SSA depends on monomer diameter, number of monomers, refractive index of soot and wavelength. For example, Liu et al (2008) found that the SSA ratio (SSA_compact/SSA_lacey) increases from 4.25 to 5.25 as soot size increases (from N = 200 to N = 400) for a monomer diameter of 15 nm at 870 nm wavelength. Our previous study also shows higher SSA ratio (1.98 versus 1.66) for larger soot (N = 150) compared to relatively smaller soot size (N = 66) for a monomer diameter of 34 nm at 450 nm wavelength. Compact soot particles exhibit lower g compared to open fractal-like nascent soot particles for visible wavelengths. In figure 3 we show that the normalized g values (normalized by the g value of nascent soot) are <1 for both ice crystal residuals and supecooled droplet with the first being the lowest. These results suggest an overall higher SSA and lower g for compact soot particles with respect to nascent soot. Both SSA and g are input parameters into radiative transfer models. A simple estimate of the top-of-the atmosphere direct radiative forcing (TOA-DRF) using the conceptual calculation (Chylek and Wong 1995, Haywood and Shine 1995, Lenoble et al 1982) shows that soot compaction due to atmospheric freeze-drying cycle typically results in a reduction of the positive soot TOA-DRF and the degree of reduction depends on the wavelength and surface albedo (supplementary figure S7). A maximum reduction of 63% in TOA-DRF ([TOA-DRF_{IC residual}-TOA-DRF_Nascent]/TOA-DRF_{IC residual}) is estimated for ice crystal residuals, while a maximum 45% reduction is estimated for supercooled droplet residuals due to compaction. Similarly, Kahnert and Devasthale (2011) found trends of reduction (up to 60%) of TOA-DRF for compact soot compared to lacey soot. Future studies should also focus on rigorous radiative forcing calculations by incorporating residual soot concentration, size distribution and other size dependent parameters.

Zoom In Zoom Out Reset image size