The effects of morphology, mobility size and SOA material coating on the ice nucleation activity of black carbon in the cirrus regime

Abstract. There is evidence that black carbon (BC) particles may affect cirrus formation and hence global climate by acting as potential ice nucleating particles (INPs) in the troposphere. Nevertheless, the ice nucleation (IN) ability of bare BC and BC coated with secondary organic aerosol (SOA) material remains uncertain. We have systematically examined the IN ability of 100–400 nm size-selected BC particles with different morphologies and different SOA coatings representative of anthropogenic (toluene and n-dodecane) and biogenic (β-caryophyllene) sources in the cirrus regime (−46 to −38 °C). Several aerosolized BC proxies were selected to represent different particle morphologies and oxidation levels. Atmospheric aging was further replicated with exposure of SOA-coated BC to OH. The results demonstrate that the 400 nm hydrophobic BC types nucleate ice only at or near the homogeneous freezing threshold (−42 to −46 ˚C). Deposition IN, as opposed to purely homogeneous freezing, was observed to occur for some BC types between 100–200 nm within the investigated temperature range. More fractal BC particles did not consistently act as superior deposition INPs over more spherical ones. SOA coating generated by oxidizing β-caryophyllene with O3 did not seem to affect BC IN ability. However, SOA coatings generated from OH oxidation of various organic species did exhibit higher IN onset supersaturation ratio with respect to ice (SSi) compared with bare BC particles, with toluene SOA coating showing an increase of SSi by 0.1–0.15 while still below the homogeneous threshold. n-dodecane and β-caryophyllene-derived SOA only froze in the homogeneous regime. We attribute the inhibition of IN ability to the filling of the pores on the BC surface by the SOA material coating. OH exposure levels of all SOA coating experiments, from an equivalent atmospheric 10 days to 90 days, did not render significant differences in IN potential. Our study suggests that BC particles with large sizes and/or oxidized surfaces generally exhibit better IN ability, and that the organic coating materials can inhibit ice formation.



Introduction
Cirrus clouds affect the global energy balance predominantly by more effectively trapping terrestrial long-wave terrestrial radiation than reflecting solar energy (e.g., Heymsfield et al., 2017;Kä rcher, 2018;Kä rcher et al., 2007). In cirrus clouds, ice crystals can form via two pathways, i.e. homogeneous and heterogeneous ice nucleation (IN) (Heymsfield et al., 2017;Kanji et al., 2017;Vali et al., 2015). Homogeneous freezing is the spontaneous freezing of solution droplets without any foreign 40 surfaces aiding the process (Heymsfield et al., 2017;Koop et al., 2000;Vali et al., 2015). Heterogeneous IN occurs more readily than homogeneous IN due to the presence of an ice nucleating particle (INP) at a lower supersaturation with respect to ice (SSi) or warmer temperature (DeMott et al., 2003;Kanji et al., 2017;Vali et al., 2015). Deposition IN is one heterogeneous IN mode, in which solid ice is formed by direct water vapor deposition on to an INP surface.
Laboratory experiments have been carried out to simulate the atmospheric environment to study the effects of isolated processes on BC IN ability in detail. Both well-characterized commercially available BC (e.g., Brooks et al., 2014;DeMott et al., 1999;Fornea et al., 2009;Mahrt et al., 2018;Nichman et al., 2019) and soot particles from combustion sources (e.g., 55 Crawford et al., 2011;Diehl and Mitra, 1998;Dymarska et al., 2006;Friedman et al., 2011;Kanji and Abbatt, 2006;Kanji et al., 2011;Koehler et al., 2009;Kulkarni et al., 2016;Möhler et al., 2005b;Mahrt et al., 2018;Nichman et al., 2019) have been used to investigate the IN ability of BC particles, with a particular focus on the deposition IN mode below -38 °C . According to previous studies (e.g., Friedman et al., 2011;Hoose and Möhler, 2012;Koehler et al., 2009;Kulkarni et al., 2016;Mahrt et al., 2018;Nichman et al., 2019), the following physicochemical properties of particles may play vital roles in determining BC 60 deposition IN activity: a) mobility diameter (dm), b) morphology, c) surface oxidation state, and d) organic material coating. It is widely acknowledged that larger particles act as more efficient INPs (e.g., Pruppacher and Klett, 2010). Although the mechanism remains uncertain, one common theory is that nucleation probability and rate are positively correlated to particle surface area (e.g., Hoose and Möhler, 2012); therefore, larger particles may offer more surface sites for nucleation. The IN ability of monodisperse BC particles with the size range of 100-800 nm has previously been characterized (Friedman et al.,65 particles act as active deposition INPs below -38 °C varied between 100 nm and 400 nm. However, the size threshold below which BC cannot nucleate ice in deposition mode and the underlying mechanism is still uncertain. Laboratory experiments and field observations confirmed that BC morphology and surface chemistry may change significantly during the atmospheric aging processes, leading to changes in particle surface area, shape, and chemical 70 composition (e.g., China et al., 2015;Fu et al., 2012;Li et al., 2017;Li et al., 2016;Moffet et al., 2016;Slowik et al., 2007;Tritscher et al., 2011;Wang et al., 2017). Commonly-used BC morphology characteristics are those derived from 2-D projected electron microscopy images, including fractal dimension (Df), roundness, aspect ratio (AR), and convexity (e.g., China et al., 2013;China et al., 2014;China et al., 2015;Kulkarni et al., 2016;Lee and Kramer, 2004;Ramachandran and Reist, 1995).
The surface chemistry of the emitted particles is governed by the source and the host environment in which the particles 85 evolve. Nascent BC particles can interact with volatile species such as sulfates and unburnt hydrocarbons in the aircraft cooling exhaust plume and grow (e.g., Anderson et al., 2011;Kä rcher, 2018;Lefebvre, 1998;Onasch et al., 2009). These particles can remain suspended in the atmosphere for days to weeks, during which the exposure to atmospheric biogenic and anthropogenic species, as well as oxidation, can lead to complex secondary organic aerosol (SOA) coatings (Kulkarni et al., 2016). Numerous experiments have been conducted to investigate the effects of surface coating on BC deposition IN ability. Hygroscopic BC 90 particles (Koehler et al., 2009), or BC particles coated by hygroscopic materials, such as sulfuric acid (Crawford et al., 2011;DeMott et al., 1999;Möhler et al., 2005b), water-soluble organic acids (Friedman et al., 2011;Nichman et al., 2019) and SOA (Kulkarni et al., 2016), tended to enhance BC water uptake ability and form aqueous solutions on BC surface, moving IN onset SSi towards the homogeneous freezing threshold. Hydrophobic organic coatings tended to impede surface interaction between BC and water molecules. Möhler et al. (2005b), Crawford et al. (2011), andMahrt et al. (2018) reported a transition from 95 heterogeneous to homogeneous freezing mode for combustion BC with increasing OC content. Ozone (Friedman et al., 2011) and hydroxyl (OH) radical (Chou et al., 2013;Kulkarni et al., 2016) oxidation can change surface functional groups of BC particles and enhance hydrophilicity, but no distinguishable BC IN activity change has been observed. Despite these previous https://doi.org/10.5194/acp-2020-809 Preprint. Discussion started: 2 September 2020 c Author(s) 2020. CC BY 4.0 License. efforts, the influence of particle morphology, chemistry, and aging, as well as the microphysical mechanism behind BC deposition IN ability, remains ambiguous. 100 In this work, we examine the effects of particle mobility diameter, morphology, and SOA coating on the IN ability of several aerosolized BC proxies as a function of SSi in a cirrus relevant temperature regime (from -46 °C to 38 °C).
Representative species of anthropogenic (toluene and n-dodecane) and biogenic (β-caryophyllene) volatile organic compounds were chosen to simulate potential photochemical atmospheric aging processes of BC. Different aging durations in equivalent atmospheric times were simulated by controlling the OH radical exposure. Our results help to clarify the effects of 105 physicochemical properties and SOA formation on BC IN ability and cirrus formation in the upper troposphere. Colorants Division), corresponding to different surface chemistry and morphology regimes were studied as proxies of atmospheric BC. Table 1 summarizes the characteristics of these BC proxies. R2500U and R330R are carbonaceous black pigment powder generated by incomplete combustion. COJ300 is a highly dispersible ink due to the 4-carboxyphenyl-modified surface (Johnson, 1999). COJ300 is selected for its high degree of oxidation, which is confirmed by the Particle Analysis by 115 Laser Mass Spectrometry (PALMS) chemical analysis (see Fig. A1), classifying it as the most oxidized BC proxy in this study. R2500U and R330R are unoxidized but differ in morphology, which was confirmed by morphology characterization and PALMS analysis (see Sect. 3.1). R2500U, R330R, and COJ300 were chosen as proxies of freshly emitted BC, atmospheric compacted BC, and atmospheric oxidized BC, respectively. IN properties of 800 nm R330R and R2500U particles were previously studied (Nichman et al., 2019). This work addresses the remaining questions raised in the previous study and focuses 120 on the impact of particle size, morphology, and surface oxidation.

SOA coating materials
Three organic species, toluene, n-dodecane, and β-caryophyllene, were selected to represent atmospheric SOA precursors from anthropogenic and biogenic sources (Table S1). Toluene and n-dodecane are often selected as surrogate jet fuel components to investigate combustion and emission characteristics because they have been proven well-suited to represent 125 tens of hundreds of components found in mainstream jet fuels (e.g., Dooley et al., 2010;Dooley et al., 2012;Zhang et al., 2016;Zhao et al., 2017). Field aircraft emission studies also confirm the presence of these unburnt aliphatic and aromatic organic compounds in aircraft engine exhaust (e.g., Beyersdorf et al., 2012;Kinsey et al., 2011;Pison and Menut, 2004;Timko https://doi.org/10.5194/acp-2020-809 Preprint. Discussion started: 2 September 2020 c Author(s) 2020. CC BY 4.0 License. et al., 2014). These organic compounds may coat BC particles, forming BC-containing aerosols in engine plume. Moreover, toluene is considered a dominant aromatic SOA precursor due to anthropogenic activities (e.g., Pandis et al., 1992) and serves 130 as a proxy for other light aromatic species in atmospheric aromatic-seeded SOA formation models (e.g., Hildebrandt Ruiz et al., 2015). n-dodecane is one of the most studied long-chain aliphatic SOA precursor (e.g., Loza et al., 2014;Presto et al., 2010;Yee et al., 2013) representing less volatile aliphatic species. β-caryophyllene has been found to be one of the most atmospherically abundant sesquiterpenes (Ciccioli et al., 1999;Guenther et al., 2012;Henrot et al., 2017). Due to its high reactivity towards ozone and hydroxyl radical to form oxidized products with low volatility, β-caryophyllene has a strong 135 potential to form biogenic SOA in the atmosphere (Calogirou et al., 1997;Griffin et al., 1999;Hoffmann et al., 1997;Lee et al., 2006).  for bare BC and BC-SOA mixing experiments, respectively. The relative humidity (RH) of the aerosol stream entering the DMA measured by the BMI built-in RH sensor was ~16 %. During the BC ice nucleation experiments, the size-resolved 155 particle number concentration was monitored with a BMI condensation particle counter (BMI CPC, Model 1700; Brechtel Manufacturing Inc.).

Characterization of BC morphology
The 200 nm, 300 nm and 400 nm R2500U, and 400 nm R330R and COJ300 BC particles were collected on 300-mesh carbon film copper grids (Ted Pella, Inc.) with a MOUDI impactor (Model M135-10; TSI Inc.) for offline morphology analysis. 160 The flow rate through the impactor was controlled by a MFC (Model MC-5SLPM-D; ALICAT Scientific) at 2 SLPM so that the cut-off size of the impactor was 100 nm. The samples were analyzed offline in a Zeiss Merlin High-resolution Scanning Electron Microscopy (HRSEM; Carl Zeiss Microscopy GmbH). Table 1. Characteristics of selected BC proxies in this study.
− is the BET specific surface area based on N2 adsorption isotherms; dm is the particle mobility diameter; ̅̅̅̅ denotes the mean 2-D projected area-equivalent aggregate diameter derived 165 from SEM images; mean aspect ratio ( ̅̅̅̅ ), roundness ( ̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ ) and circularity ( ̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ ) are the geometric mean morphology parameters derived from several aggregates and are defined in Sect. 2.2.2; ̅̅̅̅̅ denotes the mean geometric diameter of primary particles measured from SEM images, and N the number of primary particles analyzed for each BC type and size; Df denotes the 3-D fractal dimension derived from 2-D SEM images; dva is the particle vacuum aerodynamic diameter measured by the Particle Analysis by Laser Mass Spectrometry (PALMS) instrument; values in parenthesis are the 170 corresponding standard deviation. (400 nm in this study) and times the reference density 1 g/m 3 (Cziczo et al., 2006). g Calculated from PALMS spectra area. 175 Table 1 summarizes the morphological characteristics, including the projected area-equivalent diameter (da), aspect ratio (AR), roundness, circularity, and 3-D fractal dimension (Df), for different BC types and sizes derived from high resolution SEM images (×30,000 to ×150,000). Primary particle diameter ( ̅̅̅̅̅ ) is the geometric average of the length and width of a https://doi.org/10.5194/acp-2020-809 Preprint. Discussion started: 2 September 2020 c Author(s) 2020. CC BY 4.0 License. clear primary particle ( Fig. A2 and A3).

BC
= √4 ⁄ is the diameter of a spherical aggregate that has the same projected area ( ) as the BC aggregate (China et al., 2014). = ⁄ is the ratio between the longest dimension ( ) of an 180 aggregate periphery to the perpendicular maximum width ( , Fig. A2). = √4 2 ⁄ is used as a BC aggregate shape descriptor (e.g., China et al., 2013;China et al., 2015;Kulkarni et al., 2016). Both AR and roundness are used to represent shape deviation from a circle, whose AR and roundness equal 1. is a parameter used to describe the rugged level of an aggregate periphery, with rugged irregular periphery causing circularity smaller than 1. Df depends on primary particle number (N) and radius of gyration (Rg) of the aggregate (Mandelbrot, 1982). By using an ensemble 185 approach, N is found to be scaled with ( ) ⁄ 1.09 , where Aa and Ap are projected area of aggregate and primary particles, respectively (China et al., 2014;Köylü et al., 1995;Oh and Sorensen, 1997;Samson et al., 1987). The approximate relation 2 ⁄ = 1.50 ± 0.05 iss used to substitute Rg, (Brasil et al., 1999), and yield ( ̅̅̅ ⁄ ) = ( ) ⁄ 1.09 . Df can then be derived by a power law fit of scattered points between ̅̅̅ ⁄ and ( ) ⁄ 1.09 for each aggregate (Fig. A4).

Chemical composition characterization of single BC particle 190
Qualitative chemical composition of monodisperse BC particles was determined by PALMS. The detailed description of PALMS can be found elsewhere in literature (Cziczo et al., 2006;Zawadowicz et al., 2015). PALMS is an online single particle mass spectrometer in which inlet particles are first aligned by an aerodynamic lens. Two Nd:YAG green (532 nm) laser beams separated by 33.6 mm are arranged at the bottom of the inlet, measuring particle velocity based on time gap between the scattering signals. The velocity can be converted into vacuum aerodynamic diameter (dva) from the measured time of flight 195 (Cziczo et al., 2006, Fig. A5). A 193 nm ultraviolet (UV) excimer laser is then triggered, ablating and ionizing the particle.
The ions of both refractory and volatile particle components are classified based on their mass to charge (m/z) ratio. PALMS provides either positive or negative polarity spectra for each particle. Particle ionization is often not quantitative. However, average ion ratios across many spectra allows a qualitative compositional comparison between two similar aerosol populations.
Hundreds of spectra were collected for each soot sample to account for ionization difference caused by particle orientation 200 difference (Murphy et al., 1998).
Chemical composition of the SOA-coated BC particle stream was analyzed online by PALMS and an Aerosol Mass Spectrometry (AMS; Aerodyne Research Inc.). The AMS offers quantitative average mass spectrum of an ensemble of aerosols. Particles entering AMS first go through an aerodynamic lens inlet to form a particle beam. A mechanical chopper is used downstream the inlet to control sampling particle or particle free period. The AMS employs a heated 600 °C tungsten 205 surface to vaporize nonrefractory aerosols. Ionization is achieved using a universal 70 eV electron ionization technique.

SOA material coating on BC particles
The 350 nm COJ300 BC was chosen to be the seed particle in all SOA coating experiments because of its effective IN 210 activity as well as its higher particle concentration (~1 × 10 6 # L -1 ) at the selected size in comparison with other BCs (1-3 × 10 4 # L -1 ).
Particle generation during the SOA coating experiments was identical to bare BC experiments. The SOA coating experimental setup section is enclosed in the yellow dashed box of Fig. 1. COJ300 BC particles were nebulized in an air flow of 2.2 SLPM and dried in two consecutive 43 cm silica gel diffusion dryers, and then 350 nm BC particles were size-selected 215 by a TSI 3081 DMA, and directed to a potential aerosol mass (PAM) oxidation flow chamber (Kang et al., 2007;Lambe et al., 2011a;Liu et al., 2018). In the PAM reactor, gas phase volatile organic compound (VOC) reacts with OH radical and/or O3 (Lambe et al., 2011a;Zhang et al., 2018a), and subsequently form SOA-coated BC particles. All flow rates were controlled by i.e. 10 V and 3V, were tested in this study (indicated in Table 2 as suffixes -10 and -3), corresponding to different OH exposure levels and atmospheric aging time, ~10-15 days and ~70-90 days based on previous calculation (Lambe et al., 2011a). The VOC was injected into a heated bulb by a syringe pump and mixed with 0.7 SLPM dry air. The particle size distributions downstream of the PAM were measured by a BMI scanning mobility particle sizer (BMI SMPS, comprising a Model 2002 230 DMA and a Model 1700 CPC; Brechtel Manufacturing Inc.). The injection rate was controlled so that the mode diameter of the particles shifted from 350 nm bare BC particles to 400 nm SOA-coated BC particles, as illustrated in Fig. 2. The 400 nm SOA-coated BC particles were then dried to ~16% RH and kept below 25% by passing through two consecutive 43 cm silica gel diffusion dryers (DDU 570/H, Topas). The PAM chamber was cleaned by flushing 10 SLPM clean air overnight after each experiment. In order to confirm the cleanliness of the chamber, particle concentration was measured before and after each 235 experiment.
https://doi.org/10.5194/acp-2020-809 Preprint. Discussion started: 2 September 2020 c Author(s) 2020. CC BY 4.0 License.  nm and an increase of the 400 nm particle concentrations was observed for all experiments (Fig. 2 and Fig. B1), implying SOA coating on BC particles. The name prefixes BG, T, D, B in Table 2 stand for background test, toluene SOA coating experiments, n-dodecane SOA coating experiments, and β-caryophyllene SOA coating experiments, respectively. The name suffixes BC, 0, 3, 10, and s denote seed BC only, O3 oxidation only, low OH exposure level (3 V), high OH exposure level (10 V) and SOA self-nucleation experiments, respectively. All three organic species were exposed to both low and high OH concentrations to 245 investigate the effect of oxidation level on SOA formation and IN activity. An extra O3 oxidation experiment (B-0) was performed for β-caryophyllene because it is highly reactive towards O3 and may form SOA absent of OH. Self-nucleation IN experiments (-s) were performed for pure SOAs generated from each organic species to exclude the effect of nucleated pure SOAs mixing with SOA-coated BC particles.  (Garimella et al., 2016;Nichman et al., 2019;Wolf et al., 2019), and a brief description is given here. 260 SPIN is a continuous flow diffusion chamber style instrument comprising two flat parallel stainless-steel walls whose temperatures are controlled independently. The sampling flow rate of SPIN is 1.0 SLPM. Particles fed into SPIN are constrained by a ~9.0 SLPM sheath gas within a lamina near the centerline. Turbulent mixing at the injection point causes some particles to spread outside of the aerosol lamina centerline. Since particles experience lower RH as they spread outside of the lamina, correction factors ranging from ~1.9 to 8.0 were considered in previous studies (Garimella et al., 2017;Nichman 265 et al., 2019;Wolf et al., 2019). Both walls are coated with ~1 mm ice prior to experiments. At the beginning of each experiment, a linear temperature gradient and water vapor partial pressure field are established between the warm and cold walls.
Supersaturation with respect to ice is achieved because of the exponential relationship between temperature and saturation vapor pressure. For all the experiments in this study, SPIN was operated in a SSi scanning mode (1.0 to 1.6) while keeping the lamina temperature (-46 to -38 °C) constant for each scan. The SSi increased from 1.0 at a rate of 0.03 per minute by increasing 270 temperature gradient between the walls above homogeneous IN threshold and then lowered to ice saturation. https://doi.org/10.5194/acp-2020-809 Preprint. Discussion started: 2 September 2020 c Author(s) 2020. CC BY 4.0 License.
An optical particle counter (OPC) collects scattering signals for number counting and sizing, and a forward scattering depolarized signal for phase discrimination at SPIN chamber outlet. The size detection range of the OPC is 0.5 to 15 μm. A machine learning algorithm using the OPC scattering and laser depolarization signal (Garimella et al., 2016) was used to classify each particle as an inactivated aerosol or ice crystal over the course of an experiment. 275 We define the IN onset as 1% of particles activating, i.e. AF = 1%, for a period of 10 s as activation correction factor of 3.4 and 2.2 was applied for R2500U and R330R . Here the AF is defined as the number concentration of ice crystals identified by the machine learning algorithm divided by the total particle number concentration entering SPIN. For the size-selected bare BC experiments, the total particle number concentration was measured by a CPC operating simultaneously with SPIN, while for the SOA coating experiments, the total particle number concentration was integrated from 280 the SMPS measurement. As shown in Table 1 and Fig. 3, the three test BC types are substantially different in particle morphology. 400 nm R2500U has the smallest Df (~1.9), and COJ300 and R330R have larger Df (~2.3); R2500U is the most fractal BC while COJ300 and R330R are more spherical and compact. Meanwhile, R2500U and COJ300 have similar ̅̅̅̅̅ (34-35 nm), and R330R has larger 290 (~45 nm) primary particles. The larger ̅̅̅̅̅ of 200 nm R2500U might result from the fusion of primary particles under high magnification. Single particle surface area can be inferred by combining fractal level and ̅̅̅̅̅ together, and the decreasing order of single particle surface area is R2500U > COJ300 > R330R, which is in agreement with BET specific surface area data. Negative polarity mass spectra collected for 400 nm BC particles with PALMS are presented in Fig. 3C. The spectra of all three BC types exhibit typical consecutive carbon peaks (m/z = 12, 24, 36, etc.). The spectra of COJ300 shows presence of 295 oxidized ions, such as O -(m/z = 16), OH -(m/z = 17), and COOH -(m/z = 45), which are highlighted in red in Fig. 3C(b). The PALMS O:C ratio result confirms that COJ300 is more oxidized than R2500U (Fig. A1).

Ice nucleation on bare BC particles
https://doi.org/10.5194/acp-2020-809 Preprint. Discussion started: 2 September 2020 c Author(s) 2020. CC BY 4.0 License.   The results in Fig. 3A demonstrate that the particle size is relevant to particle IN ability. The 400 nm R2500U and R330R BC particles were able to nucleate ice below the homogeneous freezing threshold within the representative uncertainty in the temperature range of -46 to -38 °C . The COJ300 BC particles exhibited deposition IN activity regardless of particle size and 310 temperature in this study. The IN onset SSi of all depositional active BC particles increases with increasing temperature. The trend is in agreement with previous studies on in cirrus temperature regime (Chou et al., 2013;DeMott et al., 1999;Koehler et al., 2009;Kulkarni et al., 2016;Möhler et al., 2005a;Mahrt et al., 2018;Nichman et al., 2019 (Nichman et al., 2019). The COJ300 BC is more IN active compared with R2500U and R330R. The COJ300 particles show deposition IN ability below homogeneous freezing 320 threshold down to 100 nm within the temperature range in this study; the lower size threshold for COJ300 is below 100 nm.
This finding agrees with the lower size limit between 100 nm and 200 nm for BC particles to act as an active INP reported by Mahrt et al. (2018).
The IN onset results show no clear dependence on particle fractal level and surface area. Even though the more fractal and branching feature of R2500U BC particles may imply that there are more potential surface defects to initiate IN, the 325 R2500U particles do not clearly exhibit superior IN activity over R330R. Koehler et al. (2009) showed that IN was favored for oxidized hydrophilic BC, but too many hydrophilic active sites may bond water molecules, impeding ice embryo formation and thus impair IN (Pruppacher and Klett, 2010). The surface modified, highly dispersible and spherical COJ300 with smaller ̅̅̅̅̅ shows better IN efficiency than fractal BC, which is consistent with the results of Mahrt et al. (2018) and Nichman et al.
(2019) based on PCF mechanism. The smaller ̅̅̅̅̅ offers smaller cavities on particle surfaces that can accommodate liquid 330 water below bulk water saturation by the inverse Kelvin effect (David et al., 2019;Koop, 2017;Marcolli, 2014Marcolli, , 2020. Water saturation pressure drop as a function of cavity radius is shown in Fig. C1 (Marcolli, 2020).  There exists no distinguishable difference between bare COJ300 and BC coated with highly oxidized toluene SOA (T-10 in Table 2) from -46 to -44 °C. Ice crystals may form on the carbonaceous part of partially coated particles, whose IN onset 345

Ice nucleation on BC coated with SOA material
SSi should be the same as bare COJ300. At temperatures above -43 °C, toluene SOA-coated BC particles nucleate ice at SSi ~0.1 to 0.15 above bare 350 nm COJ300, but still ~0.15 below the homogeneous freezing threshold. This is in agreement with Wang et al. (2012) that pure aromatic SOA nucleate ice at SSi 0.1-0.15 below the homogeneous freezing limit. BC coated by ~10-15 equivalent days atmospherically oxidized toluene SOA (T-3 in Table 2) particles nucleate ice in deposition mode at higher SSi than highly oxidized toluene SOA-coated BC within the investigated temperature range. Previous studies reported 350 a molar weight range of 58-135 g mol -1 for toluene SOA (e.g., Bohn, 2001;Ji et al., 2017). The toluene SOA mass spectrum in Fig. 6(a) exhibits higher m/z = 44 and lower m/z = 43 fraction signal, indicating more oxidized organic species were generated during T-10 and T-3 experiments (Lambe et al., 2011b), agreeing with the previous study on toluene SOA (Liu et al., 2018).
On the one hand, the higher O/C ratio (Fig. 7) of toluene-derived SOA when compared with the other two types of SOA may enhance the hygroscopicity of the particle (Lambe et al., 2011b;Liu et al., 2018;Zhao et al., 2016) and thus may reduce the 355 deposition IN ability of BC particles. On the other hand, Hinks et al. (2018) shows that toluene-derived SOA also contains a significant amount of oligomers under dry laboratory conditions, similar to what we conducted in the PAM chamber in this study, potentially reducing the hygroscopicity and altering the phase state of the SOA to be semi-solid or solid (Li et al., 2020;Zhang et al., 2018b), under which the SOA can still nucleate ice (Zhang et al., 2019c). Overall, these two competing factors make our toluene SOA coating IN onset move towards but not fully in the homogeneous freezing regime, agreeing with the 360 results of Kulkarni et al. (2016).
However, O3 oxidized β-caryophyllene SOA shows no significantly alternation of IN ability. The mass spectra in Fig. 6(b) and experiments (Lambe et al., 2011b). The H/C and O/C values of n-dodecane and β-caryophyllene SOA coating in Fig. 7 are smaller than that of toluene SOA, which are in agreement with previous studies (Li et al., 2019;Pereira et al., 2019;Simonen et al., 2017). The slopes of H/C and O/C values of these two types of SOA and their respective two precursors (Fig. 7) are in the range between -1 and 0, which is consistent with the simultaneous formation of carboxylic acid functional groups and C-C bond breakage (Heald et al., 2010;Lambe et al., 2011b). Addition of carboxylic acid group may enhance the hygroscopicity 380 of n-dodecane and β-caryophyllene SOA, and the hygroscopicity is further enhanced with more OH exposure (Bé et al., 2017;Frosch et al., 2013;Schilling et al., 2015;Yee et al., 2013). We conclude that BC with OH oxidized n-dodecane and βcaryophyllene SOA coatings, regardless of oxidation level, may condense on BC surface and forms organic films, leading to nucleation in the homogeneous regime. The experimental results are attributed to two factors: organic coating and volatility. Previous studies controlling the 390 combustion fuel-air-ratio produced BC particles occupying different organic content fractions, with higher organic content resulting in amorphous organic surfaces (Crawford et al., 2011;Möhler et al., 2005b;Mahrt et al., 2018). Shifts from heterogeneous to homogeneous freezing with increasing organic content have been observed. Kulkarni et al. (2016) reported that α-pinene SOA coating suppressed the ice nucleation ability of BC particles. However, studies show that as the volatility of the organic coating decreases below certain threshold, especially near glass transition temperature, these organic coatings 395 might be able to heterogeneously nucleate ice (Berkemeier et al., 2014;Murray et al., 2010;Zhang et al., 2019c). The suppression of BC IN ability by organic coating was attributed to coverage of surface-active sites and filling of pores on BC surface when the volatility of the organic coating is relatively high. Certain SOA coatings in this study are less oxidized and thus may similarly impair BC IN ability due to their relatively high volatility, as Docherty et al. (2018) andHildebrandt Ruiz et al. (2015) showed an inverse correlation between the volatility and oxidation state. Our results suggest that less oxidized 400 SOA (n-dodecane and β-caryophyllene derived SOA from photooxidation), despite their high mass loadings in PAM chamber, are more likely to condense on seed particle and forms fully coated BC particles, moving IN onset SSi to the homogeneous regime, while β-caryophyllene SOA oxidized by O3 did not alter the SSi of the soot particles. In addition, more oxidized SOA (toluene derived SOA from photooxidation) with potentially more oligomer formation, moving IN onset SSi towards, but still below, homogeneous freezing. 405

Atmospheric implications
BC particles emitted from combustion sources (such as aero-engines) are carbonaceous nanoscale fractal aggregates with primary particle diameter of 20-50 nm Vander Wal et al., 2014). These BC particles can remain suspend in the atmosphere for days, and might undergo compaction and atmospheric aging, such as oxidation and mixing with atmospheric aerosols. This study focuses on the impact of morphology, particle size and mixing state on the IN ability of BC-410 containing aerosols. Three BC proxies were chosen to represent freshly emitted (in other words, unoxidized and more fractal) BC (R2500U), unoxidized compacted BC (R330R), and atmospheric chemically aged BC (COJ300). The morphological characteristics, such as dpp, circularity, roundness, and Df, are within the value range of typical BC emitted from aircraft engines, vehicles, biomass burning, laboratory flames, and field observation (e.g., China et al., 2013;China et al., 2014;China et al., 2015;Lapuerta et al., 2007;Vander Wal et al., 2014;Zhang et al., 2019b). Findings in this study can be relevant to 415 airborne aircraft emissions and ground emissions carried by updrafts to tropopause.
The IN results for bare BC particles show dependence on particle size and surface chemistry, but the role of fractal level seems to be of limited importance. The lower size limit of bare BC to exhibit IN activity is between 300-400 nm for R2500U at -46 °C. This is important for freshly emitted BC from aircraft engines and vehicles, which are usually fractal with dm<200 nm (Kittelson, 1998;Moore et al., 2017). It is unlikely that small, freshly emitted BC will activate as INP in aircraft plumes 420 https://doi.org/10.5194/acp-2020-809 Preprint. Discussion started: 2 September 2020 c Author(s) 2020. CC BY 4.0 License.
below the homogeneous freezing threshold if they possess similar physicochemical properties as R2500U. The smallest size for compacted BC (R330R) to activate as INPs lies between 200-400 nm at -46 °C. This means that the IN ability of small BC particles may be enhanced after cloud cycles, during which fractal BC geometries may collapse and forms PCF favoring morphology (Mahrt et al., 2020). The COJ300 IN results imply that ice crystal formation may favor oxidized hydrophilic surfaces. The dpp of COJ300 is appropriate for mesopores to accommodate ice crystal formation below water saturation. 425 Particles down to 100 nm can act as efficient INP. This implies that for long-lived atmospheric BC particles, after being oxidized and compacted, may act as efficient INP.
To simulate atmospheric aging, toluene, n-dodecane and β-caryophyllene were chosen to represent anthropogenic and biogenic SOA precursors (Atkinson and Arey, 2003;Ding et al., 2014;Hu et al., 2008). Toluene-derived SOA coating impede BC heterogeneous IN activity slightly while n-dodecane and β-caryophyllene-derived SOA coatings caused BC particles to 430 nucleate ice homogeneously. BC emitted from aircraft and vehicles are likely to be coated by toluene and n-dodecane derived SOA (e.g., Beyersdorf et al., 2012;Beyersdorf et al., 2014;Timko et al., 2014). According to our experimental results, even though such coating can facilitate particle growth, coated particles are more likely to nucleate ice near the homogeneous freezing threshold.
The conclusions drawn here for BC proxies may deviate from genuine BC collected from combustion sources. 435 Nonetheless, BC surrogates are often used in research to mimic aircraft emitted BC for their similarity and availability (e.g., Persiantseva et al., 2004). Additional IN studies, over a wider temperature range would also be required for the proxies to firmly verify the PCF mechanism; the question whether the studied IN is depositional or in fact homogeneous IN of liquid water in pores and cavities, remains to be answered due to the limited temperature range investigated in this study.

Summary 440
The IN ability of size-selected (100-400 nm) monodisperse BC particles with different morphologies and surface chemistry and BC particles coated with toluene, n-dodecane, and β-caryophyllene-derived SOA has been systematically investigated in the cirrus temperature regime (-46 to -38 °C). Three aerosolized BC proxies were selected to represent particle morphology at different atmospheric aging stages, i.e. freshly emitted (R2500U), atmospheric compacted (R330R), and atmospheric oxidized (COJ300). The IN activity was investigated in relation to particle size, morphology, surface chemistry, 445 SOA precursor type and OH exposure level. combustion particles (aircraft, vehicles, and biomass burning, etc.) and advanced single particle characterization for validation of the PCF mechanism.