Interactive comment on “Impact of Isolated Atmospheric Aging processes on the Cloud Condensation Nuclei-activation of Soot Particles”

The manuscript reports a laboratory study of CCN activity of soot particles during their atmospheric aging processes and implementation of the lab results into a global aerosol-climate model. The aging processes mainly consider the heterogeneous ozone oxidation at atmospherically relevant condition. The combination of chamber work with an aerosol-climate model is major strength of this study, as it highlights the importance of accurate CCN treatment in a GCM. The paper is well written overall, so I only have some minor comments for the authors to address.


Introduction
Aerosols are defined as fine solid particles or liquid droplets suspended in a gas phase. Aerosol particles impact the Earth's 30 radiative budget both directly (e.g. through scattering of shortwave and absorption of shortwave and longwave radiation) (Haywood and Boucher, 2000) and indirectly (e.g. by changing the properties of clouds; Ackerman et al., 2000;Lohmann and Feichter, 2005;Seinfeld et al., 2016;Twomey, 1977). Furthermore, they demonstrate significant impacts on air quality and human health (Anenberg et al., 2012;Janssen et al., 2011). The chemical and physical properties of atmospheric aerosol particles are highly variable and depend on surface (land vs. ocean), region (urban vs. remote), source (anthropogenic vs. 35 biogenic), and many more aspects. Additionally, ambient aerosol particles undergo physicochemical modification processes throughout their atmospheric lifetime (Monks et al., 2009). Condensation of gas phase volatile material or heterogeneous oxidation are general examples of these processes that are referred to as aging. An example of the multiple ways these processes modify the physicochemical properties of the particles is the change in the water affinity of an initially hydrophobic particle.
Accumulation of hygroscopic material on the surface can cause such a particle to become hydrophilic (Dalirian et al., 2018;Henning et al., 2012;Khalizov et al., 2009). Depending on the surrounding conditions, the particle can accumulate water vapor and form a droplet as in a cloud or in fog.
The process of forming a cloud droplet is called cloud droplet activation and the respective particles are called Cloud Condensation Nuclei (CCN). The particles' ability to act as CCN depends on its properties, such as size, morphology, and 5 chemical composition (Köhler, 1936;Sorjamaa and Laaksonen, 2007). While particles consisting of hygroscopic compounds such as sea salt have a high CCN-activity, other particles such as soot (also referred to as black carbon, BC) show extremely low CCN-activity (Petzold et al., 2013). Nevertheless, BC particles have been found in cloud droplet and ice crystal residuals in ambient measurements, indicating that within their atmospheric lifetime these particles are incorporated into hydrometeors (Cozic et al., 2008;Hiranuma et al., 2013). 10 Soot is a by-product of incomplete combustion. Depending on its origin, soot varies greatly in chemical composition, size, and co-emitted substances. Soot particles have an atmospheric lifetime of up to one week, which is long compared to other aerosol particles (Textor et al., 2006). Currently, the impact of soot particles on human health, environment, and climate is of scientific and economic interest. Understanding their carcinogenic nature (WHO, 2016) or their impact on crops (Burney and Ramanathan, 2014) are only some examples of the increased interest in soot particles during the last few decades. Regarding 15 their atmospheric impact, a good understanding has been gained with respect to their direct effect on visibility and air quality but their indirect climate impact, i.e. on clouds and cloud formation, remains highly uncertain (IPCC, 2013).
The pathway of how soot particles end up in hydrometeor residuals remains a major topic of discussion. On the one hand, these hydrophobic particles show little interaction with water and are reported to be poor CCN and ice-nucleating particles, respectively (INPs; Friedman et al., 2011;Koehler et al., 2009;Kulkarni et al., 2016). On the other hand, field measurements 20 show that soot particles are enriched in cloud droplets and ice crystals compared to interstitial particles (Cozic et al., 2008;Hiranuma et al., 2013). These findings indicate that soot particles can become incorporated into hydrometeors beyond impaction scavenging potentially by an increase in hygroscopicity upon atmospheric aging turning them into CCN or INPs, respectively. The details and relevance of atmospheric aging processes potentially causing such a significant change in CCNactivity of soot particles are not well understood yet. Besides the complexity of aerosol particles, one of the challenges lies in 25 investigating these processes in the laboratory at atmospheric conditions. Furthermore, modeling studies show that even though soot particles are poor INPs compared to dust particles (Kanji et al., 2017) they are still relevant INPs in the atmosphere due to their high abundance (Hoose et al., 2010;Savre and Ekman, 2015).
While there is a broad consensus that coating with hygroscopic substances e.g. sulfuric acid (Dalirian et al., 2018;Henning et al., 2012;Khalizov et al., 2009) increases the particle-water interaction of soot particles, the influence of oxidation processes 30 is less well understood. The impact of oxidation processes can be investigated by simulating atmospheric aging under controlled laboratory conditions. One of the experimental challenges is to achieve extended aging time periods because the average atmospheric lifetime of soot particles is approximately one week (Textor et al., 2006). A common approach is the application of (photo-) Oxidation Flow Reactors (OFR) like the Potential Aerosol Mass (PAM) chamber (Kang et al., 2007), the Toronto Photo-Oxidation Tube (TPOT; George et al., 2007), the Micro Smog Chamber (MSC; Keller and Burtscher, 2012), 35 the TUT Secondary Aerosol Reactor (TSAR; Simonen et al., 2017) or the Photochemical Emission Aging flow tube Reactor (PEAR; Ihalainen et al., 2019). Within these devices, the residence time ranges from 3 to 170 s and the OH-radical concentration ranges from 4.9 × 10 8 to 130 × 10 8 molec•cm -3 , while the average atmospheric OH-radical concentration is orders of magnitude lower with 1.5 × 10 6 molec•cm -3 (Mao et al., 2009). The exposure conditions are recalculated to an equivalent atmospheric aging time of 0.4 to 10 days (Lambe et al., 2015). This approach implies that the oxidation speed is 40 linearly dependent on the concentration of OH-radicals which is supported by the findings of .
Deploying larger aerosol chambers with several cubic meters of volume operated in batch mode allows for longer experimental durations at more atmospherically relevant oxidant concentration levels. For example, Wittbom et al., (2014) achieved aging durations of up to 5 h in a 6 m 3 aerosol chamber at OH-radical concentrations ranging from 1 × 10 6 to 2 × 10 6 molec•cm -3 which is approximately equal to one day of atmospheric aging. Both, the OFRs and the batch-aerosol chamber methods show that equivalent atmospheric aging time spans of several hours to days are required to make soot particles CCN-active at atmospherically relevant super saturations (SS) of below 0.8 % (Pruppacher and Klett, 2010).
Another very important atmospheric oxidant is ozone (O3). The effect of O3 oxidation on the CCN-activity of soot particles 5 has been investigated extensively in various laboratory studies. Despite these efforts, no CCN activation at atmospheric O3 concentration and atmospherically relevant SS has been reported to the authors' knowledge. However, in the range from 1200 ppb to 20,000 ppb O3 a significant increase in CCN-activity of soot particles was reported for SS above 0.8 % for exposure times between 100 s and 2 h, which should correspond to atmospheric aging times of up to 3.5 days (Friedman et al., 2011;Grimonprez et al., 2018;Lambe et al., 2015). Contrary to the agreement regarding the linearity of aging and OH-10 exposure within the scientific community (Renbaum and Smith, 2011), there is no such consensus concerning aging and O3 exposure. While the results of the studies mentioned above are interpreted on the assumption that the oxidation speed is directly proportional to the O3 concentration, Kotzick et al., (1997) reported that no impact of concentration could be detected in the range from 25 ppb to 1000 ppb O3.
Studies focusing on the uptake of O3 by soot particles suggest that the reaction might not follow first-order kinetics with respect 15 to the O3 gas-phase concentration e.g. (Ammann et al., 2003;Kamm et al., 1999;Lelievre et al., 2004;McCabe and Abbatt, 2009;Zelenay et al., 2011). Similar results have been found for the decomposition of Polycyclic Aromatic Hydrocarbon (PAHs) and other organic compounds condensed on aerosol particle surfaces McNeill et al., 2007;Pöschl et al., 2001;Shiraiwa et al., 2011). These findings combined with soot particles found in hydrometeor residuals question the validity of extrapolations from non-atmospheric reaction conditions being used as the basis to infer atmospheric 20 implications. To evaluate the atmospheric relevance of O3 oxidation on the CCN-activity of soot particles and the impact of the O3 concentration in the atmospherically relevant range, laboratory experiments should preferably be performed at atmospherically relevant conditions with respect to oxidant concentration, relative humidity (RH), particle number concentration as well as reaction time.
Different experimental setups provide different advantages with respect to mimicking atmospheric aging processes. OFRs 25 have the advantage of operating at particle number concentrations prevalent in the atmosphere. Therefore, the aerosol particles can be size-selected before entering the chamber, and changes in the CCN-activity can be investigated excluding artifacts from potential size-related effects. However, because those chambers are operated at oxidant conditions that are up to 4 orders of magnitude higher than in the atmosphere (Bruns et al., 2015) and might follow atmospherically non-relevant reaction pathways (McNeill et al., 2007). In contrast to the OFR`s, there are large aerosol chambers with several m 3 of volume in which aerosols 30 can be exposed to atmospherically relevant concentrations of oxidants and trace gases (Cocker et al., 2001;Leskinen et al., 2015;Nordin et al., 2013;Paulsen et al., 2005;Platt et al., 2013;Presto et al., 2005;Rohrer et al., 2005). Because those chambers are typically operated in batch-mode, they require elevated particle number concentrations in the input flow in order to reach the desired aerosol concentration inside the chamber within a reasonable time. Therefore, these types of chambers are often filled with non-size-selected aerosol particles, hampering the separation of CCN activation due to chemical 35 transformation from potential particle size effects.
In this paper, we present the results from a study investigating the effect of heterogeneous O3 oxidation at atmospherically relevant conditions on the CCN-activity of soot particles derived from a co-flow propane diffusion flame. The experiments were performed within a ~3 m 3 stainless steel aerosol chamber operated in continuous-flow stirred tank reactor (CSTR) mode. 40 This approach allowed us to 1) achieve aging durations of up to 12 h, 2) utilize atmospherically relevant O3 concentrations of up to 200 ppb and varying levels of humidity, and 3) execute the experiments with 100 nm size-selected soot particles. The experimental results were then implemented into the global climate model ECHAM6.3-HAM2.3 Tegen et al., 2019). Based on these results, we discuss the impact of aging processes on the change in CCN burden and cloud droplet number concentration (CDNC) on the global as well as regional scale.

Overview
Multiple sets of experiments were performed in the laboratory using a 2.78 m 3 stainless steel aerosol chamber at ETH Zurich, Switzerland during summer 2016 and summer 2018. To allow for extended aging durations of up to 12 h, the aerosol chamber was operated following the concept of the CSTR (Levenspiel, 1999). In accordance with this concept, soot particles were 10 continuously added to the aerosol chamber inlet, simultaneous to the continuous withdrawal of aerosol samples from the chamber outlet for analysis. All sample lines used within the set-up consisted of stainless steel with an inner diameter of 4 mm.
The sample lines were configured such that the soot aerosol could be sampled either directly upstream, i.e. bypassing, or downstream of the aerosol chamber. Friebel and Mensah (2019) introduced a new analysis technique that allows for the retrieval of CCN activation data from experiments executed in aerosol chambers operated in CSTR mode. In addition to the 15 investigation of the CCN-activity, data were recorded for the investigation of a broad range of physical and chemical parameters, e.g. the size and mass distribution of the particles, the INP potential, the chemical composition, and the lung deposited surface area (LDSA). Further details on the experimental set-up, instrumentation, and specific settings beyond the information given in the following section can be found in the appendix.

20
Soot particles were produced by a propane-fueled Jing Ltd., miniature Combustion Aerosol STandard (miniCAST 4200) generator. Such type of burners and specifically the miniCAST burner have been used widely for the production of soot particles in laboratory studies e.g. (Durdina et al., 2016;Kim et al., 2015;Malmborg et al., 2018;Mamakos et al., 2013;Maricq, 2014;Mueller et al., 2015;Török et al., 2018). The miniCAST was operated at two different settings for the generation of soot samples with different organic carbon (OC) contents. The first sample, hereafter referred to as CAST brown (CBW), was generated under fuel-rich conditions (fuel-air ratio; FAR = 1.03). The second sample, hereafter referred to as CAST black (CBK), was generated under fuel-lean condition (FAR = 0.95) . Further details on the miniCAST set-points used during the study are listed in appendix Table 4.
The miniCAST has been used as surrogate for soot emissions from vehicle internal combustion engines (Maricq, 2014;Moore 5 et al., 2014;Mueller et al., 2015) and aircraft engines (Bescond et al., 2014;Durdina et al., 2016). According to Marhaba et al., (2019) high engine thrust levels can be mimicked by CBK soot, while CBW soot better represents engine emissions at lower thrust levels .
The gases used were nitrogen (N2) of grade 6.0 for mixing and quenching, and in-house filtered and compressed air for 10 oxidation and dilution. The compressed air was purified by passing through a particle filter resulting in a particle concentration below the detection limit of the particle counting instrumentation. The air was further passed through a charcoal filter for the removal of volatile organic compounds (VOC). The remaining VOC content was tested by mixing the filtered air with 200 ppb of O3. As no new particle formation could be detected we considered the filtered air particle-and VOC-free with respect to our instrumentation. 15 After starting the burner, it was run for at least 2 h before the operating conditions were considered stable and the output was sampled. The output of the miniCAST burner was diluted by a factor of 10 using a Palas VKL 10 diluter. 6 L•min -1 of the diluted sample was introduced into a pre-mix chamber of 0.125 m 3 volume. The stainless steel pre-mix chamber was air-tight and equipped with a continuously stirring fan. To allow for the selection of 100 nm particles of sufficient concentration, the particles were allowed to agglomerate within the pre-mix chamber. After an average residence time of 21 min, the mode 20 diameter of CBW and CBK particles was 90 nm and 150 nm, respectively.

Aerosol chamber
A 2.78 m 3 stainless steel aerosol chamber was used as a reaction vessel. As a detailed description of the stainless steel aerosol chamber has been previously presented by Kanji et al., (2013), only a brief description follows. The aerosol chamber is equipped with a pitched blade fan of 30 cm diameter at its bottom. The fan was operated at 1000 rpm to ensure a homogenous 25 distribution of the aerosol throughout the aerosol chamber. Based on the operational experience acquired in the summer 2016, the aluminum fan was gold plated prior to the campaign in summer 2018 to increase its conductivity and thereby reduce particle loss on its surface. Temperature, pressure, and RH inside the aerosol chamber were monitored by sensors mounted on a diagonally oriented taut wire. While pressure and RH were controlled by the conditions of the input flow, the double-wall design of the aerosol chamber allowed for direct temperature control, which was utilized in some of the experiments. Soot 30 aerosol and O3 were introduced through individual ports on stainless steel aerosol chamber. Another port was used for the withdrawal of sample aerosol, which then was distributed to various measurement instruments.
In general, the total volumetric flow through the aerosol chamber while filling and steady state was set to 23 L•min -1 . In some experiments, a reduced flow rate of 13 L•min -1 was applied during the overnight flushing regime to maximize the duration of particle concentration above the detection limit. The experimental conditions allowed the exposure time of the soot aerosol 35 during oxidation and humidification experiments to be extended up to 12 h.

Sample selection and conditioning
The soot aerosol was conditioned in multiple ways prior to entering as well as within the aerosol chamber. Following the premix chamber, a home-built charcoal denuder was placed in-line for the removal of remaining gas-phase VOCs from the combustion process within the miniCAST burner. The denuder consisted of a glass tube of 40 cm in length and 10 cm in 40 diameter filled with approximately 0.7 kg of activated charcoal (Sigma-Aldrich). A metal mesh of 1.5 cm diameter connecting the inlet and outlet of the denuder allowed the aerosol stream to pass through the center of the denuder without direct exposure to the charcoal. The denuder was bypassed (i.e. the sample was not denuded) for some of the experiments to evaluate the potential impact of the remaining VOCs on the CCN-activity of the particles. A TSI 3081L Differential Mobility Analyzer (DMA) was used downstream of the denuder for the selection of 100 nm soot particles. The DMA was operated with a sample airflow rate of 1.7 to 1.9 L•min -1 and a sheath air flow rate of 10 L•min -1 . After diluting the sample airflow with 21 L•min -1 of 5 particle-and VOC-free compressed air, a particle concentration of ~1200 cm -3 was achieved. A TSI 3772 Condensation Particle Counter (CPC) was used to monitor the number concentration of the soot aerosol particles entering the aerosol chamber. The input concentration remained stable for the entire duration of an experimental run.
Conditioning of the soot particles with O3 or elevated humidity took place within the aerosol chamber. Gas streams containing O3 and water vapor were fed into the aerosol chamber through individual ports. O3 was produced by a continuously running 10 corona discharge O3 tube operating on high purity 5.6 synthetic air. The output of the O3 generator was diluted by a factor of 100 using a Palas VKL 100 diluter with particle-and VOC-free in-house compressed air. The flowrate of the O3 stream into the stainless steel aerosol chamber was maintained at 0.040 -0.070 L•min -1 . The O3 concentration within the aerosol chamber was monitored by an Aeroqual series 940 transmitter (0 -0.5 ppm) mounted on an additional port. For sample humidification, particle-and VOC-free compressed air was split into multiple streams. One stream was passed through a silica gel diffusion 15 dryer resulting in a RH of less than 5 %. The second stream was split and led through two Nafion-humidifier coil tubes surrounded by thermostated water resulting in a RH of up to 95 %. The temperature of the water was controlled by using an Ecoline Immersion thermostat E300 with Stainless Steel bath 006. The flow rates of the dry and the humidified air streams were regulated by individual mass flow controllers (MKS, 0 -20 L•min -1 ) and mixed within a 5 L glass volume. This set-up allowed for the stable production of air at a pre-set RH level and a flow rate of 20 L•min -1 . In addition to the sensors monitoring 20 the humidity within the aerosol chamber, a Vaisala HMT337 humidity sensor was used to monitor the humidification air before entering the aerosol chamber.

Sample characterization
A suite of instruments was deployed for the characterization of the soot particle samples. Besides stationary centerpieces for the determination of the particle size distribution and CCN-activity, the specific configuration of instruments varied depending 25 on availability. Data on the chemical composition, the INP activity, and the single-particle mass distribution was acquired in many but not all experiments.
A TSI Scanning Mobility Particle Sizer (SMPS) consisting of a TSI 3081L DMA and a TSI 3772 CPC was used to record the particle size distributions in the range of 8 -280 nm at a scanning frequency of 135 s. The DMA was operated with a sample flow rate of 1 L•min -1 and a sheath air flow rate of 10 L•min -1 . The CCN-activity of the soot particles was determined using a 30 continuous flow Cloud Condensation Nuclei Counter (CCNC) from Droplet Measurement Technologies (DMT; Roberts and Nenes, 2005). As size-selected particles were investigated, the CCN-activity was investigated by exclusively modulating the SS in the range from 0.2 % to 1.4 %. An additional data point at a SS of 1.6 % was acquired in CBK experiments. With an acquisition duration of 6 -10 min at each set-point, the sampling interval for a full scan was approximately 66 min.

2.5
Experimental procedure and experimental conditions

the right. The labels: A (bypass), B (filling), C (bypass), and D (flushing) indicate specific periods within the experiment.
The experimental procedure and therefore the data that can be acquired differs from experiments conducted in OFRs and batch reactors. As an example, a typical data set of an O3 experiment is displayed in Figure 2 (experiment #3 in Table 1, no denuder, 5 % RH). The CCN-activity is presented as colored crosses (right bottom axis), the particle number concentration as a black line (left bottom axis), the O3 concentration as a blue line (left top axis), and RH as a red line (right top axis), all as a function 10 of experimental duration (bottom axis).
While the experiment was conducted for about 18 h in total, individual time frames can be distinguished and are indicated by the capital letters A to Dthe four modes of operation -in Figure 2. Before the start of each experiment it was ensured that the aerosol chamber was particle-free, i.e. a particle concentration of < 1 cm -3 at the CPC downstream of the aerosol chamber.
Then the aerosol chamber was filled with the size-selected particles at a flow rate of 23 L•min -1 . During the first hour, a subset 15 of the aerosol stream (approx. 4 L•min -1 ) bypassed the aerosol chamber for the determination of the baseline characteristics of the aerosol. This period is indicated by the letter A (bypass). Data acquisition from the bypass was completed after completion of one full SS scan in the CCNC. From then on, the sample was extracted from the aerosol chamber, while the filling of the aerosol chamber continued. This period is indicated by the letter B (filling). Due to running the aerosol chamber in CSTR mode, a dynamic equilibrium was established within the aerosol chamber after a certain time. Once sufficient data of the 20 aerosol in that stage was acquired (at least 3 full scans with the CCNC) another bypass sampling period was started. This period is indicated by the letter C (bypass) in Figure 2. Data of this period were used to ensure that no changes in the particle production caused changes in the particle properties since the start of the experiment. Similar to the procedure in period A, one full SS scan in the CCNC was executed before this sampling period was completed by returning to sampling from behind the aerosol chamber. Simultaneous to changing the sampling extraction location, the supply of fresh soot particles into the 25 aerosol chamber was stopped. The particle-containing inlet flow was replaced by particle-and VOC-free compressed air. This period is indicated by the letter D (flushing).
The experimental procedure is reflected by the change in particle number concentration (black line, left bottom axis) presented in Figure 2. The values present the concentration within the sampling line just in front of the measurement instruments.
Therefore, values recorded within the bypass periods A and C present the particle number concentration in the bypass section and values recorded within the periods B and C present the particle concentration in the reaction chamber. At the beginning of period B, an increase in particle number concentration is recorded asymptotically approaching a plateau in the dynamic equilibrium. The particle loss rate within the aerosol chamber is significantly higher compared to the bypass section, therefore, the particle number concentration within the plateau is lower compared to the periods A and C. Throughout the flushing period 5 D, an exponential decay of the particle number concentration is recorded in accordance with theoretical expectations as no fresh particles are supplied to the aerosol chamber. The experimental conditions within this period can be considered similar to standard batch experiments. Monitoring and active control of the particle number concentration, RH, and temperature within the feed-in flow as well as O3 concentration, RH, and pressure within the aerosol chamber ensured consistent experimental conditions within the aerosol chamber. The O3 concentration (blue line) as well as RH (red line) within the aerosol chamber 10 are shown in the top panel in Figure 2, illustrating the constant conditions throughout the duration of the entire experiment.
The experimental conditions investigated within this study span a multidimensional space as O3 (0 or 200 ppb), RH (5 or 75 %), and gas-phase VOC content (sample denuded or undenuded) were varied. Each of these settings was repeated at least twice. A summary of the experimental conditions is shown in Table 1.
The experiments conducted in summer 2016 were executed at room temperature, which varied from day to day. During the 15 experiments in summer 2018, the temperature in the aerosol chamber was actively controlled and held at 25 °C in addition to the room temperature being controlled and maintained at 23 °C. The chamber temperature ensured constant reaction conditions and the room temperature ensured constant operating conditions of the measurement instruments. Table 1: Summary of all experimental results. The type of soot sample is given in the first column. The experimental conditions, the experiment number, and the experiment date are given in the second, third, and fourth columns, respectively. Experimental activation times at the respective SS are given in the following 9 columns. The reaction temperature and the O3 concentration are given in the last two columns. The values in brackets denote the standard deviation (SD).

CSTR mode versus batch mode
As can be seen in Figure 2, the transformation of fresh soot particles at an atmospherically relevant O3 concentration demands multiple hours of reaction time before CCN-activity of the particles can be detected (after ~2 h). In batch mode operation, a 5 reaction volume is first filled with the sample aerosol as fast as possible to achieve high homogeneity of the sample. After the desired starting concentration is achieved, further addition of the sample aerosol is stopped and the aging is initiated e.g. by addition of the oxidant. This point in time is generally referred to as t = 0 in such experiments. Analysis of the sample takes place while the reaction volume is flushed with sample-free gas. The aerosol chamber available at ETH Zurich has a volume of ~3 m 3 . Therefore, it was not possible to achieve sampling times of up to 12 h at the flow rates demanded by the suite of 10 instruments deployed if the aerosol chamber was operated in batch mode. With the aim to perform aging experiments at atmospherically relevant oxidant concentrations and to allow for atmospherically representative aging durations, the aerosol chamber was operated in CSTR mode. As mentioned previously, this mode of operation is characterized by a continuous addition of fresh aerosol simultaneous to a continuous extraction of the sample while the reaction conditions (e.g. oxidant concentration) are kept constant in the reaction volume. 15 While the entire aerosol is uniformly aged in batch experiments, the aerosol within a CSTR setup consists of a homogeneous mixture of differently aged aerosol particles. The continuous extraction of particle sample taking place concurrently to the addition of fresh particles causes fresh and old particles of varying residence times to be present simultaneously. Supported by the active mixing of the fan, the extracted sample consists of a homogeneous mixture of the particles in the aerosol chamber.
The distribution of the particles in terms of their residence time within the aerosol chamber is well defined as it solely depends 20 on the characteristics (e.g. volume) and operating conditions (e.g. flow rates) of the aerosol chamber operated in CSTR mode.
The variability in residence times is referred to as Residence Time Distribution (RTD) under ideal conditions and Particle Age Distribution (PAD) under real conditions as will be discussed in the following section. A more detailed description of the experimental approach used here can be found in Friebel and Mensah (2019).

Activated Fraction 25
The CCN-activityof the soot particles is presented as activated fraction (AF), which is calculated by dividing the number of When the experimental settings are switched to flushing (phase D), i.e. no fresh particles are supplied to the aerosol chamber any longer, a steep increase in AF can be observed while the particle number concentration decreases exponentially. In theory, a maximum AF of 1 should be reached at all SS levels if sufficient experimental aging time was permitted. In the case of the experiment shown in Figure 2, the experimental duration is sufficient to allow for an AF of 1 at SS of 1.4 % and 1.2 % only.
Due to the CSTR concept, the evolution of AF over time is significantly different from experiments conducted in batch-mode aerosol chambers, therefore a different data analysis concept is required.

Activation time tact
Generally, the critical supersaturation (SScrit) is reported from batch experiments to present the CCN-activity of the particles.
The SScrit is defined as the SS where an AF of 0.5 is reached at a specific time after the start of the experiment. This parameter 5 cannot be extracted directly from CSTR data as presented herein. Instead, the new parameter the activation time (tact) will be used as a reference parameter as has been introduced by Friebel and Mensah, (2019). Although tact and its corresponding activation supersaturation (SSact) are not identical to the SScrit, after a defined aging time, both data sets are comparable.
While the transformation caused by the O3 oxidation can be considered a continuous process, the change in CCN-activity of an individual soot particle at a defined SS is discontinuous and can be referred to as a non-gradual transition or a transition 10 within a binary system as a particle is either inactive or active. In this context, tact represents the minimum aging time a single soot particle requires to cross a certain transformation threshold. The tact-concept is valid for any transformation process involving a threshold. In the specific case presented herein, this process corresponds to a change in CCN-activity. As can be seen in Table 1, tact is dependent on the SS. The higher SS, the shorter tact. In other words, the higher SS, the less transformation and therefore the less time is needed to cause CCN activation of a particle. The AF can, therefore, be defined as the fraction of 15 particles that is older than tact. Assuming ideal conditions, tact in steady state can be calculated following eq. (1), with CSTR being the hydrodynamic residence time which is defined as the ratio of the volume of the CSTR (VCSTR) to the total flow rate through the volume (V̇) (Friebel and Mensah, 2019).

Particle losses
Knowledge of the particle age distribution (PAD) inside the aerosol chamber is required for the extraction of tact. In case particle losses are negligible, the PAD within the aerosol chamber is identical to the residence time distribution (RTD) of the particles as shown in the equation below (eq. (3)). 25 If particle losses occur, the PAD deviates from the RTD. As apparent by the reduced particle number concentration within the aerosol chamber in steady state compared to the bypass measurements, significant particle losses occurred in the aerosol chamber. In fact, there were two processes occurring simultaneously which cause a reduction in particle number concentration.
First, the particle loss to any surface within the aerosol chamber e.g. the aerosol chamber walls. Since this loss process can be 30 described by a first-order loss kinetic, the loss rate (kloss) is directly proportional to the particle number concentration. Second, the particle removal due to sample extraction (kCSTR), which can be considered a loss process as well. Since both processes follow the same kinetic they can be combined by introducing the effective particle loss rate kage and its reciprocal, the particle lifetime τ age . To obtain kage for the two first-order particle loss processes the individual loss rate constants have to be summed up as shown in eq. (4) below. 35 PAD(t)= e -t τ age (5) Here, the particle wall loss rate constant is kloss. The particle flush rate constant (kCSTR) is the inverse of the hydrodynamic residence time CSTR. The particle age distribution (PAD) can finally be calculated by substituting τ CSTR in eq. (1) by the real particle lifetime (τage) from eq. (4) leading to eq. (5) as shown above. The individual loss rates were determined in every single 5 experiment according to the following procedure. The decay in particle number concentration recorded during flushing (period D) was defined as the total loss rate kage. Assuming ideality of the set-up, the experimental flush rate is expected to be equal to the theoretical flush rate (kCSTR), which can be calculated based on the flow rates (eq. (2)). Therefore, the difference between the experimental value and the theoretical value corresponds to the wall loss rate kloss.
During the first campaign (Summer 2016) when the majority of CBW experiments were performed, τ age ranged from 96 to 10 102 min and τ loss from 500 to 800 min. During the second campaign (Summer 2018) when the CBK experiments were performed, τ age ranged from 100 to 108 min. The increased particle lifetime was a result of a reduced wall loss rate and therefore an increased τ loss ranging from 1000 to 2000 min. We attribute this pronounced change in wall loss rate within the aerosol chamber to the fact that the aluminum fan was gold-plated prior to the second campaign.

CCN-activity
Following the discussion in the previous section, the real tact can be calculated by replacing the ideal hydrodynamic residence time (τCSTR) in eq. (1) with the real particle lifetime (τage) from eq (4) leading to eq (6) shown below.
20 Table 1 provides an overview of all experiments performed, including the various experimental conditions employed.
Significant CCN-activation was observed only in experiments with an O3 concentration of ~200 ppb. Contrary to the impact of O3, neither elevated humidity conditions nor VOC denuding had an effect detectable with the instrumentation deployed. In Figure 3, the activation time (tact, left axis) as a function of the activation supersaturations (SSact, bottom axis) is presented for the same experiment as shown in Figure 2 (100 nm CBW, 200 ppb O3, RH 5 %, no denuding; experiment #3 in Table 1).
The right axis shows the cumulative O3-exposure in molec•s•cm -3 , which is the product of the O3 concentration and the exposure 5 time. In our case, the exposure time corresponds to tact. While an activation time of 152 min was determined at an SSact of 1.4 %, the activation time was more than three times higher at an atmospherically relevant SSact of 0.3 % (524 min). An activation time of 617 min (>10 h) was determined at an SSact of 0.2 % highlighting the capability of achieving atmospherically relevant aging durations within an aerosol chamber run in CSTR mode. The activation times are calculated from the AF in steady state according to eq. (6). The vertical error bars represent an instrumental uncertainty of ± 12 min calculated by error 10 propagation from the instrumental uncertainties of the CCNC and CPC. The horizontal error bars represent a 5 % uncertainty in the SS inside the CCNC following the recommendations of Rose et al. (2007). Details on the error calculation are presented in Friebel and Mensah, (2019).
As can be seen in Figure 2, the increase of AF from zero appears later with decreasing SS. Similarly, lower AFs are determined for lower SS in steady state. Both aspects correspond to an increase in activation time with decreasing SS. In other words, the 15 modification caused by O3 oxidation needs a longer time to allow for CCN activation at lower SS s.
As mentioned previously, no CCN activation could be determined without exposure to O3. Nevertheless, the SSact and tact for CBW particles without O3 exposure can be estimated using kappa-Köhler theory (Petters and Kreidenweis, 2007). A spherical particle with a diameter of 100 nm that is non-hygroscopic (kappa = 0), but fully wettable (contact angle = 0°) should activate at a SS of 2.1 % (blue cross in Figure 3). As can be seen, this theoretical data point aligns well with the experimental results. 20 Overall, an almost linear decrease in tact was determined with increasing SSact as can be seen in Figure 3. However, this is a theoretical approach as there is no scientific evidence that would support a linear correlation between oxidation with O3 and SSact. Furthermore, it is unclear if the soot used here is fully wettable or only partially wettable (contact angle > 0°) which would demand a higher SSact at tact = 0 min.
In all experiments with an O3 concentration of 200 ppb, the same trend of decreasing tact with increasing SSact was observed 25 independent of soot type, RH and VOC conditions. Despite the similarity in the trend, the individual values of tact at the same SSact change by up to a factor of 2 between experiments of the same soot type and at the same experimental conditions in terms of O3 concentration, RH and VOC concentration. For example, looking at tact for CBW particles at 200 ppb O3, 5 % RH, without denuding, and at an SSact of 1.4 % leads to a value of 152 min for experiment #3 and 267 min for experiment #5. These two values differ by a factor of 1.8. This deviation is significantly higher than the instrumental uncertainty of ±12 min discussed 30 above. Further analysis of the experimental conditions in summer 2016 and additional test experiments indicate that the average reaction temperature inside the aerosol chamber had a significant impact on the activation time. With increasing reaction temperature, shorter tact's were determined. Because attempts to control the room temperature by air conditioning in summer 2016 were not sufficient to keep the reaction temperature stable, the temperature of the aerosol chamber itself was actively controlled in the experiments performed in summer 2018. 35

CAST-Black
During the second measurement campaign in summer 2018 a second soot type (CAST black; CBK) was investigated. The particles of this soot type are characterized by a significantly reduced OC content compared to CBW particles as presented in Table 4. CBK particles were exposed to 200 ppb O3 at RH of 5 % with a charcoal denuder in line. As no impact of RH and VOC could be determined in the campaign in summer 2016, these parameters were kept constant in all CBK experiments in 40 summer 2018. Nevertheless, the experimental setup was improved by implementing a direct temperature control of the chamber (see section 2.5) and by reducing particle losses (see section 3.4). A significant difference in CCN-activity upon O3 exposure was determined between the two soot types -CBW and CBK. As can be seen in Table 1, CBK particles show much lower CCN-activity than CBW particles. CBK particles had to be oxidized for 725 to 742 min in order to show CCN-activity at an SSact of 1.4 %, which corresponds to an increase in tact by a factor of 2 to 4 times compared to CBW particles(experiment #20 and #21 vs #1 and #2 in Table 1). Considering similar minimum aging 5 durations/tact's, CBW particles activate at an SSact of 0.4 % after 552 and 523 min (#1 and #2) while CBK particles demand an SSact of 1.6 % for activation after 552 and 584 min (#20 and #21), respectively. Overall, no CCN-activity of CBK particles could be detected at atmospherically relevant SS (0.3% -0.8 %; Pruppacher and Klett, 2010) within the maximum aging time of up to 12 h..

O3 Spike Experiments 10
For the experiments #17 and #19, soot particles were not exposed to O3 while the aerosol chamber was filled. Only after switching to the flushing mode the O3 concentration was ramped to 200 ppb within approximately 30 min. Once this concentration threshold was reached no further O3 was added. The O3 concentration decayed exponentially reaching a value of 50 ppb within 120 min after the O3 supply to the aerosol chamber was switched off. Despite the temporary exposure to O3, no CCN-activity at any SS could be detected within the remaining experimental duration of 6 hours. However, an increase in 15 the particle mean diameter of 3 nm was detected while the O3 was added to the chamber.

Discussion
In an attempt to attribute the change in CCN-activity to the heterogeneous oxidation with O3 we investigated various parameters. These parameters include particle size, reaction temperature, relative humidity, and VOC content of the sample.
The particle size was determined by DMA measurements. Size distribution measurements of the particles before and after 20 aging in the aerosol chamber revealed no substantial restructuring such as compaction of the particles. To the contrary, a slight growth upon O3 exposure was detected in the range of 3 nm. Such growth of particles has already been reported by Fendel et al. (1995) for metal and spark discharge graphite particles and by Kotzick et al., (1997) for spark discharge graphite particles.
A detailed analysis of this aspect is beyond the scope of this paper.
Experiments performed during the measurement campaign in summer 2016 were executed at room temperature without active 25 temperature control of the aerosol chamber. Despite an air conditioning unit being installed, the difference between the coldest and the warmest average daily temperature measured throughout the campaign was greater than 5 K. Referring to the results of two CBW experiments executed at the same conditions (200 ppb O3, 5 % RH, without denuder; experiment #3 to #5 in Table 1) , it can be seen that a decrease in the average chamber temperature is associated with an increase in activation time.
Such temperature dependency is in accordance with the expected impact of temperature on the reaction speed following the 30 van't Hoff rule. It is known from model simulations and experimental studies that the O3 oxidation of polycyclic aromatic hydrocarbons (PAH) and organic molecules with C=C double bonds require an activation energy of 40 to 80 kJ•mol -1 (Berkemeier et al., 2016;Lee et al., 2009;Pöschl et al., 2001;Stephens et al., 1989). Even though many different compounds can be found on soot surfaces, PAH's are considered to be a good reference compound (Slowik et al., 2004). A temperature change by 5 K would change the reaction speed and therefore tact by a factor of 2. The deviations in tact determined 35 experimentally are within the same order of magnitude as the theoretical calculations supporting the presumed impact of reaction temperature.
Investigation of the RH conditions revealed neither a short-term nor a long-term effect within the range 5 -75 %. Changes in the particle morphology could be considered as a short-term effect. Contrary to the impact of O3, no significant change of the particle diameter could be detected upon exposure of the particles to elevated RH conditions. Overall, our findings are 40 supported by Mahrt et al., (2018) who showed that the water uptake on CBW and CBK particles does not exceed the adsorption of one monolayer at RH below 90 %. Long-term exposure of the particles to elevated RH conditions showed no impact on the CCN-activity even after up to 12 h, which is independent of the soot type investigated within this study.
Homogeneous O3 oxidation of VOCs can lead to semi-volatile reaction products which in turn can condense onto pre-existing particles and thereby modify the particle's CCN-activity (Wittbom et al., 2014). Because the VOC concentration within the 5 aerosol chamber could not be determined directly, the impact of VOCs emitted by the miniCAST was evaluated by implementing a charcoal denuder into the experimental setup. No impact on CCN-activity or particle size could be determined for experiments with and without the denuder in line.

CCN-activity
In Figure 3 the tact as a function of SSact is presented. As can be seen, increasing SSact are associated with decreasing tact's . The 10 uncertainty in the determination of tact is ± 12 min and originates from the instrumental errors of the CPC and CCNC as reported by Friebel and Mensah (2019). Therefore, relative uncertainties in tact and the calculated O3-exposure are below 10 %.
Compared to the uncertainties reported for the OH-exposure from different OFRs which are on the order of a factor of 5 (Lambe et al., 2011;Simonen et al., 2017), the uncertainties for the approach used here are significantly smaller.
While the distinct mechanism that leads to the significant change in CCN-activity of oxidized soot (e.g. inverse Kelvin effect, 15 formation of soluble or surface-active compounds) cannot be identified, it can be ruled out that the change is due to a growth of the particle diameter. The average diameter increase (CBW: + 3 nm; CBK: + 1.5 nm) is too small to have a decisive impact on the CCN-activity. Furthermore, the growth of the diameter occurs on a time scale of max. 30 min and is therefore much faster than the change in CCN-activity which occurs over a time scale of multiple hours.
Overall, the soot particles show more pronounced CCN-activation after exposure to O3 than has been reported previously in 20 the literature. It should be mentioned that this assertion is qualitative, because the particle sizes and particle compositions vary across the different studies. Nevertheless, the cumulative O3-exposure, the product of the O3 concentration and the exposure time, can be taken as a metric for comparison. On that basis, 100 nm diameter CBK particles in our study (SSact = 1.6 % after 4.9 × 10 16 molec•s•cm -3 O3-exposure) show CCN-activity within the same order of magnitude as 150 nm kerosene diffusion flame soot particles investigated by Grimonprez et al., (2018) (SScrit = 1.4 % after 10 × 10 16 molec•s•cm -3 O3-exposure) and as 25 222 nm ethylene premix flame soot particle (SScrit = 1.5 % after 5 × 10 16 molec•s•cm -3 O3-exposure; Lambe et al., 2015).
The differences could be attributed to the different chemical compositions of the soot particles as well as the different experimental setups but indisputable statements cannot be made on the basis of the data currently available. Experiments by Lambe et al., (2015) were performed at an O3 concentration of up to 20 ppm and exposure times of 100 s. In contrast to that, the approach presented herein allows for atmospherically relevant oxidant concentrations (200 ppb) and exposure times (up to 30 12 h). Note the comparison approach in terms of the cumulative O3-exposure performed here is valid only if the reaction speed is directly proportional to the O3 concentration i.e. follows a first-order reaction kinetic. The validity of this assumption cannot be verified on the basis of the data presented herein.
The differing activation times of CBW and CBK particles investigated in the same experimental setup indicate an impact of the chemical composition. O3-exposures higher by a factor of 2 to 4 are required to cause the same level of activation for CBK 35 particles compared to CBW particles of the same size and experimental conditions. In view of the abundance of soot particles in the atmosphere, the increase in CCN-activity of CBW and CBK particles due to heterogeneous oxidation of soot particles can be considered as atmospherically relevant. A linear extrapolation to atmospheric O3 background concentration levels of 20 to 45 ppb (Hough and Derwent, 1990;Vingarzan, 2004) shows that CBW and CBK particles would become CCN-active at 0.3 % SS after 2 to 4 days and 4 to 16 days, respectively. These values lie within the range of the average atmospheric 40 lifetime of one week (Textor et al., 2006) and indicate that this aging pathway could be a significant source of CCN-active soot particles within the atmosphere.

6
Atmospheric relevance Similar as in a CSTR aerosol chamber, particles are constantly emitted into the atmosphere as well as constantly removed from the atmosphere except in case of plume events. As a result, a mixture of particles at different aging stages is present in the atmosphere. From this perspective, the atmosphere can be approximated to be a CSTR in steady state. This approximation indicates that CSTR data is at least as suitable for parameterizations in global climate models data obtained from other 5 experimental setups. We performed three experiments with the global aerosol-climate model ECHAM6.3-HAM2.3 Tegen et al., 2019) to evaluate if the change in CCN-activity of soot particles due to heterogeneous O3 oxidation has an impact on the cloud droplet number concentration (CDNC) and therefore on cloud properties from a global perspective.
The size distribution of atmospheric aerosol particles in ECHAM6.3-HAM2.3 is described by seven log-normal modes (four internally mixed modes and three externally mixed modes), the particle number concentration, and the mass mixing ratio of 10 up to five aerosol components (sulfate, BC, particulate organic matter (POM), sea salt, mineral dust). While the structure of the size distribution is prescribed, the particle number concentration and the mixing ratio of each component are computed prognostically for each mode (for details see Tegen et al., (2019)). All BC emissions (fossil fuel, bio-fuel, biomass burning) and the POM emissions from fossil fuel are emitted into the externally mixed Aitken mode. Sixty-five percent of POM emissions from bio-fuel, biomass burning and biogenic secondary organic aerosols are emitted into the internally mixed modes 15 and 35 % of these emissions are considered insoluble and emitted into the externally mixed Aitken mode (Zhang et al., 2012).
In the standard setting of ECHAM6.3-HAM2.3, which was used for the reference ( was extended by allowing BC and POM particles in the externally mixed Aitken mode to activate to cloud droplets using a parameterization developed based on the results from the CSTR aging experiments. Details of this parameterization will be described briefly in the next section and further details can be found in the appendix (section 8.2.). 30

Parameterization of experimental results
Following the scheme of Abdul-Razzak et al., (1998), the parameter B in eq. (10) and (12)  concentration of 200 ppb. In the first case tact,ref is equal to 10 h, which is derived from the experimentally determined tact for CBW. In the second case tact,ref is equal to 50 h, which is set based on an extrapolation from experimentally determined tact values for CBK. The effective tact is calculated from tact,ref and the O3 concentration in the grid box at each time step assuming a first-order reaction kinetics with respect to O3 in accordance with Friedman et al., (2011), Lambe et al., (2015 and Grimonprez et al., (2018) as shown below in eq. (7). More information can be found in the appendix 8.2.
Adapting eq. (3), which describes the residence time distribution in a CSTR, allows for the estimation of the particle age 5 distribution in the atmosphere as presented in eq. (8). The ideal mean particle lifetime ( CSTR ) is replaced by the average atmospheric life time ( atm ) of soot particles, i.e. 7 days according to Textor et al., (2006). PAD atm (t) = e -t τ atm (8) Integration of the atmospheric particle age distribution PADatm(t) from t = tact to t = infinity (eq. (9)) yields XCCN, the fraction of CCN-active BC and POM particles in the externally mixed Aitken mode.
Please note that the increase of the soot particle's hygroscopicity due to oxidation with O3 increases the soot particle removal 10 rate from the atmosphere, e.g. due to a higher wet-deposition rate. As a result, the average atmospheric lifetime of soot particles decreases. However, the reduction of the soot particle lifetime was below 2% in both scenarios. Since this lifetime reduction is statistically not significant, its impact on the CCN burden and CDNC was not considered within this study.

6.2
Resultsmodeling   (Table 2). This increase of 12.7 % is statistically significant. At this tact,ref, the maximal regional increase can be determined in the latitudes north of 60 ° with a relative increase of more than 70 %. Taking the transition behavior of black carbon particles into account (tact,ref = 50 h), which has been determined by the investigation of CBK particles, the global mean CCN burden is determined to be 24.6 × 10 10 m -2 . This increase of 9.7 % is statistically 15 significant, but 3 percentage points less than in the case of tact,ref = 10 h. Investigating the regional impact, the relative increase still maximizes in the latitudes north of 60 ° but is about half as strong as in the case of tact,ref = 10 h. Regional changes in CCN burden are pronounced where either the emissions and atmospheric burden of BC and POM are large (e.g. tropics) or where CCN concentrations are relatively low (e.g. central to northern Europe and Asia). We hypothesize that in regions where many CCN are available in the internally mixed modes, the additional CCN in the externally mixed Aitken mode compete with the 20 CCN from the internally mixed modes for the available water vapor. This competition is also considered in the parameterization  Figure 4 and S1 in the supplement). However, over the tropics the increase in CDNC is much lower than the increase in the CCN burden. The difference is caused 30 by the higher abundance of stratiform liquid clouds in mid latitudes compared to the tropics, which is indicated by the higher CDNC burden in mid latitudes (Figure 4b). Please note that ECHAM6.3-HAM2.3 accounts for CDNC from detrained cloud water of convective clouds but otherwise the convective cloud parameterization does not consider CDNC . Therefore, our simulations could underestimate the impact of heterogeneous ozone oxidation of soot particles on CDNC, in particular where convective clouds are common like the tropics. The largest changes in CDNC occur below about 700 hPa, i.e. for low-level clouds ( Figure S1). Again, the impact at tact,ref = 10 h is much more pronounced than at tact,ref = 50 h with a global mean CDNC burden of 3.8 × 10 10 m -2 (+ 17.8 % compared to REF) and 3.5 × 10 10 m -2 (+ 8.9 % compared to REF), respectively (

Conclusion
We successfully applied the CSTR approach for the investigation of the change in CCN-activity of two soot types. Here we present the results of experiments in which soot particles were exposed to 200 ppb O3 and varying levels of humidity for up to 10 12 h. The CSTR approach allows for a low particle input concentration (1000 to 1500 cm -3 ) and size-selection of particles (e.g. at 100 nm).
We show that the heterogeneous O3 oxidation is a process that can make soot particles CCN-active at atmospherically relevant SSs of 0.3 to 0.8 %. The general finding agrees with literature results underlining the applicability of the CSTR approach.
Nevertheless, the SSact in our experiments is significantly lower at the same O3-exposures compared to results obtained in other 15 experimental setups (Grimonprez et al., 2018;Lambe et al., 2015). The soot rich in OC (CBW) demanded 2 to 4 times less aging time (tact = 3 -6 h at 1.4 % SSact) than soot low in OC (CBK, tact = 12 h at 1.4 % SSact). In contrast to O3, no effect of RH (up to 75 %) or denuding of the gas phase was observed. Instead, it was found that temperature fluctuations of 5 K inside the aerosol chamber have a strong impact on the activation time tact and were the largest single contributor to the experimental uncertainties. 20 A test with a global aerosol-climate model, where a first-order kinetic was assumed, showed that the change in CCN-activity of soot particles that are not taken into account in the standard configuration can lead to statistically significant increases in CCN burden and CDNC burden. The strongest increases were observed where the soot burden was large, and/or the initial CCN concentration was rather low for both reference activation times investigated. In the case of the CDNC burden it is additionally beneficial if CCN do not have to compete for water vapor and stratiform liquid clouds are frequent. 25 Both, the discrepancy in activation levels between studies using different experimental approaches and the initial O3 adsorption detected by a particle diameter increase within minutes suggest that the underlying reaction mechanism might not be sufficiently well described by assuming first-order kinetics. Therefore, we suggest performance of tailored experiments with a focus on the effect of different O3 concentrations as well as different temperatures. This might allow for further insight into the reaction kinetics and improvement of the accuracy of extrapolations to atmospheric conditions. 30 Data availability: The data presented in this publication can be downloaded from http://dx.doi.org/10.5281/zenodo.2541937 . The scripts to produce Figures 4 and S1 can be downloaded from http://dx.doi.org/10.5281/zenodo.3452036 . We thank Zamin A. Kanji, Oliver F. Bischof, and Thomas Peter for their valuable discussions, and Ulrike Lohmann's group for their support.  Within ECHAM6.3-HAM2.3 seven log-normal modes are defined for the aerosol particle size distribution. The CCN-activity of each mode is characterized by component specific parameters as well as the particle size distribution within each mode according to Abdul-Razzak and Ghan, (2000). However, the model does not contain "soot particle" as a category but contains 20 BC and POM as separate categories. The two categories BC and POM together represent the properties of soot with a lesser or higher amount of organic material, respectively. Therefore, they are modified together to represent the change in CCNactivity of soot particles due to heterogeneous O3 oxidation.
So far BC and POM particles are considered to be not CCN-active within ECHAM6.3-HAM2.3, unless they are internally mixed with soluble components such as sulfate. Particles are transferred from the insoluble to the soluble mode when sufficient 25 sulfuric acid gas condenses on insoluble particles to form a monolayer of coating, or by coagulation with soluble particles. The product of the van't Hoff factor and the osmotic coefficient describes the solubility of the particles in water. Within ECHAM6.3-HAM2.3, • is set to 0, because BC and POM particles are considered to be not CCN-active. However, based on the experimental results we presented, it is possible to calculate a new value of • for BC and POM particles, which is then used in the model experiments. POM, respectively. Note, within the actual model experiments is not set to 1 but it represents the mass fraction of BC/POM in the externally mixed Aitken mode in order to calculate the average SS in the externally mixed Aitken mode.
Please note that the treatment of secondary organic aerosol (SOA) is simplified in ECHAM6.3-HAM2.3. During emission the soluble and insoluble fractions of SOA are assumed to condense immediately on the soluble Aitken and accumulation and 5 insoluble Aitken mode respectively. However, He et al. (2016) accounted in their experiments for BC aging by condensation of SOA and found that their chemical aging mechanism still accounted for more than 50% of the BC aging rate at high latitudes (polewards of 60°N/S) and above 900 hPa.