Effects of oligomerization and decomposition to the nanoparticle growth, a model study

The rate at which freshly formed secondary aerosol particles grow is an important factor in determining their climate impacts. The growth rate of atmospheric nanoparticles may be affected by particle phase oligomerization and decomposition of condensing organic molecules. We used Model for Oligomerization and Decomposition in Nanoparticle Growth (MODNAG) to investigate the potential atmospheric significance of these effects. This was done by conducting multiple simulations with varying reaction-related parameters (volatilities of the involved compounds and reaction rates) using both 15 artificial and ambient measured gas phase concentrations of organic vapors to define the condensing vapors. While our study does not aim at providing information on any specific reaction, our results indicate that particle phase reactions have significant potential to affect the nanoparticle growth. In simulations where one-third of a volatility basis set bin was allowed to go through particle phase reactions the maximum increase in growth rates was 71% and decrease 26% compared to base case where no particle phase reactions were assumed to take place. These results highlight the importance of investigating and increasing our 20 understanding of particle phase reactions.


Introduction
Aerosols are ubiquitous in the atmosphere and they affect our climate in multiple ways. Directly they can affect the radiative forcing by reflecting, refracting, and absorbing sunlight, and indirectly by acting as cloud condensation nuclei (CCN) and forming clouds (Boucher et al., 2013). The effect of aerosol-cloud interactions in the Earth's radiative balance is one of the 25 biggest uncertainties we have in recent climate models and studies (IPCC, 7. chapter, Boucher et al., 2013).
For an aerosol particle to act as CCN, it needs to be large enough in size, at least some tens of nanometers in diameter (Pierce and Adams, 2007;Reddington et al., 2017). This can be, on one hand, a limiting factor for climate impacts of small primary aerosols as Aitken-mode sized primary particles such as soot particles are quite often non-hygroscopic, which hinders their activation as CCN (e.g. Zhang et al. 2008, PNAS). However, atmospheric aging typically enhances their solubility and alter their morphology towards being CCN active (Trischer et al. 2011;Lambe et al. 2015,). On the other hand, the secondary aerosols that are formed in the atmosphere via gas-to-particle conversion (e.g., Kulmala et al. 2014,.) need to undergo substantial growth until they reach sizes relevant for CCN activation (Kerminen et al. 2012). Regardless, it is estimated that approximately half of the particles acting as CCN are formed in the atmosphere by nucleation from atmospheric gases (Merikanto et al., 2009;Paramonov et al. 2015). 35 The first part on right in Eq. (1) describes the transition regime mass flux of condensation to/evaporation from a particle and is based on the difference in the gas phase and equilibrium concentrations of the compound. βi is the transition regime 95 correction factor defined as (Fuchs and Sutugin, 1970 where αm,j is mass accommodation coefficient and Knj is the Knudsen number. In our model, for Knudsen number we use (Lehtinen and Kulmala, 2003): 100 where λj is the free mean path of the condensing compound j. The mean free path is defined as (Lehtinen and Kulmala,2003): where cp and cj are the mean thermal speed of the particle and condensing compound, respectively. The equilibrium vapor 105 concentration of j is calculated as , = * exp( 4 ) where γj is the activity coefficient, χj the mole fraction, C*j the saturation concentration, νj the molar volume of compound j, σ the surface tension of the particle, R the gas constant and T the temperature. In this study we assume an ideal solution and therefore the activity coefficient γj is equal to 1.
The second part on the right-hand side of Eq. (1) describes the production and loss of the compound j from oligomerization 110 and decomposition reactions. Lolig,j describes the loss rate of compound j due to oligomerization reactions with other compounds i (Lolig,j ≠ 0 only for organic group II) and is calculated as where kolig is the oligomerization rate coefficient, Vp the volume of the particle and Nj and Ni the particle phase molecular concentrations of compounds j and i, respectively. Ldec,j describes the loss rate of compound j due to decomposition to smaller 115 molecules (Ldec,j ≠ 0 only for organic group III) and is calculated as where kdec is the decomposition rate coefficient of compound j. Polig,j and Pdec,j describe the production of compound j by oligomerization and decomposition, respectively, (≠ 0 only for organic group IV) and they are calculated as , = where f, y, i indexes describe the two oligomerizing compounds from the group II and the fragmenting compound in the group III, respectively.
Eq. (1) is used both for organics and sulfuric acid. Water is assumed to be constantly in equilibrium between gas and particle phase and the amount of ammonia (by mole) is assumed to equal the amount of sulfuric acid in the particle phase. The model 125 is built with an option of including particle phase acid-base chemistry according to E-AIM (http://www.aim.env.uea.ac.uk;Clegg et al., 1992;Clegg and Seinfeld, 2006a, b;Wexler and Clegg, 2002) similar to MABNAG model (Yli-Juuti et al., 2013).
However, here an option of ideal solution assumption without acid-base chemistry was applied as the focus is on the organics, although we acknowledge that acidity can enhance the oligomerization (Tolocka et al, 2004).
The particle is assumed to be liquid like and have no particle phase diffusional limitations. The viscosity of the particle has 130 been suggested to possibly have an effect on the particle growth (Virtanen et al., 2010), however the effect may not be significant at atmospheric boundary layer relative humidities at least at warm environments (Renbaum-Wolff et al., 2013;Yli-Juuti et al., 2017;Li et al., 2019), and here such effect was neglected to focus on oligomerization and decomposition.

Simulation setup 135
In MODNAG it is possible to include multiple oligomerization and decomposition reactions in the same simulation. In recent studies it has been shown that oligomerization is often reversible (Trump and Donahue, 2014). In our simulations we have mostly assumed irreversible reactions but conducted simulations with reversible oligomerization for few example cases to explore potential influence of this process. Also, for simplicity we have assumed that only one reaction happens at a time and that it happens only between two compounds, although in reality the reaction chains are observed to be longer and include 140 multiple reactions and compounds (Tolocka et al.,2004, Kolesar, Heaton et al., 2007. In all simulations the initial particle diameter was 2 nm, and it consisted solely of sulfuric acid. Our analysis included "oligomerization simulations" and "decomposition simulations". In the oligomerization simulations one pair of compounds in organic group II was allowed to react and form a dimer. Other compounds in group II and all compounds in group III were assumed to be non-reactant similarly to group I compounds. We run such simulations for 145 reactions between all possible compound pairs in organic group II. For each pair of reacting compounds, several simulations https://doi.org/10.5194/acp-2021-571 Preprint. Discussion started: 14 July 2021 c Author(s) 2021. CC BY 4.0 License.
were run by assuming the product to have a saturation concentration between C*=10 -6 µg m -3 and 10 1 µg m -3 , however the volatility of the product was restricted to be always at least one order of magnitude lower than the volatility of the less volatile reacting compounds. For each combination of the pair of reacting compounds and C* of the product, the analysis included simulations where oligomerization rate coefficient kolig ranged from 10 -27 m 3 s -1 to 10 -18 m 3 s -1 . Sensitivity tests showed that 150 with higher or lower oligomerization rate coefficients there were not any significant changes in the results compared to these upper and lower limits, respectively.
In the decomposition simulations the initial compound from organic group III fragments forming two smaller product compounds. These two product compounds could be identical or have different properties. Other compounds in group III and all compounds in group I and group II were assumed to be non-reactant. For each decomposing compound from organic group 155 III, simulations were run with volatilities of the product compounds ranging from C*=10 -3 µg m -3 to 10 2 µg m -3 . The volatility of each product compound was limited to be always at least one order of magnitude higher than the initial compound's volatility. For each combination of the initial compound and the pair of product compounds, simulations were run with decomposition rate coefficient kdec ranging from 10 -5 s -1 to 1 s -1 .
In the simulations with reversible reaction oligomerization of compounds was done similarly as in irreversible oligomerization 160 simulations described above, with the exception where the formed oligomers could decompose back to their initial group II bins after oligomerization. Ranges for kolig and kdec were similar to the irreversible reactions.
The total gas phase concentration of compounds in different VBS bins was divided evenly between groups I, II and III, which means, that one-third of a bin was reacting in a given simulation. We chose this in order to investigate effects of particle phase reactions in a more moderate case compared to assuming that all or majority of the compounds of any volatility would undergo 165 reactions. An assumption for this was required since relevant particle phase reactions of organics are not well known. We tested the sensitivity of results to this assumption by performing additional simulations with the assumption that all molecules of a bin can react (see Results,Sect. 3.2).
In this study the aim was to get an overview of how much oligomerization and decomposition can affect the growth of atmospheric nanoparticles. For this, the above sensitivity runs were performed for two scenarios being representative of 170 environments where the nanoparticle are growing. In Case 1, an artificial gas phase composition was given as an input to the model. The vapor concentrations were selected in a way that vapor concentrations and the resulting particle growth rates are of similar magnitude as observed in the boreal forest atmosphere (Mohr et al., 2019). Additionally, the less volatile organic compounds were set to have lower concentrations compared to the more volatile compounds following atmospheric observations but in a simplified way (Mohr et al., 2019). Properties of the seven model compounds (VBS bins) in all 175 condensing groups (I-III) are illustrated in Table 1. In addition, we have 280 product compounds (group IV). Here the diffusivity is assumed to be similar with condensing components in similar volatility bin. For oligomerization product compounds the molar mass of the compound was the sum of molar masses of reacting compounds, and for decomposing product compounds the molar masses were calculated by dividing the molar mass of decomposing compound relative to the logC* of the product compounds. Table 1: The properties of organic model compounds in groups I, II and III, for Case 1, where the properties are artificial, mimicking atmospheric conditions. C* is saturation concentration, M molar mass, D particle phase diffusion coefficient and C gas phase concentration. C* is expressed both in units of µg m -3 and molec cm -3 .

185
In Case 2, simulations were run with gas composition more directly restricted by atmospheric observations. Vapor concentrations and molecular composition measured with a chemical ionization mass spectrometer at Hyytiälä measurement station (Mohr et al., 2019) in spring 2014 during a new particle formation (NPF) event were grouped in a VBS based on their C* estimated with the parameterization by Li et al. (2016) and temperature dependence of C* estimated based on the method 190 by Epstein et al. (2010). This VBS representation of the gas concentrations of the organics was used as an input for the model.

210
3 Results and discussion

Simulations based on artificially generated gas phase concentrations
The particle growth in simulations with artificially generated gas phase concentrations (Case 1) are presented in Figure 1. In general, our results show that oligomerization increases, and decomposition decreases the particle growth rate. At maximum, 215 the growth rate was increased 139% by oligomerization and decreased 20% by decomposition. In some simulations the growth rate is decreased also by oligomerization. These are simulations where two low volatile (C* < 10 -2 µg m -3 model compounds are forming a dimer and the product is only one order of magnitude less volatile than the initial compounds. In this situation, due to small difference in volatilities between the reacting monomers and the product dimer and zero gas phase concentration of the product compound, the evaporation rate of product compound exceeds the enhancement of condensation due to 220 oligomerization. bin 1 10 2 / 3.04*10 11 0.198 5.51*10 -6 7.10*10 6 bin 2 10 1 / 2.77*10 10 0.217 5.20*10 -6 5.90*10 6 bin 3 10 0 / 2.51*10 9 0.240 5.06*10 -6 3.97*10 6 bin 4 10 -1 / 2.35*10 8 0.256 5.00*10 -6 2.85*10 6 bin 5 10 -2 / 2.08*10 7 0.290 4.56*10 -6 1.83*10 6 bin 6 10 -3 / 1.86*10 6 0.323 4.26*10 -6 1.41*10 6 bin 7 10 -4 / 1.60*10 5 0.376 3.99*10 -6 2.61*10 6 https://doi.org/10.5194/acp-2021-571 Preprint. Discussion started: 14 July 2021 c Author(s) 2021. CC BY 4.0 License. The effects of different parameters to the growth with oligomerization can be seen in Figure 2, where simulations with different 230 oligomerization rate coefficient (Fig. 2 a, b), saturation concentrations of oligomerization product (Fig. 2 c, d) and saturation concentration of one of the oligomerizing compounds ( Fig. 2 e, f) are presented. Each subfigure shows the base case simulation with no reactions (red dashed line) and the simulation with the fastest growth, where compounds from bins 2 (C* = 10 1 µg m -3 ) and 3 (C* = 10 0 µg m -3 ) form a compound with two orders of magnitude lower saturation concentration than in bin 7 (C* = 10 -6 µg m -3 ) with oligomerization rate coefficient kolig of 10 -18 m 3 s -1 (black dashed line). Growth rate (GR) is the changing 235 rate of the particle diameter and it was calculated based on differences in simulated diameter between each time step. The growth rates increase with increasing kolig and with decreasing volatility of oligomerization product. Generally, the growth rate also increases with increasing volatility of oligomerizing compounds. However, with very high volatilities (C* > 10 0 µg m -3 ) the tendency of these compounds to evaporate can hinder the oligomerization reaction. Also, if two compounds from same bin react, the growth is hindered for two reasons: 1) total gas phase concentration of reacting compounds is lower (one-third of 240 one bin) than in the case of compounds of different bins reacting with each other (one-third of each bin) and 2) the particle phase concentration reduction due to oligomerization is greater than in simulations with reactive compounds from two separate bins. When interpreting our simulation results for the small particle sizes it should be noted, that the initial assumption of particle containing only sulfuric acid may affect the results at the beginning of the simulation. As the initial particle contains no organics, some organics will condense in the particle fast during the first-time steps (due to the solution effect in Ceq) causing artificially high GR for the beginning of the simulation. For this reason, in Figs 2 b, d and f we present the GR only after the diameter reaches 3 nm. At this point the mass of the particle is about twice the initial mass. 265 Figure 3 shows effects of different parameters to the particle growth in simulation where decomposition is allowed. Simulations with different decomposition rate coefficient (Fig. 3a, b), saturation concentrations of decomposing compound (Fig. 3c, d), and saturation concentrations of one product compound (Fig. 3e, f) are shown. Each subfigure shows the base case simulation (no reactions, red dashed line) and the simulation with the slowest growth, where compound from bin 7 (C* = 10 -6 µg m -3 ) decomposes into two product compounds in bin 1 (C* = 10 2 µg m -3 ) with decomposition rate coefficient kdec of 1 s -1 (black 270 dashed line). Our results show that all varied parameters affect the growth of the particle. For kdec the effect is quite straightforward; with increasing kdec, the growth rate is decreased. The growth rate slows down with a decreasing C* of decomposing compound due to larger contribution of lower volatility compounds to the particle growth and with an increasing C* of the product as a consequence of the product evaporating faster for the higher C* compounds.  In Figure 4 we present the growth rate of the particle in all different simulations with oligomerization for particles under 5 nm in diameter (Fig 4a) and over 5 nm in diameter (Fig 4b). Each colored dot represents one simulation, and the color describes the growth rate. Growth rates are calculated by fitting a straight line in diameter as a function of time, i.e. assuming linear growth. This is important to notice especially with particles under 5 nm in diameter for in that size range the growth is not 290 usually linear (see Fig. 2 and 3).  For small values of kolig (< 10 -24 m 3 s -1 ) the increase in growth rate due to oligomerization is small, especially for under 5 nm particles, where any notable increase can be seen only after kolig > 10 -24 m 3 s -1 . Even for simulations where the most volatile compounds in our setup (bin 1 and 2) oligomerize, growth does not increase much with these low kolig, since compounds that condense to the particle phase will evaporate back to the gas phase before they have time to form less volatile product 305 compounds. With larger kolig however, oligomerization happens so fast that even these higher volatility molecules will oligomerize before evaporation. Without oligomerization reaction these compounds would not contribute to the growth almost at all, so with their oligomerization the growth is enhanced greatly. The gas phase concentrations of three higher volatility compounds are higher than those of lower volatility compounds, which enhances the growth rates even further in simulations where oligomerization takes place between high volatile compounds. 310 A clear result here is that the decomposition decreases the GR across the board. In more details, with a small kdec (< 10 -4 s -1 ) 325 the decomposition does not affect the growth, since the rate of reactions is slow compared to the condensation mass flux and therefore only a relatively small fraction of molecules reacts. The decomposition starts to have an impact if kdec is at least 10 -4 s -1 and the impact is dependent on the volatilities of the decomposing compound and the product compounds. If decomposing compound is one of the most volatile compounds in our setup, i.e., from three most volatile bins with C* > 10 -1 µg m -3 , the effect of decomposition on GR is very small, because of their low contribution to particle mass. Instead, if the decomposing 330 component is from least volatile bin (C* = 10 -4 µg m -3 ), the effect on GR is large, even if the C* of the product compounds would be as low as 10 -2 µg m -3 .
In most of our simulations and in all the results presented this far, the oligomerization and decomposition reactions are assumed to be irreversible. In Figure 6 we present a few cases, where we tested the effect of reversible reactions.
In subfigures a, c, e and g (Reaction A) compounds from two most volatile bins (1 and 2, C*= 10 2 µg m -3 and 10 1 µg m -3 335 respectively) form an ELVOC (bin 7, C*=10 -4 µg m -3 ) and in subfigures b, d, f and h (Reaction B) compounds from bin 2 and 5 (C*=10 1 µg m -3 and 10 -2 µg m -3 ) form a similar ELVOC as in the left-hand side reactions (bin 7, C*=10 -4 µg m -3 ). In both cases the oligomerization product can decompose into the initial compounds. In Reaction A the effect of oligomerization is large, since without it the reacting compounds would contribute to the growth very little. In Reaction B the effect of oligomerization is smaller, for especially the low-volatile reacting compound would condense to the particle phase even 340 without oligomerization. It is worth noting that the difference in gas phase concentrations between higher and lower volatility bins also contributes to the extent that the oligomerization enhances the growth rate. In Reaction B the effect of reversibility is seen only with small values of oligomerization rate coefficients (kolig < 10 -23 m 3 s -1 ) while in Reaction A the effect of different decomposition rate coefficients can be seen already with larger values of oligomerization rate coefficients (kolig > 10 -18 m 3 s -1 ). In reaction B the reversibility of the reaction has less effect on the growth because lower volatility reacting compound tends 345 to stay in the particle phase and thus helps driving the oligomerization reaction.

Simulations based on measured gas phase concentrations
Here we explore the thermodynamic parameters in the model and contrast the results to observations in an aerosol formation event observed at Station for Measuring Ecosystem-Atmosphere Relations (SMEAR-II, Hari and Kulmala, 2005) in Hyytiälä, Finland. Nanoparticle growth after nucleation has been extensively studied at Hyytiälä and nanoparticle GR is relatively well characterized there. At this location, GR values ranging from below 1 nm h -1 to several tens of nm hhave been observed (Dal Maso et al., 2005;Yli-Juuti et al., 2011) average GR for 3-25 nm particles being 2.5 nm h -1 (Nieminen et al., 2014). While sub-20 nm particle composition measurements are missing, seasonal variation of GR with maximum in summer indicates importance of organic vapors with biogenic origin (Dal Maso et al., 2005;Yli-Juuti et al., 2011). The importance of organics 365 is supported, e.g., by the positive correlation found between GR of 7-20 nm particles and monoterpene concentration (Yli-Juuti et al., 2011). Further, sulfuric acid condensation can explain only a small fraction of particle growth even down to sub-3 nm size range (Nieminen et al., 2014;Yli-Juuti et al., 2016) and the composition observations of 20 nm particles indicate that organics would cover more than half of particle mass growth (Pennington et al., 2013). On the other hand, GR of particles has been observed to increase with particle size and GR of sub-3 nm particles does not exhibit similar seasonal variation as GR of 370 larger particles which together suggest that there may be different factors affecting growth at different sizes (Yli-Juuti et al., 2011). Particle growth model constrained by observed gas phase concentration of organics captures the observed growth rate fairly well without need for assuming particle phase reactions (Mohr et al., 2019). However, due to uncertainties in gas phase concentrations and properties of organics, possibility of particle phase reactions cannot be completely overruled. Ehn et al. (2007) observations behind a thermodenuder determined size distributions showed that the growth rate was one-third of the 375 GR measured with a normal DMPS-system. Based on a long-term volatility measurements, Häkkinen et al. (2012) found out that soot is not able to explain the residual in the particulate phase and speculated for oligomerization to contribute to the nonvolatile cores of nanoparticles growing in the boreal environment.
An evolution of aerosol number size distribution in Hyytiälä on 23.4.2014 is presented in Fig. 7. The observed growth rate was 1.7 nm h -1 . In the subsequent simulations we explored the capability of the model and parameter selection to explain the 380 observed aerosol growth in the boreal environment.
Growth of the particle in simulations with measured gas phase concentrations (Case 2) can be seen in Figure 7a. The blue area envelopes simulation results with oligomerization reaction and yellow area envelopes simulation results with decomposition reaction. Without oligomerization the growth in the model is slower than the observed growth (GR in base case simulation 1.54 nm h -1 , observed GR 1.7 nm h -1 ). When oligomerization is allowed in the model, it is possible to reach similar GR as 385 observed. However, this can be achieved with multiple combinations of parameters, and therefore it would be challenging to try to estimate what kind of reactions take place in the growing particles by optimizing the model respect to the observed growth. For example, fitting growth rate can be achieved with simulation where compounds from bin 3 and bin 7 form an ELVOC (C* = 10 -6 µg m -3 ) with oligomerization rate coefficient of 10 -27 m 3 s -1 and also with simulation where two compounds from bin 1 form a LVOC (C* = 10 -2 µg m -3 ) with oligomerization rate coefficient of 10 -19 m 3 s -1 . Similar problem has also been 390 noted by Roldin et al. (2014) when analyzing particle evaporation in laboratory and by Trump and Donahue (2014) when comparing their model to the SOA formation measurements by Presto and Donahue (2006).  Uncertainties in saturation concentrations of organic compounds are another issue that makes it difficult to approximate which of the assumed oligomerization reactions would fit best with the observed growth. In this study, we used the parametrization of Li et al. (2016) to calculate the C* values based on molecular formula. Multiple other parametrizations have also been proposed (e.g., Donahue et al., 2011;Stolzenburg et al., 2018;Mohr et al., 2019) and these lead to somewhat different simulated 405 growth rates (Mohr et al., 2019). It should be noted that while the observational data used here was part of the analysis in Mohr et al. (2019), a different C* parametrization was used there as the base case and with that parameterization the growth was overestimated even without any particle phase reactions. Overall, the C* values of organics can vary over several orders of magnitude between different calculation methods (Mohr et al., 2019). Therefore, the reaction that would produce the best fit between observed and simulated GR may vary from oligomerization to decomposition between the different C* 410 parameterizations.
In the results presented so far, in each simulation one-third of a VBS bin was allowed to react. To see how this assumption affects model results, we made additional simulations where we allowed all molecules of a bin to react. The growth of the https://doi.org/10.5194/acp-2021-571 Preprint. Discussion started: 14 July 2021 c Author(s) 2021. CC BY 4.0 License. particle in these simulations is presented in Figure 7b. The possible contribution of oligomerization or decomposition reaction is remarkable. If only one-third of a bin was allowed to react the growth rate of the particle was in maximum increased by 415 oligomerization 71% and decreased 26% by decomposition, but if whole bin reacted, the maximum increase was 138% and decrease 80%.

Conclusions
A wide range of model simulations were conducted to study effect of particle phase oligomerization and decomposition on the nanoparticle growth. Based on our model results, these reactions have potential to affect particle growth. However, the extent 420 of the effect was strongly dependent on the assumed properties of the organics (volatilities of the initial and product compounds, reaction rate coefficients and fraction of molecules that are reactive) and the sensitivity of particle growth on one property depended on the other properties. In the simulations constrained by observed gas phase concentrations, the agreement between simulated and observed particle growth rate changed considerably when the assumptions of the organic properties were varied. However, simulated and observed growth rate can be brought to a good agreement with multiple combinations of 425 assumptions of the properties which would make it challenging to try to estimate which combination describes the condensing organic properties best.
When considering agreement between observation constrained growth model simulations and observations of particle growth, uncertainties in gas phase concentration measurements and in estimation of saturation vapor pressure of organics need to be considered. For example, the C* values vary over several orders of magnitude between different parametrizations (Mohr et al.,430 2019) and there are discrepancies in C* of organics bases on different measurement techniques (Bilde et al., 2015). Mohr et al. (2019) estimated the uncertainty for gas phase concentrations of organics, which were same as used in our study, to be 53% and considered an uncertainty of two orders of magnitude for saturation concentrations. In their model simulations these uncertainty limits lead to 46% and 64% increase and 41% and 27% decrease in growth rates for uncertainties in gas phase concentrations and saturation vapor pressures respectively. Compared to 71% increase by oligomerization and 26% decrease 435 by decomposition calculated in our study if one-third of a bin is allowed to react, these effects are similar in magnitude, but if whole bin is allowed to react, the effect by oligomerization and decomposition are greater. Within uncertainties, it is possible to explain the detected atmospheric nanoparticle growth based on the observed gas phase concentrations even without particle phase oligomerization and decomposition, as shown by Mohr et al. (2019) and our base case simulation. Nevertheless, oligomers are found in abundance in SOA and although some of it is oligomers condensed 440 straight from the gas phase, it is presumable that also particle phase oligomerization and decomposition occur as have been shown by multiple studies (e.g. Zhao et al, 2005;Zhao et al., 2006;Krizner at al., 2009 andWang et al., 2010). Hence, it remains open to what extent particle phase reactions take place in nanoparticles and how much is particle growth rate affected by them. Our results suggest that including these processes in models that describe atmospheric particle dynamics may be required, however, as the simulated growth is sensitive to the assumptions of reactions and reaction rates, also investigations