Bounce characteristics of α-pinene-derived SOA particles with implications to physical phase

Transcription

1 Boreal Environment Research 18: ISSN (print) ISSN (online) helsinki 28 June 213 Bounce characteristics of α-pinene-derived SOA particles with implications to physical phase Jonna Kannosto 1), Pasi Yli-Pirilä 2), Li-Qing Hao 2), Jani Leskinen 3), Jorma Jokiniemi 3)4), Jyrki M. Mäkelä 1), Jorma Joutsensaari 2), Ari Laaksonen 2)5), Douglas R. Worsnop 2)5)6), Jorma Keskinen 1) and Annele Virtanen 1)2) 1) Tampere University of Technology, Department of Physics, P.O. Box 692, FI-3311 Tampere, Finland 2) University of Eastern Finland, Department of Applied Physics, P.O. Box 1627, FI-7211 Kuopio, Finland 3) University of Eastern Finland, Department of Environmental Science, P.O. Box 1627, FI-7211 Kuopio, Finland 4) VTT Technical Research Centre of Finland, Fine particles, P.O. Box 1, FI-244 VTT, Espoo, Finland 5) Finnish Meteorological Institute, P.O. Box 53, FI-11 Helsinki, Finland 6) Aerodyne Research, 45 Manning Road, Billerica, MA , USA Received 13 Feb. 212, final version received 12 June 212, accepted 25 May 212 Kannosto, J., Yli-Pirilä, P., Hao, L. Q., Leskinen, J., Jokiniemi, J., Mäkelä, J. M., Joutsensaari, J., Laaksonen, A., Worsnop, D. R., Keskinen, J. & Virtanen, A. 213: Bounce characteristics of α-pinenederived SOA particles with implications to physical phase. Boreal Env. Res. 18: Electrical Low Pressure Impactor (ELPI) was used to study the physical phase of fresh secondary-organic-aerosol (SOA) particles formed by the ozonolysis of pure α-pinene in a chamber. The particles bounced from smooth impactor plates towards lower impactor stages, indicating a solid physical state. These particles had a similar bouncing ability as Scots pine-derived particles in previous studies. The measured bounce factor of large particles (diameter > 4 nm) did not significantly change during the particle growth, indicating no changes in the particle solidity. For the smallest particles, the calculated bounce factor increased as the particles grew. The smallest particles were less solid (i.e. having lower viscosity) than the larger ones. The maximum value of the bounce factor decreased for consequent impactor stages. According to a simplified model, this can be explained by a combination of bounce probability and charge transfer between the particles and the impactor surface if at least 5% particles are bouncing. Introduction Atmospheric aerosol is a complex mixture of primary and secondary material. Primary particles are emitted directly from sources such as traffic, industrial activity, volcanic eruptions etc. Secondary particles are formed in the atmosphere Editor in charge of this article: Veli-Matti Kerminen by gas-to-particle conversion processes such as nucleation, condensation, oxidation and other chemical reactions (Atkinson 2, O Dowd et al. 22, Kulmala et al. 24). Secondary aerosol contains ammonium, sulfates and different organic compounds. Vegetation, such as boreal forests, can produce a wide variety of volatile

2 33 Kannosto et al. Boreal Env. Res. Vol. 18 organic compounds (VOC) that participate in the formation and growth process of newly-formed secondary-organic-aerosol (SOA) particles. Recent studies estimated that the SOA particles cover 6% 7% of the organic aerosol mass on the global scale and locally even more (Kanakidou et al. 25, Hallquist et al. 29). Numerous modeling studies suggest that the SOA particles significantly affect the climate (e.g. Spracklen et al. 26). Until recently, the SOA particles have been assumed to be in liquid state in atmospheric conditions (Pankow 1994, Odum et al. 1996, Marcolli et al. 24). Virtanen et al. (21) reported bounce properties of SOA particles formed from the oxidation products of the VOCs emitted by Scots pines in a smog chamber and of the VOCs formed in the boreal forest. Based on the bounce behavior of particles and electron microscopy analyses, they concluded that the studied SOA particles were amorphous semi-solid or solid particles. The bounce was studied using an Electrical Low Pressure Impactor (ELPI). Virtanen et al. (21) defined a so-called bounce factor, which can be calculated from the measured ELPI currents. In the particle size range of > 3 nm, the bounce of particles was almost independent of particle size. In a later study (Virtanen et al. 211), the bounce behavior of Scots pine-derived particles was further studied by an ELPI with an additional low-cut diameter impactor stage. It was found that the bounce decreased with a decreasing particle diameter for particles below 3 nm. The bounce factor measured using the additional stage was approximately.15, as compared with the higher value of.3 for the original instrument setup. Recently, Cappa and Wilson (211) and Vaden et al. (211) reported observations of solid-phase α-pinene SOA particles based on particle evaporation characteristics studies. Due to possible kinetic limitations related to the mixing of adsorbed molecules in the particle bulk, the semi-solid or solid phase might have influence on several atmospheric processes, such as the water uptake, partitioning between the gas and particle phase and, for example, oxidative reactions taken place in the particle phase (Zahardis and Pertucci 27, Zobrist et al. 28, Murray 28, Mikhailov et al. 29, Shiraiwa et al. 211, Pfrang et al. 211). Chamber studies have been widely used as model studies for atmospheric SOA systems, so that conclusions related to the atmospheric SOA particle formation and characteristics have been drawn based on chamber measurements. It is thus important to study the physical phase of SOA particles formed from common precursor VOCs used in chamber studies. A widely-used precursor is α-pinene because it is globally the most abundant atmospheric monoterpene and covers one fourth of the total terpene emissions (Kanakidou et al. 25). In boreal forest air, α-pinene contributes to about 48% of total emitted monoterpenes (Kulmala et al. 21). In this paper, we concentrate on studying the physical properties of SOA particles formed via the dark ozonolysis of pure α-pinene in a smog chamber. The bounce behavior of SOA particles was studied during the particle growth process and the results obtained were compared with particles formed from pine emissions. The bounce factor of small particles was studied using an ELPI with the additional impactor stage. We also made a first attempt to explain the impactor stage-dependent bounce factor in terms of the true bounce probability of the particles. Experiments The chamber setup used in this study has been described in detail by Hao et al. (29). Briefly, the setup consists of a reaction chamber (made of Teflon TM FEP film with a volume of 6 m 3 ), and gas and particle measurement systems that are described in greater detail below. Apart from using real Scots pines, measurements were made using the same experimental setup as reported by Virtanen et al. (21). Pure α-pinene was used here to form the SOA particles by the dark ozonolysis. First, α-pinene was injected into the chamber in order to obtain a chamber concentration of 7 to 45 ppb. Next, air with 7 ppb of ozone, produced by a UV lamp O 3 -generator, was added to the chamber at a flow rate of 4 l min 1 for approximately 8 min. In all experiments, the chamber temperature (T ) was kept at 22 ± 1 C and the relative humidity (RH) was kept at 3% ± 5% (measured with Vaisala Humidity and

3 Boreal Env. Res. Vol. 18 Phase of SOA particles 331 Temperature Probe HMP5). Ozone (DASIBI 18-RS O 3 analyzers, Dasibi Environmental Corporation), NO x (AC 3M NO x analyzer Environment s.a.), SO 2 (AF21M SO 2 analyzer Environment s.a.) and VOC concentrations were measured inside the chamber during experiments. VOC samples were collected on Tenax- TA adsorbent and the samples were analyzed by gas chromatography-mass spectrometry (Vuorinen et al. 24). Three separate experimental runs (A, B and C; see Table 1) were performed in order to study the physical state of SOA particles generated from pure α-pinene. Experimental runs A and B had different initial VOC concentrations, while the impactor plates in both runs were smooth. Since only one ELPI with the new impactor stage was available, we needed two measurements to compare the different impactor plates. Therefore, run C was a repeat of run B but with porous impactor plates. The chamber parameter in experimental runs B and C were very close to each other. For comparison, corresponding data for an experiment run reported by Virtanen et al. (21) are also presented (see Table 1). In that experiment, particles were produced from VOCs emitted by a living Scots pine. The particle mobility size distributions were measured using two scanning mobility particle sizers (SMPS), one with a Differential Mobility Analyzer (DMA; model 381, TSI, Inc.) coupled to a condensation particle counter (CPC; model 322, TSI, Inc.), and the other one with a nano DMA (model 385, TSI, Inc.) coupled to an ultrafine CPC (model 327, TSI, Inc.). The two instrument combinations covered mobility diameter ranges of 1 7 nm and 3 6 nm, respectively. Furthermore, the chemical composition and mass size distributions of aerosol particles formed during the experiments were measured with a quadrupole-based Aerosol Mass Spectrometer (AMS, Aerodyne Inc.). Particle aerodynamic size distributions and bounce behavior of the particles were determined using an electrical low pressure impactor (ELPI, Dekati Ltd.). In the ELPI, particles are charged by a unipolar corona charger, and after the charger particles enter a 12-stage cascade impactor (Keskinen et al. 1992). Recently, a new impactor stage was developed to improve the basic impactor setup, and to enhance the resolution for the smallest particles (Yli-Ojanperä et al. 21). The lowest cut-point of a classic impactor is 28 nm and the cut-point of the new impactor stage is 17 nm. The new stage was inserted between the first stage and the filter stage. To disambiguate the names of used impactor stages, henceforth the filter stage is called stage zero, the new stage is stage one, and so on until stage 12. The new ELPI impactor setup has been tested and calibrated (Yli-Ojanperä et al. 21). In this paper, the new ELPI impactor setup was used to study the physical state of α-pinene derived particles. Particle bounce characteristics were investigated using the currents measured from the ELPI stages. When a particle collides with an impactor plate, a part of its kinetic energy is dissipated in the deformation process, while another part is converted elastically into the kinetic energy of rebound. The particle will bounce from the impactor stage if the rebound energy of the particle exceeds the adhesion energy. Thus, the elastic properties of the particles affect their Table 1. Description and conditions of experiment runs A, B and C. For comparison, the respective data from the experiment of the Virtanen et al. (21) study are also given. The chamber was humidified prior to the experiment run. Experiment rh (%) VOC (ppb) O 3 (ppb) Description of experiment ELPI impactor plate type A :52 α-pinene injection smooth, greased 1:14 1:22 O 3 addition B :55 α-pinene injection smooth, greased 1:15 1:25 O 3 addition C :22 α-pinene injection porous, greased 1:4 1:49 O 3 addition (Virtanen et al. 21) :51 VOC of pine addition smooth, greased 14:22 14:33 O 3 addition

4 332 Kannosto et al. Boreal Env. Res. Vol. 18 bounce probability (Rogers et al. 1984). Generally, easily-deformed particles tend to adhere to the impactor plates upon collision. Solid particles bounce more easily from the impactor plate, resulting in a lower collection efficiency. In addition, the surface properties of an impactor substrate affect the particle bounce probability. For example, porous or roughened impactor plate surface is known to reduce the bounce probability (Chang et al. 1999, Marjamäki and Keskinen 24). In this study, two plate materials were used: smooth greased aluminum foil and sandblasted, greased stainless steel. The roughness of the latter, as characterized by the arithmetic mean deviation of the profile, was 6.9 ±.7 µm (Marjamäki and Keskinen 24). Excess current in the lowest stages indicates particle bounce. The amount of particle bounce can be analyzed by comparing the measured bounced currents to the simulated ideal (nonbounced) currents. The ideal currents can be simulated using measured SMPS number distribution and the mathematical model of ELPI (Marjamäki et al. 2, 25, Virtanen et al. 21), assuming no bounce in the impactor. This procedure is described in detail by Ristimäki et al. (22). To be able to simulate the currents, a value for the density of particles is needed. In this paper, we used the density value derived by comparing the data from SMPS volume and AMS mass size distributions (De Carlo et al. 24). The density of α-pinene-derived particles in chamber experiment runs A and B was found to be about 1.1 g cm 3. The bounce factor of impactor stage n, BF n, is here defined as:, (1) where I j and I j id are the measured and simulated normalized currents of stage j, respectively. Note that this definition is similar to that used by Virtanen et al. (21), generalized to all the stages. It gives the fraction of current which was transported from stage n to the lower stages due to the particle bounce. The density of the particles used in the current simulation affects the value of the bounce factor of < 4 nm particles, but it does not eliminate or explain the bouncing phenomena. At small particle sizes, a higher density value used in simulation (as compared with the real density value) produces a higher bounce factor, whereas a lower density gives a lower bounce factor (Virtanen et al. 211). This is because higher density values used in simulations decrease the second term (I id ) in the numerator in Eq. 1. The value of the bounce factor also changes depending on how many lowest stages (n) are added up. The α-pinene-derived SOA particles were not only measured with size distribution measurement instruments but also collected onto an electron microscopy grid (Agar, holey carbon/ carbon grid 3 Mesh Cu) by using an aspiration-type TEM sampler (Lyyränen et al. 29). The collection took place at the room temperature and pressure. The aerosol was lead through the grid at the flow rate of.2 l min 1, so particles were collected onto the carbon film mainly by diffusion. The samples were analyzed by a field emission gun scanning electron microscopy (FEG-SEM, Zeiss ULTRAplus). Before the SEM analysis, the sample was covered with thin carbon coating. The size distribution of collected particles was determined by a SEM image analysis (see Virtanen et al. 21). Results α-pinene experiments and comparison with Scots pine results The development of particle size (characterized by geometric mean diameter, GMD) and concentration for runs B (greased smooth impactor plates) and C (porous impactor plates) as a function of time was first considered (Fig. 1). Zero time marks the time when the first particles were detected with the CPC. During the closerstudied period (thin, grey section in Fig. 1), the mode GMD, total particle concentration and the age of particles age in experimental runs B and C were almost identical. The number size distributions measured with the SMPS for the two cases agreed very well (Fig. 1b). In run B, a clear excess current was registered at the impactor stages 1 and 2 when the smooth impactor plates were used (Fig. 1c, white bars), which is a clear

5 Boreal Env. Res. Vol. 18 Phase of SOA particles 333 GMD (nm) Normalized current : :3 1: 1:3 2: 2:3 3: 3:3 Particle age (h:mm) GMD, smooth impactor plate exp. B GMD, porous impactor plate exp. C concentration, smooth impactor plate exp. B concentration, porous impactor plate exp. C a c ELPI currents SEM-sample collection porous impactor plates (exp. C) smooth impactor plates (exp. B) 9.E+4 8.E+4 7.E+4 6.E+4 5.E+4 4.E+4 3.E+4 2.E+4 1.E+4 Concentration per cm 3 dn/dlogdp (number cm 3) 2.E+5 1.8E+5 1.6E+5 1.4E+5 1.2E+5 1.E+5 8.E+4 6.E+4 4.E+4 2.E+4 Normalized current b.1.1 d (nm) d smooth impactor plates (exp.b) SMPS exp. B SMPS exp. C simulated smooth impactor plates (exp. B) d p (µm) d p (µm) Fig. 1. (a) Particle history during experiments B (greased smooth impactor plates) and C (porous impactor plates). (b) SMPS size distributions during experimental runs B and C. (c) Measured ELPI currents with smooth impactor plates (run B) and porous impactor plates (run C). (d) Measured ELPI currents with smooth impactor plates compared with simulated bounce-free currents (both for run B). evidence of the bounce of the α-pinene-derived particles. The porous impactor plates prevented the bounce of the particles (Fig 1c, grey bars). Due to the broadened instrument response functions for the porous plate case, the measured currents differed from the simulated ideal currents. However, the difference between measured and simulated currents was much smaller than in the case of the smooth plates (Fig. 1d). For the α-pinene-derived particles in experimental runs A and B, the particle bounce factor BF 2 remained almost constant during the particle growth process and was approximately.32 in the size range of 4 15 nm (Fig. 2). The calculation of BF 2 was based on the sum of the impactor stages and 1 (see Eq. 1) and corresponds to the procedure used by Virtanen et al. (21). For comparison, their bounce factor values from a Scots pine experiment (Virtanen et al. 21) are also shown in Fig. 2. In short, the bounce behavior of the α-pinene-derived SOA particles was practically identical to that of the Scots-pine derived SOA particles. The α-pinene-derived particles are almost spherical (see insert in Fig. 2). The particle size was analyzed from the SEM image and it was compared with size distribution measured with the SMPS. The size distribution was measured simultaneously with the SEM sample collection. The size distributions analyzed from the SMPS measurement and from SEM images matched closely (3% difference in peak sizes), indicating

6 334 Kannosto et al. Boreal Env. Res. Vol. 18 Bounce factor GMD (nm) α-pinene: exp. A (density 1.1) α-pinene: exp. B (density 1.1) pine: Virtanen et al. 21 (density 1.1) Fig. 2. Bounce factor of α-pinene-derived particles during the growth process for two different experiment runs A and B. For comparison, also the bounce factor for Scots pine experiments using the same instrument setup (Virtanen et al. 21). The insert shows a SEM image of the α-pinene-derived particles. that the particles had not substantially evaporated or flattened, and that the SEM sample was a representative sample of the particles. The findings based on the SEM image analysis support those of the bounce analysis, indicating a solid phase of α-pinene SOA particles, and are also identical to those revealed in the Scots pine SOA experiments (Virtanen et al. 21). Overall, it can be concluded that the physical state of the α-pinene-derived particles equals that of the Scots-pine-derived SOA particles (Virtanen et al. 21), hence they can be assumed to be in a semi-solid or solid state. Aerosol mass spectrometer results of α-pinene and Scots pine experiments were almost identical, which indicates that the chemical composition of the particles was quite similar. Thus, α-pinene-derived SOA appear to be a good model system as far as the physical properties of the particles are concerned. Effect of particle size and age In order to study the bounce factor of the α-pinene SOA particles more closely, the bounce factors for different impactor stages (BF 1 to BF 4 ) were calculated as a function of the GMD of particle mode (Fig. 3a). All the curves first increased with an increasing GMD and then seemed to achieve a plateau. However, the plateau values for the different stages differed. The bounce factors were normalized by dividing the individual values of BF n by the highest value of the corresponding BF n curve (see Fig. 3b). The values of GMD were normalized by dividing the individual values of GMD by the cut-point diameter of the corresponding impactor stage (stage n). Bounce factors BF 2 to BF 4 all showed a decrease for normalized GMD values lower than approximately unity. This is believed to be an instrumental effect and can be readily explained by the impactor operation characteristics. As particle size decreases close to the stage cut-point diameter, the collision velocity of the particles on the impactor plate decreases. At the cut-point diameter, 5% of the particles do not even collide with the plate. Thus, by definition, the median collision velocity at the cut-point diameter is zero. According to various fundamental studies of particle bounce, the bounce probability decreases with a decreasing collision velocity (John 1995). This has been shown in various impactor bounce studies as a local maximum in the collection efficiency for particle sizes of approximately the cut-point diameter. It is therefore evident that the drop of the BF curves, for distributions with a GMD close to the stage cut-point diameter, does not represent any changes in the particle properties but is an inherent property of the measurement device. On the other hand, the normalized BF values presented as a function of normalized GMD values reach a plateau soon above unity. This reflects the fact that the median collision velocities reach values high enough for the bounce values to become practically constant. We estimate that there is a critical GMD value for each stage that is so high that the particle bounce is not affected by the vicinity of the cut-point diameter. Here, the normalized GMD value of 1.5 was taken as an estimate for this to happen. In other words, for the stages 2 to 4, the normalized BF values for GMD values above 1.5 times the cut-point diameter were expected to be free of the instrumental effect (this limit is shown by the dashed vertical line in Fig. 3b). For the impactor stage 1, the 1.5 times the corresponding cut-point value was approxi-

7 Boreal Env. Res. Vol. 18 Phase of SOA particles 335 mately 25 nm. BF 1 started to decrease at GMD values well above this value (Fig. 3). From the geometrical and flow dynamic points of view, the lowest impactor stage is practically similar to the other ones presented in Fig. 3. Furthermore, the jet velocity is not higher than the jet velocities of the previous stages. Therefore, the lowest impactor stage should not be more prone to bounce than the previous stages. The basic difference between the stages is the lowering pressure, and consequently the lowering cut-point diameter, as the stage numbering decreases. Our results indicate that the decrease in the bounce factor of the smallest particles bouncing first time in the first impactor stage is caused by changes in the particle bounce or charge transfer characteristics, not by the instrument functions. This result is in line with our previous study (Virtanen et al. 211). Next, the bounce factor was studied as a function of quantities related to the particle growth (see Fig. 4). We observed that particles of different size bounce differently, and this feature is not a device artifact (see Fig. 3). The time evolution of the particle number concentration was relatively similar between runs A and B, i.e. particle formation rates were similar, but the GMD of the mode increased faster during run B due to the higher VOC concentration (i.e. the higher particle growth rate) in that run (Fig. 4a). The initial increase in the value of BF 1 for GMD values from 25 nm to approximately 4 nm was quite identical for the two runs (Fig. 4b). The plateaus in the values of BF 1, reached at 4 nm, were also practically identical. Note that a given GMD value was reached much slower in run A, and therefore particles of the same GMD were older in run A than in run B. The particle age was estimated simply as the time elapsed from the time when the first particles were measured by the CPC, and the particle growth rate was estimated from the slope of five consecutive mode sizes and particle ages measured with the SMPS (two minutes between the SMPS scans). Neither the particle age (Fig. 4c) nor the particle growth rate (Fig. 4d) explained the change in the bounce factor or the mode diameter. The bounce factor of growing particles, at sizes below 4 nm, seemed to follow mostly the mode GMD. This could be related to the dependence of the chemical composition Bounce factor Normalized bounce factor a b GMD (nm) BF1 BF2 BF3 BF4 BF1.2 BF2 BF3 BF Normalized size d b (by cut-size) Fig. 3. (a) The bounce factors BF 1 to BF 4 (Eq. 1) as a function of mode GMD. (b) The normalized bounce factors as a function of normalized GMD. of the particles on particle size, or to changes in bounce or charge transfer processes in this size range. Unfortunately, no chemical information of particles in the experiment below 4 nm was available. Bounce probability for different impactor stages The value of bounce factor decreased with the decreasing stage numbers, i.e. further down the impactor (see Fig. 3a). Thus, BF 1 had the lowest values and BF 4 had the highest values. When moving towards the lower stages in the cascade impactor, the impactor jet velocity increases or, for the last ones, remains close to the sonic velocity (Yli-Ojanperä et al. 21). It is well known that when the particle size is larger than the cut-point diameter of the stage, the impact velocity of the particle is close to the jet velocity (e.g. Cheng and Yeh 1979). This has been

8 336 Kannosto et al. Boreal Env. Res. Vol. 18 GMD (nm) a α-pinene: exp. A (GMD) 2 2 α-pinene: exp. B (GMD) α-pinene: exp. A (concentration) 1 α-pinene: exp. B (concentration) : :3 1: 1:3 2: 2:3 3: 3:3 Particle age (h:mm) Concentration per cm 3 Bounce factor b α-pinene: exp. A (density 1.1) α-pinene: exp. B (density 1.1) GMD (nm).2 c.2 d Bounce factor.15.1 Bounce factor α-pinene: exp. A (density 1.1) α-pinene: exp. B (density 1.1) : :3 1: 1:3 2: 2:3 3: 3:3 Particle age (h:mm).5 α-pinene: exp. A (density 1.1) α-pinene: exp. B (density 1.1) GR (nm s 1 ) Fig. 4. Particle characteristics for two experimental runs (A and B). (a) GMD of the distribution and the total number concentration of the particles as a function of time. Bounce factor as a function of (b) GMD, (c) particle age, and (d) particle growth rate. verified for the present impactor type by CFD modeling (Virtanen et al. 211, Arffman et al. 211). Therefore, the actual bouncing probability should also increase or remain very high as the particles bounce down to lower stages of the impactor. Overall, if the particle bounces from one impactor plate, it is very probable that it would bounce from all the following plates as well, and finally the particle should end up in the filter stage. Therefore, the values for the bounce factors BF 1 BF 4 for > 1 nm particles should be approximately the same if the particles carry all their charge to the filter stage. Obviously, this is not the case, and in addition to the particle bounce characteristics also the particle charge transfer properties affect the measured BF values. We next present the particle bounce process in the impactor using a simple model taking into account the charge transfer during the impact. John (1995) describes the charge transferred from the particle to the impactor surface (q T ) during the bounce of one particle as being comprised of two independent factors: q T = q C + βq (2) where q C is the contact charge (i.e. the charge transferred from a neutral particle), q is the precharge of the particle (i.e. the initial charge of the particles entering the impactor stage), and β is the fraction of the pre-charge transferred. The first factor, q C, is assumed to be independent of the particle pre-charge, q. The contribution of this factor can be tested by turning the charger of the ELPI off. We investigated how the current of stage 2 and the sum of the currents to 2 behaves during such a test (Fig. 5). During the charger-off period, the measured currents were smaler than.1% of the charger-on values. The

9 Boreal Env. Res. Vol. 18 Phase of SOA particles 337 fact that the stage current was close to means that the average value of the contact charge was. Incidentally, because the sum of the currents was also close to, SOA particles were also in charge equilibrium. Probably majority of the particles were neutral. Because the current in stage 2 was, we could further assume that the contribution of the contact charge process was negligible, so we can keep only the latter term of Eq. 2. In case when the stage cut-point is clearly smaller than GMD, it can be assumed that the particles do not pass the stages without impacting the plate. In any stage n, on average a single particle will leave a charge of q n given by q n = (1 P n )q,n + P n β n q,n, (3) where q,n is the pre-charge of the particle (charge of the particle entering stage n), and P n is the true bounce probability within the stage n, i.e. the fraction of particles that bounce from the stage. The charge escaping the impactor stage with the bouncing particles is then q,n q n = P n (1 β n )q,n, (4) Multiplying both sides of Eq. 4 by N and Q where N is the particle number concentration entering the stage and Q is the volumetric flow rate we get the corresponding currents: (5) Here the sum of I j from stage 1 down to stage n equals the total current entering stage n. Finally, dividing by the total current entering the stage gives the measured current penetration through the stage. As this is the apparent fraction of particles bounced if calculated simply from the currents, it is here named the apparent bounce probability AP n : (6) Current (fa) charger off The apparent bounce probabilities were calculated for the lowest impactor stages (i.e. n values of 1, 2, and 3) for several distributions with GMDs ranging from 85 to 1 nm. The values for different stages were different (Table 2), but were rather constant for the GMD range. The pre-charge transfer term cannot be negative, so the minimum value of β n is zero. Therefore, the value of AP is also the minimum value for the true bounce probability of the particles. The minimum true bounce probability is 5%, reached at the lowest impactor stage (stage 1). The maximum bounce probability P is 1, giving the maximum value for the pre-charge transfer term β. The pre-charge transfer term should be close to for insulating particles and 1 for conducting particles. For sodium chloride, the measured pre-charge transfer was.42 (John 1995). Most likely, the charge transfer term in the present case is nonzero and, thus, the true.4765 Time (h:m:s) stage 2 sum of stages to 2 Fig. 5. Current in stage 2 and the sum of the stages 2 during the charger-off test in experiment B; GMD = 74 nm. Table 2. The maximum value of the fraction of particle pre-charge transferred during the bounce, β, and apparent bounce probability, AP, for the lowest impactor stages. The values of AP are also the maximum bounce probabilities for the same stages. Stage β AP

10 338 Kannosto et al. Boreal Env. Res. Vol. 18 bounce probability is larger than the minimum values presented in Table 2. The different values for different stages (see Table 2) mean that either the pre-charge transfer term (β) or the fraction of particles that bounce (P) change as the aerosol travels to the lower stages. The pre-charge transfer term (β) increases with increasing impact velocity (John 1995). The particles reach somewhat higher impact velocities at lower stages (Virtanen et al. 211). Therefore, it is likely that the decrease in the apparent bounce probability AP for the lower stages is at least partly caused by an increase in thee pre-charge term. It is also possible that the true bounce probability P decreases for the lower stages. This could be caused by secondary deposition of the particles within the impactor cascade. The secondary deposition could be higher for the largest particles, changing the size distribution that reaches the lower stages. Finally, one should note that the bounce characteristics above are for the greased impactor plates. Future studies of the bounce process should be performed (also) with the ungreased plates. The bounce characteristics of similar particles may then be different. Conclusions We studied the bounce behavior of particles derived from α-pinene using an ELPI impactor. The particles were observed to bounce from the smooth impactor plates towards the lower impactor stages, indicating a solid physical state. According to the results, the SOA particles formed by the ozonolysis of pure α-pinene had a similar bouncing ability as the Scots-pine-derived particles in previous studies. This indicates that the phase behavior of the α-pinene-derived SOA particles is comparable to the phase behavior of the pine-derived SOA particles in the plant chamber conditions, which was also supported by the SEM image analysis. Thus, α-pinene appears to be a good model component for the pine-emitted VOCs when the physical properties of the particles are considered. It is even possible that α-pinene-derived oxidation products are the ones responsible for the solidification behavior of the SOA particles produced by boreal forests. The measured bounce factor of large particles (diameter > 4 nm) did not significantly change during the particle growth process and it was not seen to depend on particle size, growth rate or age, indicating no changes in the particle solidity in this size range. On the other hand, for the small particles the measured bounce factor showed clear changes: the amount of bounce increased as particle size increased, indicating that particles get more solid during their early growth/ageing process. Our results are well in line with the earlier ones reporting the size-dependent bounce behavior of pine VOCderived SOA particles in the size range < 3 nm. This indicates that during the early growth process, the particles can be treated as liquid droplets in models describing the growth process in the atmosphere. As the particle age and size increases, particles solidify and the particle interaction with surrounding vapors might differ from thermodynamic equilibrium behavior. How large is the difference (i.e. how important role the kinetic limitations play in the growth process) depends on the diffusion coefficient of the diffusing molecule in the particle bulk. This actually can be estimated, if the particle viscosity is known. Thus, it is obvious that the future work should be aiming to quantify the viscosity/diffusion coefficient values of diffusing molecule in atmospherically relevant SOA particles. The maximum value of the bounce factor (BF n ) decreased for the subsequent impactor stages. We presented a simplified model to describe charge transfer behavior in the bounce process. According to the model, the measured values can be explained by a combination of the bounce probability and charge transfer between the particles and the impactor surface. The results can be explained if at least 5% of the particles bounce from each stage. The value is rather independent of particle diameter once the diameter is well over the stage cut-point diameter. Acknowledgements: Jonna Kannosto received financial support from the Doctoral Programme ACCC Atmospheric Composition and Climate Change: From Molecular Processes to Global Observations and Models (funding decision number ). Jorma Joutsensaari received financial support from the Academy of Finland (grants 11763, 13119, , 25298), the Centre of Excellence Programme and the University of Eastern Finland (spearhead project CABI).

11 Boreal Env. Res. Vol. 18 Phase of SOA particles 339 Li-Qing Hao received financial support from the Emil Aaltonen Foundation. References Atkinson R. 2. Atmospheric chemistry of VOC and NO x. 2. Atmos. Environ. 34: Arffman A., Marjamäki M. & Keskinen J Simulation of low pressure impactor collection efficiency curves. J. Aerosol Sci. 42: Cappa C.D. & Wilson K.R Evolution of organic aerosol mass spectra upon heating: implications for OA phase and partitioning behavior. Atmos. Chem. Phys. 11: Chang M., Kim S. & Sioutas C Experimental studies on particle impacton and bounce effects of substrate design and materials. Atmos. Environ. 33: Cheng Y.S. & Yeh H.C Particle bounce in cascade impactors. Environ. Sci. Technol. 13: De Carlo P.F., Slowik J.G., Worsnop D.R., Davidovits P. & Jimenez J.L. 24. Particle morphology and density characterization by combined mobility and aerodynamic diameter measurements. Part 1: Theory. Aerosol. Sci. Technol. 38: Hallquist M., Wenger J.C., Baltensperger U., Rudich Y., Simpson D., Claeys M., Dommen J., Donahue N.M., George C., Goldstein A.H., Hamilton J.F., Herrmann H., Hoffmann T., Iinuma Y., Jang M., Jenkin M.E., Jimenez J.L., Kiendler-Scharr A., Maenhaut W., McFiggans G., Mentel Th.F., Monod A., Prévôt A.S.H., Seinfeld J.H., Surratt J.D., Szmigielski R. & Wildt J. 29. The formation, properties and impact of secondary organic aerosol: current and emerging issues. Atmos. Chem. Phys. 9: Hao L.Q., Yli-Pirilä P., Tiitta P., Romakkaniemi S., Vaattovaara P., Kajos M., Rinne J., Heijari J., Kortelainen A., Miettinen P., Kroll J.H., Holopainen J.K., Smith J.N., Joutsensaari J., Kulmala M., Worsnop D.R. & Laaksonen A. 29. New particle formation from the oxidation of direct emissions of pine seedlings. Atmos. Chem. Phys. 9: John W Particle surface interactions: charge transfer, energy loss, resuspension, and deagglomeration. Aerosol Sci. Technol. 23: Kanakidou M., Seinfeld J.H., Pandis S.N., Barnes I., Dentener F.J., Facchini M.C., Van Dingenen R., Ervens B., Nenes A., Nielsen C.J., Swietlicki E., Putaud J.P., Balkanski Y., Fuzzi S., Horth J., Moortgat G.K., Winterhalter R., Myhre C.E.L., Tsigaridis K., Vignati E., Stephanou E.G. & Wilson J. 25. Organic aerosol and global climate modelling: a review. Atmos. Chem. Phys. 5: Kulmala M., Hämeri K., Aalto P.P., Mäkelä J.M., Pirjola L., Nilsson E.D., Buzorius G., Rannik U., Dal Maso M., Seidl W., Hoffman T., Janson R., Hansson H.C., Viisanen Y., Laaksonen A. & O Dowd C.D. 21. Overview of the international project on biogenic aerosol formation in the boreal forest (BIOFOR). Tellus 53B: Kulmala M., Vehkamäki H., Petäjä T., Dal Maso M., Lauri A., Kerminen V.-M., Birmili W. & McMurry P.H. 24. Formation and growth rates of ultrafine atmospheric particles: a review of observations. J. Aerosol. Sci. 35: Keskinen J., Pietarinen K. & Lehtimäki M Electrical Low Pressure Impactor. J. Aerosol Sci. 23: Lyyränen J., Backman U., Tapper U., Auvinen A. & Jokiniemi J. 29. A size selective nanoparticles collection device based on diffusion and thermophoresis. J. Phys.: Conf. Ser. 17, 1211, doi:1.188/ /17/1/1211. Marcolli C., Luo B.P. & Peter T. 24. Mixing of the organic aerosol fractions: liquids as the thermodynamically stable phases. J. Phys. Chem. A 18: Marjamäki M. & Keskinen J. 24. Effect of impaction plate roughness and porosity on collection efficiency. J. Aerosol Sci. 35: Marjamäki M., Lemmetty M. & Keskinen J. 25. ELPI response and data reduction I: response functions. Aerosol Sci. Technol. 39: Marjamäki M., Keskinen J., Chen D.R. & Pui D.Y.H. 2. Performance evaluation of the electrical low pressure impactor (ELPI). J. Aerosol Sci. 31: Mikhailov E., Vlasenko S., Martin S.T., Koop T. & Pöschl U. 29. Amorphous and crystalline aerosol particles interacting with water vapor: conceptual framework and experimental evidence for restructuring, phase transitions and kinetic limitations. Atmos. Chem. Phys. 9: Murray B.J. 28. Inhibition of ice crystallisation in highly viscous aqueous organic acid droplets. Atmos. Chem. Phys. 8: O Dowd C.D., Aalto P., Hämeri K., Kulmala M. & Hoffmann T. 22. Aerosol formation: atmospheric particles from organic vapours. Nature 416: Odum J.R., Hoffmann T., Bowman F., Collins D., Flagan R.C. & Seinfeld J.H Gas/particle partitioning and secondary organic aerosol yields. Environ. Sci. Technol. 3: Pankow J.F An absorption model of the gas/aerosol partitioning involved in the formation of secondary organic aerosol. Atmos. Environ. 28: Pfrang C., Shiraiwa M. & Pöschl U Chemical ageing and transformation of diffusivity in semi-solid multicomponent organic aerosol particles. Atmos. Chem. Phys. 11: Rogers L.N. & Reed J The adhesion of particles undergoing elatic-plastic impact with a surface. J. Phys. D: Appl. Phys. 17: Ristimäki J., Virtanen A., Marjamäki M., Rostedt A. & Keskinen J. 22. On-line measurement of size distribution and effective density of submicron aerosol particles. J. Aerosol Sci. 33: Shiraiwa M., Ammann M., Koop T. & Pöschl U Gas uptake and chemical aging of semisolid organic aerosol particles. PNAS 18: Spracklen D., Carslaw K., Kulmala M., Kerminen V.-M., Mann G. & Sihto S.-L. 26. The contribution of bound-

12 34 Kannosto et al. Boreal Env. Res. Vol. 18 ary layer nucleation events to total particle concentrations on regional and global scales. Atmos. Chem. Phys. 6: Vaden T. D., Imre D., Beránek J., Shrivastava M. & Zelenyuk A Evaporation kinetics and phase of laboratory and ambient secondary organic aerosol. PNAS 18: Virtanen A., Marjamäki M., Ristimäki J. & Keskinen J. 21. Fine particle losses in electrical low-pressure impactor. J. Aerosol Sci. 32: Virtanen A., Joutsensaari J., Koop T., Kannosto J., Yli-Pirilä P., Leskinen J., Mäkelä J.M., Holopainen J., Pöschl U., Kulmala M., Worsnop D. & Laaksonen A. 21. An amorphous solid state of biogenic secondary organic aerosol particles. Nature 467: Virtanen A., Kannosto J., Joutsensaari J., Saukko E., Kuuluvainen H., Hao L., Yli-Pirilä P., Tiitta P., A. Kortelainen A., Holopainen J.K., Worsnop D.R., Smith J.N. & Laaksonen A Bounce behavior of freshly nucleated biogenic secondary organic aerosol particles. Atmos. Chem. Phys. 11: Vuorinen T., Nerg A.M., Ibrahim M.A., Reddy G.V.P. & Holopainen J.K. 24. Emission of Plutella xylostellainduced compounds from cabbages grown at elevated CO 2 and orientation behavior of the natural enemies. Plant Physiol. 135: Yli-Ojanperä J., Kannosto J., Marjamäki M. & Keskinen J. 21. Improving the nanoparticle resolution of the ELPI. Aerosol and Air Quality Research 1: Zahardis J. & Petrucci G.A. 27. The oleic acid-ozone heterogeneous reaction system: products, kinetics, secondary chemistry and atmospheric implications of a model system: a review. Atmos. Chem. Phys. 7: Zobrist B., Marcolli C., Pedernera D.A. & Koop T. 28. Do atmospheric aerosols form glasses? Atmos. Chem. Phys. 8: