Strong atmospheric new particle formation in winter, urban Shanghai, China

Particle size distributions in the range of 1.34–615.3 nm were recorded from 25 November 2013 to 25 January 2014 in urban Shanghai, using a combination of one nano Condensation Nucleus Counter system (nCNC), one nano-Scanning Mobility Particle Sizer (SMPS), and one long-SMPS. Measurements of sulfur dioxide by an SO 2 analyzer with 5 pulsed UV ﬂuorescence technique allowed calculation of sulfuric acid proxy. In addition, concentrations of ammonia were recorded with a Di ﬀ erential Optical Absorption Spec-troscopy (DOAS). During this 62-day campaign, 13 NPF events were identiﬁed with strong burst of sub-3 nm particles and subsequent fast growth of newly formed particles. The observed nucleation rate ( J 1.34 ), formation rate of 3 nm particles ( J 3 ), and 10 condensation sink (CS) were 112.4–271.0 cm − 3 s − 1 , 2.3–19.2 cm − 3 s − 1 , and 0.030– 0.10 s − 1 , respectively. Subsequent cluster/nanoparticle growth showed a clear size dependence, with average values of GR 1.35 ∼ 1.39 (from the bin of 1.34–1.37 nm to the bin of 1.37–1.41 nm), GR 1.39 ∼ 1.46 (from 1.37–1.41 to 1.41–1.52 nm), GR 1.46 ∼ 1.70 (from 1.41– 1.52 to 1.52–1.89 nm), GR 1.70 ∼ 2.39 (from 1.52–1.89 to 1.89–3.0 nm), GR 2.39 ∼ 7 (from 15 1.89–3.0 to 7 nm), and GR 7 ∼ 20 (from 7 to 20 nm) being 1.6 ± 1.0, 1.4 ± 2.2, 7.2 ± 7.1, 9.0 ± 11.4, 10.9 ± 9.8, and 11.4 ± 9.7 nm h − 1 , respectively. Correlation between nucleation rate ( J 1.34 ) and sulfuric acid proxy indicates that nucleation rate J 1.34 was proportional to a 0.64 power of sulfuric acid proxy. Correlation between nucleation rate ( J 1.34 ) and gas-phase ammonia suggests that ammonia was associated with NPF events. The 20 calculated sulfuric acid proxy was su ﬃ cient to explain the subsequent growth of 1.34– 3 nm particles, but insu ﬃ cient for particles exceeding this size range. Qualitatively, NPF events in urban Shanghai likely occur on days with low levels of PM 2.5 .

River Delta, including Hong Kong (e.g., Guo et al., 2012;Yue et al., 2013). As one of the most industrialized area of China, one of the most populated area in the world, and one of the hotspots for particle pollution, Yangtze River Delta (YRD) has only seen a few research activities on atmospheric nucleation (Du et al., 2012;Herrmann et al., 2014). Among the few studies, measurements at the station for Observing Regional Processes 5 of the Earth System, Nanjing University (SORPES-NJU) offered a first insight for new particle formation in the western part of YRD (Herrmann et al., 2014). On the other hand, atmospheric nucleation research in China is still in its infant stage and the latest experimental techniques are yet to be applied in China. For example, data on freshly nucleated particles are really sparse, except for those from an air ion spectrometer 10 employed at SORPES-NJU (Herrmann et al., 2014). To the best of our knowledge, the employment of a Particle Size Magnifier (PSM), which is able to measure atmospheric nucleation at the size (mobility diameter) down to 1.5 ± 0.4 nm (Kulmala et al., 2012), has not been reported in a Chinese location in literature. The lack of key information greatly hinders a better understanding of nucleation mechanisms in China. 15 In this study, we measured atmospheric nucleation from 25 November 2013 to 25 January 2014 in urban Shanghai with nCNC and two sets of SMPS. Nucleation frequency, nucleation rate (J 1.34 ), condensation sink (CS), and growth rates (GR) have been reported and compared with previous studies with similar or dissimilar atmospheric environments. In addition, the potential nucleation mechanism was explored by 20 correlating sulfuric acid proxy calculated from sulfur dioxide precursor and gas-phase ammonia to nucleation rate (J 1.34 ). The competition between background particle levels and available condensable sulfuric acid that determines whether an atmospheric nucleation occurs or not has also been discussed. Introduction

Nucleation measurements
Nucleation measurements were carried out on the rooftop of a teaching building (31 • 18 N, 121 • 30 E) that is about 20 m above ground on the campus of Fudan University between 25 November 2013 and 25 January 2014. This monitoring site is mostly 5 surrounded by commercial properties and residential dwellings. The Middle Ring Road, one of main overhead highways in Shanghai, lies about 100 m to the south of the site. Hence, the Fudan site can be treated as a representative urban site influenced by a wide mixture of emission sources Ma et al., 2014). Ambient particle size distributions in the range of 1.34-615.3 nm were measured 10 using a combination of one nano Condensation Nucleus Counter system (model A11, Airmodus, Finland), one nano-SMPS (consisting of one DMA3085 and one CPC3776, TSI, USA), and one long-SMPS (consisting of one DMA3081 and one CPC3775, TSI, USA). The instruments were continuously running except for maintenance and minor instrument breakdown during the campaign. 15 Ambient air was drawn into a stainless steel manifold of 5.0 m length and 4 inch inner diameter at a flow rate of 153 CFM using a low-volume blower (Model DJT10U-25M, NUSSUN, China). From this main manifold, 1.75 lpm ambient air was drawn through a 1/4 inch inner diameter stainless tube of 18 cm length, and diluted with a zero air flow generated by a zero air generator (Model 111, Thermo, USA) at a ratio of 1 : 1 to reduce Introduction The nCNC system consists of one PSM (model A10, Airmodus, Finland) and one butanol Condensation Particle Counter (bCPC, model A20, Airmodus, Finland), and was used to measure size distributions of 1.34-3 nm clusters/particles. Briefly, PSM activates the smallest particles using diethylene glycol as a working fluid and condensationally grow nanoparticles up to larger than 90 nm in mobility equivalent diameter, 5 after which an external CPC is used for further growing the particles to optical sizes and counting the grown particles (Vanhanen et al., 2011). In this study, PSM was used in the scanning mode in which the saturator flow rate is changed continuously, giving an activation spectrum of the measured particles to obtain size distribution of sub-3 nm clusters/particles. A scanning cycle of 100 steps between saturator flow rates 0.1-1 lpm 10 and a time resolution of 220 s were chosen. Sub-3 nm clusters/particles were classified into 5 bins, i.e., 1.34-1.37, 1.37-1.41, 1.41-1.52, 1.52-1.89, 1.89-3.0 nm, respectively. Geometric mean values of upper and lower limits of the five bins, i.e., 1.35, 1.39, 1.46, 1.70, and 2.39 nm, respectively, were used to refer to the five bins in the growth rate calculation. 15 The nano-SMPS measured particles in the size range from 3 to 64 nm and the long-SMPS recorded particles from 14 to 615.3 nm. For both SMPSs, 64 size bins and 5 min time resolution were chosen. The sample flow to sheath flow ratios for both SMPSs were set at 1 : 10. A comparison analysis on the total particle concentrations between 14 and 64 nm measured by both nano-SMPS and long-SMPS showed less than 10 % 20 difference in the size range of 55-64 nm between two SMPSs. Hence, number concentrations of particles in the size range of 3-615 nm, N 3∼615 , were calculated from the sum of N 3∼55 obtained from nano-SMPS, N 55∼64 from the arithmetic average of nano-SPMS and long-SMPS, and N 64∼615 from long-SMPS.
At the same site, sulfur dioxide (SO 2 ) was measured by an SO 2 analyzer with pulsed 25 UV fluorescence technique (Model 43i, Thermo, USA) and calibration of this SO 2 analyzer was performed every month. A DOAS system was used to measure the integrated concentration of NH 3 along the optical path between a transmitter telescope using a 35W Deuterium lamp as the light source and a receiver telescope (53 m), and then to yield the average concentration of NH 3 through dividing the integrated concentration by the absorption length (Platt and Stutz, 2008). In this study, the concentration of NH 3 was determined by fitting the reference spectra to the atmospheric spectra in a given window (205-220 nm) using a nonlinear least-squares method, similarly to a previous measurement of HONO and NO 2 . where CoagS d p =2 nm represents coagulation sink of 2 nm particles, an approximation for that in the size interval of 1.34-3 nm; and GR 1.34∼3 represents the apparent clusters/particle growth rate between 1.34 and 3 nm. Formation rate of 3 nm particles was calculated in a similar way (Sihto et al., 2006;Kulmala et al., 2012), providing a comparison with previous studies, where CoagS d p =4 nm represents coagulation sink of 4 nm particles, an approximation 5 for that in the size interval of 3-6 nm. Growth rate (GR) is defined as the rate of change in the diameter of a growing particle population, using the maximum-concentration method (Kulmala et al., 2012), where d p 1 and d p 2 are the representative particle diameters at times t 1 and t 2 , respectively. Condensation sink (CS) describes the condensing vapor sink caused by the particle population (Kulmala et al., 2012), where D is the diffusion coefficient of the condensing vapor, usually assumed to be sulfuric acid (0.104 cm 2 s −1 used in this study); and β m,d p is the transitional regime correction factor. 20 Sulfuric acid has been accepted as a key gas-phase precursor for atmospheric nucleation and contributes to the subsequent growth of newly-formed particles ( , 1996;Sipila et al., 2010). The accurate measurement of gas-phase sulfuric acid requires application of chemical ionization mass spectrometry using nitrates as reagent ions (Eisele and Tanner, 1993), which is not possessed by this research group during this study. Instead, the sulfuric acid proxy [H 2 SO 4 ] was estimated based on local solar radiation level Radiation, SO 2 concentration [SO 2 ], condensation sink CS, and rela-5 tively humidity (Mikkonen et al., 2011),

Sulfuric acid
where k is the temperature dependent reaction rate constant (Mikkonen et al., 2011). Condensation of sulfuric acid contributes to the growth of newly-formed particles. 10 The growth of clusters/particles due to condensation of sulfuric acid, GR H 2 SO 4 , can be estimated by the following equations (Nieminen et al., 2010), where ρ v and d v are the condensed phase density and molecule diameter of H 2 SO 4 , respectively; m p and m v are particle and H 2 SO 4 vapor molecule masses, respectively; Kn is the Knudsen number (Lehtinen and Kulmala, 2003). For this calculation, ρ v = 20 ρ p = 1.83 g cm −3 , d v = 0.56 nm, and m v = 98 amu were used.

Classification of new particle formation (NPF) events
Figure 1 presents a contour plot for particle size distributions of 3-615.3 nm and a number concentration plot of sub-3 nm clusters/particles N 25 January 2014. Data were occasionally missing because of maintenance and minor breakdown of instruments. From the figure, frequent bursts of sub-3 nm particles were evident, with concentrations up to 8.0 × 10 4 cm −3 around noon time. However, similarly to previous studies (Kulmala et al., 2007(Kulmala et al., , 2013Yu et al., 2014), not all sub-3 nm particles eventually underwent a continuous growth to larger sizes. In this study, we 5 define an observation day with appearances of sub-3 nm clusters/particles over a time span of hours and their subsequent growth to larger sizes for a few hours that presents a banana-shaped contour plot of particle size distributions obtained from SMPS (Dal Maso et al., 2005) as a NPF event day, and focus on characteristics and potential mechanisms of these events.
According to the classification, there were 13 event days during the 62-day campaign, as illustrated by the shadow in Fig. 1. Although nCNC data were partially unavailable on 26 December 2013 and completely unavailable on 24 January 2014, these two days are still defined as NPF days since a distinctive banana-shaped contour plot for particle distributions between 3-615.3 nm existed. 18 December 2013 was not re- 15 garded as a NPF day since N 1.34∼3 was not recorded and the growth of 3 ∼ 20 nm particles was relatively short-lived.
Among these NPF events, 5 NPF events occurred in November, 3 in December, and 5 in January. The averaged frequency for NPF events was 21.0 % during the 62day campaign. Our NPF frequency at Shanghai is larger than the average value in 20 winter 1996, SMEAR II station, Hyytiälä, Finland (Dal Maso et al., 2005, likely because nucleation events at Hyytiälä have recently been related to oxidation products of biogenic emissions (Kulmala et al., 1998;Schobesberger et al., 2013;Riccobono et al., 2014) and photochemistry of volatile organic compounds is less intensive in winter months. This frequency is also higher than that in winter, semi-rural Kent, OH, to that measured in winter 2012, Nanjing, which is also located in Yangtze River delta (Herrmann et al., 2014). Number concentrations of particles in different size ranges, i.e., N 1.34∼3 , N 3∼7 , and N 7∼30 , respectively, on a NPF day (11 December 2013) and an obvious non-NPF day (7 January 2014) are further examined to illustrate features of a NPF event, as shown in 5 Fig. 2. On the NPF day, 1.34-3 nm particles appeared as early as 7 a.m. in the morning that was right after sunrise (6:42 a.m. on 11 December 2013), reached its maximum just before noontime, and spanned for almost the whole daytime (sunset at 4:52 p.m., 11 December 2013). This size distribution of atmospheric neutral and charged clusters/particles by a scanning PSM is identical to that measured at Hyytiälä, Finland 10 (Kulmala et al., 2013), suggesting that photochemistry products likely contribute to formation of smallest particles. On the same NPF day, 3-7 nm and 7-30 nm particles appeared much later, resembling previous findings only with SMPS measurements (e.g., Yue et al., 2010). The lag in peaking times of N 1.34∼3 , N 3∼7 , and N 7∼30 on the NPF day clearly indicated the continuous growth of clusters/particles, the loss of particles due to 15 coagulation during the growth, and the diverse sources of particles in the size range of 7-30 nm. In contrast, N 1.34∼3 and N 3∼7 showed a flat profile on the non-NPF day. The minor enhancement in N 7∼30 between 10 a.m. and 5 p.m. on the non-NPF day was not due to growth of newly formed particles. Instead, direct emission of 7-30 nm particles from transportation activity likely explained their appearance.  their results, we conclude that strong nucleation events occur geographically widely in the YRD region. Formation rate of 3 nm particles J 3 has been more routinely reported. Similarly to J 1.34 , J 3 at Shanghai is significantly larger than 0.61 cm −3 s −1 at Hyytiälä, Finland . The fast reduction from J 1.34 to J 3 was likely due to the presence of a large background particle number as shown in Table 1. The large background particle number concentrations corresponded to the high con-   , and our GR 7∼20 is close to the upper bound of those at urban Beijing (Wu et al., 2007;Yue et al., 2009;Zhang et al., 2011;Gao et al., 2012), and generally larger than 1.5-8.4 nm h −1 in Hong Kong (Guo et al., 2012), indicating that the condensable vapor was intense. In addition, our growth rates suggest that the smallest clusters (the bin of 1.34-1.37 nm), if not scavenged by larger particles, will 20 grow to 3 nm within ∼ 12 min, and to 20 nm within ∼ 2 h.

Potential mechanisms for NPF events
As shown in Table 1, nucleation rate (J 1.34 ) in this study is typically larger than 100 cm −3 s −1 , suggesting that the ion-induced nucleation was not a main mechanism for observed fast nucleation (Hirsikko et al., 2011;Riccobono et al., 2014). The 2012 25 winter study at the SORPES-NJU station that is also located at YRD shows that the ratio of J 2 between ions and total particles (ions plus neutral particles) was 0.002, also indicating the minor role of ion-induced nucleation (Herrmann et al., 2014). Hence, it is likely that nucleation of neutral precursor molecules actually largely determined the observed NPF events. Correlations between log J 1.34 and log [H 2 SO 4 ] (Fig. 4) and between log J 1.34 and log[NH 3 ] (Fig. 5) have been examined to elucidate potential mechanisms for our NPF 5 events. Since J 1.34 could not be accurately determined on some of the NPF days, the data points in both figures are less than the actual number of NPF events that have been observed. Daily daytime (6 a.m.-6 p.m.) averages of sulfuric acid proxy and ammonia were used as approximations for their effective concentrations on a NPF day. Kupiainen -Määttä et al. (2014) recently reported that the number of molecules in a critical cluster cannot be determined by a slope analysis in atmospherically relevant applications, underscoring the need to further explore the exact nucleation mechanism. Herrmann et al. (2014) also calculated the sulfuric acid proxy, related it to observed nucleation rates, and speculated that the sulfuric acid exponent might be well over 2, 25 which underscores the reliability of calculation methods in a Chinese location. Hence, our preliminary result should be further tested with actual measurements of gas-phase sulfuric acid, but does indicate the key role of sulfuric acid in NPF events. On the other hand, linear correlation between log J and log [NH 3 ] was observed in a previous nu-

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | cleation study in Atlanta, GA, with a slope of 1.17 (McMurry et al., 2005), but a clear relationship was not perceived at Kent, OH (Erupe et al., 2010). This discrepancy may come from the level of ammonia that has been predicted to have a saturation threshold (Napari et al., 2002) and/or the co-occurring sulfuric acid concentration (Benson et al., 2009). Nevertheless, our correlation between log J and log[NH 3 ] suggests that 5 ammonia also participated in the nucleation. The subsequent growth of newly formed particles can be partially attributed to condensation of sulfuric acid. The theoretical maximum growth rate of 1.34-3 nm clusters/particles due to condensation of sulfuric acid (GR H 2 SO 4 (1.34∼3) ), calculated according to Eqs. (6) and (7), was 2.5 ± 0.82 nm h −1 on average. This rate is larger than the observed growth rates of clusters/particles from the bin of 1.34-1.37 nm to the bin of 1.89-3.0 nm (GR 1.35∼2.39 ), being 2.0±2.7 nm h −1 , indicating that sulfuric acid proxy was enough to explain the observed growth for particles under 3 nm. Similar calculation of GR H 2 SO 4 (3∼7) and GR H 2 SO 4 (7∼20) yielded 1.9 ± 0.62 and 1.6 ± 0.54 nm h −1 , respectively.
In Fig. 6, relative contributions of sulfuric acid to growth of particles in the range of 15 3-7 and 7-20 nm, respectively, on each NPF day is presented. Since 7 nm particles reached their maximum earlier than 3 nm particles on 9 and 15 January 2014, there was no calculated GR 3∼7 and hence no ratios available on these two days. In addition, condensation of sulfuric acid was more prominent for 3-7 nm particles on 6 NPF days (25, 26, 30 November, 10, 11, and 12 December 2013), whereas it was more signifi-20 cant for 7-20 nm particles on the other 5 NPF days (28, 29 November 2013, 13, 21, and 24 January 2014). On average, condensation of gas-phase sulfuric acid explained 26.5 % of GR 2.39∼7 , and 20.9 % of GR 7∼20 , respectively. The rest of growth might be largely attributed to condensation of extremely low volatility organic compounds (Ehn et al., 2014), and potentially heterogeneous reactions of organics on nanoparticle sur-Introduction 214.8 µg m −3 , although similar sulfuric acid proxies existed between the two days.
Hence, we conclude that, qualitatively, NPF processes in urban Shanghai occurred with low levels of PM 2.5 and that high sulfuric acid favored NPF events when PM 2.5 was low. This conclusion is identical to that drawn from a Mexico City study where NPF events observed in the city correlated with elevated concentrations of sulfur dioxide and 5 low particulate matter mass concentrations in the afternoon hours (Dunn et al., 2004). Concentrations of PM 2.5 showed correlations with particle growth rates in our study. Figure 8 denotes particle growth rates of particles in various size ranges and the realtime hourly-averaged mass concentrations of PM 2.5 when particles grew to the corresponding size. GR 1.35∼2.39 and GR 2.39∼7 were anti-correlated with PM 2.5 as expected, since preexisting particles coagulate with smallest new clusters/particles and scavenge them. GR 7∼20 was positively correlated with PM 2.5 . Although this positive correlation is not exactly understood, we speculate that high PM 2.5 concentrations in Shanghai were at least partially associated with secondary processes in the atmosphere leading to low vapor pressure organic and inorganic species, which could increase condensation 15 rate or react with larger particles therefore increasing PM.

Summary and conclusions
Atmospheric new particle formation is a significant source of atmospheric aerosol particles. Understanding NPF under the current levels of air pollution in China is of special scientific interests because the exact nucleation mechanism remains elusive.

20
From 25 November 2013 to 25 January 2014, a combination of one nano-CNC, one nano-SPMS, and one long-SPMS has been utilized to investigate atmospheric nucleation by measuring particle size distributions in the range of 1.34-615.3 nm at urban Shanghai, located in the east Yangtze River Delta. During this 62-day campaign, 13 NPF events were identified with strong burst of sub-3 nm particles and subsequent 25 fast growth of newly formed particles. Together with nucleation frequency (21 %), the obtained nucleation rate J days clearly indicate that strong atmospheric new particle formation occurred in winter, urban Shanghai with competition between promotion from condensable vapors and scavenging by preexisting particles. The absolute values of J 1.34 and CS are one to two orders of magnitude lager than those at locations with a pristine atmosphere (e.g.,

5
(Kulmala et al., 2012) and semi-rural locations with very low pollution levels such as Kent, OH (Yu et al., 2014), as a reflection from the significantly-altered atmospheric background.
Our preliminary exploration on nucleation mechanism indicates that nucleation rate J 1.34 was proportional to a 0.64 power of sulfuric acid proxy. It is hence likely that 10 observed NPF events could be explained by the activation theory. As Herrmann et al. (2014) doubted reliability of sulfuric acid proxy, accurate measurements of gasphase sulfuric acid instead of calculation of a proxy is necessary to achieve an unambiguous conclusion. The positive correlation between J 1.34 and gas-phase ammonia hints the involvement of ammonia in new particle formation, but its exact role cannot 15 be determined without measurements of nucleating clusters, either. A size-depend particle growth in the range of 1.34-20 nm has been observed in this study, consistent with predictions from nano-Köhler theory (Kulmala et al., 2004), and Kelvin-limited diffusion, surface, and volume growth laws (Kuang et al., 2012). Sulfuric acid proxy was enough to explain the observed growth for particles under 3 nm, 20 and contributed to 26.5 % of GR 2.39∼7 , and 20.9 % of GR 7∼20 , respectively. The rest of growth can be largely attributed to condensation of extremely low volatility organic compounds (Ehn et al., 2014), and potentially heterogeneous reactions of organics on nanoparticle surface (Wang et al., 2010.