Differing Mechanisms of New Particle Formation at Two Arctic Sites

New particle formation in the Arctic atmosphere is an important source of aerosol particles. Understanding the processes of Arctic secondary aerosol formation is crucial due to their significant impact on cloud properties and therefore Arctic amplification. We observed the molecular formation of new particles from low‐volatility vapors at two Arctic sites with differing surroundings. In Svalbard, sulfuric acid (SA) and methane sulfonic acid (MSA) contribute to the formation of secondary aerosol and to some extent to cloud condensation nuclei (CCN). This occurs via ion‐induced nucleation of SA and NH3 and subsequent growth by mainly SA and MSA condensation during springtime and highly oxygenated organic molecules during summertime. By contrast, in an ice‐covered region around Villum, we observed new particle formation driven by iodic acid but its concentration was insufficient to grow nucleated particles to CCN sizes. Our results provide new insight about sources and precursors of Arctic secondary aerosol particles.


Introduction
The Polar environment is currently undergoing rapid change as manifested in decrease in snow cover (Callaghan et al., 2011), thawing of permafrost (AMAP, 2017), greening, and especially sea ice loss (Meier et al., 2014). The Arctic Ocean emits a variety of precursor gases for new particle formation (NPF) and particle growth to the sizes of cloud condensation nuclei (CCN). These gases include dimethylsulfide (DMS; Levasseur, 2013), organic compounds (Mungall et al., 2017), and iodine species (Cuevas et al., 2018;Raso et al., 2017). Coastal bird colonies form a primary source of ammonia (NH 3 ) in the Arctic (Croft et al., 2016) and Arctic tundra is a source of volatile organic compounds (VOC; Lindwall et al., 2016). The NPF frequency in the Arctic atmosphere was found to be correlated with the diminishing sea ice extent (Dall'Osto et al., 2017, probably via increased phytoplankton productivity, but so far the details of NPF and CCN production over ice-covered and open Arctic waters have remained largely unknown. The absence of detailed chemical understanding down to molecular scale observations hinders estimations of the emissions of the precursor gases, or their contribution to the initial steps of NPF and subsequent particle growth (S. H. Lee et al., 2019). More importantly, those will change in response to the changing climate. Further decrease of sea ice extent may lead to enhanced open ocean phytoplankton productivity and DMS emissions , while increased terrestrial primary production due to warming has already caused enhanced biogenic volatile organic compound emissions (Lindwall et al., 2016) that can undergo autoxidation and lead to elevated gas-phase highly oxygenated organic molecule (HOM) concentrations (Bianchi et al., 2019;Ehn et al., 2014). Homogeneous nucleation of iodic acid (HIO 3 , IA) occurs over the ice-covered Arctic Ocean (Baccarini et al., 2020;Sipilä et al., 2016), while sulfuric acid (H 2 SO 4 , SA, formed through oxidation of marine DMS via SO 2 ), NH 3 , and air ionization by galactic cosmic radiation drive NPF in coastal Antarctica (Jokinen et al., 2018).
To bridge the connection between sea ice decline and associated changes in oceanic precursor emissions and related effect on aerosol and CCN populations, we conducted a suite of measurements at the two sides of the Fram Strait -at Villum Research Station, North Greenland, and at Ny-Ålesund on Svalbard. The stations are located at similar latitudes within a distance of about 600 km but they have distinctly different environmental conditions. Western Svalbard (Ny-Ålesund) is surrounded by open waters throughout the whole year due to advection of warm Atlantic water, whereas the ocean next to the North-Eastern coast of North Greenland (Villum) is permanently covered by ice with some larger polynyas occurring in the late summer.
Here we use state-of-the-art instrumentation, which enables observation of the NPF mechanism starting from molecular level and growing further to CCN sizes. Our study shows significant differences in particle formation and growth between these two locations ( Figure S1) and over different seasons. We show the relevance of low-volatility vapors, such as SA together with NH 3 , IA, HOM, and methane sulfonic acid (CH 3 HSO 3 , MSA) in Arctic NPF.

Materials and Methods
We conducted a measurement campaign at Villum Research Station in North Greenland (81°36′N, 16°39′W) from mid-February 2015 until the end of August 2015. At Villum, the particle number size distribution in the diameter range of 10-900 nm has been measured with a Scanning Mobility Particle Sizer (SMPS, Wang & Flagan, 1990) since 2010Nguyen et al., 2016). Additionally, we conducted a second campaign in Ny-Ålesund, Svalbard (78°55′N, 11°56′E), from the end of March 2017 until the end of August 2017. The measurements took place in the atmospheric observatory, Gruvebadet, located 2 km southeast of Ny-Ålesund.
During our measurement campaigns, we used a nitrate chemical ionization atmospheric pressure interface Time-Of-Flight mass spectrometer (CI-APi-TOF, Jokinen et al., 2012) to measure low-volatility va-BECK ET AL. 10.1029/2020GL091334 2 of 11 be reflected in the chain of processes linking the Arctic biosphere and cryosphere to atmospheric aerosol particles, clouds, and climate. pors, APi-TOF (Junninen et al., 2010) to measure ambient ions and clusters, and Particle Size Magnifiers (Vanhanen et al., 2011) to measure the concentration of sub-3 nm particles. At Villum, we used an Air Ion Spectrometer (AIS, Mirme & Mirme, 2013) to measure the charged particle number size distribution in the diameter range of 0.8-40 nm, and at Ny-Ålesund, we measured with a Neutral cluster and Air Ion Spectrometer (Manninen et al., 2009) which is measuring same sized charged particles as well as neutral particles (2-42 nm). Additionally, we used SMPS data, which are continuously measured at Gruvebadet. Furthermore, we measured marine chlorophyll a and dimethylsulfoniopropionate (DMSP) concentrations in nearby waters. The samples were collected between April 12, 2017 and May 26, 2017 at 10 m depth in Kongsfjorden, in a midfjord location in front of Ny-Ålesund.

Overall Behavior of Aerosol Precursor Gases
Following the end of the Polar night in February at Villum, we found a rise in the IA concentration with increasing intensity of solar radiation (Figure 1a). After the IA maximum close to 5 × 10 7 molecules cm −3 in early May, the IA concentration started to decay likely due to changes in sea ice properties and extent or ice-borne microalgae productivity (Saiz-Lopez et al., 2015). SA and MSA concentrations were rather low, and we did not detect gas-phase HOM during our campaign.
At Ny-Ålesund, the springtime IA concentration was lower and exhibited less pronounced temporal changes compared to Villum. In contrast, the concentrations of SA and MSA started to increase rapidly with an increasing radiation intensity, reaching daily average values larger than 10 6 molecules cm −3 for SA and larger than 10 7 molecules cm −3 for MSA ( Figure 1b). This rise coincided with the start of the phytoplankton bloom in mid-April, observed as an increase in chlorophyll a concentration in seawater (Figure 1c). During the peak of the bloom in early May, highly elevated DMSP concentrations were observed. During the 2017 bloom, the dominant plankton species was observed to be Phaeocystis pouchetii, a strong DMSP producer (Stefels et al., 2007). The species dominance, however, may vary regionally and from year to year (Hegseth et al., 2019). The high SA and MSA concentrations prevailed throughout June, which agrees with previous BECK ET AL.  studies of the observed phytoplankton activity in the Arctic Ocean (Assmy et al., 2017). HOM concentrations at Ny-Ålesund were very low during spring but started to increase steeply after the snow melt in the end of May ( Figure S2). Due to the complexity of the spectrum and the insufficient resolution of the instrument, it was not possible to identify the elemental composition of the HOM.  Figure 3c) could mostly be explained by IA, even though SA was found in both iodine clusters and in some of the smallest clusters either alone or together with NH 3 . The particle growth, however, was dominated by IA condensation (Figure 2a.5, see also Figure S3). Particles only grew up to 10 nm in diameter or slightly above (Figure 2a  Negative ion-induced nucleation drives particle formation with a contribution from positive ion-induced nucleation as well as potentially neutral nucleation during summer. The shaded area indicates the uncertainties. (a6) & (c6)-(e6): Total daily particle growth, calculated from particle measurements (AIS & SMPS), as well as based on gas concentrations of SA, IA, MSA, and HOM assuming kinetic condensation. For SA growth rate (GR), the dipole-dipole enhancement was taken into account (Halonen et al., 2019). Other GR were calculated based on Nieminen et al. (2010) due to lack of knowledge about enhancement of their collision rates. only a few weak NPF events were observed (Figure 2b.1) while negative ion-induced nucleation seemed to initiate the clustering (Figures 2b.3 and 2b.4). The identified nucleating clusters at Villum were composed of SA and NH 3 (Figure 3d). At Villum, the concentrations of particles larger than 10 nm were higher during summertime compared to spring, and according to our AIS measurements, the larger particles detected by SMPS were not nucleating in the vicinity of Villum but were advected there and continued to grow on site ( Figure S4). Previous studies have observed that larger, accumulation mode particles consisted mainly of sulfate and some organic material Nielsen et al., 2019).

Particle Formation at Ny-Ålesund and Villum
At Ny-Ålesund, the concentrations of SA and MSA were clearly higher ( Figure S5). On 4 May, they peaked at about 10 7 molecules cm −3 (Figure 2c.1) while IA concentration showed no clear temporal behavior, and the HOM concentration was low. Associated with the high SA concentration, we observed SA-NH 3 ion clusters (Figure 3a), indicating a clear negative ion-induced nucleation. Simultaneously, we observed the appearance of negative ion clusters between about 1 and 3 nm (Figure 2c.3), and the formation rate of 1.5 nm negative ions (J negative , J neg,max = 0.33 cm −3 s −1 ) was close to that of total 1.5 nm particles (J total , J tot,max = 0.27 cm −3 s −1 ; Figure 2c.5). The overall growth of newly formed particles to larger sizes was modest on this day, with particles reaching a maximum size of 20 nm in diameter (Figure 2c.6). During 8 May, we observed another example of negative ion-induced nucleation initiating the particle formation on both days (Figures 2d.3 and 2d.5, Figure S5), and with positive ion-induced nucleation being absent (Figures 2d.4 and  2d.5). While SA, IA, and HOM concentrations were similar to those on 4 May, the MSA concentration was higher by a factor of 5-6, leading to a higher total particle growth rate (Figure 2d.6). Pronounced growth with higher MSA was observed throughout the spring period ( Figure S5). On 5 and 6 August, the HOM concentrations were two orders of magnitude higher than in the springtime (Figure 2e.1), whereas SA and MSA concentrations were low. During summertime, the early growth (1.5-3 nm) was visible similarly in The mass defect is presented versus the ion/cluster mass/charge ratio depicting the abundance and atomic composition of nucleating ion clusters. The area of the dots is logarithmically proportional to the observed signal (counts per second). At Ny-Ålesund, MSA, IA and nitrate/nitric acid are detected in some of the smallest clusters but SA-NH 3 nucleation is a sole pathway for new particle nucleation. In springtime Ny-Ålesund, the largest detected cluster contained 8 molecules of SA and 8 molecules of NH 3 on bisulfate core ion (HSO 4 -) formed via interactions of SA and air ions. During summertime at Ny-Ålesund, SA-NH 3 clusters are not as abundant anymore as in springtime. At Villum in springtime, some SA-NH 3 clusters are seen but the nucleation is dominated by iodic acid clusters. The majority of the IA clusters contain also SA and/or sulfur trioxide. The observed I 2 O 5 clusters likely form from restructuring of two IA molecules in the clusters and recycling of water (Sipilä et al., 2016). During summertime in Greenland, there are more SA-NH 3 clusters abundant in the atmosphere while iodinecontaining clusters are seen to less extent than in springtime. both polarities (Figures 2e.3 and 2e.4), suggesting that the dominant particle formation mechanism was different from that in spring.
Direct measurements of ion cluster composition in springtime Ny-Ålesund showed the abundance of SA and NH 3 in small negative clusters (Figure 3a), in line with the observed nucleation rates. A comparison of observed (total) formation rates of 1.5 nm particles to the parameterizations based on experimental results (Dunne et al., 2016) suggests regional springtime NH 3 concentrations of 10 8 -10 9 cm −3 ( Figure S6a) in line with a recent modeling study (Croft et al., 2016), meaning that the nucleation rate is sensitive to both SA and NH 3 . Those observations align with a recent study by Giamarelou et al. (2016), indicating that 12-nm particles consist mainly of ammonium sulfate at Zeppelin Observatory in Ny-Ålesund. Furthermore, the nucleation rate is also sensitive to ion concentration and thus to the level of ionizing radiation due to the dominance of ion-induced pathway in the SA + NH 3 + H 2 O system with nucleation rates below ion-pair production rate (∼2 ion pairs per second) (Dunne et al., 2016;Kirkby et al., 2011). MSA and IA appeared to have only a small role in the initial stages of particle formation. However, any change in the SA, MSA, and IA concentration distribution could lead to a more complex nucleation mechanism, already reflected here as the smallest clusters containing MSA or IA (Figures 3a and 3b). The identified negative clusters during summertime events with high neutral HOM concentrations showed a similar composition, with clearly detectable SA-NH 3 clusters (Figure 3b), but unlike in springtime, malonic acid was present. Since HOM concentrations were high during summertime, it is highly probable that besides SA-NH 3 ion-induced nucleation, also (ion-induced) SA-HOM nucleation (Riccobono et al., 2014), pure biogenic (ion-induced) nucleation of HOM (Kirkby et al., 2016), and/or multicomponent SA-NH 3 -HOM nucleation (Lehtipalo et al., 2018) contributed to the total particle formation rate (Figure 2e.5). HOM concentrations of just a few 10 6 molecules cm −3 are sufficient to promote nucleation together with SA (Riccobono et al., 2014), whereas HOM concentrations of a few 10 7 molecules cm −3 can initiate NPF in the absence of SA, and nucleation can be seen equally effectively with both polarities (Kirkby et al., 2016).

Particle Growth and Subsequent CCN Production
Our observations at Ny-Ålesund suggest that MSA frequently had a significant contribution to the growth of newly-formed particles (Figures 2b and S6b). This is quite a surprising result, considering that MSA is more volatile than SA (Berresheim et al., 2002), and that the observed methane sulfonate-to-non-sea-salt-sulfate ratios (MSA − /nss-SO 4 2− ) in the bulk summer Arctic aerosol are usually (also in our case) well below unity (L. Chen et al., 2012; Figure S7). The reason for low MSA − /nss-SO 4 2− in bulk aerosol samples is most probably associated with heterogeneous reactions producing sulfate in wet aerosol particles or cloud droplets (Pozzoli et al., 2008) more efficiently than methane sulfonate (Q. Chen et al., 2018;Hodshire et al., 2019), or with long-range transportation of anthropogenic sulfate (L. Chen et al., 2012). Thus, bulk aerosol measurements clearly do not reflect the growth mechanisms or chemical composition of the smallest particles. In chamber studies, MSA has been shown to form particles in mixtures (Dawson et al., 2012) and our observations suggest that MSA is an important component of the Arctic secondary aerosol. During summer, the particle growth was obviously dominated by HOM condensation at Ny-Ålesund (Figure 2c.6), as the particles grew up to 40 nm in spite of low concentrations of SA, MSA, and IA. The high concentration and predominant role of HOM in the particle growth is also an unexpected result, considering that the terrestrial biomass in the Ny-Ålesund area is scarce. HOM concentrations in excess of 10 8 molecules cm −3 are typical for forests (Ehn et al., 2014), and therefore finding similar concentrations in Svalbard was a surprise. An explanation for this could be the very low condensation sink of ∼4 × 10 −4 s −1 at Ny-Ålesund (∼3 × 10 −4 s −1 at Villum), which allows a more efficient buildup of the gas-phase HOM concentration tied to the slower scavenging rate of HOM by pre-existing aerosol particles.
The climatic effects of atmospheric NPF are tied closely with two issues: the growth of newly-formed particles to sizes large enough to be able to act as CCN (Gordon et al., 2017;Kerminen et al., 2012), and the susceptibility of cloud properties to additional CCN (e.g., Moore et al., 2013;Zhao et al., 2018). The susceptibility of clouds to CCN tends to be highest under clean atmospheric conditions, such as those typically observed over the Arctic outside the Arctic haze period. Cleaner conditions also tend to cause higher water vapor supersaturations inside liquid clouds (Hudson & Noble, 2014), reducing the minimum size at which newly formed particles are able to act as CCN. At the Zeppelin Observatory (480 m a.s.l. above Ny-Ålesund), activation of nucleated particles into droplets of low-level liquid clouds was observed once these particles had grown to sizes larger than about 20 nm ( Figure S8). Activation of Aitken mode aerosol particles as small as 20 nm into cloud elements was shown to occur on a regular basis at Zeppelin, especially in the absence of accumulation mode particles and when the particle number size distribution was dominated by recent NPF . Based on these analyses, we speculate that atmospheric NPF observed at Ny-Ålesund, despite a rather slow consequent growth to Aitken mode size aerosol, may also contribute to the regional CCN budget.
At Villum, the lower concentration of SA and MSA compared to Ny-Ålesund can be explained by the extensive presence of sea ice in the region around Villum, limiting the formation of DMSP producing phytoplankton blooms and preventing gas exchange between the ocean and atmosphere ( Figure S9). In our study, at Villum, IA alone seems insufficient to grow particles effectively into CCN size regimes. However, in other parts of ice-covered oceans with stronger iodine sources growth to CCN sizes can occur as recently shown by Baccarini et al. (2020). IA-containing particles might also serve as seeds for SA, MSA or organic matter condensation if advected over areas with stronger DMS or VOC emissions. Thus, iodine emissions from sea ice might still be important for Arctic CCN budgets close to the sea ice edge and in regions with non-consolidated ice. Furthermore, in other parts of ice-covered oceans, where iodine emissions and thus IA concentrations are higher, the iodine-mediated formation and aerosol growth to CCN size particles will be more effective. Iodine-containing CCN-sized particles have been detected in measurements done in the Greenland Sea (Allan et al., 2015).

Conclusions and Implications for the Changing Arctic Environment
Our novel measurement techniques enabled the observation of the very first steps of nucleation and growth at a molecular level, complementing the current understanding of NPF in the Arctic (Baccarini et al., 2020;Dall'Osto et al., 2017. Our observations show great disparity in particle precursor vapor concen-BECK ET AL. 10.1029/2020GL091334 7 of 11 trations and NPF events between the two Arctic locations with divergent environmental settings (Figure 4, Figure S1). Over the open water surrounding Ny-Ålesund, we observed that ion-induced nucleation of SA and NH 3 is an important pathway for secondary aerosol formation during the Arctic spring, with nucleation rates being sensitive to both, SA and NH 3 concentration. These observations are in good agreement with the study of H. Lee et al. (2020), showing that NPF is associated with enhanced DMS and NH 3 concentrations at the Zeppelin station, Svalbard. According to our observations, MSA and SA were critical for the particle growth in springtime. With high HOM concentrations during summertime, this compound class was likely to be involved in nucleation (Lehtipalo et al., 2018) and probably gave the dominant contribution to subsequent particle growth, especially because all the other low-volatility vapors had rather low concentrations. The region of sea ice-covered ocean surrounding Villum showed a much weaker particle growth throughout the measurement period. In springtime, we observed pure iodic acid nucleation at Villum, whereas in summertime, we observed weak SA-NH 3 nucleation.
Enhanced DMS emissions associated with sea ice loss and higher surface water temperatures have already been reported over the summertime Arctic Ocean Li et al., 2019). Further sea ice loss might increase the productivity of pelagic phytoplankton (Ardyna et al., 2014;Renaut et al., 2018), and/ or seawater-atmosphere exchange, which could lead to even higher atmospheric DMS concentrations in the future. In the regions dominated by sea ice, IA seems to be an important contributor to the nucleation and growth processes, although the newly formed particles seldom grow to CCN sizes where they become relevant for cloud formation, unless the air mass is transported over regions with strong emissions of DMS or gaseous precursors to SA. Thinning sea ice will likely lead to enhanced emissions of iodine due to more efficient transpassing of radiation and gases through the ice layer (Cuevas et al., 2018), causing elevated atmospheric IA concentrations. Those enhanced emissions will likely be a transient effect while the sea ice continues diminishing. Emissions of VOC are expected to increase with warming climate and lesser snow coverage, at least from peat and tundra vegetation (Lindwall et al., 2016) but possibly also from the Arctic Ocean (Mungall et al., 2017), which would be reflected in higher concentrations of HOM and accelerated secondary aerosol formation. Based on our results on mechanisms of aerosol formation, all these changes would lead to an increased production of new aerosol particles and their potential growth to CCN, thus affecting cloud microphysical properties in the summer Arctic atmosphere (Mahmood et al., 2019;Ridley et al., 2016). It remains, however, uncertain how, for example, phytoplankton will respond to other components of Arctic change (Hoppe et al., 2018) or how bird populations respond to global change and concomitant changes in the marine food web (Howard et al., 2018).
We have shown the process chain from the phytoplankton metabolite production to CCN-size aerosol formation, and confirmed the importance of oceanic DMS and NH 3 emissions as well as galactic cosmic radiation in Arctic secondary aerosol formation ( Figure 4). Furthermore, we have illustrated the role of iodine and HOM emissions in this process. Definite predictions on how CCN concentrations and radiative forcing will respond to Arctic change cannot yet be made. Still our findings are critical for assessing the relevance of the famous, yet debated CLAW-hypothesis (Charlson et al., 1987;Quinn & Bates, 2011), which suggests a negative feedback of climate warming via enhanced phytoplankton productivity. The mechanistic understanding obtained here will help to improve large-scale models to shed more light on the pre-industrial, present and future aerosol state and on the magnitude of total radiative forcing associated with the inevitable Arctic warming and evanescence of sea ice.