Wintertime sub-arctic new particle formation from Kola Peninsula sulphur emissions

Metallurgical industry in Kola peninsula, North-West Russia, form a second largest source of air pollution in the Arctic and sub-Arctic domain. Sulphur dioxide emissions from the ore smelters are transported to wide areas including Finnish Lapland. We performed investigations on concentrations of SO2 and aerosol precursor vapours, aerosol and ion cluster size distributions together with chemical composition measurements of freshly formed clusters at SMEAR I station in Finnish 15 Lapland relatively close (~300 km) to Kola peninsula industrial sites during winter 2019-2020. We show that highly concentrated SO2 from smelter emissions is converted to sulphuric acid (H2SO4) with sufficient concentrations to drive new particle formation hundreds of kilometres downwind of the emission sources even with very low solar radiation intensities. Observed new particle formation is primarily initiated by H2SO4 – ammonia (negative-) ion induced nucleation. Particle growth to cloud condensation nuclei (CCN) sizes was concluded to result from sulphuric acid condensation. However, airmass 20 advection had a large role in modifying aerosol size distributions and other growth mechanisms cannot be fully excluded. Our results demonstrate the dominance of SO2 emissions in controlling winter-time aerosol and CCN concentrations in the subarctic region with heavily polluting industry.

The metallurgical industry with large scale smelter complexes in Kola peninsula, North-West Russia, form the second largest source of air pollution in the Arctic and sub-Arctic region. Smelters emit large quantities of SO2, metals and particulate matter to the atmosphere. These pollutants, especially SO2, largely impact the atmosphere and biosphere in the area, including eastern parts of Finnish and Norwegian Lapland. In the close proximity of industrial plants these pollutants have literally destroyed 35 ecosystems creating "industrial deserts" Paatero et al. (2008). Though emissions have significantly decreased from ca. 600 kilotons yr -1 in early 1990's (Tuovinen et al., 1993;Ekimov et al., 2001), partly because of the collapse of Soviet Union and Eurasia and therefore understanding their role in atmospheric chemistry and physics is of great importance.
SO2 can be photochemically oxidized to sulphuric acid (H2SO4) in the gas phase. Though most of SO2 reacts in liquid phase in 45 cloud droplets and precipitate as acid rain, with very high concentrations of SO2 in the Kola peninsula area also high production rate of gas phase H2SO4 can be expected. H2SO4 vapour can, in turn, contribute to atmospheric new particle formation (NPF) via nucleation and subsequent particle growth even up to sizes of cloud condensation nuclei (CCN) by further condensation of H2SO4 and potentially some other vapours (e.g. Weber et al., 1995;Kirkby et al., 2011;Jokinen et al., 2018). NPF is an important process because, according to model simulations, it accounts for more than a half of atmospheric CCN formation 50 globally (Merikanto et al., 2009;Gordon et al., 2017). In the high latitudes the contribution of NPF is estimated to be even larger, reaching >90% of the cloud level CCN in the high Arctic and approximately 70-80% in our study area, the sub-Arctic zone of Northern Finland and North-Western Russia (Gordon et al., 2017). Vehkamäki et al. (2004) were first to report observations of NPF (> 8nm particles) at Värriö SMEAR I field station in eastern 55 Lapland, Finland, relatively close to Kola peninsula smelter complexes. Their results on the contribution of SO2 pollution were not completely definitive, out of four years of measurements and 147 observed NPF events 15 were concluded to be explained by the pollution plume. Kyrö et al. (2014) who recorded particle number size distributions down to 3 nm in diameter showed that NPF is connected to high concentrations of SO2. They observed NPF even during winter in almost dark conditions, indicating that during episodes of vast concentrations of SO2 sufficient fraction is converted to H2SO4 in the gas phase even in 60 very low solar radiation levels to initiate NPF. However, to date, no reports on quantification of sulphuric acid concentration by direct measurement or detailed mechanisms and chemical compounds involved in NPF in the area exists.
While observation of atmospheric NPF has been reported in hundreds of publications since the times of John Aitken (1900), the details, i.e. the dynamics and contributing compounds, of NPF have been experimentally resolved only in a limited number of atmospheric environments. Pioneering studies include observations by Weber et al. (1995) on the connection of between sulphuric acid and atmospheric nucleation, and the first report on ion-induced nucleation and simultaneous detection of sulphuric acid anion clusters by a mass spectrometer by Eisele et al., (2006). Among the field work, several laboratory investigations by the same research groups probed the properties of sulphuric acid -water and sulphuric acid -ammoniawater clusters and their potential role in new particle formation (Ball et al., 1999;Hanson and Eisele, 2002).

70
Later advances in understanding the molecular steps of nucleation and growth in the atmosphere include the discovery that iodic acid (HIO3) is primarily responsible on nucleation and growth in coastal areas and in the vicinity of the Arctic sea ice (Sipilä et al., 2016;Baccarini et al., 2020). Jokinen et al. (2018) demonstrated that in close coastal Antarctica, H2SO4 from oxidation of dimethyl sulphide (DMS, emitted by pelagic phytoplankton) and ammonia (NH3, from penguin colonies) nucleate 75 via negative ion-induced mechanism with sulfuric acid condensation accounting for most of the subsequent particle growth.
Further observations on nucleation mechanisms indicate the key role of highly oxidized organic molecules HOM (Ehn et al., 2014) in NPF in the spring-summer time boreal forest (e.g. Kulmala et al., 2013;Rose et al., 2018) and in the mid-latitude continental free troposphere (Bianchi et al., 2016) in parallel with sulfuric acid -ammonia nucleation (Bianchi et al., 2016;Schobesberger et al., 2015;Yan et al., 2018). Dimethyl amine has been found to contribute to initial nucleation in polluted 80 urban air (Yao et al., 2018). In addition to these, yet rare molecular level atmospheric observations several laboratory studies have investigated the details of these nucleation mechanisms (e.g. Kirkby et al., 2011;Almeida et al., 2013;Kürten et al., 2014;Kirkby et al., 2016).
Mass spectrometers (Junninen et al., 2010;Jokinen et al., 2012) and air ion spectrometers (Mirme & Mirme, 2013), have 85 largely facilitated the recent progress in the field of atmospheric NPF. By utilizing them in conjunction with aerosol and meteorological observations, this work aims to shed light on the molecular steps of NPF resulting from (sub-)Arctic air pollution during wintertime. Investigations were carried out at SMEAR I research station in Värriö strict nature reserve in Finnish Lapland close to the industrial plants of Kola Peninsula, north-west Russia, during the winter 2019 -2020.

Site and time of the study
Measurements were carried out at Värriö SMEAR I research station (Hari et al., 1994)  closest, relatively small, coal burning plant is 550 km away. SMEAR I station was set up in 1991 for monitoring the air pollution, especially sulphur dioxide (SO2) originating from Kola peninsula smelters. In this work we present 4.5 months data from winter time, 1 st November 2019 until 16 th March 2020.

100
Aerosol number size distribution between 3 and 750 nm was recorded by a twin differential mobility particle sizer (DMPS) (Aalto et al., 1999) comprising Hauke-type differential mobility analyzers (lenghts 110 and 280 mm) and TSI-3776 and TSI-3772 condensation particle counters (TSI Inc., Shoreview, MN, USA) as detectors. Another DMPS malfunctioned during 9 th -10 th and 14 th -27 th January resulting in the loss of data on 3 -10 nm particles.

Nucleation rate calculation
120 Negative (-) and positive (+) ion-induced nucleation rates of 1.5 nm particles, J -/+ 1.5, were calculated assuming a steady-state between formation and loss of particles in the size range of 1.5 and 2.5 nm: where N -/+ 1.5-2.5 is the total concentration of negative or positive ions in the size range between 1.5 and 2.5 nm, filtered using Matlab's Savitzky-Golay 2 nd order filtering method with a frame size of 51 to remove instrument noice, krec is the recombination coefficient between negative and positive small ions which is here approximated by a size independent constant of 1.6×10 -6 cm 3 s -1 (Tammet, 1995), N +/-<1.5 is the concentration of positive or negative sub-1.5 nm cluster ions. GR2 is the 2 nm particle growth rate, width of the size interval for which concentration is defined Δdp = 2.5 nm -1.5 nm = 1 nm and CoagS is the 130 coagulation sink of 2 nm particles to background aerosol: where Ni is the concentration of particles in the channel i of DMPS and coagulation coefficients K2nm,i between 2nm, and i nm 135 particles are calculated based on Seinfeld and Pandis, 1998.
Accurate determination of particle growth rate for 2 nm particles from the size distribution is challenging, and therefore GR2 as a sole mechanism of growth (Jokinen et al., 2018). Justification for the approach will be discussed later. More standard method for GR determination is to approximate GR2 by the average growth rate of the formed particle population, including mainly particles grown far above the 2 nm size, during a few hours starting from the beginning of the event as demonstrated 145 in Supplement Figure S1. However, this approach neglects the effect of airmass advection which, as will be discussed later, may largely determine the time development of the size distribution and thus also the apparent growth. These two methods lead to a factor of ~17 difference in GR2 on the example day depicted in Supplement Figure S1. For example, on example day of 29 th January 2020 the factor of ~17 in GR2 , is reflected in a factor of ~1.9 difference in the calculated nucleation rate. Ioninduced nucleation rate calculation is thus not very sensitive to GR2 because ion-ion recombination term dominates the loss in 150 our conditions.

Sulphuric acid proxy calculation
Because of significant gaps in the measured data, [H2SO4] was also calculated using a proxy developed by Dada et al. (2020).
Calculation accounts for oxidation of SO2 to H2SO4 both by OH (proxied by global radiation) and stabilized Criegee Intermediates (Sipilä et al., 2014) proxied by monoterpene and ozone concentrations, as well as loss off H2SO4 by dimerization 155 (negligible in observed concentrations) and condensation on pre-existing aerosol (the primary loss term). Unfortunately, there are no VOC measurements available at SMEAR I, but because the data were collected during winter well outside of the growth period, we assumed monoterpene concentration to be zero. Global radiation measurement showed unexplained fluctuation (maybe caused by low solar zenith angles or freezing of the sensor) during the measurement period and therefore we used UVB radiation and the relation between UVB and Global radiation determined by Dada et al. (2020).

Trajectory analyses
Trajectories were calculated by using the Hybrid Single-Particle Lagrangian Integrated Trajectory model HYSPLIT (Stein et al., 2015) with GFS 0.25 degree meteorology as an input. We calculated 24-hour backward trajectories arriving at 250 meters above ground level, for the period 28.1.2020 00UTC to 30.1.2020 00 UTC, arriving every 6 hours. This is approximately the 165 altitude at which SMEAR I station is and thus should represent the air that is also observed at the station even in a strong temperature inversion.
3 Results and Discussion 3.1 New particle formation during the measurement period.

180
Not all events occurred during eastern winds. The events observed close to mid-winter, from early December until early January, occurred with western winds or during the transition of wind direction from west to east, in relatively low concentration of SO2 and virtual absence of day light and H2SO4. These low-H2SO4 mid-winter events are observed to start from sizes larger than few nm, i.e. nucleation does not take place in-situ in the surroundings of SMEAR I. Those particles thus have formed elsewhere and transported to the measurement site by horizontal advection or vertically down from above the 185 mixed layer. While new particles, with diameters of few nanometres, can survive a while against loss by coagulation, H2SO4 is lost much more rapidly to pre-existing particles after its production ceases. Therefore, the lack of H2SO4 is not excluding its primary role in NPF though not supporting that role either. Some of the mid-winter events coincide with elevated SO2, suggesting that sulphuric acid may have be formed in the airmass earlier, while some, especially the relatively strong event on December 3 rd presented in the Supplement (Figures S8-S10) occurs in the virtual absence of SO2 suggesting that sulphuric acid 190 has not been formed to significant extent in that airmass. Currently we thus cannot explain the mechanism of NPF that day.
However, the NO2 concentration is slightly ( Figure S8) elevated in the air mass, which might be connected to the event or source of particles. Nevertheless, most of the events seem to be connected to the presence of H2SO4.

Case study 28 th -29 th January 2020
To resolve the details of new particle formation and growth, we focus on 3 time periods with clearly occurring nucleation and 195 particle growth. Here we show results from analysis of a 2-day period from 28 th -29 th January 2020. To demonstrate that the

215
Despite the low UVB radiation, required for O3 photolysis that initiates H2SO4 production via OH radical formation, H2SO4 concentration increases from close to lowest detection limit values of ~10 5 molecules cm -3 up to 8×10 5 molecules cm -3 during 28 th and up to 1.5×10 6 molecules cm -3 on 29 th (Figure 5e). Because OH production rate must be low, high SO2 concentration is a perquisite for the H2SO4 production during cold and dark winter months. Though stabilized Criegee Intermediates (sCI) formed in alkene ozonolysis can oxidize SO2 to produce H2SO4 during summertime Sipilä et al., 2014), 220 alkene (terpene) emissions from the vegetation and thus sCI production rate are negligible during the winter season. Proxy calculations agree well with the measured sulphuric acid concentration on 29 th but show clearly higher values on 28 th January.
The cause of the disagreement on 28 th January is probably the stable and shallow boundary layer. Temperature gradient close to surface is almost +0.2 °C m -2 at noon on 28 th (Figure 3b). Solar radiation from close the horizon is not penetrating efficiently inside the canopy and thus UVB measured above canopy and used in the proxy calculation does not reflect the situation at the 225 ground level. Sulphuric acid produced above the canopy, on the other hand, does not mix downwards due to strong temperature inversion and calm wind. On 29 th January the gradient is absent or slightly negative allowing surface air to mix with the air above canopy.
Besides H2SO4, also minute signals of iodic acid (HIO3) is observed during the day (Figure 5e). The exact production 230 mechanism of HIO3 remain globally unknown despite the emerging evidence on its critical role in new particle formation especially in the Arctic regions (Sipilä et al., 2016;Baccarini et al., 2020). Methane sulphonic acid (MSA), that has been observed in larger aerosol particles (refs.) and that could potentially contribute also to NPF, hardly exceeds the detection threshold. This was expected since MSA originates from dimethyl sulphide (DMS) photo-oxidation. DMS end up in the air mainly from metabolism of pelagic phytoplankton during summer months, not during dark winter. No other condensable 235 vapours, such as HOM which dominate the new particle growth in summertime boreal forest (Ehn et al., 2014), were observed during this case study period or during other periods depicted in Supplement.

New particle formation
Ion size distribution 240 Figures 5a&b show the NPF events on 28 th and 29 th January as observed by NAIS operated in ion mode. Omnipresent small < 1.5 nm ions are continuously produced by galactic cosmic radiation and radon decay. Approximately at 11 o'clock on 28 th January, coinciding with the increase of H2SO4 concentration, small negative cluster ions start to grow, which is seen as small increase in ~1.5 -2 nm negative ion concentration. During their growth beyond ~2 nm in diameter, those cluster are neutralized by collisions with positively charged ions and thus they disappear from the spectrum. They still obviously continue their growth since charged particles reappear in the spectrum after reaching some 5 nm in diameter when diffusion charging becomes effective enough; equilibrium charging state for 2 nm particles is 0.8% while 5 nm particles are charged with an efficiency of 2.3% and out of 20 nm particles 11% are negatively charged (Wiedensohler et al., 2012). Opposite to negative ions, positive cluster ions do not grow. Larger, > 5nm positive particles (charged by diffusion charging during the course of their growth) grow similar to negative particles. On 29 th January, with clearly higher H2SO4 concentration, the appearance of >1.5nm 250 negative clusters is more pronounced suggesting higher rates of nucleation and critical role H2SO4 in the initial steps of NPF.
Positive cluster ions are again only bystanders and do not contribute to nucleation. This observation suggests negative ioninduced nucleation as a primary pathway to new particle formation similar to H2SO4 -NH3 (

Nucleation rates
Even though weak growth of the small negative ions around noon on 28 th January is visually observable in Figure 5a the concentration of clusters in the 1.5 -2.5 nm size range (N -1.5-2.5) is hardly distinguishable from the noise (Figure 5c). Nucleation rate, calculated using filtered concentration data, only slightly exceeds the baseline (caused by presence of minute, almost 260 omnipresent signal from ion clusters extending above 1.5 nm but which is not connected to sulphuric acid nucleation), and is approximately 0.005 cm -3 s -1 with high relative uncertainty (Figure 5d). On 29 th , with 2.3-fold sulphuric acid concentration, the concentration of 1.5 -2.5 nm negative clusters is well above the instrument noise reaching 20 cm -3 s -1 around noon. Nucleation rate peaks at 0.064 cm -3 s -1 . Temperatures during nucleation (~ noon) were close identical, around -22°C, in both days and therefore they can be directly compared. Approximately 10-fold difference in nucleation rate between the two days and a factor 265 of 2.3 difference in sulphuric acid concentration is in line with the results from the CLOUD-chamber experiment on sulphuric acid -ammonia -water nucleation (Kirkby et al., 2011). The so-called "slope" that approximately (not exactly in real atmospheric situations) equals to the number of sulphuric acid molecules in the critical cluster (Vehkamäki et al., 2012 Figure 5f). Since H2SO4 is a stronger acid than HNO3, proton transfers from H2SO4 to NO3explaining the observed behaviour when [H2SO4] starts to rise. When [H2SO4] still increases during the course of the day, (NH3)m(H2SO4)nHSO4 --clusters start 285 to form. Cluster signals peak around noon coinciding with the highest [H2SO4] and N -1.5-2.5 after which they start to decay. On 29 th January, the same behaviour is observed, but with somewhat stronger cluster signals due to the higher [H2SO4].

290
To get more insight into the chemical composition of clusters, the ion-cluster mass spectrum was integrated over 4 hours (2 hours effective data collection due to switching between CI and ion-inlet). The resulting spectrum is presented in Figure 6 by means of a mass defect plot, where the mass-to-charge ratio (m/z, unit Th) corresponds -with only singly charged ion clusters present in the air -to the mass of the cluster (m, unit Da, equal to unified atomic mass unit, u). Mass defect is the mass difference (in Th or Da) between the exact mass of the cluster and the integer mass defined as the sum of nucleons in the 295 atomic nuclei of the cluster; for example, exact mass of a HSO4 --ion that has 97 nucleons is 96.960103 Da and the mass defect is thus 0.039896 Da. Area of the dot is proportional to the logarithm of the observed signal intensity. In the mass defect plot each addition of a molecule or atom is represented by a vector. Addition of, e.g., H2SO4, with a negative mass defect leads to increasing mass and decreasing total mass defect, while an addition of a positive mass defect NH3 molecule leads to increasing total mass defect. Successive addition of certain molecules to an ion results into a straight line in the defect plot so that the 300 different cluster formation pathways are readily distinguishable from the plot.
Largest signals are associated with omnipresent nitrate ion and its cluster with nitric acid (NO3and HNO3×NO3 -). Rest of the small (<180 Da) ions are mainly different sulphur species, with bisulphate ion partly clustered with nitric acid (HSO4and HNO3×HSO4 -) being most abundant. Other small sulphur ions present in the spectrum are SO4 -, SO5 -, HNO3×SO3and 305 HNO3×SO4 -. Deprotonated iodic acid (IO3 -) and its nitric acid cluster, (HNO3×IO3 -) are also abundant. Despite multiple different types of these core ions, initial growth of them is solely caused by the attachment of sulphuric acid molecules. We observed clusters with 1-4 H2SO4 molecules attached to SO4 --ion, one H2SO4 molecule to SO5and SO3 --ions and 1-3 H2SO4 molecules to IO3 --ion. For simplicity, we assume that the negative charge remains in the core ion. This is not necessarily true, but H2SO4 may lose a proton e.g. to IO3resulting in composition of HIO3× (H2SO4)n-1×HSO4instead of -(H2SO4)n×IO3 -. Furthermore, water, if present in the clusters, efficiently evaporates in the vacuum of the mass spectrometer and therefore information on the role of water in the cluster formation is lost.
None of the above discussed clusters seem to adopt ammonia efficiently enough for their signals to exceed the detection threshold of the APi-TOF (mass dependent, ~10 -3 to few 10 -3 ions/second for 2 hour integration for m/z > 400 Th). Only 315 clusters made solely of sulphuric acid with a bisulphate ion (HSO4 -) as a core seem to efficiently attach ammonia resulting in the formation of (NH3)m× (H2SO4)n×HSO4 --clusters (n>=3). This sequential addition of NH3 and H2SO4 has been shown to be an effective (ion-induced) cluster formation and growth mechanism in coastal Antarctica (Jokinen et al., 2018) as well as a secondary pathway in the free troposphere (Bianchi et al., 2016) and in the spring/summer time southern Finland boreal forest .

320
Our results on negative cluster composition can be compared to the results from CLOUD experiment at -25°C for varying NH3/H2SO4 -ratio (Schobesberger et al., 2015). Based on those data, with the NH3/H2SO4 -ratio exceeding approximately 100 the cluster composition, and also the nucleation rate (Kirkby et al., 2011), saturate and become unaffected by the increase of NH3-concentration. In those conditions a cluster comprising 3 molecules of sulphuric acid on a bisulphate ion, (NH3)n× 325 (H2SO4)3×HSO4 -, contains on average approximately n~1 molecule of ammonia and cluster composed of 4 molecules of sulphuric acid and a bisulphate ion, (NH3)n× (H2SO4)4×HSO4 -, carries on average approximately n~1.5 NH3 molecules (Schobesberger et al., 2015). In our case (Figure 6), corresponding average ammonia numbers are n~0.4 and n~0.8 for (NH3)n×(H2SO4)3×HSO4and (NH3)n×(H2SO4)4×HSO4 -, respectively, suggesting that NH3/H2SO4 -ratio in our case is well below 100, likely below 10 (Schobesberger et al., 2015). That would indicate ammonia concentration of the order of ~10 7 molecules 330 cm -3 , or ~1 pptv. Low NH3/H2SO4 -ratio would mean that the system is not saturated with respect to NH3 and the nucleation rate should therefore be sensitive to both H2SO4 and NH3 similar to Jokinen et al. (2018). This highlights the importance of understanding NH3 sources, transportation and atmospheric mixing ratios down to sub-ppt levels for a proper description of new particle formation also in the subarctic region. Unfortunately, NH3 concentrations in the range of 1 pptv are not (reliably) detectable with any present-day measurement technology.

335
The present analysis shows that sulphuric acid -ammonia ion-induced nucleation can trigger new particle formation in the winter time sub-arctic / boreal environment with high level of anthropogenic SO2 pollution but low UV-radiation intensity.
Data on neutral 1.5 -3nm particles are not available and neutral nucleation rates could not be derived. However, based on all evidence obtained from field (mainly Jokinen et al., 2018) and especially from CLOUD experiments (Kirkby et al., 2011; 340 Schobesberger et al., 2015), in absence of significant amounts of compounds other than H2SO4 and NH3 and with nucleation rate below the ion pair production rate (typically 2-4 ion pairs cm -3 s -1 in the Earth's surface layer) ion induced nucleation dominates over the neutral process. In our case, HOM were below the detection limit, and amines, if important, would appear also in the anion spectrum in H2SO4 clusters. HIO3 and MSA were present, but significant neutral homogeneous nucleation of HIO3 would require ~100-fold concentration (Sipilä et al., 2016).

345
The observation of clusters containing IO3or HIO3 together with H2SO4 is, however, highly interesting. HIO3 has been shown to nucleate homogeneously but no reports on mixed clusters containing both HIO3 and H2SO4 exists though Sipilä et al. (2016) found unidentified sulphur compounds in the ion cluster population dominated by iodic acid in the Atlantic coast. If the SO2 rich pollution plumes are advected to iodine source areas (arctic ocean and especially sea ice zone as well as macroalgae rich 350 coasts) or vice versa, this mixed mechanism may become important.

Particle growth and relevance for CCN-concentrations
Based on the above analysis, particle nucleation is clearly driven by sulphuric acid and ammonia with nucleation rate sensitive to concentrations of both. But how do the freshly nucleated clusters grow? Assuming irreversible condensation, even the peak sulphuric acid concentration of 1.5×10 6 molecules cm -3 can explain only a small fraction of the observed growth rate. Consistent 355 with an earlier report on wintertime particle growth rates at Värriö (Kyrö et al., 2014), the apparent average growth rate on 29 th January is approximately 4.5 nm h -1 ( Figure S1), which, based on Stoltzenburg et al. (2020) would require a steady concentration of ca. 2.6×10 7 molecules cm -3 throughout the growth process which continues long after the sunset and the decay of [H2SO4]. Obviously, there are two possibilities for the disagreement. Either sulphuric acid is not responsible for most of the growth or the air is not homogenous and the apparent growth is caused by the airmass advection.

360
Besides sulphuric acid, only condensable vapours detected are MSA and HIO3 (and NH3). However, their concentrations are clearly lower than that of sulphuric acid, and therefore condensation of those in a homogeneous airmass cannot explain the apparent growth either. It could be speculated that compounds not recorded by CI-APi-TOF, such as SO2 or some volatile or semi-volatile organic compounds, (S)VOC, react in particle phase forming low volatile compounds therefore contributing to 365 growth but we have no evidence on such a process. NO2 concentration was moderate, up to 7 ppb, and therefore also nitric acid concentrations were likely insufficient to have a measurable effect on growth (Wang et al., 2020). However, temperature was low during the studied time period and therefore HNO3 or some contribute to growth, if they were present. Ammonia was detected in small ion clusters but its contribution to particle volume, assuming the cluster NH3/H2SO4 -ratio reflects the composition of also larger particles, is marginal. Assuming particle composition to be ammonium bisulphate, i.e. NH3/H2SO4 370 -ratio of unity, ammonia would contribute 17% to the particle volume and 5% to particle diameter growth rate.
The most plausible explanation for the observed growth is that the particle growth is driven by H2SO4 condensation but its concentration is not uniform over the source area. In that case, particles would nucleate and grow to their final size during the few hours of sunlight. Particles formed and grown in the close proximity of emissions source in high SO2 and thus H2SO4 concentration environment grow to larger sizes than particles that are formed near the measurement site. Airmass advection would then transport particles through the dark hours leading to steadily increasing nucleation (and later Aitken) mode diameter at SMEAR I observed as apparent steady growth even through the night. Modelling efforts and measurement of chemical composition or hygroscopicity of growing mode would be required for unambiguous explanation of the particle growth.

380
New particle formation in the sub-Arctic winter would be irrelevant if formed particles would not grow to sizes (above few ten nm) where they can act as CCN. CCN concentrations (in different supersaturation) were not recorded but the airmass arriving from the Murmansk -Kandalaksha region from 28 th January ca. 3:00 onwards (Figure 4) contains elevated concentration of Aitken and accumulation mode particles mainly in the size range of ~20 -500 nm (Figure 7). New particle formation clearly increases the concentration of >3 nm particles but also the concentration of particles larger than 50 nm show 385 increase especially on 29 th January. >100 nm particle concentration is relatively constant and apparently unaffected by the NPF in the time scale the events could be observed. Airmass advection, and particle loss processes, however, naturally have an impact on measured concentrations and are largely responsible on development of particle populations. Figure 8 presents the average particle number size distribution during the ca. 1-week period of eastern winds (28 th Jan -3 rd 390 Feb 2020) when the two clear NPF events presented above occurred. Concentrations of particles in all size classes were remarkably higher, even an order of magnitude for 10-200 nm particles, than the average concentrations during the preceding and succeeding time period with western winds. Concentrations during that 1-week period were also clearly higher the average concentrations between 1 st November and 29 th February suggesting that new particle formation may be a significant source of particles in the eastern airmasses. However, primary emissions from smelters and the surrounding cities would naturally show 395 up in the size distribution plot as well. More thorough analysis is needed to separate the roles of secondary NPF and primary emissions in the aerosol and CCN budgets. For clarity March, with almost constant NPF is excluded from this analysis.
Since the observed secondary particles are probably highly acidic, they are also highly hygroscopic extending the CCN active fraction toward smaller particle sizes. For accurate assessment of contribution of secondary aerosol formation in CCN 400 concentrations at SMEAR I or regionally, meteorological situation, including boundary layer dynamics, wet deposition of particles, etc. should be accounted for. However, our observations point toward a major contribution of Kola peninsula SO2 emissions to winter time CCN concentrations in the region.

Conclusions
Winter time secondary new particle formation and growth was investigated at SMEAR I station, in Värriö strict nature reserve, 405 Finnish Eastern Lapland. Sulphur dioxide concentration in the airmasses arriving from Kola peninsula were often very high, occasionally over 30 ppb. In these high concentrations enough sulphuric acid formed even in very low solar radiation intensity to initiate new particle formation and growth.
New particle formation was observed mostly, but not solely, with eastern winds and in airmasses arriving from the direction 410 of Kola peninsula smelters. Newly formed (4-10 nm, concentration > 50 # cm -3 ) particles were observed in altogether 51 days between 1 st November 2019 and 31 st March 2020. Excluding March, these signs of new particle formation were observed in 30 days. Nucleation was observed in-situ at SMEAR I during events associated with H2SO4 concentration exceeding ca. 1×10 6 molec. cm -3 . These cases were identified based on the appearance of ~1.5-2 nm ion clusters. Other events were observed as appearance of few nm particles which gradually grew over time. Nucleation at SMEAR I was shown to proceed primarily via 415 negative ion-induced sulphuric acid -ammonia (-water) channel. Closer to SO2 emission sources where H2SO4 concentrations are likely remarkably higher, nucleation can proceed also via neutral channel and could, theoretically, involve compounds other than H2SO4, NH3 and water.
Larger particles, few nm and up, observed at SMEAR I, were formed out of the immediate vicinity of the site and grown during 420 the airmass advection. Secondary aerosol formation from Kola emissions together with primary particle emissions impact the aerosol size distribution, clearly increasing the concentrations of particles in each size class, and therefore unavoidably also CCN concentrations. Detailed assessment of the contribution of Kola SO2 emissions to local and regional CCN concentrations and upscaling our results to cover the whole (sub)-arctic Eurasia with vastly polluting industrial cities such as Norilsk, require more measurements -complemented by CCN or cloud residual measurements -ideally in more than only one location 425 (SMEAR I) around the Kola peninsula, as well as regional chemical transport and aerosol dynamic modeling.
Data availability. Mass spectrometer and air ion spectrometer data related to this article are available upon request to the corresponding author. Rest of the data are available for download from https://avaa.tdata.fi/web/smart/smear. Competing interests. The authors declare that they have no conflict of interest.