Shrinkage of Newly Formed Particles in an Urban Environment

New particle formation (NPF) significantly influences the number concentrations and size distributions of an atmospheric aerosol and is often followed by the rapid growth of the newly formed particles. Recently a reversal process leading to particle shrinkage was reported and was found to occur under specific atmospheric conditions, which are unfavourable for the growth of nucleated particles. In this study we present an analysis of particle shrinkage events following prior NPF at an urban background station in Prague, Czech Republic. The study is based on two-year measurements of particle number size distributions using Scanning Mobility Particle Size (SMPS). A total of 22 shrinkage events were identified and the vast majority of the events was observed in the spring and summer seasons. The determined shrinkage rates span the range from –2.5 to –12.5 nm/h. During the most intensive events, the grown particles shrank back to the smallest measurable size of 10 nm. The particle shrinkage was attributed to the evaporation of previously condensed volatile and semi-volatile species from the particulate phase to the gas phase. When analysing the dependence on meteorological conditions, the particle shrinkage was found to occur under peak global radiance intensity, which was gradually decreasing, or under a sharp drop of global radiance intensity. The shrinkage events were also related to high ambient temperature and low relative humidity. The occurrence of atmospheric mixing was found to promote particle shrinkage very effectively, because the decrease of vapour concentration due to changing atmospheric conditions is more rapid than the decrease of vapour concentration caused by lower photochemical activity due to a drop in solar radiation intensity.


INTRODUCTION
Atmospheric aerosol particles affect the climate owing to their major role in atmospheric chemistry and interaction with solar radiation.Their direct impact on the climate is caused by scattering and absorbing solar radiation and their indirect impact by acting as cloud condensation nuclei (CCN) and thus determining the optical properties of clouds and cloud lifetime.The effect of aerosols on the climate is currently the largest source of uncertainty in estimating and interpreting the Earth's changing climate (Merikanto et al., 2009;IPCC, 2013).New particle formation (NPF), which is frequently observed in the atmosphere (Kulmala et al., 2004), is an important source of ambient aerosol.The climatic effects of aerosol particles formed during NPF events are governed by their subsequent growth, which determines the fraction of nucleated particles growing to sizes where they can interact effectively with solar radiation or participate in cloud formation processes (Dusek et al., 2006;Andreae and Rosenfeld, 2008;Kulmala and Kerminen, 2008).The atmospheric conditions, which are favourable for NPF events and the subsequent growth of newly nucleated particles, include low concentrations of pre-existing particles (thus leading to negligible scavenging of nucleated particles by pre-existing ones), an excess of condensable low-volatility vapours, and high values of global radiation (Birmili and Wiedensohler, 2007;Hamed et al., 2007).The contribution of different chemical species to particle growth depends on the type of atmospheric environment (Boy et al., 2005;Stolzenburg et al., 2005;Yue et al., 2010) and particle size -the non-volatile or low-volatile compounds are responsible for the initial growth of particles in the smallest size ranges and the condensation of more volatile components increases with increasing particle size (Zhang et al., 2004).Sulphuric acid (H 2 SO 4 ) has been found to be the key component in atmospheric nucleation (Riipinen et al., 2007;Kuang et al., 2008;Sipilä et al., 2010;Brus et al., 2011), however its contribution to subsequent particle growth is only minor (Smith et al., 2008;Nieminen et al., 2010;Riipinen et al., 2011).Low-volatile organic species were found to be responsible for up to 90% of the observed growth of newly formed particles (Makëla et al., 2001;Smith et al., 2008).
Recently, several studies have reported that under specific atmospheric conditions, which are unfavourable for the growth of freshly nucleated particles, a reversal process leading to particle shrinkage can occur.This phenomenon has been observed in Brazil (Backman et al., 2012), Taiwan (Young et al., 2013), Hong Kong (Yao et al., 2010) and Spain (Cusack et al., 2013).This behaviour is attributed to the evaporation of condensed species from the particles due to a decrease of the precursor vapour concentrations caused by changes in atmospheric conditions and the evaporation of ammonium nitrate (NH 4 NO 3 ) due to a shift in the nitric acid (HNO 3 ) -ammonia (NH 3 ) -ammonium nitrate (NH 4 NO 3 ) equilibrium.The reported shrinkage rates span the range from -5.1 to -11 nm/h (Yao et al., 2010;Cusack et al., 2013;Young et al., 2013).
The aim of this study is the qualitative and quantitative analysis of the particle shrinkage following NPF events at an urban background station in Prague -Suchdol, Czech Republic.More specifically, we will address the following issues: a) to study the seasonal variation of particle shrinkage events b) to investigate the connection of particle shrinkage to ambient conditions and to find the processes favouring this phenomenon.

Measurement Site
The measurements were performed between May 2012 and April 2014 at the urban background station Prague -Suchdol (50.13°N, 14.38°E).The site is located in a northwest suburb approximately 5 km from the city centre (see Fig. 1).The Prague region is densely populated (2357.6 inhabitants per km 2 (Czech Statistical Office, 2013)) and industrialized.The station is located on the edge of a plateau above Prague with an altitude 277 m above sea level on the campus of the Institute of Chemical Process Fundamentals (ICPF).The site is exposed to emissions from local sources including residential houses (~30 m from the site), traffic from the nearest road -distance approximately 250 m (daily traffic ~15,000 cars) and the Václav Havel Airport Prague, located 9 km to the southwest.

Sampling and Instrumentation
The concentrations of atmospheric aerosol were measured using a Scanning Mobility Particle Sizer (SMPS), determining the particle number size distribution (PNSD) in the submicron size range, and providing information on the raw mobility size distributions and on the diagnostic parameters of the instrument as well.The diagnostic file consists of the values of temperature, relative humidity, absolute pressure, and flows of both aerosol and sheath air.The SMPS 3034 (TSI, USA) has been used in an upgraded version, so it meets the quality standards for ultrafine aerosol measurement networks described in Wiedensohler et al. (2012).The quality assurance is based on temperature and relative humidity in the SMPS system, and on the aerosol and sheath flow rates.From the raw and diagnostic files, the PNSD were computed using BY-HESS LabView shareware application software.The BY-HESS inversion routine is based on multiple charge inversion (Pfeifer et al., 2014) and contains also corrections on the Condensation Particle Counter (CPC) counting efficiencies (inbuilt CPC 3010, TSI, USA was used), particle losses in the connecting tubes and inside the SMPS, based on an equivalent pipe length (Wiedensohler et al., 2012), and Differential Mobility Analyser (DMA) transfer function (TSI 3034 DMA with 32.7 cm active length).At the given flows (1 L/m of aerosol flow and 4 L/m of sheath flow), the PNSD in the size range between 6.5 and 540 nm was measured every 5 minutes; due to the measurement uncertainties at both edges of the distribution, only PNSD between 10 and 500 nm were considered.
In addition to the SMPS data, basic meteorological data (temperature, global radiation, relative humidity, wind speed and wind direction) from the meteorological station located in the same measurement container of automated immission monitoring (AIM) on the campus of ICPF were available.Also the concentrations of gaseous pollutants, namely sulphur dioxide (SO 2 ), nitric monoxide (NO), nitric dioxide (NO 2 ), ozone (O 3 ), carbon monoxide (CO), toluene, benzene, PM 10 and PM 2.5 measured at AIM station by the Czech Hydrometeorological Institute (CHMI), were added to the analysis.

Determination of Particle Shrinkage Rate, Condensational Sink and H 2 SO 4 Proxy Concentration
Condensational sink (CS) is a measure of the loss rate of condensable vapours onto the pre-existing aerosol.An elevated concentration of pre-existing aerosol particles can supress atmospheric nucleation due to the scavenging of freshly nucleated particles and condensable vapours.CS was calculated using size distributions measured with the SMPS as follows (Kulmala et al., 2001): where D is the diffusion coefficient of the vapour considered, D p,i is the diameter of the particles in the size bin i and N i is the number concentration of these particles.The term βm presents the transition regime correction factor for mass transfer and was calculated according to the equation given in Fuchs and Sutugin (1970) with a value of 1 for the mass accommodation coefficient.When calculating CS, the condensable species were assumed to have molecular properties of H 2 SO 4 and thus negligible saturation vapour pressure corresponding to the assumption that H 2 SO 4 behaves as a non-volatile vapour.
H 2 SO 4 is the key component of atmospheric NPF (Sipilä et al., 2010;Brus et al., 2011).To investigate the connection of particle shrinkage to ambient sulphuric acid, the proxy of H 2 SO 4 concentration presented in Petäjä et al. (2009)  The shrinkage rate was determined using the mode fitting method presented in Dal Maso et al. (2005).The number size distributions obtained from the SMPS measurement were fitted with log-normal distribution and the geometric mean diameter (GMD) of the nucleation mode was determined.The particle shrinkage rates were then obtained from the change of the modal GMD as a function of time:

Particle Shrinkage Events Characteristics and Frequency
The results presented here are based on measurements conducted between May 2012 and April 2014 and are thus representative for all seasons in the Central Europe region.During this period, there were 183 days with new particle formation (NPF) events and 22 particle shrinkage events following the NPF.The classification of NPF events was conducted according to the criteria proposed in Dal Maso et al. (2005).NPF events are characterized by the appearance of freshly formed particles in the nucleation mode and the continuous growth of nucleated particles for at least several hours.Particle shrinkage presents the continuous decrease of the geometric mean diameter of the grown particles following a previous NPF event for at least 2 hours.The identification of particle shrinkage was performed visually from the contour plots of the days with an NPF event.The identification was based on several distinct features of the temporal changes of the particle size distributions and particle sizes.
Table 1 summarizes the frequencies of the NPF events and particle shrinkage identified during the measurement period.The NPF event frequency exhibits strong seasonal variation, which influences also the occurrence of particle shrinkage events.NPF events were most abundant during spring (38% from the total number of event days) and least in winter (14%).The frequency of NPF in summer was 32% and in autumn 16%.The corresponding particle shrinkage frequencies are 50% in spring, 41% in summer and 9% in autumn.No particle shrinkage was observed in the winter season.However, the occurrence of particle shrinkage is strongly influenced by the seasonal variation of NPF events.

Observed Particle Shrinkage Rates
The shrinkage rates were calculated using the Eq. ( 3).The determined shrinkage rates span the range from -2.5 to -12.5 nm/h.In the most intensive cases, the grown particles shrank back to the smallest measurable particle size of 10 nm.Both the minimum and the maximum shrinkage rates show slightly higher span of values from the ones reported in the earlier studies.The previously reported shrinkage rates were between -8.6 and -11 nm/h in Hong Kong (Yao et al., 2010), -5.1 and -7.6 nm/h in Taiwan (Young et al., 2013), -11 nm/h in Montseny, Spain (Cusack et al., 2013) and -5.2 nm/h in Sao Paulo, Brazil (Backman et al., 2012).Based on the analysis of 7 years of particle size distribution measurements, the growth rates of newly formed particles during NPF events were found to be dependent on particle size and have a clear seasonal pattern with the highest values during summer (Yli-Juuti et al., 2011).In our study, the seasonal variation of particle shrinkage was strongly influenced by the seasonal variation of NPF events.Although our results indicate that the shrinkage rates were highest during summer, it is not feasible to quantify this tendency owing to the limited number of observations (in particular for autumn and with no data available for the winter season).

Particle Concentrations and Size Distributions
The temporal change of the particle number size distribution of NPF directly followed by particle shrinkage has several characteristic features.At the beginning of NPF, a distinctly new particle mode in the sub-25nm size range is observed, prevails for at least several hours and the newly formed particles grow to larger sizes by the condensation of various vapours.Under specific atmospheric conditions, which do not favour the further growth of the nucleated particles, a reversal process leading to particle shrinkage can occur.When the geometric mean diameter of the nucleated particles mode reaches its maximum, the size of the grown particles starts to decrease.In the most intensive cases, the particles shrank back to the smallest measurable sizes.These events, observed in Taiwan (Young et al., 2013) and Spain (Cusack et al., 2013), were characterized by an "arch-like" shape in the size distribution contour plots.An example of an intensive event, when the grown particles shrank back to the smallest detectable size of 10 nm is presented in Figs.2(a)-2(c).
Intensive NPF events are characterised by a sharp increase of particle number concentration in the nucleation and Aitken modes.In our study, most of the particle shrinkage events following a previous NPF were found to exhibit a decrease of the particle number concentration in N 10-100 with a simultaneous decrease of particle size.However, in several cases an increase of the concentration of the particles with sizes < 30 nm with a decreasing geometric mean diameter was recorded during shrinkage.Yao et al. (2010) attributes this behaviour to a lower scavenging rate of the < 30 nm particles than the shrinkage rate of the particles with sizes > 30 nm.Young et al. (2013) who also recorded this phenomenon concluded that this behaviour was observed because of the arrival of new air masses (characterized by low condensational sink and SO 2 -enrichment) due to atmospheric advection and mixing near the end of the particle shrinkage period, which might have caused a new NPF event.When comparing the particle number concentration at the end of the shrinkage period to the concentrations recorded at the onset of the NPF event, the vast majority of newly nucleated and grown particles was found to evaporate during intensive shrinkage events.In Figs.3(a)-3(c) the particle shrinkage recorded on 1 October 2013 is presented, where it is evident that the particle number concentrations at the beginning of the NPF and at the end of particle shrinkage are almost identical.

Particle Shrinkage in Relation to Air Pollutants Concentration, H 2 SO 4 Proxy Concentration and Condensational Sink
The 10-minute averaged PM 10 , PM 2.5 , CO, NO x , O 3 , SO 2 , benzene and toluene concentrations obtained from the automatic immission monitoring station (AIM) during particle shrinkage event days were investigated.The O 3 concentration had a distinct daily pattern with a rapid increase of [O 3 ] after sunrise followed by a decrease to a minimum concentration after sunset.In general, the [O 3 ] was often found to decrease during particle shrinkage, with the lowest concentration at the end of the shrinkage event.This decrease of [O 3 ] was accompanied by lower global radiation causing decreased photochemical activity.No clear pattern in the PM 10 , PM 2.5 , NO x , CO, benzene and toluene concentrations variability during particle shrinkage was found -in some cases these concentrations were slightly increasing or decreasing, rapid hourly variations were observed several times and in few cases the concentrations were almost stagnant.The variation in concentrations can be attributed to the possible dilution of the air masses or arrival of polluted air during particle shrinkage; however, our observations indicate that the change in PM 10 , PM 2.5 , CO, NO x , benzene and toluene concentrations does not affect particle shrinkage.SO 2 is the precursor of gas phase H 2 SO 4 , which plays a key role in atmospheric nucleation and is also one of the major condensing vapours during new particle growth.Previous studies from Budapest (Salma et al., 2011) and Taiwan (Young et al., 2013) report that SO 2 concentration is not the limiting factor for NPF in urban environments, because SO 2 is always present in excess of the concentration needed for new particle formation.In this study, the SO 2 concentration was found to be rather stagnant during shrinkage events as compared to the observed concentration variability of other air pollutants (PM 10 , PM 2.5 , CO, NO x , O 3 , benzene and toluene) except for several events, which were accompanied by the arrival of polluted air masses with a subsequent distinct increase of [SO 2 ].On the contrary, the proxy [H 2 SO 4 ] determined using Eq. ( 2) was found to have a distinct daily pattern since gas phase H 2 SO 4 formation involves photochemical processes.The NPF burst was almost always preceded by an increase of the proxy [H 2 SO 4 ].On the other hand, [H 2 SO 4 ] was frequently observed to decrease during particle shrinkage, which is mainly due to lower global radiation, which was recorded during most of the particle shrinkage events.This trend is similar to the observed [O 3 ] variation during shrinkage.
Condensational sink (CS) describes the loss of condensable vapours from gas phase onto the pre-existing particles and is proportional to the surface area of the aerosol particles.High CS might inhibit new particle formation and the growth of newly nucleated particles because a large fraction of the vapour could condense on larger particles.In our study, the beginning of NPF was often observed to coincide with the decrease of the condensational sink.CS then gradually increased over the course of the NPF and subsequent growth of the newly formed particles.On the contrary, particle shrinkage was often found to commence with decreasing CS, which is suggestive of air mixing leading to a lower concentration of volatile and semi-volatile vapours.A decrease of the concentration of these species might promote a subsequent evaporation of the volatile and semi-volatile vapours from the grown particles.This behaviour was also observed in previous studies of particle shrinkage (Cusack et al., 2013;Young et al., 2013).

Meteorological Conditions Favouring Particle Shrinkage
Particle shrinkage following a previous NPF occurs under specific atmospheric conditions, which do not favour further particle growth.This phenomenon has been observed in both urban (Yao et al., 2010;Backman et al., 2012;Young et al., 2013) and regional background (Cusack et al., 2013) stations.All particle shrinkage observations have so far been made in subtropical (Yao et al., 2010;Backman et al., 2012;Young et al., 2013) or a hot Mediterranean region (Cusack et al., 2013) and the measurements in a mild climatic region have not been reported yet.
Particle shrinkage reported in studies, which provide precise meteorological information during shrinkage events (Backman et al., 2012, Cusack et al., 2013;Young et al., 2013), have several common features.In general, particle shrinkage events appeared to start under intensive solar radiation, high ambient temperature and atmospheric dilution.A consistent feature of our observations is the global radiance (GR) dynamics.The vast majority of particle shrinkage events observed in our study started either at peak global radiation intensity, decreasing gradually over the course of shrinkage or at peak GR followed by a sharp drop in intensity.Similar variability has been reported from Brazil (Backman et al., 2012) and Spain (Cusack et al., 2013).Lowering global radiance intensity is also likely the reason for the decreasing concentrations of [O 3 ] and the proxy [H 2 SO 4 ] recorded during most of the shrinkage events.Intensive solar radiance may initiate particle shrinkage due to the evaporation of condensed species from particles (Cusack et al., 2013), whereas a significant drop of GR intensity may interrupt a previously initiated growth process due to lower photochemical activity.The observed values of wind speed were quite variable -some shrinkage events occurred under almost stagnant air with wind speeds as low as 0.5 m/s while in other cases the wind speed reached up to 4 m/s.Previous observations have reported particle shrinkage taking place under relatively high ambient temperatures of ~18°C and above.In our study, the temperatures during shrinkage events ranged from 12°C to 27°C, however most events occurred at temperatures from 20 to 24°C.This temperature dependency also explains the observed seasonal variability -only 1 event was observed in autumn and none in winter.Evaporation of both organics (Denjean et al., 2015) and NH 4 NO 3 (Dassios and Pandis, 1999) was observed to increase with increasing temperature.Our observations suggest that the evaporation of organics and NH 4 NO 3 might be significantly hindered under ~12°C and temperature is very likely one of the limiting factors for shrinkage events seasonal variability.Relative humidity was measured in a relatively wide range from 30 to 77% and was found to both increase and decrease during particle reverse growth.Growing RH may indicate air mixing, while lowering RH suggests an increasing temperature.
Several studies have recently focused on the chemical composition of particles grown after NPF.Bzdek et al. (2012) reports that NO 3 -, NH 4 + and organics contribute to 54-59% of new particle growth, whereas sulphates are responsible for the remaining 41-46% of the observed growth.Smith et al. (2008) showed that the major compound of the newly formed particles are organics (~84%) and include organic acids (formate, acetate), hydroxyl organic acids and nitrogen-containing organic compounds.The vast majority of these organics are volatile or semi-volatile and can evaporate from the particles when the phase equilibrium favours the gas phase.Also the decrease of relative humidity can cause morphological changes resulting in particle size change.An example of such a behaviour is the displacement reaction of alkylamines with (NH 4 ) 2 SO 4 resulting in formation of alkylaminium sulphates.This reaction causes the transition from the crystalline to the amorphous phase (Qui and Zhang, 2012), which enhances water uptake and consequently particle growth with increasing RH.The transition from the crystalline to the amorphous phase can be reversible with a lowering of the relative humidity.Yao et al. (2010), who report particle shrinkage in Hong Kong, attribute this behaviour to the evaporation of organics and NH 4 NO 3 from the particles since these compounds can repartition between the gas and particle phases.Backman et al. (2012) explained the observed particle shrinkage in Sao Paulo, Brazil by the evaporation of previously condensed species from particles when the precursor vapour concentration was decreased (likely due to dilution of air masses) and by the evaporation of NH 4 NO 3 .Although it was not possible to identify the chemical composition of the grown and shrinking particles from the available data, these processes are very likely the cause of the particle shrinkage observed in our study.
Organic species and NH 4 NO 3 have different timescale of evaporation.The mass transfer of atmospheric submicron NH 4 NO 3 particles between the aerosol and gas phase was estimated to proceed in the time range from a few seconds to 15 minutes depending on the ambient temperature (Dassios and Pandis, 1999).The evaporation of organics was observed to proceed much slower.Grieshop et al. (2007) studied the dilution-induced evaporation of laboratory produced secondary organic aerosol (SOA) formed by α-pinene ozonolysis in a chambrer study.The system was observed to reach evaporation equlibrium within two and half hour.The obtained relatively slow evaporation rate was attributed to the formation of low volatility oligomers.Another study (Vaden et al., 2011) studied evaporation kinetics of laboratory SOA produced by reaction of α-pinene with O 3 and atmospheric SOA.They concluded that the SOA evaporation has two separates regimes: a fast regime, which takes ~100 minutes and second, very slow regime, which lasts more than a day.The SOA evaporation was also observed to be almost size independent and did not follow the evaporation of a liquid droplet indicating that the SOA particles were coated.However, the atmospheric particles contain wide range of organic and inorganic compounds and the chemical composition governs aerosol properties i.e., hygroscopicity or volatility.Several recent studies (Yli-Jutti et al., 2013;Hakkinen et al., 2014) report that the volatility of organic acids is influenced by the presence of inorganic salts in aerosol particles.Deeper investigation is therefore needed to gain a better insight into the particle phase chemistry and its influence on aerosol particles partitioning and aerosol mass loadings.Backman et al. (2012) speculate that the decrease of vapour concentration due to changing atmospheric conditions is more rapid than the decrease of vapour concentration caused by lower photochemical activity.Intensive atmospheric mixing would thus promote evaporation from particles more effectively when compared to lower global radiation and would also enhance shrinkage rates.In the following segment, we present 2 particle shrinkage events, one which occurred under relatively stagnant atmospheric conditions and the other under atmospheric mixing.

Case study of Particle Shrinkage Events
On 13 July 2013, the start of morning NPF occurred at ~8:30 and the initial GMD of the particles was 11 nm (Figs.4(a)-4(f)).The NPF burst was followed by particle growth to 21 nm from 9:00 to 11:00 with a growth rate of 4.5 nm/h.The peak hourly concentration of particles in a size range of 10-30 nm reached 2.2 × 10 4 at 12:30.Between 11:00 and 15:00, the GMD decreased to 11 nm with a shrinkage rate of -2.5 nm/h and the N 10-30 decreased between 12:30 and 20:00 from 2.2 × 10 4 to 2.2 × 10 2 .It is notable that the shrinkage of GMD occurred at ~11:00 when there was a sharp drop of global radiance intensity from 880 to 260 W/m 2 likely due to a short period with cloud cover.This sharp decrease of GR intensity may have hindered the photochemical processes, thus lowering the condensable vapour concentrations and causing the reversal in the growth process and evaporation of particles.During the period of decreasing GMD between 11:00 and 15:00, there was a low wind speed of ~1.5 m/s, T increased from 20.6 to 22.7°C and RH decreased from 52 to 42%.These conditions further enhanced the particle evaporation.There were only slight changes in the air pollutants concentrations, implying no significant changes in air masses.The meteorological conditions were therefore very likely the driving force of the observed particle shrinkage event.
On 7 July 2013 the nucleation particle burst started at 9:00 and was followed by intensive particle shrinkage (Figs.5(a)-(f)).After the start of NPF, the GMD grew rapidly from 13 nm to 50 nm with a growth rate of 12 nm/h.The N 10-30 decreased sharply at the beginning of NPF as the particles were growing to sizes over 30 nm and increased again significantly when the particles started to evaporate and lower in size.The GMD decreased during particle shrinkage from 50 nm to 25 nm between 12:00 and 14:00 with a shrinkage rate of -12.5 nm/h and N 10-30 reached its minimum value of 2.7 × 10 3 at 18:00.Similarly to the previously described event, the particle shrinkage occurred during a substantial drop of GR intensity from 844 to 287 W/m 2 suggesting a period of a cloudy sky between 12:00 The dynamics of T and RH were similar to the previously described event -T increased from 23 to 24.9°C, RH decreased from 46 to 39% and the wind speed varied between 0.7 and 2 m/s.There were reductions in the air pollutants concentrations during the shrinkage period, which indicate atmospheric mixing leading to a decrease of the condensable vapour concentrations, which would further enhance particle evaporation.
In general, particle shrinkage seems to appear when a significant shift in meteorological conditions interrupts the process of particle growth and in addition the reversal process leading to particle shrinkage is favoured by the ambient conditions.Furthermore, our results indicate that particle evaporation is very effectively enhanced under the conditions of atmospheric mixing.When comparing the 2 previously described shrinkage events, which occurred under similar meteorological conditions, we calculated the dew point temperature, which is a good indicator of the air mass origin (Kleinman et al., 2012).In both cases the dew point changed only slightly during particle shrinkage indicating no significant air mass changes.When comparing the concentrations of air pollutants, there were significant changes in concentrations during particle shrinkage in the second case study.We conclude that in the second case study the particle shrinkage occurred under vertical mixing which very effectively enhanced particle evaporation and caused significantly higher shrinkage rates as compared to first case study.This observation is in agreement with the previously mentioned idea outlined in Backman et al. (2012) that the decrease of vapour concentration due to changing atmospheric conditions is more rapid than the decrease of vapour concentration caused by lower photochemical activity.

CONCLUSIONS
In this paper we have presented a description of particle shrinkage following previous new particle formation events.The study is based on an analysis of a two-year long dataset of SMPS particle size distributions obtained from an urban background station in Prague, Czech Republic.A total of 22 shrinkage events following a previous NPF were identified between 1 May 2012 and 30 April 2014.They exhibit very strong seasonal variability with the highest occurrence during spring (50% of all identified events) and summer (41% of all identified events), however the occurrence was strongly influenced by the seasonal variability of NPF events.The observed shrinkage rates ranged from -2.5 to -12.5 nm/h.In the most intensive cases, the grown particles shrank back to the smallest measurable size of 10 nm and the number concentration dropped to values comparable to concentrations observed at the beginning of the NPF burst.
When analysing the dependence on the meteorological conditions, the observed shrinkage events were almost always preceded either by a sharp drop of global radiance intensity or occurred under peak global radiance intensity, which was gradually decreasing over the course of the particle shrinkage.The shrinkage events were also found to occur under high ambient temperature and low relative humidity.Also the occurrence of atmospheric mixing was found to promote particle shrinkage very effectively.Although the observed events of particle shrinkage were few, from this limited data we can conclude that a large fraction of the originally condensed volatile and semivolatile species evaporated from the particle phase during the shrinkage events.Although the identification of the potential evaporating chemical species was not possible from the data available, we speculate that the evaporating compounds include NH 4 + , NO 3 -and organics.Our results are indicative that the growth and shrinkage of newly formed particles are very sensitive to changing atmospheric and meteorological conditions and both the concentration and geometrical mean diameter of these particles may evolve significantly in a short time period.Deeper investigation is therefore needed to gain a better understanding of the processes governing new particle growth and shrinkage in the atmosphere.

Fig. 1 .
Fig. 1.The measurement site position.The location of the station is depicted as a black square.

Fig. 4 .
Fig. 4. The evolution of (a) particle size distribution, (b) total number concentrations, (c) modes location, (d) meteorological parameters and (e) air pollutants concentrations, (f) condensational sink and [H 2 SO 4 ] during the shrinkage event observed on 13 July 2013, where meteorological conditions are assumed to be the driving force of particle shrinkage.

Fig. 5 .
Fig. 5.The evolution of (a) particle size distribution, (b) total number concentrations, (c) modes location, (d) meteorological parameters and (e) air pollutants concentrations, (f) condensational sink and [H 2 SO 4 ] during the shrinkage event observed on 7 July 2013, where particle shrinkage is assumed to be initiated by a meteorological condition change and further promoted by atmospheric mixing.
Since this calculation of H 2 SO 4 proxy concentration is based on measurements in a boreal forest in Finland (Petäjä et al., 2009), the absolute values of the proxy [H 2 SO 4 ] should be considered as indicatory [H 2 SO 4 ] only.