The aerosol particle size distribution measurements are carried out using Wide Range Aerosol Spectrometer (WRAS) located at CSIR-National Physical Laboratory (CSIR-NPL), Delhi (28.64° N, 77.17° E) during December 2011 to January 2013. The system comprises of a Scanning Mobility Particle (SMPS + C) and an Optical Particle Counter (OPC). The GRIMM SMPS + C system consists of a high resolution condensation particle counter (CPC) attached to the GRIMM “Vienna Type” M-DMA (Monodisperse Differential Mobility Analyser). DMA classifies the particles according to its electrical mobility which is then measured using a CPC. Both components together measure in the ultrafine size range from 5 to 350 nm. OPC which work on the light scattering technology for particle counting gives the particle size distribution in 31 channels ranging from 250 nm to 32 µm (Grimm and Eatough, 2009). Semiconductor laser serves as the light source for OPC. Thus integrating both, SMPS + C and OPC data using a GRIMM software, the WRAS system gives the particle size distribution in 72 channel ranging from 5.5 nm to 32 µm. There is an overlapping region from 250 to 350 nm due to the integration of SMPS and OPC, here we take the average number concentration for the analysis. A Nafion dryer is used inside the sampling probe in order to avoid the loss of volatile components and dehumidify the sample during the measurement period. The GRIMM-WRAS data is supposed to have an uncertainty in total flow rate up to 5% as per manufacturer specification. SMPS is reported to have an uncertainty of less than 10% for particles in the size range of 20–200 nm, while for particles above and below this range uncertainty increases (Wiedensohler et al., 2012). Uncertainty of OPC is estimated to be about 9.9% (Grimm and Eatough, 2009). The detailed description and principle of the instrument are discussed elsewhere (Grimm and Eatough, 2009). The system also has an integrated meteorological sensor which provides temperature, relative humidity and wind vector data.

The incoming global (direct + diffused) solar radiation were measured on a horizontal surface in the UV (280–400 nm) range using the Kipp & Zonen CUV-4 radiometer respectively. The radiation flux measurements have an estimated experimental error of 3% in the shortwave and <10% in the UV range. More details about the radiation flux measurements are reported elsewhere (Bano et al., 2013). Ambient SO₂ data is obtained using the self-calibrated Horiba APSA-370 SO₂ analyzer which uses Ultraviolet Fluorescence (UVF) method as the operating principle for measurements. The detailed working principle is reported elsewhere (Suneja et al., 2019).

Continuous sampling was done from December 2011 to January 2013, however, some data gaps exist on a few occasions whenever technical errors were encountered during the observation. A total of 222 days’ data has been studied and analyzed seasonally, categorized as, winter (DJF), summer (MAMJ), monsoon (JAS) and post monsoon (ON). Considering the regional climatic conditions June has been grouped into summer as south west monsoon hits Delhi by the end of June or early July. The instrument provides aerosol particle size distribution at a span of 5 min interval. Percentage data availability for winter, summer, monsoon and post-monsoon are ∼36%, ∼49%, ∼35 and ∼63% of the total days respectively. The solar flux data was obtained in Wm⁻² every 2 min, during the observation period and averaged hourly and seasonally.

During the preliminary stages of the analyses the data has been flagged and removed if any discrepancy in the instrument or measurement is noted. Out of the entire experimental period, a total of 222 days of data has been found proper and the same has been used for further analysis. NPF events have been identified from the collected data sets based on the following criteria, as also suggested by (Dal Maso et al., 2005): 1) A distinctly new mode of particles must appear in the size distribution; 2) the mode must start in the nucleation mode; 3) the mode must prevail over a time span of few hours; 4) the new mode must show signs of growth. Further, the NPF events have been identified according to the following scrutiny techniques. 1) The visual identification of the contour plots of particle size distribution over the 24 h period. 2) The variation of mode diameter across 24 h period. 3) Individual mode analysis of the size distribution at each 5 min interval.

In this study, we have classified the modes into three categories, Nucleation mode (5–30 nm), Aitken mode (30–100 nm) and accumulation mode (100–1,000 nm) (Ueda et al., 2016). The data has been classified into event days, non-event days, and unidentified event days according to a number of criteria discussed in various peer reviewed papers (Dal Maso et al., 2005; Kanawade et al., 2014; Kamra et al., 2015; Tröstl et al., 2016). Based on this, three categories of events (classifications) emerge with their distinct features. First, 1) NPF Event days, which are those days in which there is a sudden outburst of number concentration in the size range between 5 and 30 nm for a prolonged time period of 2–3 h (Kanawade et al., 2014). Another category, 2) Non Event days are those for which no nucleation and particle growth is observed in these days. The contour plots show no particular trend or variations in the number concentration. Finally, third category 3) Unidentified events days are the ones where we are unable to deduce the features of the number concentration from the contour plot and other scrutiny methods due to high level of noises in the data or the gaps in the data. This can be due to the background pollution and the haze conditions prevailing in the atmosphere.

The rate of change of diameter of the particle (dD_p/dt) is termed as growth rate (GR), where D_p is the diameter of the particle and t is time (Tröstl et al., 2016). The growth rate can be found out from the linear regression analysis of the mode diameter (D_p) with time (Tröstl et al., 2016; Salimi et al., 2017). The slope of the regression gives the growth rate. The linear regression equation is given by D p   =   a 0 + a 1 t , (1) where a₀ is the intersect at the y axis and a₁ is the slope of the line pointing to the growth rate.

Condensation sink (CS) is the scavenging speed of the gaseous molecule due to the condensation onto particles. It mainly depends on the particle size and the condensation (Ueda et al., 2016), and the equation is given by Kulmala et al. (2001) as C S =     2 π D v ∑ i D P , i β m , i N i , (2) D v is the diffusion coefficient of H₂SO₄ and is assigned as 0.104 cm² s⁻¹ (Kanawade et al., 2014). D P , i and N i is the center particle diameter and number concentration of the size class i respectively. β m , i denotes the size dependent transition correction factor given by Fuchs and Sutugin (1971) and is expressed as β m , i =   1 + K n , i 1 + 0.377 K n , i + 4 3 α − 1 K n , i + 4 3 α − 1 K n , i 2 , (3) where K n , i is the Knudsen number for the size class i and the mass accommodation coefficient ( α − 1 ) is assumed as unity.

Figure 1 shows the box and whiskers plot of monthly averaged aerosol particle number concentration for different modes. The maximum aerosol number concentration for an individual day is recorded in December with a value of 16.67 × 10³ cm⁻³. The mode wise analysis of number concentration shows that Aitken mode (30–100 nm) dominates in the study region throughout the year with a maximum mean concentration during the post monsoon season (9.6 × 10³ cm⁻³). Accumulation and nucleation mode particles are minimal compared to the Aitken mode particles during the entire study period at the sampling site. However, seasonal patterns showed variations in both the modes. Nucleation mode particles are high in number than accumulation mode particles during summer and monsoon seasons, whereas accumulation mode particles dominate during the other two seasons in comparison with nucleation mode. High mean concentration of nucleation mode particles (4.67 × 10³ cm⁻³) are noticed during July, when the minimum mean concentration of accumulation mode particles (1.79 × 10³ cm⁻³) are registered. Wet deposition of accumulation mode particles could be a plausible reason for the decreased concentration of accumulation mode particles and increased concentration of nucleation mode particles in the monsoon season. Heat map analysis of aerosol particle size distribution revealed a considerable shift in the size of particle as it moves from winter to pre monsoon.

The seasonal mean number concentrations of aerosol particles in different modes are tabulated in Table 1. Here, N_5-30 indicates number concentration in nucleation mode, N_30-100 in Aitken mode and N_100-1000 in accumulation mode. Highest average total number concentration of 15.80 ± 4.26 × 10³ cm⁻³ is noted during the post monsoon season, this can be attributed to the biomass burning events during the post monsoon seasons from nearby agrarian states like Punjab and Haryana. Mean total number concentration for the entire experimental period is found to be 14.08 ± 7.60 × 10³ cm⁻³. Fractional share of each mode to the total number concentration is shown in Figure 2. The highest concentration of nucleation mode particles were observed during monsoon season with the mean number concentration reaching up to 4.06 ± 0.53 × 10³ cm⁻³ with a fractional share of 0.315. The maximum number concentration of accumulation mode particle was found to be 5.90 ± 3.17 × 10³ cm⁻³ in post monsoon season with a fractional share of 0.36. A considerable increase in the accumulation mode particles is seen during post monsoon and winter season as the major contributor for accumulation mode particles come from biomass burning activity (Chen et al., 2017; Ghosh et al., 2019).

Figure 3 depicts the diurnal variation in mean number concentration for different seasons. Aitken mode is the primary peak and follows a bimodal pattern for all the seasons. Starting from a gradual increase from early morning Aitken mode reaches a peak value at about 09:00–10:00 AM and then decreases. Aitken mode again shows an increase in concentration during the evening hours which later dies out at late night. This high particle concentration during the morning and evening hours is associated with the high rush of traffic during these hours. Among the two Aitken mode peaks observed, evening peak is seen to be prominent with higher concentration of Aitken mode particles, which may be due to the boundary layer dynamics associated with the diurnal meteorology. The decrease in the concentration of Aitken mode particle after the morning peak hours can be attributed to the mixing of air due to increased convection as the temperature increases and the increased concentration at the evening hours may be attributed to the stable layer formation during the evening, trapping the pollutants inside the boundary layer (Mönkkönen et al., 2005). The pattern shows a similar trend in all seasons except for monsoon season, where the peak concentration is observed in the midday. The morning peak Aitken mode is succeeded by a small increase in nucleation mode particles during midday on few days and some of it later grows to nucleation events which are discussed in the following session.

Figure 4 shows the aerosol volume size distribution for different seasons. Bimodal pattern is observed for volume size distribution, characterizing fine (<1 µm) and coarse mode (>1 µm). For volume size distribution fine mode dominates and peaks at 0.38 and 0.27 µm for winter and post monsoon respectively. Coarse mode dominates during summer season and peaks at 2.25 µm. Increase in coarse mode particles during summer could be due to the high dust episodes (Mishra and Shibata, 2012) in the study location. Fine mode dominates for monsoon with a tiny peak at 0.38 µm. High volume concentration is noticed during the winter and post monsoon seasons with maximum volume concentration during November. Increased concentration of accumulation mode particles due to the surge in biomass burning activities could be the possible reason for the high volume concentration during post-monsoon season. It has also been noted that coarse mode particles are less predominant in these seasons and constitute a very feeble peak in the size distribution.

There haven’t been many studies on long term measurements of aerosol number size distribution over Delhi. Hyvärinen et al. (2010) studied Gual Pahari, a semi urban area near Delhi and reported increase in nucleation and Aitken mode particles in summer and monsoon seasons, whereas accumulation mode was seen dominant during winter season. They also reported high particle concentration during the winter season. A study by Kumar et al., 2018 looked at the seasonal variation of size distribution in size segregated aerosols and found PM_0.95 as the dominant fraction in PM₁₀ concentration. Another recent study by Gani et al. (2020) has reported the long term measurements of aerosol size distribution over the megacity Delhi. They suggest that accumulation mode particle contributes the most to the PM_2.5 mass concentration over Delhi. Our result is in consistent with the above studies with high concentration of Aitken mode particle during all the four seasons of the Delhi with increase in accumulation mode particles during winter season.

As discussed in the previous section, diurnal variation of aerosol particles showed an increase in nucleation mode particles during midday and warranted for further investigation. Hence, contour plots of aerosol size distribution have been done. Sudden burst of nucleation mode particles, revealing a banana type of growth were noticed from the contour plots. Further analyzing the mode diameter during the event shows a sudden dip towards the nucleation mode confirming the initiation of New Particle Formation (NPF). Contour plots of selected NPF event days of different seasons are shown in Figure 5 along with their diurnal variations in the mode diameter and mode wise number concentration. It shows that NPF events are generally observed at midday during the winter, summer and monsoon periods. A clear sudden burst of nucleation mode particles for a period of few hours is visible at all these days pointing to the NPF events (Dal Maso et al., 2005). The NPF starts at around 10:00 LT in the morning and continue to grow till 14:00 LT. Associated increase in number concentration on nucleation mode can also be seen in Figure 5. Aitken and nucleation mode are the dominant modes over the sampling site during the event days except for winter. The NPF event is preceded by an elevated Aitken mode particle concentration which drops as the NPF commences. During the NPF event time, N_5-30 shows a peak growth. In the representative figures for NPF events highest increase in nucleation mode number concentration events are seen during summer, which reaches ∼3.1 × 10⁴ cm⁻³, following which the particle concentration decreases and continues to grow in Aitken mode which reaches a maximum concentration of ∼4.5 × 10⁴ cm⁻³.

From a total of 222 days of data, 17 (8.4%) NPF event days have been identified along with 28 (∼12.4%) no NPF events days. 177 unidentified events have also been noted which corresponds to ∼78.2% of the total data. Unidentified events are those days in which high level of background noise are present making it difficult to comprehend the features of number-size distribution in the data. This background noise could be due to the pollution or haze condition in the atmosphere. Also, the days on which if there were any doubts regarding the identification of clear NPF events then such days have been classified as unidentified days. An important point to be noted here is that there might be hidden NPF events in this unidentified event days however, the background noise hinders it from quantifying the particle growth. Among the 17 NPF events, 13 occurred in summer season, 2 in winter season and 2 during monsoon season. No events were observed in the post monsoon season. All the NPF event days showed a banana type of growth that continued for several hours and matched the previously reported NPF events in literature (Junninen et al., 2008; Lee et al., 2008). To understand the characteristics of new particle formation during all the seasons, the growth rate and condensation sink parameters have been studied.

Table 2 summarizes the time period of the nucleation event, growth rate and condensation sink of the particle for different seasons. Almost all events start at much lower range in nucleation mode (<20 nm). Growth Rate for nucleation mode (GR_nuc) particle is determined for the event days. In summer it ranges from 1.8 nm/h to 21.66 nm/h with a mean of 8.77 ± 6.25 nm/h. On a few occasions, a comparatively high GR_nuc values were observed during the summer. Such high values of GR indicate the lesser time for the growth of nucleation mode particles. In monsoon GR_nuc is noted to be about 4.8 nm/h and 5.7 nm/h with a mean of 5.25 ± 0.64 nm/h for the two events. Two daytime NPF events were observed during winter period whose GR_nuc were 4.4 nm/h and 3.91 nm/h respectively. The mean GR_nuc for summer and monsoon are comparatively higher than winter nucleation rates. Further, the GR_nuc values of the present study at Delhi are slightly higher when compared with the regions such as Idaho hill (Weber et al., 1997); Ontario (Verheggen and Mozurkewich, 2002); Germany (Birmili et al., 2003) and Hyytiala, Finland (Dal Maso et al., 2007) but it is quite comparable with the GR values at most of the Indian locations reported, like Pune (Kamra et al., 2015); Kanpur (Kanawade et al., 2014) and Hanle, (Kompalli et al., 2014). At the coastal location of Trivandrum, however, the GR values are reported to be high (Kompalli et al., 2014) than that at Delhi.

Being a very polluted environment, the frequency of NPF events is notably less in Delhi. Out of 222 days, only 17 NPF events have been observed with maximum events (73.6%) occurring during the summer months and no NPF events reported during the post monsoon season. A number of factors influence the occurrence of NPF events at regional level including the abundance of precursor gases, pre-existing particle, solar radiation and the local meteorological conditions (Kerminen et al., 2018).

One of the influential factors for local NPF event is the amount of ambient gaseous H₂SO₄ concentration, owing to its inherent connection to the gas phase chemistry associated with aerosol nucleation and subsequent growth (Lee et al., 2019). The solar radiation also acts as a driver for the new particle formation as it controls the formation of OH radicals and the concentration of H₂SO₄ in the air (Pirjola et al., 1998; Dada et al., 2017). A decrease in the intensity of UV radiation tends to decrease the nucleation and the NPF events (Hamed et al., 2011; Baranizadeh et al., 2014). Stanier et al. (2004) used data of UVxSO₂ as a proxy to H₂SO₄ and correlated it with the condensation sink (CS) data to see the conditions that favor nucleation. He found that for the same value of UVxSO₂, nucleation took place at higher CS value during summer.

To analyse the possible mechanism for NPF events over the study region, we have studied different governing factors conducive for NPF events at selected dates from June 12, 2012 to June 18, 2012 for summer and February 13, 2012 to February 17, 2012 for winter. The time period is chosen in such a way that at least two NPF events are present for both the seasons and in accordance with the availability of SO₂ data. Monsoon events are not studied here, due to the unavailability of SO₂ data. As H₂SO₄ data is unavailable for the study site, product of SO₂ and UV is used as a proxy to understand the chemistry on event and non-event days. In the present study hourly data of UVxSO₂ and condensation sink (CS) is correlated for event and nearby non-event days and plotted for the time window of 8:00 LT to 16:00 LT, keeping in view the occurrence of NPF events (Figure 6). It may be noticed that high CS and high UVxSO₂ is observed for NPF event day in summer and low CS and high UVxSO₂ was favorable for the NPF events in winter. We have also found that for the similar UVxSO₂ value nucleation takes place at higher CS values during summer as compared to winter. Unlike Stanier et al. (2004), where nucleation is usually more prominent on the right side of the diagonal, our study do not show such clear preferences, possibly due to the presences of other predictive variables such as relative humidity, concentration of pre-existing particles and prevailing meteorological conditions.

Over our study site, the amount of incoming solar radiation (both in shortwave and UV range) varies significantly during the summer and winter seasons. In summers, the day-time (7 AM–5 PM) shortwave and UV flux increase by 52 and 61% respectively, from the corresponding winter values. Thus, the abundant solar insolation during summer season (high UV) makes summer months more conducive for new particle formation, and this could be one of the plausible reasons for the increased NPF events during summer season. We have also looked at the diurnal variation of UVxSO₂ for event days and non-event days as plotted in Figure 7. It shows the diurnal variation in UVxSO₂ for June (summer) and February (winter). Considerable increase in UVxSO₂ is noticed during summer NPF days compared to the non-event days in summer. A similar result is seen for winter NPF event days also. However, the difference between NPF event and non-event days is much less during winter month. This also suggests that the abundant UV radiation along with high concentration of SO₂ (Figure 8) during the summer months could be the reason for the increased NPF event over the study region. A recent study by Jokinen et al. (2017) revealed the significance of photochemistry in NPF by studying the NPF event during a partial solar eclipse day. He demonstrated the decrease in H₂SO₄ due to decrease in UV during the eclipse time resulting in decreased nucleation events. Thus, the strong gas phase chemistry owing to the high UV radiation coupled with abundant precursor gas concentration (SO₂) could be the reason for the high NPF events reported during the summer period over the study site.

Some of the previous studies have also shown that RH is negatively correlated with continental NPF events (Hamed et al., 2011; Dada et al., 2017). Increase in condensation and coagulation sink have been observed generally during high RH days, owing to the hygroscopic growth by pre-existing aerosol particles due to high intake of water vapor, leading to larger loss of condensable vapor such as H₂SO₄ (Hamed et al., 2011). This may lead to much smaller nucleation rates or nucleation events. In our study, however, we have found that most of the NPF event occurred at RH less than 40%. As the number of NPF events are also less in the present case and mostly observed below 40% RH, a correlation between NPF event and the RH could not be established here.

In order to see the distribution of preexisting particles over the study area we have looked at the PM₁₀, PM_2.5 and SO₂ concentration over the study region. Figure 8 shows the distribution of PM₁₀, PM_2.5 and SO₂ over the selected time window for the NPF and non-NPF days. Comparatively less PM₁₀ concentration is observed for NPF days than the non NPF event days for both seasons which is in accordance with previous literatures. However, high PM₁₀ concentration is noticed for the summer NPF event on June 13, 2012 reaching up to 900 µg/m³. SO₂ concentration is also seen high on this day reaching to a concentration of 6 ppb. Delhi experiences high dust activity during summer months, could be a possible reason for the high PM₁₀ on the particular day. Compared to non-event days in summer, high CS and high PM₁₀ concentration have been noticed for NPF event days. This is a deviation from the usual NPF mechanism. However high CS along with high preexisting particle has also been noticed during NPF events in other regions as well (Yue et al., 2009; Nie et al., 2014). Study by Nie et al. (2014) suggest a possible mechanism for the dust induced NPF event, through dust induced heterogeneous photo catalytic reactions producing OH radicals enhancing the SO₂ production. Our study using proxy data for H₂SO₄ suggests the strong gas phase chemistry owing to the high UV radiation coupled with abundant precursor gas concentration (SO₂) could be the reason for the high NPF events reported during the summer period over the study site.