Influence of semi-volatile aerosols on physical and optical 1 properties of aerosols in the Kathmandu Valley

10 We conducted field study during pre-monsoon season 2015 in the urban atmosphere of the Kathmandu Valley to study 11 the influence of the semi-volatile aerosol fraction on physical and optical properties of aerosols. Our experimental 12 setup consisted of a single ambient air inlet from which the flow was split into two sets of identical sampling 13 instruments. The first set connected directly with an ambient sample while the second set received the air sample 14 through a thermodenuder (TDD). Four sets of experiments were conducted to understand aerosol number, size 15 distribution, absorption, and scattering properties using. Condensation Particle Counters (CPCs), Scanning Mobility 16 Particle Sizers (SMPSs), Aethalometers (AE33) and Nephelometers, respectively. The influence of semi-volatile 17 aerosols fraction was calculated based on the difference of aerosol properties at room temperature, 50°C, 100°C, 18 150°C, 200°C, 250°C and 300°C through set TDD temperatures on ambient samples. Our results show that with 19 increasing TDD temperatures, the evaporated fraction of semi-volatile aerosols also increased. At room temperature, 20 the semi-volatile fraction of aerosol number was 12% of ambient aerosol, while at 300°C it was as high as 49%. 21 Aerosol size distribution analysis from SMPS shows that with an increase in temperature from 50°C to 300°C, the 22 peak mobility diameter of particles shifted from around 60nm to 40nm. However, no distinct change in the effective 23 diameter of the aerosol size distribution was observed with an increase in set TDD temperature. The change in size of 24 aerosols due to loss of semi-volatile component had a stronger influence (~70%) at larger size bins when compared 25 (~20%) to smaller bins of SMPS. At 300°C, the semi-volatile aerosol fraction amplified BC absorption by 26 approximately 28% while scattering by the semi-volatile aerosol fraction contributed up to 71% of total scattering. 27 The Scattering Angstrom Exponent (SAE) of the semi-volatile aerosol fraction was found to be more sensitive at 28 lower temperatures (<100°C) than at higher temperatures. However, the Absorption Angstrom Exponent (AAE) of 29 the semi-volatile aerosol fraction did not show any significant temperature dependence. 30 31


Introduction
Aerosols are suspended particles in the air that are solid, liquid or a mixture of both states, ranging in size from a few nanometers to several micrometers (Warneck, 2000;Zellner, 1999).Studies have shown that aerosols have profound effects on climate (Pöschl, 2005), human health (Kampa and Castanas, 2008;Mauderly and Chow, 2008), and visibility of the atmosphere (Dzubay et al., 1982).Atmospheric aerosols are among the key factors that influence the earth's radiative energy balance.Aerosols affect climate directly by absorption and scattering incoming solar radiation (Haywood and Shine, 1997;Yu et al., 2006) and indirectly by acting as cloud condensation nuclei and affecting the optical properties and life cycles of cloud (Ishizaka and Adhikari, 2003).
Aerosols are classified as primary or secondary depending upon their origin.Primary aerosols are particles that are emitted directly from the combustion of fossil fuels and biomass, such as black carbon as well as sea salts and windblown mineral dust.Secondary aerosols are formed due to condensation, oxidation and chemical transformation (Seinfeld and Pandis, 2006).Secondary aerosols tend to be semi-volatile in nature (Hennigan et al., 2008).The aerosol components that do not condense under normal atmospheric conditions are considered as volatile aerosols while those aerosols that remain in condensed phase under certain atmospheric conditions are classified as semi-volatile.Whereas, the non-volatile aerosols have negligible vapor pressure and remain in condensed phase under normal atmospheric conditions (Fuzzi et al., 2006), semi-volatile aerosols are believed to contribute most significantly to the toxicity of particles (Stevanovic et al., 2015).The volatility of an aerosol provides an indication about its emission sources, history, and chemical composition (Capes et al., 2008).The semi-volatile fraction of an aerosol largely depends on the source of aerosol generation and the atmospheric conditions around the sampling site (Robinson et al., 2007).
Past field and laboratory measurements show that the volatility of aerosols is largely influenced by reaction temperature and precursor gases.Previous studies show that a large fraction of aerosols is highly volatile under 150°C (Ishizaka and Adhikari, 2003;Murugavel and Chate, 2011 and references therein).Measurements made by Lee et al. (2010) and Murugavel and Chate (2011) indicate that the semi-volatile fraction in ambient aerosol was between 50% and 80% of the number concentration.Lin (2013) reported that the potential radiative effect of secondary organic aerosols (SOA), which are a major fraction of semi-volatile aerosols, on climate was around one third of total aerosols.Chung & Seinfield (2002) reported that organic carbon (OC) has a global radiative forcing of -0.09 to -0.17 Wm -2 wherein 50% of the radiative forcing was contributed by SOA.The above studies show that semi-volatile aerosols play a significant role in air quality and energy budget from local to global scales.
Various epidemiological studies have reported the impact of semi-volatile aerosols on human health (Dalton et al., 2001;Ronai et al., 1994).Some of these studies find that in higher concentrations these semi-volatile aerosols can act as potential carcinogens.For example, the semi-volatile polycyclic-aromatic hydrocarbons (PAHs) are not just carcinogenic but also cause genetic susceptibility and oncogene activation (Ronai et al., 1994).The exposure of semivolatile components, such as dioxins, induces heart disease leading towards mortality (Dalton et al., 2001).These studies point toward the need for critical assessment of the semi-volatile aerosol fraction, which can lend to better understanding of human health end points.
The Kathmandu Valley in Nepal is a polluted area in South Asia and generates much interest due to its rapid urbanization, emission sources, topography and proximity to the Himalayas.Emissions from heavy traffic movements, brick-kilns, open burning of solid waste, as well as from households and industrial activities in particular are primary sources of pollution within the valley.Several studies have been conducted in the Kathmandu valley to quantify the mass of black carbon (BC), PM2.5 and PM10 and to identify the seasonality of air pollution (Aryal et al., 2009;Majumder et al., 2012;Putero et al., 2015;Sharma et al., 2012).These studies showed the highest aerosol loading occurs during winter season: BC in winter was ~14 µg/m 3 (Sharma et al., 2012) while PM10 reached ~320 µg/m 3 (Putero et al., 2015).This is due primarily to strong inversion and calm weather conditions during the winter months.
Studies on source apportionment of ambient air quality in the Kathmandu valley based on NMVOC's indicate that brick kilns (~10%), traffic (~17%), residential biofuel and waste disposal (~11%), and industries (32%) are the major sources of pollution (Sarkar et al., 2017).Similar findings but with slight variation in percentages were also observed when estimating sources of EC and OC from PM10 (Kim et al., 2015), PAH's from TSP (Chen et al., 2015), PM2.5 and bulk aerosol studies (Shakya et al., 2010(Shakya et al., , 2017)).Although none of the studies quantified the contribution of the semivolatile aerosol fraction to the ambient atmosphere directly.Source apportionment studies of PM2.5 and PM10 within the valley indicate ~50% contribution from semi-volatile aerosols composed primarily of OC, NH4 and SO4.This fraction, however, varies from time to time depending upon the sampling period and sampling method.In this context, it important to study the impact of the semi-volatile aerosols on physical and optical properties of aerosols in the Kathmandu Valley.To the best of our knowledge, this is the first study of its kind in this area.

Experimental site and general meteorology
The Kathmandu Valley (Figure 1a) sits at 1,300m above mean sea level and is surrounded by hills as high as 2,500m as shown in Figure 1b.We conducted our experiments to characterize the semi-volatile fraction of aerosols contributing to the particulates of the valley.Experiments were conducted on the rooftop of the Integrated Centre for International Mountain Development (ICIMOD) headquarters in Lalitpur, Nepal (27.6464°N, 85.3235° E).As shown in the Figure 1b, the sampling site is located at 7 km southwest of the city center.The sampling site is primarily surrounded by residential dwellings, hospitals, educational institutes, and brick kilns (the nearest brick kiln 2 km).
There are no obstructions around the sampling site within a 50 meter radius.During the measurement period, instruments sampled ambient air from an inlet located 2m above the roof of a four story building (~15 m above the ground).
The general meteorology in Kathmandu during the observation period (https://www.wunderground.com/)included a mean temperature of 21.8±3.1 °C and average relative humidity of 75±10.2%.The daily average wind speed was approximately 1.3 m s -1 indicating prevalence of light air conditions during the sampling time period.The dominant wind direction during this period was westerlies (South West-North West) and easterlies (South East-North East) due to the presence of high mountain peaks on the northern and southern fringes of the Kathmandu valley (Panday et al., 2009;Regmi and Maharjan, 2015).Atmospheric pressure was also observed to be ~868 hPa.In these weather conditions, all measurements were carried out within the span of few months (Table 1).Care was taken so that no experiments were conducted during any occasional rain events.

Experimental Setup
The measurements presented in this study were made during pre-monsoon season of 2015.Four sets of experiments were conducted with this set up, using four different pairs of instruments.Because the TDD flow rate is restricted to 3 Lpm, we could not connect multiple instruments to maintain the TDD flow rate.In each experiment, the TDD's temperature was changed over time to examine the fractional loss of the semi-volatile aerosol fraction at increasing temperatures.A summary of the four sets of experimental setup and respective sampling dates are summarized in Table 1.
We note that the experiments had to be halted due to massive earthquake in Nepal, a period lasting from late April to late May.In the preceding sections the term "wet" sample is used to refer to ambient measurements while the term "dry" sample refers to measurements carried out with instruments coupled with TDD.TDD temperatures were set at room temperature, 50 0 C, 100 0 C, 150 0 C, 200 0 C, 250 0 C and 300 0 C in experiments 1, 3 and 4. Similar experimental temperatures were used in previous studies (Ishizaka and Adhikari, 2003;Jennings et al., 1994;Murugavel and Chate, 2011).In experiment 2, we examined only the relationship between particle size and volatility at room temperature, 50 0 C, 100 0 C, 200 0 C and 300 0 C.All the equipment used in this study were brand new and factorycalibrated.

Instrument description
The TDD used in this experiment is a Low-Flow (4 L min -1 ) Thermodenuder Model 3065, manufactured by Topas GmbH, Germany (Madl et al. 2003).The TDD consists of two sections: one for desorption and the other for adsorption.The TDD removes the semi-volatile fraction of an ambient sample by thermal desorption using a heating element.The semi-volatile fraction that is evaporated by thermal desorption is then adsorbed by the activated carbon which is used as the working material in adsorption section.We operated the TDD only up to 300 0 C though the instrument has a capacity to work at temperatures up to 400 0 C. The activated carbon was changed regularly to ensure an optimal working state of the instrument.
The Condensation Particle Counters (CPC) used in this study were model 3775 manufactured by TSI Inc., USA.This instrument can detect particles as small as 4 nm in diameter and over a wide range of 0 to 10 7 particles per cm 3 .The CPC was operated with flow rate of 1.5 L min -1 .Butanol was used as the working fluid and the instrument was used in the auto drain mode.Efficiency and operation of the CPC model 3775 has been further discussed by Hermann et al. (2007).
We operated a Scanning Mobility Particle Sizer (SMPS 3034, TSI Inc., USA) to measure the size distribution of aerosol particles.The SMPS 3034 works by separating fine particles within a range of 10 to 487 nm based on their electrical mobility.SMPS 3034 is a compact instrument with inbuilt Differential Mobility Size Analyzer (DMA) and Condensation Particle Counter (CPC).We used neutralizer model number 308701 which is x-ray based.Operation of SMPS 3034 has been further discussed by Hogrefe et al. (2006).
We used a Magee Scientific Aethalometer model AE33 to study the black carbon (BC) concentration and aerosol absorption in this study.During the entire measurement period the maximum attenuation limit was set at 100.The instrument was operated at flow rate 2 L min -1 .The Aethalometer model AE33 measures BC concentration at seven different wavelengths (Drinovec et al., 2015) using filter-based light attenuation due to aerosol loading.The manufacturer calibrated instrument measured light attenuation with soot particle loading.However, in real conditions light attenuation reported by the instrument represents all absorbing aerosols including dust, BC, and organic carbon.
We followed the methodology described by Drinovec et al. (2015) to convert BC concentration to absorption by using the relationship: BC Absorption in mm -1 = BC (in ng m -3 )* 10 -3 * MAC (the mass absorption cross sectional values).
MAC values for the specific wavelengths are given by Drinovec et al. (2015).The absorption at 880nm wavelength is usually represented by BC absorption, as other particles like dust and organic carbon do not absorb or their contribution to absorption is negligible at this wavelength (Singh et al., 2014).However, at the 370nm wavelength, aerosol absorption is represented by all three BC, organic carbon and dust (Lim et al., 2014).
We employed a TSI integrating Nephelometer 3563 to measure aerosol scattering coefficient at three different wavelengths: 450nm, 550nm and 700nm.We followed the correction methodology given by Anderson & Ogren (1998).A Nephelometer records both total scattering and backscattering coefficients; however, we only report results from total scattering.We operated the instruments for 24 hours at all previously reported TDD set temperatures.We followed the methodology explained by Anderson and Ogren (1998) to minimize angular truncation error in the dataset using the relationship: σcorrected = Correction factor (C) * σneph; where, C is correction factor, σneph is the scattering coefficient reported by the instrument, and σcorrected is the corrected scattering coefficient.

Quality Control
Prior to the experiments, we conducted collocated inter comparison studies to estimate any biases within each set of identical instruments.Instruments were operated with a single inlet using a Y connector for a 24 hour period with a one minute time resolution.Thereafter, correlation between the identical instruments were calculated.The slope was approximately equal to 1 and the r 2 values 0.99.No correction factors were required.(see Figure S1).
We conducted leakage tests on the inlet pipelines and TDD.The main inlet (Figure 2) was connected to a high efficiency particulate arrestance (HEPA) filter to verify any leakage in sampling system.HEPA filters are known to be highly efficient to produce zero aerosol concentration with a minimum 99.7% of contaminants greater than 0.3 micron (MIL-STD-282 method 102.9.1).During the beginning of each set of experiments, a HEPA filter was connected to the main inlet to check for any leakages in the setup.When the HEPA filter was connected, particle concentration readings in both the identical instruments dropped to zero as shown in Figure S2.These tests gave us confidence that the instrument setup was proper and that there were no leaks in the system.

Results and discussion
Experimental results are summarized according to aerosol number concentration, size distribution, and optical properties in this section.

Influence of volatility on aerosol number concentration
As discussed in Section 3, the influence of semi-volatile aerosol fraction on aerosol number concentration was identified using the CPC and CPC-TDD setup.During the first experiment, the TDD operated at room temperature by not providing power to TDD thermal desorption section as TDD works with only set temperatures as opposed to ambient.This setup provided information about the semi-volatile aerosol fraction loss due to the dry activated carbon column in the TDD.Activated carbon in the adsorption section of the TDD changes the equilibrium state of the semivolatile aerosol fraction and leads to some evaporation even at room temperature (Huffman et al., 2009).We observed a 12% contribution to particle loss from the semi-volatile aerosol fraction at room temperature.Similar losses of 10% to 15% have also been reported in previous experiments (Fierz et al., 2007;Lee et al., 2010;Stevanovic et al., 2015).
These particle losses are governed by various other factors that cannot be completely avoided, such as sedimentation in micro-sized particles as explained by Burtscher et al. (2001) and thermophoretic and diffusional loss in submicron particles as detailed by Stevanovic et al. (2015).
The CPC and CPC-TDD setups shown in Figure 2 operated for 24 hours at each set temperature.The comparison of wet and dry particle number concentrations at the set temperatures of 50°C and 300°C are shown in Figure 3a.The slope of the scatter plot gives the fraction of dry particles at a given temperature.We can see in Figure 3a that there was 16% and 49% particle loss at 50°C and 300°C, respectively.A strong correlation between wet and dry CPC indicates that the fraction of the semi-volatile aerosols at different ambient particle concentrations is similar.However, Figure 3a also shows that the correlation between wet and dry becomes weaker when the ambient particle number concentration is very high (>50000 #/cm 3 ).Similar wet and dry particle number comparisons were obtained at the other TDD set temperatures between 50°C and 300°C and these are provided in Figure 3b. Figure 3b also shows the temperature dependence of the semi-volatile fraction of aerosols.Using one minute data, we calculated the particle percentage loss in dry sampling, displayed in the box plots.The semi-volatile aerosol number fraction was observed to be 12%, 16%, 18%, 23%, 28%, 46% and 49% at room temp., 50, 100, 150, 200, 250, 300°C set TDD temperatures respectively (Table 2).
Comparing these findings with other studies, we note that Murugavel & Chate (2011) reported from Pune, India, that at temperatures below 150°C, 51%-71% of the particles evaporated out of the ambient aerosol.They also reported a 13%-26% loss between 150-300°C and a 7%-13% loss of particles at temperatures greater than 300°C.These results show that the evaporated fraction of the semi-volatile aerosols at different temperature ranges is comparatively less in Kathmandu than Pune.Whereas, Murugavel & Chate (2011) used SMPS for their study, we used CPC for this study, which measures particle number concentrations from 4 nm to a few microns.The SMPS used by Murugavel & Chate (2011) only measures particle number concentrations between 10 nm to 487 nm.Kathmandu and Pune have different source characteristics, topography, and local meteorological conditions.Thus, these differences in results with Murugavel & Chate (2011) should not be surprising.
We compared the semi-volatile aerosol fraction from our experiment with the standard chemical compounds that have evaporating temperatures equivalent to the TDD set temperatures.This comparison would provide vital information about aerosol chemical composition and volatility.A similar comparison technique has been adopted by several others studying this topic (Burtscher et al., 2001;Ishizaka & Adhikari, 2003;Murugavel & Chate, 2011).
For further analysis, we sorted aerosol volatility into two categories: (I) highly volatile for those components which vaporize at temperatures ≤150 °C, and (II) moderately volatile for those which vaporize between 150 °-300 °C.We found 23% of aerosols in our study to be highly volatile and 26% to be moderately volatile.The average aerosol number concentration during our experimental period was 16,136 #/cm 3 .Out of this number, 2,038 #/cm 3 are highly volatile in nature.These particles may represent some of the highly volatile aerosol components cited in Ishizaka & Adhikari (2003), such as ammonium chloride, ammonium sulphate, terpene, organic nitrogen, ethyl benzene, and sulfuric acid, among others.Among the moderately volatile components cited in Ishizaka & Adhikari (2003), including diesel exhaust and ammonium bisulphate, we found 6319#/cm 3 aerosol particles to be moderately volatile in nature.In   concentration.These spikes may represent different air masses or fresh nearby sources.Thus, the semi-volatile aerosol number fraction is significantly higher during peak events.However, compared to the aerosol number diurnal variation, semi-volatile aerosol fraction does not exhibit strong diurnal variation (Figure 4) and the fraction remains almost constant (Figure 3).
By comparing the diurnal variation of highly and moderately volatile aerosol fractions, we noticed highly volatile aerosol contribution was nearly consistent throughout the day while the moderately volatile aerosol fraction changed significantly during peak events (Figure S3).As mentioned above, diesel combustion is a moderately volatile aerosol source and this combustion (primarily from vehicle traffic) may account for the minor spikes in semi-volatile aerosol fraction.

Influence of volatility on aerosol size distribution
Our experimental setup using SMPS was different from other instrumental setups.In first experiment using CPC, we identified that semi-volatile aerosol number fraction which did not show any strong diurnal variation.From this point, we wanted to understand particle size loss due to the semi-volatile aerosol fraction rather than diurnal variability.
Hence, we operated identical SMPSs (as described in the instrument setup section), but changed the TDD set temperature every hour.We readily acknowledge that this decision also reveals a limitation of our study.
The results of SMPS measurements are summarized in Table 2.The semi-volatile aerosol fraction was observed to be 62% and 49% during SMPS and CPC experiments respectively at TDD set temperature 300°C (see table 2).Similar behavior was seen at other temperatures as well.One possible reason for the difference between the two measurements might be that CPC reported aerosol number concentrations from 4 nm to a few microns whereas the SMPS only monitors 10nm to 487nm.The semi-volatile aerosol fraction may be higher in smaller diameter particles compared to larger diameter particles.Figure 5 shows the number size distribution for TDD at room temperature and 300°C.Even though the dry aerosol number size distribution at room temperature was significantly lower compared to wet aerosol, their peak mobility diameter did not shift significantly (from 85nm to 80nm).A similar comparison at 300°C shows a different result.At this temperature, the peak diameter shifted from approximately 60nm to 40nm.This correlates with a previous study (An et al., 2007) that reported that when individual ammonium sulphate particles of different sizes were evaporated at different temperatures, the particle sizes decreased significantly.Thus, we hypothesize that the decrease in particle size of individual components of aerosol contributed in the shift in peak diameter towards smaller diameter as observed in our study.
In Figure 6, we show individual size bin semi-volatile aerosol fraction loss as a function of the particle diameter at all experimental temperatures.The data show a greater reduction of aerosol size for larger diameter aerosols compared to smaller diameter aerosols.This may be due to the fact that the semi-volatile aerosol fraction at larger diameters may be in the form of a coating or internally-mixed state.By losing this fraction, the aerosol size is expected to decrease.This conclusion is corroborated by the observation that, as the size of the aerosols decrease, the peak diameter also shifts towards smaller diameters at higher temperature.The shift in peak diameter was observed to be approximately 5-7 nm between wet and dry sampling at TDD set temperature ≤100°C, and shifted significantly to 20-22 nm at TDD set temperature 200°C to 300°C.Murugavel and Chate (2011) reported variable particle loss percentage at different sizes and at different TTD temperatures over a yearlong study.The difference between their study and ours may be attributed to differences in data collection.Murugavel and Chate (2011) reported data that represents monthly and annual averages, whereas our present study represents an event sampling (one hour).Our results show that the mixing process of ambient aerosols with highly and moderately volatile aerosols/precursors are different due to differing emission sources and requires further investigation.

Influence of volatility on aerosol absorption
In actual atmospheric conditions, BC exists in both elemental and mixed states.An aethalometer-derived aerosol absorption represents both these states.Previous studies show that non-absorbing material coatings over elemental carbon (EC) enhances absorption due to the lensing effect (Lack & Cappa, 2010;Schnaiter et al., 2005;Shiraiwa et al., 2010;Zhang et al., 2008).By removing this coating, elemental carbon absorption will change.Dust aerosol displays absorption features in the ultraviolet (UV) through the VIS wavelengths due to its mineralogical composition; however dust aerosol is non-volatile in nature.Elemental carbon can be volatilized above 600°C (Shrestha et al., 2014), but below 300°C it is stable.In our study, semi-volatile aerosol fraction absorption is represented either by light absorbing organics or material (organic/inorganic) coated on EC.
Dust absorption is generally low in the spectral regime above 600 nm or tends to have a constant background absorption value for wavelengths larger than 600 nm (Gillespie and Lindberg, 1992;Lindberg et al., 1993;Sokolik and Toon, 1999;Cao et al., 2005;Kumar et al., 2008).Hence, wet and dry absorption measured by the aethalometer at 880nm wavelength is representative of either EC or mixed state of EC absorption (Lack & Cappa, 2010).
Comparison of wet and dry aerosol absorption at 880nm for 50°C and 300°C is shown in Figure 7a.The semi-volatile aerosol fraction absorption contribution to the total aerosol absorption was observed to be 20% and 28% at 50°C and 300°C, respectively, at 880nm wavelength.Loss of absorption from 50°C to 300°C increased only 8%, which is less compared to particle loss of 33%.The particle loss was not proportional to the absorption loss at 880nm wavelength mainly due to EC and its mixed state.Since EC and dust cannot be volatilized under 300°C thus, the measured semivolatile aerosol fraction absorption contribution at 880nm is mainly from EC mixed state with semi-volatile aerosol fraction.
Highly and moderately volatile aerosol fraction contributed 21% and 7%, respectively, to aerosol absorption at 880nm wavelength.As shown in Table 3, one fourth of aerosol absorption at 880nm wavelength is contributed by the semivolatile aerosol fraction.Wet and dry aerosol absorption at 370nm is shown in Figure 7b.The semi-volatile aerosol fraction absorption was slightly higher at 370nm compared to 880nm (Table 3).The correlation for wet and dry absorption at both wavelengths stays similar at a higher range unlike aerosol number concentrations reported by CPC in section 4.1.Results for other TDD set temperatures are summarized in Figure 7c.
If we assume BC mixing state absorption effects are similar at 370nm and 880nm wavelengths, then the difference between 370nm to 880nm wavelength absorption may be attributed to changes in the intrinsic properties of the semivolatile aerosol, the size of aerosols, mixing state or brown carbon (BrC), which is unknown at present.This brown carbon absorption is approximately 3% and 9% at 50°C and 300°C, respectively.As shown in Table 3, highly volatile aerosol fraction does not enhance aerosol absorption at lower wavelengths (0-3%).This finding indicates that highly volatile aerosols are not truly representative of brown carbon aerosols.Furthermore, as our results show, the moderately volatile aerosol fraction absorption does enhance 4-9% at 370nm compared to 880nm.This indicates that brown carbon aerosols are moderately volatile in nature.We conclude that the brown carbon contribution is relatively less (0-9%) compared to absorption enhancement (16-28%) due to EC mixing state with the semi-volatile fraction of aerosol (Table 3).
In Figure 8a and 8b, we report the diurnal variation of absorption using the aethalometer and the aethalometer coupled with a TDD setup at 50 °C and 300 °C, respectively, for 520nm wavelength.As expected, both figures show an increase in BC absorption during the early morning hours, less during afternoon, and then building up again toward eveninga finding similar to results in Backman et al. (2012).Figure 8c shows that at 50 °C both wet and dry aerosol absorption demonstrate similar magnitude as well as diurnal variation.The semi-volatile aerosol fraction absorption is approximately 20% at 50°C TDD set temperature.However, as shown in Figure 8d, although the diurnal variability is similar at 300 °C, the semi-volatile aerosol absorption increases to nearly 30%.Unlike the CPC testing, the semivolatile aerosol absorption is more variable throughout the day at TDD set temperature 50°C.It is also noteworthy that the semi-volatile aerosol absorption shows similar variability at both TDD set temperatures.
Further, the data obtained from aethalometer were used to illuminate the wavelength dependency of absorption which is usually expressed as an Absorption Angstrom Exponent or AAE.AAE is simply the negative slope of the log of absorption by the log of two different wavelengths.In past studies, AAE has been used to make inferences about the dominant composition of absorbing aerosols in the atmosphere (Bergstrom et al., 2007).Several studies report AAE values close to 1 for fossil fuel sources and close to 2 for biomass sources (Kirchstetter et al., 2004).Our results, as summarized in Table 3 show wet aerosol AAE ranges from 0.97 to 1.30 with a median value around 1 implying sampled aerosols are dominated by fossil fuel sources.The AAE results for 300 °C shown in Figure 9 was around 1.5 indicating the influence of biomass burning source(s).
Dry absorption at different TDD set temperatures was deducted from simultaneous measurement of wet absorption values to compute the semi-volatile aerosol fraction absorption (details given in S1).Thereafter, semi-volatile aerosol fraction absorption at seven different wavelengths was used to compute AAE for the semi-volatile fraction.We computed AAE over a range of wavelengths (370nm -970nm) for wet, dry and semi-volatile aerosol fractions individually (Figure 9).As also shown in Table 3, the semi-volatile aerosol fraction AAE ranged between 1.10 to 1.43.
The maximum semi-volatile aerosol fraction AAE value was observed when the sample was influenced by biomass sources.Highly volatile aerosol AAE was recorded at 1.1 whereas moderately volatile aerosol AAE was ranged between 1.1-1.4.As discussed above, the highly volatile aerosol absorption was primarily from the BC mixing state and the mixing state AAE is 1.1.We compared our findings with Lack and Langridge (2013) who reported brown carbon AAE ranging between 2-10.Our observed semi-volatile aerosol fraction AAE was significantly below this range further corroborating the results that absorption is mainly influenced by mixing states as compared to brown carbon.

Influence of volatility on aerosol scattering
In this section, we discuss the influence of volatility on the scattering properties of the aerosols.As shown in Figure 10a, the semi-volatile aerosol fraction scattering contribution at 700nm wavelength was observed to be 8% and 66% of wet aerosol scattering at 50°C and 300°C TDD set temperatures, respectively.Whereas at 450nm wavelength, the semi-volatile aerosol fraction scattering contribution increased to 17% and 71% at 50°C and 300°C TDD set temperatures, respectively (Figure 10b).The influence of wavelength on scattering loss is evident for all set temperatures (Figure 11a).Even though the CPC and scattering experiments were conducted on different days, by assuming the urban air mass characteristics remains similar, we infer that particle loss is not proportional to scattering loss (Figure 3a and Figure 10a & 10b).However the scattering loss at 700nm wavelength is somewhat similar (66% to 62% versus 66% to 49%) to the particle loss in the SMPS experiment (Table 2 and Table 4).Thus, the smaller particle semi-volatile aerosol fraction has greater influence in total aerosol scattering.More tests are required to statistically validate these inferences.We summarize the results for all three wavelengths (450nm, 550nm and 700nm) at different TDD set temperatures in Figure 11a.Diurnal variability of wet and dry aerosol scattering (figure not shown) was observed similar to CPC experiment.
Using the methodology described in section 4.3 we computed the Scattering Angstrom Exponent (SAE).The semivolatile aerosol fraction SAE results differ compared to AAE (Figure 11b).The semi-volatile aerosol fraction AAE was similar at all TDD set temperatures while SAE shows higher values for room temperature and 50°C compared to higher TDD set temperatures.The semi-volatile aerosol fraction SAE values were observed to be greater than 4 at room temperature and 50°C.We repeated the same experiments several times to cross-check the SAE values and observed a consistent trend.The scattering contribution at 700nm wavelength might be representing bigger particles and they may have less contribution of highly volatile aerosols.However, this low scattering loss at 700nm, indicates the necessity for significant in-depth future studies.This indicates that the scattering contribution of the semi-volatile fraction is nearly 8 times higher at 450nm wavelength compared to the 700nm wavelength at room temperature and 50°C.For other TDD set temperatures, the semi-volatile aerosol fraction contribution to scattering is around three times at 450nm wavelength compared to 700nm wavelength, which is slightly higher than that of wet aerosol SAE (Table 4).

Influence of volatility on aerosol single scattering albedo
Single Scattering Albedo (SSA) is the ratio of scattering coefficient to extinction coefficient, which provides an indication of how absorbing or scattering the sampled aerosol is.An SSA value greater than 0.95 represents aerosol with a net effect of cooling whereas an SSA value less than 0.85 will have a warming effect.SSA values between 0.85 and 0.95 may represent warming or cooling effect depending upon surface albedo and cloud cover (Ramanathan et al., 2001).We computed SSA (more details in S2) values for different semi-volatile aerosol fraction at different TDD set temperatures by assuming wet aerosol SSA as 0.9 and 0.95.By assuming wet aerosol SSA at 0.9 we can derive scattering and absorption coefficient values as 100X and 11X (X can be any arbitrary value).We know from sections 4.3 and 4.4 the amount of scattering and absorption contribution from the semi-volatile aerosol fraction.By applying these fractions to the 100X and 11X values we can retrieve scattering and absorption coefficients of the semi-volatile aerosol fraction and calculate the SSA.By using this method, we calculate the semi-volatile aerosol fraction SSA values (Table 5).
Our results show that the semi-volatile aerosol fraction SSA was observed lower at room temperature and 50°C compared to other TDD set temperature at all wavelengths.In addition, the semi-volatile aerosol fraction is more absorbing at 700nm compared to the 450nm wavelength.This may due to the fact that scattering and absorption loss are not similar at different wavelengths.Our results show at TDD set temperature 50°C, the semi-volatile aerosol fraction SSA was observed to be minimal at all wavelengths.The semi-volatile aerosol fraction scattering loss was observed relatively high compared to its absorption TDD set temperature at 50°C.This led to lower SSA values at this temperature.If this process is applicable in atmospheric conditions, rising noon time temperature may influence the net aerosol optical properties and make the atmosphere more absorbing in nature.

Conclusion
Ours is the first of its kind study to quantify the semi-volatile aerosol fraction influence on aerosol physical and optical properties over the Kathmandu Valley.Experimental results show that the semi-volatile aerosol number fraction ranged from 12% to 49% at TDD set temperatures from room to 300°C, respectively.During our experiment, we observed that highly volatile aerosols do not exhibit diurnal variability while the contribution of moderately volatile aerosols increases during peak concentration events.In addition, SMPS experiment results show that the reduction of the aerosol size was high for larger diameter aerosols compared to smaller diameter aerosols due to removal of the semi-volatile aerosol fraction.We also note that the semi-volatile aerosol fraction mixing state contributed around 20% to total aerosol absorption.Aerosol absorption by the semi-volatile aerosol fraction were observed to be in between 16% to 28% at 880nm wavelength whereas calculated brown carbon contribution to aerosol absorption ranged from 0 to 9%.The scattering contribution was observed to be in the range 18% to 71% and 8% to 66% at 450nm and 700nm, respectively.Our results show that the semi-volatile aerosol fraction contribution to aerosol scattering was significantly higher compared to aerosol absorption and number.Since the semi-volatile aerosol fraction scattering contribution was found to be two times higher than its absorption, this implies that removal of the semi-volatile aerosols will lead to a more absorbent atmosphere.
In short, our study shows that the semi-volatile aerosols play important role in characterizing aerosol physical and optical properties over the Kathmandu Valley which will further aid in understanding health and climate impacts of aerosols.The results discussed are based on limited and unique aerosol sampling in the Kathmandu Valley and can be improved to better characterize aerosols in the valley.27 experiments carried out byIshizaka & Adhikari (2003) andShrestha et al. (2014) the remaining particles which did not volatilize up to 300°C were soot carbon, polymerized organic compounds, calcium carbonate, sea salt and mineral dust which may represent non-volatile aerosols in our experiment.Characterizing aerosol volatility as a function of chemical composition in the Kathmandu Valley needs further investigation.The diurnal variation of wet and dry aerosol number concentrations and the semi-volatile aerosol fraction at 50°C and 300°C are shown in Figure4.We show results from only these two temperatures to represent the minimum and maximum TDD set temperatures.Results from other TDD set temperatures lie between these two temperature results.

Figure 4
Figure 4 (a & b) show diurnal variation in wet aerosol number concentrations which shows one peak during morning hours around 9 a.m. and the other during evening hours around 8 p.m. Previous studies by Panday & Prinn (2009) and Putero et al. (2015) reported a similar diurnal profile.

Figure 4
Figure 4 (c & d) shows the semi-volatile aerosol number fraction at 50°C and 300°C TDD set temperatures.The semivolatile aerosol fraction did not show any diurnal variation like wet aerosols at 50°C, which indicates that the semivolatile aerosol number fraction is uniform throughout the sampling period.However, as shown in Figure 4d at TDD set temperature 300°C, the semi-volatile aerosol fraction fluctuates with minor spikes in wet aerosol number

FiguresFigure 1 .
Figures Figure 1.(a) Satellite image of South Asia showing the location of Kathmandu Valley (indicated by red square symbol).(b) Elevation contour map displaying the Kathmandu valley and the ICIMOD sampling site indicated by symbol "*".Color bar indicates elevation above mean sea level in meters.The red square in the top figure (a) has same coordinates as the bottom figure (b).

Figure 2 .Figure 3 .
Figure 2. Schematic of instrumental setup.CPC, aethalometer, SMPS and nephelometer were operated at flow rates 1.5 lpm, 2 lpm, 5 lpm and 5 lpm respectively.Identical instruments were maintained with same flow rates.At a time, the experiment was carried out with only one set of instruments, i.e. either CPC or aethalometer or so on.

Figure 4 .
Figure 4. Comparison of diurnal variation of particle number concentration of wet and dry aerosol at TDD set temperatures 50 °C (a) and 300 °C (b).Diurnal variation of fraction of dry particles at TDD set temperatures 50 °C (c) and 300 °C (d).

Figure 5 .
Figure 5. Particle size distribution of wet and dry aerosol at room temperature (a) and 300°C (b).

Figure 6 .
Figure 6.Percentage of particle reduction from wet aerosol in their respective size bins at different TDD set temperatures.

Figure 7 .Figure 8 .
Figure 7.Comparison of dry and wet BC absorption at TDD set temperatures-50°C and 300°C at wavelengths-880nm (a) and 370nm (b).(c) Boxplot of loss of absorption at different TDD set temperatures of 880nm and 370nm wavelengths.

Figure 9 .
Figure 9. Boxplot of Absorption Angstrom Exponent (AAE) of wet, dry and semi-volatile aerosol at different TDD set temperatures.AAE values were computed over the range of 370nm-970nmwavelengths.

Figure 11 .
Figure 11.(a) Boxplot of scattering loss at different TDD set temperatures at 450nm, 550nm and 700nm wavelengths.(b) Boxplot of Scattering Angstrom Exponent (SAE) of wet, dry and semi-volatile aerosol at different TDD set temperatures.SAE was computed over the range of 450nm-700nm wavelengths.

Table 1 .
Summary of four sets of experiments carried out with their respective sampling dates

Table 3 .
Summary of influence of volatility on absorption at various temperatures.
# The fraction represented in the table are derived from linear interpolation of slopes

Table 4 .
Summary of influence of volatility on scattering at various temperatures.
*Average absorption Angstrom coefficient of wet aerosols (ambient aerosol) while the simultaneous dry experiment was being conducted at TDD set temperatures.#The fraction represented in the table are derived from linear interpolation of slopes

Table 5 .
Summary of semi-volatile aerosol fraction Single Scattering Albedo (SSA) assuming wet aerosol SSA as 0.9 and 0.95 at different wavelengths and