Evaluation of Artifacts Generated during Collection of Ultrafine Particles Using an Inertial Filter Sampler

Two artifacts were observed during collection of atmospheric ultrafine particles (UFPs; < 100 nm diameter) using an inertial filter (INF) sampler recently developed for this purpose. These artifacts were evaluated. First, the adsorption of organic and ionic gas onto a quartz fiber filter installed in the INF sampler was evaluated to provide information on the positive artifact; this information is important for accurate analysis of the UFP components. Gas adsorption by the INF sampler was similar to or less than that of other devices for collecting UFPs. The gas adsorption of organic carbon (OC) fractions, and ionic components of NH4, NO3 and SO4 was confirmed, and the adsorption of OC1 and ionic components were highly dependent on the concentration of the gas and the UFP concentration under ambient conditions. These results suggest that it is necessary to know the concentration of the UFPs on the filter in order to evaluate the exact concentrations of UFP components under high flow rate conditions. Second, the efficiency of UFP collection on a polytetrafluoroethylene (PTFE) filter installed in the INF sampler was evaluated to clarify the negative artifact. The data confirmed that the efficiency of collection changed with changing structure such as pore size, porosity and thickness of the filter. The structure of these filters and UFPs collected on there were observed by a scanning electron microscopy. The highest particle collection efficiency (almost 100%) was obtained by installing two thick membrane PTFE filters. The collection of UFPs using a sampler comprising several filter stages is a convenient and useful method for evaluating positive and negative artifacts and for quantifying the concentration of components of UFPs.


INTRODUCTION
The particle size and chemical components of atmospheric particles are important for evaluating their influence on air pollution and human health.In particular, ultrafine particles (UFPs, < 0.1 µm) have a greater tendency to be deposited in the pulmonary alveoli than do fine particles (PM 2.5 , < 2.5 µm) and thus UFPs may act as important carriers of toxic compounds into the pulmonary alveoli (Ferin et al., 1992;Oberdörster et al., 2000;Ibald-Mulli et al., 2002;Li et al., 2002;Oberdörster and Utell, 2002;Oberdörster et al., 2002;Li et al., 2004;Oberdörster et al., 2004;Myojo et al., 2010;Xia et al., 2010).Although the contribution of UFPs to mass concentration is low, their contribution to number concentration and surface area is higher than that of larger particles.Therefore, UFPs are believed to be more harmful than PM 2.5 at the same mass concentration.For example, roadside measurements of nitroarene particles with UFP or accumulation mode particle (particle size, 0.15-0.5 µm) diameters showed that the mutagenicity per unit mass of UFPs was significantly higher (Kawanaka et al., 2006(Kawanaka et al., , 2008)).The main sources of UFPs include diesel and gasoline vehicle emissions in urban areas; in particular, many nuclear mode particles (particle size, 0.01-0.03µm) are emitted from diesel engines (Kittelson, 1998;Geller et al., 2005;Vouitsis et al., 2009).Secondary particles formed by photochemical reactions of gaseous compounds are also sources of UFPs (Shi et al., 1999(Shi et al., , 2001;;Hitchins et al., 2004;Minoura and Takekawa, 2005).However, the atmospheric behavior of UFPs remains unclear.The evaluation of the health effects and atmospheric behavior of UFPs requires their collection on a filter, classification, and analysis of the UFP components.Low pressure impactor (LPI) and micro orifice uniform deposition impactor (MOUDI) have previously been used to collect UFPs (Pakkenen et al., 2001;Fushimi et al., 2008;Gietl et al., 2010;Fushimi et al., 2011), but unstable components collected on the filter can volatilize using these techniques, and the low pressures used during collection can give rise to non-uniform collection on the filter (Michael et al., 2002;Zhu et al., 2010).
An inertial filter (INF) sampler (Fig. 1) was recently developed to address these problems (Otani et al., 2007;Furuuchi et al., 2010a, b).The advantages of the INF sampler for sampling UFPs include uniform sampling onto the filter and collection under ambient pressure.The INF sampler consists of four impaction stages that remove particles with cut-off diameters of 10, 2.5, 1.0 and 0.5 µm by impaction onto the quartz fiber filters comprising each impaction stage.After these four impaction stages, particles pass through a cartridge packed with stainless steel fibers at a high flow rate (40 L min -1 ) to promote inertial collection with decreasing diffusion force; the UFPs are separated and led to a backup polytetrafluoroethylene (PTFE) filter, then are collected uniformly onto a 47 mm φ quartz fiber filter.The pressure loss in this sampler is approximately 20 kPa, which is lower than that of a LPI sampler (about 70 kPa) (Furuuchi et al., 2010a).The low pressure loss and uniform collection of UFPs suppresses the volatilization of unstable components and improves the accuracy of component analysis after filter punching.
The classification and collection performance of an INF sampler and a Nano-MOUDI for general atmospheric UFPs was compared by analyzing the carbon and ionic components of the particles (Kim et al., 2010).The INF sampler was also used in a study in which UFPs were collected during the day and at night (Kim et al., 2013) and demonstrated excellent time resolution.These studies showed that the INF sampler is suitable for the collection of UFPs.However, the positive and negative artifacts potentially arising from gas adsorption onto the quartz fiber filter and particles passing through the PTFE filter have not been evaluated.
An INF sampler collects UFPs under high flow rate onto a high-surface-area quartz fiber filter; therefore, gas components such as volatile organic compounds (VOCs), nitrate, and ammonia might adsorb onto the filter, resulting in overestimation of the concentration of UFPs and generating a positive artifact.Although the volatilization of volatile components from the filter is also considered as negative artifact simultaneously, many researchers (Sihabut et al., 2005;Chen et al., 2010;Cheng et al., 2010;Zhu et al., 2012) have reported that the influence of arttifact during PM 2.5 collection using a cyclone or a MOUDI is remarkably affected by gas adsorption than volatilization.Therefore, only gas adsorption was evaluated for a quartz fiber filter as positive artifact in this study.PTFE filters are typically used to collect fine particles for metal component analysis.Their collection efficiency has been investigated, and the penetration of fine particles has been reported (Benjamin et al., 1976;Chin et al., 2005), resulting in underestimation of the particles and generation of a negative artifact.The purpose of this study was to estimate the influence of gas adsorption onto the quartz fiber filters as a positive artifact, and particle penetration through a PTFE filter as a negative artifact, on UFP collection using an INF sampler.This INF sampler has been used for atmospheric sampling by many researchers (Otani et al., 2007;Furuuchi et al., 2010a, b;Kim et al., 2013).

Gas Adsorption Evaluation of a Quartz Fiber Filter
Gas adsorption onto a quartz fiber filter was estimated using three 47 mm φ quartz fiber filters separated by 5mm-thick PTFE O-ring spacers in collecting stage of an INF sampler and UFPs were collected from the ambient atmosphere.UFPs are collected on only the first filter, but organic and inorganic gases are adsorbed onto the first, second and third filters.If there is no difference in the concentration of organics and the concentration of inorganics on the second and third filters, then gaseous components should adsorb onto each filter at a constant ratio.Concentrations measured on the second filter may include some volatile gaseous components vaporized from the first filter but are likely to be too low to significantly influence the total concentration on the second filter.Therefore, the second and third filters were used to evaluate the gas components and the concentrations of their chemical components were compared.The gas adsorption ratio of the filters was calculated using Eq.(1) as the percentage of gas components present in collected UFPs.
Gas adsorption ratio (%) Second or third filter concentrarion μg m 100 First filter concentrarion μg m Two sampling sites were used to examine the concentration of air pollutants.The main site was on the roof (about 40 m above ground level) of the 10-story Research and Project Building at Saitama University (Saitama, Japan).This building is located about 200 m from a road (National Road 463).The distance of this collection site from the road and from the ground should mitigate direct sources of UFPs, such as automobile exhaust and soil particles.The second site was the entrance gate of Saitama University, adjacent to National Road 463.This site chosen to examine the influence of high gaseous and particulate concentrations.Atmospheric sampling for evaluation gas adsorption was conducted during three periods: i) from Aug. 7 to 10, 2011 (summer) on the roof; ii) from Dec. 26 to 31, 2011 (winter) on the roof; and iii) from Mar. 31 to Apr. 5, 2014 (spring) at the gate, adjacent to the road.Each sampling duration was 23.5 h, corresponding to a total sampling volume of 56.5 m 3 for UFPs.The samples were analyzed for OC and ions (NH 4 + , NO 3 -and SO 4 2-) as gas adsorption components, and for elemental carbon (EC), which was only emitted as primary particles and used to confirm the passage of UFPs through the first filter.

Evaluation of the Collection Efficiency of PTFE Filters
Three types of PTFE filters (PF040, ADVANTEC; P5PJ047, PALL; R2PJ047, PALL shown in Table 1; 47 mm φ) were tested in the INF sampler for collecting UFPs.PF040, R2PJ047, and P5PJ047 are described in this paper as S, M R and M P, respectively.The duration of sampling was 23.5 h and sampling was conducted on the roof during three periods: i) from July 3 to 6, 2013 for S filter; ii) from Nov. 12 to 15, 2013 for M R filter; and iii) from Nov. 19 to 22, 2013 for M P filter.The S filter is a sheet filter with a porosity of 75%.The M P and M R filters are membrane filters with a 2 µm pore size.The INF sampler cannot function if the porosity or pore size of an installed filter is too small due to high pressure loss, whereas it is difficult to collect UFPs if the porosity or pore size is too big.From these points and previous research by Chin et al. (2005), the S, M P and M R filters were appropriate for the INF sampler.The structures of the quartz fiber filter, and of the S, M P and M R filters after collecting UFPs, were confirmed by scanning electron microscopy (SEM; S-4100, Hitachi, Japan).For SEM analysis, each filter was cut to 2 mm sample using a stainless steel knife.The PTFE filter providing the highest collection efficiency on the roof was used to collect UFPs at the university gate, adjacent to the road, during a period from Dec. 17 to 20, 2013 in order to confirm its performance.Two INF samplers were used to measure the collection efficiency of the PTFE filters and contained both a PTFE filter and a quartz fiber filters as shown in Fig. 2. As shown in Fig. 2(A), in one INF sampler the PTFE filter was installed at the top and any UFPs that passed through this filter were collected by the quartz fiber filter below.In the second INF sampler, the quartz filter was installed at the top, as shown in Fig. 2(B), and all UFPs were collected by the quartz fiber filter.A PTFE filter was installed below the quartz fiber filter to collect UFPs at same pressure condition.Atmospheric UFPs were simultaneously collected using both the INF samplers and the EC component of the UFPs collected by each quartz fiber filter was analyzed to evaluate the passage of UFPs through the PTFE filter.Therefore, collection efficiency of PTFE filter was calculated by Equation 2.

Analysis of Carbonaceous and Ionic Components
OC and EC concentrations in UFPs collected on the quartz fiber filter were analyzed using a Desert Research Institute (DRI) Model 2001 Carbon Analyzer (Atmoslytic Inc., Calabasas, CA, USA) following the Interagency Monitoring of Protected Visual Environments (IMPROVE) TOR method (Chow et al., 2001).A sample (0.503 cm 2 ) punched from the filter was placed in the sample load position of the analyzer and heated to produce four OC fractions (OC1, OC2, OC3 and OC4) at temperatures of 120, 250, 450 and 550°C, respectively, in a non-oxidizing helium atmosphere, as well as three EC fractions (EC1, EC2 and EC3) at 550, 700 and 800°C, respectively, in an oxidizing atmosphere of 2% O 2 /98% He.At the same time, pyrolysis OC (POC) was measured by the transmittance of a He-Ne laser (λ = 633 mm) signal to correct the OC and EC concentrations.In this study, OC and EC were determined using Eqs.( 3) and ( 4), respectively, and OC1, OC2, OC3, and OC4 were defined as having been corrected for the POC concentration (Judith et al., 2007).
Ionic components were extracted and analyzed by placing the remaining quartz fiber filter (after sample removal for carbon analysis, above) in 5 mL of Milli-Q water in a 20 mL glass vial, which was then sealed with a cap and PTFE tape to prevent the volatilization of ionic species.The ion species were extracted in an ultrasonic bath maintained below 10°C with ice packs for 30 min, and then the ionic species were analyzed using an ion chromatograph (DX-100, Dinoex Corp., Sunnyvale, CA, USA) (Kim et al., 2011).Before sampling, the quartz fiber filters were baked in an oven at 350°C for 1 h to remove OC according to the standard method reported from the Japanese government.After sampling, each filter was immediately stored in a cleaned petri slide, sealed in a resalable plastic and aluminum bag, and placed in a freezer (-40°C).The values of the detection limit and quantitation limit were defined as the average concentration of blank (control) filters plus three and ten standard deviations (3σ and 10σ), respectively.The standard deviation of OC component was 2.72 × 10 -1 µg m -3 and both of EC and ionic components were not taken into account in this study because those concentrations were not detected on the blank filter.

Evaluation of Gas Adsorption on Quartz Fiber Filters Carbon Components
The average concentration and the gas adsorption ratio of carbonaceous components of particles collected on the roof of the Research and Project Building at Saitama University during the summer and winter 2011, and at the university gate adjacent to the road during spring 2014, are shown in Figs.3-5, respectively.The mean value and error bars obtained using the standard deviation of each sample are shown in the bar graph and the respective gas adsorption ratio is shown as a solid line.UFPs concentrations near the road were higher than those at the roof, probably because particles were both directly generated by automobiles on the road and by condensation to provide secondary particles, whereas these source effects were less on the roof (Kim et al., 2010).No EC was detected on the second and third filters, indicating that all UFPs were collected on the first filter at both sampling sites.
The measured OC concentrations in samples collected on the roof (Figs. 3 and 4) were slightly higher on the second filter compared with the third filter.This trend indicates that volatilization on the first filter was high in the summer and some low-boiling components, such as OC2 and OC3, adsorbed onto the second filter after volatilizing from the first filter.Volatilization during the summer of lower-boiling components in OC1 on the first filter was also observed, but was not observed in the winter.The correlation between OC concentration on the second and third filter following each collection is shown in Fig. 6.High correlation was confirmed in all cases, but there was a slight bias for the second filter, especially in summer.These results also confirmed that high temperatures during collection in the summer resulted in low-boiling point components on the first filter volatilizing and absorbing onto the second filter, and volatilization was much less pronounced in winter.However, the influence of volatilization, demonstrated in the difference between the second and third filter, was not very high given the amount of gas adsorption.It was previously reported that the INF sampler and Nano-MOUDI perform comparably (Kim et al., 2010).Consequently, the amount of gas adsorbed on the second and third filter was assumed to be same, and the positive artifact due to gas adsorption on the first filter was calculated from the second or the third filter to the first filter.The maximum value of the gas adsorption ratio on the first filter was evaluated using the OC concentration on the second filter, including the OC component from the first filter.On the other hand, the minimum value of the gas adsorption ratio, taken as being due to only gas adsorption, was evaluated using the OC concentration on the third filter.These maximum and minimum values of the gas adsorption ratio in each OC fraction showed a similar change.The average values of the gas adsorption ratio of the total OC components collected on the roof during summer and winter were 33.0 ± 11.7% and 27.8 ± 12.0%, respectively (Figs. 3-5).Samples collected using MOUDI (Cheng et al., 2010) did not show as pronounced a seasonal variation in gas adsorption as observed here using the INF sampler, indicating that the effect of gas adsorption on mass concentration was equal for the UFPs collection even when atmospheric concentrations and situations were different.About 40% of the OC2 and OC3 fractions comprised gaseous components whereas OC1 and OC4 were essentially not adsorbed onto the filters, and about 90% of the OC gas adsorption was due to OC2 and OC3.A previous study of positive artifacts used an URG sampler to study gas adsorption following the collection of    PM 2.5 and PM 1.0 and showed that the most abundant species in the adsorbed gas were likewise OC2 and OC3 and accounted for about 80% of the OC gas adsorption (Cheng et al., 2010).Those OC2 and OC3 fractions comprised semi-volatile compounds such as carboxylic acids, hopanes, and alkanes and their adsorption could become large because of the low molecular weight species (Sihabut et al., 2005).Also, lowboiling components such as OC1 absorb little onto the filters because of the high flow rate and the slight vacuum effect by in the INF sampler.
The adsorption of gaseous carbon components near the road in spring is shown in Fig. 5.The effect of volatilization from the second filter to the third filter was small compared to the results for samples collected on the roof in winter, as shown in Fig. 6.The gas adsorption ratio near the road in spring was 30.7 ± 1.19% (Fig. 5).Comparison of gas adsorption near the road and on the roof (Figs.3-5) shows that 96.1% of high gas adsorption ratio of OC1 was confirmed near the road whereas the ratios of the other OC fractions were similar, whether the samples were collected on the roof or near the road.We had anticipated that there would be more gaseous components near the roadside because of the influence of the direct emission sources, and OC1 was adsorbed on filters without volatilization.The precise evaluation of OC1 concentration requires confirmation of the OC1 concentration when sampling near polluted locations.However, gas adsorption by OC1 contributed little to the total OC concentration because OC1 has a very low mass concentration.Therefore, the gas adsorption ratios of the total OC sampled at different locations and time of year showed a similar tendency (33.0 ± 11.7% and 27.8 ± 12.0% on the roof in summer and winter, respectively, and 30.7 ± 1.19% near the road).
A positive artifact in gas adsorption was observed when a MOUDI sampler was used to collect PM 0.1 near a roadside and in a tunnel; the values were similar and ranged between 20-50%, and the gas adsorption effect on the filters were altered by the ambient conditions of the gas and the particles (Chen et al., 2010;Zhu et al., 2012).Therefore, the positive artifact due to gas adsorption when using an INF sampler demonstrates a performance equal to conventional commercial samplers, despite the high flow rate in the INF sampler during collection.The passage of particles and the gas adsorption must be evaluated for accurate sampling of UFPs under diverse conditions.This method, which uses three quartz fiber filters, could be a simple and useful technique for that evaluation.

Ionic Components
The average concentration of water-soluble ionic species in samples collected in winter on the roof and near the road in spring are shown in Figs.7 and 8, respectively.The ion equivalent ratios between NH 4 + and NO 3 -on the second and third filter at each sampling site are shown in Figs. 9 and 10, respectively.Comparison of the carbon components (Fig. 4) and ion components (Fig. 7) shows a slightly larger difference in ion concentration between the second and third filters.The concentration of ionic components derived from inorganic gaseous components in these samples is much lower than the concentration of OC components in the atmosphere, suggesting that a large part of the inorganic gases adsorbed onto the first and second filters.However, since Fig. 7.The average concentration of ionic components in UFPs and from gas adsorption of inorganic gases in samples collected on the roof in winter.the concentrations on the second and third filters were sufficiently determined for evaluating artifacts, the maximum and minimum values of the adsorption ratio for ionic components derived from inorganic gaseous components were investigated, as were the carbon components.
For samples collected on the roof, the adsorption ratio of the ionic components derived from inorganic gaseous components on the first filter to each total ionic component was 8.5% for NH 4 + and 23.8% for 3 -.There was no effect of gas adsorption for SO 4 2-, due to the amount of atmospheric gas components (Kim et al., 2010).Environmental data provided by Saitama City indicate that the average concentration of inorganic gaseous components during the sampling period was 0.035 ppm for NO x and 0.002 ppm for SO 2 (Air Environmental Division of Saitama Environmental Department, 2011).Ammonia gas in the atmosphere reacts preferentially with acid to form stable ammonium salt particles (Chang et al., 2000).As noted above, the concentrations of NH 4 + , NO 3 -and SO 4 2-were low.A positive artifact due to gaseous components could result from the reaction of NH 3 and HNO 3 (NH 3 + HNO 3 → NH 4 NO 3 ).We therefore investigated the ion equivalent ratio between NH 4 + and NO 3 -on the roof and the result is shown in Fig. 9.The data indicate that the ionic components were completely neutralized and that some gaseous HNO 3 and some NH 3 were converted into NH 4 NO 3 particles.The ion equivalent ratio between NH 4 + and NO 3 -on the second and third filters of samples collected on the roof indicated approximately equal amounts of NH 4 + and NO 3 -on the two filters.
The inorganic gas adsorption ratio of samples collected near the road in spring were 23.6% for NH 4 + , 67.4% for NO 3 -and 3.1% for SO 4 2-; these values are higher than those of samples collected on the roof in winter (Figs. 7  and 8).This means the influence of direct sources near the roadside was similar to the case of OC1.Although (NH 4 ) 2 SO 4 was possibly produced by reaction between gaseous SO 2 and NH 3 on the second and third filters, the concentration  of SO 4 2-was much lower than the concentrations of other components, and thus the main reaction occurring near the road was between HNO 3 and NH 3 .The ion equivalent ratios between NH 4 + and NO 3 -near the road were investigated; the results are shown in Fig. 10 and confirm that ammonium nitrate was formed by adsorption of HNO 3 and NH 3 gas on the filters, similar to the case of samples collected on the roof.
We concluded that gas adsorption effects cause a large positive artifact at the sampling location where direct sources had a strong effect.It has been suggested that this problem can be solved by using a denuder system to remove gaseous components; this will not cause loss of UFPs and will allow the accurate determination of UFP concentrations (Sekiguchi et al., 2009).In addition, it should be considered that this study was carried out in winter of Japan at the limited sampling point with low SO 2 gas concentration, and the influence of volatilization of gas components is also low due to low ambient temperature and humidity.

Evaluation of the Collection Efficiency of a PTFE Filter
The results of the collection efficiency of each PTFE filter for UFPs are shown in Table 2 and SEM images of the structure of each filter and of the UFPs on the PTEF filters are shown in Fig. 11.The M P filter showed the highest collection efficiency of the three filters, and the collection efficiency of the S (sheet) filter was lower than that of the M R and M P membrane filters (Table 2).Therefore, UFPs on the S filter were not observed clearly by SEM (Fig. 11(d)).It has been reported that fine particles and UFPs may undergo a diffusion process during collection whereas larger particles are mainly collected by both inertial and diffusion processes (Furuuchi et al., 2010).A quartz fiber filter consists of fine fibers and has low porosity (Figs. 11(a) and 11(b)) and is therefore ideal for collecting UFPs by a diffusion process around the fibers.In contrast, some UFPs pass through the PTFE filters (Table 2), probably due to the higher porosity of a PTFE filter compared to a quartz fiber filter (Fig. 11).confirmed that a sheet S filter is not a fibrous structure and was inappropriate for the collection of UFPs.However, the M R and M P membrane filters have a fibrous structure, allowing diffusion processes and the collection of a relatively large number of UFPs.Since the M P filter is thicker and has lower porosity (Figs.11(e)-11(h)) than the M R filter (Table 1), the M P filter could collect UFPs more effectively due to a higher diffusion process.However, an area of the MP filter with aggregated UFPs was observed, as shown in Fig. 11(h), indicating that UFPs were also collected by an inertial process on the same M P filter.These sample collection experiments using an INF sampler confirmed that the M P filter is the most suitable for the collection of UFPs because it collects UFPs effectively by diffusion and inertial processes.The performance of the M P filter was confirmed by collecting UFPs using the INF sampler for 24 h near the road.This experiment confirmed that the collection efficiency of the M P filter was 78%, the same as when UFPs were collected on the roof.When two M P filters were installed in two stages of the INF sampler, the collection efficiency for UFPs increased to almost 100%.The results indicate that the choice of a suitable PTFE filter and its stacking are important to decrease negative artifacts and allow adequate collection of UFPs at sampling sites with high particulate concentrations.

CONCLUSIONS
Two artifacts encountered during the collection of UFPs, a positive artifact due to gas adsorption and a negative artifact due to the passage of UFPs through the filters, were investigated using INF samplers under conditions of high flow rate.These artifacts were evaluated by analyzing the organic and ionic components of the UFPs.The results of this work are summarized as follows: 1.The positive artifact for the total OC gas adsorption onto a quartz fiber filter during the sampling of UFPs using an INF sampler at collection points in this study was about 30%, comparable to that of conventional samplers.2. The results of ionic components measured from the filters installed each three stage suggested that NH 4 NO 3 particles were potentiality formed by the reaction of atmospheric NH 3 and HNO 3 gaseous components.3. The gas adsorption effect of OC1 and inorganic gases increased at the sampling location where direct sources had a strong effect, so it is necessary to reduce gas adsorption by, for example, using a denuder system.4. It is important to use the appropriate PTFE filter for the collection of UFPs.The best filter was identified by quantifying the concentration of the components of the UFPs.This is especially important when sampling areas with high particulate concentrations.5.The choice of a suitable PTFE filter and its stacking are important to collect sufficient quantities of UFPs.However, the number of filters must be limited to prevent pressure loss.6.The evaluation of positive artifacts using three quartz fiber filters and negative artifacts using a PTFE filter with a quartz fiber filter is a simple and useful technique.This method could evaluate not only gas adsorption onto the filter, but also the particles passing through the filter during the collection of UFPs.

Fig. 3 .
Fig. 3.The average concentration and the gas adsorption ratio of carbonaceous components in samples collected on the roof in summer.

Fig. 4 .
Fig. 4. The average concentration and the gas adsorption ratio of carbonaceous components in samples collected on the roof in winter.

Fig. 5 .
Fig. 5.The average concentration and the gas adsorption ratio of carbonaceous components in samples collected near the road in spring.

Fig. 6 .
Fig. 6.Correlation between OC concentration on the second filter and third filter for each sample collection period.

Fig. 8 .
Fig. 8.The average concentration of ionic components in UFPs and from gas adsorption of inorganic gases in samples collected near the road in spring.

Fig. 9 .
Fig. 9. Equivalent ratios between NH 4 + and NO 3 -for the second filter (left) and the third filter (right) of samples collected on the roof in winter.

Fig. 10 .
Fig. 10.Equivalent ratios between NH 4 + and NO 3 -for the second filter (left) and the third filter (right) samples collected near the road in spring.

Fig. 11 .
Fig. 11.SEM images of UFPs collected on the each filter.UFPs are indicated by circles in the images.
The effect of an inertial process likely increases during high flow rate collection, and UFPs passed through the PTFE filters of the INF sampler.The images shown in Figs.11(c)-11(f)

Table 1 .
Characteristics of each PTFE filter.

Table 2 .
The efficiency of each PTFE filter in an INF sampler for collecting UFPs.Samples were collected on a rooftop approximately 40 m (10 stories) above the ground.