Sampling efficiency of low-volume PM₁₀ inlets with different impaction substrates

Abstract

The louvered 16.7 L min⁻¹ PM₁₀ inlet is commonly used in PM₁₀ and PM_2.5 FRM samplers or FEM monitors. Its sampling efficiency is influenced by particle bounce, re-entrainment, and overloading since the PM₁₀ inlet contains a PM₁₀ impactor with an uncoated impaction surface. In this study, a modified PM₁₀ (M-PM₁₀) inlet with an oil-soaked glass fiber filter substrate supported by an oil-soaked porous metal disc was developed to eliminate the particle bounce and overloading effects. The oiled M-PM₁₀ inlet and the traditional PM₁₀ inlets with and without grease coating were collocated at the field for long-term comparison tests. The results show that the traditional uncoated PM₁₀ inlet which is cleaned initially but not cleaned daily afterwards oversamples PM₁₀ concentration after one 24-h sampling day and has the high positive average sampling bias during 14 sampling days due to particle bounce and re-entrainment. The grease-coated PM₁₀ inlet without daily cleaning shows a better performance with a smaller sampling bias, but it still oversamples PM₁₀ concentrations after the first three 24-h sampling days and then undersamples after 10 sampling days due to particle bounce and overloading effects, respectively. In comparison, the M-PM₁₀ inlet shows a good performance with a small average sampling bias during 35 sampling days since vacuum oil wicks up through the deposit to eliminate particle bounce and overloading. It is suggested that the oiled M-PM₁₀ inlet can be used to replace the traditional EPA PM₁₀ inlet and for long-term sampling of over 1 month without the frequent maintenance need.

Copyright © 2019 American Association for Aerosol Research

1. Introduction

PM₁₀ and PM_2.5 (particles with aerodynamic diameter ≤10 and ≤2.5 µm, respectively) are among regulated air quality metrics because they can deposit in the tracheobronchial and alveolar regions and pose health risks to humans (Rückerl et al. 2011). To determine the compliance with the national air quality standard, PM₁₀ and PM_2.5 concentrations are measured by using federal reference method (FRM) samplers or federal equivalent method (FEM) monitors, each of which consists of a size-selective inlet followed normally by a filter to collect particles for further analysis (Chow and Watson 2008). The low-volume PM₁₀ inlet with the flow rate of 16.7 L min⁻¹ is one of the most widely deployed inlets in FRM and FEM devices for aspirating and classifying particles smaller than 10 µm in aerodynamic diameter (US EPA 2017). It is because that the low-volume PM₁₀ inlet is low cost, light weight, small footprint, and portability (Chow and Watson 2008). The low-volume PM₁₀ inlet is first developed by McFarland and Ortiz (1984), which is called “flat-topped” PM₁₀ inlet, and subsequently modified by Tolocka et al. (2001), which is called “louvered” PM₁₀ inlet, to become an EPA-approved (Environmental Protection Agency) PM₁₀ inlet. The louvered PM₁₀ inlet, henceforth referred to as “PM₁₀ inlet,” is designed to have the cut-size (d_pa50, aerodynamic diameter corresponding to 50% of the sampling efficiency) of 10 ± 0.5 µm and the sharpness of the sampling efficiency curve (calculated as the square root of d_pa16/d_pa84) of 1.5 ± 0.1 (Wang et al. 2005; US EPA 2017). The sampling efficiency of the PM₁₀ inlet is determined mainly by the classification ability of its lower part, a single-nozzle PM₁₀ impactor. While its upper part, an aspirator, helps aspirate the representative aerosol sample in the ambient air and remove some insects, rain, snow, airborne debris, water droplets and very-large particles to protect the impactor from high mass loading and severe ambient conditions (Liu and Pui 1981; McFarland and Ortiz 1984; Tolocka et al. 2001; Vanderpool et al. 2018).

The performance characteristic of the PM₁₀ inlet should match with the requirements of 40 CFR Part 53 (US EPA 2017). However, several studies found that the PM₁₀ inlet has uncertainty in the measured sampling efficiency. Buser et al. (2008) indicated that the PM₁₀ inlet oversampled PM₁₀ concentration as compared to the highest theoretical PM₁₀ concentration which was the product of the particle size distribution (PSD) and the characteristic of PM₁₀ inlet with the cut-size of 10.5 µm and the curve sharpness of 1.6. It is because that the PM₁₀ inlet was used to sample the agricultural particles (cotton gins) which had a high fraction of large particles. Wang et al. (2005) also revealed that the sampling performance of the PM₁₀ inlet varied with the various particle sources. Similarly, Park et al. (2009) found that when the sampling source was fly ash which represented the industrial particles the measured PM₁₀ concentrations of the PM₁₀ inlet were higher than those of the Andersen cascade impactor, but no explanation was given. Faulkner, Smith, and Haglund (2014) explained that a small fraction of particles larger than the cut-size in the measured PM₁₀ concentrations could cause a significant positive error when large particles were dominant in sampling particle sources. Watson et al. (2011) showed that the BGI PQ200 sampler equipped with the PM₁₀ inlet was equivalent to a high-volume (Hi-Vol) PM₁₀ sampler when sampling ambient aerosols. However, the PM₁₀ inlet undersampled PM₁₀ concentrations after sampling for a few days at the source-dominated environment with a high TSP/PM₁₀ ratio (total suspended particulates).

The overestimation and underestimation of the PM₁₀ inlet could be due to particle bounce and re-entrainment, and particle overloading of the PM₁₀ impactor (Wang et al. 2005; Buser et al. 2008; Zhao et al. 2009). The PM₁₀ inlet contains a PM₁₀ impactor with an uncoated surface which may not absorb the kinetic energy of incoming particles completely resulting in the particle bounce of large particles (John, Winklmayr, and Wang 1991; John and Wang 1991; Tsai and Cheng 1995; Peters, Vanderpool, and Wiener 2001). Moreover, if the uncoated impaction surface is not cleaned regularly, the deposited particles which do not stick well on the uncoated surface can be de-agglomerated and re-entrained to the downstream leading to oversampling (Rao and Whitby 1978; John, Winklmayr, and Wang 1991). Then, if more particles are accumulated on the substrate to form a particle mound, the cut-size of the impactor is reduced due to the particle overloading effect resulting in undersampling. These effects observed in the PM_2.5 well impactor ninety-six (WINS, Vanderpool et al. 2001; Le and Tsai 2017) are also expected to occur in the PM₁₀ inlet which contains the PM₁₀ impactor. Although the likelihood of the impaction-well like design helps the PM₁₀ impactor eliminate the initial particle bounce in the wind tunnel study with monodisperse particles (Kim, Kim, and Lee 1998; Tolocka et al. 2001), the particle bounce and loading effects of the PM₁₀ inlet have never been addressed and tested fully in the field study. Moreover, it is necessary to determine the routine cleaning frequency for the PM₁₀ inlet to maintain its good sampling performance. But for the long-term sampling need, it takes lots of man power if the cleaning frequency is too high. In addition, disassembling the PM₁₀ inlet for cleaning is hard since the inlet cylinder is big (diameter = 69.3 mm) with a very tight screw junction. Therefore, for the long-term monitoring and sampling need, it is desirable that the PM₁₀ inlet should not be cleaned too frequently.

Some impactors are designed to overcome the particle bounce and overloading effects for unattended, continuous operation. The application of viscous substrate coating such as grease on the impaction surface of an inertial impactor has been studied earlier (Rao and Whitby 1978; Turner and Hering 1987). Since the impaction surface is coated with a thin layer of grease, the impactor can eliminate the initial particle bounce successfully, but the particle bounce still occurs when the particles impact on the first several layers of the deposited particles formed on the substrate after a few sampling days (John, Winklmayr, and Wang 1991; John and Wang 1991; Tsai and Cheng 1995; Tsai and Lin 2000; Zhao et al. 2009). The impactor with the oil-soaked impaction substrate is widely used to replace the impactor with the grease-coated substrate since vacuum oil with low viscosity (30–300 mm² s⁻¹) can wick up through the deposit to capture particles more readily (Turner and Hering 1987; Peters, Vanderpool, and Wiener 2001; Tsai et al. 2012). Filter mats, such as membrane filter and glass fiber filter (GFF), are usually used as the substrate to retain vacuum oil. Based on the design of inverted conical cavity of Tsai and Cheng (1995), an impaction well is used to contain a given amount of oil to prevent the oil from drying (Peters, Vanderpool, and Wiener 2001). Based on the design of the well impactor of Peters, Vanderpool, and Wiener (2001), Le and Tsai (2017) developed a non-bouncing PM_2.5 impactor in which a small amount of water is injected upward to wash the deposited particles off the impaction surface to avoid particle overloading. The porous metal substrate is used to reduce particle bounce and increase the particle loading capacity, but the cut-size is decreased, and the curve is less sharp due to air penetration into the pores of the porous metal disc (Huang, Tsai, and Shih 2001; Huang et al. 2005; Huang and Tsai 2003; Marjamäki and Keskinen 2004). If the pores are filled with vacuum oil, the cut-size and the curve sharpness may not be changed, but these remain to be studied. In summary, if the impaction surface is designed in such a way that the kinetic energy of the incident particle can be absorbed, the particle bounce effect can be minimized. In addition, the ability to capture more particles without particle overloading effect is likely if a particle mound does not form.

The accuracy of the PM₁₀ inlet is also affected by the performance of the aspiration section of the PM₁₀ inlet as shown in previous studies (Liu and Pui 1981; Kenny et al. 2005; Vanderpool et al. 2018). Kenny et al. (2005) revealed that aspirated particles were lost on the louvered plate of the aspirator (particle settling, inertial deposition, and diffusion) which varied from 58% to 92% depending on the mass median aerodynamic diameter (MMAD, from 13 to 46 µm) but was independent of the wind speeds (1.8 km h⁻¹ and 3.6 km h⁻¹). Vanderpool et al. (2018) reported that the performance of the aspirator was affected by the wind speeds (2, 8, and 24 km h⁻¹) and particle sizes. The results of the wind tunnel test showed that the sampling effectiveness of the aspirator was increased when the wind speed was increased from 2 to 8 km h⁻¹. The situation at the wind speed of 24 km h⁻¹ was complicated since the sampling effectiveness was higher than 100% when particles were smaller than 14 µm, but it became lower than that at 8 km h⁻¹ for particles ranging from 14 to 30 µm. Tolocka et al. (2001) tested the PM₁₀ inlet with the aspirator and the impactor and showed that the cut-sizes (9.9, 10.3, and 9.9 µm) were not different regardless of the wind speeds of 2, 8, and 24 km h⁻¹. However, the sampling efficiency curve became sharper if the wind speed was higher than 8 km h⁻¹ (sharpness = 1.4, 1.38, and 1.31 corresponding 2, 8, and 24 km h⁻¹, respectively). Liu and Pui (1981) also showed that the sampling characteristic of the PM₁₀ inlet was independent of the wind speed up to 9 km h⁻¹.

To decrease particle bounce and increase particle loading capacity, this study developed a modified PM₁₀ (M-PM₁₀) inlet in which an oil-soaked GFF substrate supported by an oil-soaked porous metal disc (PMD) is used to replace the uncoated impaction surface used in the traditional PM₁₀ inlet. The uncoated PM₁₀ inlet which was cleaned initially but not cleaned daily afterwards (i.e., uncleaned, uncoated inlet) was first tested at the field for examining the particle bounce and re-entrainment effects for 14 days (period #1) as compared to the reference PM₁₀ inlet which was cleaned daily and coated with grease (cleaned, greased inlet). The traditional uncleaned, greased inlet was then tested for the particle bounce and overloading effects for 16 days (period #2) and 23 days (period #3), respectively. Finally, the M-PM₁₀ inlet was tested to evaluate the sampling performance for 23 days (period #3) and 35 days (period #4).

2. Experimental method

Figure 1 shows the schematic diagram of the modified PM₁₀ (M-PM₁₀) impactor with an oil-soaked GFF substrate supported by a PMD. A circular cavity (diameter (D) × depth (H) = 25 × 4.5 mm or 25 × 3.5 mm) is drilled at the center of the impaction plate to contain the 25-mm PMD and the GFF. A small amount of vacuum oil of 1.3 mL (TT-DM-100, Lisons Chemical Co., Ltd., Taiwan) for the cavity depth of 4.5 mm is added to fill up the cavity. The oil-soaked GFF substrate is expected to overcome particle bounce and capture more particles as the oil with low viscosity (∼100 mm² s⁻¹) can be retained on the GFF against air flow very well and wick up through the particle deposit to maintain a viscous impaction surface. The PMD is used to enhance the capillary action and increase the particle loading on the substrate. The M-PM₁₀ impactor with the new impaction substrate as well as the conventional PM₁₀ impactors were first tested at the laboratory for determining the cut-size and the penetration curve sharpness and then at the field after assembling with the aspirator section for evaluating the sampling performance.

Figure 1. Schematic diagram of the M-PM₁₀ impactor.

2.1. Laboratory tests

The schematic diagram of the experimental setup for determining the penetration curve of the PM₁₀ impactor is shown in Figure 2. In this study, the penetration curves of the impactors were then combined with the sampling effectiveness of the aspirator from the wind tunnel test resulted by Vanderpool et al. (2018) to obtain the sampling efficiency curves of the PM₁₀ inlet whose estimated biases were then calculated.

Figure 2. Experimental setup of the particle penetration test.

The detailed experimental method is described in Section A1 of the online supplemental information (SI). Only the impactor section (the PM₁₀ inlet without the aspirator) was tested, which was a static test instead of the wind tunnel test which is required for testing the PM₁₀ inlet (US EPA 2017). Briefly, the test was carried out using a Vibrating Orifice Aerosol Generator (VOAG, TSI Model 3450, USA) to generate monodisperse solid ammonium fluorescein (AF; the particle density is 1.35 g cm⁻³) particles with 19 discrete particle sizes ranging from 5.0 to 19.8 µm (Vanderpool and Rubow 1988). An aerodynamic particle sizer (APS, TSI model 3321, USA) was used to measure the particle number concentration passing through a metal tube (upstream) and the impactor (downstream). The penetration (P, %) of the impactor for each particle aerodynamic diameter is calculated as (1) $$P\mathrm{ }\left(\mathrm{\%}\right)=\mathrm{ }\frac{\mathrm{ }{C}_{\text{down}}}{C_{\text{up}}}\mathrm{ }\times 100,$$(1) where C_up and C_down are the upstream and downstream aerosol number concentrations (# cm⁻³), respectively. The discrete penetration data were fitted by using Equation (2)(2) $$P\left(\mathrm{\%}\right)=a+\mathrm{ }b\left[1-{\left[1+\exp \left(\frac{\left(d_{\text{pa}}+d\mathrm{l}\mathrm{n}\left(2^{1/e}-1\right)-c\right)}{d}\right)\right]}^{-e}\right]\mathrm{ }\times 100,$$(2) to determine the d_pa50, d_pa16, and d_pa84 to calculate the sharpness of the penetration curve (Peters, Vanderpool, and Wiener 2001; Le and Tsai 2017) as (2) $$P\left(\mathrm{\%}\right)=a+\mathrm{ }b\left[1-{\left[1+\exp \left(\frac{\left(d_{\text{pa}}+d\mathrm{l}\mathrm{n}\left(2^{1/e}-1\right)-c\right)}{d}\right)\right]}^{-e}\right]\mathrm{ }\times 100,$$(2) where a, b, c, d, and e are fitting parameters. The sampling efficiency (η, %) of the PM₁₀ inlet is calculated as follows: (3) $$\eta \mathrm{ }\left(\mathrm{\%}\right)=\mathrm{ }P{E}_{\mathrm{a}},$$(3) where E_a (%) is the sampling effectiveness of the aspirator obtained from the previous study by Vanderpool et al. (2018). The estimated PM₁₀ concentrations of the PM₁₀ inlet were then calculated to determine the estimated biases based on the ideal particle distributions (US EPA 2017) (see Section A1 of the SI).

2.2. Field comparison tests

The field comparison tests of the PM₁₀ samplers equipped with different PM₁₀ inlets were conducted at National Chiao Tung University (NCTU) from November 2017 to October 2018. The sampling station was located on the 6th floor of the building of the Institute of Environmental Engineering (IEV) at NCTU which is about 1 km away from the roads. Samplers were set up at the same height (2 m from the platform of the station), and all samplers were spaced less than 2 m from each other. During the sampling period, temperature (T), relative humidity (RH), and wind speed (WS) ranged from 13 to 33 °C, 57 to 88%, and 2.4 to 15.2 km h⁻¹ (0.67 to 4.2 m s⁻¹), respectively.

Four PM₁₀ samplers each of which contains a filter cassette with a 37 mm Teflon filter (Teflo R2PL037, Pall Corp., New York, USA) were collocated. The sampling flow rate of each of the four samplers was controlled by a mass flow controller (MFC). Four PM₁₀ inlets are: one standard PM₁₀ inlet (Model 20XXA457780712, Thermo Sci., USA) and three other PM₁₀ inlets manufactured in China (Model B2151250015, Focused Photonics Inc., China). One of the Chinese PM₁₀ inlets was modified to be the M-PM₁₀ inlet with the oiled impaction substrate while the other three PM₁₀ inlets remained to have the traditional flat impaction surface. The filters were conditioned before and after sampling with the procedure described in Section A2 of the SI. Four cleaned PM₁₀ inlets which were an oiled M-PM₁₀ inlet and three conventional PM₁₀ inlets coated with grease to eliminate the initial particle bounce were first tested for five 24-h sampling days to evaluate the sampling performance of the Chinese PM₁₀ inlets based on the reference EPA PM₁₀ inlet.

Then, based on the traditional PM₁₀ inlet with grease coating and daily cleaning (i.e., cleaned, greased inlet), the traditional PM₁₀ inlet which was cleaned initially but not cleaned daily afterwards, and not coated with grease (i.e., uncleaned, uncoated inlet) was tested in period #1 for 14 days to examine the particle bounce and re-entrainment effects. Next, the traditional inlet which was coated with a thin layer of grease and not cleaned daily (i.e., uncleaned, greased PM₁₀ inlet) was tested in 2 periods (period #2: 16 days and period #3: 23 days) to evaluate the particle bounce and overloading effects. In period #2, the performance of the uncleaned, greased PM₁₀ inlet was evaluated based on the cleaned, greased PM₁₀ inlet as the reference for 16 sampling days. In period #3, the uncleaned, greased PM₁₀ inlet was tested for a longer period of 23 days at higher PM₁₀ concentrations. Meanwhile, the prototype of the oiled M-PM₁₀ inlet (cavity depth H = 3.5 mm, oil amount = 1.0 mL) was also tested in period #3 for 23 days based on the cleaned, greased inlet as the reference to evaluate the sampling performance. Since the oil of the M-PM₁₀ inlet was found to be blown off the GFF after 6 days during period #3, the M-PM₁₀ inlet with a deeper cavity (H = 4.5 mm, oil amount = 1.3 mL) was tested over a longer period of 35 days (period #4) to examine if the oil blowing off problem found in period #3 was resolved. In period #4, the PM₁₀ concentrations of the 24-h samples were measured daily in the first 5 days, then the sampling pumps were kept running for sampling but without sampling filters for 4–7 days before measuring 24-h average PM₁₀ concentration again for 1–3 days. During the non-sampling period of 4–7 days, particles larger than PM₁₀ were loaded on the impactor of the PM₁₀ inlet continuously to increase the particle loaded mass while PM₁₀ particles were filtered by a HEPA filter before the sampling pump.

After the tests of period #2 and #3 were over, the uncleaned PM₁₀ inlets (greased and oiled) were tested for the cut-size and the penetration curve in the laboratory following the method described in the previous section. Both M-PM₁₀ (H = 4.5 mm) and uncleaned, greased inlets were also tested every 6 h, twice every day, and continued for 8 days which is equivalent to four 24-h sampling days. The purpose of this test is to find out exactly when particle bounce starts to occur in the uncleaned, greased inlet. The results of this test are shown in details in Section B3 of the SI.

3. Results and discussion

3.1. Laboratory results of the sampling efficiency of cleaned PM₁₀ inlets

The PM₁₀ impactor parts of three PM₁₀ inlets with different impaction substrates (traditional uncoated, traditional greased, and oiled M-PM₁₀ impactors) were tested for the penetration curve in the laboratory. The sampling efficiencies of the PM₁₀ inlets were then calculated using Equation (3)(3) $$\eta \mathrm{ }\left(\mathrm{\%}\right)=\mathrm{ }P{E}_{\mathrm{a}},$$(3) and compared to those of the traditional uncoated PM₁₀ inlet in Tolocka et al. (2001). Figure 3 shows the particle penetrations of the three cleaned PM₁₀ impactors (uncoated, greased, and oiled). The results show that the cut-size of the oiled impactor (d_pa50 = 10.42 µm) is not different from those of the greased impactor (d_pa50 = 10.44 µm) and the uncoated impactor (d_pa50 = 10.3 µm), while the penetration curve of the former (sharpness = 1.4) is slightly sharper than those of the later (sharpness = 1.48 for greased impactor and 1.53 for uncoated impactor). The cut-size is a function of the Stokes number (Stk) which is the ratio of the particle’s stopping distance to the characteristic dimension of the nozzle (Hinds 1999). Because the characteristic dimensions of the three impactors are identical in the present study, the cut-sizes are the same. The oiled impactor collects fewer particles smaller than the cut-size and more particles larger than the cut-size compared to the uncoated and greased impactors, leading to a sharper curve. It is because that the oiled impactor can capture large particles effectively. On the other hand, the uncoated impactor shows a smoother curve than those of the oiled and greased impactors due to higher penetrations of large particles. It indicates that the uncoated impactor may have the particle bounce effect, but the likelihood of the impaction-well like design helps to mitigate this effect resulting in the gradual decrease in the penetrations with the increased particle sizes. However, the estimated biases in PM₁₀ concentrations of the oiled impactor (fine: −2.51%, typical: −0.7%, and coarse: +0.69%) and the uncoated impactor (fine: −0.35, typical: 0.43%, and coarse: +1.09%) are very small for three idealized distributions based on the greased impactor as the reference. It indicates that the difference in the curve sharpness of these impactors does not contribute to the sampling error between the three PM₁₀ inlets with different impaction substrates.

Figure 3. Particle penetration curve of the cleaned PM₁₀ impactors with different impaction substrates.

The M-PM₁₀ impactors with the other two types of oiled substrates which are oil-soaked GFF supported by a solid metal disc (oiled + SMD), and oil-soaked GFF without the support were also studied. The results shown in Figure 3 indicate that the M-PM₁₀ impactors with different oiled substrates have similar curve sharpness (varying from 1.4 to 1.44) but different cut-sizes. The cut-size of the M-PM₁₀ impactor with the oiled substrate without the support (d_pa50 = 11.44 µm) is larger than that of the M-PM₁₀ impactor with the oiled substrates with the support (d_pa50 = 10.42 µm for oiled + PMD support and d_pa50 = 10.26 µm for oiled + SMD support) due to its larger jet-to-plate distance. The oiled + PMD impactor (H = 4.5 mm) can contain 1.3-mL oil soaked in the PMD and GFF, while the oiled + SMD impactor can contain 0.3-mL soaked in the GFF only. Therefore, the PMD is chosen to be the support for the oil-soaked GFF substrate in the M-PM₁₀ impactor to retain as much vacuum oil as possible.

The cut-sizes and the sharpnesses of the sampling efficiency curves of the three PM₁₀ inlets at different wind speeds calculated by using Equation (3)(3) $$\eta \mathrm{ }\left(\mathrm{\%}\right)=\mathrm{ }P{E}_{\mathrm{a}},$$(3) are shown in Table 1. The results show that only the calculated cut-sizes of the three PM₁₀ inlets at the wind speeds of 2 and 8 km h⁻¹ and the curve sharpnesses at the wind speeds of 2 and 24 km h⁻¹ are very close to those of the uncoated PM₁₀ inlet in the previous study (Tolocka et al. 2001). The comparison is discussed in details in Section B1 of the SI. It indicates that the calculated sampling efficiency is in a good agreement with the measured sampling efficiency at the low wind speed of 2 km h⁻¹. The interaction of the aspirator and the PM₁₀ impactor may be complex at higher wind speeds, 8 and 24 km h⁻¹.

Table 1. Cut-size and curve sharpness of PM₁₀ inlets.

			PM10 inlet
		PM10 impactor	2 km h^−1	8 km h^−1	24 km h^−1
dpa50 (µm)	Oiled substrate (present^a)	10.42	9.48	10.42	10.94
	Greased substrate (present^a)	10.44	9.4	10.44	11.0
	Uncoated surface (present^a)	10.3	9.36	10.37	10.99
	Uncoated surface (Tolocka et al. 2001)		9.9	10.3	9.7
Sharpness	Oiled substrate (present^a)	1.4	1.4	1.41	1.3
	Greased substrate (present^a)	1.48	1.48	1.5	1.34
	Uncoated surface (present^a)	1.53	1.47	1.5	1.31
	Uncoated surface (Tolocka et al. 2001)		1.4	1.38	1.31

^aEquation (3)(3) $$\eta \mathrm{ }\left(\mathrm{\%}\right)=\mathrm{ }P{E}_{\mathrm{a}},$$(3) was used to calculate the sampling efficiency of the PM₁₀ inlet.

3.2. Field comparison results of the PM₁₀ inlets with different impaction substrates

After testing the penetration curves of the PM₁₀ impactors in the laboratory for calculating the sampling efficiencies of the PM₁₀ inlets with different impaction substrates, the field comparison tests were conducted. Figure 4 shows that the linear relationship of four cleaned PM₁₀ inlets (a reference US EPA greased PM₁₀ inlet, two Chinese greased PM₁₀ inlets, and an oiled M-PM₁₀ inlet) when the range of measured PM₁₀ concentrations is wide (50–126.8 µg m⁻³). The measured PM₁₀ concentrations of four PM₁₀ inlets agree well with each other with very close slopes, intercepts and R² as shown in Figure 4, which meet the US EPA requirements of slope = 1.0 ± 0.1, intercept = 0.0 ± 5.0 µg m⁻³, and R² > 0.97 (US EPA 2017). The average sampling bias of the three Chinese inlets shown in Figure S2 of the SI is very small (-0.7 ± 3.2% for the 1st greased inlet, 2.3 ± 2.3% for the 2nd greased inlet, and 1.8 ± 2.2% for the oiled inlet) based on the reference inlet. These cleaned Chinese PM₁₀ inlets and the M-PM₁₀ inlet are identical to the US EPA inlet in measured PM₁₀ concentrations.

Figure 4. Field comparison results of the cleaned PM₁₀ inlets.

The traditional uncleaned, uncoated PM₁₀ inlet (cleaned initially but not cleaned daily afterwards) which was collocated with the cleaned, greased inlet as the reference was first tested in the field in period #1 for 14 days. In Figure 5, it can be seen that the uncoated PM₁₀ inlet oversamples PM₁₀ concentration during 14 sampling days with PM₁₀ concentrations (cleaned, greased) ranging from 8.06 to 25.73 µg m⁻³. The positive average sampling bias of 14.58 ± 21.25% does not meet the US EPA requirement with the bias smaller than ±10% (US EPA 2017) indicating that the daily uncleaned, uncoated inlet has the particle bounce effect. It is observed that the sampling bias fluctuates from the negative bias of –7.37% (the concentration difference is −2.02 µg m⁻³) at day 1 to the positive bias of +19.68% (the concentration difference is +2.57 µg m⁻³) at day 2 and attains the high values of +69.3% and +41.08% at day 7 and 12, respectively (the concentration differences are +9.53 and +8.16 µg m⁻³, respectively). It indicates that the particle bounce occurs at the second 24-h sampling day with the loaded mass of 474 µg and the particle re-entrainment causes the very high positive error at day 7 and day 12 with the loaded masses of 1,804 and 3,316 µg, respectively. Note that the loaded mass is the product of the average (TSP - PM₁₀)/PM₁₀ ratio of 0.72 ± 0.1 and measured PM₁₀ concentrations. The average (TSP - PM₁₀)/PM₁₀ ratio is calculated from the monthly TSP data collected in 2 years (2017 and 2018) obtained by Hsinchu EPA (2018). The monthly data of (TSP - PM₁₀)/PM₁₀ are shown in Table SI-1, Section B2 of the SI.

Figure 5. Sampling bias of the uncoated PM₁₀ inlet without daily cleaning in period #1.

The traditional PM₁₀ inlet was then coated with a thin layer of grease to evaluate the particle bounce and overloading effects for 16 days in period #2. Figure 6 shows the sampling bias of the greased PM₁₀ inlet which is not cleaned during sixteen 24-h sampling days based on the reference cleaned, greased PM₁₀ inlet. It is seen that the negative sampling bias varies from −0.5 to −7.4% from day 1 to day 3 with PM₁₀ concentrations (cleaned, greased) ranging from 17.1 to 19.7 µg m⁻³, and the bias becomes positive ranging from +0.8 to +7.2% in day 4 to day 16 with PM₁₀ concentrations ranging from 14.2 to 27.7 µg m⁻³. Although the sampling bias of the uncleaned, greased PM₁₀ inlet is reduced as compared to the uncleaned, uncoated inlet and meets the US EPA requirement with the bias of < ±10%, overestimation of PM₁₀ concentrations is apparent after 3 sampling days. The linear regression parameters (slope = 1.09, intercept = −1.44 µg m⁻³, and R² = 0.96) do not very good as shown in Figure S3 of the SI). T, RH, and WS did not vary too much during period #2 (T = 30.5 ± 0.7 °C, RH = 68.4 ± 5%, and WS = 6.6 ± 3 km h⁻¹, respectively), which might not be the cause of positive bias after day 3. The overestimation of PM₁₀ concentrations of the uncleaned, greased PM₁₀ inlet can well be due to the particle bounce effect. The cleaned, greased substrate seems to capture particles in the first three days effectively and then as deposited particles form several layers locally on the substrate, the likelihood of particle bounce occurs in the next 13 days. The estimated loaded masses after 3 days and 16 days are 965 and 5,979 µg, respectively ((TSP - PM₁₀)/PM₁₀ = 0.72 ± 0.1).

Figure 6. Sampling bias of the uncleaned, greased PM₁₀ inlet in period #2.

In period #3, similar trend of positive sampling bias for the uncleaned, greased PM₁₀ inlet (varying from 0 to +11.2%) is observed from day 3 to day 10 as shown in Figure 7, when the loaded particle mass reaches to 6,397 µg ((TSP - PM₁₀)/PM₁₀ = 0.72 ± 0.1) at the end of day 10. Since higher PM₁₀ concentrations (cleaned, greased) are measured in period #3 (ranging from 15.9 to 101.8 µg m⁻³), the particle bounce effect occurs after just 2 sampling days and persists for the next 8 sampling days. After that, the sampling bias of the uncleaned, greased PM₁₀ inlet suddenly drops to become negative (varying from −1.5 to −5.5%) during the last 13 sampling days. Lower T (20.3 ± 3.6 °C), higher RH (75 ± 8.3%), and similar WS (5.4 ± 3.4 km h⁻¹) were recorded during 23 sampling days in period #3 as compared to those of period #2. The statistical analysis shows that there is no clear relationship between the sampling bias and ambient conditions. The underestimation of PM₁₀ concentrations of the uncleaned, greased PM₁₀ inlet after day 10 can be explained by the particle overloading effect. The loaded mass increases from 6,397 µg at day 10 to 15,386 µg after 23 sampling days. When the substrate is overloaded, the particle collection efficiency is increased by localized particle pile-ups resulting in a smaller cut-size and a smoother sampling efficiency curve. These particle bounce and overloading effects of the uncleaned, greased PM₁₀ inlet were further examined by conducting the particle penetration test in the laboratory after period #2 and #3, and the results are shown in Section 4.

Figure 7. Sampling bias of the uncleaned, greased PM₁₀ inlet and the uncleaned M-PM₁₀ inlet in period #3.

The sampling bias of the oiled M-PM₁₀ inlet during twenty-three 24-h days in period #3 is also shown in Figure 7. It shows that in 23 days the average sampling bias of the M-PM₁₀ inlet is small (0.31 ± 4.56%), and the measured PM₁₀ concentrations of the M-PM₁₀ inlet have a very good linear relationship with those of the cleaned, greased PM₁₀ inlet with the slope, intercept, and R² of 0.97, 0.77 µg m⁻³, and 0.99, respectively (the linear regression curve is shown in Figure S3 of the SI). It is found that the positive biases of +9.85%, +8.46%, and +7.84% occurred in day 7, 8, and 10, respectively, because the oil was blown off the GFF leaving dried spots on the GFF after 6 sampling days. That is, the cavity with the depth of 3.5 mm contains 1.0 mL oil which is not enough for a long sampling period. After day 10, a small amount of oil was added to wet the substrate of the M-PM₁₀ inlet, the bias soon became small and mostly negative (average: −2.3 ± 2.3%, range: −5.7 to 1.7%) from day 11 to day 23. Generally, it can be observed that the average sampling biases of the M-PM₁₀ inlet during the first 10 days and the last 13 days are +2.9 ± 4.7% and −2.3 ± 2.3%, respectively, which are smaller than those of the uncleaned, greased PM₁₀ inlet (+3.9 ± 5.46% and −3.1 ± 1.3%, respectively). The laboratory results of the sampling efficiency of the uncleaned, oiled M-PM₁₀ inlet after 23 sampling days, which are discussed in Section 4, further shows a good performance of the M-PM₁₀ inlet as compared to the uncleaned, greased inlet.

To keep enough oil on the GFF for the long-term sampling need, the cavity depth of the M-PM₁₀ inlet was increased to 4.5 mm to contain 1.3 mL oil and form an extra 1-mm oil layer on top of the GFF. Then the M-PM₁₀ inlet was again tested for another long sampling period of 35 days in period #4. Figure 8 shows that the sampling bias of the M-PM₁₀ inlet ranges from −3.8 to +4.45% with the average of 0.27 ± 2.9%, which is very small, indicating that the M-PM₁₀ inlet had eliminated the problem of oil blowing off the GFF occurred in period #3. Note that the days without the sampling bias data are non-sampling days but with continuous particle loading as described in Section 2.2. The 1-mm oil layer is enough to keep the GFF surface wetted with oil which helps particle capture effectively and provides more particle loading capacity as compared to the greased substrate of the conventional PM₁₀ inlet. Therefore, the M-PM₁₀ inlet can be used in FRM samplers and FEM monitors for long-term sampling without the needs of frequent maintenance. Because the measured PM₁₀ concentration ranging from 13.3 to 34.6 µg m⁻³ is lower during period #4 than that during period #3, the loaded mass is 12,496 µg after 35 sampling days, which is equivalent to that of 23 sampling days in period #3.

Figure 8. Sampling bias of the uncleaned M-PM₁₀ inlet in period #4.

Because the uncleaned, greased inlet is found to have particle bounce after three 24-h sampling days in period #2 and two 24-h sampling days in period #3, it was further tested for short-term sampling of 6 h in 8 days, two tests per day, to determine the exact time or loaded particle mass beyond which particle bounce occurs. In Figure 9, it is found that the uncleaned, greased PM₁₀ inlet show a good performance with the sampling bias ranging from −10.6% to +6.5% in the first 60 sampling hours but oversamples PM₁₀ concentrations after 60 h, equivalent to 24-h sampling for 2.5 days, with the positive bias ranging from +0.37 to +20.9%. The loaded mass is 1,113 µg after 60-h sampling. In comparison, the M-PM₁₀ inlet still shows a good performance with the sampling bias ranging from −11.4 to 7.4% during overall 96 sampling hours. Therefore, the greased PM₁₀ inlet should be cleaned and re-greased after three 24-h sampling day with the moderate PM₁₀ concentrations (15.1–38 µg m⁻³). The detailed results are shown in Section B3 of the SI.

Figure 9. Six-hour sampling bias of the uncleaned, greased PM₁₀ inlet and the uncleaned M-PM₁₀ inlet.

4. Discussion

To explain the particle bounce and overloading effects of the uncleaned, greased inlet occurred in period #2 (16 days) and #3 (23 days), respectively, the penetration curves of the uncleaned, greased PM₁₀ impactor in period #2 and period #3 were obtained in the laboratory after the long-term tests. The results shown in Figure 10 indicate that the cut-size and the curve sharpness of the impactor in period #2 are 10.2 µm and 1.57, respectively. It can be seen that the impactor in period #2 has the similar cut-size but less sharp curve than the cleaned, greased impactor. The penetration of particles larger than 12.8 µm is non-zero, indicating that particle bounce had occurred in the impactor in period #2 after 16 days. Although the smoother penetration curve of the impactor in period #2 may result in lower PM₁₀ concentrations, but the bounce effect of particles larger than the cut-size leads to higher PM₁₀ concentrations in comparison to the cleaned, greased impactor. These opposite effects explain the slight positive sampling bias of the uncleaned, greased PM₁₀ inlet from day 4 to day 16 in period #2 shown in Figure 7.

Figure 10. Particle penetration curve of the uncleaned PM₁₀ impactors with different impaction substrates.

The cut-size and the curve sharpness of the impactor in period #3 are 9.77 µm and 1.61, respectively which are different from those of the cleaned, greased impactor. Obviously, the particle overloading effect had occurred in the impactor in period #3 resulting in a smaller cut-size and less sharp curve. The sampling efficiency curves of the uncleaned and cleaned inlets calculated from the penetration curves of impactors are used to calculate PM₁₀ concentrations based on the typical mass distribution measured at the NCTU station by using an NMCI (NCTU micro-orifice cascade impactor, Chien et al. 2015). The distribution data show that the MMAD and the geometric standard deviation (GSD) of the fine mode of 0.4 µm and 1.6, respectively, and the MMAD and the GSD of the coarse mode of 6 µm and 1.9, respectively. The estimated bias of the inlet in period #3 is calculated to be −7.6% based on the daily cleaned, greased inlet as the reference indicating that the greased PM₁₀ inlet not cleaned during a long sampling period undersamples PM₁₀ concentrations because of the particle loading effect. After the particle loading effect occurred from day 11 to day 23 in period #3, the sampling bias becomes negative with the average value of −3.1 ± 1.4% (range: −1.53 to −5.48%) as shown in Figure 8. The smaller negative bias observed in the field test of period #3 than the estimated bias could also be due to the opposite particle bouncing effect for particles larger than the cut-size.

The penetration of the M-PM₁₀ impactor after 23 sampling days in period #3 is also shown in Figure 10. The cut-size and the curve sharpness are 10.41 µm and 1.43, respectively, which are exactly the same as those of the cleaned M-PM₁₀ impactor. That is, the oiled M-PM₁₀ inlet is again shown to be able to retain particles effectively without particle bounce and overloading effects during long-term sampling.

5. Conclusion

This study investigated the effects of particle bounce and overloading on the sampling efficiency of the EPA louvered 16.7 L min⁻¹ PM₁₀ inlet. The field comparison test results show that the EPA louvered 16.7 L min⁻¹ PM₁₀ inlet which is cleaned initially but not cleaned daily afterwards, and not coated with grease oversamples PM₁₀ concentration during 14 sampling days with the average sampling bias of 14.58 ± 21.25% as compared to the cleaned, greased PM₁₀ inlet due to particle bounce and re-entrainment. The overestimation starts at the second sampling day with the sampling bias of +19.68% which does not meet the US EPA requirement with the bias smaller than ±10%. The sampling bias becomes severe at day 7 and day 12 with the high positive sampling bias of +69.31% and 41.08%, respectively. The EPA louvered 16.7 L min⁻¹ PM₁₀ inlet is then coated with the grease substrate to eliminate the particle bounce. The results show that the sampling bias is reduced during 16 sampling days in period #2 and 23 sampling days in period #3 as well. However, the uncleaned, greased inlet still oversamples PM₁₀ concentrations after three sampling days with the loaded particle mass of ∼1,113 µg and persists during 10 days due to the particle bounce effect. When the loaded mass is increased to 6,397 µg, the uncleaned, greased PM₁₀ inlet undersamples PM₁₀ concentrations due to particle overloading effect resulting in the decrease in the cut-size (from 10.44 to 9.77 µm) and a less sharp curve (the sharpness increases from 1.49 to 1.61).

The M-PM₁₀ inlet in which an oil-soaked GFF supported by an oil-soaked 25-mm PMD was used to replace the flat impaction surface in the traditional PM₁₀ inlet was developed to eliminate particle bounce. The field comparison test results show that the M-PM₁₀ inlet has the same sampling performance as the cleaned, greased PM₁₀ inlet with a very small average sampling bias (0.27 ± 2.9%) during 35 sampling days. The laboratory test results show that the cut-size and the curve sharpness of the M-PM₁₀ inlet remain the same (d_pa50 = 10.41 µm and sharpness = 1.43) after 23 sampling days as the cleaned M-PM₁₀ inlet (d_pa50 = 10.42 µm and sharpness = 1.4). The results of 6-h PM₁₀ concentrations show that the M-PM₁₀ inlet is able to use for short-term PM₁₀ monitoring and sampling without significant sampling error as compared to the greased PM₁₀ inlet. In summary, the present M-PM₁₀ inlet can be used for unattended continuous operation over 1 month with a good sampling accuracy without the need of frequent maintenance.

Additional information

Funding

The financial support of this work by the Taiwan Ministry of Science and Technology via the contracts MOST 105-2622-8-009-007-TE4, MOST 105-2221-E-009-005-MY3, and MOST 107-2622-8-009-004-TE5 is gratefully acknowledged.