Effects of Three-partitioned Horizontal Inlet and Clean Air on Collection Efficiency and Wall Loss of Slit Virtual Impactors

A numerical and experimental investigation comparing two configurations for a slit nozzle virtual impactor, one using a three-partitioned horizontal inlet and the other using a typical vertical inlet, was conducted to improve collection efficiency and reduce wall loss. All parameters, such as the length and width of the acceleration nozzle, the width of the collection nozzle, the distance between the two nozzles, and the inlet flow rate, were kept constant, and only the inlet configuration was changed. The ratio of the minor to the total flow rate was fixed at 0.1. Parametric analysis was performed to study the flow-rate influence of incoming clean air and aerosol on the performance of the three-partitioned horizontal inlet virtual impactor. The square root of the Stokes number, (Stk), was used to characterize the collection efficiency and wall loss of both configurations. It was observed that by using the three-partitioned horizontal inlet, the cut-off diameter was reduced from (Stk) = 0.80 to (Stk) = 0.41, and the wall loss near the cut-off diameter decreased from 25%–45% to below 4.5%. Moreover, using the three-partitioned horizontal inlet reduced the contamination by particles smaller than the cut-off size in the minor outflow section of the virtual impactor to almost 0%.


INTRODUCTION
Inertial impactors, because of its simple design and operation, are widely used for separation of aerosol particles by their sizes.A lot of studies have been carried out for performance improvement of inertial impactors by changing its geometric configurations and operating conditions, such as nozzle position, dimensions, and aerosol flow rate (Marple and Liu, 1974;Huang and Tsai, 2002a).Gómez-Moreno et al. (2002) studied the influence of Reynolds number and nozzle-to-collector distance on impactor collection efficiency.The cut-off size of inertial impactors corresponds to the particle size, for which the collection efficiency is 50%.Ideally, particles larger than the cut-off size are supposed to be collected on the impaction plate while particles smaller than the cut-off size escape without being trapped.Hence, an ideal inertial impactor should have sharp collection efficiency curve.Huang and Tsai (2002b) used a porous substrate in inertial impactors to show the effect of Reynolds number and porous substrate resistance factor on the sharpness of separation efficiency curve.Jurcik and Wang (1995) showed the effect of impactor geometry on the shape and sharpness of collection efficiency curve.Zhang et al. (2017) used the combined configuration of an inertial filter and an impactor to get a steeper collection efficiency curve.Arffman et al. (2012) compared the performance of round and slit nozzle impactors both numerically and experimentally and showed the effect of nozzle length on the performance of slit impactor.Cheon et al. (2017) placed an additional punched impaction plate between the existing impaction plate and the nozzle outlet for enhancing the separation characteristics of an inertial impactor.However, these impactors may have problems of particle bounce and overloading.Numerous efforts have been made to solve these problems.Rao and Whitby (1978) and Lee et al. (2005) suggested the use of oil or a glass filter layer over the surface of impaction plate for reducing the particle bounce.Kim et al. (2013) used an elliptical concave plate to minimize the particle bounce.
These problems can also be reduced by using a virtual impactor invented by Hounam and Sherwood (1965) and further developed by Conner (1966), Dzubay andStevens (1975), andForney (1976).In this type of impactor, impaction plate is replaced with a collection nozzle.The virtual impactor has two outflow sections, namely major flow and minor flow, and works on the same principle of inertia as the plate impactor does.Particles with smaller size and inertia follow the streamline and are collected at the major flow section, whereas larger particles due to their inertia cannot follow the streamline and hence collected at the minor flow section.The aerosol particles captured at the minor flow section determine the collection efficiency.The particles that get stuck to the wall between major and minor flow section causes wall loss.Hassan et al. (1979); Marple and Chien (1980) studied the effects of various parameters including major to minor flow ratio and Reynolds number, on the collection efficiency of round nozzle virtual impactor.Wada et al. (2016) compared the performance of virtual impactor and the conventional impactor, and have shown that virtual impactor has better separation capability and sampling accuracy.In virtual impactors, particles smaller than the cut-off size need to be collected at the major outflow section only and no such particles should be collected at the minor outflow section.However, in general, virtual impactors contain a certain fraction of particles smaller than the cut-off size in the minor outflow section, which is undesirable for particle size sampling.Masuda et al. (1979) and Chen et al. (1986) proposed an improved design of virtual impactor having a clean air core, which reduced the problem of contamination by particles smaller than cut-off size in the minor outflow, but caused a substantial increase in the wall loss.Chen and Yeh (1987) carried out a parametric analysis of virtual impactor and proposed new design for reducing the wall loss.Loo and Cork (1988) carried out parametric analysis of the improved virtual impactor and proposed a new design of virtual impactor having high collection efficiency and low wall loss; however, this proposed design had a problem of contamination by particles smaller than cut-off size in the minor outflow section.Chein and Lundgren (1993) used a clean air core at the inlet of virtual impactor and showed that the performance of virtual impactor could be enhanced by varying the clean air and aerosol flow rates.Ding and Koutrakis (2000) performed an experimental study on virtual impactor by varying different geometric and operating parameters, such as the width ratio of collection nozzle to accelerating nozzle, minor to major flow ratio, and Reynolds number.They came up with a conclusion that most of the theoretical principles, applicable to round nozzle virtual impactors, could also be applied to slit nozzle virtual impactors.Lee et al. (2014) used an orifice in the upstream of slit virtual impactors for improving the collection efficiency and minimizing the wall loss.However, the presence of an orifice caused high wall loss for large particles.
Most of these configurations of virtual impactors discussed so far have tried to reduce the wall loss and cut-off size.In addition, few studies have been conducted for removing the contamination by particles smaller than cut-off size in the minor outflow section of virtual impactors.However, there is still need of a virtual impactor that should have all good characteristics including a smaller cut-off size, very little or no wall loss and no contamination by particles smaller than cut-off size in the minor outflow section.Therefore, in this study, a new configuration of slit virtual impactor having three-partitioned horizontal inlet is used for reducing wall loss and improving collection efficiency.In other words, three partitions are provided to introduce two sheath flows for sandwiching the aerosol flow to prevent the fine particles from entering the minor flow section and to avoid their deposition on the wall between major and minor flow section.Hence, the objective is to achieve zero collection efficiency for particles having smaller diameter than the cut-off size, that is, to obtain a minor outflow free from contamination by particles smaller than cut-off size and to investigate the wall loss and collection efficiency of this new configuration.

NUMERICAL METHOD
A cross-sectional view of a slit virtual impactor having simple vertical inlet is shown in Fig. 1(a).The width of aerosol inlet was represented as D 1 and length of acceleration nozzle as T. W a , W c , and S denoted the width of acceleration nozzle, width of collection nozzle, and distance between the two nozzles, respectively.In this study, all geometric parameters were taken as a multiple of acceleration nozzle width, that is, W c /W a = 1.4,T/W a = 2.5, S/W a = 1.5, and D 1 /W a = 6, by following the studies conducted by Loo and Cork (1988), Ding and Koutrakis (2000), and Lee et al. (2014).The dimension of nozzle span was ten times the nozzle width, that is, l/W a = 10, to make sure the application of two-dimensional analysis.Fig. 2(b) shows a cross-sectional view of three-partitioned horizontal inlet slit virtual impactor developed in this study.The upper and lower inlets were used for clean air flow, while aerosol was introduced through the mid-inlet.The total inlet flow rate (Q) was the combination of clean air flow rate at upper inlet (Q 1 ), aerosol flow rate at midinlet (Q 2 ), and clean air flow rate at the lower inlet (Q 3 ).The optimum ratio of flow rate at given inlet to the total flow rate was determined through parametric analysis.The total horizontal length of inlet was represented as H, the length of partition as L, the partition width as P and width of inlet passages as D 2 .The dimensions of these parameters were set at H/W a = 58, L/W a = 9, P/W a = 4, and D 2 /W a = 5.Besides the inlet shape, the other geometric parameters (W a , W c , T, S) were kept the same for the two virtual impactors shown in Figs.1(a) and 1(b).The different values of acceleration nozzle width along with the corresponding Reynolds number (Re) and inlet flow rate (Q) for which analysis were performed on both impactors are listed in Table 1.
A computational fluid dynamics (CFD) code, ANSYS FLUENT Release 16.1 was utilized for fluid flow simulation in slit virtual impactor.Grid independence test was performed on impactor geometry to find out the optimum number of computation cells.As a result, the number of computational cells used for the analysis ranged from 16,000 to 65,000 based on the width of acceleration nozzle.It was assumed that the air flow was two-dimensional, steady, laminar, and incompressible.The nozzle Reynolds number (Re) used was in the range of 800 to 2000 and was calculated from the given equation.

Re
where ρ is the air density, U is the average velocity of airflow through acceleration nozzle, and µ is the air viscosity.The operating temperature and pressure of air were 20°C and 101.3 kPa.A SIMPLE algorithm for coupling velocity and pressure was used.The convergence criterion for iteratively solving continuity, momentum, and energy equations was set at 10 -6 .The boundary conditions utilized were velocity inlet at the inlet section, and outflow conditions at both major and minor outflow sections.The impactor walls were set at no-slip condition and the central axis line at symmetry.The ratio of minor to total flow was kept at 0.1 and major to total flow at 0.9.The discrete phase model (DPM) available in the FLUENT software was utilized for calculating particle trajectories.The factors considered to affect the particles were Stokes' drag force along with slip correction, gravitational force, and Brownian diffusion.Spherical particles with the same density, that is, 1000 kg m -3 , and at a constant space between them, were injected at the aerosol inlet section.
Uniform flow velocity profiles were assumed at all inlets of slit virtual impactor.The particles colliding with the impactor wall were assumed to be permanently captured by the wall without any reflection.The collection efficiency (η) and wall loss (WL) were calculated as 100 (%) where N min is the number of particles coming out of minor flow section, N maj is the number of particles received at major flow section, N in is the total number of particles injected at the inlet section, and N trap is the number of particles captured by the impactor walls.The η and WL for different widths of slit acceleration nozzle along with other operating conditions mentioned in Table 1 were determined and compared in terms of Stokes number's square root.The Stokes number was calculated from the given equation: where ρ p is the particle density, d p is the particle diameter, C c is the slip correction factor, and U is the average velocity of airflow through acceleration nozzle.

EXPERIMENTAL METHOD
Virtual impactors with desired inlet configurations, that is, vertical inlet and three-partitioned horizontal inlet, were fabricated to verify the numerically predicted results with experiments.The width of slit nozzle, W a , and the span, l, of both fabricated virtual impactors were kept at 1 mm and 10 mm respectively.The total flow rate (Q) was 9.1 L min -1 at a Reynold's number of 1000, for which the average velocity at acceleration nozzle came out to be 15.2 m s -1 .The schematic of experimental setup used for measuring the η and WL of virtual impactors having vertical inlet and horizontal inlet with three partitions for clean air flow are shown in Fig. 2. The Arizona Test Dust (ISO 12103-1, A4 type) was aerosolized using a solid aerosol generator (SAG410, TOPAS, Dresden, Saxony, Germany) and supplied to a large mixing chamber, where it was diluted with clean air.As the concentration of dust particles was very high, the excess amount of dust particles was exhausted to the atmosphere through HEPA filters.
The flow rates of clean air and aerosol particles were controlled by flow meters and were introduced to their respective inlets in virtual impactor.The major flow rate was 8.19 L min -1 and the minor flow rate was 0.91 L min -1 .The particle number concentration at major flow outlet, minor flow outlet, and inlet of the virtual impactor was measured with an optical particle counter (OPC, Model 1.109, GRIMM, Ainring, Bayern, Germany).Because the suction capacity of OPC was 1.2 L min -1 , 0.29 L min -1 of filtered room air was added to 0.91 L min -1 of sampled aerosol flow rate.The Arizona Test Dust particle sizes were converted into aerodynamic particle sizes by using a particle density of 2.56 g cm -3 and a dynamic shape factor of 1.4 (Krug et al., 2017).Stainless steel material was used for impactor fabrication.The lengths of the tubes and the bends in upstream and downstream were kept identical to neglect the particle transport loss.The experiments on each impactor were repeated five times and the following equations were used for experimentally determining collection efficiency (η) and wall loss (WL).

(%)
In these equations, C min and C maj are the particle number concentrations at minor and major exit sections, and C in is the particle number concentration at the impactor inlet section.Similarly, Q min , Q maj , and Q in are the flow rates at minor flow, major flow, and inlet section, respectively.

RESULTS AND DISCUSSION
Numerical parametric analysis was performed on threepartitioned horizontal inlet slit virtual impactor for particles having diameters in the range 0.1 to 10 µm, for different combinations of inlet flow rates as enumerated in Table 2, and the corresponding η and WL were plotted against the particle diameter in Fig. 3(a).The flow rate combination was arranged as Q 1 -Q 2 -Q 3 and the total flow rate was kept at 9.1 L min -1 .It was observed that the flow rate combination 2-3-4.1 (case 8) has shown small cut-off diameter, low wall loss, and zero collection efficiency for particles smaller than the cut-off diameter.All other flow rate combinations showed either a larger cut-off diameter or higher wall loss, especially for particles larger than 4 µm.As the inlet flow rate combination with Q 1 < Q 2 < Q 3 has shown better results, analysis was additionally performed for particles having diameters in the range of 0.1 to 10 µm for the values listed in Table 3, to get the best combination of inlet flow rates.The η and WL curves obtained are given in Fig. 3(b), which show that the flow rate combination of Table 2. Volumetric flow rates at horizontal inlet sections of slit virtual impactor for parametric analysis.

Case No.
Q 1.2 1.8 6.1 14 0.9 1.9 6.3 1.2-1.8-6.1 (case 13) had zero η for particles smaller than cut-off size, that is, there was no fine-particle-contamination in the minor outflow section.Moreover, it had a smaller cut-off diameter and much smaller wall loss for all particle sizes.Thus, the flow rate ratio at each inlet section, which gave higher η and low WL were given as Q 1 /Q = 0.13, Q 2 /Q = 0.20, and Q 3 /Q = 0.67.Fig. 4 shows the particle trajectories in both vertical inlet and three-partitioned horizontal inlet virtual impactors for particle sizes near the cut-off diameter.Here, the flow rate at the inlet sections were Q 1 = 1.2 L min -1 , Q 2 = 1.8 L min -1 , and Q 3 = 6.1 L min -1 (case 13).Some of the particles after accelerating in the slit nozzle collided with the wall of collection nozzle, as it turned towards the major flow section and caused wall loss.The number of particles colliding with the collection nozzle wall were more in vertical inlet virtual impactor as compared to the threepartitioned horizontal inlet virtual impactor.As there were three inlet sections in the horizontal inlet, that is, the upper and lower sections for clean air flow and the mid-section for aerosol flow, this configuration caused the particles to flow between the two layers of clean air, reducing the number of particles turning towards the major outflow section and also preventing the particles from colliding with the collection nozzle wall.As the three-partitioned horizontal inlet reduced the number of particles colliding with the collection nozzle and also increased the number of particles moving towards the minor outflow section, the wall loss could be reduced, and the collection efficiency could be improved.Fig. 5 shows the comparison of η and WL between numerical and experimental data for vertical inlet virtual impactor and three-partitioned horizontal inlet virtual impactor.The inlet flow rate (Q) for both configurations was 9.1 L min -1 .The minor to total flow ratio was kept at 0.1.The dimensions of impactors were fixed at W c = 1.4 mm, W a = 1 mm, T = 2.5 mm, S = 1.5 mm, D 1 = 6 mm, D 2 = 5 mm, and l = 10 mm.The lines in Fig. 5 indicate the numerically obtained results and the symbols with error bars represent the experimental outcomes.Overall, the numerical and experimental results agreed well with each other.In case of vertical inlet, as shown in Fig. 5(a), the cut-off diameters obtained were approximately 2.5 µm from both simulation and experiment, and the wall loss near the cut-off diameter was approximately 25%.However, by using the threepartitioned horizontal inlet with the flow rates of Q 1 = 1.2 L min -1 , Q 2 = 1.8 L min -1 , and Q 3 = 6.1 L min -1 , as shown in Fig. 5(b), the cut-off diameters obtained from both simulation and experiment were reduced to approximately 1.2 µm, whereas the wall loss near the cut-off size decreased to less than 4.5%.Moreover, particles smaller than 0.7 µm were collected only at the major outflow section, that is, there was almost no contamination by particles smaller than 0.7 µm in the minor outflow section.
Furthermore, simulations were performed on both configurations of virtual impactors by varying the width of acceleration nozzle from 1 to 4 mm and keeping the other geometric parameters fixed at W c /W a = 1.4,T/W a = 2.5, S/W a = 1.5, D 1 /W a = 6, D 2 /W a = 5 and l/W a = 10.The flow rates corresponding to different acceleration nozzle widths, given in Table 1, were used for performing fluid flow simulation.The collection efficiency and wall loss for different nozzle widths were determined and plotted as a function of square root of Stokes number (i.e., Stk 1/2 ), as shown in Fig. 6.The numerically obtained η and WL curves, of both vertical and horizontal inlet slit virtual impactors having different widths overlapped on one curve, showing that the slit virtual impactor performance with both inlet configurations can be characterized by using Stk 1/2 .The value of square root of Stokes number corresponding to the cut-off diameter, that is, (Stk 50 ) 1/2 , was determined to be 0.80 for vertical inlet, as shown in Fig. 6(a).However, by using three-partitioned horizontal inlet with the flow rate ratios of Q 1 /Q = 0.13, Q 2 /Q = 0.20, and Q 3 /Q = 0.67, the (Stk 50 ) 1/2 reduced to 0.41, that is, by about 49%, as shown in Fig. 6 inlet slit virtual impactor were up to 43%, whereas by using the three-partitioned horizontal inlet, it was found to be lower than 3%.However, for particles having diameter larger than 9 µm, the wall loss raised to almost 4.5%.Hence by using a three-partitioned horizontal inlet, a virtual impactor having a smaller cut-off diameter, very little wall loss and almost no contamination by particles smaller than cut-off diameter in the minor flow can be obtained.

CONCLUSIONS
A numerical and experimental investigation of slit nozzle virtual impactors using either a three-partitioned horizontal inlet configuration or a typical vertical inlet configuration was performed to evaluate the effect of inlet configuration on collection efficiency and wall loss.The simulation correctly predicted the performance of both configurations.The geometric parameters were fixed at W c /W a = 1.4,T/W a = 2.5, S/W a = 1.5, D 1 /W a = 6, D 2 /W a = 5, and l/W a = 10.The minor to total flow ratio was kept at 0.1.It was found that by using the three-partitioned horizontal inlet and introducing clean air through the upper and lower inlet sections, the square root of the Stokes number corresponding to the cut-off diameter, (Stk 50 ) 1/2 , was reduced by about 49%-from 0.80 to 0.41-and the wall loss near the cut-off diameter decreased from 43% to below 4.5%.Moreover, with the three-partitioned horizontal inlet, a zero-collection efficiency was obtained when (Stk) 1/2 was less than 0.25.Hence, the three-partitioned horizontal inlet significantly improved the performance of the slit virtual impactor by reducing the wall loss and minimizing the error in particle collection from the minor flow port, thereby increasing the accuracy of the aerosol-particle sampling.

Fig. 3 .
Fig. 3. Comparison of η and WL for different combinations of inlet flow rates on a three-partitioned horizontal inlet slit virtual impactor for: (a) parametric analysis; (b) optimum flow rate ratio.

Fig. 6 .
Fig. 6.Comparison of numerically obtained η and WL of virtual impactors for different nozzle widths as a function of square root of Stokes number: (a) with vertical inlet; (b) with three-partitioned horizontal inlet.

Table 1 .
Width of a slit virtual impactor nozzle and the corresponding operating conditions.

Table 3 .
Volumetric flow rates at horizontal inlet sections of slit virtual impactor for optimum flow rate ratio.