A comparison study of the filtration behavior of air filtering materials of masks against inert and biological particles

Inert particles are widely used as surrogates for biological particles in filtration tests. However, there is always a concern that the different physical properties between inert particles and biological particles may affect the filtration efficiencies of filter media. In the present study, the inert particles, e.g. NaCl and polystyrene latex (PSL), and biological particles, e.g. E.coli , B.sublitis , bovine serum albumin (BSA), and endotoxin, were used to evaluate the filtration efficiencies of the filtering materials of a surgical mask and a N95-rated respirator. The results showed that the difference in the filtration efficiencies of the N95-rated respirator for the tested inert particles and biological particles was smaller than 0.02%. The filtration efficiency of the surgical mask for PSL particles was lower than that for E.coli and B.subtilis particles with the same aerodynamic diameter because of the larger interception lengths of E.coli and B.subtilis particles. The average filtration efficiencies of the surgical mask for the tested NaCl particles and biological particles in the size range of 20 – 600 nm were similar. For the monodisperse particles, the filtration efficiency of the surgical mask for NaCl, BSA and endotoxin particles in the size range of 50 – 250 nm decreased in the above order due to the different shapes, dielectric properties, and multiply-charged fractions of the particles. The surgical mask presented the highest filtration efficiency for BSA particles and the lowest filtration efficiency for NaCl particles when the particle size was larger than 300 nm, which could be related to the difference in the electrostatic deposition and bounce probabilities. The present work revealed that the influence of several particle characteristics such as shape, dielectric property, and charge state on filtration efficiency should be considered when using common inert particles to replace biological particles to measure efficiencies of air filters.


Introduction
The environmental and health issues induced by bioaerosols have always attracted public attention.Bioaerosols include living and dead organisms, such as algae, viruses, archaea, bacteria, fungal spores, pollen, and plant fragments [1][2][3].As one of the air particulate pollutants, bioaerosols accelerate the spread of airborne organisms, deteriorate air quality, and increase health hazards.Numerous studies focused on the filtration efficiency of air filters or filtering facepiece respirators (FFRs) for bioaerosols in recent years [4], especially since the outbreak of the COVID-19 pandemic which has been reported to transmit via the airborne route [5,6].To investigate the filtration performance of air filtering materials against human pathogens is a challenging task because the generation of human-pathogenic particles is difficult and risky to be simulated in the experiment [7].Thus, inert particles are widely used as surrogates for human pathogens in filtration tests [8].
Many researchers hypothesized the capacity of air filters / FFRs to capture the human pathogens can be assessed based on the filtration performance of the filters for the inert particles (e.g.NaCl) with the same size as the biological particles.Leung and Sun [9] tested the filtration efficiency of nanofiber filters by using NaCl particles to simulate viruses in the range of 50-500 nm.By using the salt particles in the size range of 0.3-10 μm, Sousan et al. [10] assessed if the respirator filtering material of Halyard H600 was enough for respiratory protection during the COVID-19 pandemic.Chen et al. [11] also evaluated the filtration efficiency of the antibacterial nanofibrous membrane by using NaCl particles to simulate human pathogens.The polymersomes, which possess similar size and structure as viruses, were employed as a new type of virus-surrogate particles for filtration efficiency testing in Wang et al. [12].However, some researchers believe that the penetration of biological particles through air filters / FFRs might differ from the penetration of inert particles, because the shape, surface and composition of biological particles are different from inert particles [13,14].
Aerosolization of human pathogens was typically discouraged because of possible health hazards and the need for qualified biosafety laboratories [13].Therefore, some researchers proposed using nonpathogenic biological particles to evaluate the filtration performance of air filters / FFRs for human pathogens.Zacharias et al. [15] assessed the capacity of mobile air purifiers to remove virus-containing aerosols using bacteriophages as a surrogate for SARS-CoV-2.In Zhang et al. [16], in order to understand the effectiveness of residential HVAC filters during COVID-19, MS2 was utilized as a surrogate for SARS-CoV-2 virion in the filtration test.Moreover, the low-potency pathogenic A (H5N2) influenza virus, fungal spores, mycobacterium abscessus, and an attenuated vaccine strain of A-type influenza virus were employed as the challenge particles for filtration efficiency measurement [13,[17][18][19].In addition, Saccani et al. [7] stated that the capability of commercial filters to stop a virus propagation should be tested by using the virus itself.They evaluated the capacity of air filters to block SARS-CoV-2 using inactivated SARS-CoV-2 virion.In each of the above studies, only one type of inert particles or biological particles was used for measuring the filtration performance of air filters / FFRs; there was no comparison of the filtration behaviors of inert particles and biological particles.If the inert particles can well represent the biological particles for the filtration test is still under discussion.
Only a few studies have investigated the influence of inert particles and biological particles on filtration efficiency.Rengasamy et al. [20] tested the filtration efficiency of N95 FFRs by using both inert particles and biological particles.The results showed that the filtration efficiency of N95 FFRs was 98.15-99.68%for NaCl particles compared to 99.74-99.99%for PSL particles, 99.62-99.9%for Staphylococcus aureus, and 99.8-99.9%for bacteriophage phiX174.They concluded that using NaCl particles in filtration tests was conservative and was able to identify poorly performing filtration devices because the filtration efficiency of tested N95 FFRs for NaCl particles was significantly (p ≤ 0.05) lower than that for the used biological particles.However, the particle sizes of inert particles and biological particles in their study were different, in which the particle sizes of NaCl particles, PSL particles, Staphylococcus aureus, and bacteriophage phiX174 were 0.022-0.259μm, 0.1 μm, 3 μm, and 3 μm, respectively.In McCullough et al. [13], a variety of respirator filters and surgical masks were challenged with three aerosolized bacteria (Mycobacterium abscessus, Staphylococcus epidermidis, and Bacillus subtilis) and one inert particle (PSL).The authors found that the penetration of the PSL particles was greater than that of the biological particles, however, the PSL particles were smaller than the three used biological particles.Whether inert particles could be the surrogate particles for human pathogens in filtration tests remains uncertain.
A comprehensive comparative study is required in terms of the filtration efficiency of air filtering materials for biological particles and inert particles with the same size.In the present study, the filtration efficiencies of a N95 FFRs and a surgical mask were evaluated by using both inert particles (NaCl and PSL) and the biological particles including bovine serum albumin (BSA), endotoxin, and the lysate of E. coli and B. subtilis.In the measurement, both the inert particles and biological particles were controlled to have monodisperse distribution and the same particle size.By analyzing the influence of physical characteristics (shape, multiple charges and dielectric property) of the utilized particles on the filtration efficiency, the feasibility of using inert particles to replace biological particles in filtration tests was discussed.

Materials
A brand of surgical masks (Foliodress, EN 14683 Type II) and a N95 rated respirator (Filtering face piece class 2 according to the standard of EN149:2001 (FFP2), UVEX) on the Swiss market were selected.E. coli (K12, a gram-negative bacteria) and B. subtilis (B 4056, a gram-positive bacteria) were purchased from the Culture Collection of Switzerland (CCOS).NaCl, PSL, BSA, and endotoxin were obtained from Sigma-Aldrich.

Generation of inert particle and bioaerosol
As shown in Fig. 1a, NaCl, BSA, endotoxin and the lysate of E. coli and B. subtilis were dispersed in deionized water.Then, the solutions / suspensions with different concentrations were atomized by highly pressurized air.The atomized droplets passed through a drier to facilitate water evaporation and to form solid particles.Particles produced by the atomizer obtained a specific size distribution, which was measured by a scanning mobility particle sizer (SMPS).The SMPS consisted of a differential mobility analyzer (DMA, TSI 3081) and a condensation particle counter (CPC, TSI 3775).The concentrations of suspensions and the flowrate of applied pressurized air are shown in Table 1.

Filtration efficiency test
The setup of the filtration efficiency test was shown in Fig. 1b.The solid particles were neutralized by a neutralizer (Kr-85), and the DMA was used to select the monodisperse particles to challenge the filter.Two CPCs were used to measure the particle concentrations of upstream and downstream of a circular sample from the mask.For the filtration efficiency test of particles larger than 1 μm, an aerodynamic particle sizer (APS, TSI 3321) was used to measure the particle concentrations upstream and downstream (Fig. 1c).A face velocity of 5.3 cm/s was set in all the filtration efficiency tests.During the classification of monodisperse particles, the larger particles with multiple charges could pass through the DMA thus potentially affecting the test result of filtration efficiency.In order to analyze the fractions of the multiply charged particles, the SMPS was employed to scan the size distribution of the particles coming out from the first DMA (Fig. 1d).

Size distribution of generated particles
The compressed-air atomizer used in the present study is the most common way to generate droplet aerosols, which breaks the liquid into droplets from liquid solutions or suspensions.When the atomizer is used with a volatile solvent containing soluble or insoluble solid material, e. g., salt and PSL, the solvent evaporates quickly with the assistance of the diffusion drying column and solid aerosol particles are produced.As shown in Fig. 2a, b, c, when the same concentration of solution / suspension was applied, the count median diameter (CMD) of particles decreased with the increase of compressed-air pressure.The higher compressed-air pressure increased the flow speeds of gas and liquid, which mean the air can provide more energy for the liquid to overcome the surface tension and viscous force [21].The increased energy broke the liquid into smaller droplets, and the droplets were reduced in size by the same factor after drying [22], therefore the number of smaller solid aerosol particles increased consequently.At the same time, the higher compressed-air pressure enabled more liquid to be drawn and atomized.As a result, the higher compressed-air pressure induced smaller CMD, wider size distribution and higher number concentration of the total particles.
As shown in Fig. 2d, e, f, the solute concentration of solution / suspension significantly affected the size distribution of the solid aerosol particles.The CMD and the total number concentration of the solid aerosol particles increased with the increasing solute concentration, which was also observed in previous studies [23][24][25].Higher solution concentration may result in bigger particle sizes due to more solute being present in each droplet [26].Moreover, the coagulation rate was proportional to the square of particle concentration [27], thus the increase in particle concentration promoted that the particles collided with each other and grew into larger droplets.As a result, it was observed that the CMD shifted to right with the increase of solute concentration.
The increasing total particle number concentration with the higher dissolved solute concentration was not intuitively understood if one assumed the atomization process was unaffected by the solute concentration.However, several internal and external factors might contribute to this observation.Internal factors mainly referred to the alteration of solution density, viscosity and surface tension.These factors changed the atomization itself such as increasing the number of droplets during atomization or influencing the solvent evaporation.Amaral et al. tested the physical properties of NaCl solutions at different solute concentrations, and the results showed that the density, viscosity and surface tension slightly increased with the increase of solute concentration in the range of 0.01%-6% [24].Salt has high affinity with water and the higher NaCl concentration discourages evaporation, thus could lead to more droplets in the atomizer [22].
The coagulation and wall loss of particles during drying and transportation were considered as external factors influencing the total number concentration.The effective coagulation coefficient and wall loss decreased with increasing CMD values, which meant the total number concentration increased with CMD increases.In other words, very small particles were scavenged by coagulation or lost by diffusion, therefore the particles from the low solute concentration lost more particles because they had more very small particles.There is a complex interplay between the CMD and total number concentration.For example, the higher particle concentration may result in a larger CMD by increasing the coagulation rate of particles, at the same time, larger CMD may increase the particle concentration by decreasing effective coagulation coefficient.However, the results of [23] showed that neither coagulation nor wall loss was the dominant factor for the increased total number concentration when the solute concentration shifted to larger sizes.
In the present study, we mainly focused on the difference of inert particles and biological particles in filtration efficiency tests.In the further experiments, the solutions of 1%wt NaCl, 5 mg/ml endotoxin, and 25 mg/ml BSA were atomized to generate the test particles by 2.5 bar pressure.In addition, E. coli and B. subtilis were suspended in deionized water, and the suspensions were atomized for testing the filtration efficiency of filters for micron particles.According to the particle distributions of E. coli and B. subtilis particles, PSL particles with similar sizes were selected for the comparison of filtration efficiencies (Fig. 3).As shown in Fig. 3, the influence of the droplets generated from the deionized water on the particle distributions of E. coli, B. subtilis and PSL particles was negligible, because all the particles were fully dried by the diffusion dryer.

Filtration efficiencies for NaCl, BSA, endotoxin, PSL, E. coli, and B. Subtilis particles
As shown in Fig. 4a and 4b, when the different kinds of particles with the same size were selected in the range of 20-1000 nm, the difference in the filtration efficiencies of the FFP2 respirator for NaCl, BSA, endotoxin, PSL, E. coli, and B. subtilis particles was less than 0.02%.The most penetrating particle size (MPPS) is an important parameter for each filter medium.The standardized efficiency tests for filter media measure

Table 1
The concentrations of suspensions and the pressure of pressurized air used for particle generation.the minimum efficiency at MPPS, which presents the worst-case scenario of filter media [4,28].In the present study, the MPPS of the FFP2 respirator were 40, 40, and 50 nm when NaCl, BSA, and endotoxin particles were used as the challenge particles, respectively.The filtration efficiencies and MPPS of the FFP2 respirator against NaCl, BSA, and endotoxin particles were similar, which indicated that using NaCl particles as the surrogate for biological particles is reasonable in the filtration test of FFP2 respirators or the same level filter media.Both NaCl and PSL particles are easy to purchase on the market at reasonable prices, more important, the filtration experiments using NaCl and PSL particles can be operated in general lab environment, removing the need of a biosafety laboratory.
The filtration efficiencies of the surgical mask for 0.7 μm E. coli and 1 μm B. subtilis were higher than those for PSL particles with the corresponding sizes (Fig. 4c).In the size range of 20-600 nm, the average filtration efficiencies of the surgical mask for NaCl, BSA, and endotoxin particles were 89.18 ± 0.15%, 91.08 ± 0.17%, and 88.39 ± 0.14%, respectively.The NaCl particles can be an alternative to biological particles to test surgical masks or same-level filter media, when only the average filtration efficiency is considered.For monodisperse particles, Relationship between the pressure of compressed air and generated particle number distribution in the size range of 15-700 nm in terms of electrical mobility diameter, the solution concentrations were 0.04 wt% for NaCl, 0.2 mg/ml for endotoxin, 1 mg/ml for BSA.(d), (e), (f): Relationship between solution/suspension concentration and generated particle number distribution in the size range of 15-700 nm in terms of electrical mobility diameter, the pressure of compressed air was 2.5 bar.
when the particle size was less than 250 nm, the filtration efficiencies of the surgical mask for NaCl particles were higher than those for BSA and endotoxin particles, and the trend reversed when the particle size was larger than 300 nm.Moreover, the MPPS of the surgical mask for NaCl, BSA, and endotoxin particles were observed at 300, 200, and 250 nm, respectively.
The filtration efficiencies of FFP2 respirator for all particles in the size range of 20-600 nm were larger than 99.9%, so the influence of particle types on filtration efficiency was not significant.In contrast, the surgical mask possessed lower filtration efficiency and the different particle parameters presented larger influence on its filtration efficiency.For example, the filtration efficiency of the surgical mask changed much more with the shift of particle size compared to the FFP2 respirator, which indicated that the surgical mask would exhibit larger efficiency difference for particles with different shapes, i.e. different interception lengths.The different filtration behaviors of the filter media for inert particles and biological particles will be further analyzed in terms of the shapes, multiple charges and dielectric properties of the test particles.

The influence of particle shape on filtration efficiencies
Particle shape is an important parameter that strongly affects the transportation and deposition of particles in filter media [29,30].Gao et al. [30] tested the filtration efficiency of Nuclepore filters (with 3 μm pore size) for NaCl particles and graphene nanoplatelets (GNPs) at 5 cm/ s with the same aerodynamic diameters.Their results showed that the capture efficiency of GNPs was much higher than that of NaCl particles due to the platelet-shaped GNPs had larger geometric diameter.The influence of the particle shape on filtration efficiency should be carefully considered for the biological particles, because the biological particles possess various shapes, such as approximately spherical shapes (spores, coccal bacteria, some pollen), rod-or fiber-like geometries (bacilli, hairs), and disk-or platelet-like geometries (plant fragments) [31].E. coli and B. subtilis were known to be cylinder shape.The NaCl, BSA, and endotoxin particles showed cubic, mixtures of sphere and oblate  spheroid, and spherical shape, respectively (Fig. 5).
Usually, different equivalent diameters relevant to each deposition mechanism were used to account for particle shape effects.The aerodynamic diameter d a , geometric diameter d g and mobility diameter d m were used to analyze the filtration efficiencies attributed to impaction, interception and diffusion, respectively [30].The interception was often the dominant deposition mechanism for particles in the size range of 0.3-1 μm.The geometric diameter is related to the volume equivalent diameter d e , therefore, d e of particles was calculated for analyzing the particle shape effect.In the present study, the measured diameters of E. coli, B. subtilis, and PSL particles were aerodynamic diameters.The aerodynamic diameter is the diameter of a unit-density sphere that has the same terminal settling velocity as the particle under consideration.For a unit-density sphere, d a is equal to d e .The terminal settling velocity (V TS ) for a spherical particle with unit density is and the terminal settling velocity for a non-spherical particle or a spherical particle with non-unit density (V where ρ 0 is unit density (1 g/cm 3 ), g is the acceleration of gravity, μ is air viscosity.ρ P is the density of the particle.The densities of E. coli and B. subtilis were approximated to the density of typical protein (1.3 g/ cm 3 ), and the density of PSL particle is 1.05 g/cm 3 .C c is the slip correction factor which can be calculated by where λ is the air mean free path.× is the dynamic shape factor of the particle.Both the E. coli and B. subtilis are cylinders, and their SEM images were provided in supplementary materials (Figure S1).In the present study, the × of E. coli and B. subtilis were calculated based on a shape resistance factor model which is in good agreement with cylinders and ellipsoids.The shape resistance factor, K, is equal to 1/x [22,32].K was calculated as: D s , D max , and D n are the volume equivalent diameter, maximum body dimension, and projected area diameter depended on the motion direction of the particle obtained from SEM pictures, respectively.Ψ is the surface sphericity, which was defined as [33]: ψ = surface area of the volume equivalent shpere surface area of the particle (3.5) As a result, the dynamic shape factors of the used E. coli in horizontal, vertical, and averaged orientation were 1.08, 1.38, and 1.28, respectively.The dynamic shape factors of the used B. subtilis in horizontal, vertical, and averaged orientation were 1.05, 1.32, and 1.23, respectively.PSL particle is a sphere and its × is 1.Then, the d e of E. coli, B. subtilis, and PSL particles corresponding to d a can be calculated according to equations (3.1) to (3.3).In Kim et al. 2009 [34], they defined the dimensionless interception parameter based on the maximum projected length of the non-spherical particles.Herein, the maximum projected length of used particles was defined as interception length.Because the d e and particle shapes of PSL, E. coli, and B. subtilis were known, the interception lengths of the particles can be calculated.The de and the interception lengths of the particles are shown in Table 2, and the detailed calculation process was shown in the Supplementary Information.
The measured diameters of NaCl, BSA, and endotoxin were mobility diameters.The mobility diameter is the equivalent diameter of a spherical particle that has the same terminal electrostatic velocity in the electric field as the particle under consideration.The particle reached terminal electrostatic velocity as the electrostatic force balanced the drag force.The drag force (F DS ) of spherical particle with a single charge is and the F ′ DS of a non-spherical particle is  As shown in Fig. 5, NaCl particles are cubic shape, and the × of a cubic particle is 1.08 [22].The endotoxin particles are spherical, with × of 1.Most of the BSA particles are oblate spheroid, and the × of the BSA particles can be also calculated based on the shape resistance factor model.The detailed calculation equations were shown in the Supplementary Information.As a result, the dynamic shape factor of BSA for horizontal axis, vertical axis, and averaged orientation was 1.31, 1.14, and 1.19, respectively.Assuming the drag forces of two particles with the same electrical mobility diameter were the same, the d e of NaCl, BSA, and endotoxin particles corresponding to d m can be calculated according to equations (3.6) and (3.7).The calculated de of the used particles are shown in Table 3.
The larger dimension of E. coli and B. subtilis were more prone to be intercepted during filtration process.As a result, the surgical mask presented higher capture efficiency for E. coli and B. subtilis than that for spherical PSL particles with the same aerodynamic diameter (Fig. 4c).With the same mobility diameter, the order of the interception length of the three particles was NaCl > BSA > endotoxin, which may be one of the reasons for the difference of filtration efficiencies of the three particles in the size range of 50-250 nm (Fig. 4d).However, the interception length cannot explain the difference among the filtration efficiencies for NaCl, endotoxin, and BSA particles in the size range of 300-600 nm.According to a previous study, the filtration efficiency of cubic particles was lower than that of spherical particles, because the bounce probability of cubic particles was higher than spherical particles during the particles tumbling along the fiber [35].Moreover, the bounce probability increased with the increase of kinetic energy which grew with the increasing particle size [35].Thus, the difference in filtration efficiency for bigger particles was larger than that for smaller particles (Fig. 4d).Additional discussion of the different filtration efficiencies based on electrostatic mechanism can be found in the next section.

The influence of multiple charges and dielectric property of particles on filtration efficiency
In the present study, the test filter medias were electret filters, in which diffusion and electrostatic absorption were the dominant deposition mechanisms.The simplest means for calculating the filtration efficiency due to diffusion (E d ), coulomb forces (E c ), and polarisation forces (E p ) may be written as [36,37] where P e is the Peclet number, N Qq is the dimensionless parameter describing capture of charged particles by a charged fiber, and N Q0 is the dimensionless parameter describing the capture of neutral particles by a charged fiber.N Q0 and N Qq are defined as where Q is the charge per unit length of the fiber, q is the charge held by a particle (q = ne, e is the elementary charge, n is the number of elementary charges on the particle), d p and d f are the diameter of the particle and fiber, respectively.ε 0 is the permittivity of free space, μ is viscosity, U is air velocity, and D p is the dielectric constant of the particle material.The dielectric constant D p and the number of elementary charges on the particle were the main factors affecting the filtration efficiency, when the same filter media were tested at the same conditions.The number of multiple charged larger particles was measured by using the setup shown in Fig. 1d.The mobility diameter was set to a predefined value in the first classifier.Then, SMPS was employed to scan the particle size distribution.Other peaks appearing at larger particle sizes than the pre-set one indicated multiply charged particles [38].The size distributions of the classified particles in the size range of 50-250 nm were shown in Fig. 6, and the fraction of multiply charged particles (f mc ) was calculated.
At the same classified particle size, the BSA particles had a higher fraction of larger particles with multiple charges compared to NaCl and endotoxin particles, and the order was f mc (BSA)＞f mc (NaCl)＞f mc (Endotoxin).The f mc of the 100 nm BSA particles was obviously higher than other classified BSA particles and other types of particles of 100 nm, which might result from agglomerates of biological residues.In Modesto-Lopez et al. [39], they concluded that the agglomerates of biological residues might induce the other larger peaks of the monodispersed biological particles.In addition, the amount of multiple charged particles is depended on the relationship between the classified particle size and the peak of the original size distribution.Normally, classifying the particle size which is on the right side of the peak of the size distribution leads to a low concentration of multiple-charged particles [38].In the present study, the original size distribution of the generated BSA particles had peak size larger than 100 nm, therefore not favorable for classifying 100 nm BSA particles, in contrast, the endotoxin size distribution was better for classifying 100 nm particles (Fig. 2).According to equations (3.8) to (3.12), several conclusions can be summarized if only one of the dimensionless parameters related to diffusion, coulomb forces or polarisation forces is considered at a time.On the one hand, P e is a dimensionless parameter related to the diffusional motion of particles.Equation (3.8) indicates that the E d decreases as P e increases, and P e increases with the increase of particle diameter.Therefore, the different f mc of NaCl, BSA, and endotoxin led to E d (Endotoxin)＞E d (NaCl)＞E d (BSA), where E d (Endotoxin) means the filtration efficiency of filter media for endotoxin particles by diffusion.On the other hand, the higher f mc represents larger value of q which increases the value of dimensionless parameter N Qq .Thus, the different f mc of NaCl, BSA, and endotoxin induced E c (BSA)＞E c (NaCl)＞E c (Endotoxin),  where E c (Endotoxin) means the filtration efficiency of filter media for endotoxin particles by coulomb forces.In addition, the dielectric constant of NaCl is 5.9, while the minimal value of the dielectric constant for proteins is about 20 [40,41].The larger dielectric constant of particle caused the larger value of the dimensionless parameter N Q0 .The different dielectric constants D p of NaCl, BSA, and endotoxin resulted in E p (BSA)≈E p (Endotoxin)＞E p (NaCl), where E p (Endotoxin) means the filtration efficiency of filter media for endotoxin particles by polarisation forces.
The influence of interception length, f mc and D p on the filtration efficiency by interception, diffusion, coulomb forces and polarisation forces was summarized in Table 4. BSA particles had larger interception length and stronger coulomb force than endotoxin particles, leading to higher filtration efficiency for BSA and endotoxin observed in the experiments.In the size range of 50-250 nm, the higher efficiency of NaCl particles compared to BSA particles was probably due to the combined effects of diffusion, electrostatic deposition, and interception.On the one hand, E d (NaCl) was higher than E d (BSA) because the lower fraction of bigger particles with multiple charges.On the other hand, the interception length of NaCl particles was slightly bigger than that of BSA particles, therefore the efficiencies by interception (E r ) of NaCl particles were higher than that of BSA particles.In the size range of 300-600 nm, the influence of stronger coulomb and polarization forces on the BSA particles than the NaCl particles seemed to outweigh the difference in the interception lengths, contributing to the larger filtration efficiency of BSA than NaCl.In addition, the higher bounce probability of cubic particles might also be a factor for the lower filtration efficiency of NaCl in 300-600 nm.

Conclusion
The filtration efficiencies of a surgical mask and a N95 rated respirator for inert particles (NaCl, PSL) and biological particles (BSA, endotoxin, E.coli, and B.subtilis) were evaluated, and the influence of the physical properties of the particles on the filtration efficiencies was analyzed.The results showed that there was no significant difference in the filtration efficiency and the most penetrating particle size (MPPS) of the N95 rated respirator, when different types of particles were used as the test particles.The filtration efficiencies of the surgical mask for 0.7 and 1.0 μm PSL particles were lower than those for E.coli and B.subtilis particles with the same aerodynamic diameter, because the used E.coli and B.subtilis particles had larger interception lengths.In the size range of 20-600 nm, the NaCl particles can be a biological particles surrogate for testing the average filtration efficiency of the surgical mask or the same level filter media.For the monodisperse particles in the size range of 300-600 nm, the filtration efficiencies of the surgical mask for NaCl particles were lower than those for BSA and endotoxin particles.When the monodisperse particles were smaller than 250 nm, the difference of the shape, density, dielectric properties, and fractions of multiply charged particles of the particles caused the filtration efficiency of the surgical mask for NaCl, BSA, and endotoxin to decrease in the above order.
It should be noted that the achieved results of filtration efficiency are valid for the investigated particles and electret filters in the present study.The filtration performance of other air filter media for different biological particles may be different due to the complex nature of various bioaerosols and air filters.However, the analysis process is applicable to other types of particles and air filters.The particle capture mechanisms should be identified based on the types of air filters and the charge states of test particles.Then, analysis should be performed on the influence of particle properties such as the shape, bounce probability, density, and dielectric property on the filtration efficiency induced by different particle capture mechanisms.The present work is of significance for analyzing the effect of particle types on the filtration efficiencies of air filters.Based on the analysis results, a proper particle can be selected to replace pathogenic biological particles in the air filtration test.For example, the particle with the similar shape as the pathogenic biological particles should be selected if interception is the dominant filtration mechanism for the tested air filter; when the particle filtration mainly depends on the electrostatic absorption, the particle with similar dielectric properties would be better.

Fig. 1 .
Fig. 1.The set up for (a) particle generation; (b) filtration efficiency test of submicron particles; (c) filtration efficiency test of micron particles; (d) the measurement of multiply charged particles.
E. coli and B. subtilis 25 mg/ml 2.5 bar W.He et al.

Fig. 2 .
Fig. 2. (a), (b), (c):Relationship between the pressure of compressed air and generated particle number distribution in the size range of 15-700 nm in terms of electrical mobility diameter, the solution concentrations were 0.04 wt% for NaCl, 0.2 mg/ml for endotoxin, 1 mg/ml for BSA.(d), (e), (f): Relationship between solution/suspension concentration and generated particle number distribution in the size range of 15-700 nm in terms of electrical mobility diameter, the pressure of compressed air was 2.5 bar.

Fig. 3 .
Fig. 3.The size distribution of E. coli, B. subtilis, and PSL particles used in the filtration test.

Fig. 4 .
Fig. 4. (a) The filtration efficiency of the FFP2 for 0.7 and 1 μm particles; (b) the filtration efficiency of the FFP2 for 20-600 nm particles; (c) the filtration efficiency of the surgical mask for 0.7 and 1 μm particles; (d) the filtration efficiency of the surgical mask for 20-600 nm particles.

. 7 )Fig. 5 .
Fig. 5.The SEM images of NaCl, BSA, and endotoxin particles loaded on the test filter media.

Fig. 6 .
Fig.6.The size distributions of particles exiting from the first DMA.The peaks appearing at larger particle sizes than the pre-set one indicated multiply charged particles, f mc is the fraction of multiply charged particles.

Table 2
The volume equivalent diameter de and interception length of E. coli, B. subtilis, and PSL particles corresponding to measured aerodynamic diameter da.

Table 3
The volume equivalent diameter d e and interception length of NaCl, endotoxin, and BSA particles corresponding to measured electrical mobility diameter d m ; the d e for BSA particles are demonstrated in different projected orientations, the format is horizontal axis/vertical axis/orientation averaged.