An elastomeric 3M 6000 series half-mask respirator equipped with two 3M 2091 P100 filters (3M Corp., St Paul, MN, USA) was chosen for the testing to assure the continuity of our previous research (He et al., 2013a,b). The rationale for selecting the above respirator was described in detail by He et al. (2013a). An 11-mm-long flush probe with a 14-mm-diameter flange and a 4-mm-diameter inlet was mounted on the surface of the respirator centerline between the manikin’s nose and upper lip. The end of the probe (14-mm flange) was flush with the interior surface of the half-mask. The probe was located ~25 mm from the manikin’s nose/mouth.

The challenge aerosols were separately generated by burning the following three materials inside a test chamber: wood (24-cm-long and 0.4-cm-diameter pellets, 1.9 ± 0.5 g), paper (23 × 24 cm brown multifold paper towel, 2.1 ± 0.2 g), and polyethylene (23 × 20 cm, 1.7 ± 0.3 g—further referred to as plastic) (He et al., 2013a). Each material held by a caliper was ignited by a long-reach lighter and completely burnt inside the chamber. All burnt materials were captured in a water-filled basin placed on the floor. The measurements were initiated 15 min after burning to allow the combustion aerosol to reach a spatial uniformity. As our previous measurements revealed that 90% of particles so generated were within the range 20–200 nm (He et al., 2013a), we focused on this size range.

The specifications of the advanced manikin headform chosen for this study were reported in detail by Bergman et al. (2013). Briefly, the headform is of the medium size defined by the NIOSH Principal Component Analysis panel created using data from a large-scale anthropometric survey of US workers conducted in 2003 (Zhuang et al., 2007). A human-like skin with locally correct thicknesses was mounted on the headform skull. The material used to generate the skin is called Frubber™ (Hanson and White, 2004), a fluid-filled cellular matrix composed of an elastomer that compresses, elongates, and otherwise deforms in ways that simulates human skin (Hanson and White, 2004).

The experimental setup is shown in Fig. 1. The respirator was donned on the advanced manikin headform, which was then challenged with one of three combustion aerosols (wood, paper, and plastic). The donning was performed according to the manufacturer’s user instructions. After donning, the respirator was not redonned or repositioned until the completion of the study, which allowed maintaining the size and shape of the face seal leaks. Obviously, the static headform design did not allow fit testing of the respirator prior to the study. The P100 filters were changed after every 15 runs to minimize the effect of loading of combustion products on the filter medium. The filters were changed carefully to minimize the effect of this procedure on the respirator face seal leakage. After each filter change, the particle penetration was measured to assure that the size of the leak was consistent with the one existing before the change under the same experimental conditions. Temperature and relative humidity were kept at 17–22°C and 30–50%, respectively.

The headform was connected to a Breathing Recording and Simulation System (BRSS, Koken Ltd, Tokyo, Japan) with a high-efficiency particulate air filter placed in between to keep particles from reentering the respirator cavity in exhaled air. The BRSS consists of an electromechanical drive-cylinder coupled with two air cylinders. As the electromechanical cylinder stroke moves back (inspiratory duration, half a period) and forth (expiratory duration, half a period), a sinusoidal air flow is generated (Haruta et al., 2008). The stroke moving distance and frequency can be adjusted with a resolution of 0.1 mm and 0.01 Hz, respectively, thus allowing for precise changes to breathing flow rate and frequency when human breathing is simulated (Haruta et al., 2008).

The test were conducted under three cyclic breathing flows (MIF—time weighted average flow rate over the width of an inspiration, MIF = 30, 55, and 85 l/min) and five breathing frequencies (10, 15, 20, 25, and 30 breaths/min). A completely randomized factorial study design was chosen with four replicates for each combination of the tested breathing flow rate and frequency. A summary of the experimental conditions is presented in Table 1. Aerosol concentrations inside (C_in) and outside (C_out) the respirator were measured using a nanoparticle spectrometer (Nano-ID NPS500, Naneum Ltd, Kent, UK) size selectively in 10 channels between 20 and 200 nm at a sampling flow rate of 0.2 l/min. Each concentration measurement took 3 min. The corresponding mean sizes for the 10 chosen channels were 22.4, 28.2, 35.5, 44.7, 56.2, 70.8, 89.1, 112.2, 141.3, and 177.8 nm. TIL was determined for each particle size (d_p) as the ratio of inside to outside concentration: (1)TILdp=Cin,dpCout,dp×100%

In addition, by combining the 10 channels, (overall) TIL was calculated for all combustion particles ranging from 20 to 200: (2)TIL=∑i=110Nin,i∑i=110Nout,i×100%where N_in is the number of particles measured in a specific channel inside the respirator, N_out is the number of particles measured in a specific channel outside the respirator, and i is the ith particle size channel.

Data analysis was performed using SAS version 9.3 (SAS Institute Inc., Cary, NC, USA). For the size-selective TIL, the effect of particle size was analyzed using the one-way analysis of variance (ANOVA) on the pooled data. Three-way ANOVA was used to study the significance of combustion material, breathing flow, and breathing frequency for the size-independent TIL. One-way ANOVA was performed to quantify the effect of breathing frequency on the size-independent TIL after stratifying the data by the combustion material and breathing flow. All pairwise comparisons were performed using Tukey’s range test. P-values < 0.05 were considered significant.

The size-selective TIL values determined within the chosen 10-channel size range are shown in Fig. 2. The graphs represent three MIFs, five breathing frequencies, and three combustion aerosols (error bars are not shown as they make it difficult to see the data lines for each breathing frequency).

The reported TIL data were compared with the filter penetration data. The latter were obtained in our earlier study, which examined the performance of a fully sealed half-mask equipped with two P100 filters while challenged with wood, paper, and plastic combustion particles of size 20–200 nm (He et al., 2013a). For most of the particle sizes, TIL was much greater than the filter penetration (<0.03%), suggesting that the face seal leakage was the primary particle penetration pathway for the tested elastomeric half-mask (an additional experiment was conducted to confirm that exhalation value did not leak).

The average size-selective TIL values obtained from this study were comparable to those reported in our previous study with a partially sealed (nose or nose–chin area) half-mask donned on a hard plastic manikin headform (He et al., 2013a,b), and 5- to 10-fold lower than those obtained with the unsealed half-mask donned on the plastic headform. This improvement can be attributed to the softness of the manikin skin, which deforms under stress, thus increasing the contact surface area and consequently forming a better seal.

Figure 2 shows that particle size affected TILdp (P < 0.0001) only for particles <40 nm. For all three aerosols, the TILdp values obtained at the lowest tested sizes, 20–40 nm, fell below those obtained at >40 nm regardless of the breathing flow and frequency. This can be explained by diffusional deposition, which is more pronounced for smaller particles (consequently, a smaller fraction could penetrate inside the respirator more readily). For particles 40–50 nm and larger, the TILdp curves showed no consistent increasing or decreasing trend. Furthermore, ANOVA results showed no significant differences in TILdp (P = 0.36) among the six channels between 45 and 200 nm. The relatively flat curves obtained for the half-mask indicate a wide range for the MPPS. No TIL MPPS data were available in the literature for elastomeric respirators until our recent study (He et al., 2013a), which measured an MPPS of 120–140 nm for the same half-mask donned on a plastic manikin headform and challenged with plastic combustion aerosol. The cited study did not detect an MPPS for wood and paper aerosols due to multiple peaks in the TILdp curves. Previous TIL studies have focused on N95 FFRs or surgical masks, for which the particle size effect was found significant and the TIL values were close to the penetration levels observed for the filters only (Myers et al., 1991; Cho et al., 2010a; Rengasamy and Eimer, 2011). At the same time, little information has been generated with respect to the MPPS for face seal leakage. When testing N95 FFRs with artificially created cylindered leaks, Rengasamy and Eimer (2011) found that for leak diameters ≤1.65 mm, the MPPS was ~50 nm; we believe that this value likely represents the N95 filter penetration, whereas the TIL of the respirator used in this study (equipped with P100 filters) is primarily governed by the face seal leakage penetration. This likely explains the suppressed peaks or the plateaus seen in the curves presented in Fig. 2.

The effect of breathing frequency on TILdp was complex. When challenged with wood combustion aerosol, frequencies of 30, 20, and 15 breaths/min produced the highest TILdp values at MIF = 30, 55, and 85 l/min, for most size channels between 20 and 200 nm. MIF = 30 l/min produced the most notable effect of the breathing frequency on the TILdp. The TILdp curves obtained from testing with paper combustion aerosol showed some peaks at 30 and 10 breaths/min at MIF = 30 and 85 l/min, but no clear peak was identified for MIF = 55 l/min. The graphs representing TILs for plastic aerosol revealed separations between the TILdp curves at MIF = 30 l/min, which essentially diminished at higher flow rates.

Increasing the MIF resulted in a decrease in the average TIL values for all challenge aerosols (wood, paper, and plastic). For example, for the wood combustion aerosol, the average TIL values were 0.6–1.0% at MIF = 30 l/min, 0.5–0.8% at 55 l/min, and 0.3–0.6% at 85 l/min. Several published FFR studies also reported that face seal penetration decreased with increase in breathing flow when challenged with particles >500 nm (Chen et al., 1990; Cho et al., 2010b). On the other hand, another FFR study observed no significant increase or decrease in face seal penetration by 8–400 nm particles associated with increasing breathing flow (Rengasamy and Eimer, 2011).

The size-independent (overall) TIL values for the half-mask respirator challenged with the three tested combustion aerosols are presented in Fig. 3. The highest TIL values were determined at MIF = 30 l/min and 30 breaths/min for all three challenge aerosols (peak TIL = 1.08, 1.33, and 1.87% for wood, paper, and plastic, respectively). The lowest TIL values (0.35, 0.60, and 0.75% for wood, paper, and plastic, respectively) were obtained at MIF = 85 l/min combined with 25 breaths/min. ANOVA with Tukey’s range test was performed to study the effects of combustion material, breathing flow, and breathing frequency on the TIL, and the data analysis results are presented in Tables 2 through 5.