Mask Material Filtration Efficiency and Mask Fitting at the Crossroads: Implications during Pandemic Times

The COVID-19 pandemic triggered the widespread use and need for respirators and face masks for the healthcare workers and public. In this study, several generally available respirators and mask designs were fit tested, and their materials were evaluated for filtration efficiency using 250 nm polystyrene latex particles. Efficiency testing was performed for 2 and 0.5 h at low (2.6 L min–1) and high (7.4 L min–1) airflows, respectively, using ~17.4 cm2 material area. As expected, all N95 and KN95 respirators passed the fit test, and their materials showed efficiencies > 95% for the entire experiment at both airflows. Of the three air filters used in the 3D-printed Montana masks, only the HEPA filter had a filtration efficiency > 95% at both airflows. Regardless of the insert material, the Montana mask failed all fit tests. Homemade duckbill masks made of Halyard H600 sterilization wrap and WypAll X80 reusable wipe also failed the fit test, and both filter materials had an average filtration efficiency < 95% at high airflows. To explain the filtration efficiency results, the structure and composition of all filter materials were determined using FE-SEM, and IR and Raman spectroscopy. In conclusion, when highly efficient materials are used in masks that do not fit the users properly, the potential of these materials to protect the users from aerosols is compromised. Therefore, the mask design is as important as the filtration efficiency of the mask material.


42
The Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) caused the pandemic 43 known as Coronavirus Disease 2019 , infecting >97 million people around the globe 44 with >2 million deaths worldwide as of January 20, 2021(Worldometer, 2020.  in the air as respiratory droplets and aerosols that are produced when an infected person coughs, 46 sneezes, or talks (Jianyun et al., 2020;Liu et al., 2020;Morawska and Cao, 2020;van Doremalen et al., 2020;CDC, 2020a). The aerosol transmission via speaking involves generation of a large 48 number of small (<2 µm) droplets per second (Morawska et al., 2009;Asadi et al., 2019). These 49 aerosols play an important role in spreading the infection (Morawska and Cao, 2020;Stadnytskyi 50 et al., 2020;Wang and Du, 2020). 51 Wearing a mask significantly reduces and prevents transmission of human coronaviruses from 52 both symptomatic and asymptomatic individuals (Leung et al., 2020). The most penetrating particle 53 size (MPPS) of most mechanical filters used as masks is between 0.1 and 0.3 µm, while filters with 54 electrostatic charge (e.g., N95) have smaller MPPS, typically below 0.8 µm (Ou et al., 2020). 55 These two flow rates are equivalent to respiration at rest (~30 L min -1 ) and during exertion (~85 L 152 min -1 ) (Lee et al., 2005;Bałazy et al., 2006a;He et al., 2014;Pei et al., 2020), considering the 153 average surface area of an N95 (~200 cm 2 ) (Roberge et al., 2010) and the surface area of our test 154 materials. In order to provide a fair comparison of the ability of each material to capture 250 nm 155 PSL particles, all materials were tested at the same flow rate. Since the filter material efficiency 156 depends on the velocity of the flow passing through the filter (TSI Application Note ITI-041,2020), 157 our data provide a fair comparison of the filtration efficiency of each material with the same surface 158 area, at the same flow velocity. However, respirators and masks investigated here had different 159 surface areas (~40, ~180, and ~270 cm 2 for the Montana masks, respirators, and duckbill masks, 160 respectively). Hence, the Montana mask would have a higher velocity of air flowing through its 161 insert materials (12.5 and 35.4 cm s -1 under light and heavy workloads, respectively); whereas the 162 respirators would have about the same (2.8 and 7.9 cm s -1 ); and the duckbill mask would have 163 smaller (1.9 and 5.2 cm s -1 ) compared to our experimental flow velocities (2.5 and 7.1 cm s -1 ). 164 Since filter material efficiency and flow velocity are inversely proportional (Bałazy et al., 2006a;165 He et al., 2013), the Montana masks would have a smaller filtration efficiency (i.e., less efficient 166 filtration); respirators would have similar (i.e., nearly the same filtration), and the duckbill masks 167 would have larger filtration efficiency (i.e., more efficient filtration) compared to the results 168 determined in our experiments. 169

Filtration Efficiency Calculation 170
The filtration efficiency was calculated every 1 min using Eq. (1), and then averaged for the 171 entire test time. Efficiency was also calculated based on particle size using particle counter size 172 bins. 173

Mask Fit Testing 174
Mask fit tests were performed using a TSI Portacount 8030 unit. The Portacount is a quantitative 175 analyzer which measures the concentration inside the mask, in the breathing zone, and outside of 176 the mask. The ratio of the number of particles inside the mask to the outside of the mask gives a 177 "fit factor" where higher values indicate a better fit. To supplement the number of particles in the 178 environment, a TSI NaCl particle generator 8026 was used during all fit test experiments. Both the 179 real time and OSHA fit tests were conducted on each mask. 180

Materials Characterization 181
The field emission scanning electron microscope (FE-SEM) images were acquired on a Hitachi 182 S-4700 FE-SEM (accelerating voltage 2 kV, beam current 10 µA, working distance 12 mm) in the 183 secondary electron imaging mode. The fiber diameters were measured using Image Processing and 184 Analysis in Java (ImageJ, NIH). The Raman spectra of individual fibers were recorded on a Bruker 185 Optics Senterra dispersive Raman microscope spectrometer with a spectral resolution of ~3-5 cm -186 1 , using 50x microscope objective and 785 nm laser excitation. The infrared (IR) spectra of the 187 KN95 had lower average filtration efficiency (96.8 ± 0.7%) than that measured by NIOSH (CDC, 205 2020d); however, it is still greater than the threshold level (95%) required for KN95 (3M Technical 206 Bulletin, 2020). All N95 and KN95 examined in this study passed fit testing (Table 3). 207

A C C E P T E D M A N U S C R I P T
Although Montana design has been shown to pass fit tests (The Montana mask, 2020b), the 208 Montana masks used in this work did not pass the fit test and failed each of the test parameters 209 (Table 3). Thus, our data suggest these masks fit poorly to our subject's face and given the printed 210 material is rather rigid, these masks can only fit a limited percentage of the population with a 211 specific face shape. In order to confirm that the efficiency of the insert materials did not contribute 212 to the poor fit test results, we also evaluated the average efficiency of these materials (Table 2). 213 Both the MERV 13-H and HEPA filters had average filtration efficiency >95% at low flow rate. 214 The HEPA filter performed >95% at high flow rate, but MERV 13-H efficiency decreased to ~85% 215 at this flow rate. The MERV 13-AIRx was the least efficient with average filtration efficiency 216 ~85% and ~65% at the low and high flow rate, respectively. According to its manufacturer (AIRx 217 Health MERV 13 Pleated Air Filter, 2020), the MERV 13-AIRx filter traps 90% of particles in the 218 1-3 µm size range, including bacteria and viruses, which can be <0.5 µm in size. The particles used 219 in this work are smaller than the size guaranteed by the manufacturer, hence our lower filtration 220 efficiency results (<90%). Although none of the Montana masks passed the fit test (Table 3), some 221 of the filter materials (especially the HEPA filter) used in these masks were >95% efficient in 222 blocking the passage of particles. This eliminates the filtration efficiency of the filters as a 223

A C C E P T E D M A N U S C R I P T
13 contributing factor to low fit test results in Montana masks and confirms our hypothesis. Our 224 investigation suggests 3D printed masks are typically not flexible enough to allow fit to a wide 225 range of face shapes. While the mask material may be efficient, the mask itself might not protect 226 its user. 227 Neither of the duckbill masks made of WypAll X80 or Halyard H600 passed the fit test (Table 3); 228 however, we suspected the poor fit test results were attributable to the design of the masks itself. 229 To test this hypothesis, we determined the average filtration efficiency of both materials at two 230 flow rates (Table 2). At low flow rate, the WypAll X80 average filtration efficiency was ~53%, 231 and it decreased to ~38% at the high flow rate. The latter value is consistent with the Michigan 232 mask response (2020) report, which found filtration efficiency of 30.6% at a flow rate of 60.8 L 233 min -1 . The Halyard H600 average filtration efficiency was >95%, and ~89% at low and high flow 234 rate, respectively. The material tested in this work was a new (unused) Halyard H600 sterilization 235 wrap but many masks used by the healthcare workers are made from used Halyard H600 wrap. We 236 hypothesized that efficiencies of the new and used Halyard H600 wraps might be different. We Regardless of the respirator/mask material tested, filtration efficiencies were lower at higher flow 242 rate compared to lower flow rate (Table 2). Likewise, many studies reported decrease in efficiency 243 at higher flow rates (Bałazy et al., 2006a;He et al., 2013). In most cases, the difference between 244 the efficiencies at low and the high flow rate was small (a few percent at most). For the MERV 13-245 AIRx and MERV 13-H filters, and the WypAll X80 this difference was much larger (~10-20%). 246 The average pressure drop (ΔP), a measure of material breathability, across all respirator/mask 247 materials at both flow rates are displayed in Table 2. All materials showed average pressure drops 248 <343 Pa, which is the maximum allowable pressure drop for N95 or equivalent respirators during 249 inhalation (3M Technical Bulletin, 2020; CDC, 2020e). This suggests acceptable breathability of 250 the masks made from the materials tested using the similar number of sheets reported in this study. 251 As expected, the average pressure drop across each material increased with an increased flow rate 252 (3M Technical Bulletin, 2020). Mask thickness varied from 0.1 to 0.4 mm ( Table 2)

Filtration efficiency as a function of particle size 256
The variation in filtration efficiency as a function of particle size was investigated for some of 257 the respirator/mask materials (Fig. 2). Both the low and high flow rate efficiencies were averaged 258 based on the particle counter bin sizes. The filtration efficiency as a function of the particle size for 259

A C C E P T E D M A N U S C R I P T
15 the HDX N95 did not vary significantly, and was nearly identical, at both the low and high flow 260 rates. However, the filtration efficiency of both the Wypall X80 and the Halyard H600 at both flow 261 rates was lower for smaller particles and increased with larger particle size. This is consistent with 262 the literature showing that the filtration efficiency of fibrous filters depends on the size of test 263 particles (Zangmeister et al., 2020; TSI Application Note ITI-041, 2020). The filtration efficiency 264 of the Halyard H600 was significantly larger than that of the Wypall X80 when particles with the 265 same size were tested. Some of the variations of the efficiency versus particle size shown in Fig. 2  266 in the 1-5 μm range is likely due to the low concentration of particles in this size range. 267

Filtration efficiency over time 268
Most studies do not provide information on the duration of the filtration efficiency tests. He Change in the filtration efficiency over time is an important factor for the mask's performance. 275 Since many mask users including healthcare workers may use masks for a prolonged time, the 276 present study evaluated the temporal changes in the filtration efficiency of the tested mask materials. 277 We were interested in determining whether the high efficiencies remain stable over a prolonged 278 time. The filtration efficiency of each respirator/mask material was first evaluated for 2 h at low 279 flow rate and then for 0.5 h at high flow rate. To the best of our knowledge, this is the first 280 continuous filtration efficiency experiment for particles larger than 200 nm. Temporal variations 281 in filtration efficiency, reported every 5 min, for both flow rates are shown in Fig. 3. 282 The N95 and three KN95s exhibited stable efficiency throughout the experiment duration at both 283 flow rates. Most respirators had stable and very high (>99%) filtration efficiencies at both flow 284 rates; however, at the high flow rate the ARUN KN95 and AOXING KN95 filtration efficiency 285 was stable but slightly lower (>96.8%). The filtration efficiency of the HEPA filter decreased only 286 by 1.5% (from 99.6 to 98.1%) at the low flow rate and was stable (~99%) at the high flow rate. The 287 MERV 13-H filtration efficiency was quite stable and decreased by only 2.5% (from 97.1 to 94.6%) 288 and 3.2% (from 87.5 to 84.3%) at the low and high flow rate, respectively. The efficiency of MERV 289 13-AIRx filter was lower and less stable over time; it decreased by 15.3% (from 88.9 to 73.6%) 290 and 7.9% (from 68.1 to 60.2%) at low and high flow rates, respectively. At low flow rate, the 291 WypAll X80 filtration efficiency decreased by 8.7% (from 62.7 to 54.0%) during the first 0.5 h 292 and by 16.0% (from 62.7 to 46.7%) after 2 h. At the high flow rate, however, it was much lower 293 but more stable over time with a decrease of only 2.8% (from 39.0 to 36.2%) during the 294 corresponding period. The Halyard H600 had smaller temporal change in filtration efficiency at 295 both flow rates with a decrease of 2.2% (from 97.9 to 95.7%) and 0.8% (from 89.5 to 88.7%) at 296 the low and high flow rates, respectively. During prolonged testing with continuous sampling, the 297 pressure drop values of almost all tested materials increased (average of 1.3 and 0.8 Pa for low and 298 high flow rates, respectively), as expected (Eryu et al., 2011). 299

Mask materials characterization 300
To explain the results of filtration efficiency and the average pressure drop across the materials, 301 structure of respirator and masks (e.g. number of sheets, fiber diameters) was examined using FE-302 SEM. Chemical composition of mask layers was determined using IR and Raman spectroscopy. 303 The results are summarized in Table 4

What will make an efficient mask 328
Many factors including the flow velocity and particle (size distribution, concentration, and 329 charge) as well as mask material characteristics (chemical composition, fiber diameters, 330 electrostatic charge, mask thickness, and packing density) can affect the filtration efficiency of a 331 given material (see Huang et al., 2013 andTcharkhtchi et al., 2021 for more details). Here we 332 provide explanation of some of these factors. Since we tested each material under stable and 333 constant flow conditions and under the same flow velocity, with the same particle type, size, 334 concentration, and charge, we conclude that our filtration efficiency results reflect the performance 335 of the mask and the respirators. We characterized the tested mask materials to further discuss what 336 could make a mask efficient. It has been suggested that masks comprised of polypropylene and

A C C E P T E D M A N U S C R I P T
19 polyester will be effective in filtrating particles (Lustig et al., 2020;Zangmeister et al., 2020). 338 Chemical characteristics of the respirators examined here supports this hypothesis; however, other 339 factors including but not limited to packing density, thickness, orientation, and fibers diameters in 340 the individual layers also appeared to play a role in the filtration efficiency. Thicker masks had 341 higher filtration efficiencies, with the exception of WypAll X80 and Halyard H600, which suggests 342 that factors other than thickness contributed to the filtration efficiency.

A C C E P T E D M A N U S C R I P T
20 vast difference between the filtration efficiency of WypAll X80 and Halyard H600 can be attributed 356 to the different filtration mechanisms of these fabrics. Whereas both made of polypropylene, 357 Halyard H600 is thinner but has the advantage of electrostatic charge (Ou et al., 2020) as well as 358 the finer and more densely packed fibers in the middle layer. All these features are characteristic 359 of the efficient N95 and KN95s examined here. Therefore, it appears there is not just one factor but 360 a combination of factors that contributes to a high mask filtration efficiency. Although the filtration 361 efficiency of the mask material is important, poor mask fit may lead to leakage. Thus, mask fit is 362 another important factor that should be considered (Tcharkhtchi et al., 2021). This work highlights 363 the importance of both filtration efficiency and mask fit as two major determinants of mask 364 performance. Poor fit of the mask to the user's face may render a highly efficient mask filter 365 material ineffective, as demonstrated by three air filters used in the 3D-printed Montana masks. All 366 three filters had an overall good filtration efficiency; but none of these masks passed the fit test and 367 they failed in each of the test parameters. The Montana mask was rigid and probably will only fit 368 a small population with a specific face shape. New designs with improved malleability should be 369 developed. As an example, nose bridge strips or additional layer (Huang et al., 2013;Zangmeister 370 et al., 2020) might improve the mask's ability to protect the users by leveraging the filtration 371 efficiency of high performing materials like Halyard H600 while still meeting the requirements for 372 air resistance. 373

A C C E P T E D M A N U S C R I P T
21 CONCLUSIONS 375 376 The efficiency of several mask materials in filtering 250 nm PSL particles was evaluated. All 377 examined respirators exhibited initial filtration efficiencies >95% (characteristic of the N95 or 378 equivalent respirators) that remained stable (>95%) during prolonged test durations (2 h at low and 379 0.5 h at high flow rates). All respirators passed the fit test. Only one (HEPA) of three (MERV 13-380 AIRx, MERV 13-H, and HEPA) air filters as potential filters in the 3D-printed Montana masks had 381 efficiency >95% at the high flow rate; the other two filters were less efficient (~65-86%). At the 382 low flow rate, the MERV 13-AIRx was the filter that performed poorly (~85%). The Montana 383 masks did not pass fit testing with any of these filters. The WypAll X80 and the Halyard H600 384 were tested for a potential use in the homemade duckbill masks. The Halyard H600 had efficiency 385 of ~97 and ~89% at low and high flow rates, respectively. The average filtration efficiency of 386 WypAll X80 was much lower (~53 and ~38% at the low and high flow rates, respectively). 387 Duckbill masks made from both the Halyard H600 and WypAll X80 fabrics failed fit testing. 388 Temporal variation of the filtration efficiency of all filter materials was evaluated. Most 389 materials showed stable filtration efficiencies (maximum change ~3.3% over time) except for two 390 materials (MERV 13-AIRx and WypAll X80) whose efficiency decreased by ~8-16%. These 391 results may guide the users in deciding which mask materials will be used for a prolonged time. 392 M A N U S C R I P T 22 The structure and composition of all filter materials were determined using FE-SEM, IR, and 393 Raman spectroscopy to examine the filtration performance. All high performing materials were 394 found to be electrostatically charged (with the exception of the HEPA filter) with at least one layer 395 comprised of densely and randomly distributed fibers with fine diameters in the 1-8 μm range. The 396 high filtration efficiency of a material contributes to the mask's potential to effectively protect the 397 user but the mask's ability to fit the user is undeniably as important. Hence, as shown in our study, 398 when masks made of highly efficient materials do not fit the user, the potential of these materials 399 for protecting the user is compromised.   Sci. Environ. Epidemiol. 27: 554 352-357, DOI: 10.1038/jes.2016.42. 555 Shokri, A., Golbabaei, F., Seddigh-Zadeh, A., Baneshi, M.-R., Asgarkashani, N., and Faghihi-556 Zarandi, A. (2015. Evaluation of physical characteristics and particulate filtration efficiency of 557 surgical masks used in Iran's hospitals, Int. J. Occup. Hyg. 7: 10-16. 558 Tcharkhtchi, A., Abbasnezhad, N., Zarbini Seydani, M., Zirak, N., Farzaneh, S., Shirinbayan, M. 559 (2021)   Minimum acceptable values were 20, which represent ~95% efficiency. Real Time results were 604 performed in the N95 compatibility mode (it limits the particle size recorded) to obtain the highest 605 fit factor prior to the OSHA test -29 CFR § 1910.134. If a value of at least 20 was not obtained in 606 the Real Time test, no further studies were performed. The fit test was performed over the entire 607 particle size range of the instrument. Each column corresponds to a specific action defined in the 608 OSHA fit test (e.g., talking, head movement, etc.) averaged over the experiment duration. 609 Table 4. Structure and composition of masks and mask materials.     Figure S9). Dimensions shown for soft pulp fibers are those of their cross-sections that were measured for fiber orientations observed in the FE-SEM images.