key: cord-0308719-rono5rb0
authors: Pan, J.; Harb, C.; Leng, W.; Marr, L. C.
title: Inward and outward effectiveness of cloth masks, a surgical mask, and a face shield
date: 2020-11-20
journal: nan
DOI: 10.1101/2020.11.18.20233353
sha: 51c23ea29b73006d0fad4f99c2203b018ea22e3f
doc_id: 308719
cord_uid: rono5rb0

We evaluated the effectiveness of 11 face coverings for material filtration efficiency, inward protection efficiency on a manikin, and outward protection efficiency on a manikin. At the most penetrating particle size, the vacuum bag, microfiber cloth, and surgical mask had material filtration efficiencies >50%, while the other materials had much lower filtration efficiencies. However, these efficiencies increased rapidly with particle size, and many materials had efficiencies >50% at 2 m and >75% at 5 m. The vacuum bag performed best, with efficiencies of 54-96% for all three metrics, depending on particle size. The thin acrylic and face shield performed worst. Inward protection efficiency and outward protection efficiency were similar for many masks; the two efficiencies diverged for stiffer materials and those worn more loosely (e.g., bandana) or more tightly (e.g., wrapped around the head) compared to a standard earloop mask. Discrepancies between material filtration efficiency and inward/outward protection efficiency indicated that the fit of the mask was important. We calculated that the particle size most likely to deposit in the respiratory tract when wearing a mask is ~2 m. Based on these findings, we recommend a three-layer mask consisting of outer layers of a flexible, tightly woven fabric and an inner layer consisting of a material designed to filter out particles. This combination should produce an overall efficiency of >70% at the most penetrating particle size and >90% for particles 1 m and larger if the mask fits well.

and Schooley 2020), universal masking has emerged as one of a suite of intervention strategies for 29 reducing community transmission of the disease. There is a correlation between widespread mask 30 wearing (The Economist 2020), or at least interest in masks (Wong et al. 2020) , and lower 31 incidence of COVID-19 by country and between mask mandates and county-level COVID-19 32 growth rates in the US (Lyu and Wehby 2020), but a causal relationship has not been confirmed. 33

Due to a shortage of medical masks and respirators, some public health agencies have 34 recommended the use of cloth face coverings. While there have been numerous studies on the 35 ability of surgical masks and N95 respirators to filter out particles, far less is known about the 36 ability of cloth masks to provide both inward protection to reduce the wearer's exposure and 37 outward protection for source control. Ideally, a randomized controlled trial would be conducted, 38 but in the absence of such evidence, we can evaluate the ability of masks to block particles under 39 controlled conditions. 40

Reviews on the use of masks in both healthcare and non-healthcare settings to reduce transmission 41 of other respiratory diseases mostly show a protective effect. A systematic review and meta-42 analysis of interventions against respiratory viruses found that wearing simple masks was highly 43 effective at reducing transmission of severe acute respiratory syndrome (SARS) in five case 44 control studies (Jefferson et al. 2008) . In contrast, a review of 10 randomized controlled trials of 45 mask wearing in non-healthcare settings concluded that there was not a substantial effect on 46 influenza transmission in terms of risk ratio, although most of the studies were underpowered and 47

Methods 109

We tested nine materials that were fashioned into masks, one surgical mask, and one face shield, 111 shown in Figure 1 . To make the masks, we cut materials into 15.5 cm × 10 cm rectangles and 112 securely taped them to a frame tailored from a surgical mask, except for two designs that followed 113 instructions from the US Centers for Disease Control and Prevention (CDC). These included a 114 sewn mask made of two layers of a 200-thread count cotton pillowcase and a non-sewn mask cut 115 from a cotton t-shirt (Centers for Disease Control and Prevention 2020). The instructions for the 116 non-sewn mask used in this study have been supplanted with an updated design involving a large 117 square of fabric and rubber bands. The surgical mask had a single layer and was advertised to meet 118 ASTM level 1 specifications, which require ≥95% filtration efficiency of particles larger than 1 119 µm. We characterized the texture and structure of the masks using a scanning electron microscope 120 (FEI Quanta 600 FEG). Because it is not possible to generate or characterize particles spanning a 121 wide range of sizes with a single experimental setup, we designed several different protocols for 122 testing masks, optimizing among different types of equipment and detection limits, as described 123 below. 124 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 20, 2020. ; https://doi.org/10.1101/2020.11.18.20233353 doi: medRxiv preprint

Evaluation of the materials for filtration efficiency followed a protocol based on National Institute 130 of Occupational Safety and Health (NIOSH) testing procedures. Using a Collison 3-jet nebulizer 131 (BGI MRE-3, BGI Inc., MA, USA), we generated challenge particles of size 0.04 to 1 µm from a 132 2% NaCl solution. The particles filled a 280 L polyethylene chamber (Sigma AtmosBag, Sigma-133 Aldrich, ON, Canada), in which we placed a small fan to promote mixing. The temperature and 134 humidity inside the chamber were 22 °C and 25-35% RH, respectively. We measured particle 135 concentrations and size distributions using a scanning mobility particle sizer (SMPS 3936, TSI  136 Inc., MN, USA), with the particle density set to 2.165 g/cm 3 (NaCl) to convert from mobility 137 diameter to aerodynamic diameter. We cut out circular pieces of each material to mount in a 25 138 mm stainless steel filter holder (Advantec, Cole Parmer, IL, USA) that was connected to a vacuum 139 line whose flow rate was maintained at 2.7 L/min by a mass flow controller (32907-53, Cole 140

Parmer, IL, USA). The SMPS sampled from this line at a rate of 0.3 L/min, producing a total flow 141 rate of 3.0 L/min and a corresponding face velocity of 10 cm/s through the material. Clean make-142 up air flow to the chamber was provided through a high-efficiency particulate air filter capsule 143 (12144, Pall Corporation, MA, USA). We checked the material filtration efficiency of an N95 144 respirator and the microfiber cloth with and without a Kr-85 radioactive neutralizer (3012, TSI  145 Inc., MN, USA) or soft x-ray neutralizer (XRC-05, HCT CO., Ltd, Republic of Korea) after the 146 nebulizer, and did not find significant differences (Figures S1-S3), so we did not employ a 147 neutralizer in subsequent tests. To calculate the size-resolved filtration efficiency, we compared 148 measurements with the material in the filter holder to those made with an empty filter holder, as 149 shown in equation (1), where FE is the material filtration efficiency; DP is the particle diameter; 150

Cblank is the concentration of challenge particles measured downstream of the empty filter holder, 151

and Cmaterial is the concentration of particles downstream of the material: 152

We conducted these experiments in triplicate using three different pieces cut from each material. 153

In addition to challenging the masks with submicron particles generated by the Collison nebulizer, 154

we also tested larger particles ranging in size from 2 to 5 μm. We generated these from a 2% NaCl 155 solution using a flow focusing monodisperse aerosol generator (FMAG, TSI Inc., MN, USA). We 156 measured the particles using an aerodynamic particle sizer spectrometer (APS 3321, TSI Inc., MN, 157 USA). Because the APS samples at a flow rate of 1.0 L/min, we adjusted the vacuum line to 2.0 158 L/min to produce a total flow rate of 3.0 L/min, the same as used for testing smaller particles. 159

Clean make-up air was also applied as described above. We calculated the filtration efficiency 160 according to equation (1) in triplicate. We also measured the pressure drop of each material in the 161 filter holder using a differential pressure gauge (Minihelic II 2-5005, Dwyer Instruments, IN,  162   USA) . 163

We evaluated both inward and outward protection efficiency of face coverings using two manikins 165 mounted on opposite sides of a 57-L acrylic chamber (51 cm × 34 cm × 33 cm), mimicking the 166 situation of close talking, with a mouth-to-mouth distance of 33 cm (Figure 2a, b) . The "exhaling" 167 manikin was connected to a medical nebulizer (AIRIAL) filled with 2% NaCl solution, that 168 produced a flow rate of 10 L/min through 0.79 cm i.d. tubing. The "inhaling" manikin was 169 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 20, 2020. ; https://doi.org/10.1101/2020.11.18.20233353 doi: medRxiv preprint connected to both the APS and a vacuum line, with flow rates of 1 L/min and 14 L/min, 170 respectively, resulting in a total flow rate of 15 L/min through 1 cm i.d. tubing. Make-up air entered 171 the chamber around the top perimeter, to minimize disruption to air flow that might be introduced 172 by a port, and had a background particle concentration of at most 0.5% of that generated in the 173 chamber by the nebulizer. The air velocity at both manikin's mouths was 3. To evaluate inward protection efficiency, we attached face coverings to the inhaling manikin 177 ( Figure 2a , Figure S4 ) and tested two scenarios. In scenario 1, we ran the medical nebulizer for 3 178 s through the exhaling manikin, generating particles of size 0.5-2 μm. Using a three-way valve, 179 we set up the APS to sample either through the inhaling manikin's mouth or through tubing whose 180 inlet was placed outside the face covering, near the manikin's mouth. The flow rate through the 181 inhaling manikin remained constant at 15 L/min. We then waited 30 s for particle concentrations 182 to decay below the upper limit of detection of the APS, switched the valve to sample from outside 183 the mask, and measured the size distribution in the chamber for 5 s, denoted Cc1. We then switched 184 the valve so that the APS sampled through the inhaling manikin's mouth and measured particles 185 that penetrated the mask, denoted Cm. To account for the continually decaying particle 186 concentration in the chamber, we then switched back to measuring particles in the chamber again, 187 denoted as Cc2. The difference between Cc1 and Cc2 was less than 10% in all cases. Therefore, we 188 used the average of Cc1 and Cc2 to represent Cc at the time when we measured Cm. We calculated 189 the inward protection efficiency based on equation (1), replacing the numerator with Cm(DP) and 190 the denominator with Cc(DP). The temperature and humidity inside the chamber were 22 °C and 191 50-70% RH, respectively. In a separate experiment, we demonstrated that the three-way valve and 192 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

To evaluate outward protection efficiency, we removed the three-way valve and connected the 200 APS and vacuum line directly to the inhaling manikin ( Figure 2b ). In each test, we ran the medical 201 nebulizer for 30 s and then allowed particle concentrations to decay, as in scenario 2 of the inward 202 protection protocol, and we measured the chamber concentration (Cc1) using the APS at 1-s 203 resolution. After introducing the HEPA-filtered air to flush particles from the chamber, we then 204 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 20, 2020. ; https://doi.org/10.1101/2020.11.18.20233353 doi: medRxiv preprint put the mask or face shield on the exhaling manikin and ran the medical nebulizer for 30 s again 205 to measure the concentration (Cm). Then we flushed the chamber again, ran the nebulizer to 206 measure the chamber concentration Cc2, and calculated the average Cc as described in scenario 1. 207

We calculated the outward protection efficiency according to equation (1) as well. We conducted 208 all measurements in triplicate. 209

We evaluated the ability of the face coverings to block droplets larger than 20 μm, which is the regulator set at 165.5 kPa, resulting in a total flow rate of 10 L/min, the same as the flow rate of 216 the medical nebulizer. We filled the air brush with 2% NaCl solution and red food dye at a ratio 217 of 4:1. We taped five glass slides (75 mm × 25 mm) to the face of the inhaling manikin. We pre-218 cleaned each slide using 70% isopropyl alcohol wipes. 219

First, we sprayed the air brush for 3 s without the face covering on the exhaling manikin. We then 220 removed the glass slides from the inhaling manikin and inspected them under an optical 221 microscope at 10× magnification (EVOS FL Auto, Life Technologies, CA, USA). We put the face 222 covering on the exhaling manikin and repeated the same steps. To identify droplets on the slides, 223 we processed the images using ImageJ and then manually counted the stains and measured their 224 size with a limit of detection of 12.3 μm/pixel. Because the droplets spread upon impaction with 225 the slides, we corrected their size assuming a spread factor of 1.5, the ratio of the size of the stain 226 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 20, 2020. ; https://doi.org/10.1101/2020.11.18.20233353 doi: medRxiv preprint to the original diameter of the droplet (Johnson et al. 2011) . We conducted all measurements in 227 triplicate. 228

We used four different types of aerosol generators to cover a broad size range and to accommodate 231 different setups. The Collison nebulizer and FMAG, used to determine material filtration 232 efficiency, generated particles ranging in size from 0.04 to 1 μm and from 2 to 5 μm, respectively 233 ( Figure 3a 

We tested the material filtration efficiency of nine common homemade mask materials and one 248 surgical mask. We did not test the face shield because it does not allow air flow through it. Figure  249 4 shows results obtained using the Collison nebulizer and SMPS over the size range 0.04 to 1 µm. 250

The efficiency curves exhibit the expected U shape with a minimum in most cases in the range 251 0.1-0.3 µm, where no collection mechanism is especially efficient (Hinds 1999) . 252

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 10% for particles at 0.17 μm, lower than the reported value of 34.4% in another study (at a face 264 velocity of 6.3 cm/s) (Zangmeister et al. 2020), but its efficiency rapidly increased with particle 265 size to 75% for particles at 1 μm. The MERV 12 filter reached its lowest efficiency of 25% at 0.1 266 μm and had an efficiency > 50% at the extremes shown in Figure 4 . Common fabrics, including 267 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 20, 2020. ; https://doi.org/10.1101/2020.11.18.20233353 doi: medRxiv preprint the thin cotton and bandana (2 ply), had low efficiencies, mostly between 30% and 50%. The 268 fabrics fashioned into the CDC non-sewn and CDC sewn masks, bandana (1 ply), and thin acrylic 269 had even lower efficiencies of 5-40% for submicron particles. 

Most of the materials exhibited a much better material filtration efficiency for particles >1 µm than 275 for smaller ones, as shown by the black solid line in Figure 5 . The vacuum bag, microfiber, surgical 276 mask, and MERV 12 filter achieved 90% or higher efficiency at 2 μm, and thin cotton and coffee 277 filter were around 80% efficient at this size. The 2-ply bandana performed much better than the 1-278 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 20, 2020. ; https://doi.org/10.1101/2020.11.18.20233353 doi: medRxiv preprint ply bandana, with efficiencies of ~75% and <40% at 2 μm, respectively. The CDC non-sewn and 279 CDC sewn mask materials had efficiencies of ~50% at 2 μm, and their efficiencies increased with 280 particle size to up to 75% at 5 μm. The thin acrylic still ranked at the bottom. Its efficiency was 281 <30% at 2 μm but reached 75% at 5 μm. 282 

SEM images of the materials' structure can partly explain the differences in the performance. The 306 vacuum bag, which had the highest material filtration efficiency, had the smallest-diameter fibers 307 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 20, 2020. ; https://doi.org/10.1101/2020.11.18.20233353 doi: medRxiv preprint and such a compact structure that the pores or intervals between fibers were the least perceptible 308 among all the materials (Figure 1a) . The fibers of the microfiber cloth were also more tightly 309 woven than those of other materials (Figure 1b) , resulting in good filtration efficiency. The 310 materials with low efficiency were generally loosely woven, such as the bandana (1 ply), 200-311 thread-count pillow case used for the CDC non-sewn mask, cotton t-shirt used for the CDC sewn 312 mask, and thin acrylic (Figure 1g-j) . However, the tightness of the weave was not the only factor 313 influencing the filtration efficiency. For example, the fiber intervals were large for the surgical 314 

In this study, the inward protection efficiency (IPE) quantifies the capability of a mask, as worn 325 on a manikin, to protect the wearer by filtering out particles moving in the inward direction through 326 the mask, from the surrounding air to the wearer's respiratory tract. The outward protection 327 efficiency (OPE) quantifies the capability of a mask for source control, to filter out particles 328 moving in the outward direction through the mask, from the wearer to the surrounding air. After 329 being made into a mask, the vacuum bag still ranked first for protection efficiency in both 330 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 20, 2020. ; directions, with its IPE and OPE curves close to the material filtration efficiency curve ( Figure  331 5a), especially for particles larger than 1 μm. Both IPE and OPE were >50% at 0.5 μm and >90% 332 for particles larger than 2 μm. However, there were large variations in OPE for particles smaller 333 than 0.7 μm. The IPE and OPE were also similar to the respective material filtration efficiency for 334 the CDC-sewn and thin acrylic masks (Figure 5i, j) , though their performance was much worse 335 than that of the vacuum bag. The OPEs of the CDC sewn mask and thin acrylic mask were ~75% 336 and ~50%, respectively, for particles larger than 2 μm, and both masks were not effective at 337 blocking particles smaller than 0.7 μm. Notably, the OPE of the CDC sewn mask was slightly 338 higher than its IPE at 2.0 μm, whereas no significant differences (p>0.05) between OPE and IPE 339 were observed across all sizes for the thin acrylic mask. 340

In contrast, the microfiber and coffee filter masks had a much worse IPE and OPE than their 341 material filtration efficiency (Figure 5b, d) , indicating leakage and a poor fit. The OPE for the 342 microfiber mask was <25% for particles smaller than 2 μm, a difference of >50 percentage points 343 compared to its material filtration efficiency. Its IPE was slightly better but still 20-50 percentage 344 points lower than its material filtration efficiency for particles smaller than 2 μm. Similar trends 345 were also observed for the coffee filter, except that its OPE was slightly higher than its IPE at 346 particle sizes larger than 2 μm. 347

For the surgical mask, thin cotton, and MERV 12 filter, the differences between OPE or IPE and 348 material filtration efficiency were moderate, usually within 25 percentage points (Figure 5c, e, f) . 349

The OPEs of the surgical mask and thin cotton mask were higher than their IPEs but not 350 significantly; and these efficiencies were lower than the corresponding material filtration 351 efficiency. In particular, the average OPE of the surgical mask was substantially better than its IPE 352 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 20, 2020. ; https://doi.org/10.1101/2020.11.18.20233353 doi: medRxiv preprint at particle sizes ranging from 0.7 to 2 μm, but given the large variability in OPE, such differences 353

were not statistically significant (p>0.05). There were no significant differences (p>0.05) between 354 IPE and OPE for the MERV 12 filter across all sizes. 355

The bandana, CDC non-sewn mask, and the face shield had unique forms. The bandana was folded 356 in half in a triangle to mimic how people would normally wear it; its IPE and OPE fell in between 357 the single-layered and double-layered material filtration efficiency (Figure 5g) , with the OPE 358 higher than IPE at a particle size of 1 μm (p<0.05). The CDC non-sewn mask, whose fit can be 359 adjusted by tightening or loosening the straps, had an OPE that was significantly (p<0.05) higher 360 than the material filtration efficiency at sizes ranging from 1 to 2 μm. It is likely that stretching or 361 loosening the fabric altered its filtration efficiency. Its average OPE was also higher than the IPE, 362 whereas no significant difference was found between its IPE and material filtration efficiency. The 363 face shield did not block almost any aerosols smaller than 0.7 μm, as expected, for it did not fit 364 closely to the manikin and thus allowed virus-laden aerosols to travel freely around the shield. 365

However, it exhibited a decent OPE for particles at 5 μm (~75%) and an IPE of ~25% for such 366 particles. 367 Figure 7 compares the IPE and OPE across all masks. The vacuum bag mask had the best 368 performance in both directions, while the coffee filter mask, thin acrylic mask, and face shield 369 ranked at the bottom. The CDC non-sewn mask and surgical mask followed the vacuum bag 370 closely for OPE but not IPE. Interestingly, the OPE values for masks tested spanned a wide range, 371 whereas their IPE values were closer, except for the vacuum bag. In addition, direct comparison 372 of the two panels in Figure 7 reveals that OPE tended to be higher than IPE, illustrating that many 373 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

In response to a study that suggested that neck gaiters offer very little protection (Fischer et al. 379 2020), we measured the OPE of two neck gaiters, one made of thin 100% polyester and another 380 made of a double layer of microfiber fabric that was 87% polyester and 13% elastane. Their 381

average OPEs were at least 50% at 1 µm and >90% at 5 µm ( Figure S5, S7) , similar to the results 382 for the CDC non-sewn mask. When doubled over, the thin polyester neck gaiter achieved an OPE 383 of >90% over the size range of 0.5-5 µm ( Figure S6 ). Due to the late addition of these face 384 coverings, we were not able to measure their material filtration efficiency or IPE. 385

Droplet deposition analysis found no stains in the slides for all face coverings, indicating that all 386 of them were able to prevent droplets larger than 20 μm from spreading 33 cm away. 387 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 20, 2020. ; https://doi.org/10.1101/2020.11.18.20233353 doi: medRxiv preprint

For most of the face coverings tested, those with a high material filtration efficiency also had a 389 better OPE and IPE. One example is the vacuum bag, which achieved outstanding performance 390 compared to other materials with regards to material filtration efficiency, IPE, and OPE. It was 391 able to filter out at least 60% of particles under perfect conditions and had an OPE and IPE of at 392 least 50% and 75%, respectively, for particles 0.5 μm and larger. The MERV 12 filter, surgical 393 mask, thin cotton, and CDC sewn mask also had decent material filtration efficiencies, OPEs, and 394

IPEs, whereas the thin acrylic mask performed worst or near-worst on all three metrics. However, 395 there were some exceptions, such as the microfiber cloth and coffee filter. The material filtration 396 efficiencies of these two masks was much higher than their OPEs and IPEs (Figure 5b, d) . The 397 coffee filter and microfiber were thick and stiff, resulting in a poor fit with larger gaps between 398 the manikin and the mask, through which particles could short circuit the mask. In contrast, the 399 vacuum bag was thin and soft, which allowed it to conform to the face easily and achieve a high 400 IPE and OPE. Hence, we propose that the stiffness of the material impacts the fit of the mask and 401 can be responsible for large discrepancies between the material filtration efficiency and OPE and 402 IPE. Additionally, differences in mask use among individuals will lead to variability in fit and thus 403 effectiveness. 404

The CDC non-sewn mask was another exception. Generally, the IPE or OPE should be lower than 405 the material filtration efficiency because the latter was tested in a filter holder with no opportunity 406 for leaks. Nonetheless, the CDC non-sewn mask had a higher OPE than its material filtration 407 efficiency. This unexpected result may be due to its unique form, resulting in a different way of it 408 being stretched. Its two straps can be adjusted to fit it more tightly to the manikin face, especially 409 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 20, 2020. ;

to the mouth opening. Hence, the increased pressure caused by the expiratory flow was not able to 410 push the CDC non-sewn mask outwards to create gaps between masks and the manikin like other 411 conventional masks do (Lei et al. 2013; Liu et al. 1993; Mittal, Ni and Seo 2020) , minimizing air 412 leakage and bypass through the gaps. The stretching of the fabric may have caused a change in 413 pore size and woven structure, which further impacted the filtration efficiency. In addition, the 414 masks themselves also reduced the expired air velocity, which caused the particles to deposit 415 before they could reach the sampling device, as shown in other studies (Hsiao et al. 2020; Mittal, 416 Ni and Seo 2020; Tang et al. 2009 ). The combined effects of reduced gaps and reduced air velocity 417 resulted in a uniquely high OPE for the CDC non-sewn mask. For other masks with a conventional 418 shape, however, these two effects seemed compensatory during evaluation of OPE. While the 419 masks caused a decrease in the expiratory air velocity, they were also pushed outwards by the 420 outgoing flow, creating larger gaps between the masks and manikin. The contradiction in part 421 explained why the differences between OPE and IPE were not as large as expected for the masks 422 with conventional shapes, and why the bandana achieved an OPE better than expected, because it 423 created a larger plenum between itself and the manikin that provided additional containment of the 424 flow to lower the pressure drop and slow the flow jets through the gaps. 425

During the testing of IPE, we noticed that the vacuum through the inhaling manikin can suck the 426 mask tightly against inlet opening, thus reducing the size of any gaps. This can explain the small 427 differences between the material filtration efficiency and IPE, except for the coffee filter and 428 microfiber as they were stiff and hard to move. However, this phenomenon also illustrates the 429 tradeoff between breathability and filtration efficiency. Therefore, it is important to select fabrics 430 that can achieve both high filtration efficiency and low pressure drop for making masks. 431

We also observed variable hydrophobicity of the mask material during the testing of IPE and OPE. 432

The fabrics (e.g., thin cotton and thin acrylic) and coffee filter were wetted easily by droplets, . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 20, 2020. ; https://doi.org/10.1101/2020.11.18.20233353 doi: medRxiv preprint 446 Figure 8 . Masked deposition rate of 10 masks and a face shield as a function of the aerodynamic diameter. 447 Figure 8 shows the masked deposition rate as a function of particle size. Here, lower values are 448 better. The vacuum bag performed best, with a deposition rate of <10% across all sizes. The thin 449 acrylic mask, the coffee filter mask, and the face shield were the worst, with a 50% or higher 450 deposition rate at a particle size of 2 μm. Although there is considerable concern about exposure 451 to virus in the smaller particles, the particles with the highest deposition rate were those around 2 452 μm. For example, SARS-CoV-2 RNA has been detected in particles in the size range of 1-4 μm 453 (Chia et al. 2020). The smallest particle size considered in this analysis was 0.5 µm, but the 454 deposition efficiency of 0.3 µm particles in the respiratory tract is even lower, so it is possible that 455 concerns about mask efficiency at this size are overstated. 456 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 20, 2020. ;

This study was designed to test masks under tightly controlled conditions, which necessitate the 457 use of mechanical particle generation and manikins instead of humans. However, this approach 458 presents several limitations. The manikins are much more rigid than human skin, so masks may 459 not fit as tightly. A study involving a head form with pliable, elastomeric skin found that fit factors 460 of respirators were comparable to those measured on humans (Bergman et al. 2015) , whereas in 461 prior studies with head forms made of more rigid material, the fit factors were not as good 462 (Bergman et al. 2014 ). In addition, our manikins did not perfectly mimic human respiratory 463 activities because the aerosol flow traveled in only one direction in the inhaling manikin and the 464 exhaling manikin. As discussed above, inhalation and exhalation will alter the plenum between the 465 mask and the manikin, thus resulting in changes of the pressure drop and expiratory jets. We 466 investigated only one flow rate out of the possible spectrum from gentle breathing to vigorous 467 sneezing. Additionally, masks fit differently on different head shapes. Therefore, the performance 468 of the masks on a human face under real-world conditions will certainly vary from the 469 experimental results in this study. We did not test masks constructed of multiple layers of fabric, 470 as prior work has shown that overall filtration efficiency is readily predicted by combining 471 individual layers in series (Drewnick et al. 2020) . 472

Based on these results and other studies (Drewnick et al. 2020), we recommend a three-layer mask 473 consisting of two outer layers of a very flexible, tightly woven fabric and an inner layer consisting 474 of a material designed to filter out particles. The inner layer could be a high efficiency particulate 475 air (HEPA) filter, a MERV 14 or better filter (Azimi, Zhao and Stephens 2014), a good surgical 476 mask, or a vacuum bag. This approach produces a good fitting mask with high performance in 477 both directions. If the filter material is 60% efficient at the most penetrating particle size and the 478 outer layers are 20% efficient (Figure 1) , the mask would have a minimum efficiency of 74%. At 479 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 20, 2020. ; a particle size of 1 µm, where filter materials can easily have an efficiency of 75% and common 480 fabrics 40%, the overall efficiency would be greater than 90%. 481

In this study, we evaluated the material filtration efficiency, inward protection efficiency, and 483 outward protection efficiency of 10 masks and a face shield on a manikin, using NaCl aerosols 484 over the size range of 0.04 µm to >100 µm. The vacuum bag performed best on all three metrics; 485 it was capable of filtering out 60-96% of particles, and achieved an outward protection efficiency 486 of 50%-95%% and an inward protection efficiency of 75%-96%% for particles of aerodynamic 487 diameter 0.5 μm and greater. The thin acrylic performed worst, with a material filtration efficiency 488 of <25% for particles at 0.1 μm and larger, and inward and outward protection efficiencies of 489 <50%. The material filtration efficiency was generally positively correlated with either inward or 490 outward protection effectiveness, but stiffer materials were an exception to this relationship as they 491 did not fit as closely to the manikin. Factors including stiffness of the material, the way of wearing 492 the mask (e.g., earloops vs. tied around the head), and material hydrophobicity affected the fit of 493 the mask and thus its performance. Future studies may focus on the influence of material properties 494 on the fit of the mask, and how the transmission of real viruses, including SARS-CoV-2, is altered 495 by wearing the masks. 496

Flow Focusing Monodisperse Aerosol Generator 1520 to the Marr lab. This work used shared 501 facilities at the Virginia Tech National Center for Earth and Environmental Nanotechnology 502 Infrastructure (NanoEarth), a member of the National Nanotechnology Coordinated Infrastructure 503 (NNCI), supported by NSF (ECCS 1542100 and ECCS 2025151) . 504 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 20, 2020. ; https://doi.org/10.1101/2020.11.18.20233353 doi: medRxiv preprint

Recognizing and controlling airborne transmission of sars-cov-506 2 in indoor environments

The coronavirus pandemic and 508 aerosols: Does covid-19 transmit via expiratory particles?

Estimates of hvac filtration efficiency for fine and ultrafine 511 particles of outdoor origin

Correlation of respirator fit measured on human subjects and a static 515 advanced headform

Development of an advanced 519 respirator fit-test headform

How to make masks

Detection of air and surface contamination by sars-cov-526 2 in hospital rooms of infected patients

Physical distancing, face masks, and eye protection 530 to prevent person-to-person transmission of sars-cov-2 and covid-19: A systematic review 531 and meta-analysis

Aerosol 533 filtration efficiency of household materials for homemade face masks: Influence of 534 material properties, particle size, particle electrical charge, face velocity, and leaks

Flexible nanoporous 537 template for the design and development of reusable anti-covid-19 hydrophobic face masks

Low-cost 540 measurement of face mask efficacy for filtering expelled droplets during speech

Characterizing exhaled airflow from breathing and talking

A letter about the airborne transmission 545 of sars-cov-2 based on the current evidence

Aerosol Technology: Properties, Behavior, and Measurement of Airborne 548 Particles

Covid-19: An aerosol's 550 point of view from expiration to transmission to viral-mechanism

Physical interventions to interrupt or reduce the spread of respiratory viruses: 554 Systematic review

Modality of human expired 558 aerosol size distributions

Aerosol filtration 561 efficiency of common fabrics used in respiratory cloth masks

Simulation and evaluation of respirator faceseal 564 leaks using computational fluid dynamics and infrared imaging

Respiratory virus shedding in exhaled breath 568 and efficacy of face masks

A cough aerosol 571 simulator for the study of disease transmission by human cough-generated aerosols

Respirator leak detection by ultrafine 574 aerosols: A predictive model and experimental study

Aerodynamic analysis of sars-cov-2 in two Wuhan hospitals

Community use of face masks and covid-19: Evidence from a 580 natural experiment of state mandates in the us

Effectiveness of surgical masks 583 against influenza bioaerosols

Mechanistic insights into the 586 effect of humidity on airborne influenza virus survival, transmission and incidence

Influenza virus 589 aerosols in human exhaled breath: Particle size, culturability, and effect of surgical masks

The flow physics of covid-19

598 The effectiveness of respiratory protection worn by communities to protect from volcanic 599 ash inhalation. Part i: Filtration efficiency tests

Surgical mask filter and fit performance

Evaluation of decontamination methods 604 for commercial and alternative respirator and mask materials -view from filtration aspect

Application of nanofibers to improve the filtration 607 efficiency of the most penetrating aerosol particles in fibrous filters

Reducing transmission of SARS-Cov-2

Simple respiratory protection-evaluation of the 612 filtration performance of cloth masks and common fabric materials against 20-1000 nm 613 size particles

A schlieren optical study of the human 616 cough with and without wearing masks for aerosol infection control

Masks probably slow the spread of covid-19

Professional and home-made face masks reduce 622 exposure to respiratory infections among the general population

Covid-19 and public interest in face mask use

2020. 628 Nonpharmaceutical measures for pandemic influenza in nonhealthcare settings-personal 629 protective and environmental measures

Exhaled droplets due to talking and coughing

Infectious virus in exhaled breath of symptomatic seasonal influenza cases 635 from a college community

Concentrations and size distributions of airborne 638 influenza a viruses measured indoors at a health centre, a day-care centre and on aeroplanes

Filtration efficiencies of 641 nanoscale aerosol by cloth mask materials used to slow the spread of sars-cov-2

Household materials selection for homemade cloth face coverings and 645 their filtration efficiency enhancement with triboelectric charging

Defining the sizes of airborne particles that mediate 649 influenza transmission in ferrets