Benchtop assessment of sealing efficacy and breathability of additively manufactured (AM) face masks

The onset of the 2019 novel coronavirus disease (COVID-19) led to a shortage of personal protective equipment (PPE), medical devices, and other medical supplies causing many stakeholders and the general public alike to turn to additive manufacturing (AM) as a stopgap when normally accessible devices were not available. However, without a method to test these AM constructs, there continued to be a disconnect between AM suppliers and the community’s needs. The objective of this study was to characterize the pressure drop and leakage of four different publicly available AM face mask models with two filter material combinations, as well as to investigate the impact of frame modification techniques including the use of foam strips and hot-water face forming to improve fit when the masks are donned on manikin head forms. AM face mask frame designs were downloaded from public repositories during the early stages of the COVID-19 pandemic. AM face masks were fabricated and tested on manikin head forms within a custom chamber containing dry aerosolized NaCl. Pressure drops, particle penetration, and leakage were evaluated for various flow rates and NaCl concentrations. Results indicated that filter material combination and frame modification played a major role in the overall performance of the AM face masks studied. Filter material combinations showed improved performance when high filtration fabric was used, and the cross-sectional area of the fabric was increased. AM frame modifications appeared to improve AM face mask leakage performance by as much as 69.6%.


Introduction
The onset of the 2019 novel coronavirus disease  led to a shortage of personal protective equipment (PPE), medical devices, and other medical supplies [1][2][3][4][5][6]. Various communities including healthcare facilities, schools, and other essential workers required additional PPE to fulfill their functions, but supply chains were unable to meet these demands. Inadequate PPE supply and usage appears to have been linked to an increased risk of healthcare workers contracting COVID-19 [7][8][9]. As a result, these stakeholders and the general public alike turned to additive manufacturing (AM), commonly called 3D printing, to bridge these gaps when normally accessible devices or wearables were not readily available [2,[10][11][12]. While some 3D printed PPE such as door openers or face shields were seen by members of the public and other stakeholders to have relatively low risk, other wearables such as face masks may pose additional risks when donned for applications other than source control. 3D printed face masks generally have an undefined performance in regard to the level of protection they offer against viral aerosol particles and droplets. While hospitals were in urgent need of PPE, in some cases they were unable to accept 3D printed face mask donations because of performance concerns such as material properties and structural integrity [13]. Without a method to test these AM constructs, there continues to be a disconnect between AM suppliers and the community's needs. Additionally, evaluating what methods or modifications may improve 3D printed face masks would be beneficial.
There may be numerous considerations for designing and fabricating AM face masks for source control, or even AM barrier face coverings for known and validated levels of protection [14]. Some of the current standards for barrier face masks, medical/surgical masks, and respirators include ASTM F2299/2299 M [15], F1862/F1862M [16], F2101 [17], F3502-21 [14], ISO 16900-1 [18], and EN 14683 [19]. While multiple testing standards exist for these traditional PPEs, there may be additional considerations for AM face masks due to the lack of AM technology, material, and testing standards. This necessitates a need for robust testing methods for AM face masks and/or AM barrier face coverings to determine how they compare to other face masks or traditional controls, such as N95 respirators. These test methods may include evaluations of important quantities of interest (QOI) such as filtration efficiency (FE), leakage, fit factor, pressure drop, and liquid barrier protection. Unlike current U.S. Food and Drug Administration (FDA) cleared surgical masks/respirators which are generally single use and disposable, AM face masks may have 3D printed pseudo-rigid structural frames and filter retaining components which may allow the user to reuse some fabricated components. For these AM devices, standards to test the mask components and/or final mask forms are needed to evaluate their performance and evaluate if reuse adversely affects performance.
Donning face masks which attenuate viral transmission is an important source control strategy during scenarios such as the COVID-19 pandemic. Understanding the filtration efficacy of the various materials that can constitute a face mask is equally important. Guha at al. performed a systematic study on a variety of common household materials to see which ones performed the best during submicron NaCl aerosols, water droplets, and mucous mimicking macro droplets [20]. His team found that some filter material combinations had submicron filtration efficiencies above 40% although the pressure drop across those combinations was high, indicating potential considerations for breathing difficulties. Rogak et al. performed a similar study with micron based aerosols and found that many of the filter materials had a filtration of over 90% [21]. Rengasamy et al. evaluated five categories of common fabrics with both poly-and mono-dispersed aerosols ranging from 20 to 1000 nm [22]. They found that the materials had a wide penetration range (40-90%) against poly-dispersed NaCl aerosols and an even wider penetration range (9-98%) when discretely testing mono-disperse NaCl aerosols. Bagheri et al. conducted filtration experiments on commonly accessible materials for the general public and hospital settings [23]. They used particles occurring in ambient air in a method adapted from ASTM F2299 and found that four layers of dry baby wipes had a filtration efficiency of around 85%.
Respirators need to adhere to at least three performance metrics to perform reasonably well given their validated state: particle filtration efficiency, breathability, and fit factor. The U.S. Army is currently investigating AM Respirators as, among other things, a stop-gap measure to ensure the Department of Defense has the capability to manufacture PPE supplies during scenarios such as pandemics [24]. Efforts are even being made by He et al. to 3D print PLA struts on PLA electro spun nanofiber webbing to create mask filter structures, which the authors state can have similar performance to respirators [25]. He et al. state that their structures can achieve around 79% filtration efficiency with a single layer of their material using air particulate matter in their laboratory similar to dry NaCl. Monzamodeth et al. fabricated AM face masks and face shields using filament material extruded PLA to evaluate if simulating the PPE donned on mannequin head forms was possible [26]. They concluded that computational fluid dynamics (CFD) modeling was possible and that their filament printed PPE can be more efficient than home-made fabric masks, despite potential gaps in the printed layer spacing. Another issue that researchers have investigated was reducing the time to manufacture needed medical PPE using AM [3,27] which also fits into a current hurdle for AM in general: mass manufacturing.
At the start of the COVID-19 pandemic, there were stakeholders who immediately invested efforts into fabricating protective face masks for healthcare settings. Of the 7 current ISO / ASTM 52900 [28] defined AM technologies, material extrusion / FFF and polymer powder bed fusion / SLS may best accommodate AM face masks fabrication. As FFF systems are widely available and generally inexpensive [29], community available AM face mask design and instructions were typically influenced by FFF system capabilities. While SLS is more expensive and less available to the general community, it appears to be chosen due to its build volume capacity and ability to accommodate designer requirements while also accelerating fabrication time compared to FFF. Since these AM technologies appeared to be prevalent for personal protective wear and AM face mask fabrication, the material capabilities of these systems were inherently recommended. These include AM materials such as PETG, TPU, ABS, and PLA for FFF systems and nylon for SLS systems [30,31]. Other AM technologies such as vat-photopolymerization and material jetting may be capable of fabricating AM face masks, however at the time of this manuscript's efforts, this did not seem as common. Reasons could include longer print times, residual material residue, and support material waste.
AM has the capability to fabricate complex geometries that may be near impossible for some traditional fabrication technologies. However, AM still requires more research and innovation to improve potential material defects in the final finished components, such as porosity. It is well known that large quantities of voids and pores can be present in the final components of material extrusion systems, which in turn, may be detrimental to the components performance [32,33]. AM PPE performance could be impacted if these porous flaws are present in sufficient quantities and create detriments such as unintended fluid pathways which allow the infiltration of contaminated blood or air. Wang et al. investigated the effects of porosity on FFF fabricated PLA ASTM D638 specimens [34]. They quantified the number of pores in the specimens based on pore size and found that the porosity for the samples ranged from 4.05% to 6.32% depending on how the samples were fabricated. About 99% of the pores were 0.2 mm or below in size while about 1% were 0.2-1 mm in size. Flodberg et al. researched the effects of porosity on Selective Laser Sintered (SLS) tubes from PA12 powder with and without carbon fibers [35]. They state that PA12 without carbon fiber had about 4.7% porosity with the 20 largest pores accounting for 43% of the total porosity and had an average pore volume of 75 × 10 4 µm 3 . However, they also state that adding carbon fiber reduced the porosity in their tube samples to 0.68% with the 20 largest pores having an order of magnitude less average pore volume of 72 × 10 3 µm 3 .
Swennen, et al. proposed a proof-of-concept prototype AM face mask which was intended to supplement commercially available FFP2 and FFP3 respirators in cases of PPE shortage crises [36]. The mask consisted of two re-usable 3D printed components: the face mask frame and the filter support. Disposable components to the prototype mask included a polypropylene non-woven melt-blown particle filter membrane and a head fixation band. They also individualized the device by matching the wearers face to the frame design by using 3D facial scanning software. Swennen states that further clinical testing of the prototype is essential prior to widespread use. Another similar face mask, currently under evaluation in a clinical setting, is the StopGap Surgical Face Mask [37].
Workflows or processes applied to AM face mask designs may help improve the fit to a user's facial profile. To improve the performance of AM face masks, some designers and users suggest matching the mask to the users via 3D facial scanning [36], adding foam strips to the face-frame interface [38], or molding or face-forming [39] the AM frames using methods such as forced hot-air, hot-water, or microwave [2,40,41]. If done successfully, then it may be reasonable to conceptualize why and how these individualizations, augments, and post-processing routines could improve the fit of the AM masks and reduce aerosol leakage.
While there are efforts to produce AM wearables that may provide some level of personal protection, current designs, fabrication choices, and workflows still require progress. Duda et al. previously found that the AM face masks they evaluated did not measure up to the particle filtration efficiencies of commercially available FFP2/3 respirators [42]. Chichester et al. stated that AM face masks may be able to provide a stopgap measure in crisis condition when traditional N95 masks become unavailable, although cautioned that users should be made aware of their risks and extend of their testing [38]. Bezek et al. compared a select set of AM "respirators" and concluded that many of them had filtration efficiencies equivalent or inferior to cloth masks [43].
In this study, the performance of some publicly available AM face mask designs was evaluated. This was done to assess how well these masks may perform under certain simulated use conditions, such as when donned on a National Institute for Occupational Safety and Health (NIOSH) standardized manikin head form. The objective of this study is to characterize the pressure drop and leakage of four different AM face mask models with two filter material combinations, as well as to investigate the impact of using foam strips and hot-water face forming to improve fit for the chosen models. To the best of our knowledge, these are the first efforts to quantitatively evaluate and compare multiple AM face mask designs, filter materials, and leak reduction strategies using benchtop assessment methods. Additionally, this appears to be the first time AM face masks and N95 filtering facepiece respirators are compared using a custom designed and 3D printed head form with an elastomeric skin.

AM face masks
AM face mask designs for this study were selected from a range of open-source model repositories. These models were selected relatively early on during the COVID-19 Pandemicin May of 2020 -and were selected based on the number of downloads from their respective hosting website, their perceived ease of fabrication, and/or popularity of the model. While the authors acknowledge that there are many other mask designs that were not selected, it is believed the frames chosen were representative of the most popular designs available during the initiation of this project.
Four frame designs were considered for this study: Frame A [37], Frame B [44], Frame C [45], and Frame D [46] (Supplement 2) which are shown in Fig. 1. Frame sizes used throughout this effort are medium for Frame A and large for Frame B, while Frames C and D only had one size. Frame A was fabricated in Nylon 12 (PA2200) on an EOS P396 Selective Laser Sintering (SLS) printer. Frames B, C, and D were fabricated in polyethylene terephthalate glycol (PETG) and thermoplastic polyurethane (TPU) on PRUSA i3 MK3S fused filament fabrication (FFF) printers (PRUSA Research; Czech Republic). Efforts were made to use uniform print parameters for all FFF masks, but different considerations and printing parameters were required for rigid and elastomeric filaments. Print settings for the AM materials / technologies is shown in Supplement 1.
While evaluating the performance of each AM face mask design in various AM materials / technologies would be an interesting endeavor, this was not explored. The reason this was not explored is that designers of the masks generally prescribed AM materials / technology for their fabrication. An example of this is Frame A, which is prescribed by the designers to be fabricated in powder bed fusion nylon. Preliminary evaluations of the selected designs in different materials revealed that certain AM materials / technology combinations revealed complications, such as Frame A fabricated in PETG using FFF resulted in poor quality and significantly increased fabrication time. Additionally, time and funding limitations of the project did not allow for more experimental combinations.
Two filter material combinations were used: one layer of 1000 thread count cotton with three layers of a mask bandana [20], and two layers of melt blown polypropylene (MBP) also called Technostat 70 (Hollingsworth & Vose; East Walpole, MA) with one layer of 0.8 OSY spun bound polypropylene (Air Filters Inc.; Houston, TX). These are referred to as filter combination 1 and 2, respectively. To fully construct the AM face masks, ¼" cotton elastomer blend elastic was used as straps. Some test combinations used ¼" self-adhesive polyurethane foam strips 8709K42 (McMaster-Carr; Elmhurst, IL). The term "AM Frame" in this manuscript refers only to the fabricated AM frame including the filter cover, while the term "AM face mask" refers to the fully assembled mask (AM frame, filter combination, elastic straps, and any applicable frame modifications).
Four different frame modification groups were considered for PETG masks: As Printed, As Printed with Foam Strips, Form Fit, and Form Fit with Foam Strips (Fig. 2). Two frame modification groups were studied for TPU and Nylon masks as form fitting was not a studied variable for these manufacturing materials. The two test groups were As Printed and As Printed with Foam Strips.
Form fitting of the PETG AM frames was performed by pouring hot water over the masks and allowing them to soften in the hot water (~86 • C) for 30 s. Then the mask was pressed and held against a nylon static NIOSH medium face form until the frame cooled. For some models, this process was repeated 2-3 times to ensure optimal form fitting.
After the AM frames were printed, filter material was inserted and restrained in the frame per the frame's designed filter holding mechanisms. Then, 4 elastic straps were attached to the frames. Mask test groups which required foam strips had 1/4" inch adhesive foam placed around the mask edge. Form Fitted AM frames underwent the same assembly process after the AM frames were dry.

Aerosol exposure and measurement chamber
A custom aerosol exposure test chamber was fabricated. Standardized NIOSH head forms and other testing fixtures used to don and assess the PPE were placed inside the chamber with isolated plumbing used to sample aerosols that penetrate and leak through the PPE. The size and concentration of aerosols were used to assess filtration performance. Aerosolized sodium chloride (NaCl) particle generation was achieved using methods outlined previously [13]. A schematic and details of the aerosol exposure test chamber is shown in Supplement -Aerosol Leakage and filtration performance were evaluated using the test setup shown in Supplement 2. Particle leakage of the PPE were quantified using the metric defined by where PC D,BC is the downstream sampling-bias corrected particle count within a specified size bin, PC U is the upstream particle count within a specified size bin, k is a particular size bin, and n is the total number of particle size bins evaluated. This equation is similar to that used in works such as Tcharkhtchi et al. [47] and defines leakage as the ratio of the particle counts downstream to the particle counts upstream. Particle penetration, a term generally reserved for performance where face-frame leakage is not a key analysis factor, is also evaluated using Eq. 1. Pressure drop was measured using differential pressure transducers connected to the fluid ports outside the PPE in the ambient chamber (upstream) and inside the PPE past the filtration material (downstream). Further information on the pressure drop measurement system can be found in Supplement -Aerosol Exposure and Measurement Chamber.

Face mask donning and material performance assessment
Three test fixtures were used to assess AM face mask performance: a static standardized NIOSH head form, a static standardized NIOSH head form with an elastomeric skin, and a baseline fixture which removes the effect of face-frame interface leakage. The NIOSH medium head form was chosen due to its simplicity of fabrication and its geometric fidelity, which is consistent with the AM system used to fabricate it (EOS P396). To improve upon the facial skin compliance not demonstrated by a rigid surface, a NIOSH medium head form with a uniform elastomeric skin made from silicone was also designed and constructed. Lastly, to observe the performance of the face masks with no leakage from the face-frame interface (i.e., simulate a perfect fit), a flat plate baseline test fixture was constructed (Fig. 3c).

Face mask baseline performance
Obtaining performance metrics for the AM face masks and control respirators without the influence of the face-frame interface factor (i.e., perfect fit) was accomplished by simulating constant flow rate inhalation through the face masks and respirator control samples sealed on a flat surface (Fig. 3c). The baseline test fixture has pressure, particle sampling, and vacuum sheathing ports to allow for particle sampling at a given flow rate. This experimental method is called the "baseline" setup since it establishes the baseline particle penetration and pressure drop performance in the absence of face-frame leakage.
Since the AM frames were too rigid to seal on a flat acrylic surface, flat cut models of each mask frame design were created, Fig. 3d. Baseline frame models were obtained by cutting the original models in Pru-saSlicer (PRUSA Research; Czech Republic) to have a flat base. Due to the variance in mask design, each model was cut at a different z -height (direction perpendicular that of the filter material plane) but effort was taken to ensure internal volume of each mask shape remained as close to the uncut form as possible.
Baseline "cut" models were printed and assembled with the filter materials. The baseline models were then attached to acrylic plates and sealed with a layer of Gaff tape and RTV Silicone. The plates were then attached to the baseline test fixture which in turn was placed inside the aerosol chamber and connected to the sampling setup using conductive tubing and quick release connectors. These baseline models were then tested using the setup described in Supplement 2.

Leak performance using NIOSH head forms
To circumvent costly and time-consuming Institutional Review Boards (IRB)s and human fit-testing, one method to assess fit of PPE such as masks, respirators, and face shields is to don them on manikin head forms [48]. Standardized digital models provided by NIOSH, which represent a select percentage of the U.S. workforce [49], may be appropriate for this reason. In this work, two head forms were designed and fabricated to assess the performance of the face masks described above. One head form was relatively simple to fabricate and had a rigid facial profile while the second had a more complex fabrication workflow but attempts to mimic a more realistic, elastic, facial skin profile using  silicone. It was anticipated that this may facilitate a more accurate assessment of how the face masks would perform on a human subject.
Standardized head digital models were obtained from NIOSH's website [50]. For the scope of this manuscript, the medium head model was selected as it was representative of the widest range of face/head morphologies [51]. Models were downloaded as CAD files and modified in SolidWorks 2020 (Dassault Systemes; Waltham, Massachusetts) to accommodate instrumentation. The model was exported as an STL file and prepared for additive manufacturing in Magics 24.1 (Materialise NV; Leuven, Belgium). For light-weighting and material saving purposes, the models were hollowed to a uniform 5 mm shell which maintained the outer facial contours. The model was then printed in Nylon (polyamide 2200) on an EOS P396 (Electro Optical Systems GmbH; Krailling, Germany). Fig. 3a shows the rigid NIOSH head form.
The elastomeric head form base was constructed in a similar fashion to the rigid NIOSH Medium head form. However, during the digital design workflow, a 5 mm uniform shell was removed from the outer contour of the solid model to create a constant thickness depression and allow room for the silicone "skin" to be added. The resulting depressed head form structure was hollowed out as previously described and printed in nylon. The elastomeric skin layer was constructed by molding the full-sized NIOSH medium head form in MoldStar 15-Slow (Smooth-On; Easton, PA) and then creating a cast using the elastomeric head form base in EcoFlex 00 -20 (Smooth-On; Easton, PA). This skin cast was designed to be a uniform 5 mm thickness and was cast in two pieces (front and back) and adhered to the underlying nylon form, Fig. 3b. Additional information on the construction and validation of the NIOSH head forms is detailed in Supplement -Leak Performance Using NIOSH Head Forms.

Inherent filter material performance
The performance of various filter material combinations was evaluated independent of the face mask frames using a smaller exposure chamber with sampling ports similar to the larger chamber described in this study [20]. This was done so that significantly shorter aerosol steady state times were achieved to record particle penetration values. All filtration combination samples had a constant exposed cross section of 12 cm 2 across which a vacuum flow was pulled. Two NaCl concentrations were used for the inherent filter material performance assessment: 50 mg NaCl/100 mL H 2 O and 1000 mg NaCl/100 mL H 2 O, referred to as 50 and 1000 mg from here on. The higher concentration was selected to determine if the aerosol loading would impact the filtration as it will expose the filter material to a higher number of dry NaCl particles and broaden the particle size distribution in the chamber. The flow rate for the vacuum was set so that the face velocity imposed on the filtration combinations was equivalent to that of the AM face mask filter area designs under 9 SLPM conditions. The largest and smallest AM face mask filter cross-sectional areas, frames A and B, respectively, were used for this evaluation. In the smaller chamber, to keep the velocity through the filter materials equal, Frame A and B's 9 SLPM flow rate were achieved by using 2.7 SLPM and 6.5 SLPM, respectively. Three samples for each combination were tested. Optical imaging, Fig. 4, of the filter material was conducted using a HiRox RH-2000 microscope, a MXB-2500REZ Zoom Lens at 35X (4.26 µm resolution), and RH-2000 Ver 2.0.40 software (Hirox Co Ltd.; Tokyo, Japan). The filtration QOI for the inherent filter material performance is the filter particle penetration, Pen Fil . Furthermore, the pressure drop of the filter material was investigated at the respective flow rates.

Test parameters and performance assessment
Various tests to assess the source of AM face mask leakage was investigated. By isolating potential leakage sources for the masks, a better understanding of the performance losses may be obtained. Inherent filter material tests establish the particle penetration and pressure drop of the filter material combination at given face velocities similar to that of face masks in the baseline test. The baseline set up established a leakage value due to the filter material particle penetration, the filter material constraining mechanism of the frame (e.g., filter cap), and potential leakage through flaws in the AM frame (Eq. 2). The head form models established leakage of the filter material, AM frame, and frame-face mismatch (Eq. 3).
After the AM face masks were mounted in the baseline fixture, aerosol testing was performed at a constant vacuum flow rate of 9 SLPM and measured using the SMPS. This corresponds to light breathing by a human subject. Two NaCl concentrations were used for filter combination 2, 50 mg and 1000 mg. Only the 50 mg concentration was used for filter combination 1. This was because filter combination 1 was observed to clog which perturbed the performance results of the baseline test over time. At flow rates beyond 9 SLPM, preliminary experiments revealed high flow resistance and NaCl clogging effects for filter combination 1 which could not be adequately measured using the current test setup. Due to these experimental limitations, NaCl leakage data was only collected at 9 SLPM. It should be noted that the NaCl solid particle filtration evaluation method is not in accordance with U.S. CFR Title 42, Part. 84 [52].
Three upstream and three downstream particle classification samples were collected for each frame/filter material combination. The first samples for each stream were discarded to account for possible particle size separation errors in the first collection. For filter combination 2, only AM mask frames with the largest and smallest filter material cross sections were tested. Pressure drops were evaluated at 9 SLPM and 20 SLPM for filter combination 1 and 9 SLPM, 20 SLPM, and 85 SLPM for filter combination 2. While 85 SLPM is a relevant flow parameter to test as per NIOSH [52], the systems flow capacity for the baseline setup was exhausted beyond 74 SLPM, so pressure drop was only evaluated up to 74 SLPM for the baseline setup. 3 M 1860 N95 respirators (3 M; Saint Paul, MN) were also assessed in the baseline setup as a reference. Three replicates were tested for each baseline test. The filtration QOI for the baseline test is the baseline leakage where Leakage Frame is the leakage from the AM frame and AM components. This could be from the filter retaining cap or potential porous flaws in the frame itself. Pen Fil and Leakage Frame are measured experimentally using the definition specified in Eq. 1. All assembled AM face mask designs and modification combinations were donned on the medium rigid NIOSH head form. 3 M 1860 N95 respirators were also used as a reference. Placement of the face masks and respirators on the head form was done in a manner to visibly estimate the best fit with minimal gaps. The elastomeric straps on the AM face masks were then tied with a constant tension to the best of the engineers' ability. Nose bridge band adjustment for the N95 was also adjusted to minimize gaps. Dry NaCl particle filtration testing was performed at a constant vacuum flow rate of 9 SLPM and measured using the SMPS. Pressure drops were evaluated at 9 SLPM and 20 SLPM. Two replicates were tested for each rigid NIOSH head form test. The filtration QOI for the head form tests is where Leakage Fit is the leakage from the face-frame fit effect. A smaller selection of AM face mask frames was evaluated on the uniform elastomeric head form so that a larger test set of filter material and NaCl concentration could be assessed. All frame modifications for these mask designs were tested. Frame A and Frame B were chosen since they represented the largest and smallest designed filter material cross section areas, respectively. Again, 3 M 1860 N95 respirators were used as a reference, placement of the masks was done to minimize gaps, and elastomeric straps were tied as consistently as possible. Like the baseline test, one concentration was used for filter combination 1 (50 mg) and two NaCl concentrations were used for filter combination 2 (50 mg and 1000 mg). Pressure drops were evaluated at 9 SLPM and 20 SLPM for filter combination 1 and 9 SLPM, 20 SLPM, 74 SLPM, and 85 SLPM for filter combination 2. Two replicates were tested for each uniform elastomeric head form test. The test matrix for the AM face masks, modifications, and head form combinations is shown in Table 1.
Supplementary tests were performed with 3 M 1860 N95 filter material coupons (Supplement 6b) in Frame A to demonstrate the tradeoff in filtration performance and breathability for the selected AM face mask (n = 3). Preliminary tests and observations showed that a silicone gasket (Supplement 6a) was needed to acquire low particle leakage levels expected of the inherent N95 filter material. Additionally, the 3D printed nylon frames had two different amounts of applied RTV silicone sealant (Supplement 6c and d) to test if there was a significant contribution of leakage from the nylons potential print porosity. If there was a significant print porosity effect on the leakage, the standard deviation of the leakage results would be high when considering all 3 baseline samples (i.e., the RTV silicone would significantly assist in preventing leakage).

Inherent filter material performance
Particle penetration through the filter material was lower for filter combination 2 for all testing conditions (Fig. 5). Pressure drops across filter combination 1 were always higher than those for filter combination 2. Additionally, increasing the NaCl concentration from 50 mg to 1000 mg appeared to slightly increase the particle penetration through the filter material. Since the 1000 thread count cotton layer was known to clog at higher NaCl concentrations, filter combination 1 was not evaluated at the higher NaCl concentration conditions. For the equivalent face velocity of the A Medium filter area (at 9 SLPM), the do-ityourself (DIY) filter combination 1 had 15.5 and 4.2 fold more pressure drop and NaCl penetration respectively than filter combination 2.

Baseline Filtration and Pressure Drop
Leakage and pressure drop across all AM frames were recorded for all AM frame types and material combinations to estimate the leakage contribution of the frame-face interface (baseline). Due to restraints inherent to the baseline test fixture, 85 SLPM was not obtainable and the highest flow of 74 SLPM was used instead. At a vacuum flow rate of 9 As Printed x x x x As Printed Foam x x x x SLPM, results indicated that leakage and pressure drops were higher for filter combination 1 than filter combination 2 (Fig. 6a) which is consistent with the inherent filter material results shown in Fig. 5. Additionally, an increased NaCl concentration did not appear to consistently affect leakage or pressure drop. Increasing vacuum flow rate increased the pressure drop for all testing conditions, however frames with filter combination 2 maintained noticeably lower pressure drops than filter combination 1 at 20 SLPM (Fig. 6b). When AM frames were tested with filter combination 2 at 74 SLPM, the pressure drops were still generally less than those for filter combination 1 at 20 SLPM. Filter combination 1 was not evaluated at flow rates higher than 20 SLPM because the resulting pressure drops were greater than the maximum range for the sensors used in these experiments. Using 50 mg of NaCl exposure at 9 SLPM, the A Medium frame with filter combination 2 had 3 orders of magnitude more leakage than the experimentally measured N95 reference using the baseline setup. If a 95% N95 filtration efficiency reference is used, then the AM A Medium face mask has 32.6% more particle penetration. The baseline pressure drop for the same AM frame and filter combination was 38.6% and 18.3% less than the N95 for flow rates of 9 and 74 SLPM, respectively. This comparison shows what might be expected for the best observed AM face mask performance compared to an N95, if a good fit could be obtained.
With the assistance of a laser cut silicone gasket, N95 filter material coupons in the A Frame did show respectable leakage performance compared to N95 filtering face piece respirators (Supplement 6e), but at the expense of significant pressure drop (Supplement 6 f). The pressure drop increase of the A Medium baseline test with the N95 filter coupons was 201.8%, 231.3%, and 257.2% more compared to the commercial 3 M 1860 N95 respirator. This observation could be an indicator that breathability may be compromised. The small leakage standard deviations of the A Medium frame with the N95 filter material coupons (0.017 ± 0.007) implies that the leak contribution of the 3D printed nylon frame (with minimal and significant RTV silicone) was potentially insignificant. Had the standard deviation been vastly larger, then that may have implied that the RTV silicone covering large portions of the frame assisted in reducing leakage of the nylon print.

AM frames on head forms
Leakage and pressure drop at 9 SLPM were recorded for the N95 and Frame A mask with and without foam strips and with both filter combinations. Both were tested on the rigid (Fig. 3a) and uniform elastomeric (Fig. 3b) head forms. Results show that leakage was higher on the rigid NIOSH Medium head form for most frame designs regardless of frame modifications. The addition of foam strips and/or form fitting was generally observed to decrease leakage. Pressure drops generally increased for the elastomeric head conditions compared to the rigid head form except for when foam strips were applied to the AM face masks. Minimizing both leakage and pressure drop would generally indicate a better face mask performance, although the effects are linked since a reduction in leakage for these face masks inherently increases pressure drop across the filter. It was observed that the leakage of the N95 on the elastomeric head form was less than 0.05 while the N95 leakage on the rigid head was approximately 0.55. These results lend evidence that the elastomeric head form may simulate real-world performance more accurately. Therefore, the elastomeric head form was chosen for further AM face mask testing.
Using 50 mg of NaCl exposure at 9 SLPM, the A Medium frame with filter combination 2 had 14.3 fold more leakage and 59.7% less pressure drop than the measured N95 reference using the uniform elastomeric head form setup. If foam strips were utilized with the same AM face mask and setup, then the AM face mask only had 3.6 fold more leakage and 37.8% less pressure drop when compared to the N95. Increasing the NaCl concentration and flow rate to 1000 mg and 85 SLPM, the A Medium Frame with filter combination 2 as-printed showed a pressure drop decrease of 33.2% when compared to the N95. When foam strips are added to the A Medium face mask with filter combination 2, the pressure decreased by 69.6%.
An initial counter-intuitive observation from Fig. 7 is that the pressure drop trend for "A Medium Foam" with filter combination 1 is reversed (i.e., the pressure drop is higher on the Rigid NIOSH Medium head form than on the Uniform Elastomeric head form). Additional replicates were fabricated and tested which appeared to verify this trend. Given that the foam strips are open cell, they may have variable filtration properties as a function of flow through the foam and how compressed the open cell foam strips are. Also, an increase in foam compression may cause a decrease in overall flow leakage (e.g., Fig. 5. Filter material combinations were tested independently by adhering each fabric sample combination directly to an acrylic plate with a 12 cm 2 hole and using vacuum flow rates (6.5 SLPM and 2.7 SLPM for the Frame B Large and Frame A Medium, respectively) that match the face velocity of the AM frame filter cross sectional area to 9 SLPM. Error bars indicate standard deviation (n = 3).
increased pressure drop via reducing/closing the amount of fluid pathways) but allow for an increase in overall leakage (e.g., increasing NaCl infiltration by widening pores in the open cell foam).
Pressure drops were also recorded at 20 SLPM for filter combination 1 and 20 SLPM and 85 SLPM for filter combination 2 on the head forms, Fig. 8. Pressure drops tended to increase with an increase in vacuum flow rate. Again, pressure drops generally increased for the elastomeric head conditions except for when foam strips were applied to the AM face masks. However, no correlation was observed between pressure drops at different NaCl concentrations.

Uniform elastomeric head form testing
Post processing the as-printed frames (i.e., form fitting) and adding foam strips generally reduced the leakage in the rigid PETG face masks (Fig. 9). In all AM face mask frames, regardless of filter material, a leakage comparable to a certified N95 was not able to be replicated. However, with adequate fit using a flexible frame material, foam strips, and/or heat forming, some frames did achieve relatively low leakage of around 0.18. While the high thread count cotton in filter combination 1 may provide moderate filtration if the frame is adequately sealed, 0.60 leakage Fig. 6a (Frame A Medium), the fabric often resulted in higher pressure drops outside the acceptable range indicating a decrease in breathability. Overall, there did not appear to be a correlation between NaCl concentration used and changes in leakage or pressure drop.
The improvements of the frame modifications were quantified using the uniform elastomeric head form setup, a NaCl concentration of 50 mg, and a flow rate of 9 SLPM. The frame modifications of adding foam strips, face forming, and the combination of foam strips and face forming decreased the mean leakage of the PETG B Large face mask with filter combination 1 by 5.7%, 3.6%, and 19.9%, respectively, when compared to the as-printed state. When the same setup uses filter combination 2 the mean leakage is reduced 46.1%, 28.5%, and 38.8%, respectively, when compared to the as-printed state. Foam strip additions to the A Medium nylon mask decreased leakage by 58.0% and 69.6% for filter combination 1 and 2, respectively. Printing the B Large face mask in TPU instead of PETG decreased the leakage by 1.9% for filter combination 1 but decreased the leakage by 48.4% for filter combination 2. This may show that compliant elastomeric materials have different considerations regarding forces exerted on the mask potentially due to pressure differences and elastomeric strap forces. Adding foam strips to the TPU B Large face mask did not improve leakage but instead increased the leakage by 7.2% for the filter combination 1.

General observations
Filter material had a noticeable influence on pressure drops and overall performance of the final AM face mask construct. Although tightly woven filtration materials may have low penetration compared to other DIY materials, the associated pressure drops when installed in the AM frames may make these materials undesirable. This appears to be due to the reduced area for filter material in the chosen AM face mask designs, when compared to general cloth or surgical masks. Therefore, it may be best to pick a filtration material which is breathable for the AM face mask design and provides some barrier protection against relevant particulates.
Filter material combinations in these experiments changed the filter performance as a function of flow rate. High flow rates may deform the fabric of polyester or loosely woven fabrics allowing for additional leakage. Certain fabrics, such as the high thread count cotton in filter combination 1, were also observed to clog at high salt concentrations and have been shown to artificially increase filtration and reduce breathability during testing [20]. Therefore, care should be taken with such fabrics to ensure that filtration tests are conducted at low salt concentrations (e.g., 50 mg NaCl/100 mL water).
While 42 CFR 84 stipulates that the maximum permissible inhalation resistance for an air purifying particulate respirator is 35 mmH 2 O at 85 SLPM, clinical studies [53] have found that long term donning of respiratory protective devices with pressure drop greater than 10 mmH 2 O can often lead to significant side effects. For face coverings, ASTM F3502-21 has stipulated two cutoff values of pressure drops: 5 and 15 mmH 2 O. As seen in Fig. 6b when using tightly woven filter material (e. g., one-thousand thread count cotton) most AM face mask configurations would likely have pressure drops greater than 35 mmH 2 O at flow rates of 85 SLPM, indicating potential decreased breathability. This is likely because the total filter area through which flow is intended to occur for the chosen AM face masks are typically small compared to the total filter area of a traditional respirator or face covering. Indeed, the filter area of the AM face mask models studied in these experiments ranged from 16 to 39.4 cm 2 (Supplement 7) compared to a typical respirator's filter area of roughly 150 -200 cm 2 . Therefore, when a fabric with an intrinsically high-flow resistance is used as a filter for an AM face mask, designing the filter component with a larger cross-sectional area may help reduce the pressure drop. This small filter area design decisions may be due to, as an example, keeping the mask as least cumbersome as possible for the wearer, to conserve filter material stock, or because it facilitates ease of fabrication. Another alternative might be to use other low pressure drop filter materials such as polypropylenetraditionally used in respirators. For the same filter area, the pressure drops across commercial filter materials may be significantly lower (Fig. 6b) than the alternative DIY fabrics and may be a better choice for AM face masks to mediate high pressure drops for small filter areas.
Form fitting PETG AM face mask frames to the head form generally provided a better seal and therefore prevented more particles from entering the mask without passing through the filter. This improved overall fit and reduced leakage. When foam strips were added to the face-contacting section of the AM face mask, leakage was generally reduced when compared to AM face masks without foam strips. Though it should be noted that there may be a limitation to this since the foam strips used in this study are open cell and could create diminishing improvements at low levels of leakage. Overall, AM frame materials that can be form fit to the face using heat or pressure were shown to be more ideal for altering the frame to a custom fit for the wearer, especially when foam strips were added. The TPU material used to fabricate the B Large face masks did not show an improvement when foam strips were added. This could be due to the foam strips sliding with the compliant elastomeric frame and filter holder, the frame and foam slightly buckling/creasing to form gaps around the nose bridge region, or even due to smaller and unobserved print flaws in the frame itself. Further investigation into this cause is needed.
Gaps in the frame-face interface reduced the total aerosol flow through the filter material and thus the pressure drop in the manikin experiments tended to decrease ( Fig. 7 and Fig. 8). For the same flow rate and filter material, a lower pressure drop was recorded for a specific AM mask model on the head forms compared to the corresponding sealed flat-plate experiments (also referred to as baseline) indicating increased leakage. As an example, the A Medium nylon face mask with filter combination 2 was observed to have a mean leakage of 6.6% and 61.8% in the baseline and elastomeric head form test, respectively. If it can be assumed that the porous flaws in the frame, leaks through the filter holder interface, and all other variables affecting leakage are constant, then this 55.2% difference may be attributed to the leakage effects of the face-frame interface. When foam strips are added to the A Medium face mask then the elastomeric head form test shows a mean leakage of 18.8%, which is only a 12.2% difference.
Overall, AM face mask frames were able to form a better seal on the elastomeric head form than on the rigid NIOSH medium head form. This is likely due to the increased compliance of the 5 mm thick silicone skin layer used to mimic the elastomeric properties of human skin. This explains why the N95 mean leakage on the elastomeric head form was approximately 4.0% compared to 56.0% on the rigid NIOSH head form. However, the AM face mask frames were able to shift more easily on the elastomeric head form than on the rigid head form due to the compliance and smoothness of the silicone. This characteristic may explain the higher observed variations in pressure drop on the uniform elastomeric head form (Fig. 8).
AM fabrication methods contain inherent variabilities. Depending on the risk of the end use, it may be advisable that every frame and AM component be inspected for flaws (layer shifting, gaps, excess manufacturing debris, etc.) which may hinder their performance. These types of flaws were observed during this effort and an example is shown in Supplement 8. Causes for observed flaws could be explained by flaw mechanisms detailed in [54]. Verification tasks that could bolster confidence that the frame and AM components are reasonably leak free may help mitigate this risk.
Mannequin head forms such as the ones designed and fabricated in these works could also be used to initially assess the performance of wearables such as surgical masks, face shields, and more. This may also be true when evaluating how reuse methods may impact the useable life of the wearables. Additionally, since the head forms are mostly fabricated using AM technology, they may allow for more accessible and widespread testing capabilities of stopgap PPE.

Mask frame design and fabrication considerations
When considering material options, a balance between material properties should be considered. Flexible elastomeric materials, such as TPU, may stretch to better accommodate the face of the user but may not be as easily modified or form fitted as some rigid materials. Depending on the AM technology used, it is also important to ensure that all manufacturing debrissuch as un-sintered powder, filament stringing, or uncured resinare removed from 3D printed parts before final assembly. Stock material should be stored properly, and contaminated or defective stock material should not be used to fabricate AM frames. For example, TPU absorbs moisture from the air which can affect the printability of the filament and, therefore, requires specific storage conditions to mitigate these risks and ensure consistent print properties.
AM material selection may have an interesting leakage performance difference. From the baseline tests results, Fig. 6a, PETG masks which used a filter combination 1 appeared to consistently have lower mean leakages compared to the same face mask designs printed in TPU (e.g., B Large TPU had lower mean leakage than B Large PETG). This observation is switched for the filter combination 2 (i.e., B Large TPU had higher leakage than B Large PETG). It is hypothesized that the resulting higher pressure drop of filter combination 1 challenges the leakage of the AM face mask design and material combination so that fabrication defects, such as thru-porous pathways and filter retaining mechanism efficacy, are more readily apparent. Higher pressure differentials may drive higher leakage from potential print flaws.
Filter material selection is an important consideration to maximize filtration without creating unacceptable breathability. Layers of materials that offer additional filtering or droplet protection without increasing pressure drop may be useful. Cross sectional area of the filter material may be considered to mitigate unacceptable pressure drops since large pressure drops may indicate decreased breathability. Additionally, any modifications to improve fit between the mask frame and the wearer's face may be beneficial as this study showed decreased leakage for some frame and material combinations when form fitting and foam strips were applied.
Leakage effects from the 3D printed material should be considered: Comparisons between the inherent filter material performance and baseline tests demonstrate what the leak contribution of the frames porosity and filter retaining mechanism may be. The A Medium frames leakage performance in the baseline test with the Combination 2 material (0.11, Fig. 6) compared to the inherent Combination 2 materials performance (0.09, Fig. 5) shows minor differences. This could indicate that the leakage contribution from A Medium's printed nylon porosity and/or from the filter retaining mechanism is negligible. Experiments on the same frame with differing amount of RTV silicone applied to the printed material support this conclusion. A similar observation is made for the leakage of the B Large PETG frame with the Combination 2 material (0.22, Fig. 6) compared to the inherent Combination 2 materials performance at an equivalent face velocity (0.22, Fig. 5). However, print porosity of filter retaining leakage effects may not be negligible for the B Large TPU frame with the same filter material. This is because the frames fabricated in the TPU material showed a higher leakage in the baseline test (0.30, Fig. 6) versus the inherent filter materials performance (0.22, Fig. 5). It is hypothesized that this potential additional leakage may be reduced by improving either the masks design, the TPU materials printability, or optimizing the printers build parameters.

Limitations
Although the uniform elastomeric head form developed for this study appeared to be more representative of a human face than the rigid head form, it is likely not fully representative of human static (e.g., sweat) and dynamic physiology (e.g., cyclic breathing profiles). Therefore, extrapolating these results to human subjects is likely to require further validation.
Additionally, the silicone used to construct the "skin" layer of the uniform elastomeric head form may not fully represent the texture of human skin in all conditions such as when dry or sweating or the presence of facial hair. The texture of the silicone may have allowed some of the AM face masks used in this study to shift during the experiment despite best effort to create consistent testing conditions. A silicone skin head form with varying thickness that closely mimics a human's facial topology and tissue profile may provide more realistic results.
Findings from this work are heavily reliant on the selected AM frame designs and filter material choices. Other AM frame designs or filter materials may produce different findings as the pressure drop and fit is dependent on fabric choices, flow rates [20], and face mask design. While the NIOSH head forms have an established use in testing and certifying N95s, the authors acknowledge that the AM face masks tested in this study may have different performances on other face shapes such as any of the other four standardized head shapes from NIOSH.
Because the filter materials would have excessive pressure drop even at low flow rates, we had to significantly modify NIOSH standardized test methods and use lower flow rates. Since higher flow rates tend to yield the lowest filtration efficiencies, hence from a filtration efficiency perspective our test method was not worst case.
Exhalation breathing modes were not evaluated in the current test methods. The size of the aerosols, and their nature (e.g., charges, or wetness) may be different for exhaled aerosols. However, inhaled aerosols are likely to be smaller (caused by evaporation of larger droplets). These differences may also prompt even poor-quality masks to often offer greater protection for exhaled droplets [55] compared to smaller aerosols in inhalation studies [20] and hence constitute worst case testing in this regard.
Currently there exists a gap in the standard communities when it comes to evaluation of specific AM face coverings, and none existed for AM face masks at the time of this study. The authors chose to use standard practices for evaluating traditionally manufactured face masks but are aware that there are additional considerations for AM face masks that were unable to fully be assessed or addressed within the scope of this study such as material off gassing, in-depth analysis of leakage due to print inconsistencies, and dermal irritation effects due to AM material over short and long periods of time.
The effects of human factors were not evaluated in this study. While the general trend was that foam strips, especially combined with form fitting, decreased leakage, variations still may occur due to human error such as foam strip misplacement. Further testing with users of varying skill levels on the head forms with additional designs and filtration materials may be necessary to better ascertain face mask performances and potential variabilities.

Future needs
Using elastomeric static head forms may be an appropriate initial evaluation strategy for future AM face mask designs, leak reduction strategies, and even common PPE such as N95 ′ s and surgical masks. Future research that could benefit and further facilitate AM face masks as an acceptable stop gap measure include: establishing sufficient manufacturing verification tasks (i.e., process control) that bolster confidence the printed component lots are sufficiently leak free, human factor trials that assess how repeatable AM face mask performance is given different skill levels of the fabricator, studies on how the constructed face masks may impact the user from remaining manufacturing residuals (e.g., feedstock powder / particulate) or outgassing, best practices for evaluating the performance of AM face masks, best washing and reuse methods for the AM components, best practices for the end user of an AM face mask to verify that a selected design does what it is intended to do, optimal leak reduction strategies (e.g., what foam is best, how to apply the foam, solvent smoothing), and a systematic study on what commercial and do-it-yourself filter materials may work best for the AM face masks.

Conclusions
In future airborne pandemics, AM face masks may play an important role in curbing the spread of pathogens. However, the community's understanding of these masks is still in infancy. Our study highlights the following: a. The test methodologies employed in this effort may better help evaluate the performance of AM face masks. This is done by separately understanding the inherent filter material performance, the baseline leakage performance of the AM face mask (removing the face to frame fit effect), and the leakage that may be observed from the AM face masks while donning them on manikin head forms. b. AM face mask frame design, processing, and frame modifications are critical to face mask sealing efficacy. Filter material cross-sectional area should be carefully considered and designed to establish a balance between breathability and filtration performance; closely knit cotton fabrics such as the one used in this study are unsuitable for use in the chosen AM face masks. Especially with the small filter cross sectional areas present in the AM masks, as the pressure drop of the fabric is outside acceptable breathability ranges. c. The elastomeric head form may be more representative of static human physiology. Using a manikin head form with an elastomeric skin derived from a NIOSH model had better expected N95 performance observations compared a rigid manikin head form. This elastomeric head form may produce more accurate results when evaluating AM face masks. d. Depending on the AM technology and material, porous flaws may be present in the AM face mask frame and could potentially affect performance of the face mask, thus verification tasks to mitigate this risk may be needed to ensure minimal leakage. e. AM face mask modifications such as hot water face forming and adding elastomeric foam strips may decrease aerosol leakage in PETG masks when tested on an elastomeric manikin head form. Similar improvements were also observed for foam strip additions on the nylon frame face masks with observed decreases in mean leakage of 69.6%. The TPU face masks with foam strip modifications tested on the elastomeric head form only observed increases in leakage of up to 7.2%. f. Although the leakage obtained in the AM face mask designs used in this article generally fell short of the leakage that's expected from commercial N95s, future work in choosing appropriate materials with high filtration efficiencies, minimal pressure drops, and accomplishing better fits may help address this shortcoming. High filtration efficiencies could be achieved with the tested AM face masks, but at the potential expense of significant pressure drops.