Visualizing face mask data.
How good are common materials?

Header image

The primary function of any mask is to filter particles out of the air that passes through the mask, which is relevant to preventing the spread of the COVID-19 and other respiratory diseases. Filtration occurs as a result of several mechanisms, including impaction, interception, diffusion, and electrostatic attraction.

Impaction and interception Diffusion and electrostatic

Impaction tends to dominate for larger particles, while diffusion is most significant for smaller particles. Electrostatic charges in the mask, as are typically present in N95-compliant masks, tend to enhance particle filtration but also tend to hold up rather poorly when the mask is sanitized. For example, washing most N95-compliant masks with soap and water will significantly reduce their filtration properties.

This web app seeks to visualize the overall filtration measured for a range of common materials, using data from Rogak et al. (2021). In accordance with that paper, we consider relatively large particles in the typical aerosol range, roughly from half to several micron. This compliments other studies that consider smaller particles. Limited information is also provided in terms of sanitization treatments (e.g., laundering) and their effect on material properties.

Before continuing, we make several general notes:

This is a key question!

We note that the focus of this data is on the materials themselves. As such, these curves do not consider leakage around the mask. This is another very important feature in constructing an effective masks but requires different kinds of tests that are wearer-specific. There are studies by other aerosol researchers and health professionals considering this aspect of mask construction.

Konda et al. (2020), for example, present the difference in filtration between candidate materials with and without leakage. This research found a gap representing only 0.5-2% of the surface area reduced the performance of an N95-compliant mask from 99% to less than 15% filtration. It is worth noting that those meausurements correspond to where a majority of the particles are near the lower end of the particle size range considered here. The largest human-generated particles (that is, the particles you can see, which are above the size range even considered here) will travel ballistically (that is, more of less on straight trajectories), such that leakage should trend downwards, even if some of the particles from the deflected spray will still escape the masks in practice.

Regardless, the mask will function better if one reduces the impact of gaps. We recommend maximizing the surface area of air exchange to improve effective ventilation. A cup-shape or duck bill design may be preferred. Other solutions to reduce gaps include external braces, sponge nose pads, or reinforced nose wires.

Overall, any mask is still better than no mask as (1) the remaining filtration efficiency (~15% for the N95-compliant masks in the Konda et al. (2020) study) will filter some of the smaller particles and (2) the mask should significantly reduce the fraction of larger droplets released into the air.

We used a flow rate of 1 LPM, yielding a face velocity of 4.9 cm/s, below that used in the NIOSH N95 standard (8-9 cm/s).

There are multiple kinds of particle diameters that aerosol scientists use when making these kinds of measurements, which may differ a bit from the physical diameter.

We phrase things in terms of the aerodynamic diameter, which is related to the drag force on the particles. This is more relevant for the impaction and interception mechanisms, which are dominant for the size range considered here.

We occasionally characterize the materials by their physical colour. This is done as there is often limited information available about the specific material properties (e.g., thread count). In this regard, the colour is simply used to denote different materials within a similar material class. Accordingly, this data should be interpretted as indicating the potential variability of filtration and pressure drop that could be associated with a given type of material.

A pertinent example is the knit cottons, which have variable quality despite being a similar material to the consumer.

Simply put, these are codes we assign to each material, composed of characters identifying a material, its structure, number of layers used during measurement, and sanitization treatment applied (e.g., laundered), if any. While we provide these codes for reference back to the associated paper, in most instances the reader can ignore these codes.

For those more interested:

1. The first letter or pair of letters generally denotes the material structure, using W for woven materials, K for knit materials, CP for cut pile materials (e.g., corduroy), and nW for non-woven materials. Composite masks, such as the N95-compliant materials tested, have different codes (e.g., ASTM for surgical masks that adhere to the ASTM standard). The different material structures can be added or removed from the plot using the checkboxes in the controls above the first plot.

2. The second component is a number used to identify the different materials within a given material structure. As the number is only used to identify the material, it does not contain any information about the material makeup or structure.

3. Next, some codes append x* to denote if multiple layers were used during the test. For example K4x2 is four layers of the K4 material (Double-knit jersey, yellow). If the x* component is missing, a single layer of the material was used.

4. Finally, some codes append a space followed by a code denoting if a sanitization treatment was applied to the material. Examples include, HS for heat treatment, IPA for isopropyl alcohol treatments, WD for laundering, and SW for washing with soap and water. If no such code is appended, the material is in its original condition. If santization was repeatedly applied, this latter part of the code will also be appended with x* (e.g., WDx10 denotes materials laundered ten times).

For example, a code denoting a knit material that has been laundered multiple times is K4x4 WDx4. A list of the codes is included in the table at the bottom of this page.


This first graphic uses data from Rogak et al. (2021) to examine the filtration properties of face masks alongside a second important property: pressure drop. Filtration has rather intuitive consequences for face masks. If a face mask has poor filtration properties, particles (perhaps containing virus) are allowed to pass through the fabric and reach the wearer, reducing the utility of the mask. The consequences of a poor pressure drop are less direct. Pressure drop denotes resistance. In this respect, a low pressure drop is preferable for two reasons: (1) it makes for a more breathable mask, which is less likely to irritate the wearer, and (2) it results in less leakage, improving the overall utility of the mask.

A good mask has a balance of both of these properties. The following plot considers these properties for some common materials. Uncertainties in the data are generally larger for low fitration efficiencies (negative filtration efficiencies occasionally reflect this; they are artifacts of the measurements and are not physical). Generally materials in the upper-left region of the plot (high filtration, low pressure drop) are the best. N95 masks are located on the upper axis, effectively filtering 100% of particles at these sizes.

Circles with thicker outlines indicate materials with some form of sanitization treatment applied (e.g., laundered or heated).

In the plot below, dashed lines correspond to lines of constant quality.

A good (or high) quality denotes a hybrid of good filtration while maintaining a low pressure drop. Infinite quality would correspond to a material that filters all of the particles, with no pressure drop. Naturally, real materials have trouble realizing this ideal. Masks with the highest quality in this study include N95-compliant and surgical masks. N95-compliant masks in some cases essentially blocked 100% of particles (within the limits of the current measurement apparatus), such that the quality approached ∞. Calculations of quality, Q, use the natural number as a base (this varies in the literature and should be checked between studies); pressures drop, Δp, in kPa; and the filtration efficiency, η, as a fraction; in the following expression:

Q = −ln(1 − η)/Δp

For more details on how quality is defined and what it means, we refer the reader to Rogak et al. (2021) and the references therein.

This classification refers to whether or not the mask meets the breathability requirements of the ASTM F3502 standard.

Note that the face velocity used for these measurements differs from the standard, such that a correction is applied (approximately halving the pressure drop requirements). Part of that standard also restricts the filtration efficiency at 0.3 micron, which is not considered in the provided classification.

Also note that the N95 Δp refers to a measured pressure drop, rather than to the NIOSH N95 standard.

The aforementioned quality (see the corresponding Q&A entry above) is a useful surrogate for determining how well a material performs in optimizing both pressure drop and filtration. The graphic below shows the range of quality observed for each material structure, in the form of a box-whisker plot. Here, the box corresponds to the 25th and 75th quantiles for a given structure, while the whiskers correspond to the 5th and 95th quantiles. In this plot, quality is capped at 1,000 and corresponds to the filtration at an aerodynamic dimaeter of 2.76 μm.


Considering size-resolved, multilayer mask filtration.

Not all particle sizes are filtered in the same way. For example, small particles tend to pass through masks more easily, resulting in lower filtration efficiencies. This second graphic examines the size-resolved mask filtration of the data from Rogak et al. (2021). Generally, curves higher on the plot are preferred; though, these materials may result in unreasonable pressure drops that make the mask uncomfortable to wear and lead to larger amounts of leakage.

In addition, this graphic estimates the filtration properties of masks composed of multiple layers. In this regard, solid, coloured lines indicate individual filter layers, while the black, dashed line corresponds to an estimate combining all of the selected feature layers.

Individual feature layers can be added, removed, or changed using the dropdown boxes and can include up to three layers, the number recommended by the WHO.

Estimates for multilayered masks are only approximate and are computed using the product of the particle penetration (one minus the filtration efficiency as a fraction) for individual layers.

It is worth noting that some recent work (e.g., Zangmeister et al. (2020) and Rogak et al. (2021)) suggests that the first layer may filter a disproportionate number of particles. In these cases, the presented estimates are likely to be overly optimistic.

In general, yes. It is well-established that the filtration efficiency consistently decreases as the particles get smaller for this size range. The most penetrating particle size (MPPS), that is, the minimum in these kinds of curves, is located to the left of the shown domain and occurs when the balance of the diffusion and impaction mechanisms are at a minimum. This typically occurs in the 50-500 nm range and is the focus of several other similar studies (see references in Rogak et al. (2021)). Here we focus on the larger particles, which remain relevant for bioaerosol applications.

Not necessarily! This graphic does not give an indication of the feasibility of the different layers, as they do not have a standardized weight, pressure drop, thickness, or compatibility. We again refer the interested reader to Rogak et al. (2021) for a more detailed discussion.

A pertinent example is gauze, which, when combined in multiple layers (up to 48 layers if selected from the dropdowns, where each filtration layer itself is composed of 16 layers of gauze), yield face masks that are extremely bulky.

Another example is the dried baby wipes, which would work fine as an insert but may be suboptimal in constructing the outer layers.

The colours of the solid lines correspond to the material classes for the individual filter layers, with the same colours as those used in the previous graphic (for example, yellow is for knitted materials).

While some materials contribute very little to the filtration properties, they may have other advantages. For example, some layers may absorb moisture. However, it is important that these layer do not add a lot of pressure drop to the mask, for reasons noted above.

These give a sense of the significance of the pressure drop, using the N95-compliant mask as the standard (or moderate) pressure drop. Pressure drops sufficiently below the measured N95-compliant pressure drop are considered low, while pressure drops higher than the N95-compliant masks are considered high. The real significance of the pressure drop will depend on other factors, such as how hard an individual is breathing. As such, these estimates are only an initial guide to the user.

Generally, for improvised masks, low pressure drops are preferred, as the masks do not fit as well as N95-compliant masks.

This classification refers to whether or not the mask meets the breathability requirements of the ASTM F3502 standard, as per the note above.

Clicking this button will copy a link to the clipboard for the current mask configuration. One can then paste this link anywhere to reintialize the web app to this combination.


Source data in CSV format is available online at:


This includes full details including the the quality and other properties for all of the materials and santization treatments. This package is released under a GNU GPLv3 license. Code supporting this web app is also openly available in the associated GitHub repository.