Factors Affecting Reduction of Infectious Aerosols by Far-UVC and Portable HEPA Air Cleaners

Abstract

Technologies that can reduce concentrations of airborne microorganisms through either particle capture or inactivation are important tools for reducing the risk of disease transmission and improving overall indoor air quality. The effectiveness of these technologies is tested in different ways, and as a result, it is challenging to compare results and optimize their use in applied settings. In this study, experiments were conducted in a large bioaerosol chamber to evaluate the efficacy of far-UVC and portable HEPA air cleaners against the bacteriophage MS2 as a surrogate for human viral pathogens. For both technologies, changing the media used to aerosolize the microorganism from deionized water to a simulated saliva doubled effectiveness metrics (both log₁₀ reductions and clean air delivery rates). Because reductions did not follow first order, log–linear dynamics, using different segments of the test period to calculate efficacy also significantly impacted reported performance. Evidence shown here indicates that both microbiological and particle dynamics likely play a role in impacting test outcomes under current methods, and more research is needed to improve repeatable and reliable standardized approaches for determining technology performance against infectious aerosols.

Graphical Abstract

1. INTRODUCTION

The COVID-19 pandemic has raised awareness about airborne disease transmission and the importance of indoor air quality (IAQ) on human health.^1–3 Increasing ventilation and source control (e.g., masking) are effective ways to reduce concentrations of infectious aerosols in indoor environments, but these strategies are not always feasible, for example, due to engineering, cost, or outdoor air pollution constraints.^4,5 Accordingly, there has been an increasing interest in utilizing technologies that can actively reduce concentrations of pollutants, including infectious aerosols, in occupied spaces.

Filtration-based technologies have long been used to reduce concentrations of particulates indoors.^4,6,7 In addition to upgrading filters in building heating, ventilation, and air conditioning (HVAC) systems, portable air cleaners have been widely recommended and used in recent years for reducing exposure to virus-laden aerosols^8–10 and other pollutants, including wildfire smoke,¹¹ but they can have high operational and energy costs and be noisy to operate.^6,7,12 Devices that emit ultraviolet radiation in the UVC spectrum (200–280 nm) have been used for decades to reduce pathogens in air, water, and on surfaces.¹³ Historically, low-pressure mercury vapor lamps emitting at a peak wavelength of ~254 nm have been used to treat the unoccupied upper-air zone of occupied indoor spaces, as direct exposure to this wavelength is harmful to human eyes and skin. Recent research has focused on far-UVC technologies (200–230 nm), which are less damaging to mammalian tissues but have been shown to be effective against microorganisms.^14–17

Recent standards and guidelines have introduced new targets for reducing the risk disease transmission in occupied spaces based on delivering specified rates of infectious-aerosol-free air to occupied spaces.^18–20,23 Air cleaning technologies offer an important pathway to achieving these targets and improving public health protections against a variety of diseases, but a significant challenge to selecting and optimizing their use has been a lack of standardized test methods for evaluating their effectiveness in a manner that is repeatable, reliable, and relevant to real-world settings. Laboratory testing is often conducted under conditions that are ideal for microbial reduction, making it difficult to extrapolate results to different applications with wide-ranging conditions. Recently published test methods for evaluating technology performance against bioaerosols do not specify certain details of the experimental procedure, including what aerosolization medium should be used, and these methods assume log–linear decay behavior in calculating technology efficacy.^21–23 In this study, we evaluated the impact of a simple change to test methodology–changing the aerosolization media for the test microorganism–on the calculated effectiveness of two different air cleaning technologies (far-UVC and HEPA devices) and compared different methods for quantifying technology performance. We also evaluated how measured performance metrics can change depending on what portion of the test period is used to calculate them.

2. MATERIALS AND METHODS

The test chamber and microbiological methods utilized in this study are described in detail in Ratliff et al.²⁴ A brief overview and additional information relevant to these specific experiments are stated below.

2.1. Test Chamber and Environmental Conditions.

Experiments were conducted in a bioaerosol test chamber measuring 7.6 × 3.6 × 3.0 m³ (85 m³) in EPA’s Aerosol Test Facility in Research Triangle Park, NC (Figure 1A). The chamber was outfitted with a mock HVAC system constructed from galvanized steel duct materials that served to recirculate chamber air (i.e., no fresh air introduction) throughout the test period at rate of 350 ft³/min (CFM, or 595 m³/h) or seven air changes per hour through the HVAC system. Two metal mixing fans were also operated throughout the entirety of each test to facilitate mixing, and chamber air was filtered through high efficiency filter banks until negligible particles were present following the completion of each test. Duplicate background samples were taken prior to the initiation of each test to ensure that no bioaerosol was present in the chamber (Figure 1B). Temperature and relative humidity (RH) were set at target conditions and monitored throughout each test; all experiments were conducted at 22–23 °C. Tests conducted with deionized (DI) water as the aerosolization medium occurred with an RH of 30–31%, and tests with simulated saliva (hereafter “saliva”) were conducted at an RH of 50–53%. Particle size and concentrations were measured throughout testing (including during chamber flushing) using an Aerodynamic Particle Sizer (APS, Model 3321, TSI, Inc., Shoreview, MN) for the 0.5–18 μm fraction and a Scanning Mobility Particle Sizer (SMPS, Model 3080 Electrostatic Classifier and 3022 Condensation Particle Counter, TSI, Inc., Shoreview, MN) for the 0.01–0.6 μm fraction. Ozone concentration was monitored by an Aeroqual Series 500 portable air quality monitor (Aeroqual Limited, Auckland, NZ) fitted with a 0–0.5 ppm sampling head.

Figure 1.

A) Plan-view schematic of test chamber layout. Direction of airflow through the HEPA units noted by arrows, and direction of far-UVC source orientation also noted by arrows (sources oriented with centerline parallel to the ground). B) Experimental timeline.

2.2. Microbiological Methods.

The test microorganism used in this study was the Biosafety Level (BSL)-1 bacteriophage MS2 (ATCC 15597-BI), a nonpathogenic RNA virus that infects Escherichia coli (ATCC 15597). As a small, nonenveloped virus, MS2 is generally more resistant to inactivation than enveloped viruses, such as influenza,^24–26 and has been shown to be more resistant to far-UVC than SARS-CoV-2.²⁷ The size and type of the genome has also been shown to impact viral susceptibility to UV inactivation.²⁸ For each experiment, inoculum was prepared by diluting the MS2 stock into either filter-sterilized DI water or a simulated saliva prepared following the recipe in Woo et al.²⁹ and aerosolized over a 10 min period (Figure 1B) using four 6-jet Collison Nebulizers (CH Technologies, Westwood, NJ). The saliva was prepared using Type II porcine mucin (M2378, Sigma-Aldrich, Allentown, PA). Duplicate bioaerosol samples were collected over sampling windows that started at specified times (0, 15, 30, 60, and 90 min; Figure 1) from the breathing zone using SKC BioSamplers (SKC Inc., Eighty-Four, PA) with phosphate-buffered saline (PBS, P0196, Teknova, Hollister, CA) as the collection medium. Each sample was collected using air sampling pumps operating at 12.5 L/min over a 10 min period, with the sampling time reported herein representing the start of each 10 min sampling window, and the time = 0 min sample beginning immediately following the aerosolization period (Figure 1B). All bioaerosol samples reported herein were above the limit of detection (53 PFU/m³). In select tests, a 10-stage micro-orifice uniform deposit impactor (MOUDI, model 110-R, TSI, Inc., Shoreview, MN) was used to measure the fraction of viable bioaerosols in particle size cuts between 0.056 and 18 μm. All samples were plated at an onsite microbiology laboratory the same day of testing, incubated overnight at 35 ± 2 °C, and enumerated the following day.

2.3. Test Devices.

Two different air cleaning technologies, each with different mechanisms to reduce infectious aerosol concentrations, were used in this study. The first was a commercially available, far-UVC filtered krypton chloride excimer source that emits UV radiation at a 60° beam angle with a peak wavelength of 222 nm and an output center irradiance of 14 μW/cm² at 1 m distance from the source. Per the manufacturer’s recommendations, two units, each with a single far-UVC module mounted on a tripod, were used in the chamber during testing. Both units were placed on the centerline of the long axis of the chamber facing inward toward the middle of the chamber with the emission centerline parallel to the ground and the lower opening of each module window at 2.24 m above the chamber floor (Figure 1A). The module closest to the HVAC return was placed 1.0 m from the chamber short wall, and the module farther away was 1.4 m inward from the short wall. Both units were placed to minimize the impacts of shadowing in the chamber. UV irradiance measurements were taken using an ILT2500 Light Measurement System (International Light Technologies, Peabody, MA) fitted with a SED220 detector using a W diffuser (International Light Technologies, Peabody, MA) calibrated to 222 nm to ensure the source was operating consistent with the manufacturer’s specifications (Figure 1A) and were continually recording during testing from a location in the center of the test chamber. In the far-UVC tests, both units were powered at 10 min, immediately following completion of the initial time = 0 min bioaerosol sample window (Figure 1B).

The second technology tested was a commercially available portable HEPA air cleaner with clean air delivery rates (CADRs) of 99 CFM for smoke, 118 CFM for dust, and 125 CFM for pollen (as reported by the manufacturer). The air cleaner stands 0.56 m tall and contains an internal 5 W low-pressure mercury UVC bulb. Two units were used during testing, each operating on the highest fan speed setting with the UVC component on, and they were placed within the chamber so that the filtered air was delivered away from sampling ports and in a manner that minimized impact on other testing processes (Figure 1A). The units were powered prior to the collection of the background sample for these tests (i.e., before the chamber was sealed for the duration of testing), as the units required manual input to select the settings that were used.

2.4. Calculations and Statistics.

Efficacy of each technology type was calculated as a log₁₀ reduction and by calculating CADRs. For these calculations, bioaerosol concentrations for each experimental replicate were normalized to the initial time = 0 min sample window concentration from each replicate, then the normalized recoveries at each sampling period were averaged at each sampling time window across the replicate set, and these mean values were used for the efficacy calculations. Log₁₀ reductions were calculated for each sampling window as the difference between the normalized, averaged log₁₀ MS2 recoveries from control tests (no technology operating) and the technology (far-UVC and HEPA) tests. CADRs were calculated using normalized, average bioaerosol recoveries following Stephens et al.³⁰ as the difference between the first-order loss rate constants derived from the tests, multiplied by the chamber volume. The Supporting Information contains additional information regarding how uncertainty was calculated for each type of data presented.

Calculations and statistics were performed using Python (version 3.9.12) with packages NumPy (version 1.23.4) and SciPy (version 1.9.1). Differences between data from each set of test replicates were evaluated at each sampling time window using Wilcoxon rank sum tests and Welch’s t tests (for log₁₀-transformed data that met a normality check following a Shapiro-Wilk test), with significance determined at the α = 0.05 level.

3. RESULTS AND DISCUSSION

A total of 10 chamber tests were conducted with DI water as the aerosolization medium (three control tests, four far-UVC tests, and three HEPA unit tests), and nine tests were conducted with simulated saliva (three of each control, far-UVC, and HEPA tests). Saliva tests were conducted at a target of 50% RH, in line with the 40–60% “Goldilocks” zone that has been recommended for both comfort and epidemiological reasons.^31–33 Because MS2 is generally less stable at intermediate RH,³⁴ and scoping tests demonstrated that bioaerosol decay was greatest over the test period when aerosolized in DI at 50% RH (Figure S1), the DI experiments were conducted at 30% RH to increase the dynamic range of the experiments.

3.1. Bioaerosol Dynamics.

MS2 recoveries at each sampling time window for each test replicate are shown in Figure S2. Generally, bioaerosol recoveries were variable between test replicates, with MS2 concentrations ranging up to an order of magnitude during the same sampling window within a given replicate set. This variability is inherent in microbiological testing (particularly aerosol experiments) and highlights the need to conduct replicate tests for each condition for more reliable characterization of technology efficacy.²⁴

Figure 2 shows the average normalized MS2 concentration over the test period (also listed in Table S1), where the value plotted at each sampling window is the mean recovery from the individual test replicates conducted at each condition, for both types of aerosolization media. For all test types (control, far-UVC, and HEPA), bioaerosol concentrations were consistently higher over the duration of the test period when MS2 was aerosolized in saliva compared to DI water. The largest difference within a test type resulting from aerosolizing in different media occurred between the control tests, where recoveries were 1.6 log₁₀ higher for tests conducted with saliva compared to DI water at the end of the test period (Table S2). Toward the end of the test period, differences in bioaerosol concentrations resulting from changing the aerosolization media were greater with the far-UVC tests (0.9 log₁₀ difference) compared to the HEPA tests (0.3 log₁₀ difference) at the 90 min sampling window (Table S2), potentially due to the interference of the additional soil components in saliva on UVC efficacy.^35–37

Figure 2.

Average normalized concentrations of MS2 at each sampling window during testing for control, far-UVC, and portable air cleaner (HEPA) tests with MS2 aerosolized in A) deionized (DI) water and B) simulated saliva as the aerosolization medium. Error bars represent standard deviations of bioaerosol concentrations from each experimental condition’s set of replicate tests.

3.1.1. MS2 Log₁₀ Reductions.

Calculated log₁₀ reductions for both technologies and aerosolization media, which are calculated as the time-matched difference between bioaerosol recoveries for the control and technology experiments at each sampling window (Figure 2) are shown in Figure S3. Consistent with efficacy testing for other air treatment technologies,^24,25 log₁₀ reductions generally increased over the 90 min test period. Although absolute MS2 concentrations for both far-UVC and HEPA tests conducted with DI were lower compared to those with saliva (Figure 1; Tables S1, S2), the time-matched log₁₀ reductions were generally higher for the saliva tests (Figure S3). The enhanced bioaerosol stability in the saliva control tests (Figure 2B) compared to the DI control tests (Figure 2A) is the primary driver of these larger log₁₀ reductions with saliva, despite lower bioaerosol concentrations in the DI tests. During the final 90 min sampling window, the log₁₀ reductions were approximately double for both far-UVC and HEPA when tested in saliva compared to DI (1.4 vs 0.76 for far-UVC and 2.2 vs 0.96 for HEPA, in saliva vs DI). This finding demonstrates that a greater overall bioaerosol reduction does not necessarily equate to a higher calculated log₁₀ reduction efficacy. Using representative time-matched control experiments to quantify technology efficacy (rather than using the difference between initial and final bioaerosol concentrations within a given experiment, Table S1) isolates the technology’s impact on bioaerosol reduction by decoupling it from the natural decay and settling losses, which are quantified by the appropriate control tests.

3.1.2. MS2 Clean Air Delivery Rates.

The increased bioaerosol stability in saliva also impacts the calculated MS2 CADRs for both technologies (Table 1). Figure S4 shows the first-order loss rate constant estimates for each test type that were used to calculate CADRs (as the sum of air cleaning from two devices in the chamber during testing for each technology) over the entirety of the test period. For the two far-UVC devices, the calculated CADRs were 62 CFM and 116 CFM in DI water and saliva, respectively, and for the two HEPA units, the calculated CADRs were 98 CFM and 193 CFM. As with the log₁₀ reductions, aerosolizing in saliva nearly doubles the calculated CADR of both technologies, driven by the enhanced bioaerosol stability in the saliva control tests. Assuming that a per-unit CADR is the calculated value divided by two, the bioaerosol CADR for the HEPA unit from the saliva tests (97 CFM) aligns closely with the manufacturer-reported CADR for smoke (99 CFM). The majority of the viable MS2 is in the particle size fraction that overlaps with the smoke used in air cleaner performance testing³⁸ (Figure S5).

Table 1.

Clean Air Delivery Rates^a Calculated for MS2 over Different Test Time Periods for Both the Far-UVC and HEPA Devices^b When Aerosolized in Simulated Saliva (SS) or Deionized (DI) Water

	MS2 CADR, CFM
	Far-UVC		HEPA
Time (min)	DI	SS	DI	SS
0–30	48 ± 43	128 ± 29	264 ± 69	243 ± 166
0–60	67 ± 33	132 ± 12	139 ± 109	230 ± 58
0–90	62 ± 45	116 ± 23	98 ± 83	193 ± 47
15–90	86 ± 55	109 ± 25	74 ± 92	124 ± 41
30–90	59 ± 58	106 ± 45	−47 ± 32	178 ± 80

^aCADRs, ± standard deviation.²¹

^bWith two of each device used for each of the respective test types.

Calculating clean air delivery rates following the method described in Stephens et al.³⁰ and used in standard test methods^21–23 assumes that the contaminant loss rates follow a simple first-order exponential decay model, yet the MS2 decay curves in Figure S4 exhibit additional nonlinearities. The CADRs can be calculated for specific parts of the test period, and the results from doing this calculation using different segments of the test period are shown in Table 1. For the far-UVC tests, the CADRs range 48–86 CFM with DI and 106–128 CFM with saliva, depending on the segment of the test period used to calculate the CADR. The highest far-UVC CADRs occur when calculated using the 0–60 min (saliva) and 15–90 min (DI) sampling windows, where each of those time periods overlaps with intermediate parts of the test period. For the HEPA tests, the CADRs range from −47–264 CFM with DI and 124–243 CFM with saliva. The highest CADRs occur during the earliest part of the test period and generally decrease when calculated over later segments, resulting in even a negative CADR (−47 CFM) for the last segment of the DI tests. The highest calculated HEPA CADR (132 CFM per unit, for the DI test, 0–30 min test period) was slightly greater than the highest manufacturer-reported CADR of 125 for pollen.

The average bioaerosol decays in all test conditions are generally more rapid during the initial test period and tail off toward a slower decay rate during later parts of the test (Figure 2), consistent with prior observations that microorganisms can exhibit biphasic stability with varying decay rates over time that can be impacted by a variety of factors, including RH and droplet volume.^39–41 Evidence of a nonlinear decay during the control tests (when no treatment is active) indicates that this behavior is inherent to the bioaerosols themselves, even without treatment technologies active, but the technologies themselves may also act nonlinearly to reduce concentrations of viable bioaerosols. For example, UV inactivation scales with dose and has also been shown to elicit a biphasic response.^36,42,43 In the far-UVC tests reported here, CADRs could be larger when calculated over the first hour of the test rather than the first 30 min because a threshold exposure time is required for a substantial fraction of bioaerosols to be exposed to a high enough UV dose to demonstrate efficacy.

3.2. Aerosol Particle Dynamics.

3.2.1. Particle Size and Bioaerosol Distributions.

Particle size and concentration measurements were taken throughout the duration of each test to evaluate both the impacts of changing aerosolization media and the technologies themselves on particle dynamics. The overall particle size distributions (from merged APS and SMPS data) weighted by number and mass from the beginning and end of the test period from control tests with both media are shown in Figures S6 and S7. For both DI water and saliva, the mass mode diameter is constant at 0.86 μm over the duration of the entire test period. The count mode diameters for both media were also similar at 50 nm for DI water and 41 nm for saliva at time = 0 min, and they increase by ~10 nm over 90 min to 85 and 95 nm. Accordingly, changing the aerosolization medium does not markedly impact the aerosol particle size distributions, but the distribution of viable MS2 does shift with changing the aerosolization media (Figure S5), with a larger fraction of bioaerosol recovered from slightly larger particle sizes with saliva compared to DI water. When aerosolized in DI, a greater fraction of the viable MS2 is closer to the most challenging particle size for filtration (0.3 μm),⁴⁴ which could be a factor as to why the MS2 CADRs from the HEPA experiments are higher when MS2 was aerosolized in saliva compared to DI (Table 1), except when calculated over the initial 0–30 min portion of the test period.

3.2.2. Particle Concentrations and Clean Air Delivery Rates.

The particle mass concentrations over time, as measured by the APS and SMPS, are shown in Figure 3. For both sets of measurements and aerosolization media, the particle concentrations from the far-UVC tests and the control sets are essentially the same, indicating that the far-UVC devices had no appreciable impact on particle concentrations over the test period. In contrast, the portable HEPA units decreased the APS concentration by 98.4% and 98.6% and the SMPS concentration by 98.6% and 98.9% for the DI water and saliva tests, respectively. These similar overall particle reductions driven by filtration are expected given the similar particle size distributions of the two aerosolization media (Figures S6, S7). Particle CADRs can also be calculated based on the change in aerosol mass concentrations (Table 2, Figure S8). Because the HEPA units are reducing aerosol concentrations relative to the controls over the test period, their particle CADRs are much higher than the far-UVC particle CADRs, as far-UVC does not remove particles (Figure 3).

Figure 3.

Particle concentrations measured over the test period weighted by mass as measured by the APS (A, B) and the SMPS (C, D) when MS2 was aerosolized in DI water (A, C) or simulated saliva (B, D) for the control, far-UVC, and HEPA tests.

Table 2.

Clean Air Delivery Rates^a Calculated for Particles Measured by Either the APS or SMPS over Different Test Time Periods for Both the Far-UVC and HEPA Device Tests^b When Aerosolized in Simulated Saliva (SS) or Deionized (DI) Water

	APS CADR, CFM				SMPS CADR, CFM
	Far-UVC		HEPA		Far-UVC		HEPA
Time (min)	DI	SS	DI	SS	DI	SS	DI	SS
0–30	4 ± 1	3 ± 2	206 ± 1	188 ± 1	0 ± 2	16 ± 4	195 ± 3	194 ± 4
0–60	2 ± 0	2 ± 1	190 ± 2	184 ± 2	1 ± 2	9 ± 3	183 ± 5	183 ± 4
0–90	2 ± 0	2 ± 1	161 ± 4	164 ± 3	0 ± 2	6 ± 3	157 ± 7	162 ± 6
15–90	0 ± 1	1 ± 1	143 ± 4	155 ± 3	1 ± 1	0 ± 2	144 ± 8	149 ± 6
30–90	1 ± 0	1 ± 0	118 ± 4	139 ± 3	0 ± 1	1 ± 1	123 ± 8	133 ± 7

^aCADRs, ± standard deviation.³⁸

^bWith two of each device used for each of the respective test types.

Like the MS2 CADRs, the portion of the test period used to calculate particle CADRs impacts the value, indicating that the particle decays also do not follow the first order decay assumption in these tests. The particle CADRs are greater for the initial part of the test period and are lower when calculated over later time segments, when the slopes of the particle decay curves flatten and become increasingly parallel (Figure 3, Figure S8), similar to the MS2 decay curves. These nonlinear particle dynamics also play a role in impacting the calculated MS2 CADRs, as the fate of the infectious aerosol, including the conditions under which it is exposed to treatment, is intrinsically linked to the aerosol particles that transport MS2 through the air.

The APS and SMPS can also be used to track the mass median aerodynamic diameter (MMAD) of particles over time (Figure 4). As measured by the APS, the initial MMAD were slightly larger for tests using saliva compared to DI, although similar trends are observed in changes to the MMAD over time for each technology and both media types. The APS MMAD decreases in a similar fashion for the control and far-UVC tests, but the trajectory of the MMAD differs for the HEPA tests; it experiences an initial decrease, but after approximately 60 min, it exhibits a flatter (or, in the case of DI water, even a slightly upward) trend (Figure 4A,B). The initial decrease in the MMAD exhibited by all three test types potentially results from larger particles settling more rapidly compared to smaller particles or from evaporative processes.^33,45 In contrast, the MMAD as measured by the SMPS is relatively stable throughout the test period at around 0.3 μm, which is the most penetrating particle size for filtration.⁴⁴ The SMPS-measured MMAD is also more variable toward the end of the test period, compared to the control and far-UVC tests. The deviations MMAD behavior during the later portions of the HEPA tests suggests that particle dynamics differ during those periods, and more research is needed to understand these dynamics, as well as how these particle dynamics generally impact technology efficacy.

Figure 4.

Mass median diameters as measured by the APS (A, B) and by the SMPS (C, D) when MS2 was aerosolized in DI water (A, C) or simulated saliva (B, D) for the control, far-UVC, and HEPA tests.

3.3. Ozone Concentrations.

Ozone concentration was monitored throughout all DI water tests, as well as during one control test and all three far-UVC tests conducted with saliva (Figure S9). The initial ozone concentration was higher for the DI tests than for the saliva tests, likely due to differences in background concentrations resulting from seasonal ozone fluctuations. During the DI tests, the ozone concentrations increased more so over the course of the far-UVC tests compared to the control and HEPA tests. This observation is consistent with recent studies that have demonstrated the generation of ozone from 222 nm UVC sources.^46–48 Ozone concentrations did also increase during the far-UVC saliva tests, although it also increased slightly during the control saliva test during which it was monitored. The ozone sensor used here is relatively low-cost compared to more accurate monitoring equipment, and studies have shown that correction factors can be necessary for accurate quantitative concentration reporting.^49,50

3.4. Implications for Quantifying Efficacy and Test Method Development.

The results presented herein demonstrate challenges that exist related to conducting tests to quantify the effectiveness of air cleaning and treatment technologies against infectious aerosols. The efficacy of antimicrobial products has traditionally been quantified through log₁₀ reductions (e.g., U.S. EPA’s Antimicrobial Efficacy Test Guidelines⁵¹), but extrapolating log₁₀ or percent reductions to spaces of different sizes is problematic.³⁰ CADR is a useful measure for quantifying air cleaner effectiveness, but this metric has historically been calculated for other indoor pollutants such as pollen, dust, and smoke.³⁸ Here, we show that the CADR for both infectious aerosols and particle measurements in efficacy tests with both portable HEPA air cleaners and far-UVC devices can be highly sensitive to the test segment over which it is calculated, because both the particles and bioaerosols exhibit nonlinear dynamics that are not consistent with the first order decay assumption. Data presented here indicate some portion of the particles are changing size over time, effectively changing the particle challenge over the course of the test, and the performance of portable HEPA units has been shown to be sensitive to relatively small changes in particle size.⁵² Moreover, microorganisms often exhibit a biphasic stability, commonly featuring an initial rapid decline of more susceptible organisms, followed by a slower decay rate from a more resistant fraction. Changing particle sizes could also affect bioaerosol stability over time⁵³ and impact susceptibility to far-UVC inactivation.^37,41,54 Microorganism CADRs vary even more widely than the particle CADRs, suggesting that this nonlinear bioaerosol stability likely plays an additional role (in addition to the particle dynamics) in impacting decay rates over time and the resulting CADRs. Current limitations in bioaerosol sampling methods, including relatively long sampling periods (where decay rates may be varying over time within a given sampling window), compared to more instantaneous measurements for particle and other pollutant concentrations, also likely contribute to the challenges in calculating CADRs using the current methodology.

The results shown here also show how a simple change to test methodology–using a simulated saliva instead of DI water as the aerosolization medium–can significantly impact calculated efficacy. For both log₁₀ reductions and CADRs calculated over the entire test period, efficacy roughly doubles when using saliva compared to DI because of the enhanced bioaerosol stability exhibited in the saliva control tests compared to the DI controls (even though more overall bioaerosol reduction occurs for the DI technology tests). This result has important implications for both informing the development of standardized test methods and for using these technologies to meet emerging IAQ standards to protect against disease transmission, which can, in turn, impact energy use and costs. For example, comparing results from the DI water and saliva tests, one would need to deploy roughly twice as many units of both technologies to meet the same clean air delivery target if using test results produced using DI water compared to saliva.

There are numerous limitations that must be considered when extrapolating technology performance as evaluated and calculated here (and any laboratory test) to application settings. Between the HVAC system and mixing fans, the chamber air is likely better mixed than most indoor environments, and the degree of mixing in the chamber could also impact CADRs, and, ultimately, performance. By design, the tests were also conducted in a manner such that a high viral load was introduced at the beginning of the test, but no additional microorganism was released following the initial aerosolization period, which differs from how virus might be shed. Different saliva recipes using different types and concentrations of mucin have also been shown to have different UV absorbencies,³⁵ which could impact the performance of the far-UVC device, and the saliva composition (as well as environmental conditions including temperature, RH, and CO₂ concentration⁵⁵) could also impact bioaerosol stability. Moreover, the test organism used here (MS2) is not a human pathogen, and the relative susceptibility of different microorganisms to different technologies will vary.

Further research is needed to improve development of test methods that effectively quantify technology performance, as well as to improve extrapolations of laboratory findings, to optimize technology use for different applications. Additional test parameters that likely impact performance of these technologies against infectious aerosols include the chamber size, aerosolization and sampling methods, sampling frequency and duration, generated particle size distribution, environmental and airflow conditions, use of surrogate organisms, directionality of the device effluent outflow and biological sampling devices, orientation and proximity of the UV devices to sampling devices, and the bioaerosol concentration. Understanding how these factors impact calculated performance and corresponding sensitivity to these factors are critical for generating applicable and reliable test methods. Both log₁₀ reductions and CADRs are commonly used to report technology performance, but here, the highest values for each occur during opposite parts of the test period. CADRs are typically highest during the initial part of the test period, but fewer statistically significant differences are observable between the bioaerosol concentrations measured in control and technology tests during that period. Conversely, log₁₀ reductions are highest at the end of the test, with clearer differences between MS2 recoveries, but lower rates of concentration change, leading to smaller (and even negative) calculated CADRs. More work is needed to understand how best to evaluate and characterize performance of these technologies in a manner that is most relatable to use under real-world scenarios, which is necessary to optimize their use to reduce the risk of disease transmission.