Wind tunnel‐based testing of a photoelectrochemical oxidative filter‐based air purification unit in coronavirus and influenza aerosol removal and inactivation

Abstract Recirculating air purification technologies are employed as potential means of reducing exposure to aerosol particles and airborne viruses. Toward improved testing of recirculating air purification units, we developed and applied a medium‐scale single‐pass wind tunnel test to examine the size‐dependent collection of particles and the collection and inactivation of viable bovine coronavirus (BCoV, a betacoronavirus), porcine respiratory coronavirus (PRCV, an alphacoronavirus), and influenza A virus (IAV), by a commercial air purification unit. The tested unit, the Molekule Air Mini, incorporates a MERV 16 filter as well as a photoelectrochemical oxidating layer. It was found to have a collection efficiency above 95.8% for all tested particle diameters and flow rates, with collection efficiencies above 99% for supermicrometer particles with the minimum collection efficiency for particles smaller than 100 nm. For all three tested viruses, the physical tracer‐based log reduction was near 2.0 (99% removal). Conversely, the viable virus log reductions were found to be near 4.0 for IAV, 3.0 for BCoV, and 2.5 for PRCV, suggesting additional inactivation in a virus family‐ and genus‐specific manner. In total, this work describes a suite of test methods which can be used to rigorously evaluate the efficacy of recirculating air purification technologies.


| INTRODUC TI ON
The coronavirus disease 2019 (COVID- 19) pandemic has led to considerable public attention on methods to mitigate viruses that transmit through aerosols. There is evidence that aerosol-based transmission, either via direct inhalation of infectious particles, or indirect via deposition of particles and subsequent infection from surface transfer, is an important infection route for severe acute respiratory syndrome coronavirus type 2 (SARS-CoV-2), the coronavirus responsible for COVID-19. [1][2][3][4][5][6][7] Sufficiently small aerosol particles (distinguished in the medical community from "droplets," ie larger entities which gravitationally or inertially deposit seconds after release 8,9 ), once emitted, have longer lifetimes in indoor spaces unless they are actively removed via ventilation. While the most direct method of addressing aerosol clearance is to modify heating ventilation air conditioning (HVAC) systems to operate at higher air change rates, such modifications are often costly, particularly in older residential and commercial buildings, which can operate without or with limited forced air heating and cooling systems. A possible alternative to effectively increase ventilation rates, or to locally reduce particle and airborne virus concentrations in an indoor space, is use of recirculating air purification units, which can incorporate filters, UV-light, electrostatic precipitation, and photoelectrochemical oxidation technologies in order to collect, react, or inactivate pollutants and pathogens in indoor aerosols.
At the time of writing this manuscript, there is a tremendous number of recirculating air purification technologies, largely filter based, on the consumer market. However, most of these commercial systems have only been coarsely tested in terms of their abilities to mitigate either non-infectious or infectious aerosol exposure.
It is not common to evaluate the size-dependent collection efficiencies of these devices, as the Association of Home Appliance Manufacturers (AHAM) testing standard AC-1-2015 only describes testing with particles in three broad classes without size distribution measurement: cigarette smoke, fine dust, and pollen. 10 The purpose of the study presented here was to develop and apply an alternative method of evaluating a recirculating air purification unit, the Molekule Air Mini, in terms of both its size-dependent particle collection efficiency as well as in its efficacy in removing and inactivating virus-laden aerosols with three different viruses: an influenza A virus, an alphacoronavirus, and a betacoronavirus. The tested recirculating air purification unit utilizes photoelectrochemical oxidation in addition to typical mechanical filtration in pollutant removal.
It was thus important to evaluate collection of virus-laden particles as well as to evaluate the inactivation of virus-laden particles passing through the unit. The wind-tunnel method, which has not been applied previously in evaluating recirculating air purifiers, involves mounting and sealing the air purification unit into a single pass wind tunnel, and examining both size-dependent particle collection efficiency as well as virus removal efficiency in single pass measurements. In this regard, the developed approach heavily hinges upon requirements of the ASHRAE 52.2 method of evaluating filters for HVAC ducts. For viruses, we distinguish between collection and inactivation via a combination of upstream and downstream sample titer and reverse transcription-quantitative polymerase chain reaction (RT-qPCR) measurements. Physical collection efficiency is quantified as a function of size as the penetration (1-collection efficiency), while virus removal is quantified via log reduction (based 10 log of the upstream-to-downstream concentration ratio). The most common figure of merit for recirculating air purification technologies is the clean air delivery rate (CADR), which is the product of the devices collection or removal efficiency and its operating volumetric flow rate. Wind tunnel testing at prescribed flow rates enables determination of the CADR in a size-dependent manner. The methods employed and results with the tested air purification unit are presented in the following sections. Subsequently, we discuss the implications of measurement results, including future prospects in studying virus aerosol control technologies.

| MATERIAL S AND ME THODS
Three types of experiments were performed to examine the air purification unit's capabilities in removing viable viruses from an aerosol. To our knowledge, only with electrospray-based aerosolization have isolated viruses been measured in an aerosol. 14,15 The particular • In adapting wind tunnel testing to recirculating air purification units, we are able to accurately determine their size-dependent collection efficiencies, and their clean air delivery rates (product of collection efficiencies and operating flow rates).
• By placing the wind tunnel within a biosafety level II facility we are able to directly examine the efficacy in collecting and inactivating airborne viruses, with collection and inactivation distinguishable from one another via use of both virus titration and reverse transcription quantitative polymerase chain reaction assays.
coronaviruses and the influenza virus used were selected because of recent concern of aerosol routes to infection 3,16,17 by viruses in these families (coronaviridae and orthomyxoviridae), but at the same time, their uses do not require Biosafety Level III facilities for experiments.
In examining two different types of coronaviruses, we specifically investigate whether different coronavirus genera display different inactivation behavior. Virus titration, RT-qPCR, and fluorimetry (of a fluorescent tracer) were utilized to characterize virus removal. Third, virus titration and RT-qPCR assays were performed on BCoV directly loaded onto the photoelectrochemical filter utilized in the test air purification unit, along with selected control filters. We briefly describe the methods employed in each of these measurements.

| Wind tunnel penetration tests
Both non-infectious particle and viable virus penetration tests were performed using a custom-designed wind tunnel, depicted in F I G U R E 1 A schematic diagram of the wind tunnel system utilized in penetration and log reduction measurements with the air purifier unit installed (A). Labeled photograph of the wind tunnel system within a biosafety level II room (B) Figure 1, and described in detail in Qiao et al (2020). 18 The air purification unit (Molekule Air Mini, Molekule Inc.,) was inserted into the test wind tunnel and sealed with a silicone adhesive to a mounting plate designed specifically for the air purification unit. The tested unit control technology incorporates a minimum efficiency reporting value-16 (MERV-16)-based filter laminate including a photoelectrochemical oxidation (PECO) layer, which is exposed to UV-A light during operation to drive oxidation of incoming particles and vapor phase species. The filter laminate is structured such that the outer surface of the filter (the upstream layer) features a carbon and electrostatic media, providing the bulk of the physical particle filtration.
The inner surface of the filter (the downstream layer) features the media upon which the PECO reaction takes place. The UV-A source is only turned on when the unit is closed with a manufacturerproduced filter installed, and the tested unit used in this study was not modified in any way. Sealing around the unit effectively drove all flow in the wind tunnel through the air purification unit (entering its inlet and exiting its outlet, as would be the case in recirculating operation), enabling penetration tests akin to ASHRAE Standard 52.2 tests for duct filtration units. For non-infectious particle penetration tests, we operated the wind tunnel at flow rates of 566 L/ min −1 , 1416 L/min −1 , and 2350 L/min −1 . We utilized a custom pneumatic nebulizer 19,20 to nebulize 5% weight aqueous KCl solution, yielding polydisperse KCl particles in the 10 nm to 10 µm diameter range. The produced droplets were entrained into the wind tunnel flow; droplet drying yielded dry KCl particles at a relative humidity in the 57%-62% range near 300 K. The size distribution function of KCl particles was measured upstream of the tested unit by sampling with isokinetic probes into both a TSI 3034 scanning mobility particle sizer (SMPS), which integrates a differential mobility analyzer and condensation particle counter for particle size distribution measurements in the 10 nm-487 nm range, and a TSI 3330 optical particle spectrometer (OPS), optically measuring particle size distributions in the 500 nm-10 µm range (we note down to 300 nm is possible, but we applied the OPS only at sizes beyond the SMPS range). A separate Po-210 bipolar ion source was used upstream of the SMPS to bring particles to a steady-state charge distribution prior to measurement. Similar size distribution measurements were  Figure S1 and its caption. All tested flow rates were accessible directly with the air purifier unit operating as a stand-alone instrument with 2350 L/min −1 achieved at its highest speed setting (there are five settings); however, because of the additional pressure drop brought about by wind tunnel components, an external blower was used to boost the system flow rate to the desired value for each test, with the flow rate monitored using a calibrated orifice meter. For both the upstream and the downstream sampling, customized mixing plates were installed before the sampling points to redistribute the particles across the radius of the wind tunnel. Prior to testing, via the Log-Tchebycheff method for both flow velocity and particle concentration, the coefficient of variation in the flow velocity was found to be below 0.10 at the location of the first upstream sampling point, and for particles below 5 µm in diameter, the coefficient of variation in particle concentration was below 0.15. Background correction was not applied in penetration calculation because background particle levels were multiple orders of magnitude below concentrations when aerosolizing KCl particles. hence, combining results here and in prior work, it does not appear ionization of virus-laden particles has a demonstrated influence on virus viability in aerosols. In total, the applied aerosolization procedure yielded bipolarly charged, virus-laden highly polydisperse particles which were near 3 µm in mean diameter by volume; hence, the virusladen particles were largely supermicrometer in diameter but primarily below 5 µm, as shown in Qiao et al. 18 For virus aerosolization suspensions, the three test viruses were prepared as described in the Supplementary material. Viruses were grown to titers of 10 7 -10 7 . 75 TCID 50 ml −1 (50% tissue culture infectious dose) for BCoV, PRCV, and IAV. All virus suspensions were also spiked with 0.3 g/L −1 fluorescein dye, used as a physical tracer for particle penetration. Upstream and downstream sampling were carried out for 30 min in triplicate for each test virus, both in the presence (penetration tests) and absence (correlation tests) of the air purification unit in the wind tunnel. For correlation tests, flow was controlled entirely by the external blower in the wind tunnel, with flow rate held at the same value as in penetration tests. Following each 30-minute aerosolization test, the blower was turned off, the syringe pump stopped, and the particles which deposited on the impactor plates were extracted with a cell scraper using 3 ml of appropriate virus growth media on each stage as collection liquid. All three stages were pooled for a single titer and RT-qPCR measurement per impactor (though with the filters kept separate from the impaction stages). This was done in an effort to maximize signal in all measurements, and avoid any issues in data interpretation brought about by particle bounce at the elevated impactor flow rates. Virus titration, RT-qPCR, and fluorimetry of impaction plate samples and total filter samples were carried out as described in the "Virus Titration,RT-qPCR, and Fluorimetry" subsection.

| Virus viability on filter media
In addition to carrying out measurements on aerosol samples, in tests with the air purification unit, samples were collected from the upstream and downstream filter laminate layers by wiping the filter with a 3 inch × 3 inch gauze piece wetted with virus growth media. 33 During the air purification unit operation, the inside of the filter laminate is exposed to a UV-A light source; UV-A photons and the PECO layer material drive photoelectrochemical oxidation of deposited particulate matter and material passing through the device. 34 Subsequent to swab tests, we elected to perform a systematic study of the influence of both the filter material and UV-A irradiation on the viability of viruses on filter surfaces. 0.5 ml of a high titer (10 7.25 TCID 50 ml −1 ) suspension of BCoV was inoculated onto 1 inch × 2 inch cut pieces of (1) the PECO filter conventionally used in the air purification unit (2 sets, 1a and 1b), (2) the filter conventionally used in the air purification unit, but without the PECO layer, and (3) commercial HEPA filtration media. We remark that the Molekule Air Mini will not operate without manufacturer-designated filters installed; hence, type (1) and type (2) filters were provided by the manufacturer, with type (2) made specifically for these experiments. Type (1a) filter pieces were re-inserted into the filter laminate, placed within the air purifier unit and the device operated for 4 h. At time intervals of 0, 30, 60, 120, and 240 min, three of the filter samples were extracted.
Following extraction, the removed filter samples were replaced by new filter strips in an effort to minimize any changes to the flow pattern in the air purification unit. Type (1b) filters were similarly extracted at these time intervals, but were not placed within the air purification unit and hence were not exposed to UV-A light. Type (2) filter samples were placed within the device and followed the procedure for type (1a) samples; however, photoelectrochemical oxidation is not anticipated in this instance as the PECO material is excluded in type (2). Type (3) HEPA filter samples were handled in an identical manner to type (1b) samples. Extraction was carried out by placing each filter sample into a 50 ml conical tube with 5 ml BCoV growth media, vortexing for 2 min at 3,000 rpm, and centrifuging at 3000 × g for 10 min at 4°C. The supernatant containing the virus was then aliquoted into tubes. Samples were titrated on the same day as collection, and the remaining aliquoted samples were frozen at −80°C until RT-qPCR and fluorimetry tests.

| Non-infectious particle penetration tests
Particle penetration, that is, the ratio of the downstream size distribution function to upstream size distribution function (1-collection efficiency), is shown in Figure 2 in the 10 nm to 10 µm diameter range for three selected flow rates. Solid lines denote data obtained with an SMPS, while dashed lines denote OPS measurement results. While the two instruments focus on different size ranges (with 0.5 µm to 10 µm plotted for the OPS), evident for data at all three test flow rates is that the penetration results from both instruments combine to form near-singular curves across the entire measured size range, with little discontinuity at the SMPS-OPS interface near 0.5 µm. As expected for a MERV 16-based filter, 39 the most penetrating size falls below 300 nm and is in the 40-60 nm size range.
Even in this size range, the penetration is below 10 −1 and the tested air purifier penetration is below 5 × 10 −2 (> 95% efficient) at nearly all sizes examined. Above 0.5 µm, with the exception of the highest flow rate setting tested, the penetration is below 10 −2 . The observed variation in penetration curves with flow rate is consistent with traditional fibrous filtration efficiency. For supermicrometer particles, collection is facilitated by particle inertial impaction and interception, leading to lower penetrations at increasing flow rate. 40,41 When inertial impaction becomes a significant mechanism, as is the case for higher flow rates and larger particles, the penetration decreases to 10 −4 -10 −5 . Meanwhile, submicrometer particles are predominantly collected via diffusion, 42,43 leading to higher penetrations with increasing flow rate in the submicrometer size range. The size-dependency of these disparate collection mechanisms leads to the most penetrating particle size falling into the 30-200 nm size range. 44 F I G U R E 2 Particle penetration versus particle diameter for the air purification unit, measured in a sealed wind tunnel. Solid lines denoted measurements utilizing a scanning mobility particle sizer, while dashed lines denote optical particle sizer results. The error bars are the standard deviation according to the four penetration values at each particle size determined following ASHRAE Standard 52.

| Virus aerosol removal
Viruses in aerosols, which are attached to sub-and supermicrometer aerosol particles, 3,17,23,47 will be removed by recirculating air purification units in a similar fashion to non-infectious particles, that is, biological activity has little-to-no influence on the physics governing particle penetration. Nonetheless, it is important to examine virus aerosol removal by air purification units for several reasons. This stated the similarity in performance of the two tested coronaviruses, from different genera, strongly suggests that similar results would be obtained for SARS-CoV-2 and others in the coronavirus family. We also note that because of the inherent variability in virus titer measurements, and to a lesser extent RT-qPCR measurements, small differences in log reduction (e.g. 0.1-0.2) are not strong evidence of inactivation.

| Virus viability on filter media
Swab extractions from the outside and inside of the PECO filters following virus aerosol penetration tests yielded the virus titers TA B L E 1 A summary of fluorimetry, virus titer, and RT-qPCR test results for pooled Anderson impactor stages upstream and downstream of the air purification unit in the wind tunnel  Figure 4, which also displays images of PECO and HEPA filter samples. Evident in Figure 4B, the PECO filter material alone, the filter without the PECO layer both inside and outside the air purification unit, and the HEPA filter all yielded similar changes in virus titer over time; a log reduction near 1.0 is achieved after 240 min on the filter.
Meanwhile, for the PECO filter exposed to UV-A light within the air purification unit, a log reduction in virus viability near 2.0 is observed over the same time interval, with discernable increases in the log reduction evident for test periods longer than 60 min. This provides evidence that for collected viruses, photoelectrochemical oxidation can drive virus inactivation at a faster rate, presumably through the product of short-lived reactive oxygen species (which also drive oxidation of volatile organic compounds (VOCs)) passing through the PECO filter system. 34 Shown in Figure 4C, reduction in BCoV viability is not associated with decay in the RNA fragment used in RT-qPCR amplification, which is similar for all tested virus types. This is presumably because the RNA fragment chosen for amplification is rather short (number of bases), and the oxidation reactions driving inactivation act non-specifically.

| CON CLUS IONS and LIMITATIONS
We have developed and applied a medium-scale wind tunnel toward examining the size-dependent physical collection efficiency and virus removal efficiency (accounting for removal and inactivation) of a recirculating air purification unit, the Molekule Air Mini. Wind tunnel testing was performed using IAV, BCoV, and PRCV. Based on the performed measurements, we make the following concluding remarks and note the following limitations of our study: ing. While this can be rectified through regular filter replacement, temporal changes in pump/fan performance, as well as any components designed for particle electrostatic collection, virus FI G U R E 4 Images of 1 in × 2 in cutout strips of photoelectrochemical oxidation (PECO) filter laminates and HEPA filter media used in virus surface inactivation studies (A). Log reduction over time on various filter media based upon bovine coronavirus titer (B). Log reduction over time on various filter media based upon bovine coronavirus RT-qPCR (C). Log reduction was calculated based on the average values of three replicates at different sampling times, comparing with initial time (t = 0) in both titer and RT-qPCR measurements. Error bars were calculated using the root sum square error propagation method with the standard deviations for triplicate measurements at the initial time and the sampling time in titer and RT-qPCR considered, respectively inactivation, or oxidation (eg, electrostatic precipitators and UVsources) may change in performance over time.

ACK N OWLED G EM ENTS
We thank the staff of the University of Minnesota Veterinary Isolation Facility for their support in ensuring proper experimental setup in this study. This work was supported by Molekule Inc.

CO N FLI C T O F I NTE R E S T
None of the authors have any financial or personal interests related to the results of this study.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/ina.12847.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data have been provided as tables directly within the manuscript, and raw data are available via e-mail upon request from the corresponding authors.