Comparative characterization of bio-aerosol nebulizers in connection to atmospheric simulation chambers

The interplay of bio-aerosol dispersion and impact, meteorology, air quality is gaining increasing interest in the wide spectrum of atmospheric science. Experiments conducted inside confined artificial environments, such as the Atmospheric Simulations Chambers (ASCs), where atmospheric conditions and composition are controlled, can provide 10 valuable information on bio-aerosol viability, dispersion, and impact. We focus here on the reproducible aerosolization and injection of viable microorganisms into an ASC, the first and crucial step of any experimental protocol to expose bio-aerosol at different atmospheric conditions. We compare the performance of three nebulizers specifically designed for bioaerosol applications: the Collison nebulizer, the Blaustein Atomizing Modules (BLAM) and the Sparging Liquid Aerosol Generator (SLAG), all manufactured and commercialized by CH TECHNOLOGIES. The comparison refers to operating conditions and 15 the concentration of viable bacteria at the nebulizer outlet, with the final goal to measure the reproducibility of the nebulization procedure and assess the application in experiments at ASCs. A typical bacterial test model, Escherichia coli (ATCC® 25922TM), was selected for such characterization. Bacteria suspensions, with a concentration around 10 CFU ml, were first aerosolized at different air pressures and collected by a Liquid Impinger, to obtain a correlation curve between airflow and nebulized bacteria, for each generator. Afterwards, bacteria were aerosolized inside the atmospheric simulation chamber 20 ChAMBRe (Chamber for Aerosol Modelling and Bio-aerosol Research) to measure the reproducibility of the whole procedure. An overall reproducibility of 11% was obtained with each nebulizer through a set of baseline experiments.

Prior to experiments, bacteria are cultivated on a non-selective Tryptic Soy Broth (TSB) medium until the mid-exponential phase (Optical Density, OD, at λ = 600 nm around 0.5) and then centrifuged at 4000 g for 10 minutes. Afterwards, bacteria are resuspended in sterile physiological solution (NaCl 0.9 %) to prepare a suspension of approximately 10 8 CFU ml -1 (Colony Forming Units), as verified by standard dilution plating. For the experiments performed inside the simulation chamber, the bacteria concentration was around 10 7 CFU ml -1 (see Massabò et al., 2018 for details). The average on CFU counting is used 65 to estimate the uncertainty range of the bacterial concentration in the solutions.

Nebulization systems
Many natural sources of bio-aerosol arise from wet environments; bacteria and viruses are commonly found in liquid suspensions and are hence aerosolized from liquids (Alsved et al., 2019). Among the liquid bio-aerosol generators, the pneumatic nebulizers, as the Collison device, are the most frequently used. Each atomizer considered in this study works with 70 different pressure range and aerosolization flow rate, as described below.
The Collison nebulizer has widespread applications, produces high concentrations of aerosol, but can cause damage to microorganisms due to strong impaction and shear forces. The recirculation of the cell suspension increases fragmentation of bacteria during prolonged nebulization as well (Reponen et al., 1997, Zhen et al., 2014. This device generates droplets by physical shearing and impaction onto a vessel wall. The solution to be sprayed is positioned directly in the glass jar. The 75 compressed air is used to aspirate the liquid from the reservoir into a sonic velocity air jet, wherein the liquid is sheared into droplets. The resulting liquid jet impacts against the wall of the jar, removing the larger fraction (in size) of the droplets. The resulting smaller droplets are carried out by the airflow while the larger particles return to the liquid reservoir are then reaerosolized. In this work, the 1 nozzle version of Collison was used. The upstream pressure can span in the 1 -6 bar range, which corresponds to an airflow rate from 2 to 7 lpm, for the 1-jet model. The main disadvantage of this device is the 80 recirculation of the liquid: the repetitive exposure to shear forces during atomization and impaction against the vessel wall can progressively cause damage and loss of viability to biological entities (Zhen et al., 2014). Several literature studies on the Collison performance report high particle concentrations but with a resulting cell damage (Mainelis et al., 2005;Thomas et al., 2011;Zhen et al., 2014).
The single-jet BLAM is used in one-pass mode: the liquid medium is subjected to the sonic air jet only one time. The atomizing 85 head is composed of two main parts: nozzle body and expansion plate (Fig. 1). The atomizer features a modular design, composed of five interchangeable plates, with different cavity depth and cone diameter, to accommodate liquids with different properties (viscosity, mainly). The atomization process is generated by a vacuum effect produced in the cavity between the body of the nozzle and the expansion plate, when pressurized air passes at sonic velocity through a precisely laser cut ruby crystal (fixed size 0.010 in. diameter) located into the nozzle body (Fig. 1). This effect pushes the liquid hosted in the cavity 90 into the air jet, which breaks up the liquid into droplets. Only the droplets smaller than a certain critical size can follow the airflow to the outlet tube located on the top of the BLAM unit: this critical size is determined by the speed of the airflow through the nebulizer. The jar should be filled with 20 ml of test solution, which serves only as a soft impaction surface for https://doi.org/10.5194/amt-2020-490 Preprint. Discussion started: 8 January 2021 c Author(s) 2021. CC BY 4.0 License. the larger droplets and it is not used for atomization. The liquid is delivered to the nozzle body with a desired flow rate (range of liquid feed rate: 0.1 -6 ml min-1) using a precision pump New Era Pump Systems,95 Inc.). The upstream air pressure determines the resulting airflow rate in the range 1 to 4 lpm which is kept constant by a mass flow controller. The properties of the aerosol generated by the single-jet BLAM are, nominally, a function of the jet hole size, depth of the liquid cavity, expansion cone size, and liquid viscosity. In this work, the expansion plate with a cavity depth and a cone diameter of 0.001 and 0.020 in., respectively, was used.
So far, the bubbling mechanism has been studied as a naturally occurring phenomenon and has been recognized as a significant 100 factor in aerosolization of seawater and suspended contaminants from breaking waves (Mainelis et al., 2005). The SLAG model is a single pass bubbling generator where a suspension of particles or microorganisms is pumped at a certain flow rate to the top surface of a porous stainless-steel disk where it forms a liquid film. Then, the airflow is delivered through the porous disc creating fine bubbles in the liquid film that subsequently burst, releasing particles into the air. Particles are carried out of the device by the same air stream. We used a SLAG with a 0.75" diameter porous disk and nominal pore size of 2 µm. The 105 recommended airflow ranges between 2 and 6 lpm and it is set by a mass flow controller. This principle of gentle bubbling aerosolization is expected to reduce stress and damage to microorganisms compared to pneumatic nebulization (Simon et al., 2006).

Experimental set up
In the first phase we used the experimental setup schematically shown in Fig. 2. The aerosol was sampled directly at the output 110 of the nebulizer, through a flanged connection, by an impinging system (liquid impinger by Aquaria Srl) filled with 20 ml of sterile physiological solution and operated at a constant airflow of 12.5 lpm. The bacteria suspension (concentration about 10 8 CFU ml -1 , see section 2.1.), was sprayed and directly collected by the liquid impinger. The number of cultivable cells inside the impinger was then determined as CFUs by standard dilution plating: 100 μl of serial dilutions of the solution was spread on an agar non-selective culture medium (trypticase soy agar, TSA), and incubated overnight at 37 °C before the CFU counting. 115 For each nebulizer, different airflows were tested, using a mass flow controller (Bronkhorst, model F201C-FA), and the nebulization efficiency was determined in terms of culturable fraction of aerosolized bacteria (i.e., percentage ratio of the concentration of viable bacteria inside the liquid impinger and in the sprayed solution).
The further tests took place at ChAMBRe (Chamber for Aerosol Modelling and Bio-aerosol Research), a 2.2 m 3 stainless steel atmospheric simulation chamber specifically designed for the research on atmospheric bio-aerosol. At ChAMBRe, particles 120 in the dimensional range of bacteria (1-2 µm) have a lifetime of several hours (Massabò et al., 2018). Atmospheric conditions and composition (i.e., type and concentration of gaseous species and PM) can be monitored and controlled. Water vapour can be directly injected into ChAMBRe thus adjusting the relative humidity inside the chamber from 0 to about 99 %. Temperature and relative humidity (RH%) inside the chamber are monitored using a HMT334 Vaisala® Humicap® transmitter. In the operative range (from 15 to 25 °C), the RH accuracy is ± 1 % RH (0 to 90 % RH) and ± 1.7 % RH (90 to 100 %RH), the 125 temperature accuracy is ± 0.2 at 20 °C. A set of two pressure gauges is used to measure the atmospheric pressure inside and https://doi.org/10.5194/amt-2020-490 Preprint. Discussion started: 8 January 2021 c Author(s) 2021. CC BY 4.0 License. outside the chamber. A MKS Instruments 910 DualTrans™ transducer is installed inside (measuring range from 5 x 10 -4 to 2 x 10 3 mbar; accuracy of ±0.75% of reading in the range 15 -1000 mbar). A Vaisala BAROCAP® Barometer PTB110 is installed outside the chamber with a measuring range from 5 x 10 2 to 1.1 x 10 3 mbar and accuracy of ±0.3 mbar at 20 °C. Data Acquisition) application allows the user to interact with the system by a user-friendly graphical interface. ChAMBRe is equipped with a sterilization system too: a 58 cm long UV lamp (UV-STYLO-F-60H, Light Progress srl) is inserted through a lateral flange. The lamp produces a 60 W UV radiation at λ = 253.7 nm which is used to sterilize the chamber volume without producing ozone before and after any experiment with bio-aerosol. Before each test with the nebulizers, the chamber was cleaned by evacuating the internal volume down to 10 -5 mbar thanks to a composite pumping system (a rotary pump model 145 TRIVAC® D65B, Leybold Vacuum, followed by a root pump model RUVAC WAU 251, Leybold Vacuum and a Leybold Turbovac 1000). Then, the chamber was vented again to atmospheric pressure throughout a 5-stage filtering/purifying inlet

Tests with Impinger 165
In the first set of experiments, we measured the nebulization efficiency, in terms of culturable fraction of aerosolized bacteria for each device and at different airflows. We adopted the ratio between the CFU counted in the impinger liquid and the CFU introduced in liquid solution of the nebulizer as operative definition of efficiency. With BLAM and SLAG, the latter corresponds to the product of the concentration (i.e., CFU ml -1 ) in the bacterial solution for the volume of liquid (2 ml) introduced in the nebulizer. To have a comparable metric, in the experiments with the Collison the volume of the liquid was 170 substituted with the injection time (5 min). Even if this choice does not meet a strict metrological criterium it makes possible a direct comparison of the three devices in well-defined operative conditions (see Table 1). The aerosolization air flow varied in the range of 1.4 -3.5 lpm for BLAM and 2 -5 lpm for SLAG and Collison. The bacteria suspension was supplied to the BLAM and SLAG devices at the same liquid flow rate of 0.4 ml min -1 (see Table 1). The tests started after a suitable warming time (about five minute), to get a stable nebulizer output. Afterwards, the aerosol was extracted for a further 5-minute time 175 with an impinger flow of 12.5 lpm. This way, 2 ml of bacteria suspension at the flow rate of 0.4 ml min -1 were aerosolized both by BLAM and SLAG.
Figures 3-5 show the nebulization efficiency of the BLAM, SLAG and Collison, respectively. The average on CFU counting was used to evaluate the uncertainty range of the bacterial concentration in the nebulized solution, while the uncertainty on the air flow was determined as the 1% of the flow controller full scale. At fixed air flow, the BLAM shows the highest nebulization 180 efficiency, followed by Collison and SLAG (e.g., at 3.5 lpm the BLAM efficiency is about 2 and 4 times higher than the Collison and SLAG ones, respectively). Our experimental procedure did not allow a direct control of the fraction of damaged bacteria during the nebulization phase, but, in the specific case of the Collison nebulizer (Fig. 5), the nebulization efficiency of the culturable fraction increases linearly with the airflow until about 3 lpm, after that the curve bends likely because the cell damage becomes more and more relevant. However, with the described injection conditions (5 min, air flow ≤ 5 lpm) the 185 output of viable bacteria turned out to be quite high.
At the same time, with the BLAM the flow of liquid supplied to the nebulizer can be accurately tuned. The SLAG requires a lower upstream pressure and, according to the producer claim, results in a softer injection (and then less bio-damage) of viable bacteria. Therefore, the SLAG looks best suited for experiment with fragile bacteria that can be nebulized in large numbers even with its extremely gentle nebulization system. The BLAM efficiency seems subjected to a higher variability: such feature 190 is likely due to the coupling between the nebulizer and the impinger set-up since the experiments with injection directly into the simulation chamber resulted much more stable (see section 3.2). https://doi.org/10.5194/amt-2020-490 Preprint. Discussion started: 8 January 2021 c Author(s) 2021. CC BY 4.0 License.

Tests at ChAMBRe
In the second set of experiments, we focused on the performance of the three aerosol generators when used to nebulize bacteria directly inside an atmospheric simulation chamber. Four experiments were performed with BLAM and Collison and five with 195 SLAG, all between November 2019 and July 2020. Experimental conditions and results are reported in Tables 2-4. The uncertainties quoted on both injected and collected bacteria are just those deriving from the Poisson fluctuation (i.e., the square root of the number of colonies counted in the Petri dishes) and they do not include any other systematic or statistical term. The values of the collected CFU are the average of the counts of the four Petri dishes exposed in each experiment; each group of four turned out to be statistically compatible (i.e., within the interval delimited by the statistical uncertainty, the counts in the 200 four Petri dishes agreed). Inside the chamber, the working condition adopted for the Collison produces an initial PM10 concentration of about 200 mg m -3 (Table 4), like the BLAM output ( Table 2). The initial PM10 concentration, as determined by the OPC, was taken as a rough reference for the aerosolization efficiency and quantity of aerosol generated (bacteria plus NaCl particles).
The injected bacteria correspond to the product of the concentration (i.e., CFU ml -1 ) in the bacterial solution for the volume of 205 liquid (2 ml) introduced in the nebulizer for BLAM and SLAG. In the experiments with the Collison, the injection time was considered instead of the liquid volume to calculate the number of inject bacteria and thus to make possible a direct comparison with the BLAM and SLAG performance. At ChAMBRe, considering the range of inlet air flows for the three devices, the typical figure for the ratio between the CFU on petri dishes (diameter: 10 cm) placed inside the camber to collect the bacteria by a gravitational settling and the injected CFU, is 10 -6 for each nebulizer. A good and stable correlation between the number 210 of injected and collected CFU was obtained for each nebulizer, as shown in Fig. 6-8, which refer to BLAM, SLAG and Collison respectively. The uncertainty on the slope of the correlation curves always turned out to be < 5% and the overall standard deviation around the average ratio (collected/injected CFU) was 11%. The experimental reproducibility appears to be adequate to design experiments within an ASC: it roughly corresponds to the sensitivity of the whole procedure to changes in the viability, for instance when bacteria will be exposed to different air quality conditions. 215 The absolute value of the aerosolization efficiency depends on the pressure at the nebulizer outlet (i.e., inside the atmospheric chamber, Feng et al., 2020). The results presented in this work were performed in a specific pressure regime i.e., with internal pressure about 2 mbar lower than the ambient pressure. This condition favors the bacteria confinement inside the chamber and was explored in view of experiments with pathogenic strains. With each specific set-up (i.e., simulation chamber or other downstream expansion volumes), the actual nebulization efficiency should be determined following the same steps above 220 reported. At ChAMBRe, the internal pressure can be maintained up to ± 5 mbar greater/lower than the ambient pressure. At ChAMBRe we could verify that with a 3-5 mbar overpressure and an internal pressure ranging from 1011 to 1026 mbar, the Collison efficiency in nebulizing physiological solution remain stable within 9%.

Conclusions
We compared the performance of three commercial nebulizer (BLAM, SLAG and Collison) Table 4. Bacteria concentration in the aerosolized solution, average number of colonies counted on the petri dishes and the meteorological parameters (P, T, RH) in ChAMBRe in the experiments with the Collison nebulizer operated at: liquid feed rate = 0.4 ml min -1 , volume of injected solution = 2 ml, injection time = 4 min, air flow = 3.0 lpm.