Non-thermal plasma for decontamination of bacteria trapped in particulate matter filters: plasma source characteristics and antibacterial potential

A novel plasma source (PlasGlAiry^®, Plasmatec GmbH, Adelebsen, Germany) was used throughout this study. As depicted in figure 1(a), it is based on the concept of volume DBD, where 20 individual electrodes arranged in one plane at an air gap of 1.5 mm form a DBD stack. Each individual electrode is made from a hollow glass cylinder with an outer diameter of 2 mm and a length of 128 mm. The cavities inside the glass cylinders are filled with a conductive material. A power supply (TRE510C, Cofi srl, Treviso, Italy) provides two connectors, each of which outputs a sinusoidal signal with a frequency of 50 Hz at a phase shift of 180° to each other. The electrodes of the DBD stack were alternately connected to the two outputs thereby maximizing the electric field in the gas gap between the glass surfaces of two neighbouring electrodes. Arranged orthogonally to an air flow, the DBD stack of 12.8 cm in width and 6.95 cm in height covers a total area of 89 cm² in which a streaming gas is subject to the NTP ignited in the 19 gas spaces between two adjacent electrodes. Figure 1(b) shows a close-up photograph of plasma propagation in the gas gaps.

Zoom In Zoom Out Reset image size

As depicted in figure 1(a), U-I-characteristics of the gas discharge were recorded by two voltage probes (P6015A, Tektronix, Wilsonville, OR, USA) and a current monitor (Model 411, Pearson Electronics, Palo Alto, CA, USA) connected to a digital storage oscilloscope (Model RTM3004, Rhode und Schwarz, Munich, Germany). The mean electric power was calculated by the U-I-method for a set of 77 consecutive waveforms applying the oscilloscope [17].

According to figure 2(a), an Echelle spectrograph (Aryelle Butterfly, LTB Lasertechnik Berlin GmbH, Berlin, Germany) equipped with an optical fibre facing the NTP was applied to obtain the optical emission spectroscopy (OES). The spectrograph covered the range from 190 nm to 450 nm at a spectral resolution of $\lambda /{{\Delta }}\lambda$ = 15 000. Intensity calibration was performed by a calibrated deuterium lamp (DH-2000, Ocean Optics, Dunedin, USA). Given the low intensity of radiation, spectra were continously recorded at a constant integration time of 30 s per spectrum over a period of 4 h and finally accumulated to acquire sufficient signal-to-noise-ratio. Utilizing a set of methods described by Peters et al rotational and vibrational temperature as well as mean electron energy $\varepsilon$ were derived from the accumulated spectral data [18]. Electron density ${n_e}$ was estimated by using the equation (1) with the measured current $I$ as proposed by Keller et al [19]

$$\begin{equation}{n_e} = \frac{I}{{A \cdot e \cdot {{\left( {\mu \cdot N} \right)}_\varepsilon } \cdot {{\left( {E/N} \right)}_\varepsilon }}}.\end{equation} \tag{ 1 }$$

Zoom In Zoom Out Reset image size

The cross-sectional area $A$ of the plasma in m² was calculated based on circular area with a radius of 85 µm according to micro photography (D90 + AF-S Micro NIKKOR 85 mm 1:3.5 G ED, Nikon Corp., Minato, Japan) analysis of multiple microdischarge channels depicted in figure 1(b). Values for mobility coefficient ${\left( {\mu \cdot N} \right)_\varepsilon }$ in m⁻¹ V⁻¹ m⁻², reduced electric field $E/{N_\varepsilon }$ in V m² as well as the electron velocity distribution function $f\left( E \right)$ in eV^−3/2 were derived from the Boltzmann solver BOLSIG⁺ (12/2019) for a gas mixture of N₂ (80%) and O₂ (20%) at the mean electron energy $\varepsilon$ derived from OES measurements [20]. Equation (2) was subsequently used to calculate reaction rates $v$ for selected inelastic collisions of electrons with the gas molecules of the ambient air, considering equation (3):

$$\begin{equation}v = \mathop \int \limits_{{E_{{\text{threshold}}}}}^\infty \sigma \left( E \right) \cdot \sqrt {\frac{{2 \cdot e}}{{{m_e}}}} \cdot E \cdot f\left( E \right){\text{d}}E \cdot {n_e} \cdot n\end{equation} \tag{ 2 }$$

$$\begin{equation}\mathop \int \limits_0^\infty f\left( E \right){\text{d}}E = 1.\end{equation} \tag{ 3 }$$

Data on cross sections $\sigma \left( E \right)$ for electron collisions leading to dissociaton and ionization of O₂ were obtained from [21], while data on vibrational exication, dissociation and ionization of N₂ can be found in [22]. Furthermore, ${m_e}$ denotes the electron mass in ${\text{kg}}$, $e$ the elementary charge in $A{\text{ }}s$ and $E$ the kinetic electron energy in ${\text{eV}}$. The species concentrations $n$ of N₂ and O₂, respectively, were calculated on the basis of a gas temperature equal to the rotational temperature (${T_{{\text{gas}}}} \approx {T_{{\text{rot}}}})$ and atmospheric pressure for a gas mixture of 80% N₂ and 20% O₂.

Figure 2(a) shows the scheme of the experimental setup mimicking a typical arrangement of an air purifier. A cylindric ventilation duct of 100 mm in diameter and 160 mm in length comprises a duct insertation filter cartridge.

As depicted in figure 2(a), this cartridge is equipped with a single layer of plane PM filter material (synthetic F9 filter, JS Filtersysteme, Sinsheim, Germany) and activated carbon fleece (figure 2(c)) (C 20 C, Riensch & Held Hamburg GmbH & Co KG, Hamburg, Germany) at a strength of 16.5 mm. Downstream the cartridge, a fan (F12, Arctic GmbH, Braunschweig, Germany) was plugged to the duct. Operating at a power uptake of approx. 2 W, it induced a constant volume flow of 59.5 m³h⁻¹ at an air flow velocity of 2.1 m s⁻¹ in the duct according to differential pressure measurements (testo 512-1, Testo SE & Co. KGaA, Titisee–Neustadt, Germany) based on DIN EN ISO 5801 and fan characteristics. The same method was used to determine the pressure drop at the DBD stack to 0.2 Pa. At a distance of 100 mm upstream of the filter cartridge, the DBD stack was positioned in a square box comprising two holes of 100 mm in diameter. While one hole serves as air inflow, the other hole was connected to the cylindric ventilation duct.

Concentrations of reactive species in the ventilation duct were analysed by means of an ozone monitor (Model 106-L, 2B Technologies Inc., Broomfield, CO, USA) and a nitrogen oxide monitor (Model 405 nm NOx Monitor, 2B Technologies Inc., Broomfield, CO, USA). According to figure 2(a), the inlet of a hose with outer diameter of 6 mm was centred either at 2 mm distance upstream or downstream the filter cartridge, while the other end was split into two separate hoses each connected to one of the monitors. As a consequence, the constant volume flow through the hose close to the filter surfaces was approx. 0.15 m³ h⁻¹. With this arrangement, O₃ and NO_x concentrations could be measured at the same time and space.

All experiments took place inside a laboratory fume hood (L × W × H = 1.19 m × 0.53 m × 0.74 m = 0.46 m³) connected to an exhaust system expelling 50.4 m³ h⁻¹ according to Vane Anemometric measurements (EM54, FLIR Systems Inc., Wilsonville, OR, USA). The room air temperature and relative humidity (HL-1D, Rotronic Messgeräte GmbH, Ettlingen, Germany) were recorded during each experiment.

As a proof of principle, the gram-negative bacteria E. coli was used in this study. The E. coli strain DSM 11250 was purchased from DSMZCellDive (DSMZ-German Collection of Microorganisms and Cell Cultures GmbH, Braunschweig, Germany). In brief, E. coli was cultured in 5 ml Luria–Bertani broth (LB medium) with continuous shaking (300 rpm) overnight at 37 °C for 20 h. The bacteria were pelleted, washed twice and resuspended in 1 ml of a 1X phosphate buffer solution (PBS). The bacteria were next counted with the use of a haemocytometer. The final mixture was resuspended in 10 ml of 1X PBS mixture with a final concentration of 1 × 10⁷ bacteria ml⁻¹.

Contamination of the filter material was performed with detached plasma module in figure 2. Instead, a humidifier (Super Fog II, Lucky Reptile, Waldkirch, Germany) was attached to the duct system with the PM filter installed in the filter cartridge. Prior to bacteria contamination, the PM filter was exposed to humid air conditions by operating the humidifier for 20 min. Afterwards, the solution containing E. coli as a contaminant, was aerosolized with the use of a nebulizer (Inhalator IN 500, Medisana, Neuss, Germany) until all 10 ml of the bacteria solution was used (approx. 20–30 min). In order to allow bacteria droplets to reach and be adjusted on the PM filter surface, 20 extra minutes without additional moisture supply were given as an adjustment phase before proceeding to the next procedures. After the 20 min adjustment phase, samples were taken by cutting two square pieces of the contaminated filter (2 × 2 cm² each) as a control (no treatment) from the centre of the PM filter.

For the NTP experiments, the remaining filter piece was mounted in the filter cartridge and the plasma module was re-attached to the duct system (figure 2). PM filters were then exposed to the reactive species in the air flow for either 5, 10 or 15 min of operation of the DBD stack. Following each treatment, two square pieces of each exposed filter were excised as previously described for the control sampling. Each experiment was performed in triplicates.

The 2 × 2 cm² square filters (either control or DBD-treated pieces) were submerged into a 50 ml conical tube containing 1X PBS (10 ml) and 5 sterile glass beads, 4 mm in size (Solid glass beads, Merck, Darmstadt, Germany). The filters were then incubated in this solution for 10 min. In order to extract the bacteria from the filter pieces, the tubes were continuously vortexed for 1 min. Serial dilutions were next made and the bacteria were plated on agar plates and incubated overnight at 37 °C for 18 h. The CFU were counted and CFU density on the filter material was calculated. The comparison of control and treated samples was statistically analyzed using a 2-tailed student's t-Test. A $p$ value < 0.05 was considered statistically significant.

The U-I-characteristics of the entire plasma module are given in figure 3. Each pole of the power supply outputs a sinusoidal voltage at a frequency of 50 Hz and a maximum voltage of ${\hat u_p}$ = 5.5 kV. Since the voltage profile of the two poles is 180 degrees out of phase with each other and the DBD electrodes are alternately contacted with the two poles, a maximum potential difference of ${\hat u_{pp}}$ = 11 kV in the gas spaces between two adjacent DBD electrodes can be observed. From the current characteristics depicting a multitude of short and successive charge transfer events it is evident that the discharge operates in the multi-filamentary mode. The mean electric power uptake by the plasma module is (4.3 ± 0.16) W and corresponds to an area power density of (118 ± 4.4) mW cm⁻² in the gas.

Zoom In Zoom Out Reset image size

Figure 4 depicts an overview of the radiation emitted by the plasma of the DBD stack. The spectral lines derive from the second positive system ${N_2}\left( {{C^3}{{{\Pi }}_u} - {B^3}{{{\Pi }}_g}} \right)$ as well as the first negative system $N_2^ + \left( {{B^2}{{\Sigma }}_u^ + - {X^2}{{\Sigma }}_g^ + } \right)$ of molecular nitrogen. The absence of emissions from the $\gamma$-system $NO\left( {{A^2}{{\Sigma }} - {X^2}{{\Pi }}} \right)$ of nitric oxide corresponds well with the comparatively low gas temperature and is a strong indicator, that the discharge operates in the O₃-regime rather than in the NO_x -regime [23].

Zoom In Zoom Out Reset image size

According to table 1, the NTP produced by the DBD stack operates close to room temperature and comprises energetic electrons. Among the electron-impact reactions with the heavier gas particles, vibrational excitation of nitrogen molecules is most frequent followed by dissociation and ionization of N₂ which is the reason for the high vibrational temperature of the plasma. In terms of the development of a reactive gas phase chemistry, the presence of atomic oxygen is an important prerequisite. Even though its direct production by electron-impact dissociation of O₂ occurs at a rate 6 times lower than vibrational excitation of N₂, other reactions such as electron-impact dissociation or electronic excitation of N₂ contribute indirectly to atomic oxygen formation via heavy particle reactions [24].

Table 1. Data for rotational temperature ${T_{{\text{rot}}}}$, vibrational temperature ${T_{{\text{vib}}}}$, mean electron energy $\varepsilon$, electron density ${n_e}$ as well as selected reaction rates $v$ for electron collisions with molecules.

${{{T}}_{{{rot}}}}{{\,}}\left({\approx \! {{{T}}_{{{gas}}}}} \right)$	${{{T}}_{{{vib}}}}$	${{\varepsilon }}$	${{{n}}_{{e}}}$
$290\,\text{K}$	$2\,382\,\text{K}$	$9.4\,e\text{V}$	$1 \cdot {10^{19}}\,{\text{m}^{ - 3}}$
${{v\,}}\left( {{{e}} + {{{N}}_2} \to {{e}} + {{{N}}_{2\left( {{{vib}}} \right)}}} \right)$		$1.9 \cdot {10^{30}}\,{{\text{m}}^{ - 3}} \cdot {{\text{s}}^{ - 1}}$
${{v\,}}\left( {{{e}} + {{{N}}_2} \to {{e}} + 2{{N}}} \right)$		$1.3 \cdot {10^{30}}\,{{\text{m}}^{ - 3}} \cdot {{\text{s}}^{ - 1}}$
${{v\,}}\left( {{{e}} + {{{N}}_2} \to 2{{e}} + {{N}}_2^ + } \right)$		$8.7 \cdot {10^{29}}\,{{\text{m}}^{ - 3}} \cdot {{\text{s}}^{ - 1}}$
${{v\,}}\left( {{{e}} + {{{O}}_2} \to {{e}} + 2{{O}}} \right)$		$3.3 \cdot {10^{29}}\,{{\text{m}}^{ - 3}} \cdot {{\text{s}}^{ - 1}}$
${{v\,}}\left( {{{e}} + {{{O}}_2} \to 2{{e}} + {{O}}_2^ + } \right)$		$2.0 \cdot {10^{29}}{ }{{\text{m}}^{ - 3}} \cdot {{\text{s}}^{ - 1}}$

The concentrations of O₃ and NO₂ as well as NO at two different locations in the duct system during exposure of microorganisms are presented in figure 5. In close vicinity to the PM filters (upstream filter cartridge), quasi-stationary concentrations of reactive species establish 1 min after plasma ignition. Time-averaged concentrations until the end of the monitoring period of 32 min amount to (10 812 ± 551) ppb for O₃ and (286 ± 13) ppb for NO₂. Data for NO is close to or below the detection limit of 1 ppb and is therefore not analysed.

Zoom In Zoom Out Reset image size

The concentrations of reactive species are significantly lower downstream the activated carbon layer due to catalytic reactions of the reactive gas species. According to figure 5(b), it takes 2.5 min for NO₂ to reach stable concentrations, while the concentrations of NO and O₃ could not achieve stability within the observation period. The time-averaged concentrations between 2.5 min and 30 min after plasma ignition are (455 ± 48 ppb) for O₃, (119 ± 5) ppb for NO₂, and (62 ± 9) ppb for NO. These values were significantly reduced compared to upstream of the PM where O₃ showed the strongest reduction by a factor of 24. Further adjustments to the ozone filter in our system will need to be considered for future investigations in order to control the ozone emission in compliance with the WHO Air Quality Guidelines (maximum of 50 ppb in an 8 h period per day) [25].

The antimicrobial potency of the DBD stack was tested and figure 6 shows the survivability of E. coli after exposure to the gas stream enriched with reactive gas species.

Zoom In Zoom Out Reset image size

The increase in the length of treatment corresponds with a higher mortality rate in E. coli. The reduction rate rises with increasing exposure time of the inoculated PM filters to the constant volume flow containing reactive species. We observed a maximum reduction rate of 99.96% after 15 min exposure. Even though the mechanisms by which NTP inactivates bacteria are not yet completely understood, it is widely accepted, that reactive gas species play an important role and interaction with the cell membranes and cell interior can ultimately lead to cell inactivation [26]. For our experimental conditions, only long-lived species such as O₃ and NO_x should contribute to these interactions. Since the temperature of the microdischarges as the hottest element in our experimental setup is close to room temperature according to the results in table 1, we conclude that heat energy plays a minor role in our experiments.

Table 2 summarizes the characteristics and results of studies with similar objectives. Comparison of different plasma devices in regards to energy density. However, these were conducted with different plasma sources, bacterial species, exposure times, or experimental conditions and microbiological methods. The energy densities were calculated to enable a comparative discussion.

A study by Kelly–Wintenberg et al obtained similar results for an effective sterilization of the air filters against S. aureus. Main differences to our approach, however, were the use of a planar plasma electrode configuration based on surface DBD directly integrated directly on the surfaces of a high efficiency particulate air (HEPA) [27]. With their setup, they achieved 99.99% bacterial reduction after 5 min. When comparing practical aspects of both concepts, our system consists of modular components and a change of the PM filter after its service lifetime can be performed without affecting the plasma electrode arrangement. In terms of energy efficiency, our setup obtains almost identical inactivation results at only a quarter of the energy requirement. However, it is well established that gram-negative species are more susceptible to NTP than gram-positive species, which might explain this observation [28, 30].

Park et al proposed a packed-bed reactor comprising dielectric beads in combination with a HEPA filter and an ozone catalyst installed in some cm distance downstream the reactor for disinfection of circulating air [28]. They observed a reduction of 98.35% after 20 min operation for E. coli and 70.69% for M. luteus, respectively. While we operated our setup at only half the energy density of Park et al we could achieve a comparable inactivation of E. coli.

A recent study has applied a system that was similar to our approach but in a larger scale and also applying simultaneous filtration and decontamination with NTP. As opposed to our system, they used a combination of a corona discharge array, a metal foam as a PM filter, a mesh electrode and two ozone filters. Methodically, they collected bacteria from a 50.4 m³ office on the PM filter during one-hour-operation of their air purifier and refrained from artificial inoculation of PM filter surfaces. According to table 2, the authors showed a 68.65% reduction of undefined airborne bacteria at 27 J cm⁻² of energy density [29]. At an energy density comparable to Li et al, we could achieve 97.25% inactivation of E. coli. Even though the results cannot directly be compared due to differences in the experimental design and uncertainties concerning bacterial species, the results of Li et al can serve as a platform for future applications of our system in real indoor environments.

Up to now, no international threshold values for acceptable bacterial aerosol concentrations in indoor environments exist due to the lack of data on exposure-response relationships [31]. Still, numerous studies have accessed the bacterial aerosol concentrations in different indoor environments and found levels between 170 CFU m⁻³ in operating theatres in hospitals to 3224 CFU m⁻³ in primary schools [32, 33]. Considering continuous operation of our experimental setup in such environments, we can estimate a continuous depositon of bacterial aerosols on the PM filter at a rate of 10⁶–10⁷ CFU m⁻² h⁻¹. From the data presented in figure 5, we determined the inactivation rate in units if CFU m⁻² h⁻¹ for each exposure time separately and subsequently calculated the mean value and standard deviation from the individual rates. The mean inactivation rate induced by our setup and for our experimental conditions can be estimated to (1.7 ± 0.6) 10⁹ CFU m⁻² h⁻¹. Even though this ratio is encouraging, the inactivation rate is valid only for gram-negative E. coli under our experimental conditions. Dominant bacterial species found in indoor air of hospitals, museums, offices, residences and classrooms are often from the genus Bacillus, Micrococcus and Staphylococcus [34]. As these gram-positive bacteria were generally found to be less susceptible to NTP than gram-negative bacteria, the inactivation rate of our apparatus calculated in this study cannot directly be transferd to realistic indoor environments [30].