Ozone Generation from a Germicidal Ultraviolet Lamp with Peak Emission at 222 nm

Abstract

Recent interest in commercial devices containing germicidal ultraviolet lamps with a peak emission wavelength at 222 nm (GUV222) has focused on mitigating virus transmission indoors while posing minimum risk to human tissue. However, 222 nm light can produce ozone (O₃) in air. O₃ is an undesirable component of indoor air because of health impacts from acute to chronic exposure and its ability to degrade indoor air quality through oxidation chemistry. In seven four-hour experiments we measured O₃ produced from a single filtered GUV222 lamp in a 31.5 m³ stainless steel chamber. Using an emission model, we determined an O₃ generation rate of 19.4 ppb_v h⁻¹ ± 0.3 ppb_v h⁻¹ (equivalent to 1.22 mg h⁻¹ ± 0.02 mg h⁻¹). We estimated the fluence rate from the lamp using two methods: (1) chemical actinometry using tetrachloroethylene (actinometry) and (2) geometric projection of the irradiance field from radial and angular distribution measurements of the GUV222 lamp fluence (irradiance). Using the estimated lamp fluence rates of 2.2 μW cm⁻² (actinometry) and 3.2 μW cm⁻² (irradiance) we predicted O₃ production in our chamber within 20 % of the average measured mixing ratio. Future studies should evaluate the indoor air quality impacts of GUV222 technologies.

Introduction

The on-going COVID-19 pandemic has highlighted the need for effective, in room, low energy air cleaning devices to enable safer in-person interactions in indoor environments.^1,2,3 Portable cleaning devices use a range of technologies that have uncharacterized impacts on indoor air quality.⁴ These impacts could result in human exposure to pollutants that are at odds with the intended benefit of the technology.⁵ One such technology is germicidal ultraviolet lamps that operate with a peak emission wavelength at 222 nm (GUV222). This wavelength is appealing as research to date indicates it does not significantly penetrate human skin and is effective at inactivating pathogens.^6,7,8

Air cleaning devices equipped with GUV222 are of particular importance when considering the potential for ozone (O₃) formation. In the range of 175 nm to 242 nm, molecular oxygen (O₂) will absorb light and dissociate with a quantum yield of unity to produce two ground state oxygen atoms (O; Reaction 1) that can then go on to recombine with O₂, in a termolecular reaction involving a collisional body (M = N₂ or O₂), to form O₃ via Reaction 2.^{9, 10,11}

$$O_2+hv\left(\lambda =175nmto242nm\right)\to 2O$$ (R1)

$$O+O_2+M\to O_3+M$$ (R2)

Reactions 1 and 2 are known to occur in the stratosphere where deep UV radiation is available.¹² Characterization of spectral output and potential for O₃ production is necessary when considering the application of GUV222 devices for mitigating virus transmission while maintaining good air quality in indoor environments.

While O₃ itself can be a harmful byproduct of air cleaner operation¹³, it can also react with gases and surfaces indoors¹⁴—including human skin¹⁵—leading to the formation of other potentially concerning byproducts such as gas-phase aldehydes and ultra-fine particulate matter¹⁶. Of particular concern is the exposure to O₃ and O₃-generated indoor pollutant byproducts from application of multiple GUV222 units in small and/or poorly ventilated indoor spaces.¹⁷ Here we present measurements of O₃ generation from a commercial GUV222 lamp in a stainless steel laboratory chamber, support our O₃ formation observations with a chemical kinetic model, and determine O₃ generation rates for this GUV222 lamp that can be used in future evaluations of GUV222 technologies in indoor spaces.

Methods

Measurement of the GUV222 Lamp Emission Spectrum.

Spectral irradiance measurements of a krypton chloride (KrCl) excimer GUV222 lamp were performed with a commercial UV spectrometer with detection sensitivity in the spectral range of 200 nm to 415 nm. GUV222 emission light was collected by an integrating sphere detector that was connected to the spectrometer by a UV-transmitting optical fiber patch cable. We estimate the uncertainty (k=2) of the spectral irradiance measurements at 222 nm to be 18 %. Additional details are provided in the supplemental information.

Operation of Chamber and Experiment Design.

We operated the commercial GUV222 lamp in a 31.5 m³ environmentally controlled walk-in chamber instrumented to measure O₃ and sulfur hexafluoride (SF₆; proton-transfer mass spectrometry) which was used to measure the chamber air change rate (Figure S8). A series of seven experiments were conducted to measure O₃ production from the GUV222 lamp. A metal fan was placed in the chamber to facilitate mixing. The GUV222 lamp was positioned in the upper corner of the chamber pointed down and towards the center of the chamber opposite of the fan (Figure S4). Two days prior to the seven GUV222 experiments the chamber was sealed and passivated with at least 100 ppb_v of O₃ for a total of 24 hours.

Prior to each experiment we operated the chamber to achieve a temperature of 20 °C and 50 % relative humidity. The chamber maintained a pressure of 101 kPa. Low levels of O₃, particulate matter, and volatile organic compound contaminants were achieved by passing outdoor air through MERV-13, carbon, and chemisorbant filters. This filtered outdoor air was recirculated in the chamber prior to each experiment. At the beginning of each experiment the chamber was sealed. The GUV222 lamp was then turned on for four hours over which O₃ concentration was measured. SF₆ was injected into the chamber at the start of each experiment and air change was determined from the first order loss constant (Figure S8). Tetrachloroethylene (C₂Cl₄) was vaporized and introduced to the chamber at the beginning of four of the experiments to measure the effective photon flux via actinometry (e.g., Peng, et al. 2023)¹⁸.

Results and Discussion

GUV222 Lamp Emission Spectrum.

Figure 1a shows the spectral irradiance versus wavelength of the GUV222 lamp measured directly under and at several distances from the lamp.

Figure 1.

(a) GUV222 emission spectra measured at six different distances. (b) The total irradiance versus distance. (c) Relative irradiance from GUV222 as a function of projection angle from 0 degrees to 75 degrees.

The main emission peak is at 222 nm, as reported by other studies examining emission spectra of KrCl lamps.^19,20 The lamp used in this study was a “filtered” lamp that the manufacturer states filters radiation from the emission band around 190 nm. We determined the total irradiance by integrating under the spectral irradiance curve over the entire emission range (Figure 1b). The total irradiance in the immediate vicinity of the lamp is high (105 W m⁻² at 0 cm and 27 W m⁻² at 5 cm) but drops very quickly with distance. The drop-off in total irradiance follows the relationship, E ∼ 1/d^1.52, at least up to 40 cm where E is irradiance and d the distance from the lamp. We find that the irradiance projection from the lamp is attenuated by 50 % at a projection angle of 35 ° and becomes negligible at an angle of 55 ° (Figure 1c).

Measurement and Modeling of O₃ Production from GUV222.

We measured elevated levels of O₃ in our chamber after four hours of GUV222 operation as shown in Figure 2.

Figure 2.

The average measured O₃ mixing ratio from seven GUV222 experiments is shown as the solid black line with the standard deviation (1σ) shown by the gray shaded area. The average O₃ mixing ratio predicted from the actinometry-informed model is shown as the dashed blue with the variability (1σ) too small to see on this graph. The average modeled O₃, predicted from the irradiance method, is shown in yellow with an 18 % error propagated from the uncertainty in the irradiance measurement. The O₃ measured from the experiment where the light was blocked is shown in purple.

Four hours after turning the GUV222 lamp on, we observed 53 ppb_v ± 1 ppb_v (106 μg m⁻³ ± 2 μg m⁻³) of O₃ in the chamber. To rule out the influence of other physical phenomena related to operation of the GUV222 lamp (e.g., electrical arching¹³) that could be responsible for O₃ production we operated the lamp, for one experiment, with the output covered to prevent light from illuminating the chamber. No O₃ generation was observed in that experiment (Figure 2, purple trace) providing evidence that photolysis of O₂ at 222 nm was responsible for production of O₃.

At the end of each experiment the lamp was turned off and the decay of O₃ was measured (Figure S6). We assume that O₃ is lost to stainless steel chamber surfaces and homogeneous gas-phase reactions via a first order process (k_loss). Additionally, some O₃ is lost via air change which was quantified from SF₆ decay measurements (e.g., removal via air change accounts for ≈ 6 % of total observed O loss; Figure S8). We determine the rate constant for the combined loss processes (k_decay) from a linear fit of the natural log of the O₃ mixing ratio versus time (Equation 1).

$$\ln \left(\frac{[O_3{]}_t}{[O_3{]}_0}\right)={-k}_{decay}t=-(k_{\dot{V}}{+k}_{loss})t$$ (Eqn. 1)

In Equation 1, ${\mathrm{k}}_{\mathrm{l}\mathrm{o}\mathrm{s}\mathrm{s}}$ is the first order rate constant for loss of O₃ to the chamber surfaces and homogeneous gas-phase reactions, ${\mathrm{k}}_{\dot{\mathbf{V}}}$ is the air change rate (h⁻¹), and $\mathrm{t}$ is time. We determine ${\mathrm{k}}_{\mathrm{l}\mathrm{o}\mathrm{s}\mathrm{s}}$ by subtracting the k_V̇ term (determined from decay measurements of SF₆) from the measured ${\mathrm{k}}_{\mathrm{d}\mathrm{e}\mathrm{c}\mathrm{a}\mathrm{y}}$. The rate constants for O₃ decay (${\mathrm{k}}_{\mathrm{d}\mathrm{e}\mathrm{c}\mathrm{a}\mathrm{y}}$) remained constant throughout the experiments varying by 2 % (Table 1).

Table 1.

Summary of O₃ decay constants, air change rates (k_V̇), and O₃ generation rates.

Experiment	kdecay (h−1)	kV̇ (h−1)	GR, O3 generation rate*
			μg h−1	ppbv h−1
1	0.172	0.010	1250	19.9
2	0.176	0.014	1220	19.5
3	0.169	0.010	1210	19.3
4	0.168	0.011	1210	19.3
5	0.167	0.012	1200	19.2
6	0.167	0.014	1200	19.2
7	0.171	0.012	1230	19.6
Average (±1σ)	0.170 ± 0.003	0.012 ± 0.002	1220 ± 20	19.4 ± 0.3

^*O₃ generation rates will be dependent on GUV222 path length, and thus may vary as a function of the size of a chamber or indoor space. Generation rates were calculated using the measured chamber volume, temperature, and pressure.

We calculate theoretical O₃ production from GUV222 using chemical production and loss and physical loss terms in Equation 2.

$$\frac{d\left[O_3\right]}{dt}=2j_{O2}\left[O_2\right]-\frac{2k_{1}j{O}_3[O_3{]}^2}{k_2\left[O_2\right]\left[M\right]}-(k_{\dot{V}}{+k}_{decay})[O_3]$$ (Eqn. 2)

The first term on the right hand side of Equation 2 is the O₃ production from photolysis of O₂ at 222 nm, the second term accounts for loss of O₃ through the odd-oxygen (O_x = O₃ + O) steady-state (k₁ = 7.96 × 10⁻¹⁵ cm³ molecule⁻¹ s⁻¹; k₂ = 6.10 × 10⁻³⁴ cm⁶ molecule⁻² s⁻¹) where [M] is the number density of air (M = N₂, O₂), and the third term accounts for depositional loss to chamber walls and homogeneous gas-phase reactions.

As shown in Equation 2, the photolysis rate of O₂ drives O₃ production from GUV222 and the first-order photolysis rate constant (j_O2) is strongly dependent on the photon flux (F; Equation 3) from the lamp.

$$j_{O2}=\int {\sigma }_{O2}{\Phi }_{O2}Fd\lambda$$ (Eqn. 3)

Using the measured irradiance spectrum (Figure 1a) from the lamp we calculate an effective O₂ absorption cross section (${\mathrm{\sigma }}_{\mathrm{O}2}$)⁹ of 4.30 × 10⁻²⁴ cm² across a wavelength ($\mathrm{\lambda }$) range between 210 nm and 230 nm (compared to 4.09 × 10⁻²⁴ cm² at 222 nm). The photolysis quantum yield of O₂ (${\Phi }_{O2}$) between 210 nm and 230 nm is unity.²¹ We estimate an effective photon flux (F) from GUV222 from two different methods: (1) by determining the average of the measured irradiance projected into a cone (irradiance method) and (2) following the method of Peng, et al. (2023)¹⁸, using chemical actinometry²² with C₂Cl₄ as the actinometer (actinometry method).

For the irradiance method, we generated an irradiance field within a 31.5 m³ cone by expanding the GUV222 irradiance point source axially following the relationship, E ∼ 1/d^1.52, and angularly following a relatively tight half-angle of ≈ 55° (Equation S4). We then averaged the projected irradiance over the emission volume to get the effective photon flux. For the actinometry method, C₂Cl₄ was introduced to the chamber and the GUV222 lamp was turned on for four hours to measure the C₂Cl₄ photolysis rate. Using the measured C₂Cl₄ photolysis rate, effective C₂Cl₄ absorption cross section (σ_C2Cl4)²³, and reported photolysis quantum yield (Φ _C2Cl4), we determined the effective photon flux (Equation S8). Between 210 nm and 230 nm, effective GUV222 fluence rates of 3.2 μW cm⁻² and 2.2 μW cm⁻² were determined from the irradiance method and actinometry, respectively. Using our average measured O₃ from the experiments and solving for the photon flux through our chemical model, we estimate an experimentally determined effective fluence rate of the GUV222 lamp to be 2.6 μW cm⁻². Details of the effective photon flux determination methods are discussed in the supplemental information.

For the irradiance method, O₃ levels are over-predicted by 18 %. We expect over-estimation of the effective photon flux using this irradiance method because we are not accounting for attenuation of the incident radiation by interactions with the chamber walls. Ma, et al. (2023) recently demonstrated that different types of stainless steel reflect 222 nm light with an efficiency on the order of 20 %.²⁴ The 31.5 m³ modeled conical irradiance field slightly extends beyond the chamber walls, but the lamp was positioned in a corner of the chamber such that a large volume of the chamber air was irradiated by the UV light. Our calculations indicate most of the O₃ is created within 2 m of the lamp (Table S1), so the cone extending beyond the chamber walls results in small overestimation of O₃ production. Accurately accounting for reflectance and exact chamber dimensions would decrease the effective photon flux and thus modeled O₃ production.

In contrast to the irradiance method, the model underpredicts O₃ levels by 18 % using the actinometry method. An effective fluence rate of 2.6 μW cm⁻² (k_decay= 0.19 h⁻¹) would be needed to reconcile the 18 % deficit in modeled O₃ production. Despite some discrepancies between modeled and measured O₃, our calculations provide evidence to suggest the mechanism of O₃ production from our GUV222 lamp is photolysis of O₂ from 222 nm light, and not from other physical phenomena (e.g., electrical arching).

Determination of O₃ Generation Rates from GUV222 Lamps.

In the chamber experiments O₃ was generated from GUV222 while simultaneously being lost through air change, gas-phase reactions, and deposition to surfaces. Thus, the O₃ production rate from the lamp can be determined by solving for the generation rate (GR) in the transient solution to the mass balance equation presented in Equation 4.

$$[O_3{]}_t=[O_3{]}_i{e}^{-(k_{\dot{V}}+k_{decay})t}+\frac{\frac{GR}{V}}{(k_{\dot{V}}+k_{decay})(1-e^{-\left(k_{\dot{V}}+k_{decay}\right)t})}$$ (Eqn. 4)

Where $[{\mathrm{O}}_3{]}_{\mathrm{i}}$ and $[{\mathrm{O}}_3{]}_{\mathrm{t}}$ are the initial and time t O₃ mixing ratios, V is the volume of the chamber (31.5 m³), and GR is the O₃ generation rate (μg h⁻¹).

Calculated GUV222 O₃ generation rates (converted from μg h⁻¹ to ppb_v h⁻¹ using the 31.5 m³ chamber volume), presented in Table 1, varied within 2%.

From an average of seven experiments, we measured an O₃ generation rate from the GUV222 lamp of 19.4 ppb_v h⁻¹ ± 0.3 ppb_v h⁻¹ (equivalent to 1.22 mg h⁻¹ ± 0.02 mg h⁻¹) that, from our chemical modeling, implies a lamp fluence rate of 2.6 μW cm⁻².

We measured the spectral irradiance of a commercial GUV222 lamp from 210 nm to 230 nm showing a peak emission at 222 nm. Results from seven replicate experiments of the single commercial GUV222 lamp used in this study yielded a mean O₃ generation rate of 19.4 ppb_v h⁻¹ and, from our chemical modeling, implies a lamp fluence rate of 2.6 μW cm⁻². The results observed in this study apply to this lamp and may vary between unit, manufacturer, and test conditions. For instance, when comparing the GUV222 fluence rate normalized O₃ production rate (i.e., O₃ production rate/fluence rate = O₃ production efficiency) measured in this study to two other recent studies^18,25 we find that the O₃ production efficiencies can vary within 30 % when measured from the same model lamp (Table S2).

O₃ generation rates determined in this study could be used to predict O₃ production and accumulation in indoor spaces from commercial GUV222 lamps like the one used in this study. We note that the effective fluence rate (i.e., average amount of light irradiating a given volume; Table S2) will depend on the GUV222 path length and so the O₃ generation values reported here are most applicable to indoor spaces with volumes equal to, or larger than, 31.5 m³. Like the losses of O₃ to chamber walls and gas-phase reactions we observed, much higher reactive losses of O₃ generated from GUV222 devices would be expected in real indoor environments (e.g., k_loss = 2 h⁻¹ (surfaces) and k_loss = 0.09 h⁻¹ per person (human envelope))²⁶ with potential impacts for byproduct formation that would affect indoor air quality²⁷. Additionally, we note that the radiation dosing necessary to achieve high levels of virus deactivation may require multiple GUV222 lamps²⁸ and thus our measured O₃ generation rate could be used to simulate the effects of multiple lamp installations on indoor air quality. We suggest more measurements of O₃ production should be made from commercial air cleaning devices that use GUV222 lamps to assess the impacts on indoor air quality in both real indoor and laboratory settings.

Synopsis.

Devices using 222 nm light have potential benefits in virus transmission mitigation. We highlight unintended ozone production from a 222 nm lamp.