Ozone Generation and Chemistry from $222\mathsf{nm}$ Germicidal Ultraviolet Light in a Fragrant Restroom

Abstract

Devices using $222\text{nm}$ germicidal ultraviolet light ($\text{GUV}222$) have been marketed to reduce virus transmission indoors with low risk of occupant harm from direct UV exposure. $\text{GUV}222$ generates ozone, an indoor air pollutant and oxidant, under constrained laboratory conditions, but the chemistry byproducts of $\text{GUV}222$-generated ozone in real indoor spaces is uncharacterized. We deployed $\text{GUV}222$ in a public restroom, with an air change rate of 1 h⁻¹ one weekend and 2 h⁻¹ the next, to measure ozone formation and byproducts generated from ozone chemistry indoors. Ozone from $\text{GUV}222$ increased background concentrations by 5 ppb on average for both weekends and reacted rapidly (e.g., at rates of 3.7 h⁻¹ for the first weekend and 2.0 h⁻¹ for the second) with gas-phase precursors emitted by urinal screens and on surfaces. These ozone reactions generated volatile organic compound and aerosol byproducts (e.g., up to 2.6 μg m⁻³ of aerosol mass). We find that $\text{GUV}222$ is enhancing indoor chemistry by at least a factor of two for this restroom. The extent of this enhanced chemistry will likely be different for different indoor spaces and is dependent upon ventilation rates, species and concentrations of precursor $\text{VOCs}$, and surface reactivity. Informed by our measurements of ozone reactivity and background aerosol concentrations, we present a framework for predicting aerosol byproduct formation from $\text{GUV}222$ that can be extended to other indoor spaces. Further research is needed to understand how typical uses of $\text{GUV}222$ could impact air quality in chemically diverse indoor spaces and generate indoor air chemistry byproducts that can affect human health.

Graphical Abstract

Introduction

In the early stages of the COVID-19 pandemic, challenges associated with mitigating airborne virus transmission in public spaces motivated the development of new ventilation and disinfection standards in the United States.¹ In particular, the American Society of Heating, Refrigerating, and Air-Conditioning Engineers (ASHRAE) created Standard 241-2023 “Control of Infectious Aerosols”. This standard prescribes minimum effective per person clean airflow rates to be achieved using mechanical ventilation supplemented by air disinfection technologies when the system is in an infection risk management mode—if necessary.² Germicidal ultraviolet light with a peak emission wavelength at $222\text{nm}$ ($\text{GUV}222$) has been promoted as a possible airborne pathogen transmission mitigation tool. $\text{GUV}222$ could be an important tool because it possibly can efficiently deactivate airborne viruses^3-6 and directly irradiate human occupied spaces—potentially without causing acute or chronic harm to skin or eyes^7-9. Public transportation¹⁰, restrooms^{11, 12}, elevators¹³, and environments where occupants are in close contact (e.g., classrooms and conference rooms) are examples of public spaces where $\text{GUV}222$ could be useful.¹⁴

An unintended characteristic of $\text{GUV}222$ is that it produces ozone (${\mathrm{O}}_3$), an indoor air pollutant, which can oxidize chemicals in the air, on particles, and on surfaces; resulting in degraded indoor air quality.^15-17 Several studies have determined, through calculations^{9, 18, 19}, that ${\mathrm{O}}_3$ concentrations produced from $\text{GUV}222$ would not be expected to exceed the Environmental Protection Agency 8-hour National Ambient Air Quality Standard for outdoor air of 70 parts-per-billion (70 nmol of ${\mathrm{O}}_3$ per mol of air) when properly installed (i.e., appropriate irradiance levels and ventilation rates) in indoor spaces. However, no minimum threshold currently exists for which ${\mathrm{O}}_3$ exposure may produce adverse health outcomes^20-22. Further, the reactive chemistry that ${\mathrm{O}}_3$ initiates indoors²² can produce gas-phase and particulate pollutants that may represent greater exposure risks than ${\mathrm{O}}_3$ itself^{23, 24}. Our knowledge of real-world impacts of $\text{GUV}222$ on indoor air quality is limited, in part, by a lack of field measurements in indoor spaces where $\text{GUV}222$ may typically be used.

In most indoor spaces the only source of ${\mathrm{O}}_3$ is from outdoor air infiltration which creates a background amount of byproduct formation from ${\mathrm{O}}_3$ chemistry²⁵. Introducing an additional source of ${\mathrm{O}}_3$, such as with $\text{GUV}222$, can potentially enhance byproduct formation above background levels¹⁵. Although chemically identical, we differentiate the ${\mathrm{O}}_3$ that is generated by $\text{GUV}222$ ($\text{GUV}222$-generated ${\mathrm{O}}_3$) from the ${\mathrm{O}}_3$ that is brought indoors from outdoor air infiltration to discuss the enhanced impacts of $\text{GUV}222$ on indoor ${\mathrm{O}}_3$ chemistry. We note that the ${\mathrm{O}}_3$ chemistry that we discuss in this work is applicable to any other device that would generate indoor ${\mathrm{O}}_3$ and there is nothing unique about the ${\mathrm{O}}_3$ that is generated from $\text{GUV}222$. When ${\mathrm{O}}_3$ reacts on surfaces, or with volatile organic compound ($\mathsf{VOC}$) precursors in the gas-phase, it can form oxygenated $\mathsf{VOC}$ byproducts including aldehydes and reactive oxygen species—like peroxides²⁶. Hydroxyl radicals (OH) are a stronger gas-phase oxidant than ${\mathrm{O}}_3$ and importantly are produced from reactions of ${\mathrm{O}}_3$ with unsaturated $\mathsf{VOCs}$. Formed OH can then react with other $\mathsf{VOCs}$ indoors that do not readily react with ${\mathrm{O}}_3$.¹⁶ Reactions of both ${\mathrm{O}}_3$ and OH can form gas-phase byproducts with sufficiently low vapor pressures that they can nucleate aerosol and/or condense to existing aerosol and increase aerosol mass concentrations. While we focus on ${\mathrm{O}}_3$ chemistry here, $222\text{nm}$ light has been shown to induce other reactions in the gas-phase²⁷ and in aqueous solution²⁸ suggesting chemistry other than that initiated by ozone may also impact indoor air quality.

Most studies suggest $\text{GUV}222$ should not be used in indoor spaces with “low” levels of ventilation^{15, 18, 29}. However, the conditions of “low” ventilation have not been defined. In addition, the role of ambient precursor $\mathsf{VOC}$ and particulate matter concentrations in regulating byproduct production from $\text{GUV}222$-generated ${\mathrm{O}}_3$ has not been addressed. In short, there is no real-world data to understand the relationship between ventilation (quantified here from the air change rate, $\text{ACR}$), precursor concentrations, and byproduct production potential from $\text{GUV}222$ that could be used to inform operational guidance.

Here we used a 59 m³ unoccupied restroom as a case study for understanding indoor air quality impacts of $\text{GUV}222$ in a real-world application. We deployed $\text{GUV}222$ in the restroom over two weekends. We investigated the role of ventilation in mitigating byproduct formation from $\text{GUV}222$ by providing no mechanical ventilation the first weekend ($\text{ACR}\approx 1{\mathrm{h}}^{\text{-}1}$) and with mechanical ventilation the second weekend ($\text{ACR}\approx 2{\mathrm{h}}^{\text{-}1}$). We used the removal of $\text{GUV}222$-generated ${\mathrm{O}}_3$ to understand how much ${\mathrm{O}}_3$ was reacting with surfaces and precursor $\text{VOCs}$ to form observed $\text{VOC}$ and aerosol ${\mathrm{O}}_3$ chemistry byproducts. We find that while increased ventilation can help decrease byproduct formation from $\text{GUV}222$, $\text{VOC}$ and aerosol byproducts are still measurable at an $\text{ACR}$ of approximately 2 h⁻¹. Informed by the changing precursor $\text{VOC}$ concentrations in the air in the restroom over the two weekends, we identify what conditions of precursor ${\mathrm{O}}_3$ reactivity, ventilation, and pre-existing aerosol concentrations are likely to produce measurable $\text{VOC}$ and aerosol byproducts from $\text{GUV}222$-generated ${\mathrm{O}}_3$ chemistry. Finally, we present a framework that could be applied to other indoor spaces to help determine the extent that $\text{GUV}222$ may degrade indoor air quality.

Materials and Methods

Experimental design

Measurements were conducted in an unoccupied 59 m³ restroom (Fig. 1) in a building on the campus of the National Institute of Standards and Technology (NIST) in Gaithersburg Maryland, USA on two consecutive weekends (weekend 1, Wk1, and weekend 2, Wk2). The floor and walls of the restroom are tile with a painted ceiling. Polished metal partitions separated the urinals and painted partitions separated the toilets. The surface area of the restroom, including partitions, is 142 m². For each weekend, the experiments started Friday evening at 8:00 p.m. and ended Monday morning at 6:00 a.m. During the experiments the entrance to the restroom was kept closed and remained unoccupied. Otherwise, the restroom was open for use by NIST staff and guests throughout the week. While the door was kept closed, the ingress of sampling lines prevented the door from fully closing. The average temperature in the restroom for the two weekends was 25 °C ± 1 °C with an average relative humidity of 45 % ± 3 %. $\text{ACRs}$ were measured using acetone as a tracer during Wk1 and sulfur hexafluoride (SF₆) as a tracer in Wk2 (Supplementary Materials, Section S3).

Fig. 1. Schematic of the tested restroom.

Toilets and stalls are shown on the bottom part of the schematic. Stall doors were opened inwards and are not shown. A segment of wall separates two parts of the restroom where each side contains two sinks and two urinals. The entrance to the restroom is shown in the bottom left corner of the room. The exhaust vents for the mechanical ventilation are located above the urinals and one is above the middle of the stalls. The window recess on the top left is where the scanning mobility particle sizer (SMPS) instrument was located. $\text{GUV}222$ lamps are labelled in the figure. The mixing fan was placed below one of the $\text{GUV}222$ lamps. The lamps and fan were controlled automatically using a digitally controlled relay.

For Wk1, outside air and air from the hallway passively entered the restroom providing moderate ventilation ($\text{ACR}\approx 1{\mathrm{h}}^{\text{-}1}$). For Wk2, the mechanical ventilation system was turned on which created a negative pressure in the restroom and drew air in from the hallway then exhausted it out through ventilation ducts ($\text{ACR}\approx 2{\mathrm{h}}^{\text{-}1}$). Outdoor air infiltration was also possible, but we did not measure the chemical composition of outdoor air. Instead, we use measurements of $\text{VOCs}$ and ${\mathrm{O}}_3$ in the hallway to quantify the influence of hallway air ventilation on the restroom.

On weekdays custodial staff cleaned the restroom in the morning around 10:00 a.m. Four porcelain urinals are in the restroom, and each held a urinal screen. As part of scheduled routine maintenance, custodial staff replaced old urinal screens with new ones approximately 38 hours before the first $\text{GUV}222$ on/off cycle on Wk1. At the start of the first $\text{GUV}222$ on/off cycle for Wk2 the urinal screens were approximately 206 hours old. Using online proton-transfer mass spectrometry coupled to gas-chromatography (GC-PTR-MS) we measured the concentrations of ${\mathrm{O}}_3$-reactive $\text{VOCs}$ emitted from the urinal screens in the restroom including α-terpinene, linalool, limonene, ocimene, linalyl acetate, and terpinolene. The concentrations of reactive $\text{VOCs}$ decreased by approximately an order of magnitude from the beginning of Wk1 to the beginning of Wk2 due to the combined effects of lower emission rates from the urinal screens and the increased air change rate (Fig. S1-4).

Three $\text{GUV}222$ lamps were installed at different locations in the restroom with two projecting light down at a 20° angle normal to the floor from a height of 2.4 m and one projecting light parallel to the floor at a height of 2.0 m and unobstructed for 1.4 m. From measurements of the radial and angular distribution of the light from the lamps, informed by modeling in our previous study¹⁷, we expect that > 75 % of the ${\mathrm{O}}_3$ is generated within 2 m of the lamp (Figure S9). All of the lamps experienced some attenuation of the light by restroom walls and/or stall partitions. Lamp, mixing fan, and do-it-yourself (DIY) air cleaner on/off cycles were automated using a custom computer program and a digitally controlled relay. Our measurements of per-lamp ${\mathrm{O}}_3$ generation rates (determined from mass-balance modeling of ${\mathrm{O}}_3$ concentrations and discussed later) suggest a room-averaged fluence rate of 3.2 μW cm⁻² which is more than a value of 1.5 μW cm⁻² suggested by a recent study³⁰ to accomplish disinfection while limiting human exposure to radiation. Operational guidance on recommended fluence rates from $\text{GUV}222$ currently does not exist.

The background (“$\text{GUV}222$ OFF”) air quality of the restroom during Weekend One (Wk1) was characterized by high concentrations of reactive terpenoid $\text{VOCs}$ (20 ppb of total terpenoids) whereas lower reactive terpenoid $\text{VOC}$ concentrations (5 ppb of total terpenoids) defined background conditions for Weekend Two (Wk2).

Instrumentation

A scanning mobility particle sizer was located in the restroom and sampled particles from 10 nm to 420 nm with 1.5 min resolution. All chemical instrumentation was located down the hall from the restroom in a nearby conference room. A 9.5 mm (3/8”) outer diameter PFA sampling line for $\text{VOCs}$ and a 6.3 mm (1/4”) outer diameter sampling line for trace gases (${\mathrm{O}}_3$, ${\text{NO}}_{\mathsf{x}}$, and formaldehyde) were run from the conference room to the restroom to two separate multi-port valve systems that switched between three locations (hallway, restroom center, and restroom ventilation duct) on three-minute intervals. Two bypass pumps were used to draw air at a flowrate of 15 L min⁻¹ through the $\text{VOC}$ line and 10 L min⁻¹ through the trace gas (${\mathrm{O}}_3$, ${\text{NO}}_{\mathsf{x}}$, and formaldehyde) line. Sampling lines for both the trace gases and the PTR-MS were located approximately 2.5 m directly in the line with the radial projection of light from one of the lamps. We do not expect enhanced local production of ${\mathrm{O}}_3$ or other byproducts at the gas sampling locations because most of the light is attenuated after 2 m. ${\mathrm{O}}_3$ was measured with a dual-beam absorption spectrometer, ${\text{NO}}_{\mathsf{x}}$ was measured with a cavity-attenuated phase-shift spectrometer, and formaldehyde was measured with an infrared laser absorption spectrometer.

A Time-of-flight Proton-Transfer Mass Spectrometer (PTR-MS) equipped with a gas chromatograph (GC) was used to sample $\text{VOCs}$. Most of the time the PTR-MS sampled in real-time acquisition at a rate of 1 Hz. GC samples were collected several times during the two weekends. Hourly calibrations were performed using two separate multi-component $\text{VOC}$ standard calibration cylinders during Wk1 and two NIST-certified monoterpene calibration cylinders during Wk2 (Tables S2 and S3). Hourly backgrounds were measured using ultra zero air. PTR-MS signals that were not calibrated for directly were converted to concentrations using a parameterization of the instrument sensitivity with the proton-transfer rate constant^31-33.

Terpenoids were an important source of gas-phase reactivity in the restroom. We provide an in-depth discussion of terpenoid quantification in the Supplementary Materials. Briefly, we use both GC-PTR-MS and real-time PTR-MS data to quantify terpenoids. We quantified 1,8-cineloe and linalool from GC-PTR-MS measurements. Additionally, we quantified monoterpenes including α-pinene, β-pinene, limonene, α-terpinene, ocimene, terpinolene, and camphene from GC-PTR-MS measurements when available. When GC-PTR-MS samples were not available we used the real-time measurement of the C₁₀H₁₇⁺, for monoterpenes, and C₁₀H₁₉O⁺ ion signals to estimate increases or decreases in terpenoid concentrations from the GC-PTR-MS measurements. Linalyl and isobornyl acetate were quantified from the real-time C₁₂H₂₁O₂⁺ ion signal. We assumed that decreases in the C₁₂H₂₁O₂⁺ ion signal that occurred when the lamps were on indicated reaction of linalyl acetate with ${\mathrm{O}}_3$.

${\mathbf{O}}_3$ Reactive Loss Calculations

Recently, the difference between outdoor and indoor ${\mathrm{O}}_3$ concentrations, “${\mathrm{O}}_3$ loss”, has been suggested to be a surrogate for predicting gas and particle concentrations of ${\mathrm{O}}_3$ chemistry byproducts indoors²⁵. Similar to the ${\mathrm{O}}_3$ loss framework presented by Weschler and Nazaroff (2023), we compare the background ${\mathrm{O}}_3$ reactive loss from hallway infiltration to the ${\mathrm{O}}_3$ reactive loss from $\text{GUV}222$-generated ${\mathrm{O}}_3$ and quantify the enhanced byproduct formation induced by $\text{GUV}222$, compared to background conditions, in the restroom.

The ${\mathrm{O}}_3$ chemistry that occurs in the restroom during both weekends without the influence of the $\text{GUV}222$ lamps defines background conditions. Under background conditions in the restroom, the primary source of ${\mathrm{O}}_3$ is passive ventilation from the hallway. However, $\text{GUV222}$ introduces an additional source of ${\mathrm{O}}_3$ to the restroom. Once in the restroom, either introduced from $\text{GUV}222$ or from the hallway, ${\mathrm{O}}_3$ is removed by ventilation and reactive losses. The ${\mathrm{O}}_3$ concentration and reactive loss in the restroom can thus be modeled following the mass-balance in Equation 1,

$$\begin{gathered} {[{\mathrm{O}}_3]}_{\text{rest},\mathrm{t}}={[{\mathrm{O}}_3]}_{\text{rest},0}{\mathrm{e}}^{-{\mathrm{k}}_{\mathrm{O}3\text{removal}}\cdot \mathrm{t}}+ \\ \frac{(\frac{{\mathrm{N}}_{\text{lamp}}\cdot {\text{GR}}_{03}}{\mathrm{V}}+[{\mathrm{O}}_3]{}_{\mathrm{hall}}\cdot \text{ACR})}{{\mathrm{k}}_{\mathrm{O}3\text{removal}}}(1-{\mathrm{e}}^{-{\mathrm{k}}_{\mathrm{O}3\text{removal}}\cdot \mathrm{t}}) \end{gathered}$$ (1)

where [${\mathrm{O}}_3$] is the ${\mathrm{O}}_3$ concentration in the restroom (rest) or hallway (hall), $\text{ACR}$ is the air change rate (h⁻¹), ${\mathrm{N}}_{\text{lamp}}$ is the number of $\text{GUV}222$ lamps in the restroom (${\mathrm{N}}_{\text{lamp}}=3\text{lamps}$), ${\text{GR}}_{03}$ is the per-lamp ${\mathrm{O}}_3$ generation rate (μg h⁻¹ lamp⁻¹), $\mathrm{V}$ is the volume of the restroom (59 m³), and ${\mathrm{k}}_{03\text{removal}}$ is the first-order loss rate constant representing the sum of all processes removing ${\mathrm{O}}_3$ from the restroom (h⁻¹). We modeled ${\mathrm{O}}_3$ concentrations in the restroom using Equation 1 for Wk1 and Wk2 to determine the ${\text{GR}}_{03}$ and ${\mathrm{k}}_{03,\text{removal}}$. We calculated an ${\text{GR}}_{03}$ of 850 μg h⁻¹ lamp⁻¹ for Wk1 and an ${\text{GR}}_{03}$ of 1000 μg h⁻¹ lamp⁻¹ for Wk2 resulting in an average ${\text{GR}}_{03}$ for Wk1 and Wk2 in the restroom of 930 μg h⁻¹ lamp⁻¹ ± 110 μg h⁻¹ lamp⁻¹ (≈ 8 ppb h⁻¹ lamp⁻¹ ± 1 ppb h⁻¹ lamp⁻¹ in the 59 m³ restroom). The ${\text{GR}}_{03}$ we measured in the restroom is approximately 25 % lower than the ${\text{GR}}_{03}$ we measured from one of the lamps in our chamber (1220 μg h⁻¹)¹⁷ likely due to attenuation of the light by the walls and the stall partitions.

Ventilation and total reactive losses (${\mathrm{k}}_{03,\text{reactive loss}}$) define ${\mathrm{k}}_{03,\text{removal}}$,

$${\mathrm{k}}_{\mathrm{O}3.\text{removal}}={\mathrm{k}}_{\mathrm{O}3.\text{reactive loss}}+\text{ACR}$$ (2)

We distinguish ${\mathrm{O}}_3$ removal from ${\mathrm{O}}_3$ reactive loss because ${\mathrm{O}}_3$ removed via air change does not contribute to the chemistry in the restroom. ${\mathrm{O}}_3$ removal is the sum of all processes removing ${\mathrm{O}}_3$ from the restroom and ${\mathrm{O}}_3$ reactive loss is the first-order rate constant representing the sum of all ${\mathrm{O}}_3$ gas-phase and surface reactive loss processes,

$${\mathrm{k}}_{\mathrm{O}3.\text{reactive loss}}={\mathrm{k}}_{\text{loss}.\text{surface}}+{\mathrm{k}}_{\text{loss}.\text{NO}}+{\mathrm{k}}_{\text{loss}.\text{VOC}}$$ (3)

where ${\mathrm{k}}_{\text{loss},\text{surface}}$ is the first-order ${\mathrm{O}}_3$ reactive loss rate constant to surfaces (h⁻¹), ${\mathrm{k}}_{\text{loss},\text{NO}}$ is the first-order rate constant for reaction with $\text{NO}$ (h⁻¹), and ${\mathrm{k}}_{\text{loss},\text{VOC}}$ is the first-order rate constant for reaction with gas-phase $\text{VOCs}$ (h⁻¹).

When modeling ${\mathrm{O}}_3$ concentrations in the restroom, using Equation 1, we constrained the $\text{ACR}$, ${\mathrm{k}}_{\text{loss},\text{VOC}}$, and ${\mathrm{k}}_{\text{loss},\text{NO}}$ with measurements—leaving ${\mathrm{k}}_{\text{loss},\text{surface}}$ as an unknown variable. We determined the $\text{ACR}$ using Equation S2, from the decay of acetone, for Wk1, and the decay of SF₆, for Wk2 (Fig. S7). ${\mathrm{k}}_{\text{loss},\text{NO}}$ is determined by multiplying the bimolecular rate constant for the reaction of $\text{NO}$ with ${\mathrm{O}}_3$ (${\mathrm{k}}_{\text{NO}+\mathrm{O}3}\approx 1.9\times {10}^{\text{-}14}{\text{cm}}^3$ molecules⁻¹ s⁻¹ at 298 K)³⁴ with the $\text{NO}$ concentration (molecules cm⁻³) measured in the restroom.

$${\mathrm{k}}_{\text{loss}.\text{NO}}={\mathrm{k}}_{\text{NO}+\mathrm{O}3}\cdot {[\text{NO}]}_{\text{restroom}}$$ (4)

Reactions with terpenoids were the most important gas-phase sinks of ${\mathrm{O}}_3$ in the restroom and we calculated the contribution of terpenoid + ${\mathrm{O}}_3$ reactions to ${\mathrm{k}}_{\text{loss},\text{VOC}}$ following Equation 5,

$${\mathrm{k}}_{\text{loss}.\text{VOC}}=\sum ({\mathrm{k}}_{\text{VOC}.\mathrm{i}+\mathrm{O}3}\cdot {[\text{VOC}]}_{\mathrm{i}})$$ (5)

where $\mathsf{i}$ denotes an individual terpenoid, ${\mathrm{k}}_{\text{VOC}}+\mathrm{O}3$ is the bimolecular rate constant for the reaction of ${\mathrm{O}}_3$ and a terpenoid (cm³ molecules⁻¹ s⁻¹), and [$\text{VOC}$] is the terpenoid concentration in the restroom (molecules cm⁻³). α-Terpinene, linalool, and linalyl acetate contributed the most to the gas-phase ${\mathrm{O}}_3$ + $\text{VOC}$ reactive loss with ocimene, terpinolene, and limonene making smaller contributions (Fig. S8). Other terpenoids made negligible contributions to ${\mathrm{k}}_{\text{loss},\text{VOC}}$ because they were either present in low concentrations (α-pinene, β-pinene) and/or were not reactive enough to ${\mathrm{O}}_3$ (e.g., camphene). Many $\text{VOCs}$ emitted from the urinal screens were effectively unreactive towards ${\mathrm{O}}_3$ (e.g., 1,8-cineole, isobornyl acetate), but could be reactive to OH generated from ${\mathrm{O}}_3$ + $\text{VOC}$ reactions. These $\text{VOCs}$ may also contribute to $\text{VOC}$ or particle byproduct formation from OH reactions, but potential OH chemistry was indistinguishable from ${\mathrm{O}}_3$ chemistry in our experiments. We assume any ${\mathrm{O}}_3$ removal not explained by air change or by reaction with $\text{NO}$ or $\text{VOCs}$ is attributable to surface loss (${\mathrm{k}}_{\text{loss},\text{surface}}$). Our measured and modeled ${\mathrm{O}}_3$ concentrations, using Equations 1 through 5, are different by no more than 20 % for each weekend (Fig. S11 and S12). More details on the ${\mathrm{O}}_3$ modeling are provided in the Supplementary Materials (Section S6).

Respirable particle modeling

An open-source multiple-path particle dosimetry model was used to calculate deposition efficiencies of respired particles in the head, pulmonary, and tracheobronchial regions of the respiratory tract (Fig. S14). An inhalation rate of 0.01 m³ min⁻¹ was assumed for an upright adult at rest. The total number of respired particles was determined by multiplying the total particle number concentration size distribution, measured when total particle number concentrations were at maximum value while $\text{GUV}222$ was on, by the size-dependent deposition efficiencies and then integrating across the particle size range.

Results and Discussion

$\mathrm{GUV}222$ generates ${\mathbf{O}}_3$ and ${\mathbf{O}}_3$ chemistry byproducts in a restroom

We summarize the ${\mathrm{O}}_3$ generation and chemistry leading to $\text{VOC}$ and aerosol byproduct formation from $\text{GUV}222$ operation in Fig. 2. We express ${\mathrm{O}}_3$ and $\text{VOC}$ concentrations in parts-per-billion (ppb) defined as the number density (molecules cm⁻³) of ${\mathrm{O}}_3$ or a $\text{VOC}$ divided by the air number density (≈ 2.43 x 10¹⁹ molecules cm⁻³ at Standard Ambient Temperature and Pressure). For every oxygen molecule that is photolyzed by $222\text{nm}$ radiation, two ground-state oxygen atoms are formed, $\mathrm{O}({}^3\mathrm{P})$, that then combine with molecular oxygen (in a termolecular reaction with a collisional body, M) to form two molecules of ${\mathrm{O}}_3$^{16, 17, 29}.

Fig. 2. Chemical processes promoted by $\mathrm{GUV}222$ leading to ${\mathsf{O}}_3$ chemistry byproducts in the restroom.

Urinal screens (pink) are the primary source of gas-phase reactive precursor $\text{VOCs}$ in the restroom. Purple colors show ${\mathrm{O}}_3$ reactive loss mechanisms ($\text{VOCs}$, $\text{NO}$, and surfaces). $\text{VOCs}$ react with ${\mathrm{O}}_3$ to form oxidized $\text{VOCs}$ (${\text{VOC}}_{\text{Ox}}$) that we categorize as either ${\mathrm{O}}_3$ oxidation $\text{VOC}$ byproducts (${\text{VOC}}_{\mathrm{O}3\text{Ox}}$) or gas-phase condensable $\text{VOCs}$ (${\text{VOC}}_{\text{cond},\mathrm{g}}$). Green colors show ${\text{VOC}}_{\text{cond},\mathrm{g}}$ loss mechanisms. ${\text{VOC}}_{\text{cond},\mathrm{g}}$ can nucleate particles that then can undergo coagulation and condensation to form aerosol mass (${\mathrm{M}}_{\mathrm{p}}$). ${\text{VOC}}_{\text{cond},\mathrm{g}}$ can directly condense on aerosol and surfaces. All species shown here are subject to removal or replenishment via air change.

$${\mathrm{O}}_2+\text{hv}(\lambda =222\text{nm})\to 2\mathrm{O}({}^{3}P)$$ (R1)

$$2\mathrm{O}({}^{3}P)+2{\mathrm{O}}_2+2M\to 2{\mathrm{O}}_3+2M$$ (R2)

Once ${\mathrm{O}}_3$ is generated by $\text{GUV}222$ it can undergo several fates including removal by air change as well as reactions with nitric oxide ($\text{NO}$), reactive $\text{VOCs}$ (e.g., emissions from urinal screens in the case of this restroom study), and surfaces. NO can be introduced indoors via outdoor air infiltration³⁵ and can react rapidly with ${\mathrm{O}}_3$. In many indoor spaces, reactive uptake of ${\mathrm{O}}_3$ to surfaces is a major loss process that can occur as rapidly as ${\mathrm{O}}_3$ removal via the $\text{ACR}$.²² On the other hand, the reaction of ${\mathrm{O}}_3$ with many gas-phase $\text{VOCs}$ is too slow to be considered as a major ${\mathrm{O}}_3$ sink indoors unless there is a strong $\text{VOC}$ emission source from human occupancy, consumer products, cooking, and/or cleaning.²²

The reaction of $\text{GUV}222$-generated ${\mathrm{O}}_3$ with unsaturated $\text{VOCs}$ can form OH that can then rapidly oxidize $\text{VOCs}$. Oxidized $\text{VOCs}$ that are effectively unreactive with ${\mathrm{O}}_3$ in indoor spaces can also be oxidized by OH.¹⁶ Oxidation of $\text{VOCs}$ by ${\mathrm{O}}_3$ and OH produce oxidized $\text{VOCs}$ that we classify either as ${\mathrm{O}}_3$ $\text{VOC}$ oxidation byproducts (${\text{VOC}}_{\mathrm{O}3\text{Ox}}$) or gas-phase condensable $\text{VOCs}$ (${\text{VOC}}_{\text{cond},\mathrm{g}}$). Detailed representations of $\text{VOC}$ oxidation product volatility distributions exist^36-38, but characterization of the distributions from chemistry in the restroom is beyond the scope of this study. ${\text{VOC}}_{03\text{Ox}}$ byproducts include chemicals like acetone, acetic acid, and aldehydes which we measured with our $\text{VOC}$ instrumentation. ${\text{VOC}}_{\text{cond},\mathrm{g}}$ byproducts can partition to existing aerosol and increase aerosol mass concentrations (${\mathrm{M}}_{\mathrm{p}}$). However, ${\text{VOC}}_{\text{cond},\mathrm{g}}$ can also partition to surfaces, including walls and restroom stalls. Given sufficiently low vapor pressures, ${\text{VOC}}_{\text{cond},\mathrm{g}}$ may also achieve supersaturation at low gas-phase concentrations and thus nucleate new particles upon formation of molecular clusters with other low-volatility $\text{VOCs}$, sulfuric acid, and/or ammonia.³⁹ We had a limited ability to measure ${\text{VOC}}_{\text{cond},\mathrm{g}}$ with our instrumentation.

${\mathrm{O}}_3$ and ${\mathrm{O}}_3$ chemistry byproducts were generated from $\text{GUV}222$ operation in the restroom during both weekends (shaded area under the curves in Fig. 3 and 4). $\text{GUV}222$ was operated for five three-hour on/off cycles during Wk1, and five four-hour cycles followed by four two-hour cycles during Wk2. The major driver of changes in indoor air chemistry from operation of $\text{GUV}222$ was production of ${\mathrm{O}}_3$. Production of ${\mathrm{O}}_3$ from $\text{GUV}222$ is discernible from non-$\text{GUV}222$ influences on ${\mathrm{O}}_3$ concentrations, like infiltration of hallway air to the restroom and titration of ${\mathrm{O}}_3$ by nitric oxide ($\text{NO}$; Fig. 3A and 4A). The average increase in the ${\mathrm{O}}_3$ concentration in Wk1, when $\text{GUV}222$ was on, was 4.0 ppb ± 1.1 ppb. The average increase in the ${\mathrm{O}}_3$ concentration in Wk2 when $\text{GUV}222$ was on was 5.8 ppb ± 1.2 ppb. After $\text{GUV}222$ was turned off, ${\mathrm{O}}_3$ concentrations in the restroom returned to background concentrations that vary as a function of the hallway ${\mathrm{O}}_3$ concentrations. Lower background ${\mathrm{O}}_3$ concentrations in the restroom, compared to the hallway, indicate considerable reactive ${\mathrm{O}}_3$ losses from reactions on surfaces and with gas-phase $\text{VOCs}$.

Fig. 3. Weekend 1 (Wk1) time series of ${\mathrm{O}}_3$, $\text{NO}$, and ${\mathrm{O}}_3$ chemistry byproducts.

The $\text{ACR}$ in Wk1 was ≈ 1 h⁻¹. Byproducts include ${\mathrm{O}}_3$ (purple), $\text{VOC}$ ${\mathrm{O}}_3$ oxidation products (${\text{VOC}}_{\mathrm{O}3\text{Ox}}$, black), particle number (${\mathrm{N}}_{\mathrm{p}}$, red), particle volume (${\mathrm{V}}_{\mathrm{p}}$, green), and particle mass (${\mathrm{M}}_{\mathrm{p}}$, green) concentrations. Pink shaded regions show when $\text{GUV}222$ was on. The striped, pink regions later in the weekend show shorter duration and more frequent $\text{GUV}222$ on/off cycling. Dotted lines indicate the 8-hour periods where a mixing fan was on. Shaded areas below the traces show the estimated increase in byproduct concentrations generated by $\text{GUV}222$. Hallway concentrations of $\text{VOC}$ ${\mathrm{O}}_3$ oxidation products (${\text{VOC}}_{\mathrm{O}3\text{Ox}}$) and ${\mathrm{O}}_3$ are used in ${\mathrm{O}}_3$ and ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ generation rate calculations. (A) A decrease in ${\mathrm{O}}_3$ concentrations (dark purple for restroom and light purple for hallway) with increased $\text{NO}$ (orange) concentrations is shown. (B) ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ is the sum of 34 $\text{VOCs}$ observed to increase in concentration when $\text{GUV}222$ was turned on. (C) The y-axis is ${\mathrm{N}}_{\mathrm{p}}$ divided by 1000. (D) ${\mathrm{V}}_{\mathrm{p}}$ and ${\mathrm{M}}_{\mathrm{p}}$ y-axis ranges are adjusted so the two traces are coincident.

Fig. 4. Weekend 2 (Wk2) time series of ${\mathsf{O}}_3$, $\mathsf{NO}$, and ${\mathbf{O}}_3$ chemistry byproducts.

The $\text{ACR}$ in Wk2 was ≈ 2 h⁻¹. The descriptions in the caption for Fig. 3 apply to this figure also (note the y-axis scaling for particle number). A MERV-13 do-it-yourself (DIY) air cleaner was used for two $\text{GUV}222$ on/off cycles at the end of the weekend (teal dotted lines and dots).

We measured increases in $\text{VOC}$ ${\mathrm{O}}_3$ oxidation products (${\text{VOC}}_{\mathrm{O}3\text{Ox}}$) and particle number (${\mathrm{N}}_{\mathrm{p}}$), volume (${\mathrm{V}}_{\mathrm{p}}$), and mass (${\mathrm{M}}_{\mathrm{p}}$) concentrations (particle diameter from 10 nm to 420 nm) from chemistry initiated by $\text{GUV}222$-generated ${\mathrm{O}}_3$. We expect the chemical composition of newly-formed aerosol to be mostly organic, resulting from terpenoid oxidation by ${\mathrm{O}}_3$ in the restroom, and thus we multiply ${\mathrm{V}}_{\mathrm{p}}$ by a density of 1.32 g cm⁻³ (average density of terpene secondary organic aerosol)⁴⁰ to quantify ${\mathrm{M}}_{\mathrm{p}}$. Higher concentrations of ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$. (Fig. 3B and 4B), ${\mathrm{N}}_{\mathrm{p}}$ (Fig. 3C and 4C), and ${\mathrm{M}}_{\mathrm{p}}$ (Fig. 3D and 4D) were formed from $\text{GUV}222$ in Wk1, under conditions of lower (≈ 1 h⁻¹) ventilation, versus Wk2, when the $\text{ACR}$ was higher (≈ 2 h⁻¹).

During both weekends, ${\mathrm{N}}_{\mathrm{p}}$ initially peaks to a maximum value after turning on $\text{GUV}222$ and then decreases to steady-state concentrations for the duration of the on cycle, unless the mixing fan was operating. In Wk2 when the mixing fan was operating we did not observe a discernible increase in ${\mathrm{N}}_{\mathrm{p}}$ during $\text{GUV}222$ operation compared to background ${\mathrm{N}}_{\mathrm{p}}$ variation. Steady-state production of ${\mathrm{M}}_{\mathrm{p}}$ in Wk1 was achieved approximately an hour after turning on $\text{GUV}222$ regardless of if the mixing fan was operating. Average ${\mathrm{M}}_{\mathrm{p}}$ produced from $\text{GUV}222$ during Wk1 was 1.8 μg m⁻³ ± 0.7 μg m⁻³ with a maximum observed value of 2.6 μg m⁻³. We note that in Wk2, an increase in ${\mathrm{M}}_{\mathrm{p}}$ was observed for approximately 10 hours (starting at t = 14 hours until t = 24 hours), but we hypothesize infiltration of aerosol from the hallway was the source of the increase in ${\mathrm{M}}_{\mathrm{p}}$. We only measured aerosol in the restroom and thus could not constrain the influence of infiltration on ${\mathrm{N}}_{\mathrm{p}}$ or ${\mathrm{M}}_{\mathrm{p}}$ with a simultaneous measurement of hallway concentrations.

We observed some effects on particle formation with use of a mixing fan and from using short duration high frequency on/off cycling of the lamps. The impacts MERV-13 DIY air cleaner use on ${\mathrm{N}}_{\mathrm{p}}$ and ${\mathrm{M}}_{\mathrm{p}}$ formation from $\text{GUV}222$ are not clear from our two tests and we will not discuss them here. Increasing the air mixing with the use of a fan suppressed increases in ${\mathrm{N}}_{\mathrm{p}}$ in Wk2 (Fig. S5 and Fig. S6) that were observed when the mixing fan was not used and $\text{GUV}222$ was on. We did not observe a suppression of new particle formation when the fan was on in Wk1. However, in Wk1 we observed decreases in ${\mathrm{N}}_{\mathrm{p}}$ when the mixing fan was first turned on and while the fan was on during $\text{GUV}222$ on cycles suggesting ${\mathrm{N}}_{\mathrm{p}}$ loss was enhanced by increased rates of coagulation or deposition to walls with fan mixing. We hypothesize fan mixing increased the loss of small particles (which otherwise would coagulate and serve as condensation nuclei) to walls in Wk2 thereby suppressing new particle growth. Recently, fan mixing prevented the growth of newly-formed particles < 3 nm to larger sizes in a chamber where occupants were exposed to ${\mathrm{O}}_3$⁴¹. Continued research into the influence of air mixing on particle formation dynamics from ${\mathrm{O}}_3$ reactions is warranted.

One potential application of $\text{GUV}222$ in a restroom is on/off cycling that is dependent on if an occupant is using a restroom stall. For an indoor space, like a restroom, where occupants generally use the space for short periods of time, short duration high frequency applications may be appropriate—assuming pathogen deactivation is also efficient on this timescale. In Wk1 we employed an intermittent use scheme where $\text{GUV}222$ was cycled on for two minutes and off for five minutes for an hour (on 16 min h⁻¹) and on for five minutes and off for 15 minutes for three hours (on 20 min h⁻¹). The shorter duration cycle produced less byproducts (including ${\mathrm{O}}_3$) compared to the longer duration cycle. The longer cycles resulted in measurable increases in ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$, ${\mathrm{N}}_{\mathrm{p}}$, and ${\mathrm{M}}_{\mathrm{p}}$. Our results indicate that intermittent use schemes may reduce byproduct formation from $\text{GUV}222$ compared to sustained periods of illumination but will not eliminate undesirable byproduct formation. Additionally, if we were to extend the total time the on/off cycling occurred, a buildup of byproducts may occur at low ventilation rates.

In the following sections we relate the reactive sinks of ${\mathrm{O}}_3$ to the production of $\text{VOC}$ byproducts—some of which go on to form aerosol byproducts.

Apportionment of ${\mathbf{O}}_3$ reactive loss

Fig. 5A shows the ${\mathrm{k}}_{\mathrm{O}3\text{removal}}$ determined from modeling of restroom ${\mathrm{O}}_3$ concentrations. Although average measured ${\mathrm{k}}_{\mathrm{O}3\text{removal}}$ for Wk1 (4.8 h⁻¹) and Wk2 (4.4 h⁻¹) were within 10 %, ${\mathrm{O}}_3$ ${\mathrm{k}}_{\text{loss},\text{VOC}}$ was five times higher in Wk1 (1.5 h⁻¹) compared to Wk2 (0.3 h⁻¹). The amount of ${\mathrm{O}}_3$ lost to gas-phase reactions with $\text{VOCs}$ in Wk2 decreased because initial reactive $\text{VOC}$ concentrations measured in the restroom decreased by approximately a factor of four from Wk1 to Wk2. ${\mathrm{k}}_{\text{loss},\text{surface}}$ was constant between the two weekends when the fan was off (Wk1 = 1.5 h⁻¹ and Wk2 = 1.7 h⁻¹), but use of the mixing fan in Wk2 increased ${\mathrm{k}}_{\text{loss},\text{surface}}$ by 40 % (2.4 h⁻¹). Mixing fan use only increased ${\mathrm{k}}_{\text{loss},\text{surface}}$ in Wk1 from 1.5 h⁻¹ to 1.6 h⁻¹. For simplicity, we only present and discuss ${\mathrm{O}}_3$ removal results from when the mixing fan was off.

Fig. 5. ${\mathsf{O}}_3$ removal rate constants and ${\mathsf{O}}_3$ removal apportionment.

(A) Average of measured ${\mathrm{O}}_3$ total removal rate constants (${\mathrm{k}}_{\mathrm{O}3\text{removal}}$) for each $\text{GUV}222$ on/off cycle, when the mixing fan was off, for each weekend apportioned by the contributing sinks. The error bars show the standard deviation of the total ${\mathrm{k}}_{\mathrm{O}3\text{removal}}$ averages. The apportionment of the ${\mathrm{O}}_3$ total removal rate constant attributed to $\text{ACR}$, $\text{NO}$, surface, and $\text{VOC}$ loss are highlighted by varying colors. (B) Calculated steady-state ${\mathrm{O}}_3$ concentrations either present in the gas-phase (${[{\mathrm{O}}_3]}_{\text{gas-phase}}$) or consumed by a reaction when $\text{GUV}222$ was on and when $\text{GUV}222$ was off. The assumed source of ${\mathrm{O}}_3$ when $\text{GUV}222$ is off is passive flow from the hallway. $\text{GUV}222$ on indicates source contributions from the hallway plus generation from the $\text{GUV}222$ lamps.

Assuming steady-state (i.e., t → ∞), and using Equation 1, we calculated how much ${\mathrm{O}}_3$ we expect to see in the gas-phase in the restroom, where $\text{GUV}222$ is off and ${\mathrm{O}}_3$ from the hallway is the only source (i.e., ${\text{GR}}_{\mathrm{O}3}=0$ μg h⁻¹ lamp⁻¹), compared to when $\text{GUV}222$ is on (peach bars in Fig. 5B). The sum of the ${\mathrm{O}}_3$ concentrations lost to $\text{NO}$, surfaces, and $\text{VOCs}$ at steady-state (sum of green, light gray, and blue bars in Fig. 5B) when $\text{GUV}222$ is off is the background ${\mathrm{O}}_3$ reactive loss (${[{\mathrm{O}}_3]}_{\text{reactive loss},\text{GUV}222\text{OFF}}$). Similarly, because a new source of ${\mathrm{O}}_3$ is present in the restroom, when $\text{GUV}222$ is on (i.e., ${\text{GR}}_{\mathrm{O}3}=930$ μg h⁻¹ lamp⁻¹) the sum of ${\mathrm{O}}_3$ concentrations lost to $\text{NO}$, surfaces, and $\text{VOCs}$ at steady-state is the total ${\mathrm{O}}_3$ reactive loss from both background and $\text{GUV}222$-generated ${\mathrm{O}}_3$ (${[{\mathrm{O}}_3]}_{\text{reactive loss},\text{GUV}222\text{ON}}$). Consequently, the increased concentration of ${\mathrm{O}}_3$ lost to gas-phase and surface reactions from $\text{GUV}222$ compared to background ($\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV}222}$) can be expressed as,

$$\begin{gathered} \Delta {[{\mathrm{O}}_3]}_{\text{loss}.\text{GUV222}}= \\ {[{\mathrm{O}}_3]}_{\text{reactive loss}.\text{GUV222 ON}}-{[{\mathrm{O}}_3]}_{\text{reactive loss}.\text{GUV222 OFF}} \end{gathered}$$ (6)

Additionally, we can determine the relative amount of ${\mathrm{O}}_3$ reactive loss ($\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV}222,\text{sink}}$) to each reactive sink (reactions with $\text{NO}$, $\text{VOCs}$, and surfaces), ${\mathtt{f}}_{\mathrm{O}3,\text{reactive}\text{sink}}$, via Equations 7 and 8,

$${\mathrm{f}}_{\mathrm{O}3,\text{reactive sink}}=\frac{{\mathrm{k}}_{\text{loss},\text{reactive sink}}}{{\mathrm{k}}_{\mathrm{O}3\text{loss}}}$$ (7)

$$\Delta {[{\mathrm{O}}_3]}_{\text{loss}.\text{GUV222}.\text{sink}}=\Delta {[{\mathrm{O}}_3]}_{\text{loss}.\text{GUV222}}\cdot {\mathrm{f}}_{\mathrm{O}3.\text{reactive sink}}$$ (8)

For both weekends the amount of ${\mathrm{O}}_3$ measured in the gas-phase is only a fraction of the ${\mathrm{O}}_3$ that is lost to $\text{VOC}$ and surface reactions. The amount of ${\mathrm{O}}_3$ reactive loss increases by almost a factor of three for WK1 and almost a factor of two for Wk2 compared to the amount lost under background conditions. This indicates that the $\text{GUV}222$ is enhancing indoor chemistry by at least a factor of two for this space. The extent of this enhanced chemistry will likely be different for different indoor spaces as it is dependent upon ventilation rates, species and concentrations of precursor $\text{VOCs}$, and surface reactivity.

In the following sections we use $\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV}222}$ to quantify yields of byproducts from $\text{GUV}222$-generated ${\mathrm{O}}_3$. We do not expect aerosol or $\text{VOC}$ byproducts to be formed from $\text{NO}$ reactive ${\mathrm{O}}_3$ loss and thus we only use the fraction of $\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV}222}$ that can be attributed to gas-phase and surface loss of ${\mathrm{O}}_3$.

Generation of $\mathbf{VOC}$ ${\mathbf{O}}_3$ oxidation byproducts from $\mathrm{GUV}222$

We quantified 34 $\text{VOC}$ oxidation byproducts (${\text{VOC}}_{\mathrm{O}3\text{Ox}}$) that increased in concentration when $\text{GUV}222$ was on during both weekends (Fig. 6A). We categorized the 34 ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ into lumped species of aldehydes, terpenoid oxidation products, and other oxidized $\text{VOCs}$. Several saturated aldehydes that are commonly reported in the literature to be produced from oxidation of monoterpenes by ${\mathrm{O}}_3$ were generated while $\text{GUV}222$ was on. These include formaldehyde, acetaldehyde, and hexanal. We categorized a suite of structurally ambiguous ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$, many of which were composed of seven to ten carbon atoms, as likely terpenoid oxidation products. Previous reports of ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ production from the ${\mathrm{O}}_3$ oxidation of ocimene^{42, 43} and other terpenes⁴⁴ help inform our categorization of ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ to terpenoid oxidation products (example mechanism shown in Fig. S13). ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ that we lumped into the other oxidized $\text{VOC}$ category have been reported to originate from both gas-phase oxidation of $\text{VOCs}$ as well as surface oxidation processes. These include acetone, acetic acid, hydroxyacetone, and formic acid, among others (Table S1).

Fig. 6. ${\mathsf{VOC}}_{\mathsf{O}3\mathsf{Ox}}$ concentrations, generation rates, and potential sources from $\mathrm{GUV}222$-generated ${\mathsf{O}}_3$.

(A) Average time series of ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ produced from $\text{GUV}222$ during Wk1 and Wk2. (B) Average and standard deviation of ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ generation rates from $\text{GUV}222$ (${\text{GR}}_{\text{VOC O3 Ox}.,\text{GUV}222}$) for Wk1 and Wk2. (C) ${\text{GR}}_{\text{VOC O3 Ox}.,\text{GUV}222}$ normalized to $\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV}222}$ with $\text{VOCs}$ (blue), surfaces (gray), and $\text{VOCs}$ + surfaces (black). Error bars represent the standard deviation propagated from the ${\text{GR}}_{\text{VOC O3 Ox}.,\text{GUV}222}$ measurement.

${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ are produced from background ${\mathrm{O}}_3$ chemistry as well as $\text{GUV}222$-generated ${\mathrm{O}}_3$ chemistry. To quantify how much ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ is attributable to the increase in ${\mathrm{O}}_3$ chemistry from $\text{GUV}222$, above background, we calculate ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ generation rates (${\text{GR}}_{\text{VOC}\mathrm{O}3\text{Ox}}$). We assume negligible losses from reactions and partitioning (i.e., ${\mathrm{k}}_{\text{loss}}=\text{ACR}$) and thus the ${\text{GR}}_{\text{VOC}\mathrm{O}3\text{Ox}}$ reported here may represent a lower bound and is specific to this space. Additionally, our reported values may be most biased to molecules with higher volatility that will not appreciably participate in surface portioning or reactive chemistry. We use the steady-state solution to Equation 1 shown below in Equation 9 to calculate (${\text{GR}}_{\text{VOC}\mathrm{O}3\text{Ox}}$),

$${[{\text{VOC}}_{\mathrm{O}3\text{Ox}.}]}_{\text{restroom}}=\frac{{\text{GR}}_{\text{VOC}\mathrm{O}3\text{Ox}.}+\text{ACR}\cdot {[{\text{VOC}}_{\mathrm{O}3\text{Ox}.}]}_{\text{hallway}}}{\text{ACR}}$$ (9)

where ${\text{GR}}_{\text{VOC}\mathrm{O}3\text{Ox}}$ is expressed as a total generation rate (i.e., sum of three lamps) in ppb h⁻¹. We determined ${\text{GR}}_{\text{VOC}\mathrm{O}3\text{Ox}}$ by minimizing the difference between measured and calculated ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ concentrations in the restroom.

We define a ${\text{GR}}_{\text{VOC}\mathrm{O}3\text{Ox}}$ from $\text{GUV}222$ chemistry (${\text{GR}}_{\text{VOC}\mathrm{O}3\text{Ox}.,\text{GUV}222}$) by subtracting a background source rate ($\text{GUV}222$ OFF) from the generation rate measured by $\text{GUV}222$ was on.

$${\text{GR}}_{\text{VOC}\mathrm{O}3\text{Ox}.\text{GUV222}}={\text{GR}}_{\text{VOC}\mathrm{O}3\text{Ox}.\text{GUV222 ON}}-{\text{GR}}_{\text{VOC}\mathrm{O}3\text{Ox}.\text{GUV222 OFF}}$$ (10)

We use the average hallway and restroom ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ concentrations over a 30-minute period prior to when $\text{GUV}222$ was on for the “$\text{GUV}222$ OFF” calculation and over a 30-minute period two hours after the lamps were on for the “$\text{GUV}222$ ON” calculation.

Average ${\text{GR}}_{\text{VOC}\mathrm{O}3\text{Ox}.,\text{GUV}222}$ was slightly higher in Wk1 (9.1 ppb h⁻¹ ± 2.7 ppb h⁻¹) compared to Wk2 (5.8 ppb h⁻¹ ± 3.0 ppb h⁻¹). However, the variability in average ${\text{GR}}_{\text{VOC}\mathrm{O}3\text{Ox}.,\text{GUV}222}$ was approximately 30 % for Wk1 and > 50 % for Wk2 and demonstrates the challenge of quantifying ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ production from $\text{GUV}222$ in this real environment. Two ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ that comprised approximately half of the total ${\text{GR}}_{\text{VOC}\mathrm{O}3\text{Ox}.,\text{GUV}222}$ during both weekends were acetone and acetic acid. The ${\text{GR}}_{\text{VOC}\mathrm{O}3\text{Ox}.,\text{GUV}222}$ for terpenoid oxidation products decreased from 0.9 ppb⁻¹ in Wk1 to 0.2 ppb h⁻¹ in Wk2 likely due to a decreased gas-phase terpenoid oxidation source. Aldehyde GRs stayed nearly constant (≈ 2 ppb h⁻¹) between the two weekends potentially pointing to differing contributions from both gas-phase and surface reactions of ${\mathrm{O}}_3$.

We investigate the potential sources of ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ by normalizing the ${\text{GR}}_{\text{VOC}\mathrm{O}3\text{Ox}.,\text{GUV}222}$ by $\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV}222}$ from $\text{VOCs}$, surfaces, and the sum of $\text{VOCs}$ and surfaces (Fig. 6C). This normalized generation rate is a way of investigating the potential ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ sources in the restroom and is not generalizable to other spaces. Going from Wk1 to Wk2 the ${\text{GR}}_{\text{VOC}\mathrm{O}3\text{Ox}.,\text{GUV}222}$ normalized by $\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV}222,\text{VOC}}$ increases from approximately 1 ppb ppb⁻¹ to 14 ppb ppb⁻¹ suggesting that, in order for the source of ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ to be from ${\mathrm{O}}_3$ reactions with $\text{VOCs}$, the $\text{VOC}$ source would have to produce 14 times more ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ in Wk2 compared to Wk1. Thus, we do not think ${\mathrm{O}}_3$ reactions with gas-phase $\text{VOCs}$ are the only source of ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ for both weekends. The consistency of the ${\text{GR}}_{\text{VOC}\mathrm{O}3\text{Ox}.,\text{GUV}222}$ normalized to $\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV}222,\text{surface}}$ suggests that surface reactions may be a stronger source of ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ when $\text{GUV}222$ is operating. However, there was enough variability in measured ${\text{GR}}_{\text{VOC}\mathrm{O}3\text{Ox}.,\text{GUV}222}$ such that the source of ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ could also reasonably be explained from a combination of surface and gas-phase ${\mathrm{O}}_3$ reactive loss.

${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ can be formed from both reactions of ${\mathrm{O}}_3$ on surfaces and in the gas-phase. In particular, studies have noted the reactions of ${\mathrm{O}}_3$ on surfaces soiled with human skin flakes or oils to be productive sources of ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$.^{45, 46} The absence of key ${\mathrm{O}}_3$ + skin oil oxidation products^{46, 47} from our measurements, like 6-methyl-5-hepten-2-one (6-MHO), geranyl acetone, and decanal, suggest that surface reactions of skin oils may not be a primary source of ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ byproducts in the restroom. Although the gas-phase ${\mathrm{O}}_3$ reactivity was impacted by varying urinal screen emissions between the two weekends, restroom use and surface cleaning were consistent in the days prior to each weekend. Hence, surface reactions are more likely than gas-phase reactions to be consistent between the weekends.

Typically, in the absence of a strong reactive $\text{VOC}$ source (e.g., human occupants, air fresheners, cleaning products, etc.), reactions of ${\mathrm{O}}_3$ with $\text{VOCs}$ in the gas-phase are not thought to be a major source of ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ in many indoor spaces.²² However, the correlation in reduction of ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ generation rates with the reduction in gas-phase precursor chemicals indicates this restroom is a unique instance where gas-phase production of ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ from $\text{GUV}222$-generated ${\mathrm{O}}_3$ is likely important. However, our data also suggest that partitioning of gas-phase $\text{VOC}$ precursors to surfaces and/or direct application of terpenoid cleaning products to surfaces may also serve as a persistent source of ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ from $\text{GUV}222$-generated ${\mathrm{O}}_3$. In fact, previous work has shown terpenoids to exhibit increased reactivity on surfaces, compared to gas-phase reactions^{48, 49} and also these reactions can generate ultrafine particles⁵⁰.

Ultrafine particle formation from $\mathrm{GUV}222$

The ${\mathrm{O}}_3$ produced from $\text{GUV}222$ generates ultrafine particles (i.e., particle diameter < 100 nm) from $\text{VOC}$ oxidation during both weekends (Fig. 7A). New particle formation is the process where particles nucleate from condensable vapors then grow in size via condensation and coagulation. New particle formation is the process that drives the initial increase in particle number concentrations (${\mathrm{N}}_{\mathrm{p}}$), after $\text{GUV}222$ is on, in both Wk1 and Wk2. A maximum in ${\mathrm{N}}_{\mathrm{p}}$ (Max $\Delta {\mathrm{N}}_{\mathrm{p}}$) is reached after approximately 30 minutes since $\text{GUV}222$ lamps are turned on. After that, coagulation of the newly formed particles leads to a decrease in Np. Without active mixing (i.e., mixing fan off) Np reach steady-state after approximately two hours. During steady-state, ${\mathrm{N}}_{\mathrm{p}}$ are maintained through the formation of new particles that grow to maximum sizes which are limited by the rates of coagulation and condensation of $\text{VOC}$ oxidation products to existing particles. Use of a fan affected Np by increasing the coagulation and deposition rates as indicated by the rapid decrease in Np, shown in Fig. 3C and 4C, immediately after the mixing fan is turned on. Additionally, we find in Wk2 with use of a mixing fan new particle formation is suppressed possibly from increased particle mixing that scavenges condensable vapors before they can achieve supersaturation and nucleate and/or enhanced deposition of vapors to surfaces.

Fig. 7. ${\mathsf{N}}_{\mathsf{p}}$ production from $\mathrm{GVU}222$, average Max $\Delta {\mathsf{N}}_{\mathsf{p}}$, and background versus $\mathrm{GUV}2222$-produced respired particle deposition.

(A) ${\mathrm{N}}_{\mathrm{p}}$ concentrations (top) for Wk1 and Wk2 and the corresponding size distributions (bottom, color gradients represent the log-normal ${\mathrm{N}}_{\mathrm{p}}$ distribution (dNdlogDp) and are different for Wk1 and Wk2). Time when $\text{GUV}222$ was on is indicated in the image plots by a white box. Max $\Delta {\mathrm{N}}_{\mathrm{p}}$ is determined from the maximum measured ${\mathrm{N}}_{\mathrm{p}}$ when $\text{GUV}222$ is on minus the ${\mathrm{N}}_{\mathrm{p}}$ just prior to the on cycle. (B) Average Max $\Delta {\mathrm{N}}_{\mathrm{p}}$ measured for both weekends. (C) Total number of deposited respired particles, assuming five minutes of breathing in the restroom, while ${\mathrm{N}}_{\mathrm{p}}$ was at the max fractioned by deposition zone in the respiratory tract including the head, pulmonary, and tracheobronchial zones. The amount of deposited particles when $\text{GUV}222$ is $\text{ON}$ represents the increase in deposition during $\text{GUV}222$ operation and includes background contributions (gray).

In the restroom new particle formation is driven by the formation of condensable gas-phase $\text{VOCs}$ (${\text{VOC}}_{\text{cond},\mathrm{g}}$) resulting from ${\mathrm{O}}_3$ oxidation of precursor $\text{VOCs}$, that will form molecular clusters which can then spontaneously condense at supersaturation^51-53. Increases in ${\mathrm{N}}_{\mathrm{p}}$ occur more quickly (approximately 20 minutes) after $\text{GUV}222$ is turned on in Wk1 compared to Wk2 (approximately 45 minutes after $\text{GUV}222$ is turned on). We hypothesize that the longer delay in ${\mathrm{N}}_{\mathrm{p}}$ production in Wk2 is because less ${\text{VOC}}_{\text{cond},\mathrm{g}}$ is being formed as a result of less $\text{VOCs}$ reacting with ${\mathrm{O}}_3$ which means it takes longer for ${\text{VOC}}_{\text{cond},\mathrm{g}}$ to achieve supersaturation in Wk2 and nucleate particles. On average, higher ${\mathrm{N}}_{\mathrm{p}}$ were generated at the beginning of the new particle formation events in Wk1 (≈ 11300 cm⁻³) compared to Wk2 (≈ 2600 cm⁻³) (Fig. 7B).

The ${\mathrm{M}}_{\mathrm{p}}$ we measure in this study represent a subset (particle diameter between 10 nm and 420 nm) of particulate matter with a diameter of 2.5 μm or less (PM_2.5). PM_2.5 concentrations are the most-commonly used measurement to understand aerosol health effects⁵⁴, but limited evidence also suggests exposure to ultrafine ${\mathrm{N}}_{\mathrm{p}}$ may be associated with aerosol-related health concerns^55-57. We put ${\mathrm{N}}_{\mathrm{p}}$ generation measured in this study into a potential exposure context by estimating the number of ultrafine particles inhaled and deposited in the respiratory tract from an upright person breathing for five minutes in the restroom while ${\mathrm{N}}_{\mathrm{p}}$ are at a maximum (Fig. 7C). Using inhaled particle deposition efficiencies, calculated from a multiple-path particle dosimetry model (Fig. S14), we also fractionate the deposition of inhaled particles along the head, pulmonary, and tracheobronchial regions of the respiratory tract. When ${\mathrm{N}}_{\mathrm{p}}$ production from $\text{GUV}222$ is at a maximum during Wk1 the number of particles deposited in the respiratory tract increases by a factor of three compared to when $\text{GUV}222$ is off. In contrast to Wk1, almost an order of magnitude more respired particles generated by $\text{GUV}222$ are deposited compared to background levels in Wk2.

The total number of deposited particles generated when $\text{GUV}222$ is on in Wk2 is nearly identical to the deposited particles from background aerosol in Wk1 when there was no $\text{GUV}222$, and the $\text{ACR}$ was lower with precursor concentrations higher. This result demonstrates a case where increasing ventilation to the restroom improved the background air quality, but turning on $\text{GUV}222$ increased ${\mathrm{N}}_{\mathrm{p}}$ and negated the air quality improvement created from increased ventilation. Additionally, this result demonstrates that ${\mathrm{O}}_3$ that originates from $\text{GUV}222$ sources can produce similar impacts on indoor air quality as ${\mathrm{O}}_3$ from non-$\text{GUV}222$ sources. New particle formation occurring at the higher $\text{ACR}$ and lower precursor concentrations during Wk2 $\text{GUV}222$ operation suggests that increasing the ventilation rate (at least up to ≈ 2 h⁻¹) in indoor spaces equipped with $\text{GUV}222$ may not eliminate exposure to byproducts.

Aerosol mass formation dynamics from $\mathrm{GUV}222$-initiated ${\mathbf{O}}_3$ chemistry

Informed by our measurements of $\text{GUV}222$-initiated increases in ${\mathrm{M}}_{\mathrm{p}}$ in Wk1, we propose a framework, based on ${\mathrm{O}}_3$ loss to reactions with $\text{VOCs}$ and equilibrium partitioning, for estimating a range of potential ${\mathrm{M}}_{\mathrm{p}}$ production from $\text{GUV}222$ in the restroom. ${\mathrm{M}}_{\mathrm{p}}$ production from $\text{GUV}222$ ($\Delta {\mathrm{M}}_{\mathrm{p}}$) can be predicted using Equation 11,

$$\Delta {\mathrm{M}}_{\mathrm{p}}=\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV222},\text{VOC}}\cdot {\mathrm{Y}}_{\text{Mp}}$$ (11)

where $\text{GUV}222$-generated ${\mathrm{O}}_3$ will react with gas-phase $\text{VOC}$ precursors ($\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV}222,\text{VOC}}$) and produce a characteristic yield of ${\mathrm{M}}_{\mathrm{p}}$ (${\mathrm{Y}}_{\text{Mp}}$).

We assume gas-phase $\text{VOC}$ oxidation by ${\mathrm{O}}_3$ (including any subsequent oxidation from OH) is the chemical mechanism for ${\mathrm{M}}_{\mathrm{p}}$ formation from $\text{GUV}222$ chemistry:

$${\mathrm{O}}_3+\text{VOC}\stackrel{\gamma }{\to }{\text{VOC}}_{\text{cond},\mathrm{g}}\stackrel{{\mathrm{M}}_{\mathrm{p}}}{\rightleftharpoons }{\text{VOC}}_{\text{cond},\mathrm{p}}$$ (R3)

where ${\mathrm{O}}_3$ will react with a $\text{VOC}$ producing a stoichiometric mass yield, $\gamma$, of ${\text{VOC}}_{\text{cond},\mathrm{g}}$ that can then condense (${\text{VOC}}_{\text{cond},\mathrm{p}}$) to ${\mathrm{M}}_{\mathrm{p}}$ that is present in the restroom prior to $\text{GUV}222$ operation. The value of $\gamma$ is dependent on the identity of the precursor $\text{VOC}$, but here $\gamma$ represents the effective yield of ${\text{VOC}}_{\text{cond},\mathrm{g}}$ from the lumped reactive $\text{VOCs}$ present in the restroom. We lump both low-volatility and semi-volatile $\text{VOCs}$ into a single product, ${\text{VOC}}_{\text{cond},\mathrm{g}}$, that will condense to aerosol via equilibrium partitioning⁵⁸.

Only the fraction of ${\text{VOC}}_{\text{cond},\mathrm{g}}$ that condenses to aerosol (${\mathtt{f}}_{\text{cond},\mathrm{p}}$), as opposed to being lost to air change (${\mathrm{k}}_{\text{ACR}}$) or condensation to walls (${\mathrm{k}}_{\text{cond},\mathrm{w}}$), will form ${\mathrm{M}}_{\mathrm{p}}$.

$${\mathrm{f}}_{\text{cond},\mathrm{p}}=\frac{{\mathrm{k}}_{\text{cond},\mathrm{p}}}{{\mathrm{k}}_{\text{cond},\mathrm{p}}+{\mathrm{k}}_{\text{cond},\mathrm{w}}+\text{ACR}}$$ (12)

We calculate the condensation rate constant for ${\text{VOC}}_{\text{cond},\mathrm{g}}$ to aerosol (${\mathrm{k}}_{\text{cond},\mathrm{p}}$) from the aerosol condensation sink (the integral in Equation 13) to determine how much ${\text{VOC}}_{\text{cond},\mathrm{g}}$ will condense to aerosol in the restroom (as opposed to being lost to air change or condensing to walls)⁵⁹,

$${\mathrm{k}}_{\text{cond},\mathrm{p}}=4\pi {\mathrm{D}}_{\mathrm{g}}\int \mathrm{r}\beta (\mathrm{r}){\mathrm{N}}_{\mathrm{p}}(\mathrm{r})\text{dr}$$ (13)

where ${\mathrm{D}}_{\mathrm{g}}$ is the gas diffusion coefficient of a terpenoid oxidation product with a molecular weight of 200 g mol⁻¹ (${\mathrm{D}}_{\mathrm{g}}=7.0$ x 10⁻⁶ m² s⁻¹)⁶⁰, $\mathrm{r}$ is the particle radius (m), and $\beta$ is the Fuchs–Sutugin correction for gas diffusion to a particle surface in the transition regime. The integral is determined for the ${\mathrm{N}}_{\mathrm{p}}$ size distribution from 10 nm to 420 nm discretized across 100 size bins. We used the ${\mathrm{N}}_{\mathrm{p}}$ size distribution measured at steady-state for this calculation. A discussion of the ${\text{VOC}}_{\text{cond},\mathrm{g}}$ wall loss calculation for ${\mathrm{k}}_{\text{cond},\mathrm{w}}$ is included in the Supplementary Materials (Section S9).

We use a single-product organic aerosol equilibrium partitioning model⁶¹ that accounts for the competitive loss of condensable vapors to air change or walls compared to condensation, to determine the aerosol mass yield (${\mathrm{Y}}_{\text{Mp}}$) as a function of ${\mathtt{f}}_{\text{cond},\mathrm{p}}$ and pre-existing aerosol mass concentrations (${\mathrm{M}}_{\mathrm{p}}$);

$${\mathrm{Y}}_{\text{Mp}}=\gamma \left(\frac{{\mathrm{K}}_{\mathrm{p}}\cdot {\mathrm{M}}_{\mathrm{p}}\cdot {\mathrm{f}}_{\text{cond},\mathrm{p}}}{1+{\mathrm{K}}_{\mathrm{p}}\cdot {\mathrm{M}}_{\mathrm{p}}\cdot {\mathrm{f}}_{\text{cond},\mathrm{p}}}\right)$$ (14)

where ${\mathrm{K}}_{\mathrm{p}}$ is a partitioning coefficient representing ${\text{VOC}}_{\text{cond},\mathrm{g}}$ equilibrium between the gas-phase and absorbing aerosol mass, and $\gamma$ is the mass yield of ${\text{VOC}}_{\text{cond},\mathrm{g}}$ from the $\text{VOC}$ precursors in the restroom. $\gamma$ and ${\mathrm{K}}_{\mathrm{p}}$ are determined from the fitting of ${\mathrm{Y}}_{\text{Mp}}$ to ${\mathtt{f}}_{\text{cond},\mathrm{p}}$ from Wk1 data using Equation 14. Although it is unlikely true equilibrium of ${\text{VOC}}_{\text{cond},\mathrm{g}}$ with ${\mathrm{M}}_{\mathrm{p}}$ was achieved while $\text{GUV}222$ was on, the dependence of ${\mathrm{M}}_{\mathrm{p}}$ production on background aerosol mass concentrations suggests equilibrium partitioning is an appropriate first-approximation for understanding the potential for ${\mathrm{M}}_{\mathrm{p}}$ production from $\text{GUV}222$ chemistry in the restroom.

We calculate ${\mathrm{Y}}_{\text{Mp}}$ measured in Wk1 from the increase in ${\mathrm{M}}_{\mathrm{p}}$ ($\Delta {\mathrm{M}}_{\mathrm{p}}$), while $\text{GUV}222$ is on, per ppb of ${\mathrm{O}}_3$ generated from $\text{GUV}222$ that is lost to reactions with $\text{VOCs}$ ($\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV}222,\text{VOC}}$). ${\mathrm{Y}}_{\text{Mp}}$ in the restroom can be expressed following Equation 15^{58, 62},

$${\mathrm{Y}}_{\text{Mp}}=\frac{\Delta {\mathrm{M}}_{\mathrm{p}}}{\Delta \text{VOC}}=\frac{\Delta \text{Mp}}{\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV222},\text{VOC}}}$$ (15)

where $\Delta {\mathrm{M}}_{\mathrm{p}}$ is the ${\mathrm{M}}_{\mathrm{p}}$ (μg m⁻³) produced at steady-state while $\text{GUV}222$ is on and $\Delta \text{VOC}$ is the concentration of the reactive $\text{VOCs}$ (μg m⁻³) consumed.

A higher amount of ${\mathrm{O}}_3$ reacting with $\text{VOCs}$ in Wk1 (Fig. 8A) created more ${\text{VOC}}_{\text{cond},\mathrm{g}}$ than in Wk2. Wk1 ${\text{VOC}}_{\text{cond},\mathrm{g}}$ condensed to aerosol to a greater extent than Wk2, in part, because of a higher background condensation sink (i.e., a higher ${\mathtt{f}}_{\text{cond},\mathrm{p}}$, Fig. 8B). In Wk1, on average, approximately 10 % of ${\text{VOC}}_{\text{cond},\mathrm{g}}$ generated by $\text{GUV}222$ were condensing to aerosol whereas in Wk2 only 2 % condensed to aerosol. We expect that reactions of ${\mathrm{O}}_3$ that entered the restroom from the hallway were the source of background aerosol number concentrations that served as the condensation sink for $\text{GUV}222$-generated ${\text{VOC}}_{\text{cond},\mathrm{g}}$. Because of decreasing emissions of reactive $\text{VOCs}$ from the urinal screen in Wk2, background ${\mathrm{O}}_3$ reactions resulted in lower background aerosol number and mass concentrations compared to Wk1. However, varying levels of aerosol entering the restroom from the hallway could also impact the condensation sink. Ultimately, these lower aerosol concentrations lead to a diminished ${\mathtt{f}}_{\text{cond},\mathrm{p}}$ in Wk2 compared to Wk1.

Fig. 8. Prediction of $\Delta {\mathsf{M}}_{\mathsf{p}}$ from $\mathrm{GUV}222$ chemistry as a function of $\Delta {[{\mathsf{O}}_3]}_{\mathsf{loss},\mathrm{GUV}222,\mathsf{VOC}}$, background aerosol concentrations (${\mathsf{M}}_{\mathsf{p}}$), and ${\mathsf{f}}_{\mathsf{cond},\mathsf{p}}$.

(A) Average and standard deviation of ${\mathrm{O}}_3$ generated from $\text{GUV}222$ that is lost to $\text{VOCs}$ ($\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV}222,\text{VOC}}$) at steady-state for Wk1 and Wk2. (B) Fractional contributions of the ${\text{VOC}}_{\text{cond},\mathrm{g}}$ losses ($\text{ACR}$, condensation to surfaces, and condensation to aerosol) to calculated ${\text{VOC}}_{\text{cond},\mathrm{g}}$ total loss. (C) Determination of ${\mathrm{M}}_{\mathrm{p}}$ yield (${\mathrm{Y}}_{\text{Mp}}$) as a function of ${\mathtt{f}}_{\text{cond},\mathrm{p}}$ from fitting a partitioning model (red) to ${\mathrm{Y}}_{\text{Mp}}$ measurements in Wk1. The stoichiometric mass yield, $\gamma$, is 0.34 and ${\mathrm{K}}_{\mathrm{p}}$ is 8.77 m³ μg⁻¹ of ${\mathrm{M}}_{\mathrm{p}}$ (D) This graph is generated by using the $\gamma$ and ${\mathrm{K}}_{\mathrm{p}}$ values determined from the fitting of Wk1 data in panel (C) and solving Equation 14 for various ${\mathtt{f}}_{\text{cond},\mathrm{p}}\cdot {\mathrm{M}}_{\mathrm{p}}$ to determine ${\mathrm{Y}}_{\text{Mp}}$ and then using the calculated ${\mathrm{Y}}_{\text{Mp}}$ value and various $\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV}222\text{VOC}}$ to solve Equation 11 for $\Delta {\mathrm{M}}_{\mathrm{p}}$. Dashed lines show the combinations of $\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV}222,\text{VOC}}$ and ${\mathtt{f}}_{\text{cond},\mathrm{p}}\cdot {\mathrm{M}}_{\mathrm{p}}$ that produce a given concentration of $\Delta {\mathrm{M}}_{\mathrm{p}}$. The red markers show the average and standard deviation of $\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV}222,\text{VOC}}$ and ${\mathtt{f}}_{\text{cond},\mathrm{p}}\cdot {\mathrm{M}}_{\mathrm{p}}$ values for Wk1 and Wk2.

We propose the relationship between $\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV}222,\text{VOC}}$ and the product of ${\mathtt{f}}_{\text{cond}}$ and ${\mathrm{M}}_{\mathrm{p}}$ (${\mathtt{f}}_{\text{cond},\mathrm{p}}\cdot {\mathrm{M}}_{\mathrm{p}}$) is a more useful predictor of potential ${\mathrm{M}}_{\mathrm{p}}$ formation from $\text{GUV}222$ chemistry than the $\text{ACR}$. The role of the $\text{ACR}$ in potential aerosol byproduct formation is dynamic as the $\text{ACR}$ simultaneously regulates steady-state concentrations of reactive $\text{VOC}$ precursors, timescales for gas-to-particle partitioning, and introduction of infiltrated ${\mathrm{O}}_3$. For instance, increasing the $\text{ACR}$ can decrease steady-state concentrations of reactive $\text{VOCs}$ from the urinal screens and affect the particle concentration in the restroom from mixing with hallway air. Increasing the $\text{ACR}$ can also affect the amount of ${\mathrm{O}}_3$ reacting with surfaces and $\text{VOCs}$ by increasing the hallway ${\mathrm{O}}_3$ source supplied to the restroom. Regardless of what the $\text{ACR}$ is, the amount of ${\mathrm{O}}_3$ that reacts with $\text{VOCs}$ to produce ${\text{VOC}}_{\text{cond},\mathrm{g}}$ ($\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV}222,\text{VOC}}$) and the relative strength of the condensation sink (${\mathtt{f}}_{\text{cond},\mathrm{p}}$) will determine what concentrations of aerosol mass can be generated from $\text{GUV}222$ at steady-state.

Using our Wk1 measurements of ${\mathrm{M}}_{\mathrm{p}}$ generated from $\text{GUV}222$ ($\Delta {\mathrm{M}}_{\mathrm{p}}$) and corresponding values of $\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV}222,\text{VOC}}$ we calculated ${\mathrm{Y}}_{\text{Mp}}$. We then fit ${\mathrm{Y}}_{\text{Mp}}$ to ${\mathtt{f}}_{\text{cond},\mathrm{p}}\cdot {\mathrm{M}}_{\mathrm{p}}$ using an equilibrium partitioning model shown by Equation 14 (Fig. 8C). We then predicted $\Delta {\mathrm{M}}_{\mathrm{p}}$ in Wk2 from measurements of ${\mathtt{f}}_{\text{cond},\mathrm{p}}\cdot {\mathrm{M}}_{\mathrm{p}}$ (to predict ${\mathrm{Y}}_{\text{Mp}}$) and $\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV}222,\text{VOC}}$ using Equation 11.

The combined effect of both decreased ${\mathrm{O}}_3$ reactions ($\Delta {[{\mathrm{O}}_3]}_{\text{loss},\text{GUV}222,\text{VOC}}<1\text{ppb}$) and $\text{VOC}$ loss via condensation (${\mathtt{f}}_{\text{cond},\mathrm{p}}\cdot {\mathrm{M}}_{\mathrm{p}}<0.2$) explain the lack of measurable $\text{GUV}222$-generated ${\mathrm{M}}_{\mathrm{p}}$ in Wk2 (Fig. 8D). The predicted steady-state ${\mathrm{M}}_{\mathrm{p}}$ concentrations for Wk2 were less than 0.20 μg m⁻³ for the seven $\text{GUV}222$ on/off cycles. We estimate the minimum detectable change in ${\mathrm{M}}_{\mathrm{p}}$ concentration (measured as three times the standard deviation of background ${\mathrm{M}}_{\mathrm{p}}$ concentrations by the scanning mobility particle sizer) to be 0.15 μg m⁻³ and thus we could not confidently detect ${\mathrm{M}}_{\mathrm{p}}$ formation (largest predicted ${\mathrm{M}}_{\mathrm{p}}$ formation = 0.17 μg m⁻³) from $\text{GUV}222$ operation in Wk2. Fig. 8D shows that a relatively high amount of ${\mathrm{O}}_3$ reactivity is needed (e.g., > 2 h⁻¹ in this study, $\gamma$ and ${\mathrm{K}}_{\mathrm{p}}$ values may be different in other environments) to produce more than 3.0 μg m⁻³ at steady-state in the restroom. While ${\mathtt{f}}_{\text{cond},\mathrm{p}}$ and background ${\mathrm{M}}_{\mathrm{p}}$ are important in potentiating ${\mathrm{M}}_{\mathrm{p}}$ formation, the ${\mathrm{O}}_3$ reactivity and characteristic ${\text{VOC}}_{\text{cond},\mathrm{g}}$ yield from reactive $\text{VOC}$ precursors will limit maximum ${\mathrm{M}}_{\mathrm{p}}$ production. Our calculations indicate that in Wk2 even if an infiltration source of aerosol were to increase ${\mathrm{N}}_{\mathrm{p}}$ such that ${\mathtt{f}}_{\text{cond},\mathrm{p}}$ was similar to Wk1, less than 0.25 μg m⁻³ of ${\mathrm{M}}_{\mathrm{p}}$ would be generated from $\text{GUV}222$-initiated ${\mathrm{O}}_3$ chemistry.

While previous work has comprehensively evaluated the relationship between indoor pollutant concentrations and the air change rate⁶³, generation of ultrafine particles from chemical reactions of terpenoids⁶⁴, and organic aerosol formation from partitioning of condensable gases to aerosol⁶⁵, we find that organic aerosol formation indoors is also regulated by the dynamic loss of condensable vapors to surfaces. The framework for predicting ${\mathrm{M}}_{\mathrm{p}}$ formation from $\text{GUV}222$ chemistry we present here treats the loss of ${\text{VOC}}_{\text{cond}}$ to surfaces in a simplified way compared to other sophisticated modeling studies that have evaluated surface partitioning of semi-volatile $\text{VOCs}$ by quantifying loss rates from deposition velocities and air-octanol partitioning coefficients³⁶ as well as other approaches that determine loss rates as a function of vapor pressure⁶⁶. Future work should focus on understanding how condensable vapor loss varies between indoor environments, where the ventilation and background air quality can be different, and how that can impact aerosol formation from indoor chemistry.

Conclusions

The $\text{ACR}$ was not the most important predictor of byproduct formation from $\text{GUV}222$ in our study and may not be a predictive metric for potential byproduct formation from $\text{GUV}222$ in other indoor spaces. Because byproduct formation is driven by ${\mathrm{O}}_3$ chemistry, ${\mathrm{O}}_3$ reactive loss was a better predictor of byproduct formation than $\text{ACR}$ in the restroom. In fact, the $\text{ACR}$ contributed to less than half of total ${\mathrm{O}}_3$ removal rate for both weekends. In Wk1, when more ${\mathrm{O}}_3$ was lost to reactions with $\text{VOCs}$, compared to Wk2, we measured higher concentrations of gas-phase and aerosol byproducts from $\text{GUV}222$ operation. Thus, ${\mathrm{O}}_3$ reactive loss rates were a better indicator than the $\text{ACR}$ of the potential for byproduct-forming chemistry from $\text{GUV}222$.

When trying to predict health-impacting aerosol mass (${\mathrm{M}}_{\mathrm{p}}$) generation from $\text{GUV}222$ it is important to understand how likely it is that condensing gases produced from ${\mathrm{O}}_3$ reactions with reactive $\text{VOCs}$ (${\text{VOC}}_{\text{cond},\mathrm{g}}$) will condense to pre-existing aerosol versus being lost to air change or condensation to walls. The predictive calculations of ${\mathrm{M}}_{\mathrm{p}}$ production in Fig. 8 are specific to the ${\mathrm{O}}_3$ reactivity profile of the restroom. In other words, different precursor $\text{VOCs}$ can generate higher or lower yields of ${\text{VOC}}_{\text{cond},\mathrm{g}}$ that will condense to aerosol than what we measured in this study. Terpenoids were a major source of reactive $\text{VOCs}$ for byproduct formation from $\text{GUV}222$-initiated ${\mathrm{O}}_3$ chemistry in the restroom. In a simplified calculation incorporating secondary organic aerosol yields measured in laboratory studies^{44, 67, 68}, we estimate that > 95 % of the ${\mathrm{M}}_{\mathrm{p}}$ produced from $\text{GUV}222$ measured in Wk1 was from ${\mathrm{O}}_3$ oxidation of α-terpinene (Fig. S15). α-Terpinene is the most ${\mathrm{O}}_3$ reactive terpenoid (${\mathrm{k}}_{\mathrm{O}3+\alpha -\text{terpinene}}=8.7$ x 10⁻¹⁵ cm³ molecule⁻¹ s⁻¹) and has a high organic aerosol yield (0.5 μg m⁻³ of aerosol per 1 μg m⁻³ of reacted α-terpinene). Identifying spaces with key ${\mathrm{O}}_3$ reactive precursors that can affect indoor air quality, like α-terpinene, can help in identifying spaces where $\text{GUV}222$ installations may be problematic. Additionally, labeling products with fragrance formulations could help inform consumers who want to decrease the indoor air quality impacts of terpenoid oxidation by ${\mathrm{O}}_3$.

Our method of calculating indoor space-specific aerosol yields (${\mathrm{Y}}_{\text{Mp}}$) from $\text{GUV}222$-initiated ${\mathrm{O}}_3$ chemistry considers how much condensable gases are lost to air change and surface reactions versus condensation to aerosol. In practice we have shown that if the $\text{ACR}$, size-dependent particle concentrations, and ${\mathrm{O}}_3$ reactive loss rates can be measured then the partitioning framework presented in this study could be used to estimate potential ${\mathrm{M}}_{\mathrm{p}}$ formation from $\text{GUV}222$ in other indoor spaces. The ${\mathrm{Y}}_{\text{Mp}}$ we measured is characteristic for the reactive terpenoid precursors in the restroom but will likely be different in indoor spaces with different reactive precursors. Further work modeling and measuring the impact of vapor-wall loss on aerosol formation from chemical reactions indoors would be helpful in further understanding the potential of $\text{GUV}222$ chemistry to generate aerosol pollution in chemically diverse indoor environments.

We have demonstrated that ${\mathrm{O}}_3$ and ${\mathrm{O}}_3$ chemistry byproduct formation can occur from operation of $\text{GUV}222$ in a restroom. Because the air composition in public spaces, like museums⁶⁹, offices, recreation centers^{70, 71}, and classrooms^{72, 73}, is considerably diverse⁷⁴ other field studies of air quality impacts of $\text{GUV}222$ are warranted. Deposition of $\text{GUV}222$-generated ${\mathrm{O}}_3$ to human occupants is expected to be the most important loss process in a densely occupied indoor space⁷⁵. ${\mathrm{O}}_3$ reactions in occupied spaces have been observed to produce oxidized $\text{VOC}$ byproducts^{45, 76}, OH⁷⁷, and aerosol⁷⁸. Indoor air quality impacts from $\text{GUV}222$ are thus likely to be modulated by human occupancy. Our measurements of ${\text{VOC}}_{\mathrm{O}3\text{Ox}}$ from an unoccupied restroom only capture a subset of the $\text{VOC}$ byproducts likely produced from $\text{GUV}222$-initiated chemistry. Multiple simultaneous real-time mass spectrometry measurements have been shown to be useful in capturing the range of possible oxidation products produced in the continuum that exists between reactive $\text{VOC}$ precursor and condensed-phase organics^{73, 79, 80} and would be informative for $\text{GUV}222$ air quality studies. Measurements of enhanced surface oxidation by $\text{GUV}222$-generated ${\mathrm{O}}_3$ and direct irradiation from $222\text{nm}$ light are also warranted.

When installed in the restroom, which represented a real-world application, $\text{GUV}222$ produced ${\mathrm{O}}_3$ at concentrations significant enough to induce chemistry resulting in the formation of $\text{VOC}$ and aerosol byproducts. Research has highlighted the association of chronic exposure to low levels of PM_2.5 in outdoor air to excess mortality^{54, 81, 82} and thus potential generation of ${\mathrm{M}}_{\mathrm{p}}$, and the associated exposure risk, by $\text{GUV}222$ should be considered in an assessment of the technology to improve public health. Further research is needed to evaluate indoor air quality implications of $\text{GUV}222$ in a variety of indoor spaces to inform operational guidance. Specifically, the influences of occupancy on air chemistry initiated by $\text{GUV}222$ could affect byproduct formation in ways not captured by our study of an unoccupied restroom.