Ozone for SARS-CoV-2 inactivation on surfaces and in liquid cell culture media

This study evaluated the inactivation of SARS-CoV-2, the virus responsible for COVID-19, by ozone using virus grown in cell culture media either dried on surfaces (plastic, glass, stainless steel, copper, and coupons of ambulance seat and floor) or suspended in liquid. Treatment in liquid reduced SARS-CoV-2 at a rate of 0.92 ± 0.11 log10-reduction per ozone CT dose(mg min/L); where CT is ozone concentration times exposure time. On surface, the synergistic effect of CT and relative humidity (RH) was key to virus inactivation; the rate varied from 0.01 to 0.27 log10-reduction per ozone CT value(g min/m3) as RH varied from 17% to 70%. Depletion of ozone by competitive reactions with the medium constituents, mass transfer limiting the penetration of ozone to the bulk of the medium, and occlusion of the virus in dried matrix were postulated as potential mechanisms that reduce ozone efficacy. RH70% was found plausible since it provided the highest disinfection rate while being below the critical RH that promotes mould growth in buildings. In conclusion, through careful choice of (CT, RH), gaseous ozone is effective against SARS-CoV-2 and our results are of significance to a growing field where ozone is applied to control the spread of COVID-19.


Introduction
The Coronavirus Disease 2019  is an infectious disease caused by the spread of the Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) virus. With an average mortality rate of 2.2 death per 100 cases of COVID-19, the pandemic has caused over 3 million deaths (WHO, 2021) and resulted, throughout 2020, in a loss of 4.5% of the world gross domestic product (GDP) equivalent to almost 3.94 trillion US dollars (Szmigiera, 2021). Although vaccines are exceptionally efficient at preventing this serious disease, the selection of novel variants that are less well controlled by vaccination, and the slow rollout of vaccines in low-and-middle-income countries, mean that the virus remains a public health hazard worldwide.
SARS-CoV-2 spreads from person-to-person through inhalation of contaminated droplets and aerosols or by touching contaminated surfaces (i.e. fomites) (Mohan et al., 2021;Alimohammadi and Naderi, 2021). The virus remains viable in aerosols for up to 3 h, while on surfaces, it can remain viable up to 24 h . By changing from winter to spring/fall environmental conditions, Kwon et al. showed that SARS-CoV-2 remains infectious on surfaces for up to 21 days and 7 days respectively, while it remained infectious for up to 3-4 days under indoor conditions (Kwon et al., 2021). In another study, Riddell et al. (2020) isolated viable virus from common surfaces after 28 days at 20 • C. Thus, these studies demonstrate the remarkable persistence of SARS-CoV-2 on surfaces under different environmental conditions highlighting the potential risk of contaminated surfaces as a driver of virus spread. Controlling the transmission of SARS-CoV-2 by eliminating the virus on surfaces and in liquid droplets and aerosols in the air is an important measure to curb COVID-19 disease.
In the recent year, a variety of methods have been studied to eliminate SARS-CoV-2 from the environment, including heat sterilisation, chemical disinfection, non-thermal plasma, and ultraviolet irradiation (Kwok et al., 2021;Martins et al., 2021;Volkoff et al., 2021;Welch et al., 2021;Oral et al., 2020;Murata et al., 2021). In these studies, the virus was inoculated onto a variety of hard and porous surfaces or in liquids before being exposed to the treatment. However, airborne virus studies are rare, due to the difficulty in safely handling aerosolised virus. Generally, the chemical disinfectants were either used in gas (e.g. ozone, hydrogen peroxide vapour) or liquid (e.g. chlorine-based agents, hydrogen peroxide) forms, with each offering advantages and disadvantages (Table 1). Being a gas and a strong oxidant, which can easily be produced from oxygen, ozone is more effective as compared to other chemical disinfectants. Ozone can also easily be applied in large and small areas and decompose back to safe oxygen after treatment. Ozone is particularly lethal against viruses through peroxidation of their surface lipids and subsequent damage to the lipid envelope and proteins  (Bisag et al., 2020;Chen et al., 2020;Capelli et al., 2021) Heat treatment • Common method of disinfection in an autoclave  (Mileto et al., 2021;Goyal et al., 2014;Schinkothe et al., 2021) (Murray et al., 2008), and can also damage the capsid and genome (Kim et al., 1980;Wigginton and Kohn, 2012). Generally, the susceptibility of viruses to ozone depends on the type of virus, with enveloped viruses being more susceptible to ozone attack than non-enveloped viruses; this is due to the high reactivity of ozone with the lipid layer of the envelope (Tseng and Li, 2006;Tizaoui and Ozone, 2020). Being an enveloped virus, SARS-CoV-2 is therefore vulnerable to ozone attack through reactions with reactive oxygen species (ROSs) including molecular ozone and its decomposition products such as hydroxyl radicals and singlet oxygen (Tizaoui and Ozone, 2020;Li et al., 2016). The efficacy of ozone gas to inactivate surface or airborne viruses depends on several operating factors including the product ozone concentration times exposure time (i.e. CT value -see Supplementary Information S1), the relative humidity (RH) (Hudson et al., 2009;Dubuis et al., 2020), the chemical composition of the media carrying the virus (i.e. water, biological fluid, aerosol, or dried/wet virus-adhered surface) (Araújo et al., 2013), and the type and texture of the surface (Szpiro et al., 2020). Although ozone has recently been studied to eliminate SARS-CoV-2 (Martins et al., 2021;Volkoff et al., 2021;Murata et al., 2021;Zucker et al., 2021;Yano et al., 2020;Uppal et al., 2021), these studies have only concerned with a narrow range of ozone dose and RH, and have mostly used surrogates instead of authentic SARS-CoV-2 virus (Zucker et al., 2021;Uppal et al., 2021). In addition, the effect of the media on virus inactivation has not been studied, and the CT values reported varied widely; for example, for a virus inactivation of 93% ± 3%, the reported CT values spanned over the range 0.1-40 g.min/m 3 (Murata et al., 2021;Zucker et al., 2021;Percivalle et al., 2021;Criscuolo et al., 2021). Furthermore, the effect of RH on inactivation has not been fully clarified, and there are no clear kinetic data on ozone-mediated inactivation of the virus.
To address these knowledge gaps and support possible large-scale field implementation of ozone for the elimination of SARS-CoV-2 from surfaces and in liquid medium, this paper aims to provide novel quantitative evidence on ozone inactivation of the causative virus of COVID-19. In particular, we used authentic SARS-CoV-2 virus to evaluate the synergistic effect of both CT and RH, and developed a unified basis to quantify their joint effect on virus inactivation by ozone. We evaluated the kinetics of virus inactivation and calculated rate constants in liquid and at different RH values. The effect of virus matrix was evaluated in liquid and dried media through reaction and mass transfer studies benchmarked against a probe of known reactivity with ozone, and the findings allowed us to suggest possible routes explaining the role of virus medium in interfering with the inactivation process.

Cell lines and virus
VeroE6-ACE2 cells were generated by transducing VeroE6 cells (gifted by the University of Glasgow/MRC Centre for Virology, UK) with Lentivirus vectors expressing ACE2 followed by antibiotic selection. The England2 strain of SARS-CoV2 was provided by Public Health England. The virus was passaged at a low multiplicity of infection (MOI) of 0.01 in VeroE6 cells in Dulbecco's Modified Eagle Medium (DMEM). Virus containing supernatant was harvested at 72 h post-infection (hpi), before being stored at − 80 • C. The initial virus concentration was typically between 1 × 10 7 and 4 × 10 7 PFU/mL.

Quantification of virus
Virus infectivity was assessed by plaque assays. 10-fold serial dilutions were used to infect VeroE6 or VeroE6/ACE2/TMPRSS2 cells for 1 h, in duplicate. The cells were then overlaid with DMEM containing 2% FBS, and 1.2% Avicel® and incubated for 72 h. Following this, the overlay was removed, the cells were washed, fixed with 100% methanol and stained with 25% (v/v) methanol and 0.5% (v/v) Crystal Violet. The cell monolayer was then washed with water and plaques were counted.

Inactivation of SARS-CoV-2 by ozone
In the dried virus experiments, the tests involved drying a suspension of the virus (100 µL) onto a surface (mainly polystyrene plastic) under a gentle air stream in the biosafety cabinet for approximately 1 h. The obtained dried virus layer (approximately 1.5 cm × 1.5 cm) was then exposed to gaseous ozone inside a reactor (see 2.4 below). At the end of the experiment, the sample was reconstituted in 1 mL DMEM and infectious SARS-CoV-2 was quantified by plaque assay. The number of virus particles capable of forming plaques per unit volume (i.e. plaqueforming unit PFU/mL) were counted after ozone application and compared to a control exposed to air in the absence of ozone.
In the liquid experiments, 100 µL of DMEM virus solution was mixed with 100 µL of aqueous ozone solution or water (control) and was left to react before being titrated and plaques counted; typically, the initial ozone concentration was ~1 mg/L. In addition, liquid ozone solutions (1 mL, ~ 1 mg/L) were directly applied to dried virus samples; resuspension of a dried virus sample in 1 mL water was used as control. The suspensions were then titrated, and plaques counted.
The virus log 10 -reduction was calculated from the count of the surviving virus after exposure to ozone and the count of the control after exposure to air (or pure water) according to Eq. (1).
Where: N is the count of the surviving virus per unit volume after ozonation and N 0 is the control virus count per unit volume.

The experimental reactor for viral exposure to ozone
The reactor used for exposing the virus to ozone was made of a 3 L plastic box fitted with a fan, a gas sampling port, a manual humidifier, a temperature and humidity probe, and an ozone supply canister (see Supplementary Information S2). The ozone canister was prepared by adsorbing ozone on silica gel and stored in a freezer at − 18 • C (see below Section 2.5). In an inactivation experiment, the ozone canister was collected from the freezer and immediately placed inside the reactor with the canister lid off and reactor cover on. As the opened canister is exposed to ambient conditions inside the reactor, ozone desorbs and occupies the headspace of the reactor while the fan ensures homogeneous distribution of the gas. The ozone concentration in the reactor was measured as outlined below (2.5).

Supply of ozone and gas concentration analysis
In this study, we developed new practical methods for ozone supply and analysis to avoid the use of specialised and complicated ozone equipment (i.e. ozone generator, a source of oxygen, an ozone analyser, valves, flow meters, pipes etc.), which is not practical in a category 3 biosafety laboratory (BSL-3). The supply of ozone was made from canisters containing ozone loaded on silica gel (typical mass of silica gel ~3 g) (Tizaoui and Slater, 2003). Briefly, the preparation of the canisters was made as follows: a 0.5 kg of silica gel (Sigma Aldrich, UK) of particle size of about 2-3 mm were packed in a glass column with gas inlet and outlet ports. Ozone was produced from pure oxygen (1 L/min) using an electrical ozone generator (BMT 803 N, BMT Messtechnik, Germany) and its gas concentration was measured with an ozone gas analyser (BMT964, BMT Messtechnik, Germany). The ozone adsorption was made at 0 • C and ozone gas concentration was typically 40 or 80 g/m 3 NTP (NTP: normal temperature and pressure). The ozone-loaded silica gel was then, carefully and rapidly, transferred into 8 mL glass vials (the canister) and immediately stored in a freezer at − 18 • C. The vials were transported in a car freezer (− 18 • C) from Swansea University to Cardiff University, where they were stored in a lab freezer (− 18 • C) in readiness for the virus inactivation tests. Preliminary tests showed that ozone was stable while adsorbed on silica gel and stored under freezing conditions for over a month.
The ozone concentration in the gas phase (i.e. reactor headspace) was measured using a modified indigo trisulphonate method. Briefly, a 2 mL potassium indigo trisulphonate solution (Nobbs and Tizaoui, 2014) were withdrawn in a 10 mL syringe followed by the withdrawal of 8 mL of the ozone-containing gas from the reactor. The gas/liquid mixture was then thoroughly shaken and the change in absorbance at 600 nm of the indigo solution before and after exposure to ozone was measured using a UV/Vis spectrophotometer (Cary 60, Agilent). The ozone concentration was calculated using an indigo emissivity constant of 20,000 M -1 cm -1 . The ozone exposure CT value was calculated by integration of the measured ozone concentration as function of time.

Interference of virus culture media
To establish the extent of ozone quenching by the virus culture media in liquid and dry forms, DMEM samples (without virus) were exposed to ozone in the same manner as the inactivation experiments, but with slight modifications to account for larger sample analysis requirements. UV/Vis spectra were measured before and after ozonation using a UV/ Vis spectrophotometer. A solution of indigo trisulphonate dye, used as a reference compound, was also exposed to ozone under the same conditions. Indigo dye was chosen as a reference because its reactivity with ozone is known (k O3-indigo = 9.4 ×10 7 M -1 s -1 ; 1:1 stoichiometry) (Muñoz and von Sonntag, 2000), and also because it showed comparable reactivity to DMEM (see Section 3.2). In liquid form, a 1 mL DMEM solution (or indigo trisulphonate) was mixed with 1 mL ozone solution (~1 mg/L) and left to react for 5 min. In preliminary experiments, reaction times of 5 min were largely sufficient for this study due to the extremely rapid kinetics between ozone and DMEM (or indigo). In dry form, 600 µL of DMEM solution (or indigo trisulphonate solution) was dried on a glass petri dish overnight in the dark before being exposed to ozone in the gas reactor (or air as a control) for 20 min (4 canisters high ozone levels were used). After ozonation, the samples were reconstituted with 600 µL of DI water before their UV/Vis spectra were measured. In addition, ozone diffusion from the gas phase to the surface of a dried DMEM (or indigo) was evaluated and compared to a surface containing a liquid solution. For this, 8 mL of the DMEM solution were placed in a glass petri dish (diameter 6 cm) and exposed to ozone (or air as a control) for 20 min (1 ozone canister was used). Ozone concentration in the reactor's headspace and CT values were determined as described in 2.5 above. Photographs of dried samples were also taken to determine the surface area exposed to ozone, which permitted calculation of ozone flux and dose received by the dried sample.

Statistical analysis
The number of SARS-CoV-2 (PFU/mL) were counted and, after statistical analysis, the data were transformed to base10 logarithmic values. The mean and the standard error of the mean (i.e. standard deviation divided by the square root of the number of observations) were calculated and are reported graphically. The statistical significance of the results was evaluated at a confidence level ≥ 95% (i.e. significance level p ≤ 0.05). The One-Way ANOVA analysis of variance was used to compare the statistical differences in PFUs/mL numbers between control and ozone exposed samples. Differences were considered statistically significant when the p-value was ≤ 0.05. The function anova1 of the Statistics and Machine Learning Toolbox in Matlab (Mathworks®, release R2020b) was used to perform the statistical analysis.

Ozone analysis and supply
The method used in this study to determine ozone gas concentration in the reactor headspace by indigo trisulphonate (see 2.5) was validated against a calibrated ozone analyser (B12 Transmitter Ozone 0-5/ 200 ppm, ATI, UK); the difference between the indigo method and the calibrated analyser was less than 10% (see Supplementary Information S3). Ozone concentrations, by consequent CT values, were easily controlled from low to very high by simply changing the size (or number) of canisters ( Fig. 1) or changing the ozone concentration used to prepare the canisters (data not shown). Typical changes of headspace ozone concentration as function of time are depicted in Fig. 1 at different ozone loadings in the absence (a) and in the presence (b) of virus sample. Fig. 1 shows that ozone transferred rapidly to the reactor headspace to reach a maximum concentration, C 0 , followed by a slow decline in concentration in the absence of virus sample ( Fig. 1(a)) but a rapid decline was observed when the reactor contained the virus sample ( Fig. 1(b)). The rapid increase in ozone concentration in the reactor headspace was due to rapid release of ozone (i.e. desorption) from silica gel to the surrounding gas, while the depletion, observed in the absence of virus sample, was due to ozone self-decay in the gas phase. In the presence of virus sample, besides ozone depletion by self-decay reaction, ozone also diffuses from the headspace to the surface of the virus sample where it reacts, thereby its depletion rate was high as observed in Fig. 1 (b). A model accounting for these processes was developed and is presented in Supplementary Information (S4). Ozone depletion in the gas phase was assumed to proceed according an overall first-order reaction with an overall depletion rate constant, K, which was determined by fitting the experimental data with the model. Fig. 1 shows good agreement between the model (continuous red lines) and the experimental data (symbols) (R 2 > 0.995). The values of the ozone depletion rate constant in the absence and in the presence of virus samples were 0.020 ± 0.003 and 0.17 ± 0.016 min -1 , respectively, indicating that the ozone consumption was about 8 times higher in the presence of the virus solution. The high depletion rate in the presence of the virus sample was due to ozone consumption by the virus itself but also by constituents in the culture media that carry the virus; interference of the virus media is currently overlooked by published studies but can be significant (see next Section 3.2).

Interference of virus media on inactivation
The antiviral action of chemical disinfectants can be influenced by dissolved and surface-adhered organic and inorganic substances (Araújo et al., 2013;Gram et al., 2007), though, little is known about the interference of virus media on the antiviral action of ozone. Therefore, this study evaluated the extent of such effect on ozone reactivity, and mass transfer in both liquid and dried forms using indigo trisulphonate dye as a chemical reference compound.
In liquid solutions, the UV/Vis spectrum of DMEM changed significantly following mixing with the ozone solution at a similar removal percentage (25%) to that of indigo trisulphonate (29%) (Fig. 2 and Table 2). Both reactions were extremely rapid with no further reduction in absorbance at either 560 nm (for DMEM) or at 600 nm (for indigo trisulphonate) following analysis of the first sample (t < 1 min) indicating rapid uptake of ozone by the constituents of the DMEM. This suggests that in liquid form, the virus culture medium DMEM exerts a strong ozone demand and has a reactivity similar to that of indigo trisulphonate (k = 9.4 ×10 7 M -1 s -1 ) (Muñoz and von Sonntag, 2000). However, in the dry experiments, the removals of both DMEM and indigo trisulphonate were extremely low at percentages of 2% and 8% respectively (Table 2) indicating that sample drying resulted in significant inhibition of ozone reactions. To compare the uptake of ozone by a dried surface to a liquid surface, the results indicate that the removals were much higher for a liquid surface at 50% DMEM and 53% indigo trisulphonate. These results further indicate that DMEM and indigo trisulphonate have similar reactivity with ozone. Comparing the diffusion rate of ozone transfer to a dried surface and a liquid surface, it was found that the rate of ozone diffusion to a dried surface was significantly lower by approximately 100 times to that of a liquid surface (Table 2). This suggests that viral inactivation of dried virus will require significantly high ozone doses that allow sufficient ozone transfer to reach the virus particles.
DMEM is a complex mixture containing inorganic salts (e.g. NaCl, CaCl 2 , ferric, bicarbonate, phosphate etc.), amino acids (e.g. arginine, phenylalanine, tryptophan, tyrosine, glutamine etc.), vitamins (e.g. folic acid, thiamine, etc.), and glucose etc. The indigo trisulphonate solution also contained inorganics (phosphoric acid and sodium dihydrogen phosphate) in addition to the organic indigo trisulphonate dye. Thus, upon drying, the inorganic salts in solution crystallise (see Supplementary Information S5) and occlude the organic substances, providing protection from gaseous ozone attack. This is in line with disinfection studies, which showed that crystalline-type materials such as inorganic salts provide great protection to inactivation due to occlusion of microorganisms in crystals (Alfa et al., 1998(Alfa et al., , 1996Höller et al., 1993). According to the obtained ratio of ozone mass transfer fluxes to dried and liquid surfaces, it could be suggested that for the same virucidal effect, gaseous ozone doses would be in the order of 100 times higher for inactivating dried virus than inactivating virus in liquid solutions. This suggestion was corroborated by a ratio of 60 for CT values, which is within the same order of magnitude as the ratio reported for ozone fluxes, obtained for dried and liquid media inactivation of φ6 bacteriophage by ozone Li, 2006, 2008). Ozone reacts rapidly with most of the constituents in DMEM solution, particularly the aromatic and sulphur-containing amino acids (k O3 > 10 4 M -1 s -1 ) (Pryor et al., 1984), and as suggested earlier, the rate constant can be as high as that of indigo ~9.4 × 10 7 M -1 s -1 . Accordingly, the virus matrix would also impair virus inactivation through competition reactions.
Moreover, in liquid phase disinfection with gas, ozone must first transfer from the gas phase to the liquid phase before attacking the virus particles. Using liquid DMEM in conjunction with the reference indigo trisulphonate dye solution, our investigation showed that due to the high reactivity of ozone (demonstrated earlier), mass transfer controls the overall process and the reaction occurs entirely in the mass transfer liquid film with limited or no gas reaching the bulk fluid (Charpentier and Transfer, 1981).
In summary, our results suggest that interference of organic and inorganic substances on virus inactivation could occur through three different processes including mass transfer limitation, competitive

Table 2
Ozone reactivity with liquid and dried media in the absence of virus (for gas above dried sample or above liquid solutions: ozone exposure time = 20 min, RH~50%).

Sample
Ozone in liquid sample (C = 1 mg/L) where Abs0 is the initial absorbance and Abs is the absorbance after ozonation. For liquid mixtures of DMEM or indigo trisulphonate, Abs0 was determined by mixing the stock solution of either DMEM or indigo trisulphonate with DI water (without ozone) at the same volume ratio as the ozone experiment.
reactions, and occlusion of the virus in dried media. The first two processes are particularly relevant to virus inactivation in liquid solutions while the latter process is relevant to dried samples. Despite interference of the virus medium, ozone disinfection still occurs, albeit with reduced efficiency as demonstrated next.

Ozone inactivation of SARS-CoV-2 3.3.1. Virus inactivation in liquid DMEM
Aqueous ozone solutions were used to inactivate SARS-CoV-2 either suspended in liquid solutions (DMEM) or dried on a polystyrene plastic surface. At a low liquid ozone CT value (~0.1 mg.min/L), the application of aqueous ozone to suspended virus in DMEM led to only a small inactivation in virus titres (27%, p = 0.022) (Fig. 3). Others have achieved higher virus inactivation (99%) at similar CT value (Murata et al., 2021). However, they used a larger volume of ozone solution, which would also dilute out interfering substances. To increase virus inactivation, we increased the ozone dose by adding ozone-loaded silica gel (~0.04 g) directly to 2 mL virus solution for five minutes. Although the experiment continued for 5 min, the ozone concentration in solution depleted very fast (within 30 s); the CT value was estimated at 2.2 ± 0.2 mg.min/L. At this applied CT value, the virus inactivation increased significantly to 99%, p = 0.018 (Fig. 3). Having demonstrated effects on virus in solution, we next examined effects of ozonated water on surface-dried virus solution (i.e. a DMEM solution that contained the virus was dried in the biosafety cabinet by a gentle air stream followed by either resuspending the dried virus sample in aqueous ozone solution or in DI water or DMEM control experiments). The application of ozonated water on the surface containing initially dried virus solution at approximately CT = 0.5 mg.min/L yielded virus inactivation of 71%, p = 0.010 (Fig. 3) indicating that liquid ozone was also capable of inactivating dried virus on a surface. This latter experiment was in fact equivalent to exposing the virus in an aqueous ozone solution. This is because upon exposing the dried matrix to the liquid solution and with a quick swivel, the dissolution and resuspension of the substances was very fast, resulting in a well-mixed liquid solution where all constituents (i.e. virus and medium constituents) are exposed to ozone (also see Supplementary Information S6). Considering the results of virus inactivation in liquid form, the inactivation rate constant value (commonly used in water disinfection and is defined as the ratio log 10 (N 0 /N) over CT) was approximately 0.92 ± 0.11 log 10 -reduction per CT value (Fig. 3). In drinking water disinfection, this ratio is typically 10 log 10 -reduction per CT value (Rakness, 2005), which is approximately 10 times higher than the ratio obtained in our study. This could be explained by stronger ozone demand exhibited by the virus medium (i.e. DMEM) as compared to drinking water. Thus, high virus inactivation in solutions will require high ozone doses, to overcome the demand imposed by interfering substances.
We then used a modified Chick-Watson model (Lambert and Johnston, 2000), which assumes that the virus is inactivated by ozone at a rate governed by a rate constant K 1 and at the same time, ozone is quenched at a rate governed by a rate constant k Q (see Supplementary  Information S7). The rate constants K 1 and k Q were determined by fitting of experimental data. Their values were (7.2 ± 0.98)× 10 5 M -1 s -1 and 30.3 ± 4.1 s -1 respectively (see Supplementary Information S7). The value of the virus inactivation rate constant, K 1 , is in good agreement with rate constants reported for the inactivation of other viruses with ozone, which varied in the range 10 5 to 10 6 M -1 s -1 (Wolf et al., 2018). On the other hand, the magnitude of the ozone quenching rate constant indicates that the quenching reaction is very rapid and any increased inactivation times will lead to a maximum value of log 10 -reduction, which is dependent on the initial concentration of ozone, C 0 , given by K 1 C 0 /(2.3k Q ). Accordingly, to achieve, for example, 3 and 4 log 10 -reductions in the liquid solutions, the initial ozone concentrations need to be 15 and 20 mg/L, respectively. This simple model is of particular importance with respect to the disinfection of SARS-CoV-2 by ozone in liquid solutions since, for known kinetic parameters K 1 and k Q (k Q is media dependent), it allows predictions of virus inactivation and the effect that quenching substances impart to the disinfection process.

Ozone inactivation of dried virus.
To examine the ability of gaseous ozone to inactivate dried virus, 100 µL of the virus solution was dried onto the bottom of a polystyrene plastic well, then exposed to either ozone gas or air (control), for 3 min (CT = 0.5 g.min/m 3 ). Virus was then recovered and titrated, revealing that ozone gas inactivated the virus by only 23%, p = 0.033 (~0.12 log 10 reduction) (Fig. 4). To enhance inactivation, the exposure time was increased to 5 and 20 min, consequently CT values increased to 1.0 and 4.7 g min/m 3 , respectively. The measured virus inactivation values were 30%, p = 0.022% and 55%, p = 0.015 (Fig. 4), indicating the positive effect of CT. The dried matrix and relative humidity (RH ~40% in this experiment) affect the activity of ozone, which explains this modest inactivation. These inactivation percentages are comparable to recent results for the inactivation of SARS-CoV-2 surrogates (Zucker et al., 2021). To improve virus inactivation, we increased RH to ~70% and the CT values to ~5.0 g min/m 3 and ~15.0 g.min/m 3 (this was done by activating the manual humidifier inside the reactor and by increasing the number of ozone canisters), while keeping the experimental time short (< 1 h). It was crucial to keep the experimental time short, so virus inactivation due to "natural" decay of SARS-CoV-2 (Morris et al., 2020) did not interfere with the ozone inactivation. As CT increased to ~5.0 g min/m 3 and ~15.0 g.min/m 3 , virus inactivation also  significantly increased to 95%, p = 0.0097% and 99%, p = 0.01, respectively (Fig. 5). These results highlight the importance of CT values in inactivation and correspond well to those reported by other recent studies on coronaviruses and surrogates to SARS-CoV-2 who used similar virus matrix to ours (Murata et al., 2021;Zucker et al., 2021;Clavo et al., 2020). However, significantly lower CT values were reported by Yano et al. (2020). At average RH 70%, they reported that ozone CT values of only 0.1 and 0.7 g min/m 3 inactivated SARS-CoV-2 by 97% (1.5 log 10 -reduction) and 99.95% (3.3 log 10 -reduction) respectively. However, it's notable that they purified virus by ultrafiltration and three washing steps. Differences could therefore be attributed to virus medium composition variability, which, as demonstrated in this study, interfere significantly with ozone reactions in liquid and dried phases.

Effect of humidity.
To evaluate the effect of humidity on the inactivation of surface-adhered virus, we treated dried samples of SARS-CoV-2 with ozone at different RH (17% to ~70%), while the ozone CT value was approximately 5.8 g min/m 3 . The experimental results showed that as the relative humidity increased, the virus inactivation also increased ( Fig. 6a-b), which is in agreement with literature (Tseng and Li, 2008). Mechanistic principles describing the influence of relative humidity on virus inactivation by gaseous ozone are not well developed. However, we drew upon the known effect of relative humidity on the decay of SARS-CoV-2, which follows an exponential decay model Marr et al., 2019), to suggest a minimal empirical correlation describing the effect of RH at a constant CT~5.8 g. min/m 3 (Eq. 2). The inactivation-RH plot (Fig. 6a) can be extrapolated to a zero inactivation to give a critical RH value of 13% below which ozone cannot inactivate the virus; a result which is consistent with literature (Dubuis et al., 2020). Although our experimental results highlight that an increase of RH yielded increased inactivation rates, other studies suggested that extremely high RH could reduce or increase the efficacy of ozone disinfection (Volkoff et al., 2021;Aydogan and Gurol, 2006). Thus, Eq. (2) is limited to RH <~70% only.
The increased efficacy observed for RHs higher than the critical RH value could be explained by rehydration of the desiccated layer surrounding the virus, thereby enhancing ozone diffusion and virus exposure to the disinfectant; this is in line with the fact that ozone mass transfer rate was 100 times higher to a liquid surface as opposed to dried surface (see Section 3.2). Humidity, also promotes the decomposition of ozone to highly reactive species such as • OH, H 2 O 2 , O 2 •-, and singlet oxygen, potentially increasing virus inactivation (Dubuis et al., 2020;Hudson et al., 2007). It can thus be concluded that humidity has a strong effect on disinfection of dried virus samples and its combination with ozone yields increased virus inactivation (Hudson et al., 2009;Elford and van den Ende, 1942). However, since extremely high RH could lead to either enhanced or reduced efficacy of ozone (Volkoff et al., 2021), careful choice of RH should be made to ensure high inactivation rates. For example, Ishizaki, Shinriki and Matsuyama (Ishizaki et al., 1986) reported that an RH above 50% was required to inactivate Bacillus spores, while Bayarri et al. Bayarri et al. (2021) concluded that the best viral inactivation rates with ozone are achieved at RH between 70% and 90% only.
To establish the effect of RH on the virus inactivation kinetics, we determined the inactivation rate constant, k, at different relative humidities; k was calculated by k = -log 10 (N/N 0 )/CT (Supplementary Information S8). Starting from low values and increasing RH to approximately 70%, RH accelerated the inactivation rate by approximately 25 times. However, a further increase of RH above ~70% yielded a reduction in the inactivation rate constant (Supplementary Information S8). This might be due to accelerated ozone decomposition at high RH ( Supplementary Information S9), which reduces the ozone concentration that reaches the surface where the disinfection process takes place. It could thus be concluded that humidification has the benefit of accelerating the ozone inactivation kinetics of dried SARS-CoV-2 but only up to a certain level, which was found RH= 70% in our experiments. This RH not only gives the highest inactivation rate constant, but it is also a plausible relative humidity since it is below the critical RH that promotes mould growth in buildings (>80% Viitanen et al. Viitanen et al., 2010).

Synergistic effect of CT and RH.
Our previous results show that ozone treatment of dried SARS-CoV-2 sample was mostly affected by the cooperative effect of CT and RH. To describe this synergy, we correlated the impact of CT and RH by isolines of inactivation percentages displayed in the contour plot on Fig. 7. Experimental results were complemented by model data points and Matlab (Mathworks®, release R2020b) was used for data processing and plotting. The results on Fig. 7 were determined for a temperature of 26 • C and virus dried in cell culture medium (DMEM). This condensed information can guide the selection of operational (CT, RH) to assist applications and research. Although DMEM was the medium used in this study, the conclusions of this investigation could be extrapolated to real fluids. For instance, both DMEM and saliva are made of 98-99% water and both media share a large number of organic and inorganic components (Engelen et al., 2007;Gibson and Beeley, 1994). Namely, both fluids contain sodium, potassium, calcium, bicarbonate, chloride and phosphate and share several organic molecules such as glucose, amino acids, cystine, glycine, and histidine compounds. Even though the concentrations of these substances vary, their occurrence in real biological fluids will, as demonstrated in DMEM, interfere with the virus inactivation process through competitive reactions, mass transfer limitation, and occlusion of the virus in crystallised materials (e.g. salts) upon drying. Thus, high RH (~70%) will remain an essential variable when treating dried real biological fluids with ozone while high CT doses will also remain an important requirement for high inactivation rates.

Inactivation of SARS-CoV-2 dried on different surface materials (gaseous ozone)
The effect of gaseous ozone on dried SARS-CoV-2 inactivation on different surface materials was also evaluated. Rigid nonporous (copper, stainless steel, and glass) and porous (coupons of ambulance seat and ambulance floor) surfaces were used. The surfaces were cleaned with water and acetone and dried before use. With copper, ambulance seat and ambulance floor, no viable virus could be recovered after treatment, even from the control sample. Copper is inherently antimicrobial and has been shown to inactivate SARS-CoV-2 within minutes (Bryant et al., 2021), while the porous ambulance seat and floor may have inactivated the virus or trapped the remaining virus particles, preventing their elution during virus recovery. However, ozone effectively reduced infectious virus on both glass and steel by 99% (p-values < 0.05) (Fig. 8); this result is comparable to that on polystyrene plastic at the same CT and relative humidity (Fig. 7). Thus, virus inactivation on the rigid inert surfaces was not affected by the surface type, but inactivation on the porous materials or surfaces known to show antimicrobial effect should be examined.

Conclusions
This study demonstrated that ozone is effective for the inactivation of SARS-CoV-2 in liquid cell culture media and adhered on surfaces. In liquid solution, ozone was found to attack the virus rapidly at a rate constant of about 7 × 10 5 M -1 s -1 . The surface experiments demonstrated that both CT and RH have a synergistic effect on virus inactivation, with RH 70% was suggested as a plausible working relative humidity. Ozone mass transfer experiments demonstrated that the ozone flux to a liquid surface was 100 times higher than to a dried surface, which suggests that   7. Synergistic effect of (CT, RH) on dried SARS-CoV-2 inactivation by ozone gas. SARS-CoV-2 in cell culture media was dried on polystyrene plastic surface and exposed to ozone at different CT and RH% values (T = 26 • C). rehydration of dried virus medium at high RH enhanced virus exposure to ozone, thereby its inactivation. The results of virus medium effect suggested that the composition of the medium could limit viral inactivation through depletion of ozone by competitive reactions, mass transfer limiting the penetration of ozone to the bulk of the matrix, and potential occlusion of the virus in crystalline materials. In addition, this study demonstrated that the type of surface material affected virus inactivation. Specifically, rigid inert surfaces (stainless steel, glass, and plastic) gave similar virus inactivation by ozone whilst porous materials (ambulance seat and ambulance floor) or surfaces known to show antimicrobial effect (copper) inactivated the virus even in the absence of ozone. Moreover, this study correlated the synergistic effect of CT and RH in a contour plot to guide the selection of values for these important variables in ozone disinfection research and application studies.
As demonstrated in this study, the virus medium plays an important role in inhibiting virus inactivation, which suggests that future studies should further evaluate the elimination of the virus under different media conditions. The study also raised additional fundamental questions that are worthy investigating in the future. It is important to understand how constituents of the virus matrix affect virus shielding and how they affect ozone mass transfer to the bulk of the medium. Fundamental studies on the interaction between biological fluids and surface materials and how such interaction impacts virus inactivation by ozone in the presence of humidity are also worth pursuing in the future. In addition, studies on the inactivation of aerosolised SARS-CoV-2 by ozone, which are currently lacking, are worth pursuing given the importance of the airborne transmission route.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Fig. 8. SARS-CoV-2 inactivation by ozone on stainless steel and glass (100 µL of virus solution in DMEM was dried on SS or glass then exposed to ozone at RH 81%, T = 26 • C, CT~15.0 g min/m 3 ). Virus inactivation percentages and pvalues are shown between brackets. Standard error is shown by vertical bars.