Investigation of the Effect of Aerosolized TiO2 Nanoparticles on the NO2 Concentrations under UVA Exposure

Engineered nanomaterials (ENM) are utilized and produced during many engineering processes. However, they may generate airborne nanoparticles during different phases of their life cycles. TiO2, which is known for its photocatalytic properties, especially in terms of degrading NO2, is one of the most widely produced ENM. Ambient NO2 stems mainly from vehicular and industrial combustion processes and potentially harms human health upon inhalation. This study investigated the photocatalytic reduction of NO2 due to varying concentrations of aerosolized TiO2 nanoparticles during varying levels of ultraviolet A (UVA) exposure. Additionally, we evaluated the effects of low (5%), moderate (35%), and high (70%) relative humidity. Our results demonstrated that the decrease in NO2 depended on both the TiO2 concentration and the UVA intensity. However, when the former was high (> 3.23E+06 cm–3), the contribution of the latter was negligible. On the other hand, when the latter was high, the influence of the former was minimal. Furthermore, as the UVA exposure increased in duration, its effect became dominant over that of the TiO2 concentration. Finally, we found the relative humidity to be an insignificant factor during conditions with strong UVA (6328 W s m–2).


INTRODUCTION
Engineered nanomaterials (ENM) can be released to the environment, including the atmosphere, during many instances throughout the product life cycle (Gottschalk et al., 2011). They can be released into the air, e.g., with the flue gas from an industrial facility where they are produced or processed, during mechanical treatment of a composite with embedded ENM (e.g., sanding of a coating) or during waste incineration. Once airborne, due to their small size and low inertia, ENM can easily disperse over long distances from their emission sources. Only little is known about the release of ENM to the atmospheric environment (John et al., 2017). It was estimated that the amount of ENM used in different industrial applications would rise up to 1,663,000 tons by 2020. Keller et al. (2013) reported that 0.1-1.5% of produced ENM are estimated to be released into the atmosphere during their life cycle. While airborne, ENM may interact with gaseous components of the air and affect atmospheric chemistry. Due to the generally high surface-area-to-mass ratio of ENM, even rather low mass concentrations of ENM may have a stronger effect on atmospheric chemistry than their larger counterparts.
Titanium dioxide (TiO2) is one of the most produced and used ENM worldwide (Piccinno et al., 2012). TiO2 is chemically inert in the dark, inexpensive, and non-toxic. Nano-sized TiO2 is widely used for its photocatalytic activity, whereas micron-sized TiO2 is a common white pigment. Nanosized TiO2 is mostly implemented in a solid (e.g., polymer) or liquid (e.g., sunscreen) matrix material, from where strong release into the atmosphere is rather unlikely. It was estimated in 2012 that the global annual production of TiO2 nanoparticles would be between 500 and 5500 tons, out of which between 55 and 3000 tons would be produced in the European Union (EU;Piccino et al., 2012). Mueller and Nowack (2008) models showed that the amount of TiO2 nanoparticles in the atmosphere above Switzerland is around 1.5 ng m -3 . The strongest release of TiO2 into the atmosphere can therefore be expected from the manufacturing plants, where the materials are produced. When TiO2 is exposed to UVA light with wavelengths between 320 and 390 nm, photochemical reactions can be initiated at the TiO2 surface. These reactions have the potential to change the chemical balance of the atmosphere (John et al., 2017). TiO2 has three major phases, i.e., anatase, rutile, and brookite. Anatase TiO2 generally provides the highest photocatalytic activity (Augustynski, 1993). However, adding a fraction of rutile TiO2 in an anatase sample further increases the catalytic activity (Chen et al., 2012). The material used here, i.e., P25 TiO2 (Evonik AG, Germany), is composed of anatase and rutile phases. Their ratio has been reported to be between 70:30 and 80:20 (Tobaldi et al., 2014).
Nitrogen dioxide (NO2) is a very common air pollutant, especially in urban areas. NO2 has been identified to pose a health risk to humans when inhaled, in particular on the respiratory and the cardiovascular system (Kampa and Castanas, 2008). A main source of NO2 is the transportation sector (diesel engines); around 4% of the urban population of the 28 EU member states (EU-28) was exposed to NO2 concentrations above the EU annual limit value. According to estimations, the population exposure to NO2 in the EU-28 in 2018 resulted in around 54,000 premature deaths (Air quality in Europe, 2020). Several regulations are in place to control the exposure of the general public and workers to NOx. In the EU, the annual limit value for the ambient NO2 concentration is 40 µg m -3 .
It is well known that TiO2 is photocatalytically active in presence of UV light and is capable of reducing nitrogen dioxide (Febo et al., 1995). Recently, titanium dioxide has been widely used in building material, as self-cleaning surfaces, in concrete or glass (Ohko et al., 2008;Langridge et al., 2009;Monge et al., 2010), or as photoactive paints (Devahasdin et al., 2003;Maggos et al., 2007;Laufs et al., 2010) to reduce nitrogen dioxide from the atmosphere. Sodium lamps emitting UV light can be coated with TiO2 to reduce and decompose NO2 emissions from vehicles passing through tunnels (Fujishima et al., 2008). Building exteriors coated with TiO2 are widely used around the world. Ballari et al. (2011) experimentally studied NOx conversion over a photocatalytic concrete stone containing TiO2. Their results show that NO2 conversion can increase up to 75%, depending on the UVA irradiance and relative humidity. Laufs et al. (2010) studied the decomposition of nitrogen oxides on the surface painted with the commercial photocatalytic paints in a flow tube reactor. Their results show that covered surfaces are an effective sink for NO2. The reported uptake coefficient in their work under a UVA irradiance of 57 W m -2 was in the range of 0.5 to 3E-05. Langridge et al. (2009) experimentally studied the photocatalytic conversion of NO2 on commercial self-cleaning window glass panes at different relative humidity values. Their experimental results show that the NO2 reduction increases from 50% to 70% by increasing the relative humidity from 8% to 45%. Monge et al. (2010) reported that the uptake coefficient of NO2 increased from 2.2E-09 to 1.1E-07 by changing the ratio of TiO2/SiO2 mixture film used over coated surfaces from 10% to 100%, which demonstrated the important role of titanium dioxide on the reduction of NO2. Zein and Bedjanian (2012) investigated the effects of different parameters such as NO2 concentration, UV irradiance intensity, and relative humidity on the uptake coefficient of NO2. Zein and Bedjanian (2012) used a tube reactor in which the outer tube surface was coated with TiO2 to investigate the effects of different parameters such as NO2 concentration, UV irradiance intensity, and relative humidity on the uptake coefficient of NO2. Their results showed that the initial reaction rate for relative humidity of 40% is higher than at a relative humidity of 0%. However, by increasing the reaction time, the effect of relative humidity became negligible. The effect of airborne submicron TiO2 on NO2 concentration for different ranges of relative humidity was investigated by Gustafsson et al. (2006). They reported uptake coefficients in the range of 1.2E-04 and 9.6E-04 for relative humidity between 15% and 80%, intensity of 16 W m -2 , reactive surface of 0.43 m 2 m -3 , and NO2 mixing ratio of 100 ppbv. All showed a clear effect of TiO2 on NO2 concentration. However, in the case of aerosolized nanoscale TiO2, still more investigations are needed. Especially, the effect of reactive surfaces of TiO2, UVA exposure and exposure time still need to be considered.
The aim of this work was therefore to experimentally study the effect of airborne TiO2 nanomaterial on NO2 concentrations. To do so, different concentrations of TiO2 were examined under different UV light intensities. Besides, the effect of relative humidity on the NO2 reduction was investigated by considering different TiO2 concentrations and UV light exposure. Fig. 1 shows a schematic of the experimental set-up. The experimental set-up consists of a quartz glass tube reactor with an inner diameter of 10.17 cm and length of 200 cm, coaxially installed inside a PVC tube with a diameter of 30 cm and length of 200 cm. Three sequential groups of UV lamps (λmax = 365 nm; TL-D 18W/08; Philips) are fixed on the inner side of PVC tube. Each group consists of six parallel lamps (in total 18 lamps). The lamps can be individually switched on and off to mimic different exposure times and light intensities. The eventual exposure of the aerosol to UV light is determined by the product of the exposure time and the UVA light intensity. In total, for a given flow rate 18 different but equally spaced exposures can be adjusted with this set-up, determined by the number of active UV lamps.

Description of Experimental Set-up
For producing NO2, 0.5 L min -1 filtered air was passed through a U-tube in which a permeation tube (VICI Metronics, Inc.) was placed. The temperature of the U-tube with the permeation tube was maintained constant at 30°C by placing it inside a water bath. The 0.5 L min -1 NO2 flow was mixed with 9.5 L min -1 filtered air before being introduced into the quartz glass tube. The air flow rates were controlled by two mass flow controllers (GFC17; Aalborg, Orangeburg, NY, USA). To supply air with different relative humidity, a cabin air filter test rig according to ISO/TS 11155 was utilized to provide relative humidity levels of 5%, 35%, and 70%. The concentrations of NO/NO2 at the inlet and outlet of the reactor were measured using two NOx analyzers (AC32M; Environnement S.A.). Based on a European air quality standard (Volz-Thomas et al., 2003), the 1-hour NO2 values must not exceed 120 ppbv. Therefore, a concentration near this value was chosen for the experiments. The TiO2 ENM were aerosolized from aqueous suspensions of P25 (Evonik Industries AG, Germany) using an atomizer (ATM 226; Topas GmbH), followed by a homemade coaxial, silica-gel-based diffusion dryer. Particle size distributions were measured with a Scanning Mobility Particle Sizer (SMPS; Model 3996; TSI Inc.). The SMPS consists of an 85 Kr aerosol neutralizer (Model 3077A; TSI Inc.), differential mobility analyzer (DMA; Model 3081; TSI Inc.), a water-based condensation particle counter (CPC; Model 3787; TSI Inc.), and an electrical classifier (Model 3080; TSI Inc.). Under UV light illumination, TiO2 aerosols can reduce NO2 to HONO. As HONO is detected by the NOx analyzers as NO2, a Na2CO3 denuder was used between the exit of the quartz glass tube and the detector to remove any HONO that might be generated during the experiments to reduce false-positive results. The denuder consists of a quartz glass tube with a length of 55 cm and a diameter of 0.7 cm, that was coated with a solution of: 1% Na2CO3 (w/v) + 1% glycerol (w/v) in water-methanol 1:1 solution. The solution was poured into the glass tube, rotated and moved in upward-downward direction for several times at different angles (Febo et al., 1993(Febo et al., , 1995Monge et al., 2010). The total NO2 trapped in the denuder in the absence of UV exposure was around 3%, which was similar than the value reported by Monge et al. (2010).
To estimate the total UV exposure of the aerosol inside the reactor, it is necessary to know both the residence time of the aerosol inside the reactor as well as the UV intensity. The latter was measured using a radiometer (RM-21; Opsytec Dr. Gröbel GmbH, Germany). The sensor measures the UVA intensity in a spectral range from 315 to 400 nm. The measurement was conducted along the center line of the reactor glass tube with only one UV lamp active.

UVA Exposure inside the Reactor
The experimental results show that the average UVA exposure over the length of one lamp is 12 W m -2 . The measurements were repeated by adding the number of active lamps per group and it could be shown that the exposure is additive. For the given flow rate, the actual exposure of the aerosol, i.e., the product of exposure time and UVA intensity, is therefore only determined by the number of active lamps.
Several experiments were carried out to investigate the effect of UVA intensity, exposure length and exposure time. For instance, our experiments showed that for an exposure length of 120 cm (two groups) and two sequential active UVA lamps (one per group), the reduction of NO2 is identical as with an exposure length of 60 cm (one group) and two parallel active UV lamps . These experiments were repeated for one active lamp in different groups, as well as two, three, and six active lamps in different combinations. Therefore, our results show that the conditions inside the reactor are homogeneous and that the NO2 reduction as a function of UVA exposure can be easily determined by the number of active lamps.
With the measured data, the UVA exposure (expressed in W s m -2 ) is calculated as follows: where EUVA represents the UVA exposure that is the UVA radiant energy emitted over a surface per unit of time, NL the number of active lamps, I0 the intensity of one lamp (here 12 W m -2 ) and t0 is the exposure time for a single active lamp. The exposure time is calculated as follows: where Ll represents the length of one group of lamps and v0 the mean airflow velocity: where Q represents the air flow rate and A the cross-sectional area of the quartz glass tube. With a flow rate of 10 L min -1 , a length per group of lamps of 60 cm and a reactor tube diameter of 10 cm, the residence time per group of lamps is 29.2 s and the UVA exposure per active lamp 351.6 W s m -2 . The UVA exposure in this study is between 351.6 and 6328 W s m -2 . The number of lamps and related intensity and UVA exposure are presented in Table 1.

Experimental Set-up Evaluation
The NO2 reduction in absence of TiO2 in terms of NO2 photolysis frequencies (j) was considered to evaluate the experimental set-up. Photolysis frequency was expressed as the rate of NO2 concentration loss due to the UVA exposure. The photolysis frequency is defined as: where [NO2] and t represent the NO2 concentration and exposure time, respectively. The photolysis frequencies (j) of our experimental results are presented in Fig. 2. Similar to Zein and Bedjanian (2012) results, the trend of NO2 loss was rising along with increasing the number of UVA lamps in absence of TiO2. In our experiments, j(NO2) ranged from 0.0015 to 0.011 s -1 (one to six lamps) which showed a good agreement with Zein and Bedjanian (2012) who reported j(NO2) from 0.002 to 0.012 s -1 . Bohn et al. (2005) reported j(NO2) values between 0.0001 and 0.01 s -1 using a simulation chamber as a chemical actinometer under the influence of natural sunlight and Laufs et al. (2010) reported this quantity to be below 0.008 s -1 in central Europe.
The range of photolysis frequencies in our work is comparable with the corresponding values in the atmosphere as well as previous studies which investigated the reaction between NO2 and UVA exposure over a coated surface with TiO2 films. The time series of NO and NO2 concentration variations under UVA exposure is presented in Fig. S1 (in the supplementary materials).

Particle Size Distribution
The effect of TiO2 on the NO2 concentration was investigated for different UVA exposures at a relative humidity of 5%. The number size distributions of aerosolized TiO2 particles were measured with the SMPS. The modal diameters of the distributions were 140 nm. The size distribution shown in Fig. 3 represents the average distribution for particle concentrations of 0.47, 1.03, 1.72, and 3.23E+06 cm -3 , normalized with respect to the total concentration. For each concentration, size distributions were at least measured in triplicate. (The error bars in this study represents the standard deviation of the results obtained in each experiment). By using the SMPS measurements, the reactive surface area concentration (m 2 m -3 ) (Uekawa et al., 2003) was calculated. Table 2 contains information about the number concentration and related reactive surface area. (The method to calculate the reactive surface area and mass size distribution of TiO2 were presented in details in the supplementary materials).

Uptake Coefficient
Results are presented in two different formats: Fig. 4 presents the reduction of the NO2 at the outlet of the reactor as a function of the UVA exposure for different TiO2 reactive surface area concentrations. The results are further presented in terms of the uptake coefficient (γ) in Fig. 5. The uptake coefficient reflects both physical and chemical processes and represents the rate of effective collisions between TiO2 nanoparticles and NO2 gas molecules (Bartolomei et al., 2014). The uptake coefficient as a dimensionless parameter, which considers the effect of exposure time and reactive surface effects on NO2 reduction, is an appropriate number for reporting photochemical reaction and also for comparison with previous results. The uptake coefficient is calculated according to Eq. (5): where k is the first-order reaction rate (s -1 ), v is the mean molecular speed of NO2 (m s -1 ) and As (m 2 m -3 ) is the reactive surface area concentration. The first-order reaction rate, which is used to calculate the rate of NO2, was calculated as follows: where c0 represents the NO2 concentration in the absence and ci in the presence of UVA light, each measured at the reactor outlet. Δt is the exposure time which is 29.2 s if only one group of lamps is active. If lamps from two groups are switched on, Δt is 58.4 s and with lamps from all three groups 87.6 s, respectively.

Effect of TiO2 Concentrations and UVA Exposure on the NO2 Concentration
The reduction and uptake coefficient are presented in Figs. 4 and 5, respectively, for reactive surface area concentrations between 0.102 and 2.86 m 2 m -3 , which were illuminated under UVA exposure in the range of 351.6-2109 W s m -2 (one to six lamps from the first group with an exposure time of 29.2 s).
In the presence of TiO2 nanoparticles, the NO2 reduction gradually increases with increasing UVA exposure (the effect of UVA exposure on the photochemical reactions in the presence of TiO2 is presented in the supplementary materials in the section "The effect of UVA exposure on the photocatalysis reaction"). For instance, the reduction increased from 35.7% to 63.2% by increasing the UVA exposure from 351.6 W s m -2 (one active lamp) to 2109 W s m -2 (six active lamps) with a reactive surface area concentration of 0.79 m 2 m -3 . This trend is in good agreement with previous studies. Bartolomei et al. (2014) showed that increasing light intensity as well as exposure time leads to a higher decrease of the NO2 concentration. Our experimental results further show that with increasing reactive surface area concentration, the reduction also grew for each tested UVA exposure. For UVA exposure of 1055 W s m -2 (three active lamps), the NO2 reduction increased from 27.5% to 97.3% by enlarging the reactive surface area concentration from 0.102 to 2.86 m 2 m -3 . (It is necessary to explain that the reported reduction of nearly 100% only applies to the experimental conditions used in the present experiment. The reduction depends on the NO2 concentration, UVA exposure and reactive surface area concentration.) Zein and Bedjanian (2012) similarly reported that by increasing the surface length coated with TiO2 in their experiment by a factor of 4, the ratio of the NO2 concentration at the outlet to the inlet reduced by a factor of 3.
By increasing the reactive surface area, the dependency of NO2 loss on UVA exposure decreased (chemical reactions and explanation related to the role of TiO2 on the photochemical reactions are presented in the section "Governing photochemical reaction inside the reactor" in the supplementary materials). For example, for a reactive surface area concentration of 0.102 m 2 m -3 , the reduction of NO2 increased by a factor of 2.7 times by adding UVA exposure from 351.6 to 2109 W s m -2 . However, for the reactive surface area concentrations of 0.79 and 2.8 m 2 m -3 the NO2 concentrations were reduced by a factor of 1.7 and 1, respectively, for the same range of UVA exposure. Our experimental results also show that for the highest UVA exposures, the reactive surface area concentration only plays a minor role. For instance, by changing the reactive surface from 0.102 to 2.8 m 2 m -3 the reduction increased by a factor of 7 for the UVA exposure of 351.6 W s m -2 . This increase was reduced to a factor of 1.75 for the UVA exposure of 6328 W s m -2 . (Results in details are presented in Table S1 in the supplementary materials.) Fig. 5 shows the uptake coefficient as a function of UVA exposure for different TiO2 reactive surface area concentrations. (The first-order reaction rate is calculated, and results are presented in Fig. S3.) The trend of the uptake coefficient, for each reactive surface area concentration, is increasing with increasing UVA exposure. For instance, the uptake coefficient raised from 4.9E-04 to 1.5E-03 by increasing the UVA exposure from 351.6 W s m -2 (one active lamp) to 2109 W s m -2 (six active lamps) with a reactive surface area concentration of 0.102 m 2 m -3 . For a reactive surface area concentration of 0.79 m 2 m -3 , an increase of the UVA exposure from 351.6 to 2109 W s m -2 (6-fold) resulted in an increase of the uptake coefficient by a factor of 2.25. This result is in line with those reported by Zein and Bedjanian (2012). In their study, they also report that a 6-fold increase of the irradiation intensity (increasing j = 0.002 to 0.012 s -1 ) increased the uptake coefficient by a factor of approximately 2.
Investigating the effect of TiO2 reactive surface area concentration in our experiments showed that adding the TiO2 nanoparticle concentration increases the photocatalytic reaction of NO2 (Fig. 4) and consequently resulted in a higher reaction rate (Fig. S3). However, the trend of the uptake coefficient decreases with increasing TiO2 reactive surface area concentration (more information is added to the section "The effect of TiO2 on uptake coefficient" in the supplementary materials. The graph of uptake coefficient versus reactive surface area concentration (Fig. S4) is also presented that section.) Similar results were reported by Zein and Bedjanian (2012), who reported that the uptake coefficient decreased with increasing the TiO2 mass over the illuminated surface. The upper limit of the uptake coefficient in their research was 1.9E-03 which was for the smallest TiO2 mass for coating the reactive surface.
The highest uptake coefficient resulting from our experiments was 1.5E-03 for a reactive surface area concentration of 0.102 m 2 m -3 and a UVA exposure of 2109 W s m -2 (j = 0.012 s -1 ), while the lowest uptake coefficient was 2E-04 for a reactive surface area concentration of 0.79 m 2 m -3 and a UVA exposure of 361.5 W s m -2 (j = 0.002 s -1 ). This is again in good agreement with the results reported by Zein and Bedjanian (2012) who found uptake coefficients between 0.0019 and 0.00012 for 0.002 s -1 ≤ j ≤ 0.012 s -1 . Gustaffson et al. (2006) also reported the range of uptake coefficient for aerosolized TiO2 to be in the range between 1.2E-04 and 9.6E-04 for different relative humidity levels and with a reactive surface area concentration of 0.43 m 2 m -3 . In summary, the comparison of our data with those obtained by Gustaffson et al. (2006) and Zein and Bedjanian (2012) show that results from aerosolized TiO2 nanoparticles compare well with those from TiO2-coated surfaces when the Brunauer-Emmet-Teller (BET) surface area concentration, provided by the manufacturer, is used as the reactive surface area concentration in the data evaluation of the aerosol experiments. (The details of reduction, first-order reaction rate and uptake coefficient of reported results are presented in Table S1.)

Effect of Exposure Time on the NO2 Reduction
To investigate the effect of exposure time on the reduction of the NO2 concentration and the uptake coefficient, two different intensities of 12 W m -2 (one lamp) and 72 W m -2 (six lamps) for two reactive surface area concentrations of 0.102 and 0.79 m 2 m -3 , are presented in Figs. 6 and 7, respectively. For the constant intensity of 12 W m -2 (or 72 W m -2 ), the exposure time is 29.2 s for one lamp (or six lamps) in the first group. The exposure time of 58.4 and 87.6 s was adjusted by adding one lamp (or six lamps) in the second and third group, respectively. The arrangement of lamps is presented for one and six lamps in Figs. S5 and S6, respectively. Fig. 6 shows that illumination of the reactor with a high intensity (72 W m -2 ) resulted in the highest reduction of the NO2 concentration, but the effect of the exposure time gets weaker as an increase of the UVA exposure time by a factor of 3 only resulted in an approximately 30% higher decrease in the NO2 reduction. For instance, by increasing the exposure time from 29.2 to 87.6 s with a reactive surface area concentration of 0.79 m 2 m -3 , the reduction increases only by a factor of 1.26. In contrast, the enhancement of the reduction for the lower intensity is higher and showed a nearly linear trend. Comparison of results also revealed that for the longer exposure time the effect of UVA intensity is dominant compared to the effect of the reactive surface area. For instance, one can note that the difference between the NO2 concentration reduction increased between two cases of reactive surface area concentration of 0.102 m 2 m -3 and UVA intensity of 72 W m -2 and reactive surface area concentration of 0.79 m 2 m -3 and intensity of 12 W m -2 . For the same exposure time, the case of the nearly 8 times higher reactive surface area concentration resulted in a lower NO2 reduction compared with the case of 6 times higher UVA exposure time. A similar trend was observed between two cases of "reactive surface area concentration of 0.102 m 2 m -3 and UVA intensity of 12 W m -2 " and "without TiO2 nanoparticles and UVA intensity of 72 W m -2 " where the higher UVA intensity resulted in higher NO2 reduction, even in absence of TiO2 nanoparticles. (Results of first-order reaction rate are shown in Fig. S7.)  Our results (Fig. 7) show that the uptake coefficient decreases by increasing the exposure time. A similar descending trend for the uptake coefficient versus exposure time has been reported in the literature (Laufs et al., 2010;Zein and Bedjanian, 2012;Bartolomei et al., 2014). According to Zein and Bedjanian (2012) this is caused by the fact that the experimental conditions approach steady state condition, in which the TiO2 nanoparticles start to deactivate (more explanation is presented in the section "The effect of exposure time on the rate of NO2 reduction" in the supplementary materials). Fig. 7 also shows that for each investigated exposure time, a higher intensity results in a higher uptake coefficient and thus a stronger photocatalytic reaction. Comparing the experimental results showed that the dependence of the uptake coefficient on UVA intensity is higher for lower reactive surface area concentrations. In the same way, the effect of reactive surface area concentration is dominant for low UVA intensity.

Effect of Humidity on the NO2 Reduction
The effect of humidity on the reduction of NO2 and uptake coefficients was investigated for relative humidity values of 5%, 35% and 70%. The NO2 reduction and uptake coefficients under UVA exposure 351.6, 1055, and 6328 W s m -2 for two different TiO2 reactive surface area concentrations of 0.102 and 0.79 m 2 m -3 are presented in Fig. 8.
Figs. 8(a) and 8(b) show the reduction for the reactive surface area concentration of 0.102 and 0.79 m 2 m -3 , respectively. For the lower reactive surface area concentration of 0.102 m 2 m -3 , there is no clear difference between the relative humidity of 5% and 35%. However, by increasing the relative humidity from 35% to 70%, the reduction of the NO2 concentration shows a slight increase. For the highest UVA exposure (6328 W s m -2 ), the effect of humidity is low and the dominant source of the NO2 reduction is UVA exposure.
By increasing the reactive surface area concentration to 0.79 m 2 m -3 ( Fig. 8(b)), the relative humidity also has an effect on the NO2 reduction at low UVA exposure; however, for the high UVA exposure (6328 W s m -2 ), noticeable differences were not observed and the reduction of NO2 for all three different relative humidity values were around 80%. The results for a reactive surface area concentration of 0.79 m 2 m -3 and lower UVA exposure are in general agreement with the findings of Bartolmie et al. (2014), who reported that the NO2 reduction was higher at a relative humidity of 70% than at 50% for an exposure time up to 3.5 s and a UVA intensity of 10.6 W m -2 over a coated surface. Zein and Bedjanian (2012) reported the initial first-order reaction rate to be higher at 40% relative humidity than at 0%. However, for long UVA exposure time, the difference was not noticeable anymore. (The first-order reaction rate for both reactive surface area concentrations of 0.102 and 0.79 m 2 m -3 are presented in Fig. S8.) The uptake coefficient for the reactive surface area concentrations of 0.102 and 0.79 m 2 m -3 are presented in Figs. 8(c) and 8(d), respectively. For the lower UVA exposure (351.6 and 1055 W s m -2 , intensity of 12 and 36 W m -2 and an exposure time of 29.2 s), the effect of increasing relative humidity is negligible for reactive surface area concentrations of 0.102 m 2 m -3 in comparison with 0.79 m 2 m -3 , especially for low relative humidity levels (5-35%). In previous research (Pradhan et al., 2010T;ang et al., 2014), it was reported that the uptake coefficient is constant for relative humidity levels below 40%, since the amount of adsorbed water is quite low. However, for higher relative humidity, they reported an increasing trend for the uptake coefficient. They explained that increasing the water layer on the TiO2 nanoparticle surface leads to an increase of the uptake coefficient (more details are presented in the section "The effect of humidity on uptake coefficient" in the supplementary materials). Our experimental results also show that the uptake coefficient is increased at a relative humidity level of 70% for the reactive surface area concentrations of 0.102 m 2 m -3 . In the case of reactive surface area concentration of 0.79 m 2 m -3 , results showed an increasing trend with increasing relative humidity up to 40%. This difference is due to the relatively large reactive surface area, which made it possible to adsorb more water and thus leads to a higher uptake coefficient than at the low relative humidity level of 5%.
For the UVA exposure of 6328 W s m -2 (72 W m -2 with exposure time of 87.6 s), the explanation about exposure time is added in the supplementary materials; the uptake coefficient for all three relative humidity cases showed similar results for both reactive surfaces of 0.102 and 0.79 m 2 m -3 . Zein and Bedjanian (2012) also reported that the uptake coefficient is not sensitive to relative humidity in the case of high exposure time which agrees with our results for the high UVA exposure. They reported that for NO2 concentration of 10 13 molecules cm -3 (around 406.5 ppb) under UVA with photolysis frequency 0.006 s -1 (similar to two lamps in our study, 24 w m -2 ) with TiO2 sample mass of 0.4 mg cm -1 (2320 m 2 m 3 , by considering TiO2 density with 1.236 g cm -3 and tube diameter of 2.4 cm). For our experimental condition, the amount of intensity is 9 times higher and the amount of reactive surface is around 2937 times lower than Zein and Bedjanian (2012) experiment conditions. Therefore, it can be expected that the speed of approaching TiO2 nanoparticles to steady state condition, in our experimental condition, occurred quite faster. Gustaffson et al. (2006) investigated the reaction between NO2 (with mixing ratio of 100 ppbv) and aerosolized TiO2 (with the concentration of 0.43 m 2 m -3 ) under UVA illumination with the intensity of 16 W m -2 for relative humidity levels between 15% and 80%. They reported the lower and upper limit to be 1.2E-04 and 9.6E-04, respectively, which is in good agreement with our results (between 1.9E-04 and 1.5E-03). However, the results presented by Gustaffson et al. (2006) showed a negative correlation with relative humidity which is different from our results as well as results reported by Zein and Bedjanian (2012) and Bartolmie et al. (2014). Zein and Bedjanian (2012) explained this difference to be possibly due to the short exposure time. The different results obtained by Gustaffson et al. (2006) cannot be explained at this point, because their paper does not provide information on the flow rate, exposure time and geometry of their experimental set-up.

CONCLUSION
We investigated the reduction of NO2 using aerosolized TiO2 nanoparticles, which ranged from 0.102 to 2.8 m 2 m -3 in reactive surface area concentration (and from 0.47E+06 to 3.23E+06 cm -3 in number concentration, respectively), and UVA, which ranged from 351.6 to 6328 W s m -2 . In general, the UVA alone caused the largest decrease in the ambient NO2. Only high concentrations of TiO2 produced a noticeable effect (> 0.102 m 2 m -3 [> 0.47E+06 cm -3 ]), although the maximum value (2.8 m 2 m -3 [3.23 E + 06 cm -3 ]) drove a reduction in the NO2 that was almost dissociated from the UV intensity when the exposure was continuous. Thus, we established that the drop in NO2 depended on both the TiO2 concentration and the UV exposure when the former was high. The upper limit of the uptake coefficient was 1.5E-03 when the reactive surface area concentration was 0.102 m 2 m -3 and the UVA exposure was 6328 W s m -2 , and the lower limit was 1.9E-04 when the concentration was 0.79 m 2 m -3 and the exposure was 351.6 W s m -2 . Moreover, we discovered that the percentage of reduction attributable to the UVA intensity grew with the exposure time, possibly owing to the deactivation of the TiO2.
Finally, we also assessed the effect of the relative humidity on the NO2 reduction under varying conditions. The influence of this factor, although significant when the UVA intensity and the TiO2 concentration were high and low, respectively, became negligible when the UVA intensity was high, regardless of the TiO2 concentration.