During the European Life+ project PhotoPAQ (Demonstration of Photocatalytic remediation Processes on Air Quality), photocatalytic remediation of nitrogen oxides (NO_x), ozone (O₃), volatile organic compounds (VOCs), formaldehyde (HCHO) and particles was studied both, in the Leopold II tunnel in Brussels, Belgium, and in an artificial street canyon in Bergamo, Italy using photocatalytic cementitious coating materials.

Der Fachvortrag zur Veranstaltung ist im Volltext verfügbar. Das PDF enthält alle Bilder und Formeln.

Graphical Abstract

Abstract

During the European Life+ project PhotoPAQ (Demonstration of Photocatalytic remediation Processes on Air Quality), photocatalytic remediation of nitrogen oxides (NO_x), ozone (O₃), volatile organic compounds (VOCs), formaldehyde (HCHO) and particles was studied both, in the Leopold II tunnel in Brussels, Belgium, and in an artificial street canyon in Bergamo, Italy using photocatalytic cementitious coating materials.

In contrast to former field studies, no significant photocatalytic remediation of NO_x, O₃, VOCs and particles was observed. For NO_x an upper limit remediation of ≤2% was derived for both studies. For the tunnel experiments this is explained by strong deactivation of the photocatalytic materials applied under the highly polluted tunnel conditions. Simple model calculations using reactivity data for the deactivated surfaces from the laboratory indicate only a theoretical reduction of 0.4 % in good agreement with the experimental results. For the canyon experiment the low reduction is explained by transport limitations. Reasons for higher remediation results reported in former field studies are discussed leading to a more consistent picture after extrapolation to realistic main urban street canyon conditions. In contrast to non-measurable photo-catalytic effects for most pollutants, undesired photocatalytic formation of formaldehyde (HCHO) was identified, which is explained by degradation of organic additives in the used cementitious materials.

If photocatalytic materials are optimized for high reactivity and low emissions of harmful degradation products they can generally be recommended for urban air remediation. However, the expected NO_x reductions under typical urban conditions of a few % at maximum are much lower than promised by industry. Finally, these reductions have to be compared on a cost-benefit analysis basis with other measures to improve urban air quality (e.g. low emissions zones), for which, e.g. with respect to urban NO₂ levels, also only negligible improvements were obtained in past.

Keywords: TiO₂; photocatalysis; air purification; tunnel study, canyon study

1. Introduction

Urban air quality is of high importance for human health since the majority of the world’s population lives inside metropolitan areas. Different methods to improve urban air quality have been implemented in the past, for example, improved combustion and exhaust after-treatment technologies and implementation of low emission zones, which generally show a positive impact on the environment. However, while for example concentrations of nitrogen oxides (NO_x) have been significantly reduced during the last three decades in European cities, annual averaged levels of nitrogen dioxide (NO₂) are stagnating. In addition, these levels are still commonly higher than the actual annual threshold limit value of 40 µg m^-3 (~20 ppb) under urban conditions [2, 3]. Besides NO_x, threshold limits for particles and ozone (O₃) are also often exceeded in Europe [4, 5] as a result of the high levels of precursors like volatile organic compounds (VOCs) [6, 7].

In view of the adverse health effects of air pollution, further improvement of the urban air quality is necessary. Here, over the two last decades, titanium dioxide (TiO₂) based photocatalytic surfaces have been developed and tested, both, at a laboratory scale and in the open atmosphere as an alternative technical means for the remediation of NO_x [8, 9, 10, 11, 12], VOCs [12, 13, 14, 15], O₃ [15] and particles [15, 16, 17].

However besides the desired reduction of primary pollutants by photocatalytic surfaces, the potential formation of harmful by-products is controversially discussed. While it is generally assumed that for example NO_x is quantitatively converted into nitrate [18], some recent studies on pure photocatalysts and on self-cleaning window glass also observed the formation of the intermediate nitrous acid (HONO) [19, 20, 21, 22], which is even more harmful than the primary reactants NO and NO₂ [23]. In addition, re-noxification and ozone formation originating from photocatalytic decomposition of adsorbed nitrate was recently observed in laboratory experiments [22, 24]. Also, harmful oxygenated species, for example aldehydes [25, 26, 27], were detected during the photocatalytic degradation of VOCs.

Besides potential formation of harmful reaction products, controversial results also exist concerning the extent of NO_x reduction in the real urban atmosphere. Two field trials in artificial model street canyons showed high NO_x remediation of 25-30% [28] and 40-80% [8, 29], respectively. These studies are in general agreement with two experiments in real urban street canyons for which NOx reductions of 19% [30] and 26-66% [31] were reported. In contrast, the measured NO_x reduction was below precision errors in recent field projects using photocatalytic noise protection barriers at Putten, Netherlands [32] and Grenoble, France [33], in agreement with a recent study using active pavement blocks on the sideways of an urban street in Fulda, Germany [34]. Reasons for these contradictory results are still under discussion and further field studies are necessary to better account for the total impact of this technology on air quality.

In addition to the application on outdoor surfaces, photocatalysis might also contribute to air quality improvement inside road tunnels, but requires the photocatalytic tunnel surfaces to be irradiated by UV light. The expected benefits would be the reduced exposure of the passengers to air pollution, potential reductions in ventilation requirements and a reduction of tunnel exhaust related pollution on the surrounding areas. However so far, only one study has been conducted on the photocatalytic NO_x reduction in a road tunnel in Rome, Italy, for which a significant abatement of 20 % was reported [35]. Caused by the limited available experimental data, further studies are necessary to quantify the impact of photocatalytic air remediation inside road tunnels.

The European Life+ project PhotoPAQ (Demonstration of Photocatalytic remediation Processes on Air Quality) [1] was aimed at demonstrating a possible usefulness of photocatalytic construction materials for air purification purposes in the urban environment. Both, tunnel [36, 37] and street canyon studies [38] were performed to test photocatalytic cementitious coating materials in the real atmosphere.

2. ExperimentalTunnel Study

The tunnel study was performed in a section of one tunnel tube of the Leopold II tunnel, in Brussels, Belgium. The experimental set-up is shown schematically in Fig. 1 and details can be found elsewhere [36, 37]. After a first pre-campaign which took place in June 2011 in the untreated tunnel, the side walls and ceiling of a 70 m test section were coated with a commercial photocatalytically active mortar (Italcementi, TX-Active^® Skim Coat, hereafter: TX) in August 2011, followed by a monitoring campaign in September 2011. In addition, in January 2013 a third and final tunnel campaign was conducted in the same section which was extended to 160 m and covered by a newly developed photocatalytically more active mortar (Italcementi, TX-Active Skim Coat Boosted, hereafter: TX-Boosted). During the September 2011 and January 2013 campaigns, two different switchable UV lighting systems were installed in the corresponding tunnel sections (see graphical abstract) with average irradiance levels (range 315-420 nm) on the active surfaces of 0.6±0.3 and 1.6±0.8 W m^-2, respectively.

The testing strategy applied during the three tunnel trails will be described in detail elsewhere [37] and is only summarized here. In short, three different strategies have been applied to quantify the NO_x abatement:

(1) Monitoring of the pollutants before and after the application of the active material.

(2) Monitoring of the pollutants at the same time at two measurement sites up- and down-wind of the active tunnel section with lamps on.

(3) Monitoring of the pollutants at the downwind site with the UV irradiation on and off.

Due to the highly variable nature of atmospheric composition, traffic flow and dilution inside a road tunnel, the monitored NO_x concentrations were normalized to the photocatalytically inert tracer CO₂, which is simultaneously emitted by the vehicles together with the pollutants under investigation (NO_x, VOCs, etc.). From the slope of a plot of a selected pollutant against CO₂, emission ratios (pollutant/CO₂) were obtained, which are independent of the absolute pollution level and thus of the vehicle source strength and the variable dilution in the tunnel. Changes in this ratio, for example, between up- and downwind of the active section can thus be attributed to the photocatalytic remediation.

Fig. 1: Experimental set-up in the Leopold II tunnel in Brussels, Belgium

To follow the three strategies mentioned above, two identical sets of scientific instruments were deployed at two different measurement sites in a large control room above the active tunnel section, where the trace gases were sampled from the tunnel [37]. Under normal axial flow direction, which was controlled by the one-directional traffic flow, the station upwind of the active section was defined as site 1 and the one downwind as site 2. Thus, under normal traffic conditions, only the air studied at site 2 was in contact with the photocatalytic surfaces.

2.1.1 Monitored Parameters

At each of the two measurement sites two similar sets of instruments allowed quantification of the photocatalytic remediation. All instruments were carefully intercalibrated during the campaigns, details of which are described elsewhere [36].

NO_x and HONO were detected by the chemiluminescence and LOPAP techniques, respectively. Formaldehyde (HCHO) was measured by a Hantzsch instrument and for carbon dioxide (CO₂) the Non-Dispersive Infra-Red (NDIR) absorption technique was used. In addition, also meteorological parameters, i.e. wind speed (WS) and direction (WD), temperature (T) and relative humidity (r.h.) were measured using two weather stations installed close to the inlets of the trace gases below the ceiling of the tunnel.

2.2 Canyon Study

2.2.1 Field site

Originally, the PhotoPAQ consortium intended to study photocatalytic air remediation in a real street canyon situation. However, during the course of the project it turned out difficult to find a field site for which a) both, the street surfaces and the walls of the buildings could be covered with the photocatalytic materials to obtain high active surface to volume ratios (S_active/V) and a quantifiable pollutant reduction, and b) a co-located non-active reference site of similar geometry and pollution level was present to allow for correction of the meteorological variability in pre- and after-application measurement campaigns. Thus finally, the consortium decided to set-up two artificial street canyons of the dimensions 5×5×53 m (width×height×length) at an industrial site from the industry partner Italcementi in Petosino – Sorisole, a few kilometers north of Bergamo, Italy. To create conditions as similar as possible in both canyons, the walls were constructed with fibre cement boards mounted on metal scaffoldings in front of the original buildings (see graphical abstract and Fig. 2).

During the control campaign (11^th – 17^th April 2013) the surfaces of both untreated canyons were photocatalytically inactive and consisted of the fibre cement walls and the original concrete street surfaces. During the main campaign (30^th April – 7^th May 2013) walls and ground surfaces only of one canyon (hereafter: active canyon) were coated with a photocatalytically active mortar (Italcementi, TX-Active Skim Coat Boosted, hereafter: TX-Boosted) specially designed for application in the PhotoPAQ project. The other canyon (hereafter: reference canyon) was not modified.

Fig. 2: Schematic presentation of the model street canyon site in Petosino.

2.2.2 Monitored Parameters

The two street canyons were separated by two buildings (see graphical abstract and Fig. 2) which housed the analytical instruments, which are explained in detail elsewhere [38]. All instruments were carefully intercalibrated before and after each campaign. NO_x and HCHO were determined by two chemiluminescence and Hantzsch instruments, respectively. For the VOCs a SRI-PTR-TOF-MS 8000 (Selective Reagent Ionization Proton Transfer Reaction Time of Flight Mass Spectrometer) was used. Ozone concentrations in both canyons were measured by two UV absorption ozone analyzers. Particle composition and size distribution were determined by Berner impactors, a Scanning Mobility Particle Sizer (SMPS) and a High Resolution Time-of-Flight Aerosol Mass Spectrometer (AMS). Finally, also metrological parameters and UV light intensity (irradiance, actinic flux) were quantified.

3. Results and Discussion

3.1 Tunnel Study

3.1.1 Nitrogen oxides (NO_x)

Based on a) laboratory data on clean samples, b) corresponding simple estimations (see section 0) and c) results from the only available photocatalytic tunnel study [35], a significant difference in the NO_x concentration between the two sites was initially expected under UV irradiation. To test this hypothesis the UV lamps were periodically switched on and off in the tunnel. As an example, one day from the campaign in January 2013 with the longer active tunnel section (160 m), the more reactive “TX-Boosted” material and the higher irradiance level (1.6 W m^-2) is shown in Fig. 3. However, switching on and off the lamps did not result in any systematic steps in the concentration difference between the sites (see Fig. 3).

Fig. 3: Plot of the 10 min NO_x and lamps on/off data at sites 1 and 2 during one day of the campaign in January 2013.

To improve the significance of the data evaluation and eliminate dependencies from the variable pollution level in the tunnel, the photocatalytic remediation was evaluated from the average NO_x/CO₂ ratio using the three following approaches.

Before/After Approach

One approach to quantify the photocatalytic remediation in the tunnel is to compare the NO_x/CO₂ ratio from the first campaign (original tunnel surface) with the ratios from the later two campaigns (active surfaces). For the campaigns with active surfaces only the NO_x/CO₂ data from the downwind site 2 has been analyzed where photocatalytic remediation should be visible. Within the experimental uncertainties, no difference between the campaigns could be observed for the NO_x/CO₂ ratios when using all the campaign data, indicating a low photocatalytic remediation in the tunnel, see Tab. 1. Since the photocatalytic remediation will only be active under UV irradiation, for the latter two campaigns NO_x/CO₂ ratios were also calculated for the periods with the lamps switched on. Again the similar ratios obtained compared to the pre-campaign indicate a low photocatalytic remediation in the tunnel.

Tab. 1: Average NO_x/CO₂ ratios determined for all three campaigns. The given errors reflect only the 2s precision errors from the linear regression analysis.

Upwind/Downwind Approach

To better quantify the photocatalytic remediation in the tunnel, simultaneous data of both measurement sites were compared when the lamps were switched on. Under these conditions, exactly the same air mass and vehicle emissions are studied at both sites, which further reduces the precision errors of the pollutant reduction, limited here mainly by the precision errors between the two similar instruments at both sites (±2 %) and the precision errors of the NO_x/CO₂ ratio.

When comparing all upwind and downwind data for both campaigns a minor formation of NO_x is visible in the September 2011 campaign (70 m test section), whereas a small reduction may be inferred at the downwind site 2 for the January 2013 campaign with the extended test section, see Tab. 2. These results are qualitatively in agreement with laboratory experiments on the sample plates exposed to tunnel air (see section 0). In these experiments, photocatalytic formation of NO_x was observed on “dirty” tunnel samples under the conditions of the September 2011 campaign (TX), whereas only a very small reduction in NO_x was measured on sample plates exposed to tunnel air under the conditions of the January 2013 campaign (TX-Boosted). However, considering the combined errors of the NO_x/CO₂ ratios and the precision of the duplicate instruments used (±2 % for NO_x), the observed differences in both tunnel campaigns are not significant. Similar results were also obtained when only low wind speed data (longer reaction time) were used, for which the highest reduction was expected.

In conclusion, also when using the upwind and downwind approach no significant photocatalytic remediation was visible. As an upper limit of the pollutant reduction a value of ~3 % is estimated by error propagation of the NO_x/CO₂ ratio (2 %) and the precision between the two instruments used (2 %).

Tab. 2: Average NO_x/CO₂ ratios from both sites for the two campaigns with photocatalytically active tunnel surfaces and with lamps on. The given errors reflect only the 2s precision errors from the linear regression analysis.

Lamps On/Off Approach

Finally, in the on/off approach only the trace gas concentrations at the downwind site 2 were used and the results were compared for active (lamps on) and inactive (lamps off) photocatalytic remediation. This approach further minimizes precision errors since only one instrument for each species is used.

Comparison of the NO_x/CO₂ ratios for the active (on) and non-active (off) experimental conditions again does not show any quantifiable remediation beyond the experimental uncertainty (see Tab. 3). This result is also obtained when only low wind speed (long reaction time) data were used, for which the highest reduction was expected. As an upper limit, taking into account the precision errors of the data analysis, a value of ~2 % has been estimated for the photocatalytic remediation in the longer more active tunnel section from the campaign in January 2013.

Tab. 3: Average NO_x/CO₂ ratios at the downwind site 2 with lamps on or off for the campaign in January 2013. The given errors reflect only the 2s precision errors from the linear regression analysis.

3.1.2 Laboratory NO_x Experiments on Samples Exposed to Tunnel Air

To understand the low NO_x remediation, laboratory experiments were performed using samples collected from the tunnel, details of which are explained elsewhere [37].

In laboratory experiments using NO as reactant and samples from the September 2011 campaign (TX), significant formation of NO_x could be observed under UV irradiation. This behaviour contrasted sharply with the clean reference samples, for which reasonable uptake of NO and net loss of NO_x was obtained. This unwanted NO_x formation could be explained by (a) desorption of adsorbed NO_x and/or (b) photocatalytic formation of NO_x from adsorbed nitrates [22, 24]. These TX samples could not be significantly reactivated even when irradiated with 4 W m^-2 UVA for 121 h. The laboratory results on the deactivation are in excellent agreement with the missing NO_x-remediation in the September 2011 tunnel campaign and demonstrate that existing standard test methods for photocatalytic materials, in which deactivation issues are not considered, like ISO [39] or UNI [40] methods, should not be used alone to estimate potential pollutant reduction in the real environment.

Consequently, a more active photocatalytic material (TX-Boosted) was produced by the industry partner and applied during the January 2013 tunnel campaign. However, also when these samples were stored in the dirty tunnel prior to the laboratory experiments, an order of magnitude reduced photocatalytic activity compared to the clean reference samples was observed at the low irradiance and high humidity levels of the tunnel study (photocatalytic deposition velocities [41] clean reference sample: 0.22 cm s^-1; tunnel sample 0.019 cm s^-1). In addition, a much higher NO₂ yield of 50 % was observed for the dirty sample compared to the 16 % yield for the reference sample. The slow uptake kinetics of NO and the observed NO₂ formation show that a significant NO_x remediation was also not to be expected in the last campaign.

3.1.3 Upper Limit Calculations

Based on the reactivity data of the investigated material, simple model calculations were performed to estimate the expected remediation in the tunnel. The calculations which are explained elsewhere [36] were based on a first order kinetics in a well mixed tunnel situation, for which besides the uptake kinetics on the deactivated samples only information about the geometry of the tunnel and the wind speed are necessary input parameters. Using the conditions for the campaign in January 2013 in the 160 m tunnel section on the tunnel exposed TX-Boosted samples, a NO reduction of only 0.40 % is calculated, which is clearly below the experimental uncertainties. This simple upper limit model describes very well the obtained experimental tunnel results. Only when the observed passivation of the material was ignored in the model and when a very high UVA irradiance of 10 W m^-2 and a low wind speed of 1 m/s was assumed, a theoretical reduction of ~20 % was obtained, in agreement with the 20 % reduction observed in a 350 m long tunnel in Rom [35]. The model tool is available also for other future applications of photocatalysis in road tunnels [1]

3.2 Canyon Study

3.2.1 Nitrogen Oxides (NO_x = NO + NO₂)

Fig. 4 shows all 10 min averaged NO_x and UVA data from the main measurement campaign with the photocatalytically active surfaces during May 2013 in Petosino.

Fig. 4: 10 min averaged NO_x concentration in the active canyon, difference between both canyons and UVA irradiance during the main field campaign in Petosino.

NO_x levels varied from 1 to 30 ppb and thus, the field site represented only a moderately polluted urban background situation. During the first part of the campaign (1^st – 4^th May 2013) the weather was typically good with only partial cloud coverage and high UVA irradiance of up to 40 W m^-2 inside the canyon. From the morning of the 5^th May to the end of the campaign the weather turned less favourable for photocatalysis with higher cloud coverage and lower UVA levels. The wind speed (WS) measured inside the active canyon was typically very low (<1 m s^-1). Hence, at least for the first four days, the low WS and high UVA favoured photocatalytic remediation inside the canyon. However, from the difference of the 10 min averaged NO_x data between both canyons, on average no significant remediation is visible (see Fig. 4).

The weather during the last two days of the main campaign was partially rainy and compared to the first days much lower UVA levels were observed (see Fig. 4). Therefore, the period 1^st-4^th May 2013 with a higher expected photocatalytic remediation was separately evaluated. All 10 min averaged NO, NO₂ and NO_x data from the active canyon were plotted against corresponding data from the reference canyon for day- and night-time separately (see Fig. 5).

Fig. 5: Correlation plots of the active against the reference site for NO, NO₂ and NO_x during night-time and daytime of the period 1^st - 4^th May 2013.

While deviations of the individual slopes from unity may explained by a) systematic differences between the instrument’s responses and b) potentially general pollution differences between both canyons, the comparison of the slopes between daytime and night-time should only reflect the photocatalytic effect independent of the artificial differences a) and b). However, even using this more precise correlation data, no photocatalytic reduction of NO₂ and NO_x is observed. It is only for NO that a small remediation may be inferred (see Fig. 4; slope night = 0.988; slope day = 0.953; resulting remediation: ca. -3.5%). In conclusion, when comparing day- and night-time correlation data, an upper limit NO_x remediation of ≤2% can be derived from the present field data.

Subsequent laboratory studies comparing the reactivity of untreated reference samples with those directly collected from the canyon after the field campaign showed no deactivation of the reactivity in the open atmosphere in contrast to the tunnel study. Thus, potential fast deactivation of the photocatalytic surfaces as observed in other field studies [30, 36, 37] cannot explain the low photocatalytic remediation results from the present field trial.

3.2.2 Volatile Organic Compounds (VOCs)

The following VOCs were selected as typical organic pollutants and summed to quantify the photocatalytic degradation of VOCs: butene, pentene, benzene, toluene, xylene, trimethylbenzene, diethylbenzene, acetone and propanol. Correlation plots of the sum of the VOC concentrations from both canyons obtained during the control campaign with non-active surfaces reveal no significant difference between the two sites (slope active against reference: 0.966 ± 0.028) confirming the expected general similarity of the two co-located canyons. However, exactly the same results were obtained also for the main measurement campaign with the active canyon (slope active against reference: 0.970 ± 0.029) indicating no significant photocatalytic degradation of the selected VOCs by the photocatalytic material. In between the experimental errors, same results were also obtained when single VOCs were considered. From the combined 2s precision errors of the correlation data and of the PTR-TOF-MS an upper limit of the photocatalytic remediation of the selected VOCs of ≤5% was derived.

3.2.3 Ozone (O₃)

The correlation of the O₃ data from both canyons collected during the control campaign shows that the two artificial street-canyons were comparable also with respect to the ozone levels. When all campaign data were used, only small difference between the two sites was visible, with slightly higher values in the active canyon (slope active against reference: 1.046 ± 0.016). When only the daytime data was considered even smaller differences were observed (slope: 1.026 ± 0.024). Due to the fast Leighton equilibrium between O₃ and NO_x during daytime [6], it can be concluded that both canyons were under the influence of similar pollutant sources.

During the main campaign with active photocatalytic surfaces no remediation of O₃ was again observed. The correlation using all campaign data showed exactly the same slope compared to the control campaign (slope active against reference: 1.038 ± 0.021). When using only daytime data, the correlation slope was even higher compared to the control campaign (slope active against reference: 1.067 ± 0.030), with the difference in between the combined precision errors. Higher O₃ levels in the active canyon during daytime may be explained by photocatalytic O₃ formation as recently observed in laboratory experiments [24]. Based on the present study, an upper limit average photocatalytic remediation of ≤3% was derived from the combined 2s precision errors of the correlation data.

3.2.4 Particles

During the control campaign, no significant differences in the particle number, mass, size distribution and chemical composition between the two canyons were observed indicating again the general similarity of the two sites also with respect to the particles. However, similar results were also found during the main campaign when the active material was applied. Correlations of the obtained data between the two canyons for organics, nitrate, sulfate and ammonium reveal slopes indistinguishable from unity for all compounds. In conclusion, the impactor, SMPS and AMS results indicate no significant influence of the used photocatalytic material also on the particle concentrations and composition.

3.2.5 Discussion Canyon Study

In the present canyon study no significant photocatalytic remediation of NO_x, VOCs, O₃ and particles could be derived under atmospheric conditions at an urban background site. For NO_x only an upper limit of ≤2% could be derived from comparison of day- and night-time correlation plots. For VOCs and particles the non-significant reductions could be expected caused by a) slower photocatalytic uptake kinetics of VOCs, typically one order of magnitude smaller compared to NO_x [42], b) resulting also in a limited impact on secondary aerosol (SOA) formation by VOC oxidation and c) obviously, no direct influence of heterogeneous photocatalysis on primary airborne particles (e.g. soot) as proposed recently [15], since only adsorbed particles can be photocatalytically oxidized.

However, at least for NO_x much higher remediation was originally expected based on most other field trials in the open atmosphere [8, 28-31]. In contrast, non-quantifiable NO_x reductions were also observed at two motorways [32, 33] and at an urban street site [34]. The latter results may be explained by the low surface to volume ratios (S_active/V) of the open structured motorway field sites with only photocatalytically active noise protection walls [32, 33] or where only the sideways of a street were active [34]. However, in other studies in real [30, 31] or model street canyons [8, 28, 29] similar to the present study, much higher NO_x reductions in the range of 19-80% were reported. The much lower reduction of the present study compared to these former field trails can neither be explained by potential deactivation of the studied photocatalytic surfaces (see above) as observed in other investigations [30, 36, 37], nor by the geometry of the present canyon site with an even higher S_active/V ratio (0.6 m^-1) compared to other photocatalytic field experiments [28, 30, 31]. In contrast, the unrealistically high S_active/V ratio during the PICADA model canyon experiments using only 2.4 m wide canyons is one potential reason for the high NO_x reductions of up to 80% stated [8, 29]. As recently discussed [18], the S_active/V ratio of a field site is one important parameter limiting heterogeneous uptake on photocatalytic surfaces. Thus, high remediation results from experiments in smaller model sites ([8, 28, 29] and present study) have to be scaled down to real urban street canyon conditions. In the former study [18], extrapolation of results from smaller model canyon studies [8, 28, 29], resulted in an expected reduction of only ~5% in a typical main urban street.

Besides the unrealistic geometries used in some previous field trials, in all former experiments in which significant photocatalytic remediation was observed [8, 28-31] only daytime data were evaluated, while in the three other studies with non-measurable NO_x reduction [32-34], the whole diurnal data were considered. Since photocatalytic remediation is aimed as one measure to reach the annual threshold limit for NO₂ of 40 µg m^-3 under urban conditions, results from studies in which daytime data is used should only be considered as upper limits. NO_x levels during night-time are often comparable to those during daytime since higher emissions are compensated by stronger convective vertical dilution during daytime. Hence, relative reductions observed in studies using only daytime data can be roughly divided by a factor of two to estimate the average reduction in an urban environment. For example, since the estimated 5% reduction (see above and [18]) is still based on daytime data, only an annual NO_x reduction by photocatalysis of ~3% is expected in a typical main urban street.

In addition to the sampling intervals considered, also the distance of the sampling inlet to the active surface is of high importance to understand differences in the published photocatalytic NO_x remediation results. While a measurement height of 3 m is recommended for urban network stations in which annual NO₂ threshold limit exceedance is verified, photocatalytic remediation was typically measured much closer to the active surfaces in most former studies. For example, in PICADA [8, 29] samples were taken only a few cm away from the active canyon walls, again explaining the exceptionally high NO_x reduction results from this study. Also in other field trials in which high NO_x remediation was observed, the sampling heights above the active surfaces were much lower than 3 m (0.05-1.5 m in [30], 0.3-1 m in [31], 0.5 m in [28]). Caused by the expected and also observed [30] gradients of the photocatalytic reduction, the high remediation results of these former studies again have to be considered as upper limits. Unfortunately, extrapolation to the recommended 3 m sampling height is only possible using at least 1D model calculations to realistically describe the turbulent vertical mixing. In contrast, in the present study samples were directly collected at 3 m altitude, resulting in a more reasonable estimate of the expected reductions at urban measurement stations. Lower sampling heights may still be important for specific environmental issues, e.g. health effects for pedestrians or entrainment of pollutants to indoor air via air-exchange through windows mounted on the photocatalytically active walls. However, concerning the urban NO₂ problem, remediation should be quantified at 3 m height.

Another potential reason for the high reductions observed in some former field studies are general differences between the active and reference sites used. Here, similarity was presumed to quantify the photocatalytic remediation [8, 28-34], which may not always be the case, for example, in model street canyon experiments in which NO_x was artificially injected into the active and reference canyons [8, 28, 29]. In contrast, in the present study the homogeneous urban NO_x background was investigated, for which the similarity between both canyons is more reasonable. In addition, in real street canyon experiments in which the two sites are located at a significant distance from each other [30, 31], different emission strength and pollutant dispersion may cause high uncertainties in the remediation results. For example, in a recent model study [43, 44] results from an experimental street canyon campaign in Bergamo [31] were re-evaluated, resulting in an expected upper limit photocatalytic remediation during daytime in the range of only 4-14%, compared to the published values of 26-66% [31]. Reasons for this discrepancy were suggested to be a) strong differences in the vehicle NO_x emissions at the active and reference sites with much higher vehicle fleet density at the reference site and b) different dispersion conditions (geometry of the sites, micrometeorology). These model results can even be considered as an upper limit, since transport limitations represented by the turbulent and quasi laminar resistances [45] were neglected in the model. Thus, the calculated NO_x reduction during daytime of e.g. 4% for a reasonable average wind speed of 1 m s^-1 and when using a realistic photocatalytic deposition velocity of 0.3 cm s^-1 during daytime [44] will in reality be reduced to <2% using a) diurnal averages and b) considering also transport limitations. This result is again in excellent agreement with present and other studies [32-34] highlighting the necessity of careful interpretation of field data.

In conclusion, when considering all these differences in the geometry of the field sites, in sampling positions, in sampling periods and general differences between the active and reference sites, a much more consistent picture of the expected NO_x remediation by photocatalysis can be drawn. Based on the discussion presented above and considering results from different field and modelling studies, a realistic annual averaged NO_x reduction of ~2% can be estimated when photocatalytic active surfaces are used in main urban street canyons, in which the annual NO₂ threshold limit value of 40 µg m^-3 is typically exceeded.

However, even this number has to be considered as an upper limit since a) often commercially available photocatalytic surfaces show lower photocatalytic deposition velocities than the 0.3 cm s^-1 used in the model described above [44] and b) deactivation of the surfaces after longer use in the field may further reduce the remediation [30, 36, 37]. Here on the one hand, high threshold photocatalytic activities against NO₂ should be defined in future CEN (European Committee for Standardization) and ISO (International Organization for Standardization) standards to test photocatalytic surfaces, in which presently only NO is considered [41]. On the other hand, potential deactivation should be studied at the field site of interest in small scale experiments, before application of larger surfaces [36, 37]. Only when surfaces show high activity against NO₂ and when no strong deactivation under the specific field conditions is observed, should photocatalysis be recommended as one measure to reduce urban NO₂ levels.

The low reduction by photocatalysis may be considered as a disappointing result; however, it should be compared with other measures used to improve urban air quality. Here, also improved emission standards [46, 47, 48, 49] or implementations of low emission zones [50, 51] showed only small reductions of the urban NO₂ levels by a few percent at maximum. Hence, a complete costbenefit analysis of all discussed measures is highly recommended, for which photocatalysis may be still more attractive compared to other expensive methods if extra costs for photocatalytic surfaces compared to the application of normal urban surfaces (roads, paints, roof tiles, etc.) are minimized by industry in the future and when photocatalytical materials are applied when urban surfaces are renewed anyway.

3.3 Photocatalytic Formaldehyde (HCHO) Formation

Both, in the tunnel and in the canyon studies, undesired photo-catalytic formation of formaldehyde (HCHO) was observed on the photocatalytic surfaces used in PhotoPAQ which was scaling with the UV irradiance. This result was also confirmed in laboratory studies in which emission fluxes of HCHO were determined for samples directly collected from the canyon and from the tunnel. These HCHO emission fluxes were significant and e.g. one order of magnitude higher compared to theoretical deposition fluxes of NO_x under the prevailing atmospheric canyon conditions. Further laboratory experiments demonstrated that HCHO was formed at least in part by photocatalytic degradation of organic additives in the cementitious materials used. Although the emissions were found to decrease with irradiation time, they were still significant after ca. 1000 h irradiation time. Thus for the future, photocatalytic materials should be optimized also for minimum emissions of harmful pollutants.

4. Conclusions

4.1 Tunnel Study

During the period June 2011 – January 2013, three field campaigns were organized in the Leopold II tunnel in Brussels, Belgium, in which photocatalytic cementitious coating materials were applied on the side walls and ceiling of tunnel test sections of 70 m and 160 m length using artificial UV lighting systems in two of the campaigns. In contrast to first estimations based on laboratory data of untreated photocatalytic samples and the only available tunnel study to date, the results indicate no significant reduction of NO_x. As an upper limit a photocatalytic NO_x reduction of ≤2 % was observed for the 160 m long active section. Additional laboratory experiments showed a serious de-activation of the photocatalytic material under the heavily polluted tunnel conditions. In addition, the UVA irradiance (0.6 and 1.6 W m^-2) was below the targeted values, which contributed to the de-activation phenomenon and further reduced the photocatalytic activity. Simple model calculations based on a first order reaction kinetics using experimental reactivity data on tunnel samples indicate an upper limit photocatalytic NO_x remediation of only 0.4 %. This result is in excellent agreement with the experimental tunnel results of ≤2 %. In contrast, under ideal experimental conditions (UVA: 10 W m^-2, WS: 1 m s^-1, r.h.: 50 %) and ignoring the observed passivation of the surfaces, a NO_x reduction of 20 % might have been possible in the 160 m tunnel section. However, especially the passivation is a serious problem, which should be considered for future standard test methods concerning photocatalytic materials to ensure the successful application of photocatalysis in the urban environment. In conclusion, careful characterization of the tunnel conditions, quantification of possible deactivation of the photocatalytic material in the tunnel, simple upper limit model calculations of the expected NO_x reduction and a cost-benefit analysis is recommended, before the application of photocatalytic materials and concomitant UV lighting systems in road tunnels.

4.2 Canyon Study

In the present canyon study photocatalytic remediation of nitrogen oxides (NO_x), ozone (O₃), volatile organic compounds (VOCs) and particles was studied in an artificial street canyon in comparison to a similar co-located reference site. No significant photocatalytic remediation could be observed for the investigated trace species. For NO_x an upper limit remediation of ≤2% was derived. This result cannot be explained by potential deactivation of the used surfaces as observed in the tunnel, but is due to transport limitations of the pollutants towards the active surfaces. The high photocatalytic NO_x remediation results from some previous studies can be explained by differences in the geometry of the field sites, in sampling positions, in sampling periods and by general differences between the active and reference sites. If all these factors are considered and results from different studies are extrapolated to realistic urban conditions, a diurnal averaged NO_x reduction in the same range as the upper limit of the present study is estimated (~2%), leading to a consistent picture of the possible urban air pollution remediation by photocatalysis. Thus, this technique alone will not solve the urban NO₂ problem in European cities and should be considered only as one measure besides the others. The lower photocatalytic remediation obtained in the present study should not be considered as stop criteria for this innovative technique, but should be compared to other measures aimed to improve urban air quality on a cost-benefit analysis basis. Here, also more modern combustion technologies (e.g. EURO standards) or implementations of low emissions zones resulted in a reduction of the urban NO₂ levels by only a few percent at maximum in the past. 

Acknowledgements

The authors gratefully acknowledge the financial support of the European Commission through the Life+ grant LIFE 08 ENV/F/000487 PHOTOPAQ. 

References

[1] PhotoPAQ. European Life+ project PhotoPAQ. 2010-2014, http://photopaq.ircelyon.univ- lyon1.fr/

[2] Carslaw DC, Beevers SD, Bell MC. Risks of Exceeding the Hourly EU Limit Value for Nitrogen Dioxide Resulting from Increased Road Transport Emissions of Primary Nitrogen Dioxide. Atmos Environ 2007; 41:2073-82.

[3] Kurtenbach R, Kleffmann J, Niedojadlo A, Wiesen P. Primary NO₂ Emissions and their Impact on Air Quality in Traffic Environments in Europe. Environ Sci Europe 2012; 24(21):1- 8.

[4] Air pollution by ozone across Europe during summer 2013: Overview of exceedances of EC ozone threshold values: April–September 2013. EEA Technical report No 3/2014, ISSN 1725-2237.

[5] Air quality in Europe — 2014 report. EEA Report No 5/2014, ISSN 1725-9177.

[6] Finlayson-Pitts BJ, Pitts Jr JN: Chemistry of the Upper and Lower Atmosphere, Academic Press, 2000.

[7] Beekmann M, Vautard R. A modelling study of photochemical regimes over Europe: robustness and variability. Atmos Chem Phys 2010, 10:10067-84.

[8] Maggos Th, Plassais A, Bartzis JG, Vasilakos Ch, Moussiopoulos N, Bonafous L. Photocatalytic Degradation of NO_x in a Pilot Street Canyon Configuration using TiO₂-Mortar Panels. Environ Monit Assess 2008; 136:35-44.

[9] Chen J, Poon C-S. Photocatalytic construction and building materials: From fundamentals to applications. Build Environ 2009; 44:1899-1906.

[10] Ohama Y, Van Gemert D. Application of Titanium Dioxide Photocatalysis to Construction Materials. State-of-the-Art Report of the RILEM Technical Committee 194-TDP, Springer, XII; 2011, p 48.

[11] Chen H, Nanayakkara CE, Grassian VH. Titanium Dioxide Photocatalysis in Atmospheric Chemistry. Chem Rev 2012; 112:5919-48.

[12] Schneider J, Matsuoka M, Takeuchi M, Zhang J, Horiuchi Y, Anpo M, Bahnemann DW. Understanding TiO₂ Photocatalysis: Mechanisms and Materials. Chem. Rev. 2014, 114: 9919-86.

[13] Pichat P, Disdier J, Hoang-Van C, Mas D, Goutailler G, Gaysse C. Purification/deodorization of indoor air and gaseous effluents by TiO₂ photocatalysis. Catal Today, 2000, 63:363–9.

[14] Strini A, Cassese S, Schiavi L. Measurement of benzene, toluene, ethylbenzene and oxylene gas phase photodegradation by titanium dioxide dispersed in cementitious materials using a mixed flow reactor. Appl Catal B, Environ, 2005, 61(1-2):90-97.

[15] De Richter R, Caillol S. Fighting global warming: The potential of Photocatalysis against CO₂, CH₄, N₂O, CFCs, tropospheric O₃, BC and other major contributors to climate change. J Photochem Photobiol C Photochem Rev, 2011; 12:1-19.

[16] Costa A, Chiarello GL, Selli E, Guarino. Costa, A. Effects of TiO₂ based photocatalytic paint on concentrations and emissions of pollutants and on animal performance in a swine weaning unit. J Environ Manag, 2012, 96(1):86-90.

[17] Chang KL, Sekiguchi K, Wang QY Zhao F. Removal of Ethylene and Secondary Organic Aerosols Using UV-C_{254 + 185 nm} with TiO₂ Catalyst. Aerosol Air Qual Res, 2013, 13(2):618- 26.

[18] Laufs S, Burgeth G, Duttlinger W, Kurtenbach R, Maban M, Thomas C, Wiesen P, Kleffmann J. Conversion of Nitrogen Oxides on Commercial Photocatalytic Dispersion Paints. Atmos Environ 2010; 44:2341-9.

[19] Gustafsson RJ, Orlov A, Griffiths PT, Cox RA, Lambert RM. Reduction of NO₂ to Nitrous Acid on Illuminated Titanium Dioxide Aerosol Surfaces: Implications for Photocatalysis and Atmospheric Chemistry. Chem Commun 2006; 3936-8.

[20] Ndour M, D’Anna B, George C, Ka O, Balkanski Y, Kleffmann J, Stemmler K, Ammann M. Photoenhanced Uptake of NO₂ on Mineral Dust: Laboratory Experiments and Model Simulations. Geophys Res Lett 2008; 35:L05812.

[21] Beaumont SK, Gustafsson RJ, Lambert RM. Heterogeneous Photochemistry Relevant to the Troposhere: H₂O₂ Production during the Photochemical Reduction of NO₂ to HONO on UV-Illuminated TiO₂ Surfaces. ChemPhysChem 2009; 10:331-3.

[22] Monge ME, D’Anna B, George C. Nitrogen Dioxide Removal and Nitrous Acid Formation on Titanium Oxide Surfaces – An Air Quality Remediation Process? Phys Chem Chem Phys 2010; 12:8991-8.

[23] Pitts Jr JN. Formation and Fate of Gaseous and Particulate Mutagens and Carcinogens in Real and Simulated Atmospheres. Environ Health Perspect 1983; 47:115-40.

[24] Monge ME, George C, D’Anna B, Doussin J-F, Jammoul A, Wang J, Eyglunent G, Solignac G, Daële V, Melluki A. Ozone Formation from Illuminated Titanium Dioxide Surfaces. J Am Chem Soc 2010; 132:8234-5.

[25] Salthammer T, Fuhrmann F. Photocatalytic Surface Reactions on Indoor Paint. Environ Sci Technol 2007; 41:6573-8.

[26] Auvinen J, Wirtanen L. The Influence of Photocatalytic Interior Paints on Indoor Air Quality. Atmos Environ 2008; 42:4101-12.

[27] Geiss O, Cacho C, Barrero-Moreno J, Kotzias D. Photocatalytic degradation of organic paint constituents – formation of carbonyls. Build Environ 2012; 48:107-12.

[28] Fraunhofer. Clean Air by Airclean^®. 2010,
http://www.ime.fraunhofer.de/content/dam/ime/de/documents/AOe/2009_2010_Saubere%20Luft%20durch%20Pflastersteine_s.pdf.

[29] PICADA. European PICADA Project, GROWTH Project GRD1-2001-40449. 2006,
http://www.picada-project.com/domino/SitePicada/Picada.nsf?OpenDataBase.

[30] Ballari MM, Brouwers HJH. Full scale demonstration of air-purifying pavement. J Hazard Mat 2013; 254-255:406-14.

[31] Guerrini GL, Peccati E. Photocatalytic Cementitious Roads for Depollution. In: International RILEM Symposium on Photocatalysis, Environment and Construction Materials; 8-9 October 2007, Florence, Italy, 179-86.

[32] IPL. Dutch Air Quality Innovation Programme concluded: Improved Air Quality with Coating of Titanium Dioxide not Demonstrated. 2010, http://laqm.defra.gov.uk/documents/Dutch_Air_Quality_Innovation_Programme.pdf.

[33] Tera. In situ study of the air pollution mitigating properties of photocatalytic coating, Tera Environement, (Contract number 0941C0978), Report for ADEME and Rhone-Alpe region, France. 2009, http://www.air-rhonealpes.fr/site/media/telecharger/651413.

[34] Jacobi, S. NO₂-Reduzierung durch photocatalytisch wirksame Oberflächen? Modellversuch Fulda. Hessisches Landesamt für Umwelt und Geologie – Jahresbericht 2012; http://www.hlug.de/fileadmin/dokumente/das_hlug/jahresbericht/2012/jb2012_059- 066_I2_Jacobi_final.pdf.

[35] Guerrini GL. Photocatalytic Performance in a City Tunnel in Rome: NO_x Monitoring Results. Constr Build Mater 2012; 27:165-75.

[36] Gallus M, Akylas V, Barmpas F, Beeldens A, Boonen E, Boréave A, Cazaunau M, Chen H, Daële V, Doussin JF, Dupart Y, Gaimoz C, George C, Grosselin B, Herrmann H, Ifang S, Kurtenbach R, Maille M, Mellouki A, Miet K, Mothes F, Moussiopoulos N, Poulain L, Rabe R, Zapf P, Kleffmann J: Photocatalytic Nitrogen Oxides Abatement Results from the Leopold II Tunnel in Brussels. Build Environ 2015; 84:125-33.

[37] Boonen E, Akylas V, Barmpas F, Boréave A, Bottalico L, Cazaunau M, Chen H, Daële V, De Marco T, Doussin J-F, Gaimoz C, Gallus M, George C, Grosselin B, Guerrini GL, Herrmann H, Ifang S, Kleffmann J, Kurtenbach R, Maille M, Manganelli G, Mellouki A, Miet K, Mothes F, Moussiopoulos N, Poulain L, Rabe R, Zapf P, Beeldens A. Construction of a Photocatalytic De-polluting Field Site in the Leopold II Tunnel in Brussels. 2014, J Environ Manage, submitted.

[38] Gallus M, Ciuraru R, Mothes F, Akylas V, Barmpas F, B