Prior to the detailed characterization and photocatalytic investigation, a screening photocatalytic study was carried out, covering an extensive range of minority oxides and their concentrations. The parameters were chosen according to the ISO standard (22,197–1:2007) apart from the length of the experiment, which was shortened from 5 h to 1.5 h (modified ISO conditions). The results of the screening study are shown in Fig. 3 (full results in Table S1 in Supplementary Information).

Results of initial screening of various composite photocatalysts (modified ISO conditions, as described in Section 2.5)

Composites containing MnO2 and WO3 < 100 nm were excluded from further testing due to their poor performance. From each group of composites containing individual oxides, the best performing composites were chosen for further testing based on these premises: a) the secondary photocatalyst is present in the composite material in a concentration lower than 20% and b) the composite material has the best photocatalytic activity possible.

Prediction of the photocatalytic reactions is as follows[3]:

$$NO \stackrel_}_\right)}_+_O \rightarrow HN_\ ,$$

(1)

$$3 _+ _O \rightarrow 2 _+NO,$$

(2)

$$2 NO+ _\rightarrow 2 _$$

(3)

$$2 _+ _O \stackrel_} HONO+ _$$

(4)

$$HONO \stackrel ^+NO.$$

(5)

The photocatalytic oxidation of nitric oxide produces nitrogen dioxide, which is subsequently oxidized by the attack of water or OH radicals into the final product, HNO3. The final product of photocatalytic oxidation captured on the photocatalytic surfaces was exclusively nitrites (\(}}_^\)) while the concentrations of HONO and \(}}_^\) were below the detection limit [10].

The main criteria for evaluation of photocatalytic activity were the rate of NOx degradation and selectivity. The less NO2 was formed during the experiment, the better the performance. The selected composite photocatalysts along with the concentration of the secondary photocatalyst contained and comparison of their photocatalytic activity in steady state are shown in Fig. 4 (full results in Table S2 in Supplementary Information).

Results of repeated optimized screening experiments (modified ISO conditions, as described in Section 2.5)

Based on the repeated experiments, four composites with the best performance were chosen, whose composition is shown in Table 2. Concerning the composite photocatalyst containing WO3, only the better performing type was chosen for further testing, namely the one with 23–65 nm particle size. Figure 5 shows the photocatalytic performance of the selected materials.

The measurement of the 4 selected composites showing a the depletion of NO and formation of NO2, b depletion of NOx (modified ISO conditions, as described in Section 2.5)

The optimal concentration chosen is in agreement with previously published research. Chen et. al. [11] reported an optimal concentration of WO3 in a WO3/TiO2 catalyst prepared by ball-milling to be 3 wt%. Liao et. al. [12] prepared a doped TiO2/ZnO photocatalyst using a sol–gel method and reported maximal activity at 4 wt% of ZnO. Chang et. al. [13] reported maximal activity for monolithic Fe2O3/TiO2 photocatalyst prepared by sol–gel method when Fe/Ti molar ratio was 0.03 (cca 3.5 wt%).

Phase composition of pristine P25 and used metal oxides was verified by XRD (Table 3, Figure S1—Supplementary Information). The results show that P25 powder consists of two allotropic modification of titanium dioxide anatase and rutile. From used metal oxide powder only V2O5 oxide is mono-phase, consisting of mineral phase scherbinaite. The rest of used metal oxide contain two phases. The main mineral phase of Fe2O3 powder was surprisingly maghemite (γ-Fe2O3), and the second phase was hematite (α-Fe2O3). Maghemite is thermodynamically less stable compared to hematite [14], the dominance of maghemite in the used metal oxide is resulting from nano-size of its crystals [15]. WO3 powder consists mainly of monoclinic WO3 (m-WO3), the second phase was orthorhombic form of WO3 (o-WO3). The dominant mineral phase in ZnO powder was zincite, and minor phase hydrozincite (Zn5(CO3)2(OH)6). The values of wt% composition of studied pristine additive MOx are shown in Table 3.

The phase analysis of composite photocatalysts is summarized in Table 4 and depicted in Fig. 6. The dominant phases in all composites are anatase and rutile. In the P25/Fe2O3 composite photocatalyst only maghemite was present. Hematite was under detection limit in the composite photocatalyst, because of its very low concentration in pristine Fe2O3 dopant. The same situation was found with P25/WO3 and P25/ZnO photocatalysts, where the present phases were only monoclinic WO3 and zincite, respectively. In the P25/V2O5 photocatalyst, scherbinaite was identified.

Phase identification and comparison of diffraction patterns of P25 and composite photocatalysts. Only the MOx peaks are marked

The crystallite size of anatase and rutile was in the range between 50 and 60 nm both in the pristine P25 and as the part of the composite photocatalyst. The largest crystals were identified in V2O5 powder (4 × larger in comparison with rutile and anatase crystallites), while the smallest ones were found for ZnO powder (20–30 nm). The crystallite sizes of mineral phases in Fe2O3 and WO3 powders were comparable with the range between 60 and 80 nm. As a result of the preparation of composite photocatalysts a reduction in the crystallite size was observed. This effect was the most visible for Fe2O3 and ZnO mineral phases maghemite and zincite, respectively. The average crystallite size was about 10 nm.

On Figure S2 (Supplementary information) are shown the SE images of the surface morphology of composite photocatalysts, SEM–EDS chemical analysis of metal oxide particle and SEM–EDS chemical distribution maps of composite photocatalysts. P25/Fe2O3 and P25/WO3 show rather fine-grained homogeneously distributed particles, form small clusters. In contrast, P25/V2O5 and P25/ZnO form relatively large agglomerates on the surface, which was confirmed by the research of Devi et. al. [16] and Ong et. al. [17].

The diffuse reflectance spectra were used for the evaluation of the optical properties of used metal oxides and composite photocatalysts (Fig. 7a, b). According to Ohtani [18] and Zanatta [19], the obtained spectra were plotted using Kubelka–Munk function (K–M) and the optical band gap was determined using Eq. (1).

$$(K-M \, h\nu)^} ( - _)$$

(1)

where K–M represent the Kubelka–Munk function, h Planck constant, ν oscillation frequency, n constant relating to a mode of transition and Eg optical band gap. The optical band gap energy was determined from the plot of \(}^}}}\) against \(\) (using \(} \, = \, \frac}\) for direct transition and \(} \, = \, \) for indirect transition), where the linear part of the plot intersects the x-axis.

UV–Vis diffuse reflectance spectra of a pristine P25 and metal oxides, b pristine P25 and composite photocatalysts. Comparison of band gap energy c pristine P25 and Fe2O3 and d pristine P25 and composite photocatalyst P25/Fe2O3

The adsorption band edges of pristine metal oxides and P25 are on Fig. 7a. The adsorption edge for P25 can be found around 390 nm. The adsorption edge for ZnO was observed around 410 nm followed by WO3 (470 nm) and V2O5 (590 nm). Two adsorption band edges were found for Fe2O3 (640 and 710 nm).

Except for the P25/ZnO photocatalyst, a significant decrease in the adsorption edge position was observed for the rest of the composite photocatalysts. Similar position of the adsorption edge of the pristine ZnO with P25 caused almost no change in the adsorption edge position of P25/ZnO (415 nm). For the remaining composites, however, a change in optical properties was observed, even when the concentrations of secondary oxides were rather low, only 1–2 wt%. P25/WO3 composite photocatalyst provide the value of the adsorption edge around 418 nm, which assumed the similar optical properties with P25 and P25/ZnO. In contrast, two adsorption edges were provided by V2O5 (440 and 560 nm) and Fe2O3 (460 and 610 nm), suggesting the increase of the activity or selectivity of composite photocatalysts also in the area of visible light [2, 5].

On Fig. 7c is shown the comparison of the band gap energy value of pristine P25 and Fe2O3. Both P25 and Fe2O3 are direct gap semiconductor materials [20]. The band gap energy of Fe2O3 is observed at the position of 2.20 eV. Figure 7d shows the shift in the bang gap energy value in the composite photocatalyst P25/Fe2O3. The band gap energy of P25/Fe2O3 increased up to 2.95 eV as a result of the majority concentration of P25 (99 wt%) compared to Fe2O3 (1 wt%) in the composite photocatalyst. However, even small concentration of metal oxide affected the optical properties of the composite photocatalyst.

The results from the photocurrent study show non-negligible differences between the surface photoactivity of studied P25/MOx materials, which is reflected in their photocatalytic properties. Figure 8 shows the dark current–voltage (I–V) characteristics of the P25, P25/ZnO, P25/V2O5, P25/Fe2O3 and P25/WO3 in N2 gas. All samples show asymmetric I–V characteristic. The observed transition in the power dependence of current in the I–V curve is likely to arise from charge carriers trapped in the surface defect states that contribute to the high current at higher bias voltage. The high current at higher bias voltage can be explained by the space-charge-limited (SCLC) current mechanism [21]. This behavior is consistent with previous reports on doped metal oxide thin film based resistive switching devices [22,23,24,25] and it is explained on the basis of presence of oxygen vacancy-related traps within the bandgap of the materials. Under a negative bias voltage, oxygen vacancies with positive charges migrate away from the interface between contact and the P25/MOx, which widens the depletion layer, resulting in high resistivity. On the other hand, with positive bias voltage, the oxygen vacancies start moving toward the interface, resulting in higher current [22, 25, 26].

Dark current–voltage characteristic of P25/MOx layers

Based on the above results and analysis, we further investigated the photogenerated charge kinetics of electron–hole pairs in the generation, transport, and recombination processes. Figure 9 presents transient photocurrent responses of P25/MOx layers. Semi-logarithmic plot time-dependent photo-response to light excitation using a focused LED beam (370 ± 10 nm) as an irradiation source under dry synthetic air gas flow. The bias voltage between Pt-IDEs was kept constant at 2 V. The irradiation source was switched light and dark periodically at 10 ms intervals. The photocurrent initially grows very fast and then slowly increased with time and saturated. The photo-response of P25, P25/ZnO, P25/Fe2O3 and P25/WO3 shows a bi-exponential growth and bi-exponential decay behavior. The time-dependent growth behavior of the photocurrent curve is fitted with Eq. 2.

$$}_}\left(}\right)=}_+}_\left(1 - }^}_}}\right)-}_}}^}_}},$$

(2)

where I1, A1 and A2 are positive constants. Calculated time constants from fittings are t1 = 19 ms and t2 = 45 ms for the P25 layer. For the P25/ZnO layer, the photocurrent growth and decay time constants are t1 = 21.3 ms and t2 = 44 ms, t1 = 19 ms and t2 = 43 ms for P25/Fe2O3 and t1 = 11 ms and t2 = 31 ms for P25/WO3 indicating a very rapid photocurrent growth initially followed by slow decay process. The photocurrent of P25/MOx layers consists of two parts: a) a rapid process of photogeneration and recombination of electron/hole pairs and b) slow process attributed to the surface adsorption and photodesorption of oxygen molecules [27]. In this process, the electron–hole generation is the only source for current carriers and all other processes decreased the carriers (chemisorption of oxygen). Physisorbed oxygen causes almost no change in the material’s electrical properties due to weak electrostatic interaction through van der Waals forces.

Transient photocurrent responses of P25/MOx layers on a Pt-IDEs: a P25/Fe2O3, b P25/V2O5, c P25/WO3, d P25/ZnO, e P25

In Table 5 are summarized the results from the current measurement (values of dark and light current and current light/dark ratio) for pristine P25 and studied composite photocatalysts.

Figure 10 shows the photocatalytic activity of selected composites under various conditions in gas phase depicted as relative reaction rate rrel. Relative reaction rates were calculated as the ratio of the reaction rate of the composite to the reaction rate of P25 under the given reaction conditions (Eq. 3).

The results of additional experiments depicted as relative reaction rates rrela P25/Fe2O3, b P25/V2O5, c P25/WO3, d P25/ZnO

$$}_}}_}= \text \frac}}_}}}}_}}\times 100\%.$$

(3)

For the degradation of NO and total degradation of NOx, rrel of more than 100% is favorable. On the other hand, for the formation of NO2, rrel of less than 100% is favorable.

When visible light source was used, none of the composites showed higher r(NO). Regarding the comparison of r(NOx), the P25/ZnO, P25/Fe2O3 and P25/WO3 composites show roughly comparable activity, about half of P25 values (Fig. 10). However, P25/WO3 composite shows different selectivity since the degradation of NO is slower but the formation of NO2 is not as high as with the other composites (r(NO2):r(NO) 0.60 for P25/WO3 vs. 0.78 for P25). Overall, the comparison shows that the addition of small amounts of transition metal oxide nanoparticles did not improve the photocatalytic efficiency in visible light. The P25/V2O5 composite shows low NO conversion and a relatively low removal rate r(NOx), but virtually no NO2 is formed on them. This favorable change in selectivity is probably related to the suppression of the most active centers.

P25/ZnO and P25/Fe2O3 composites show better photocatalytic behavior than pristine P25 under simulated indoor conditions (rrel(NOx) 143.7% and 115.1%, respectively). For these composites, the increase in the reaction rate r(NO) is not accompanied by an increase in the undesired formation of NO2, which leads to an increase in the overall reaction rate of NOx degradation. The P25/WO3 composite shows a reduction in r(NO) and NO2 production compared to P25, leading to a lower r(NOx). However, its selectivity is improved (r(NO2):r(NO) 0.50 for P25/WO3 vs. 0.58 for P25). The activity of P25/V2O5 composite was severely diminished (rrel(NOx) 15.1%) with worsened selectivity (r(NO2):r(NO) 0.63).

When simulated outdoor conditions were applied, the P25/Fe2O3 composite showed improvement in comparison to pristine P25 in both higher r(NO) and lower r(NO2), resulting in a better photocatalytic performance in total reduction of NOx (115.5%, r(NO2):r(NO) 0.48 for P25/Fe2O3 vs. 0.54 for P25). The P25/WO3 and P25/ZnO composites both showed lower photocatalytic activity than pristine P25; however, P25/WO3 was more active than P25/ZnO under these conditions in comparison with simulated indoor experiments. The selectivity under the simulated outdoor conditions significantly worsened for both P25/WO3 (r(NO2):r(NO) 0.65) and P25/ZnO (r(NO2):r(NO) 0.68). These results are in good agreement with optical properties of studied composites. The P25/V2O5 composite showed activity at the detection limit, which can be caused by dissolution of V2O5 in the presence of 50% humidity, resulting significant decrease the photoactivity [28].

Overall, P25/Fe2O3 turned out to be the most promising, with improved selectivity and higher NOx degradation rate than P25 under both simulated setups. However, all the composites have potential to be further used with necessary optimization.

The absolute measured values and calculated reaction rates can be found in Table S3 for VIS experiments, Table S4 for simulated indoor experiments, and Table S5 for simulated outdoor experiments (Supplementary Information).