FLU (purity > 98%) was provided by J&K Chemical Ltd. Methanol and formic acid (FA) were purchased from Merck (Darmstadt, Germany) and were of HPLC grade. Other chemicals used in this study such as Ti(SO₄)₂, HCl, NaOH, p-benzoquinone (p-BQ), urea (CO(NH₂)₂, U), ammonium nitrate (NH₄NO₃, AN), ammonium chloride (NH₄Cl, AC), AgNO₃ and tert-butyl alcohol (TBA) were analytical reagent grade and used without further purification. A Milli-Q Plus system (Millipore, Bedford USA) was employed to prepare ultrapure water (> 18.2 MΩ cm), which was used throughout all experiments.

N-doped TiO₂ catalysts were synthesized by using precipitation combined with thermal decomposition method. The specific steps are as follows: First, using Ti(SO₄)₂ (purity ≥ 96%) as a Ti precursor and NaOH (2 M) as precipitator, after the pH of the mixed solution rose to 10 in the precipitation process, then the precipitation cakes were filtrated at vacuum condition, washed using ultrapure water four times and dried at 80°C overnight. The resultant sample was crushed to obtain fine powder. Second, 1.0 g of the sample powder obtained above and 1.0 g of nitrogenous compounds including CO(NH₂)₂, NH₄Cl or NH₄NO₃ were mixed uniformly and then milled thoroughly in an agate mortar, respectively. The resultant mixture was transferred into porcelain crucible for calcination at 450°C for 2 h at a heating rate of 4 °C/min in a muffle furnace, and then cooled down to room temperature. The formed product was further crushed. Thus, N-doped TiO₂ catalysts were obtained and signified as TiO₂/U, TiO₂/AC, or TiO₂/AN, respectively. For comparison, the undoped TiO₂ was also synthesized using the same method in the absence of dopants.

Scanning electron microscopic (SEM) instrument (FEI, Netherlands) was used to observe the morphology of the prepared samples. The powder crystalline structures of the samples were obtained using Smart lab 9 X-ray diffractometer (Rigaku, Japan) with Cu Ka (0.15406 nm) X-ray source. Raman spectra of these samples were collected using HORIBA HR800 Raman spectrometer and the intensity data were collected from 100 to 800 cm⁻¹. The chemical states of the samples were investigated by carrying out X-ray photoelectron spectroscopy (XPS) determination employing a PHI 5000 Versa Probe system (ULVAC-PHI, Japan) with Al Kα radiation source. The C 1s level at 284.6 eV was used as an internal standard to correct the shift of binding energy due to relative surface charging. The UV–Vis diffuse reflectance spectra (UV-Vis-DRS) of the samples were obtained via UV-visible spectrophotometer (PerkinElmer, Lambda-950) employing BaSO₄ as a reference material.

All experiments were carried out in a XPA-7 multi-tube agitated photochemical reactor (Xujiang Technology Co., Ltd. Nanjing, China). The photocatalytic activity of the samples was assessed by degrading FLU in aqueous solution under 500 W xenon light irradiation, and its light intensity at the sample position was about 48.0 mW cm⁻². The photochemical reactor is equipped with water circulation cooling device to control the reaction temperature at 25 ± 1°C and further details of XPA-7 can be found in our previous work [36]. In a typical experiment, a 10 mg of sample was added into a quartz photo-reactor containing 50 mL FLU solution (10 mg L⁻¹). Prior to irradiation, the obtained suspension was stirred magnetically for 30 min in the dark to ensure the establishment of adsorption/desorption equilibrium. At prescribed time intervals, 1 mL sample solution was extracted and immediately filtered through a 0.22 μm filter into a brown HPLC vial for further analysis. In the experiment, the initial pH value of the reaction solution was adjusted to pH = 7 ± 0.2 using HCl and NaOH. All experiments were performed in triplicate. To estimate the catalytic stability and reusability of the prepared N-doped TiO₂ catalysts, the recycling experiments for photocatalytic activity measurements of the catalyst with the highest catalytic performance were conducted. The catalyst was collected by centrifugation after each run and then washed with ultrapure water three times. The catalyst solids were left and the fresh FLU reaction solution was added to the reaction tube, followed by the next run.

An HPLC UV (Agilent Technologies, Series 1200) system equipped with an Agilent ZORBAX SB-C18 column (5 μm, 4.6×150 mm) was used for monitoring FLU concentration. The maximum absorbance wavelength for FLU was 238 nm. The isocratic elution was methanol and 0.3% formic acid in water (v/v = 70:30) at a flow rate of 1 mL min⁻¹. Total organic carbon (TOC) determinations were carried out in a Shimadzu 5000A TOC analyzer.

The degradation products of FLU were analyzed by high resolution liquid chromatography time of flight mass spectrometry (LC-QTOF-MS) using ultra performance liquid chromatography tandem-mass spectrometry (UPLC-MS/MS, Waters Corp., USA). A BEH C18 column (2.1×50 mm, 1.7 μm) was used for chromatographic separation. The mobile phase consisting of 0.1% formic acid in water (A) and acetonitrile (B) was eluted at 0.3 mL min⁻¹. Mass spectrometric analysis was carried out with a G2-XS-QTOF MS operating in a positive ion mode by an electrospray ion source. Mass range of TOF MS was set at m/z from 100 to 600. The other mass parameters were set as follows: Capillary voltage: 3.0 kV, source temperature: 100°C, desolvation gas flow rate: 800 L h⁻¹ and desolvation temperature: 350°C.

In order to evaluate the role of the main reactive species including hydroxyl radical (^•OH), superoxide radical (O₂^•−), photogenerated electrons (e⁻) and positive hole (h⁺) during the photocatalytic reactions, four scavengers including TBA, p-BQ, AgNO₃ and FA were added to the FLU solutions containing catalyst to capture ^•OH, O₂^•−, e⁻, and h⁺, respectively [37–39]. The detailed step was as follows: 10 mg of as-prepared catalyst sample was dispersed in the 50 mL of FLU aqueous solution (10 mg L⁻¹), and then 38.3 mM TBA, 0.38 mM p-BQ, 4.0 mM AgNO₃ or 1.0 g L⁻¹ FA was added into reaction system, respectively. The choice of the concentrations of TBA, p-BQ, AgNO₃ and FA was referred to a previous report [36]. The analytical method was the same as that used in the evaluation of photocatalyst activity mentioned above.

The surface morphologies of the four as-synthesized samples were examined using SEM technique and the results are shown in Fig. 1. The surface topography of these samples varied widely. The surface of pure TiO₂ particles appeared rough and not uniform (Fig. 1(a)). Compared to pure TiO₂, the TiO₂/U sample particles were tightly packed and layered, resulting in bad adsorption capacity (Fig. 1(b)). For TiO₂/AC sample, although its particle size was obviously smaller than those of pure TiO₂ and TiO₂/U, there was agglomeration in it (Fig. 1(c)). Compared to the three samples mentioned above, the particles of TiO₂/AN sample (Fig. 1(d)) exhibited spherical, highly dispersed and evener particle size distribution in the range of 9–13 nm. Thus, it was expected to have high photocatalytic activity.

XRD measurements were performed to identify the crystal structures of the four samples (pure TiO₂, TiO₂/U, TiO₂/AC and TiO₂/AN) and the results are shown in Fig. 2(a). The XRD spectra results of these samples indicated that the main diffraction peaks at 25.3°, 37.9°, 48.1°, 54.0°, 55.1° and 62.9° are attributed to (101), (004), (200), (105), (211) and (204) lattice planes of anatase of TiO₂, agreeing well with the standard JCPDS No. 21-1272, respectively. Specifically, all diffraction peaks of pure TiO₂ and TiO₂/U samples belong to the crystalline phase of anatase of TiO₂. However, for TiO₂/AC and TiO₂/AN samples, some new diffraction peaks were observed in their XRD patterns. The new diffraction peaks at about 27.4°, 31.7°, 45.4°, 56.5° and 66.2° were observed along with the anatase phase peaks for TiO₂/AC sample, which is attributed to the (111), (200), (220), (222) and (400) planes of NaCl (JCPDS No. 75-0306), respectively. At the same time, for TiO₂/AN sample, new diffraction peaks at about 29.4°, 31.8°, 35.4° and 42.5° were observed along with the anatase phase peaks and they may belong to the planes (104), (006), (110) and (202) of NaNO₃ (JCPDS No. 72-1213), respectively. It is worth noting that NaNO₃ and NaCl were generated by the reactions between Na⁺ and the anions in the dopants NH₄NO₃ and NH₄Cl, respectively, where Na⁺ comes from the precursor NaOH used in the preparation of TiO₂. In addition, two diffraction peaks of NH₄NO₃ (JCPDS No. 74-0972) were observed at 22.8 ° and 27.8° for (111) and (210) planes, respectively. These give the evidence that NH₄NO₃ was doped successfully into the crystal structure of TiO₂, including N⁵⁺ (NO₃⁻) and N³⁻ (NH₄⁺), which is consistent with the conclusion of XPS analysis discussed later. It may be one of the reasons for the improvement of the photocatalytic activity of TiO₂/AN sample because NO₃⁻ and NH₄⁺ can trap photogenic electrons (e⁻) or photogenic holes (h⁺) to prevent e⁻/h⁺ pairs recombination and this helps to promote the charge separation. Based on the Scherrer’s formula and XRD peak at about 25.3° of (101) plane, the average crystallite sizes of TiO₂, TiO₂/U, TiO₂/AC and TiO₂/AN samples were calculated, and they were 19.9, 16.1, 14.2 and 11.0 nm, respectively. These results showed that the average particle size of TiO₂/AN sample is less than that of other three samples. Thus, it illustrates that NH₄NO₃ doping can inhibit better the crystal growth of anatase TiO₂, which helps to improve the photocatalytic activity of TiO₂.

Fig. 2(b) displays the Raman spectra of TiO₂, TiO₂/U, TiO₂/AC and TiO₂/AN samples. Raman peak shifts at about 145, 397, 516, 639 cm⁻¹ stem from the characteristic Raman peaks of anatase phase [36, 40, 41], demonstrating that the main phase structure of the prepared samples was anatase. No peak attributed to nitrogen-containing phases was detected in the doped TiO₂ catalysts, indicating that nitrogen was highly dispersed on the TiO₂ surface. It can be observed that the peak at about 145 cm⁻¹ for the doped TiO₂ samples became wider in the following order TiO₂ < TiO₂/U < TiO₂/AC < TiO₂/AN, suggesting that their particles size decreases correspondingly and the results of Raman analysis are in good agreement with the XRD results (see Fig. 2(a)).

The surface composition and chemical states of elements in the NH₄NO₃-doped TiO₂ sample were further characterized by XPS and shown in Fig. 3. Fig. 3(a) presented the XPS full spectrum survey of the sample and shows that the species like C, N, O, Ti and Na are predominantly present in the wide XPS spectrum. Taking the C 1s peak (286.4 eV) as a benchmark, the obtained binding energies were corrected. The C 1s peak may come from exterior hydrocarbon from XPS instrument itself [1]. Based on the XPS data, the atomic concentrations of Ti, O, Na and N for the TiO₂/AN sample were estimated to be about 9.8, 47.8, 12.5 and 9.2 atom%, respectively. Na 1s peak at about 1,071.3 eV comes from the precipitant NaOH used during the synthesis of TiO₂ [42]. As shown in Fig. 3(b), two peaks located at 458.1 and 463.9 eV in Ti 2p XPS spectrum belong to Ti 2p_3/2 and Ti 2p_1/2, respectively. The two peaks have slightly lower binding values than those of classical spectrum of Ti⁴⁺ in TiO₂ [43], suggesting NO₃⁻ doping has negligible effect on the chemical valence status of Ti. Fig. 3(c) indicates that two N 1s peaks are observed at 407.32 eV and 403.35 eV, implying that there are two kinds of doped N species. The peaks at 407.32 eV is assigned to the binding energy of N⁵⁺ in NO₃⁻ ions, which confirms that NO₃⁻ can be incorporated into the TiO₂ lattice, the other at 403.35 eV is attribute to N³⁻ in NH₄⁺. The relative magnitude of their peak areas suggested that N⁵⁺ is main N doping species. As shown in Fig. 3(d), the XPS spectra of O 1s can be fitted into three different peaks at binding energy values of 532.9, 531.3 and 529.3 eV corresponding to surface chem-isorbed oxygen, oxygen vacancies and lattice oxygen respectively. The peak at 532.9 eV belongs to O atoms of NO₃⁻ ions, demonstrating successful preparation of NO₃⁻-doped TiO₂.

UV–Vis diffuse reflectance spectra (UV-Vis-DRS) of as-synthesized samples are indicated in Fig. 4. A weak absorption band between 400 and 520 nm was found for TiO₂/U, TiO₂/AC and TiO₂/AN, suggesting that the light absorption performance of TiO₂/U, TiO₂/AC and TiO₂/AN is better than that of TiO₂. On the other hand, the forbidden band width values (E_g) for TiO₂, TiO₂/U, TiO₂/AC and TiO₂/AN can be estimated using a Tauc plot based on the UV-Vis-DRS data, and the corresponding values are 3.22, 3.15, 3.10 and 3.01 eV, respectively. Therefore, N-doping can narrow the band gap of TiO₂ and enhance its performance for photocatalytic degradation of FLU, especially for TiO₂/AN. The same result was also found for urea or thiourea-doped TiO₂ prepared using the sol–gel technique, and the shift of the light absorption profile in N-doped TiO₂ may be associated to the presence of manifold surface states close to the valence band edge [35].

Four recycling experiments for degradation of FLU were performed to assess the photocatalytic stability and reusability of the TiO₂/AN catalyst. The specific method of reusing TiO₂/AN catalyst is as follows: 50 mL of FLU solution (10 mg L⁻¹) and 10 mg fresh catalyst was added to the reaction vessel, and the suspension was left to equilibrium adsorption for 30 min in the dark and then subject to 500 W Xe light irradiation for 4 hours. FLU degradation was monitored at prescribed time intervals and the first cycle was completed. The catalyst was collected by centrifugation and then washed with ultrapure water three times. After washing, the water was removed as much as possible, only the catalyst solids were left and the fresh FLU reaction solution of 50 mL was added to the reaction tube, followed by the second run and this procedure was repeated for three cycles. The results are depicted in Fig. 6. FLU was almost completely removed by using TiO₂/AN as catalyst after 4 hours irradiation at the first cycle, and FLU conversion efficiency remained at 91.8% with TiO₂/AN catalyst after four cycles. The reduction in catalytic performance of TiO₂/AN catalyst may originate from the catalyst loss due to washing process. Thus, the photocatalytic degradation of FLU by TiO₂/AN was affected slightly during four consecutive runs under 500 W Xe light irradiation, indicating good photocatalytic stability of the TiO₂/AN catalyst.

The reactive species quenching tests were performed to distinguish the major reactive components including ^•OH, O₂^•−, photogenerated holes (h⁺) and photogenerated electron (e⁻) during the photocatalytic degradation process of FLU with TiO₂/AN as catalyst. In the experiments, TBA, p-BQ, FA and AgNO₃ were used as scavenges to quench the related active species [37–39] and were separately introduced into the photocatalytic system (TBA for only free ^•OH, p-BQ for O₂^•−, FA for both h⁺ and total ^•OH, AgNO₃ for e⁻, respectively). As shown in Fig. 7(a), the removal of FLU was all inhibited by the addition of these scavengers separately in the FLU solution. Compared with the blank test, the addition of AgNO₃ in the FLU solution has slight inhibitory effect on the removal of FLU. However, the removal of FLU was obviously inhibited by the addition of FA and stronger inhibitory effect were also both observed after the addition of TBA and p-BQ. Their k_app values for photocatalytic degradation of FLU with or without scavenger were also calculated and displayed in Fig. 7(b). As observed in Fig. 7(b), the efficiency for photocatalytic degradation of FLU was inhibited in the order of FA > p-BQ > TBA > AgNO₃. These experimental results indicated that both h⁺ and ^•OH are major reactive species in this photocatalytic degradation system and O₂^•− also play an important role, while e⁻ play negligible roles.

The primary intermediates for photocatalytic degradation of FLU were determined by employing the MS² scan technique included in LC-TOF-MS. Based on MS data analysis, eight plausible degradation intermediates were detected in the positive mode. Fig. S2 indicates the proposed molecular structures of these intermediates and their fragmentation patterns. Table S1 summarizes their MS data and corresponding retention time. It can be found that small errors exist between the calculated and experimental m/z for these suggested molecular formulas. Therefore, the structural assignment for these intermediates is highly reliable, which makes it possible to propose the pathways for photocatalytic degradation of FLU.

Three plausible pathways for photocatalytic degradation of FLU employing TiO₂/AN as photocatalyst were proposed based on the geometric structures of the identified degradation intermediates and displayed in Fig. 8. In the first pathway (pathway I), a propane molecule was removed from the upper saturated heterocyclic ring, resulting in formation of the degradation intermediate P1, which occurs most probably due to attack of h⁺ on the saturated ring of FLU. The intermediate P1 was also detected in the previous report on direct photolysis of FLU in water [44]. The next pathway (pathway II) was initiated by the attack of ^•OH at C7 position of the saturated ring, leading to the formation of hydroxylated intermediate P2. The subsequent hydrogen abstraction of P2 produces another intermediate P3. The intermediates P2 and P3 were also found in the recent study on oxidative degradation of FLU using K₂FeO₄ as oxidant [45]. The third pathway (pathway III) was also triggered by ^•OH attack, but the attack position is most likely to be on the C3 atom of FLU according to the theoretically calculated electron densities of FLU and our previous research [36, 45], and the resultant intermediate P4 is the isomer of P1. Subsequent substitution of fluorine atom by ^•OH may occur for P4, resulting in the formation of dihydroxylated intermediate P5. In addition, P4 can also undergo further decarboxylation, demethylation, the opening of the quinolone ring or their combination to yield the intermediates P6, P7 or P8, respectively. The intermediates P4, P6, P7 and P8 were also identified in a recent report [45]. It should be noted that the above decarboxylation processes are likely to occur by a photo-Kolbe pathway with the direct participation of h⁺ as shown by the reaction (RCOO⁻ + h⁺ → RCOO^• → R^• + CO₂), while the opening of the quinolone ring may be initiated by the attack of superoxide radical (O₂^•−) on the C2–C3 double bond of the quinolone ring.

The ECOSAR program [46] was used to predict acute and chronic toxicities of FLU and proposed degradation intermediates (P1~P8) to fish, daphnids and green algae, and the results are summarized in Table S2. It should be noted that the predicted toxicity data of P1 were marked with “*” in the output result, suggesting that it might be too insoluble to predict the toxicity effect. Therefore, the predicted toxicity data of P1 are not listed in Table S2. It can be seen from Table S2 that FLU has chronic toxicity to three test organisms (1 mg L⁻¹ < ChV ≤ 10 mg L⁻¹) according to the Globally Harmonized System of Classification and Labelling of Chemicals, while the degradation intermediates of FLU don’t show toxicity in general. Thus, NH₄NO₃-doped TiO₂ could be employed as an adequate catalyst for the photocatalytic degradation of FLU in water and wastewater.

The photocatalytic degradation of FLU in the aqueous phase have been investigated employing commercialized or modified TiO₂ photocatalysts [8, 35, 47–49]. A comparison for photodegradation of FLU between the present study and these reports was made and is displayed in Table S3. The degradation efficiencies of FLU in these researches using pure anatase TiO₂, thiourea-TiO₂, urea-TiO₂, graphene-TiO₂ or TiO₂/ZnO/Sepiolite catalysts were found to be in the range of 55~100% in 60~240 min of solar-simulated irradiation or UV irradiation, and the TOC removal efficiencies ranged over 35.8~74%. While in the present study, the efficiencies for FLU or TOC removal reached 100% or 81.2 % respectively after 4 h of irradiation. In general, the results of this study are comparable to literature studies.