Photocatalytic water treatment has shown great potential as a cost-efficient technique for the removal of recalcitrant organic pollutants in water [1-6]. Although hundreds of new photocatalysts have been developed by using approaches such as engineered nanoarchitectures or band energy structures [7-13], TiO₂ is still the most promising material for water purification because of its advantages such as robustness, low cost, and safety [14, 15]. Photocatalytic processes and their reaction kinetics depend greatly on the surface properties of TiO₂, therefore, surface modification has been intensively investigated in TiO₂-mediated photocatalysis [16-21]. Among species that can be used to alter the surface properties of TiO₂, Nafion and phosphate are two readily available candidates. Nafion, which is an anionic perfluorinated polymer consisting of a hydrophobic polytetrafluoroethylene backbone and side chains with hydrophilic sulfonate groups, is chemically and photochemically inert under both oxidative and reductive conditions [22-24]. Nafion-modified TiO₂ (Nf-TiO₂) has shown significantly enhanced photocatalytic activity in the degradation of cationic substrates (e.g., herbicides, dyes, and tetramethylammonium) [25-30]. This could be because of improved adsorption of cationic species at ion-exchange sites in Nafion [25-30]. The photocatalytic degradation of organic compounds in water can be influenced by coexisting inorganic anions (e.g., phosphate) that can strongly bind to the TiO₂ surface and alter the surface charge properties [31]. It has been reported that phosphate-modified TiO₂ (P-TiO₂) positively influences degradation of compounds that adsorb poorly on TiO₂ by increasing the production of hydroxyl radicals; however, it decreases the degradation of strongly adsorbing compounds [31-33]. The enhanced performance of phosphate-modified TiO₂ has also been attributed to facilitated electron transfer and improved oxygen adsorption [21, 34, 35]. Although photocatalytic reactions involving band gap excitation of phosphate-or Nafion-modified TiO₂ have been reported, there have been few studies of the photoreactions of these modified TiO₂ materials under ambient conditions with coexisting dissolved natural organic matter (NOM) and under daylight irradiation.

NOM, which is pervasive in water, always decreases the degradation efficiency of target prior pollutants by photocatalysis or other advanced oxidation processes [32, 36-39]. This inhibitory effect is mainly attributed to competitive reactions with reactive oxygen species (ROS) and competitive adsorption between NOM and target pollutants [32, 39]. Surface modification to alter the adsorption of NOM and modulate photocatalytic processes has shown great potential in counteracting inhibitory effects [32]. However, ROS can be photochemically produced in aerobic water containing NOM, and this facilitates the photosensitized degradation of waterborne organic pollutants [40-45]. Photoionization of NOM by light of wavelength 300–500 nm can produce triplet states (³NOM^*) and hydrated electrons (e^-) [43, 46]. Vinodgopal et al. [47] reported charge injection from excited NOM into semiconductor colloids; this indicates that electrons photoproduced from NOM can be trapped by semiconductor nanoparticles, to produce superoxide radicals (O₂^·-) through a single-electron oxygen reduction reaction (ORR). NOM therefore favorably affects TiO₂ photocatalysis by serving as sensitizers that extend the photocatalytic function to the visible-light (VL) range. It is a challenge to strengthen this favorable effect but avoid the inhibitory effect of NOM, and achieving this is important in pilot studies of TiO₂ photocatalytic water treatment.

Humic acid (HA), a representative NOM containing abundant phenolic, hydroxyl, and carboxylic groups, adsorbs strongly on TiO₂ with a high Langmuir binding constant (0.12–0.90 L mg^-1) [48]. In this study, we compared the sensitizing effects of HA on Nafion-and phosphate-modified TiO₂ materials under VL irradiation. Phenol was used as a model pollutant to probe the photocatalytic activity because of its non-ionic properties and negligible adsorption on TiO₂. The mechanism of the effect of HA on the two modified TiO₂ materials was explored based on analyses of the degradation kinetics, adsorption, main active species, and semiconductor properties. Clarification of the effects of surface species on the VL activity of TiO₂ in water containing NOM is important in understanding the photochemical activity and toxicity of nanoparticles, and will help in the development of effective photocatalytic materials for pilot studies of water purification.

Humic acid sodium (HA), furfuryl alcohol (FFA), phenol, NaH₂PO₄, isopropyl alcohol (IPA), p-benzoquinone (BQ), and N, N-diethyl-p-phenylenediamine (DPD) were all obtained from Sigma Aldrich, USA. Nafion (5 wt% solution in a mixture of ethanol, 1-propanol, and water) was purchased from the Dupont Co., Ltd., USA. 5, 5-Dimethy-l-pyrroline N-oxide (DMPO) was obtained from the Adamas Co., Ltd., China. Peroxidase (POD, horseradish, RZ > 1.5) was obtained from the Sangon Biotech Co., Ltd., China All other reagents were purchased from the Shanghai Chemical Reagent Co., Ltd., China.

Commercial Degussa P25 TiO₂ (a mixture of 71% anatase and 29% rutile) was used for all tests. To avoid desorption, P-TiO₂ was obtained by photocatalytic degradation in a suspension of TiO₂ and 2 mmol/L NaH₂PO₄. Nf-TiO₂ was prepared using a previously reported method [28]. Briefly, Nafion solution and TiO₂ powder, at a ratio of 50 mg (Nafion)/g (TiO₂), were added to deionized water (10 mL). The suspension was stirred for 6 h to ensure homogeneous coating of the TiO₂ nanoparticles with Nafion. The product was dried in a vacuum oven at 30 ℃ for 24 h to give Nf-TiO₂.

The light source was a 500-W xenon lamp with a 420 nm cutoff filter and a filter to remove infrared light. In the phenol degradation tests, a suspension of the photocatalyst (0.5 g/L) was mixed with phenol (10 mg/L) in a 50 mL quartz vessel; various amounts of HA were then added. In all the tests, the initial pH values of the suspensions were carefully adjusted to 3.0 with HClO₄ to minimize desorption of phosphate or Nafion from TiO₂. The suspension was kept in the dark at room temperature for 30 min to achieve adsorption–desorption equilibrium, and was then illuminated to start the photocatalytic degradation. At predetermined times, samples were collected and filtered through a 0.45 μm polyethersulfone (PES) filter. The residual phenol concentrations were determined using a high-performance liquid chromatography system (LC-2010AHT, Shimadzu) equipped with a C-18 column. A binary mixture of phosphoric acid solution (0.1 wt%) and methanol at a volumetric ratio of 70:30 was used as the mobile phase, and the detection wavelength was 270 nm. Apparent first-order degradation rate constants (k) were obtained for all photocatalytic processes. The ratio (R) between the reaction rate constants in the absence (k₀) and presence of HA (k) was used as a measure of the inhibitory or favorable effect of HA. The concentration of HA in all tests, unless specified, was 20 mg/L.

The concentration of H₂O₂ in the illuminated suspension was determined using a modified DPD–POD method [49, 50]. Briefly, the solution obtained after filtration (1 mL) was transferred to a 10 mL volumetric flask and mixed with phosphate buffer (3 mL, 0.5 mol/L, pH = 6). Then DPD solution (50 μL, 10 mg/mL) and POD solution (50 μL, 1 mg/mL) were pipetted into the mixture and deionized water (10 mL) was added. The H₂O₂ concentration was obtained by measuring the absorbance at 551 nm (UV-vis spectrophotometer, T6-New Century, Purkinje General).

The electron paramagnetic resonance (EPR) signals for DMPO–O₂^•- adducts were recorded using an EPR spectrometer (MS 5000, Magent Tech), with DMPO as the spin trap. A catalyst suspension (0.5 g/L) in methanol was irradiated for 5 min under VL in the presence of DMPO (20 mmol/L). A sample of the suspension (about 0.5 mL) was removed, filtered through a 0.22 μm PES filter, immediately transferred to a flat quartz EPR cell, and the EPR spectrum was recorded. The EPR settings were center field, 336 mT; sweep width, 10 mT; sweep time, 60 s; modulation, 0.2 mT; resonance frequency, 9.464 GHz; and microwave power, 20 mW. X-ray photoelectron spectroscopy (XPS) was performed using an AXIS Ultra DLD spectrometer (Kratos Analytical-A, Shimadzu) with a monochromatic Al K_α source (1486.6 eV). P-TiO₂ samples for XPS were prepared by dispersing TiO₂ in a NaH₂PO₄ solution (2 mmol/L). The solid was separated by filtration and freeze-dried.

Electrochemical measurements were performed using a CHI 760E electrochemical system with a three-electrode cell. The light source was a 500-W xenon lamp with a cutoff filter (λ > 420 nm). A Pt plate and Hg/HgCl₂ electrode were used as the counter electrode and reference electrode, respectively. The working electrode was prepared by coating a slurry of the sample on a cleaned indium tin oxide glass substrate by the doctor blade method. The slurry was prepared by dispersing the powder sample (0.2 g) in absolute ethanol (1 mL). The coated electrodes were heated at 200 ℃ in air for 2 h to improve adhesion. In all the tests, the available surface area of the working electrode was 1 cm².

Impedance–potential tests were performed in a three-electrode cell, using previously reported procedures [51]. The ac amplitude and frequency were set at 5 mV and 1.5 kHz, respectively. Electrochemical impedance spectroscopy (EIS) was performed by applying an open circuit voltage bias (0.2 V) and were recorded over the frequency range 0.1–10⁵ Hz at an ac amplitude of 5 mV. The electrolyte for current–time and EIS tests was a 0.1 mol/L sodium perchlorate solution of pH = 3.5, and for the impedance–potential tests it was a 0.5 mol/L Na₂SO₄ aqueous solution of pH = 5.5. The electrolyte was purged with nitrogen for 20 min prior to each test and continuously purged during the measurements.

TiO₂ (25 mg) or Nf-TiO₂ (25 mg) was added to HA aqueous solutions (50 mL, 20, 40 or 60 mg/L), and the suspension pH was adjusted to 3.0. To determine the effect of phosphate modification, HA solutions were prepared in the presence of 2 mmol/L phosphate solutions. The suspensions were stirred at 30 ℃ for 1 h to reach adsorption equilibrium. In time-dependence tests, samples were withdrawn at predetermined times. The concentration of HA was determined by measuring the absorbance at 254 nm and calibrating using standard samples.

Fig. 1(a–c) show the effects of HA on VL degradation of phenol over TiO₂ and surface-modified TiO₂. About 26% (±0.9%) phenol was removed on pristine TiO₂ by VL irradiation for 120 min. This is attributed to the formation of phenolic complexes on the TiO₂ surface, which enabled ligand-to-metal charge transfer in the VL range [52]. Unlike the inhibitory effect of HA in phenol degradation over TiO₂ under UV irradiation [32], VL degradation of phenol was greatly enhanced in the presence of HA. The degradation rate constants for phenol increased to (0.0104±0.0001) min^-1 at a HA concentration of 20 mg/L, about four times that in the absence of HA. This can be ascribed to the sensitizing effect of HA, which can absorb light below 500 nm and inject excited electrons into TiO₂ to produce ROS. Even in the absence of TiO₂ or other semiconductors, ROS can be produced by illuminating a HA solution [43, 45], and the presence of electron donors such as hydroxyaromatic compounds accelerate the generation of O₂^·- [42]. This explains why a 17% (±0.5%) decrease in the amount of phenol was achieved in the presence of 20 mg/L HA alone after irradiation for 120 min. The VL degradation rate of phenol on TiO₂ (Fig. 1(b)) increased with increasing HA concentration. However, the improvement lessened with increasing HA concentration from 20 to 40 mg/L. This can be ascribed to an inhibitory effect of HA and the saturation of occupied adsorption sites.

The Nf-TiO₂ and P-TiO₂ samples slightly changed the VL activity in phenol degradation compared to that of unmodified TiO₂ (Fig. 1(a)). The phosphate and Nafion loadings were also investigated (Fig. S1). The results indicate that the catalysts with 50 mg Nafion/g TiO₂ and 2 mmol/L phosphate had the best VL activities. However, the effects of both loadings on the photocatalytic performance of TiO₂ were low. HA strongly inhibited the VL degradation of phenol on P-TiO₂, with an R value of 0.51, which is much less than 1. This contradicts previous reports that under UV irradiation HA facilitated charge transfer and enhanced phenol degradation on phosphate-modified TiO₂ [32]. This indicates that the mechanism of phenol photocatalytic degradation depends on the light source. The phosphate in the solution did not inhibit sensitization of degradation of phenol by HA alone (Fig. S2). However, HA (20 mg/L) significantly increased the VL degradation rates of phenol on Nf-TiO₂. The reaction rate constant increased from (0.003±0.001) min^-1 in the absence of HA to (0.025±0.003) min^-1 in the presence of 20 mg/L HA. The factor R is 7.8, which is much larger than that for pristine TiO₂ (R = 4.0 at 20 mg/L HA). Moreover, the phenol removal efficiency was 96% (±0.3%) after VL irradiation for 120 min. The results of control tests performed at an initial pH of 6.0 suggest that the trends in the effect of HA were similar for Nf-TiO₂ and P-TiO₂ under acidic and neutral conditions (Fig. S3). However, the phenol degradation rate decreased when the HA concentration was increased to 40 mg/L (Fig. 1(c)). This can be attributed to a concurrent inhibitory effect.

Enhancement of the VL activity of Nf-TiO₂ by HA modification was confirmed by photocurrent measurements. The current–time curves for electrodes with various on–off cycles of intermittent irradiation are shown in Fig. 1(d). The inset in the figure shows that the photocurrent signal was negligible for all the electrodes in the absence of HA; this is consistent with the large band gap of TiO₂ (~3.2 eV). For the P-TiO₂ electrode in the presence of HA, the photocurrents were also low, less than 1 μA/cm². Sensitization by HA induced a significant photocurrent for TiO₂ (90 μA/cm²), and an even higher photocurrent for Nf-TiO₂ (117 μA/cm²). The better photocurrent response by Nf-TiO₂ indicates enhanced photoelectron generation and mobilization, which is consistent with a higher activity in phenol degradation under VL irradiation.

XPS was used to prove the presence of phosphate and Nafion on P-TiO₂ and Nf-TiO₂. The P/Ti and F/Ti ratios at the TiO₂ surface were 0.05 and 1.3, respectively. For P-TiO₂, the asymmetric P 2p spectrum (Fig. 2(a)) can be fitted by two peaks, at 132.8 and 133.7 eV. These can be assigned to monodentate coordination between surface Ti sites and PO₄^3-, with P in the pentavalent state [35]. For F-TiO₂, the band centered at 688.8 eV (Fig. 2(b)) is assigned to the F signal from Nafion molecules [27]. The presence of adsorbed Nafion is confirmed by a peak at 535.4 eV in the O 1s spectrum (Fig. S4), which corresponds to oxygen in the sulfonic acid group.

One possible explanation for the differing effects of HA on the VL activities of P-TiO₂ and Nf-TiO₂ in phenol degradation and the photocurrent responses is that the amounts of HA adsorbed are different. An increase or decrease in HA adsorption would strengthen or weaken the favorable sensitizing effect of HA. The adsorption of HA on various catalysts was investigated; the results are shown in Fig. 3. The time to reach adsorption balance for Nf-TiO₂ was about 40 min (Fig. 3(a)), slightly longer than those for TiO₂ and P-TiO₂ (less than 10 min). However, the equilibrium concentrations of HA on P-TiO₂ and Nf-TiO₂ were similar (Fig. 3(b)). The adsorption capacities at an initial concentration of 20 mg/L HA were (30.2±1.4) and (29.9±0.1) mg/g for P-TiO₂ and Nf-TiO₂, respectively; these are slightly lower than that for TiO₂ (33.9±0.1) mg/g). The adsorption of HA on the oxide surface involves electrostatic interactions, ligand exchange, hydrophobic interactions, entropic effects, hydrogen bonding, and cation bridging [53, 54]. The modified TiO₂ materials both had negative zeta potentials at the studied pH (Fig. S5), indicating similar electrostatic interactions between HA and the modified TiO₂. However, phosphate modification would increase the hydrophilicity of the TiO₂ surface and strengthen adsorption via ligand exchange or cation bridging. Nafion modification of TiO₂ would enhance the hydrophobicity and improve hydrophobic interactions with HA. Both modified TiO₂ materials therefore had adsorption capacities similar to that of pristine TiO₂ at a low initial HA concentration. At higher initial concentrations of HA, the adsorption capacity of TiO₂ was much higher than those of P-TiO₂ and Nf-TiO₂. However, the amount of HA adsorbed on Nf-TiO₂ was always lower than that on P-TiO₂, which indicates that the poor sensitizing effect of HA on P-TiO₂ cannot be ascribed to blockage of HA adsorption by surface-modifying phosphate anions, and the better photocatalytic activity of Nf-TiO₂ is irrelevant in terms of adsorption.

Three chemicals, namely FFA, BQ, and IPA, were used as scavengers for singlet oxygen, superoxide, and hydroxyl radicals, respectively. Because phenol degradation in the P-TiO₂/HA suspension was poor, the effects of scavengers on phenol degradation in TiO₂/HA and Nf-TiO₂/HA suspensions were investigated. Fig. 4 shows that addition of BQ significantly decreased the phenol degradation efficiencies of both the TiO₂ and Nf-TiO₂ systems, suggesting that superoxide radicals are the main ROS in HA-sensitized TiO₂ photocatalysis. The mechanism can be described by the sequential one-electron reactions shown in Eqs. (1)–(4). Under VL irradiation, electrons are generated by excitation of HA (Eq. (1)) and transferred to TiO₂ (Eq. (2)). The electrons are then scavenged by oxygen, producing superoxide radicals (Eq. (3)) and then H₂O₂ (Eq. (4)). Recombination can also take place when electrons are trapped by positively charged HA radicals (Eq. (5)). Unlike UV irradiation, which can induce band gap excitation of electrons and produce holes that are strong enough oxidants to produce ·OH, the active species produced by VL irradiation of sensitized TiO₂ originate exclusively from electron transfer, and superoxide radicals are the dominant ROS for organic degradation.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

EPR spectroscopy was used to confirm the production of O₂^•- in the illuminated suspensions. Fig. 5(a) shows that six characteristic peaks from DMPO–O₂^•- spin adducts were clearly observed for the TiO₂/HA and Nf-TiO₂/HA suspensions; however, no signal was detected for P-TiO₂/HA, which indicates the absence of O₂^•- in the P-TiO₂/HA suspension. Because O₂^•- is the photochemical precursor of H₂O₂ and accumulation of H₂O₂ formed as an intermediate is always observed in photocatalysis, the concentrations of H₂O₂ in different photocatalytic systems were monitored using the DPD–POD method to further investigate the effect of HA sensitization of modified TiO₂. The concentration of H₂O₂ was negligible in the illuminated P-TiO₂/HA suspension, but was significant in the other two suspensions (Fig. 5(b)). After VL irradiation for 120 min, the concentration of H₂O₂ had increased to 93 and 99 μmol/L for the TiO₂/HA and Nf-TiO₂/HA suspensions, respectively. The amount of H₂O₂ in the P-TiO₂/HA suspension was almost negligible, indicating low production of O₂^•-, in agreement with the EPR analysis. The results suggest that P-TiO₂ enabled the ORR to proceed via a direct four-electron pathway (Eq. (6)). This can be rationalized by assuming that the HA molecules and anchored phosphate anions served as proton relays to provide adequate localized protons for proton-coupled electron transfer in a multi-electron ORR [55]. Although singlet oxygen could be generated through energy transfer processes (Eqs. (7) and (8)), its contribution to organic degradation is limited because of its low quantum yield under VL irradiation and relatively weak oxidation power. The poor VL activity of P-TiO₂/HA in phenol degradation and the light-source-dependent effect of HA on P-TiO₂ photocatalysis can therefore be explained.

The production of a significant amount of H₂O₂ in the illuminated Nf-TiO₂/HA suspension indicates that ORR on this catalyst followed the single-electron pathway. Moreover, the phenol degradation activity of Nf-TiO₂/HA was better than that of TiO₂/HA, although less HA was adsorbed on Nf-TiO₂ than on TiO₂. The improved electron transfer efficiency was investigated by measuring the conductivities of the catalysts. Fig. 6(a) shows Mott–Schottky curves for the three electrodes. For n-type semiconductors such as TiO₂, the relationship between the depletion layer capacitance (C_sc) and applied bias (E) can be described by Eq. (9),

(9)

wherein e is the electronic charge, ε₀ is the permittivity of a vacuum, ε_r is the dielectric constant of TiO₂, A is the contact area of the electrode in the electrolyte, E_FB is the flat-band potential of the semiconductor, and N_D is the carrier (electrons for n-type) density. The term kT/e can be ignored because of its low value. The E_FB and N_D of the semiconductor can be obtained from the slope and intercept, respectively, of the fitted curves. The fitted equations in Fig. 6(a) show that the estimated E_FB potentials for TiO₂, Nf-TiO₂, and P-TiO₂ were -0.54, -0.85, and -0.36 V, respectively, vs the normal hydrogen electrode at pH = 7. Nafion modification led to clear cathodic shifts of the energy band positions, whereas phosphate modification had the opposite effect on energy band shifts. Electron transfer is expected to be facilitated when the semiconductor has more cathodic conduction band positions. The N_D values can be compared based on the curve slopes. P-TiO₂ had the lowest N_D value, about 62.4% of that for TiO₂, and Nf-TiO₂ had the highest N_D, about 62% higher than that for TiO₂. The increased N_D for Nf-TiO₂ improves the electron mobility and reduces electron resistance; this is supported by the EIS results (Fig. 6(b)). Nyquist plots for TiO₂ and the modified TiO₂ materials were obtained at 0.2 V vs the standard calomel electrode. The charge transfer resistance (R_ct), which is the resistance to electron transfer from the electrode to the electrolyte, directly correlates with the semicircle diameter at high frequencies. The EIS plots show that the Nf-TiO₂ electrode had the lowest R_ct and P-TiO₂ had the highest R_ct. This indicates that resistance to electron transfer through the solid/liquid interface was lowest for the Nf-TiO₂. This would result in fewer possibilities for charge recombination and higher efficiency of ROS generation and pollutant degradation.

In this study, the sensitizing effects of HA on P-TiO₂ and Nf-TiO₂ were investigated. The results indicate the following. (1) Sensitization by HA was negligible for P-TiO₂, and active species such as superoxide radicals were not observed in the illuminated suspension; this can be ascribed to a possible direct four-electron ORR. (2) Sensitization by HA was strong for Nf-TiO₂; the reaction rate constant for phenol degradation in the presence of 20 mg/L HA was 7.8 times that in the absence of HA; this can be attributed to a cathodic shift of the flat-band potential, increased electron density, and decreased resistance between TiO₂ and the electrolyte after modification with Nafion. The results of the present work suggest that phosphate modification could be used to reduce the photochemical activity of TiO₂ nanoparticles in water containing NOM, and the VL activity of sensitized TiO₂ can be further improved by tuning the semiconductor properties through surface modification.