Visible-Light-Mediated TiO2 Photocatalysis of Fluoroquinolone

Department of Civil and Environmental Engineering, University of Illinois at ... This study reports on the photocatalytic transformation of fluoroquin...
1 downloads 0 Views 214KB Size
Environ. Sci. Technol. 2007, 41, 4720-4727

Visible-Light-Mediated TiO2 Photocatalysis of Fluoroquinolone Antibacterial Agents T I A S P A U L , †,§ P E N N E Y L . M I L L E R , ‡,§ A N D T I M O T H Y J . S T R A T H M A N N * ,†,§ Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, Urbana, Illinois 61801, Department of Chemistry, Rose-Hulman Institute of Technology, Terre Haute, Indiana 47803, and Center of Advanced Materials for the Purification of Water with Systems, Urbana, Illinois 61801

This study reports on the photocatalytic transformation of fluoroquinolone antibacterial agents (ciprofloxacin, enrofloxacin, norfloxacin, and flumequine) in aqueous titanium dioxide (TiO2) suspensions irradiated with ultraviolet (UV; λ > 324 nm) or visible light (λ > 400, > 420, or > 450 nm). Visible-light-mediated fluoroquinolone degradation is unexpected from direct photolysis or established TiO2 band gap photoexcitation mechanisms, which both require UV light. Visible-light-mediated photocatalysis requires an appropriate conduction band electron acceptor (e.g., O2, BrO3-), but is not dependent upon hydroxyl radical, superoxide, or other reactive oxygen species generated upon TiO2 band gap excitation. The process slows considerably when fluoroquinolone adsorption is inhibited. Whereas fluoroquinolone decomposition in UV-irradiated TiO2 suspensions is accompanied by mineralization, no changes in dissolved organic carbon occur during visible-lightphotocatalyzed degradation. Results are consistent with a proposed charge-transfer mechanism initiated by photoexcitation of surface-complexed fluoroquinolone molecules. Complexation to the TiO2 surface causes a red shift in the fluoroquinolone absorption spectrum (via ligandto-metal charge transfer), enabling photoexcitation by visible light. Fluoroquinolone oxidation then occurs by electron transfer into the TiO2 conduction band, which delivers the electron to an adsorbed electron acceptor. The lack of organic carbon mineralization indicates formation of stable organic byproducts that are resistant to further degradation by visible light. In UV-irradiated TiO2 suspensions, the charge-transfer mechanism acts in parallel with the semiconductor band gap photoexcitation mechanism.

Introduction Fluoroquinolones (Figure 1) are broad-spectrum antibacterial agents widely used for treating bacterial infections. Until recently, large quantities were also used as feed additives in livestock production to prevent disease and promote rapid growth. Fluoroquinolones are incompletely metabolized, and * Corresponding author e-mail: [email protected]; phone: 217244-4679; fax: 217-333-6968. † University of Illinois at Urbana-Champaign. ‡ Rose-Hulman Institute of Technology. § Center of Advanced Materials for the Purification of Water with Systems. 4720

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 13, 2007

FIGURE 1. Structures of ciprofloxacin and related analogues. Dashed box contains the core oxo-quinolinecarboxylic acid substructure also present in enrofloxacin, flumequine, and norfloxacin. Dashed oval contains the fluorophenyl substructure present in some analogues. a significant fraction is excreted in the pharmaceutically active form (1). Sorption to sludge has been reported to be the major process controlling fluoroquinolone removal during wastewater treatment (2). However, sorption is incomplete and fluoroquinolones have been widely detected in treated wastewater effluent (3). Furthermore, fluoroquinolones sorbed to land-applied sludge may still be released into aquatic environments via surface runoff or leaching into the underlying groundwater. The widespread use of fluoroquinolones and the lack of treatment processes targeting these compounds is resulting in their ubiquity in surface waters (3, 4). The presence of fluoroquinolones in natural environments raises several human health concerns. First, continuous release of fluoroquinolones into aquatic environments may promote antibiotic resistance within native bacterial populations in impacted environments. The increased resistance can then be transferred to other bacterial populations (e.g., human pathogens) via plasmids, as has been reported for tetracyclines (5, 6). Martinez-Martinez and co-workers identified a broad-host range plasmid that readily transfers fluoroquinolone resistance genes between bacterial species (7). Resistance to fluoroquinolones is especially undesirable because medical practitioners often prescribe them as “last resort” drugs when other antibiotics are ineffective (8). The release of fluoroquinolones in wastewater effluent is also a concern because the receiving waters are often used as sources of drinking water. Although the chronic effects of sub-therapeutic doses of fluoroquinolones have not been established, higher doses are known to elicit a toxic response in some individuals (9). Concerns about the human and ecological effects of fluoroquinolones and other wastewater-derived micropollutants are increasing interest in the development of technologies that can efficiently treat waste streams contaminated 10.1021/es070097q CCC: $37.00

 2007 American Chemical Society Published on Web 06/05/2007

with these compounds. One process that holds promise for transforming and degrading fluoroquinolones and other pharmaceutical micropollutants is titanium dioxide (TiO2) photocatalysis. TiO2 photocatalysis using ultraviolet (UV) light has been extensively studied and shown to degrade a wide range of organic contaminants, including pharmaceuticals (10-12). UV irradiation of TiO2 photoexcites electrons from a filled valence band to an empty conduction band, giving rise to electron-hole pairs. The valence band holes (h+vb) can migrate to particle surfaces and react with adsorbed OH-/ H2O to produce hydroxyl radicals (‚OH), whereas the conduction band electron (e-cb) reacts with adsorbed electron acceptors (e.g., O2 in aerated systems) (13). Due to the nonselectivity and high reactivity of ‚OH, UV-TiO2 photocatalysis is effective at oxidizing a wide range of organic contaminants. However, the nonselectivity of ‚OH also results in inefficient use of photon energy when treating trace contaminants in mixed waste streams where much higher concentrations of nontarget constituents (e.g., natural organic matter) are present. Some organic compounds have also been found to be susceptible to photocatalytic degradation by TiO2 irradiated with visible light, including chlorophenols (14, 15), catechols (16), organic dyes (17), and metal phthalocyaninesulfonates (18). The mechanisms controlling visible-light photocatalysis and the molecular characteristics that impart susceptibility to these processes are less well understood than those for UV processes. However, visible-light photocatalysis processes are potentially advantageous because they can be used to selectively target specific contaminants in mixed waste streams. As part of an ongoing investigation of the (photo)catalytic transformation of pharmaceutical micropollutants, it was revealed that fluoroquinolones are degraded in aqueous TiO2 suspensions irradiated with visible light (λ > 400 nm), an observation that was contrary to common expectations. This contribution reports for the first time on the visiblelight- and UV-mediated TiO2 photocatalytic transformation of fluoroquinolones. The major goals of the study are to characterize the mechanism, kinetics, and byproducts of fluoroquinolone transformation via visible-light photocatalysis, and compare results with UV photocatalysis under the same conditions.

Materials and Methods Reagents and Photocatalysts. A list of all reagents is provided in the Supporting Information (SI). Photocatalysis studies were conducted using two different nanophase TiO2 powders. Hombikat UV100 (provided by Sachtleben Chemie, Germany), referred to as TiO2#1, is a pure anatase phase with a specific surface area of 250 m2/g and average particle size 324 nm) and visible-light irradiation experiments (λ > 400, > 420, or > 450 nm). Potassium ferrioxalate actinometry measurements (20) showed that the incident photon flux on the reactor was 9 × 10-5 einsteins/min for the UV portion of the spectrum (324-400 nm) and 1.6 × 10-4 einsteins/min for the visible-light portion of the spectrum (>400 nm). Kinetics Experiments. Batch photocatalysis experiments were typically conducted in 250 mL glass jacketed beakers connected to a circulating constant-temperature water bath (25 ( 0.1 °C). Unless otherwise indicated, the standard

photocatalysis experiments were conducted by irradiating 50-mL aqueous suspensions containing 100 µM ciprofloxacin (cipro), 0.5 g/L TiO2#1, pH 3.0 (set using HClO4), and 10 mM ionic strength (set using NaClO4 + HClO4). The pH was held constant during reactions either by conducting experiments under acidic conditions (where production or consumption of small amounts of H+ during reactions have negligible effect on the initial [H+]) or by using an automatic pH stat instrument at higher pH conditions (Radiometer Analytical, TIM854; NaOH titrant added to keep pH from drifting downward during reactions). Prior to each batch reaction, suspensions were stirred under darkness and sparged with hydrated CO2-free air for 30 min. Reactions were then initiated by exposing the suspensions to the light source. Irradiated suspensions were continuously stirred and sparged with air throughout each reaction and 1 mL aliquots were collected at regular intervals. Prior to analysis, the aliquots were adjusted to pH 11 with NaOH to desorb cipro from the TiO2 particles, and supernatant was collected after centrifugation. Larger volume samples used to characterize the evolution of reaction intermediates and byproducts were obtained by irradiating a series of equivalent suspensions for different time periods, then collecting the entire 50 mL suspension for analysis. A portion of the collected suspensions used for analysis of cipro, its organic transformation products, and inorganic anions was adjusted to pH 11 to ensure desorption of these substances from TiO2 prior to centrifugation and analysis of the supernatant. Complete desorption of organic analytes was confirmed by total organic carbon (TOC) analysis. A second portion used for NH4+ analysis was filtered (0.45 µm, cellulose acetate) under acidic conditions to prevent losses due to water/air partitioning of the neutral conjugate base; negligible NH4+ adsorption to TiO2 at these conditions was confirmed. Anoxic batch reactions were prepared in a similar manner inside an oxygen-free glovebox chamber (95% N2, 5% H2; 25 °C; Pd catalyst; Coy Laboratory Products) using water that was deoxygenated by boiling and sparging with nitrogen for >30 min. A series of equivalent 50-mL suspensions were prepared, one for each reaction time point, in 160-mL glass serum bottles crimp-sealed with butyl stoppers. Individual serum bottles were then removed from the glovebox and irradiated for the desired period of time; bottles were placed on their sides and mixed with magnetic stir bars. Equivalent air-sparged serum bottle reactions were also carried out for comparison with anoxic serum bottle reactions. The rate for reactions performed in air-sparged serum bottles were approximately 50% faster than comparable reactions conducted in the jacketed beaker setup described above, indicating a small effect of reactor geometry. The extent of fluoroquinolone adsorption to TiO2 was determined by measuring the supernatant concentration prior to irradiation (30 min equilibration time) and calculating the adsorbed concentration by difference. Separate analyses revealed adsorption kinetics reached equilibrium within 5 min. Analysis. The concentration of fluoroquinolones was determined by high-performance liquid chromatography (HPLC) with photodiode array detection (Shimadzu VP system). Separation was achieved using a Nova-pak column (C18, 3.9 × 150 mm, 4 µm particle size) and guard column of the same material (20 mm). An isocratic mobile phase (1.5 mL/min) consisted of a 15:85 ratio of acetonitrile and an aqueous eluent containing 2.5 mM sodium 1-heptanesulfonate adjusted to pH 2 with H3PO4 (21). The molecular weights and structures of organic intermediates and products were determined by liquid chromatography with tandem mass spectrometry (LC-MS/MS), performed on a ThermoFinnigan LCQ DecaXP ion trap VOL. 41, NO. 13, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4721

> 324 nm) and visible light (λ > 400 nm, > 420 nm, and > 450 nm). Cipro degradation is observed under all lighting conditions, with rates increasing as the cutoff wavelength is lowered. No cipro degradation is observed in the absence of light, but 43% percent of the compound is reversibly adsorbed under the conditions shown (can be desorbed by adjusting to pH 11). In TiO2-free solutions, slow direct photolysis of cipro occurs under UV light (∼10% degraded in 1 h), but no direct photolysis occurs under visible light. According to the conventional UV-TiO2 photocatalysis mechanism, cipro degradation in TiO2 suspensions irradiated with UV light is anticipated, whereas visible-light-promoted degradation is contrary to expectations. Earlier reports and results from our own studies demonstrate that such a finding is not common for most organic contaminants.

FIGURE 2. Photocatalytic degradation of ciprofloxacin under different lighting conditions (A) and photocatalysis of related compounds at λ > 400 nm (B). Initial conditions: 100 µM fluoroquinolone, 0.5 g/L TiO2#1, pH 3.0, I ) 0.01 M, 25 °C. Rate coefficients for each reaction are provided in Table S1 in the Supporting Information. instrument using positive electrospray ionization (ESI+). The column used was the same one described for HPLC analysis. A gradient method was used to separate analytes, beginning with 100% solvent A (95% water, 5% acetonitrile, 0.2% formic acid) at a flow rate of 1.5 mL/min and linearly increasing solvent B (95% acetonitrile, 5% water, 0.2% formic acid) to 60% over the first 35 min, then increasing to 90% over the next 10 min. MS/MS spectra were obtained by fragmentation with 35 V. Organic carbon mineralization was determined by TOC analysis, performed on a Dohrmann Phoenix 8000 UVpersulfate TOC analyzer. The evolution of ionic products was quantified using ion chromatography with conductivity detection (nitrate, nitrite, fluoride; Dionex ICS-2000 system) and colorimetric analysis (ammonia; Nessler’s reagent). Diffuse reflectance ultraviolet-visible spectra (DRUVS) of TiO2 powders were collected neat on a Varian Cary 500 instrument equipped with a Praying Mantis diffuse reflectance accessory and microsampler. TiO2#1 suspensions (0.5 g/L) were prepared at pH 6.7 in the absence and presence of 500 µM ciprofloxacin. After equilibrating overnight, solids were collected by vacuum-filtration (0.45 µm) and dried for 2 days under darkness at room temperature. The dried powders were then homogenized using a mortar and pestle and kept under darkness until analysis.

Results and Discussion Reaction Kinetics. Kinetics studies conducted under a wide range of conditions demonstrate that fluoroquinolones are susceptible to photocatalytic degradation. Figure 2A shows cipro degradation in TiO2 suspensions irradiated with UV (λ 4722

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 13, 2007

Multiple lines of evidence indicate that the observed visible-light-mediated photocatalysis of cipro is the result of a fluoroquinolone-specific mechanism and not simply due to stray ultra-band gap UV radiation passing through cutoff light filters, as follows: (i) Radiometer readings indicate that minimal UV radiation passes through the >400 nm longpass filter. (ii) Degradation of cipro via visible-light photocatalysis is also observed with higher visible-light cutoffs (λ > 420 nm, λ > 450 nm), where the likelihood of stray UV light passing through the cutoff filters is further diminished. (iii) Experiments conducted with other organic compounds in the same experimental setup (e.g., sulfamethoxazole, N-nitrosodimethylamine), which are susceptible to UV-TiO2 photocatalysis (22, 23), show no degradation when incident light is passed through the >400 nm, >420 nm, or >450 nm long-pass filters. (iv) The relative rates of photocatalysis measured for two different TiO2 phases differs for λ > 324 nm versus λ > 400 nm (to be discussed later). (v) Other evidence, that will be discussed later, which supports an alternative mechanism. Fluoroquinolones in general are susceptible to this unique photocatalytic degradation process. Figure 2B shows the visible-light photocatalytic degradation of cipro and three structural analogues (flumequine, enrofloxacin, and norfloxacin) that retain the oxo-quinolinecarboxylic acid core structure (portion of cipro structure depicted inside dashed box in Figure 1). In contrast, no visible-light degradation is observed for 1-(2-fluorophenyl)piperazine, an analogue that lacks this core structure, but retains a portion of the cipro structure depicted within the dashed oval in Figure 1. For the range of conditions examined in batch reactions, fluoroquinolone photocatalytic degradation follows a pseudofirst-order rate law. Measured pseudo-first-order rate coefficients (kobs; min-1) and initial reaction rates (ro; µM min-1) are provided for all batch reactions in Table S1 in the SI. Reaction rates increase with increasing initial cipro concentration and increasing TiO2 loading (see Figure S1 in the SI). Observed trends are consistent with a LangmuirHinshelwood model for surface-mediated reaction kinetics (19, 24, 25). Effects of Radical Scavengers and Electron Acceptors. A number of diagnostic tests were conducted with cipro to probe the mechanism responsible for visible-light-mediated photocatalysis. Figure 3A shows the effects of adding methanol, an established hydroxyl radical scavenger, on cipro degradation in both visible- and UV-irradiated TiO2 suspensions. Results show that an excess concentration of ‚OH scavenger has no effect on cipro degradation under visible light. This contrasts with results observed for UV irradiation, where the observed rate coefficient decreases by 37% when methanol is added. These observations indicate that the mechanism responsible for cipro degradation under visible light does not involve ‚OH, whereas both ‚OH-dependent and ‚OH-independent mechanisms are active under UV irradiation. Separate experiments also show that adding

FIGURE 3. Effect of radical scavenger on UV and visible-light photocatalysis of ciprofloxacin (A) and effect of electron acceptors, O2 and 10 mM BrO3-, on ciprofloxacin visible-light photocatalysis (B). Initial conditions: 100 µM ciprofloxacin, 0.5 g/L TiO2#1, pH 3.0, I ) 0.01 M, 25 °C. Visible ) λ > 400 nm, UV ) λ > 324 nm. Reaction rate coefficients are provided in Table S1. superoxide dismutase has no affect on the rate of cipro degradation under visible light, demonstrating that superoxide (O2‚-) also does not play a crucial role in the controlling reaction mechanism. Results presented in Figure 3B show that visible-light photocatalysis of cipro, like UV photocatalysis, requires an appropriate TiO2 conduction band electron acceptor. In airsaturated TiO2 suspensions, ∼250 µM dissolved oxygen is present to act as the ecb- acceptor (26). Visible-light photocatalysis of cipro is markedly inhibited in anoxic suspensions that lack an external ecb- acceptor; similar results are observed in UV-irradiated suspensions. However, cipro photocatalysis is observed again when 10 mM bromate (BrO3-), an alternate ecb- acceptor (27), is added to anoxic TiO2 suspensions; no reaction between cipro and bromate is observed in either dark or TiO2-free irradiated controls. Furthermore, no cipro degradation is observed in visiblelight-irradiated suspensions of metal oxides that lack the semiconductor properties of TiO2 (SiO2, γ-AlOOH) (Figure S2 in the SI), indicating that the controlling reaction mechanism is not simply a direct photolysis reaction that is catalyzed by fluoroquinolone adsorption to metal oxide surfaces. Instead, these collective observations suggest that the TiO2 conduction band plays a key role in the mechanism by mediating electron transfer from the fluoroquinolone to an appropriate electron acceptor. Importance of Fluoroquinolone Adsorption. Results also suggest that cipro adsorption to the TiO2 surface is a key requirement for visible-light photocatalysis. This can be demonstrated by comparing measured rates of cipro deg-

FIGURE 4. Effect of NaF addition at pH 6.0 (A) and effect of pH (B) on visible-light photocatalysis rates and the extent of ciprofloxacin adsorption to TiO2. Initial conditions: 100 µM ciprofloxacin, 0.5 g/L TiO2#1, I ) 0.01 M, 25 °C, λ > 400 nm. Percentages reflect the extent of ciprofloxacin adsorption to TiO2. Reaction rate coefficients are provided in Table S1. radation to the corresponding percentage of cipro adsorbed to TiO2 at different solution conditions. Figure 4A shows that both the reaction rate and the extent of cipro adsorption progressively decrease when increasing concentrations of fluoride ion, a competitive adsorbate, are added to TiO2 suspensions at pH 6; measured kobs values are directly proportional to the amount of cipro adsorbed. Figure 4B shows that reaction rates and the extent of cipro adsorption also show similar trends when pH varies. However, measured kobs values do not vary in direct proportion to changes in the amount of cipro adsorption, suggesting that other pHdependent factors are also important. For example, total adsorbed cipro can include multiple molecular species (e.g., inner-sphere surface complexes and hydrogen bonded outersphere complexes) that exhibit considerable differences in photocatalytic reactivity, and the relative contribution of the individual species to total adsorbed cipro can vary with pH. Previous studies have reported spectroscopic evidence for the simultaneous presence of inner- and outer-sphere complexes of ionogenic organic compounds at aqueousmetal oxide interfaces (28, 29). Future studies are aimed at exploring the surface speciation of adsorbed fluoroquinolone molecules and the relationship between surface speciation and rates of visible-light photocatalysis. Product Identification. Figure 5 summarizes the results of product analysis studies. Figure 5A shows the change in total organic carbon that accompanies visible and UV photocatalyzed degradation of cipro. Although cipro is degraded to below the HPLC-UV detection limit within 1 h in visible-light-irradiated TiO2 suspensions, the total organic carbon (TOC) does not show any decrease within 3 h. This VOL. 41, NO. 13, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4723

FIGURE 5. Summary of results on product formation. Change in TOC during visible and UV photocatalysis of cipro (A), inorganic product formation during visible-light (B) and UV irradiation (C), organic product formation during visible-light (D) and UV irradiation (E), and proposed product structures (F). Initial conditions: 100 µM ciprofloxacin, 0.5 g/L TiO2#1 pH 3.0, I ) 0.01 M, 25 °C. result contrasts with UV-TiO2 photocatalysis, where TOC is reduced by 55% over the same time period (and by 88% in 12 h). The stability of TOC measurements under visible light indicates that ciprofloxacin is being transformed into stable organic byproducts. The evolution of inorganic products in visible- and UV-irradiated TiO2 suspensions is shown in Figure 5B and C, respectively. These measurements show that fluorine and nitrogen atoms within the cipro structure are incompletely mineralized under visible light (Figure 5B). After 3 h, the inorganic nitrogen species that were monitored ([NH4+] + [NO3-]) account for approximately 2/3 of the ciproderived nitrogen atoms (∼200 µM), and half of the structural fluorine atoms are converted to fluoride ions. In comparison, structural nitrogen and fluorine atoms are completely mineralized in UV-irradiated TiO2 suspensions over the same time period (Figure 5C). Nitrite was also monitored, but none was detected. Figure 5F shows the proposed structures of 6 organic intermediates/products identified by LC-MS/MS analysis (full MS/MS data are provided in Table S2 in the SI), and Figure 5D and E show timecourses for the LC-MS peak areas associated with these products in visible- and UV-irradiated TiO2 suspensions, respectively. Such timecourses are useful 4724

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 13, 2007

for showing the time dependence for formation and further decay of individual compounds, but the magnitudes of the peak areas should not be used as the basis for estimating the relative abundances of the detected analytes. Examination of the data shows that nitrogen- and fluorine-containing organic products remain after irradiating with visible light for 3 h, whereas those same intermediates are completely degraded under UV irradiation. In fact, only one of the identified intermediates (m/z ) 334) is detected after 1 h of UV irradiation. Although exact concentrations of the intermediates cannot be determined accurately without authentic standards, comparison of their HPLC-UV peak areas with the peak area recorded for 100 µM cipro (λ ) 278 nm) suggests that the six identified products quantitatively represent major organic intermediates generated by visible-light photocatalysis, which then react further to produce stable organic products not detected by HPLC-UV or LC-MS/MS. Together, analysis of organic and inorganic products indicates that fluoroquinolones are completely mineralized by UV photocatalysis, albeit at a slower rate than degradation of the parent compound. In contrast, during visible-light photocatalysis the organic products identified by LC-MS/ MS retain at least two nitrogen atoms as well as the fluorine

FIGURE 6. Proposed mechanism for visible-light-mediated photocatalytic degradation of ciprofloxacin (A) and effect of ciprofloxacin adsorption to TiO2#1 on DRUVS spectra (B). Numbers in (A) correspond to steps in the mechanism described in the text. atom, possibly preserving the fluoroquinolone core structure. Determining the site of nitrogen loss provides useful information for assessing whether or not the products will retain antimicrobial activity. Nitrogen release from the quinoline ring destroys the fluoroquinolone core structure, while loss of nitrogens from the piperazine group can potentially leave the antimicrobially active moieties intact (1). The lability of the quinoline nitrogen during visible-light photocatalysis processes was probed by measuring the formation of inorganic nitrogen products derived from flumequine, which contains a single nitrogen atom within the core quinoline group and lacks the piperazine ring nitrogens (Figure 1). Tests show that 22% of the flumequine-derived nitrogen is converted to NH4+ within 3 h (62% in 6 h), a finding which demonstrates that the core antimicrobially active structure of fluoroquinolones can be degraded by visible-light photocatalysis. Cipro product data presented in Figure 5B and D indirectly support this conclusion. Of the identified organic intermediates, 4 retain all 3 structural nitrogen atoms while the others retain 2 structural nitrogen atoms. Therefore, if these intermediates collectively represent any more than a very small fraction of organic mass balance, then all 3 nitrogen atoms (including the quinoline N) must be liberated from the remaining organic products to account for the NH4+ production observed under visible light () 2/3 of cipro-derived N atoms; Figure 5B). Photocatalysis Mechanisms. We propose that fluoroquinolone degradation proceeds by a visible light-sensitive charge-transfer process at the aqueous-TiO2 interface (Figure 6A). The key steps in the mechanism include the

following: (1) formation of a coordination complex between fluoroquinolone and the photocatalyst surface, (2) photoexcitation of the surface-complexed fluoroquinolone molecule, (3) the excited-state fluoroquinolone molecule returns to the ground state by transferring an electron into TiO2 conduction band, (4) the electron in the conduction band either recombines with fluoroquinolone donor molecule or is transferred to an adsorbed conduction band electron acceptor, and (5) the unstable 1-electron oxidation product of the fluoroquinolone is further transformed into more stable organic and inorganic products. As discussed earlier, only cipro analogues that retain the oxo-quinolinecarboxylic acid group (see dashed box in Figure 1) are subject to visible-light photocatalytic degradation (Figure 2B). Similarly, intermediates identified by LC-MS/MS analysis retain this group, enabling further degradation by the same mechanism. As depicted in the Figure 6A, the Lewis basicity and favorable chelation geometry of the ortho oxo-carboxylic groups suggest a likely mode for complex formation with surface-bound Ti(IV) ions (30). Gu and Karthikeyan attributed changes in the infrared spectra of cipro adsorbed to Al and Fe oxides to formation of surface complexes between these same functional groups and surface-bound Al and Fe ions (31). The cipro analogue 1-(2fluorophenyl)piperazine lacks the oxo-carboxylic chelate group (Figure 1) and does not adsorb to TiO2, presumably one of the major reasons why this compound is not sensitive to visible-light photocatalytic degradation (Figure 2B). Interpreting the reactivity pattern of the other fluoroquinolone analogues is more difficult because the observed reaction rates for individual compounds may be the net result of several (possibly offsetting) factors that affect individual steps in the proposed reaction mechanism, including varying degrees of surface complexation and varying quantum efficiencies for photoexcitation by visible light. For example, one analogue could adsorb to TiO2 to a greater extent than another, but have a lower quantum efficiency for photoexcitation. Fluoroquinolone surface complexation prior to photoexcitation is crucial for visible-light photocatalysis because neither dissolved fluoroquinolones nor TiO2 absorb visible light independent of each other. Figure 6B shows the diffuse reflectance UV/vis spectra of TiO2 before and after cipro adsorption. The presence of adsorbed cipro molecules causes a red shift into the visible region of the spectrum, presumably through ligand-to-metal charge transfer (LMCT) between the fluoroquinolone and a surface-bound Ti(IV) ion. Similar red shifts have been reported for chlorophenols (15) and catechols (16), which are also reported to undergo visible-light photocatalyzed degradation by a similar mechanism. Degradation of cipro by UV-A irradiation (324 nm < λ < 400 nm) results from two mechanisms occurring in parallel: (1) the charge-transfer process described above and (2) semiconductor charge separation by photons with ultra-band gap energy, leading to formation of ‚OH and other reactive oxygen species at the aqueous-TiO2 interface. Evidence for the parallel mechanisms is provided by the hydroxyl radical scavenger experiments presented in Figure 3A. Addition of 1 M methanol partially inhibits UV photocatalytic degradation of cipro, but the degradation rate still exceeds the rate observed for λ > 400 nm (plus any contributions from direct UVA photolysis). Therefore, photons from both the visible and UV regions of the spectrum can initiate the photoexcited charge-transfer process when suspensions are irradiated with λ > 324 nm. Operation of this mechanism under UV conditions also helps to explain why the same transient organic intermediates are observed at all conditions. Presumably, intermediates and products resulting from the ‚ OH-dependent pathway are not amenable to the LC-MS/ MS method used. VOL. 41, NO. 13, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4725

The relative contribution of the two photocatalysis mechanisms under λ > 324 nm is dependent upon the TiO2 catalyst phase. Comparable overall rates of cipro degradation are observed in suspensions of TiO2#1 (Hombikat UV100) and TiO2#2 (Degussa P25) irradiated by light with λ > 324 nm (see Figure S3 in SI). However, addition of methanol, an ‚OH scavenger that inhibits the band gap excitation pathway, has a much more pronounced effect on reaction rates with TiO2#2 than with TiO2#1. This suggests that the band gap excitation pathway is a greater contributor to overall cipro degradation in UV-irradiated suspensions of TiO2#2. It is also worth noting that TiO2#1, the phase that is most active in visible-light-irradiated suspensions where charge transfer is presumed to be the only active pathway, is also more reactive in UV-irradiated methanol-amended suspensions. TiO2#1 likely favors the surface complexation-dependent charge -transfer pathway more than TiO2#2 because a much greater quantity of cipro adsorbs to this phase (43%) in comparison to TiO2#2 (5%) at the conditions examined, presumably due to the higher specific surface area of the former. Environmental Implications. Ideal water treatment technologies for fluoroquinolones or other pharmaceutically active compounds will be (i) effective at deactivating the target compound’s antibacterial activity, (ii) selective for the target compounds in the presence of nontarget water constituents, and (iii) operated in an environmentally sustainable fashion. The identification of fluoroquinolone sensitivity to visible-light TiO2 photocatalysis suggests a treatment strategy that could meet all of these criteria. First, results demonstrate that visible and UV photocatalysis processes are capable of degrading the core oxo-quinolinecarboxylate structure believed to be responsible for antibacterial activity (the nitrogen within the quinoline ring structure of cipro must be at least partially liberated to account for the quantity of ammonia produced during visiblelight photocatalysis). Because active pharmaceutical products are often selected from many structurally related analogues based on heightened relative activity, it might be expected that even small structural transformations will substantially reduce pharmaceutical activity. That said, further research is needed to quantify the residual antibacterial activity of the photocatalysis products in comparison to both the parent fluoroquinolones and products resulting from other treatment processes. Removing trace levels of micropollutants from waste streams and natural water matrices that contain other nontarget constituents at much higher concentrations (e.g., NOM) is difficult, especially when using UV-TiO2 photocatalysis or other nonselective advanced oxidation processes. More compound-selective treatment processes, such as the visible-light TiO2 photocatalysis of fluoroquinolones, may be much more efficient strategies for treating trace levels of micropollutants since they target the contaminant of concern. The promise of this technology needs to be further evaluated in more complex matrices with environmentally relevant fluoroquinolone concentrations (i.e.,