Efficient Photocatalytic Bilirubin Removal over the Biocompatible Core/Shell P25/g-C3N4 Heterojunctions with Metal-free Exposed Surfaces under Moderate Green Light Irradiation

Highly-monodispersed g-C3N4/TiO2 hybrids with a core/shell structure were synthesized from a simple room temperature impregnation method, in which g-C3N4 was coated through self-assembly on the commercially available Degussa P25 TiO2 nanoparticles. Structural and surface characterizations showed that the presence of g-C3N4 notably affected the light absorption characteristics of TiO2. The g-C3N4/TiO2 heterojunctions with metal-free exposed surfaces were directly used as biocompatible photocatalysts for simulated jaundice phototherapy under low-power green-light irradiation. The photocatalytic activity and stability of g-C3N4/TiO2 were enhanced relative to pure P25 or g-C3N4, which could be ascribed to the effective Z-scheme separation of photo-induced charge carriers in g-C3N4/TiO2 heterojunction. The photoactivity was maximized in the 4 wt.% g-C3N4-coated P25, as the bilirubin removal rate under green light irradiation was more than 5-fold higher than that under the clinically-used blue light without any photocatalyst. This study approves the future applications of the photocatalyst-assisted bilirubin removal in jaundice treatment under moderate green light which is more tolerable by humans.


Results and Discussion
Schematic illustration of the synthesis process. Figure 1 shows this simple self-assembly impregnation route for synthesis of g-C 3 N 4 /TiO 2 heterojunctions. During the process, the commercially available P25 TiO 2 nanoparticles was considered as a host and template for the self-assembly of g-C 3 N 4 nanosheets, while the cheap and recyclable melamine was used as the sole dispersion liquid. As reported, in methanol, stack-layered g-C 3 N 4 can easily be exfoliated into thin nanosheets, which are prone to a rolling and regrowth process 14 . In this impregnation, the g-C 3 N 4 nanosheets could curl up and wrap around the P25 TiO 2 nanoparticles to minimize the total interfacial energy, and then reassemble into a homogeneous coating layer after the methanol was removed by the air stream. The main advantage of this method is that the entire composite process is waste-free and achievable at room temperature. Nevertheless, introducing a very low amount of g-C 3 N 4 (< 10 wt.%) can stabilize the metal-free surfaces of photocatalysts. Because of environmental friendliness, low cost and high yield, this method is appealing for future application of photocatalyst-assisted phototherapy of neonatal hyperbilirubinemia. The g-C 3 N 4 /TiO 2 heterojunctions as-prepared were named as PCNx, where x is the weight percentage of g-C 3 N 4 . Crystal structure. Figure 2 shows the X-ray diffraction (XRD) patterns of PCNx. The diffraction peaks at 2θ = 25.3, 37.7, 47.9, 53.8, 62.6 and 74.9° can be ascribed to (1 0 1), (0 0 4), (2 0 0), (1 05), (2 0 4) and (2 1 2) reflection of anatase TiO 2 , respectively, while the peaks at 2θ = 27.8, 35.9, 54.9 and 68.8° can be ascribed to (1 1 0), (1 0 1), (2 1 1) and (3 0 1) reflection of rutile TiO 2 , respectively 18,19 . The patterns clearly illustrate that the crystals of PCN2, PCN4 and PCN8 are all composed of 80% anatase TiO 2 and 20% rutile TiO 2 , which are nearly the same as P25 TiO 2 . The predominant anatase phase in P25 TiO 2 was sustained after g-C 3 N 4 modification. The unique mixcrystal structure of P25 TiO 2 is considered to be favorable for photocatalysis applications. Since the anatase-to-rutile phase transformation of TiO 2 occurs at around 600 °C, the low-temperature treatment benefits the crystal stability of P25 TiO 2 -based photocatalyst. Pure g-C 3 N 4 shows two distinct diffraction peaks at 2θ = 13.2° and 27.4°, corresponding to the (1 0 0) and (0 0 2) peaks of the graphitic phase, respectively 20 . However, the XRD patterns of PCNx are not changed notably after coating with g-C 3 N 4 because of the low content (< 8%) and the low XRD intensity of g-C 3 N 4 . Nevertheless, the co-presence of TiO 2 and g-C 3 N 4 was confirmed by X-ray photoelectron spectroscopy (XPS) ( Figure S1). Besides, PCNx have the similar average crystal sizes as P25, demonstrating the encapsulation of g-C 3 N 4 layers also suppresses the growth of TiO 2 crystals under the impregnation coating treatment.
Nitrogen adsorption analysis. The specific surface areas (SSAs) of PCNx were determined from Nitrogen adsorption and desorption isotherms (Fig. 3). The insets in Fig. 3 show the Barrett-Joyner-Halenda (BJH) pore distributions of pure P25 TiO 2 and PCNx. Unlike pure P25 TiO2, the isotherm curves of the PCN2, PCN4 and PCN8 exhibit a distinct uptake of N2 as a result of capillary condensation in a wide relative pressure (P/P0) range of 0.4-0.95, which indicates the existence of multiform pore distributions. The low-pressure hysteresis loop (0.4 < P/P 0 < 0.8) is related to the intra-aggregated pores of g-C 3 N 4 -TiO 2 , while the high-pressure hysteresis loop (0.8 < P/P 0 < 0.95) is probably associated with the larger pores formed between secondary particles 21 . The Brunauer-Emmett-Teller SSAs (S BET ), pore volumes, and pore sizes of all samples are summarized in Table 1. The S BET of PCNx as well as the pore volumes decreases slightly compared to pure P25 TiO 2 , which is ascribed to the adhesion and self-aggregation of P25 nanoparticles in liquids during impregnation. Nevertheless, the S BET of the representative PCN4 is up to 45.79 m 2 ·g −1 , which is favorable compared with other reports of TiO 2 based  TiO 2 /g-C 3 N 4 photocatalysts 22,23 . The S BET of show slightly variations with g-C 3 N 4 mass ratio increased, in which PCN8 sample with a highest g-C 3 N 4 content processed a lower S BET compared with that of PCN4, the possible reason was that the excessive g-C 3 N 4 nanosheets cannot effectively hybrids with P25 nanoparticles and agglomerate into g-C 3 N 4 cluster with low specific surface areas.
Morphologic characterization. SEM images of P25 TiO 2 and PCN4 in Fig. 4 suggest that the highly monodispersed particle distribution has been maintained after g-C 3 N 4 self-assembly coating. PCN4 preserves a macroscopic network structure with a relatively regular array of macropores (Fig. 4d). The macropores also have similar sizes as pure TiO 2 . It is interesting that the surface of PCN4 is roughened after g-C 3 N 4 coating. From the morphology comparison of bare P25 TiO 2 and PCN4, similar highly-integrated composite systems can be observed, indicating a homologous surface area of PCN4 with P25, which is in good agreement with the BET results. In general, the particles of PCN4 are highly-monodispersed, which benefits its application in phototherapy.

Sample
Mass ratio of g-C 3 N 4 and TiO 2 (%) S BET (m 2 /g) Pore size (nm) Pore volume (cm 3 /g) Eg (eV)  The morphological and structure of PCNx heterojunctions were further investigated by transmission electron microscopy (TEM) and high-resolution TEM (HRTEM). As showed in Fig. 5a,b, the particle sizes of TiO 2 in pure P25 TiO 2 and PCN4 both range from ca. 25 to 35 nm, which well agrees with the sizes of P25 TiO 2 determined from XRD. The well-dispersed g-C 3 N 4 layers are deposited all over the surface of P25 TiO 2 nanoparticles. As showed on the high-resolution images of P25, PCN2, PCN4 and PCN8 ( Fig. 5c-f), the lattice pitch of TiO 2 in PCN4 is 0.351 nm, which is in accordance with (1 0 1) lattice plane character of anatase TiO 2 24 and similar to that of bare P25 TiO 2. There is no difference on lattice pitch among the PCNx composites, indicating our synthetic method is moderate and does not affect the lattice stability of P25 TiO 2 . It should be noted that a close coating layer appears uniformly on the surface of TiO 2 nanoparticles, which is probably formed by the stacking of g-C 3 N 4 nanosheets (Fig. 5b). The g-C 3 N 4 coating shells of PCN2, PCN4 and PCN8 are about 1, 2 and 3 nm thick, respectively. Hence, the layered shells are thickened with the increase of g-C 3 N 4 :TiO 2 mass ratio. These g-C 3 N 4 shells not only keep a biocompatible metal-free surface, but also form g-C 3 N 4 /TiO 2 heterojunctions with excellent VLD photocatalytic performance. However, an excess of g-C 3 N 4 coating shell will reduce the light harvest of g-C 3 N 4 / TiO 2 heterojunctions and block the transfer channel of mass and free radical during photocatalysis, which will be discussed in detail later.
FT-IR spectra. Figure 6 shows the FT-IR spectra of pure g-C 3 N 4 , PCN2, PCN4 and PCN8. For pure P25 TiO 2 , the main peaks at 400-700 cm −1 are assigned to the stretching vibrations of Ti-O-Ti and Ti-O in anatase crystals, while the other two wide peaks at 1650 and 3400-3500 cm −1 correspond to hydroxyl group and physically-adsorbed water, respectively. The spectrum of pure g-C 3 N 4 shows the strong bands within 1200-1650 cm −1 , with peaks at 1238, 1320, 1406, 1459, 1547, 1572 and 1639 cm −1 , which correspond to the typical stretching vibrations of the sp3 C-N bonds and sp2 C = N heterocycles 25 . Additionally, the peak at 807 cm −1 is due to the out-of-plane skeletal breathing of triazine. Besides, a wide band between 3000 and 3400 cm −1 corresponds to the N-H stretching vibration of residual NH 2 attached to the sp2-hybridized carbon 26 . The NH 2 band is not obvious in PCNx. Nevertheless, the stretching vibration intensification of C-N and C = N heterocycles in the spectra clearly indicates the presence of g-C 3 N 4 with the increase of the g-C 3 N 4 /TiO 2 ratio. Ultraviolet-visible diffuse reflectance spectroscopy (UV-vis DRS). Figure 7 shows the UV-vis DRS spectra of PCNx. The strong absorption throughout the UV region is characteristic of TiO 2 . For bare P25 TiO 2 and pristine g-C 3 N 4 , the estimated band gaps are 2.90 and 2.75 eV, respectively. It can be inferred from DRS that in addition to the typical absorption band of TiO 2 , a second small shoulder appears at higher wavelength in the spectra of PCNx. This means PCNx photocatalysts exhibit broader absorption and narrower band gap, owing to the formation of g-C 3 N 4 /TiO 2 heterojunctions 27 . The absorption area of PCNx in the green light range (490-550 nm) is enlarged, which provides a basic possibility for catalyst-assisted phototherapy under green light irradiation. The band gap energies of indirect semiconductors can be estimated by Kubelka− Munk transformation. The calculated band gap values also change with the increase of g-C 3 N 4 content ( Table 1). PCN4 has the most optimized band gap of 2.51 eV. Clearly, the resulting g-C 3 N 4 /TiO 2 functional material shows an enhanced capacity of green light harvest due to the narrow band gap of g-C 3 N 4 /TiO 2 heterojunctions, especially for PCN4.
Photocatalytic activity and photostability under green light irradiation. The photocatalytic activities of the PCNx were evaluated by bilirubin degradation experiments using a 300 umol·L −1 bilirubin solution, in which the bilirubin concentration was similar to that in the blood of neonatal jaundice patients. The irradiation intensity was set at 5 mW/cm 2 . Figure 8a shows the bilirubin degradation curves in aqueous solutions with the presence of a photocatalyst under 495 nm green light irradiation, where C 0 and C are the bilirubin concentrations under equilibrium and after visible-light irradiation, respectively. Prior to irradiation, the bilirubin solutions added with catalyst were kept in the dark for 40 min to reach adsorption equilibrium. Also a bilirubin solution without addition of any catalyst was set as a blank control. Results show the bilirubin degradation was very low in the dark or under green light irradiation without any photocatalyst, indicating the self-degradation of bilirubin was not obvious. Meanwhile, pure P25 and g-C 3 N 4 also present rather low bilirubin degradation rates. However, the thin g-C 3 N 4 -modified TiO 2 photocatalysts were much more photoresponsive to green light irradiation. Bilirubin photoalteration was significantly promoted with the presence of PCNx under green light, in which bilirubin concentration steeply decreased by ~70% after 120 min. The bilirubin degradation efficiencies after 60 min were 15.3%, 13.7%, 29.2%, 55.7%, 75.9% and 65.9% in the blank and with the presence of P25, g-C 3 N 4 , PCN2, PCN4 and PCN8, respectively, indicating that bilirubin was degraded more efficiently by PCNx than by the pure P25 or g-C 3 N 4 . Over the clinically relevant normal limit of 100 umol·L −1 , the decrease of bilirubin concentration was drastic and significant, indicating the reaction follows a pseudo-first-order procedure. Among all the PCN  catalysts, the optimum carbon nitride content envisaged was found in PCN4, which nearly photoreduced half of bilirubin only after 40 min. It is noticeable that the physical mixing of TiO 2 and g-C 3 N 4 led to a similar photoactivity of pure TiO 2 . This fact clearly denotes a synergetic effect to impregnation preparation, which provides a better junction between the two catalysts. Provided that the reaction follows a pseudo-first-order procedure, we estimated the bilirubin decomposition rate of PCN4 was 2.38 umol·min −1 , much higher than the blank, P25 and g-C 3 N 4 (0.59, 0.57 and 1.09 umol·min −1 , respectively). Therefore, photocatalytic bilirubin degradation was successfully achieved by g-C 3 N 4 /TiO 2 heterojunctions under illumination of 495 nm green light.
Regarding the safety requirement in biochemical applications, we further evaluated the stability of pure g-C 3 N 4 , bare P25 TiO 2 and PCN4 through reuse experiments after bilirubin photodegradation ( Figure S2). After five cycles of bilirubin photocatalytic degradation, PCN4 remained high activity, confirming that the g-C 3 N 4 /TiO 2 heterojunctions were effective and stable during the photocatalysis. The comparison of XRD patterns between fresh PCN4 and used PCN4 ( Figure S3) shows almost no difference, which demonstrates the excellent structural stability of the g-C 3 N 4 /TiO 2 heterojunctions.

Comparative photocatalytic activity under different irradiations.
To compare the practical values in bilirubin degradation between photocatalyst-assisted green light phototherapy and the currently-used blue light-induced phototherapy, we recorded the photocatalytic performances of all systems using different light sources and photocatalysts (Fig. 8b). In addition, a 20 mW/cm 2 blue light source was introduced to simulate the high-intensity blue light phototherapy condition. Clearly, the photodegradation rate under 5 mW/cm 2 green light without any catalyst was rather low. The bilirubin photodegradation rate under 5 mW/cm 2 blue light was two times of that under 5 mW/cm 2 green light irradiation. However, the photodegradation effect under 5 mW green light irradiation was promoted significantly by PCN4. The bilirubin degradation activity with the presence of PCN4 under 5 mW green light was almost 6 times that under 5 mW/cm 2 blue light irradiation without catalyst, and was even better than the effect under 20 mW/cm 2 high-intensity blue lightirradiation without any photocatalyst. This result confirms that the PCNx can affect the efficacy of bilirubin phototherapy, which provides a new promising path based on moderate green light irradiation.

Mechanisms of photocatalytic activity enhancement.
Generally, photocatalytic activity is mainly governed by surface properties, light-absorption ability and photogenerated charge-separation efficiency. The BET experiments above indicate that the coating of g-C 3 N 4 does not substantially affect the SSAs of PCNx compared with P25 TiO 2 . However, the light-absorption ability was improved by g-C 3 N 4 coating treatment in forming g-C 3 N4/TiO 2 heterojunctions, as indicated by UV-vis DRS. The photogenerated charge separation process was investigated by electrochemical impedance spectroscopy (EIS). The arc radius on EIS Nyquist plot of PCN4 under green light irradiation is smaller than that of g-C 3 N 4 and P25 ( Figure S5), indicating the effective separation of photogenerated electron-hole pairs and fast interfacial charge transfer in g-C 3 N 4 /TiO 2 heterojunctions under green light excitation. Therefore, the close g-C 3 N 4 and TiO 2 interaction is thus an important influence factor on the photocatalytic activity of TiO 2 . Meanwhile, the amount of g-C 3 N 4 in PCNx should be prudently controlled. When the g-C 3 N 4 coating was very low as in PCN2, the photogenerated charges cannot efficiently migrate due to the poor electroconductivity of g-C 3 N 4 , thus reducing the heterojunction effect. However, with excessive g-C 3 N 4 as in PCN8, a dense g-C 3 N 4 shell around TiO 2 was gradually formed, which prevented TiO 2 from harvesting light energy and led to a serious recombination of electron-hole pairs. The optimum coating of g-C 3 N 4 was 4 wt.%, which enhanced photocatalytic activity and inhibited photocorrosion. This mechanism of photocatalytic activity enhancement was approved by former reports 15 .
To investigate and identify the main active oxidation species generated in the photocatalytic process responsible for the degradation of bilirubin under green light, radical trapping experiments were performed in the presence of KI (a trapping scavenger for h + and ⋅ OH radicals on the catalyst surface), methanol (CH 3 OH, a trapping scavenger for ⋅OH radicals in the solution) and 1,4-benzoquinone (C 6 H 4 O 2 , a trapping scavenger for O⋅ 2− ), respectively. The results shown in Fig. 9 indicate that the degradation efficiencies of bilirubin significantly decreased from 94.1% to 23.4% after 120 min in the presence of 4 mM KI. When 4 mM C 6 H 4 O 2 was added to the reaction system, the degradation efficiency of bilirubin slightly decreased to 52.9%. However, 100 mM CH 3 OH just affect the degradation efficiency a little. These results indicated that h + on the catalyst surface is the most important oxidising species during the photocatalytic process, that O⋅ 2− radical is also responsible for the degradation of bilirubin, and that ⋅ OH radical in the solution is not the main active species.
As far as we know, due to the narrow band gap of 2.67 eV and a relative more negative CB position of approximately − 1.1 eV of g-C 3 N 4 , the photogenerated charges separation and transfer model in g-C 3 N 4 /TiO 2 can follow heterojunction-type or Z-scheme mechanism [28][29][30] . Comparing these two models, the active species in the Z-scheme model are photogenerated hole, superoxide anion radical and hydroxyl radical, but the active species of heterojunction-type model only involves superoxide anion radical. In this case, the reactive species trapping experiment results clearly demonstrated that the holes play the leading role in photocatalytic remove of bilirubin, while no obvious ⋅ OH radical can be found. The main reason was that the VB potential of CN is lower than that of the normal potential of the OH − /⋅ OH ( + 2.4 V versus NHE), thus the photo-generated holes on the surface of g-C 3 N 4 cannot react with OH − /H 2 O to form ⋅ OH radical. Therefore, we propose the separation and transfer model of as-prepared Core/Shell P25/g-C 3 N 4 under green light was a Z-scheme, in which photogenerated holes directly consumed by the bilirubin molecules attract on the surface of g-C 3 N 4 /TiO 2 . This is in accordance with previous reports 31,32 . Prospects of clinical application. In this work, the more important issue was to clarify the mechanism and potential of the therapeutic protocols of photocatalyst-assisted bilirubin removal under moderate green light irradiation compared with the high-intensity blue-light-driven phototherapy. Previous animal experiments showed that skin bilirubin of jaundiced rats under blue or green light irradiation was converted to metastable geometric isomers, which were then transported in the blood and excreted in the bile 33,34 . The same reaction probably occurred in light-exposed jaundiced babies, particularly during phototherapy. Since blue light irradiation can also be strongly absorbed by competing hemoglobin which is rich in human blood, the optical intensity in blue light therapy must be high enough to ensure the bilirubin molecules receive enough energy for efficient photoalteration 35 . The undesirable high-intensity irradiation causes health problems such as hyperthermia, insensible water loss, rash and even DNA damages. The use of bio-tolerant green light is attractive regarding the health protection of newborn babies which is the most popular in neonatal jaundice patients (Fig. 10). In this work, photocatalysis experiments show g-C 3 N 4 /TiO 2 heterojunction can improve the efficacy of bilirubin phototherapy under relatively moderate green light irradiation. Our preliminary toxicology experiments in mouse fibroblast cell line L-929 showed that the toxicity of PCNx was between grade 0 and 1, and the material lacked hemolytic activity, which indicates the high biocompatibility of this g-C 3 N4/TiO 2 heterojunction. Therefore, the g-C 3 N4/ TiO 2 hybrids here are suitable for further biochemical applications. This provides a new pathway to design photocatalyst-assisted neonatal jaundice therapeutic protocols based on moderate green light as an ideal irradiation source in the future, such as services as photosensitizer in photodynamic therapy, undissolved injectable in-situ forming implants or non-invasive bleaching membrane 36,37 .

Conclusions
The g-C 3 N 4 /TiO 2 heterojunctions with metal-free surfaces were synthesized through a simple impregnation route. Multiple characterization techniques revealed that the close integration of g-C 3 N 4 and TiO 2 in g-C 3 N 4 /TiO 2 heterojunction extended its visible light response and promoted its charge-separation efficiency. The g-C 3 N 4 /TiO 2 nanocomposites were used as high-performance photocatalysts for neonatal jaundice phototherapy by photocatalytic removal of bilirubin under green light irradiation. The 4 wt.% g-C 3 N 4 /TiO 2 heterojunction demonstrated a remarkable photocatalytic performance of 50% bilirubin removal rate under 495 nm green light within 40 min, as well as an excellent cycle performance. The enhancement of photocatalytic activity after coating may be mainly attributed to the Z-scheme synergic effect between g-C 3 N 4 and TiO 2 , which enhances the green light harvest and greatly accelerates the separation of photogenerated carriers. The bilirubin removal effect with the presence of Scientific RepoRts | 7:44338 | DOI: 10.1038/srep44338 g-C 3 N 4 /TiO 2 photocatalysts was clearly higher than that under clinically-used high-intensity blue light without any photocatalyst. These findings provide insight of an instant, noninvasive and cheap neonatal jaundice treatment protocol of semiconductor photocatalyst-assisted phototherapy under moderate green light irradiation.
Preparation of g-C 3 N 4 /TiO 2 photocatalysts. The g-C 3 N 4 /TiO 2 heterojunctions were prepared according to a reported procedure except using P25 TiO2 as the substrate. Firstly, bulk g-C 3 N 4 was prepared by simple calcination of melamine at 550 °C for 4 h in a covered alumina crucible. Then 0.1 g of g-C 3 N 4 was added into 100 mL of methanol and sonicated for 2 h to make the bulk g-C 3 N 4 exfoliated into g-C 3 N 4 thin sheets, forming suspension A. Meanwhile, 1 g of P25 TiO 2 power was added into 100 mL of methanol and sonicated for 10 min, forming suspension B. The g-C 3 N 4 /TiO 2 heterojunctions were synthesized by simply mixing appropriate amounts suspension A and suspension B together and stirring them at room temperature in the dark for 24 h. Afterwards, the mixtures were evaporated at room temperature to remove the methanol, forming composite photocatalysts.
Characterization. XRD patterns were recorded on a Rigaku D/MAX-2495VB/PC diffractometer under the following conditions: θ -2θ mode, CuKα 1 radiation (λ = 1.5406 Å), 40 kV, 100 mA, and scanning step 0.02 ° per sec. The average crystal size was determined from XRD pattern parameters according to the Scherrer equation. SEM images were recorded using an FEI XL-30 scanning electron microscope, operated at an acceleration voltage of 25 kV. TEM images were taken on a JEOL JEM-2010 TEM device with an acceleration voltage of 200 kV. The Brumauer-Emmett-Teller (BET) specific surface areas were calculated based on nitrogen sorption isotherms which were recorded on a BeiShiDe 3H-2000PS4 device at − 196 °C, pore size distributions were deduced via the Barrett-Joyner-Halenda (BJH) method. Total pore volume (V t ) was determined at a relative pressure of 0.98. XPS data were recorded by a PHI 5000 C ESCA X-ray photoelectron spectrometer with Al Kα source at 14.0 kV and 25 mA. All the binding energies were referenced to the contaminant C 1 s peak at 284.6 eV. The ultraviolent-visible (UV-vis) diffuse reflectance spectra (DRS) were achieved using a Shimadzu UV-2401 UV-vis spectrometer equipped with integrating sphere. Fourier transform infrared (FTIR) spectra were tested on a Nicolet Nexus 470 FTIR Spectrometer. Electro-chemical impedance spectroscopy (EIS) tests were performed at open circuit potential over the frequency range between 100 kHz and 0.1 Hz. Evaluation of photocatalytic activities. The photocatalytic performances of g-C 3 N 4 /TiO 2 photocatalysts were determined by degrading bilirubin under 595 nm monochromatic green lights from a 100 W lamp which was powered by an LED system (CEL-LED100, Beijing CEAULIGHT Co. Ltd.). The excitation intensity can be controlled by tuning the applied current of the lamp and determined by a LED luminous intensity meter (ALTT-86LB, Beijing CEAULIGHT Co. Ltd.). Photocatalytic degradation of bilirubin was performed in a bilirubin solution (50 mL) with an initial concentration of 300 umol·L −1 . The bilirubin solution was stirred in the dark after the photocatalysts were added for 40 min until adsorption-desorption were balanced between the catalyst and contaminants. Then the resulting mixtures were exposed to green light. Afterwards, 1 mL of aliquots were periodically withdrawn from the reaction vessel every 20 min, diluted and centrifuged at 10,000 rpm for 5 min to separate the catalysts. Bilirubin levels in solution were measured by a UV-vis spectrophotometer (Spectrumlab 752 s, Xunda, Shanghai) at λ = 483 nm. The photocatalytic decolorization is a pseudo-first-order reaction: ln(C 0 /C t ) = kt, where the k is the apparent rate constant, C 0 and C t represent the concentration of bilirubin at initial stage and after irradiation for some time, respectively. For comparison, the bilirubin removal rate under blue light was investigated by using a monochromatic LED light of 440 nm from the 100 W LED lamp by the same method. In the photocatalytic stability evaluation, the reaction photocatalysts were washed by centrifugation at the end of each cycle to remove the organic residues.