Metal-free graphitic carbon nitride (g-C₃N₄) has attracted great interest in photocatalytic degradation and H₂ production owing to its low cost, nontoxicity, good thermal and chemical stability, visible-light response, and proper electronic band structure. However, fast carrier recombination, low optical absorption in the visible region, and small surface area dramatically impact its overall performance [1, 2].

Over the past few years, many approaches have been investigated to improve the photocatalytic activity of g-C₃N₄ [2–4]. Doping metallic or nonmetallic elements is an efficient strategy to modify the surface properties and electronic structures [5–8]. Reducing the particle size or forming a hierarchical structure is also an effective method for addressing the drawbacks of g-C₃N₄ by enlarging the specific surface area, shortening the diffusion pathway of charge carriers, and further promoting the separation and transfer of electron-hole pairs [4]. Besides, constructing a heterojunction is another way of improving the photocatalytic performance of g-C₃N₄. Various semiconductors have been reported to form heterostructures with g-C₃N₄, such as TiO₂, MoS₂, Bi₂WO₆, Ag₃PO₄, NiS₂, and carbon quantum dots [9–15], which can suppress the recombination of photoinduced electrons and holes.

CdS is a fascinating visible-light photocatalyst for H₂ evolution owing to the narrow bandgap (2.42 eV) and suitable band edge positions [16, 17]. However, CdS is unstable owing to the photocorrosion effect of the photogenerated holes, fast recombination of the photoexcited charge carriers, and aggregation of CdS nanoparticles, which limit its application as a photocatalyst to a great extent. These problems can be solved by coupling CdS with various semiconductors (TiO₂, BiVO₄, ZnO, Fe₂O₃, etc.) [18–21]. These heterostructures show enhanced photocatalytic activities. The band edge offsets between g-C₃N₄ and CdS are suitable for charge transfer, and therefore, the coupling of CdS with g-C₃N₄ can effectively accelerate the separation of photoinduced charge carriers and alleviate the photocorrosion occurring in CdS due to photogenerated holes [16, 17, 22–24]. However, as Ge et al. [25] reported that the H₂ evolution efficiency was not as high as expected, being only ~9 times that of pure g-C₃N₄. More efforts are therefore required to further improve the photocatalytic activity of g-C₃N₄ and CdS hybrid.

The porous one-dimensional nanostructure offers a great advantage in terms of improving both the specific surface area and charge carrier mobility [26–28]. Herein, we designed a porous tubular g-C₃N₄ by using ethylene glycol as a template, which could be removed easily. The porous structures provide higher specific surface areas than pure g-C₃N₄, which make it an ideal substrate for loading CdS nanoparticles. The g-C₃N₄ nanotube can inhibit the photocorrosion of CdS and promote the separation of the photogenerated electrons and holes. The photocatalytic activities of porous g-C₃N₄/CdS nanotubes for the degradation of RhB and the photocatalytic hydrogen evolution reaction (HER) under visible light irradiation were determined. This study may provide a new insight into the design of photocatalysts with high visible light activities.

All the reagents used were purchased from Aladdin Chemical Reagent Co., Ltd and Sinopharm Chemical Reagent Co., Ltd. Deionized water was used throughout this study. All the chemicals were analytical grade and used as-received without any further purification.

Briefly, 1 g of melamine was dissolved in 30 mL ethylene glycol at 60 ℃ with continuous magnetic stirring to form a transparent solution. Afterwards, 60 mL of 0.15 mol/L aqueous nitric acid was added drop by drop. After cooling to room temperature, the generated white precipitate was collected and washed with ethanol three times and dried in a vacuum oven at 60 ℃ overnight.

The as-prepared MNFs were sealed in a quartz boat and enfolded by a tin foil with a few pinpricks. Then the sample was heated to 550 ℃ for 2 h at the ramp rate of 10 ℃/min. The obtained product was washed with deionized water three times and dried in the vacuum oven at 60 ℃ overnight. For comparison, bulk g-C₃N₄ (B-CN) was prepared by direct calcination of melamine at 550 ℃ for 2 h by employing the same heating rate.

Typically, 0.1 g of T-CN was sonicated and dispersed in 80 ml deionized water. Afterwards, 1 mmol (5, 10, and 20 mmol) of cadmium chloride and thioacetamide with cadmium/sulfur stoichiometric ratio of 1:1 were added into the above solution. Then, the solution was heated and stirred at 60 ℃ for 30 min, and the resulting yellow precipitate washed with ethanol and deionized water several times and dried at 60 ℃ overnight. The prepared samples were named T-CN/CdS-x (x = 1, 5, 10, and 20), where x represents the molar weight of CdS. The entire synthesis process is illustrated in Scheme 1.

The Fourier transform infrared spectroscopy (FT-IR) patterns of the photocatalysts were obtained in a transmission mode between 4000 and 400 cm^–1. The crystalline phases were characterized by using an X-ray diffractometer (Bruker D8 Advance with Cu K_α1 radiation at 40 kV and 30 mA). The morphology and sample size were determined by scanning electron microscopy (SEM; HITACHI S-4800) and transmission electron microscopy (TEM; JEM2100), respectively. X-ray photoelectron spectroscopy (XPS) measurements were carried out on a VG ESCALAB 220i-XL system. The Brunauer-Emmett-Teller (BET) surface area was measured by an ASAP 2010 V5.02H analyzer. The UV-vis diffuse reflectance spectra (DRS) of the samples were obtained by using a UV-vis spectrophotometer (UV-3600, Shimadzu, Japan). The photoluminescence (PL) spectra were measured with a Hitachi F-7000 fluorescence spectrophotometer at room temperature.

Electrochemical tests were performed by using a three-electrode system with an electrochemical workstation (CHI660C Instruments, China). Fluorine-doped tin oxide conductive glass, platinum wire, and saturated Ag/AgCl electrode were used as the working electrode, counter electrode, and reference electrode, respectively. Here, 0.5 mol/L aqueous Na₂SO₄ solution acted as the electrolyte. A 300 W xenon lamp with a UV-cutoff filter (λ ≥ 420 nm) was used as the visible light source. The photocurrent responses were obtained by using a photochopper to simulate the light/dark conditions at 25 s intervals. The electrochemical impedance spectra (EIS) were recorded over the frequency range 1–10⁵ Hz with an amplitude of 10 mV at the same bias voltages. The Mott-Schottky plots were recorded by using the impedance-potential mode in the dark.

For all the photocatalytic experiments, the distance between the xenon lamp (300 W with a 420 nm UV-cutoff filter) and the surface of the reacting solution was maintained at 10 cm. A mass of 0.04 g of the catalyst was added to 100 mL of RhB solution (10 mg/L). Prior to irradiation, the suspension was magnetically stirred in the dark for 30 min to establish adsorption-desorption equilibrium. At specific time intervals (15 min), 3 mL of the suspension was collected and centrifuged (8000 rpm, 10 min) to remove the photocatalyst particles thoroughly. The absorbance of RhB was determined by the UV-vis spectrophotometer at the maximum absorption wavelength of 554 nm.

According to Lambert-Beer theory A = ε × b × c, where A refers to the absorbance of the sample, ε is the molar absorption coefficient, b is the width of the sample cell, and c is the molar concentration of RhB, we know that the absorbance of RhB is directly proportional to its concentration, and the degradation efficiency (X) can be expressed as X = (C₀–C_t)/C₀ × 100%, where C₀ is the initial adsorption equilibrium concentration and C_t is the molar concentration of RhB at different times.

The photocatalytic H₂ evolution tests were conducted in a 250 mL sealed quartz flask at a constant temperature and negative pressure. A 300 W xenon lamp was used as the light source with a cutoff filter (λ ≥ 420) nm to trigger the photocatalytic reaction. In a typical experiment, 0.1 g of the photocatalyst with 3 wt% platinum loading was dispersed in 100 mL of an aqueous solution containing 10 vol% triethanolamine (TEOA). Before irradiation, the system was bubbled with N₂ to remove air completely. The H₂ was analyzed by a gas chromatograph (Agilent, 7820A) equipped with a thermally conductive detector (TCD) and a 5 Å molecular sieve column, with high-purity N₂ serving as the carrier gas. In the cyclic tests, the system was bubbled with N₂ to completely remove the residual H₂ before the next cycle.

The morphological images of the as-prepared samples were collected by SEM and TEM. It can be observed from the SEM image (Fig. 1a) that vimineous and uniform MNFs exhibit high aspect ratios, with the sectional width being less than 2 μm. T-CN with numerous porous structures can be confirmed through SEM in Fig. 1b. Owing to the high reactivity of the amine of the melamine molecule, it can be easily protonated by treating with dilute aqueous HNO₃. Thus, it can be deduced that the fibrous crystal is a class of salt that is composed of protonated melamine and nitric ions. The existence of both H⁺ and NO₃^– is prerequisites for fibrous self-assembly, which can be maintained by electrostatic attraction, p–p stacking interactions, and H bonding. After the thermal treatment, a porous structure can be formed with the release of NH₃ molecules [29]. It can be seen from Fig. 1c and the inset of Fig. 1c that CdS nanoparticles with an average size of around 20 nm disperse uniformly on the surface of T-CN/CdS-10 composite. HRTEM (Fig. 1d) shows that lattice spacings are 0.332 and 0.302 nm, corresponding to the (002) lattice planes of g-C₃N₄ and (111) lattice planes of CdS, respectively, which indicate the coexistence of g-C₃N₄ and CdS. The elemental mapping images (Fig. 1e) verify the uniform distribution of carbon, nitrogen, cadmium, and sulfur elements. The weight rate of CdS is about 11.0%, as shown in Fig. 1f.

The crystal structure and phases of the samples were investigated through X-ray diffraction (XRD) analysis, and the results are shown in Fig. 2a. There are two similar diffraction peaks for bulk and tubular g-C₃N₄. The former peak located at 13.1° corresponds to the in-plane structural packing of tri-s-triazine units ((100) crystal planes), while the peak at 27.5° can be ascribed to the interplanar stacking of conjugated aromatic system ((002) crystal planes) [30, 31]. It is remarkable that, because of the less ordered stacking of the tri-s-triazine motifs, the peak corresponding to the (100) crystal plane of tubular g-C₃N₄ is much weaker than that of bulk g-C₃N₄. The XRD data of pure CdS reveal characteristic 2θ peaks at 24.9°, 27.1°, 36.8°, 44.3°, 48.2°, and 52.4°, which correspond to the hexagonal wurtzite structure (JCPDS no. 65-3414). Besides, the main characteristic diffraction peaks of T-CN/CdS composites gradually shift from T-CN to CdS with the increase in the proportion of the CdS hybridized to T-CN.

The chemical structures were analyzed by using FT-IR. As shown in Fig. 2b, the sharp peak centered at ~810 cm^–1 corresponds to a breathing mode of the s-triazine units of g-C₃N₄. On the other hand, the absorption bands in the range 1200–1750 cm^–1 can be attributed to the aromatic C–N stretching vibration mode. The broad band located at ~3150 cm^–1 corresponds to the stretching mode of the N–H bond [32–34]. Meanwhile, the typical absorption peaks of CdS emerge in all the composites, which indicate that CdS was successfully deposited with T-CN by the present synthetic route without any structural changes. As for the pure CdS sample, the bands at 1400, 1290, 1106, 963, and 701 cm^–1 are attributed to the Cd–S bond. The broad band centered at 3420 cm^–1 and the band observed at 1617 cm^–1 originate from the water molecules absorbed on the surface of the sample [35, 36]. However, no characteristic peaks related to the Cd–S bond appear in the patterns of the T-CN/CdS-x composites owing to the low contents of CdS.

XPS was carried out to analyze the surface elementary compositions and chemical states. The C 1s spectrum of T-CN/CdS-10 (Fig. 3a) could be deconvoluted into three peaks centered at 288.6, 286.4, and 284.8 eV; the first two peaks correspond to the binding energy of the carbon atom in the C–(N)₃ and N–C=N bonds of g-C₃N₄, whereas the third peak could be attributed to the adventitious hydrocarbon of the instrument [37]. There are three main peaks in the high-resolution N 1s spectra (Fig. 3b): the characteristic peak at 400.2 eV can be assigned to the delocalized π-electron of the C–N heterocycle, the peak at 398.5 eV can be assigned to the binding energy of the sp³-hybridized nitrogen atom in the N–(C₃N₃)₃ bond, and the peak at 397.8 eV can be assigned to the binding energy of the sp³-hybridized nitrogen atom in the C=N–C bond. It can also be observed from the high-resolution C 1s and N 1s spectra that the peaks of T-CN/CdS-10 slightly shift to higher binding energies when compared with those of pure T-CN, indicating the presence of strong interfacial interactions between g-C₃N₄ and CdS. Additionally, in the Cd 3d orbital scan spectrum (Fig. 3c), the peaks located at 412.3 and 404.9 eV could be attributed to the Cd²⁺ of CdS. In the S 2p orbital scan spectrum (Fig. 3d), the peaks located at 162.4 and 161.2 eV could be attributed to the S^2– of CdS [35]. Based on the above analysis, it can be confirmed that the graphic phases C₃N₄ and CdS coexist in T-CN/CdS-10 sample with a strong interfacial force.

Low-temperature N₂ adsorption-desorption experiments were carried out to analyze the specific surface area and pore size distribution of the photocatalyst because these two factors play essential roles in determining the properties of photocatalysts. The N₂ adsorption-desorption isotherms of B-CN and T-CN displayed in Fig. 4 confirm that both these samples exhibit type Ⅳ isotherms, which indicate the presence of mesoporous and macroporous structures. The small hysteresis loop of B-CN at high P/P₀ is caused by the aggregation of the B-CN particles. In contrast, T-CN exhibits an obvious hysteresis loop at high P/P₀, indicating the presence of slit-shaped mesopores and macropores. It can be seen from the BJH pore size distribution diagram (inset of Fig. 4) that B-CN reveals a small peak centered at 4 nm. As for T-CN, a wide peak at approximately 20 nm can be ascribed to the abundant small mesopores. Additionally, the pore volumes of B-CN and T-CN are 0.032 and 0.214 cm³/g (Table 1), respectively, and a higher pore volume is more beneficial for the absorption of more substrates. The BET specific surface area of T-CN was calculated to be 38.8 m²/g, which is 6.2 times higher than that of B-CN (6.3 m²/g). The enlarged specific surface area can offer more active sites for the reaction, thus further increasing the photocatalytic activity. In addition, the porous structure can effectively reduce the migration distances of the electrons and holes, thereby reducing the recombination probability of the carriers as well as facilitating their migration to the surface of the material for the occurrence of the corresponding reactions. The specific surface area of the CdS-modified g-C₃N₄ nanotube was determined to be 38.2 m²/g, which is close to that of bare T-CN. The slight decrease in the specific surface area could be attributed to the blockage of the porous channel in the g-C₃N₄ nanotube by the CdS nanoparticles.

UV-vis diffuse reflectance spectroscopy was used to analyze the photoabsorption properties of the samples. As shown in Fig. 5a, the characteristic absorption edge of T-CN is located at 460 nm, and T-CN presents a wider absorption edge at 525 nm. It is noteworthy that the photoabsorption abilities of T-CN and T-CN/CdS-x composites are stronger than that of the CdS sample in the visible region. This phenomenon may originate from the multiple reflections of incident light within the hierarchical nanotube structure.

The bandgap energies of T-CN and CdS can be estimated by using Kubelka-Munk theory according to the formula αhv = A (hv-E_g)^n/2, where α represents the absorption coefficient, h is Planck's constant, v is the light frequency, A is a constant of the conductor, n is a constant dependent on the transition type of the semiconductor (for direct transition, the value is 1; for indirect transition, the value is 4), and E_g is the bandgap energy. Thus, the bandgaps estimated from the intercepts of the tangents to the plots of (ahv)^n/2 vs. photon energy (hv) are 2.70 and 2.36 eV (Fig. 5b) for T-CN and CdS, respectively. The band energy shows an obvious blue shift with the increase in T-CN content, compared to that of CdS, due to the enhanced photon absorption of T-CN.

PL emission spectra can be obtained when the photogenerated electron-hole pairs combine together. Hence, the intensity of the PL emission peak can directly reflect the separation rate of photoexcited carriers. As observed in Fig. 5c, the PL spectra of the samples show different luminescence intensities at the excitation wavelength of 320 nm. The pristine B-CN and CdS show stronger PL emission peaks, which suggest that the electrons and holes of B-CN and CdS undergo facile recombination. However, because of the structural effect, the PL intensity of T-CN is much lower than those of B-CN and CdS, which can decrease the diffusion distance of the electrons and holes to the surface of the photocatalyst. More importantly, the T-CN/CdS-10 composite exhibits the lowest fluorescence intensity, which indicates that the recombination of electron-hole pairs can be further inhibited by constructing a T-CN/CdS heterojunction by coupling with CdS nanoparticles.

Fig. 6a shows the photocurrent responses for five on-off cycles under visible light irradiation. When the light is turned on, reproducible and steady photocurrents are recorded when using T-CN, CdS, and T-CN/CdS-10 as the working electrodes. It is noticeable that the T-CN/CdS-10 composite exhibits a much higher photocurrent density compared with the other samples, which suggests that effective separation of the photogenerated electron-hole pairs can be achieved between T-CN and CdS. When the light is turned off, the photocurrent responses of the samples can be negligible. Fig. 6b shows the EIS of T-CN, CdS, and T-CN/CdS composite. Under the same conditions, the radius of the T-CN/CdS-10 heterojunction is smaller than that of pristine T-CN and CdS, indicating that the interfacial resistance between T-CN and CdS is weak. These results are consistent with the photocurrent response results. To deduce the conduction band (CB) position as well as to verify that the synthesized T-CN/CdS samples are suitable for H₂ production, the flat band positions of T-CN and CdS were investigated through Mott-Schottky experiments in the dark. As shown in Fig. 6c and 6d, both T-CN and CdS exhibit positive slopes, which indicate that both are typical n-type semiconductors. Moreover, the flat band potential of CdS (–0.69 eV vs. Ag/AgCl at pH = 6.6) is more positive than that of T-CN (–1.31 eV vs. Ag/AgCl at pH = 6.6), but more negative than that of water reduction (H⁺/H₂: –0.59 eV vs. Ag/AgCl at pH = 6.6), indicating that photogenerated electrons can transfer from T-CN to CdS to further reduce H⁺ into H₂. The valence band (VB) positions of T-CN and CdS can thus be calculated based on the CB positions and bandgap energies, and are 1.39 and 1.67 eV vs. Ag/AgCl at pH = 6.6, respectively.

The photocatalytic degradation efficiency of RhB was measured under visible light irradiation (λ ≥ 420 nm). As presented in Fig. 7a, RhB can barely be degraded in the absence of a photocatalyst, indicating that RhB exhibits stable chemical properties and cannot be self-degraded under visible light irradiation. T-CN exhibits enhanced photocatalytic activity compared with that of B-CN; approximately 59.2% of RhB can be degraded within 60 min of irradiation. CdS exhibits a higher photocatalytic performance; approximately 78.5% of RhB can be degraded over the same duration. It is noteworthy that when T-CN and CdS combined and formed binary composites, the T-CN/CdS-x composites, especially T-CN/CdS-10, showed enhanced photocatalytic degradation abilities, which can be ascribed to the establishment of a heterojunction between T-CN and CdS. Fig. 7b illustrates that all the samples obey apparent first-order reaction rate. Fig. 7c illustrates the variation in the absorbances at the excitation wavelength of 554 nm when using T-CN/CdS-10 as the photocatalyst. The absorbances obviously decrease with increasing irradiation time. No absorbance peak can be observed after irradiation for 60 min, which suggests that RhB dye decomposed completely.

The photocatalytic performance was also evaluated for water splitting to produce H₂ in a TEOA aqueous solution under visible light irradiation. As depicted in Fig. 7d, after 5 h of visible light irradiation, B-CN and T-CN show a relatively low H₂ production, with only 19.7 and 69.2 μmol of H₂ being generated, respectively. This can be ascribed to the poor charge transport ability. On the other hand, T-CN/CdS composites exhibit enhanced apparent water splitting activities, especially T-CN/CdS-10 composite with an average H₂ production of 362.7 μmol. It can also be observed in Fig. 7e that the H₂ production rate over T-CN/CdS-10 is up to 71.6 μmol/h, which is about 16.3 and 4.6 times higher than those of B-CN and T-CN, respectively. The enhanced photocatalytic activity can be primarily attributed to the synergistic effect between T-CN and CdS. Furthermore, as shown in Fig. 7f, the excellent H₂ production rate over T-CN/CdS-10 can be maintained over five cycles of photocatalytic reactions, which indicates its promise in practical applications.

On the basis of the above experimental results and discussion, the catalytic mechanism of T-CN and T-CN/CdS composites toward photocatalytic degradation and H₂ evolution was inferred, as illustrated in Scheme 2. According to previous research [26, 27], a one-dimensional tube-like photocatalyst displays a higher charge carrier mobility along the longitudinal direction, which is beneficial for the separation and transfer of electrons and holes; this, together with the high specific surface area, accounts for the excellent photocatalytic performance of T-CN compared with that of B-CN. For the T-CN/CdS composites, because the CB position of T-CN is more negative than that of CdS and the VB position of CdS is more positive than that of T-CN, the photogenerated electrons are prone to transfer from T-CN to CdS while the holes are prone to transfer from CdS to T-CN; thus, electrons and holes are spatially separated. In the RhB photocatalytic degradation system, holes with strong oxidizing properties can oxidize H₂O to ·OH, which can further oxidize and degrade RhB; meanwhile, the holes can also directly oxidize and degrade RhB. In the photocatalytic H₂ evolution system, the electrons present in the CB of CdS are transferred to the surface of the platinum particles, where they reduce the H⁺ absorbed on the surface of the catalyst and release H₂; meanwhile, holes are consumed by the TEOA added to the catalytic system.

A newly designed porous g-C₃N₄ tube with a reasonable loading capacity of CdS has been proved to be a promising photocatalyst for the degradation of contaminants and for H₂ evolution. The T-CN/CdS composites display obviously enhanced photocatalytic performance for the degradation of RhB and water splitting compared with those single-component photocatalysts. The highest RhB degradation efficiency was realized when using the T-CN/CdS-10 composite as the photocatalyst. The optimum H₂ evolution efficiency is up to 71.6 μmol/h, which is about 16.3 times higher than that of B-CN. This study may broaden views on the design of new morphologies and structures of g-C₃N₄-based semiconductor photocatalysts for better utilization of solar energy.

Acknowledgments

The authors acknowledge the financial support from the National Natural Science Foundation of China (51602297 and U1510109), Major Research Project of Shandong Province (2016ZDJS11A04), Fundamental Research Funds for the Central Universities (201612007), Postdoctoral Innovation Program of Shandong Province (201603043), Australia Research Council (ARC) under the Project DP160104089, and Start-up Foundation for Advanced Talents of Qingdao University of Science and Technology (010022919).