Crystalline C3N4/CeO2 composites as photocatalyst for hydrogen production in visible light

Crystalline carbon nitride (C-C3N4) doped with cerium oxide (CeO2) was synthesized using ionothermal method to increase the photocatalytic activity for H2 production. Graphitic carbon nitride (g-C3N4) obtained from direct pyrolysis of urea at 550°C was subsequently annealed with a mixture of KCl and LiCl to obtain C-C3N4. CeO2 was doped onto C-C3N4 and g-C3N4 via calcination at 550°C. XRD analysis showed the formation of high intensity C3N4 and CeO2 peaks in C-C3N4/CeO2, meanwhile g-C3N4/CeO2 only showed CeO2 peaks. FTIR analysis confirmed all the samples contained C3N4 polymeric structure. The specific surface area of g-C3N4 was measured at 61 m2/g. The surface area increased to 92 m2/g when g-C3N4 transformed into C-C3N4, and further increased to 106 m2/g on C-C3N4/CeO2. The photocatalytic activity for H2 gas production showed significant increase of H2 rate on C-C3N4/CeO2 compared to g-C3N4/CeO2 and g-C3N4. The high crystallinity and high surface area were suggested to enhance photocatalytic activity of C-C3N4/CeO2 in visible light presumably due to the increase of electron and hole lifetimes.


Introduction
The increasing demands for energy have ultimately intensified the usage of fossil fuel, resulting in serious environmental issues cause by production of greenhouse gas (GHG). The depletion of fossil fuel and the growing environmental issues from fossil fuel consumption, have risen a concern to explore other sustainable alternative sources of energy. Hydrogen (H2) have created some traction as a highly efficient clean energy carrier since the combustion of hydrogen produced only water as by-product [1]. H2 has a higher energy content (120-142 MJ kg -1 ) compared to fossil fuels [2]. However, around 90% of global hydrogen is attained by steam reforming of fossil fuel, which yet defeats the purpose of being a sustainable source of energy [3]. Subsequently the production of green H2 by utilizing the abundant solar energy can resolve the environmental issues and the future energy source complications [4]. One approached frequently investigated for the green energy route is photocatalytic water splitting (reaction involved: 2H2O → 2H2 + O2). There are several advantages from this method such as using the abundant solar energy and water to produce H2 with the help of low-cost semiconductor photocatalyst.
The development of a low-cost, sustainable and highly effective photocatalyst in visible light (>420nm) for water-splitting reaction have been investigated for over 50 years [5]. Metal oxide is considered as one of the materials well studied as semiconductor photocatalyst. TiO2 have been intensively investigated due to its ability to produce hydrogen from water, additionally it possesses attractive quality such as being inexpensive, abundantly available and stable [6]. However, similar to other metal oxide which generally occupies a large band gap (~3.2 eV), photogeneration of energy 2 carriers only taking place in ultra-violet range, hence limits the ability to fully utilizing 96% of visible light region [7]. Generation of photocatalyst to fully utilize the abundant photon energy in visible light energy would increase the efficiency and viability of production of green hydrogen energy [8].
A metal-free graphitic carbon nitride photocatalyst consisting of tri-s-triazine repeating unit, have gained considerable recognition due to the capacity to harness photon in visible region for water splitting [9]. Interestingly, g-C3N4 has a high thermal stability (~600˚C), a flexible electronic structure, cost effective, non-toxic, abundantly available and a lower band gap of 2.7 eV which corresponded to the 460 nm visible light region [10], [11]. Comparison to metal base photocatalysts which are only adsorbed photon in the UV region, hence g-C3N4 is considered a promising contender to fully utilised sunlight energy for water splitting [12]- [14]. Despite the positive features of g-C3N4, the performance is hampered by small specific surface area (10-15 m 2 g -1 ) and high photogenerated electron-hole recombination rate [10], [15], [16]. Improving the limiting factors of g-C3N4 have been studied substantially. Structural modification of g-C3N4 would affect the activity of photocatalyst by decreasing the band gap, increasing the specific surface area and eliminating defects on the surface of g-C3N4 [15].
Urea is nitrogen-rich precursor commonly used to synthesis g-C3N4 due to the availability, inexpensive price and ability to produce large surface area. Several methods have been studied for the synthesis of g-C3N4 such as chemical vapour deposition [19], sol-gel method [20], hydrothermal method [21] and microwave method [22]. Thermal condensation or direct pyrolysis of nitrogen-rich precursors is one of the main method due to its simplicity and low-cost production [4], [23]. However, direct pyrolysis of urea only produced ~6% of g-C3N4 [17], [18]. Apart from that, bulk g-C3N4 synthesized from direct pyrolysis of nitrogen-containing precursors results in low crystallinity, as the consequences of incomplete deamination during thermal condensation of intermediates. Incomplete deamination increased the formation of terminal defects such as -CNH2 and -C2NH which acted as a recombination centres [24]. Subsequently instead of producing g-C3N4, they are producing melem monomers which are connected by hydrogen bond to form the tri-s-triazine-based melon structure [25]. These hydrogen bonds lead to low conductivity caused by the blocking electron conductivity of the plane [4].
To improve the thermal condensation and the crystallinity of the g-C3N4, ionic salts (KCl and LiCl) were used in the synthesis process referred as ionothermal method [24]. The molten salts combination was considered as suitable solvent to facilitate thermal condensation at high temperatures [26]. By improving the structure of the crystalline g-C3N4, the photocatalytic activity for H2 generation in the visible light region has increased compared to the bulk g-C3N4 [26]. Rapid photogenerated electronhole recombination is one of the limitations of g-C3N4. In order to gain optimum efficiency of photocatalyst g-C3N4, the electron-hole recombination needs to be deferred as long as possible for complete oxidation/reduction reaction to takes place. The alteration by doping of metal is one method to delay the electron hole recombination [27]. Metal acts as electron-hole trapping sites which effective facilitates the electron-hole separation for maximum photocatalytic activity [28], [29]. Some commonly used noble metal are Ag, Pt, Rh, Au and Pd, although this addition is beneficial to increasing the activity of photocatalyst, the material production cost would significantly increase [27].
Apart from that, the fast recombination of electron-hole can also be delayed by doping with transitional metal and metal oxide base semiconductor. These nanocomposites will create a heterojunction structure which act as a separation for the electron and hole of the photocatalyst [6]. Two common heterojunctions investigated are traditional type-II heterojunctions and Z-scheme heterojunctions. Doping cerium oxide (CeO2) with g-C3N4 was predicted to improve the limitations of the photocatalyst. CeO2 has gained popularity due their abundance, inexpensive, inert, non-toxic, high stability in absorption of photon energy, and excellent transfer of charged electron promoting electronhole separation which would enhance photocatalytic activity [30]- [33]. The g-C3N4/CeO2 composite would form a heterojunction structure supporting the separation for photogenerated electron-hole. In a heterojunction situation the photoexcited electron would move from valence band to the conduction band of g-C3N4, to the conduction band of CeO2, this creates a delay of the electronhole recombination which improve the photocatalytic activity [34]. Currently there is no definitive technique of synthesizing the ideal g-C3N4 base photocatalytic, which resulted in various methodologies and outcomes. There is still a need to determine the right combination of semiconductor material to synthesis the best of g-C3N4 photocatalyst. For our research, we have focused on two main objectives. First, we aimed to improve the crystallinity of gC3N4 by applying ionothermal method using the molten salts (KCL and LiCl). Secondly, we aimed to create a heterojunction structure for g-C3N4 by doping with metal oxide CeO2. We expected the semiconductor composite to have an improved photocatalytic activity for hydrogen production in visible light region.

Materials
Urea fertilizer (CO(NH2)2) was used as carbon and nitrogen-rich precursors. Potassium Chloride (KCl) and Lithium Chloride (LiCl) were used as the ionic solvent for preparation of crystalized carbon nitride. Cerium Oxide (CeO2) was used for the doping of carbon nitride. Triethanolamine (TEAO) was used as a sacrificial agent in the H2 production experiment. All chemicals were used without any purification. Distilled water was used in all experiment.

Preparation of bulk g-C3N4 (g-C3N4)
. In a typical procedure, 25 grams of urea fertilizer (CO(NH2)2) in a covered alumina crucible was heated to 550˚C held for 3 hours with a ramping rate of 1˚C/min in a muffle furnace. The powder product was collected and ground in a mortar. The obtained sample is referred as Graphitic Carbon Nitride (g-C3N4) and produced 6% yield.

Preparation of crystalized C3N4 (C-C3N4
). 1 gram of g-C3N4 were grounded with 3.3 grams of KCl and 2.7 grams LiCl in a glass dish. 25 ml of distilled water was added to the powder mixture and magnetically stirred for 1 hours. The sample was poured into an alumina crucible and heated to 120˚C for 2 hours in oven. Then, the mixture was heated with a covered alumina to 550˚C for 3 hours at ramping rate of 1˚C/min under an air atmosphere in a muffle furnace. The product was washed 3 times with boiling water and using centrifuge at 40 000 rpm for 10 minutes to collect the sediments. The product was dried at 60˚C overnight. The sample obtained was referred as Crystallized Carbon Nitride (C-C3N4)

Preparation of graphitic C3N4 doped with CeO2 (g-C3N4/CeO2
). 1 gram of g-C3N4 and 1 gram of CeO2 were mixed and dissolved in 25 ml H2O. The sample heated at 65 ˚C on the hot plate to evaporate water and stirred for 3 hours. The sample collected was grounded with mortar and transferred to a covered alumina crucible. The mixture was heated with a covered alumina to 550˚C for 3 hours at ramping rate of 1˚C/min under an air atmosphere in a muffle furnace. The sample collected was ground using mortar, this sample was denoted as Graphitic Carbon Nitride Doped Cerium Oxide (g-C3N4/CeO2).

Preparation of crystalized C3N4 doped with CeO2 (C-C3N4/CeO2
). 1 gram of g-C3N4 and 100 mg of CeO2 samples were ground with 3.3 grams of KCl and 2.7 grams LiCl in a glass dish. 25 ml of distilled water was added to the powder mixture and mixture was magnetically stirred for 1 hours. The sample was moved into an alumina crucible and heated to 120˚C for 2 hours in oven. Then, the mixture was heated with a covered alumina to 550˚C for 3 hours at ramping rate of 1˚C/min under an air atmosphere in a muffle furnace. The product was washed 3 times with boiling water and using centrifuge at 40 000 rpm for 10 minutes to collect the sediments. The product was dried overnight at 60˚C overnight. The sample obtained was referred as Crystallized Carbon Nitride Doped Cerium Oxide (C-C3N4/CeO2)

Characterization
The crystal structural properties of the powder photocatalyst was analyzed via X-ray diffraction (XRD) on a Shimadzu XRD-7000 at a scanning rate of 2θ/min in the 2θ range from 10θ to 80θ. The detailed morphology and structure were investigated by high-resolution transmission electron microscopy (HRTEM) using Tecnai G2 F20 STWIN. Fourier transform infrared (FTIR) spectra were obtained using Agilent Cary 630 FTIR Spectrometer from 600 -4000 cm -1 . The diffuse reflection spectra (DRS) of the samples were recorded on a UV-vis spectrophotometer using Agilent Cary 4000/5000/6000i Series UVVis-NIR. The wavelength recorded was in the range of 200 to 800 nm. The Brunauer-Emmett-Teller (BET) specific surface areas and Barret-Joyner-Halenda (BJH) pore structures were obtained via measurements from the volumetric N2 absorption and desorption isotherms at liquid nitrogen temperatures using Micromeritics ASAP2020. The samples were degassed under vacuum at 300˚C for 2 hours at the ramp of 10˚C/min before measurements.

Photocatalytic activity
Photocatalytic activities of the photocatalysts were determined for H2 generation under visible-light. Typically, 20 mg of catalyst powder was dispersed in a 25 ml of aqueous solution containing 10 vol.% triethanolamine (TEOA) in a 64 ml cylindrical quartz vial and covered using a rubber septum. Prior to the irradiation, the reaction suspension was magnetically stirred for 10 minutes to ensure equal dispersion of photocatalyst and purged using N2 for 5 minutes to remove residual air in the reaction tube. The photocatalytic test was implemented under irradiation of Osram Powerstar light with UV filter providing visible light irradiation. At 30 minutes intervals, 0.5 ml of the gas suspension was collected by gas syringe and analyzed via a gas chromatography, Shimadzu GC-2014 with a thermal conductive detector (TCD).

Result and Discussion
3.1. Physiochemical Properties 3.1.1. XRD analysis. The X-ray diffraction analysis in figure 1 shows all the carbon nitride samples and CeO/C3N4 composites. g-C3N4 appears to have two different peaks at 13.4˚and 27.7˚ indicating successful synthesis of carbon nitride by thermal polymerization of urea (JCPDS 87-1526) [35]. The minor diffraction peak at 2θ = 13.4˚ was indexed to (100) plane of tri-s-triazine unit in-planar structural packing motif [36], [37]. The dominant diffraction peak at 2θ = 27.7˚ is the characteristic interlayer staking of aromatic ring system repeating units of that correspond to (002) plane. This corresponds with the d-space of 3.22 Å with an interlayer spacing distance of 0.323 nm [38], [39]. In previous research the dominate peak of g-C3N4 was ascribed as 27.4˚, the slight shift from 27.4˚ to 27.7˚ indicates the sample has a shorter interlayer spacing distance hence the aromatic ring stacks are denser due to higher polymerization [8], [18].  Figure 1. XRD pattern of g-C3N4, C-C3N4, g-C3N4/CeO2 and C-C3N4/CeO2. The asterisk (*) denotes the peak that represents CeO2 peaks.
When g-C3N4 was treated with molten salts, the resulting peaks showed the main peak shift to 26.8˚ and 28.0˚, while the minor peak shifted to 12.2˚ which corresponded to the intercalation of K + ion in the plane. The use of molten salts decreased the overall peak intensity of C-C3N4. However, the appearances of different crystalline planes suggested the formation of other crystalline structure. New peaks were visible at 21.0˚, 24.7˚ and 32.5˚. The main diffraction peak at 2θ = 26.8˚ is considered to be in a nematic-discotic phase, similarly it is the interlayer staking that corresponds to (002) plane [26]. The peak shifted have resulted in the enlargement of the interlayer stacking distance of the C3N4 nanosheet at 0.336 nm [40]. Interestingly another peak ascribed to the (002) plane appears in C-C3N4, the diffraction peak shifted to a higher angle at 2θ = 28.0˚ have led to the decreased of interlayer distance to 0.317 nm, presenting a denser and tighter packing layer of C3N4 nanosheet [41]. The coexisting of both peaks represents a slight undulation of the layer of C3N4 nanosheet hence producing two interlayer stacking distance, which suggest a mixture of both crystalline and amorphous structure in C-C3N4. The minor diffraction peak at 2θ = 12.2˚ corresponding to the (100) plane, represent the interplanar structural packing motif. The shift to a lower angle may indicate smaller tri-s-triazine units hence increasing the interplanar distance to 0.73 nm.
When CeO2 was added into g-C3N4, the peaks corresponded to g-C3N4 observed at 13.4˚ and 27.4˚was not visible due to low intensity compared to CeO2 peaks in g-C3N4/CeO2. The peak associated with CeO2 with the respective crystal planes were observed at 28.4˚ (111) [34]. The XRD of g-C3N4/CeO2 indicate the present of higher number of amorphous g-C3N4 structure in catalyst when obtained via direct calcination. Without the addition of molten salts, there is an increased possibility of thermal decomposition of g-C3N4. Decomposition of g-C3N4 led to formation of defects which act as a recombination site and reduced specific surface area hence detrimental towards photocatalytic activity of g-C3N4/CeO2 [8]. When C-C3N4/CeO composite was prepared using molten salt method, the peaks corresponded to C-C3N4 were still visible at 12.2˚, 21.0˚, 24.4 ˚, 26.8 ˚ and 32.24˚. Significantly low intensity of peak corresponding to CeO2 was also observed at 28.7˚, 47.2 ˚ and 56.2˚ indexed to (111), (220) and (311), respectively.

Transmission electron microscope (TEM) analysis.
The synthesis of C-C3N4/CeO2 using ionothermal method by utilizing molten salts (KCL and LiCl) was expected to enhance the polymerization process hence increasing its crystallinity. From the XRD result, multiple crystalline peaks were revealed in C-C3N4/CeO2, therefore further investigation in order to understand the morphological structure was performed using transmission electron microscope. At the lower magnification on figure 2 (a)-(d) it is observed that C-C3N4/CeO2 consisted of large nanorods of C-C3N4 and agglomerated CeO2 on C-C3N4. The CeO2 is not evenly distributed on the C-C3N4 surface and show sign of agglomeration. The metal oxide was not dispersed homogeneously on the surface of C-C3N4, suggesting although the heterojunction structure is established, the efficiency may be slightly reduced. This may affect the photocatalytic activity of C-C3N4/CeO2.
On further investigation at a higher magnification in figure 2 (e), two distinct structures attributed to C3N4 were perceived. The 2D layers of conjugated sheet of tri-s-triazine specifies the interlayer stacking distance of 0.329 nm -0.333 nm, and corresponds to the (002) plane of 2θ = 26.6˚ denoted as C3N4 structure [42]. Apart from that, the amorphous structure of C3N4 was also observed which displayed as a corrugated structure [30]. It was suggested that the formation of amorphous structure responsible to the reduction of XRD peak intensity for (100) and (002) planes of C3N4. Structures identified as CeO2 were observed in figure 2 (e-f). CeO2 displayed a varied interlayer spacing which corresponds with their XRD peaks. The interlayer spacing of 0.275 nm was attributed to the (200) plane of 2θ = 31.3˚, while interlayer spacing of 0.324 nm were assigned to the (111) plane of 2θ = 28.6˚ and interlayer spacing of 0.199 nm were assigned to the (220) plane of 2θ = 47.5˚ which representing CeO2. The amorphous structure of C3N4 and the agglomeration of CeO2, displayed a disorganized structure of C-C3N4/CeO2, which led to the lower intensity of crystalline peak at (002) plane in XRD analysis.
There is an abundance of porous structure of C-C3N4 due to the higher polymerization thus increase the surface area and the available active sites for photoreaction to occur [43] [44]. The layered structure provide pore channels that promotes the mobility for photogenerated charges and provide abundant photoreaction sites for the production of electron hole pairs [26]. The exfoliation of layered C-C3N4 sheet is considered beneficial as it increase the specific surface area and reduces the transport distant for charged electrons, this can be achieved by enhancing the polymerization [18]. The HRTEM analysis showed a varied structure for C-C3N4 and CeO2, which was supported by the XRD analysis. The addition of molten salts in the calcination process have given rise to new crystalline peaks, which can be interpreted as the present of multiple different structure C-C3N4 and CeO2.

BET isotherm plot.
The nitrogen adsorption-desorption isotherms and Barret-Joyner-Halenda (BJH) pore-size distributions curve of all catalyst are shown in figure 3. According to the adsorptiondesorption isotherms BDDT classification, all samples exhibit type III isotherm. The samples exhibit type H3 hysteresis loop according to IUPAC classification which is associated with slit-shaped pores present in the aggregated of sheet-like particles, complying to the morphology of C3N4 [38]. Different precursors such as melamine, dicyanamide and urea would produce different morphology features of C3N4, it was concluded that urea produce the larger specific surface area due to the light and fluffy texture of C3N4 [43]. Based on previous results calcination temperature of 550°C is considered ideal, providing a enlarge specific surface area by exfoliation without decomposing the samples [8]. Hence increasing the photocatalytic activity due to higher number of available active sites for photoreaction to occur. The specific surface area of g-C3N4 was measured at 61.3 m 2 /g. When CeO2 was added via direct calcination, the surface area of g-C3N4/CeO2 was significantly reduced to 35.9 m 2 /g. It is interesting to see that the surface area of C-C3N4 and C-C3N4/CeO2 were enhanced to 92.1 m 2 /g and 106.4 m 2 /g when molten salt method was applied in the calcination process. The increase in surface area indicates that the C-C3N4 produced when using molten salt method increased the stability of carbon nitride thus contributed to the increase in specific surface area. However, for g-C3N4/CeO2 obtained from direct calcination without the ionic liquid solvent, we suggest that the C3N4 has decomposed which may resulted in a larger ratio of CeO2 in the catalyst, hence resulted in a smaller specific surface area of 35.9 m 2 /g. The BJH pore-size distribution curve, shows all samples are composed of small mesopores (~2.6 nm) which are associated to the porosity formed in between the nanoparticles [38]. A larger area of hysteresis loop and a shift to a lower relative pressure of C-C3N4/CeO2 was observed from the adsorption curve in figure 3(b), suggesting the formation of bigger mesopores in the catalyst [18]. Apart from that, the C-C3N4/CeO2 also formed larger mesopores (~15 nm -23 nm) which are attributed to the pores formed between packed C3N4 layers. Summary from table 1 shows CC3N4/CeO2 have the largest pore volume of 0.4462 cm³/g and pore-size average of 16.8 nm, indicating a higher porosity of this sample.
It is reasonable to conclude that the molten salt method has enhanced the C-C3N4/CeO2 photocatalytic activity by increasing the porosity and specific surface area of the composite. A larger specific surface area is considered favourable to photocatalytic activity due to the higher number of active sites; this acts as site for the separation of photogenerated electron and holes, hence improving the photocatalytic reaction rate.
The absorption bands appeared at 1200-1650 cm -1 assigned to the typical characteristic of aromatic C-N heterocyclic [37], [45], [46], [49], [50]. The peaks at 1203-1357 cm -1 are attributed to the aromatic C-N stretching vibration. It is assigned to the stretching vibration of connected units of CN [22]. The peaks at 1400 -1632 cm -1 are attributed to the stretching vibration of heptazine-derived repeating unit [46] which originated from the C=N stretching [22]. With the addition of CeO2, the peaks gained increase intensity compared to bulk g-C3N4. The small band observed at 2128 cm -1 was due to the dissociated CN network, forming both C≡N and N=C=N bonds, formed by broken aromatic unit of C3N4 [49].
The broad adsorption band at 3000-3250 cm -1 is attributed to the stretching vibration of N-H (primary and secondary amine), which indicates that uncondensed amino functional group is present in the sample [22], [37], [49], [52]. A highly intense peak attributes to higher number of -NH/-NH2 groups were observed on g-C3N4 and CeO/g-C3N4, which indicates a larger number of particles are presence as smaller polymer contained fewer heptazine units [53]. The increase of amine terminals in these smaller segment are concluded as defects, which acts as the recombination sites in the photocatalytic reaction [53]. The N-H stretching bands were disappeared on C-C3N4 and CeO/C-C3N4, which implied crystallisation in the presence of ionic salt removed defects. The broad peak 3250-3313 cm 1 is attributed to the vibration of intermolecular O-H hydrogen bonds or hydroxyl groups from the absorption of water in CN [22], [49]- [51].
As seen from the figure 2, g-C3N4 and g-C3N4/CeO2 undergoes direct calcination without molten salts resulted in a higher number of defects. Defects such heptazine deformation, dissociated CN network and higher number of -NH/-NH2 groups, which ultimately would decrease the photocatalytic performance of the two samples. Generally the peaks observed by the catalysts were similar to other reported literatures on the CN synthesis from urea [17], [50], [54]. All these characteristic FTIR peaks suggest that the overall structure of g-C3N4 maintained the original form even after CeO2 doping. As stated previously in XRD analysis, g-C3N4/CeO2 have a lack of carbon nitride peaks, however from the FTIR result, it is certain that carbon nitride is still present in the photocatalyst.

UV-vis (optical absorption studies).
Photocatalytic reaction depends on the ability of catalyst to harvest photon from the light. The photon energy absorbed by the photocatalysts would then excite the electrons on the valance band to the conduction band, generating holes and ultimately create a chain 10 redox reaction. Factors influencing the photocatalytic activity generally depends on catalyst ability to harvest light energy and reduced the recombination rate of the electron and hole in the valance band. Structural modification of photocatalyst carbon nitride is one of the methods used to improve the optical absorption. UV-visible diffused reflectance spectra reveal the optical property and the band energy of the samples synthesized by direct calcination and ionothermal calcination. The light absorbance graph was constructed using the Kubelka Munk calculation. From figure 5, all of the samples have an absorption edging towards the visible light region with absorbance between 395 nm to 418 nm. Following direct calcination of g-C3N4 and g-C3N4-CeO2, the absorbance was red-shifted slightly to 416 nm and 418 nm with band gap energies were determined at 2.94 eV and 2.97 eV. When the samples produced with molten salt addition during the calcination, C-C3N4 and C-C3N4-CeO2 exhibited a lower absorbance wavelength at 395 nm with band gap energy of 3.12 eV.
Theoretically the band gap energy from all the samples are adequate for endothermic reaction for water splitting [55]. However as seen in figure 5, the DRS curve contained several loops due to the defects arising from the polymerization of the samples [40]. This is supported by the FTIR of direct calcined samples containing more defects compared to the crystalline samples. The loop is attributed to multiple reflection of incident light within photocatalyst, subsequently varied the absorbance in the photocatalyst. This would result in inefficient transfer of electron which led to high recombination rate of photocatalyst resulting in low photocatalytic activities.

Photocatalytic activity
Triethanolamine (10 vol%) was used as the sacrificial agent in water splitting reaction. The hydrogen evolution reaction (HER) was performed under visible region with a UV filter to provide only photon energy in visible light region. Figure 6 shows the HER values of the samples g-C3N4 and C-C3N4/CeO2 in 180 minutes. The C-C3N4 and g-C3N4/CeO2 are not displayed in the graph due to only a trace of H2 produced from both samples. The hydrogen generated seen in figure 6 for g-C3N4 after 180 minutes is 170.84 µmolg -1 , whereas CC3N4/CeO2 generates a higher HER value of 268.47 µmolg -1 . Interestingly the outcome from the optical analysis predicted that g-C3N4 would performed better due to a smaller band gap energy of 2.94 eV compared to 3.12 eV for C-C3N4/CeO2. However, the activity of C-C3N4/CeO2 is higher compared to g-C3N4. There are several influences that play a role in this. g-C3N4 was presumed to have higher number of defects as revealed in FTIR analysis compared to C-C3N4/CeO2. An excess number of defects may contribute to the lower HER of g-C3N4, as defects often acts as a recombination site for electron. The doping of metal oxide CeO2 may contribute to the higher HER value by creating a heterojunction structure, delaying the recombination of electro-hole in CC3N4/CeO2, hence prolonging the redox reaction in the valance and conduction band. Lastly, the specific surface area of C-C3N4/CeO2 was much larger at 105.4 m 2 /g with higher porosity density hence enhancing the number of available active sites for the photogenerated reaction to occur.
Nevertheless, the photocatalytic activity for both catalyst is still considered low. It was suspected that several factors may contribute to this predicament. As seen in the UV-vis DRS, there are several loops where absorbance varies in all the photocatalyst. These loops caused by defects during polymerization of catalyst attributed to inefficient transfer of photogenerated charge carriers which eventually led to fast recombination of electron and hence factoring the low photocatalytic activities. The absorbance in UV-vis DRS was measured at 395 nm -418 nm, which is considered not favourable under visible light with a wavelength of 420 nm above. This may contribute to the low HER value of all samples as the photon energy available in the visible region is not efficiently harvested for the excitation of electron during hydrogen production.

Conclusion
In summary, a highly active C-C3N4/CeO2 composite was synthesized by heating g-C3N4 and CeO2 with molten salt mixtures at 550˚C for 3 hours. The C-C3N4/CeO2 resulted in high crystalline structure and large specific surface area when compared with the g-C3N4. The photocatalytic activity of C-C3N4/CeO2 was significantly higher than g-C3N4, g-C3N4/CeO2 and C-C3N4. The high H2 production was due to a higher number of active sites as a result of a larger specific surface area.