Highly efficient hydrogen production and selective CO₂ reduction by the C₃N₅ photocatalyst using only visible light

Abstract

The production of energy sources by metal-free photocatalysts based on graphitic carbon nitride (g-C₃N₄) has garnered substantial attention. In this study, nitrogen-rich carbon nitride (C₃N₅) was successfully synthesized through the thermal polycondensation of 3-amino-1,2,4-triazole. The structural and physical characterization has suggested that a portion of the triazine rings, which constitute the structural framework of g-C₃N₄, may be substituted with five-membered rings in C₃N₅. Furthermore, the polymerization of C₃N₅ proceeded more extensively than that of g-C₃N₄ from melamine precursors. The increased nitrogen content in C₃N₅ resulted in a heightened number of π-electrons and a narrowed energy bandgap, with the potential of the valence band maximum being negatively shifted. Additionally, photocatalytic assessments encompassing nitro blue tetrazolium reduction, H₂ production from triethanolamine aqueous solution, and CO₂ reduction in the liquid phase were performed. All findings demonstrated that C₃N₅ exhibits significantly superior photocatalytic properties compared to g-C₃N₄. It is particularly noteworthy that C₃N₅ selectively generates methanol and H₂ from oversaturated CO₂ solutions under visible light irradiation, while g-C₃N₄ selectively generates formaldehyde. These outcomes strongly indicate that C₃N₅ serves as a metal-free, visible-light-responsive photocatalyst, capable of contributing to both the production of renewable energy sources and the reduction of greenhouse effect gases.

---

Introduction

Since the first report of photocatalytic hydrogen generation using graphitic carbon nitride (g-C₃N₄) in 2009, g-C₃N₄ has attracted attention as an inexpensively, semi-permanently and easily synthesized metal-free photocatalytic material.^1–3 Prior research endeavors have reported more efficient H₂ production and CO₂ reduction by, for example converting g-C₃N₄ from a two-dimensional material to a three-dimensional one and loading noble metal co-catalysts on it.^4,5 At present, the reaction efficiency of g-C₃N₄ is still notably inferior to that of metal compound photocatalysts. As long as continuing g-C₃N₄ studies just follow conventional approaches and methodologies that have been applied for metal compound photocatalysts with overwhelmingly high performance, it seems quite difficult to make carbon nitride materials exceed metal compound photocatalysts. However, if carbon nitride can exhibit photocatalytic properties comparable to those of metal compound photocatalysts, it would be an economically-viable choice and highly valuable for use from the viewpoints of sustainability and resource saving. Therefore, a fundamental structural reformulation of g-C₃N₄ is first required to create a metal-free photocatalyst that has capabilities like metal photocatalysts.

In the latest study, a new type of nitrogen-rich carbon nitride (C₃N₅) has been considered as an emerging photocatalytic material, because of its attractive features such as the smaller BG (ca. 2.1 eV) than that of g-C₃N₄, easier adsorption of organic compounds, faster charge transfer to metal cocatalysts, and robustness against secondary contamination.^6–9 Furthermore, C₃N₅ was combined with other metal photocatalytic materials, which resulted in efficient H₂ production, CO₂ reduction, and organic decomposition.^10–12

On the other hand, the reported C₃N₅ studies used metal-containing materials such as KBr during the synthesis process, and its photocatalytic properties are also evaluated in combination with other metal photocatalysts. This cancels out the attractive features of carbon nitrides that are easy-to-synthesize and metal-free. To effectively utilize the functions of carbon nitride, it is important to evaluate and compare the inherent natures of existing materials (g-C₃N₄) and new ones (C₃N₅) and find important directions for material development, rather than focusing on immediate property improvements. To date, we have not been able to find any papers that experimentally prove that C₃N₅ alone has better photocatalytic properties than g-C₃N₄ alone.

In this study, first, C₃N₅ was synthesized by only thermal polymerization from a precursor, without the use of metal-containing materials. Next, the detailed crystal structure and energy bands were evaluated. Finally, the photocatalytic properties of C₃N₅ were evaluated by H₂ production and CO₂ reduction, which were theoretically indicated as possible photocatalytic reactions in visible light over C₃N₅ owing to its conduction band (CB) and valence band (VB) edge positions. Similar experiments were also performed for g-C₃N₄ and the results for both C₃N₅ and g-C₃N₄ were compared and discussed. These experimental studies show that C₃N₅ is a better material than g-C₃N₄ for solving environmental and energy issues such as H₂ production and CO₂ reduction, and provide new guidelines for photocatalyst material design based on carbon nitride.

Experimental methods

Materials and reagents

In this work, the following chemicals and reagents were purchased and used without any further purification: 3-amino-1,2,4-triazole (AT, C₂H₄N₄, Tokyo Chemical Industry), melamine monomer (C₃H₆N₆, Tokyo Chemical Industry), triethanolamine (TEOA, N(CH₂CH₂OH)₃, Nacalai Tesque; >98%), nitro blue tetrazolium (NBT) chloride (C₄₀H₃₀Cl₂N₁₀O₆, Tokyo Chemical Industry), terephthalic acid (C₆H₄(COOH)₂, Nacalai Tesque), and sodium hydroxide solution (NaOH, Nacalai Tesque; 5 M).

Synthesis of g-C₃N₄ and C₃N₅

C₃N₅ was synthesized by the following method. First, AT (3 g) was dissolved in 30 ml of pure water and stirred for 30 min. After that, the pure water was evaporated using an oil bath at 100 °C for 3 h and the remaining powder was dried under vacuum at 70 °C overnight. Finally, it was heated at a rate of 5 °C min⁻¹, kept at 550 °C for 2.5 h, and cooled naturally to room temperature. g-C₃N₄ was synthesized by thermal polycondensation of melamine (3 g) at a heating rate of 5 °C min⁻¹ and at a holding temperature of 550 °C for 2.5 h.

Characterization

The morphology of the samples was investigated with a transmission electron microscope (TEM, Tecnai G2,FEI). Specific surface area measurements were performed using a commercially available adsorption analyzer (ASAP2020, Micromeritics). The chemical state information was obtained by an X-ray photoelectron spectroscopy instrument (XPS, JPS-9010TR, JEOL) with Al-Kα radiation and a Fourier transform infrared spectroscopy (FTIR) system (ALPHA, Bruker). The crystal structure was specified by an X-ray diffractometer (XRD, D8 ADVANCE, Bruker) with Cu-Kα radiation. To examine the detailed energy band structure, diffuse-reflectance ultraviolet-visible (UV-Vis) absorption spectra were monitored by a UV-visible-near-infrared (UV-Vis-NIR) spectrophotometer (UV-3600Plus, Shimadzu) with an integration sphere. In addition, the determination of the flat band (FB) potential was performed by Mott–Schottky analysis with an electrochemical impedance analyzer (VersaSTAT3, AMETEK). For electrochemical impedance (EI) measurements, a 0.5 M Na₂SO₄ aqueous solution was employed as a liquid electrolyte, and an Ag/AgCl and a platinum wire were used as the reference and counter electrode, respectively.

Photocatalytic hydrogen production using TEOA solution and photoreducing power assessment with NBT solution

45 ml of pure water, 5 ml of TEOA, and 300 mg of carbon nitride were put into a 50 ml beaker and stirred for 30 min. After the beaker was placed inside a home-made measurement cell with gas circulation, the gas lines were purged with argon and gas was circulated in a closed system. Subsequently, visible light (λ > 385 nm) from a xenon lamp (MAX-303, Asahi Spectra) was irradiated onto the photocatalyst samples through the quartz window of the measurement cell, and the generated gas was analysed every 1 hour after the onset of the light irradiation, by a gas chromatograph (GC-8A, Shimadzu) equipped with a thermal conductivity detector (TCD) and packed columns (ShinCarbonST, Shinwa Kako).

30 ml of 5.0 × 10⁻⁵ M NBT aqueous solution and 10 mg of the synthesized carbon nitrides were added into a 50 ml beaker and stirred in the dark for 1 h. After visible light (λ > 385 nm) from a xenon lamp (MAX-303, Asahi Spectra) was irradiated for 5 min, 3 ml of the solution was centrifuged to separate the carbon nitride powder from the NBT solution. Then, the absorbance of NBT at 260 nm was measured with a UV-Vis-NIR spectrophotometer. NBT is oxidized to formazan by superoxide anions (˙O₂⁻) formed by photocatalytic reduction of dissolved oxygen in solution.¹³ NBT has a maximum absorption wavelength at around 260 nm, while the formazan reveals a maximum absorption at around 530 nm.¹⁴ Since formazan is prone to be adsorbed on the catalyst surface and cannot be easily removed from the catalyst sample, the change in the absorbance of NBT was employed as an indicator of the photoreducing power of the catalyst in this study.¹⁵

CO₂ photoreduction in the liquid phase

50 ml of pure water and 300 mg of carbon nitrides were added into a 50 ml beaker and bubbled with pure CO₂ (99.9%) for 8 min. After that, a similar procedure as mentioned in the photocatalytic H₂ production measurement was employed to detect the products generated during CO₂ photoreduction. The solution in the beaker after the experiments was analysed by another gas chromatograph (GC2014, Shimadzu) with a TCD and packed columns (Sunpak-H, Shinwa Kako) using helium carrier gas. Gas analysis was also performed by the same gas chromatograph using packed columns (ShinCarbonST, Shinwa Kako) with argon carrier gas. The OH radicals formed on the photocatalyst surface were measured using the photoluminescence (PL) method. For this experiment, 10 mg of carbon nitrides, 0.04 g of terephthalic acid as a probe molecule, and 0.3 ml of NaOH, which is a reagent dissolving the terephthalic acid and preventing quenching of the fluorescent material, were added to 30 ml of an oversaturated CO₂ solution. Subsequently, visible light (λ > 385 nm) was irradiated for 10 minutes, and only 3 ml of the solution was collected after centrifugation. When the OH radicals generated by the photocatalytic reaction interact with terephthalic acid, a highly fluorescent compound, 2-hydroxyterephthalic acid, is formed. The PL emission spectrum of this solution containing 2-hydroxyterephthalic acid was analysed using a fluorescence spectrophotometer (RF-6000, Shimadzu), with excitation and emission wavelengths set at 315 nm and 425 nm, respectively.

Results and discussion

Morphology and material characterization

The TEM observations presented in Fig. 1 show that the synthesized C₃N₅ and g-C₃N₄ have plate-like layer structures (Fig. 1a and b) and that C₃N₅ has a much larger plate size (ca. 5 μm) than g-C₃N₄ (ca. 1 μm) (Fig. 1c and d). This means that the polymerization in C₃N₅ proceeded more prominently than in g-C₃N₄. Brunauer–Emmett–Teller (BET) specific surface areas of the C₃N₅ and g-C₃N₄ samples were measured to be 2.6 and 5.4 m² g⁻¹, respectively, which are well-correlated with the plate sizes estimated from Fig. 1.

Fig. 1 TEM images of (a) C₃N₅ and (b) g-C₃N₄ particles. Low magnification TEM images of the same (c) C₃N₅ and (d) g-C₃N₄ particles are also presented.

In the FTIR transmission spectra of the synthesized C₃N₅ and g-C₃N₄ (Fig. 2a), some characteristic absorption bands for carbon nitride species appeared. An absorption peak at 800–900 cm⁻¹ originates from triazine rings, broad bands in the range from 1100 to 1700 cm⁻¹ derive from C–N and C=N bonds, and the bands at 2900–3400 cm⁻¹ are due to the end groups such as C–N–H, N–H, and O–H.^16,17 Although the broad absorption bands from 1100 to 1700 cm⁻¹ did not show any significant difference between these two samples, the peak at 800–900 cm⁻¹ in C₃N₅ was smaller than that in g-C₃N₄, presumably because a portion of the triazine rings was replaced with five-membered rings derived from the AT precursor in C₃N₅. Several literatures have reported that some of the precursor backbones exist in the synthesized samples, and the C₃N₅ precursor (AT) molecule owns a five-membered ring.^18–20 Also, the absorbance of the bands from 2900 to 3400 cm⁻¹ in C₃N₅ seems somewhat smaller than that in g-C₃N₄. As we discussed with the TEM images (Fig. 1) and the BET specific surface areas, the C₃N₅ plate size was larger than that of g-C₃N₄, resulting in the smaller number of end groups in the C₃N₅ sheets.

Fig. 2 (a) FTIR spectra and XPS spectra of (b) N1s and (c) C1s signals measured for C₃N₅ and g-C₃N₄ samples. (Peaks at 284.6 eV were used for charge correction.) (d) Structural diagrams of polymerization for C₃N₅ and g-C₃N₄. (e) XRD patterns for C₃N₅ and g-C₃N₄.

The compositional ratio of carbon and nitrogen atoms (C/N) in the synthesized C₃N₅ and g-C₃N₄ was calculated from the XPS spectra (Fig. 2b and c), where the N1s and C1s signals were deconvoluted by using a Gaussian–Lorentzian function. The C/N ratio was given as the ratio between the peak area of the C1s signal divided by the sensitivity of 4.079 and that of the N1s signal divided by the sensitivity of 7.041. As a result, the calculated C/N ratio was 3 : 4.93 for C₃N₅ and 3 : 3.96 for g-C₃N₄, respectively, supporting that the chemical composition is almost stoichiometric for both C₃N₅ and g-C₃N₄.

In the narrow-scan N1s spectra (Fig. 2b), the C–N–C peaks are assigned to be triazine frameworks, the N–(C)₃ peaks are a heptazine ring nucleus and the bridges between the heptazine rings, and the N–H and N–OH are end groups.^21–23 The ratios among the respective signals (C–N–C : N–(C)₃ : N–H and N–OH) were 14 : 67 : 19 for C₃N₅ and 28 : 50 : 22 for g-C₃N₄, respectively. The larger ratio of N–(C)₃ and the smaller ratio of N–H and N–OH for C₃N₅ indicate the grain size enlargement, which was in line with the TEM observation results (Fig. 1) and measured specific surface areas. Additionally, the relative intensity of C–N–C in C₃N₅ was smaller than that in g-C₃N₄, which probably reflects the partial replacement of six-membered rings in the triazine frameworks with five-membered rings, as we discussed in the FTIR spectra (Fig. 2a). Furthermore, a small π-electron peak was observed at around 404 eV in only C₃N₅, where n–π* transition can be promoted by the increase in the number of unshared electron pairs with an increasing nitrogen content.²⁴ Besides, the main peak position of N1s binding energy in C₃N₅ (398.2 eV) is shifted negatively from that of g-C₃N₄ (398.7 eV). Since C₃N₅ has more π electrons, the electron density in C₃N₅ becomes larger than that in g-C₃N₄. As a result, the binding energy is considered to become shifted negatively.^25,26 As presented in Fig. S1 (ESI†), the peak area of the O1s signal in C₃N₅ (4.55 × 10⁴) is smaller than that of g-C₃N₄ (5.70 × 10⁴). Since the oxygen peaks are derived from adsorbed oxygen species and end OH groups,^27,28 this decrease in the O1s peak intensity of C₃N₅ correlates with the larger C₃N₅ particle size.

Based on these FTIR and XPS data, the structures of C₃N₅ and g-C₃N₄ are schematically drawn in Fig. 2d. XRD profiles in Fig. 2e reveal that both samples displayed 100 and 002 reflections. The 100 peak indicates the in-plane ordering of the two-dimensional direction, and the 002 peak indicates interphase stacking of the carbon nitride sheets.^29,30 The 002 peak position of C₃N₅ (27.68°) was shifted to a higher diffraction angle than that of g-C₃N₄ (27.43°). Considering that C₃N₅ has more π-electrons, the C₃N₅ layers are attracted to each other by π–π interactions more strongly than the g-C₃N₄ layers. Therefore, the 002 peak of C₃N₅ was shifted to a higher diffraction angle.

Energy band structure

Tauc plot analysis was carried out for estimating the band gap (BG) energy of the samples. Fig. 3a presents the Tauc plots obtained from the measured diffuse-reflectance UV-Vis absorption spectra under the assumption that C₃N₅ and g-C₃N₄ have indirect bandgaps.^31,32 In Fig. 3a, α is the light absorbance, h the Planck constant, and v the light frequency, respectively. The estimated BG values of C₃N₅ and g-C₃N₄ are 2.15 and 2.75 eV.

Fig. 3 (a) Tauc plots obtained from UV-Vis absorption spectra for C₃N₅ and g-C₃N₄ (the obtained BG values are given in the figure). Mott–Schottky plots measured for (b) C₃N₅ and (c) g-C₃N₄ (the obtained FB potentials were provided in the figures). XPS valence band spectra of (d) C₃N₅ and (e) g-C₃N₄. (f) Schematic drawing of energy band diagrams of C₃N₅ and g-C₃N₄.

Mott–Schottky (M–S) analysis was performed to further investigate the energy band structure. The carbon nitrides were deposited onto fluorine-doped tin oxide (FTO) substrates in the same way as for the powder synthesis, except that an FTO substrate was placed on top of the powder sample during the high-temperature treatment. The applied potential (V) and measured space charge capacitance (C) at various modulation frequencies produced M–S plots (1/C²–V curves), as shown in Fig. 3b and c. The positive slopes in the obtained M–S plots indicate that the synthesized C₃N₅ and g-C₃N₄ are n-type semiconductors.^33,34 If we assume that the conduction band (CB) edge is about 0.2 V more negative than the FB level of n-type semiconductors, the CB edge positions of C₃N₅ and g-C₃N₄ films were calculated to be −0.77 V (vs. normal hydrogen electrode (NHE)) and −0.78 V (vs. NHE), respectively.^35,36 The valence band (VB) edge positions of C₃N₅ and g-C₃N₄ were also determined to be 1.38 V (vs. NHE) and 1.97 V (vs. NHE), respectively, by taking their BG energy values into account. XPS valence band spectra (Fig. 3d and e) revealed that the VB edge of C₃N₅ was more negative than that of g-C₃N₄, which was consistent with the above-described results. The energy band diagrams of the synthesized C₃N₅ and g-C₃N₄ are schematically depicted in Fig. 3f. The BG of carbon nitride materials is determined by the N2p orbital for VB and the C2p orbital for CB.^37,38 The narrower BG of C₃N₅ is attributed to the shallower VB level because C₃N₅ is richer in nitrogen than g-C₃N₄ and has more nitrogen-derived π electrons.

Photocatalytic hydrogen production over C₃N₅ and g-C₃N₄

Fig. 4 shows the photocatalytic H₂ production from TEOA solution over C₃N₅ and g-C₃N₄, respectively. The H₂ production rate of C₃N₅ (ca. 1.2 μmol h⁻¹) was almost twice as high as that of g-C₃N₄ (ca. 0.6 μmol h⁻¹). Photocatalysis always involves a similar number of holes used for oxidation reactions and electrons used for reduction reactions. Both C₃N₅ and g-C₃N₄ meet the oxidation potential (1.23 V (vs. NHE)) and reduction potential (0 V (vs. NHE)) for hydrogen production from water.^39,40 But when water contains sacrificial reagents, they are preferentially oxidized and this photooxidation reaction proceeds fast.⁴¹ In this case, the rate-determining process must be a two-electron reduction of protons produced by the oxidation reaction. Therefore, the reason why C₃N₅ could generate hydrogen more efficiently than g-C₃N₄ can be ascribed to the improved reducing power of C₃N₅.

Fig. 4 Amount of H₂ production from TEOA aqueous solution over C₃N₅ and g-C₃N₄ under VIS irradiation.

NBT measurements (Fig. 5) also suggested that C₃N₅ has better reducing power than g-C₃N₄. The reducing power of C₃N₅ is enhanced by the large reduction in BG by the negative shift of the VB edge and by the increase in the number of excited electrons by visible light irradiation, despite the similar structures and almost identical CB edge positions between C₃N₅ and g-C₃N₄.

Fig. 5 Efficiency of NBT reduction caused by O₂ radicals produced over C₃N₅ and g-C₃N₄.

Liquid phase CO₂ photoreduction over C₃N₅ and g-C₃N₄

The experimental results of CO₂ photoreduction in water under oversaturated conditions are provided in Fig. 6a and b. C₃N₅ selectively produced methanol (CH₃OH) and H₂, while g-C₃N₄ produced only formaldehyde (CH₂O). No other gases and liquids, except CO₂, were observed (Fig. 6c). In Fig. 6c, the pH of CO₂ aqueous solution was 3.9 immediately after preparing the oversaturation condition and was gradually raised up to 5.6 after 2 h in the dark. In general, the pH of pure water is 6 to 6.5, and when CO₂ is dissolved, it tends to be more acidic depending on its concentration.^42,43 In other words, the detection of CO₂ gas and the change in pH verified that CO₂ was present in excess in the water.

Fig. 6 Amounts of (a) CH₃OH or CH₂O and (b) H₂ production from oversaturated CO₂ aqueous solution over C₃N₅ and g-C₃N₄ under VIS irradiation. (c) Temporal changes in CO₂ amount detected by gas chromatograph and pH of the CO₂ aqueous solution with no catalyst in the dark.

As shown in Fig. 6, the time variation of the gas production (CH₃OH, CH₂O, H₂) amounts, pH of the solution, and CO₂ amount become almost unchanged after 2 h passed from the start of the measurements. The reaction kinetics depends on temperature, activation energy, and concentration of gases.⁴⁴ In the present experiment, the temperature was constant during 3 h after the start of the measurements. On the contrary, the concentration of CO₂ dissolved in the solution decreased significantly over time. Hence, the chemical reaction rate of CO₂ reduction was relatively high at the initial stage of the photocatalytic experiment due to the presence of CO₂ in an oversaturated state, while the reaction rate became lowered with the rapid decrease in the CO₂ concentration in the solution.

In the case of simple CO₂ reduction, CO is usually detected as an intermediate.^45,46 However, CO was not detected in our measurements (Fig. 6). This indicates that CO₂ directly produces CH₃OH and CH₂O by multi-electron reduction.

CO₂ + 6H⁺ + 6e⁻ → CH₃OH + H₂O (1)

CO₂ + 4H⁺ + 4e⁻ → CH₂O + H₂O (2)

Six electrons are required for the generation of CH₃OH, and four electrons are required for the generation of CH₂O. Nevertheless, the quantity of CH₃OH produced by C₃N₅ (0.17 μmol) is greater than that of CH₂O produced by g-C₃N₄ (0.12 μmol) (Fig. 6a). Some possible reasons why these reactions could occur are provided as follows.

It has been reported that when the CO₂ concentration near the catalyst is high, the final substance is formed directly from CO₂.⁴⁷ Additionally, both g-C₃N₄ and C₃N₅ have narrow BGs, and they have a larger number of excited electrons. These situations may have induced multi-electron reduction and allowed direct reduction to the final material.⁴⁸ C₃N₅ may have produced CH₃OH by six-electron reduction because of its narrower BG than g-C₃N₄.⁴⁹ It is also worth noting that no byproduct radical species were generated. In general, the generation of radical species could be one of the key factors that promote the formation of intermediates such as CO and CH₄ in the reduction of CO₂.⁵⁰ However, OH radicals were not produced under our experimental conditions (Fig. S2, ESI†). This is because the VB edges of C₃N₅ and g-C₃N₄ are far from the redox potential to generate OH radicals [E(H₂O/OH˙) = 2.72 V (vs. NHE)], and therefore, they would not be able to produce OH radicals from water.⁵¹ In addition, under the conditions of the current experiment, O₂ radicals are not produced because dissolved oxygen is no longer present due to CO₂ bubbling.

CO₂ dissolved in water is known to exist in the form of CO₂ (aq), carbonic acid (H₂CO₃), and carbonate ions (HCO₃⁻).⁵²

CO₂ + H₂O ⇄ CO₂ (aq) + H₂CO₃ ⇄ H⁺ + HCO₃⁻ (3)

It has been reported that the CO₂ concentration in CO₂ (aq) is approximately 500 times as large as that of H₂CO₃, and the concentration of H₂CO₃ in the solution is from 23 to 71 times greater than that of HCO₃⁻.⁵³ Although the concentration of H₂CO₃ is quite low, the reaction pathway via H₂CO₃ formation should be regarded. Taking thermodynamics into consideration, as the reduction potential of the H₂CO₃/CH₃OH pair (0.044 V (vs. NHE)) is more positive than that of CO₂/CH₃OH (−0.38 V (vs. NHE)), H₂CO₃ reduction will be more favorable than CO₂ reduction.^54,55 The same is true for the CH₂O formation because the reduction potentials of H₂CO₃/CH₂O and CO₂/CH₂O are −0.05 V (vs. NHE) and −0.52 V (vs. NHE), respectively.⁵⁵ The reaction pathways of H₂CO₃ reduction are not routed through carbon monoxide (CO) as an intermediate, which is consistent with our results without CO generation (Fig. 6).^53,55 From Fig. 3f, the CB edges of both C₃N₅ and g-C₃N₄ meet the reduction potentials to produce CH₃OH and CH₂O from H₂CO₃. Accordingly, CO₂ reduction via H₂CO₃ formation is one of the plausible scenarios in this work.

The oxidation reaction route can generate protons from pure water, that are essential for the formation of CH₃OH and CH₂O.

2H₂O + 4h⁺ → 4H⁺ + O₂ (4)

Here, we would like to note that protons can be slightly produced by reversible reaction of CO₂ in water (eqn (3)). Also, no change in the amount of CH₃OH and H₂ produced after 2 and 3 h (Fig. 6a and b) reflects that the oversaturated CO₂ solution is the source of H₂, and that CH₃OH does not contribute to the production of H₂.

Additional control experiments (Fig. S3, ESI†) clearly show that this photocatalytic CO₂ reduction was caused by C₃N₅ and g-C₃N₄. As for the photocatalytic cycling test, a CO₂ reduction experiment over C₃N₅ was performed three times, which resulted in no change in photocatalytic activity (Fig. S4, ESI†). Furthermore, XRD and TEM observations of the samples after the photocatalytic measurements revealed no significant changes in the morphology and crystal structure of C₃N₅ (Fig. S5 and S6, ESI†). The above results indicate that C₃N₅ is chemically stable while g-C₃N₄ is reported to be chemically unstable. The increased number of π-electrons may allow orbital interactions and improve the chemical stability of C₃N₅.⁵⁶

Conclusions

We have successfully synthesized C₃N₅, a nitrogen-abundant carbon nitride, by thermal polycondensation using 3-amino-1,2,4-triazole as a precursor. TEM images and specific surface area measurements showed that the synthesized C₃N₅ has a larger grain size than g-C₃N₄ prepared from melamine precursors, and that polymerization proceeds more easily in C₃N₅. XPS and FTIR studies suggested that part of the triazine frameworks in nitrogen-rich C₃N₅ is composed of five-membered rings. Then, the increase in nitrogen atoms with unshared electron pairs leads to the generation of more π-electrons in C₃N₅. This increase is reflected in the shorter interlayer distance of C₃N₅ with stronger π–π interactions. Furthermore, experimentally obtained energy band alignments revealed that C₃N₅ has a narrower band gap than g-C₃N₄ owing to a large negative shift of the valence band maximum dominated by N2p orbitals, which may also be attributed to the increased π-electrons in C₃N₅.

In the photocatalytic H₂ production from TEOA solution, the rate of H₂ production for C₃N₅ was about twice as high as that for g-C₃N₄. Considering the results of the NBT experiment, the promotion of the rate-determining proton reduction might lead to more efficient hydrogen production. Finally, in the liquid phase CO₂ photoreduction, C₃N₅ selectively produced CH₃OH and H₂, while g-C₃N₄ selectively produced CH₂O. The amount of CH₃OH produced by C₃N₅ was greater than that of formaldehyde produced by g-C₃N₄, indicating that C₃N₅ is more capable of multi-electron reduction than g-C₃N₄. To understand the observed product selectivity in the CO₂ photoreduction, two possible reaction pathways were considered and discussed based on the reversible changes of CO₂ in the water.

Eventually, all photocatalytic experiments in this study supported our idea that C₃N₅ alone has better photocatalytic properties than g-C₃N₄ alone. The results of this research will guide the synthesis of novel metal-free photocatalysts and contribute, in part, to triggering new breakthroughs for their practical applications.

Author contributions

Conceptualization: K. I. and K. N., methodology: K. I. and K. N., investigation: K. I., data curation: K. I., supervision: K. N., project administration: K. N., writing – original draft: K. I. and K. N., writing – review & editing: K. I. and K. N.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was supported by a Grant-in-Aid for Scientific Research (KAKENHI No. 19H02174) of the Japan Society for the Promotion of Science (JSPS). K. I. is very grateful for the support from JST SPRING Grant Number JPMJSP2123.

Notes and references

1. X. Wang, K. Maeda, A. Thomas, K. Takanabe, G. Xin, J. M. Carlsson, K. Domen and M. Antonietti, Nat. Mater., 2009, 8, 76–80.
2. P. Chen, X. Dong, M. Huang, K. Li, L. Xiao, J. Sheng, S. Chen, Y. Zhou and F. Dong, ACS Catal., 2022, 12, 4560–4570.
3. N. Jiang, L. Lyu, G. Yu, L. Zhanga and C. Hu, J. Mater. Chem. A, 2018, 6, 17819–17828.
4. X. Wang, J. Gong, Y. Dong, S. An, X. Zhang and J. Tian, Mater. Today Phys., 2022, 27, 100806.
5. S. Tasleem and M. Tahir, Int. J. Hydrogen Energy, 2021, 46, 20995–21012.
6. P. Kumar, E. Vahidzadeh, U. K. Thakur, P. Kar, K. M. Alam, A. Goswami, N. Mahdi, K. Cui, G. M. Bernard, V. K. Michaelis and K. Shankar, J. Am. Chem. Soc., 2019, 141, 5415–5436.
7. J. Zhang, B. Jing, Z. Tang, Z. Ao, D. Xia, M. Zhu and S. Wang, Appl. Catal., B, 2021, 13, 24907–24915.
8. H. Che, J. Wang, X. Gao, J. Chen, P. Wang, B. Liu and Y. Ao, J. Colloid Interface Sci., 2022, 627, 739–748.
9. J. Liu, S. Wang, C. Zhao and J. Zheng, Nanomaterials, 2023, 13, 499.
10. L. Wang, R. Chen, Z. Zhang, X. Chen, J. Ding, J. Zhang, H. Wan and G. Guan, J. Environ. Chem. Eng., 2023, 11, 109345.
11. S. Vadivel, M. Fujii and S. Rajendran, Chemosphere, 2022, 307, 135716.
12. S. Li, M. Cai, Y. Liu, J. Zhang, C. Wang, S. Zang, Y. Li, P. Zhang and X. Li, Inorg. Chem. Front., 2022, 9, 2479–2497.
13. W. Yang, K. Sun, J. Wan, Y.-A. Ma, J. Liu, B. Zhu, L. Liu and F. Fu, Appl. Catal., B, 2023, 320, 121978.
14. C. Zhua, Y. Zhanga, Z. Fana, F. Liua and A. Li, J. Hazard. Mater., 2020, 393, 122395.
15. C.-H. Chang, C.-L. Wang and B.-R. Li, Biosens. Bioelectron., 2023, 236, 115403.
16. J. Xu, L. Zhang, R. Shi and Y. Zhu, J. Mater. Chem. A, 2013, 1, 14766–14772.
17. Z. Xiong, Y. Liang, J. Yang, G. Yang, J. Jia, K. Sa, X. Zhang and Z. Zeng, Sep. Purif. Technol., 2023, 306, 122522.
18. A. Krishnan, M. Yoosuf, K. Archana, A. S. Arsha and A. Viswam, J. Energy Chem., 2023, 80, 562–583.
19. W. J. Ong, L. L. Tan, Y. H. Ng, S. T. Yong and S. P. Chai, Chem. Rev., 2016, 116, 7159–7329.
20. Y. Zhao, J. Zhang and L. Qu, ChemNanoMat, 2015, 1, 298–318.
21. X. Wu, R. Zhong, X. Lv, Z. Hu, D. Xia, C. Li, B. Song and S. Liu, Appl. Catal., B, 2023, 330, 122666.
22. J. Ding, Q. Tang, Y. Fu, Y. Zhang, J. Hu, T. Li, Q. Zhong, M. Fan and H. H. Kung, J. Am. Chem. Soc., 2022, 144, 9576–9585.
23. X. Guan, M. Fawaz, R. Sarkar, C.-H. Lin, Z. Li, Z. Lei, P. D. Nithinraj, P. Kumar, X. Zhang, J.-H. Yang, L. Hu, T. Wu, S. Chakraborty, J. Yi and A. Vinu, J. Mater. Chem. A, 2023, 11, 12837–12845.
24. X. Li, D. Shen, C. Liu, J. Li, Y. Zhou, X. Song, P. Huo, H. Wang and Y. Yan, J. Colloid Interface Sci., 2019, 554, 468–478.
25. L. Yang, X. Liu, Z. Liu, C. Wang, G. Liu, Q. Li and X. Feng, Ceram. Int., 2018, 44, 20613–20619.
26. L. V. Goncharova, Basic Surfaces and their Analysis, Morgan & Claypool Publishers, USA, 2018, 6, 1–17.
27. Q. Zhang, Z. Zhang, S. Xu, A. Liu, L. Da, D. Lin and C. Jiang, Anal. Chem., 2023, 95, 4536–4542.
28. X. Huang, X. Xu, R. Yang and X. Fu, Colloids Surf., A, 2022, 643, 128738.
29. X. An, Q. Tang, H. Lan, H. Liu, X. Yu, J. Qu, H. Lin and J. Ye, Angew. Chem., Int. Ed., 2022, 134, e202212706.
30. B. Yang, J. Han, Q. Zhang, G. Liao, W. Cheng, G. Ge, J. Liu, X. Yang, R. Wang and X. Jia, Carbon, 2023, 202, 348–357.
31. P. V. Viet, T.-D. Nguyen, D.-P. Bui and C. M. Thi, J. Materiomics, 2022, 8, 1–8.
32. M. Mohammadikish and N. Mosleh, Appl. Organomet. Chem., 2023, e7177.
33. Z. Cai, Y. Huang, H. Ji, W. Liu, J. Fu and X. Sun, Sep. Purif. Technol., 2022, 280, 119772.
34. R. Ren, G. Liu, J. Y. Kim, R. E. A. Ardhi, M. X. Tran, W. Yang and J. K. Lee, Appl. Catal., B, 2022, 306, 121096.
35. X. Ma, Z. Ma, H. Zhang, D. Lu, J. Duan and B. Hou, J. Photochem. Photobiol., A, 2022, 426, 113772.
36. Y. Liang, X. Wu, X. Liu, C. Li and S. Liu, Appl. Catal., B, 2022, 304, 120978.
37. W. Hoa, Z. Zhanga, M. Xub, X. Zhang, X. Wang and Y. Huang, Appl. Catal., B, 2015, 179, 106–112.
38. J. Chen, Y. Xiao, N. Wang, X. Kang, D. Wang, C. Wang, J. Liu, Y. Jiang and H. Fu, Sci. China Mater., 2023, 66, 3165–3175.
39. D. Dai, X. Liang, B. Zhang, Y. Wang, Q. Wu, X. Bao, Z. Wang, Z. Zheng, H. Cheng, Y. Dai, B. Huang and P. Wang, Adv. Sci., 2022, 9, 2105299.
40. Y.-J. Yuan, N. Lu, L. Bao, R. Tang, F.-G. Zhang, J. Guan, H.-D. Wang, Q.-Y. Liu, Q. Cheng, Z.-T. Yu and Z. Zou, ACS Nano, 2022, 16, 12174–12184.
41. V. Kumaravel, M. D. Imam, A. Badreldin, R. K. Chava, J. Y. Do, M. Kang and A. A. Wahab, Catalysts, 2019, 9, 276.
42. H. Jun, S. Choi, M. Y. Yang and Y. S. Nam, J. Mater. Chem. A, 2019, 7, 17254–17260.
43. J. Byun, W. Huang, D. Wang, R. Li and K. A. I. Zhang, Angew. Chem., Int. Ed., 2018, 57, 2967–2971.
44. B. Liu, X. Zhao, C. Terashima, A. Fujishima and K. Nakata, Phys. Chem. Chem. Phys., 2014, 16, 8751–8760.
45. N. F. Khusnun, A. A. Jalil, T. A. T. Abdullah, S. S. M. Latip, C. N. C. Hitam, A. A. Fauzi, N. S. Hassan, M. A. H. Aziz, A. F. A. Rahman, F. F. A. Aziz, M. Bahari, R. H. Adnan and R. Saravanan, J. CO2 Util., 2022, 58, 101908.
46. X. Wanga, Y. Wanga, M. Gaob, J. Shenb, X. Puc, Z. Zhangb, H. Linb and X. Wang, Appl. Catal., B, 2020, 270, 118876.
47. A. B. Anderson and H. A. Asiri, Phys. Chem. Chem. Phys., 2014, 16, 10587–10599.
48. X. Wu, Y. Li, G. Zhang, H. Chen, J. Li, K. Wang, Y. Pan, Y. Zhao, Y. Sun and Y. Xie, J. Am. Chem. Soc., 2019, 141, 5267–5274.
49. A. Li, T. Wang, C. Li, Z. Huang, Z. Luo and J. Gong, Angew. Chem., Int. Ed., 2019, 58, 3804–3808.
50. M. Ma, Z. Huang, R. Wang, R. Zhang, T. Yang, Z. Rao, W. Fa, F. Zhang, Y. Cao, S. Yu and Y. Zhou, Green Chem., 2022, 24, 8791–8799.
51. K. Ito, R. Uchida and K. Noda, J. Photochem. Photobiol., A, 2023, 443, 114824.
52. T. Zhao, J. Li, J. Liu, F. Liu, K. Xu, M. Yu, W. Xu and F. Cheng, ACS Catal., 2023, 13, 4444–4453.
53. S. Pocoví-Martínez, I. Zumeta-Dube and D. Diaz, J. Nanomater., 2019, 6461493.
54. A. L. Soli and R. H. Byrne, Mar. Chem., 2002, 78, 65–73.
55. X. Yang, T. Xiao and P. P. Edwards, Int. J. Hydrogen Energy, 2011, 36, 6546–6552.
56. S. Shen, Y. Zeng, X. Li, L. Meng and X. Zhang, Int. J. Quantum Chem., 2018, 118, e25521.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3cp04431a

---