Facile in situ reductive synthesis of both nitrogen deficient and protonated g-C₃N₄ nanosheets for the synergistic enhancement of visible-light H₂ evolution

Abstract

A new strategy is reported here to synthesize both nitrogen deficient and protonated graphitic carbon nitride (g-C₃N₄) nanosheets by the conjoint use of NH₄Cl as a dynamic gas template together with hypophosphorous acid (H₃PO₂) as a doping agent. The NH₄Cl treatment allows for the scalable production of protonated g-C₃N₄ nanosheets. With the corresponding co-addition of H₃PO₂, nitrogen vacancies, accompanied by both additional protons and interstitially-doped phosphorus, are introduced into the g-C₃N₄ framework, and the electronic bandgap of g-C₃N₄ nanosheets as well as their optical properties and hydrogen-production performance can be precisely tuned by careful adjustment of the H₃PO₂ treatment. This conjoint approach thereby results in improved visible-light absorption, enhanced charge-carrier separation and a high H₂ evolution rate of 881.7 μmol h⁻¹ achieved over the H₃PO₂ doped g-C₃N₄ nanosheets with a corresponding apparent quantum yield (AQY) of 40.4% (at 420 nm). We illustrate that the synergistic H₃PO₂ doping modifies the layered g-C₃N₄ materials by introducing nitrogen vacancies as well as protonating them, leading to significant photocatalytic H₂ evolution enhancements, while the g-C₃N₄ materials doped with phosphoric acid (H₃PO₄) are simply protonated further, revealing the varied doping effects of phosphorus having different (but accessible) valence states.

---

Introduction

Graphitic carbon nitride (g-C₃N₄), a prototypical metal-free semiconductor, has been intensively studied owing to its excellent photocatalytic applications in H₂ production, environment remediation, CO₂ reduction and photosynthesis.^1–5 Bulk g-C₃N₄ with high thermal, chemical and photocatalytic stabilities can be readily synthesized via direct thermal polymerization of nitrogen-rich precursors. However, the visible-light photocatalytic efficiency of such directly polymerized g-C₃N₄ material is far from satisfactory stemming from its limited visible-light absorption range, low surface area, confined active sites, and rapid charge-carrier recombination. Thus, various techniques including morphology engineering,^6,7 constructing Z-scheme heterostructure or unique surface bonding states,^8–10 doping with both nonmetal and metal elements,^11,12 introducing elemental defects or vacancies and protonation have been applied in order to improve the photocatalytic performance of g-C₃N₄.^13–15 Inspired by the two-dimensional graphene-like nanosheet of the material, efforts have also been made to transform bulk g-C₃N₄ into porous structures composing thin nanosheets which help deliver superior photocatalytic activities.⁷

Recently, it has been demonstrated that the photocatalytic activity of g-C₃N₄ under visible-light irradiation can be significantly enhanced by introducing nitrogen defects into the basic g-C₃N₄ “melon” structure as well as doping with compounds of phosphorus.^2,12 To date, nitrogen deficient g-C₃N₄ was synthesized through alkali-assisted polymerization, hydrogen reduction, high-temperature thermal polymerization and hydrothermal routes.^2,13,16,17 Unlike other methods-which routinely suffer from control difficulties-the so-called alkali-assisted polymerization achieves a highly effective control of both the type and the abundance of the introduced nitrogen defects leading to enhanced photocatalytic H₂ evolution.² However, this alkali-assisted approach, using KOH, NaOH or Ba(OH)₂ is unlikely to work in the “bottom-up” strategy which utilizes NH₄Cl as a so-called dynamic gas template for scalable production of g-C₃N₄ nanosheets, due to the non-coexistence of NH₄Cl and alkali metal hydroxides.⁷ Moreover, it has been shown that the alkali-assisted process progressively decreases the layer stacking distance and reduces the specific surface area of g-C₃N₄ nanosheets when excessive KOH is added.² This suggests the over-addition of alkali metal hydroxides would cause a severe decrease in photocatalytic activity because of the unfavorable crystal morphology changes.

Phosphorus, has emerged as a likely contender to be introduced into the g-C₃N₄ matrix to achieve enhanced photocatalytic performance. In most of the prior work,^15,18–20 pentavalent phosphorus such as H₃PO₄, (NH₄)₃PO₄, BmimPF₆ ionic liquid, Na₂HPO₄, and CO(NH₂)₂·H₃PO₄ figures strongly as the phosphorus doping agents. To the best of our knowledge, there is little work reported on the synthesis of phosphorus-doped g-C₃N₄ using phosphorus compounds having lower valence states and the photocatalytic ability of such low-valent phosphorus doped g-C₃N₄ has hardly been studied.⁴

Here, we present a one-step, in situ reduction method utilizing the strategy of using NH₄Cl as dynamic gas template and H₃PO₂ doping for the synthesis of nitrogen deficient g-C₃N₄ nanosheets. The photocatalytic activity of g-C₃N₄ was greatly enhanced with the formation of both protonated and porous nanosheets by using NH₄Cl as dynamic gas template. With the co-addition of H₃PO₄ as dopant, g-C₃N₄ nanosheets were further protonated and delivered superior photocatalytic activities. However, besides protonation, the g-C₃N₄ nanosheets doped by the lower-valent phosphorus compound H₃PO₂ also led to a reduced electronic bandgap in the material and enhanced visible light absorption. The differences observed between the H₃PO₂ and H₃PO₄ doped g-C₃N₄ nanosheets promotes a discussion of using dopants with phosphorus at different valence states. It demonstrates the possibility of using low-valent phosphorus compounds to both synthesize and modify g-C₃N₄ in one single step, thereby offering a comprehensive enhanced and efficient visible-light photocatalytic performance. Thus, the preparation strategy reported here represents an effective approach to optimize the morphology, chemical composition, optical response and resulting activity of g-C₃N₄ photocatalysts.

Experimental

The standard bulk g-C₃N₄ reference (B-CN) was prepared by the direct pyrolysis polymerization of dicyandiamide. Preliminary, 4 g dicyandiamide powder was placed in a 50 mL alumina crucible with cover and then was calcined in static air at 550 °C for 4.5 hours with a ramping rate of 2.3 °C min⁻¹. The g-C₃N₄ nanosheets were prepared by the modified “bottom-up” strategy with NH₄Cl as dynamic gas template,⁷ and the as prepared g-C₃N₄ nanosheets without and with addition of phosphorus dopants were labeled as G-CN and M-Pn-CN, respectively. Here, M represents the phosphorus dopant usage (weight percentage of the added phosphorus element to the dicyandiamide precursor, ranging from 0.2 to 3.2) and Pn (n = 1 or 2) meant that the corresponding dopant was H₃PO₄ and H₃PO₂, respectively. Firstly, 4 g dicyandiamide and 20 g NH₄Cl were premixed with 5 mL solution containing the designed amount of dopant (for G-CN, the addition amount of dopants was zero). For example, in the preparation of 0.8-Pn-CN samples, the addition amounts of 85 wt% H₃PO₄ and 50 wt% H₃PO₂ were 119.0 and 136.3 mg, respectively. Then, the powder and solution mixture were stirred to form homogeneous slurry. Finally, the slurry was calcined with the same procedures in B-CN preparation. All the as prepared samples were collected and stored at room temperature in normal atmosphere.

Full experimental details including reagents, characterization, transient photocurrent measurement, RTK-Solar visible-light H₂ evolution system (Fig. S1†) and AQY measurement can be found in the ESI.†

Results and discussion

Morphology

Fig. 1A–D show representative g-C₃N₄ morphologies characterized by field emission electron scanning microscope (SEM). Compared with the B-CN powder which shows a large agglomerate appearance (Fig. 1A), the g-C₃N₄ synthesized using the NH₄Cl-assisted “bottom-up” strategy shows characteristics of thin nanosheets with crinkly structures (Fig. 1B–D). It can be seen that G-CN is mainly consisted of micrometer scale petals-like nanosheets with relatively smooth surfaces and edges (Fig. 1B). The 0.8-P1-CN nanosheets (Fig. 1C) remain a flower-like structure and part of the nanosheets become much smaller with serrated edges, which could be ascribed to the reaction of g-C₃N₄ nanosheets with H₃PO₄.¹⁸ In contrast, few flower-like structures are found in the H₃PO₂ doped powder (Fig. 1D) and even smaller nanosheets can be observed with some large nanosheets surrounded. Further to transmission electron microscope (TEM) images, the dense stacking structure of bulk B-CN is illustrated in Fig. 1E, while the as prepared g-C₃N₄ nanosheets using NH₄Cl as dynamic gas template (Fig. 1F–H) present porous structures. In the N₂ physisorption measurements, the specific surface areas, pore volumes and pore size distributions of the g-C₃N₄ samples were obtained (Fig. 1I). All the tested g-C₃N₄ samples possess well-defined mesopores. The specific areas and pore volumes of g-C₃N₄ nanosheets are nearly 3 and 4.5 times enlarged than those of B-CN, respectively. Thus, the porous g-C₃N₄ nanosheets assemblages appeared to us to provide more active sites for an enhanced photocatalytic performance.

Fig. 1 (A, B, C and D) SEM images of B-CN, G-CN, 0.8-P1-CN and 0.8-P2-CN, respectively. (E, F, G and H) TEM images of B-CN, G-CN, 0.8-P1-CN and 0.8-P2-CN, respectively. (I) N₂ physisorption measurement results.

Structure characterization

In Fig. 2A we present the X-ray diffraction (XRD) patterns of the g-C₃N₄ samples. All samples show the two characteristic diffraction peaks at around 13.0° and 27.4°, which were assigned to the in-plane (100) and interlayer-stacking (002) crystal planes of g-C₃N₄, respectively.² Compared with the peaks in B-CN, diffraction peaks of g-C₃N₄ nanosheets samples (G-CN and M-Pn-CN) became weaker and the peak intensities of the phosphorus doped g-C₃N₄ nanosheets were continuously weakened with increasing addition of phosphorus dopants. Notably, all H₃PO₂ doped g-C₃N₄ samples shows lower peak intensities compared to the corresponding H₃PO₄ doped g-C₃N₄ nanosheets at the same amount of phosphorus usage, which indicates H₃PO₂ could more easily cause the loss of ordered structures within g-C₃N₄ framework.² The magnified XRD patterns (inset in Fig. 2A) depict that the (002) diffraction peak of G-CN nanosheets shifted to 27.8°, suggesting a smaller gallery distance between the basic layers in the nanosheets. Earlier pioneering work^6,21 demonstrated that so-called planarizing the potentially undulated layers in g-C₃N₄ would result in denser stacking, thereby resulting in shifts in the characteristic (002) diffraction peaks to higher 2θ angles. Thus, the (002) diffraction peaks shifts in XRD patterns also indicate the formation of planarized g-C₃N₄ nanosheets as also demonstrated in our SEM studies. Interestingly, the (002) peaks of the H₃PO₂ doped g-C₃N₄ nanosheets continuously shifted back to around 27.5° with increasing dopant addition, while those of the H₃PO₄ doped samples were nearly centered at the same position with G-CN (ESI, Fig. S2†). This (002) peak shift in the H₃PO₂ doped nanosheets should be attributed to the increased disorder in the g-C₃N₄ in-planar matrix and consequently resulted in larger layer distance compared with G-CN and H₃PO₄ doped g-C₃N₄ nanosheets.^1,2 Thus, this interesting observation suggests that this H₃PO₂ doping protocol could help to relieve the layer stacking problems with remaining nanosheets structures.

Fig. 2 (A) XRD patterns of representative g-C₃N₄ samples. (B) FTIR spectra of these same samples.

In the Fourier Transform Infrared (FTIR) spectroscopy (Fig. 2B), the typical peaks of g-C₃N₄ at 810 cm⁻¹ and in the range of 900–1800 cm⁻¹ were observed among all the tested g-C₃N₄ samples, which originated from the heptazine ring out-of-plane bending and the N–C=N heterorings in the “melon” framework.² Compared with the pristine B-CN, three distinct changes (highlighted by the orange shaded regions) occurred with varying dopants and their usages. One was the progressive loss of peak intensities in the phosphorus doped nanosheets located between 3000 and 3300 cm⁻¹ which were attributed to the N–H stretching vibrations. Another difference observed at 2184 cm⁻¹ as a new peak appeared in H₃PO₂ doped g-C₃N₄ nanosheets representing the stretching vibration mode of cyano groups (–C≡N).^2,22,23 Moreover, a weak and broad peak centered at around 3460 cm⁻¹ was developed in the g-C₃N₄ nanosheets samples and this peak could be assigned to the stretching vibration mode of hydroxyl groups (–OH).^24,25 It is noteworthy that the hydroxyl group stretching vibration peaks in the H₃PO₂ doped nanosheets are relatively stronger than those in the H₃PO₄ doped samples, suggesting more exposure of hydroxyl groups in the H₃PO₂ doped g-C₃N₄ nanosheets, which will be further evidenced by the after-mentioned X-ray photoelectron spectroscopy analysis.

The FTIR results reveal that the addition of H₃PO₂ and H₃PO₄ in g-C₃N₄ nanosheets synthesis could significantly decrease N–H concentration and cyano groups can be very effectively introduced by using H₃PO₂. The introduction of oxygen containing groups like hydroxyl group should be attributed to the high temperature (550 °C) synthesis and the atmospheric sample storage. Notably, recent research results claimed that the photocatalytic activity of g-C₃N₄ can be greatly enhanced by introducing cyano groups into the melon framework.^2,24 Thus, the H₃PO₂ doped g-C₃N₄ nanosheets are expected to possess superior photocatalytic activities as would be demonstrated in the H₂ production test.

Besides XRD and FTIR studies, the chemical structures of the g-C₃N₄ samples were further characterized by solid-state ¹³C, ³¹P and ¹H magic angle spinning (MAS) NMR spectra. Fig. 3A presents the ¹³C NMR spectra, in which all samples show two strong peaks at 156.3 and 164.6 ppm corresponding to the chemical shifts of C_3N (1) and C_2N–NHx (2) in the corrugated g-C₃N₄ melon networks, respectively.^26–28 Compared with B-CN, the C_3N (1) peak in the g-C₃N₄ nanosheets powders was intensified, which indicates the loss of NH_x groups, in accordance with the FTIR results. Moreover, G-CN and 0.8-P1-CN were found very similar, suggesting that the nonoxidative H₃PO₄ would not strongly disintegrate the g-C₃N₄ heptazine units. However, the C_3N (1) peak was further intensified in the 0.8-P2-CN sample and three new peaks (peak 3, 4 and 5) emerged. These new peaks with a chemical shift of 119.9, 112.3 and 208.1 ppm could be assigned to the carbon atoms in cyano groups, sp² hybridized carbon atoms and the carbon atoms in the carbonyl group (–C=O) containing species, respectively.^2,28–30

Fig. 3 (A) Solid-state NMR spectra of ¹³C and basic g-C₃N₄ triazine unit structure. (B and C) Solid-state ³¹P and ¹H NMR spectrums, respectively.

No obvious differences are observed on the ³¹P NMR spectra (Fig. 3B) of 0.8-P2-CN and 0.8-P1-CN nanosheets. Both samples show single sharp peak centered at −0.49 ppm which corresponds to the P–N coordinate bonds with phosphorus atom being connected to two adjacent pyridinic N atoms from two separated g-C₃N₄ triazine units,³¹ indicating that the phosphorus atoms sourced from H₃PO₂ and H₃PO₄ were in the same final state and interstitially doped in the g-C₃N₄ matrices.³² As the synthesis was conducted in the high temperature air, these doped phosphorus atoms would most likely be in pentavalent oxidation state and absorb moisture from air and eventually transform into the form of phosphoric acid. In terms of the reasons for the differences between the H₃PO₄ and H₃PO₂ doped g-C₃N₄ nanosheets, a detailed discussion will be presented later.

Considering NH₄Cl and H₃PO₂ would generate HCl and H₃PO₄ which were commonly used for g-C₃N₄ protonation,^14,18 the as prepared g-C₃N₄ could be protonated. As seen in Fig. 3C, the five peaks at 2.2, 5.1, 8.7, 11.5 and 13.5 ppm in the ¹H NMR spectra were assigned to the protons in aliphatic groups, residual water, NH_x groups, the hydrogen bonds in acid groups (such as carboxylic acid and phosphoric acid) and g-C₃N₄ framework,^33–35 respectively. Thus, the intensification of the 8.7 or 11.5 ppm peak can be regarded as evidence for the protonation of g-C₃N₄ nanosheets. Compared with B-CN, the 8.7 ppm peak in G-CN was obviously strengthened because of the formation of hydrogen bonds between the proton and nitrogen atoms in the g-C₃N₄ framework. Despite of the weakened peaks at 8.7 ppm over the phosphorus doped nanosheets, the intensified peaks at 11.5 ppm in 0.8-P1-CN and 0.8-P2-CN should be attributed to the protonation with protons connected with phosphoric acid groups.

The measured zeta potentials (ESI, Fig. S3†) reveal that the surface charge properties were significantly changed in the g-C₃N₄ nanosheets. The zeta potentials of G-CN, 0.8-P1-CN and 0.8-P2-CN were increased from −16.3 mV (B-CN) to −6.1, −3.1 and −4.5 mV, respectively, and these zeta potential changes further confirm the protonation of the g-C₃N₄ nanosheets.¹⁵ Besides the positive potential on the dispersibility, electronic bandgap structure and surface area of g-C₃N₄, one of the other important advantages of protonation is to induce higher ionic conductivity to the g-C₃N₄ framework and then enable the acceleration of charge carrier migration, which would benefit for better photocatalytic performances.^14,18,36 Thus, higher H₂ evolution ability could be expected for the protonated g-C₃N₄ nanosheets.

In addition, combining with the newly emerged carbonyl group peak in the ¹³C spectra and the enhanced acid groups hydrogen bonds ¹H NMR signal in 0.8-P2-CN nanosheets, it is reasonable to propose that carboxyl groups (–COOH) were introduced into the H₃PO₂ doped nanosheets, thereby resulting in a stronger 11.5 ppm peak in 0.8-P2-CN than that of 0.8-P1-CN, and this peak intensification is in agreement with the observation that the H₃PO₂ doped g-C₃N₄ nanosheets showed stronger hydroxyl group (–OH) peak intensities in FTIR spectra.

Chemical compositions

The g-C₃N₄ compositions obtained from organic elemental analysis (OEA) and X-ray photoelectron spectroscopy (XPS) can be found in Table 1 (ESI, XPS spectra in Fig. S4†). As can be seen in Table 1, the increased H/C ratios in g-C₃N₄ nanosheets can also be regarded as evidences for protonation.¹⁵ The N/C atomic ratio of B-CN obtained from OEA and XPS were 1.467 and 1.318, respectively. According to XPS results, the N/C ratio of G-CN slightly decreased from 1.318 to 1.283 when B-CN was transformed into G-CN, and this was ascribed to higher degree polymerization with more NH₃ release during the formation of nanosheet structures.⁷ Although the N/C ratio of H₃PO₄ doped nanosheets dropped to about 1.249, it was still very close to that of G-CN. However, the N/C ratio of H₃PO₂ doped g-C₃N₄ nanosheets fell to 0.904 as a result of the intensive loss of nitrogen caused by the addition of H₃PO₂. Meanwhile, the O/C ratio of the H₃PO₂ doped g-C₃N₄ nanosheets drastically increased to 0.279. Thus, it was most likely that the H₃PO₂ induced the opening of g-C₃N₄ heptazine rings and caused nitrogen vacancies. The exposed defective edges would then react with oxygen to form oxygen containing groups such as carboxyl groups etc.²⁴

Table 1 Element atomic ratios of the g-C₃N₄ samples^a

Sample	OEA			XPS
	N/C	O/C	H/C	N/C	O/C	P/C
a OEA and XPS represent the results obtained from organic elemental analysis and X-ray photoelectron spectroscopy, respectively.
B-CN	1.467	0.007	0.023	1.318	0.038	0
G-CN	1.421	0.018	0.027	1.283	0.064	0
0.2-P1-CN	1.389	0.047	0.029	1.287	0.079	0.018
0.8-P1-CN	1.381	0.064	0.031	1.266	0.105	0.038
3.2-P1-CN	1.365	0.073	0.032	1.249	0.112	0.043
0.2-P2-CN	1.293	0.091	0.030	1.204	0.127	0.010
0.8-P2-CN	1.117	0.135	0.035	1.037	0.223	0.017
3.2-P2-CN	1.002	0.162	0.037	0.904	0.279	0.024

---

In accordance with the fact that H₃PO₂ will decompose to release both PH₃ and H₃PO₄,⁴ the differences highlighted between the H₃PO₂ and H₃PO₄ doped g-C₃N₄ nanosheets provides strong evidence that the introduction of nitrogen vacancies can be attributed to the strongly reductive PH₃ rather than H₃PO₄. In the reduction synthesis of nitrogen defective g-C₃N₄,¹⁶ the energy changes for removing a lattice nitrogen atom located at N1, N2, N3 and N4 (referring to Fig. 3A) and terminating the dangling bonds of C atoms with H atoms using H₂ as reduction agent were 0.65, 0.83, 1.40 and 2.39 eV, respectively,¹⁶ and the H₂ reduction caused homogeneous sp² hybridized nitrogen atoms losses on the g-C₃N₄ heptazine rings. Changing H₂ to PH₃, we calculated the corresponding energy changes and the four energy changes decrease to 0.61, 0.75, 1.28 and 2.27 eV (ESI†), respectively, indicating the stronger reduction agent PH₃ would be more favorable for opening heptazine rings and result in significant nitrogen atoms in situ reduction removal. Thus, it was reasonable to propose that intensive nitrogen vacancies could be introduced into the H₃PO₂ doped g-C₃N₄ nanosheets because of that the strong reducing agent PH₃, produced via H₃PO₂ decomposition.

The narrow scan C 1s, N 1s, P 2p and O 1s XPS spectra were collected and deconvoluted into their components (Fig. 4A–D). The four peaks in C 1s spectra are located at 284.8, 286.2, 288.1 and 289 eV, corresponding to the adventitious hydrocarbons (C–C or C=C), C–NH_x (x = 1, 2), N–C=N coordination and carboxyl groups, respectively.^2,37,38 It is obvious that the 288.1 eV signal lost intensity, while the 286.2 eV peak is significantly intensified in the 0.8-P2-CN powder. As cyano group has similar C 1s binding energy to C-NH_x,² the intensified 286.2 eV signal together with the weakened N–C=N coordination peak could be further evidence for the heptazine rings opening and the formation of cyano groups in the H₃PO₂ doped nanosheets. Besides, the enhanced 289 eV peak in 0.8-P2-CN confirms the considerable introduction of carboxyl groups into H₃PO₂ doped g-C₃N₄ nanosheets. The three peaks centered at 398.6, 400 and 401 eV in the N 1s XPS spectra are assigned to the sp²-hybridized (N_2C), sp³-hybridized (N_3C) nitrogen atoms and NH_x groups, respectively.² For the H₃PO₂ doped nanosheets, the drastically weakened NH_x peak and shift of N_3C to lower binding energy also indicate the NH_x loss and formation of cyano groups, which are in agreement with the results obtained by FTIR and NMR spectra.

Fig. 4 (A) C 1s XPS narrow scan. (B) N 1s XPS narrow scan. (C) P 2p XPS narrow scan. (D) O 1s XPS narrow scan.

Both the P 2p XPS spectra of the H₃PO₄ and H₃PO₂ doped nanosheets show one single pentavalent phosphorus peak centered at 133.8 eV, indicating that no P–C bonds formed.¹⁸ Moreover, the elemental mapping conducted on scanning transmission microscope (ESI, Fig. S5†) reveal that phosphorus is distributed evenly in both 0.8-P1-CN and 0.8-P2-CN with an actual content of around 0.5 wt%. The parallel intensification of the 533.5 and 532.2 eV peaks representing –OH and C=O groups^39–41 in the 0.8-P2-CN O 1s spectra are consistent with the hypothesis that carboxyl groups was introduced into the H₃PO₂ doped g-C₃N₄ nanosheets. According to the summarized XPS data (ESI, Table S1†), the N_2C/C ratio in the g-C₃N₄ nanosheets remarkably dropped from 0.917 to 0.568 and the corresponding O/C ratio increased from 0.062 to 0.217 with the addition of H₃PO₂, these results also suggest the introduction of nitrogen defects and oxygen species into the H₃PO₂ doped g-C₃N₄ nanosheets.

Proposed schematic structures

Further to our XPS and OEA analysis, electron paramagnetic resonance (EPR) experiments were conducted to provide fingerprint evidence for probing the surface nitrogen vacancies introduced to the g-C₃N₄ nanosheets. Several different samples having closely similar weights were examined to ensure a meaningful comparison of the peak intensities (areas). As shown in Fig. 5, all samples display a single Lorentzian line with an electronic g value of 2.0043 in the magnetic field from 3460 to 3560 G, which represents the unpaired electrons of sp² hybrid carbon atoms in π-conjugated aromatic rings.^42,43 Thus the formation of two-coordinated nitrogen vacancies in the heptazine rings of g-C₃N₄ would donate unpaired electrons to the sp²-carbon atoms. Therefore, compared with the negligible EPR signals of B-CN, G-CN and 0.8-P1-CN, the significantly enhanced EPR intensity in 0.8-P2-CN nanosheets adds weight to the idea of the opening of heptazine rings and the formation of two-coordinated nitrogen vacancies.

Fig. 5 EPR spectra of the representative g-C₃N₄ samples.

Thus, we believe that combining the various techniques including FTIR, NMR, zeta potential, XPS and EPR analysis, the controllable generation of nitrogen vacancies, protonation and newly introduced functional groups in g-C₃N₄ framework are soundly confirmed. From all this information, the proposed schematic molecule structures and evolution processes of the g-C₃N₄ samples can now be advanced (Fig. 6).

Fig. 6 The g-C₃N₄ evolution process and proposed chemical structures.

In brief, compared with the directly polymerized bulk g-C₃N₄ agglomerate, protonated g-C₃N₄ nanosheets could be readily fabricated with the help of dynamic gas template NH₄Cl. With co-addition of H₃PO₄ as dopant, the g-C₃N₄ nanosheets could be further protonated with phosphorus atoms interstitially doped into the frameworks. However, apart from the additional protonation, the addition of the reducing dopant H₃PO₂ could also induce significant nitrogen loss to generate nitrogen vacancies and variously introduce, cyano and carboxyl functional groups, into the g-C₃N₄ nanosheets.

Optical properties

The optical properties and light harvesting abilities of g-C₃N₄ samples were significantly modified by the specific addition of NH₄Cl and phosphorus dopants. The obvious changes were well observed as the yellow dense B-CN powder was transformed into porous and fluffy states having different colours (Fig. 7A). Compared with B-CN and G-CN, the colours of the M-P1-CN nanosheets became lighter with the increasing addition of H₃PO₄, whereas those of the H₃PO₂ doped nanosheets turned from dark yellow to brown. The absorption edge of G-CN shifted to a lower wavelength as compared with B-CN and it continuously shifted to the lower direction with increasing H₃PO₄ addition. The blueshifts in the G-CN and H₃PO₄ doped nanosheets were attributed to the enhanced protonation.^5,14 Unlike H₃PO₄, a progressive redshift was achieved by increasing H₃PO₂ usage, which indicates the bandgap structures of the H₃PO₂ doped nanosheets could be easily tuned by adjusting the amount of H₃PO₂.

Fig. 7 (A) UV-Vis DRS spectra and. (B) Plots of transformed Kubelka–Munk function versus proton energy for a variety of g-C₃N₄ samples.

According to the transformed Kubelka–Munk function determined electronic bandgaps shown in Fig. 7B, the bandgaps of H₃PO₄ doped g-C₃N₄ nanosheets were slightly increased to a narrow range of 2.70–2.74 eV from 2.67 eV (G-CN), while that of the H₃PO₂ doped nanosheets can be significantly narrowed to 2.13 eV. The narrowed bandgaps reveal the enhanced visible light harvesting ability of the nitrogen deficient g-C₃N₄ nanosheets which were doped by H₃PO₂.

To understand the influence of protonation (including both the protons which connect with nitrogen atoms and the interstitially doped phosphoric acid groups), nitrogen vacancies and carboxyl groups (–COOH) on the bandgaps of the g-C₃N₄ samples, partial density of states (PDOS) and density-functional theory (DFT) calculations were performed (ESI for details, Fig. S6†).

As shown in Fig. 8A and B, the calculated bandgap for B-CN is 2.67 eV which becomes slightly enhanced to 2.74 eV in the protonated g-C₃N₄ nanosheets, indicating that the protonation (no matter induced by NH₄Cl gas template or interstitially doped phosphorus) does not significantly impact on the magnitude of bandgap. To clarify the effects of the nitrogen vacancies and carboxyl groups, g-C₃N₄ unit cells only containing nitrogen vacancies or carboxyl groups were built, respectively. The calculations in Fig. 8C and D show that the nitrogen vacancies and the carboxyl groups decrease the bandgap to 2.52 and 2.20 eV, respectively. The defect energy level observed in the g-C₃N₄ only containing nitrogen vacancies, composing of both C 2p and N 2p orbitals, is about 1.5 eV above valence band (VB). The narrowed bandgap of the g-C₃N₄ containing nitrogen vacancies and carboxyl groups agrees with the trend observed in the UV-Vis DRS analysis above. As the PDOS seen in Fig. 8E, the conduction band (CB) of B-CN is composed of C 2p and N 2p orbits, while C 2p and N 2p orbits mainly contribute to valence band; this result is consistent with previous work.^2,44 According to Fig. 8F, the significantly narrowed bandgap width of the H₃PO₂ doped g-C₃N₄ nanosheets is due primarily to a lowering of CB minimum by about 0.4 eV with the emergence of defect and carboxyl states. These results therefor confirm that the coexistence of nitrogen vacancies and carboxyl groups would decrease the width of g-C₃N₄ bandgap. As confirmed in previous work,² the coexistence of cyano groups and nitrogen vacancies would also lower the CB minimum and result in narrower electronic energy bandgaps. By analogy, the coexistence of cyano groups, carboxyl groups and nitrogen vacancies would synergistically decrease the bandgap width of the as prepared g-C₃N₄ nanosheets which were doped by H₃PO₂.

Fig. 8 (A–D) Calculated band structures of B-CN and g-C₃N₄ with protonation, nitrogen vacancies and carboxyl groups, respectively. (E and F) corresponding PDOS for B-CN and g-C₃N₄ containing nitrogen vacancies and carboxyl groups, respectively.

As the photoluminescence (PL) spectra illustrated in Fig. 9A, B-CN shows an intense fluorescence signal at around 460 nm under visible-light irradiation and this signal in G-CN is significantly decreased. Because the PL spectra can be interpreted as the radiative recombination of surface trapping states, thus the weakened PL intensity could indicate the enhanced separation of photo-excited electrons and holes.^45,46 By adding H₃PO₄, the peak is further weakened. It is noted that much lower peak intensities are found over H₃PO₂ doped g-C₃N₄ nanosheets and the peak continuously shifts to around 510 nm with the increasing addition of H₃PO₂. The PL peak red-shift in the H₃PO₂ doped g-C₃N₄ nanosheets is associated with the decreased bandgap energy of the samples and consistent with the trend as seen in UV-Vis DRS.⁴⁶ As illustrated in Fig. 9B, all samples exhibit sensitive photocurrent responses during the visible-light on/off irradiation. The photocurrent density values for the four representative samples (B-CN, G-CN, 0.8-P1-CN and 0.8-P2-CN) are ≈2.5, 7.5, 11.1 and 17.0 μA cm⁻², respectively.

Fig. 9 (A) Photoluminescence (PL) spectra and (B) photocurrent curves for B-CN and representative g-C₃N₄ nanosheet materials. (C) Time-resolved photoluminescence decay spectra for representative g-C₃N₄ samples.

To investigate the dynamic electron immigration process, time-resolved fluorescence spectra were collected. As seen in Fig. 9C, the lifetime of the photo-induced charge carriers in the H₃PO₂ doped g-C₃N₄ nanosheets were significantly prolonged compared with B-CN, G-CN and H₃PO₄ doped nanosheets sample, revealing the accelerated charge transfer performance.⁴⁶

Both photoluminescence and photocurrent tests indicate that the g-C₃N₄ nanosheets possess much better performance in the critical process of photo-excited charge carrier separation and visible-light response than the corresponding directly polymerized B-CN. Regarding G-CN and H₃PO₄ doped g-C₃N₄ nanosheets, the enhanced separation of charge carriers could be attributed to their large specific surface areas and protonation.^7,14 Besides, the interstitially doped phosphorus atoms in the H₃PO₄ doped g-C₃N₄ nanosheets would likely facilitate optimization of the π-conjugated heptazine rings to improve the carrier mobility and offer a new channel for carrier migration,^18,32 thereby offering enhanced charge carriers separation. Besides the interstitial phosphorus doping, the nitrogen vacancies in the H₃PO₂ doped g-C₃N₄ nanosheets would also induce unpaired sp²-carbon atoms within the π-conjugated heptazine rings. These electron defected sp²-carbon atoms and newly introduced electron-withdrawing groups (cyano and carboxyl groups) can redistribute the π-electrons and result in improved visible light absorption and photo-excited charge carriers separation.^4,47–49 Thus, a confluence of the synergistic effect of protonation, interstitial phosphorus doping, nitrogen vacancies and newly introduced functional groups in the H₃PO₂ doped g-C₃N₄ nanosheets was achieved and expected to offer a comprehensive enhancement of visible-light photocatalysis. We now turn to these experiments.

Photocatalytic H₂ evolution performances

The photocatalytic performances of the g-C₃N₄ samples were evaluated by measuring the visible-light (λ ≥ 400 nm) H₂ evolution in 20 vol% triethanolamine (TEOA) aqueous solution with 1.5 wt% of Pt loading. As shown in Fig. 10A, the g-C₃N₄ nanosheets exhibited much higher photocatalytic activities than the standard reference B-CN. The H₂ generation rate of G-CN reached 99.1 μmol h⁻¹, whilst only 41.6 μmol h⁻¹ was obtained over B-CN. Importantly, the H₂ evolution rate of g-C₃N₄ nanosheets was significantly improved with the doping of H₃PO₄ and H₃PO₂, respectively. The H₃PO₄ doped g-C₃N₄ nanosheets exhibited a H₂ evolution rate of 144.2 μmol h⁻¹ by the 1.6-P1-CN sample. Interestingly, 0.8-P1-CN and 3.2-P1-CN gave a very similar H₂ evolution rate as 1.6-P1-CN, which could be ascribed to that the limited visible light absorption of the H₃PO₄ protonated g-C₃N₄ nanosheets confines the highest H₂ evolution rate when a certain level of H₃PO₄ addition was reached. For the H₃PO₂ doped nanosheets, 0.8-P2-CN showed the highest H₂ evolution rate of 255.3 μmol h⁻¹ which was 6.14, 2.58 and 1.77 times of that over B-CN, G-CN and 1.6-P1-CN, respectively. The bandgap structures determined by the valence band XPS spectra and UV-Vis DRS results indicate that the narrowed bandgap of the H₃PO₂ doped g-C₃N₄ nanosheets originates from the conduction band decrease (ESI, Fig. S7 and Table S2†). Thus, the excessively lowered reduction driving force for H₂ evolution may result in lower H₂ evolution activity even though the visible light harvesting ability of g-C₃N₄ was progressively intensified by increasing H₃PO₂ usage. These findings could explain why the highest H₂ generation performance was achieved on the Pt loaded 0.8-P2-CN.

Fig. 10 (A) Rates and time-on-stream of the photocatalytic H₂ evolution over 1.5 wt% Pt-loaded g-C₃N₄ samples in 20 vol% TEOA solution under visible-light irradiation (λ ≥ 400 nm). The g-C₃N₄ usage in each experiment was 15 mg. (B) H₂ evolution rates over the 1.5 wt% Pt-loaded B-CN, G-CN, 1.6-P1-CN and 0.8-P2-CN in the 20 vol% TEOA solution with optimum addition of K₂HPO₄ (left). Time course of the long-term H₂ evolution over the 0.8-P2-CN nanosheets loaded with 1.5 wt% Pt in 20 vol% TEOA and the optimum TEOA/K₂HPO₄ mixture solution (right).

Because our RTK-Solar H₂ evolution system was used for the first time for photocatalytic H₂ evolution tests, its accuracy was verified in two steps. First, the evolved gas was confirmed to be only H₂ with the help of Hiden HPR 20 gas chromatograph/mass spectrometer system (ESI, Fig. S8†). Second, H₂ evolution experiments over 1.5 wt% Pt loaded B-CN, G-CN, 1.6-P1-CN and 0.8-P2-CN in 20 vol% TEOA solution were further conducted under the same conditions as those in the RTK-Solar system. With the quartz reactor being connected to a gas circulation system, the generated H₂ amount was quantified by a calibrated gas chromatography equipped TCD detector (Clarus 580, PerkinElmer, helium as carrier gas) for every 1 hour. The gas chromatography determined H₂ evolution rates over the 1.5 wt% Pt loaded B-CN, G-CN, 1.6-P1-CN and 0.8-P2-CN were 42.8, 106.6, 140.1 and 261.6 μmol h⁻¹, respectively (ESI, Fig. S9 and Table S3†). The ratios of the H₂ evolution rates determined by gas chromatography method and the RTK-Solar system over the four representative g-C₃N₄ samples were located in a very narrow range of 0.972 to 1.076, confirming the reliability of RTK-Solar system for measuring the amount of generated H₂. The photocatalytic H₂ evolution results reveal the significantly enhanced g-C₃N₄ photocatalytic performance over the nitrogen deficient and protonated g-C₃N₄ nanosheets produced by this facile in situ reductive synthesis strategy.

As proposed by Ye,¹ using the nature-inspired strategy of adding K₂HPO₄ to TEOA solution, the H₂ evolution performance of g-C₃N₄ could be significantly enhanced. In Ye's important work,¹ it was shown that the added HPO₄²⁻ can function both as a proton-pump in natural photosynthesis (to facilitate proton transport in the reaction solution) and also act as a mediator to give rise to a new proton-reduction pathway. The use of HPO₄²⁻ instead of H₂O provides the necessary protons to react with photo-generated electrons to produce H₂ and PO₄³⁻ on the surface of Pt co-catalyst at the very beginning. Following the H₂ evolution from HPO₄²⁻, the resulting PO₄³⁻ entity immediately combined with H⁺ from H₂O to regenerate HPO₄²⁻ and finally complete the proton-reduction cycle.

In other words, the added K₂HPO₄ would not be consumed during the photocatalytic H₂ evolution. In addition, K₂HPO₄ would also promote the oxidation of TEOA. The synergy of enhanced proton reduction and improved photooxidation of TEOA boosted the visible-light photocatalytic H₂ evolution over the Pt loaded g-C₃N₄ catalysts. Thus, enhanced H₂ evolution over the Pt loaded 0.8-P2-CN could also be expected by adding K₂HPO₄ to the TEOA solution.

As illustrated in Fig. 10B, with the optimum addition of K₂HPO₄ (32 mmol) to the 20 vol% TEOA solution, the H₂ evolution of the Pt loaded 0.8-P2-CN was drastically enhanced to a rate of 881.7 μmol h⁻¹, and impressive H₂ generation could be observed (ESI, Video S1†). Similarly, the H₂ generation rates of the Pt loaded B-CN, G-CN and 1.6-P1-CN were also promoted by the optimum TEOA/K₂HPO₄ solution, but the 0.8-P2-CN with 1.5 wt% Pt loading showed the highest increase among others. Though, G-CN and 1.6-P1-CN nanosheets possess well improved charge carrier separation abilities through protonation and phosphorus interstitial doping, and their H₂ evolution rates could greatly enhanced by adding K₂HPO₄ to facilitate both proton reduction and the photooxidation of TEOA. However, their limited visible light absorption abilities with a wide bandgap of around 2.7 eV will confine the further improvement of visible-light H₂ evolution.

In contrast, the 0.8-P2-CN nanosheets could harvest much more visible light to generate photo-excited electrons and holes with a significantly narrowed bandgap of 2.42 eV (ESI, Fig. S7†). Thus, together with the enhanced charge carrier separation, proton reduction and photooxidation of TEOA, the significantly intensified visible-light absorption ability could help the Pt loaded 0.8-P2-CN nanosheets to boost visible-light H₂ evolution rate to an extremely high level.

In our 25 hours long-term H₂ evolution experiments, it was found that the Pt loaded 0.8-P2-CN generated nearly the same amount of H₂ for each 5 hours period in the TEOA solution, while the H₂ generated in the last 5 hours period could still remain a 76.5% amount of that in the first 5 hours for the optimized TEOA/K₂HPO₄ mixture solution. As stated above, the K₂HPO₄ would not be consumed during the photocatalytic H₂ evolution. The gradual decrease of the long-term H₂ evolution rate in the TEOA/K₂HPO₄ solution should be attributed to the significant consumption of TEOA during the photocatalytic H₂ evolution process. Thus, both the long-term H₂ evolution experiments conducted in TEOA and TEOA/K₂HPO₄ solutions indicated the excellent photocatalytic stability of the H₃PO₂ doped nanosheets. The FTIR, elemental analysis conducted using STEM and XPS narrow scan results of the fresh and used 0.8-P2-CN suggest that there are no obvious changes which may induce adverse effect on the photocatalytic performances, confirming the stability of the as-prepared g-C₃N₄ nanosheets (ESI, Fig. S10–S12†).

In addition, the spent 0.8-P2-CN catalyst still showed observable H₂ generation ability (ESI, Video S2†). The 0.8-P2-CN with 1.5 wt% Pt loading achieved an AQY of 10.7% and 40.4% at 420 nm in the TEOA and TEOA/K₂HPO₄ (ESI, Fig. S13†), respectively, indicating the extremely efficient visible-light photocatalytic H₂ evolution over the H₃PO₂ doped g-C₃N₄ nanosheets compared with the previously reported work (Table S4†). The AQY measurement results at other wavelengths including 400, 450, 500, 550 and 600 nm were also supplemented in the ESI (Fig. S14 and Table S5†).

Conclusions

Nitrogen deficient and protonated g-C₃N₄ nanosheets were successfully synthesized through a one-step, in situ reduction process combining both a NH₄Cl-assisted strategy and H₃PO₂ doping. It has been demonstrated that the protonated g-C₃N₄ nanosheets have superb photocatalytic activities and can be robustly fabricated by using NH₄Cl as a dynamic gas template. Compared with the g-C₃N₄ nanosheets which were simply further protonated by interstitial H₃PO₄ doping, the electronic bandgaps of the nitrogen deficient g-C₃N₄ nanosheets can be easily controlled by adjusting the added H₃PO₂. The enhanced protonation and reduced electronic bandgaps of H₃PO₂ doped g-C₃N₄ nanosheets can not only readily harvest visible light but also serve to separate the excited state charge carriers more effectively, thereby offering extremely efficient H₂ production under visible-light irradiation.

With the addition of K₂HPO₄ to TEOA solution, the H₂ evolution rate over the 1.5 wt% Pt loaded H₃PO₂ doped g-C₃N₄ nanosheets could be boosted to 881.7 μmol h⁻¹ with an apparent quantum yield of 40.4% at 420 nm. In addition, this comprehensive investigation on different phosphorus compounds for g-C₃N₄ nanosheets modification indicates that H₃PO₂ is a promising dopant to synergistically modify the g-C₃N₄ photocatalysts in a single step. We believe this work will provide guidance for the facile synthesis of nitrogen defective and protonated g-C₃N₄-based materials for further synergistic enhancements of visible-light photocatalysts performance.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

The authors acknowledge the financial support from State Key Laboratory of Chemical Engineering (Tianjin University) and EPSRC. We appreciate the valuable suggestions provided by Dr Xiangyu Jie (Chemistry department, University of Oxford) for manuscript preparation. Weisong Li and Litong Jiang also gratefully thank the China Scholarship Council (CSC) for scholarships.

Notes and references

1. G. Liu, T. Wang, H. Zhang, X. Meng, D. Hao, K. Chang, P. Li, T. Kako and J. Ye, Angew. Chem., 2015, 127, 13561.
2. H. Yu, R. Shi, Y. Zhao, T. Bian, Y. Zhao, C. Zhou, G. I. N. Waterhouse, L. Wu, C. Tung and T. Zhang, Adv. Mater., 2017, 29, 1605148.
3. L. Ge, C. Han and J. Liu, Appl. Catal., B, 2011, 108, 100.
4. X. Liu, P. Wang, H. Zhai, Q. Zhang, B. Huang, Z. Wang, Y. Liu, Y. Dai, X. Qin and X. Zhang, Appl. Catal., B, 2018, 232, 521.
5. G. Zhang, S. Zang, L. Lin, Z. A. Lan, G. Li and X. Wang, ACS Appl. Mater. Interfaces, 2016, 8, 2287.
6. P. Niu, L. Zhang, G. Liu and H. M. Cheng, Adv. Funct. Mater., 2012, 22, 4763.
7. X. Lu, K. Xu, P. Chen, K. Jia, S. Liu and C. Wu, J. Mater. Chem. A, 2014, 2, 18924.
8. C. Li, S. Yu, H. Dong, C. Liu, H. Wu, H. Che and G. Chen, Appl. Catal., B, 2018, 238, 284–293.
9. C. Li, Y. Du, D. Wang, S. Yin, W. Tu, Z. Chen, M. Kraft, G. Chen and R. Xu, Adv. Funct. Mater., 2017, 27, 1604328.
10. C. Li, Y. Xu, W. Tu, G. Chen and R. Xu, Green Chem., 2017, 19, 882–899.
11. Z. Li, C. Kong and G. Lu, J. Phys. Chem. C, 2015, 120, 56.
12. Y. P. Zhu, T. Z. Ren and Z. Y. Yuan, ACS Appl. Mater. Interfaces, 2015, 7, 16850.
13. P. Niu, G. Liu and H. M. Cheng, J. Phys. Chem. C, 2012, 116, 11013.
14. Y. Zhang, A. Thomas, M. Antonietti and X. Wang, J. Am. Chem. Soc., 2008, 131, 50–51.
15. M. Wu, T. Ding, J. Cai, Y. Wang, H. Xian, H. Zhang, Y. Tian, T. Zhang and X. Li, ACS Sustainable Chem. Eng., 2018, 6, 8167.
16. P. Niu, L. C. Yin, Y. Q. Yang, G. Liu and H. M. Cheng, Adv. Mater., 2014, 26, 8046.
17. Z. Hong, B. Shen, Y. Chen, B. Lin and B. Gao, J. Mater. Chem. A, 2013, 1, 11754.
18. L. Shi, K. Chang, H. Zhang, X. Hai, L. Yang, T. Wang and J. Ye, Small, 2016, 12, 4431.
19. Y. Zhang, T. Mori, J. Ye and M. Antonietti, J. Am. Chem. Soc., 2010, 132, 6294.
20. J. Yuan, Q. Gao, X. Li, Y. Liu, Y. Fang, S. Yang, F. Peng and X. Zhou, RSC Adv., 2014, 4, 52332.
21. M. Groenewolt and M. Antonietti, Adv. Mater., 2005, 17, 1789.
22. Y. Cui, Z. Ding, X. Fu and X. Wang, Angew. Chem., 2012, 124, 11984.
23. W. Lei, D. Portehault, R. Dimova and M. Antonietti, J. Am. Chem. Soc., 2011, 133, 7121.
24. J. Fu, B. Zhu, C. Jiang, B. Cheng, W. You and J. Yu, Small, 2017, 13, 1603938.
25. G. Dong, Z. Ai and L. Zhang, RSC Adv., 2014, 4, 5553.
26. J. Sehnert, K. Baerwinkel and J. Senker, J. Phys. Chem. B, 2007, 111, 10671.
27. B. Jürgens, E. Irran, J. Senker, P. Kroll, H. Müller and S. Wolfgan, J. Am. Chem. Soc., 2003, 125, 10288.
28. V. W. H. Lau, I. Moudrakovski, T. Botari, S. Weinberger, M. B. Mesch, V. Duppel, J. Senker, V. Blum and B. V. Lotsch, Nat. Commun., 2016, 7, 12165.
29. Y. Fu, J. Zhu, C. Hu, X. Wu and X. Wang, Nanoscale, 2014, 6, 12555.
30. S. J. Makowski, D. Gunzelmann, J. Senker and W. Schnick, Z. Anorg. Allg. Chem., 2009, 635, 2434.
31. M. Zhu, S. Kim, L. Mao, M. Fujitsuka, J. Zhang, X. Wang and T. Majima, J. Am. Chem. Soc., 2017, 139, 13234.
32. X. Ma, Y. Lv, J. Xu, Y. Liu, R. Zhang and Y. Zhu, J. Phys. Chem. C, 2012, 116, 23485.
33. G. R. Goward, M. F. Schuster, D. Sebastiani, I. Schnell and H. W. Spiess, J. Phys. Chem. B, 2002, 106, 9322.
34. T. Emmler, S. Gieschler, H. H. Limbach and G. Buntkowsky, J. Mol. Struct., 2004, 700, 29.
35. Y. Chen, B. Lin, H. Wang, Y. Yang, H. Zhu, W. Yu and J. M. Basset, Chem. Eng. J., 2016, 286, 339.
36. C. Ye, J. X. Li, Z. J. Li, X. B. Li, X. B. Fan, L. P. Zhang, B. Chen, C. H. Tung and L. Z. Wu, ACS Catal., 2015, 5, 6973–6979.
37. M. Rong, Z. Cai, L. Xie, C. Lin, X. Song, F. Luo, Y. Wang and X. Chen, Chem.–Eur. J., 2016, 22, 9387.
38. H. Gao, S. Yan, J. Wang, Y. A. Huang, P. Wang, Z. Li and Z. Zou, Phys. Chem. Chem. Phys., 2013, 15, 18077.
39. K. B. Yatsimirskii, V. V. Nemoskalenko, V. G. Aleshin, Y. I. Bratushko and E. P. Moiseenko, Chem. Phys. Lett., 1977, 52, 481.
40. M. A. Mohamed, M. F. M. Zain, L. J. Minggu, M. B. Kassim, N. A. S. Amin, W. N. W. Salleh, M. N. I. Salehmin, M. F. M. Nasir and Z. A. M. Hir, Appl. Catal., B, 2018, 236, 265.
41. J. Liu, T. Zhang, Z. Wang, G. Dawson and W. Chen, J. Mater. Chem., 2011, 21, 14398.
42. G. Zhang, M. Zhang, X. Ye, X. Qiu, S. Lin and X. Wang, Adv. Mater., 2014, 26, 805.
43. W. Tu, Y. Xu, J. Wang, B. Zhang, T. Zhou, S. Yin, S. Wu, C. Li, Y. Huang, Y. Zhou, Z. Zou, J. Robertson, M. Kraft and R. Xu, ACS Sustainable Chem. Eng., 2017, 5, 7260.
44. G. Dong, K. Zhao and L. Zhang, Chem. Commun., 2012, 49, 6178.
45. J. Qin and H. Zeng, Appl. Catal., B, 2017, 209, 161–173.
46. F. Wang, Y. Wang, Y. Feng, Y. Zeng, Z. Xie, Q. Zhang, Y. Su, P. Chen, Y. Liu, K. Yao, W. Lv and G. Liu, Appl. Catal., B, 2018, 221, 510–520.
47. Y. Zhao, M. Shalom and M. Antonietti, Appl. Catal., B, 2017, 207, 311.
48. J. Zhang, M. Zhang, R. Sun and X. Wang, Angew. Chem., Int. Ed., 2012, 51, 10145.
49. G. Liu, G. Zhao, W. Zhou, Y. Liu, H. Pang, H. Zhang, D. Hao, X. Meng, P. Li, T. Kako and J. Ye, Adv. Funct. Mater., 2016, 26, 6822.

---

Footnote

† Electronic supplementary information (ESI) available: Sample characterization, H₂ evolution equipment, supplementary XRD, XPS and elemental mapping data, photocatalytic H₂ evolution experimental results. See DOI: 10.1039/c9sc05060d

---