A Cocrystal Precursor Strategy for Carbon-Rich Graphitic Carbon Nitride toward High-Efficiency Photocatalytic Overall Water Splitting.

Direct and efficient photocatalytic overall water splitting is crucial for the sustainable conversion and storage of renewable solar energy. Herein, we present the design of a carbon-rich graphitic carbon nitride (Cco-C3N4), prepared from a layered molecular cocrystal precursor. The cocrystal microsheets were synthesized using a facile hydrothermal process. Following two-step thermal treatment and liquid exfoliation, the product maintains the 2D morphology owing to the toptactic transformation process. The Cco-C3N4 exhibits an enhanced photogenerated electron-hole separation, high charge transport capacity, and prolonged lifetime of the carriers, relative to the g-C3N4 system. In the absence of any sacrificial reagent or co-catalyst, the Cco-C3N4 microsheets exhibit a high photocatalytic activity. The work presented in this report supplies a cocrystal route for the orderly molecular self-assembly of precursor materials to tailor the chemical compositions and electronic structures. Moreover, the generation of a highly efficient water-splitting photocatalyst has larger implications for sustainable energy applications.

Introduction

Recent efforts toward the direct conversion of solar energy into chemical fuels have been faced with challenges with regard to effective energy conversion and environmental protection (Kudo and Miseki, 2009, Walter et al., 2010, Chen et al., 2010). Of the currently employed green strategies for energy conversion, the use of sunlight to directly split water into H2 and O2 is believed to have the most potential (Tada et al., 2006, Kageshima et al., 2016, Liang et al., 2016). The sunlight-driven water-splitting process sequentially undergoes three steps: (1) light harvesting, (2) separation and transport of the photogenerated carrier, and (3) surface redox (Kudo and Miseki, 2009, Xiang et al., 2012). Consequently, efficient photocatalysts usually feature favorable band gap energies and minimal recombination of the photocarriers (Shown et al., 2018, Asahi et al., 2001, Yamada et al., 2014). However, many of the current photocatalytic systems require the use of sacrificial reagents and co-catalysts to bolster catalytic performance, presenting a liability in the case wherein the photogenerated electrons or holes react preferentially with the sacrificial reagents or co-catalysts (Yao et al., 2016). Therefore it is highly desirable to develop suitable photocatalysts, which are capable of facilitating the overall water splitting in the absence of sacrificial reagents and co-catalysts (Zhang et al., 2016).

Although several well-developed inorganic semiconductors (such as TiO2, Ni et al., 2007; BiVO4, Kim and Choi, 2014; and Fe2O3, Sivula et al., 2011) have drawn great interest in the field of photocatalysis, the efficiency and stability of these materials is still far from the requirement of overall water splitting. Efforts to develop robust visible-light-active photocatalysts identified a polymeric semiconductor derived from small molecules (such as melamine and urea), namely, graphitic carbon nitride (g-C3N4), which has elicited ripples of excitement in the research communities as a promising next-generation metal-free photocatalyst, owing to its facile synthesis, favorable electronic structure, high physicochemical stability, and earth-abundant nature (Cao et al., 2015, Zhang et al., 2011, Wang et al., 2009, Liu et al., 2015). The unique properties of metal-free g-C3N4 allow for its use in a diverse range of applications, including photocatalytic hydrogen evolution (Duan et al., 2015), the oxygen reduction reaction (Yu et al., 2016), and the oxygen evolution reaction (Arif et al., 2018). However, the photocatalytic performance of g-C3N4 has been restricted by its low-wavelength absorption, fast recombination of carriers, and weak electronic conductivity (Dai et al., 2014, Niu et al., 2012, Che et al., 2017). Strategies that have been employed to address these challenges include the introduction of mesopores (Hao et al., 2016), engineering of micro- or nanostructure (Guo et al., 2016), hybridization with semiconductor (Liao et al., 2012), deposition by noble metal (Zhu et al., 2016), and doping with metals or non-metals (Liu et al., 2010). Moreover, stitching together two-dimensional (2D) domains featuring similar aromatic structures (e.g., a graphitic carbon ring unit) via continuous π-conjugation provides a method for tuning the band structure and electronic transport properties and results in unimpeded in-plane electron-hole separation and charge transfer processes (Che et al., 2017, Liao et al., 2012, Du et al., 2012, Gong et al., 2014, Huang et al., 2015). Despite great advancements in the g-C3N4-based photocatalysts, the tailoring of the π-conjugated sections in-plane in carbon nitride for achieving minimized recombination and the ordered movement of photocarriers to suitable redox sites remains elusive.

The term cocrystal usually refers to a molecular solid composed of at least two types of molecular units, organized by non-covalent interactions (Etter et al., 1989). We hypothesized that polymerization of a cocrystal precursor could be employed to generate carbon nitride materials featuring desirable molecular sections and suitable C/N ratios, aided by the modular design and facile preparation of the cocrystallization process (Fang et al., 2017, Yan and Evans, 2014). Moreover, the toptactic conversion of the orderly molecular assembly from the cocrystal to carbon nitride may result in the high dispersion of catalytic sites at the molecular level requisite for high-efficiency water-splitting applications. In this work, we selected terephthalic acid (TPA) and melamine (MA) to build cocrystals, as the hydrogen-bonding interaction between –COOH functionality in TPA and the –NH2 group in MA facilitates the formation of 2D molecular sheets (Figure 1). Furthermore, the capacity of aromatic TPA for π-conjugation is amenable to the introduction of ordered carbon-rich frameworks for effective photogenerated carrier separation and transport. The resulting carbon-rich graphitic carbon nitride yields a high photocatalytic activity with simultaneous production of H2 and O2 at rates up to 530 and 255 μmol·g−1·h−1, respectively, which exceeds those of most reported current metal-free photocatalysts. The work presented herein not only establishes a cocrystal precursor method for extending the diversity of the chemical structure and composition of graphitic carbon nitride materials but also demonstrates their applicability as efficient water-splitting photocatalysts.

Figure 1

The Formation Process of the Pristine g-C3N4 and C-C3N4

Results and Discussion

The cocrystal composed of a 2:1 ratio of MA and TPA was prepared using a hydrothermal method at 140°C. Details of the crystal are summarized in Table S1. The X-ray diffraction (XRD) pattern of the resultant 2MA.TPA sample is consistent with the simulated result, confirming high purity of the cocrystal product (Figure S1A). The crystal structure illustrates that, in a single layer, the electrostatic and hydrogen bonding interactions drive the cooperative assembly of protonated melamine, terephthalic acid, and H2O into the 2D supramolecular pattern. Besides, the molecular sheets orient perpendicularly through π-π interaction and hydrogen bonds (Figures S1B and S1C), with an interlayer spacing of approximately 1 nm. Scanning electron microscopic (SEM) imaging of the surface (Figures 2A and S1D) reveals the presence of typical 2D microsheet morphologies in the 2MA.TPA cocrystal, consistent with the layered structure observed by XRD. Following a two-step thermal treatment, the cocrystal product presents the same XRD pattern and Fourier transform infrared bands with the typical g-C3N4 structural features (Figures S5 and S6A), indicating the successful formation of graphitic carbon nitride (denoted as C-C3N4 hereafter). The color of C-C3N4 is brown, which is different from that of the pristine g-C3N4 (Figure S2). Elemental analysis and energy-dispersive X-ray spectra (Figure S7) measured a C/N ratio of approximately 7:3 for the C-C3N4 product, suggesting its carbon-rich characteristics through the cocrystal precursor method. The C/N ratio is also in agreement with the idealized model derived for the C-C3N4. The SEM images (Figures 2A, 2B, and S3) indicate that the C-C3N4 maintains sheet-like characteristics, suggestive of a topotactic transformation from the initial 2D microsheets of the cocrystal.

Figure 2

Morphology Images for Cocrystal Precursor and C-C3N4

(A and B) SEM images for cocrystal precursor (A) and C-C3N4 (B).

(C–F) High-resolution TEM image for C-C3N4 nanosheets in different scales at the same area.

After grinding the synthesized cocrystal into fine powders and dispersing the resultant particles using ultrasonic waves, homogeneous C-C3N4 microsheets can be observed, as illustrated in the transmission electron microscopic (TEM) image (Figure S4A), which is almost the same as the pristine g-C3N4 (Figure S4B). A high-magnification image reveals that there is no obvious nanoparticle formation on the smooth surface. The high-resolution TEM images (Figures 2C–2F) exhibit considerable apparent lattice fringe regions randomly incorporated throughout the disordered basal plane domain of the g-C3N4 matrix. The corresponding lattice fringes of 0.21 nm are ascribed to the (100) planes of graphite. The observed hexagonal pattern is attributed to the incorporated carbon ring and/or triazine frameworks (Che et al., 2017). The presence of crystal defect sites, created during thermal treatment, is manifest in breaks in the hexagonal pattern.

X-ray photoelectron spectroscopy was used to analyze the composition and chemical environment of both the prepared C-C3N4 and pristine g-C3N4 (Figure 3). In the C 1s spectrum of C-C3N4, two major peaks at 288.17 and 284.82 eV were assigned to the sp2-hybridized carbons in the heterocyclic aromatic ring (N‒C=N) and the graphitic carbon (C‒C), respectively (Hammer et al., 1998). Compared with the pristine g-C3N4, the graphitic carbon (C‒C) units in the spectrum of C-C3N4 have been highly enhanced, and the ratio of triazine units to carbon-ring can be estimated at around 3:1. In the N 1s spectrum of C-C3N4, evidently, the intense and broad peak ranging from 396 to 402.7 eV can be deconvoluted into three peaks, including 398.59 eV for sp2-hybridized nitrogen in triazine rings (C‒N=C), 400.0 eV for tertiary nitrogen N‒(C)3 groups, and 401.21 eV for amino functions carrying hydrogen (C‒N‒H) (Hammer et al., 1998). A small peak observed at 404.66 eV in the N 1s spectrum is attributed to a positive charge localization in the heterojunction (Che et al., 2017). Usually, several peaks can be detected in bulk g-C3N4 (Li et al., 2012). As shown in the Raman spectra (Figure S6B), an obviously wide peak from g-C3N4 is observed at ca. 1,550 cm−1, which indicates that the bulk g-C3N4 has been well exfoliated into thin nanosheets (Jiang et al., 2014). Meanwhile, a similarly wide band can be observed in the C-C3N4, which is also due to the exfoliated nanosheets as observed by TEM above. Moreover, two peaks at 1,343 and 1,590 cm−1 were detected, corresponding to the D and G bands of graphene-based nanomaterials, respectively (Ma et al., 2017). Collectively, these spectra demonstrate the successful formation of the g-C3N4-based heterostructures via the incorporation of carbon ring.

Figure 3

X-ray Photoelectron Spectroscopy for C-C3N4 and g-C3N4

(A and B) C 1s high-resolution X-ray photoelectron spectroscopy (XPS) spectra for C-C3N4 (A) and pristine g-C3N4 (B).

(C and D) N 1s high-resolution XPS spectra for C-C3N4 (C) and pristine g-C3N4 (D).

To test the photocatalytic activity, we evaluated the water-splitting performance of the C-C3N4 microsheets dispersed in pure water in the absence of sacrificial reagents or co-catalysts under standard reaction and measurement conditions. Simultaneous evolution of H2 and O2 gases was detected in a ratio of approximately 2:1 for the C-C3N4, and the corresponding H2 and O2 evolution rates are up to 530 and 255 μmol·g−1·h−1, respectively (Figure 4A). As is shown in Figure S9, the quantum efficiency of C-C3N4 reaches up to 5.28% at 400 nm. As a control experiment, the photocatalytic activity of pristine g-C3N4 was tested under the same experimental condition. In this case, no H2 or O2 evolution was detected, which may be attributed to the gas production below the detection limit.

Figure 4

The Characterization of Overall Water Splitting for C-C3N4

(A) Typical time course of H2 and O2 evolution for C-C3N4.

(B) UV-vis absorption and corresponding Tauc/Davis-Mott plots.

(C) Mott-Schottky plots.

(D) Band structure diagram of C-C3N4 and g-C3N4 polymers calculated by optical absorption and typical electrochemical Mott-Schottky methods.

To gain insight into the superior photocatalytic activity of C-C3N4, the electronic structure of this material was studied. Ultraviolet-visible (UV-vis) light absorption spectra were analyzed to determine the band gap (Figure 4B). The pristine g-C3N4 exhibits an intrinsic absorption edge near 460 nm, corresponding to a calculated band gap of about 2.75 eV according to the Kubelka-Munk function. A similar absorption edge was observed for the C-C3N4 microsheets; however, an extended visible light absorption over 500–800 nm is also detected, which suggests the accessibility of low-energy visible light. The UV-vis spectrum appearing between 480 and 600 nm signifies the existence of an Urbach band tail. This suggests that the integration of carbon-rich units into g-C3N4 has imported a new intermediate energy level, which is consistent with the decreased band gap. Moreover, a good linear fit is obtained using (αhv)2 for Tauc plots, indicating direct band gap semiconductor behavior for C-C3N4. The band gap of C-C3N4 is evaluated as 2.55 eV, corresponding to the extended visible light absorption. The decrease in band gap energy relative to the pristine g-C3N4 is consistent with the introduction of the carbon-rich units in the C-C3N4 matrix. This is because the graphitic structure possesses strong absorption in the visible light region, contributing to the great light-harvesting efficiency of the heterostructure. Consequently, the range of UV-vis absorption has been broadened (Che et al., 2017, Liao et al., 2012, Du et al., 2012). Based on electrochemical Mott-Schottky plots (Figure 4C), the conduction band minimum and valence band maximum of the C-C3N4 can be determined as −1.1 and 1.45 eV versus reversible hydrogen electrode, respectively. The redox potential of C-C3N4 is reductively shifted relative to pristine g-C3N4 (−1 eV), which supports the redox shuttling behavior of C-C3N4 in the water-splitting reaction (Figure 4D).

Photoluminescence (PL) emission is useful to probe the efficiency of electron-hole pair trapping, migration, and transfer in the semiconductor material. Relative to the pristine g-C3N4, which produces an emission at ca. 440 nm (Figure 5A), the C-C3N4 exhibits an obvious red-shifted emission occurring at 468 nm. The observed energy shift of emission is attributed to the enhanced graphitic carbon sections and extended visible absorption as described above. The PL intensity of C-C3N4 is diminished relative to pristine g-C3N4, which confirms a decreased electron-hole recombination rate for C-C3N4. To gain more insight into the information on charge separation ability in the C-C3N4 matrix, the PL decay spectra were performed on both the pristine g-C3N4 and C-C3N4 (Figure S8). The PL lifetimes of pristine g-C3N4 and C-C3N4 are retrieved to be 3.41 and 18.14 ns, respectively. Obviously, C-C3N4 has a longer PL lifetime, suggesting that the formed carbon-rich units and heterostructure interface speed up the photogenerated charge transfer and effectively depresses charge recombination (Yue et al., 2017a, Yue et al., 2017b). Moreover, a comparison of the photocurrent-time curves obtained by subjecting g-C3N4 and C-C3N4 to typical on-off cycles of visible light irradiation shows an enhanced photocurrent response for C-C3N4 (Figure 5B), nearly two times that of the pristine g-C3N4. Notably, the pristine g-C3N4 exhibits an anodic photocurrent spike upon initial irradiation, followed by decay of the photocurrent until a constant current is reached. The observed decrease in photocurrent is attributed to the accumulation of photogenerated holes on the g-C3N4 surface and the competitive recombination of these holes with electrons or reduced species in the electrolyte. Following equilibration of the competitive separation and recombination of electron-hole pairs, a constant photocurrent is achieved. Similar behavior has been observed in a previously reported g-C3N4 system (Ran et al., 2018). In contrast, little to no anodic photocurrent spike is observed for C-C3N4. The alternation of the photoelectron transfer behavior suggests that the C-C3N4 could obviously enhance the separation of photogenerated carriers. We hypothesize that the introduced carbon rings of C-C3N4 act as a highway for electron transfer, allowing the photogenerated electrons to rapidly traverse the material while the photogenerated holes rest on melem units. In short, the mobility of the photoexcited charge carriers is highly promoted.

Figure 5

The Performance Comparison between C-C3N4 and g-C3N4

(A) Photoluminescence emission curves.

(B) Transient photocurrent responses of C-C3N4 and g-C3N4 under visible light irradiation.

(C) Transient open-circuit voltage decay (OCVD) measurement, and inset depicting the average lifetimes of the photogenerated carriers (τ) obtained from the OCVD measurement.

(D) Electrochemical impedance spectra (EIS) and insets showing the equivalent circuit impedance mode.

The transient open-circuit voltage decay measurements (Figure 5C) indicate that the C-C3N4 holds a higher open-circuit voltage and slower photovoltage decay behavior upon light cycling. These properties suggest a higher energy and longer lifetime of the photoexcited carriers. Moreover, the average lifetime of the photocarriers in C-C3N4 is prolonged by nearly three orders of magnitude relative to that of pristine g-C3N4 (the inset of Figure 5C). The prolonged lifetime of the photogenerated carriers, in combination with the enhanced electron mobility and higher work function, enables the C-C3N4 to trap photoexcited electrons and suppress electron-hole recombination more effectively than the pristine g-C3N4. Furthermore, the Nyquist plots for C-C3N4 catalysts illustrate low charge transfer resistance relative to that of g-C3N4 (Figure 5D). The aforementioned results indicate that the electronic structure of C-C3N4 has been greatly modified by grafted carbon rings, leading to efficient photoexcited electron-hole separation and longer lifetime of the photocarriers.

Conclusion

In summary, we have prepared a C-C3N4 heterostructure with uniform introduction of graphitic carbon rings into the typical g-C3N4 using a facile cocrystallization method. The resultant material achieves highly efficient photoexcited electron-hole separation and superior charge transport behavior relative to the pristine g-C3N4. In addition, the C-C3N4 can effectively facilitate the overall water-splitting reaction under light irradiation with the prominent H2 and O2 production rates up to 530 and 255 μmol·g−1·h−1, respectively. It is noted that the photocatalytic water-splitting activity surpasses majority of the currently reported photocatalysts, with hydrogen evolution under the full solar spectrum (Table S2). Moreover, the water-splitting activity is achieved in the absence of sacrificial reagents and co-catalysts, supporting metal-free photocatalysis. The highly efficient and simultaneous production of O2 and H2 is attributed to the enhanced generation and mobility of the photoexcited charge carriers upon introduction of the carbon-rich graphitic carbon units. The extension of the cocrystal precursor method presented in this report for the fabrication of other energy-matched semiconductor systems will have larger implications in the development of materials for applications in solar energy conversion.

Limitations of the Study

More systematic and detailed theoretical calculations on the relationships between the structure and its overall water-splitting performance are required for further study. To achieve practical application of the metal-free materials, considerable work needs to be done.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.