Report

# Metal-free efficient photocatalyst for stable visible water splitting via a two-electron pathway

See allHide authors and affiliations

Science  27 Feb 2015:
Vol. 347, Issue 6225, pp. 970-974
DOI: 10.1126/science.aaa3145

## An enduring catalyst built from carbon

Splitting water into its constituent elements, hydrogen and oxygen, generally requires the assistance of metal catalysts. Liu et al. now show that a metal-free hybrid material composed of carbon and nitrogen can promote this reaction all on its own, with the help of some visible light. The photocatalyst combines one material (C3N4) known to split water into hydrogen and peroxide with a second material (CDot) that breaks the peroxide down before it can damage the first. The robust stability of this hybrid bodes well for practical implementation of optimized analogs in solar energy storage schemes.

Science, this issue p. 970

## Abstract

The use of solar energy to produce molecular hydrogen and oxygen (H2 and O2) from overall water splitting is a promising means of renewable energy storage. In the past 40 years, various inorganic and organic systems have been developed as photocatalysts for water splitting driven by visible light. These photocatalysts, however, still suffer from low quantum efficiency and/or poor stability. We report the design and fabrication of a metal-free carbon nanodot–carbon nitride (C3N4) nanocomposite and demonstrate its impressive performance for photocatalytic solar water splitting. We measured quantum efficiencies of 16% for wavelength λ = 420 ± 20 nanometers, 6.29% for λ = 580 ± 15 nanometers, and 4.42% for λ = 600 ± 10 nanometers, and determined an overall solar energy conversion efficiency of 2.0%. The catalyst comprises low-cost, Earth-abundant, environmentally friendly materials and shows excellent stability.

Production of hydrogen and oxygen from water is a promising means of storing solar energy in a way that compensates for the intermittency of sunlight as a primary source of power (1, 2). It can be realized by applying a hybrid system in which a solar cell powers an electrolyzer [photovoltaic (PV) electrolysis]. Photoelectrolysis (PE) uses photocatalyst electrodes with additional electrical power provided by a photovoltaic element. Photocatalysis (PC) involves light-irradiated catalysts (typically catalyst powders suspended in water) for water splitting (3). Recently reported “solar-to-hydrogen” (STH) efficiencies for PV electrolysis systems exceed 10% (46). State-of-the-art PE systems yield STH values of 2 to 3% (7) but are believed to provide a cheaper solution for H2 production. PC is the simplest water-splitting approach, more amenable to cheap, large-scale applications of H2 generation. Unfortunately, despite intense efforts during the past 40 years (815), current direct photocatalysts for water splitting still face several challenging issues: (i) Currently reported catalysts suffer from low quantum efficiency (QE) in the visible range, with STH efficiencies less than 0.1% (1618); (ii) many photocatalysts are made of rare and expensive materials; (iii) various photocatalysts show poor stability [e.g., inorganic sulfide and (oxy)nitride-based photocatalysts are less stable and more susceptible to oxidation than water]; (iv) O2 release from semiconductor catalysts is difficult, so that the use of sacrificial reagents is required; (v) the overall four-electron water oxidation to O2 encounters a large overpotential; and (vi) the kinetically competing two-electron reaction to H2O2 often poisons the photocatalysts (19).

Overall water splitting to H2 and O2 requires a high free energy of 113.38 kcal/mol (20, 21). The challenge lies mainly in the release of diatomic O2, which involves four electron and four proton transfers for the eventual formation of an O-O bond. The concerted four-electron process for oxygen evolution (1.23 eV) is thermodynamically more favorable than the two-electron process for H2O2 formation (1.78 eV). However, detailed analysis (see supplementary text) shows that a higher reaction rate may be achieved in a system where water is first oxidized via a two-electron reaction to H2O2 and H2, followed by H2O2 decomposition to O2 and H2O. For this stepwise two-electron/two-electron water splitting to H2 and O2 to be viable and practical, the photocatalyst applied should be capable of promoting generation as well as subsequent decomposition of H2O2 with high efficiencies and low overpotentials, so as to allow considerable reduction in the energy cost for production of H2 and O2 via overall water splitting. Here, we report that nanocomposites of carbon nanodots embedded in a C3N4 matrix perform as an excellent photocatalyst fulfilling the above requirements.

C3N4 is commercially available (e.g., from Carbodeon) and can be easily fabricated (e.g., from urea) (19). It is an Earth-abundant and low-cost photocatalyst capable of generating H2 and H2O2 from water even in the absence of catalytic metals, albeit with a low QE (19, 2226). C3N4 belongs to the oldest reported polymer materials prepared by chemists (by Berzelius in ~1830) and first termed “Melon” (27). In 2006 it was determined that the visible light activity of TiO2 after treatment with urea was due to “Melon” (28). In 2009, Antonietti and colleagues described in detail the optical properties, electronic structure, and photocatalytic activity of C3N4 (29). Following this work, many groups attempted to optimize the catalytic properties of C3N4, motivated by its relatively low band gap energy Eg of 2.7 eV, and high valence band and conduction band positions (29) [1.8 and –0.9 eV versus reversible hydrogen electrode (RHE)]. Many heterojunction composites with oxide semiconductors as well as photocatalyst systems were investigated. The latter included systems with a variety of oxides (30) and sulfides (31) along with pure metals (19) and even graphene (32) and carbon nanotubes (33). Different preparation methods were studied in an effort to increase the surface area of C3N4 and to improve its catalytic activity. The QE values obtained using C3N4 as a catalyst for water splitting to H2 and O2 have been low (maximum 3.75% at 420 nm and less than 0.5% for 500 nm), and generally the use of a sacrificial reagent has been necessary (19, 24, 26, 3032, 34). The efficiency at 700 nm can be largely increased by applying dye molecules, but again a sacrificial reagent is a must (35). During water splitting C3N4 suffers from poisoning by the produced H2O2, which is difficult to remove from the C3N4 surface (19). Different methods including stirring, bubbling, and/or addition of chemical agents have been attempted for regeneration of the poisoned C3N4 catalyst (19).

Carbon nanodots (CDots; monodisperse graphite particles less than 10 nm in diameter) exhibit unique photo-induced electron transfer, photoluminescence, and electron reservoir properties (36). In particular, CDots possess high catalytic activity (by chemical catalysis; no light is needed) for H2O2 decomposition (37). Given the photocatalytic properties of CDots and C3N4, we hypothesized that a combination of CDots and C3N4 could constitute a high-performance composite photocatalyst for water splitting via the stepwise two-electron/two-electron process: (i) 2H2O → H2O2 + H2; (ii) 2H2O2 → 2H2O + O2.

CDots were synthesized by a typical electrochemical method followed by hydrothermal treatment with ammonia (37). CDots-C3N4 composites were then prepared by heating a mixture of ammonia-treated CDots and urea powder at 550°C for 3 hours (see supplementary material). Characterization of the as-produced CDots-C3N4 composites by transmission electron microscopy (TEM) showed highly porous grains (Fig. 1A) consisting of CDots (2 to 10 nm in diameter) embedded in a porous C3N4 matrix (Fig. 1B). The CDots were nonuniformly distributed, with apparent regions of dots denser by one order of magnitude than the average concentration in the C3N4 matrix. A high-resolution TEM (HRTEM) image of a CDot crystallite (Fig. 1C) showed an interplanar spacing of 0.202 nm, which corresponds to the 〈101〉 spacing of graphitic carbon. The corresponding 2D fast Fourier transform (FFT) pattern (Fig. 1D) exhibits the hexagonal crystalline structure of the CDots. Linewidth analysis of x-ray powder diffraction patterns (fig. S1) of the grains of the CDots-C3N4 composite suggests that the C3N4 matrix comprises nanocrystallites with an average diameter of ~4 nm (29, 38, 39). The diameter of the grains of the CDots-C3N4 composite deduced from atomic force microscopy (fig. S2) ranges between 90 and 400 nm (fig. S3).

The incorporation of CDots into the C3N4 matrix leads to an increase in the ultraviolet-visible (UV-vis) absorption over the entire wavelength range investigated (Fig. 2A). The optical band gap of a semiconductor can be estimated from the Tauc plot [i.e., the curve of converted (αhν)r versus hν from the UV-vis spectrum, in which α, h, and ν are the absorption coefficient, Planck constant, and light frequency, respectively, and r = 2 for a direct band gap material and r = 1/2 for an indirect band gap material]. Figure 2B shows a good linear fit when using r = 2, in accord with previous work (25) claiming C3N4 to be a direct band gap material (no good linear fit is obtained for r = 1/2). The Eg value of CDots-C3N4 (CDots concentration of 1.6 × 10−5 gCDots/gcatalyst) was thus determined to be 2.77 eV by measuring the x-axis intercept of an extrapolated line from the linear regime of the curve (Fig. 2B, red curve), which is almost identical to that of pure C3N4 (Fig. 2B, 2.75 eV, black curve) within experimental error. The Tauc plot curve (Fig. 2B, red curve) of CDots-C3N4 shows an apparent tail between 2.0 and 2.7 eV, which is helpful for improving the light absorbance and the photocatalytic efficiency. Aside from an appropriate band gap, the proper matching of conduction band and valence band levels of a photocatalyst with the redox potentials of the photocatalytic reactions is also important for water splitting. We used ultraviolet photoelectron spectroscopy (UPS) to determine the ionization potential [equivalent to the valence band energy (Ev)] of CDots-C3N4, which was calculated to be 6.96 eV by subtracting the width of the He I UPS spectra (Fig. 2C) from the excitation energy (21.22 eV). The conduction band energy Ec is thus estimated at 4.19 eV from EvEg. The Eg, Ev, and Ec values of CDots-C3N4 in electron volts are converted to electrochemical energy potentials in volts according to the reference standard for which 0 V versus RHE (reversible hydrogen electrode) equals –4.44 eV versus evac (vacuum level). Figure 2D further shows that the reduction level for H2 is positioned below the conduction band of CDots-C3N4, and the oxidation level for H2O to H2O2 or O2 is above the valence band; these bands are properly positioned to permit transfer of electrons and holes, respectively, for water splitting, thus corroborating the potential of CDots-C3N4 as a photocatalyst for overall water splitting. Additional characterization of the CDots-C3N4 composites included Raman spectroscopy (fig. S4), Fourier transform infrared spectroscopy (FTIR, fig. S5), energy-dispersive spectra (EDS, fig. S6), x-ray photoelectron spectroscopy (XPS, fig. S7), and x-ray absorption near edge structure (XANES, fig. S8).

Figure 3A shows the evolution of H2 and O2 from 150 ml of water containing 0.08 g of nonoptimized CDots-C3N4 composite under visible light irradiation. H2 and O2 were both quantified by gas chromatography (GC); a typical sample curve (GC signal) is shown in fig. S9. H2 and O2 evolution proceeded continuously in a molar ratio of H2/O2 of 2.02, effectively identical to the theoretical value of 2 for overall water splitting, but ceased immediately when the light was turned off. The constant H2 evolution rate was ~8.4 μmol/hour and that of O2 ~4.1 μmol/hour, and no other gases than H2 and O2 (e.g., CO2 or N2) were detected by GC. Control experiments showed that no O2 evolution was detected by gas chromatography when pure CDots, pure C3N4, or a macroscopic mixture of the two was used as photocatalyst over a 24-hour reaction period. This indicates (see supplementary material) that proximity between the CDots and the generation sites of H2O2 (achieved in the composite CDots-C3N4 structure) is necessary for the CDots to decompose H2O2 and generate O2.

We further verified that the detected O2 was indeed generated by water splitting. When we used H218O as reagent under the same water-splitting conditions described earlier, 18O2 (mass 36) was the only product detected by GC-MS (gas chromatography–mass spectrometry). We next measured the QE of overall water splitting by CDots-C3N4 as a function of the incident light wavelength λ0 (Fig. 3B). QE decreased with increasing wavelength, and the longest wavelength capable of inducing water-splitting coincided with the red absorption edge of the CDots-C3N4 composite, suggesting the reaction proceeds via photoabsorption by the catalyst. We proceeded to optimize the catalyst composition by measuring QE for different concentrations of CDots in a fixed mass of composite (Fig. 3C). With increasing CDots concentration, the QE increased to a maximum value of 16% around 4.8 × 10−3 gCDots/gcatalyst, after which it decreased upon addition of more CDots. Next, we optimized the amount of composite catalyst added to water at the optimum CDots/C3N4 ratio of 4.8 × 10−3 gCDots/gcatalyst (Fig. 3D). We found that the QE increased to a maximum value of 16% as the catalyst was added, and then stayed the same upon further addition.

We interpret the dependence of QE on CDots concentration to an enhancement of the decomposition rate of H2O2 until the rate is sufficient to remove all the generated H2O2 (by the photocatalytic effect of C3N4). Afterward, further increase of the CDots concentration could enhance the light absorption by introducing subband states, thus raising the QE to a maximum, after which addition of more CDots seems to increase energy losses (light absorbed in CDots and not in C3N4, e/h+ recombination, lower catalytic efficiency of C3N4 due to CDots, etc.), thus decreasing the QE. On the other hand, the initial increase (Fig. 3D) of the composite catalyst load (keeping the CDots concentration constant) increases the generation rate of H2 (more catalyst → more light absorbance) until reaching a maximum. Additional increase of the composite CDots-C3N4 load cannot further increase the light absorbance (all incident light is absorbed), so that the QE remains unchanged (Fig. 3D). We believe that the large values of QE obtained by the CDots-C3N4 composite relate to the highly porous structure of the C3N4 grains (resulting from the preparation method of the catalyst), which yields a large water-catalyst interface area [97 m2/g as measured by the Brunauer-Emmett-Teller (BET) method]. Our further studies systematically confirmed that water-splitting photocatalysis by CDots-C3N4 indeed proceeds via the stepwise 2e/2e two-step process, in which H2O oxidation to H2O2 is the first and rate-limiting step, followed by the second and fast step of H2O2 disproportionation to O2, which is chemically catalyzed by CDots (detailed experiments and analysis are given in the supplementary material).

The experiments above focused on catalytic properties at λ = 420 nm; however, solar water splitting in the vicinity of the solar spectrum peak (λ = 550 to 650 nm) region is more relevant to efficient harvesting of solar energy. The QE of the standard CDots-C3N4 catalyst (i.e., CDots concentration of 1.6 × 10−5 gCDots/gcatalyst, catalyst load 0.53 g/liter) at λ = 580 nm was relatively low (~0.3%). Upon optimization, we succeeded in preparing catalysts with a high QE = 16% at λ = 420 ± 20 nm by increasing the quantity of CDots in the C3N4. The higher CDots concentration and the associated larger total C fraction in the composite (Fig. 4A) concurrently increased the absorbance in the solar spectrum peak region (black versus red curve in Fig. 4B). This is most likely due to the effect of C addition leading to the formation of more subband states in the band gap. Figure 4C shows that the H2 generation rate at our standard conditions (80 mg of catalyst in 150 ml of water; 300-W Xe light radiation with a long-pass cutoff filter allowing λ > 420 nm) increased by a factor of 5.4 for CDots-C3N4 with a higher CDots concentration. The QE of the catalyst with the optimum amount of CDots (4.8 × 10−3 gCDots/gcatalyst) remarkably increased (Fig. 4D, black) relative to the QE of the catalyst with a CDots concentration of 1.6 × 10−5 gCDots/gcatalyst (Fig. 4D, red). It reached 6.29% at λ = 580 ± 15 nm (20 times the QE for the 1.6 × 10−5 gCDots/gcatalyst catalyst) and 4.42% at λ = 600 ± 10 nm. For irradiation wavelengths λ > 650 nm, both catalyst systems showed zero QE (Fig. 4D).

The solar energy conversion was evaluated in the following studies by using an AM 1.5G (air mass 1.5 global conditions) solar simulator as the light source (see fig. S10 for ~output spectral distribution) and CDots-C3N4 (4.8 × 10−3 gCDots/gcatalyst) as the catalyst (80 mg catalyst in 150 ml of water). After 6 hours of illumination, the total incident power over the irradiation area of 9 cm2 was 0.63 W, so that the total input energy was 1.36 × 104 J. During the photocatalytic reaction, 1150 μmol of H2 was detected by GC, which indicated that the energy generated by water splitting is EF = 274 J. The STH value of CDots-C3N4 with the higher concentration of CDots was determined to be 2.0%, which is at least one order of magnitude larger than previously reported values (40, 41). Indeed, the STH values were very low (0.3% for 1.6 × 10−5 gCDots/gcatalyst) for low CDots concentration and reached 2% only at the optimum CDots concentration (fig. S11). An independent alternative STH calculation based on the CDots-C3N4 (4.8 × 10−3 gCDots/gcatalyst) QE curve (Fig. 4D, black curve) and the spectral irradiance of the AM 1.5G solar simulator (fig. S10) was also carried out, yielding a STH value of 1.78%, in good agreement (89%) with the direct solar simulator value of 2% (see supplementary material for detailed calculations).

A recent work (40) claimed a STH efficiency of 5.1% for CoO nanoparticles, but the catalyst system was unstable and corroded after 1 hour of operation. In striking contrast, the CDots-C3N4 composite of 1.6 × 10−5 gCDots/gcatalyst exhibited long-term stability of at least 200 days for a catalyst dried and reused 200 times (Fig. 3A), whereas the catalyst with the larger CDots concentration of 4.8 × 10−3 gCDots/gcatalyst has shown no obvious decay of QE after 50 1-day cycles of reuse (fig. S12). The stability of the CDots-C3N4 catalyst system was further studied from several different vantage points: (i) the structural stability of the catalyst over time, (ii) the catalytic functionality (generation rate of H2 and O2) over time, (iii) the mass loss or gain after long-term operation, and (iv) the gases released during operation (to detect possible degradation by-products). These factors were probed for two CDots concentrations (1.6 × 10−5 and 4.8 × 10−3 gCDots/gcatalyst) and two catalyst loads (80 mg/150 ml and 10 mg/150 ml). All photocatalytic water-splitting experiments conducted under different conditions for a variety of reaction times manifested the same H2 and O2 generation rates and QEs, within experimental error (Fig. 3A and figs. S12 to S15). These different experiments included (i) 45 days of continuous operation (figs. S13 and S14), (ii) 200 cycles of 24 hours each (Fig. 3A), (iii) 50 cycles of 24 hours each (fig. S12), and (iv) 15 cycles of 24 hours (fig. S15). Raman, FTIR, and XPS spectra (figs. S16 to S18) of CDots-C3N4 catalysts before and after 50 24-hour cycled reactions show no obvious differences, confirming the structural stability of the CDots-C3N4 catalyst under water-splitting conditions. No CO2 or N2 gas could be detected in the reaction system by GC, which suggests that the CDots-C3N4 catalyst is stable and did not decompose during the photocatalytic process. There was only negligible mass loss or gain in all the experiments in which the catalyst was weighed before and after use (tables S1 and S2).

A recent U.S. Department of Energy (DOE)–solicited technoeconomical analysis of H2 generation by solar water splitting (3) suggested that PC systems with STH = 5% (not far away from the 2% efficiency reported above) would allow a H2 production cost of $2.30/kg, which meets the DOE target of$2 to $4/kg H2. The cheapest PE configuration with STH = 10%, in comparison, allows a H2 production cost of$5.60/kg H2, more than twice the cost of the PC system (although the efficiency of the PE system is twice the efficiency of the PC system). The present PC catalyst thus offers a simpler and cheaper approach to extract H2 from water in large scale. The main disadvantage of PC systems is the generation of a potentially explosive gas mixture of oxygen and hydrogen, which requires separation to ensure safety. Modern industrial technology offers a variety of mature methods for realization of large-scale hydrogen separation and extraction from other gases (nitrogen, argon, oxygen). This point is considered and is part of the above H2 production cost evaluation of the DOE-solicited work. The separation systems evaluated (3) include pressure swing adsorption (PSA), temperature swing adsorption (TSA), palladium membrane separation, nanoporous membrane separation, and electrochemical pumps (42). The active wavelength region of CDots-C3N4, λ < 620 nm, would allow a theoretical STH efficiency of ~15% for sunlight (AM 1.5G), which thus leaves substantial room for technical optimization. Our catalyst is also mildly active for the overall seawater photocatalytic splitting. Using CDots-C3N4 (4.8 × 10−3 gCDots/gcatalyst) in seawater, we obtained QE (420 nm) of 3.86% and STH = 0.45% (fig. S19). This is a preliminary result, and future studies should probe the reason for the reduction of water-splitting efficiency of sea water versus pure water [QE (420 nm) = 16%, STH = 2%].

We have shown that CDots-C3N4 composites can be made of low-cost, environmentally benign materials and can split water into H2 and O2 with QEs of 16% for λ = 420 ± 20 nm and 6.3% for λ = 580 ± 15 nm. The 2.0% STH efficiency obtained is at least one order of magnitude larger than that previously reported for any stable water-splitting photocatalysts (41). It is close to 5% STH, which allows achievement of the DOE price target for H2 generation. In contrast to the conventional one-step four-electron reaction, CDots-C3N4 catalyzes water splitting to hydrogen and oxygen via the stepwise two-electron/two-electron two-step pathway under visible light irradiation. C3N4 is responsible for the first step (photocatalysis), and CDots are responsible for the second step (chemical catalysis). CDots also increase the light absorbance and thus the values of QE and STH. The composite nature of the catalyst provides sufficient proximity between the H2O2 generation sites on the C3N4 surface and the Cdots so that H2O2 decomposition and O2 generation in the second stage become efficient. Moreover, CDots-C3N4 maintains a high rate of hydrogen and oxygen production (for λ > 420 nm) with robust stability in 200 runs of recycling use over 200 days. The results demonstrate CDots-C3N4 as a highly efficient and stable photocatalyst for visible light–driven water splitting.

## Supplementary Materials

www.sciencemag.org/content/347/6225/970/suppl/DC1

Materials and Methods

Figs. S1 to S36

Tables S1 to S3

References (4354)

## References and Notes

1. Acknowledgments: We thank A. Rothschild for helpful discussions. We thank Beijing and Shanghai synchrotron radiation facility for XANES and EXAFS data collection. Supported by Collaborative Innovation Center of Suzhou Nano Science and Technology, the National Basic Research Program of China (973 Program) (2012CB825803, 2013CB932702), the National Natural Science Foundation of China (51422207, 51132006, 21471106), the Specialized Research Fund for the Doctoral Program of Higher Education (20123201110018), a Suzhou Planning Project of Science and Technology (ZXG2012028), and a project funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD).
View Abstract