Modified Nanopillar Arrays for Highly Stable and Efficient Photoelectrochemical Water Splitting

Abstract Atomically modified graphitic carbon nitride quantum dots (QDs), characterized by strongly increased reactivity and stability, are developed. These are deposited on arrays of TiO2 nanopillars used as a photoanode for the photoelectrochemical water splitting. This photoanode shows excellent stability, with 111 h of continuous work without any performance loss, which outperforms the best‐reported results by a factor of 10. Remarkably, our photoanode produces hydrogen even at zero bias. The excellent performance is attributed to the enhancement of photoabsorption, as well as to the promotion of charge separation between TiO2 nanopillars and the QDs.

The modified g-C 3 N 4 QDs were synthesized on TiO 2 nanopillars by using the one-pot quasi-chemical vapor deposition (CVD) method [18] (more details in the Supporting Information). While in the processing in a crucible system [18] the amount of dicyandiamide (DCD) was fixed (7 g), four different amounts of barbituric acid (BA) were used: 0, 0.15, 0.3, 0.5 g. This led to the corresponding four different composites called in our simplified notation as follows: CNQD@TiO 2 , CNB 0.15 QD@TiO 2 , CNB 0.3 QD@TiO 2 , and CNB 0.5 QD@TiO 2 , respectively.
The electron microscopic study of the CNQD@TiO 2 is shown in Figure 1. The scanning electron microscopy (SEM) images are shown a) and b), and demonstrate that our TiO 2 arrays consist of highly dense and vertically aligned nanopillars, which provides a large specific surface area for QD adhesion. The nanopillars have cross-sectional dimensions ranging from 100 to 200 nm, and lengths of ≈3 µm. There is no significant morphology difference between modified CNQD@TiO 2 and pristine TiO 2 (see Figure S1, Supporting Information). The transmission electron microscopy (TEM) image shown in Figure 1c reveals that the surface of TiO 2 nanopillars is uniformly decorated with QD, each with diameter of 3-6 nm. The high-resolution TEM (HRTEM) image is shown in Figure 1d, revealing further nanoscopic details of the structure. The inset in this figure shows further magnification (atomic resolution) image of the TiO 2 nanopillar fragment (marked with a square), which shows that the interatomic lattice spacing is 0.250 nm, which corresponds to the (101) planes of a rutile TiO 2 structure.
The photoluminescence (PL) property of modified CNQDs is also characterized and investigated in this work (see Figure S2, Supporting Information). Several modified CNQDs@TiO 2 substrates are immersed in water, and the transparent solution of CNQDs is obtained after sonication for 20 min and filter process. The modified CNQDs show excitation-dependent PL spectra, and the broad peak can be attributed to the abundant N defects in CNQDs, which is similar to the previous report. [19] Figure 2a shows the X-ray diffraction (XRD) patterns from all samples FTO, TiO 2 nanopillars, and the four modified composites. In addition to the eight diffraction peaks of FTO, there are three characteristic peaks from TiO 2 , indexed to typical crystal planes. The lack of obvious diffraction peaks from g-C 3 N 4 is due to insufficient volume and poor crystallinity of the CNQD. The X-ray photoelectron spectroscopy (XPS) signal for C (1s) is shown in Figure 2b, and for N (1s) in c) for modified CNQD@TiO 2 sample. It demonstrates that the modified composites are composed of C, N, Ti, and O elements (see also Figure S3, Supporting Information). Specifically, a high-resolution XPS spectrum of C (1s) in Figure 2b shows that the peak centered at 284.8 eV is exclusively attributed to the accidental contamination with the carbon from XPS instrument. The peak at 286.1 eV is assigned to CO bonds resulted from the pyrolysis of precursors, and the one at 288.0 eV is from carbon atom in the NCN group. The XPS spectrum of the N (1s) shown in Figure 2c region can be fitted with three peaks centered at 398.6, 399.7, and 401.2 eV. These are ascribed to the sp 2 -hybridized nitrogen (CNC), nitrogen in tertiary NC 3 groups, and amino functional groups with a hydrogen atom (CNH), respectively. Compared with pristine TiO 2 , the Ti 2p spectrum of modified CNQDs@TiO 2 sample shows a 0.4 eV shift to lower binding energy (see Figure S4, Supporting Information). Those results confirm the presence of g-C 3 N 4 , and the robust interaction between TiO 2 and CNQDs.
The UV-vis diffuse absorbance spectra displayed in Figure 2d show that the absorption of modified CNBxQDs@TiO 2 gradually shifts toward longer wavelength with the increasing BA contents. As compared with pristine TiO 2 , CNB 0.5 QDs@TiO 2 shows a broadest absorption edge of about 440 nm. That can be ascribed to BA introduction of carbon atoms into the melonbased carbon nitride structures during copolymerization process, which changes the electronic structure of g-C 3 N 4 and thus extends its optical absorption range. Obviously, the combination of TiO 2 and modified CNQDs contributes to the enhancement of the visible absorption.
The samples have been used as photoanodes in PEC measurements, which is performed in a three-electrode electrochemical system, with 0.5 m Na 2 SO 4 electrolyte (pH = 7.62) under simulated solar light illumination at 100 mW cm −2 . The linear sweep voltammograms displayed in Figure 3a clearly show that both CNBxQD@TiO 2 and CNQD@TiO 2 samples show higher photocurrent values and negative shift of the onset potential for water splitting than the pristine TiO 2 does. CNB 1.5 QDs@TiO 2 (with the content of 1.5 g BA) shows the highest photocurrent density (J), and demonstrates only slight enhancement as compared with CNB 0.1 QDs@TiO 2 (see Figure S5, Supporting Information). The photocurrent density    gradually decreases with an increase of the BA content, even though the corresponding absorption in the visible increases. This can be attributed to the formation of defects in the melonbased carbon nitride structures during the substitution of N in the copolymerization process. The resulting defects facilitate charge separation.
The linear sweep voltammetry measurement performed under intermittent illumination (Figure 3b) shows that the CNQDs@TiO 2 composite shows remarkable enhancement of photocurrent density as compared with pure TiO 2 . After copolymerization with BA, the optimal sample (CNB 0.15 QDs@TiO 2 ) achieves high photocurrent density of 0.57 mA·cm −2 at 1.23 V (versus reversible hydrogen electrode (vs RHE)), which is 4.75 times higher than that for the pristine TiO 2 (0.12 mA·cm −2 ) and 1.9 times higher than CNQDs@TiO 2 (0.3 mA·cm −2 ), at identical conditions. These results not only demonstrate great advantages of the extended absorption range (into the visible range) and better charge separation, but also suggest that the small size and good dispersion of CNQD on the TiO 2 nanopillars induce abundant active sites, which enhance the PEC performance. Figure 3c displays the chronoamperometry curves of samples at 1.15 V versus RHE under discontinuous (chopped) light illumination. The samples show high stability of the photoresponse. More importantly, as shown in Figure 3d, the optimized photoelectrode (CNB 0.15 QDs@TiO 2 ) demonstrates excellent stability with only 0.72% decay of photocurrent density (from 0.56 to 0.556 mA·cm −2 ), under continuous illumination of a simulated solar light at 1.15 V versus RHE for more than ≈111 h. This electrode retains the remarkable stability even at 0.65 V versus RHE ( Figure S6, Supporting Information) and also retains a relatively stable photocurrent value of 0.065 mA·cm −2 after irradiation for 72 000 s (12 h). We attribute this outstanding performance to a strongly reduced photocorrosion by the CNQD decorating the TiO 2 nanopillars.
The photocatalytic hydrogen generation ability of the samples was also investigated (see experimental details in the Supporting information). As shown in Figure 4, the pristine TiO 2 exhibits the HER of only of 0.055 µmol·h −1 ·cm −2 . A significant enhancement of HER up to 0.225 µmol·h −1 ·cm −2 occurs already for the CNQD@TiO 2 sample, but a really dramatic increase, up to 0.8525 µmol·h −1 ·cm −2 is achieved in the sample of the optimal composite (CNB 0.15 QDs@TiO 2 ).
Based on the above results, we provide the following, microscopic explanation of the excellent performance, which we attribute to the atomic level change in the g-C 3 N 4 QDs. This is shown schematically in Figure 5. The exposure to dicyandiamide and BA at 550 °C during our process leads to the substitution of N atoms, shown in blue color in the top-right inset in Figure 5, with C atoms (N-defects), shown in green color in both insets. The N-defect sites are more reactive, which improves the PEC performance, and in consistent with the previous reports of induction defects in photocatalytic system. [20] In conclusion, we developed atomically modified graphitic carbon nitride QDs, characterized by strongly increased reactivity and stability. These have been deposited on arrays of TiO 2 nanopillars, forming composites used as a photoanode for the PEC water splitting. We demonstrate that these photoanodes are highly stable and are characterized by a highly efficient PEC performance. The photoanode based on the best (optimized) composite (CNB 0.15 QDs@TiO 2 ) exhibits the highest photocurrent density of 0.57 mA·cm −2 at 1.23 V, and an excellent photocatalytic stability of ≈111 h, under continuous illumination. It also demonstrates remarkable hydrogen production, with a rate of 0.8525 µmol·h −1 ·cm −2 , which is 15.5 times higher than that Global Challenges 2019, 3, 1800027