Borocarbonitride Layers on Titanium Dioxide Nanoribbons for Efficient Photoelectrocatalytic Water Splitting

Abstract

Heterostructures formed by ultrathin borocarbonitride (BCN) layers grown on TiO₂ nanoribbons were investigated as photoanodes for photoelectrochemical water splitting. TiO₂ nanoribbons were obtained by thermal oxidation of TiS₃ samples. Then, BCN layers were successfully grown by plasma enhanced chemical vapour deposition. The structure and the chemical composition of the starting TiS₃, the TiO₂ nanoribbons and the TiO₂-BCN heterostructures were investigated by Raman spectroscopy, X-ray diffraction and X-ray photoelectron spectroscopy. Diffuse reflectance measurements showed a change in the gap from 0.94 eV (TiS₃) to 3.3 eV (TiO₂) after the thermal annealing of the starting material. Morphological characterizations, such as scanning electron microscopy and optical microscopy, show that the morphology of the samples was not affected by the change in the structure and composition. The obtained TiO₂-BCN heterostructures were measured in a photoelectrochemical cell, showing an enhanced density of current under dark conditions and higher photocurrents when compared with TiO₂. Finally, using electrochemical impedance spectroscopy, the flat band potential was determined to be equal in both TiO₂ and TiO₂-BCN samples, whereas the product of the dielectric constant and the density of donors was higher for TiO₂-BCN.

1. Introduction

The current energetic model based on fossil fuels is unsustainable from an environmental perspective, as it is one of the leading causes of global warming and climate change [1,2,3]. The focus is now placed on solar and wind energy, which have the problem of being intermittent, which points to the necessity of developing new ways of storing energy.

Among all the energy storage methods, energy storage using molecular bonding stands out, such as the one in the hydrogen molecule. Hydrogen has been reported to be a suitable energy vector and a clean energy fuel if its production comes from renewable sources [2,4].

In 1972, Honda and Fujishima reported a way to obtain hydrogen by carrying out a photoassisted water splitting reaction using TiO₂ as the photoanode [5]. Since then, this effect has been considered one of the cleanest methods to obtain green hydrogen and a promising strategy to overcome the environmental and energy crises [6,7]. The water-splitting reaction consists of two partial reactions, the oxygen evolution reaction (OER) and the hydrogen evolution reaction (HER). The OER is the rate-determining step as it involves the transfer of four electrons [8], and this is the reaction we will tackle in this paper.

The low cost, stability and non-toxicity of TiO₂, as well as its adequacy to carry out the water splitting reaction, has drawn the attention of many groups who have reported the good properties of this material by synthesizing it in different structures and nanostructures [9,10].

To increase the charge transfer between the electrode and the electrolyte, metals nanoparticles, such as Pt and Ni, are commonly used as active electrocatalytic sites for water splitting [11]. The main focus is now placed on developing new metal-free compounds to be used as catalysts for the oxygen and hydrogen evolution reactions.

Graphene analogues and other 2D materials have demonstrated to be highly interesting metal-free compounds with a wide range of applications in electrocatalysis [12]. These layers have the advantage of being distributed along all the electrode surface, increasing the reaction area in comparison to the metal nanoparticles. Among these 2D materials borocarbonitride compounds (BCNs hereafter) stand out, which are low cost and highly stable materials formed by h-BN and graphite domains. BCNs have been reported to be efficient electrocatalysts for the HER [13,14] and, most interestingly, have recently been proved to be an efficient electrocatalysts for the OER, improving the properties of TiO_x substrates for this reaction [15].

In this article we have first confirmed the utility and versatility of plasma enhanced chemical vapor deposition to grow BCN on samples with different morphologies. This technique has been previously used to grow BCN on TiO_x, Cu, and other flat substrates [15,16]. In this work, BCN was grown, for the first time, on nanostructured TiO₂ samples. In particular, BCN was grown on TiO₂ nanoribbons, obtained by the thermal annealing of TiS₃ [17]. A deep characterization of TiO₂ nanoribbons, with and without BCN, as well as the starting material, TiS₃, was made with scanning electron microscopy (SEM), X-ray diffraction (XRD), Raman spectroscopy and diffuse reflectance measurements. X-ray photoelectron spectroscopy (XPS) characterization of bare and BCN-covered TiO₂ nanoribbons was performed.

Finally, we demonstrated the good properties of the BCN as an electrocatalyst of the OER by improving the charge transfer between the TiO₂ nanoribbons electrode and the KOH aqueous electrolyte, under dark and light conditions.

2. Materials and Methods

2.1. Synthesis

The starting TiS₃ material has been obtained by the sulfuration of Ti disks (15 mm diameter, Good Fellow 99.5%) in sealed Pyrex ampoules for 20 h at 550 °C, using sulfur powder as sulfur source [18,19].

TiS₃ nanoribbons were squashed in one direction and oxidized on a hot plate at 300 °C in air (the decomposition temperature of TiS₃ [17]) for 20–30 s, which allowed us to obtain the desired TiO₂ nanoribbons. Figure S1 shows the change in the color of the samples, from black to white, at a glance.

BCN was grown on TiO₂ using PE-CVD to get the TiO₂-BCN heterostructures. The TiO₂ sample and a single-source molecular precursor (methylamine-borane, BH₃NH₂CH₃) were placed inside a Pyrex ampoule immersed in liquid nitrogen and then sealed under vacuum at a pressure around 10⁻⁵ mbar. When the sealed ampoule reaches room temperature, the molecular precursor is in equilibrium with its vapor pressure in the 10⁻² mbar range. Then a plasma was activated inside the ampoule using the radiation of a conventional microwave oven. More details about this technique can be found elsewhere [15,16].

2.2. Characterization

The starting material, TiS₃, as well as TiO₂ and TiO₂-BCN heterostructures, were characterized by using different techniques.

The morphology of the samples was investigated by scanning electron microscopy (SEM) using a Hitachi S3000 instrument. Additionally, they were characterized using Raman spectroscopy with a WITec ALPHA 300AR instrument using a confocal microscope with lenses of 20× and 100×. The used laser had a power of 0.2 mW and an excitation wavelength of 532.3 nm.

The structural properties were measured using a Panalytical X’Pert Pro X-ray diffractometer at glancing angle configuration (incident angle of 1.7°, CuKα radiation).

The optical reflectance spectra were recorded in a UV/VIS/NIR Perkin-Elmer LAMBDA 950 spectrophotometer equipped with an integrating sphere to collect the reflecting flux, using a spot size of 21 mm² in the 300–2000 nm spectral range.

The composition on the surface of TiO₂ and TiO₂-BCN samples was investigated by X-ray photoelectron spectroscopy (XPS). These measurements have been carried out in an ultrahigh-vacuum (UHV) chamber, with a base pressure in the low 10⁻¹⁰ mbar range. Photoelectrons excited by an Al Kα photon source ($h\upsilon =1486.7 \mathrm{eV}$), were measured by a hemispherical electron analyzer (VG Microtech Clam-2) in a pass energy mode set at 50 eV for Ti and C, and 100 eV for B and N. Further details about the procedure are available in [15,16,20,21]. The binding energy (BE) was calibrated by acquiring the Au4f_7/2 (84.0 eV of BE) core-level after each measurement. The measurements were done after annealing the samples at 320 °C for 1 h, at 110 °C for 15 h and 320 °C for another hour in UHV.

Electrochemical measurements were done in a three-electrode cell. Our material, which had an apparent area of 1.3 cm², was placed as the working electrode (WE), a platinum sheet (9 cm²) as the counter electrode (CE), and, as the reference electrode (RE), an Ag/AgCl electrode filled with 1M KNO₃ (XR440 from Radiometer Analytical) was used. Its electrode potential was 484 mV vs. a normal hydrogen electrode (NHE). These three electrodes were immersed in 0.1 M (pH = 13.0) and 1.0 M (pH = 13.7) KOH aqueous electrolyte. The potentials are converted to the reversible hydrogen electrode (RHE) using Equations (1) and (2):

$$E_{NHE}=E_{Ag/AgCl}+E_{Ag/AgCl}^0$$

(1)

$$E_{RHE}=E_{NHE}+0.059·pH$$

(2)

where $E_{NHE}$ is the electrode potential in the NHE scale, $E_{Ag/AgCl}$ is the experimental electrode potential measured vs. Ag/AgCl reference electrode, $E_{Ag/AgCl}^0$ is the electrode potential vs. the NHE, $E_{RHE}$ is the electrode potential in the RHE scale, and pH is the pH of the solution used.

The electrochemical and photoelectrochemical (PEC) measurements were done using a potentiostat–galvanostat PGTAT302N (Autolab) provided with an integrated impedance module, FRAII. The WE was illuminated with a Xe lamp (Jobin Yvon) of 75 W in the visible-UV range. The maximum light intensity reaching the sample was 140 mW. During electrochemical measurements, argon was bubbled at a constant flow of 20.0 sccm. To characterize the electrodes under both dark and illumination conditions, linear sweep voltammetry (LSV), cyclic voltammetry (CV) and current density measurements at a fixed potential were employed. In addition, electrochemical impedance spectroscopy (EIS) was used to characterize the interface electrolyte-semiconductor, by using a sinusoidal AC voltage signal with an amplitude of 10 mV and a variable frequency between 100–1000 Hz.

More details about the photoelectrochemical experimental system can be found in the Supplementary Materials (Figure S2).

3. Results and Discussion

3.1. Morphological and Structural Characterization

The oxidation of the TiS₃ nanoribbons was performed to change the atomic structure and the composition of the material, without changing the morphology of the sample, in order to obtain TiO₂ nanoribbons. Figure 1a reports the morphology of TiS₃ before and after the oxidation. It is clear from the SEM measurements, that there was no significant change in the morphology, apart from a slight increase of the roughness on the TiO₂ surface compared to TiS₃. We also acquired optical microscopy images to have a deeper understanding of the change on the surface of these nanoribbons (Figure 1b), which allowed us to determine that the material became more transparent to the optical microscope, in good agreement with the results reported by Ghasemi et al. [17] and with our optical characterizations shown below (see Figure 4).

Structural changes were also monitored via XRD after the oxidation and after growing the BCN on top of this oxidized material (Figure 2). There were clear changes in the structure due to the thermal annealing, as there was a transition from monoclinic TiS₃ to tetragonal TiO₂-anatase with some peaks of tetragonal TiO₂-rutile.

After the BCN synthesis, the intensity of the rutile peaks increased. This was ascribed to the high temperature that the sample achieved during exposure to the plasma, which modified the structure of the bulk material by crystallizing the sample from anatase TiO₂ to rutile TiO₂.

Raman spectra of TiS₃ and TiO₂ nanoribbons, reported in Figure 3, show that there is a complete change in their structure from TiS₃ (175 cm⁻¹, 303 cm⁻¹, 373 cm⁻¹, 562 cm⁻¹) [22] to TiO₂-anatase (145 cm⁻¹, 396 cm⁻¹, 518 cm⁻¹, 643 cm⁻¹) [10,23,24]. In the case of TiO₂-BCN, there was no change in the Raman spectra of the nanoribbons, thus they maintained their anatase composition. The growth of the BCN in this sample was confirmed by the presence of D (1375 cm⁻¹) and G (1590 cm⁻¹) Raman bands, ascribed to the BCN layers graphene-like sp2 structure [15,16].

It must be mentioned that in the TiO₂-BCN sample, some TiO₂ rutile nanoribbons [25,26] were also found in the Raman spectra (Figure S3). This result is in good agreement with the increase in the rutile peaks in the XRD diffractogram for this sample. The nanoribbons were found at the borders of the samples and did not contribute to the electrochemistry as they were not exposed to the electrolyte due to the shape of the electrode holder.

3.2. Optical Characterization

Diffuse reflectance measurements were done using an integrating sphere to characterize the change in the energy bandgap. The reflectance (R) data were converted into a Kubelka–Munk function, F(R) [27] (Equation (3)), which was proportional to the optical density in the optical absorption measurements.

$$F\left(R\right) =\frac{{\left(1-R\right)}^2}{2R}$$

(3)

Results as a function of the incident light energy can be seen in Figure 4a. At first glance, the presence of 1.23 eV, 1.73 eV and 2.1 eV absorption bands, characteristic of the TiS₃ samples [19] can be observed, but they disappear after the oxidation process due to the transformation of TiS₃ into TiO₂.

To obtain the band gap energy of TiS₃ and TiO₂ samples, Tauc fitting equation for direct band gaps (${\left(F\left(R\right)·h\nu \right)}^2$) has been applied to the F(R) function (Figure 4b). A clear change can be observed from $0.94\pm 0.04 \mathrm{eV}$ in TiS₃, to $3.3\pm 0.4 \mathrm{eV}$ in TiO₂. These results are in good agreement with the values previously obtained for TiS₃ [19] and TiO₂ [28].

3.3. Surface Chemical Composition and Element Distribution

For further characterization of the chemical bonding state in the surface of TiO₂ and TiO₂-BCN samples, XPS measurements were carried out. The Ti2p level for the TiO₂ sample is shown in Figure 5a, and the B1s, C1s and Ti2p core levels for TiO₂-BCN can be seen in Figure 5b–d. A fitting analysis using Voigt line shapes (combination of Gaussian and Lorentzian curves) was performed to take into account the overall experimental uncertainty and the intrinsic linewidth, respectively. A Shirley background was introduced as a fitting parameter for all the analyzed peaks. All the fitting parameters for the graphs in Figure 5 are reported in Table S1.

From this result, it is evident that the Ti2p levels show the presence of Ti⁴⁺ in both samples, which can be attributed to the formation of TiO₂ [29]. However, in the case of TiO₂-BCN, a small additional peak at 457.55 eV of BE appears, which is ascribed to the presence of Ti³⁺ bonds [30,31], probably due to small areas with defects close to the edge of the sample. Nevertheless, this component represents the 0.5% of the relative intensity, meaning that the plasma did not significantly affect the surface of the substrate for the growth times used.

In the TiO₂-BCN sample, XPS measurements revealed that the BCN layer was composed of C and h-BN domains with high mutual doping levels of B and N in C, and of C in BN, respectively. The B1s and C1s peaks appear with a large width and are composed of more than one peak, indicating the presence of different components associated with the mutual bonding among the elements.

The B1s core level in TiO₂-BCN shows three peaks due to the B–C (190.51 eV), B–N (191.41 eV) [15,16,21,32,33,34] and B–O (192.96 eV) bonds [15,16,20,21,35]. The C1s core level exhibit four main components due to the C–C bonding at 284.6 eV, signature of the sp2-bonded carbon, C–N bonds (285.48 eV), C–B bonds (282.65 eV), and multiple components due to different bonds of C to O [15,16,21,32,33,34], which are unresolved and appear as a single broad band at 286.43 eV. We were not able to decompose the N1s peak in different Voigt functions, as we did with C1s and B1s peaks, since the signal of N1s is overlapped with a small signal of the Ta4p_3/2 peak, ascribed to the tantalum clips used to fix the sample to the sample holder. The spectrum in an energy range close to the N1s peak (385–413 eV) is shown in Figure S4.

Thus, the B and C 1s core levels present the typical features associated with the mutual chemical bonds in sp2-hybridized compounds. This is a clear signature of the formation of a ternary BCN planar layer [15,16]. The amount of BCN in the sample was estimated by XPS quantification, taking into account the cross-section, obtaining a ratio between Ti and B of 7:1, which suggests that the layer of BCN was ultrathin.

3.4. Photoelectrocatalytic Activity for the OER

The effect of BCN on the electrocatalytic activity of TiO₂ was investigated at first via linear sweep voltammetry (LSV) measurements under intermittent illumination conditions comparing the results from TiO₂-BCN heterostructure and these from bare TiO₂ one (Figure 6a). Regarding the dark condition, bare TiO₂ presented low currents and poor catalytic activity for the OER compared to those of the TiO₂-BCN sample, which points out the electrocatalytic effect of BCN. In fact, at the maximum applied potential (1.9 V vs. RHE), the dark current was 12-fold higher than that of the bare TiO₂. Photocurrents were measured at different applied stationary potentials (Figure 6b), and, as can be seen, it increases from around 1 µA (bare TiO₂) to around 40 µA (TiO₂-BCN) at 1.23 V vs. RHE. The photocurrents were stable (Figure S5) and have a positive value, in good agreement with the nature of TiO₂ as an n-type semiconductor [36]. It can be concluded that the BCN layer does not only improves the electrocatalytic properties of the TiO₂ at dark conditions, but also under illumination by increasing the photocurrents. The LSV curves, as well as CV curves, were measured for different scan rates, which made it possible to determine that the current that appears in the TiO₂-BCN sample before the OER in the LSV curve is due to the scan rate and not to another reaction (Figure S6).

The stability has been monitored by doing series of 50 cycles at 0.05 V/s, proving the high stability of our material, as can be seen in Figure S7. Raman measurements were also made before and after the photoelectrochemical measurements to prove that there was no degradation of the sample and that the BCN layer was still there (Figure S8).

In order to characterize the interface between the electrodes (TiO₂ and TiO₂-BCN) and the electrolyte (KOH), EIS measurements were done. By measuring the impedance at different frequencies, the capacitance in the spatial charge region ($C_{SC}$) could be acquired, and with the Mott–Schottky equation (Equation (4)), the value of the flat band potential ($V_{fb}$) could be obtained.

$$\frac{1}{C_{SC}^2}= \left(\frac{2}{\epsilon ·A^2·e·N_D}\right)·\left(V_{bias}-V_{fb}-\frac{k_B·T}{e}\right)$$

(4)

where $\epsilon$ is the dielectric constant of the electrode, ${\epsilon }_0$ is the vacuum permittivity, $A$ is the area of the electrode, $e$ is the charge of the electron, $N_D$ is the donor density, $V_{bias}$ is the applied potential, $k_B$ is the Boltzmann’s constant and T is the temperature. The term $\frac{k_B·T}{e}$ can be neglected due to its low value when compared to $V_{bias}$ and $V_{fb}$.

The value of the flat band potential is highly significant in the electrochemical characterization of a semiconductor–electrolyte interface, as it is related to the bottom energy of the conduction band. Together with the bandgap energy, it is used to describe the energy levels position at the electrode–electrolyte interface, determining the adequacy of a material to carry out or not any reaction, in this case, water splitting. Figure 7a,b shows the Mott–Schottky plot at three representative frequencies for the bare TiO₂ and TiO₂-BCN samples. The value obtained for the flat band potential is $0.2\pm 0.1 \mathrm{V} \mathrm{vs}. \mathrm{RHE}$ for bare TiO₂, and $0.2\pm 0.1 \mathrm{V} \mathrm{vs}. \mathrm{RHE}$ for TiO₂-BCN. These results are in good agreement with values of the flat band potential previously obtained for TiO₂ samples [9,10], and with a previous report about the good electrocatalytic activity of BCN for OER, in which it was shown that the BCN layer does not affect the flat band potential of the underlying material [15]. Table S2 summarizes the main differences between TiO₂ and TiO₂-BCN.

The slope of the Mott-Schottky fitting gives information about the dielectric constant of the material and its donor density. By using the data of the dielectric constant of TiO₂ at 20 °C reported by Wypych et al. [37] and the slope of bare TiO₂, its donor density was determined to be independent of the frequency and has a value of $\left(1.53\pm 0.02\right){10}^{19}{ \mathrm{cm}}^{-3}$ (Figure S9). This is in good agreement with the fact that the donor density is a characteristic parameter of the material and should not change with the frequency. However, as observed in [37], the dielectric constant is a function of the frequency.

By comparing the slopes of TiO₂ ($m_{{\mathrm{TiO}}_2})$ and TiO₂-BCN ($m_{{\mathrm{TiO}}_2-\mathrm{BCN}})$ at each frequency (Equation (5)) the relationship between the product of $\epsilon$ and $N_D$ for both samples can be determined (Figure 7c).

$$\frac{m_{{\mathrm{TiO}}_2}}{m_{{\mathrm{TiO}}_2-\mathrm{BCN}}}=\frac{{\epsilon }_{{\mathrm{TiO}}_2-\mathrm{BCN}}. {N_D}_{{\mathrm{TiO}}_2-\mathrm{BCN}}}{{\epsilon }_{{\mathrm{TiO}}_2}. {N_D}_{{\mathrm{TiO}}_2}}$$

(5)

It can be concluded that the TiO₂-BCN heterostructure has a greater value of factor $\epsilon ·N_D$ than the bare TiO₂. This can be partially ascribed to a higher $N_D$ in the TiO₂-BCN sample, induced by the plasma treatment during BCN growth. The higher donor density in the TiO₂-BCN heterostructure is also in good agreement with the increase in the photocurrents in this sample in comparison with TiO₂ (Figure 6b). Moreover, the dielectric constant of the TiO₂-BCN heterostructure was also modified as the factor $\frac{{\epsilon }_{{\mathrm{TiO}}_2-\mathrm{BCN}}. {N_D}_{{\mathrm{TiO}}_2-\mathrm{BCN}}}{{\epsilon }_{{\mathrm{TiO}}_2}. {N_D}_{{\mathrm{TiO}}_2}}$ is not constant with the frequency. With the value of ${\epsilon }_{{\mathrm{TiO}}_2}.$ ${N_D}_{{\mathrm{TiO}}_2}$, the product ${\epsilon }_{{\mathrm{TiO}}_2-\mathrm{BCN}}. {N_D}_{{\mathrm{TiO}}_2-\mathrm{BCN}}$ vs. the frequency can be obtained, as shown in Figure 7d. This factor also follows an exponential decay tendency, similar to that obtained for TiO₂ [37] in the region of 100–1000 Hz, ascribed to the ${\epsilon }_{{\mathrm{TiO}}_2-\mathrm{BCN}}$ as the ${N_D}_{{\mathrm{TiO}}_2-\mathrm{BCN}}$ is expected to be independent of the frequency.

To summarize our results, the photo-electrochemical response characterizations suggest that the BCN layer acts as an efficient electrocatalyst, improving electron transfer between the electrolyte and the underlying TiO₂ electrode. It has been reported before that these good electrocatalytic properties are related to the heterogeneity of these compounds, formed by highly doped C-rich and h-BN domains. Substitutional doping and grain boundaries defects act as electrocatalytic sites for OER [15]. The improvement of the electrocatalytic properties of the samples is also confirmed by EIS measurements, that show an increase in the product of the donor density and the dielectric constant induced by BCN growth. On the other hand, it must be noticed that the BCN layer is very thin and absorbs mainly in the UV region [16], so it is not expected that it contributes to the creation of electron–hole pairs due to the appearance of new energy levels

4. Conclusions

We have successfully changed the composition and structure of TiS₃ by thermal annealing at 300 °C to obtain TiO₂ without changing the nanoribbon morphology of the samples. The XRD and Raman measurements reveal a change from TiS₃ monoclinic to anatase tetragonal TiO₂, and the diffuse reflectance measurements a change in the gap from $0.94\pm 0.04 \mathrm{eV}$ to $3.3\pm 0.4 \mathrm{eV}$. On top of the TiO₂ nanoribbons, we have successfully grown a BCN layer by plasma-enhanced CVD without affecting the TiO₂ structure and morphology as it has been proved by XRD, Raman and XPS techniques. The analysis of the chemical composition and bonding scheme of the BCN layer revealed that our layer is composed of C and h-BN nanodomains with high mutual doping levels of B and N in C, and of C in BN, respectively.

The heterostructure TiO₂-BCN has been measured in a photoelectrochemical cell, corroborating the good photoelectrocatalytic properties of the BCN to carry out the OER when compared with the bare TiO₂ substrate. Cycling stability tests and Raman measurements after the experiments demonstrated that the BCN layer remained on the sample and that the structure of the sample has not changed, showing the high stability of our samples. Finally, with EIS measurements, flat band potentials for TiO₂ and TiO₂-BCN have been determined to have the same value ($0.2\pm 0.1 \mathrm{V} \mathrm{vs} \mathrm{RHE}$). By analyzing the slopes of the Mott–Schottky plots, it has been determined that the factor $\epsilon ·N_D$ is higher in the case of the TiO₂-BCN sample which is in good agreement with the increase observed in the photocurrents. In conclusion, the present results point out the excellent photoelectrocatalytic properties of the BCN as a metal-free material to be used in water splitting devices.