Synthesis and decomposition of mpg-C3N4
Figure S1 in the Supporting Information shows the isotherm acquired from the sorption experiment performed on the mpg-C3N4 template (inset) as well as the average pore size distribution calculated in the same experiment. The graph in the inset shows a hysteresis loop in the plot and reflects a type IV isotherm associated with the capillary condensation process in the structures of the mesopores, according to IUPAC classification. The specific BET surface area obtained was 248 m2 g−1, which was consistent with the formation of mesopores specifically designed by using colloidal silica as a hard template.25, 26 From the results in Figure S1 in the Supporting Information, a high value of the average pore volume (0.51 cm3 g−1) was obtained based on a Barret–Joyer–Halenda (BJH) plot, and the average desorption pore size was 64 Å with a narrow distribution, which accurately reflected the size of the original silica template used in the synthesis. Table S1 in the Supporting Information presents the textural properties discussed above and the elemental analysis obtained for the mpg-C3N4 reactive template used in this investigation. Elemental analysis gave a lower carbon to nitrogen ratio (≈0.63) when compared with the stoichiometry of C3N4 (0.75), which suggested a nitrogen-rich carbon nitride sample, possibly because the surface of the sample was mainly C-NH2, 2C-NH, and hydrogen-bonded OH groups or absorbed water molecules, as corroborated by spectroscopy studies and discussed below.
A FTIR spectrum of the mpg-C3N4 template is presented in Figure S2 in the Supporting Information; the FTIR spectrum of melamine is also displayed for comparison. The formation of the mpg-C3N4 structure was confirmed by the existence of a graphite-like sp2-bonding state and adsorption peak at 810 cm−1, which corresponded to the breathing mode of triazine (out-of-plane ring-bending vibration mode).29, 30 Characteristic stretching vibration signals of the tri-s-triazine heterocyclic rings are present in the range of 1200–1600 cm−1 with peak maxima at 1233, 1405, and 1565 cm−1.31, 32 The bands observed at 1317 and 1610 cm−1 are related to the C(sp2)N and C(sp2)N stretching modes, respectively, in a graphitic-type structure.32, 33 The weak signal at 2175 cm−1 can be assigned to the cyano group stretching band, which suggests that the condensation of the tri-s-triazine network is incomplete.34 The broad peaks observed from 2900 to 3400 cm−1 are generally related to the stretching and deformation modes of the residual NH components and their intermolecular hydrogen bonding. It is possible to compare this region with the melamine spectrum above 2900 cm−1, where the signals at 3125, 3325, 3415, and 3466 cm−1 are attributed to NH stretching vibration modes of the amino groups.35 Residual hydrogen atoms on the edges of the polymeric network bind through CNH2 and 2CNH bonds and are energetically stable.29 The absorbance of the OH band overlaps in this range and can be attributed to hydrogen-bonded OH groups and absorbed water molecules in the sample.36 Furthermore, there is no trace of the absorption bands for vSi-O-Si at 1000–1200 cm−1,37 which indicates successful removal of silica used for the synthesis of mpg-C3N4.
Figure 1 A shows the detection of the main gaseous products found when using MS during the temperature-programmed decomposition of the mpg-C3N4 template under a flow of argon. The products that remained gaseous after cooling to room temperature were observed between 705 and 1100 K, as reported previously.30, 32 At 940 K, clear peak maxima of the main signals were detected at 27, 28, and 52 amu; these were assigned to hydrogen cyanide (HCN, MW=27.02 g mol−1), nitrogen (N2, MW=28 g mol−1), and cyanogen (C2N2, MW=52.03 g mol−1), respectively. These results provide evidence for the formation of cyanogen (52 amu) as the main gaseous product formed during mpg-C3N4 decomposition. The recovered polymeric yellow film at the outlet of the reactor indicates the formation of heavier C3Nx species as a result of fragmentation of the tri-s-triazine heterocyclic rings in the carbon nitride material.32 Other nitrogen-containing gases and oxygenated hydrocarbons were not observed during the heating process. No remaining materials were recovered after the decomposition experiments, which clearly suggests that complete volatilization occurred during the decomposition of mpg-C3N4. From the findings of this study, we can derive Equation (1) for the decomposition of mpg-C3N4 at high temperatures and under inert conditions:(1)
Figure 1. Mass signals obtained A) during the temperature-programmed decomposition of mpg-C3N4 and B) during tungsten carbide nanoparticle synthesis.
Download figure to PowerPoint
Hydrogen and excess nitrogen may arise from the surface CNH2 and 2CNH groups, as confirmed by FTIR spectroscopy. Moreover, the stoichiometry of the carbon nitride template produced was calculated from elemental analysis results.
Tungsten carbide synthesis using an mpg-C3N4 template
A reactive hard-templating method using the nanometer-sized confinement encountered in mpg-C3N4 has been reported as favorable for the synthesis of metal nitride nanoparticles. TiN,25 VN,38 GaN,38 Ta3N5,26, 39 and ternary Al–Ga–N and Ti–V–N27 have been successfully obtained. The decomposition reaction of the confining matrix serves as a nitrogen source for the metal source found in the pores of the carbon nitride.28 To the best of our knowledge, carbide materials have not been reported with carbon nitride as the reactive template.
The pores of the reactive template were impregnated with a chloride dialkoxide solution of pentavalent tungsten. A blue solution was prepared from [WCl6] and ethanol, as described in Equations (2) and (3):40(2), (3)
Figure 1 B shows the mass spectra obtained during the attempted synthesis of tungsten carbide materials from the tungsten source immersed in mpg-C3N4 (1:1, WCl6 to mpg-C3N4 weight ratio). The main gaseous products indicated by the 27, 28, and 52 amu signals were assigned as HCN, N2, and C2N2, respectively, as previously observed in the experiments for mpg-C3N4 decomposition. However, the peak maxima were shifted to lower temperatures at about 845 K with the tungsten source when compared with those without tungsten species (at 920 K), as shown in Figure 1 A. This reduction of temperature clearly indicates reactions of mpg-C3N4 with the tungsten precursor [WCl3(OC2H5)2]. Two peaks were observed in the nitrogen signal; the first one corresponded to the reaction of the mpg-C3N4 template with the tungsten precursor and the second one (≈705 K; Figure 1 B) corresponded to the decomposition of intact mpg-C3N4, as indicated by the signal shown in Figure 1 A. A solid white powder recovered at the outlet of the reactor was identified as ammonium chloride (NH4Cl) by X-ray diffraction (XRD) measurements. The crystals were most likely deposited at the outlet walls of the quartz tube as a result of the reaction between HCl and NH3 and originated from the tungsten precursor solution and carbon nitride decomposition, respectively.
A series of tungsten products was synthesized from the reaction with the mpg-C3N4 template under various conditions. First, Figure 2 A shows XRD patterns of the samples synthesized by varying the starting precursor weight ratio ([WCl6]/C3N4 ratio) while keeping the temperature constant at 1223 K. Other variables studied included the volume of ethanol and heating rate, but no significant effects were observed in the product structure as a result of these changes. Alternatively, prominent effects on the product structure were observed with variations in the weight ratio of [WCl6] to mpg-C3N4. A 1:1 ratio gave hexagonal δ-WC with a P6m2 space group (ICSD-77566), as identified in the XRD pattern (1:1 in Figure 2 A), with primary diffraction peaks detected at 2θ=31.5 (d=2.858, WC [0 0 1]), 35.7 (d=2.511, WC [1 0 0]), and 48.2° (d=1.887, WC [1 0 1]). A broad peak was observed between 64.1 and 65.4°, which corresponded to d=1.450 (WC [1 1 0]) and 1.426 (WC [0 0 2]), respectively. In the literature, the δ-WC phase is commonly referred to as α-WC, and this designation is used hereafter.41 As the amount of available carbon increased in the synthesis, a mixture of α-WC and α-W2C phases was obtained when using a 1:2 [WCl6]/C3N4 ratio and only broader peaks assigned to the α-W2C phase were observed at [WCl6]/C3N4 ratios of 1:4 and 1:8 (Figure 2 A). This α-W2C phase has a hexagonal structure with a P3m1 space group (ICSD-77568), with characteristic diffraction peaks observed at 2θ=37.9 (d=2.368, W2C [0 0 2]), 61.8 (d=1.498, W2C [1 1 0]), and 74.9° (d=1.266, W2C [1 1 2]). Finally, when the amount of carbon nitride template was reduced in the synthesis (2:1), a mixture of tungsten carbide phases and tungsten metal was identified. Unstrained cubic W0 (ICSD-43667) with primary diffraction peaks at 2θ=40.2 (d=2.239, W0 [1 1 0]), 58.2 (d=1.583, W0 [2 0 0]), and 73.1° (d=1.292, W0 [2 1 1]) was detected in the diffractogram in conjunction with the characteristic diffractions of the α-WC and α-W2C phases.
Figure 2. The effects of the tungsten to carbon nitride weight ratio and synthesis temperature on the XRD patterns of the tungsten-based products. A) Variation of the [WCl6] to mpg-C3N4 weight ratio at 1223 K. B) Variation of the temperatures at a 1:1 weight ratio. C) Variation of temperatures at a weight ratio of 1:2. The asterisk (*) indicates the formation of tungsten oxide species (WO2, W18O49, and WO3).
Download figure to PowerPoint
Figure 2 B shows XRD patterns of the samples synthesized at different temperatures (1073–1373 K), while keeping the ratio of [WCl6]/C3N4 at 1:1. At relatively low temperatures (<1223 K), broad peaks characteristic of hexagonal α-W2C were observed and indicated the formation of small particles, as expected from utilizing the confinement of the mpg-C3N4 pores. Hexagonal α-WC was formed at temperatures above 1223 K, whereas α-W2C was detected only slightly at both 1223 and 1323 K, as the minor phase in the product, and was almost undetectable at 1373 K. It is interesting to note that there was a clear phase transition at 1273 K where the primary diffraction peaks assignable to cubic W0 appeared together with the peaks for the α-WC phase. Some diffraction peaks assigned to WO2 and WO3 are also highlighted with an asterisk (*) in Figure 2; these peaks must have originated from the oxidation of tungsten metal during the passivation process at the end of the synthetic procedure. It was considered that metallic tungsten formed as a result of the eutectoid decomposition of α-W2C, in agreement with the WC phase diagram, where metallic tungsten and α-WC could coexist below 1500 K.41 The peaks for α-WC became sharper at higher temperatures, which clearly indicated an increase in particle size because particle aggregation increases at high temperatures. The Scherrer equation was used to approximate the size of the crystallites in the samples42 in a range from 56 to 152 Å in the temperature range of 1223 to 1373 K, as listed in Table 1. This finding is consistent with a gradual growth of particle size with increasing temperatures. The phase assignments of the products obtained with XRD are also listed in Table 1. The results described above can lead to conceptual Equations (4)–(7):(5), (6), (7)
Table 1. Phase assignment from the XRD diffractogram and particle size evaluation from the Scherrer equation of selected samples produced at different temperatures with a constant precursor weight ratio of 1:1.
|T [K]||Phase from XRD||FWHM at 2θ [º][a]||Scherrer size [Å]||Surface area [m2 g−1][b]|
|1273||WC W2C W0||48.2||63||60|
Equation (4) should include extra amine/imide groups from C3N4, which form additional ammonia. Equation (5) indicates the disproportionation of α-W2C to form metallic W and α-WC. Metallic tungsten is not stable and it reacts with the remaining carbon or with oxygen (passivation).
The role of the W and mpg-C3N4 precursor ratio during the synthesis of tungsten carbide nanoparticles was explored in experiments in which the mass of mpg-C3N4 was doubled with respect to the amount of the tungsten precursor at different temperatures. It was observed that α-W2C initially formed at temperatures above 1073 K did not transform into α-WC even at 1273 K (Figure 2 C). For all temperatures ranging from 1073 to 1273 K, broad peaks were observed, mainly from α-W2C characteristic diffraction patterns. This result indicates that excess C3N4 (thus leading to excess carbon) suppresses the disproportionation of α-W2C to metallic W and WC [Eq. (5)]. Hereafter, the samples are denoted based on the major phases identified by XRD followed by the temperature used in the syntheses (e.g., W2C-1173, WC-1223, WC-1373).
The X-ray photoelectron spectroscopy (XPS) results of W 4f for the synthesized carbide materials obtained at different temperatures are shown in Figure 3. WO3 (99.9 %, Aldrich) was used as a reference material for the direct comparison of hexavalent oxidation states and gave spectra with binding energies for W 4f7/2 and W 4f5/2 of 35.8 and 38.0 eV, respectively, which was consistent with the values reported in the literature.43 All of the tungsten carbide samples showed W 4f signals above (31.3±0.1) eV, which corresponded to the metallic state of tungsten.21 For α-WC samples synthesized in the range of 1223–1373 K, the W 4f spectra show peaks at binding energies assigned to WC at 32.2 (W 4f7/2) and 34.4 eV (W 4f5/2), which closely agrees with reported values.15, 43, 44 These α-WC samples showed signals emitted from the partially oxidized surface of the materials because of the O2 passivation treatment and unavoidable exposure of the nanoparticles to the ambient conditions. The intensity of the W 4f signals assigned to the WO3 species (35.8 and 38.0 eV) decreased with an increase in temperature, whereas the carbidic W 4f signal intensity (32.2 and 34.4 eV) increased with increasing temperature, which corresponds to the concrete formation of WC, as observed in the XRD patterns (Figure 2 B). The α-W2C produced at lower temperatures (<1223 K) reflected a further reduced state of W with the peak location shifted 0.1 eV lower when compared with the α-WC samples.15 These XPS spectra did not show surface oxidation with any appreciable peaks at 38.0 or 35.8 eV (Figure 3). When α-WC samples were treated under substantial Ar ion sputtering, the intensity of the oxidative phase was diminished, which suggests the removal of consecutive layers of the oxide species without an effect on the W 4f carbidic signals (Figure S3 in the Supporting Information).
Figure 3. W 4f XPS spectra of tungsten-based products obtained at different temperatures. The spectrum of WO3 is included as a reference for comparison. The metallic tungsten characteristic signal is indicated with an arrow at 31.3 eV.
Download figure to PowerPoint
Table 2 shows the results (in mass percentages) of elemental analyses (C, H, N) of the synthesized W samples to confirm the formation of the carbide rather than the nitride form of the transition metal. The nitrogen content for all samples (0.1–0.8 %) was well below the possible stoichiometric values for tungsten nitride (7.07 and 3.66 % for WN and W2N, respectively). As listed in Table 2, the carbon content in the samples obtained below 1223 K (6.12 %) exceeded the stoichiometric value (3.16 %), as assigned from XRD (α-W2C). The samples produced at higher temperatures with an assigned stoichiometry of α-WC exhibited closer agreement with the theoretical carbon content. In the WC-1273 sample, the carbon content was 1.9 %, which was well under the stoichiometric value and supports earlier observations from XRD and XPS where a phase transition was observed at 1273 K and metallic tungsten led to the oxidation of the carbide phase. All of the samples produced at higher temperatures (>1223 K) and identified by XRD as α-WC showed similar carbon contents, which is a good indication of the reproducibility of the procedure and is in reasonable agreement with the results observed by XRD and shown in Figure 2 B.
Table 2. Elemental analyses and calculated BET surface areas of tungsten carbide products obtained at varying temperatures, while maintaining a precursor weight ratio of 1:1.
|Sample[a]||Weight||Element [%]||BET surface area|
| ||ratio[b]||C||N||H||[m2 g−1]|
To further characterize the tungsten content in the resulting tungsten carbide samples, thermogravimetric analysis coupled with differential scanning calorimetry (TGA/DSC) was utilized to simultaneously monitor the heat flow in the experiment in addition to the weight change. For a WC-1373 sample under a flow of air (Figure S4 in the Supporting Information), the mass of the sample increased throughout the heating process until it was 16.5 % higher than that of the original product. The final theoretical stoichiometric mass gain of 18 % (for α-WC) is represented in Figure S6 in the Supporting Information by a dashed line. The 1.5 % difference correlates to the findings observed in elemental analysis where excess carbon was present in the sample. There are two subtle shoulders in the weight signal during the weight increment process, which is also reflected in the heat flow signal as two exothermic peaks with maxima at 713 and 800 K. The first peak can be attributed to the partial oxidation of the material towards possible intermediate mixed-valence species (i.e., WO2, W2O3, and/or W18O49) with the release of COx.45 The main peak represents the full oxidation of the sample towards WO3, as observed by the color change from black to bright yellow at the end of the experiment. The tungsten mass content in the original sample could be calculated from this experiment to be approximately 92.5 %, which is in close agreement with the stoichiometric value of 94 % for WC and further confirms the results obtained from elemental analysis and XRD.
Nitrogen sorption experiments were performed on the samples to obtain the BET surface area. Table 2 compiles the values of the surface areas of the materials. The surface area of the products decreased with increasing temperature. Higher surface areas were attained for W2C products compared with the WC products. The α-WC sample with the highest surface area (≈104 m2 g−1) was obtained at 1223 K with the precursor weight ratio maintained at 1:1. However, it is worth noting that the measured BET surface area for the tungsten carbide nanomaterials may be affected by the presence of excess carbon, especially at low temperatures where the amount of carbon is higher (>12 %, Table 2). As evaluated from the XRD diffractogram by using the Scherrer equation, the nanocrystallite size was 6–15 nm at higher temperatures (1223–1373 K) and about 2–3 nm at lower temperatures (1073–1173 K; Table 1). By using these sizes and assuming nonporous spherical particles with a homogeneous size distribution, the surface area of the W2C or WC particles can be approximated, as shown in Table 1. The surface areas obtained were slightly lower than the measured BET surface areas and the difference between these calculations may have originated from the presence of carbon and the error generated from the assumption of sphericity.
Figure 4 shows the results obtained from TEM investigations on W2C-1073 (A, B), WC-1223 (C, D) and WC-1373 (E, F). Low- and high-magnification micrographs are shown for representative samples obtained at different temperatures. As observed in Figure 4 A, C, and E, the particle size increases with increasing temperature, which is in agreement with the results observed from XRD and calculated from the Scherrer equation. The W2C-1073 sample exhibited a uniform particle size distribution (≈5 nm) that accurately reflected the original pore size of the mpg-C3N4 template, as observed in Figure 4 A. The HRTEM image of the WC-1223 sample showed a greater extent of aggregation and sintering, with nanoparticle sizes up to 10 nm (Figure 4 C). Finally, the samples obtained at an even higher temperature (WC-1373) showed sizes up to 25 nm. The HRTEM images on the right (Figure 4 B, D, and F) allowed us to observe the crystalline nature of the tungsten carbide nanoparticles produced in detail. It was possible to observe the remaining carbon present in the samples, as indicated by elemental analysis and TGA. Although it was not possible to quantify the amount of carbon present, energy-dispersive X-ray (EDX) spectroscopy was performed to confirm the chemical composition of the crystalline nanoparticles. A line profile of EDX spectroscopy was performed to study the local environment for a single nanoparticle observed in scanning transmission microscope (STEM) mode; the result shows that the nanocrystal is composed of tungsten and carbon; moreover, carbon was also observed in the surroundings of the nanoparticles (Figure S5 in the Supporting Information).
Figure 4. High-resolution transmission electron microscopy (HRTEM) images of W2C-1073 (A, B), WC-1223 (C, D), and WC-1373 (E, F). The [WCl6] to mpg-C3N4 weight ratio used for the synthesis was 1:2 (A, B) and 1:1 (C–F).
Download figure to PowerPoint
In summary, we have shown that by tuning the synthesis conditions, we are able to obtain materials with the desired tungsten carbide phase and a high surface area. This synthetic method led to the formation of extra carbon in addition to the tungsten phase, which plays an important role in inhibiting transformation of the α-W2C phase to the α-WC phase (Figure 2 B). As shown in the MS experiments, the tungsten precursor [WCl3(OC2H5)2] reacted with mpg-C3N4 inside the nanoconfinement of the mesoporous template to control the particle size of the “seeds” at relatively low temperatures (800–1000 K; Figure 1 B). When the synthetic conditions were optimized to obtain α-WC at higher temperatures (>1200 K), only α-WC was present in the final product, as confirmed by XRD, XPS, TGA, and elemental analysis. The high-surface-area tungsten carbide obtained could be utilized as a catalytically active site and, in this study, we investigated the electrochemical activity in water redox reactions.
Tungsten carbide tungsten carbide as an electrocatalyst
The electrochemical stability of WC under acidic conditions at room temperature was evaluated for the WC-1223 sample. A cyclic voltammetry (CV) experiment with a wide potential window (0.05–1.05 V vs. a reversible hydrogen electrode (RHE)) was performed to evaluate the stability of tungsten carbide materials. Tungsten carbide showed a major oxidation peak starting at 0.6 V versus RHE, which was consistent with similar observations found in the literature (Figure S6 in the Supporting Information).46 The oxidation peaks completely vanished after several cycles, which indicates the irreversible nature of the WC oxidation process. Great care must be taken to retain the original nature of the carbide structure to be able to correctly evaluate and compare the electrocatalytic performance of the materials.
The WC-1223 sample was tested for HOR activity in acidic solution and the results were comparable to those obtained for 40 % Pt/CB. The polarization curve of the WC-1223 material is presented in Figure 5 A. The HOR was evaluated by using a rotating disk electrode (RDE) with the rotation speed, ω, varying from 400 to 2500 rpm. For WC-1223, the HOR current was primarily controlled by a region of mixed kinetic/diffusion control at overpotentials up to 0.15 V versus RHE. Diffusion-limiting currents were prominent above 0.2 V versus RHE for curves obtained under rotational speeds below 900 rpm. It is worth noting that a wider overpotential region (0–0.3 V vs. RHE) can be used for kinetic analysis in the 2500 rpm polarization curve (Figure 5 A) where the measured current is primary kinetically controlled. Platinum polarization curves showed that diffusion-limited currents were rapidly obtained at 0.06 V versus RHE (Figure S7 in the Supporting Information). The results indicated that the HOR anodic currents were kinetically controlled in a very narrow potential range (0–0.05 V vs. RHE), in agreement with the literature.47 The linearity and intercept to zero observed in the inset of Figure 5 B proved that at sufficient overpotentials (i.e., 150 mV vs. RHE) the current is mainly diffusion-limited and proportional to the square root of ω.48 This total mass-transfer-limited condition is described by the Levich equation in the form shown in Equation (8):
Figure 5. A) HOR polarization curves on WC-1223 at varying rotation rates obtained by using a RDE. B) HOR Koutecký–Levich plots for WC-1223 obtained at varying overpotentials (mV vs. RHE). The inset shows the Koutecký–Levich plot for 40 % Pt/CB obtained at 150 mV versus RHE. C) Tafel plots using mass-transport-corrected HOR currents of the WC-1223 sample and 40 % Pt/CB (H2-saturated 0.5 M H2SO4, 50 mV s−1, 298 K).
Download figure to PowerPoint
in which iL is the Levich current, B is the Levich constant, n is the number of electrons transferred in the half reaction, F is the Faraday constant, A is the surface area, D is the diffusion coefficient (298 K, 3.7×10−5 cm2 s−1, estimated from the product of H2 diffusivity at infinite dilution and the ratio of the dynamic viscosities of the electrolyte and pure water),47 ω is the angular rotational speed, υ is the kinematic viscosity (298 K, 1.2×10−2 cm2 s−1), and C0 is the H2 saturation concentration (298 K, 7.14×10−4 M).49 From the 40 % Pt/CB experimental data obtained at 0.15 V versus RHE (shown in the inset of Figure 5 B), the least-squares regression fit returned a slope of BC0=6.032×10−3 mA rpm−0.5. At sufficient overpotential and at a specific rotational speed it is possible to estimate the surface area of the catalyst from Equation (8). Thus, the platinum surface area calculated from the Equation (8) was 0.0304 cm2. When a kinetic limitation is involved in the electron transfer reaction, as in the case for tungsten carbide (Figure 5 A), the currents at the RDE are described by the Koutecký–Levich equation [Eq. (9)]:
in which iK represents the current in the absence of any mass-transport effects and, hence, the current is determined only by kinetic limitations. From Equation (9), the plot of the inverse of the current as a function of the inverse of the square root of the rotational speed will produce a straight line with an intercept corresponding directly to the kinetic current iK−1 and a slope of (BC0)−1. Figure 5 B depicts the Koutecký–Levich plots for tungsten carbide (WC-1223) at different overpotentials (mV vs. RHE). There is a noticeable contribution of kinetic limitation over a wide range of overpotentials (0.025–0.2 V vs. RHE). Non-zero intercepts shown in Figure 5 indicate that the rate of the electron-transport process was the limiting factor. However, at sufficient overpotentials (0.4 V vs. RHE), we can consider the kinetic component negligible and it is possible to assume complete diffusion-limiting conditions. From the least-squares regression lines obtained from Figure 5 B, the average slope observed from 0.075 to 0.4 V versus RHE was BC0=3.08×10−3 mA rpm−0.5. Similarly to the platinum calculations, the estimated tungsten carbide surface area was calculated to be 0.015 cm2. The mass-transport-corrected kinetic current was normalized by using the calculated surface area approximations from the Equation (9). Hence, diffusion-corrected Tafel plots of the polarization curves obtained from 40 % Pt/CB (2500 rpm) and WC-1223 (1600 rpm and 2500 rpm) are presented in Figure 5 C. The kinetic current was independent of the rotation rate above 1600 rpm, as observed in Figure 5 C, which ensures the accuracy of iK determination for WC-1223 in this rotation range. From the Tafel analysis (η=a+b log j) in Figure 5 C, the exchange current density (j0) can be extrapolated from the Tafel curve linear region to an intercept log j0 and represents the electrochemical reaction rate at equilibrium.48, 50 This kinetic parameter can be described as any system’s capacity to facilitate a net current (reaction rate) without incurring significant energy losses from activation.48 The Tafel slope (β) can be empirically calculated from the polarization curve between specific overpotentials (η), which provides insight into the possible mechanism of the reaction (see below). In the literature,6, 51, 52 three elementary reaction steps depict the mechanism for hydrogen oxidation/evolution on solid-state catalytic sites:
The dissociative adsorption of hydrogen molecules on the catalyst surface, known as the Tafel reaction or recombination reaction for hydrogen evolution [Eq. (10)].(10)
The direct reaction of hydrogen molecules through the Heyrovsky reaction, also known as the ion-plus-atom reaction [Eq. (11)].(11)
The ionization of the H atom to produce one electron and one hydronium ion in solution, which results in an empty active site in the reaction, known as the Volmer reaction [charge-transfer reaction; Eq. (12)].(12)
In the elementary steps mentioned above, Hads represents atomic hydrogen chemisorbed to a metal surface (i.e., platinum) and the symbol * denotes a free adsorption site. From Equations (10) to (12), it is possible to theoretically calculate the expected Tafel slope from surface coverage assumptions (see the Supporting Information). For the HER, the first common step is represented by the Volmer equation or discharge reaction [Eq. (12)]. Then there are two possibilities, either electrochemical desorption [Eq. (11)] or a recombination step [Eq. (10)]. The reversible reactions (HOR) follow a Tafel–Volmer or Heyrovsky–Volmer mechanism. As mentioned previously, platinum is one of the most studied catalysts for the HOR and HER in the literature and exchange current densities are difficult to obtain because of the fast kinetics and lack of accuracy during correction for mass-transport limitations in most cases. However, several groups have found that the Pt(110) surface is the most active for both the HOR and HER because it exhibits Tafel slopes of approximately 30 mV dec−1, which correspond to a Tafel–Volmer mechanism [Eqs. (10) and (12)].47, 50, 53 The recombination reaction or Tafel step is the rate-determining step (RDS), as reflected at low overpotentials with a measured Tafel slope of 28 mV dec−1 for the HOR in Figure 5 C (40 % Pt/CB), which is also in agreement with the findings for well-characterized polycrystalline Pt47, 54 and the theoretical value of 30 mV dec−1 obtained from Equation (10) [Eq. (13)].(13)
in which R is the ideal gas constant, T is the absolute temperature, and F is the Faraday constant. When the electrochemical desorption step [Eq. (11)] is slow then Equation (14) is valid:(14)
in which α is the transfer coefficient. Finally, if Equation (12) is the RDS or the rates of Equations (11) and (12) are comparable, then a high slope value is obtained from Equation (15):
The HOR Tafel plot obtained for tungsten carbide seems to have two apparent Tafel slopes of 43 (0.1–0.4 V vs. RHE) and 150 mV dec−1 (0.4–0.7 V vs. RHE; Figure 5 C). There is no clear linear region in the kinetic HOR polarization curve, which possibly indicates a potential-dependent reactive intermediate coverage.49 Moreover, polarization of the HOR cannot be simplified into a single linear Tafel equation and the kinetic parameters in the Tafel region are difficult to accurately determine. Nevertheless, on the basis of the Tafel slopes obtained, the WC-1223 sample exhibited a Heyrovsky–Volmer HOR mechanism with the ion–atom reaction as the RDS [Eqs. (11) and (12)]. At higher potentials, the rates of the electrochemical desorption [Eq. (11)] and the discharge reaction [Eq. (12)] may become comparable, which makes it difficult to define the RDS. However, the calculated slopes (>120 mV dec−1) indicate that at sufficiently high overpotentials the discharge reaction [Eq. (12)] may act as the RDS.
The catalysts synthesized at different temperatures were also examined for activity in the HER in 0.5 M H2SO4 at 298 K by using a RDE to achieve steady-state current conditions without diffusion effects on the measurements. Figure 6 A gives the voltammograms of tungsten carbide synthesized at different temperatures along with voltammograms of Pt and GCE. The acquired currents were tentatively normalized by the surface area calculated from the Levich analysis performed earlier for the HOR on WC-1223. The electrochemically active surface area of the platinum catalyst was calculated to be similar to the geometric area of the electrode (0.07 cm2). Cathodic currents attributed to the HER were observed for all carbide samples with an onset potential of approximately −0.1 V versus RHE. The WC-1223 sample exhibited the highest HER current among all of the samples with a hydrogen evolution overpotential below 100 mV versus RHE and it thus outperformed the rest of the samples. Figure 6 B shows Tafel plots for the HERs of the W2C-1073 and WC-1223 samples. Pt was an excellent catalyst for the HER, since it produced an exchange current density of j0≈1 mA cm−2. Among the synthesized tungsten-based nanoparticles, WC-1223 had the highest activity with j0=0.35 mA cm−2, followed by the W2C-1073 sample with an exchange current density of j0=0.28 mA cm−2, which is consistent with the polarization curves exhibited in Figure 6 A. However, it should be noted that for tungsten carbide materials, the exchange current density was normalized by the estimated surface area from Koutecký–Levich experiments rather than the electrochemically active surface area as for the platinum electrode. The WC-1223 sample exhibited a Tafel slope of 84 mV dec−1 (0.1 to 0.15 V vs. RHE), whereas the W2C-1073 sample exhibited a slope of 102 mV dec−1 (0.13 to 0.19 V vs. RHE). The results indicate a Volmer–Heyrovsky HER mechanism catalyzed by the tungsten carbide samples (WC-1223) with a rate-determining electrochemical desorption step from the ion to the atom reaction, as shown in Equations (10)–(12).51, 52, 55 When the tungsten carbide phase is different (W2C-1073), the HER mechanism changes to the extent that the discharge step or the Volmer reaction becomes rate limiting. The high values for the Tafel slope most likely resulted from variation in the surface coverage of the adsorbed hydrogen. Thus, it is possible that the rates of the discharge reaction and the electrochemical desorption are comparable, which then results in intermediate values of Tafel slopes between 40 and 120 mV dec−1. As seen from the results in Table 3, a steeper Tafel slope was obtained for the samples where WC was the major phase present, which indicates a clear difference in the reaction mechanism pathways for different carbide phases (W2C and WC). The higher slopes obtained in the HOR correlate with the values obtained for the HER, which indicates that the reversible hydrogen reaction may proceed by the same mechanism with a similar RDS. The higher current density observed for the WC-1223 sample (Figure 6 A) could be attributed to the monocarbide phase of the sample (as resolved by elemental analysis, XRD, TGA, and XPS) in combination with the high surface area of the material (Table 2). α-WC exhibits better hydrogen evolution performance than α-W2C under acidic conditions even though α-W2C has a higher surface area. This trend is explained by a volcano plot of exchange current as a function of the hydrogen binding energy of the material, developed by Nørskov and co-workers,6 and is consistent with recent DFT and experimental studies on transition-metal carbides.15 Therefore, the Pt-like electronic structure allowed α-WC to show higher activity than any other sample in this study, even though the α-W2C samples exhibited a higher surface area. However, as the synthesis temperature increased, the HER current decreased along with the decrease in the surface area of the samples because of the reduction in the number active sites available for hydrogen evolution.
Figure 6. A) HER voltammograms of tungsten carbide samples synthesized at different temperatures (W2C-1073, W2C-1173, WC-1223, and WC-1273) along with voltammograms of a Pt electrode and a glassy carbon electrode (GCE). B) HER Tafel plots for W2C-1073, WC-1223, and Pt electrode. C) The stability test of the WC-1223 sample under a wide range of pH conditions. The first and 800th polarization cycles for the WC-1223 sample are shown in the inset (0.5 M H2SO4, in Ar, 50 mV s−1, 298 K).
Download figure to PowerPoint
Table 3. HER evaluation at −0.3 V versus RHE and Tafel parameters calculated from kinetic measurements in an Ar-purged 0.5 M H2SO4 solution for selected samples at room temperature with a 50 mV s−1 scan rate.
|Sample||Current density[a] [mA cm−2]||η [mV]||log (io) [mA cm−2]||β [mV dec−1]|
The stability test of the material with the best performance for the HER is shown in Figure 6 C. The first and the 800th polarization cycles are shown in the inset of Figure 6 C and are not significantly different; hence, the WC-1223 carbide sample demonstrated negligible current loss after 800 charging–discharging cycles between −0.3 and 0.1 V versus RHE in a 0.5 M H2SO4 solution at 298 K. This finding clearly indicates the high stability of the carbide material under acidic conditions. The activity in the HER was also evaluated under neutral and alkaline conditions and compared with that of the platinum catalyst. Figure 6 C shows the polarization curves obtained under different pH conditions and the results were compared with those obtained using 40 % Pt/CB (Table S2 in the Supporting Information). The onset potential for nonacidic conditions was very similar at −0.15 V versus RHE, although the current was lower than that under acidic conditions. As presented in the Supporting Information, the trend for WC-1223 was similar to that of 40 % Pt/CB; the activity decreased to 20 % of the original performance at neutral pH and then recovered to approximately 30 % of the activity observed under acidic conditions at pH 13. This result relates the electronic structure of tungsten carbide, which is similar to that of platinum, with the HER behavior of platinum, as previously discussed.15–17, 21, 46
Figure 7 shows the experiments performed for the ORR on W2C-1173, WC-1223, and WC-1373 samples, and a platinum electrode. In this case, the recorded currents for the materials were normalized using the geometrical surface area of the electrode (jgeom). Special care must be taken when evaluating the ORR on tungsten carbides to prevent oxidation of the catalyst. As previously discussed, tungsten carbide oxidation was observed at positive potentials higher than 0.6 V versus RHE (Figure S6 in the Supporting Information). Therefore, the polarization experiments were performed only for positive potentials below 0.5 V versus RHE to ensure that the catalytic nature of the material was unaltered. As shown in Figure 7, the ORR activity for all of the tungsten-based samples was low and cathodic currents attributed to the ORR were observed at potentials as low as 0.3 V versus RHE for the best tungsten carbide sample. The WC-1223 sample showed the highest activity towards the ORR among the other samples tested. As expected, the platinum catalyst exhibited excellent performance towards the ORR with an onset potential of 0.9 V versus RHE, in contrast to tungsten carbide with 0.3 and 0.1 V versus RHE for WC-1223 and W2C-1073, respectively. The reduced ORR activity observed for tungsten carbide materials was viewed as an advantage for an overall water-splitting application, as further discussed below.
Figure 7. Polarization curves in saturated O2 obtained for W2C-1173, WC1223, WC-1373, the Pt electrode, and GCE (0.5 M Na2SO4, pH 3.6, 50 mV s−1, 298 K).
Download figure to PowerPoint
Photocatalytic overall water splitting
As described previously, Pt-group metal surfaces catalyze proton reduction, but are also highly active in inducing the back-reaction from H2 and O2 to water (ORR).56 In the literature, the Ni/NiO57, 58 core/shell structure and RuO259 have been extensively utilized as successful cocatalysts for OWS. Recently, a chromium shell was studied as a possible selective membrane to prevent O2 permeation to the metal surfaces.7 Apparently, the core/shell structure of the chromium-based layer protects the metallic nanoparticle surfaces to suppress the ORR (back-reaction), while maintaining hydrogen evolution (i.e., proton reduction).8 Moreover, photocatalytic OWS in a particulate-type system has been achieved only on metal/metal oxide surfaces with a core/shell structure. Therefore, from the results obtained from the HER and ORR experiments where WC materials catalyzed hydrogen evolution but revealed poor performance for the ORR, the samples were considered to be potential cocatalysts for the water-splitting reaction.
OWS was attempted under varying loading conditions on a Na-doped SrTiO3 (STO) photocatalyst, and the results are shown in Figure 8 A. This photocatalyst was active for OWS under UV irradiation (photon distribution in Figure S8 in the Supporting Information) when impregnated with Rh and Cr species, giving about 300 μmol h−1 H2 and about 150 μmol h−1 O2, as discussed elsewhere.56 The photocatalyst with 2.75 wt % WC/STO:Na showed the highest activity with the production of stoichiometric amounts of hydrogen and oxygen in the experiment. HRTEM images are presented in Figure 8 B and show a representative tungsten carbide particle (WC-1373) supported on the WC/STO:Na photocatalyst. When the loading of tungsten carbide on the semiconductor photocatalyst STO:Na reached 10 wt %, the hydrogen production rate decreased and oxygen gas was not observed as a product, like in the case of unmodified STO:Na. The lack of oxygen production was indicative of the oxidation of impurities or surface hydroxyl groups. OWS was achieved at low oxygen and hydrogen partial pressures in the WC/STO:Na system. Tungsten carbide showed reasonably high electrochemical hydrogen evolution activity, while exhibiting poor oxygen reduction performance, which may explain the system’s ability to catalyze OWS. WC was used as a cocatalyst for photocatalytic water splitting and functioned as a hydrogen evolution site, while preventing the back-reaction in a dual-role system. The similarity of the electronic structure of tungsten carbide to that of platinum may be reflected in its ability to improve charge separation and serve as a hydrogen-evolution active site. It is also possible, as demonstrated by XPS, that tungsten carbide undergoes oxidation on its surface to prevent back-reactions, while reflecting the low activity observed in the OWS experiments. Although further optimization of the synthetic and reaction conditions is needed, WC shows promise in the search for non-noble-metal cocatalysts for OWS. Furthermore, these findings show that tungsten carbide can be used for water splitting without a core/shell structure, which opens up new possibilities in the design of cocatalysts for photocatalytic OWS.
Figure 8. A) OWS experiments performed on a recirculating reactor unit by using the STO:Na photocatalyst with varying loadings of the WC-1373 sample (300 W Xe lamp containing UV light, 100 mL milli-Q H2O, 50 mg photocatalyst). B) HRTEM micrographs of a) 1.25 wt % WC-1373/STO:Na; b) 5 wt % WC-1373/STO:Na.
Download figure to PowerPoint