Filippo
Pota
a,
Maida Aysla
Costa de Oliveira
a,
Christian
Schröder
a,
Aran
Rafferty
a,
Clara
De Castro
b,
Ludivine
Rault
c,
James A.
Behan
d,
Frédéric
Barrière
d and
Paula E.
Colavita
*a
aSchool of Chemistry, Trinity College Dublin, College Green, Dublin 2, Ireland. E-mail: colavitp@tcd.ie
bEarth Surface Research Laboratory, Geology Department, Trinity College Dublin, College Green, Dublin 2, Ireland
cUniv Rennes, CNRS, ScanMAT – UAR 2025, Rennes, F-35000, France
dUniv Rennes, CNRS, Institut des Sciences Chimiques de Rennes – UMR 6226, Rennes, F-35000, France
First published on 21st March 2025
Electrocatalytic hydrogenation (ECH) of organics of biomass origin represents a promising strategy to enable integration of renewables and circular economy practices. However, most electrocatalysts investigated for ECH remain largely based on precious metals. Nanostructured materials based on transition metals encapsulated in a nitrogenated carbon matrix (M@C:N) offer a promising alternative. Herein, we report on the synthesis of Mo@C:N and W@C:N composites that display the same metal atomic concentrations and thus allow for a comparative study of the effect of the metal centre identity on the properties of such heterostructured materials and their performance in the ECH of benzaldehyde, a diagnostic organic substrate. A combination of structural characterisation methods indicates that the type of metal impacts carbon porosity and metal surface concentration in the synthesised structures. W displays a higher tendency to yield encapsulated nanoparticles compared to Mo, which is instead present with surface excess but at predominantly high oxidation states. Electrolysis studies at varying potentials demonstrate high product rates of benzaldehyde hydrogenation, with good selectivity for the production of the corresponding alcohol vs. the dimerization side product. Turnover frequency (TOF) estimates under the operational conditions tested suggest that replacing Mo-centres with W-centres in M@C:N architectures improves overall performance. A comparison of performance indicators with those for Pt-group metals suggests that W@C:N could be a competitive material for practical implementations of ECH.
Electrocatalysts for ECH processes are generally still based on precious metals such as Pd, Pt, Au5–7 and their alloys, or on carbon-supported precious metals (M/C) such as Pt/C, Pd/C, and Ru/C.8–11 Since the above elements can command market prices ranging from 15 to 90 USD per g,12 there is an ongoing research effort to develop abundant and low-cost electrocatalysts as viable alternatives. Recently, the use of metal nanostructures encapsulated within N-doped carbon architectures (M@C:N) has emerged as a promising strategy for the preparation of materials with ECH activity.13 M@C:N materials have received significant attention as electrocatalysts for a range of cathodic processes of high importance in energy technologies, including the hydrogen evolution reaction (HER), and the reduction of O2 and CO2.14–20 M@C:N heterostructures can display unique properties such as high surface area and tuneable surface electronic structure and chemical reactivity.18 Moreover, the presence of a carbon shell that protects the metal from oxidation due to air exposure or from corrosion in aqueous solutions can expand the range of metals used for electrocatalysis to encompass low-cost and earth abundant transition metals, enhancing the versatility and stability advantages of M@C:N architectures.18,21 However, development of M@C:N architectures for ECH still presents significant challenges and little is known about how to optimize both the metal core and the graphitised scaffold to improve ECH performances. Organic hydrogenations must progress with efficiency under electrochemical conditions at which the HER is typically a competing reaction. The Volmer reaction step leading to Hads activation at the electrocatalyst surface is a requirement for both the HER and the ECH to take place.22 However, these two reactions compete for this intermediate to yield final products: indeed the Tafel and Heyrovsky steps in the HER consume Hads species and may drastically suppress the faradaic efficiency (FE) of organic hydrogenations.11
Previous work from our group demonstrated for the first time that M@C:N electrocatalyst materials synthesised using a low cost transition metal such as iron are active in the ECH of carbonyl compounds. In this work, we show that this material design strategy can be expanded to incorporate alternative transition metals into the electrocatalyst, namely W and Mo, enabling the optimisation of ECH performance indicators in M@C:N architectures. Mo and W are highly desirable as metal cores due to their significantly lower costs compared to precious metals, with market prices of ca. 0.07 USD per g and 0.05 USD per g, respectively.12 Further, these metals and several of their compounds also possess the ability to activate chemisorbed hydrogen through the Volmer step while displaying differences in their activity towards hydrogen evolution,18,23,24 thus offering a potential route to regulating the degree of competition between the HER and the ECH. Benzaldehyde (BZH) was used as a model organic compound due to its solubility in water25,26 and because it enables quantitative benchmarking of performance indicators relative to other results in the ECH literature.27–29 Our results demonstrate that the identity of the metal core can have a profound influence on the properties of the graphitised carbon matrix and on the surface chemical composition in M@C:N materials with identical atomic metal loadings. This translates into important changes in product rates, faradaic efficiencies and turnover frequencies thus providing new insights on how to optimise transition metal-based heterostructured electrocatalysts for the ECH of organics.
Material | Metal precursor | Initial metal precursor (g) | wt. (%) | at. (%) | BET (m2 g−1) |
---|---|---|---|---|---|
a Data obtained by ICP-OES on the graphitised material. b Data obtained by ED-XRF on the graphitised material. | |||||
Mo@C:N | (NH4)6Mo7O24·4H2O | 1.78 | 10.8a | 1.51 | 229 |
W@C:N | (NH4)6W12O39·H2O | 0.61 | 17.7b | 1.41 | 450 |
RPM/BP | N/A | N/A | N/A | N/A | 600 |
The bulk structure of the materials was first investigated via XRD; Fig. 1b reports patterns collected for metal-free RPM/BP, Mo@C:N and W@C:N. RPM/BP materials yielded two peaks at 24.3° and 43.2°, indexed as (002) and (101) reflections of the graphitic carbon,38 respectively. The Mo@C:N pattern indicates the presence of crystalline phases consisting of mostly Mo2C, with contributions from MoO2 oxide. Peaks at 34.4°, 37.8°, 39.3°, 51.9°, 61.5°, 69.3°, 74.4°, 75.6°, match well those Mo2C reference (PDF#35-0787) and are assigned to reflections (100), (002), (101), (102), (110), (103), (112) and (201), respectively. The broad peak at 25.3° is assigned to (002) reflection of graphitic carbon and matches the main reflection in the pattern of metal-free RPM/BP. The small broad peak at 36.9° indicates the presence of MoO2 (PDF#65-5787) as well. The XRD pattern for W@C:N shows peaks at 37.4°, 43.6°, 63.3°, 76.1° and 80.1°, that match those of the WN reference pattern (PDF#65-2898) and are assigned to reflections (111), (200), (220), (311), and (222), respectively.39,40 The (002) reflection of graphitic carbon is also detectable at 24.6°.
Synthesised materials were found to be porous; Fig. 1c shows the N2 adsorption/desorption isotherms collected for Mo@C:N, W@C:N and the metal-free RPM/BP. Samples display similar hysteresis loops characteristic of type IV isotherms, associated with capillary condensation, and type H4 loops reported for narrow slit-like micro-/mesopores.41,42 The BET specific surface area values calculated from the isotherms are summarised in Table 1. The incorporation of Mo or W into the structure results in a decrease in BET surface areas by 15% to 67% compared to metal-free RPM/BP. Assuming that contributions to the surface area from metal-containing phases are negligible relative to those from the carbon phases, and accounting for the metal wt% in Table 1, it is possible to note that for W@C:N the BET area is only slightly lower than expected from a replacement of 1.4% of C atoms in RPM/BP with W (494 m2 g−1). This suggests that the development of the carbon microscopic area during graphitisation is not significantly affected by the presence of the W centres at these concentrations. On the other hand, for Mo@C:N the BET area is less than half the value expected based on replacement of 1.5% of C atoms alone (535 m2 g−1). This suggests that Mo centres impact the development of pores in the carbon phase to a much greater extent than W. Metal centres can have a catalytic effect on carbon gasification and graphitisation,37,43–47 and such processes can enhance pore opening and lead to reductions in BET areas.
To better understand the impact on pore size, the DFT method was used to generate pore size distributions (PSD), shown in Fig. 1d. The three materials are similar insofar as they exhibit distinct occurrences of porosity in the micro (<2 nm) and meso (2–50 nm) ranges, with broadly similar distributions, i.e. peaks at 1.5, 3.5 and 5 nm, characteristic of the parent carbon phase. W@C:N and RPM/BP share the most obvious similarities, with a significant volume of micro-pores, which are key contributors to the surface area. A closer inspection of the micropore region shows differences, in that the second peak (ca. 1.5 nm) grows according to Mo@C:N > W@C:N > RPM/BP, while the peak at 1.1 nm decreases accordingly. In other words, larger micropores are generated at the expense of smaller ones. Notably, Mo@C:N displays limited contributions from micropores relative to both RPM/BP and W@C:N. This likely explains the lower surface area observed for this sample, as well as its low pore volume (Table S1†). Further, it highlights a key difference between Mo@C:N and W@C:N: for Mo@C:N the smallest tranche of micropores have either been rendered inaccessible or have undergone enlargement. DFT pore width values (mode), in Table S1,† show that pore size increases with a decrease in surface area, as expected. Of note is the fact that the values for RPM/BP and W@C:N occur in the micropore range, where that of Mo@C:N falls in the mesopore range thus confirming that pore evolution in the N-doped carbon matrix is significantly affected by the incorporation of Mo centres.
TGA analysis in air was carried out and results are reported in Fig. S1.† Briefly, in the case of the metal-free RPM/BP a single combustion process with zero residual mass is observed over 650–750 °C, associated with oxidation of graphitised carbon. In the presence of metals, carbon oxidation occurs at significantly lower temperatures, with the effect on onset temperature being more pronounced in the case of Mo-containing materials. This is consistent with oxidation being catalysed by metal centres, in agreement with prior literature;48,49 while the enhanced effect of Mo vs. W in these reactions is also consistent with observations from BET and DFT analysis that suggest a greater effect of Mo centres on pore opening processes.
Morphology was investigated via electron microscopy. Fig. 2a shows an SEM image of Mo@C:N, revealing particles with rough and irregular surfaces. SEM-EDX analysis confirms that Mo is uniformly distributed over the carbon matrix (see also Fig. S2†). HR-TEM, in Fig. 2b, indicates that the sample consists of opaque nanoparticles (∼5 nm diameter), embedded within a graphitised carbon matrix: nanoparticles exhibit lattice d-spacing of 0.138 nm, in Fig. 2c, closely matching (103) reflections of Mo2C at 0.135 nm (PDF#35-0787). SEM of W@C:N, in Fig. 2d, displays a porous solid matrix with well-defined rod-like nanostructures with diameters in the 50–200 nm range. SEM-EDX indicates that these nanorods are rich in W (see also Fig. S2†). HR-TEM of such features, in Fig. 2e, reveals that they display well defined crystal facets and are encapsulated within a graphitised shell 10–20 nm in thickness. The selected area electron diffraction (SAED) pattern of nanorods, in Fig. 2f, confirms d-spacings of 0.238 nm and 0.205 nm, consistent with the (111) and (200) reflections of WN at 0.238 nm and 0.206 nm, respectively (PDF#65-2898).
Surface composition was studied using XPS and Table 2 summarises results obtained from best fits of high-resolution spectra for all materials. Survey spectra, Fig. S3,† display peaks associated with C 1s, N 1s and O 1s ionisations, as well as those associated with characteristic Mo and W ionisations for the respective M@C:N materials. Both the nitrogen and metal %-contents are much lower for W@C:N than for Mo@C:N, despite the two metals being present at similar atomic bulk concentrations. Indeed, a comparison between metal/carbon atomic %-ratios obtained by XPS and those from bulk determinations (Table 1) indicates that while W remains preferentially in the bulk, or below sub-surface, Mo is found predominantly at the air interface.
C 1s spectra of Mo@C:N and W@C:N are shown in Fig. 3a. The C 1s peak was deconvoluted using five components attributed to sp2 (ca. 284.4 eV) and sp3 (ca. 285.4 eV) carbon, C–O/C–N (ca. 287 eV) and CO/COOH (ca. 289 eV) groups, and π–π* excitations (ca. 292 eV).50–53 For all materials, the sp2 peak constitutes the largest contribution to the total intensity, as shown in Table 2 suggesting that materials undergo graphitisation to a similar extent during the annealing step. N 1s spectra are shown in Fig. S3;† the low N-content in W@C:N results in low overall intensity, nonetheless, peak binding energies at ca. 498 eV for both spectra suggest that pyridinic-N accounts for the majority of contributions, as was also observed for metal-free RPM/BP (Fig. S4†). Best-fits obtained using three components assigned to graphitic-N, pyrrolic-N and pyridinic-N,51,54 with details summarised in Table S2,† support this conclusion. An additional peak at 394.6 eV is also observed for Mo@C:N and is characteristic of Mo 3p3/2 ionisations.55–57 Finally, we note that differences in N/C surface content observed via XPS spectra are consistent with results from XRD characterization. XRD showed that annealing yields MoCx/MoOx and WN phases in Mo@C:N and W@C:N, respectively. This indicates that N-precursors react readily with W and are sequestered to the subsurface, as is the case for the W atoms, whereas for Mo@C:N they remain available to react with the carbon matrix.
The Mo 3d spectrum of Mo@C:N, Fig. 3b, was fitted using three doublets with an energy splitting of 3.1 eV, assigned to Mo2+ (228.5 eV), Mo4+ (229.8 eV), and Mo6+ (232.2 eV) species. Mo2+ components indicate the presence of molybdenum carbides in the form of Mo2C and/or MoC, whereas Mo4+ and Mo6+ arise from oxide species due to surface oxidation in air, in agreement with previous reports for Mo carbides.55,58,59 Peak area ratios indicate that the majority of surface Mo (>80%) is indeed oxidised after air exposure, in contrast with the predominance of carbide phases in the bulk, evidenced by XRD studies.
The W 4f spectrum of W@C:N, Fig. 3c, was fitted using two doublets with energy splitting of ca. 3 eV.50,60–62 The 4f7/2 peak at low binding energy (32.2 eV) is characteristic of either W2C or WN,63–65 and therefore consistent with XRD results. The 4f7/2 peak at high binding energy (35.2 eV) is assigned to oxidised tungsten in the form of WO3.61,65 Notably, the oxide component accounts for a minority of surface W species, while carbides/nitrides constitute 64–68% of the contributions from metal centres. These compounds have been reported to oxidise in air as demonstrated by multiple studies,58,66–69 therefore the absence of significant contributions from oxide species suggests that the majority of the carbide/nitride phases are encapsulated within a protective carbon shell that prevents/minimises oxidation.13 This further supports findings from HR-TEM, while being consistent with the surface W-content being lower than the bulk value.
The electrochemical response was investigated via voltammetry experiments in 0.1 M H2SO4 using glassy carbon disks modified with the M@C:N materials via drop casting. Fig. 4 shows linear sweep voltammograms (LSV) for Mo@C:N and W@C:N at 2 mV s−1 in the presence and absence of 30.0 mM benzaldehyde in the electrolyte. In the absence of benzaldehyde, Mo@C:N displays HER activity (η = −0.16 V at 1.5 mA cm−2) comparable to that observed at a range of Mo2C electrodes under similar conditions.59,70–72 The presence of a cathodic peak at −0.078 V, characteristic of proton-coupled reductions of MoO3 at low pH,72–74 is consistent with a significant concentration of Mo6+ at the electroactive surface of Mo@C:N and with evidence from XPS characterisation. Mo2C is reported to be activated under cathodic polarisation,72 while MoO2/MoO3 are sluggish in the HER;72 therefore the observed activity of Mo@C:N in acid electrolytes is likely to be dominated by Mo2+ centres at the electrolyte interface.
![]() | ||
Fig. 4 LSV collected on GC disk for W@C:N and Mo@C:N in 0.1 M H2SO4 (dashed line) and 0.1 M H2SO4 with 30 mM BZH (solid line), at 2 mV s−1 scan rate. |
W@C:N also shows activity in the HER (η = −0.50 V at 1.5 mA cm−2) but with lower current densities relative to Mo@C:N. Similar overpotentials have been reported for either WN75 or carbon-encapsuled WC electrodes.76 Interestingly, the LSV is devoid of peaks characteristic of proton-coupled reductions of WO3 species,77 thus supporting the conclusion from XPS that the majority of W surface sites are in a low oxidation state. Given that drop-cast films are porous, a comparison of the LSV curves was also carried out after normalisation by capacitive contributions to account for differences in microscopic area (Fig. S5†).78–80 Mo@C:N was found to still display higher HER currents than W@C:N in agreement with prior reports of better HER performance in general for MoCxvs. WCx electrodes in acid electrolytes.81 Finally, control LSV experiments with polished glassy carbon and metal-free RPM/BP electrodes, Fig. S5,† yielded higher HER overpotentials than either Mo@C:N or W@C:N after correction for microscopic area. Results, therefore, strongly suggest that M@C:N materials have improved intrinsic HER activity relative to metal-free graphitic carbons and that this can be attributed to the incorporation of Mo/W centres in the form of carbides/nitrides.
LSV in 0.1 M H2SO4 with 30 mM benzaldehyde at 2 mV s−1, in Fig. 4, show a cathodic peak associated with the reduction of the organic at W@C:N (ca. −0.7 V), while the overall cathodic current density is decreased relative to the LSV in supporting electrolyte. This indicates a general quenching of HER faradaic currents due to adsorption of the organic species.13,82,83 On the other hand, Mo@C:N displays only slight changes in current density while only a broad, ill-defined peak is discernible over the cathodic background current (ca. −0.3 V). This suggests that there is minimal competition by organic species for the HER-active sites at the Mo@C:N surface. Nonetheless, the peak at −0.078 V, visible in supporting electrolyte, is suppressed thus suggesting that benzaldehyde interacts with Mo6+ surface sites.
Benzaldehyde reactions were then studied using electrolysis experiments coupled to product detection. Carbon cloth (CC) current collectors were drop-coated using the same ink dispersions as for disk voltammetry studies. Fig. 5a and b show SEM and SEM-EDX images of CC coated with Mo@C:N and W@C:N, respectively, indicating uniform metal distribution over the CC surface. The LSV in supporting electrolyte (Fig. S6†) shows that the presence of the carbon materials increases the capacitance of CC significantly, as expected after coating with a porous particle film. Capacitance values were found to be comparable for Mo@C:N and W@C:N (Fig. S7 and Table S3†), suggesting that the two materials do not differ significantly in ECSA despite their differences in BET area. Fig. 5c shows LSVs recorded for Mo@C:N and W@C:N on CC in 30 mM BZH in 0.1 M H2SO4. W@C:N displays multiple broad cathodic peaks between −0.8 and −0.26 V whereas Mo@C:N yields a featureless response, likely due to HER being the dominant contribution to the LSV, as observed for disk electrodes. The presence of multiple peaks associated with benzaldehyde in the LSV could arise from either mixed mass transport regimes in the porous CC electrode,84,85 or from reduction at different functional sites, as observed by some even at carbon disk electrodes of well-defined geometry.86
![]() | ||
Fig. 5 SEM images and EDX mapping of CC electrodes coated with (a) Mo@C:N and (b)W@C:N electrocatalyst inks. (c) LSV collected for W@C:N and Mo@C:N on CC support in 0.1 M H2SO4 + 30 mM BZH. An expanded view over 0 to −0.9 V is shown in Fig. S8.† |
An H-cell was used for electrolysis studies, as previously described.13 Chronoamperometry experiments were carried out in a 30 mM BZH in 0.1 M H2SO4 solution at three selected potentials: −0.26, −0.5 and −0.8 V vs. RHE, as shown in Fig. 6a. Catalyst layers displayed high mechanical stability over the duration of the experiments thus yielding stable chronoamperogram curves (Fig. S9†). Repeat potential step experiments also yielded comparable chronoamperograms (Fig. S9†), thus confirming the robustness and stability of the fabricated electrodes. Log–log plots of current vs. time (Fig. S9†) show that after ca. 100–150 s, the slope of these plots transitions from a value of ca. −0.5 to a value close to zero, indicative of convective control. Evidence of convection is in agreement with prior work on similar electrode materials and is attributed to gas evolution during HER.13
Both benzyl alcohol (BA) and hydrobenzoin (HBZ) were identified as reduction products via gas chromatography, confirming the hydrogenation of benzaldehyde. Carbon balance was found to be ca. 90% (Fig. S10†) for all materials, indicating negligible mass losses due to diffusion through the membrane, loss to the headspace and/or to surface adsorption. Yields and conversion values were calculated for all materials and are shown in Fig. S11.† As the potential is shifted to more cathodic values, both conversion and yield show an increase, while the preferred product remains BA (Fig. S12†). At any given potential, better yields and conversions were observed for W@C:N vs. Mo@C:N over 2 h, thus suggesting that given, the same metal atomic concentrations, W-containing catalyst layers result in better operational performances in organic hydrogenations.
The faradaic efficiency, FEECH were calculated and are summarised in Fig. 6b. Mo@C:N shows good FEECH at the lowest overpotential, but the efficiency of reductions decreases significantly at −0.5 V before increasing slightly at −0.8 V again. The greater HER activity of Mo@C:N vs. W@C:N and RPM/BP, enables facile Hads activation even at −0.26 V which increases rates of ECH via Hads addition or concerted e-H+ additions;5,82 however, as the potential is shifted to more cathodic values, the HER becomes the dominant faradaic process decreasing the overall efficiency of ECH. At higher cathodic overpotentials proton coupled reductions of BZH can become an additional and energetically favourable pathway for the production of BA/HBZ particularly at carbon surfaces, as discussed in detail in computational studies by Cantu et al.5 We believe this is likely to explain the increase in FEECH at −0.8 V for Mo@C:N. In the case of both W@C:N and RPM/BP, the efficiencies are low at −0.26 V but superior to that of Mo@C:N at −0.5 V, which is consistent with these materials being less effective at evolving hydrogen and with the HER being suppressed as a competing reaction at −0.5 V.
Product rates normalised by geometric area and by capacitance are shown in Fig. 6c and d, respectively. Normalisation by capacitance puts in clear evidence the significant differences in specific electroactive area between the metal-free material and M@C:N catalysts, as well as the superior performance in organic hydrogenations obtained with the latter. W@C:N materials offer significant advantages at −0.5 and −0.8 V. Interestingly, while the Volmer step is facile and H2 desorption limiting at WC/WN,24,87 it is Hads activation that is rate determining at carbons. Therefore, Hads coverages should remain low at RPM/BP but increase readily with cathodic overpotential at WN, which likely explains why the product rate normalised by specific area is markedly superior for W@C:N at both −0.5 V and −0.8 V.
Fig. 6e shows product rates for Mo@C:N and W@C:N after normalisation by bulk metal contents. These results align with trends in FEECH, showing higher performance for Mo@C:N at −0.26 V and for W@C:N at −0.5 V; while at −0.8 V the two material shows similar product rates. However, their performances diverge when considering differences in surface segregation between Mo and W centres observed for the two M@C:N. This can be taken into account by estimating an effective turnover frequency (TOFeff) for W and Mo metal centres at the electrolyte interface. Assuming that the double-layer capacitance arises from the carbon matrix alone (20 μF cm−2)88,89 and accounting for the areal carbon density of graphite of 3.82 × 1015 atoms per cm2,90 TOFeff values can be calculated using the atomic %-contents of 0.43% and 4.3% for W and Mo, respectively, obtained from XPS. TOFeff values thus calculated are shown in Fig. 6f and suggest that encapsulated W centers yield significant advantages in terms of intrinsic activity.
The above results indicate that both Mo@C:N and W@C:N show activity in the hydrogenation of benzaldehyde and that W@C:N appears to be particularly promising, achieving satisfactory FEECH and good product rates at moderate potentials. At low overpotentials, FEECH values for Mo@C:N samples are comparable to those reported for Pt/C (∼60%) and exceed those of similar M@C:N architectures synthesised using Fe as a metal core.13,82 Further, W@C:N shows good FEECH at −0.5 V compared to Pt/C, as determined by Song et al. at pH 5 and similar potentials.82 Interestingly, TOFeff values estimated for W@C:N under our operational conditions are comparable to those reported for Pt/C and Rh/C catalysts at −0.5 V vs. RHE in the same work. Additionally, both M@C:N materials show high TOF values relative to our previous report using Fe@C:N.13 Our findings therefore suggest that compositional changes can be successfully leveraged to modulate and improve ECH performance at M@C architectures based on transition metals.
It is important however to highlight that, when interpreting differences in performance between Mo and W-containing electrodes it remains challenging to disentangle the effects of a change in metal identity (W vs. Mo), from those arising from morphological and compositional differences that result from metal-catalysed graphitisation. In the case of W@C:N materials, metal centres are encapsulated and protected from oxidation; the majority of them are present in the form of WN nanorods that can improve, simultaneously, Hads stabilisation24,87 and metallic character/conductivity of the porous carbon matrix. For Mo@C:N it is also possible to argue that Hads coverages are stabilised relative to RPM/BP; however, the presence of surface MoOx indicates that only a fraction of metal sites should facilitate Hads activation, while charge transfer impedances are likely to be negatively affected by surface oxidised phases.91
Finally, the calculation of TOFeff values assumes implicitly that the metal sites are responsible for BZH reduction rates. This is likely an oversimplification that neglects possible activation of multiple N-/C-sites due to e.g. proximity effects of metal in the sub-surface,18,20 as well as the fact that hydrogenation pathways are potential-dependent.5 First, several proposed hypotheses have been put forward to explain the nature of the active site at M@C:N electrodes in the HER. Both M-sites in contact with the electrolyte but with their reactivity modulated by the surrounding carbon matrix, or C/N-sites with their reactivity modulated by metals in the subsurface are considered as possible and experimentally consistent explanations for Hads activation at carbon-encapsulated architectures. However, unequivocal identification of these active sites remains a topic of ongoing debate, as highlighted recently by Yoo et al.,92 despite the HER being a considerably simpler reaction than the ECH of benzaldehyde. Nonetheless, and regardless of whether the active sites are C- or M-centers, the activity remains dependent on the degree to which the metal is in proximity to, or at, the surface. Therefore, XPS atomic coverages, which reflect metal content within the first 1–3 nm from the surface, are a relevant quantity when attempting to account for metal effects on surface reactivity. A more in-depth analysis of the active site in the ECH is likely to require further experimental work, including e.g. use of poisoning/capping agents.92 Second, potential-dependence of the active site must also be contemplated; at increasing cathodic overpotentials it is reasonable to assume that the total product rate stems at least in part from proton coupled reductions, as noted when discussing FEECH values at −0.8 V. Indeed, contributions from M-site mediated H-addition and from proton coupled reduction of BZHads must both be considered, particularly at high overpotentials and based on the current mechanistic understanding of ECH.5,93–95 Therefore, the validity of assumptions underpinning TOFeff calculations might be limited at the most negative potentials tested and comparisons between metal TOF values are best restricted to the low overpotential range where metal catalysed pathways are expected to dominate hydrogenation rates. Finally, although electrolysis experiments are performed in this work under natural convection and in an H-cell, well-defined convective conditions e.g. flow through electrolyser devices, would be desirable to fully account for mass transport effects, support comparative performance studies and investigate stability under conditions closer to those encountered in electrolyser applications.20
The synthesis and properties of materials Mo@C:N and W@C:N, with nearly identical atomic metal contents of ca. 1.5%, were investigated. Structural characterisation revealed that Mo@C:N has a bulk composition primarily consisting of Mo2C, but its surface is rich in oxides. In contrast, W@C:N has a bulk structure consisting of WN with minimal oxide present at its surface; W is localised in rod-like nanostructures protected from oxidation and likely corrosion by a graphitised shell. Both materials were found to be porous: the presence of W centres introduces only small changes in carbon porosity relative to the metal-free analogue material, whereas Mo centres catalyse pore opening.
Incorporation of Mo and W centres improves the HER activity for both materials compared to the metal free analogue RPM/BP, with Mo@C:N displaying higher HER cathodic currents. Electrolysis experiments coupled with gas chromatography analysis of products demonstrated that this translates into increased ECH activity at low overpotentials in the hydrogenation of benzaldehyde, the unsaturated compound used as a diagnostic substrate. Selectivity for the benzyl alcohol product was observed at all tested potentials; this is a promising result as achieving selectivity remains a challenge for most studied electrocatalysts, particularly at cathodic potentials.13,82,83 Mo@C:N shows high FE at anodic potentials, while W@C:N is more efficient at cathodic potentials due to suppressed competition between HER and ECH pathways. Product rates and TOFeff values were used as a basis for comparison, with W@C:N showing higher ECH performances at the investigated atomic loading, especially when considering the differences in surface excess between the two metals. Estimated TOFeff are comparable to those for Rh/C and Pt/C catalysts;82 given the four orders of magnitude difference in material cost, results suggests that W@C:N architectures are likely to be cost-competitive for ECH applications and could reduce pressure on precious and critical raw materials for ECH implementation at scale.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4ta08953g |
This journal is © The Royal Society of Chemistry 2025 |