1 Introduction

The rapid growth of fossil fuel consumption has caused significant energy crisis and environmental pollution, necessitating the development of clean and renewable energy sources. Among alternative energy resources, H₂ produces only water through its combustion and has the highest mass-energy density (142 MJ kg⁻¹). The generation of H₂ via electrocatalytic water splitting (2H₂O → 2H₂ + O₂) is probably the most environmentally friendly technique, and therefore, a tremendous amount of research is being conducted all over the world. An important issue is developing efficient and cheap electrocatalysts for the cathodic hydrogen evolution reaction (HER) because the most efficient HER electrocatalysts so far require Pt, an expensive noble metal.

Transition metal dichalcogenides (TMDs) have recently attracted attention for catalytic HER.^[1-6] TMDs are typical 2D layered materials with the chemical formula MX₂, where M is a group V–VII transition metal and X = S or Se. Each layer in TMD comprises a sublayer of M atoms sandwiched between two sublayers of X atoms. Pristine TMDs have intrinsically low electrical conductivity, and their exposed X sublayers are not often very catalytically active.^[1-3] These drawbacks are usually resolved by forming chalcogen vacancies, substitutional/adatom doping, metal/molecule ion intercalation, or alloying.^[4-6] Enhanced electrocatalytic activity toward HER or CO₂ reduction has been demonstrated for a number of binary metal-site alloy systems such as MoW,^{[7, 8]} MoRe,^[9-11] NbTa_,^[12] MoNb,^[13] MoV,^[14] NbV,^[15] WV,^[16] WNb,^[17] WRe.^[18] and ReV.^{[19, 20]} Notably, ternary alloy nanosheets like (MoNbV)Se₂ and (MoWV)Se₂ via colloidal reaction have demonstrated a synergistic effect to improve the HER performance of constitutional binary alloys.^{[21, 22]} The use of a large variety of components allows researchers to optimize the electronic structure of TMD alloys for the electrocatalytic reaction.

Semiconducting ReX₂ nanosheets exhibit excellent (photo)electrocatalytic performance.^[23-47] In a pioneering work, Yang et al. reported enhanced HER performance of (ReMo)S₂ alloy monolayers at x_Mo (mole fraction of Mo) = 0.45 synthesized via chemical vapor deposition.^[9] Our research group synthesized (ReMo)S₂ and (ReMo)Se₂ nanosheets by hydrothermal reactions, and the best HER performance occurred at x_Mo = 0.5 and 0.1, respectively, because of the more metallic phase than the ReX₂.^{[10, 11]} Lee et al. reported the colloidal synthesis of (WRe)S₂ monolayers quantum dots with enhanced catalytic HER at x_Re = 0.49.^[18] Alloying with metallic VX₂ can also improve the catalytic activity of ReX₂. We reported the synthesis of (ReV)Se₂ and (ReV)S₂ nanosheets via colloidal and solvothermal reactions, respectively, showing an enhanced HER performance at x_V = 0.2–0.8.^{[19, 20]} By introducing one more component, (ReMoV)X₂ nanosheets are expected to exhibit even better catalytic activity, but the related results have not been reported.

In this study, (ReMoV)S₂ and (ReMoV)Se₂ ternary alloy nanosheets were synthesized using solvothermal and colloidal reactions, respectively. Their composition was successfully tuned over a wide range (x_Re = 0.10 – 0.5, x_Mo = 0.10 – 0.5, and x_V = 0.16–0.8). Atomic-resolution scanning transmission electron microscopy (STEM) was employed to examine the crystal structure with the atomic distributions of metals. We analyzed thoroughly their electronic structures using X-ray photoelectron spectroscopy (XPS) and extended X-ray absorption fine structure (EXAFS). The electrocatalytic activity toward the HER in acidic electrolytes was studied for both the sulfide and selenide alloys, in comparison with the (ReV)X₂ binary alloys. We also monitored the electronic structures during the HER using ex situ EXAFS. First-principles calculations were performed to predict the crystal structures, density of states, and Gibbs free energy (ΔG_H*) of the ternary alloy nanosheets. The calculation results support the observed crystal/electronic structures as well as the HER performance of ternary alloys.

2 Results and Discussion

As depicted in Scheme 1a, (ReMoV)S₂ and (ReMoV)Se₂ alloy nanosheets were synthesized using solvothermal and colloidal reactions, respectively. The precursors were ammonium perrhenate (NH₄ReO₄), bis(acetylacetonato)dioxo molybdenum (VI) (MoO₂(C₅H₇O₂)₂), and vanadyl acetylacetonate (VO(C₅H₇O₂)₂). The procedures are the same as those of (ReV)X₂, except for adding the Mo precursor.^{[19, 20]} Starting from the three unitary compositions ReX₂, MoX₂, and VX₂, we prepared 10 samples of sulfide alloys with x_Re = 0.10–0.5, x_Mo = 0.10–0.5, and x_V = 0.16–0.8, as shown in the ternary diagram of Scheme 1b. For comparison, thermally annealed samples were also prepared by heating the as-grown samples at 400 °C. Eight selenide alloys were also synthesized with x_Re = 0.10–0.5, x_Mo = 0.10–0.5, and x_V = 0.25–0.75, as shown in the ternary diagram of Scheme 1c. In the rest of the paper, the sulfide and selenide samples are sometimes referred to as “aS” and “bSe” (a = 1–10, b = 1–8), respectively. The ternary compositions may also be denoted as (ReMo)_1-xV_xX₂ to highlight the V composition (x_V).

Figure 1a shows the scanning electron microscopy (SEM) and high-resolution transmission electron microscopy (HRTEM) images for 4S and 3Se (x_Re = x_Mo = x_V = 0.33). The layered nanosheets aggregated into flower-like nanoparticles (nanoflowers). Figures S1 and S2 (Supporting Information) show the SEM and HRTEM images of other sulfide and selenide samples, respectively. The ReX₂ nanosheets comprised 2–5 layers (thickness: ≈2 nm) that aggregated into nanoflowers with a size of ≈100 nm. By contrast, the thickness and size of the VX₂ nanosheets were ≈10 and 200 nm, respectively. The MoS₂ nanosheets (thickness: 2–5 nm) aggregated to form random-sized nanoflowers (30–50 nm), whereas the MoSe₂ nanosheets had a thickness of ≈2 nm and were bundled into nanoflowers ≈100 nm in size. After alloying, the size of the nanoflowers decreased to less than 50 nm. The interlayer distance (d₀₀₁) of sulfide nanosheets was in a wider range and larger (≈10 Å) than that of selenide nanosheets (≈6 Å).

Energy-dispersive X-ray spectroscopy (EDX) data verified the successful composition tuning and homogeneous distribution of Re, Mo, V, and S (or Se) (see Figures S1 and S2, Supporting Information). EDX elemental mapping and the corresponding high-angle annular dark-field (HAADF) STEM images of 4S and 3Se are shown in Figure 1b. In the normalized EDX spectra, the relative intensities of Re, Mo, and V peaks changed throughout the series. The composition of each sample matched that of the precursor (Table S1, Supporting Information). Seven of the sulfide samples (1S–7S) have zero S vacancy (defined as C_VX = 1- $$\frac{1}{2}\frac{{[ X ]}}{{[ {Re} ] + [ {Mo} ] + [ V ]}}$$, where X = S or Se), and 8S–10S have 3%–6% (see Table S1, Supporting Information). The ternary alloys in either the as-grown or annealed conditions had smaller C_VS than that of Re_1-xV_xS₂.^[20] By contrast, the selenide samples contained 4%–8% Se vacancies, which was comparable to that of Re_1-xV_xSe₂.^[19] These results consistently showed that the chalcogen vacancies increased with increasing x_V. We have previously proposed a model for this, in which the slower reaction of VX₂ produced X vacancies by forming V─O bonds.^{[19, 20]}

Figure 1c shows the atomic-resolution HAADF STEM images of four sulfide samples 1S (x_Re = 0.42, x_Mo = 0.42, x_V = 0.16), 4S (0.33, 0.33, 0.33), 8S (0.167, 0.167, 0.67) and two selenide samples 3Se (0.33, 0.33, 0.33) and 7Se (0.167, 0.167, 0.67). The S atoms (Z = 16) were not clearly identified because of the brighter Re (Z = 75), Mo (Z = 41), and V (Z = 23) atoms. The darker regions corresponding to the VX₂ sites become dominant at higher x_V. The 1T structure is obviously the major one in both the sulfide and selenide alloys. The corresponding fast Fourier transform (FFT) image shows the [01$$\overline{1}$$0] reflection spots of the 1T-like phase. The FFT image provides d₀₁₀ = 2.8 Å for sulfide and 2.9 Å for selenide, corresponding to the lattice constant a = 3.24 and 3.35 Å, respectively, which are close to the reference value of VS₂ and VSe₂, as discussed later. The line profiles along the marked area support the homogeneous mixing of Re, Mo, and V atoms, and the increased number of V sites with increasing x_V.

The crystal phases of the ternary alloys were examined using XRD. Figure 2a shows the data for ReS₂, MoS₂, VS₂, and their ternary alloys. All samples showed a peak at 2θ = 9°, indicating that d_L was ≈10 Å. The lattice constants of ReX₂, MoX₂, and VS₂ were determined using spin-polarized density functional theory (DFT) calculations, as listed in Table S2 (Supporting Information). The data of ReS₂ matched those of 1T″ phase, with (a, b, c) = (6.49, 6.39, 10.35) Å and (α, β, γ) = (105.10°, 91.82°, 118.82°). Except for the c value, all others are similar to those of JCPDS No. 24–0922; (a, b, c) = (6.455, 6.362, 6.401) Å and (α, β, γ) = (154.04°, 91.60°, 118.97°). For MoS₂, the XRD peaks were matched to those of the 1T′ phase with c = 9.8 Å, based on the calculated lattice constants of (a, b, c) = (6.56, 3.19, 5.88) Å and γ = 119°. The broad peak feature of ReS₂ and MoS₂ was similar. In the case of VS₂, the peaks matched the 3R-stacking 1T (referred to as 3R-1T) phase with a = 3.23 Å and c = 29.2 Å (d_L = 9.7 Å).^[20] After heating the as-grown samples at 400 °C, d_L returned to 6 Å (see Figure S3, Supporting Information). The 1T′ phase MoS₂ and 3R-1T phase VS₂ transformed into 2H MoS₂ and V₃S₄/V₅S₈ phases, respectively. The expansion of d_L was already discussed in the work of (ReV)S₂, using the intercalation of molecular ions (e.g., hydrated NH₄⁺ with an ionic radius of 3.31 Å).^[20] Spin-polarizedDFT calculations were also performed to show that intercalation of NH₃ or NH₄⁺ can occur over the entire composition range to expand the interlayer distance, and it also stabilizes the 3R-1T phase VS₂.^[20]

Samples 6S–10S show the peaks of 3R-1T VS₂, e.g., (006)_3R-1T at 2θ = 18.2°, (102)_3R-1T at 2θ = 32.5°, and (015)_3R-1T at 2θ = 35.4°. 10S exhibits a peak pattern closest to that of VS₂. These XRD patterns suggest a phase transition to 3R-1T VS₂ at x_V = 0.42 (6S), similar to the case of Re_1-xV_xS₂ (x = 0.5).^[20] The right panel shows the magnified XRD peak at 2θ = 54°–62° for (2 $$\overline{4}$$ 1)_1T′′ of ReS₂, ($$\overline{2}$$05)_1T′ of MoS₂, and (110)_3R-1T and (113)_3R-1T of VS₂. The ReS₂, MoS₂, and VS₂ samples have similar lattice constant for the basal planes (a and b): a_1T′′/2 (= 3.25 Å) and b_1T′′/2 (= 3.20 Å) of ReS₂, a_1T′/2 (= 3.28 Å) and b_1T′ (= 3.17 Å) of MoS₂, and a_3R (= 3.23 Å) of VS₂. The negligible peak shift (Δ2θ = 0.4°) of alloy samples 1S–10S is due to the similar lattice constants of these three unitary phases.

We performed spin-polarized DFT calculations for the Re_0.31Mo_0.38V_0.31S₂ composition as a model for 4S (Re:Mo:V = 1:1:1) to understand the crystal phases of the alloys. The (4 × 4 × 1) supercell monolayers (48 atoms) were stacked infinitely; however, interlayer intercalation was not considered because of the complexity. For the V atoms, the effective Hubbard U parameter (U_eff) was set to 1 eV to correlate the localized d electrons. In our previous work, we showed that U_eff = 1 eV led to a better agreement with our experimental lattice constant of VSe₂ than U_eff = 3 eV.^[20] Both 2H- and 1T-like phases were considered for various configurations, as shown in Figure S4 (Supporting Information). The lattice parameters and the total and relative energies are summarized in Table S2 (Supporting Information).

Figure 2b shows the most stable configurations of the 2H and 1T-like phases. They comprise randomly dispersed Re, Mo, and V atoms, which are consistent with the STEM images. The lattice constants of the 1T-like structure are (a, b, c) = (3.23, 3.24, 5.93) Å and (α, β, γ) = (89.72°, 90.66°, 120.18°). Because a ≈ b, α ≈ β ≈ 90°, and γ ≈ 120°, this structure is closer to that of the 1T phase than that of 1T″ or 1T′. In the 1T″ and 1T′ phases, the metal−metal distance alternates along the a/b-axes and a-axis, respectively. In the ternary alloy, there exist only locally asymmetric metal–metal distances with almost no bond alternation, and thus the metal–metal distance was closer to that of the 1T phase (so we referred to it as 1T-like). The 2H structure (a = 3.14 Å and c = 12.35 Å) is more stable than the 1Τ-like structure by 128.978 meV per atom. We plotted the calculated XRD pattern using the lattice constants of the 1T-like structure with an expanded c = 9.8 Å (see Figure 2a). The experimental XRD peaks of 4S agree well with the calculated ones, indicating that the ternary samples prefer the 1T-like structure, which is also consistent with the STEM images.

The XRD patterns of the selenide samples are shown in Figure 2c. The peaks of ReSe₂ matched the 1T″ phase ReSe₂ with (a, b, c) = (6.716 Å, 6.602 Å, 6.90 Å) and (α, β, γ) = (104.15°, 91.82°, 118.15$${\text{\textdegree}}$$), close to the calculated ones. These values are close to those of the reference (JCPDS No. 74–0611, P̄1, a = 6.716 Å, b = 6.602 Å, c = 6.728 Å, α = 104.90°, β = 91.820°, γ = 118.94°). MoSe₂ was in the 2H phase with (a, c) = (3.287 Å, 12.925 Å), corresponding to JCPDS No. 29–0914 (P6₃/mmc). The peaks of 1T phase VSe₂ matched to the reference (JCPDS No. 89–1641; P3_m1) with (a, c) = (3.3587 Å, 6.1075 Å). The d_L value of ReSe₂ and MoSe₂ is 6.5 Å, while that of VSe₂ is 6.1 Å. As the sample number increases, the (001) peak shifts gradually from 2θ = 13.7° to 14.2°, indicating a decrease of d_L due to the phase transition to 1T VSe₂.

The right panel displays the magnified patterns at 2θ = 52°–62°, showing that the peak becomes sharper after 4Se (x_V = 0.5). This peak shifts from 2θ = 55.5° to 55.1° with increasing x_V. The expansion of the basal lattice constants is due to the phase transition to 1T VSe₂. Therefore, the phase transition to 1T VSe₂ probably occurs at x_V = 0.5 (4Se), similar to the 1T″ → 1T phase transition of Re_1-xV_xSe₂ that occurs at x = 0.5.^[19] All alloy samples show the 1T phase peak at 2θ = 34°, which cannot be assigned to either 1T′′ ReSe₂ or 2H MoSe₂. Similar to the sulfide series, the selenide series probably has a 1T-like structure, which is supported by the STEM images. Therefore, as x_V increases, both the sulfide and selenide alloys undergo a transition from the 1T-like to 1T VX₂ phase. The favorable formation of the 1T-like phase in the experiment is probably related to the kinetically controlled solution reaction.

The whole XPS results are shown in Figure S5 (Supporting Information). Figure 3a presents the Re 4f, Mo 3d, and V 2p spectra of the (ReMoV)S₂ alloys and the corresponding unitary samples. The Re 4f peaks correspond to Re-S bonding structures. The alloys exhibited a more redshifted peak than ReS₂, probably owing to the higher metallicity of the 1T-like phase than the semiconducting 1T″ phase ReS₂. As the sample number increases, the Re peaks redshift continuously, indicating a more metallic nature at higher V compositions. The redshift (41.1 eV at x_V = 0.8) is more significant than that (41.4 eV at x_V = 0.8) of binary counterparts, indicating more metallic properties than the binary alloys.^[20] The Mo 3d peaks were resolved into the M1 and M2 bands that correspond to Mo─S bonds and defects, respectively. The S 1s peak appears at 226.5 eV, which is redshifted from that of S⁰ at 228.0 eV. As x_V increases, the M1 band of Mo 3d_5/2 redshifts from 229.0 eV (metallic 1T′ phase MoS₂ and 1S) to 228.8 eV (10S), indicating the increased metallicity at higher x_V. The V 2p peaks were resolved into the V1 (V─S bond with V⁴⁺) and V2 (V─O bond with V⁵⁺) bands; V1 (at 514 eV) and V2 (at 517–518 eV) bands for the V 2p_3/2 peak. In samples 1S and 2S, the band at 519 eV is ascribed to Re 4p_1/2 because of the higher Re compositions. As the sample number increases, the fraction of the V2 band increases. For VS₂, the V1 band almost disappears. We calculated the oxidation number of V atoms using the ratio of V1 and V2 bands and found that it increases from 4.3 (1S) to 4.5 (10S). In the case of Re_1-xV_xS₂, this value is 4.5–4.7. Therefore, sample oxidation is reduced in the ternary alloy samples. The XPS data of (ReMoV)Se₂ alloys showed features similar to those of (ReMoV)S₂ (Figure S5, Supporting Information). The oxidation number of V atoms is lower for the ternary alloy samples than for their Re_1-xV_xSe₂ binary counterparts.

The rightmost panel of Figure 3a corresponds to the valence band spectrum (VBS) of the ternary sulfide alloys. The onset of the VBS can be used to predict the valence band maximum (VBM) below the Fermi energy level (E_F), which corresponds to zero binding energy. All the samples showed VBM = 0 eV owing to the metallic phase. Band integration of the VBS from 0 to 10 eV provided the d-band center (ɛ_d). The ɛ_d value shifted slightly toward E_F (−3.7 eV → −3.5 eV vs E_F) with increasing x_V. The ternary alloys exhibited smaller (negative) ɛ_d values compared to Re_1-xV_xS₂ (–4.2 − –3.8 eV), indicating a more metallic character (Figure S5, Supporting Information). Ternary alloying of selenide also lifted the ɛ_d value (–3.3 − –3.2 eV) toward E_F, where Re_1-xV_xSe₂ has ɛ_d = –3.8 − –3.2 eV, with VBM = 0 eV. The XPS analysis results are summarized as follows. Metallicity increased with x_V, similar to that of the Re_1-xV_xX₂ binary alloys. The ternary alloy nanosheets are more metallic and less oxidized than the binary alloys.

Figure 3b shows the EXAFS above the Re L₃-edge, Mo K-edge, and V-K edge for ReS₂, MoS₂, VS₂, and their ternary alloy samples. The corresponding k- and R-space wavelet-transform (WT) EXAFS spectra are also plotted with the data for Re, Mo, and V foils for comparison. In the X-ray absorption near edge structure (XANES) of Re and Mo, the alloy samples exhibited a lower intensity of the white line than their unitary counterparts, which is correlated with their more metallic character. As x_V increases to 0.5 (sample 7S), the intensity of Re L₃-edge and Mo K-edge decreases, indicating an increase in metallicity. The increased intensity at x_V = 0.8 could be due to the depleted electron density by oxidation. In the XANES of V K-edge, two pre-edge peaks appear at 5.47 keV (marked by A, originated from 1s → 3d transition) and 5.48 keV (marked by B, 1s → 4s-4p hybrid orbitals). Compared to VS₂, the alloy samples exhibited a reduced intensity of the A peak, probably owing to the larger number of electrons occupying the 3d orbitals of the more metallic alloys. The increased intensity of the B peak can be explained by better structural symmetry due to less oxidation.

The Fourier-transform extended XAFS (FT EXAFS) results show the metal–S bonding peaks. The Re-S and Mo-S peaks showed lower intensities than those of the unitary samples, probably owing to their reduced number of bonds and lower crystallinity. By contrast, the higher intensity of the V-S peak resulted from lower oxidation, which increased the number of V─S bonds. Figure S6 and Table S3 (Supporting Information) summarize the fitting curves and fitted values, respectively. The value of d_Re-S decreases from 2.39 Å in ReS₂ to 2.38 Å in 10S, and the average value of d_Mo-S is 2.41 Å for all compositions. The WT EXAFS data showed no shift of the Re-S peak at k = 6.5 Å and the Mo-S peak at k = 5 Å, suggesting negligible changes in electronic structures at the Re and Mo sites, respectively. In the alloy samples, d_V-S is ≈2.32 Å, whereas that in VS₂ is 2.29 Å. In the k-space of WT EXAFS, samples 4S and 7S show the V-S peak at 5–6 Å⁻¹, which is shifted from that of 10S and VS₂ at 4 Å⁻¹. Because the S atom is heavier than the O atom, it induces a large scattering amplitude at high k. Hence, this peak shift is ascribed to lower oxidation, which is consistent with the larger d_V-S value.

Table S4 (Supporting Information) summarizes the electrocatalytic parameters of the ternary alloy and unitary samples toward the HER in an acidic electrolyte (0.5 M H₂SO₄). Figure 4a displays linear sweep voltammetry (LSV) curves of the sulfide samples ((ReMoV)S₂), namely the current density (J) plotted as a function of the applied potential (vs reversible hydrogen electrode (RHE)). The applied potential is equal to the overpotential (η), and η at J = 10 mA cm⁻² is denoted as η_{J = 10}. All alloy samples have lower η_{J = 10} values (104–138 mV) than those of ReS₂ (144 mV), MoS₂ (172 mV), and VS₂ (161 mV) (see Figure S7, Supporting Information). In Figure 4b, the Tafel plot (η vs log J) provided the Tafel slope (b) by a linear fit. Most of the alloy samples exhibited lower b values (53–76 mV dec⁻¹) than those of ReS₂ (103 mV dec⁻¹), MoS₂ (54 mV dec⁻¹), and VS₂ (77 mV dec⁻¹). As a reference, we also measured the η_{J = 10} and b values of commercial 20 wt.% Pt/C to be 17 mV and 30 mV dec⁻¹, respectively. The best HER performance was observed for sample 8S (x_V = 0.67) with η_{J = 10} = 104 mV. The enhancement effect of alloying is consistent with previous observations of (MoNbV)Se₂ and (MoWV)Se₂ ternary alloys and many binary alloys (see Table S5, Supporting Information).^[7-22]

The charge transfer resistance (R_ct) and double-layer capacitance (C_dl) were measured using electrochemical impedance spectroscopy and cyclovoltammetry (Figures S8 and S9, Supporting Information, respectively). The R_ct and C_dl values showed a good correlation with the LSV data. Therefore, the enhanced HER performance of the ternary alloys is due to more efficient charge transfer and larger double-layer capacitance. The electrochemically active surface area (ECSA) was estimated using C_dl, as described in the Supporting Information (Experimental Section). In Figure S7 (Supporting Information), the LSV data were plotted using J_ECSA (defined as J divided by ECSA), showing the enhanced HER performance of ternary alloys compared to ReS₂ and MoS₂. The ternary alloying with x_V = 0.6–0.8 (samples 8S and 9S) is consistently most effective in enhancing the HER performance. The chronoamperometry (CA) data of the best sample 8S show negligible current attenuation at η_{J = 10} after 5 days, which is like Pt/C (Figure 4d). The current level recovered when a fresh electrolyte was used after 4.6 days, indicating that the current decrease came from the pH change. The LSV curve after the 2000th scan showed negligible change (Figure 4e). The XRD, EDX, and XPS data confirmed that both the crystal structure and composition were retained after the CA test (Figure S10, Supporting Information). Furthermore, we evaluated the Faradaic yield (FY) and turnover frequency (TOF) at η = 0.15 V as described in the Supporting Information. The TOF value was estimated as 0.017 H₂ s⁻¹ based on FY = 96% (Supporting Information).

In Figure 4e, the η_{J = 10} value as a function of x_V is overlaid on the data for Re_1-xV_xS₂.^[20] As x_V increases, the HER performance is enhanced owing to the increased metallicity. However, due to sample oxidation at higher x_V, the best HER performance was probably observed for sample 8S. The ternary alloys at x_V = 0.67 (sample 8S) and 0.75 (sample 9S with η_{J = 10} = 107 mV) have lower η_{J = 10} than the values of Re_1-xV_xS₂ (117–128 mV at x_V = 0.6–0.8). At the same x_V value, the Re-rich samples are always more HER-active than the Mo-rich ones;, i.e., η_{J = 10} = 138 and 125 mV for sample 2S (0.25, 0.5, 0.25) and 3S (0.50, 0.25, 0.25), and η_{J = 10} = 132 and 124 mV for samples 5S (0.16, 0.42, 0.42) and 6S (0.42, 0.16, 0.42), respectively. The HER performance data for the annealed binary/ternary samples are plotted in Figures S8,S9 and S11 (Supporting Information), showing a similar feature to that of as-grown samples. The d-band center (ɛ_d) theory has often been used to explain optimal HER performance.^[48-50] As the ɛ_d of the catalyst shifts closer to E_F (i.e., becomes less negative), the catalyst-H interaction becomes stronger owing to the more metallic character. As shown in Figure S5 (Supporting Information), the ɛ_d of the ternary alloys is closer to E_F than that of the binary alloys, suggesting that the ternary alloying increased the metallicity than the binary alloys. Therefore, the alloying of (ReV)S₂ with MoS₂ enhanced the HER performance because of the increased metallicity.

Figure 4f shows the LSV curves of the (ReMoV)Se₂ ternary alloy samples. All alloys exhibited η_{J = 10} = 110–136 mV. The Tafel slope b was 61–70 mV dec⁻¹ (Figure 4g). Those values are lower than those of ReSe₂ (141 mV and 67 mVdec⁻¹), MoSe₂ (130 mV and 73 mVdec⁻¹), and VSe₂ (320 mV and 103 mVdec⁻¹), as shown in Figure S12 (Supporting Information), indicating the positive effects of the ternary alloying. The HER performance of selenide alloys was less dependent on the composition than that of the sulfide alloys. The R_ct and C_dl values did not change significantly (Figures S8 and S9, Supporting Information). Figure 4h compares the η_{J = 10} values at different x for the (ReMo)_1-xV_xSe₂, Re_1-xV_xSe₂, and Mo_1-xV_xSe₂ samples.^{[19, 14]} These ternary alloy samples have a lower HER performance than the Re_1-xV_xSe₂ samples, while they show much enhancement compared to Mo_1-xV_xSe₂. Figure S12 (Supporting Information) shows the J_ECSA data for these (ReMoV)Se₂ samples, consistently showing less enhancement effect than that of Re_1-xV_xSe₂.

As shown in Figure S5 (Supporting Information), the ternary alloying shifts the ɛ_d values closer to E_F than that binary alloying. According to Sabatier's rule, the catalytic performance is optimized when the interaction between H and the catalyst has an intermediate strength.^[48-50] If ɛ_d is too negative or less negative, the catalyst-H interaction would be too weak or strong. Ternary alloying lowers ɛ_d to less negative values (–3.2−–3.3 eV), and thus, the interaction becomes stronger owing to excessive metallicity. Another possible model is based on the number of active sites. As we discussed in our previous work on Re_1-xV_xSe₂, the Se vacancies on the Re atoms are the HER active sites.^[19] The lower HER performance than that of Re_1-xV_xSe₂ is due to the decreased concentration of active Re sites via the alloying with MoSe₂. The higher HER performance than that of Mo_1-xV_xSe₂ is due to the increased concentration of active Re sites via the alloying with ReSe₂.

Ex situ XAFS measurements were performed on sample 8S after applying an overpotential (0–100 mV) under the HER conditions to gain insight into the electronic structures around the metal atoms during the HER. Figure 4i shows the XANES above the Re L₃-, Mo K-, and V K-edge. The Re L₃-edge was redshifted as 𝜂 increased to 100 mV (see inset). The Mo and V peaks also showed redshifts as 𝜂 increased to 100 mV, but less significantly compared to the Re L₃-edge. Nevertheless, all metal atoms became more metallic as 𝜂 increased. The FT EXAFS data with the fitting curves and fitted values are displayed in Figure S6 and Table S3 (Supporting Information). The interatomic distance of Re-S increased due to the reduction, while that of Mo-S and V-S changed very little. This suggests that the Re atoms improve more effectively the HER activity compared to the Mo atoms. This model can explain why the Re-rich samples are more HER active than the Mo-rich ones. Furthermore, the less active Mo sites also support our model for the less enhancement of (ReMoV)Se₂ samples.

For the Re_0.31Mo_0.38V_0.31S₂ model, the total electronic density of states (TDOS) near the Fermi level (E_F) was calculated, as shown in Figure 5a. The 1T-like Re_0.31Mo_0.38V_0.31S₂ ternary alloy exhibits a significant TDOS at E_F, indicating metallicity. The 2H phase alloy is a semiconductor. For comparison, Figure 5a displayed the TDOS of binary alloys Re_0.25V_0.75S₂, Re_0.5V_0.5S₂, and Re_0.25V_0.75S₂, which were reported in our previous work.^[20] As x_V increases, the TDOS at E_F > 0 increases significantly, which could be related to the higher metallicity. Compared to the binary alloys with x_V = 0.25 and 0.5, the ternary alloy at x_V = 0.31 shows a higher TDOS at E_F > 0. Therefore, the incorporation of MoS₂ enhances the metallicity, which is consistent with the experimental results.

The Gibbs free energy change for H adsorption (|ΔG_H*|) is a well-known parameter to predict the optimal HER activity using the condition of |ΔG_H*| ≈ 0. For ReS₂, the basal S atoms are not favorable for H adsorption (ΔG_H* = 1.42 eV).^{[10, 20]} The basal S atoms of 1T VS₂ are HER active, in contrast, with ΔG_H* = 0.072 eV.^[20] In the case of VS₂, significant oxidation reduces the HER performance. As shown in Figure 5b, alloying with 1T-like phase alloy activates six different S sites with ΔG_H* = 0.047–0.31 eV. The S atoms coordinated with (Re, 2 V), (2Mo, V), (3Re), (Mo, 2V), (3Mo), and (ReMoV) are all HER-active. In the case of (ReV)S₂ binary alloys, the S atoms coordinated with (3V) or (Re, 2V) are HER-active, but many of the S atoms coordinated with (3Re) or (2Re, V) are not.^[20] The ΔG_H* values are summarized in Figure 5c, suggesting that the ternary alloys possess more HER active sites than binary alloys, probably due to the more metallicity. Furthermore, the S site coordinated with Re has a higher HER activity than that coordinated with Mo: ΔG_H* = 0.047 and 0.18 eV for (Re, 2V) and (Mo, 2V) coordination, and ΔG_H* = 0.17 and 0.28 eV for (3Re) and (3Mo) coordination, respectively. This shows that compared to Mo, Re can more effectively activate the coordinated S atoms, which is consistent with the experimental results.

3 Conclusion

We synthesized (ReMoV)S₂ ternary alloy nanosheets with full composition tuning using a solvothermal reaction and (ReMoV)Se₂ nanosheets using a colloidal reaction. An increase of x_V (from 0.16 to 0.80) induced a phase change into a metallic 1T phase. Atomically resolved STEM images revealed a homogeneous mixing of the three metals at the atomic scale in the 1T phase. Spin-polarized DFT calculations predicted consistently that ternary alloying produces the metallic 1T-like structure. The XPS and XAFS data that the alloying of (ReV)X₂ with MoX₂ produced a more metallic phase with less oxidation, and the d-band center (ɛ_d) more closely approached E_F. As x_V increases, the nanosheets become more metallic. The (ReMoV)S₂ nanosheets exhibited enhanced HER catalytic activity in 0.5 m H₂SO₄, characterized by η_{J = 10} = 104 mV versus RHE at x_V = 0.67. For the (ReMoV)Se₂ nanosheets, less enhancement of HER was observed. The enhanced HER performance of (ReMoV)S₂ was ascribed to the increased metallicity. Ex situ EXAFS data suggests that the metallicity of Re atoms improves the catalytic activity of (ReMoV)S₂. The calculated ΔG_H* showed that ternary alloying effectively optimized H adsorption on the basal S atoms, supporting the experimental HER data.

4 Experimental Section

Conflict of Interest

The authors declare no conflict of interest.