How Metal Nuclearity Impacts Electrocatalytic H₂ Production in Thiocarbohydrazone-Based Complexes

Abstract

Thiocarbohydrazone-based catalysts feature ligands that are potentially electrochemically active. From the synthesis point of view, these ligands can be easily tailored, opening multiple strategies for optimization, such as using different substituent groups or metal substitution. In this work, we show the possibility of a new strategy, involving the nuclearity of the system, meaning the number of metal centers. We report the synthesis and characterization of a trinuclear nickel-thiocarbohydrazone complex displaying an improved turnover rate compared with its mononuclear counterpart. We use DFT calculations to show that the mechanism involved is metal-centered, unlike the metal-assisted ligand-centered mechanism found in the mononuclear complex. Finally, we show that two possible mechanisms can be assigned to this catalyst, both involving an initial double reduction of the system.

1. Introduction

Hydrogen has been widely sought out as an alternative energy source [1,2,3], giving rise to a quest for its cost-effective industrial scale production. The search for obtaining optimal homogeneous green catalysts to perform hydrogen evolution reactions (HER) has taken inspiration from understanding how hydrogenases work. The inorganic cofactors of these enzymes are able to mediate the reversible conversion of protons and electrons into hydrogen [4,5,6]. This family of enzymes is very attractive for HER due to their low overpotential requirement, high turnover rates, and use of earth-abundant metals [7,8]. Conventional strategies have explored copying the structural and mechanistic features of hydrogenase active sites, attempting to mimic the same kind of chemistry found in those enzymes. This resulted in the emergence of a large number of biomimetic and bio-inspired catalysts in the past few years [9,10,11]. The best-performing catalysts able to generate molecular hydrogen in organic and aqueous media are usually cobalt- and nickel-based complexes with various ligand skeletons [12,13,14,15]. In the past few decades, a new class of complexes has been developed, where the once metal-centered reactivity was shifted toward a ligand-centered one [16,17]. Prominent examples have been found with complexes based on thiosemicarbazone ligands [18,19,20,21,22,23,24].

Thiosemicarbazone-based complexes are characterized by the redox non-innocent character of the ligand that displays several N- and S-atoms as potential proton relays, a key feature to enhance and promote catalytic activity [25,26]. Such complexes have been found to possess a rather low overpotential requirement, with the ligand playing an active role in HER, and being highly customizable [16,27]. Recently, we have characterized and studied a set of mononuclear nickel-thiosemicarbazone (NiL) complexes, with chemically tuned para substituents in phenyl groups present in the ligand [24,28]. To lower the potential of the redox-active ligand, we attempted to increase the delocalization of the electronic density. The idea was that delocalization would lead to lower overpotential requirements. The use of electron-withdrawing phenyl substituents did not give us the optimal approach we looked for, as some impact was found on the kinetic rates of H₂ formation [29]. Thus, other approaches in catalyst design need to be envisioned.

In this work, we show a new possibility of modifying thiosemicarbazone-based complexes, based on nuclearity. Herein, we introduce an original trinuclear nickel-thiocarbohydrazone Ni₃L₂ complex, being equivalent to the previously reported NiL complexes [18,24]. Compared to its mononuclear counterpart, this multiple metal centre complex displays larger turnover frequency (TOF) values, whilst utilizing higher overpotential requirements for HER. Experimental and theoretical investigations show that the trinuclear equivalent of the mononuclear NiL complex has a purely metal-centred reactivity, involving exclusively the central Ni atom, in contrast with the metal-assisted ligand-centred reactivity identified in its mononuclear form, which uses the nitrogen atoms close to the metal centre. Such change in reactivity likely explains the moderate performance of the new catalyst with respect to the key parameters for catalyst evaluation: the overpotential ($\eta$) and the rate of catalysis (TOF).

2. Results and Discussions

2.1. Synthesis and Single-Crystal X-ray Diffraction Analysis

The ligand was prepared using a one-step procedure (Figure S1). The reaction involves the addition of two equivalents of 1-isothiocyanato-4-methoxy-benzene to one equivalent of thiocarbohydrazide. The reaction was performed in ethanol at reflux for two hours. The products precipitate from ethanolic solution with a 95% yield for the ligand, corresponding to 1,5-Bis(p-methoxyphenylthiocarbamoyl)thiocarbonohydrazi-de (H₆L) and did not require any additional purification. The nickel complex was obtained in a 70% yield by reacting three equivalents of Ni(NO₃)₂ with two equivalents of H₆L ligand in a DMSO/MeOH mixture in the presence of two equivalents of NaSH. The crude compound precipitated as a brown microcrystalline powder and did not require any purification, except washing with methanol (Figure 1 and Figure S2). Both ligand and complex were characterized by elemental analysis, ESI-MS, and NMR spectroscopy (Figures S3–S6). These data are consistent with the indicated trinuclear structure that is the only one found in solution. They also show that the complex is diamagnetic, having a singlet S = 0 ground spin state (see the Section 3). DOSY NMR experiments provided a unique diffusion coefficient of 1.10 × 10⁻¹⁰ m²·s⁻¹ which supports the presence of a single species in solution (Figure S7).

Crystals suitable for X-ray diffraction analysis were obtained from DMSO solution by slow evaporation of the solvent. Details of data collection and structure refinement parameters for the nickel complex are shown in the Supplementary Material (Tables S1 and S2). The single-crystal X-ray diffraction study revealed the compound to be crystallized in the C2/c space group of the monoclinic system. The asymmetric part of the unit cell comprises 1.5 [Ni₃(SH₃L)₂(DMSO)₂] (Ni₃L₂) molecular complexes and one and 2.5 co-crystallized H₂O and DMSO molecules, respectively. The two trinuclear complexes are crystallographically independent, but chemically identical, while in one of them the central metal atom, Ni5, occupies a special position on a two-fold rotational axis. Thus, according to X-ray crystallography, the chemical composition of the compound is in accordance with the following formula: [Ni₃(SH₃L)₂(DMSO)₂] 1.67DMSO 0.67H₂O, where each SL^3- anion acts as a bridging hexadentate ligand. The symmetry equivalent Ni4 and Ni4ⁱ atoms are tetracoordinated in N₂S₂ square-planar geometry, while the central Ni5 metal atom exhibits a slightly distorted N₂O₂S₂ coordination environment. It is worth noting the presence of intramolecular hydrogen bonds that are formed between the N13-H group (donor) and the O9 atom from the coordinated DMSO molecule (acceptor), which consolidate the trinuclear structure (Figure 2).

2.2. Magnetic Studies

Figure 3 shows the dependence of the magnetic susceptibility on the temperature, χT, at 0.1, 1, and 4 T. At room temperature, χT is 1.254 emu·K·mol⁻¹ and does not show any significant temperature dependence while cooling down to about 10 K. Then, it decreases at high fields due to the thermal saturation of the ground state. The dashed line is the best fit for a paramagnetic centre of spin S = 1 and g_iso = 2, in agreement with only one magnetic Ni^II ion (S = 1) being present in the complex. This is expected, as the latter displays an octahedral coordination sphere while the other two external Ni^II ions are square planar compatible with a low-spin configuration (S = 0). Therefore, the magnetic behaviour of Ni₃L₂ is ascribed to the central high-spin Ni^II ion in an octahedral environment. The possible inter-trimer interactions are weak and mostly hidden by the static field effect.

2.3. UV-Vis Spectroscopy

The optical properties of the ligand and the nickel complex were followed by UV/Vis spectroscopy in DMF (Figure 4). The evolution of the spectra upon ligand metalation allowed us to confirm the nickel binding to the ligand and the stability of the complex in solution. The H₆L ligand displays two main absorption bands centered at 270 and 325 nm while the Ni₃L₂ complex displays two main electronic transitions at 268 and 363 nm and two weaker broad bands around 518 and 895 nm (Table S3). The absorption bands present at high energy have been previously attributed to metal–ligand-to-metal–ligand transitions [18,24], while the low-energy features are compatible with d-d transitions.

These results were further supported by TD-DFT calculations performed on a neutral, closed-shell singlet Ni₃L₂ considering no bound DMSO molecules. Indeed, the predicted spectroscopic data provide a calculated spectrum that compares reasonably well with the experimental observations (Figure S10). The low-intensity features around 575 nm are due to excitations from the outer Ni ions and ligands to the central Ni ion. At higher energies, additional excitations on the outer Ni ions are found (Table S4 and Figure S11). The computed UV-vis spectrum thus supports the dissociation of the solvent molecules from the complex once solubilized, with preference for a singlet state in solution (see details in the electronic structure section).

2.4. Electrochemistry

We investigated the redox behaviour of H₆L and Ni₃L₂ by cyclic voltammetry (CV) at room temperature using a glassy carbon electrode in anhydrous DMF with 0.1 M TBAPF₆ as the supporting electrolyte. The electrochemical responses were investigated in the potential range between 0 and −2.5 V vs. Fc^+/0 (Figure 5). The cyclic voltammogram of H₆L was recorded in the absence and in the presence of 100 mM of trifluoroacetic acid (TFA, Figure S12). A scan in the negative direction without TFA reveals one irreversible wave, while in its presence, no catalytic wave was detected. These data confirm that the ligand itself cannot mediate catalytic proton reduction, which is consistent with our previous observations on related thiosemicarbazone ligands [18,24].

In the cathodic region, the cyclic voltammogram of Ni₃L₂ displays two redox systems: one quasi-reversible at −1.35 V $\left(0.5\left(E_{pc}^{*}+E_{pa}^{*}\right), \Delta E_p=86 \mathrm{mV}\right)$ and another irreversible at −2.10 V $\left(E_{pc}^{**}\right)$ vs. Fc^+/0 (Figure S13). The plot of the peak currents versus the square root of the scan rates for the quasi-reversible cathodic wave was found to be linear, indicating that this is a diffusion-controlled process (Figure S14). The redox behaviour of Ni₃L₂ in the cathodic region is reminiscent of the one obtained with another trinuclear nickel complex previously reported by our group [30] for which the reduction events were ascribed as successive metal-based processes. Such a rationale has been further investigated using both chronoamperometry and EPR spectroscopy.

To determine the numbers of electrons involved in each electrochemical event, chronoamperometric experiments on Ni₃L₂ were performed at −1.50 and −2.25 V vs. Fc^+/0 (Figures S15 and S16). The first chronoamperogram can be fitted considering one electron being passed in the system, while at the most negative applied potential, the current response is compatible with a two-electron transfer. EPR spectroscopy was used to further characterize the one-electron reduced species of Ni₃L₂. Using the sample generated upon chronoamperometry at −1.50 V vs. Fc^+/0, we recorded the EPR spectrum on a continuous wave (cw) spectrometer, which resulted in an axial signal (Figure S17). Simulation of the spectrum provided a set of of g-values of 2.012, 2.010 and 2.114, which was found to be compatible with the presence of a single paramagnetic center that was tentatively ascribed to a Ni^I center. These data were then confronted with DFT calculations performed on the one-electron reduced form of Ni₃L₂ considering a doublet state (S = 1/2) and no bound DMSO molecules. The predicted g-values of 2.052, 2.055 and 2.113 were found to be in fair agreement with the experimental ones. The electronic structure of the computed one-electron reduced species shows that the Singly Occupied Molecular Orbital (SOMO) is centered on the central nickel with the spin density being mainly distributed on the same site (see details in the electronic structure section). These data thus support the first reduction of the complex being a metal-based process with the central nickel center going from Ni^II to Ni^I, possibly due to stereochemical reasons.

2.5. Electrocatalytic Production of H₂

To explore the activity of Ni₃L₂ for catalytic proton reduction, we investigated its electrochemical response recorded at a glassy carbon electrode in DMF under argon in the presence of TFA as the proton source (pKa = 6.00 ± 0.3 in DMF) [31,32]. The addition of increasing amounts of acid triggered the appearance of two peak-shaped waves with potentials starting at about −1.10 V vs. Fc^+/0, suggesting potential-dependent mechanisms. The current response of the two waves is proportional to the concentration of protons in the electrolyte solution (Figure S18). While the peak potential of the more negative process is shifted from −1.99 V to −2.47 V vs. Fc^+/0, the potential of the first process remains unaffected (Figure 6). We only investigated the process occurring at the more positive potential with a midwave potential $E_{1/2}^{cat}$of −1.52 V versus Fc^+/0. In this case, the peak current increases with the acid concentration, and plateaus for concentration and scan rates higher than 100 mM and 2 V·s⁻¹, respectively (Figure S19).

Control experiments were performed in the absence of the complex and the resulting CV clearly indicates the role played by the nickel complex in the observed current enhancement (Figure 6 and Figure S18). We verified that the observed catalytic response can be attributed to the complex in solution and not due to adsorbed species on the surface electrode. To do so, dip and rinse tests were performed. After each electrocatalytic measurement, the glassy carbon electrode was removed from the solution and immersed into a fresh DMF solution, followed by the addition of 100 mM TFA. The electrochemical response recorded upon sweeping the potential in the cathodic region only showed a residual background current, and upon the addition of TFA, only the acid blank was observed (Figure S20). These results thus confirm that no active adsorbed species is formed during electrocatalysis and indicate that the catalytic activity is due to the dissolved complex Ni₃L₂.

To confirm that the current response in the presence of acid does correspond to the catalytic reduction of protons into H₂, we performed a bulk electrolysis experiment on a 100 mM solution of TFA in 8 mL of DMF at a constant potential in the presence of Ni₃L₂ (1 mM). Gas products were monitored by continuous in-line GC analysis. A potential of −1.70 V versus Fc^+/0 was applied to the mercury-pool working electrode, resulting in the immediate formation of hydrogen. These experiments confirm that catalysis is homogeneous in nature, as the use of the mercury-pool electrode allowed us to rule out the possibility that proton reduction occurs due to the formation of nickel nanoparticles since the latter would amalgamate on the electrode surface. During a 4 h experiment, Ni₃L₂ mediated hydrogen evolution with an 82% faradaic efficiency and a turnover number of 31 was reached (Figure S21). In similar conditions, our reference catalyst NiL produced hydrogen during 16 h with 80% faradaic efficiency and a turnover number of 21. These data first indicate that the trinuclear catalyst Ni₃L₂ is probably less stable than the parent catalyst and likely points to degradation of the complex during turnover (Figures S8 and S22). Nevertheless, the trinuclear complex is able to produce more hydrogen on a shorter timescale when compared with its mononuclear counterpart NiL, which would support that nuclearity is playing a role in partially enhancing the catalytic activity of nickel thiosemicarbazone complexes. This is further investigated in terms of mechanistic implications using computations.

2.6. Benchmarking of Performances

A value of 534 mV was calculated for the overpotential requirement, $\eta$, of Ni₃L₂ for proton reduction in the presence of TFA in DMF (see details in the Supplementary Material) [31]. The rates and the performance of the catalyst during the hydrogen evolution reaction were quantified. The TOF value of Ni₃L₂ was evaluated using the catalytic cyclic voltammograms recorded in DMF at different scan rates ($v$ = 0.5, 1, 1.5, and 2 V·s⁻¹) for various amounts of TFA (10 to 100 mM). The TOF value can be calculated from the ratio between the peak current of the catalyst in the absence ($i_p$) and presence of substrate ($i_{cat}$) as given by Equation (1) [33]:

$$TO{F}_{max}=1.938\times \nu \times {\left(\raisebox{1ex}{$i_{cat}$}\!\left/ \!\raisebox{-1ex}{$i_p$}\right.\right)}^2$$

(1)

In the case of Ni₃L₂, the ratio $\left(\raisebox{1ex}{$i_{cat}$}\!\left/ \!\raisebox{-1ex}{$i_p$}\right.\right)$ varies linearly with the inverse of the square root of scan rate ($\sqrt{v}$), which confirms the validity of the above equation (see details in the Supplementary Material and Figure S23). Using this approach, and considering a first order reaction in acid and in catalyst (Figures S24 and S25), a maximum TOF value of 2072 s⁻¹ was obtained in the presence of 100 mM TFA.

Based on this latter value, a mid-wave potential $E_{1/2}^{cat}$ of −1.52 V versus Fc^+/0, and an apparent equilibrium potential of the H⁺/H₂ couple (−1.01 V vs. Fc⁺/Fc) [32] at 1 M TFA concentration, we could derive catalytic Tafel plots correlating the turnover frequency (TOF) to the driving force of the reaction, the overpotential ($\eta$), that is related to H₂ evolution under the conditions used [34]. Focusing on Ni₃L₂ (purple trace, Figure 7), we observe that this catalyst displays a significant catalytic activity (log(TOF) > 2) for moderate overpotential values. While it does not surpass benchmark molecular catalysts such as cobaloxime (blue trace, Figure 7) [14], tetraphenylporphyrin iron [35] or nickel bisdiphosphine (orange trace, Figure 7) [36,37,38], it performs better than the parent thiosemicarbazone complex NiL (green trace, Figure 7) as far as the turnover frequency is concerned.

When compared to each other, Ni₃L₂ and NiL feature quite different catalytic performances. For instance, while their faradaic yields appear comparable with a value around 80% for both catalysts, Ni₃L₂ displays a higher TON value (31 vs. 21) and thus produces more H₂ on a shorter timescale [18]. These data suggest that Ni₃L₂ is a faster catalyst (TOF = 2072 s⁻¹) than its mononuclear counterpart (TOF = 90 s⁻¹) [24], but this is highly detrimental to the driving force of the reaction, with increased overpotential values for Ni₃L₂. Taken together, these results support the fact that changing the nuclearity of the catalyst affects its electronic structure and, as such, its redox behaviour, which influences its catalytic performance.

2.7. Electronic Structure: Solvent Binding and Redox Processes

The electrochemical investigations shown above were performed in DMF. This raises the question as to whether DMSO molecules are stably bound to the central nickel atom, Ni_c, in DMF, or if they dissociate. To address this issue, we performed DFT calculations to assess the electronic structure of the complex (see the Section 3 for details). All calculations performed here concern Ni₃L₂ to allow for a more straightforward comparison with the related mononuclear NiL complex we have previously studied [29]. In the solid state, Ni₃L₂ has an octahedral Ni_c at the centre, distinct from the two peripheral metal centers, Ni_s, which are square planar. Such geometry can render it paramagnetic, and we indeed show this is the case upon measuring the magnetic susceptibility of the complex (Figure 3). However, if DMSO dissociation occurs, two cases might arise: a tetrahedral or a square planar geometry for Ni_c can be formed. The former will keep the Ni_c paramagnetic, whilst the latter will not carry any magnetic moment. Thus, we need to investigate the energetics of Ni₃L₂ in two cases: including and excluding the DMSO molecules initially bound. We performed free energy calculations for the complex, considering the initial neutral state as being either triplet (open shell, paramagnetic Ni_c), or singlet (closed shell diamagnetic Ni_c, Table S5).

When considering the DMSO molecules initially bound to Ni_c, geometry optimization shows they remain bound only in the triplet state, whereas they dissociate in the singlet state, leading to a distorted square planar geometry. We compared the free energies for each state, and saw a preference of 3.6 kcal·mol⁻¹ for the singlet over the triplet state. To confirm the preference of singlet over triplet, we performed the same calculations without DMSO molecules. Here, DFT shows a slight preference of 1.8 kcal·mol⁻¹ for the singlet over the triplet state, and geometry optimization again resulted in a distorted square planar geometry around the Ni_c (Table S5). Thus, our calculations show that upon solvation of the complex, one can expect DMSO dissociation and the formation of a singlet ground spin. These findings are thus in agreement with the NMR data, which showed that the complex is diamagnetic.

We proceeded with DFT calculations to assess the spin states associated at each oxidation state for Ni₃L₂ upon reduction, as well as the redox potential associated with each process. We investigated the spin states with the lowest energies and used them to compute the free energy differences, which were then converted to redox potentials versus Fc^+/0. Our calculations show that, initially, the complex starts at a neutral singlet state, proceeding through a doublet state after the first one-electron reduction (Table S5, first entry), and finishing with a doublet after a second two-electron reduction (Table S5, fifth entry). We obtained reduction potentials of −1.06 V and −1.97 V, which compare fairly well with the two redox systems observed at −1.35 V and −2.10 V in Figure 5 (Table S6). The fair agreement between the experimental and calculated redox potentials allow us to assign each peak in the voltammogram from Figure 5 to successive one- and two-electron reduction processes, which is consistent with the results obtained from chronoamperometry (Figures S15 and S16). It is worth noting that we have also considered the possibility of a two-electron reduction process and the calculated potential for this process was found to be −1.47 V (Table S6, fourth entry and Figure S26). Although no clear peak was experimentally observed at this voltage, we will later discuss that such a process can take place during catalysis, when potentials below −1.80 V are applied.

To investigate the nature of these experimentally observed processes, we plotted the spin density distributions and singly occupied molecular orbitals (SOMOs) for each of the oxidation states involved in the redox potential calculations (Figure 8). Our results show that they correspond to successive metal-centred reductions (Ni^II® Ni^I), different from the ligand-centred processes established with the mononuclear NiL complex [18,24]. Compared to the computed structure for the first one-electron reduced species, the geometry of the complex upon a subsequent two-electron reduction produced a more distorted Ni_c center whose coordination geometry was more tetrahedral. Such a rearrangement would explain the irreversibility observed experimentally for the second redox system in Figure 5.

Mulliken spin population analysis shows that Ni_c holds larger values of spin density than the other Ni sites (Table S7). Besides, as we shall see next, when comparing the energetics of protonating different sites of the system, Ni_c is indeed the most favourable site for protonation. This points toward a metal-centered reactivity, in contrast to the ligand-centered reactivity seen in the mononuclear NiL complex previously studied [29]. The presence of multiple metal centres could then not only play a role in delocalizing electrons in the complex, but also have a direct impact on catalysis.

2.8. Mechanistic Considerations: Relative pK_as and Reaction Pathways

In the context of H₂ production, studying a catalytic mechanism implies knowing where protons will bind in the catalyst. Thus, we need to probe the energetics associated with protonating every potential site. Since obtaining absolute $p{K}_a$ values can be very complicated, we calculated relative $p{K}_{a}s$from free energy differences between protonating different sites for distinct oxidation states. In line with our mononuclear studies [18,29], and taking symmetry into account, we chose nine possible sites: the three Ni centers, the three N atoms adjacent to the Ni centers, and all S atoms. We calculated the relative $p{K}_{a}s$ by performing free energy calculations of the protonated complex at each site, and took the structure with Ni_c protonated as our reference. Results are shown in

Table 1. Relative pK_a values for Ni₃L₂ after the first and second reduction steps. Values are relative to the free energy of protonating the Ni_c atom. ‘X’ corresponds to cases where geometry optimization of the protonated complex shows cleavage of the complex. Cleavage cases are not considered, as they will affect the ability to restore the catalyst at the end of the catalytic cycle.

Proton Site	Ni3L20	Ni3L2−
Nic	0.0	0.0
Nis	−1.3	−5.3
Sc	−1.3	−7.8
Ssc	−4.6	X
Ss	X	X
Sd	−13.6	X
Nc	X	X
Nsc	−6.4	−8.3
Ns	−5.5	X

and Table S8, with large positive values corresponding to the most preferable sites for protonation. UV-Vis experiments in the absence and presence of trifluoroacetic acid displayed almost identical features, indicating that the structure of the complex is unaffected by the addition of the proton source (Figure S9). Since it is a strong acid, our complex will not become protonated spontaneously. Thus, a mechanism starting with a chemical step (CE and CC) is unlikely to occur and we ultimately consider the first step as being an electrochemical one.

With that in mind, we now had to probe how and where the first protonation step occurs. To investigate the likelihood of the first proton transfer occurring upon successive reductions, we attempted to obtain the energetics associated with the reactant and product states of the proton transfer between Ni₃L₂ and TFAH. We started by putting TFAH in H-bonding positions with Ni_c, the preferred protonation in terms of $p{K}_{a}s$, and we conducted geometry optimization of the system in the product state, i.e., with the proton attached to the complex. When performing such calculations with the complex reduced up to two times, stable product states were found. We then proceeded with the analysis of this first proton transfer. To do so, we performed a potential energy surface (PES) scan where we constrained the distance between the transferred H atom and proton donor O from TFAH. The idea was to obtain an approximation to the transition state, which corresponds to the highest energy point of the scan (Figure S27). From there, we performed harmonic analysis to obtain the free energies at the reactant, transition, and product states of the proton transfer from TFAH to Ni_c. The transition state had a single negative frequency whose normal mode correspondedto the proton transfer reaction, thus indicating it was a reasonable approximation. We obtained an activation barrier ΔG^‡ = 11.5 kcal·mol⁻¹, with a reaction free energy ΔG⁰ = −0.2 kcal·mol⁻¹ after the first reduction, and ΔG^‡= 6.1 kcal·mol⁻¹ and ΔG⁰ = −0.2 kcal·mol⁻¹ after the second reduction. These energies correspond to rather fast intermolecular proton transfer reactions, meaning that they should not limit the kinetics of H₂ formation.

Our previous results show that the first event must be an electron transfer, else proton transfer becomes impractical. We needed to figure out whether the second step was an electron or proton transfer (EE vs. EC). For the former, redox potential calculations were already performed in the previous section: we obtained reduction potentials of −1.06 V and −1.83 V for successive one-electron reduction processes. When inspecting the cyclic voltammogram of the complex in the presence of TFAH (Figure S18), we notice two main features: a shoulder around −1.72 V, and a peak at −1.99 V. Now, for an EC-sequence, we have the same one-electron reduction at −1.06 V. We calculated the redox potential for a second reduction occurring after the first protonation, i.e., ECE-sequence, resulting in −1.19 V. An ECEC-sequence for HER cannot capture the main feature of the CV at −1.52 V. Thus, such a mechanism cannot be described in terms of experimental electrochemical data. Additionally, an ECCE-type mechanism would yield a second redox potential that was even less negative than the −1.19 V observed for the second reduction in the ECEC sequence. Thus, it is not a sequence to be observed under the experimental condition used for H₂ production in this work.

To further investigate the likelihood of the first proton transfer occurring between the first and second reduction (EC vs. EEC), we calculated the $p{K}_a$ difference between protonating Ni₃L₂ at Ni_c and TFAH, in line with previous works [39,40]. The idea is that a positive $\Delta p{K}_a$ means that the complex is thermodynamically more likely to take the proton from TFAH, whereas this is not the case for a negative value. The $p{K}_a$ difference was calculated as $\Delta p{K}_a=p{K}_a\left(N{i}_3{L}_{2}H\right)-p{K}_a\left(TFAH\right)$. We obtained $\Delta p{K}_a$(EC) = −4.5 and $\Delta p{K}_a$ (EEC) = 6.3. Such difference supports the fact that the catalyst should be doublyreduced before performing a proton transfer.

To finalize this study, we needed to investigate the second proton transfer event. In principle, this could occur after the first protonation (EECC), or even after a third reduction (EECEC), depending on the potential requirements for such a process. We calculated the redox potential for this third reduction, obtaining −1.47 V. Unless such a process would have a very large reorganization energy and/or very low electronic coupling with the electrode, it could take place within the experimental conditions applied for H₂ production. Since calculating such properties is beyond the scope of this work, we shall consider it as a possibility. For the second proton transfer, we attempted to initially protonate all possible sites, but we found that protonation of Ni_c resulted in the immediate formation and release of H₂. When putting TFAH in close proximity to the protonated Ni_c, geometry optimizations led to a spontaneous proton transfer to form H₂, followed by its release. This occurred for both the EECC and EECEC cases, with a favorable free energy difference (−0.47 and −0.22 eV for EECC and EECEC, respectively). We summarize all the calculated redox potentials, pK_a values and free energy differences for proton transfer events in Figure 9. Our results thus support that there are two possible reaction mechanisms to mediate proton reduction into hydrogen when using Ni₃L₂: (i) either an (EECC) pathway involving two electrochemical and two chemical events, or (ii) an (EECEC) involving three electrochemical and two chemical events. In both cases, the complex displays a metal-centered reactivity to promote HER which differs significantly from the conclusions raised with the parent complex NiL.

3. Materials and Methods

3.1. Synthesis and Characterization

All solvents and chemicals were purchased from Sigma Aldrich (Merck KGaA, Darmstadt, Germany) and used without further purification. ¹H NMR spectra were recorded on Bruker 400 MHz Avance III Nanobay (Bruker, Wissembourg, France). Chemical shifts for ¹H and ¹³C NMR spectra are referred to TMS or the residual protonated solvent. 2D DOSY experiments were acquired on a BRUKER AVANCE III 600 MHz NMR spectrometer with a Double Resonance Broad Band Probe (BBFO, Bruker, Wissembourg, France) at 300 K, using a pulse sequence that incorporated bipolar gradient pulses and a longitudinal eddy current delay (LED). A total of 32 gradient values were lineary sampled from 6% to 95%. A total of 32 scans were acquired with 32k data points, for a total acquisition time of ca. 60 min. The gradient pulse duration (∂/2) and the diffusion time (∆) were set to 1.8 ms and 150 ms, respectively, with a gradient recovery delay of 0.1 ms and a LED of 5 ms. The DOSY spectra were simply obtained by using the single-exponential fitting routine of the EDDOSY feature of the TopSpin software from Bruker. Elemental analysis was performed on a Thermo Finnigan EA 1112 instrument. The results were validated by at least two measurements. ESI-MS analyses were performed using a SYNAPT G2 HDMS (Waters) spectrometer equipped with a pneumatically assisted atmospheric pressure ionization (API) source. Samples were ionized by negative electrospray mode as follows: ion-spray voltage: −2.27 kV; sampling cone: 20 V; nitrogen flux (nebulization): 100 L·h⁻¹. High resolution mass spectra (HRMS) were performed on a QStar Elite (Applied Biosystems SCIEX) spectrometer equipped with a pneumatically assisted atmospheric pression ionization source (API). The samples were dissolved in DMF (300 μL) and placed in a methanol/3 mM ammonium acetate solution. The results were validated by three measurements. UV/Vis spectra were recorded on a Varian Cary 50 spectrophotometer. Continuous-wave EPR measurements were performed using a conventional Bruker EMX spectrometer operating in the X-band with microwave frequencies of about 9.624 GHz. This spectrometer is equipped with a He-flow cryostat (ESR900) and a cryogen-free cooler (Bruker Stinger). The measurements were performed at 50 K, a microwave power of 20 mW, and modulation amplitude of 6.7G. Particular attention was given to avoiding the saturation of the signal while keeping a good signal-to-noise ratio. The EPR spectrum simulations were performed using the Matlab program package Easyspin [41].

3.2. Synthesis of H₆L

The thiocarbohydrazide (2.5 mmol 0.265 g) was dissolved in 20 mL of ethanol, and 10 mL of 1-isothiocyanato-4-methoxy-benzene (5 mmol, 0.825 g) solution in 10 mL ethanol was added dropwise. The mixture was refluxed for 2 h. After the reaction mixture was cooled to room temperature, the white product was filtered and washed twice with ethanol. The crude product was used in the next step without further purification. Yield 95%. ¹H NMR (400 MHz, DMSO-d6, 298K, TMS), δ: 10.36–9.17 (m, 6H, NH), 7.33 (d, ²J = 8.5 Hz, 4H, o-CH), 6.9 (d, ²J = 9 Hz, 4H, p-CH), 3.75 ppm (s, 6H, O-CH₃). ¹³C NMR (75 MHz, DMSO-d6, 298K, TMS), δ = 55.2, 113.4, 126.8, 131.7, 156.80, 183.9 ppm. Elemental analysis for C₁₇H₂₀N₆O₂S₃: Calcd (%): C 46.77, H 4.62, N 19.25, S 22.03; found: C 45.71, H 4.54, N 20.29, S 22.05. HR ESI-MS [M-H]- for C₁₇H₂₀N₆O₂S₃: Calcd: 435.0737; found: 435.0741.

3.3. Synthesis of [Ni₃(SH₃L)₂(DMSO)₂] (Ni₃L₂)

To the suspension of H₆L (0.225 mmol, 0.1 g) in 5 mL of methanol was added solid sodium hydrosulfide (0.112 mmol, 0.0062 g). The Ni(NO₃)₂ (0.337 mmol, 0.098 g) dissolved in 5 mL of methanol was added dropwise to this solution. DMSO was added to fully dissolve the suspension. The solution was stirred at reflux for 2 h. Slow evaporation from this solution gave single crystals suitable for crystallographic analysis with a 70% yield. ¹H NMR (600 MHz, DMSO-d6, 298K, TMS), δ: 8.75 (s broad, 6H, NH), 7.07–6.83 (s, 16H, Ph-H), 3.85 (s, 6H, -OCH₃) 3.70 ppm (s, 6H, -OCH₃). ¹³C NMR (150 MHz, DMSO-d6, 298K, TMS), δ = 55.1, 114.2, 116.1, 116.5, 125.4, 132.2, 153.8, 157.4 ppm. Elemental analysis for C₄₂H₅₈N₁₂O₈S₁₂Ni₃: Calcd (%): C 35.49, H 4.08, N 11.83, S 27.04; found: C 35.56, H 4.10, N 11.24, S 27.11. HR ESI-MS [M-H]- for C₃₄H₃₄N₁₂O₄S₈Ni₃: Calcd: 1104.8542; found: 1104.8548.

3.4. X-ray Crystallography

X-ray diffraction measurements were carried out with a Rigaku Oxford-Diffraction XCALIBUR E CCD diffractometer equipped with graphite-monochromated MoKα radiation. The unit cell determination and data integration were carried out using the CrysAlis package of Oxford Diffraction [42]. The structures were solved by Intrinsic Phasing using Olex2 [43] software with the SHELXT [44] structure solution program and refined by full-matrix least-squares on F² with SHELXL-2015 [45] using an anisotropic model for non-hydrogen non-disordered atoms. All H atoms were introduced in idealized positions with d_CH = 0.96 Å. The positional parameters of the disordered atoms were refined using available tools (PART, DFIX, and SADI) of SHELXL. The molecular plots were obtained using the Olex2 program. The crystallographic data and refinement details, as well as bond lengths and angles, are presented in the Supplementary Material.

3.5. SQUID Magnetometry

Magnetic characterization has been performed using a conventional SQUID magnetometer MPMS-XL from Quantum Design working at a magnetic field up to 5 T and temperature down to 2 K. The samples (powder) are filled in polypropylene sleeves, then sealed in order to remove the maximum dioxygen, which gives a signal around 50 K (antiferromagnetic transition). The diamagnetic contribution of the sample holder was removed as well as the diamagnetism of Ni₃L₂. To avoid field-dependent effects (such as ferromagnetic impurities), the temperature dependence of the susceptibility was recorded for three magnetic fields: 0.1, 1, and 4T. Considering non-interacting inter-trimer paramagnetic centres, we fitted χT by 1/8.S(S + 1).giso² in CGS units.

3.6. Electrochemistry

Cyclic voltammetry experiments were performed using a BioLogic SP300 potentiostat and a three-electrode setup consisting of a glassy carbon working electrode, a platinum wire counter electrode, and a leakless (Ag/AgCl) reference electrode. Ferrocene was used as an internal standard with E⁰(Fc^+/0) = 0.53 V versus Ag/AgCl. Chronoamperometric experiments were conducted with another three-electrode setup composed of a 1 mm glassy carbon electrode as the working electrode, a platinum wire as the counter electrode, and a leakless Ag/AgCl reference electrode. All studies were performed in deoxygenated DMF containing NBu₄PF₆ (0.1 M) as the supporting electrolyte. Controlled potential electrolysis experiments were carried out in a two-compartment cell. The volume of solution (DMF, 0.1 M NBu₄PF₆) used in the working compartment of the cell was 8 mL. The working electrode used was a pool of mercury, separated from the coiled platinum wire counter electrode by a porous frit. Bulk electrolysis solutions were purged with N₂ gas for at least 20 min prior to electrolysis and stirred throughout the bulk electrolysis experiment. During the experiment, the cell was continuously purged with nitrogen (5 mL·min⁻¹) and the output gas was analyzed at two-minute intervals in a Perkin–Elmer Clarus 500 gas chromatographer using a previously described setup [46]. Dip and rinse tests were performed after electrocatalytic measurements. The glassy carbon working electrode was removed from the solution and immersed into a fresh DMF solution followed by the addition of 100 mM of acid. The electrochemical response was then recorded upon sweeping the potential in the cathodic region.

3.7. DFT Calculations

The methodology is essentially the same used in our previous work with NiL [29]. We used ORCA 4.2.1 [47,48], with BP86 as our functional of choice [49,50], and def2-TZVP(-f) as our basis set [51]. Full geometry optimizations were undertaken by taking advantage of the resolution of the identity (RI) approximation in the Split-RI-J variant [52] with the appropriate Coulomb fitting sets [53]. Increased integration grids (Grid4 in ORCA convention) and tight SCF convergence criteria were used. According to the experimental conditions, these calculations were performed in DMF solvent by invoking the Control of the Conductor-like Polarizable Continuum Model (CPCM) [54]. We considered various possible spin configurations, including open-shell singlets. For the latter, we performed broken-symmetry DFT calculations, using the “FlipSpin” feature of ORCA [55,56,57]. Free energy differences were extracted from the Gibbs free energies obtained from numerical frequencies calculations. Electronic structures were obtained from single-point DFT calculations using the hybrid functional B3LYP [58,59] together with the def2-TZVP(-f) basis set [51]. TD-DFT calculations [60,61,62] used the Tamm-Dancoff approximation [63,64] and included triplet excitations. A total of 100 roots were calculated with the B3LYP [58,65] density functional under the RIJCOSX approximation with the def2-TZVP(-f) [51] and def2/J basis sets and the CPCM solvent model. [54] The grid sizes were increased to Grid7 and GridX9 in ORCA nomenclature. While other density functionals including PBE0 [66] and TPSSh [67] were tested, B3LYP gave the most favorable agreement with experiment. All the calculated reduction potentials (E) were relative to Fc+/0. The calculated Gibbs free energy difference using the BP86 functional to account for ferrocene in DMF was ∆G = 4.87 eV. The relative acidities $(\Delta {\mathrm{pK}}_{\mathrm{a}}={ \mathrm{pK}}_{\mathrm{a}}\left({\mathrm{Ni}}_3{\mathrm{L}}_2\mathrm{H}\right)-{\mathrm{pK}}_{\mathrm{a}}\left(\mathrm{TFAH}\right)$ were computed with respect to TFAH in DMF. Spin density plots, molecular orbitals, and difference density plots for each transition were generated using the orca plot utility program and were visualized with the Chemcraft program [68].

4. Conclusions

This work reports on an original trinuclear nickel complex, [Ni₃(SH₃L)₂(DMSO₂)], (Ni₃L₂) that was experimentally characterized and evaluated as a potential electrocatalyst for hydrogen evolution. Ni₃L₂ exhibited enhanced catalytic performances for the electrochemical reduction of protons into hydrogen. Interestingly, when compared to its mononuclear analogue NiL, the trinuclear complex features a comparable faradaic yield but displays higher TON and TOF values, and is able to produce more hydrogen on a shorter timescale. Ni₃L₂ is a faster catalyst but it comes with the cost of requiring a larger overpotential, the drawback to these improved performances lying thus in the driving force of the reaction. These results support the fact that the nuclearity of the catalyst affects its electronic and redox properties, which influences its catalytic performances.

The theoretical investigations showed that the reactivity of the trinuclear complex shifted with respect to that of the mononuclear catalyst. While the NiL exhibited a metal-assisted ligand-centred reactivity making use of the N-atoms close to the metal centre [29], only the central nickel ion is involved in the case of Ni₃L₂. Such change in reactivity might explain the improved efficiency of the complex. On the other hand, the increase in potential requirements of Ni₃L₂ is probably associated with the electronic density being strictly localized at the metal centre, losing the benefit from delocalization onto the ligand, as observed with the mononuclear NiL catalyst.

Replacement of the two catalytically-inactive peripheral nickel centres of Ni₃L₂ with more electron-rich and redox-innocent metals such as zinc might improve the catalytic performance of the system by diminishing the overpotential requirements. Another important question to address is whether a metal-centred reactivity is indeed responsible for better turnover rates. For comparison, the Co-based thiosemicarbazone complex has been shown to be an efficient HER catalyst requiring smaller overpotentials than its Ni counterpart [22]. Synthetizing a trinuclear Co complex could further probe how focusing density at the metal centre actually impacts the electrochemical properties of the system and assists in the catalytic turnover rates.