Synthesis of two-transition-metal sandwich POMs. The strategy for making this unprecedented type of POM involves 3 steps (Fig. 2). All 3 POM structures in Fig. 2 are the X-ray structures. Reaction with the sodium salt of both tungstate and phosphate with nickel nitrate yields the structural analogue of Co4P2, namely [Na2(Ni)2(PW9O34)2]12− (Na2Ni2P2), a complex with exchangeable sodium centers on the outside of the central belt, and Ni(II) centers installed in the internal, buried positions. The occupancy of the Na and Ni centers is confirmed by single crystal X-ray diffraction because Na and Ni(II) have sufficiently different electron densities.29 In the second step, treatment of Na2Ni2P2 with Co(II) salts (CoCl2 works satisfactorily) at pH 5.5 in the presence of KCl, forms Co2Ni2P2 in good yield but with [Co(H2O)6]2+ counterions. These counter-cations refine well crystallographically and are evident in Fig. 2 (lower left structure). However, these counter-cations must be removed because hydrated divalent cobalt ions form multi-cobalt polyhydroxo complexes that are very good WOCs and would thus interfere with comparison of the activities of Co2Ni2P2 and other WOCs. The third step involves slow re-crystallization from concentrated (0.27 M) KCl which replaces the [Co(H2O)6]2+ counter-cations with redox-inactive K+ ones. The single crystal X-ray diffraction (vide infra), elemental analysis (Table S1) and the TGA results indicate that the complete structural formula of the ion-exchanged complex (after step 3), including final crystallization, is a mixed potassium, sodium salt: K8Na2Co2Ni2P2W18O68•30H2O (KNaCo2Ni2P2; see Methods section).
Structures. The X-ray crystal structures of Co2Ni2P2 both with and without [Co(H2O)6]2+ counterions (Fig. 2; lower left and lower right, respectively) reveal that the Co2Ni2P2 polyanion consists of two trivacant B-α-[PW9O34]9− Keggin moieties with four 3d transition metals in the central belt, the classical sandwich polyanion structure first reported by Weakley et al. in 197230 and appearing in scores of publications since. The refinement strongly suggests that this central belt contains two chemically equivalent Ni(II) centers in the internal positions of this Ci symmetry polyanion and two chemically equivalent Co(II) centers in the external, solvent accessible positions defining a rhomb-like Co2Ni2 tetrad (Fig. 1b and Fig. 2 bottom structures). Each Co2+ ion in Co2Ni2P2 coordinates to six oxygen atoms of the two B-α-[PW9O34]9− units.
Infrared spectra of conventional and two-metal sandwich POMs. To assess if FTIR spectra could distinguish these two classes of sandwich POMs, the conventional polyanion with the same four transition metals in the central belt of the complex, from the new, two-transition-element sandwich POMs, the FTIR spectra of Co4P2 and Co2Ni2P2 in the P-O, W-O, and W-O-W stretch regions were compared (Figure 3). These regions are very similar, strongly suggesting that the two complexes are isostructural to one another overall. However, the triply degenerate ν3 vibrational mode of the central PO4 unit in Co4P2 is broadened but not split; whereas, it is split in Co2Ni2P2 and in the (alkali metal)2(transition metal)2 precursor complex, Na2Ni2P2, indicating a greater structural distortion and a consequent lowering of the symmetry around this central heteroatom unit in the latter two polyanions. The peaks in the low energy (<1000 cm-1) region are attributed to the characteristic ν(W-Od), ν(W-Ob-W) and ν(W-Oc-W) absorptions, where Ob = double-bridging oxygen; Oc = central oxygen; and Od = terminal oxygen.
Synchrotron XRAS: confirmation of metal positions and occupancies. While the single crystal X-ray structure of Co2Ni2P2, supported by FTIR spectral and elemental analysis data, suggests the presence of a CoNiNiCo sequence in the central belt of the complex, this structural assignment was uncertain because Co and Ni have very similar electron densities (Z values). Unequivocally and quantitatively distinguishing between adjacent 3d elements in the periodic table in a structure that also contains 18 heavily-scattering tungsten ions as in Co2Ni2P2, is particularly problematical even with recent conventional X-ray diffractometers equipped with strong X-ray sources and improved detectors. Since the thrust of this study is the impact of changing the internal 3d transition metal adjacent to the water-oxidizing external transition metals on the properties of this WOC, the unequivocal confirmation of this CoNiNiCo structural sequence is imperative. As a consequence, we turned to the use of synchrotron XRAS data to address this conundrum, and while not applied previously to POM systems, is ideally suited to quantify both the location and the abundance of the two belt-transition-metals, Co and Ni, in Co2Ni2P2. This technique should also be definitive for structural assignments in myriad potential multiple-transition-metal-containing catalysts for the multi-electron processes central to solar fuel production, HER, OER (WO) and carbon dioxide reduction (CO2RR).
The basis of the XRAS technique, details of its application to this dual-transition-metal identification problem and data collection are given in the SI. Measurements, made at the Advanced Photon Source, were acquired over a wide range of wavelengths that included the two K-edges of cobalt and nickel. Data collections were displaced at either side of the two K-edges (Figure S1). Use of the program GSAS-II31 facilitate use of all the multiple wavelength data to refine the populations of cobalt and nickel atoms at these specific metal sites. High-resolution data (30 keV) gave us an optimal structural model for the refinement of the fractions of both metals. The results are presented in Table 1. In short, the Co(II) and Ni(II) centers are confirmed to be located in 97% and 96% in the outer, solvent-accessible, and inner solvent-inaccessible positions of the central belt of Co2Ni2P2, respectively (Fig. 1b).
Table 1. GSAS-II refinement results of Na2K8Co2Ni2P2W18O68•30H2O
Outer M atoms
Inner M atoms
Catalyzed water oxidations by a sacrificial electron acceptor.
The catalytic efficiency of Co4P2, Ni4P2, Na2Ni2P2 and Co2Ni2P2 for water oxidation was evaluated in dark homogeneous, photodriven homogeneous and electrocatalytic conditions. The dark reactions used [Ru(bpy)3](ClO4)3 as a stoichiometric oxidant, Eq. 2, and were monitored by the UV-Vis spectroscopic kinetics of [Ru(bpy)3]3+ (ε670 = 420 M− 1 cm− 1)32 consumption in 80 mM borate buffer at pH 8.0 using the stopped-flow technique.
4[Ru(bpy)3]3+ + 2H2O → 4[Ru(bpy)3]2+ + O2 + 4H+ (2)
Typical kinetic curves, shown in Figure 4, are not exponential. The addition of 1.0 μM Co2Ni2P2 results in almost complete [Ru(bpy)3]3+ consumption in less than 0.5 s, which is an order of magnitude faster than with 1.0 μM Co4P2 and more than 60 times faster than the self-decomposition rate of [Ru(bpy)3]3+, also shown in Figure 4. For comparison, we also recorded the kinetics of [Ru(bpy)3]3+ reduction catalyzed by 5 μM Co(NO3)2 (brown) and by 5 μM Na2Ni2P2(green). The oxygen yields, based on the initial concentration of the oxidant, [Ru(bpy)33+], increase with catalyst concentrations and reach a plateau of about 70-80% at 5.0 µM catalyst (Co2Ni2P2 or Co4P2). In the presence of Ni4P2 or Na2Ni2P, the rate of [Ru(bpy)3]3+ consumption is the same as in the absence of a catalyst.
Light-driven catalytic water oxidation
The activity of Co2Ni2P2 in visible-light-driven catalytic water oxidation was assessed using a standard approach with [Ru(bpy)3]Cl2 as the photosensitizer and persulfate, Na2S2O8, as a sacrificial electron acceptor (Figure 5).33, 34 The initial rate of O2 formation is commonly, but incorrectly, considered as a direct measure of the catalytic activity, but in actuality this slope is a measure of the initial quantum yield. Under the conditions in Figure 5, the O2 yields and quantum yields in the presence of Co2Ni2P2are reproducibly ~23% higher than those of in the presence of Co4P2. The O2 yields in the Ni4P2 and Na2Ni2P2 reactions are the same as those without a catalyst.The light-induced oxidative decomposition of the photosensitizer, [Ru(bpy)3]2+, by persulfate is the mainside-reaction in the absence of a water oxidation catalyst.
Stability of the Co 2 Ni 2 P 2 water oxidation catalyst in solution. Four different experiments address that the stability of the catalyst are below. All the details, including considerations and controls for each, are in the SI:
1. Co2Ni2P2 was extracted by tetra-n-heptylammonium (THpA)NO3 from the post-reaction solution into toluene. The aqueous layer was evaluated by cathodic adsorptive stripping voltammetry to quantify the amount of Co(II) from POM decomposition. The concentration of Co(II) in NaBi buffer (100 mM, pH = 8.0) and NaPi buffer (100 mM, pH = 8.0) was 5% and 10%. These concentrations are significantly too low for the observed catalytic activity of Co2Ni2P2.35, 36
2. The concentration of Co(II) present in Co2Ni2P2 solutions was also determined by 31P NMR line broadening analysis. We found that the decomposition of 5 µM Co2Ni2P2 to Co(II) after 1 h in 0.1 M NaPi at pH 8.0 did not exceed 14%.37, 38
3. The dependence of catalytic activity on the storage time of Co2Ni2P2 in stock solution in 160 mM NaBi buffer at pH 8.0 was measured by stopped flow kinetics analysis (Figure S11). After 1 h of storage, the activity of Co2Ni2P2 didn’t change, again suggesting the significant hydrolytic stability of Co2Ni2P2 in borate buffer at pH 8.0.
4. The addition of bipyridine (bpy) to the solution of Coaq2+ results in the formation of mono-, bis, and tris-bpy complexes of Co(II) with log10(βi) values of 5.65, 11.25, and 16.05.39 In the solution of 1.0 µM of Co2+ and 9.0 µM of bpy, the concentration of free Co2+ is lower than 0.02 µM. The addition of small amounts of bpy to the Co2+-catalyzed water oxidation (H2O + [Ru(bpy)3]3+) completely shuts down the reaction. If 9.0 µM of bpy is added to the reaction catalyzed by 1.0 µM of Co2Ni2P2, only a very small decrease of [Ru(bpy)3]3+ consumption is observed (Fig. 4). This confirms that Co(II) can’t be the true catalyst in the Co2Ni2P2 solution. However, because the bpy ligand notably destabilizes the Co2Ni2P2 POM framework (removes Co(II)) as it does in the case of Co4V2,38 and unlike in the case of Co4P2,36 the addition of 40 µM bpy results in a visible inhibition of the reaction.
Electrocatalytic water oxidation. Previous studies showed that prolonged exposure to the high overpotential conditions required for electrochemical water oxidation tends to decompose cobalt-containing Keggin-sandwich POMs by electrodepositing cobalt oxide species on the working electrode.35, 36, 38, 40 Embedding Co-POM WOCs in carbon paste has been reported to greatly reduce the hydrolytic decomposition of these catalysts.41, 42 However, short timescale homogeneous cyclic voltammetry experiments illuminate aspects of the catalytic water oxidation activity of Co2Ni2P2. At 1.0 μM this POM produces a rising anodic current from the catalytic oxidation of water with no corresponding reductive current (Figure 6). More importantly, atom equivalent concentrations of aqueous Co2+ and Ni2+ (2.0 μM each) result in a lower oxidative currents. Given that aqueous Co2+ is a known active WOC (active WOC precursor) and Ni2+ is not, this observation strongly suggests that Co2Ni2P2 is a much faster WOC than Co2+. These results are also consistent with the stopped-flow kinetic studies, where we see not only a much faster initial rate of reaction associated with Co2Ni2P2 but also a delayed reaction onset for aqueous Co2+ that is nonexistent in the early-time water oxidations catalyzed by POM WOCs.
Comparison of catalytic activity of Co2Ni2P2 and Co4P2
In order to explain the order of magnitude higher WOC activity of Co2Ni2P2 compared to Co4P2, we studied in detail the kinetics of catalytic [Ru(bpy)3]3+ consumption. First, we attempted to estimate the standard reduction potentials of these POMs. The common electrochemical technique does not work in this case. Neither POMs show any electroactive redox behavior prior to their water oxidation catalytic current. Consequently, we performed potentiometric titration by [Ru(bpy)3]3+ (E = 1.26 V) using a stopped flow technique and measuring the [Ru(bpy)3]2+ concentration at 450 nm (ε = 1.42x104 M− 1cm− 1). The stock solution used for titration was a mixture of [Ru(bpy)3]3+ and [Ru(bpy)3]2+ in a 6:1 ratio. The addition of 0.8-3.0 equivalents of [Ru(bpy)3]3+ to 50 µM of either Co2Ni2P2 or Co4P2 resulted in an immediate (after 0.01 s) increase of absorbance at 450 nm due to a presence of [Ru(bpy)3]2+ in a stock solution. The absorbance grows exponentially with a k ≈ 0.15 s− 1. The rate constant of [Ru(bpy)3]3+ self-decomposition is between 0.02–0.025 s− 1. Therefore, the self-decomposition cannot be ignored, and as a compromise, we measured the concentration of [Ru(bpy)3]2+ 2.0 s after mixing in the titration procedure. In the presence of 50 µM Co2Ni2P2, the yield of [Ru(bpy)3]2+ formed was about 15–20% of added [Ru(bpy)3]3+ (Figure S12). Correspondingly, the first oxidation potential of Co2Ni2P2 must be 20–40 mV higher than that of [Ru(bpy)3]3+. Similar results, within experimental error, were obtained for titration of Co4P2.
Based on this finding, we constructed a kinetic model for the catalytic reduction of [Ru(bpy)3]3+. We rule out a sequential oxidation of these POMs by 4 electrons for three reasons: (a) commonly, the oxidation potentials increase with the number of removed electrons, even with redox leveling; (b) both Co2Ni2P2 and Co4P2 already have high first Co(III)/Co(II) potentials, and (c) [Ru(bpy)3]3+ is unlikely to be able to remove three additional electrons sequentially. Therefore, we assume that two molecules of [Ru(bpy)3]3+ oxidize one POM to form a 2-electron-oxidized intermediate, which then reacts with water. The resulting peroxy-like species is rapidly oxidized subsequently by two [Ru(bpy)3]3+ to form O2 and regenerate the initial form of the POM. The simplified kinetic model is Eqs. 3–6:
POM + [Ru(bpy)3]3+ Δ POM(1) + [Ru(bpy)3]2+ ΔG1 (3)
POM(1) + [Ru(bpy)3]3+ Δ POM(2) + [Ru(bpy)3]2+ ΔG2 (4)
POM(2) → HO-O-CoPOM kc (5)
HO-O-CoPOM + 2 [Ru(bpy)3]3+ → POM + 2 [Ru(bpy)3]2+ + O2 fast (6)
Where POM(1) and POM(2) are the one- and two-electron oxidized forms of the initial POM catalyst, and HO-O-CoPOM represents the key cobalt-peroxy intermediate in the rate-determining step.
The self-decomposition of [Ru(bpy)3]3+ is a complex process. The decay of absorbance at 670 nm is exponential, but the yield of [Ru(bpy)3]2+ product is higher than 95% based on initial [Ru(bpy)3]3+. Bpy self-decomposition in oxidative processes has been thoroughly studied in previous work which shows that the oxidatively damaged bpy ligand, bpy’, which is almost always more electron-rich than bpy itself, is easily oxidized to CO2.32 To take into account the stoichiometry of bpy self-decomposition, we add the reactions 7 and 8 to the kinetic model
[Ru(bpy)3]3+ → [Ru(bpy’)(bpy)2]2+ self-decomposition (7)
Analysis of this model affords the following values (details of fitting procedure and results are described in the SI) for Co2Ni2P2 (Co4P2):
ΔG1 = 29 (21) mV, ΔG2 = -34 (-33) mV, kc = 1.1e + 3 (20) s− 1 (8)
Thus, the main reason for the significant difference in catalytic activity between the two POMs seems to be the difference in rates of Eq. 5, which includes O-O bond formation.
To support this hypothesis, we performed quantum-chemical calculations of the thermodynamic properties of the Co2Ni2P2and Co4P2 intermediates, which are most likely involved in the catalytic cycle. The details are described in SI. The simplified energy diagram is presented in Fig.7. Accordingly, the difference in activity between these POMs derives primarily from the more favorable thermodynamics of the peroxo (O-O) forming step for Co2Ni2P2than for Co4P2. This is analogous to the rate-determining step proposed in other 3d-metal-oxide-based oxidations43 − 45
One key insight from the computations is that the inner and outer metal atoms in the central belt of the POM are strongly antiferromagnetically coupled via superexchange (through the bridging oxo ligands) rather than via direct exchange. Specifically, the second order interaction energy of a bonding Co-O orbital with an inner Co d-orbital lone-pair is ~ 0.5 kcal/mol in Co4P2, while the same interaction of a bonding Co-O orbital with an inner Ni d-orbital lone-pair is ~ 1.7 kcal/mol. This critical difference in the indirect inner metal-outer metal interaction energies explains, at least in part, the observed difference in the reaction rates between Co2Ni2P2 and Co4P2.