Introduction

Molecular metal oxides, so-called polyoxometalates (POMs)1,2, are an emerging class of inorganic materials. Their well-defined molecular structure and tuneable properties and reactivity have made POMs ideal analogues for the corresponding solid-state metal oxides3,4,5,6. This has led to applications in areas including energy conversion and storage7,8, molecular magnetism9,10, and catalysis11,12. Typically, POM reactivity can be tuned by functionalization with functional metal centres:13,14 In molecular magnetism, the incorporation of lanthanides in POMs is a widely used principle15, while energy conversion and storage schemes often use transition metal functionalized POMs16, In catalysis, the introduction of Lewis-acidic sites in POMs is a well-known concept17.

Thus, robust and broad-scope synthetic concepts for the metal-functionalization of POMs are critical for the targeted development of new functional compounds. For polyoxotungstates (and, to a lesser degree, for polyoxomolybdates), controlled metal-functionalization is possible using so-called lacunary cluster derivatives where one or several of the original metal ions (W or Mo) have been hydrolytically removed from the cluster shell14. Selective binding of suitable metal ions at the resulting vacancy then leads to the targeted metal-functionalized species14,18. This approach has led to ground-breaking molecular components, e.g., for catalysis11,16,19 molecular electronics15,20,21 and medicine22,23. In contrast, in polyoxovanadate (POV) chemistry, controlled and predictable metal-functionalization approaches are still in their infancy8,24,25,26,27,28, and the field is currently dominated by a combination of empirical knowledge and serendipity8,24,25,29. Thus, the development of controlled approaches for the predetermined and selective metal-functionalization of POMs is critical to enable knowledge-based materials design.

Early studies by Streb and co-workers have used a placeholder strategy for the predictable metal-functionalization of POVs. This approach uses the dodecanuclear species {(NMe2H2)2[V12O32Cl]}3-, (= (DMA)2{V12}, DMA = dimethyl ammonium), where two vacant metal coordination sites are blocked by hydrogen-bonded DMA cations. In-situ exchange of these cations with a variety of metals is possible, leading to the mono- or (more rarely) di-metal-functionalized species ({MV12} and {M2V12}, respectively), and applications ranging from (light-driven) catalysis30,31 to energy storage32,33. Note that for most of the di-metal-functionalized {V12} species were only accessible as mixed VIV/V oxidation state clusters33,34,35,36. Recently, a similar synthetic approach to metal-functionalized molybdates has been pioneered by Yamaguchi, Suzuki and co-workers. They used pyridine moieties to coordinatively stabilize and functionalize lacunary polyoxomolybdates which are otherwise difficult to access18,37.

Most often, temperature, solvent, solution acidity and type of metal salt are the key synthetic parameters varied to facilitate metal-functionalization of POVs24. In contrast, the use of light, i.e. photons with energies in the visible spectral range has rarely been discussed as a systematic control parameter to trigger POM and POV functionalization. This is surprising, as the (visible) light photoactivity of POMs is well-documented, and the light-induced excitation of O→M ligand-to-metal charge-transfer transitions is an easy tool to selectively access reduced POM species8,38,39,40,41. The concept has been pioneered by Yamase and co-workers who demonstrated that previously unknown mixed-valent POMs and POVs can be accessed photochemically in the presence of suitable electron donors, e.g., organic amines42,43,44. POV chemistry is particularly sensitive to photoinduced reactions triggered by visible light, as POVs typically show higher visible light absorption compared with tungstates and molybdates38. This approach has recently been explored by Liu and co-workers who demonstrated the visible-light-assisted synthesis of mixed-valent POVs45. Here, we demonstrate how light-dependent, coupled reaction equilibria can be used to selectively target the partial reduction and metal-functionalization of {V12}, leading to a di-magnesium functionalized species, {Mg2V12}, as an intriguing model compound for future studies, e.g., in electrochemical energy storage8,33.

Results

Synthesis and characterization

The starting point was our study into the design of magnesium(II)-functionalized {V12} as molecular models for Mg ion batteries, where solid-state magnesium vanadates are under investigation as active electrode materials46. Initial experiments to functionalize {V12} with Mg2+ were performed by reacting Mg2+ with (DMA)2{V12} in acetonitrile under the standard placeholder-functionalization conditions described above13,34. Despite the extensive variation of the reaction and isolation conditions, reproducible formation of Mg-functionalized {V12} was not possible. Systematic study of the key reaction parameters showed that visible light irradiation and oxygen-free conditions were required to access the target compound, (nBu4N)4[(MgCl)2V12O32Cl] (={Mg2V12}, 1). Diffusion crystallization using diethyl ether as diffusion solvent gave green single crystals of {Mg2V12} in yields of 64% (based on {V12}, see Methods and Supplementary Section 2). When the reaction was performed in the dark under otherwise identical conditions, only the starting material {V12} was recovered as yellow crystals. Crystallographic analysis by single-crystal X-ray diffraction shows that {Mg2V12} crystallizes in the monoclinic space group P21/c with cell axes a = 24.3414(9) Å, b = 16.7474(7) Å, c = 24.4623(9) Å and cell angles β = 94.6107(17)°, α = γ = 90° (for crystallographic details see Methods and Supplementary Section 2.9). Note that this crystal lattice is virtually identical to the previously reported di-functionalized species {Mn2V12} (nBu4N)4[(MnCl)2V12O32Cl]34. For full characterization of {Mg2V12} see Supplementary Section 2. The metal oxo framework of {Mg2V12} is isostructural to the di-metal-functionalized {V12} species reported earlier (i.e., {Mn2V12}34, {Ca2V12}33, {K2V12}32, {Sr2V12}35, {Ce2V12}30), the two square-pyramidal Mg2+ ions reside in the metal binding sites on top and bottom of the cluster and feature a terminal chloride ligand (Fig. 1a, b).

Fig. 1: Structural and spectroscopic information on the formation of {Mg2V12}.
figure 1

a side view of {Mg2V12}; (b) top view of the Mg binding site in {Mg2V12}; (c) time-lapse UV-Vis-NIR spectroscopy of the {Mg2V12} reaction mixture containing {V12} and MgCl2 in acetonitrile. d time-dependent reduction of {V12} in the presence (green squares) and absence (orange squares) of Mg2+. Conditions: irradiation with a broadband high-power LED light source (Poptical ~ 5 W), [{V12}] = 0.05 mM, [Mg2+] = 0.21 mM.

Full size image

UV-Vis-NIR spectroscopy of {Mg2V12} confirms the mixed valent (VIV/V) character of the species, as indicated by the characteristic, broad intervalence charge-transfer (IVCT) band between ~600–1200 nm34. Further support of the mixed-valent nature of {Mg2V12} is given by continuous wave electron paramagnetic resonance (EPR) spectroscopy which unambiguously shows the presence of one VIV species with S = ½ (Supplementary Fig. 3), while a pure VV cluster would be EPR silent. Furthermore, EPR-based spin counting is in good agreement with one VIV centre per {Mg2V12}.

Photochemical studies

Irradiation of a reaction mixture containing {V12} and Mg2+ in MeCN with a broadband LED light source resulted in the emergence of the IVCT transitions characteristic for the formation of mixed-valence VIVI/V species (vide supra). Thus, the change of the UV-Vis-NIR signals over time can be used to monitor the rate of {Mg2V12} formation, see Fig. 1c, d. This provides the ideal conditions to explore the fundamentals of the light-induced formation mechanism of metal-functionalized vanadates: to this end, we compared the photoreduction of {V12} in the presence and absence of Mg2+ (Fig. 1d). Strikingly, {V12} reduction is only observed in the presence of Mg2+, while in the absence of Mg2+, no formation of VIV centres and no IVCT signal is detected. We hypothesized that this finding indicates that Mg2+ interacts with {V12} in the reaction solution to give a photoactive reactive intermediate. Based on our understanding of the system, we suggested that this intermediate could be the mono-functionalized species {MgV12} (= {(DMA)[(MgCl)VV12O32Cl]}3-). The formation of an intermediate species is indicated by UV-Vis-NIR spectroscopy, which shows distinct spectral changes in the region between 300 nm to 500 nm upon addition of Mg2+ to the {V12} reaction solution (Fig. 2a). Also, 51V NMR spectroscopy shows that immediately after addition of Mg2+ to an acetonitrile solution of {V12}, a four-signal spectrum is observed which is characteristic for the mono-metal-functionalized {MV12} species (Fig. 2b).

Fig. 2: Experimental verification for the in-situ formation of the photoactive {MgV12} intermediate under reaction conditions.
figure 2

a UV-Vis-NIR spectral changes observed upon reaction of {V12} (0.05 mM) with Mg2+ (0.21 mM) in acetonitrile, resulting in the formation of the visible-light photoactive {MgV12}. b 51V NMR spectroscopic observation of the characteristic four-line pattern of mono-functionalized {MgV12} formed by reaction of {V12} (10 mM) with Mg2+(42 mM) in acetonitrile. 51V NMR spectra of {V12} (in acetonitrile) and {Mg2V12} (in dimethyl sulfoxide) are shown for comparison. c 51V TOCSY NMR spectrum indicating that all four signals belong to one 51V spin system i.e., one {MgV12} cluster. Non-diagonal signals are marked with red circles. Conditions: [{V12}] = 5.0 mM, [Mg2+] = 21.1 mM, solvent: acetonitrile. d Negative ion-mode high-resolution ESI mass spectrum showing the observed and simulated isotopic pattern for [HMgV12O32Cl]2- (= H{MgV12}), [{V12}] = 0.05 mM, [Mg2+] = 0.21 mM, solvent: acetonitrile.

Full size image

Note that a virtually identical four-line signal pattern has been reported for the mono-Zn2+-functionalized species {ZnV12} (= {(DMA)[(ZnCl)V12O32Cl]}3-)13. In addition, 51V NMR total correlation spectroscopy (TOCSY) of the reaction solution showed, that the four 51V NMR signals assigned to {MgV12} belong to one molecular species (Fig. 2c).

Mechanistic analyses

Further evidence for the formation of {MgV12} is revealed by characteristic changes in the respective 1H and 1H DOSY spectra (Supplementary Section 3.1). Upon addition of Mg2+ to a {V12} solution in acetonitrile, DMA cations are released from their original, {V12}-bound positions into solution, resulting in a dynamic equilibrium between cluster-bound and “free” DMA cations. This results in a characteristic low-field shift of the N-H proton resonances from δ ~ 6.3 ppm to δ ~ 7.0 ppm (Supplementary Fig. 12 and Fig. 13). 1H DOSY NMR spectra were collected to further study the release of DMA cations upon Mg2+ addition to {V12} solutions in acetonitrile. Specifically, we studied the characteristic changes of the respective diffusion coefficients D based on analysis of the DMA−1H resonances (methyl groups, δ ca. 2.6 ppm; ammonium groups, δ ca. 6.3 – 8.6 ppm, (Supplementary Fig. 14)47. These analyses show the expected trend, i.e., the diffusion coefficients decrease with increasing size of the species studied in the order “free” DMA <{MgV12} <{V12}.

Further, high-resolution electrospray ionization mass spectrometry (HR ESI MS) allowed us to identify a series of signals corresponding to the mono-Mg-functionalized species, e.g., [HMgV12O32Cl]2- (observed: 591.590 m/z, calculated: 591.852 m/z, see Fig. 2d, and Supplementary Section 2.8 for further peak assignments). Note that the characteristic four-line 51V NMR signal pattern assigned to {MgV12} was observed even under the dilute concentration conditions of the HR ESI MS experiments ([Mg2+] = 0.21 mM, [{V12}] = 0.05 mM, see Supplementary Fig. 15).

Next, we explored the Mg2+ functionalization further by performing a 51V NMR spectroscopy titration, where increasing amounts of Mg2+ were added to {V12} solutions in acetonitrile. To assess the formation of {MgV12}, the characteristic 51V NMR signals were integrated, and the integral areas were plotted as a function of the Mg2+ equivalents added. As shown in Fig. 3, integration of the three non-overlapping signals unambiguously indicates, that changes of the integrated area are only observed up to 1.0 equivalents Mg2+. Higher equivalents do not change the spectra observed. This strongly suggest the presence of a 1:1 molar species, which is in line with the formation of {MgV12}. These observations are supported by an identical 1H NMR titration study which shows that upon Mg2+ binding to {V12}, release of DMA cations (indicated by characteristic shifts of the DMA proton signals) is observed, see Supplementary Fig. 13.

Fig. 3: In-situ 51V NMR spectroscopic titration to assess the {MgV12} formation.
figure 3

Top: stacked 51V NMR spectra of acetonitrile solutions containing {V12} and varying Mg2+ molar equivalents (between 0 eq. to 4 eq. relative to {V12}). Bottom: the area integrals of the three characteristic {MgV12} signals marked (a–c) are shown. In each instance, integral changes are only observed up to 1.0 Mg2+ equivalents, indicating that a 1:1 species, i.e., {MgV12} is formed. Conditions: [{V12}] = 5.0 mM, [Mg2+] = 0–21.1 mM, solvent = acetonitrile.

Full size image

The different photoactivities of {V12} and {MgV12} were probed experimentally by wavelength-selective irradiation: when the standard {V12}/Mg2+ reaction mixture in acetonitrile was irradiated with a monochromatic 405 nm LED light source, the characteristic {MgV12} reduction and formation of the characteristic IVCT band was observed. In contrast, when the same experiment was performed for a pure {V12} solution (in acetonitrile, without added Mg2+), no vanadate reduction was observed. Also, irradiation of the standard {V12}/Mg2+ reaction mixture using a monochromatic 470 nm LED light source also did not lead to reduction of the vanadate cluster, see Supplementary Fig. 18.

Further insights into the electronic structure of {MgV12} and {Mg2V12} were obtained by theoretical computations using density functional theory (DFT) using the B3LYP functional48,49 combined with the def2-SVP basis set50. Analysis of the HOMO-LUMO levels of {MgV12} and analysis of the calculated UV-Vis-NIR spectrum show an intense ligand-to-metal-charge-transfer (LMCT) transition at the UV-to-Vis border, which we attribute to the experimentally observed Vis photoactivity of {MgV12}. For {Mg2V12}, similar LMCT transitions are observed, and in addition, the broad characteristic IVCT transition in the Vis-NIR range are reproduced by the calculations. For details, see Supplementary Section 4.

Based on Le Chatelier’s principle, we hypothesized that addition of DMA to the reaction solution should shift the equilibrium to the reagent side (see Fig. 4a), thus preventing the formation of the photoactive {MgV12}. This behaviour is indeed observed: when an excess of DMACl is added to the standard {V12} reaction solution and the sample is irradiated, no reduction is observed by UV-Vis-NIR spectroscopy (Supplementary Fig. 19). This suggests that the reactive intermediate which enables photoreduction is not present under these conditions and lends further support to {MgV12} being the photoactive intermediate. Also, when the {V12} photoreduction is performed in the presence of Mg2+ and air, virtually no vanadate reduction is observed (as indicated by the absence of IVCT bands in the UV-Vis-NIR spectrum, see Supplementary Fig. 20). This provides further support that a light-induced electron transfer to the photoexcited {MgV12} is a key process in the formation of {Mg2V12} and suggests that interference between the photoexcited {MgV12} and O2 (e.g., by triplet quenching51) could prevent formation of the reduced vanadate species52.

Fig. 4: Proposed coupled solution-phase equilibria during {Mg2V12} formation.
figure 4

a The light-independent pre-equilibrium forming the photoactive intermediate {MgV12} and (b) light-driven formation of the di-Mg-functionalized {Mg2V12}.

Full size image

In sum, these data suggest that reaction of {V12} and Mg2+ results in formation of the mono-functionalized {MgV12} as photoactive, reactive intermediate, which can then be converted to the di-functionalized {Mg2V12} upon irradiation with visible light. To gain insights into the sacrificial electron donor, we used cyclic voltammetry to compare the redox potentials of the possible donors (DMA, nBu4N+, MeCN). Based on this data, DMA is the most likely sacrificial electron donor, as it is significantly easier to oxidize (Eox = 0.7 V vs Fc+/Fc) compared with the other possible electron donors, i.e., nBu4N+ and MeCN (Eox > 1.6 V vs Fc+/Fc), see Supplementary Fig. 21. Based on these considerations, the following coupled reaction equilibria are proposed, see Fig. 4.

Finally, to probe whether the observed light-induced reactivity is unique to Mg2+ or can be extended to other metal functionalizations, we performed the {V12} metal functionalization experiments using Ca2+ instead of Mg2+ (experimental details see Supplementary Section 3.4). Reaction of CaCl2 x 2H2O (21.1 mM) with {V12} (5.0 mM) in acetonitrile led to the observation of the characteristic four-line 51V NMR signal pattern assigned to {CaV12} (Supplementary Fig. 22). Visible light-irradiation of the {V12} / Ca2+ reaction solution resulted in the formation of the characteristic IVCT bands between ~600 and 1200 nm which is indicative of the formation of the reduced, di-metal-substituted vanadate species (see Supplementary Fig. 23). These findings suggest that the light-induced metal functionalization reported here is not unique to Mg2+ and can be transferred to other metal cation species, and possibly also to other vanadate cluster architectures.

Discussion

We report the first example of a light-dependent, coupled set of solution-phase equilibria, enabling the controlled metal-functionalization of molecular vanadium oxides. Light-independent reaction of {V12} with Mg2+ results in a dynamic pre-equilibrium, where one DMA placeholder cation on {V12} is replaced with one Mg2+ ion, resulting in formation of the Vis-photoactive intermediate {MgV12}. The formation of a 1:1 species was verified by 51V NMR / 1H NMR spectroscopy as well as HR ESI MS studies. Competitive binding studies using DMA and Mg2+ show, that this pre-equilibrium is sensitive to the Mg2+ / DMA molar ratio, essentially allowing an on/off switching of the metal functionalization.

Visible light-irradiation of {MgV12} solutions results in the one-electron photoreduction of the cluster, release of the second DMA placeholder cation, and binding of a second Mg2+ metal centre, yielding {Mg2V12}. The increased visible-light photoactivity is in line with recent literature reports which show that metal-incorporation in POMs leads to a lowering of the HOMO-LUMO gap and thus, increased photactivity53. The photoredox processes at {MgV12} only occur in the absence of water and oxygen, indicating possible interference of these species with the light-induced electron transfer to the cluster. Electrochemical studies suggest that DMA is the most likely electron donor based on analysis of the redox potentials of the reagents used. Mechanistic experimental and theoretical studies show the light-dependent nature of the assembly process and emphasize how supramolecular reaction control can be used to trigger or inhibit photoactivity. Finally, initial experiments show that a similar route can be followed to enable {V12} functionalization with Ca2+ using light-induced cluster assembly. Thus, the principles outlined in this report open new paths for designing multi-stimuli-responsive molecular materials.

Methods

Synthesis of 1

(nBu4N)4[(MgCl)2VIVVV11O32Cl)] x CH3CN (= (nBu4N)4{Mg2V12})

The synthesis of 1 was performed in a glovebox under argon atmosphere: in a 25 mL round-bottom flask 0.200 g (0.100 mmol) (nBu4N)3(NMe2H2)2[V12O32Cl)] x CH3CN, 0.056 g (0.200 mmol) nBu4NCl and 0.0400 g (0.420 mmol) anhydrous MgCl2 were dissolved in 12 mL water-free, deaerated acetonitrile and stirred at room temperature. After four hours of stirring, the yellow solution was filtered through a glass Pasteur pipette filled with glass wool. Diffusion crystallisation with diethyl ether was setup and the samples were exposed to light via the glovebox fluorescent lamps or a broadband LED light source (see Supplementary Fig. 1 for emission spectrum of the light source). After two days, dark green crystals of 1 were obtained, filtered, washed twice with acetone and diethyl ether and dried vacuum. Yield: 0.146 g (0.0634 mmol, 63.8% based on V). Characteristic IR bands (in cm−1): 3336; 2961 (C-H stretching, alkane); 2934 (C-H stretching, alkane); 2874 (C-H stretching, alkane); 1659; 1641; 1613; 1482; 1461; 1381;1422; 1345; 1278; 1251; 1163; 1151; 1107; 1067; 995 (symmetric V = O, terminals Oxygen); 890; 875 (anti-symmetric stretching VO unit); 817; 752 (V-O-V, symmetric); 661 (V3-Oµ3, asymmetric); 593; 412. UV-Vis-NIR spectroscopic maxima ([{Mg2V12} = 56 µM in DMF): ε338 = 11,875 M−1 cm−1; ε991 = 1,400 M−1 cm-1. 1H NMR (400 MHz, DMSO-d6): δ (ppm) = 3.44 (s, H2O); 3.18 (m, 2 H, nBu4N+); 2.45 (s, unassigned); 2.07 (s, acetonitrile); 1.56 (m, 2 H, nBu4N+); 1.30 (s, 2 H, nBu4N+); 0.92 (t, 3 H, nBu4N+). 51V NMR (105 MHz, DMSO-d6): δ (ppm) = − 557.0 (s, 4 V); −578.3 (s, 8 V).

Crystallographic data for 1

(nBu4N)4[Mg2Cl3V12O32] x CH3CN, (M = 2289.15 g/mol): monoclinic, space group P2/c, a = 24.3414(9) Å, b = 16.7474(7) Å, c = 24.4623(9) Å, β = 94.6107(17), V = 9939.9(7) Å3, Z = 4, T = 150 K, ρcalc = 1.530 g/cm3, μ(MoKα) = 1.238 mm−1, 332994 reflections measured, 22031 unique (Rint = 0.0771, Rsigma = 0.0266), R1 = 0.0518 (I > = 2σ(I)), wR2 = 0.1449 (all data). CCDC 2240239 contains the supplementary crystallographic data for this paper. These data can be obtained free of charge from The Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/structures.

Solution-phase NMR spectroscopic analyses

Mechanistic reactivity studies using 1H and 51V NMR spectroscopy were carried by dissolving the compound under study in the respective solvent. The reactions were performed under the given conditions as stated above and in the Supplementary Information. All solutions were prepared in an argon-filled glovebox unless stated otherwise.