Introduction

Single-source precursors (SSPs) are molecules which contain all the required elements for a material. Thermal decomposition provides a convenient route to a homogeneously mixed material, potentially accessible at low temperatures and with high surface areas/porosities1,2,3,4,5,6. The use of SSPs can allow access to unusual phases and doping levels, or control of physical properties, that is not easily accessed by higher temperature processes1,7,8,9,10. Multimetallic SSPs are attractive for the synthesis of complex (nanocrystalline) multimetallic oxides, offering opportunities to control composition, as well as particle morphology1,11,12,13,14. A growing library of well-defined bi and tri-metallic molecular precursors is available for study in this role2,15,16,17.

The molecule-to-material transformation that occurs on the thermal decomposition of an SSP is a complex process requiring structural rearrangement and decomposition of any organic components. Detailed studies highlight the importance of atomic level characterisation in understanding these processes18. The possibility of generating interesting and useful phases at low temperatures, before the thermodynamic oxide phase, is intriguing, yet has been rarely explored in detail. Formation of amorphous phases typically makes characterisation of intermediates difficult; therefore, advanced techniques are needed to provide local structural information beyond what is available from conventional diffraction studies.

In this project the thermal transformation of mono, bi and tri-metallic POV clusters is explored in the solid phase. These SSPs contain V, V + Bi or V + Bi + Zn, and allow for exploration of the intermediary stages of thermal treatment before the formation of oxide products (V2O5 and (Zn-doped)-BiVO4), including mixed-valence amorphous and kinetically stabilised phases.

The resultant oxide products, V2O5 and BiVO4 are important materials with applications including photoelectrochemical water oxidation to enable solar-to-fuel technologies19,20,21,22,23. V2O5 is an important material, used extensively in oxidation catalysis, battery cathodes, supercapacitors, gas sensors and electro/thermo/photochromic materials24,25. BiVO4 is a very important photoanode material with optimal band-gap properties for water oxidation19,26, and is used in state-of-the-art artificial leaf devices21,22,23. Interest in BiVO4 has rapidly accelerated in the last decade, with 860 articles published in 202427. To date (in ~8000 publications), three distinct polymorphs of BiVO4 have been described: the monoclinic scheelite structure (m-BiVO4, I2/b), tetragonal zircon type structure (tz-BiVO4, I41/amd), and tetragonal scheelite structure (ts-BiVO4, I41/a), with m-BiVO4 most commonly used for photoelectrochemical water oxidation. However, recent reports of high performance photoanode materials show that Mo-doping alters phase stability and promotes tetragonal scheelite28, and, whilst the tetragonal zircon phase has a higher band-gap, it is of interest for forming type-II heterojunctions due to its high crystal symmetry29, or, alternatively, can be used as a photocathode30, therefore, polymorph engineering of BiVO4 is of wide interest. Commonly, high-activity BiVO4 photoelectrodes are produced via an electroplated BiOI phase, and may be enhanced by doping31, however, this electrochemical method is not easily scaled whilst retaining material uniformity19. To deploy BiVO4 photoanodes at scale, alternative synthetic routes, such as the use of SSPs15, require exploration.

POV clusters make attractive SSPs, as they are typically anionic, they co-crystallise with a cation, enabling two or three different metals to be incorporated homogeneously into a crystalline structure. The thermal decomposition of V(V) decavanadate ([V10O28]6–) compounds was explored by Ulická, Baran and co-workers to make ternary oxide materials (including mixtures), and, more recently, Streb and co-workers considered the impact of thermal treatment when preparing battery electrodes from POVs32,33,34,35,36,37. Initial endothermic processes result in dehydration and molecular restructuring36,38. In most cases partial reduction of V(V) to V(IV) occurs at moderate temperatures (maximum reduction at ~200 °C)34,35. Partial re-oxidation occurs on increasing annealing temperature, as the final oxide phases crystallise, however, electron paramagnetic resonance (EPR) spectroscopy indicates that some oxygen vacancies may remain38. The thermal decomposition of ammonium metavanadate [NH4][VO3] also accesses nanostructures of V2O5. Initial restructuring to [NH4]2[V6O16] is followed by intermediary, partially-reduced, black phase at 150 °C–200 °C, before V2O5 formation at 250 °C39,40. The release of NH3 and H2O gasses aids construction of mesoporous structures41.

No decavanadate structure with an associated bismuth counter-cation is known, instead, reaction of decavanadate with Bi(NO3)3 results in the neutral compound [H3Bi4V13O40(DMSO)12]·4(DMSO) (2), in which a [ε-(VO4)V12O36]15– Keggin anion core is capped by Bi cations that are supported by DMSO ligands42. Altering the synthesis by the addition of chloride ions, allows the formation of the [Bi2V12O33Cl(DMSO)6]– anion which may be partnered with a range of metal cations ([Bi2V12O33Cl(DMSO)6]x[M(DMSO)y]z, 3-M)15,16,43. These POVs have been used as SSPs for BiVO4 thin-films; 2 or 3-M (M = Co, Ni, Cu, Zn) were dropcast onto a fluorine-doped tin oxide (FTO) substrate and annealed at 550 °C to generate (M-doped) BiVO4 photoanodes on up to a 300 cm2 scale15. It is noteworthy that the V:Bi ratio in 2 and 3 is 3.25 and 6 respectively, larger than required for BiVO4, and, therefore, thermal decomposition also forms a V2O5 byproduct. However, there are advantages of a V rich precursor:

  • A V-rich environment ensures that no Bi rich phases are formed, noting that in chemical vapour deposition approaches to BiVO4, Bi2VO5 is easily produced as a by-product, likely due to loss of some V as volatile species in the process44.

  • Recent reports indicate that heterojunctions between BiVO4 and V2O5 can be advantageous for photocatalysis45.

  • The V2O5 by-product in the decomposition of 2 or 3-M is easily removed by a simple basic washing step, which aids creating a porous BiVO4 architecture15. Furthermore, deliberately releasing V ions into the electrolyte can be advantageous for retarding photodegradation of BiVO446.

Of the doped BiVO4 photoanodes produced from 3-M, the Zn-doped material showed the highest photocurrent density for water oxidation, however, the role of the Zn atoms was not clear15. Scanning electron microscopy with energy-dispersive X-ray spectroscopy (SEM-EDS) data suggested 1.5 Zn2+ dopants replace Bi3+ sites for 9% of the Bi atoms. Subsequent studies, using alternative synthetic routes, have shown that Zn2+ dopants can replace Bi3+ sites, but in these cases there are associated oxygen vacancies. These studies suggest that Zn-doping promotes absorption of water to Bi sites47. Zn doping may also increase n-type conductivity; improve the number of oxygen chemisorption sites that enhance surface charge transfer in photoelectrocatalysis; enhance surface reaction kinetics; and can be useful for forming n-n+ heterojunctions (e.g., with Mo-doped BiVO4)47,48,49.

In this study the thermal decomposition of [NH4]6[V10O28]·6H2O (1), [H3Bi4V13O40(DMSO)12]·4(DMSO) (2) and [Bi2V12O33Cl(DMSO)6]2[Zn(DMSO)6] (3-Zn) are explored, with a focus on understanding intermediate phases that occur during these molecule-to-material transformations (Fig. 1).

Fig. 1: Metal-oxygen connectivity of the molecular precursors.
Fig. 1: Metal-oxygen connectivity of the molecular precursors.The alternative text for this image may have been generated using AI.
Full size image

a–c Crystal structures of 1, 2 and 3 (anion only), SMe2 groups of DMSO ligands coordinated to Bi, and all H atoms omitted for clarity. d Scheme of the strategy used to synthesise m-BiVO4 from well-defined molecular precursors (V2O5 by-product not shown). Colour code Bi, purple; V, blue; O, red; and Cl, green.

Results and discussion

Thermal decomposition of ammonium decavanadate (1)

With the ultimate aim of studying 2 and 3-Zn15, a monometallic cluster, 1, with a similar size and structure was chosen as an entry point to the study. While 1 contains the well-studied decavanadate anion50, to the best of our knowledge, the thermal transformations of the ammonium salt have only been briefly discussed51. Therefore, we set out to identify any intermediate stages, including amorphous ones, using X-ray diffraction, PDF and solid-state NMR analysis. Bright orange 1 was prepared as a powder from literature routes52, and its structure confirmed by IR spectroscopy (Fig. S1), H, N elemental analysis, and single crystal and powder X-ray diffraction (PXRD, Fig. S2)53. On drying under extended vacuum around half of the co-crystallised water is lost. 51V MAS NMR of 1 shows multiple V environments in the ranges of −405 to −439 ppm and −480 to −555 ppm (Fig. 2a), consistent with solution spectra54, and with full-width-half-maxima (FWHM) of 4–5 ppm. The spectrum implies that there are several different decavanadate environments in this microcrystalline material, likely due to local hydration levels or partial protonation.

Fig. 2: Thermal transformation of 1.
Fig. 2: Thermal transformation of 1.The alternative text for this image may have been generated using AI.
Full size image

a Room-temperature 51V solid-state MAS NMR spectra of 1 and the products of thermal transformations. The temperatures given correspond to the maximum heating temperature. The asterisks indicate spinning sidebands. b Variable temperature PXRD of compound 1 and the products of heating.

Thermogravimetric analysis (TGA-MS/DSC) shows several consecutive mass losses totalling ~24% mass loss on heating to 370 °C, with water and ammonia detected by mass spectrometry throughout, but with enhanced water loss between 80 °C–150 °C (Figs. S3–5). A mass loss of 22.5% would be expected based on the loss of all NH3 and H2O from 1 and formation of V2O5, therefore, some partial reduction (and loss of oxide, noting that some NO loss is observed) and/or loss of volatile V compounds may occur during this process. Figures 2b and S6–7 show X-ray diffraction patterns of 1 heated in-situ from RT to 600 °C. The initial crystalline structure is maintained until 80 °C before a new broadened diffraction pattern is observed, consistent with a dehydration step35,36,38,53. Infra-red spectroscopy of a sample heated to 150 °C–175 °C shows a shift and splitting of the terminal V = O stretching frequency from ~937 cm‒1 to ~958 cm‒1 and a minor signal at 995 cm‒1 (Fig. S8) which could imply restructuring of decavanadate into hexavanadate, [NH4]2[V6O16]55. H, N analysis of a sample heated to 150 °C is a close match for hexavanadate with a small amount of remaining decavandate (Table S2), A poorly crystalline unidentified phase is observed at 160 °C by PXRD (Fig. S9), although this pattern does not match previously reported [NH4]2[V6O16] (Fig. S10). Further heating results in a loss of crystallinity at 200 °C. By 250 °C α-V2O5 signals are emerging in the PXRD (and also IR spectra), and these sharpen on further heating (Figs. 2b, S6–8 & Table S3). The crystallite size was monitored using the double-Voigt methodology reported by Balzar et al.56; powdered 1 was estimated to have average crystallite sizes of ~10 nm at room temperature, whilst, after heating to 250 °C, V2O5 crystallites are ~103 nm and continue to grow to ~162 nm on reaching 600 °C ( ~ 160 nm estimated after cooling).

Samples of 1 were annealed ex-situ to different maximum temperatures for more in-depth analysis at each stage. The samples change from orange to brown (after 150 °C) to black (after 200 °C, 1200 °C) before turning red/brown after 300 °C (Fig. S11). The black phase 1200 °C is notable, especially considering that 1 does not contain any carbon. 51V MAS NMR spectroscopy of 1200 °C revealed one broad vanadium environment at −550 ppm (FWHM of ca. 40 ppm, i.e., ten times broader than in pristine 1), consistent with an amorphous phase (Fig. 2a). After 300 °C, a single V environment is observed at −615 ppm (FWHM of 12 ppm), consistent with V2O5, and there are no further spectral changes after 400 °C.

X-ray pair distribution function analysis was conducted on 1, amorphous 1200 °C as well as a sample heated to 400 °C (Fig. S12). The sample heated to 400 °C shows an excellent match for V2O5 (Fig. 3a). 1200 °C reveals a different PDF pattern, which was found to resemble V4O9 (Pnma57,58, Rw = 30.4% after refining thermal factors and atomic positions, Fig. S13, Table S4). In Pnma V4O9 half the vanadium atoms have an oxidation state of IV, a quarter of sites are 6-coordinate with the remainder 5-coordinate (square based pyramidal, Fig. 3c). The fit was initially improved by adding a contribution from either decavanadate or hexavanadate anions (Rw = 20.1%, (1) = 21.4 wt%, Fig. S13, or Rw = 26.7%, ([NH4]2[V6O16]) = 13.2 wt% respectively), which may be present in the material at this stage of the thermal transformation, but without long-range molecular packing that would give diffraction signals. The fit was improved even further by incorporating some ammonium vanadium bronze (NH4)V4O10 (noting similar phases have been previously observed from hydrothermal heating of [NMe]4[H2V10O28]); giving a two-component fit with Rw = 17.8%54,59,60. The combination of V4O9, (NH4)V4O10 and 1 resulted in the best fit to the PDF with Rw = 14.8% and relative phase weight fractions V4O9:(NH4)V4O10:1 50(5):35(4):15(7) (Fig. 3b, S14). No improvement in the fit occurred by adding V2O5. This suggests that 1 undergoes a series of transformations during thermal decomposition to form mixed-valence V4O9 and (NH4)V4O10 before its full conversion to V2O5 beyond 250 °C. From this mixture of materials, ~35% of V might be expected in the V(IV) oxidation state and the presence of V(IV) was confirmed by EPR (Fig. S15). However, 1200 °C was dissolved in 2 M H2SO4 and a double titration method used to identify the amount of V(IV) content61, and a lower value of 2.7 ± 1 % V(IV) was identified. X-ray Absorption Near Edge Structure (XANES) analysis confirmed that V(V) is dominant as the spectra of V2O5, 1 and 1200 °C appear almost identical, (noting that XANES is less sensitive to the presence of V(IV) than EPR spectroscopy, Fig. S16). This low V(IV) content is consistent with previous studies35, yet is surprising considering the black colour of 1200 °C. The colour must arise from efficient d-d transitions and intervalence charge transfer (IVCT) excitations that occur across the visible range25.

Fig. 3: X-ray PDF data of 1 after annealing and structure of V4O9.
Fig. 3: X-ray PDF data of 1 after annealing and structure of V4O9.The alternative text for this image may have been generated using AI.
Full size image

X-ray PDF fits (Qmax = 22.0 Å-1) of the materials obtained after annealing 1 at a 400 °C (V2O5, Rw = 12.2 %), and b 200 °C (Refined weight fractions: V4O9: 50(5) wt. %; (NH4)V4O10: 35(4) wt. %; 1: 15(7) wt. %, Rw = 14.8 %). c Snapshot of the V4O9 structure (Pnma) viewed along the b axis. Colour code: V, blue; O, red.

The reduction of some V at intermediary stages of heating is caused by a redox reaction with ammonia evolved from 1, in fact, treating V2O5 with NH3 is a synthetic route to V4O962. Partial reduction is consistent with literature reports32,34,35,39, but here we extend on these by identifying a likely structure of this initial amorphous phase. In 1200 °C, the local connectivity resembles a mixture of V4O9, and (NH4)V4O10 but the low V(IV) content implies that most sites are occupied by V(V), implying that these may be oxygen-rich kinetically stabilised phases.

1H MAS NMR of 1 reveals multiple proton environments at 9–14 ppm, attributed to strongly coordinated H2O and [NH4]+ owing to their substantial deshielding compared to bulk water (ca. 5 ppm) and [NH4]+ in aqueous solutions (ca. 7 ppm), noting that their chemical shifts strongly depend on the pH and hydrogen bonding network. After heating to 200 °C, all these species are lost, and a new peak is found at 7.5 ppm, accompanied by a substantial decrease in the signal-to-noise ratio (SNR) of the spectrum (20% of that in pristine 1), consistent with the loss of the volatile proton-containing species (Fig. S17a). H, N elemental analysis collected after annealing to 200 °C and 250 °C (Table S2), and FT-IR spectroscopy (Fig. S9) confirm the loss of most H and N over this temperature range, in good agreement with the estimated phases from PDF fitting. At 250 °C, once the PXRD reveals V2O5 is forming, only trace ammonium remains, and by 300 °C only residual protons are observed in the 1H MAS NMR with an SNR of <1% of that observed in pristine compound 1 (Fig. S17a).

In summary, upon heating 1, an initial dehydration process occurs, resulting in collapse of the crystalline structure, further restructuring of the decavanadate structure via partially crystalline phases occurs before complete loss of crystallinity at 200 °C, at which stage the sample becomes black, with ~3% V(IV) content, and is best described with a local structure resembling a mixture of Pnma V4O9 and layered (NH4)V4O10 and with some remaining isolated POV anions. Heating beyond this point begins to form V2O5 with the chemical transformation completed by 400 °C. Further heating under air reduces the oxygen vacancy content and increases the crystallinity (Fig. 4).

Fig. 4: Summary diagram of the molecule-to-material thermal transformation of 1 into V2O5.
Fig. 4: Summary diagram of the molecule-to-material thermal transformation of 1 into V2O5.The alternative text for this image may have been generated using AI.
Full size image

Colour code in structures: V, blue; O, red.

Thermal decomposition of 2 and 3

2, [H3Bi4V13O40(DMSO)12]·4(DMSO), and 3-Zn, [Bi2V12O33Cl(DMSO)6]2[Zn(DMSO)6]·12(DMSO) (N.B. co-crystallised DMSO content drops if dried under vacuum), were prepared by literature methods and the expected structures confirmed by PXRD (Figs. S18–19, see supporting note 1 for discussion of how to avoid possible impurities, including [Bi2V12O33Cl(DMSO)6]2[VO(DMSO)5]·8(DMSO), 3-VO). It is noteworthy that precipitation of 2 may generate an amorphous form (2amor) which contains the expected C, H, S %, whilst slower crystallisation generates a (micro)crystalline form (2).

The crystal structure of 2 contains four distinct V sites. 51V MAS NMR of a powdered sample of 2 reveals multiple V environments between −440 and −580 ppm (Fig. 5a, RT). The spectrum is markedly broadened by static disorder owing to the presence of co-crystallised DMSO molecules. Both cross-polarisation and direct excitation 13C MAS NMR lead to fast signal build-up and a comparable SNR, which indicates that most DMSO molecules in the structure of 2 are rigid (i.e., there is static disorder) rather than fluxional (Fig. S20). This may impact the ability of DMSO to escape the crystals and promote decomposition rather than evaporation on heating.

Fig. 5: Solid-state NMR characterisation of the thermal transformations of 2 and 3-Zn.
Fig. 5: Solid-state NMR characterisation of the thermal transformations of 2 and 3-Zn.The alternative text for this image may have been generated using AI.
Full size image

Room-temperature 51V solid-state MAS NMR spectra of the starting materials and the products of their thermal transformations (ppm = parts per million). a compound 2, b compound 3-Zn. The temperatures given correspond to the maximum heating temperature. The asterisks indicate spinning sidebands. The signals attributed to β-BiVO4 are indicated by arrows, which indicate the possible positions of the isotropic resonance and a spinning sideband. While β-BiVO4 is expected to contain only one distinct V site, the large linewidth makes the identification of the isotropic peak challenging.

TGA under air shows expected mass losses (2amor, 35.5%; 3-Zn, 40.2%) during heating as DMSO is lost via decomposition (and water is also lost from 2, Fig. S21-26)15,42. A curious mass increase is seen when heating 2amor under air (5 °C/minute) between 360 °C and 450 °C (Fig. S21), associated with an exothermic process (Fig. S22). This additional mass is then lost again upon heating to 520 °C. The mass gain may be due to oxidation reactions occurring to V or S atoms within the structure at this stage. A temperature of 520 °C or 600 °C is required to complete the mass loss and formation of the extended oxide materials (m-BiVO4 and V2O5, after cooling to room temperature) for 2amor and 3-Zn respectively. TGA-MS implies DMSO decomposition occurs, with indication of CO2 and H2CO, and some evidence of SO2 and/or H2S loss occurring (Figs. S23 & S26). 2/2amor & 3-Zn are red, however, after heating above 175 °C (under air) the resulting materials appear black, before becoming green (400 °C–480 °C) and then converting to the final orange/yellow colour at ~550 °C (Fig. S27-28). The dark colours are associated with V(IV) content due to d-d and IVCT transitions and are consistent with the thermal transformation of 1. The 1H MAS NMR spectrum of 2 heated at 300 °C shows the presence of intense spinning sidebands, not present at any other temperature, which we attribute to paramagnetic effects (anisotropic bulk magnetic susceptibility and/or pseudocontact shifts, Fig. S17d).

C, H, S elemental analysis, shows that 2amor heated to 200 °C loses DMSO (via decomposition) but retains a higher-than-expected H content, likely indicating retention of hydroxide functionality (Table S5). 1H MAS NMR of this material shows the presence of multiple sharp, liquid-like species, across the spectrum, including in the aromatic (7-9 ppm) region. Since the only solvent used in the synthesis of 2 was DMSO, we cautiously attribute these species to DMSO cyclisation products catalysed by the oxide phases. FTIR analysis reveals loss of all DMSO after heating to 400 °C (Fig. S29), corroborated by the C% content and 1H and 13C MAS NMR which confirm the loss of the DMSO signal (1H, 3.5 ppm, and 13C, 40.5 ppm) and the generation of a broad, −5 to +10 ppm 1H signal after heating to 600 °C (Figs. S17, S20), which we attribute to residual protonation of the oxide surface sites (SNR of <0.1% of that in pristine 2). Some H and S content are retained at 420 °C, indicating some incorporation of S into the oxide materials at this stage (Table S5), but after 600 °C essentially all C, H and S are lost. 3-Zn also retains higher-than-expected H & S content during heating, and after 380 °C the sample still contains 1 wt% S (Table S6). 1H MAS NMR of 3-Zn also reveals loss of DMSO signals and, first, generation of intermediate proton-containing degradation products, and then residual proton-containing species (SNR < 1% of that in pristine 3-Zn) after heating at 600 °C (Fig. S17c).

PXRD was collected on samples of 2 and 3-Zn heated ex-situ (samples pre-heated and cooled, Fig. S30) and on samples of 2amor and 3-Zn heated in-situ (Fig. 6, S31–32). The in-situ samples were pre-heated to 175 °C ex-situ to remove the majority of DMSO in order to protect the equipment. 3-Zn shows restructuring of the crystalline phase after 60 °C, giving unknown phase(s) likely to retain POV species before the formation of a mainly amorphous phase at 175 °C (Fig. S19). Starting from either 2amor or 3-Zn an amorphous phase is found at 200 °C, before signals for BiVO4 and V2O5 appear after 300°C and become sharper upon heating further. N.B. In the in-situ heated samples the high-temperature tetragonal scheelite phase, ts-BiVO4, is observed during annealing up to 600 °C but this phase transitions to the stable monoclinic form, m-BiVO4, when cooled to room temperature.

Fig. 6: Characterisation and structure of β-BiVO4 during thermal transformations of 3-Zn.
Fig. 6: Characterisation and structure of β-BiVO4 during thermal transformations of 3-Zn.The alternative text for this image may have been generated using AI.
Full size image

a Variable temperature PXRD patterns starting from 3-Zn (sample pre-heated to 60 °C, cooled, then heated to 175 °C and cooled before VT-experiment 200 °C–600 °C, then cooled to 30 °C), alongside an inset (c) of the fit at 400 °C showing the contributions of the three different phases. b Evolution of the relative percent of each crystalline phase during annealing of 3-Zn as obtained from the Rietveld analysis of VT-PXRD data of 3-Zn. d Rietveld fit of 3-Zn annealed to 380 °C then cooled (data collected at Diamond Light Source, I15-1, See ESI for details) after introducing Bi disorder. Refinement details: V2O5 (63.1(2) wt. %), Orthorhombic (Pmmn), a = 11.509(1) Å, 3.5675(2) Å, c = 4.3795(5) Å; m-BiVO4 (11.8(1) wt. %), Monoclinic (I2/b), a = 5.1500(8) Å, b = 5.1110(6) Å, c = 11.722(2) Å, β = 89.98(2)°; β-BiVO4 (25.1(2) wt. %), Cubic (P213), a = 6.9916(2) Å. Rwp = 2.26 %, Rexp = 1.45, Χ2 = 1.59.) after introducing Bi disorder. e View of the refined β-BiVO4 structure without (left) and with (right) disordered Bi atoms, disordered atoms each with an occupancy of 1/3. f Comparison of the Bi‒O coordination environment in β-BiVO4 with those observed in m-BiVO4 (8-coordinated) and in compounds 2 and 3 (both 6-coordinated). Colour code: Bi, purple; V, light blue; O, red.

In samples of in-situ heated 3-Zn and ex-situ heated (microcrystalline) 2 an extra phase is observed between 350 °C–420 °C, which is then lost on further heating (Fig. 6, S30). Comparing this phase with all reported Bi, V and Bi-V oxides in the database via Pawley and Rietveld refinements yielded unsuccessful results. Indexing of the unit cell with the softwares DICVOL63 and TREOR64, however, suggested a primitive cubic unit cell for the unidentified phase with cell parameter a ~ 7.03 Å. This kinetic phase (sample heated to 400°C), was investigated further, and was found to resemble the β-SnWO4 phase65, which is also found in the family of oxide-conducting Ln2Mo2O9 (Ln = La, Pr) LAMOX compounds66,67,68. Therefore, we carried out a Rietveld refinement (See ESI, Tables S7–S8, Figs. S33–41, for details) on synchrotron X-ray data collected in I15-1 at Diamond Light Source using the β-SnWO4 structural model but replacing Sn2+ by Bi3+ and W6+ by V5+65. This gave excellent results, confirming that this new kinetic phase crystallises in the cubic space group P213 with cell parameter a = 6.9916(2) Å with Bi and V atoms located on the 3-fold symmetry axis of the cubic unit cell. Therefore, we dub this new phase β-BiVO4. From the unit cells, the β phase is expected to be less dense (6.3 g/cm3) than m-BiVO4 (6.8 g/cm3). The structure of β-BiVO4 (Fig. 6e, S42–43) consists of [VO4]3‒ tetrahedra with an average V‒O bond distance of 1.62(2) Å and bond angles ranging between 102.4-115.5°. The [VO4]3‒ tetrahedra are directly coordinated to six different Bi3+ ions, which display a very distorted BiO6 octahedral environment (Fig. 6e, f, S42–43). In this regard, it can be viewed as a rock salt-type structure with alternating Bi3+ and [VO4]3– units. Similarly to Sn2+ in β-SnWO4, Bi3+ ions display three short (2.26(1) Å) and three long (2.662(5) Å) Bi‒O distances, due to the presence of the lone electron pair, in keeping with previous reports that indicate stabilisation of Bi3+ coordination environments in BiVO4 structures65,69. This behaviour is reminiscent of β-SnWO4, even though the lone pair in Bi3+ is less localised than in the case of Sn2+, which would explain the preference to form m-BiVO4 over β-BiVO469. Similarly to in β-SnWO4, the Bi3+ lone pair was found to be oriented towards the base of the [VO4]3‒ tetrahedra. However, the large Biso value for Bi with respect to V and O, suggests that Bi ions may be disordered. Bi was allowed to refine outside the special position, resulting in a slight improvement of the fit as well as for the Biso value for Bi, strongly suggesting that Bi atoms slightly deviate from the 3-fold axis into three closely spaced positions (Fig. 6e, S41, S43). Similar disorder has been found for Ln and Mo ions in structurally-related Ln2Mo2O9 (Ln = La, Pr) oxide conductors68, and could have important implications in the photocatalytic or ionic conductivity properties of this β-BiVO4 phase. It is noteworthy that the (6-coordinate) Bi environment in β-BiVO4 is similar to in the precursors 2 or 3-Zn, but is rather different to the (8-coordinate) environment found in monoclinic or tetragonal BiVO4, and exists with a wider range of Bi‒O distances (2.26-2.66 Å vs 2.43-2.52 Å in tetragonal BiVO4, Fig. 6f)26. The V environment also exhibits shorter V‒O bond lengths compared to other polymorphs (1.62 Å compared to 1.69-1.77 Å)26. Recent studies highlight how judicious choice of precursors can access metastable phases during solid-state synthesis70. S-doping of the BiVO4 phase may be influential on the formation and stability of β-BiVO4; S-doping has been studied in the related La2Mo2O9 phase71, but is also known in monoclinic BiVO472.

DFT calculations were performed to examine the β-BiVO4 phase and compare it to the thermodynamically stable m-BiVO4 phase. Computational details are given in the supporting information (Figs. S44–46, Tables S9–10). The β-BiVO₄ structure is found to be higher in energy than m-BiVO₄ by 19.5 meV per atom, consistent with its kinetic origin. Despite the close chemical composition and identical oxidation states, the two phases exhibit significant electronic differences. An important increase in the band gap is observed in β-BiVO₄, amounting to +1.09 eV within the r2SCAN functional and to +1.58 eV within more advanced non-empirical hybrid functionals. Figure 7 shows the density of states (DOS) for the two phases, aligned through the O 2 s levels to allow direct comparison. The band-gap opening arises from shifts in both the valence-band maximum (VBM) and the conduction-band minimum (CBM).

Fig. 7: Comparison of density of states (DOS) for the β-BiVO4 and m-BiVO4 phases.
Fig. 7: Comparison of density of states (DOS) for the β-BiVO4 and m-BiVO4 phases.The alternative text for this image may have been generated using AI.
Full size image

Calculations were performed within the r2SCAN functional and aligned through the O 2 s states.

We attribute the larger band gap in β-BiVO4 to structural differences between the two phases. β-BiVO4 exhibits more widely separated VO4 tetrahedra (shortest V-V ≈ 4.39 Å vs 3.88 Å in m-BiVO4), leading to weaker V 3d-O 2p-V 3 d interactions and a narrower, more localised conduction band (Figs. S44-45). At the same time, the Bi coordination changes from nearly uniform 8-fold Bi-O bonding in the monoclinic phase to a strongly split 6-fold environment, which reduces Bi 6s-O 2p mixing at the valence-band maximum. Together, these effects raise the CBM and lower the VBM, consistent with the larger band gap in β-BiVO4. Excitonic effects were evaluated using time-dependent hybrid DFT (Fig. S46). Both phases exhibit comparable exciton binding energies ( ≈ 0.4–0.5 eV). These values are most likely overestimated73, but the similarity between the strength of the effect should be well captured. We note that absolute band-gap values in BiVO4 depend sensitively on the level of theory and on additional physical effects such as electron–phonon coupling74. Further tests are provided in the SI.

Rietveld analysis of VT-PXRD data collected between RT-600 °C reveals the evolution of the relative amounts of oxide phases; 2amor primarily yields ts-BiVO4 at 300 °C followed by the subsequent crystallisation of V2O5 as the temperature increases (Fig. S32), in contrast, for microcrystalline 3-Zn, V2O5 appears as the major species at 300 °C before BiVO4 phases grow in (Fig. 6b), perhaps the relatively higher temperature crystallisation of BiVO4 is due to the added influence of Zn atoms on the BiVO4 phases. From 3-Zn, the β-BiVO4 phase is the dominant Bi containing phase <400 °C, however, from 440 °C, only V2O5 and ts-BiVO4 are observed. After annealing 2amor to 600°C V2O5 and ts-BiVO4 appear in a 50.1(4): 49.9(4) wt. % ratio. Heating 3-Zn to the same annealing temperature gives 66.0(7): 34.0(7) wt. % ratio of V2O5 and ts-BiVO4. Only smaller quantities ( ~ 2.5 wt%) of kinetic phase β-BiVO4 could be identified after in-situ heating of 2amor (compared to up to 40% for ex-situ heating of microcrystalline 2) – it is possible that the crystallinity of the precursor influences the kinetics of the molecule-to-material transformation (Fig. S47).

Analysis of the crystallite size (starting from 2amor) shows an average crystallite size of 18 nm for BiVO4 and 5 nm for V2O5 at 300 °C (which is much smaller than observed during similar processing of monometallic 1). Crystallite size then increases with temperature, reaching 133 nm and 161 nm, respectively, at 600 °C.

Samples of 2 and 3-Zn heated to 400 °C or 420 °C were analysed by X-ray PDF (at room temperature). The XRD patterns gave a good fit for a mixture of m-BiVO4, β-BiVO4 and V2O5 as expected (Fig. S35–38, S48–51). For 2 after 400 °C, PDF analysis suggests 43(6) wt% is β-BiVO4, however this drops significantly by 420 °C, at which point weight fractions are 43(4)% V2O5: 53(4)% BiVO4: 4(2)% β-BiVO4. The weight fractions from 2 and 3-Zn after 420 °C are broadly in line with the expected quantities of oxides from the precursor formula (Table S11), however, after annealing to 600°C ICP-MS of the final material shows the amount of V has decreased, which must be due to release of volatile V species (Table S12)75. In the previously reported preparation of thin-films of BiVO4 from 2, no crystalline V2O5 phase was observed, attributed to amorphous V2O5, but perhaps this is better explained through loss of volatile V15.

Our refinement results also showed slightly different cell parameters for m-BiVO4 from 2and 3-Zn heated to 420 °C. In the Zn-doped m-BiVO4 phase (from 3-Zn) there is a slight increase of ~0.015 Å along the a-axis and a very slight decrease in the b and c-axes, with respect to m-BiVO4 from 2 (Figs. S34 and S49). This is accompanied by a lower relative occupancy of 0.945(6) at the Bi site in the Zn-doped phase (compared to 0.985(4) when from 2). The refinement of Bi site occupancy also slightly improved the refinement of 3-Zn from Rwp = 2.30 % to Rwp = 2.26 % (Fig. S52). No ZnO phases were detected in the XRD or PDF analysis, suggesting that minor changes in lattice parameters and electron density at the Bi site may be due to the introduction of Zn in the structure of m-BiVO4 (potentially replacing Bi but also as interstitial dopants).

51V MAS NMR was collected for the ex-situ heated samples of 2. The peaks of the molecular precursor broaden at 200 °C and largely disappear at 300 °C where V2O5 becomes the dominant species visible in the spectrum. A new 51V NMR signal at −535 ppm becomes apparent after heating to 300 °C (Fig. 5a), and since these materials showed the presence of significant quantities of β-BiVO4 in the 300 °C–400 °C range in the PDF data, we attribute this new species to β-BiVO4. This signal is largely lost by 420 °C at which point only V2O5 (-615 ppm) and m-BiVO4 (-425 ppm) remain, with both signals sharpening at higher temperatures (V2O5: FWHM of 17 and 11 ppm at 400 °C and 550 °C, respectively; m-BiVO4: FWHM of 28 and 9 ppm at 400 °C and 550 °C, respectively), confirming that m-BiVO4 becomes more crystalline at elevated temperatures. 51V MAS NMR of annealed 3-Zn shows the formation of V2O5 from 300°C along with a signal at −428 ppm, which we attribute m-BiVO4 (Fig. 5b). However, in this case the m-BiVO4 peak progressively disappears as the heating temperature increases, becoming invisible after heating at 600 °C. The associated PXRD data confirms the material contains m-BiVO4, and this apparent discrepancy is due to the substantial broadening of the 51V signal of m-BiVO4 which occurs upon doping with Zn2+, as previously reported in Sr-doped BiVO476. The aliovalent substitution of Zn2+ for Bi3+, possibly with added interstitial Zn2+ sites for charge balance15, induces substantial static disorder of the V sites causing the 51V signal to broaden beyond detection under these measurement conditions47,76.

In summary, 2 and 3-Zn thermally transform through amorphous mixed-valence phases before the formation of small crystallites of BiVO4 and V2O5 (Fig. 8). Initially two phases of BiVO4 are identified at a stage when DMSO ligands have decomposed but S and some V(IV) are retained in the structure. On further heating the kinetic β-BiVO4 phase is lost and the ts-BiVO4 phase is retained (forming m-BiVO4 on cooling). Whilst no ammonia is present in these compounds, the presence of organic DMSO molecules can act as an internal reducing agent upon heating. Compared to decomposition of 1, darker colours are retained during heating 2 to higher temperatures, suggesting the mixed valence is retained at later heating stages in this case. Volatile vanadium species are lost on annealing to increase the final BiVO4:V2O5 ratios. MAS NMR data from 3-Zn is strongly supportive of increasing amounts of Zn-doping into the resultant BiVO4 phase with increasing temperature.

Fig. 8: Summary diagram of the molecule-to-material thermal transformation of 2 or 3-Zn into (Zn-doped) BiVO4 and V2O5.
Fig. 8: Summary diagram of the molecule-to-material thermal transformation of 2 or 3-Zn into (Zn-doped) BiVO4 and V2O5.The alternative text for this image may have been generated using AI.
Full size image

SSNMR = solid-state nuclear magnetic resonance. Colour code in structures: Bi, purple; V, blue; O, red; and Cl, green.

Electrochemical properties of amorphous 1 200 °C

Vanadium oxides have been extensively researched as Li-ion battery electrodes due to their ability to reversibly accommodate Li ions in their structure50,77,78,79,80. Particularly, V2O5 can reversibly intercalate up to two Li ions per unit formula due to its layered structure. However, insertion of an additional Li ion drives an irreversible phase change to a disordered rock salt structure, ω-Li3V2O5, which maintains good specific capacity, and can itself be used as a negative electrode in fast-charging Li- ion batteries due to its ability to intercalate additional Li ions81. Other vanadium oxides, like ζ-V2O5 have been reported to deliver high discharge capacities of 250 mAh g-1 in the voltage window between 4.0 and 2.0 V82, whereas vanadyl phosphates like ε-VOPO4 can deliver capacities exceeding 300 mAh g-1 between 4.5 and 1.6 V due to the ability of vanadium to carry out multi-electron redox processes83. Motivated by these properties, a sample of black, mixed-valence 1200 °C, which is easily prepared on the gram scale, was explored for its electrochemical properties as a Li-ion battery cathode. Initial slow cycling of 1200 °C at 20 mA·g-1 in the voltage window between 2.0 and 4.3 V showed a modest first cycle capacity of 150 mA·h·g-1 which stabilised around 120 mA·h·g-1 during the first few cycles (Fig. S53), with a smooth slopy voltage profile during the discharge curve, characteristic of a solid solution mechanism in disordered structures. Lowering the lower voltage cut-off to 1.0 V, however, reveals the appearance of a second plateau due to the intercalation of additional Li ions. In this voltage window, the specific discharge capacity of 1200 °C reaches ca. 300 mA·h·g-1 during the first cycle and steadily increases to 400 mA·h·g-1 during the next 15 cycles (Fig. S53). To further investigate the cyclability of 1200 °C under these conditions, we carried out cycling stability tests at 100 mA·g-1 between 1.0 – 4.3 V (Fig. S54). Our results show an increase in specific discharge capacity from 60 to 210 mA·h·g-1 during the first 100 cycles, an increase of more than three times the initial capacity, after which a slower steady decrease in specific capacity is observed (note the lower capacity due to the faster cycling of the material). This gradual increase in capacity suggests that initially amorphous 1200 °C undergoes a series of structural transformations upon consecutive cycles (Fig. S55), which allow the structure to accommodate extra Li ions up to a maximum of ca. 2.75 Li per unit formula at 100 mA·g-1. These structural changes are evident in the differential capacity curves, in which both oxidation and reduction features gradually shift to higher potentials, accompanied by an increase in capacity (Fig. S53). This is consistent with progressive, irreversible structural changes upon cycling, as observed in other vanadium oxide electrodes84,85. Unlike V2O5 or (NH4)V4O10 which can accommodate Li ions in between its layers, the suggested majority V4O9 structure indicates that 1200 °C would need to rearrange to be able to accommodate such amount of Li ions, and capacity may be further promoted by the transformation of remaining decavanadate ions in 1200 °C. A similar behaviour has previously been reported for V4O9 in aqueous Zn-ion batteries, where structure amorphization has been observed upon Zn2+ intercalation alongside a capacity increase during the first 100 cycles86. Additionally, previous studies of (NH4)V4O10 and V4O9 in Li-ion battery cathodes have shown significant structural changes upon lithium intercalation with V4O9 irreversibly transforming into a disordered rock salt structure59,62,87,88. The observed specific capacities of 1200 °C are in accordance with the reported V4O9, which can intercalate up to ~6 Li+ ( ~ 3.7-4 Li+ when cycled between 1.8 and 3.5 V)62, and with other reported mixed-valence vanadium oxides nanomaterials, which can incorporate multiple Li per formula unit and whose structure has been shown to change upon electrochemical cycling85. These possible structural transformations are also reflected in the evolution observed in the voltage profile of 1200 °C, which shows the appearance of two distinct steps in the charge curve, suggesting that at least two structural processes take place upon Li deintercalation. Even though we cannot discard the contribution of small amounts of amorphous POVs in, these results show that amorphous vanadium oxides can display intriguing electrochemical properties as Li-ion battery electrodes allowing for the reversible intercalation of an increasing amount of Li ions.

In conclusion, we have shown that thermal decomposition of mono-, bi-, and trimetallic POV molecular precursors follows complex pathways involving amorphous and intermediate phases, as characterised by a wide range of analytical methods including solid-state NMR, X-ray PDF analysis, and variable temperature X-ray diffraction. This approach enabled us to identify previously unknown mixed-valence V4O9-like amorphous intermediates exhibiting strong colouration and paramagnetism. Our results show the potential of using temperature control of the decomposition process to achieve specific vanadium oxidation states, relating to oxygen vacancy content in the final materials89,90, and to tune the degree of crystallinity and particle size.

When forming BiVO4 materials, the V-rich SSPs used mitigate vanadium loss during heating, ensuring no Bi-rich oxide byproducts are formed. Furthermore, solid-state NMR studies reveal temperature-dependent incorporation of Zn dopants into the structure of BiVO4 when a Zn containing trimetallic SSP is chosen, with associated site disorder found in the doped structure. This establishes a direct link between thermal processing and atomic-scale structural evolution.

Finally, we have discovered a kinetically stabilised polymorph of BiVO4, dubbed β-BiVO4, which is structurally analogous to β-SnWO4 and accessible exclusively via controlled thermal decomposition of SSPs (and remains stable after cooling). Interestingly, the coordination geometry of Bi in β-BiVO4 resembles that in the molecular precursor more closely than other BiVO4 phases. Theoretical analysis suggests the β-BiVO4 phase displays an increased band-gap relative to m-BiVO4, in part due to a less dense structure which weakens V 3 d orbital overlap and narrows the conduction band, whilst reduced involvement of the Bi 6 s in bonding, reduces the energy of the highest energy valence band energy states. This highlights the fascinatingly different properties in polymorphs of materials and our results show the potential of POV molecular precursors for accessing novel oxide phases with tailored properties.

More widely, this study showcases the possibilities of uncovering a multitude of material phases during molecule-to-material transformations. The use of specific and tuneable precursors which transform to materials at low temperatures will surely allow access to further metastable phases yet to be discovered experimentally and suggests many further exciting discoveries are possible in this space.

Methods

Preparation of precursors

Ammonium metavanadate, [NH4][VO3], (Sigma-Aldrich, >99%) was used directly from suppliers. [VO(OnPr)3] (Sigma-Aldrich, 98%), Bi(NO3)2·5H2O (Sigma-Aldrich, 98%). and ZnCl2 (Sigma-Aldrich) were stored in a N2-filled glovebox. Dimethyl sulfoxide (DMSO) was degassed before use and stored over 4 Å molecular sieves. All other solvents were used without further purification. All reactions using a Schlenk tube were performed under inert conditions (N2 atmosphere).

Compound 1 was synthesised based on established methods originally proposed by Lacharte et al91, full details can be found in the supporting information. Elemental analysis (predicted): % H, 3.03 (3.09); % N, 6.64 (7.16).

2 and 3-Zn were prepared following literature procedures15, full details can be found in the supporting information. For 2, elemental analysis (predicted): % C, 11.07 (11.33); % H, 2.49 (2.94); % S 15.14 (15.13), ICP-OES Bi: V metal ratio (predicted): 13: 3.6 (13: 4). For 3-Zn, elemental analysis (predicted, [Bi2V12O33Cl(DMSO)6]2[Zn(DMSO)6]·12(DMSO)): % C, 10.50 (12.88); % H, 2.54 (3.24); % S 14.52 (17.19). Note that sample has less co-crystallised DMSO that previously reported, and C, H, S analysis of this sample is a better match for [Bi2V12O33Cl(DMSO)6]2[Zn(DMSO)6]·4(DMSO) (calc. %: C, 10.63; H, 2.68; S, 14.19)15. N.B. Grinding the microcrystalline sample and placing it under vacuum overnight led to an amorphous sample which had slightly less co-crystallised DMSO (Elemental analysis, % C, 10.41; H, 2.54; S, 12.97). ICP-OES Bi: V: Zn metal ratio (predicted): 24: 3.9: 0.9 (24: 4: 1).

[VO(DMSO)5][NO3]2 and 3-VO were prepared as crystalline materials; synthetic details, crystal structure information and futher analysis is available in Supporting Note 1 in the supporting information.

Annealing of precursors to extended materials

1 to V2O5

1 ( ~ 0.25 g) was added to an alumina crucible and placed in a furnace. It was heated to varying temperatures using a heating ramp rate of 10 °C min-1. Each sample was held at the specified final temperature for two hours.

2 or 3-Zn to BiVO4 / V2O5

2 or 3-Zn (~ 0.25 g) was added to an alumina crucible and placed in a furnace. It was heated to varying temperatures using a heating ramp rate of 10 °C min-1. Each sample was initially held at 175 °C for two hours to decompose most of the associated DMSO, then the ramping rate was continued until the desired final temperature was reached and this temperature was then held for an additional two hours.

Powder X-ray diffraction (PXRD)

PXRD patterns of the materials were recorded using a Panalytical Empyrean diffractometer with a Cu target (Kα1 = 1.54056 Å) using background-free holders made of monocrystalline silicon. Variable temperature PXRD was performed on a Panalytical X’Pert Pro MPD equipped with a curved Ge Johansson monochromator, giving pure Cu Kα1radiation, and a solid state PiXcel detector running in 1D scanning mode. The sample was mounted in an Anton Paar HTK1200N furnace and aligned to the half-cut direct beam at room temperature to ensure it was in the centre of rotation of the goniometer. The height was then automatically adjusted using the Anton Paar software to account for the thermal expansion of the stage with temperature. The samples were heated at a rate of 5 °C/min with a wait time of 10 min before each measurement. Scans were made between 5°–70° 2θ with a step size of ~ 0.0131° and a count time of ~ 300 s per step. The lower energy threshold of the detector was adjusted to avoid measuring the fluorescence that comes from V under Cu radiation. Rietveld refinements were carried out using the software Topas Academic v7 (http://www.topas-academic.net/)92 and the V2O5 (ICSD 60767), m-BiVO4 (ICSD 100162), t-BiVO4 (ICSD 144629) phases and instrumental parameters were obtained from the Rietveld refinement of a Si standard92.

X-ray pair distribution function (XPDF) measurements

Total scattering data were collected at the I15-1 XPDF Beamline at Diamond Light Source, U.K. Powder samples were loaded into borosilicate glass capillaries (1.0 mm) and spun perpendicularly to the beam to improve powder averaging during data collection. Total scattering data collection was measured with a PerkinElmer area detector at an X-ray energy of 76.69 keV and data collection time of 600 s. The data was processed, and Fourier transformed using PDFGetX3 in the 0.5 < Q < 22.0 Å-1 range to obtain the final PDF as G(r)93. PDF refinements were carried out using the PDFGui software94 using the V2O5 (ICSD 60767), m-BiVO4 (ICSD 100162), V2O9 (ICSD 15041), (NH4)V4O10 (ICSD 230493) and (NH4)6V10O28 (ICSD 161217).

Solid-state NMR spectroscopy

Solid-state MAS NMR spectra of 1H (500.1 MHz) and 51V (131.5 MHz) were recorded at room temperature on a Bruker Avance III 11.7 T spectrometer equipped with 2.5 and 3.2 mm MAS probes. 1H spectra were recorded using an echo sequence. 51V spectra were recorded using single pulse excitation with a short (1 μs) pulse at 80-100 W RF power. 13C (213.8 MHz) spectra were recorded at room temperature on a Bruker Avance Neo 20 T spectrometer equipped with 3.2 mm MAS probe. 51V shifts were referenced using the unified IUPAC recommendation using solid adamantane as a secondary reference (38.48 ppm for the 13C of the CH2 carbon). All acquisition and processing details are given in Table S1.

Electrochemical measurements

1200 °C coatings for electrochemical cell testing were prepared by slurry casting onto an Al foil. The slurry was obtained by mixing 80 wt.% of active material (AM), 10 wt.% of Super C65 Carbon black, and 10 wt.% of polyvinylidene fluoride (PVDF, Merck) as binder with N-methyl-2-pyrrolidone. The resulting coating was dried at 110 °C overnight in a vacuum oven before use. Electrochemical tests were performed by preparing CR2032-type coin half cells using the coated cathode (14 mm diameter, mass loading: 2-3 mg cm2), Li metal anode (15 mm, Merck), Whatman glass microfiber separator (18 mm) and 1.0 M LiPF6 in EC:EMC (ethylene carbonate/methyl carbonate, 3:7 V/V), which were assembled inside an Ar-filled glovebox. Galvanostatic cycling tests were performed in the voltage ranges of 2.0–4.3 V and 1.0 and 4.3 V at current densities of 20 mA·g-1 and 100 mA·g-1.

Solid-state UV-vis spectroscopy

UV-Vis spectra were recorded using Shimadzu 2600i spectrometer in the diffuse reflectance configuration.

FT-IR spectroscopy

IR spectra were collected using a Agilent Technologies Cary 630 FT-IR spectrometer in the solid-state using an attenuated total reflectance setup.

C, H, N, S elemental analysis

Samples were sent for analysis at London Metropolitan University.

Inductively coupled plasma – mass spectrometry (ICP-MS)

Samples were dissolved in HNO3 (precursors) or HCl (annealed oxides) and diluted with water for V:Bi:Zn ratio analysis using an Agilent 7900 ICP-MS.

Thermogravimetric analysis – mass spectrometry (TGA-MS)

Samples were heated from room temperature to 800 °C or 1000 °C with a ramping rate of 5 °C or 10 °C min-1 using a Mettler Toledo TGA/DSC 1 instrument under N2 or air respectively. Samples were heated in the presence of N2 or air with a flow rate of 20 mL min-1.

EPR spectroscopy

X-band ( ~ 9.5 GHz) spectra were recorded on a Bruker EMX spectrometer at room temperature.

XANES

The V K-edge X-ray absorption near edge structure (XANES) data of the powder samples were collected on an easyXAFS300+ spectrometer. All measurements were performed in transmission using a silver X-ray source operated at 40 kV and 10 mA. Air scattering was minimised using a helium chamber.