2.1 Preparation of αβ-[Co₃Na(H₂O)₂(P₂W₁₅O₅₆)₂]¹⁷⁻ and α₂-[Ce(III)(P₂W₁₇O₆₂)]¹⁰⁻

The dissymmetrical tricobalt complex αβ-[Co₃Na(H₂O)₂(P₂W₁₅O₅₆)₂]¹⁷⁻ and the lanthanide complex α₂-[Ce(III)(P₂W₁₇O₆₁)]⁷⁻ were prepared according to the methods described in the literature [11,12].

2.2 NMR spectroscopy

Solution and solid-state ³¹P NMR spectra were recorded at room temperature and variable temperature on Bruker Avance 300 and Avance 400 spectrometers operating, respectively, at 7.1 and 9.4 T (Larmor frequencies 121.5 and 161.9 MHz, respectively). A 85% H₃PO₄ solution was used as a chemical shift reference.

The solution spectra were obtained from ca 0.02 mol/L solutions in unbuffered D₂O/H₂O (1:1). The one-pulse ³¹P 161.9 MHz MAS NMR spectra were recorded at spinning speeds of 12 and 34 kHz. A 0.5 μs pulse (π/9) was used with 10 ms recycling time.

The ³¹P echo–MAS NMR spectra recorded at 9.4 T were acquired at different spinning rates (from 30 to 35 kHz), with 30 ms repetition time. The first and second pulse durations were 2.6 (π/2) and 5.2 μs (π), respectively, and the delay between the pulses was calculated for synchronization with the rotor frequency (between 25 and 29 μs). For recording 7.1 T echo–MAS NMR spectra, the spinning speed was 29 kHz, the first and second pulses were 0.5 μs and 1 μs, respectively, and the recycling delay was 60 ms. In both cases, the recycling delay is sufficient to allow relaxation of the P(1) nuclei, while the P(2) resonances become more or less saturated.

3.1 ³¹P NMR of αβ-[Co₃Na(H₂O)₂(P₂W₁₅O₅₆)₂]¹⁷⁻

The “lacunary” tricobalt complex αβ-[Co₃Na(H₂O)₂(P₂W₁₅O₅₆)₂]¹⁷⁻ (shortly {Co₃P₄W₃₀}) derives formally from [Co₄(H₂O)₂(P₂W₁₅O₅₆)₂]¹⁶⁻ (shortly {Co₄P₄W₃₀}) by removal of one external Co²⁺ cation from the tetrametallic sheet, the resulting lacuna being occupied by a Na⁺ cation. [13]. In this dissymmetrical species (C_s symmetry), the two {P₂W₁₅} moieties connected to the metallic sheet are non-equivalent and they present four types of non-equivalent phosphorus atoms subdivided into two groups: the P atoms in the PW₆ subunits close to the paramagnetic centres, labelled P(1) and P(1′), and those in the PW₉ subunits far from the paramagnetic centres, labelled P(2) and P(2′) (Fig. 1b).

The ³¹P solution spectrum of the tricobalt complex αβ-[Co₃Na(H₂O)₂(P₂W₁₅O₅₆)₂]¹⁷⁻ is shown in Fig. 2. The two relatively narrow resonances of equal intensity at −10 and +21 ppm (fwhm = 110 Hz, 0.7 ppm; Fig. 2b) are attributed to P(2) and P(2′) atoms far from the cobalt atoms, while the two broader resonances at +1136 (fwhm = 560 Hz, 3.5 ppm) and +1692 ppm (fwhm = 730 Hz, 4.5 ppm; Fig. 2c) are assigned to P(1) and P(1′) atoms, which are characteristic of phosphorus nuclei close to the cobalt nuclei. The positions of the signals as well as their line widths reflect the different influence of the paramagnetic Co²⁺ cations. Actually, for both P(2) and P(2′), their relative positions with respect to the trimetallic sheet are similar with a rather large closest P–Co distance (>7 Å) resulting in moderate paramagnetic frequency shifts and broadening. The situation is more contrasted for P(1) and P(1′): while P(1), in the PW₆Co₃ subunit, is nearly equally distant from the three cobalt centres (d_P–Co ca 3.3 Å), P(1′), in the PW₆Co₂Na subunit, is far from the external Co atom (d_P–Co ca 4.5 Å). Therefore, we may tentatively assign the broadest and the most shifted signal (+1692 ppm) to P(1) and the less shifted one (+1136 ppm) to P(1′).

In order to obtain supplementary data, in particular the chemical shift tensors, solid-state ³¹P NMR spectra of the tricobalt complex were recorded (Figs. 3–5). Fig. 3a shows the wide-range one-pulse ³¹P MAS NMR spectrum of this complex obtained at 34 kHz spinning rate. This spectrum presents clearly two overlapping subsets of spinning-side bands (SSB): only two SSB are observed around the two intense signals around 0 ppm, while the second subset of low-intensity SSB extends over ca 3000 ppm (nearly 500 kHz at 162 MHz). Because of phasing errors due to the large spectral width, the spectrum is of very poor quality, which does not allow us to extract useful parameters for the second SSB subset. Fig. 3b presents the one-pulse ³¹P MAS NMR spectrum acquired at lower spinning rate (12 kHz) with a reduced spectral width. The two signals of P(2) and P(2′) present well-separated SSB patterns from which we are able to determine both the isotropic chemical shifts and the chemical shift anisotropy: δ_iso = −30 ppm, CSA ∼ 500 ppm and δ_iso = −46 ppm, CSA ∼ 700 ppm. It should be noted that the chemical shift values differ notably from those obtained in solution (+21 and −10 ppm); this may be due to bulk magnetic susceptibility effects and/or to additional paramagnetic interactions between neighbouring complexes in the crystal.

As expected, ³¹P NMR spectra recorded under echo–MAS conditions (Fig. 4) do not present the phasing errors of the single-pulse MAS spectrum. For the 9.4 T spectrum, the very short recycling delay (30 ms) leads to saturate the more slowly relaxing P(2) resonances (T₁ ca 1 s), which become nearly extinguished. In principle the separation between the P(1) and P(1′) SSB patterns will be sufficient to determine at least coarse values of the CSA of both nuclei; however, spreading of the SSB is obviously far beyond the range anticipated from the single-pulse MAS spectra (Fig. 3a). Actually the relatively long π pulse (5.2 μs) does not allow complete excitation over the bandwidth covered by the SSB. This was also confirmed by changing the offset of the excitation pulse, which leads to different experimental spectra (not presented here). Therefore it was impossible to access directly the CSA parameters from these spectra.

The problem of defective excitation may be partially circumvented by working at lower field, which reduces the effective required spectral width by the Larmor frequencies ratio. The echo–MAS spectrum obtained at 7.1 T (125 MHz) and at a rotor speed of 29 kHz is presented in Fig. 4b. The SSB patterns of the P(1) and P(1′) resonances extend approximately over the same range as for the single-pulse MAS spectrum (Fig. 3a). Note also that, while still saturated, P(2) and P(2′) nuclei give rise to signals considerably more intense than that at 9.4 T. Better excitation at 7.1 T originates, not only from the narrower spectral width, but also from the shorter pulses allowed on this spectrometer (π pulse of 1 μs). Despite the better quality of the resulting spectra, they suffer from overlap of the SSB patterns: actually the resonances from both P(1) and P(1′) are hardly resolved for some bands only and, due to the more uniform excitation of the spectral width, the P(2) and P(1) subspectra are no longer fully separated. Nevertheless a rough value of ca 3300 ppm may be estimated for the chemical shift anisotropy of both P(1) and P(1′) nuclei. Concerning their isotropic chemical shifts, no direct measurement may be done by varying the spinning speed, because of temperature effects (see below); they can be evaluated by comparison with the ³¹P NMR solution data at δ_iso ∼ +1700 and ∼+1160 ppm for P(1) and P(1′), respectively.

Although the data obtained for the four non-equivalent P nuclei are not accurate, they represent the first complete set of parameters for such a type of paramagnetic species; in particular the very large CSAs of both P(1) and P(1′) and the moderate ones for both P(2) and P(2′) agree with the different relaxation rates observed in solution as well as in the solid state.

3.2 ³¹P NMR thermometer

In order to interpret the effects observed on the solid-state ³¹P NMR spectra recorded at different spinning speeds, it is essential to determine the mean temperature as well as the temperature gradient in the rotating sample. For the purpose of measuring directly the temperature of the NMR sample, a number of NMR thermometers have been proposed in the literature.

Lead nitrate is commonly used for solid-state NMR [14–17]; it is based upon the large temperature dependence of the chemical shift of the heavy ²⁰⁷Pb nucleus.

For solution NMR, methanol and ethylene glycol are considered as standards for use at low and high temperatures, respectively; in that case one makes use of the different temperature dependence of δ (¹H) for the OH groups and the aliphatic protons [18]. Alternatively the great sensitivity of the cobalt-59 nucleus and the large temperature dependence of δ (⁵⁹Co) (up to 3 ppm/K) has allowed one to propose two diamagnetic highly symmetrical cobalt(III) complexes, namely K₃Co(CN)₆ and Co(acac)₃, as useful thermometers for multinuclear solution NMR spectroscopy [19]. For ¹³C NMR, in acetic acid–acetate buffer, the paramagnetic dysprosium nitrate induces paramagnetic shifts for the acetate resonances; the chemical shift difference is temperature dependent and for well-defined experimental conditions, the temperature sensitivity of Δδ has been reported to ca 4 Hz/°C (at 15 MHz), i.e. ca 0.3 ppm/°C [20].

In transition-metal substituted phosphorous-centered POMs, the paramagnetic shifts of the ³¹P resonances and also their sensitivity to temperature variations increase as the P–M distances decrease. Therefore, the chemical shift difference Δδ (³¹P) for two phosphorous nuclei not equidistant from the paramagnetic transition-metal cation should represent a measure of the sample temperature. As ³¹P NMR of POMs may be obtained easily in solution and the solid state, POMs can be considered as good candidates for ³¹P NMR thermometers in both states.

For these preliminary tests we have chosen a relatively simple monosubstituted cerium(III) POM complex, namely [Ce(III)(P₂W₁₇O₆₁)]⁷⁻ with only two non-equivalent phosphorous atoms (see Fig. 6) [12]. The choice of Ce(III) as substituting paramagnetic metal was directed by its moderate incidence on the ³¹P relaxation, leading to exceptionally narrow resonances as shown in Fig. 6.

³¹P spectra recorded in solution between 280 and 350 K exhibit two signals of same integrated intensity, as expected from the structure (Fig. 6). Whatever be the temperature, the two signals are, however, of unequal width, which indicates different relaxation rates. Increasing the temperature leads to a shift of both resonances to high frequency, with the highest shift observed for the broadest one. This is consistent with a closer proximity of the paramagnetic atom; this last resonance was therefore assigned to the phosphorous nucleus P(1), belonging to the PW₈Ce(III) subunit [21]. Although the chemical shifts δ_P(1) and δ_P(2) and consequently Δδ (Δδ = δ_P(1) − δ_P(2)) do not vary strictly linearly with the temperature, their variations remain monotonous and Δδ, with a mean variation of ca 5 Hz/K, can be used for temperature measurement for ³¹P NMR spectroscopy. Further experiments are in progress to quantify more precisely the calibration curve.

While [Ce(III)(P₂W₁₇O₆₁)]⁷⁻ appears as a suitable NMR thermometer for ³¹P solution NMR, it is, however, not the case for solid-state experiments. Actually, the solid-state ³¹P spectra of the hydrated potassium salt of [Ce(III)(P₂W₁₇O₆₁)]⁷⁻recorded between 300 and 350 K present two broad partially-overlapped main bands (Fig. 7). Contrary to what is observed in solution, the temperature variations are nearly the same for both signals and therefore their chemical shift differences do not vary sufficiently, as compared with their accuracy, to be a reliable temperature measurement. Moreover, some additional signals are present at higher temperature, likely from less hydrated samples.

We have, therefore, to search for other paramagnetic POMs which could present higher paramagnetic shifts and also a larger range of temperature stability; this work is currently under way.