1 Introduction

In the last few years, mesoporous materials attracted much attention due to their potential applications in several fields, including catalysis, polymerization, photochemistry…[1]. The silica-based MCM-41 structure is known to act as a bioactive material. Indeed, the intrinsic porosity of such material offers a wide range of possibilities for hosting molecules and the ability for drug delivery [2,3]. Recently, several researchers have been interested in the incorporation of bisphosphonates [4] in the pores of mesostructured silica, which may inhibit bone resorption. Confinement of such molecules could offer new opportunities such as bone reconstruction. The direct synthesis of Si–O–P mesoporous materials has been very rarely reported in the literature [5,6] and has received little attention.

In the field of materials, the local structure of nuclei can be probed efficiently by solid-state NMR spectroscopy. A large panel of techniques can be implemented for establishing connectivities between nuclei. It includes cross-polarization and J-derived experiments. The CP MAS (cross-polarization magic-angle spinning) sequence [7] relies on dipolar interaction. Such experiment establishes spatial proximities between X and Y nuclei. For X = ³¹P and Y = ²⁹Si, very few results have been reported so far in the literature. These results are related to silicon phosphide (involving direct ³¹P–²⁹Si bonds) [8] and silicophosphates involving ³¹P–O–²⁹Si groups [9]. J-derived experiments based on the isotopic scalar J coupling constants establish X–Y through-bond connectivities. In the early 1990s, such solid-state NMR sequences were implemented successfully by Fyfe et al. [10] and Eckert et al. [11]. More recently, several groups showed that a large number of J experiments could be easily transposed to solid-state NMR. Among these, we find the HMQC (heteronuclear multiple quantum coherence) sequence, which has been adapted for the following spin pairs: ¹H/¹³C, ¹H/¹⁵N, ³¹P/²⁷Al, ²⁷Al/¹⁷O, ³¹P/²⁹Si and ³¹P/⁷¹Ga [11–17]. The INEPT (insensitive nuclei enhanced by polarization transfer) sequence [18] is one of the most important pulse blocks in modern solution state experiments [19]. In solid-state NMR, the INEPT sequence was adapted for the study of mobile [20] and rigid [21] organic and hybrid systems. Examples related to inorganic components have been rarely reported in the literature [22–26].

In this paper, the complete study of silicophosphate phases such as Si₅O(PO₄)₆ and various SiP₂O₇ polymorphs is presented. The efficiency of ³¹P → ²⁹Si CP MAS, MAS-J-HMQC and MAS-J-INEPT experiments involving the {³¹P,²⁹Si} spin pair is demonstrated. The ³¹P → ²⁹Si CP MAS experiment has been then successfully applied for the structural characterization of Si–O–P derived mesoporous materials. Such materials were obtained via an aerosol process. The incorporation of phosphorus in the silica network at atomic level was clearly demonstrated.

2 Experimental: syntheses and solid-state NMR

Solid-state NMR experiments were performed on a Bruker Avance 300 spectrometer at B₀ = 7 T with ν₀ (³¹P) = 121.49 MHz and ν₀ (²⁹Si) = 59.63 MHz, using a 4-mm triple resonance Bruker MAS probe. Samples were spun at the magic angle using ZrO₂ rotors (5–14 kHz). ³¹P chemical shifts were referenced to 85% aqueous H₃PO₄. ²⁹Si chemical shifts were referenced to TMS. In the case of mesoporous materials, all experiments were obtained at low temperature (T = 238 K), thanks to a cooling unit (BCU-X), a control temperature unit, and a DVT probe. The calibration of the temperature was performed using lead nitrate, Pb(NO₃)₂. Full experimental details are given in the figure captions. The X-ray diffraction patterns of mesoporous materials were obtained using a D8 Advance Bruker diffractometer (Cu Kα radiation: λ = 1.54718 Å, 1–6° in 2θ with a step size of 0.02° in 2θ, scan rate: 5 s/step). For TEM experiments (Philips Technai 12), samples were dispersed in ethanol and dropped onto a copper grid covered with carbon.

The synthesis protocols are the following: Si₅O(PO₄)₆: TEOS (Si(OCH₂CH₃)₄), ethanol and distilled water were used as precursors (TEOS/EtOH/H₂O = 1:4:3). The phosphorus precursor H₃PO₄ (85%) was added (Si/P = 1:1) at room temperature. The reaction was slightly exothermic. The addition of 1% (molar ratio) of a paramagnetic complex (NiCl₂·6H₂O) was performed for NMR relaxation purposes, leading to a slightly green solution. After stirring at 25 °C, a transparent green wet gel was obtained (2 h). The final powder was obtained after heat treatment at 800 °C for 2 h. The structure of Si₅O(PO₄)₆ described by Mayer [27] (trigonal, R3¯, a = 7.869 Å, c = 24.138 Å, JCPDS: 70-2071) involves one unique P site and three inequivalent Si sites (two 6-fold coordinated atoms Si_VI, Si(1) and Si(2), and one 4-fold coordinated atom Si_IV, Si(3)). The Si(1)/Si(2)/Si(3) ratio is 1:2:2. The structure consists of [SiO₆] and [Si₂O₇] groups linked by [PO₄] groups. Each [PO₄] tetrahedron is surrounded by three Si_VI atoms (Si(2) (×2) and Si(1)) and one Si_IV atom (Si(3)) (Fig. 1). Si(1) is bonded to six equivalent oxygen atoms O(3) (Si(1)–O(3) = 1.756 Å, Si(1)O(3)P = 145.58°). Si(2) is bonded to six oxygen atoms, but two inequivalent oxygen atoms (O(2) and O(5)) are involved ((2) = 1.791 Å, Si(2)–O(5) = 1.756 Å, Si(2)O(2)P = 131.15°, Si(2)O(5)P = 151.38°). Si(3) is bonded to four oxygen atoms, namely O(4) (×3) and O(1) (Si(3)–O(4) = 1.611 Å, Si(3)O(4)P = 138.83°, Si(3)–O(1) = 1.593 Å, Si(3)O(1)Si(3) = 180°). The ³¹P MAS NMR spectrum reveals a unique resonance located at δ (³¹P) = −43.8 ppm, while the ²⁹Si MAS NMR spectrum exhibits three resonances located at δ (²⁹Si) = −119.3, −213.5, and −217.3 ppm (Fig. 1 and Table 1). A broad resonance located at δ (²⁹Si) ∼ −113 ppm associated with amorphous silica (side product) is also evidenced.

2.1 Mixture of SiP₂O₇ polymorphs and Si₅O(PO₄)₆

After the dissolution of NiCl₂·6H₂O (1%) in ethanol, H₃PO₄ (85%) was added, followed by TEOS (Si/P = 1:2). A gel was obtained after 2 h at room temperature, heated at 100 °C for 48 h, and further heat-treated at 1000 °C for 2 h. The X-ray diffraction (XRD) powder pattern of the sample (not shown here) indicates that besides the Si₅O(PO₄)₆ crystalline phase presented above, three polymorphs of SiP₂O₇ were synthesized as major constituents, namely a tetragonal (JCPDS: 22-1320) [28] and two monoclinic forms (JCPDS: 39-0189 and 25-0755) [29,30]. As shown in Fig. 2, the ³¹P MAS NMR spectrum reveals also the presence of the SiP₂O₇ cubic form (JCPDS: 22-1321), but as a very minor component. This particular phase was not clearly evidenced by XRD. For the SiP₂O₇ polymorphs, it is known from XRD data that the pyrophosphate groups (involving generally two non-equivalent P sites) are linked exclusively to Si_VI atoms (Fig. 2). The ³¹P and ²⁹Si isotropic chemical shifts of the various SiP₂O₇ polymorphs are reported in Table 1.

3 Results and discussion

The various pulse schemes used in this work will be first presented. The full characterization of Si₅O(PO₄)₆ and SiP₂O₇ polymorphs by CP MAS, MAS-J-HMQC and MAS-J-INEPT experiments are first shown, followed by the structural characterization of Si–O–P mesoporous materials.

3.1 Solid-state NMR methodology

3.1.1 ³¹P → ²⁹Si HETCOR CP MAS experiment

The ³¹P → ²⁹Si 2D HETCOR sequence presented in Fig. 3a is an extension of the 1D ³¹P → ²⁹Si CP MAS experiment. For the set-up, the MAS rotation frequency was fixed at 5 kHz. Even at this moderate MAS rate, the Hartmann–Hahn (H–H) profile must split into sharp spinning sidebands (because of the small heteronuclear dipolar interaction involving ³¹P and ²⁹Si). A VACP profile (variable amplitude cross-polarization) [32] was used during the contact time (t_CP), in order to broaden the H–H profile. The ²⁹Si acquisition was obtained under ³¹P CW decoupling. A saturation loop (characterized by Δ and n) was applied on the ³¹P channel before the first 90° pulse.

3.1.2 ³¹P–²⁹Si MAS-J-HMQC experiment

The 2D ³¹P–²⁹Si MAS-J-HMQC sequence is presented in Fig. 4a. T₂′ time constants [15] were measured by spin-echo experiments (90°–τ′–180°–τ′) under fast MAS for the unique ³¹P signal. Such time constants are characteristic of the lifetime of the involved coherences. The experimental curve was fitted by a single exponential decay in time domain according to:

I=I0exp(−2τ′/T2′)

(1)

leading to T₂′ (³¹P) = 65 ± 3 ms. This long T₂′ value is favourable to the creation and evolution of coherences during the HMQC pulse scheme. As the optimum τ value (Fig. 4a) for a given compound depends strongly on ²J_P–O–Si values and T₂′, various 1D ³¹P–²⁹Si MAS-J-HMQC experiments at variable τ were performed (not presented here). For one given spin pair, the evolution of the HMQC signal as a function of τ can be modelled by [12a,33]:

IHMQC=I0sin2(πJP–O–Siτ)exp(−2τ/T2′(P31))

(2)

Using Eq. (2) and taking into account a unique ³¹P–²⁹Si coupling constant and T₂′ (³¹P) = 65 ms, an averaged ²J_P–O–Si value could be extracted from the build-up data. This crude approach led to ²J_P–O–Si ∼ 15 Hz [15].

3.1.3 ³¹P–²⁹Si MAS-J-INEPT experiment

The pulse sequence for refocused 2D ³¹P–²⁹Si MAS-J-INEPT experiment is shown in Fig. 5a. The original solution state technique [18] was adapted here for rotating solids and for a new pair of nuclei {³¹P,²⁹Si}. Fast magic-angle spinning (usually 14 kHz) averages the chemical shift anisotropy and the heteronuclear dipolar couplings to zero, leaving only the scalar couplings and the isotropic chemical shifts [21]. The refocused INEPT pulse sequence consists of the following steps: a 90° pulse is applied to the ³¹P channel, followed by an evolution delay optimized to achieve ³¹P anti-phase magnetization. The refocusing of the isotropic chemical shifts is obtained by the simultaneous application of 180° pulses on both phosphorus-31 and silicon-29 channels (only the ²J_P–O–Si couplings have to be taken into account). For a pair of ³¹P and ²⁹Si nuclei involved in a ³¹P–O–²⁹Si group, anti-phase phosphorus-31 coherence with respect to silicon-29 is created after the first τ–π–τ period. The two simultaneous 90° pulses lead to anti-phase silicon-29 coherence. The refocusing period τ′–π–τ′ is then applied to obtain in-phase ²⁹Si magnetization. ³¹P decoupling is finally applied during the acquisition of the ²⁹Si signal.

In the MAS-J-INEPT experiment, τ and τ′ delays must be carefully optimized. Such parameters involve not only the value of the ²J_P–O–Si coupling constants, but also the relaxation time T₂′ associated with each site. Table 2 presents an estimation of the T₂′ (²⁹Si) and T₂′ (³¹P) values. For the refocused INEPT experiment, and considering a unique ²J_P–O–Si coupling constant, the signal intensity is given by [19,34]:

IINEPT(τ,τ′)=I0sin(2πJP–O–Siτ)sin(2πJP–O–Siτ′)cosn−1(2πJP–O–Siτ′)exp(−2τ/T2′(P31))exp(−2τ′/T2′(S29i))

(3)

where n is the bond multiplicity (i.e., SI_n spin system with S ≡ ²⁹Si and I ≡ ³¹P). An optimized τ delay of 11.4 ms will be used as a constant thereafter.

Table 2

Estimated T₂′ (²⁹Si) and T₂′ (³¹P) obtained using MAS spin echo experiments and Eq. (1) for the Si₅O(PO₄)₆ phase (Si(1), Si(2), Si(3) and P sites)

	T2′ (29Si) (ms) [26]	T2′ (31P) (ms) [15]
Si(1)–O–P	41 ± 4	65 ± 3
Si(2)–O–P	55 ± 4
Si(3)–O–P	63 ± 5

3.1.4 Results

Figs. 3b, 4b and 5b present the 2D ³¹P–²⁹Si CP MAS HECTOR (with t_CP = 40 ms), 2D ²⁹Si–³¹P MAS-J-HMQC (τ = 25 ms) and 2D ³¹P–²⁹Si MAS-J-INEPT (τ = 11.4 ms and τ′ = 4.6 ms) spectra of the Si₅O(PO₄)₆SiP₂O₇ mixture, respectively (see the synthesis above). Three cross-peaks associated Si₅O(PO₄)₆ phase are systematically observed (the unique ³¹P site at −43.8 ppm is J and D coupled to the three ²⁹Si resonances – Fig. 1). Moreover, the 2D spectra reveal the presence of numerous cross-peaks (shown in the expansions in Figs. 3b, 4b and 5b), located in the 6-fold coordinated ²⁹Si chemical shift range (∼−215 ppm). These cross-peaks are safely assigned to SiP₂O₇ polymorphs. Therefore, D and J 2D experiments are suitable to edit the ³¹P MAS spectra in terms of pairs involved in each SiP₂O₇ pyrophosphate phase (Fig. 2). Moreover, the unique ²⁹Si isotropic shift corresponding to each SiP₂O₇ polymorph is determined with great accuracy. All ³¹P and ²⁹Si isotropic chemical shifts are reported in Tables 3 and 4 (HECTOR and HMQC/INEPT experiments, respectively). Comparing the three experiments, we note that in the 2D ²⁹Si–³¹P MAS-J-HMQC and MAS-J-INEPT spectra, no cross-peaks corresponding to the monoclinic 2 form of SiP₂O₇ are observed. Such cross-peaks are observed in the case of the 2D HETCOR experiment (Fig. 3b). It is assumed that the J coupling constants are markedly different for the various SiP₂O₇ phases, leading to the underestimation of the monoclinic 2 form in J-derived experiments. The absence of observable cross-peaks for the cubic forms in the three experiments is due to the low amount of this particular phase (Fig. 2).

Table 3

³¹P and ²⁹Si isotropic chemical shifts corresponding to the projections in the 2D ³¹P–²⁹Si CP MAS HETCOR experiment for the mixture of Si₅O(PO₄)₆ and SiP₂O₇ polymorphs

Phase (JCPDS)	δiso (ppm)
31P HETCOR (Fig. 3b)	29Si HETCOR (Fig. 3b)
Si5O(PO4)6 (70-2071)	−43.8	−119.6
−213.8
−217.6
SiP2O7 tetragonal (22-1320)	−45.4	−213.2
−52.9
SiP2O7 monoclinic 1 (39-0189)	−47.6	−215.5
−55.4
SiP2O7 monoclinic 2 (25-0755)	−46.0	−215.4
−49.4	−215.6

Table 4

³¹P and ²⁹Si isotropic chemical shifts corresponding to the projections in the 2D ³¹P–²⁹Si MAS-J-HMQC and ³¹P–²⁹Si MAS-J-INEPT experiments for the Si₅O(PO₄)₆/SiP₂O₇ polymorph mixture

Phase (JCPDS)	δiso (ppm)
31P HMQC (Fig. 4b) [15]	29Si HMQC (Fig. 4b) [15]	31P INEPT (Fig. 5b) [26]	29Si INEPT (Fig. 5b) [26]
Si5O(PO4)6 (70-2071)	−43.8	−119.1	−43.9	−119.3
−213.3	−213.5
−217.0	−217.3
SiP2O7 tetragonal (22-1320)	−45.5	−212.8	−45.6	−213.0
−52.9	−52.9
SiP2O7 monoclinic 1 (39-0189)	−47.6	−214.9	−47.7	−215.3
−55.4
−55.3

Si₅O(PO₄)₆ acted also as a model compound for the measurement of individual ²J_P–O–Si coupling constants via the MAS-J-INEPT experiment. The INEPT build-up curves obtained vs τ′ (at fixed τ = 11.4 ms) for Si(1), Si(2) and Si(3) are presented in Fig. 6a, b and c, respectively. The variation of the intensity is clearly dependent on both the number of P atoms bonded to the silicon atoms (n = 6 for Si(1), Si(2); n = 3 for Si(3)) and the crystallographic characteristics of the various sites (for Si(1) and Si(2)). Eq. (3), as well as the T₂′ constants presented in Table 2, were used for the fitting of the experimental curves for Si(1) and Si(3) (with n = 6 and n = 3, respectively). For both Si sites, a unique ²J_P–O–Si coupling constant is involved. The extracted J value corresponds to ²J_P–O–Si ≈ 15 and 12 Hz for Si(1) and Si(3), respectively [26]. Obviously, the Si(2) build-up curve differs drastically from the one obtained for Si(1), even though Si(1) and Si(2) correspond both to 6-fold coordinated silicon atoms. In the case of Si(2), two different ²J_P–O–Si coupling constants have to be a priori considered corresponding to Si(2)–O(2)–P and Si(2)–O(5)–P bonding paths (Fig. 6b). Considering two distinct J coupling constants (J₁ and J₂, SI₃I′₃ spin system), the product operator formalism [19] leads to:

IINEPT(τ,τ′)=I0cos2(2πJ1τ′)cos2(2πJ2τ′)[sin(2πJ1τ)sin(2πJ1τ′)cos(2πJ2τ′)+sin(2πJ2τ)sin(2πJ2τ′)cos(2πJ1τ′)]exp(−2τ/T2′(P31))exp(−2τ′/T2′(S29i))

(4)

The fitting of Si(2) curve (Fig. 6b) leads to J₁ ≈ 14 Hz and J₂ ≈ 4 Hz [26]. The obtained J₂ value appears small, but we note that two distinct bond angles are involved in the case of Si(2): Si(2)–O(2)–P ≈ 131° and Si(2)–O(5)–P ≈ 151°. This is assumed to explain the strong difference between J₁ and J₂.

MAS-J-HMQC, MAS-J-INEPT and HETCOR CP MAS experiments are comparable in terms of experimental time. However, it must be noted that very long contact times (up to 40 ms) have to be used for efficient CP transfer (the ³¹P/²⁹Si dipolar coupling is estimated to be ∼340 Hz in Si–O–P groups). The HMQC and INEPT sequences are obviously less demanding in terms of RF power, as short pulses are involved. Nevertheless, care must be taken when choosing the evolution periods, as T₂′ relaxation processes have to be taken into account.

3.2 Mesoporous Si–O–P materials

Fig. 7a shows the X-ray diffraction pattern of the SiP_0.1-aerosol sample (see Section 2). The two observed peaks can be indexed as (100) and (110) reflections of a 2D hexagonal mesostructure. The TEM image of the powder presented in Fig. 7b shows the spherical morphology of the obtained nanoparticles (with diameters ranging from ∼50 to 300 nm).

¹H → X CP MAS experiments performed at room temperature showed systematically poor efficiency of polarization transfer. It is assumed that proton mobility strongly reduces the efficiency of the CP process. CP efficiency was indeed regained at low temperature (T = 238 K). The following spectra were recorded at T = 238 K. The ³¹P MAS spectrum of the SiP_0.1-aerosol sample is shown in Fig. 7c. Four resonances at ∼2.5, −9.5, −23.0 and −38.3 ppm are evidenced, which can be assigned to Q′_N species, N ranging from 0 to 3 (see Table 5 for the definition of Q′_N and for the assignments). Protonated phosphate groups are evidenced by ¹ H→ ³¹P CP MAS experiment at t_CP = 2 ms (Fig. 7d). At this contact time, the resonance at ∼2.5 ppm corresponding to the Q′₀ specie is underestimated due to fast molecular motion even at low temperature. The other three resonances are assigned safely to protonated species, in agreement with notations presented in the literature [35]. The ²⁹Si MAS spectrum is presented in Fig. 7e. Two broad resonances at δ ∼−103 and ∼−111 ppm correspond to Q₃ and Q₄ species (Table 5). The corresponding ¹H → ²⁹Si CP MAS experiment is presented in Fig. 7f. Q3 units are overestimated, as it is usually the case in the case of silica gels and derivatives. Moreover, a weak amount of Q₂ units is evidenced: the corresponding resonance is located at ∼−91 ppm. It has to be noticed that the notations used in Table 4 are rather loose. Indeed, various combinations of ²⁹Si–O–Si/²⁹Si–OH/²⁹Si–O–P bonds can lead to the same isotropic chemical shift value. Therefore, ³¹P–²⁹Si CP MAS experiments appear as a valuable tool of investigation for the characterization of dipolar couplings within the mesoporous structure. The corresponding spectrum of the aerosol sample is presented in Fig. 7g. The signal-to-noise ratio is rather low due to the low amount of phosphorus used in the synthesis. Nevertheless, two resonances located at δ ∼−101 (hardly discernable) and ∼−113 ppm can be detected. Such observations indicate the presence of Si⋯P dipolar contacts. These preliminary results are encouraging for an in-depth characterization of the structure of the Si–O–P mesoporous materials. ¹H → ²⁹Si → ³¹P CP MAS experiments, as well as J experiments, should permit complete editing and assignment of both ³¹P and ²⁹Si resonances in such materials.

Table 5

Proposed assignments for ³¹P and ²⁹Si isotropic resonances

δ (ppm)	Assignment
(31P)
∼0	OP(OH)3
−10/−20	OP(OH)2(OP/OSi)
−20/−25	OP(OH)(OP/OSi)2
−30/−35	OP(OP)n(OSiIV)m
∼−45	OP(OSiIV)n(OSiVI)m
(29Si)
−90/−95	Si(OSi)2(OH)2 Q2
∼−100	Si(OSi)3(OH) Q3 or Si(OSi)(OH)2(OP)
∼−110	Si(OSi)4 Q4 or Si(OSi)2(OP)(OH)
−110/−120	Si(OSi)4−x(OP)x with x = 1–3

4 Conclusions

This paper has been devoted to the methodological characterization of the {³¹P,²⁹Si} spin pair by solid-state MAS NMR techniques. Experiments based on J and heteronuclear dipolar interaction were implemented using crystalline silicophosphate phases as model compounds. Si₅O(PO₄)₆ and various polymorphs of SiP₂O₇ were studied systematically by 2D ³¹P → ²⁹Si CP MAS HETCOR, MAS-J-HMQC and MAS-J-INEPT experiments. The editing role of such experiments was clearly demonstrated. The ³¹P–²⁹Si INEPT build-up curves allowed the measurement of ²J_P–O–Si coupling constants. An unexpected dependence on both the coordination mode of the Si nuclei and the involved crystallographic paths has been demonstrated. Finally, ³¹P → ²⁹Si CP MAS experiments were adapted to the characterization of Si–O–P mesoporous materials obtained by aerosol process. The results showing ³¹P···²⁹Si dipolar contacts are encouraging and will be compared in a near future to those obtained by MAS-J-derived techniques.