XRD patterns of samples Ni01 and Ni02 exhibit hk bands and 00l reflections characteristic of phyllosilicates (Figure 1a). More specifically, the bands appearing at ca. 20, 35, 55 and 62° 2θ, can be indexed as the (02, 11), (13, 20), (15, 24, 31) and (06, 33) bands of a smectite mineral [11,12,23,24,25]. The number of octahedra occupied by metal elements defines the di- or trioctahedral character of the clay, i.e., one or no vacancy out of three octahedra respectively. The position of the (06) band is related to this character, with associated d₀₆₀ periodicity values of 1.49 and 1.52 Å respectively. Figure 1b displays the (060) bands of samples Ni01 and Ni02. The d₀₆₀ periodicity of sample Ni01 is dominated by a peak centered at ca. 1.49 Å making it a solely dioctahedral mineral. Sample Ni02 exhibits two components, a main one at 1.49 Å and a secondary one at 1.51 Å, the latter demonstrating a partial trioctahedral character of the layers. The position of the (001) peak (Figure 1a), observed at ca. 12.7 and 13.2 Å for Ni01 and Ni02 samples respectively, gives the value of the interlayer distance and is typical of the smectite mineral family. The much broader (001) peak observed for Ni01 sample is characteristic of a reduced size of the coherent scattering domains perpendicular to the layer plane. To confirm these swelling properties, sample Ni02 was further subjected to hexadecyltrimethylammonium (HDTMA) intercalation. The intercalated hybrid-mineral shows a d₀₀₁ spacing at ca. 16.8 Å (Figure 1c), which corresponds to values usually observed for a laterally intercalated monolayer of HDTMA in montmorillonite minerals [26,27,28,29].

The two samples were analyzed for their chemical content (wt.%): Ni01 (Si: 27.40; Al: 9.00; Ni: 3.20; Na: 0.40; F: 0.47) and Ni02 (Si: 25.50; Al: 8.20; Ni: 7.50; Na: 0.40; F: 0.35). The amounts of Si, Al, Na and F are relatively similar in both samples, but the amount of Ni is more than doubled in sample Ni02 compared to sample Ni01, as expected from the chemical composition ot the initial reacting mixtures. Additionally, it seems that the Al content slightly decreases as the Ni content increases, as one would expect with a partial substitution of Al by Ni in the octahedral layer. This higher Ni content in the Ni02 sample corroborates the more intense color observed for that sample with respect to sample Ni01.

Typical SEM and TEM pictures of samples Ni01 and Ni02 are shown on Figure 2. The synthetic inorganic compounds exhibit the well-known sand rose or cornflake morphology (Figure 2a), which is typically observed for natural clay minerals [30,31,32], especially for swelling phyllosilicates of the smectite family.

TEM micrographs of our samples (Figure 2b) show that the synthetic minerals seem to possess a layered structure characteristic of smectite clay minerals, either natural [27,33] or synthetic [15].

The thermogravimetric and thermal analysis curves of samples Ni01 and Ni02 are displayed in Figure 3.

The curves are very similar for both samples and exhibit a first main weight variation below 110 °C of ca. 14.4 and 15.4 wt.% for samples Ni01 and Ni02, respectively. This weight loss is associated with a large endothermic peak on the heat flow vs. temperature curve and they are both associated with clay mineral interlayer dehydration [34,35,36,37]. Both samples experience also two weight losses at ca. 500 and 680 °C, combined to endothermic peaks, which are of the order of 3.0–3.3 and 1.4–1.5 wt.%, respectively. These features are attributed to the dehydroxylation of the clay layer. The dehydroxylation appearing at the highest temperature is characteristic of montmorillonite clay minerals [34,35,36] and was also observed for a synthetic montmorillonite-like mineral, which was hydrothermally synthesized at higher pressure and temperature than in our case [15]. The dehydroxylation appearing at lower temperatures has already been observed in some natural montmorillonites [34,37], and is also a peculiar feature of our synthetic Mg-Al or Zn-Al montmorillonite-like minerals, containing only Mg and Al or Zn and Al in the octahedral layer [13,14]. These features correspond to two types of structural hydroxyls, assigned to trans- and cis-vacant sites, i.e., the two occupied octahedra are positioned perpendicularly or parallel to the line formed by the hydroxyls in the structural plane, and associated with the lower and higher temperature dehydroxylation peaks respectively [35,36]. Thus, the presence of cis- and trans-vacant sites may arise from structurally different layers, i.e., cis or trans 2:1 layers [35,36], or from a clustering of either the dioctahedral elements or the hydroxyl/fluorine groups or both during the synthesis process. Such possible clustering was observed in montmorillonite-like samples synthesized at low temperature and pressure and in slightly acidic and fluoric medium (see [13,14] and vide infra). As mentioned above montmorillonite-like minerals synthesized at higher temperature and pressure seem to contain only cis-vacant sites in their structure making them more homogeneous minerals. As a consequence of the previous discussion, the dehydroxylation observed at ca. 450–500 °C might be considered as a fingerprint of montmorillonite-like minerals synthesized in conditions similar as those presented here and in our previous studies [13,14]. Additionally, it is noteworthy that dehydroxylations of the Ni-Al montmorillonite-like minerals appear at temperatures slightly higher (ca. 20 to 40°) than those observed for the Mg-Al or Zn-Al synthetic montmorillonites [13,14], making the former slightly more thermally stable.

The specific surface areas (SSA) of the samples are 115 and 134 m² g⁻¹ for Ni01 and Ni02, respectively. These values are higher than those usually observed for natural montmorillonites [38,39,40,41], though they are not unexpected since a synthetic Mg-Al montmorillonite showed a SSA of ca. 128 m² g⁻¹ [42]. However, this synthetic montmorillonite-like clay mineral was recently studied for its dissolution kinetics [43], and a purified sample exhibited a 104 m² g⁻¹ SSA. Thus the high SSA values observed for the Ni-Al montmorillonite-like synthetic minerals are probably due to the presence of an additional silica amorphous phase detected by ²⁹Si NMR (vide infra). This assumption is sustained by the fact that the SSA value increases from sample Ni01 to Ni02, as the amount of silica seems to increase (vide infra).

In order to see whether there is also Al for Si substituted in the tetrahedral sheet, ²⁷Al and ²⁹Si solid state MAS-NMR experiments were carried out.

The ²⁹Si MAS-NMR spectra of samples Ni01 and Ni02 exhibit a main peak at ca.−93 ppm, a shoulder on its left side at ca.−88 ppm and a very broad, though relatively important, peak at ca.−108 ppm (Figure 4). These features can be readily assigned to the Si(0Al) environment [44,45,46,47], i.e., Si surrounded by three Si in the tetrahedral sheet, the Si(1Al) environment [44,45,46,47], i.e., Si surrounded by two Si and one Al in the tetrahedral sheet, and to amorphous silica [14], respectively. The Si(1Al) environment occurrence is characteristic of Al for Si substitutions in the tetrahedral sheet. A slight shielding of the Si(0Al) environment chemical shift, ca. 0.2 ppm, is observed between samples Ni01 and Ni02. This slide of the chemical shift to higher field values may be related to the partial trioctahedral character of Ni02 sample [46,47], in agreement with XRD observations (vide supra). The signal to noise ratio of the spectra is relatively low and probably due to the presence of amorphous silica. The presence of amorphous silica was also observed in the previously synthesized Mg-Al or Zn-Al montmorillonite-like phyllosilicates, but not to such an extent [14]. This demonstrates clearly that the conditions of synthesis of montmorillonite-like clay minerals, consequently their occurrence in nature, might be strongly affected by the type of metal cations present in the medium. Because of the low quality of the spectra, it was not possible to properly simulate them and gather more information.

²⁷Al MAS-NMR spectra of samples Ni01 and Ni02 (Figure 5) present a main asymmetric peak at ca. 3 ppm characteristic of the octahedral aluminum environment (Al_VI) [44,46,48,49]. The resonances observed in the range 50–70 ppm are assigned to the tetrahedral aluminum environment (Al_IV) [44,46,48,49]. The presence of these resonances confirms that there are both octahedral and tetrahedral substitutions in these synthetic montmorillonite-like Ni-Al phyllolicates. The Ni01 spectrum shows only one asymmetric peak at ca. 68 ppm, while the Ni02 spectrum shows two peaks, an asymmetric one at ca. 68 ppm and a smaller one at ca. 55 ppm. In the latter case, the existence of two different Al_IV resonances might be correlated to the clustering of the metallic elements in the octahedral sheet (vide infra).

Since fluorine atoms may have two or three octahedrally coordinated elements (Al, Ni) in their environment, fluorine was used as a local probe for the determination of the various tri-hexagonal sequences present in the octahedral sheet. Huve has shown that depending on the dioctahedral or the trioctahedral character a relationship can be established between the average of the electronegativity of the octahedral sheet elements (e_□) and the theoretical ¹⁹F chemical shift (δ_19F) [50]. In the case of dioctahedral clays the following relation can be used: δ_19F = 117 e_□ − 247, whereas for trioctahedral clays it is the following one: δ_19F = 50 e_□ − 238. Assuming the presence of Al-Al- , Al-Ni- or Ni-Ni-Ni tri-hexagonal sites, their theoretical chemical shifts should be −132, −121 and −150 ppm, respectively. However, the presence of a paramagnetic center, e.g., Ni, can cause (sometimes large) chemical shift changes, known as paramagnetic shifts, and in this case the experimental chemical shift can greatly differ from the theoretical one. ¹⁹F MAS-NMR spectra of samples Ni01 and Ni02 (Figure 6) present a main resonance at ca.−132 ppm, which is assigned to the Al-Al- environment [13,14,51], one fluorine surrounded by two aluminum and one vacancy in the three neighboring octahedra (Figure 7). Both spectra exhibit a peak at ca.−122 ppm, which may be assigned to an Al-Ni- environment [50]. Finally, both spectra show a third resonance at ca.−139 ppm, which might be assigned to the Ni-Ni-Ni environment [50], for which a chemical shift of about 10 ppm is observed with respect to the theoretical value. The chemical shift change observed in the case of the Ni-Ni-Ni environment and not for the Al-Ni- environment might be explained by the much stronger local concentration of Ni. The Ni-Ni-Ni environment is thus observed in both samples and especially in sample Ni01, which is a purely octahedral mineral according to XRD (vide supra). Since no such trioctahedral clusters are expected in sample Ni01, this is evidence of the clustering of Ni, but also of the vacancies and possibly of the fluorine in the octahedral sheet.

These NMR observations of the existence of a clustering of the metal elements in the octahedral layer support the results deduced from Ni K-edge EXAFS (vide infra).

Samples Ni01 and Ni02 were analyzed on synchrotron facilities to perform EXAFS measurements at the Ni K-edge (8333 eV). The experimental χ(k) spectra of the two samples and their Fourier transformed (FT) magnitudes are displayed in Figure 8a and b, respectively. One can easily observe that both sets of data are very similar, showing that the local environment of Ni atoms is much the same in both samples. The first and second main peaks on the FT magnitudes are assigned to the nearest neighbor (NN) and next nearest neighbor (NNN) shell contributions, respectively (Figure 8b). There are only tiny differences in the data sets and they are actually best noticed for values beyond 3 Å in the R space. As a consequence, differences between samples will only affect, slightly, the NNN shell contributions. The Ni NN shell contribution consists of oxygen atoms, and the NNN shell contains the Ni and Al atoms present in the surrounding octahedra and Si and Al nuclei present in the closest tetrahedra. While comparing this set of Ni K-edge data to those obtained at the Zn and Mg K-edges for montmorillonite-like phyllosilicates [13], one can easily observe that the NNN shell contributions of the FT spectra are more affected by the metallic elements content in the case of the Zn-Al or Mg-Al samples than in the case of the Ni-Al ones.

The best simulations of the Fourier-filtered (FF) contribution of the NN peak for both Ni-Al samples are represented in Figure 9 and the structural parameters derived from these simulations are reported in Table 1. Ni nuclei have the same octahedral environment (six oxygen neighbors) in both Ni01 and Ni02 samples, as was observed previously for the synthetic montmorillonite-like phyllosilicates containing Zn and Al or Mg and Al (see samples Zn01, Zn02, Mg01 and Mg02 in Table 1) [13]. Additionally, one may see that the M-O distance is relatively close in the samples containing the same metallic element (M), meaning that it is independent of the amount contained in the sample. Though the M-O distance increases with the diameter of the M element, especially in the case of the two largest nuclei, Mg and Zn (Table 1; atomic radii: Ni²⁺− 0.69 Å, Zn²⁺− 0.74 Å, Mg²⁺− 0.72 Å). When comparing these M-O distances to the Al-O distance determined in Zn-Al or Mg-Al montmorillonites (ca. 1.93 Å) [52], or to the theoretical Mg-O and Al-O distances (2.15 and 1.91 Å, respectively) described for an ideal montmorillonite structure (Figure 1 in this reference) [53], one can deduce that the octahedron size increase induces enhanced distortions of the layer structure.

The best simulations of the FF contribution of the NNN peaks obtained for samples Ni01 and Ni02, with the number of M element fixed to four, are represented in Figure 10 and the structural parameters derived from these simulations are reported in Table 1, together with the parameters obtained for the simulations of NNN shell contributions at the Zn K-edge of samples Zn01 and Zn02 [13]. The mean Ni-Ni and Ni-Al interatomic distances slightly increase from 3.00 and 3.02 Å respectively in Ni01 to 3.03 and 3.05 Å in Ni02 (Table 1). Ni-Ni and Ni-Al distances in the octahedral sheet are very close and even Ni-Si distances between the elements located in the octahedral and tetrahedral sheets are similar.

The proportion of Ni nuclei in the NNN shell increases with the Ni content, as one would expect (Table 1). Though these proportions, 1.2 and 2.1 in Ni01 and Ni02 respectively, are much larger than expected for a statistical distribution of the M elements in the octahedral sheet. The central Ni nucleus should indeed not witness any other Ni nucleus in its NNN shell in the case of Ni01 sample, and it should see less than one Ni nucleus in the case of the Ni02 sample, as in the case of Zn-containing phyllosilicates [13]. This is the evidence of the clustering of the M elements in the octahedral sheet of the Ni-Al montmorillonite-like phyllosilicates, containing only Ni and Al in this sheet, which confirms the ¹⁹F solid-state NMR observations. However, if there are similarities in the clustering of the M elements in both Ni-Al and Zn-Al containing samples, there are slight discrepancies in terms of the proportions of the M elements surrounding the central M nucleus, and there are significant differences concerning the values of the M-M distances. In fact, the proportion of Zn surrounding the central Zn nucleus seems to be slightly larger, for similar M elements content, than the proportion of Ni surrounding the central Ni nucleus (Table 1). This observation would lead to a greater clustering of the M elements in the case of the Zn-Al phyllosilicates, containing only Zn and Al in the octahedral sheet. Additionally, the Ni-Ni, Ni-Al and Ni-Si distances in the Ni-Al containing samples are close to the ideal M-M distance [13,55]. The Ni environment is less affected by the Ni content than the Zn environment is by the Zn content in the Zn-Al samples. The Zn-Zn distance is indeed longer than the Zn-Al distance (Table 1), consequently longer than the ideal M-M distance, and the Zn-Si distance increases with the Zn content of the samples (Table 1). These observations highlight the existence of distortions in the layer structure of the Zn-Al phyllosilicates, which may thus counterbalance the effect of the higher clustering of the elements compared to the Ni-Al phyllosilicates. Additionally, the Ni-Ni distances being significantly shorter than the Zn-Zn ones, also supports the idea of lesser clustering in Ni-Al montmorillonites than in Zn-Al montmorillonites. All these considerations explain the improved local order around Ni nuclei, which is sustained by lower Ni-Ni and Ni-Al Debye-Waller factors. As a consequence, the local environment of the Ni nucleus seems to be more organized than the local environment of the Zn nucleus in these kinds of montmorillonite-like phyllosilicates.

As a complement to the NNN shell study we would like to highlight that if the best simulations are obtained with a number of four M elements surrounding the central Ni atom, simulations using three M elements were also performed, but with less success (data not shown). Clearly, in the case of a statistical distribution of the M elements in the octahedral layer of a dioctahedral phyllosilicate, the number of hexagonal sites occupied around a given one is three out of six (Figure 7). However, if sample Ni01 is dioctahedral in nature, sample Ni02 is partially trioctahedral, as observed by XRD (vide supra). Thus, we conducted a first series of fit with three neighbors out of six around the central atom (dioctahedral character). Modifying the number of NNN elements surrounding the central Ni atom is possible since the FEFF input file used for the theoretical FEFFIT calculations (see experimental section below) contains coordinates of up to six surrounding M sites forming as many Ni-M pairs (M being Ni or Al). As expected, because of the results for Zn-Al montmorillonites [13], the simulations performed with three M neighbors did not fit well. Simulations with four neighbors were more successful, as explained above, which supports the existence of a clustering of the octahedral vacancies.

Al and Mg K-edge EXAFS analysis of synthetic montmorillonites containing Mg and Al in their octahedral sheet has revealed that the mean first shell coordination radii of Al and Mg are quite different: in fact the bond lengths have been found to be R_Al-O = 1.95 ± 0.03 Å and R_Mg-O = 2.11 ± 0.02 Å [52,54]. In the ideal montmorillonite structure, the smaller Al³⁺ cations (ionic radius: 0.54 Å) are located in M₂ sites (R_Al-O = 1.95 Å) and the bigger Mg²⁺ cations (ionic radius: 0.72 Å) are located in M₁ sites (R_Mg-O = 2.15 Å). Those values are very close to the Al-O and Mg-O bond lengths derived from Al and Mg K-edge EXAFS analysis performed on montmorillonites synthesized in fluoride medium [13,52]. Some of the octahedral sites are empty, in order to ensure a mean charge of two in the whole octahedral sheet.

In the Ni-Al montmorillonite, the mean Ni-O distance is found to be 2.05 Å. If Ni²⁺ divalent cations (ionic radius: 0.69) occupy the M₁ sites, their local environment becomes distorted and differs from the ideal montmorillonite structure. The lattice distortion becomes more and more important as with increasing the Ni content. This interpretation is consistent with the increase of σ² Debye-Waller factor in the Ni02 sample, with respect to the Ni01 sample (Table 1). Furthermore, the significant increase of the mean Ni-Ni and Ni-Al interatomic distances, from 3.00 and 3.02 Å respectively in Ni01 to 3.03 and 3.05 Å in Ni02 (Table 1), suggests that more adjoining octahedral sites are filled than in the ideal montmorillonite structure (R_M-M = 3.00 Å for both M₁ and M₂ sites). This result is consistent with the observation of Ni-Ni-Ni triads in ¹⁹F NMR together with the lack of Ni-Ni- environments. In fact, the octahedral sheet of natural montmorillonites contains transition metal cations either di- or trivalent, like Ni and Fe, and some vacancies, which ensures a suitable distribution of charges through the sheet. The presence of only Ni divalent transition metal cations in Ni-Al montmorillonite may induce a filling of empty octahedral sites as with increasing the Ni content. When divalent cations become in excess, they fill the empty octahedral sites in some parts of the octahedral sheet, leading to the splitting of the 060 reflection on XRD patterns, which is the fingerprint of di- and tri-octahedral character of the clay structure. Possible higher local charge within the octahedral layer, induced by the metal elements clustering, might be compensated by a lower local charge in the tetrahedral layer, which would be at the origin of the Al for Si substitutions in the tetrahedral layer.

The samples presented in this work were hydrothermally synthesized in a fluoride medium using a 120 cm³ PTFE-lined stainless steel autoclave. The hydrogel composition was based on the following ideal montmorillonite-like phyllosilicate half-unit-cell formula: Na_2x(Al_2(1−x)Ni_2x□)Si₄O₁₀(OH)₂, where x equals 0.10 or 0.20. Reagents were mixed in the following order: deionized water, hydrofluoric acid (HF, 40%; BDH, diluted to 5%), Na (sodium acetate, CH₃COONa, 99%; Fluka), Ni (nickel acetate, Ni(CH₃COO)₂ 4H₂O, 98%; Fluka), Al (pseudo-boehmite, Al₂O₃, 75%–78%, Pural SB1; Condea) and Si (SiO₂, 99.5%, Aerosil 130; Degussa) sources to form hydrogels with the following composition 1·SiO₂ : (1 − 2x)/4·Al₂O₃ : x/2·NiO : x/4·Na₂O : 0.05·HF : 96·H₂O. The amounts of deionized water, hydrofluoric acid and silicon oxide were the same for each synthesis. Hydrogels were matured for 2 h at room temperature and then autoclaved under autogenous pressure for 72 h at 493 K. After reaction, autoclaves were quenched to room temperature under a stream of water. The initial and final pH values were 5.5 (prior to autoclaving) and 3.5–4.0 (end of crystallization) respectively. Final products were thoroughly washed by filtration using deionized water and dried at 333 K overnight. Samples were then ground to a fine powder and placed under a relative humidity of P/P₀ = 0.80, by using a saturated solution of ammonium chloride. Samples were labeled as Ni01 and Ni02 for x equals 0.10 and 0.20 respectively. Some Ni02 sample was ion-exchanged with hexadecyltrimethylammonium bromide (HDTMABr) (98%, Avocado) to probe the swelling properties of the sample.