Introduction

Nickel ions supported on different silica, silica–alumina, natural clay, and zeolite-type porous materials are well-known catalysts for the selective dimerization [19] and oligomerization [1, 5, 1019] of olefins in both gas and liquid phases. The application of silica-supported nickel catalysts for the ethylene dimerization dates back to 1980s [2022]. Later, the large and well-ordered cavities of the Ni-exchanged MCM-41 proved highly favorable for the oligomerization of olefins [23]. For instance, high productivities of oligomers were obtained over Ni-incorporated MCM-41 catalysts [17] being higher than those reported previously with silica–alumina supports [19]. The strong interactions of Ni2+ cations residing in the mesoporous cavities with the support framework made the reduction of the nickel ions difficult [17]. Further electron spin resonance (ESR) and Fourier transform infrared (FTIR) spectroscopic data [11, 24] served as evidence to support the role of Ni2+ cations as active sites in ethylene dimerization.

The nickel ions incorporated into the ordered mesoporous materials (OMMs) have also been suggested as efficient catalysts for the gas-phase transformation of ethylene to propylene [2534]. For instance, the performance and durability of Ni/MCM-41 catalysts prepared from a nickel citrate precursor was ranked as auspicious with a promising potential for closing the gap between the propylene demand and supply [35]. Ikeda et al. [36] proposed that layered nickel-silicate species were the main players in the conversion of ethylene on NiMCM-41 catalysts. In a similar study [4], the threefold coordinated Ni2+ ions situated in the 5-membered rings of the phyllosilicate pore walls of NiMCM-41 were taken as the active sites of the reaction.

Computational studies of Ni2+ binding in FER [37], MFI [38, 39], AFI [40], and silica [4, 41] have been reported earlier. Analogous reports have addressed neutral atoms [42, 43]. The present study aims at a systematic theoretical modeling of the locations of Ni2+ species in a model MCM-41 material at the fundamental level. To our knowledge, the most relevant publication to this target is that by Neiman [41] who considered an O3(SiO)2Ni structure unit cell with a formal charge of +2 for the nickel ion. The author did not investigate all of the defect sites available for the Ni2+ siting. This work is then the first fundamental report on the molecular-level heterogeneities of the NiMCM-41 catalysts.

Computational method

The cluster model approach was employed to simulate the active sites of a nanoporous NiMCM-41 catalyst through exploring the available defect sites of MCM-41 within an extended unit cell. As a common approach [44, 45], the model nanoclusters were terminated by H atoms frozen in agreement with the geometries obtained from the crystallographic data [4651]. The divalent nickel cation and the immediate neighbors including the O atoms from two defect-site hydroxyls were allowed to relax during the optimizations. As adopted earlier [52], the remaining part of the cluster was held fixed to mimic the mechanical restrictions of the matrix.

As an approximation, the structure of cristobalite is normally taken as a good representative for the amorphous silica materials in terms of type and density of the surface hydroxyl groups [51, 5356]. Many references [42, 43, 5760] have then applied this model to ordered silica mesoporous materials such as MCM-41. The optimizations and single-point computations were implemented using hybrid functional M06 [61] and the Def2-TZVP basis set [6264]. Moreover, the natural bond orbital (NBO) population [65] as well as the quantum theory of atoms in molecules (QTAIM) [6671] analyses were carried out at the same level of theory.

The ion exchange energies (ΔE ex) were calculated for the following reaction:

$$\left[ {\text{SiO}} \right]{-}\left( {\text{OH}} \right)_{ 2} + {\text{ Ni}}\left( {{\text{NO}}_{ 3} } \right)_{ 2} \to \, \left[ {\text{SiO}} \right]{-}{\text{O}}_{ 2} {\text{Ni}} + 2 {\text{HNO}}_{ 3}$$
(1)

Moreover, the binding energies (ΔE b) for the nickel ions at the defect sites of MCM-41 were assessed for the following reaction:

$$\left[ {\text{SiO}} \right]{-}{\text{O}}_{ 2} {\text{Ni }} \to \, \left[ {\text{SiO}} \right]{-}{\text{O}}_{ 2}^{{ 2{-}}} + {\text{ Ni}}^{ 2+ }$$
(2)

NWChem 6.5 [72] and Multiwfn 3.3.8 [73] were used for the computations. Finally, the graphical outputs were generated by Mercury 3.3 [7477].

Results and discussion

The viable defect sites of MCM-41 for the binding of the divalent nickel ions were determined from exploring the available pairs of vicinal and close non-vicinal silanol groups. Different NiMCM-41 clusters were considered in terms of the number of T-atoms present in the final rings and the interatomic distances between the next-nearest-neighbor Si atoms with respect to the Ni cation. The optimized geometries are shown in Fig. 1. As can be seen, different oxide rings can be formed varying from a simple 2T (a digrafted nickel species stabilized with an interaction with two geminal-type Si–O groups) to tri, tetra, penta, and hexa membered ones. To our knowledge, this heterogeneity and its consequences are discussed theoretically for the first time. As such, the simplest cluster was a 2MR ring followed by a 3MR structure with a nickel cation chemisorbed on two vicinal OH groups located at the two ends of a siloxane bridge. The other clusters represent relatively more complex types of Ni binding with proximate non-vicinal OH groups at two defect sites of the surface linked together via two or more consecutive siloxane bridges (Fig. 1). For the comparisons made here, it will be assumed [78] that the dispersion of the Ni2+ species at the defect sites of MCM-41 is controlled by their thermodynamic stability and not by any kinetic phenomenon in the preparation method.

Fig. 1
figure 1figure 1

The optimized geometries of the NiMCM-41 clusters where the darker atoms refer to the framework oxygen, the plain bigger balls are silicon atoms and the small white balls are terminal hydrogen atoms. The largest atoms represent nickel

Table 1 reports the NBO charges of the selected atoms of the NiMCM-41 clusters. As can be seen, the partial charge of the Ni cation ranged from 0.889 e for 2MR to 1.077 e for 4MR, being almost half the formal charge of +2 for the Ni ions. Except 2MR and 3MR, the bridging O atoms of the O2Ni species did not bear identical charges, falling into the range of −1.157 e to −1.020 e being almost half the formal charge of –2 for the O atoms. Although not linearly correlated, there was a general proportional trend between q(Ni) and –q(O). Such an interconnection was not confirmed on the Mulliken atomic charges (see Table 1). Moreover, the average Mulliken charges of Ni and O atoms were generally 0.60 and 0.46 times those of the NBO calculations and then even more distant from the nominal charges.

Table 1 Calculated charges of selected atoms of NiMCM-41 for different cluster models at M06/Def2-TZVP level of theory

Table 2 contains the enthalpy, entropy, and Gibbs free energy of the exchange reaction for the NiMCM-41clusters and the binding energies of Ni at the investigated sites following Eqs. (1) and (2), respectively. As evident, the grafting reaction considered here is non-spontaneous at 298 K and atmospheric pressure on all of the clusters, connoting the necessity for devising more severe (e.g., hydrothermal) preparation conditions to overcome the exchange non-spontaneity. In any event, the thermodynamic favorability for grafting of the Ni2+ cations onto the available defect sites of MCM-41 followed the sequence of 2MR < 3MR < 4MR < 6MR-3 < 6MR-1 < 6MR-2 < 5MR (53.99–72.72 kcal/mol). This is in accordance with the previous ideas on the favorability of 5MR structures mentioned previously. Moreover, the enthalpy of the exchange reaction followed the order of 5MR < 4MR < 6MR-1 < 6MR-2 < 6MR-3 < 3MR < 2MR, all being endothermic (63.09–84.10 kcal/mol). However, the binding energy for this reaction varied in the order of 4MR < 3MR < 5MR < 6MR-3 < 6MR-2 < 6MR-1 < 2MR (780.6–831.5 kcal/mol). As a result, the 2MR site is more demanding to form despite its largest binding energy.

Table 2 The Gibbs free energy (kcal/mol), enthalpy (kcal/mol), and entropy (cal/mol/K) of the exchange reaction and the binding energy (kcal/mol) of the digrafted Ni ions in NiMCM-41 at M06/Def2-TZVP level of theory (please see the corresponding reactions in the text)

In Table 3, the nickel–oxygen bond lengths and the interbond angles of Ni have been listed for comparison. The bridging (Ni–O) bond lengths fell into the range of 1.69–1.79 Å with the highest asymmetry [a difference of 0.03 Å between r(Ni–O1) and r(Ni–O2)] found in the case of 5MR. The O–Ni–O angles changed over a markedly wider range here with the minimum angle (85.6°) on 2MR and the largest one (168.9°) on 4MR. The data reported here indicated that neither the Ni–O distances nor the O1–Ni–O2 angles correlated with the size of the siloxane rings as analogously expressed in our previous study of Cr–silicalite-2 [79].

Table 3 Nickel–oxygen bond length (Å) and interbond angle (in degree) for different optimized cluster models at M06/Def2-TZVP level of theory

The characterizing features [71, 80, 81] of QTAIM were employed to determine whether the nickel–oxide solid interactions were of shared or closed-shell nature. The calculated topological properties are shown in Table 4. The tabulated data represent moderate electron densities (ρ BCP). Moreover, positive ∇2 ρ BCP values were obtained as anticipated from a common character of metal–oxygen interactions [80]. The interactions of nickel with the siloxane rings of MCM-41 were all found to be rather polar than purely covalent. More interestingly, however, the 4MR and 5MR structures revealed third Ni–O interactions of electrostatic nature that made the nickel ion interlocked three-coordinated in the plane of the ring. Potential correlations between the properties of the NiMCM-41 active sites were probed. Figure 2 shows an obvious correlation between r(Ni–O) and ρ BCP which was also observed in our previous works on MEL structure [79, 82]. No further correlation with an acceptably high coefficient of determination was found between any other two parameters.

Table 4 QTAIM data for different optimized clusters at M06/Def2-TZVP level of theory
Fig. 2
figure 2

The correlation found between the topological properties and the Ni–O distances in the NiMCM-41 catalysts

The energy levels of electrons and their gaps provide useful data for implications on the reactivity of the active sites [79, 82, 83] according to the frontier molecular orbital (FMO) theory [84]. Table 5 reports the HOMO and LUMO levels and HOMO–LUMO energy gaps for the optimized clusters shown in Fig. 1. The HOMO–LUMO energy gaps increased in the order of 4MR < 5MR < 3MR < 6MR-3 < 6MR-1 < 6MR-2 < 2MR ranging from 3.01 to 4.15 eV. Considering the chemical hardness as η = (E LUMOE HOMO)/2 [8589], one can state that the chemical hardness of the digrafted nickel ion was maximized at 2MR and minimized at 4MR and 5MR sites.

Table 5 Calculated HOMO and LUMO and HOMO–LUMO energy gaps (ΔE HOMO−LUMO) for the investigated clusters at M06/Def2-TZVP level of theory

Conclusion

This paper investigated the diversity of the cluster models of NiMCM-41 in a systematic computational framework. Total of seven active sites (2T–6T rings) were found at the defect sites of an MCM-41 silica model. The NBO partial charge of the nickel cation was less positive on 2MR and largest on 4MR. The thermodynamic favorability of the NiMCM-41 clusters followed the order of 2MR < 3MR < 4MR < 6MR-3 < 6MR-1 < 6MR-2 < 5MR. The optimized structures indicated the Ni–O distances in the range of 1.69–1.79 Å with the highest asymmetry observed in 5MR. The highest reactivity was observed in the case of the digrafted nickel ions at 4MR and 5MR sites and the lowest one at 2MR. The 4MR and 5MR clusters showed also some intertwining features for nickel hosting. The QTAIM calculations revealed intermediate polar Ni–O bonds. Moreover, the electron densities at the BCP correlated with the Ni–O distances.