Impact of Alkali and Alkali‐Earth Cations on Ni‐Catalyzed Dimerization of Butene

The presence of alkali (Na+ or Li+) or alkali‐earth (Ca2+ or Mg2+) cations adjusting the acid‐base properties on amorphous silica‐alumina influences markedly the catalytic properties of supported Ni for 1‐butene dimerization. The low concentration of Brønsted acid sites on these catalysts reduces the double bond isomerization of butene and inhibits the formation of dimethylhexene as primary product. While the alkali and alkali‐earth cations act as weak Lewis acid sites, only Ni2+ sites are catalytically active for dimerization of 1‐butene. n‐Octene and methylheptene are formed selectively as primary products; dimethylhexene is a secondary product. The open environment of the Ni2+ sites does not induce different reaction pathways compared to Ni2+ in the pores of zeolites.


Introduction
The conversion of butenes from naphtha steam cracking is of high interest [1] to synthesize branched dimers, for gasoline additives, [1c] or into linear dimers, [2] used as feedstock, e. g. for the production of PVC plasticizers. [1b,3] For the latter purpose, Ni dispersed on amorphous silica-alumina (ASA) is used as a catalyst, obtaining selectivities of approx. 60 % to linear and single branched dimers (octene and methylheptene) and 40 % to branched dimers (dimethylhexenes). [4] The ASA support provides both Lewis and Brønsted acid sites, enabling the dispersion and strong interaction of the Ni 2 + cations. They are the active sites for the dimerization of 1butene, following a Cossee-Arlman type mechanism. [5] However, isolated Brønsted acid sites (BAS) catalyze the synthesis of branched dimers, via the formation of secondary carbenium ions, as well as by isomerization of 1-butene to 2-butene. [6] Thus, high selectivity to linear octenes requires the elimination of BAS via ion-exchange for alkali and alkali-earth cations. [7] This step has shown to increase the selectivity in zeolite supported Ni 2 + catalysts toward linear dimers with C 3 or higher olefins. [5b,8] However, the steric constraints in these catalysts have been hypothesized to selectively stabilize the transition state of linear dimers. [9] Thus, we decided to explore the role of the cations (in their task to reduce Brønsted acidity) in the absence of pore confinement. This study is aimed to differentiate between intrinsic catalytic properties of Ni 2 + cations in the absence of Brønsted acid sites without constraints of a microporous environment. The impact of alkali and alkaliearth co-cations on a catalyst based on Ni 2 + cations is explored, therefore, combining detailed acid-base characterization and kinetic measurements.

Synthesis and characterization of alkali and alkali earth modified Ni/ASA catalysts
The concentration of Ni 2 + and the respective co-cations are compiled in Table 1. Amorphous silica-alumina was impregnated with equal concentrations of Ni 2 + and two different concentrations of Na + , Ca 2 + , Li + , Mg 2 + . After calcination, NiO particles were observed by XRD ( Figure S1, S2). As derived from the width of the XRD peaks, the average diameter of NiO particles increased with increasing co-cation concentration. We hypothesize that the well-dispersed co-cations interact stronger with ASA and so in turn weaken the interaction between Ni 2 + and ASA. This is concluded to lead to a higher mobility of Ni 2 + and, hence, to the formation of larger NiO particles. Also on pure SiO 2 larger NiO particles were observed, highlighting the weak interactions between the Ni 2 + and silica. Transmission electron microscopy (TEM) showed the formation of NiO particles, ranging from 10-100 nm ( Figure S3). The impregnation solely with Ni 2 + cations decreased the BET surface area of ASA from 645 to 403 m 2 /g ( Figure S4, S5), the micropore volume from 0.034 to 0.013 cm 3 /g and the mesopore surface area from 689 to 422 m 2 /g. A minor relative decrease was observed with respect to the Ni/ASA, with the cocation concentration ( Figure S4A and D), while the pore diameter remained unchanged ( Figure S4B). The formation of larger NiO particles in 2LiÀ Ni/ASA, together with lower BET surface area and smaller mesopores, suggests a high dispersion of the Li cation on the ASA support, blocking its pores.
The IR spectra of adsorbed pyridine were used to evaluate the nature and concentration of acid sites ( Figure 1, the 1.25 mmol/g loaded samples are shown in Figure S6). Pyridine adsorption on the parent ASA support ( Figure 1B, middle) led to the formation of pyridinium ions with Brønsted acid sites (BÀ Py) at 1540 and 1637 cm À 1 . [10] In addition, pyridine hydro-gen-bound to weak hydroxyls was observed at 1575 cm À 1 (HÀ Py). [10][11] The band at 1621 cm À 1 is assigned to the 8a vibrational mode of pyridine coordinatively bound to Lewis acid sites (LAS), while the band at 1454 cm À 1 is attributed to pyridine adsorbed to Al 3 + LAS (Al-LAS). [10,12] The presence of Ni (Figure 1 B, top) hardly changed the total concentration of pyridine bound to LAS and BAS (Figure S7), but significantly influenced the nature of the Lewis acid sites. In addition to the band at 1621 cm À 1 , a new band was observed at 1610 cm À 1 , indicative of sites with lower Lewis acid strength. [11a] This band is assigned to pyridine coordinated to a Ni Lewis acid site (NiÀ LAS). A small shift was also detected from 1454 to 1450 cm À 1 . It is hypothesized that this is caused by the increase in concentration of the weaker NiÀ LAS. The adsorption of pyridine on NiO/SiO 2 (Figure 1 B, bottom) did not show coordinatively adsorbed pyridine, indicating that pyridine adsorbs less strongly on NiÀ LAS. The SiO 2 support does not possess LAS or BAS detectable by adsorbed pyridine.
The co-impregnation of additional cations ( Figure 1A) led to minor changes in the IR spectra. The 1621 cm À 1 band, assigned to the LAS of the ASA support (AlÀ Py), decreased in intensity. The band at 1610 cm À 1 (NiÀ Py) assigned to the pyridine coordinated to Ni 2 + cations did not change in intensity. However, an additional third band in that region at lower wavenumbers was observed, assigned to the interaction of pyridine with the co-cations. The highest redshift was observed with the Na + (19 cm À 1 ), followed by Ca 2 + , Li + (being both of 10 cm À 1 ) and Mg 2 + (almost no red-shift). The observed redshift is strongly correlated to the decrease in Sanderson electronegativity of the different co-cations ( Figure S8), as following: Mg 2 + (1.32), Ca 2 + (0.95), Li + (0.89) and Na + (0.84). [13] In presence of co-cations, the total acid site concentration was significantly lower than on Ni/ASA ( Figure 2). The concentration of BAS decreased from 69 μmol/g with Ni/ASA to 0 μmol/g with higher co-cation loading. Increasing concentrations of co-cations decreased the concentration of LAS. The presence of alkali cations (134 and 170 μmol/g with Na + and Li + ) had a stronger impact than the presence of alkali earth cations (239 and 346 μmol/g with Ca 2 + and Mg 2 + ). Using the bands around 1600 cm À 1 to differentiate between AlÀ LAS and NiÀ LAS ( Figure 2, Table S1), it is shown that in presence of Ni 2 + almost 2/3 of the Lewis acidity was associated with Ni sites. The increase in the concentration of co-cations led to a decrease in the concentration of both AlÀ LAS and NiÀ LAS. This suggests larger NiO particles in the presence of co-cations. A negligible concentration of BAS was observed for all 2.5 mmol/g co-cation loaded samples, suggesting the full exchange of the BAS protons by the alkali and alkali-earth cations.

Dimerization of 1-butene on supported Ni catalysts
The parent ASA was active in the dimerization of 1-butene, forming only methylheptene and dimethylhexene. The catalyst deactivated rapidly ( Figure S9). In presence of Ni, the rates increased from 0.12 to 3.17 mol But g À 1 h À 1 , (Figure S9 and Figure 3) showing that the Ni 2 + cations are highly active for  dimerization. Surprisingly, NiÀ SiO 2 was not active ( Figure S10), suggesting that the very large NiO particles are not active in 1butene dimerization. Figure 3 shows the conversion of 1-butene ( Figure 3A), as well as the selectivity within the dimer fraction ( Figure 3B) for Ni/ASA. The catalyst deactivated moderately, decreasing the conversion from 36 % to 27 % within 15 hours. The selectivity to trimers and tetramers was always below 20 % ( Figure S11). The selectivity to dimethylhexene (DMH) and methylheptene (MH) was similar within the first hour (41 % and 36 %, respectively). Over time on stream (TOS), the selectivity to DMH decreased to 23 % and the selectivity to MH increased to 52 %. The selectivity to n-octene increased slightly from 18 to 21 %. The initial higher selectivity to DMH and lower selectivity to MH and n-octene suggests the dimer formation to be catalyzed by both Brønsted and Ni acid sites. The proton catalyzed pathway occurs over a carbenium ion (Scheme S1), and leads to the formation of branched dimers, [6b] as also observed with the parent ASA ( Figure S9). [14] The rapid change in selectivity suggests that the BAS sites deactivate fast. To test this hypothesis, the acid site concentration was measured after the reaction (Figure 4) and the BAS concentration was found to have decreased from 69 to 26 μmol/g. The NiÀ LAS concentration also decreased from 547 to 422 μmol/g, indicating that both BAS and NiÀ LAS were negatively affected by carbon deposits that could be fully removed by calcination. X-ray diffraction analysis of the sample before and after reaction did not indicate significant variation in this process, except for a slight decrease in NiO particle size ( Figure S12).  Impact of co-cations on Ni/ASA on 1-butene dimerization Figure 5 shows that the dimerization rates decreased in the presence of co-cations, i. e., from 3.17 mol But g À 1 h À 1 with Ni/ASA to~0.05-0.02 mol But g À 1 h À 1 ( Figure S13). While a high deactivation was observed with 2LiÀ Ni/ASA (70 % within 7 hours), the other 2MÀ Ni/ASA showed only minor deactivation (less than 10 % in 7 hours). Hence, a high dispersion of the Ni 2 + cations on ASA and the absence of Brønsted acid sites is required for high catalyst stability. The reactivity decreased further with higher co-cation loading, independent of the nature of the co-cations. This dependence suggests on first sight that only BAS are active and are gradually eliminated with higher concentration of alkali and alkali-earth cations. On closer inspection one notes, however, that the size of the NiO particles increases with cocation concentration ( Figure S2), leading to a reduction of the concentration of NiÀ LAS. This NiÀ LAS is hypothesized to be the catalytically active site, with Ni 2 + cations at exchange positions of the ASA support. Indeed, the rate normalized to the concentration of NiÀ LAS (TOF) increased with increasing cocation concentration ( Figure 6) showing that the adjustment of the acid-base properties of the support is beneficial. We speculate that a higher base strength of the support and, hence, a higher electron density at NiÀ LAS (weaker Lewis acid strength) is beneficial for the catalytic activity in 1-butene dimerization.
The very low TOF of NaÀ Ni/ASA is attributed to a very low concentration of NiÀ LAS, the reasons for which is currently subject of a separate investigation. Mg 2 + /ASA (without Ni, Figure S14) did not show catalytic activity for 1-butene dimerization, leading to the conclusion that alkali and alkali earth cations do not catalyze the dimerization of 1-butene under the chosen reaction conditions. The selectivity to branched products ( Figure 7) increased with conversion in a subtly different manner in presence and absence of co-cations. In their absence with Ni/ASA ( Figure 7A), initial selectivities show that all three n-octene, methylheptene and dimethylhexene are primary products that change nonlinearly with conversion. In the presence of co-cations (2 .5 mmol/g, Figure 7B) only n-octene and methylheptene are primary products and dimethylhexene is a secondary product.

ChemCatChem
Full Papers doi.org /10.1002/cctc.202000349 While this indicates that the elimination of BAS also eliminates the direct formation of dimethylhexene, its rapid evolution with conversion on co-cation modified Ni/ASA suggests the presence of an alkali and alkali-earth cation catalyzed isomerization of 1butene. Considering, for example, the rates of isomerization for 2MgÀ Ni/ASA and Ni/ASA ( Figure S15), one notes a higher isomerization/dimerization rate ratio in the presence of cocations, i. e., 0.3 for Ni/ASA and 6.0 for 2MgÀ Ni/ASA, respectively. Indeed, Mg on ASA (Mg/ASA) was active in butene isomerization, demonstrating the catalytic activity of the cocations. In the case of LiÀ Ni/ASA, despite having NiO particles twice the size of the other catalysts with similar co-cation loading, its catalytic performance does not differ from the other catalysts, suggesting the decrease in BAS concentration is the main driving force changing catalytic activity.
In the absence of steric constraints, the butene dimerization rates and selectivity to linear dimers differ from the microporous environments. The butene dimerization rates for MÀ Ni/ ASA were almost an order of magnitude lower (0.05-0.015 mol But g À 1 h À 1 ) than in our previous work with MÀ Ni/LTA (0.13-0.11 mol But g À 1 h À 1 ). In addition, at 35 % conversion selectivities below 10 % to dimethylhexene were observed with MÀ Ni/LTA within the C 8 fraction, while with MÀ Ni/ASA selectivities to dimethylhexene were around 35-40 %. Therefore, we conclude that steric constraints surrounding the active Ni site are able to better stabilize the transition state of butene dimers, especially those leading to the formation of n-octene and methylheptene.
In the absence of BAS we hypothesize that 1-butene is dimerized in a Cossee-Arlman type mechanism with the adsorption of 1-butene on NiÀ H and a subsequent insertion and coordination of another butene molecule into the NiÀ C bond. [2a,15] Thereby, in the formation of n-octene and methylheptene the initial adsorption would take place as 1,2insertion into the NiÀ H (Scheme 1, left pathway). The subsequent coordination of another 1-butene to the NiÀ C bond determines whether n-octene (by 2,1-insertion) or methylheptene (by 1,2-insertion) is formed. The initial 2,1-insertion of 1-butene leads to a branched dimer. Because dimethylhexene is not a primary product in presence of alkali and alkali earth cations, this route can be ruled out.
As isomerization of 1-butene into cis-and trans-2-butene is occurring, the product 2-butene would adsorb via a 2,3insertion into the NiÀ H, opening this pathway for dimerization (Scheme 1, right pathway, red). This would also lead to the formation of dimethylhexene by subsequent 2,3-insertion. The 2,1-insertion of 1-butene into the NiÀ C (formed by 2-butene) results in this case in a secondary pathway for the formation of methylheptene.

Conclusions
The catalytically active sites for 1-butene dimerization have been identified as accessible Ni 2 + on ASA (NiÀ LAS). Ni 2 + cations in NiO supported on SiO 2 have not been found to have sufficiently high catalytic activity to lead to measurable conversions under the present reaction conditions. The presence of alkali (Na + or Li + ) and alkali-earth (Ca 2 + or Mg 2 + ) co- cations on Ni/ASA blocks BAS that catalyze 1-butene dimerization. This is attributed to an increase in the local base strength. As a negative side effect, the particle size of the supported NiO particles increased, decreasing the concentration of Ni 2 + cations at exchange positions of the ASA support active in butene dimerization. The alkali-and alkali earth cations increase the initial selectivity to n-octene and methylheptene (x < 10 %). The co-cations increase, however, also the rate of 1-butene isomerization to 2-butene. In turn, this leads to a rapid increase in the formation rate and selectivity to dimethylhexene. Thus, we conclude that alkali and alkali earth cations are suitable to block BAS catalyzed dimerization of 1-butene, but their ability to induce isomerization of 1-butene does not allow for selective dimerization at higher conversions. Investigations to eliminate this pathway by organic co-reactants (poisons) are under way.

Characterization
Atomic absorption spectroscopic measurements (AAS) were performed in a Solar M5 Dual Flame graphite furnace AAS from ThermoFisher. After drying at 250°C for 24 h, the samples were dissolved in a mixture of HF and nitric acid and injected in the graphite furnace. A previous calibration was applied to determine the concentration of each of the metals.
X-ray diffraction measurements were performed in a PANalytical Empyreal System diffractometer, equipped with a Cu-Kα radiation source (Kα 1 line of 1.54 Å; 45 kV and 40 mA)). The diffractograms were measured by the usage of a sample spinner stage in a 2θ range between 5°and 70°(step size: 0.0131303/2θ) at ambient conditions.
The BET specific surface area and pore volume of the zeolite were determined by N 2 sorption. The isotherms were measured at liquid N 2 temperature (77 K) using a PMI automatic sorptometer. The catalyst was activated in vacuum at 473 K for 2 h before measurement. Apparent surface area was calculated by applying the Brunauer-Emmett-Teller (BET) theory with a linear regression between p/p 0 = 0.01-0.15. The micro-and mesopores were determined from the t-plot linear regression for t = 5-6 Å.
The adsorption of pyridine was followed by IR in a Nicolet 5700 FT-IR spectrometer, equipped with a liquid N 2 cooled detector. For the measurement, a self-supporting wafer was loaded and activated at 450°C (rate: 10°C/min) for 1 h in vacuum. After cooling down to 150°C, pyridine was equilibrated to 1 mbar for 30 min. Subsequently, the system was evacuated for another 30 min, before measurements. Scans were taken after activation and after outgassing at a resolution of 0.4 cm À 1 with an average of 250 scans per spectrum. For the calculation of the acid site concentrations, molar extinction coefficients of 0.96 and 0.73 cm/μmol and were used for the characteristic bands of 1450 cm À 1 (Lewis) and 1540 cm À 1 (Brønsted), [16] respectively. For the deconvolution of the Lewis acid sites, the bands at 1600 cm À 1 region were fitted, assuming similar extinction coefficients for all bands of LAS.