Ni(ii) and Co(ii) bis(acetylacetonato) complexes for alkene/vinylsilane silylation and silicone crosslinking

Commercially available Ni(ii) and Co(ii) complexes – M(acac)2 and M(tmhd)2 – exhibit catalytic activity for alkene/vinylsilane dehydrogenative silylation (DS) and hydrosilylation (HS) with tertiary silanes without using any external reducing agents.


Introduction
Alkene hydrosilylation is a key reaction to manufacture silicone lubricants, adhesives, rubbers, and other high-value products. [1][2][3] Platinum compounds such as Speier's 4 and Karstedt's 5 complexes traditionally serve as efficient catalysts because of their high activity, stability and selectivity for alkene hydrosilylation. However, high and volatile prices of noble metals induce an important interest in tackling hydrosilylation using earth-abundant metal complexes. Nickel, cobalt and iron catalysts have already been reported and reviewed as relatively active and selective for alkene hydrosilylation (HS) and dehydrogenative silylation (DS). [6][7][8][9][10][11][12][13] Two types of systems have been reported in the literature: low-oxidation state and highoxidation state complexes. Among low oxidation state catalysts, Chirik's bis(imino)pyridine iron(0) 14 and cobalt (I) 15 are seminal examples, which are very active under strict inert conditions. More stable, high oxidation state nickel, iron and cobalt precursors have been developed to catalyze alkene HS (or DS) after an activation via reduction with Grignard reagents, lithium species or borohydride. [16][17][18][19][20][21][22] New development in the field are systems which do not need external reducing agents and might harness the SiH inherent to the reaction medium for reduction of iron(II) [15][16][17]19,24 and cobalt(II) [23][24][25][26][27] precatalysts, though the exact activation mechanism remains elusive. 10,27 In the 90s, Marciniec reported the use of the commercially available Ni(II) acetylacetonate Ni(acac) 2 without external reducing agent as a hydrosilylation catalyst. [28][29][30] It was claimed that the SiH from the used silane allowed for reduction of the precursor. This catalyst was not selective for alkene hydrosilylation (HS) and gave several other products such as dehydrogenative silylation (DS), reduction (Red) and dimerization products (Dim) (see Figure 1). Note that only HS and DS establish a Si-C linkage and are thus interesting in coupling or crosslinking reactions. Herein we developed readily available and easy-to-handle precursors for vinyl silane/alkene silylation without the use of external strong reducing agent, yet promoting the curing of functional silicone oils. Our study is based on Marciniec analogues. 31 We were particularly interested in comparing bis(acetylacetonato)Ni(II) and Co(II) complexes ( Figure 2) and their respective activities and selectivities for alkene/vinylsilane hydrosilylation and dehydrogenative silylation using industrially relevant tertiary silanes. We also implemented these precatalysts in the crosslinking of vinyl-and Si-H-functionalized PDMS chains under non-inert atmosphere (presence of O 2 and ambient moisture, see Experimental section).

Results and discussion
We first studied the silylation of divinyltetramethylsiloxane (dvtms), vinylpentamethyldisiloxane (vpmds) and 1-octene (see Figure 2, SI for all experiments, Sections 1-4 for dvtms and vpdms & 5 for 1-octene) with MD'M (1,1,1,3,5,5,5-Heptamethyltrisiloxane). Before quantifying the conversion and the selectivity, the major products of the reaction catalyzed by Ni(tmhd) 2 were identified by NMR spectroscopy. A combination of 1 H, 13 C and 29 Si NMR, as well as GC/GC-MS analyses (for confirmation, in 1-octene case in particular) revealed the major and minor products of the reaction depicted in Figure 1 for vinyl-silane model compounds and 1-octene (see Experimental section and SI, sections 1 to 5). Figure 3: 2D 29 Si-1 H NMR spectra (700 MHz, methylcyclohexane (MCH) d14) using INEPT sequence to attribute all signals to the various reaction products. This example uses Ni(tmhd) 2 as precatalyst (see Table 1 & corresponding scheme for conditions), selective for dehydrogenative silylation (DS).
Harnessing NMR spectroscopy, the starting substrates were characterized using 1 H and 29 Si nuclei. Then, the reaction medium was analyzed with 1 H and 29 Si NMR thanks to a 2D 29 Si -1 H INEPT. Using the 2D spectra ( Figure 3 and SI, sections 2&3), we could identify the various Si moieties, and in particular the characteristic signals of the dehydrogenative silylation (DS) and reduction products (Red). Thus, we can conclude that Ni(tmhd) 2 is selective for dehydrogenative silylation and concomitant/subsequent reduction, as reported in the literature. 1-10 On the contrary, Co proved to yield both HS and DS products, even slightly favouring the former. 1 H-NMR was particularly helpful in the case of Co, to evidence all products of catalytic silylation (see SI, sections 1-4).
To further exemplify as well as compare Ni and Co, we summarized the catalytic performances of the precatalysts (1) to (4) in Table , using model reactions depicted in corresponding scheme. Reactions took place at 90°C in neat substrates using 0.5 mol% of catalysts and with a ratio of SiH:SiVi = 1:1 (mol ratio of MD'M:dvtms = 2:1 and for ratio of MD'M:vpmds = 1:1). All catalysts mainly yield alkene/vinyl silane dehydrogenative silylation (DS) giving also reduction products (Red) (entry 1,3,5 and 7 in Table 1 for dvtms), for both dvtms and vpdms substrate (comparison with entries 2, 4, 6 and 8 in Table 1 for vpdms). Cobalt catalysts yield some hydrosilylation product, which sets them apart from DS-selective Nickel catalysts (entries 5-8 for Co compared to 1-4 for Ni), in particular for dvtms as substrate. However, only the Co(tmhd) 2 precatalyst 4 is active for the silylation of vpdms. The acac-counterpart is likely quickly deactivated through formation of Co black (black residues are quickly observed in that case during catalysis).
Please do not adjust margins Please do not adjust margins The catalytic activity for the hydrosilylation of 1-octene with MD'M was also investigated (SI, section 5) and all Ni and Co catalysts are mainly selective for isomerization. Notably, observed selectivities are different from Shimada and coworkers' publication 22 where Ni(acac) 2 is active for hydrosilylation with the use of NaHBEt 3 as external reductant. Active-catalyst generation mechanism is thus probably different when the system is activated with a stronger external reducing agent than the inherent tertiary silane featuring H-SiMe(O-R) 2 moieties. [1][2][3][4][5][6][7][8][9][10] In order to study the catalyst activation mechanism, dvtms silylation was monitored in situ using 1 H NMR at 90°C with 12.5 mol% of Ni(tmhd) 2 as catalyst and with a ratio of SiH:SiVi = 1:1 (molar ratio of MD'M:dvtms = 2:1). No catalytic intermediate has been observed when the catalyst is mixed with either MD'M or dvtms alone. Monitoring the silylation through an NMR-tube experiment has allowed us to follow the kinetics of the reaction, without significant perturbation from paramagnetism (see SI, sections 4 & 8), inferring specific environments for Ni(II) (since few environments are devoid of paramagnetism) [32][33][34][35][36] and no reduction product (either paramagnetic for Ni(I) or even ferromagnetic for some nanoparticules of Ni(0)). [37][38][39] We could evidence that the reaction is occurring under these conditions and some free ligand is released (small characteristic signals for keto/enol forms of free tmhd-derived 2,2,6,6tetramethylheptane-3,5-dione can be observed on 1 H-NMR spectra, see SI, sections 2, 4, 7 & 8). This was confirmed by precontacting the Ni complex with either Si-H or vinyl moieties under similar conditions in order to identify ligand-set alteration (see SI, section 2 & 4). Scheme 1. Global putative mechanism explaining the formations of various products. The oxidation state of the metal and the generation of active species is omitted voluntarily since subjected to speculation (adapted from Marciniec 1-3 ).
In his study, Marciniec assumed that Ni(acac) 2 was reduced by the SiH from the silane before the Chalk-Harrod mechanism. 1-3,28-30 From our study, it seems that the catalytic mechanism might be different, either needing both the silane and the alkene to reduce minute amounts of the precatalyst before either conventional or modified Chalk-Harrod mechanisms occurs (see Scheme 1). Reaction could also proceed through a concerted mechanism with neither fully-reduced species nor subsequent oxidative addition as suggested in recent reports, 7-10 or alternatively bi-nuclear Ni(I) species might be at play. 39 In light of the absence of paramagnetism suggested by NMR spectroscopy, monitoring the color change of the reaction medium could suggest the formation of non-octahedral Ni(II) species via -bond metathesis, with release of acetylacetonatetype ligand (as clearly evidenced by 1 H NMR for the tmhd ligand, see SI, sections 4, 7 & 8). The green Ni(acac) 2 (due to octahedral geometry [paramagnetic] in the more favorable trimeric structure) and pink Ni(tmhd) 2 (due to a monomeric squareplanar geometry [diamagnetic] in that case) 32-36 both rapidly turn bright orange, and NMR spectra do not display paramagnetic features (broad peaks or wide range of isotropic chemical shifts) even at high catalyst loading. This could suggest the formation of either square-planar, square-pyramidal trigonal or bipyramidal geometries with coordination of Si-H and vinyl moieties and subsequent release of one tmhd ligand as already suggested in the literature for such acetylactetonatetype ligands. [1][2][3][4][5][6][7][8][9][10][28][29][30] The release of the acac ligand seems less clear in our study, which could be explained by a higher propensity towards hydrosilylation of the released acetylacetonate in the mixture making the detection complicated. This could also suggest a much easier nanoparticule formation (via complete reduction) 37,38 in the case of Ni(acac) 2 , that could be detrimental to the reaction. That could explain the highest catalytic efficiency of the tmhd counterpart, which role would then not be limited to an enhanced solubility. It is likely that the increased sterics from the tert-butyl group play a role in the bidentate ligand release, and potentially hindered subsequent hydrosilylation, making its detection possible (see SI, sections 4, 7, 8 & 9 for pictures).
Co catalysis seems more difficult to rationalize, all systems are paramagnetic no matter the geometry and the darkening of the solution (see SI, section 9 -miscellaneous) could suggest reduction to Co(0) or Co(I) species, as previously proposed. [10][11][12]40,41 In the second part of our study, we monitored the activity of these catalysts for the crosslinking of Si-H and Si-vinyl functionalized silicone oils. Harnessing alkene (hydro)silylation to cure silicone formulations is the cornerstone of silicone materials' manufacturing: multiple applications heavily rely on this reaction. [1][2][3][4][5][6][7][8][9][10] The crosslinking is conventionally catalyzed by platinum complex that remain trapped in the final product, representing a large fraction of the cost of the material. As evidenced in Marciniec work, 1-3,10,28-30 a catalyst which is selective for dehydrogenative silylation can become active for hydrosilylation as well when it reacts with functional polymeric chain, suggesting a selectivity switch that is substratedependent. The crosslinking activity of the four precatalysts Ni(acac) 2 , Ni(tmhd) 2 , Co(acac) 2 and Co(tmhd) 2 was studied, and some curing parameters were changed to monitor the influence on Crosslinking time is determined with a stop-stirring test (SST) consisting of mixing di-vinyl and poly-SiH silicone-oils with the catalyst heating in an oil bath and measuring time when the magnetic stirrer stops (see Experimental Section). Here, we performed this test under non-inert atmosphere (presence of O 2 and ambient moisture). 31 Crosslinking of dvtms with poly-SiH oil MD' 50 M was first evaluated. The catalytic activity of complexes (1) to (4) is summarized in Table 2 (Entries 1 to 4). Reference reactions were performed at 90°C using 0.5 mol% of catalysts and with a ratio of SiH:SiVi = 1:1. Nickel catalysts are sluggish for the crosslinking of dvtms. On the contrary, cobalt catalysts are very efficient, and more particularly Co(acac) 2 which allows a crosslinking in less than 15 minutes at 90°C. The crosslinking of di-vinyl PDMS oil with poly Si-H oil was also surveyed and results are compiled in Table2. Reference reactions were done at 110°C using 1 mol% of catalyst and with a ratio of SiH:SiVi = 3:1, to ensure efficient crosslinking even through undesired reactions. All four catalysts are active for the crosslinking of di-vinyl and poly-SiH oil under non-inert atmosphere (Entries 5 to 8, Table 2). Contrary to the results with dvtms (y=0), nickel catalysts are more active than cobalt catalysts for more viscous silicone oils. Dvtms possesses a specific chelating structure and is the bidentate ligand of choice for Karstedt's type catalysts. 5,42 Difference between Co-and Nibased catalyst activities may arise from a variation of dvtms coordination on the active center, precluding activity in the case of Ni. 3,42 It is possible that the chelating nature of dvtms is detrimental to the reactivity of Ni catalysts, maybe suggesting the formation of a stable octahedral Ni(II) complex, no longer able to interact efficiently with the Si-H moiety. To evidence that crosslinking is really occurring, and to evaluate further crosslinking kinetics and compare gel properties, rheology was performed for Si-H and Si-vinyl PDMS oil curing.
The catalytic crosslinking of di-vinyl and poly-SiH PDMS oils with complexes (1) to (4) is presented in Figure 4. Reactions took place at 110°C using 1 mol% of catalyst and with a selected ratio of SiH : SiVi = 3:1. First, we systematically observed an intersect of G' and G'' suggesting that crosslinking is effective, with geltimes apparently shorter than observed during SST measurements. Ni catalysts are more active under this set of conditions and yield materials with slightly higher G' moduli (close to 10 5 Pa for a reference formulation of Table 2, see also SI, sections 7 & 8). Co(acac) 2 is less active as expected from SST, probably due to higher air instability leading to catalyst degradation, and gives lower modulus gel (~10 3 Pa). Co(tmhd) 2 has a gel time in accordance with SST and modulus higher than 10 4 Pa and consistently close to 10 5 Pa. Premature oxidation might be detrimental to Co catalysis. Adventitious contaminants in silicone oils might also be responsible for the lower efficiency of a slightly more sensitive Co catalysis, which could also explain the delay in crosslinking observed by rheometry.
Interestingly DSC can be used to monitor these crosslinking reactions. Thermal kinetics of the catalyzed crosslinking of divinyl and poly-SiH PDMS oils was monitored using DSC analysis. Crosslinking of dvtms with poly-SiH oil MD' 50 M was first studied. Reference reactions were heated at 8°C/min rate, using 0.5 mol% of complexes (1) to (4) with a ratio of SiH:SiVi = 1:1 (see SI, section 6- Figure S21).  Crosslinking of di-vinyl PDMS oil with poly-SiH oil MD' 50 M was also surveyed. Reference reactions were performed at a heating rate of 8°C/min, using 1 mol% of complexes (1) to (4) with a ratio of SiH:SiVi = 2:1 and thermograms are presented in Figure S22 (see SI). Under these conditions, crosslinking catalyzed by Co(acac) 2 starts at a lower temperature but the enthalpy is less important than with the other catalysts, suggesting incomplete reaction and possible catalyst decomposition due to air exposure. Overall, and contrary to dvtms crosslinking, Ni precatalysts seem to outperform Co counterparts for longer oilbased formulations. This could also be a consequence of the increased sensitivities of Co complexes. DSC thus unveils that Co is more reactive at lower temperatures than Ni for dvtms formulations (see SI, section 6), probably due to the aforementioned chelating nature of dvtms, deactivating Ni (pre)catalysts. However, this statement does not hold true for longer oil-based industrial formulations with spaced-apart vinyl moieties. These DSC results are consistent with rheology experiments with the observed G' modulus is lower in the case of Co(acac) 2 than with the other catalysts. Initiation temperatures are then lower for Ni catalysts than for Co(tmhd) 2 for crosslinking (see SI, Figure S22) and reaction profiles are very similar for these three catalysts confirming the rheology findings. These observations are in agreement with what was suggested by Marciniec: 1,3-28-30 reaction kinetics and selectivities are strongly substratedependent and (hydro)silylation model reactions behave differently than crosslinking of silicone-based oils. 3,43 Model reactions are thus interesting to rationally design regio-and stereoselective catalyst for silylation reactions, but crosslinking of polyfunctional oils might yield different product selectivities due to unique sterics and diffusion constraints. Therefore, another method is necessary to assess the different products when crosslinking is considered. High-Resolution Magic Angle Spinning NMR (HR-MAS NMR) spectroscopy can advantageously characterize gels displaying satisfactory mobility, bridging the gap between liquid NMR and solid-state NMR. It consists in running liquid-like sequences in a rotor spun at high frequency at the magic angle to retain proper resolution (see SI, sections 7 & 8). Very interestingly, HR-MAS NMR of crosslinking formulations allowed us to get a better understanding of the selectivities taking place in these 3D-polymer networks. If Ni still promotes DS linkage as a major crosslinking knot, Co displays a higher propensity towards HS linkage in the crosslinking of functionalized-polymer oils (see Figure 5 and SI -[HR-MAS NMR analyses of crosslinking of di-vinyl PDMS and poly-SiH oils], section 7). More specifically, Figure 5 clearly depicts a very selective Co(tmhd) 2 precatalyst yielding species promoting hydrosilylation of vinylsilane moieties from the functional silicone oil. Co(acac) 2 seems a bit less selective, since some DS product can be identified. Moreover, its lesser activity, probably due to premature deactivation precludes an efficient use in such systems. This ligand dependency could suggest that the active species might retain one acetylacetonate-type ligand, akin postulated species for Pt(acac) 2 -based precatalytic systems. 43,44 It could also be the result of Co-based nanoparticles/colloids (Co(0)/Co(I) reduced species) formed in the case of Co(acac) 2 that would induce a different catalytic pathway. 11,12,38,40,41 On the contrary, Ni precatalysts still promote the DS pathway, with identifiable characteristic olefinic protons ( Figure 5a) and carbons (Figure 5b) as well as Red associated products (aliphatic regions on both overlays). Noteworthy is that Ni(tmhd) 2 and Ni(acac) 2 seem to be leading to very similar selectivities, in agreement with the model reactions (see Table 1

Conclusions
We have evaluated the activity and selectivity of nickel and cobalt acetylacetonate-based complexes as precatalysts for silylation and crosslinking using Si-H and Si-vinyl functionalized substrates without external reducing agents and under noninert atmosphere. Those complexes promote the in-situ formation of competent catalysts featuring different activities/selectivities depending on the chosen metal and ligand set. This choice consequently affects the crosslinking kinetics/efficiency and the mechanical properties of the final materials. Ni complexes promote dehydrogenative silylation selectively. Co complexes yield both hydrosilylation and DS products for model compounds, whereas the HS product is favored for crosslinking of functionalized oils, which was evidenced harnessing HR-MAS NMR spectroscopy. Co acetylacetonate-based complexes also tend to display activity at lower temperatures than Ni, while being more sensitive to non-inert conditions. When silicone curing formulations are employed, Ni(tmhd) 2 is the most active precatalyst for crosslinking of silicone oils under non-inert atmosphere, probably due to an enhanced solubility in silicone media. The activation of these high-oxidation state complexes does not seem to proceed via Si-H-promoted reduction. Further investigations are currently underway to pinpoint the activecatalyst generation mechanism. Finally, these thermallyinduced crosslinking reactions could prove key to develop efficient "one-pack" formulations, stable at room temperature for extended shelf lives and that could be triggered to form the silicone mastics/putties when heated.

General considerations
Reaction were performed under non-inert atmosphere. All commercially available reagents were used as received. Silicon oils were kindly provided by Elkem Silicones©. Silicon oils were degassed under vacuum, to remove traces of volatiles from synthetic process. All reaction vials were loaded with precisely weighted (0.1 mg precision) substrate and catalyst amounts, without precautions regarding the atmosphere (presence of O 2 and ambient moisture). The chemicals were obtained from the following suppliers: Ni(acac) 2 and Ni(tmhd) 2 from Strem, Co(tmhd) 2 from Alfa Aesar, Co(acac) 2 from Sigma, 1-octene from Sigma, and CDCl 3 and methylcyclohexane (MCH) d14 from Eurisotop.

General procedure for catalytic silylation reactions
A 20mL reaction vial was loaded with the catalyst (0.045 mmol) together with the chosen alkene/vinyl (9mmol) and MD'M (9mmol). Dodecane (standard) was added for GC quantification. The solution was stirred and heated to 90°C for 24 hours. During the course of reaction, aliquots were withdrawn for analysis. After the reaction, NMR and and/or GC/GC-MS analyses were performed. 1 H, 13 C and 29 Si NMR Spectra were recorded on a Bruker 400MHz and 700MHz spectrometers in CDCl 3 and MCH d14. The mass spectra of the products were evaluated by Gas Chromatography coupled with Mass Spectrometry (GC-MS) and quantification was determined with Gas Chromatography (GC) analysis using calibration curves with appropriate substrates. GC was performed on an HP 6890 instrument using a HP5 column made of 5% of diphenyl and 95% dimethyl polysiloxane (32 m length, 0.32 mm diameter, 0.25 µm film) and a flame ionization detector. The injection volume was 1 µL and nitrogen was used as mobile phase with a flow of 1 mL.min -1 . The temperature program was as follow: 40°C plateau for 5 min; heating to 250°C at 10°C.min -1 ; 250°C plateau for 5 min. GC-MS analysis was performed on an Agilent 6850 instrument equipped with a HP5 column made of 5% of diphenyl and 95% dimethyl polysiloxane (32 m length, 0.32 mm diameter, 0.25 µm film) and an Agilent 5975C EI detector. The same temperature program as GC was used for GC-MS analysis. Alkene/vinyl silane conversions and products' selectivities are calculated from 1 H NMR and double-checked using GC/GC-MS analysis (see SI, Sections 1-5) in particular for 1-octene.

Stop-Stirring Time (SST)
For SST, a 20mL vial was loaded with the catalyst (0.02mmol) together with divinyl oil (viscosity 100 mPa.S, 1.58 %wt vinyl function, 4g) and poly-SiH oil (MD' 50 M, 46%wt SiH function, 0.3g). Catalyst loading is expressed in mol% of vinyl functional group. The medium was stirred and heated at 90°C or 110°C. The curing time was measured from the moment of placing the tube at the desired temperature until the magnetic stir bar stopped. Either extreme viscosity or actual crosslinking could be responsible for this observation. This SST was thus correlated to actual rheology measurements and good agreements between SST and effective crosslinking (intersect of G' & G") could be evidenced, validating the method.

Rheology and DSC
For rheology and DSC (differential scanning calorimetry), a 20mL vial was charged with the catalyst (0.014mmol) together with divinyl oil (viscosity 100 mPa.S, 1,58 %wt vinyl function, 3g) and poly-SiH oil (MD' 50 M, 46%wt SiH function, 0.25g). The medium was stirred at ambient temperature. A sample of the solution was taken for analysis. Rheometer analysis are performed on a Rheometer Thermoscientific at 110°C using parallel plate, 0.5 mm gap, 1s-1 frequencies and constraint starting at 3Pa, changing at 50Pa and 1000Pa after reaching 0.001mNm. 0.2mL of previous medium was injected to be tested. DSC thermograms are completed on a Mettler Toledo DSC 300+ at a heating rate of 8°C/min and for kinetic simulation at 4°C/min and 15°C/min. 60-80 mg of previous medium is put in aluminium 100 µL pan.

Conflicts of interest
There are no conflicts to declare.