1. Introduction
Effective catalytic activation of dinitrogen under ambient conditions remains one of the biggest challenges in synthetic chemistry. Although such a transformation is feasible, as evidenced by the biosynthesis of ammonia enabled by nitrogenase enzymes [
1,
2,
3,
4], the activation of dinitrogen on transition metal complexes is not trivial and the best man-made homogeneous catalysts for ammonia synthesis can perform only several dozens of catalytic turnovers [
5,
6,
7,
8,
9,
10,
11,
12,
13].
In this context, most of the effort has been focused on molybdenum and iron systems [
14,
15,
16,
17,
18]. Two major strategies for catalytic N
2 activation by homogeneous systems were developed: reductive protonation to yield ammonia and reductive silylation to yield silylamines (
Figure 1).
Catalytic formation of ammonia on well-defined systems was pioneered by the group of Schrock who developed a molybdenum-based system that catalyzed the formation of almost eight equivalents of NH
3 per metal atom [
19]. The group of Nishibayashi developed binuclear catalysts which were able to catalyze the formation of up to 63 equivalents of NH
3 per Mo atom [
20,
21]. Iron also proved to be catalytically active as shown by the group of Peters, whose catalysts could produce up to 64 equivalents of NH
3 at very low temperatures [
22,
23].
However, these catalysts suffer from poisoning with ammonia and dihydrogen formed as side products. This is usually not the case in catalytic silylation reactions and therefore this reaction generally yields more turnovers of N
2 fixation (
Figure 2). Catalytic reductive silylation of N
2 was pioneered by Shiina who used CrCl
3 as the pre-catalyst. This led to formation of 5.4 equivalents of N(SiMe
3)
3 when N
2 was reacted with lithium and Me
3SiCl [
24]. Other salts of various metals yielded much less of the desired product. For instance, the use of CoCl
2 which is relevant to the research presented here led to the formation of 1.2 equivalents of N(SiMe
3)
3 per metal centre. This chemistry was further developed for well-defined Mo and W systems by the groups of Hidai [
25] and Nishibayashi [
26]. Another well-defined molybdenum catalyst was recently disclosed by Mézailles and co-workers [
27]. The group of Nishibayashi also reported that simple organometallic iron compounds can be used as pre-catalysts in this reaction and up to 34 equivalents of N(SiMe
3)
3 could be formed when a ferrocene derivative was used [
28]. Interestingly, none of the studied iron compounds contain coordinated dinitrogen, and an approximately 1 h incubation period was observed, indicating that the catalytically active species is formed by reaction of the iron pre-catalyst with silylchlorides under reductive conditions. A two-coordinate iron(0) complex supported with a cyclic (alkyl)(amino)carbene ligand which revealed comparable activity was recently reported by Ung and Peters [
29].
Despite the fact that cobalt-containing catalysts [
30] are active in the Haber–Bosch reaction, only a handful of homogeneous systems based on this metal have been reported in the context of N
2 activation. Yamamoto and co-workers showed that treatment of {[Co(PPh
3)
3N
2]
−}
xM
x+(THF)
2+x (M = Li, Na, Mg) complexes with acids leads to the formation of ammonia and hydrazine in sub-stoichiometric quantities [
31]. Peters and co-workers reported methylation and trimethylsilylation of dinitrogen bound to [(PhB(CH
2P
iPr
2)
3)CoN
2]
2Mg(THF)
4 (
Figure 3) [
32]. Last year, Gagliardi, Lu and co-workers reported a dinuclear cobalt complex that catalyzes the formation of N(SiMe
3)
3 from dinitrogen with an impressive turnover number of 195 [
33], while the group of Nishibayashi showed that in the presence of 2-2’-bipyridine, Co
2(CO)
8 can catalyze the formation of up to 49 equivalents of N(SiMe
3)
3 (
Figure 2) [
34]. Homogeneous cobalt complexes with tris(phosphine)borate ligands were also shown by Peters and co-workers to catalyze the formation of over two equivalents of ammonia [
35].
2. Results and Discussion
Since Yamamoto’s [Co(PPh
3)
3N
2]
− complex promoted the reduction of the coordinated N
2 in THF to ammonia and hydrazine, it seemed plausible that it could also promote the silylation of dinitrogen. Indeed, recently the group of Nishibayashi mentioned that CoH(PPh
3)
3N
2 catalyzes the reductive silylation of N
2 [
34]. Given the fact that [Co(PPh
3)
3N
2]
− can be formed, e.g., by the reduction of CoH(PPh
3)
3N
2 with alkali metals, we decided to investigate the activity of the cobalt triphenylphosphine system in silylation of N
2 in more detail.
One could expect that [CoN
2(PPh
3)
3]
− should directly react with Me
3SiCl, resulting in the formation of Si–N bonds, similarly to the dinuclear system reported by the groups of Gagliardi and Lu [
33]. However, as shown by Nishibayashi, Yoshizawa and co-workers, many simple cobalt complexes reveal long incubation periods before the onset of the catalytic reaction [
34]. We were thus interested in the mechanism behind the catalytic dinitrogen silylation by the cobalt triphenylphosphine system. Is [Co(PPh
3)
3N
2]
− the actual catalyst like the bimetallic Lu–Gagliardi complex or merely a pre-catalyst which undergoes a transformation into the catalytically active species through reaction with chlorosilane? To answer this question we started our investigations by studying the performance of the in-situ formed [Co(PPh
3)
3N
2]
− anion in the catalytic silylation of dinitrogen.
Treating a 0.0025 M solution of CoH(PPh
3)
3N
2 in THF with 180 equivalents of Na and 190 equivalents of Me
3SiCl led to the formation of 6.7 ± 1.0 equivalents (per Co atom) of N(SiMe
3)
3 as evidenced by GC. This corresponded to approximately 12% yield based on sodium (
Table 1). The identity of N(SiMe
3)
3 was confirmed with mass spectrometry. Analysis of the ammonia content using indophenol method after saponification of the reaction mixture with 1 N H
2SO
4 was in accord with the GC analysis.
For comparison, we also investigated the behavior of cobalt compounds that do not have a coordinated dinitrogen molecule. The use of Co(acac)
3 led to the substoichiometric formation of N(SiMe
3)
3, while the use of CoBr
2 or Co powder did not result in the formation of N(SiMe
3)
3 (
Table 1, entries 2–4). The use of cobaltocene led to the formation of N(SiMe
3)
3 in a slightly higher (but comparable) yield to the one obtained with CoH(PPh
3)
3N
2 (entry 5). N(SiMe
3)
3 was not formed when no cobalt was present (entry 6). CoCl
2(PPh
3)
2 which features the triphenylphosphine ligand had significantly lower activity (entry 7), while CoCl(PPh
3)
3 (which upon two-electron reduction under dinitrogen atmosphere could form [Co(PPh
3)
3N
2]
−) was less active than CoH(PPh
3)
3N
2 (entry 8). The use of diethyl ether as a solvent led to a slightly lower yield of N(SiMe
3)
3, and in benzene no formation of this compound was detected (entries 9,10). No significant changes in yield were measured when the amount of solvent varied (entries 11,12). The use of lithium as the reductant led to a comparable yield, while for potassium no desired product was formed (entries 13,14). Potassium gave good results with other systems [
26] and the failure to obtain the desired product with this reductant is likely caused by a very fast reaction of potassium (disappearance of all solids within less than an hour) with all Me
3SiCl before the catalytically active species could be formed (see below). The use of potassium graphite instead of metallic potassium led to the formation of 3.5 equivalents of N(SiMe
3)
3. The exact reason why Me
3SiCl is consumed in unproductive reaction with metallic K at higher rates than with metallic Li or Na or KC
8 is currently not clear but remains outside of the scope of this work. For an example of a somewhat related low performance of metallic K compared to Na and KC
8 in reductive coupling of alkyl chlorides see e.g., [
36].
Overall, CoH(PPh
3)
3N
2 showed a rather moderate activity, and although it performed better than its non-dinitrogen-containing analogs CoCl(PPh
3)
3 and CoCl
2(PPh
3)
2, it was still less effective than cobaltocene. Next, we undertook mechanistic studies of its catalytic activity, and attempted to detect any possible cobalt intermediates in the trimethylsilylation reaction. The time profile of the formation of N(SiMe
3)
3 reveals an induction period of ca. 4 h followed by a period of moderate activity (ca. 16 h) during which all the sodium is consumed (
Figure 4). A somewhat shorter (1 h) induction period was reported by the group of Nishibayashi for the organometallic cobalt systems. This induction period was proposed to account for the formation of catalytically active species that features trimethylsilyl groups directly bound to the cobalt center [
34].
We attempted to shed light on the possible initial transformations of CoH(PPh
3)
3N
2 using IR spectroscopy. Yamamoto et al. reported that the reduction of CoH(PPh
3)
3N
2 with sodium results in the formation of [Co(PPh
3)
3N
2]Na(THF)
3 [
31]. Indeed, the addition of five equivalents of Na to a THF solution of CoH(PPh
3)
3N
2 resulted in the disappearance of its characteristic N≡N stretch vibration at 2091 cm
−1 and the appearance of a new peak at 1912 cm
−1 attributable to [Co(PPh
3)
3N
2]Na(THF)
3. When three equivalents of Me
3SiCl were added to a solution of CoH(PPh
3)
3N
2 in THF, the disappearance of the IR signal corresponding to the cobalt-dinitrogen complex was observed within 20 min and no peaks attributable to a reduced N
2 moiety could be observed. When the addition of Me
3SiCl was performed in the presence of metallic sodium, the disappearance of the signal of CoH(PPh
3)
3N
2 was retarded, suggesting a possible concurrent reaction of Me
3SiCl with sodium. Still, however, it seemed plausible that the attack of Me
3SiCl on [Co(PPh
3)
3N
2]
− would lead to the formation of the active catalyst. Therefore, we investigated whether the reaction of pre-formed [Co(PPh
3)
3N
2]Na(THF)
3 with six equivalents of trimethylchlorosilane would lead to the formation of N(SiMe
3)
3. GC-MS analysis of this reaction mixture did not reveal the formation of either N(SiMe
3)
3 or any organic compound containing an N–SiMe
3 moiety. The subsequent addition of 180 equivalents of sodium and Me
3SiCl to the reaction mixture resulted in the formation of N(SiMe
3)
3, albeit an induction period still was observed.
Since in the first several hours of reaction no formation of N(SiMe3)3 is observed, and the stoichiometric reaction of Me3SiCl with [Co(PPh3)3N2]− does not lead to the formation of N–Si bonds, it seemed rather unlikely that the catalytic reaction proceeds on CoH(PPh3)3N2 (or [Co(PPh3)3N2]Na(THF)3). Indeed, IR spectroscopy analysis of an aliquot of a catalytically active mixture taken after 8 h of reaction revealed no peaks that could be assigned to any Co–N2 species. This suggests that after the induction period, virtually all CoH(PPh3)3N2 is decomposed and the bulk of the cobalt is not coordinated with a terminally bound N2.
The majority of the coordination compounds of cobalt are paramagnetic, therefore we used EPR spectroscopy to probe the catalytic reaction mixture. Investigation of the reaction mixture after 3.5 h, i.e., before the onset of the formation of N(SiMe
3)
3, revealed the presence of a strong signal (
Figure 5, red line) with clearly visible hyperfine couplings with the cobalt center and a very strong and broad signal spanning through the whole measurement window. Its spectral features (
gx = 2.13;
gy = 2.12;
gz = 2.02;
Ax = 215;
Ay = 195;
Az = 180 MHz) are indicative of a low-spin cobalt(II) species [
35,
37,
38]. This species corresponded to approximately 8% of the total cobalt concentration. The amount of paramagnetic material was calculated from double integrals of the EPR signal of the catalytic reaction mixture and of a solution of Cobalt(II) meso-tetraphenylporphine used as the external standard (see
Figure S1 in the supplementary materials). An aliquot from the reaction mixture taken after 48 h revealed the presence of the same paramagnetic species (with less resolved hyperfine structure, due to the lack of formation of good glass upon freezing) corresponding to approximately 5% of the total cobalt concentration (
Figure 5, blue line). These data suggest that during the reaction of the CoH(PPh
3)
3N
2 with Me
3SiCl in THF, a small amount of paramagnetic cobalt(II) species is formed which is not a catalytic intermediate in the N
2 fixation reaction. This species can be formed, e.g., by a transfer of a chloro radical from Me
3SiCl to the cobalt(I) center, or by the attack of the trimethylsilyl radical formed after a one-electron reduction of Me
3SiCl. Formation of other cobalt species which could not be easily identified either using IR or EPR spectroscopy cannot be excluded; however, the reduction of the pre-catalyst to metallic cobalt has not been observed during the reaction.
Monitoring of the catalytic reaction with 31P NMR did not reveal any substantial signal of triphenylphosphine during the incubation period, and only after two days the signal at −6.5 ppm corresponding to free triphenylphosphine was observed, constituting less than 10% of the initial amount of CoH(PPh3)3N2 (triphenylphosphine oxide added after the reaction stopped was used as an internal standard). This points to the possibility that triphenylphosphine remains bound to cobalt during catalysis; however, a fast exchange of triphenylphosphines on the cobalt center can also account for the lack of NMR signal.
To investigate whether cobalt nanoparticles are the active species, we analyzed the supernatant of the reaction mixture for the presence of nanoparticles using dynamic light scattering (DLS) after the formation of N(SiMe
3)
3 reached plateau. Measurements of the clear, brown solution obtained after sedimentation of the insoluble material formed during the catalytic reaction revealed a bimodal distribution of particles with an average size of 20 and 100 nm (see
Figures S2 and S3 in the supplementary materials). This solution was, however, no longer active in the silylation of N
2 when additional sodium and Me
3SiCl were added. This suggests that even if cobalt nanoparticles are formed, they are an unlikely catalyst for the title reaction.
From the above results it is clear that neither CoH(PPh
3)
3N
2 nor its reduction product [Co(PPh
3)
3N
2]Na(THF)
3 activate N
2 towards the formation of N–Si bonds. Dinitrogen complexes of cobalt(II) are known [
39]; however, it is unlikely that the observed paramagnetic complex is the active species given that it is already present in the reaction mixture before the onset of N(TMS)
3 formation is reached. It is possible that this species is being transformed into the actual catalyst during the incubation period. Since good molecular catalysts for the silylation of dinitrogen can reach over 200 turnovers [
25,
33], it is plausible that the actual very active catalyst is present only in minute amounts, which prevents its spectroscopic characterization. Based on the work of Nishibayashi, Yoshizawa and co-workers, an in-situ formed metallosilane species [
28,
34] can be responsible for the catalytic activity of CoH(PPh
3)
3N
2.