Angewandte
Communications
Chemie
formation of formic acid and ethyl formate over the course of
the reaction raises the question as to whether these com-
[
12]
pounds are intermediates in the formation of methanol. To
examine this possibility, we carried out separate hydrogena-
tion experiments of both compounds. Under the optimal
reaction conditions (1008C, 70 bar of H , 5.6 mol% of [Co],
2
2
4 h; see the Supporting Information), formic acid and ethyl
formate were hydrogenated to methanol in 59 and 66% yield,
respectively. The concentration–time graph in Figure 1 shows
no significant accumulation of formic acid derivatives. If
a stepwise mechanism for CO hydrogenation to methanol is
2
assumed, this result suggests that the hydrogenation of the
intermediates is not rate-limiting. However, the relatively
slow hydrogenation of either formic acid or ethyl formate
contradicts this hypothesis. Furthermore, in control experi-
ments in which no significant amount of methanol was
formed, we still observed similar amounts of ethyl formate
31
1
Figure 2. P{ H} spectra in the 20–30 ppm region for the mixture
Co(acac) ]/Triphos/HNTf in [D ]THF under CO (10 bar) and H
[
3
2
8
2
2
(30 bar) at 1008C after: a) 5 min; b) 1 h 25 min; c) 3 h 10 min;
(
Table 1, entries 6 and 20). On the basis of these observations,
d) 4 h 10 min; e) 5 h 50 min; f) 8 h.
we propose that formic acid and ethyl formate are only
formed as part of a minor reaction pathway. Thus, methanol
formation mainly takes place through an inner-sphere mech-
species, which was formed upon heating, was consumed.
Additionally, after heating for 5 h 50 min, a new doublet at
dP = 22.3 ppm appeared. These results show that on a time-
scale of several hours, after the initial formation of a cobalt–
phosphine complex, the initial species is consumed, and
multiple new complexes are formed.
[8b]
anism. Notably, as compared to the [Ru]/Triphos system,
the cobalt catalyst did not show a decrease in activity at
008C, which might indicate a different rate-limiting step in
the inner-sphere mechanism. Previous reports support this
1
[
9,12]
difference in the rate-limiting step.
This intrinsic feature
could provide potential for more energy efficient, cobalt-
based systems. Interestingly, the addition of formic acid at the
start of the reaction inhibited catalysis (see Scheme S1 in the
Supporting Information). To exclude other intermediates/side
products, we analyzed the gas phase of the reaction after
a reaction time of 96 h (see the Supporting Information).
Only CO , H , and a small amount of N (probably originating
For a better understanding of the nature of the cobalt
complexes formed under the reaction conditions, we carried
out high-resolution electrospray ionization mass spectrome-
try (HRESIMS) experiments. In a sample taken after
a reaction time of 1 h, two major cobalt species were detected:
+
+
[Co(acac) (Triphos)] (1, [M] , m/z 881.248) and [Co(acac)-
2
+
+
(Triphos)] (2, [M] , m/z 782.203; see Figure S3a). Since no
methanol was observed after that time, we conclude that these
two complexes are not catalytically active. In a sample taken
after 16 h, 1 was no longer detected, whereas 2 was still
present (see Figure S3b). This observation indicates that 1 is
an intermediate consumed during the reaction and excludes
the possibility that 2 is only a fragment of 1. On the basis of
these findings and the in situ NMR data, the following
pathway for the formation of the active catalyst can be
2
2
2
from air) were detected, whereas CO, methane, or volatile
ethers, such as DME or ethyl methyl ether, were not observed
(
detection limit: > 0.1 vol%). With this catalyst, CO was not
hydrogenated to methanol (see Scheme S2), thus excluding
the possibility of a CO-based pathway.
Notably, the concentration–time graph (Figure 1) shows
a pronounced induction period of around 6–8 h. This feature
prompted us to study the formation of the active catalyst in
situ by high-pressure NMR spectroscopy. Thus, we followed
the reaction for 8 h at 1008C under 10 bar CO /30 bar H in
proposed (Scheme 1): After initial coordination of Triphos
+
and [Co(acac) (Triphos)] formation, the remaining acac
2
2
2
[
D ]THF (see Figure S2 for details). Figure 2 depicts the
ligands are removed stepwise to liberate the active, cationic
cobalt–Triphos species, which is stabilized by NTf2 . The slow
removal of the acac ligands explains the induction period of
8
3
1
1
À
measured P{ H} spectra in the 20–30 ppm region, which is in
the range of phosphine ligands coordinated to cobalt. Upon
[
13]
heating to 1008C, a doublet at d = 29.6 ppm appeared, along
P
with a poorly resolved feature at d = 23.5 ppm. After 1 h
P
2
2
2
5 min at 1008C, the resolution of the doublet at d =
9.6 ppm had increased, whereas the feature at dP =
3.5 ppm had increased in intensity. Furthermore, at this
P
time, two new doublets appeared at d = 29.3 and 21.9 ppm,
P
which probably correspond to the same species, as well as
a broad signal at d = 24.9 ppm.
P
When heating was continued, the intensity of the pair of
signals at d = 29.3 and 21.9 ppm increased, along with an
P
increase in intensity of the two broad features at d = 23.5 and
P
2
4.9 ppm. The signal at d = 29.6 ppm, however, appeared to
Scheme 1. Proposed pathway for the formation of the active catalyst
P
decrease in intensity, thus indicating that the corresponding
from [Co(acac) ]/Triphos.
3
Angew. Chem. Int. Ed. 2017, 56, 1 – 5
ꢀ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
3
These are not the final page numbers!