As displayed in Fig. 1, after 10 min. reaction time the signal of
the catalyst3 (labelled k, 32.3 ppm) still predominates. The
traces of sideproduct7 (peak x, 22.8 ppm) are unchanged
throughout the whole reaction time and therefore cannot be
responsible for any loss of catalytic activity. However, with
time, the signal at À4.3 ppm, which belongs to uncomplexed,
but still surface-bound phosphine, grows. Correspondingly,
the k signal intensity decreases and, towards the end, even
phosphine oxide (34.4 ppm, overlapping partly with signal k)
can be identified by its large chemical shift anisotropy
(CSA).7 These spectra are direct evidence that in the case of
carbonyl nickel catalysts the prevailing mechanism of leaching
is the detachment of the metal moiety from the phosphine lin-
kers, which remain surface-bound, as checked also carefully
with 31P NMR of the supernatant solution. This result is cor-
roborated by earlier experiments by Basset and Maitlis,11 who
could prove the loss of the metal fragment and formation of
Ni0 deposits on the surface indirectly by reactivating the cata-
lyst with CO. Furthermore, even after catalytic runs, we never
found catalytic activity in any of the supernatant solutions.
To solve this leaching problem, we explored several options:
(i) one can improve the reaction conditions further in order to
enhance the lifetime of the catalyst; (ii) the metal center can be
bound more firmly by using a genuine chelating phosphine lin-
ker, and (iii) a continuous approach can be pursued in order to
check whether the catalyst is disturbed by cooling down and
opening the reaction vessel between the cycles. Surprisingly,
even approach (i) brought substantial improvements. Using
cyclooctane as the solvent, for example, increased the yield,
even in the case of 4, up to 100% within 20 h (155 ꢀC, 0.5
mol % 4, 0.5 mol lÀ1 of 5). Interestingly, all catalysts do not
perform as well in n-octane. Analogous to the case of Wilkin-
son-type catalysts,12 the lifetime of the immobilized catalysts
can also be prolonged by reducing the surface coverage. While
the densest possible surface coverage with 8 (Scheme 1, about
16 metal centers per 100 nm2) leads to a maximal TON of
4810, dilution of the surface species to one fourth (4 per 100
nm2) gives a TON of 7616.
Fig. 2 Substrate conversion versus reaction time in batchwise
repeated cyclotrimerization of 5 with catalyst 9 at 155 ꢀC. The ratio
9:5 is 1:200; 0.5 mol lÀ1 5 in cyclooctane. The curves of cycles 3, 5–9,
and 11 have been omitted for clarity.
Fig. 3 Comparison of the performance of catalyst 4 in 9 batchwise
(hatched) and continuous runs (see text). For reaction conditions see
Fig. 2 caption and text.
tane. Then the phenylacetylene was distilled up to the cup as
an azeotropic mixture with the solvent (ratio 5:cyclooctane is
about 1:3). The non-volatile products 6 and 7 were washed
down into the bottom flask by the solvent and thus were con-
tinuously removed from the catalyst. Fresh 5 was added to the
bottom flask as soon as the previous portion was used up,
without otherwise disturbing the system. As shown in Fig. 3,
the performance of 4 is much better, if the runs are done in
a pseudo-continuous, as compared to a batchwise, manner.
Comparing the 31P{1H} MAS spectrum of fresh catalyst 4
with the one of 4 after 9 runs in the soxhlet extractor (Fig. 4)
shows that there is no major decomposition of the catalyst, but
only slightly more phosphine oxide, which most probably
stems from the batchwise addition of 5.
Regarding approach (ii), a variety of chelating phosphines5
with different bite angles and spacer lengths leads to catalysts
8–11 (Scheme 1), all of which give similar catalytic activity
and selectivity. In every case, the initial product ratio 6:7 of
about 4.5:1 changes gradually within the course of 6 cycles
to about 1:1, which persists as the final ratio. This change in
selectivity was investigated in a parallel project recently.13
Overall, with this type of flexible ligand,5 the bite angle or
the distance to the support surface do not make a significant
difference. However, with such chelating ligands the lifetimes
of all immobilized catalysts are prolonged substantially,
and even after 12 cycles the yields are still relatively high, as
displayed in Fig. 2 for 9. In contrast to 4 under non-ideal con-
ditions (Fig. 1), which shows the 31P solid-state NMR signals
of uncomplexed phosphine already in the first run, catalysts
8–11 do so only after several runs. Compared to 9, catalyst 4
with monodentate phosphine ligands gives up to about 10%
less conversion especially in later cycles (Fig. 3) under other-
wise identical (optimal) conditions. Similar success of the
chelate ligands5 has been demonstrated for immobilized
Wilkinson-type rhodium hydrogenation catalysts.12
In conclusion, we could demonstrate that in order to
enhance the lifetime of immobilized catalysts (i) it pays
Following strategy (iii) we made a first step towards contin-
uous catalysis. The silica-supported catalyst 4 was filled into
the porous cup of a soxhlet extractor, and wetted with cyclooc-
Fig. 4 162.0 MHz 31P{1H} MAS spectra of 4 prior to catalysis (lower
trace), and after (upper trace) 9 runs of continuous catalysis (see text).
Rotational speed 13 kHz, for other details of the measurement see
ref. 10.
Scheme 1 Examples of carbonyl nickel catalysts with chelating
ligands.
New J. Chem., 2003, 27, 776–778
777
Reinhard, Simone
