W. Leitner et al.
ed and the effects studied directly in the continuous-flow
system.
after dehydroxylation at 5008C confirmed similar surface
properties to the standard SG100-500 (Figure S15 in the
Supporting Information). At a nominally identical pore fill-
ing of a=0.35, the SG170-500 material yielded SILP cata-
lysts with higher mass loading and increased film thicknesses
than SG100-500 due to its larger pore volume. In the contin-
uous-flow hydrogenation of 1 with scCO2, these catalyst ma-
terials performed largely similar to the ones based on
SG100-500, albeit with slightly faster deactivation (Fig-
ure S16 in the Supporting Information).
When one equivalent of free QUINAPHOS ligand rela-
tive to 3 was added to the SILP, the deactivation surprisingly
resembled more the aged catalyst materials on the basis of 3
alone: activity ceased completely after 20 h on stream and
30000 turnovers without affecting enantioselectivity (Fig-
ure S9 in the Supporting Information). Batch kinetics in
CH2Cl2 showed that hydrogenation of 1 with 3 was signifi-
cantly slower in the presence of additional ligand (plausibly
+
À
due to reversible formation of inactive [Rh(H)
N
species), but still reached full conversion and >99% ee ((S)-
2) at longer reaction times (Figure S10 in the Supporting In-
formation). The influence of excess ligand in the SILP
system is different in that it leads to rapid and irreversible
deactivation of activated 3 shortly after startup. Similar ef-
fects have recently been reported for Rh-based hydroformy-
lation SILP catalysts,[61] which were found to work best at
moderate ligand-to-metal ratios.
The support surface as decisive factor for long-term catalyst
stability: Eventually, the surface properties of the support
material were targeted. To probe chemical interactions of
the surface SiOH groups with the substrate and/or the IL
phase, SG100-500 was deuterated by repeated dehydroxyla-
tion and immersion in D2O.[71] DRIFTS analysis of the re-
sulting material confirmed successful SiOD formation on
the surface (Figure S13 in the Supporting Information). The
To substitute the potentially labile phosphoramidite
material was coated with [EMIM]ACHTNGUTERNNU[G NTf2] and stirred with a
moiety in the ligand framework, the complex [Rh
R)-Me-DUPHOS}][NTf2] (cod=cyclooctadiene;
[(2R,5R)-2,5-dimethylphospholano]-
G
R
mixture of 1 and H2 under scCO2 at 408C for one hour to
mimic reaction conditions. Pure, non-deuterated 1 was re-
covered from this experiment, thus indicating that isomeri-
ACHTUNGTRENNUNG
DUPHOS=(À)-1,2-bis
ACHTUNGTRENNUNG
À
sation and/or H D exchange of the substrate does not occur
with the surface. However, the EMIM cation of the IL had
been deuterated by approximately 20% in the acidic 2-posi-
tion during this procedure. Exchange reactions of imidazoli-
um salts with protic solvents are well known[72] but much
less studied for surfaces.[73] This clearly indicates that in ad-
dition to the anion interaction deduced from the spectro-
scopic studies at ambient conditions, under CO2 pressure
the support might also chemically interact with the cations
of the IL.
1
in CH2Cl2 proceeded with very high initial TOFs
>80000 hÀ1 and high ee values of 97% ((S)-2) at 408C and
20 bar H2 (Figure S11 in the Supporting Information). How-
ever, this pre-catalyst showed a rather disappointing per-
formance when used in SILP materials: maximum ee values
were below 90% and quick deactivation occurred to reach
less than 30000 TONs (Figure S12 in the Supporting Infor-
mation).
To test how surface interactions might influence catalyst
stability, the support material was modified by chemical
functionalisation of the terminal silanols. Polar surface mod-
ification with imidazolium units, as often employed for the
preparation of SILP catalysts based on low-porosity support
materials,[8] yielded completely inactive SILP materials
based on 3 in [EMIM]ACTHNUTRGNEUNG[NTf2]. To provide a non-protic sur-
face, various silica gels with TMS-capped silanols were
tested (Figure S17 in the Supporting Information), but all
As imidazolium-based ILs are known to form N-hetero-
ACHTUNGTRENNUNGcyclic carbene (NHC) complexes with many late-transition-
metal centres under suitable conditions,[63–66] the EMIM
cation was replaced with 4-methyl-N-butylpyrrolidinium (4-
MBP) under otherwise identical conditions. Slightly lower
levels of steady-state conversions were observed in accord
with previous catalyst evaluation,[17] but the stability of 3
was not improved (Figure S14 in the Supporting Informa-
tion). Therefore, NHC formation appeared unlikely to be a
major source of long-term catalyst deactivation.
showed very rapid deactivation of 3 in [EMIM]ACTHUNRGTNEUNG[NTf2] when
tested continuously (Figure S18 in the Supporting Informa-
tion). Finally, a perfluoroalkyl-modified silica gel[74] (mean
pore diameter 59 ꢃ, BET surface area 207 m2 gÀ1, mesopore
volume 0.8 mLgÀ1; SGFLUO) as support material imparted
much improved stability of 3 in the continuous hydrogena-
tion of 1 with scCO2 flow (Figure 9).
The SILP catalyst based on perfluoroalkyl-silica main-
tained its very high initial activity (>99% conversion) and
selectivity (> 99% ee) for at least 10 h on stream. In the fol-
lowing period, a much slower deactivation than with any
other system tested became apparent; ee values were >90%
over 30 h and fell below 80% only after 50 h of continuous
catalysis, which corresponded to 100000 turnovers. After
80 h on stream and a TON of more than 140000, the materi-
Turning the focus towards the support materials, the mor-
phology of silica was varied to qualitatively test whether mi-
crostructure and distribution of the supported IL (film-like
versus pool-like, a matter of much debate[4,67–69]) would
affect long-term catalyst performance. Hot-spot formation,
as discussed for hydrogenation SILP catalysts based on
heat-conducting metal fibres,[70] could in principle also be af-
fected by the support morphology, but it is expected to be
of minor importance in this case in view of a maximum heat
production of only 0.16 W at full conversion. A high-porosi-
ty silica gel 170 (Sꢁd-Chemie; mean pore diameter 262 ꢃ,
BET surface area 261 m2 gÀ1, mesopore volume 1.71 mLgÀ1
SG170-500) was tested, and DRIFTS analysis before and
;
&
6
&
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 0000, 00, 0 – 0
ÝÝ
These are not the final page numbers!