Phosphine containing dendrimers for highly regioselective rhodium catalysed hydroformylation of alkenes: A positive 'dendritic effect'

Article abstract of DOI:10.1039/b206597e

Diphenylphosphine functionalised polyhedral oligomeric silsesquioxane (POSS) dendrimers are used as ligands for the rhodium catalysed hydroformylation of oct-1-ene showing unexpectedly high regioselectivity to the linear aldehyde nonan-1-al (l:b = 14:1). Comparative studies with the small molecule analogues, bis-diphenylphosphinopentane, Me₂Si[CH₂CH₂PPh₂]₂ and Si[CH₂CH₂PPh₂]₄, clearly confirm the unusual selectivity of these bidentate dendritic ligands and show that a 'positive dendritic effect' occurs. This high selectivity is obtained with only one structure within the 1st and 2nd generation dendrimer (spacer of five atoms between the phosphorus atoms, and carbon-silicon linkage) whilst other frameworks (spacer of three and seven atoms between the P atoms, or carbon-oxygen-silicon linkage) lead to lower selectivity.

Full text of DOI:10.1039/b206597e

View Article Online / Journal Homepage / Table of Contents for this issue
Phosphine containing dendrimers for highly regioselective rhodium
catalysed hydroformylation of alkenes: a positive ‘dendritic
effect’†
Loïc Ropartz,a Katherine J. Haxton,a Douglas F. Foster,b Russell E. Morris,*a
Alexandra M. Z. Slawina and David J. Cole-Hamilton*a
a School of Chemistry, University of St. Andrews, St. Andrews, Fife, Scotland, UK KY16 9ST
b Catalyst Evaluation and Optimisation Service (CATS), School of Chemistry,
University of St. Andrews, St. Andrews, Fife, Scotland, UK KY16 9ST
Received 8th July 2002, Accepted 4th September 2002
First published as an Advance Article on the web 1st November 2002
Diphenylphosphine functionalised polyhedral oligomeric silsesquioxane (POSS) dendrimers are used as ligands
for the rhodium catalysed hydroformylation of oct-1-ene showing unexpectedly high regioselectivity to the linear
aldehyde nonan-1-al (l : b = 14 : 1). Comparative studies with the small molecule analogues, bis-diphenylphosphino-
pentane, Me2Si[CH2CH2PPh2]2 and Si[CH2CH2PPh2]4, clearly confirm the unusual selectivity of these bidentate
dendritic ligands and show that a ‘positive dendritic effect’ occurs. This high selectivity is obtained with only one
structure within the 1st and 2nd generation dendrimer (spacer of five atoms between the phosphorus atoms, and
carbon–silicon linkage) whilst other frameworks (spacer of three and seven atoms between the P atoms, or
carbon–oxygen–silicon linkage) lead to lower selectivity.
show such a property.21–25 We report here that 1st and 2nd
Introduction
generation polyhedral oligomeric silsesquioxane (POSS)
The hydroformylation of long chain alkenes to form aldehydes
dendrimers containing diphenylphosphine moieties at their
is an important industrial process.1 However, the use of
periphery are successful as ligands for rhodium in the hydro-
rhodium complexes, which show high reactivity and selectivity,
formylation of oct-1-ene. Unexpectedly high regioselectivity
is restrained by the cost and difficult recovery of such transition
to the linear aldehyde (l : b = 14 : 1) was found for a specific type
metals. To our knowledge, no efficient recovery process has yet
of POSS dendritic ligand, i.e. 1st and 2nd generation dendrim-
been successfully applied on an industrial scale for long chain
ers (respectively with 16 or 48 diphenylphosphine end groups)
alkenes, although a pilot plant is currently under construction
with a spacer of five atoms between the P atoms, whilst the
for a process using low pressure distillation of the product. The
small molecule analogues and the other dendritic ligands
use of dendrimers could help to overcome this problem since
prepared (different arm length or composition) showed lower
selectivity. Dendrimers in this paper are described by their
generation number, G1 = 1st generation, G2 = 2nd generation
etc. and the type and number of terminal functionality, e.g.
their size allows recycling of the catalyst using ultra filtration
techniques.2,3 Several examples of such catalyst recovery have
recently been reported.4–6 Because two major products are
formed during the hydroformylation reaction (linear and
G1-16vinyl is a 1st generation dendrimer terminated by 16 vinyl
branched aldehydes), the design of a ligand favouring mainly
groups. Preliminary communications of some of this work have
the linear aldehyde is desirable. High regioselectivity is found
appeared.18–20
when using large excesses of e.g. PPh3 1 or with bidentate phos-
phine chelates such as bis(diphenylphosphino)methyl biphenyl
(BISBI)7 and Xantphos8 type ligands, which give the highest
Results and discussion
regioselectivities (ratio > 50 : 1 to the linear aldehyde).
There are now a number of examples of the use of
dendrimers as catalytic ligands in the literature. However, they
show varying degrees of effect on the processes, with only a
few showing properties that are different from those of small
molecule analogues. These range from total inhibition of the
reaction9 to an improvement in reactivity.4,10–12 Shape-selective
catalytic reactions have also been reported.13,14 However, to our
knowledge, improved product selectivity using these macro-
molecule ligands has only been reported by Fan et al. for
hydrogenation reactions using a BINAP type ligand,15 by
Mizugaki et al. for the stereoselective allylic amination16 and by
ourselves for the hydroformylation of linear alkenes.17–20 Other
dendrimer systems used in hydroformylation reactions did not
Synthesis
The phosphine containing dendrimers were derived from the
G1-16Cl, G1-24Cl and G1-16vinyl, G1-24vinyl, G2-propyl-
48vinyl dendrimers17,26 by nucleophilic substitution of the
chlorine atoms of the terminal chlorosilane groups, or by rad-
ical addition reactions of secondary phosphines onto the vinyl
end groups. Using these two methods, we have synthesized
a series of phosphine-based dendrimers with terminal PR2
groups (R = C6H5, C6H3(CF3)2) (Table 1). The phosphorus
atoms were separated on the same arm by a spacer of three, five
or seven atoms, the central one of which was a silicon atom.
The interest in varying the dendritic structure arose from
preliminary results obtained by van Leeuwen23 and our
group18–20 showing different selectivity for different arm
lengths. The compounds were characterized by NMR,
MALDI-TOF (Matrix Assisted Laser Desorption Ionization-
Time of Flight) mass spectrometry and microanalysis.
† Electronic supplementary information (ESI) available: plots showing
the effect of temperature and dendrimer structure on the rate of
hydroformylation of oct-1-ene. See http://www.rsc.org/suppdata/dt/b2/
b206597e/
DOI: 10.1039/b206597e
J. Chem. Soc., Dalton Trans., 2002, 4323–4334
This journal is © The Royal Society of Chemistry 2002
4323

View Article Online
Table 1 Diphenylphosphine functionalised POSS dendrimers
Entry
Dendrimer
Synthetic method
31P NMR δ/ppm
Conversiona (%)
Yielda (%)
1
2
3
4
5
G1-16ethylPPh2
G2-propyl-48ethylPPh2
G1-24ethylPPh2
G1-16propylPPh2
G1-16ethylPAr2
Ar = 3,5-C6H3(CF3)2
G1-24methylPPh2
G1-16methylPPh2
G1-16methoxyPPh2
G1-16ethoxyPPh2
G1-16propylPPh2
1b
1b
1b
1b
1b
Ϫ9.6, Ϫ9.7, Ϫ14.4 (w)
Ϫ9.9 (br)
Ϫ9.5 (br)
>99f, g, h
84f
95
77
95
96
92
60f, g
56f
Ϫ17.3
Ϫ6.6, Ϫ6.9, Ϫ7.0, Ϫ12.9
75f
6
7
8
9
10
2c
2c
3d
3d
4e
Ϫ23.9
Ϫ22.6 (br), Ϫ22.7 (w)
Ϫ13.6 (br)
Ϫ22.3, Ϫ22.5
Ϫ16.1, Ϫ16.2, Ϫ16.3 (w), Ϫ16.4
60f, g
93
60–70
76
85
75
>95f, g
>86f, g
>92f, g
>85f
a Yield refers to the isolated chemical yield; conversion refers to the percentage of vinyl or chloro groups converted to phosphines. b Addition of
HPPh2 to vinyl or allyl POSS. c Reaction of R2PCH2Li with chlorosilane derivatised dendrimer. d Reaction of Ph2P(CH2)nOH with –SiMeCl2
derivatised G1 dendrimer. e Reaction of Ph2P(CH2)3MgBr with –SiMeCl2 derivatised G1 dendrimer. f 1H NMR. g MALDI-TOF. h Microanalysis.
Following the same experimental method as previously
described for diethylphosphine type dendrimers,17 the reaction
of diphenylphosphine (excess) with the vinyl substituted POSS
was carried out using AIBN as a radical initiator to give the
corresponding diphosphine-functionalised POSS. Quantitative
yield and excellent to good conversions were obtained for G-
16ethylPPh2 (complete conversion) and G2-propyl-48ethylPPh2
(84% conversion) (Table 1, Entries 1 and 2).
the selectivity of the hydroformylation reaction, although this
depends on the configuration of the complex.27–29 The complete
radical addition of the fluorinated arylphosphine onto the G1-
16vinyl17 compound failed since 25% of unreacted vinyl groups
were still present after one month, even after numerous con-
secutive additions of AIBN. The substituted POSS compound
(G1-16ethylPAr2, Ar = 3,5-C6H3(CF3)2) showed four 31P signals
at δ Ϫ6.6, Ϫ6.9, Ϫ7.0 and Ϫ12.9 ppm, indicating a variety of
phosphorus environments in such a partially substituted com-
pound (Table 1, Entry 5). As expected these 31P NMR reson-
ances are shifted to higher frequency by an average of 2–3 ppm
compared with the diphenylphosphine substituted analogue.
Substitution of the chlorine atoms of the chlorosilane POSS
by nucleophiles allows us to modify the internal structure of
our diphenylphosphine containing dendrimer. Addition of
LiCH2P(C6H5)2-TMEDA30 in excess to G1-24Cl26 failed to give
satisfactory results since after 3–5 days of reaction, only partial
substitution (60%) was reached (Table 1, Entry 6). It is believed
that this was caused by the steric hindrance of the diphenyl-
phosphine groups. As the complete substitution of the 24-Cl
POSS failed, we prepared the homologue with 16 arms G1-
16methylPPh2. Good conversion was obtained (>95%, Table 1,
Entry 7). A major peak was found in the MALDI-TOF
spectrum at m/z = 4176 (Mϩ = 4173.5). Interestingly, fragment-
ation clearly occurred since a peak at m/z [M Ϫ (PPh2)] was
identified in the mass spectrum. Good purity products were
obtained by silica gel column chromatography with 60–70%
yields of a white solid. The 31P NMR spectrum showed two
resonances at δ Ϫ22.6 and Ϫ22.7 ppm. Shorter reaction times
led to the formation of partially substituted products with
e.g. 12–13 arms.
However, only 60% conversion was reached for the 24 arm
counterpart, G1-24ethylPPh2 (Table 1, Entry 3), probably due
to steric crowding. For this reason, this dendrimer has only
been characterised by NMR spectroscopy. The 31P NMR
spectra showed that the phosphine groups in the same
dendrimer for the three compounds were not all equivalent.
Indeed two major and one minor peak were found for the
16-branched dendrimer G1-16ethylPPh2 (Table 1, Entry 1) at
δ Ϫ9.6, Ϫ9.7 ppm (major) and δ Ϫ14.4 ppm (minor). Broad
signals centered at δ Ϫ9.9 ppm for G2-propyl-48ethylPPh2
(Table 1, Entry 2) and at Ϫ9.5 ppm for G1-24ethylPPh2
(Table 1, Entry 3) were detected. We assume that several
P environments are observed for these dendrimers mainly
because the conversion is not complete. This means that some
arms will have 3, 2 or 1 -PPh2 termini. These may well have
different 31P NMR spectral resonances. In addition, back-
folding of a phosphine terminus inside the dendrimer could
lead to a different 31P NMR shift. The excess of phosphine was
removed at 120 ЊC under vacuum or by silica gel column
chromatography. Although lower yields were obtained using
chromatography (86%), no change in the conversion of the
product was found. These compounds were low melting point
air sensitive solids, which were extremely soluble in most
1
organic solvents, except polar ones such as ethanol. H NMR
Diphenylphosphino alcohol compounds were also used to
introduce phosphine moieties into the dendrimer. Addition of
these alcohols to G1-16Cl17 in the presence of base led to
diphenylphosphine-containing dendrimers with different length
spacers between the phosphorus atoms. Diphenylphosphino-
methanol (prepared in situ)31 was added in excess (3 to 1) to
the 16-chloro POSS in the presence of triethylamine. Good
conversion (>86%, Table 1, Entry 8) to the G1-16methoxyPPh2
compound was obtained after six days (only 60% after two
days). After purification through silica gel column chrom-
atography, a colourless non-crystalline product was isolated in
spectra and the microanalysis did not show the presence of
unreacted vinyl groups in G1-16ethylPPh2 but further char-
acterization by MALDI-TOF MS showed the presence of
mainly 14, 15 or 16 phosphine groups. This suggests that
fragmentation occurred during the MALDI-TOF analysis. In
addition, the compound was found to be partially oxidized
(which probably occurred during the MS analysis process) since
mass increments of 16 due to the oxygen atoms were present in
the spectrum.
Addition of HPPh2 to the allyl POSS G1-16allyl17 to prepare
the propyl analogue G1-16propylPPh2 failed since even after
heating for two weeks and several additions of radical initiator
only 56% conversion was obtained (Table 1, Entry 4). This poor
conversion is probably due to the lesser reactivity of the allyl
silane.
Attempts to introduce arylphosphines with electron with-
drawing groups (radical addition of bis[3,5-(trifluoromethyl)-
phenyl]phosphine27) into the periphery of the dendrimer were
also carried out. Such compounds are very interesting as
ligands since the electron withdrawing groups may improve
1
76% yield. H NMR and MALDI-TOF MS showed that an
average of 14 arms (of the possible 16) contained the diphenyl-
phosphine species (conversion >86%). Partial fragmentation
may also have occurred since peaks at m/z {M Ϫ x(CH2PPh2)}
were found (x = 0, 1, 2, 3, etc.). A single signal in the 31P NMR
was found at δ Ϫ13.6 ppm confirming the presence of the
phosphine species.
To vary the chain length between the phosphorus atoms,
excess 2-diphenylphosphinoethanol32 was reacted with G1-
16Cl. The product, G1-16ethoxyPPh2, was purified by silica gel
4324
J. Chem. Soc., Dalton Trans., 2002, 4323–4334

View Article Online
Table 2 Results of hydroformylation reactions of oct-1-ene using G1-16ethylPPh2 ligands
Entry
Ligand
T /ЊC
P/bar
t/h
k/10Ϫ3 sϪ1
Conversion (%)
Isomerisation (%)
l : b ratio
Nonan-1-al (%)
1
2
3
4
5
6
7
Ab
Ab
Ab
Ab
Ab
Ba
Ab
80
80
100
120
100
120
120
20
10
10
10
20
10
10
24
19
4
2
6
2
0.2
0.05
0.08
0.42
1.1
0.33
1.2
>99.9
> 99.9
> 99.9
> 99.9
> 99.9
> 99.9
56.7
2.5
6.0
5.0
6.3
6.6
8.8
10.8
12.0
7.5
13.9
15.4
83.9
83.5
86.2
85.6
3.2
84.7
6.6
86.0
1.1
6.6, 14c
48.5, 86c
Reaction conditions: [Rh(acac)(CO)2] (2.0 × 10Ϫ5 mol), toluene (4 cm3), oct-1-ene (8.3 × 10Ϫ3 mol), CO : H2 = 1 : 1.a G1-16ethylPPh2 dendrimer with
16 arms functionalised, P : Rh = 6 : 1. b G1-16ethylPPh2 dendrimer with 12 arms functionalised P : Rh = 5.4 : 1. c Based on the amount of oct-1-ene
consumed.
column chromatography to give a non-crystalline solid in 85%
yield (Table 1, Entry 9). An average of 14–15 arms were
converted to the phosphine species (>92%). The MALDI-TOF
spectrum showed a large distribution of peaks. However,
partial decomposition occurred during the MALDI-TOF
analysis since the observed peaks correspond to {M Ϫ x(CH2-
CH2PPh2)} (x = 0, 1, 2, 3, etc.) rather than {M Ϫ x[(OCH2-
CH2PPh2) ϩ Cl]} or more probably at {M Ϫ x[(OCH2-
CH2PPh2) ϩ OH], which would be expected if only partial
substitution had occurred. The 31P NMR spectrum showed two
broad signals centered at δ Ϫ22.3 and Ϫ22.5 ppm indicating
that not all the phosphorus atoms were in identical environ-
ments. To obtain a more stable compound (than the one with
a siloxane linker) with a seven atom spacer between the
two phosphorus atoms, we reacted the Grignard species
bidentate coordination might be expected, the many different
phosphorus environments may mean that when a larger
amount of rhodium is loaded, all of the bidentate sites are
filled and some rhodium binds between two single P atoms on
different dendrimers.
The catalytic reactions were carried out using different
batches of G1-16ethylPPh2 ligand since a first synthesis led
to incomplete conversion to the desired dendrimer (12 arms
substituted out of 16). This partially functionalized ligand
will be named ‘A’. The second type of dendritic ligands used
(different batches) was shown to have more than 15 arms
functionalized (ligand named ‘B’). A first set of experiments
was carried out using ligand A at 80 ЊC under 20 bar of H2/CO
with a phosphine : rhodium ratio of 5.4 : 1. After reaction (<24
h, Table 2, Entry 1) high conversion to the aldehydes, nonan-1-
al and 2-methyloctan-1-al was observed. High selectivity to the
linear aldehyde (83.9%) with a linear to branched ratio (l : b)
of 6.6 : 1 was obtained. A first order dependence in substrate
and rhodium was determined for this reaction. Similar results
although with a slightly increased rate were obtained at lower
CO/H2 pressure (Table 2, Entry 2).
Higher temperatures, i.e. 100 and 120 ЊC (H2/CO 10 bar),
increased the reaction rate allowing completion of the reaction
respectively in four and two hours (Table 2, Entries 3 and 4).
These higher reaction temperatures combined with lower CO/
H2 pressure (10 bar) led also to an increased selectivity to the
linear aldehyde (up to 85.6%) with an improved l : b ratio
of 10.8 or 12 : 1 (cf. 7.5 at 20 bar and 100 ЊC, Table 2, Entry 5).
The gas uptake plots at different temperatures showing the first
order kinetics are shown in the ESI, Fig. S1. Even higher select-
ivity was obtained under these conditions (120 ЊC, 10 bar) with
the more crowded dendrimer B (Table 2, Entry 6), since 86.0%
of the product was nonan-1-al with a l : b ratio of 13.9 : 1. It
thus appears that, for a given P : Rh ratio (5.4 : 1), both the rate
constant (from 0.08 × 10Ϫ3 to 1.1 × 10Ϫ3 sϪ1) and l : b ratio (from
6.6 to 12) markedly increased when the pressure was lowered
(compare entries 1 and 2 or 5 and 3 in Table 2) or the temper-
ature increased (compare entries 2, 3, and 4 in Table 2). This
increased selectivity is probably related to β-H-elimination,
since isomerisation is more prevalent when the l : b ratio is
higher.1,32 In all cases, traces of hydrogenated product, nonan-1-
ol (<0.4%), and substrate, octane (0.5–1.4%), were detected, but
the major side products were isomerised octenes (Table 2).
The intrinsic l : b ratio (15.4 : 1) for the hydroformylation
of oct-1-ene by rhodium complexes of G1-16ethylPPh2 was
obtained by running the hydroformylation over a short period
(0.2 h). Longer reaction times led to less selective reactions,
probably because some of the isomerised alkenes are converted
into branched aldehydes (Table 2, Entry 7).
33
ClMg(CH2)3PPh2 with the 16-chloro POSS. The addition
of an excess (three fold) of the Grignard compound to G1-
16Cl POSS afforded
a
partially functionalized POSS
G1-16propylPPh2 (76% conversion after 36 h, Table 1, Entry
10). After four days only a slight increase of the conversion was
obtained since 86% (1H NMR) of the chlorides were
substituted. The compound was then isolated after work up as a
colourless non-crystalline solid in 75% yield. Four 31P chemical
shifts were found at δ Ϫ16.1, Ϫ16.2, Ϫ16.3 and Ϫ16.4 ppm
indicating the different environments of the phosphorus atoms.
The microanalytical data and the MALDI-TOF mass spectrum
of this complex were not in agreement with the proposed
formula, so the assignment of the structure of this dendrimer
must be regarded as only tentative.
In summary, we have synthesized POSS centred dendrimer
bound alkyldiphenylphosphines with at least two phosphine
groups per dendrimer branch. The P atoms are separated by
three, five or seven atoms. The central atom is always Si and the
other atoms may be all C or two of them may be O. When O is
present, it is always adjacent to the Si atom. Generally speaking
the Si atom is attached to the POSS core by a two carbon
spacer, although three carbon spacers are used in G2-propyl-
48ethylPPh2.
Catalytic reactions
Hydroformylation using rhodium complexes of G1-
16ethylPPh2 and small molecule analogues. G1-16ethylPPh2 was
reacted in situ with [Rh(acac)(CO)2] and the resulting solutions
used for the hydroformylation of oct-1-ene. With Rh : P > 3 : 1,
the catalytic complexation occurred in a few seconds to give
homogeneous yellow or orange solutions depending on the
concentration of rhodium and/or dendrimer. Under H2/CO
pressure these solutions turned bright yellow and they had a
similar appearance after the hydroformylation reactions. Low
phosphine : rhodium ratios (≤3 : 1) led to the formation of
an insoluble solid (via a gel). Attempts to re-dissolve the pre-
cipitate failed. This is possibly due to cross-linking of the
dendrimers through rhodium, as the P : Rh ratio in the system
decreases, leading to oligomeric dendrimer species. Although
The effect of the phosphine : rhodium ratio under fixed
reaction conditions (10 bar of H2/CO, 120 ЊC, [Rh] = 3.77 ×
10Ϫ3 mol dmϪ3) was also studied. It was found that the
dendrimer could only support a limited amount of rhodium
species since when a P : Rh ratio of 2 : 1 was used, precipitation
of a metallic residue occurred with a lowering of the rate
J. Chem. Soc., Dalton Trans., 2002, 4323–4334
4325

View Article Online
Table 3 Results of hydroformylation reactions of oct-1-ene varying the phosphine/rhodium ratios using G1-16ethylPPh2 ligandsa
Entry
Ligand
P : Rh
k/10Ϫ3 sϪ1
Isomerisation (%)
l : b ratio
Nonan-1-al (%)
8
9
10
11
B
2
0.52
0.95
1.1
35.5
7.1
6.3
6.6
3.4
11.9
12
48.4
84.3
85.6
85.2
A
A
A
3.6
5.4
10.8
1.5
12.2
a [Rh(acac)(CO)2] (2.0 × 10Ϫ5 mol), P : Rh = 6 : 1, toluene (4 cm3), oct-1-ene (8.3 × 10Ϫ3 mol), 120 ЊC, CO/H2 10 bar, total reaction time 2 h,
conversions all >99.9%.
Table 4 Comparative study on the hydroformylation reaction of oct-1-ene using small molecule analogues of the dendrimer-based ligands
Ligand
k/10Ϫ3 sϪ1
Oct-2-ene (%)
3- ϩ 4-Octene (%)
Nonan-1-al (%)
l : b ratio
1
2
3
3.6
3.0
2.1
10.6
5.8
6.2
0.83
0.90
0.40
70.1
72.7
77.4
3.4
3.8
5.2
Reaction conditions: [Rh(acac)(CO)2] (2.0 × 10Ϫ5 mol), P : Rh = 6 : 1, toluene (4 cm3), oct-1-ene (8.3 × 10Ϫ3 mol), 120 ЊC, CO/H2 10 bar, total reaction
time 2 h, conversion all > 99.5%.
constant and aldehyde l : b ratio (3.4 : 1) (Table 3, Entry 1). This
may be related to the formation of oligomeric complexes, as
discussed above, although the large amount of isomerised
alkene may indicate that significant amounts of unligated
rhodium, which could also be responsible for the rhodium
plating, are also present. Interestingly, increasing the P : Rh
ratio from 3.6 to 10.8 : 1 (ligand A) led to a higher rate constant
(Table 3, Entries 9–11) while the l : b ratio stayed virtually
unchanged at ca. 12—the upper limit that we have observed for
dendrimer A. A possible explanation for this unusual behaviour
is that rhodium is strongly bound to the dendrimer, probably
via two P atoms, and that steric constraints prevent the
coordination of extra P atoms to block vacant sites and inhibit
the reaction, as is observed when using triphenylphosphine.
The high selectivity to nonan-1-al observed with these
dendrimer-based ligands was surprising since potentially
bidentate ligands with a spacer of five atoms (or more) between
the two P atoms usually only show high linear selectivity if
the backbone is constrained in some way, as for BISBI7 and
Xantphos8 type ligands, although certain flexible analogues
of Xantphos containing P(CH2)2O(CH2)2P backbones can give
l : b ratios up to 9.2.34
in Fig. 2. The rates of reaction using the small molecules
were, however, higher than those found when using the
dendrimer-based ligand.
Fig. 2 Selectivity to nonan-1-al and l : b ratio in hydroformylation
reactions of oct-1-ene using G1-16ethylPPh2 POSS, B or small molecule
analogues. 1 = H2C(CH2CH2PPh2)2, 2 = Me2Si(CH2CH2PPh2)2, 3 =
Si(CH2CH2PPh2)4.
The selectivity using 3, which can be thought of as a first
generation dendrimer, albeit with a different core was better
than that obtained with 2. 2 was also found to give a higher
selectivity than the less hindered ligand, 1. As expected for
the more sterically crowded phosphine ligand 3, the rate
constant, which was 3.0 × 10Ϫ3 sϪ1 for 2, dropped to 2.1 ×
10Ϫ3 sϪ1 (P : Rh = 6 : 1).
Using 3 the selectivity to nonan-1-al rose from 55.9 to 79.9%
when phosphine : rhodium ratios of respectively 2 : 1 to 12 : 1
were used. This type of dependence is reminiscent of
that obtained when using unidentate rather than bidentate
phosphines. No such effect was found for the dendrimer-based
ligand, confirming the strong chelation of the dendrimer to
rhodium.
In order to determine the exact influence of the dendrimer
structure on the reactivity and selectivity of the catalytic
reaction, the hydroformylation of oct-1-ene was carried out
using small molecules as ligands with structures related to those
of the dendrimer based phosphines. 1,5-Bis(diphenylphos-
phino)pentane (1), bis(diphenylphosphinoethyl)dimethylsilane
(2), and tetrakis(diphenylphosphinoethyl)silane35 (3) (Fig. 1)
were used for the in situ formation of rhodium complexes and
Hydroformylation using rhodium complexes of G1-
24methylPPh2. In order to examine the effect of the spacer
length between phosphine groups in the dendrimer upon the
hydroformylation reactions, catalytic solutions prepared in situ
from G1-24methylPPh2 (incomplete substitution, i.e. 60% func-
tionalised and [Rh(acac)(CO)2] or [Rh2(O2CMe)4] was used for
the hydroformylation of hex-1-ene under various different con-
ditions. The l : b ratio was never higher than 2.3 at a P : Rh ratio
of 4 : 1 (Table 5, Entry 12) and dropped slightly as the amount
of dendrimer added was reduced (Table 5, Entries 12–15). Simi-
lar results were obtained at lower rhodium concentration (Table
5, Entry 16) or replacing toluene with ethanol (Table 5, Entry
17), although in this last reaction, small amounts of alcohol
Fig. 1 Small molecule analogues of the dendrimers used as ligands in
the hydroformylation reaction.
their use in catalysis under the conditions of Table 2, Entry 1.
The results of the reactions are reported in Table 4 and Fig. 2.
Whilst the dendrimer gave a selectivity to the linear aldehyde,
nonan-1-al, of 86%, the small molecules 1, 2 and 3 respectively
led to 70.1, 72.7, and 77.4% selectivity to the desired aldehyde
(Table 4) clearly showing that there is a positive dendritic effect
on the selectivity of the reaction. Selectivities obtained using
the small molecules and the dendrimer-based ligands are shown
4326
J. Chem. Soc., Dalton Trans., 2002, 4323–4334

View Article Online
Table 5 Hydroformylation of hex-1-ene catalysed by rhodium complexes formed with the G1-24methylPPh2 dendritic ligand
Entry
Solvent
[Rh]/10Ϫ3 mol dmϪ3
P : Rh
t/h
Conversion (%)
Aldehydes (%)
Alcohols (%)
l : b ratioa
12
13
14
15
16
17
Toluene
Toluene
Toluene
Toluene
Toluene
Ethanol
8
8
8
8
4
8
4
3
2
1.5
3
2.5
2.5
4
4
2.5
95
95
>99
>99
89
92
92
96
97
86
93
—
—
—
—
—
5
2.3
2.3
2.0
1.9
2.4
1.8
4
16
>99
Reaction conditions: [Rh2(O2CMe)4] or [Rh(acac)(CO)2] (entry 12 only), solvent 4 cm3, hex-1-ene 8.0 × 10Ϫ3 mol, 120 ЊC, H2/CO 40 bar.a Linear to
branched ratio of aldehyde only.
Table 6 Hydroformylation of oct-1-ene catalysed by Rh complexes of POSS derived dendrimer diarylphosphines
Entry
Ligand
t/h
Rate/10Ϫ3 sϪ1
Isomer (%)
Nonan-1-al (%)
l : b ratio
18
19
20
21
22
23
G1-16methylPPh2
G1-16ethylPAr2
G1-16methoxyPPh2
G1-16propylPPh2
G1-16ethoxyPPh2
G2-propyl-48ethylPPh2
0.3
2
2
2
2
6.2
1.1
0.7
1.5
2.0
0.6
12.1
20.0
9.0
5.1
8.3
68.5
73.0
76.2
78.0
78.1
83.8
3.9
15.0
5.7
5.0
6.4
a
3
7.5
11.5
Reaction conditions: [Rh(acac)(CO)2] = 2.0 × 10Ϫ5 mol, P : Rh = 6 : 1, substrate 8.3 × 10Ϫ3 mol, toluene (4 cm3), 120 ЊC, CO/H2 10 bar, conversion all
>99.9%.a Ar = 3,5-C6H3(CF3)2.
were formed. The best selectivity obtained to the linear alde-
hyde, heptan-1-al, was 67.5% with a linear to branched ratio of
2.3 : 1 (Table 5, Entries 12 and 13).
These selectivities are considerably lower than those obtained
with G1-16ethylPPh2, confirming that a three atom spacer
between P atoms does not afford enhanced linear selectivity.
The rate of reaction with this G1-16methylPPh2 is ca. six times
higher than for the G1-16ethylPPh2 dendritic ligand (Table 2,
Entry 6). In addition, a high amount of isomerised substrate
(12.2%) is formed with the shorter spacer. Modelling studies,
to be reported separately,36 show that the three atom spacer
does not place the P atoms in the optimal relative position
for bidentate coordination with both P atoms in equatorial
positions in five-coordinate intermediates.
The hydroformylation of oct-1-ene using the dendritic ligand
G1-16ethylPAr2 (Ar = 3,5-C6H3(CF3)2) was also carried out
under similar reaction conditions. Interestingly, the pre-formed
catalytic mixture (in toluene) was heterogeneous at room
temperature showing an extremely pale brown solution and
a brown oily residue. This heterogeneity is remarkable since
both the rhodium-based catalyst, [Rh(acac)(CO)2], and the
dendrimer were independently soluble in toluene. A homo-
geneous orange–brown solution was however obtained in warm
toluene. The hydroformylation of oct-1-ene catalyzed by this
dendritic rhodium complex gave 73.0% selectivity to nonan-1-al
(Table 6, Entry 19), however the reaction gave a high l : b ratio
(15 : 1). This ligand promotes significant isomerisation activity
(20%), but is inactive for the hydroformylation of internal
alkenes; only traces of 2-ethylheptan-1-al were formed.
Interestingly, extensive studies carried out by van Leeuwen
and co-workers showed that a compound derived from tetra-
vinylsilane containing 16 Ph2P arms, but with only one CH2
spacer between the silicon and phosphorus atoms (i.e. the same
end groups as in G1-24methylPPh2) also gave an l : b ratio of
2.3 : 1, the same as that obtained using the small molecule
analogue (Me2Si(CH2PPh2)2).23 Thus, the spacer length between
the P atoms seems to be important in determining whether or
not an enhanced linear selectivity is obtained when using
dendrimers based on phosphines as ligands for rhodium
catalysed hydroformylation reactions.
Hydroformylation using rhodium complexes of other dendrim-
ers nominally with 16 PPh2 end groups; varying the spacer length
and the atoms in the spacer group. Since the regioselectivity
of the reaction was different using the dendritic ligand,
G1-24methylPPh2 from that obtained with G1-16ethylPPh2, it
was interesting to compare the latter ligand with other
16-branched dendrimers. The reactions were carried out under
the optimal conditions found for G1-16ethylPPh2 and the
results are summarised in Fig. 3 and Table 6.
Surprisingly the dendritic ligand G1-16methoxyPPh2, which
has a similar structure that of G1-16ethylPPh2 (the methylene
group attached to the peripheral Si atom is replaced by O), and
so should form similar rhodium complexes, did not lead to high
selectivity to nonan-1-al. Indeed, whilst 86.0% of the product
was the linear aldehyde using G1-16ethylPPh2 (Table 2, Entry 6)
only 76.2% of nonan-1-al was formed using the rhodium/G1-
16methoxyPPh2 species (Table 6, Entry 20). Therefore not only
is the length of the bridge between the P atoms an important
factor in determining the selectivity of the reaction, but also the
composition of this bridge can influence the selectivity of the
catalytic species. It is important to note that the electronic
properties of the phosphine species on the two different
dendrimers differed since 31P NMR resonances were at δP Ϫ9.4
and Ϫ13.6 ppm for G1-16ethylPPh2 and G1-16methoxyPPh2,
respectively. It should also be noted that a small amount of
precipitate (yellow solid) was found at the end of the reaction
using G1-16methoxyPPh2 indicating that the dendritic ligand
Fig. 3 Selectivity to nonan-1-al and l : b ratio using different POSS
dendritic ligands,
1 = G1-16methylPPh2, 2 = G1-16ethylPAr2,
3 = G1-16methoxyPPh2, 4 = G1-16propylPPh2, 5 = G1-16ethoxyPPh2,
6 = G2-propyl-48ethylPPh2, 7 = G1-16ethylPPh2 (dendrimer B).
The G1-16methylPPh2 dendritic ligand gave similar results in
the hydroformylation of oct-1-ene (68.5% nonan-1-al, Entry 18,
Table 6) to those of its counterpart G1-24methylPPh2 for hex-1-
ene (67.5% heptanal, Entry 12, Table 5, although at 40 bar).
J. Chem. Soc., Dalton Trans., 2002, 4323–4334
4327

View Article Online
may not be completely stable under the reaction conditions (the
pre-formed catalytic solution was bright yellow with no precipi-
tate and it is not known whether the precipitate occurred during
the reaction or after cooling and depressurisation). However,
since the kinetics of the hydroformylation reaction were first
order in oct-1-ene, it is unlikely that the differences observed
in the distribution of the products arose from decomposition of
the catalyst. Another explanation considered is that there may
be a different geometry of the two dendrimers due to the oxy-
gen atom linker. Indeed if such a difference exists, as suggested
by very preliminary modelling studies on this dendrimer which
suggest that it is more compact than the G1-16ethylPPh2, it
would modify the mode of coordination of the phosphine
species in the rhodium complexes, possibly leading to different
catalytic species and so different reactivity and selectivity. The
lower rate of reaction obtained with this ligand (compared
to the G1-16ethylPPh2 system) supports such a hypothesis.
Isomerisation of the substrate was relatively high since 8.82
and 0.22% of the products were respectively oct-2-ene and
oct-3-ene. This high isomerisation rate led to a relative good
l : b ratio of 5.7 : 1. The isomerisation is not followed by
hydroformylation of the internal alkenes since only traces of
2-ethylheptanal were found (0.17%).
isomerisation process increased marginally when passing from
the partially functionalised G1-16ethylPPh2 A to the bulkier
G1-16ethylPPh2 B (Table 2, Entries 4 and 6).
Reactions with all the dendritic complexes showed a first
order dependence on substrate concentration (see Fig. S2, ESI).
The perfect straight lines of the plots of ln{[P(t) Ϫ P(min)]/
[P(0) Ϫ P(min)]} versus time of the different dendrimers clearly
confirm that only hydroformylation of oct-1-ene occurred
without other side reactions consuming H2 or CO. The least
regioselective and most compact ligand, G1-16methylPPh2, led
to the highest rate of reaction (ca. three times the rate found
with G1-16ethylPPh2) while the less sterically hindered ligands
G1-16ethoxyPPh2 and G1-16propylPPh2, showing intermediate
selectivity, led to rate constants slightly higher than the G1-
16ethyl PPh2 catalytic system (Fig. S2, ESI). However, the
correlation between selectivity and rate of reaction, i.e. high
selectivity coupled with low rate, is not straightforward. Indeed,
although G1-16methoxyPPh2 led to the lowest rate of reaction
for the 1st generation dendrimers (k = 0.7 × 10Ϫ3 Ϫ1), a low
s
regioselectivity to nonan-1-al was obtained (76.2%). The rate
constant using the second generation dendrimer, G2-propyl-
48ethylPPh2, was ca. half that for the G1-16ethylPPh2 system
whilst G1-16ethylPAr2 gave similar kinetics to those of its
unfluorinated counterpart.
The use of the dendritic ligands G1-16propylPPh2 (Table 6,
Entry 21) and G1-16ethoxyPPh2 (Table 6, Entry 22) further
emphasises the importance of the spacer length for obtaining
high linear selectivities using this type of dendrimer-based
ligand. Indeed the regioselectivity to the linear aldehyde
during the hydroformylation of oct-1-ene was lower with these
two ligands (ca. 78% for both) than with the G1-16ethylPPh2
dendritic ligand (86%) (Table 2, Entry 6). Nevertheless, these
selectivities are still higher than those obtained with small
molecules (Table 4). Interestingly, unlike for G1-16ethylPPh2
and G1-16methoxyPPh2, the different electron density of the
phosphorus atoms of these two dendrimers, G1-16ethoxyPPh2
and G1-16propylPPh2, did not seem to influence the selectivity
of the reaction. Indeed although different 31P NMR chemical
shifts were found respectively at δP Ϫ22.5 and Ϫ16.5 ppm for
G1-16ethoxyPPh2 and G1-16propylPPh2, the selectivities to
nonan-1-al were 78.1 and 78.0% (Table 6, Entries 21 and 22).
G1-16ethoxyPPh2 led to a higher linear to branched ratio
than G1-16propylPPh2 (l : b = 6.4 : 1 compared to 5.0 : 1), due
to the presence of larger amounts of isomerised alkenes (2-, 3-
and 4-octene: 8.3% for G1-16ethoxyPPh2, 5.1% for G1-
16propylPPh2) in the final product solution. Also a slightly
higher rate was observed for G1-16ethoxyPPh2 (2.0 × 10Ϫ3
Characterization of catalysts
High pressure 31P NMR studies. The generally high l : b ratios
obtained using G1-16ethylPPh2 suggest that strong bidendate
coordination occurred1 or that the high local concentration of
phosphorus atoms on the surface of the dendrimer increased
the concentration of complexes containing three P donors.37–39
The mechanism of reaction for these two hypotheses would
therefore be similar to the one proposed by Wilkinson and
co-workers.37,38 Another possibility for such selectivity could be
the formation of bimetallic species as reported by Stanley and
co-workers.40,41 We have attempted to obtain information on the
species present in solution by using 31P NMR spectroscopy,
molecular mechanics simulations and HP IR spectroscopy.
As discussed previously the dendritic species G1-16ethylPPh2
seems to chelate strongly to the rhodium since the selectivity
was only slightly influenced by a phosphine : rhodium ratio
varying from 3.6 : 1 to 12 : 1. Further confirmation that the
metal was strongly bound to the dendrimer came from 31P
NMR studies of solutions prepared from [Rh(acac)(CO)2] and
G1-16ethylPPh2 (1 : 3) under CO and H2 at atmospheric
pressure. Two resonances were observed at room temperature,
one (δ 37 ppm) for phosphine species bound to rhodium
complexes and the other from the free phosphorus atoms
(G1-16ethylPPh2 resonates at δ Ϫ9.5 ppm). Both resonances
were very broad (width at half maximum = 470 and 88 Hz,
respectively). This is possibly because of different binding
environments for the rhodium since the spectrum is the same
in different solvents (CH2Cl2, THF) and the linewidth at half
maximum of the signal from the unbound P atoms only
changes slightly between Ϫ40 (63 Hz) and ϩ 60 ЊC (150 Hz).
This shows that the rhodium was not migrating rapidly around
the surface of the dendrimer, nor dissociating on the NMR
timescale, although the broadening of the signal on heating
may suggest some slow exchange. The resonance from the
Rh-bound phosphines appeared as two broad overlapping
doublets (δ 37 and 36 ppm) at Ϫ40 ЊC, but as a single broad
doublet (δ 37 ppm, JP–Rh ≈ 130 Hz) at ϩ60 ЊC, suggesting
fluxionality within the bound complex, or that two coordin-
ation environments are present at low temperature. Two broad
hydride signals were observed at δ Ϫ10.5 ppm (width at half
maximum = 80 Hz) and Ϫ9.5 ppm (ratio 5: 1) in the 1H NMR
spectrum at room temperature. It is difficult to assign such
signals to any configuration of rhodium complexes since the
broad signal could arise from the presence of several different
instead of 1.5 × 10Ϫ3 Ϫ1). Nevertheless the rates of reaction
s
using these two ligands were higher than those obtained using
the more selective G1-16ethylPPh2 dendrimer, data which are
consistent with a weaker phosphine to rhodium interaction
and so lower regioselectivity. It is thus believed that these den-
drimers do not constrain the phosphine end groups to optimum
coordination to rhodium due to the longer chain length (spacer
of seven atoms). BISBI and Xantphos do have longer spacer
lengths, but here the backbone is heavily constrained.
Using the second generation dendritic ligand with the
optimum –(CH2)2Si(CH2)2– spacer between the P atoms, G2-
propyl-48ethylPPh2 (only functionalised to 85%) rather similar
results in terms of selectivity were obtained to those using the
G1-16ethylPPh2 dendritic ligand. Indeed, using G2-propyl-
48ethylPPh2 as ligand (Table 6, Entry 23), the regioselectivity to
the linear aldehyde, nonan-1-al, was 83.8% with a l : b ratio of
11.5 : 1 whilst the 16 arm ligand gave a selectivity of 86.0% and
a l : b ratio of 13.9 : 1 (Table 2, Entry 6). A drop in the rate of
reaction by a factor ca. 2 was nevertheless apparent (Table 6,
Entry 23), perhaps because the steric hindrance of such bulky
ligands reduced the accessibility of the rhodium centre. Steric
bulk might also explain the slight increase in isomerisation
products for the larger dendrimer-based ligand (7.5 instead of
6.6% with G1-16ethylPPh2, B), since it was also found that the
4328
J. Chem. Soc., Dalton Trans., 2002, 4323–4334

View Article Online
species although a comparison with spectra obtained for
PEtPh2 complexes of rhodium is instructive.
In HP NMR and HP IR studies of [Rh2(µ-OMe)2(cod)2] and
PEtPh2 under CO and H2 pressure42,43 Freixa et al. assigned the
31P NMR resonance at δP 31.2 ppm (doublet) to the diphos-
phine ligand in a trigonal bipyramidal geometry (equatorial–
equatorial (C) and equatorial–axial (D) isomers, Fig. 4), while
Fig. 4 Equatorial–equatorial (ee)
equatorial–axial (ea) equilibrium
for trigonal bipyramidal hydridorhodium complexes together with the
proposed tris(phosphine) and dimeric rhodium complexes.
1
the hydride region of the H NMR showed a signal (td) at δH
Ϫ9.3 ppm (P : Rh = 2 : 1 to 4 : 1, 30 bar H2/CO).42 The phos-
phorus atoms of the dimeric rhodium species (F in Fig. 4) res-
onated (31P NMR) between δP 20.3 and 14.0 ppm. In another
study (P : Rh = 4 : 1, 20 bar H2, 4 bar CO, [Rh] = 0.024 mol
dmϪ3), Claver, van Leeuwen and co-workers43 assigned a new
31P NMR resonance at δP 34.5 ppm to the tris(phosphine)
rhodium hydride (E, Fig. 4) with a 1H NMR resonance for the
hydride at δH Ϫ10.1 ppm. The extrapolation of these previous
studies to this dendritic system would suggest that the complex
did not form a dinuclear species (F in Fig. 4) under an atmos-
phere of H2/CO, a configuration expected under low pressure
of H2.1 It would be also tempting to assign the broad 31P and 1H
NMR signals (at 37 and Ϫ10.5 ppm respectively) to E (Fig. 4)
and the weak resonance in the 1H NMR spectrum at Ϫ9.5 ppm
to the equatorial and axial isomers of RhH(CO)2(P–P), C and
D (Fig. 4) where P–P is the dendrimer acting as a bidentate
ligand. If this is the case, the 31P NMR resonances of the bound
phosphines in C/D and in E must overlap; we note that for
PEtPh2, they only differ by 3.3 ppm. The different environment
of the phosphorus atoms and the number of combinations
(between arms) would give a multitude of disturbed geometries
to the rhodium/phosphine complexes, leading to the observed
broad resonances. However, since a low pressure of CO/H2
(ca. 1 atmosphere) and a different concentration of rhodium
were used, it is difficult to extrapolate the results to the species
present under catalytic conditions. In addition, comparison of
the integration of the 31P NMR signal of the free phosphines
and coordinated phosphines suggests that only two phosphine
species were coordinated to the metal centre (P : Rh = 4 : 1 in
solution).
Because the infrared timescale allows the observation of
the CO stretching vibrations of the two possible catalyst
precursors, the ee and ea isomers of [RhH(CO)2(P)2], we carried
out HP IR studies on the dendrimer bound rhodium complexes.
Since the rhodium–hydride bands are usually weak and hidden
behind the CO absorption bands,1 four absorption bands
should be observed if both isomers are present. However, if
a tris(phosphine) rhodium complex is formed only one or
two bands are expected.43 The HPIR spectroscopy study was
carried out using [Rh(CO)2(acac)] (0.02 mol dmϪ3) and G1-
16ethylPPh2 (P : Rh = 4 : 1) in toluene (10 cm3) at 20 bar of
H2/CO at various temperature (25 to 100 ЊC) (Fig. 5 and Table 7).
Interestingly at 25 ЊC three absorption bands at 2030, 1979 and
1954 cmϪ1 and a shoulder around 1940 cmϪ1 were visible
indicating that the ee and ea isomers were already present when
one would normally expect the formation of a dimeric
species.42,43 The comparative positions for the ea and ee isomers
of [RhH(CO)(PEtPh2)2] are 2037, 1990, 1979 and 1947 cmϪ1, so
Fig. 5 HP IR of the rhodium/G1-16ethylPPh2 complexes under 20 bar
of H2/CO at 25–100 ЊC.
Table 7 Selected IR absorption bands of rhodium/G1-16ethylPPh2
complexes under 20 bar of H2/CO
T /ЊC
νCO (ee)/cmϪ1
νCO (ea)/cmϪ1
νCO dimer/cmϪ1
25
40
60
80
100
2030, 1979
2035, 1978
2029, 1979
2032, 1978
2029, 1978
Hidden, 1940
1990, 1937
1990, 1939
1990, 1939
1990, 1942
1954, 1723
1725
1725
1727
1728
J. Chem. Soc., Dalton Trans., 2002, 4323–4334
4329

View Article Online
the bands are assigned as in Table 7, which also shows the effect
of temperature on the various bands. At 25 ЊC there are also
weak absorptions near 1760 and 1725 cmϪ1, which are assigned
to the bridging carbonyls of the dimer (F, Fig. 4) which will also
have a band near 1950 cmϪ1. The absorptions assigned to the
dimer become less significant on heating and there is little
evidence for the presence of the tris-phosphine species under
these high pressure conditions. By comparison with PEtPh2, a
single peak near 1978 cmϪ1 would be expected for this complex.
The formation of a dicationic bimetallic RhII species as
described by Stanley and co-workers can be ruled out since
higher energy absorption bands would be expected (2095 to
2058 cmϪ1).41 On heating, the changes shown in Fig. 5 and Table
7 occur, suggesting that the dimer becomes less significant.
van Leeuwen and co-workers have investigated44 the HP IR
spectra of a number of thiaxantphos (P–P) complexes with
different electron withdrawing and donating groups, which have
different ratios of the ee and ea isomers of [RhH(CO)(P–P)2]
and have published the spectra at 80 ЊC and 20 bar. Although
our spectra are broader than his and there is overlap of the two
bands between 1975 and 1990 cmϪ1, a visual comparison of the
relative intensities of the bands at 2030 (νsym for the ee isomer)
and 1939 cmϪ1 (νasym for the ea isomer) shows that the spectrum
is most similar to that of the hydridorhodium complexes con-
taining thiaxantphos with p-Cl or p-F substituents on each of
the phenyl groups attached to P, for which 75–90% of the
complexes are present as the ee isomer, as determined from
The results of the molecular dynamics simulations indicate
that even the 1st generation G1-16ethylPPh2 dendrimer is rela-
tively globular in nature. Almost all of the diphenylphosphine
groups are located on the external surface of the molecule,
where they are available for binding to Rh. However, the bulki-
ness of such a species is evident and the high concentration of
phosphine molecules at the surface explains the difficulty in
achieving complete conversion of all arms of G1-16ethylPPh2
and especially the G1-24ethylPPh2.
The distribution of the P–P distances in the free ligands was
then calculated from the final frames of the simulation for
BISBI, Xantphos, and G1-16ethylPPh2 and are shown in Fig. 7.
Fig. 7 P–P distances in different bidentate phosphorus ligands.
BISBI and Xantphos have P–P distances centered around 4.5
and 4.8 Å respectively, the molecular modelling of the den-
drimer showed that within one arm, the P atoms were separated
by 4–7 Å, whilst between arms there were always some dis-
tances in the 5–10 Å region. Rh–P distances are of the order of
2.5 Å.
The high regioselectivity observed using BISBI or Xantphos
type ligands has been attributed to their ability to coordinate
rhodium so as to give predominantly equatorial–equatorial (ee)
coordination in key trigonal bipyramidal intermediates. Since
there are many P–P separations for the dendrimer that are very
similar to the P–P separations in BISBI and Xantphos (Fig. 7),
it seems plausible that the crowding of the dendrimer surface
gives a natural bite angle for many of the possible diphosphines
formed very close to that required for ee binding and hence
highly regioselective hydroformylation.
Models of the small molecules showed that the lowest energy
configurations had the phosphine groups as far away from each
other as possible. Assuming that the selectivity of the reaction
depends to some degree on the amount of steric crowding at the
surface of the dendrimer, one might expect that the compound
tetra(diphenylphosphinoethyl)silane for which an X-ray crystal
structure showed that the P atoms are 6.94 and 8.33 Å
apart (Fig. 8), might show intermediate behaviour between
1
coupling in the hydride resonance of the H NMR spectrum.
This suggests that there is also a similar preference for ee
binding by the dendrimer, which could help to account for
the high selectivity to linear aldehydes in hydroformylation
reactions. However, in the same study of thiaxantphos ligands,
van Leeuwen and co-workers have shown that the ee : ea ratio
does not entirely determine the hydroformylation selectivity.
The natural bite angle seems to be more important and this has
been attributed to the ability of the ligand to span transoid
positions of the basal plane of the square pyramidal inter-
mediate formed during alkene coordination.44
Molecular dynamics simulations of the dendrimers
In order to try to investigate further whether the higher l : b
ratios observed with the dendrimer bound catalysts might in
part be attributed to the natural bite angle, we carried out
some molecular dynamics simulations of the G1-16ethylPPh2
dendrimer (Si8O12[CH2CH2SiMe(CH2CH2PPh2)]8) using the
Discover program contained in the Materials Studio Suite of
Molecular Simulations Inc.45 A molecular model from this
work is shown in Fig. 6.
Fig. 8 X-Ray crystal structure of Si(CH2CH2PPh2)4.
G1-16ethylPPh2 and diphenylphosphinodimethylsilane. Indeed
this was confirmed since a l : b ratio of 5.2 : 1 was found.
It is important to remember that simulations of this kind are
Fig. 6 Molecular model of the G1-16ethylPPh2 dendrimer.
4330
J. Chem. Soc., Dalton Trans., 2002, 4323–4334

View Article Online
very simplistic, and tell us nothing about the catalytic
species. All they tell us is that the G1-16ethylPPh2 den-
drimer has possible configurations that lead to similar P–P
distance distributions to other highly selective ligands. More
advanced calculations, which are difficult on such large
molecules, are needed if we are to elucidate details of the
Rh-bound species. Molecular modelling studies of the other
dendrimers are currently being carried out and will be reported
separately.
because they promote ee rather than ea binding of the bidentate
ligand in key five-coordinate intermediates.
Experimental
Microanalyses were carried out by the University of St.
Andrews Microanalysis service on a Carlo Erba 1110 CHNS
analyser. NMR spectra were recorded on a Bruker AM 300 or a
Varian 300 NMR spectrometer. The 1H and 13C NMR spectra
were recorded with reference to tetramethylsilane (external).
31P NMR spectra were referenced externally to 85% H3PO4.
13C and 31P NMR spectra were recorded with broad band
proton decoupling. Infrared spectra were recorded using either
a Perkin-Elmer 1710 or a Nicolet 460 Protege FTIR spec-
trometer. The high pressure infrared studies were carried out in
a cylindrical internal reflectance cell. GC analyses were carried
out using a Phillips PU4000 fitted with a capilliary column with
nitrogen as the carrier gas. GCMS analyses were carried out
using a Hewlett Packard 5890 GC interfaced with an Incos
quadrupole mass spectrometer fitted with a SGE BP1 column
or using a Hewlett Packard HP6890 GC with a 5973 mass
selective detector fitted with a 5% phenyl methyl siloxane
capilliary column. Matrix assisted laser desorption/ionisation
(MALDI) mass spectra were obtained using a micromass
TOF Spec 2E mass spectrometer system equipped with a 337
nm N2 laser operating in positive ion reflection mode. Samples
were prepared by addition of the matrix (α-cyano-4-
hydroxycinnamic acid, 2.6-dihydroxyacetophenone, or 2,5-
dihydroxybenzoic acid). The mixtures were then dissolved in a
suitable solvent (THF or CH2Cl2) before being transferred to
the sample holder and dried. All mass measurements refer to
peaks for the most common isotopes (1H, 12C, 16O, 29Si, 31P).
NMR studies were carried out using solutions containing
[Rh(acac)(CO)2] (0.06 mol dmϪ3) and the dendrimer (P : Rh =
4 : 1).46 High pressure IR experiments were carried out as
previously described using solutions containing [Rh(acac)-
(CO)2] (0.02 mol dmϪ3) and the dendrimer (P : Rh = 4 : 1).46
All manipulations were carried out under dry, deoxygenated
(CrII on silica) argon, using standard Schlenk line and catheter
tubing techniques. Solvents were degassed before use having
been dried by distillation from sodium diphenyl ketyl (THF,
diethyl ether, light petroleum (boiling range 40–60 ЊC)), sodium
(toluene, cyclohexane), CaH2 (CH2Cl2) or magnesium alkoxide
(methanol, ethanol). Water was distilled and stored under
argon. Deuteriated solvents (Cambridge Isotope Laboratories)
were degassed by repeated freeze–pump–thaw cycles and
stored under argon over molecular sieves. Compounds were
purchased from Aldrich, Acros or Strem and were used without
further purification. G0-8vinyl,26 G1-16vinyl,17 G1-24Cl,26
G1-24vinyl26 and G2-propyl-48vinyl17 were prepared by
standard literature methods.
Molecular modelling was carried out using the Discover
program contained in the Materials Studio Suite of Accelrys
Inc.45 Molecular dynamics were performed using the con-
sistent valence force field (CVFF). The dendrimers were drawn
using the draw facility on the Materials Studio viewer and
minimized using steepest descent/conjugate gradient methods
to a maximum energy derivative of less than RMS 0.001 kcal
molϪ1 ÅϪ1, relaxing bond lengths to their equilibrium distances.
The models were then subjected to the heating and annealing
process designed to reach a global minimum of energy for the
structure. The structures were first heated to 1000 K to facilitate
full expansion of the branches with a time step of 1 fs, and then
annealed by reducing the temperature in 50 K steps for 5 ps
dynamics down to 50 K. At 50 K the structures were again
minimized to a maximum energy derivative of less than 0.001
kcal molϪ1 ÅϪ1 and allowed to equilibrate at 273 K for 600 ps.
Every 1000th configuration of the structure was saved in full for
analysis. The final 500 ps (500 frames) of the equilibration
trajectory was used for analysis.
Conclusion
POSS-based dendritic ligands functionalized with diphenyl-
phosphine (G1-16ethylPPh2 and G2-propyl-48ethylPPh2)
lead to high selectivity to the linear aldehyde (86%, l : b
ratios up to 14 : 1) during the hydroformylation of terminal
alkenes. Small molecule analogues (1, 2, and 3) did not show
such selectivity indicating that a ‘positive dendritic effect’ is
occurring.
It also appeared that only one dendritic framework (spacer
of five atoms (CH2CH2SiCH2CH2) between the P atoms) led to
high selectivity. It is tentatively concluded that other structures
were either too compact (G1-16methylPPh2, three atom
spacers) or not sufficiently constrained (G1-16ethoxyPPh2 and
G1-16propylPPh2, seven atom spacers) to give a high regio-
selectivity to the linear aldehyde. The seven atom spacers gave
better selectivity than the three atom spacers. Interestingly
the 2nd generation dendrimer, G2-propyl-48ethylPPh2, with a
similar structure to that of the G1-16ethylPPh2 (five atoms
between the P atoms) maintained a high selectivity (83.8%, l : b
= 11.5) although leading to lower reactivity (rate ca. half ). The
composition of the dendritic framework was also important
since, when replacing a carbon atom by an oxygen atom β to
the phosphorus atoms (from G1-16ethylPPh2 to G1-16meth-
oxyPPh2) the selectivity dropped. It is believed that the low
selectivity could arise from a different geometry of the
rhodium/dendritic ligand. Indeed an electronic effect is unlikely
in this case since no such effect was found in the dendrimers
with a seven atom spacer (G1-16ethoxyPPh2 and G1-16propyl-
PPh2). The functionalization of the dendritic ligand with
electron withdrawing groups (G1-16ethylPAr2, Ar = 3,5-
(C6H3(CF3)2) led to lower selectivity.
The characterization of the rhodium complexes present in
solution under H2 and CO pressure was carried out using
1H and 31P NMR and HP IR techniques. It appeared that the
species formed under H2/CO was mainly the five-coordinated
hydridorhodium complex (trigonal bipyramidal structure) with
the ea and ee isomers in equilibrium, with the ee isomer
predominating. The presence of the ea coordination may help
to explain the low selectivity obtained when the POSS den-
drimer was functionalized with more electron withdrawing
arylphosphine species (G1-16ethylPAr2), since Casey and co-
workers have shown that electron withdrawing substituents
increase the selectivity if they are on an equatorial P atom, but
decrease it if they are on an axial P atom.27 It may also be that
the increased steric crowding provided by the presence of
the fluorinated substituents prevents the
P atoms from
attaining the optimum relative disposition for obtaining high
selectivity.
The presence of a tris(phosphine) rhodium complex was not
completely ruled out for G1-16ethylPPh2, but no strong evi-
dence for the formation of active dimeric species could be
found at temperatures above 25 ЊC. It is therefore believed that
the high regioselectivity obtained with the G1-16ethylPPh2
ligand is mainly due to the formation of a constrained bidentate
ligand as in BISBI or Xantphos ligands. Molecular modelling
studies tend to support this conclusion, since they show that
many of the P–P distances in the G1-16ethylPPh2 dendrimer,
but not in the small molecules, are very similar to those found
for BISBI and Xantphos, ligands that give high regioselectivity
J. Chem. Soc., Dalton Trans., 2002, 4323–4334
4331

View Article Online
Syntheses
C–P), 137.0 (CH2, Si-vinyl), 132.8 (d, JC–P = 18.8 Hz, C ortho),
132.70 (d, JC–P = 18.0 Hz, C ortho), 128.50 (s, C para), 128.45 (d,
JC–P = 6.7 Hz, C meta), 21.50 (d, 1JC–P = 14.2 Hz, CH2P), 18.52,
16.89, 9.07 (d, JC–P = 12.0 Hz, SiCH2CH2P), 4.57 (br, O3Si-
CH2CH2), Ϫ5.38 (SiCH3).
1,3,5,7,11,13,15-Octakis{2-[bis(2-diphenylphosphinoethyl)-
methylsilyl]ethyl}pentacyclo[9.5.1.13,9.15,15.17,13]octasiloxane
(G1-16ethylPPh2). G1-16vinyl (0.25 g, 0.176 mmol) was added
to a dry 20 cm3 round bottomed Schlenk flask. AIBN (0.0078 g)
was added and the flask was charged with cyclohexane (5 cm3)
and diphenylphosphine (2.1 g, 11.3 mmol). The flask was sealed
and heated to 60 ЊC for 10 days. The resulting solution was
allowed to cool and concentrated in vacuo. The excess phos-
phine was removed by vacuum distillation (120 ЊC, 0.1 mmHg)
(Yield 0.73 g, 95%) or the product was loaded into a silica gel
column (eluent: gradient of petroleum/diethyl ether) (Yield
0.66 g, 85%). The resulting crude product was a colourless low
melting point solid (conversion >99%, compound B). MALDI-
TOF: multiplets (mass increment of 16, oxidation) centered
at m/z 4582 (large multiplet); 4299 (major multiplet); 4113
(medium multiplet); 3830 (br, small); 3628 (br, very small)
corresponding respectively to 16, 15, 14, 13, 12 substituted arms
([M Ϫ n{PPh2}], n = 0, 1, 2, 3, 4) (m/z expected 4397.9). The
peaks at m/z 4582, 4299 and 4113 correspond to the molecule
with 11, 5 and 0 phosphine species oxidised, respectively. The
MALDI-TOF spectrum of the partially substituted compound
A gave a broad signal centered at m/z 3680 (ca. 12 arms substi-
tuted). Microanalysis found for compound B: C, 67.9; H, 6.4.
C248H280O12P16Si16 requires C, 67.7; H, 6.4%. 31P-{1H} NMR
(CDCl3) δP (ppm) Ϫ9.4, Ϫ9.5, Ϫ14.4 (weak). 1H NMR (CDCl3)
δH (ppm) 7.40–7.20 (br m, 160 H, P(C6H5)2), 1.86 (br m, 32 H,
PCH2), 0.64–0.32 (m, 64 H, Si–CH2), Ϫ0.16 (br, 24 H, Si–CH3).
13C-{1H} NMR (CDCl3): δC (ppm) 139.27 (d, JC–P = 14.7 Hz,
2
1,3,5,7,11,13,15-Octakis{2-{bis{bis-2-[bis(3,5-trifluoromethyl)-
phenyl]phosphinoethyl}methylsilyl}ethyl}pentacyclo[9.5.1.1.13,9.-
15,15.17,13]octasiloxane (G1-16ethylPAr2). G1-16vinyl (0.25 g,
0.176 mmol) was added to a dry 20 cm3 round bottomed
Schlenk flask. AIBN (0.0078 g) was added and the flask was
charged with cyclohexane (5 cm3) and [3,5-(CF3)2C6H3]2PH27
(3.87 g, 8.45 mmol). The flask was sealed and heated to 60 ЊC
for 30 days. The resulting solution was allowed to cool and the
excess phosphine was removed by vacuum distillation (140 ЊC,
0.02 mmHg) to give a colourless solid (1.17 g, yield 96% for a
conversion of 75%). MALDI-TOF: m/z: 6919, 6459, 6001
(major), 5543 (major), 4988 (respectively 12, 11, 10, 9, 8 arms
substituted). Other peaks at m/z 6365, 5905, 5446, 4891 (frag-
mentation). 31P-{1H} NMR (CDCl3) δP (ppm) Ϫ6.6, Ϫ6.9,
1
Ϫ7.0, Ϫ12.9. H NMR (CDCl3) δH (ppm) 8.2–7.8 (m, 72 H,
P(C H (CF ) ) ), 6.0 (br, 24 H, CH᎐CH ), 5.7 (br, 12 H, CH᎐
᎐
᎐
6
3
3
2
2
2
CH2), 2.1 (br, 24 H, PCH2), 1.0–0.4 (br m, 56 H, Si–CH2),
0.6 (br, 24 H, Si–CH3). 13C-{1H} NMR (CDCl3): δC (ppm)
140.60 (m, C quaternary), 135.75, 133.80, 132.73, 132.50,
125.00, 124.66, 123.63, 121.39, 110.16, 21.34 (SiCH2CH2P),
8.30 (SiCH2CH2P), 5.20 (O3SiCH2CH2), 4.33 (O3SiCH2CH2),
Ϫ6.08 (SiCH3).
Bis(diphenylphosphinoethyl)dimethylsilane. Dimethyldivinyl-
silane (0.267 g, 2.39 mmol) was added to a dry 20 cm3 round
bottomed Schlenk flask. AIBN (0.0078 g) was added and the
flask was charged with toluene (5 cm3) and diphenylphosphine
(2.675 g, 14.3 mmol). The flask was sealed and heated to 60 ЊC
for 4 hours. The resulting solution was allowed to cool and
taken to dryness in vacuo. The excess phosphine was removed
by vacuum distillation (120 ЊC, 1.0 mmHg). The product was
recrystallised from petroleum to give a white crystalline solid
(1.06 g, 92%). Microanalysis found C, 74.0; H, 7.0. C30H34P2Si
requires C, 74.4; H, 7.1%. 31P-{1H} NMR (CDCl3) δP (ppm)
Ϫ9.4. 1H NMR (CDCl3) δH (ppm) 7.42–7.30 (m, 20 H,
P(C6H5)2), 1.94 (m, 4 H, PCH2), 0.60 (m, 4 H, Si–CH2), Ϫ0.16
(br, 6 H, Si–CH3). 13C-{1H} NMR (CDCl3): δC (ppm) 138.95 (d,
JC–P = 17.6 Hz, C–P), 132.90 (d, JC–P = 17.5 Hz, C ortho), 128.70
P(C6H5), C–P), 139.18 (d, JC–P = 14.7 Hz, C–P), 132.80 (d, JC–P
=
18.8 Hz, C ortho), 132.70 (d, JC–P = 17.4 Hz, C ortho), 128.55 (s,
1
1
C para), 128.45 (d, JC–P = 6.71 Hz, C meta), 21.40 (d, JC–P
=
2
14.8 Hz, CH2P), 8.30 (d, JC–P = 12.1 Hz, SiCH2CH2P), 4.78
(O3SiCH2CH2), 4.24 (O3SiCH2CH2), Ϫ6.08 (SiCH3).
1,3,5,7,11,13,15-Octakis{2-[tris(2-diphenylphosphinoethyl)-
silyl]ethyl}pentacyclo[9.5.1.13,9.15,15.17,13]octasiloxane
(G1-
24ethylPPh2). G1-24vinyl (0.245 g, 0.162 mmol) was added to a
dry 20 cm3 round bottomed Schlenk flask. AIBN (0.0078 g) was
added and the flask was charged with cyclohexane (7 cm3) and
diphenylphosphine (2.90 g, 0.0156 mol). The flask was sealed
and heated to 60 ЊC for 10 days. The resulting solution was
allowed to cool and the excess phosphine was removed by
vacuum distillation (120 ЊC, 0.1 mmHg). The resulting crude
product was a white solid (Yield 0.65g, 95% for a conversion of
1
1
(d, JC–P = 6.71 Hz, C meta), 128.46 (C para), 21.62 (d, JC–P
=
13.4 Hz, CH2P), 10.40 (d, 2JC–P = 9.40 Hz, SiCH2CH2P), Ϫ3.78
1
60%). 31P-{1H} NMR (CD2Cl2) δP (ppm) Ϫ9.5 (br). H NMR
(SiCH3).
(CDCl3) δH (ppm) 7.7–7.2 (br m, 140 H, P(C6H5)2), 6.1–5.8 (br,
20 H, CH᎐CH ), 5.7–5.5 (br, 10 H, CH᎐CH ), 2.2–1.8 (br, 28 H,
᎐
᎐
2
2
(ref.
Tetrakis(2-diphenylphosphinoethyl)silane
33). Tetravinyl-
PCH2), 0.90–0.25 (br, 60 H, SiCH2). IR/cmϪ1 (KBr disc) 2956s,
2919s, 2873s, 1455vs, 1409vs (PCH2), 1260vs (SiCH2), 1120vs
(SiCH2CH2Si), 1040vs (SiOSi), 952s, 800m, 750vs, 707vs.
silane (0.20 g, 1.47 mmol) was added to a dry 20 cm3 round
bottomed Schlenk flask. AIBN (0.0078 g) was added and the
flask was charged with toluene (5 cm3) and diphenylphosphine
(2.73 g, 14.7 mmol). The flask was sealed and heated to 60 ЊC
for 4 hours. The resulting solution was allowed to cool and
taken to dryness in vacuo. The excess phosphine was removed
by vacuum distillation (120 ЊC, 1.0 mmHg). The product was
recrystallised from hot dichloromethane to give a white crystal-
line solid (1.20 g, 93%). Microanalysis found C, 76.0; H, 6.10.
C56H56P4Si requires C, 76.3; H, 6.4%. 31P-{1H} NMR (CDCl3)
δP (ppm) Ϫ9.48. 1H NMR (CDCl3) δH (ppm) 7.33 (br m, 40 H,
P(C6H5)2), 1.79 (br m, 8l H, PCH2), 0.61 (br m, 8 H, Si–CH2).
13C-{1H} NMR (CDCl3): δC (ppm) 138.95 (d, JC–P = 17.6 Hz,
P(C6H5), C–P), 132.90 (d, JC–P = 17.5 Hz, C ortho), 128.70 (d,
1JC–P = 6.7 Hz, C meta), 128.46 (s, C para), 21.62 (d, 1JC–P = 13.4
Hz, CH2P), 10.40 (d, 2JC–P = 9.4 Hz, SiCH2CH2P).
1,3,5,7,11,13,15-Octakis{2-{tris{3-[bis(2-diphenylphosphino-
ethyl)methylsilyl]propyl}silyl}ethyl}pentacyclo[9.5.1.13,9.15,15.-
1
7,13]octasiloxane (G2-propyl-48ethylPPh2). G2-propyl-48vinyl
(0.22 g, 5.2 × 10Ϫ5 mol) was added to a dry 20 cm3 round
bottomed Schlenk flask. AIBN (0.008 g) was added and the
flask was charged with cyclohexane (5 cm3) and diphenylphos-
phine (1.86 g, 0.01 mol). The reaction mixture was heated to
50 ЊC for 12 days. The resulting solution was allowed to cool and
taken to dryness in vacuo. The product was loaded into a silica
gel column (eluent: gradient of petroleum/diethyl ether). The
isolated product was a colourless low melting point solid (0.443
g, yield 75% for a conversion of 84%). 31P-{1H} NMR (CDCl3)
δP (ppm) Ϫ9.9 (br). 1H NMR (CDCl3) δH (ppm) 7.6–7.0 (br m,
408 H, P(C H ) ), 6.2–5.8 (br, 14 H, CH᎐CH ), 5.65–5.45 (br, 7
᎐
6
5
2
2
H, CH᎐CH ), 1.86 (br, 82 H, PCH ), 1.62 (br, CH ), 1.45–1.00
1,3,5,7,11,13,15-Octakis{2-[tris(diphenylphosphinomethyl)-
᎐
2
2
2
(br, CH2), 1.00–0.20 (br, SiCH2), 0.20 to Ϫ0.20 (br, 72 H,
Si–CH3). 13C-{1H} NMR (CDCl3): δC (ppm) 139.27 (d, JC–P
14.7 Hz, P(C6H5), C–P), 139.18 (d, JC–P = 14.7 Hz, P(C6H5),
silyl]ethyl}pentacyclo[9.5.1.1.13,9.15,15.17,13]octasiloxane
(G1-
=
24methylPPh2). LiCH2PPh2/TMEDA30 (0.741 g, 2.22 mmol)
was dissolved at Ϫ78 ЊC in a Schlenk flask containing THF
4332
J. Chem. Soc., Dalton Trans., 2002, 4323–4334

View Article Online
(20 cm3). The solution was transferred via cannula to a Schlenk
flask containing G1-24Cl (0.15 g, 0.0874 mmol) in THF
(10 cm3). The mixture was stirred for 60 h. The solvent was
removed in vacuo. Dichloromethane was added (20 cm3). After
settling overnight, the liquid was transferred via cannula and
taken to dryness in vacuo. The product was washed with diethyl
ether (3 × 10 cm3) and dried in vacuo. The resulting product was
a white solid (0.44 g, conversion 70%, yield 93%). MALDI-
TOF: broad peak centered at m/z 4678 (ca. 17 arms substituted)
(m/z expected 5646.8). 31P-{1H} NMR (CD2Cl2) δP (ppm)
Ϫ23.9. 1H NMR (CD2Cl2) δH (ppm) 7.5–6.6 (m, 240 H, C6H5),
1.03 (s, 48 H, –CH2P), 0.9–0.4 (m, 32 H, Si–CH2CH2-Si).
13C-{1H} NMR (CD2Cl2) δC (ppm) 141.3 (br, P-C), 132.7 (br,
C ortho), 128.4 (br, C para, C meta), 11.6 (br, SiCH2P), 5.0
1,3,5,7,11,13,15-Octakis{2-[bis(2-diphenylphosphinoethoxy)-
methylsilyl]ethyl}pentacyclo[9.5.1.13,9.15,15.17,13]octasiloxane
(G1-16ethoxyPPh2). Ph2PCH2CH2OH32 (1.12 g, 4.9 mmol, 2.5
fold excess) and NEt3 (0.75 cm3) in THF (20 cm3) were added to
a Schlenk flask containing G1-16Cl (0.198 g, 0.128 mmol) in
THF (10 cm3). The mixture was stirred for 5 days. The solvent
was removed in vacuo and the solid mixture was loaded onto a
silica column where it was eluted with a gradient of petroleum
and diethyl ether. After evaporation of the solvent under
vacuum, the product was obtained as a white solid (0.48 g,
conversion >92%, yield 85%). MALDI-TOF: m/z 4657.6 (m/z
expected 4653.8), other major peaks at 4505; 4445.2 (M Ϫ
{CH᎐CHPPh }), 4233.7 (M Ϫ 2{CH᎐CHPPh }), 4156, 4047,
᎐
᎐
2
2
3943. 31P-{1H} NMR (CDCl3) δP (ppm) Ϫ22.3, Ϫ22.5.
(br, SiCH2CH2Si). IR/cmϪ1 (KBr disc) 3049vs (C᎐C), 2908m,
1H NMR (CDCl3) δH (ppm) 7.6–7.3 (m, 160 H, C6H5), 3.84
᎐
3
2
1584w (C᎐C), 1479s (C᎐C, PPh), 1433s (SiCH P), 1088vs
(br d, JP–H = 7.6 Hz, OCH2, 32 H), 2.4 (br d, JP–H = 7.7 Hz,
32 H, PCH2), 0.75–0.55 (m, 32 H, Si–CH2CH2–Si), 0.02 (br,
24 H, SiCH3). 13C-{1H} NMR (CDCl3) δC (ppm) 141.4 (br,
C–P), 132.90 (br, C ortho), 128.56 (s, C meta, C para), 60.54
(d, JC–P = 26.5 Hz, CH2P), 32.33 (d, JC–P = 12.6 Hz, OCH2),
7.97 (br, SiCH2CH2SiO3/2), 5.00 (br, CH3SiCH2), Ϫ5.37 (br,
SiCH3).
᎐
᎐
2
(SiCH2CH2Si), 1025vs (SiOSi), 999s (PPh), 740s, 695s,
471m.
1,3,5,7,11,13,15-Octakis{2-[bis(diphenylphosphinomethyl)-
methylsilyl]ethyl}pentacyclo[9.5.1.13,9.15,15.17,13]octasiloxane
(G1-16methylPPh2). LiCH2PPh2/TMEDA47 (0.741 g, 2.22
mmol) (50% excess) was dissolved at Ϫ78 ЊC in a Schlenk flask
containing THF (20 cm3). The solution was transferred via can-
nula to a Schlenk flask containing G1-16Cl17 (0.145 g, 0.0925
mmol) in THF (10 cm3). The mixture was stirred for 4 days. The
solvent was removed in vacuo and the solid mixture was loaded
onto silica gel and eluted with a gradient of petroleum and
diethyl ether. After evaporation under vacuum of the solvent,
the product was obtained as a white solid (0. 262 g, conversion
>95%, yield 70%). MALDI-TOF: m/z 4176 (m/z expected
4173.5), other peak at 4194 (1 oxide), 4036.5, 3992.3 (M Ϫ
CH2PPh2), 3736.0. 31P-{1H} NMR (CDCl3) δP (ppm) Ϫ22.6,
Ϫ22.7. 13C-{1H} NMR (CDCl3) δC (ppm) 141.50 (d, JC–P = 12.7
Hz, C–P), 141.33 (d, JC–P = 12.7 Hz, C–P), 133.03 (d, JC–P = 19.6
Hz, C ortho), 132.78 (d, JC–P = 17.2 Hz, C ortho), 128.56 (s,
C meta, C para), 12.80 (d, JC–P = 28 Hz, SiCH2P), 7.97 (br,
1,3,5,7,11,13,15-Octakis{2-[bis(2-diphenylphosphinopropyl)-
methylsilyl]ethyl}pentacyclo[9.5.1.13,9.15,15.17,13]octasiloxane
(G1-16propylPPh2). Magnesium turnings (0.50 g, 20.7 mmol)
and Ph2PCH2CH2CH2Cl33 (4.982 g, 18.96 mmol) were mixed in
THF (20 cm3). The mixture was heated to reflux and initiated
with a small amount of reacting BrCH2CH2Br/Mg.33 Heating
and stirring were continued for 1 h. The mixture was filtered
into a Schlenk flask containing G1-16Cl (0.615 g, 0.396 mmol)
in THF (20 cm3). The mixture was stirred for 4 days. The mix-
ture was added to an aqueous solution of NH4Cl (0.1 mol
dmϪ3). The organic phase was concentrated in vacuo and the
residue was loaded onto silica gel where it was eluted with a
gradient of petroleum and diethyl ether. After evaporation of
the solvent under vacuum, the product was obtained as a
non-crystalline solid (1.30 g, conversion >85% (1H NMR), yield
75%). MALDI-TOF: m/z 3574.9 (unknown, fragmentation)
(m/z expected 4623), other major peaks (unknown fragment-
ation) at 3426, 3365, 3153. Microanalysis found C, 63.3; H, 7.2.
C264H312O12P16Si16 requires C, 68.6; H, 6.8%. 31P-{1H} NMR
(CDCl3) δP (ppm) Ϫ16.1, Ϫ16.2, Ϫ16.4. 1H NMR (CDCl3)
1
SiCH2CH2SiO3/2), 5.00 (br, CH3SiCH2), Ϫ2.67 (br, SiCH3). H
NMR (CDCl3) δH (ppm) 7.4–7.0 (m, 160 H, C6H5), 1.21 (br, 32
H, –CH2P), 0.75–0.45 (m, 32 H, Si–CH2CH2–Si), Ϫ0.36 (br, 24
H, SiCH3). IR/cmϪ1 (KBr disc) 3049vs (C᎐C), 2908m, 1584w
᎐
(C᎐C), 1479s (C᎐C, PPh), 1433s (SiCH P), 1088vs (SiCH -
᎐
᎐
2
2
CH2Si), 1025vs (SiOSi), 999s (PPh), 740s, 695s, 471m.
3
δH (ppm) 7.6–7.3 (m, 160 H, C6H5), 3.84 (br d, JP–H = 7.6 Hz,
OCH2, 32 H), 2.4 (br d, 2JP–H = 7.7 Hz, 32 H, PCH2), 0.75–0.55
1,3,5,7,11,13,15-Octakis{2-[bis(diphenylphosphinomethoxy)-
methylsilyl]ethyl}pentacyclo[9.5.1.13,9.15,15.17,13]octasiloxane
(G1-16methoxyPPh2). HPPh2 (1.2 cm3, 9.96 mmol) and para-
formaldehyde (0.2088 g) were heated in a Schlenk tube for 90
min at 120 ЊC to afford Ph2PCH2OH.31 The compound formed
in situ and NEt3 (1.4 cm3) in THF (20 cm3) were added to a
Schlenk flask containing G1-16Cl17 (0.3223 g, 0.207 mmol) in
THF (10 cm3). The mixture was stirred for 4 days. The solvent
was removed in vacuo and the solid mixture was loaded onto a
silica column and eluted with a gradient of petroleum and
diethyl ether. After evaporation under vacuum of the solvent,
the product was obtained as a white solid (0.604 g, conversion
>86%, yield 76%). MALDI-TOF: m/z 4432.3 (m/z expected
4429.8), other major peaks at 4448 (oxide), 4248 (M Ϫ {PPh2}),
4234 (M Ϫ {CH2PPh2}), 4051 (M Ϫ 2{PPh2}), 4037 (M Ϫ
2{CH2PPh2}), 3836 (M Ϫ 3{PPh2}), 3819 (M Ϫ 3{CH2PPh2}),
3636 (M Ϫ 4{PPh2}), 3622 (M Ϫ 4{CH2PPh2}). Microanalysis
found C, 66.1; H, 6.3. C232H248O28P16Si16 requires C, 62.9; H,
1
(m, 32 H, Si–CH2CH2–Si), 0.02 (br, 24 H, SiCH3). H NMR
(CDCl3) δH (ppm) 7.55–7.30 (m, 160 H, C6H5), 2.18 (br, 32 H,
PCH2), 1.60–1.40 (br, 32 H, CH2), 0.88 (br, 32 H, SiCH2), 0.80–
0.55 (br, 32 H, SiCH2), 0.15 (br, SiCH3). 13C-{1H} NMR
(CDCl3) δC (ppm) 139.0 (br, P–C), 133.00 (d, JC–P = 18.4 Hz, C
ortho), 128.75 (s, C para), 128.70 (d, JC–P = 5.8 Hz, C meta),
32.29 (br), 20.31 (d, JC–P = 17.2 Hz, PCH2), 18.16, 4.00 (br,
SiCH2), Ϫ2.10 (s, SiCH3).
Catalytic reactions
The functionalised POSS dendrimer, the rhodium source
([Rh2(O2CMe)4] or [Rh(CO)2(acac)]) and ethanol or THF
(4 cm3) were charged into a Schlenk tube and stirred until
complexation of the rhodium with the phosphine species was
complete. The catalytic solution and substrate were then
injected into a degassed autoclave heated under CO/H2 (6 bar)
to the reaction temperature for 1 h with stirring at 1000 rpm.
The substrate was injected and the pressure of CO/H2 (1/1)
was increased to the desired pressure. The pressure was kept
constant by feeding CO/H2 through a mass flow controller from
a ballast vessel (Pressure drop in ballast vessel monitored every
5 s). After reaction, the products were analysed by GC/MS
or, for quantitative work using the same GC, but an FID
detector.
1
5.6%. 31P-{1H} NMR (CDCl3) δP (ppm) Ϫ13.6 (br). H NMR
(CDCl3) δH (ppm) 7.8–7.1 (m, 160 H, C6H5), 4.35 (br, OCH2,
32 H), 0.80–0.35 (m br, 32 H, Si–CH2CH2–Si), Ϫ0.05 (br, 24 H,
SiCH3). 13C-{1H} NMR (CDCl3) δC (ppm) 143.30 (br, phenyl
P–C), 133.24 (d, JC–P = 26.5 Hz, C ortho), 128.87 (s, C para),
128.64 (d, JC–P = 6.71 Hz, C meta), 63.17 (br, OCH2), 4.74 (br,
SiCH2CH2SiO3/2), 3.14 (br, SiCH2CH2SiO3/2), Ϫ6.15 (br,
SiCH3).
J. Chem. Soc., Dalton Trans., 2002, 4323–4334
4333

View Article Online
14 Y. H. Niu, L. E. Yeung and R. M. Crooks, J. Am. Chem. Soc., 2001,
123, 6840.
15 Q.-H. Fan, Y.-M. Chen, D.-Z. Jiang, X.-M. Chen, F. Xi and A. S. C.
Chan, Chem. Commun., 2000, 789.
16 T. Mizugaki, M. Murata, M. Ooe, K. Ebitani and K. Kaneda,
Chem. Commun., 2002, 52.
17 L. Ropartz, D. F. Foster, R. E. Morris, A. M. Z. Slawin and
D. J. Cole-Hamilton, J. Chem. Soc., Dalton Trans., 2002, 1997.
18 L. Ropartz, R. E. Morris, D. F. Foster and D. J. Cole-Hamilton,
Chem. Commun., 2001, 361.
19 L. Ropartz, R. E. Morris, G. P. Schwarz, D. F. Foster and D. J.
Cole-Hamilton, Inorg. Chem. Commun., 2000, 3, 714.
20 L. Ropartz, R. E. Morris, D. F. Foster and D. J. Cole-Hamilton,
J. Mol. Catal., 2002, 182–183, 99.
21 M. Reetz and D. Giebel, Angew. Chem., Int. Ed., 2000, 39, 2498.
22 A. J. Gong, Q. H. Fan, Y. M. Chen, H. W. Liu, C. F. Chen and F. Xi,
J. Mol. Catal. A: Chemical, 2000, 159, 225.
23 D. de Groot, P. G. Emmerink, C. Coucke, J. N. H. Reek, P. C. J.
Kamer and P. W. N. M. van Leeuwen, Inorg. Chem. Commun., 2000,
3, 711.
24 P. Arya, G. Panda, N. V. Rao, H. Alper, S. C. Bourque and
L. E. Manzer, J. Am. Chem. Soc., 2001, 123, 2889.
25 P. Arya, N. V. Rao, J. Singkhonrat, H. Alper, S. C. Bourque and
L. E. Manzer, J. Org. Chem., 2000, 65, 1881.
X-Ray crystallographic structure determination of tetrakis-
(2-diphenylphosphinoethyl)silane.33
C56H56P4Si, colourless prism, 0.18 × 0.1 × 0.1 mm, tetragonal,
space group I-4, Z = 2, a = 20.4806(14), c = 5.7429(6) Å, V =
2408.9(3) Å3, ρc 1.215 Mg mϪ3, 2θmax = 46.62Њ, graphite mono-
chromated Mo-Kα radiation (λ = 0.71073 Å), µ(Mo-Kα) =
0.218 mmϪ1, T = 293 K. Data were measured on a Bruker
SMART CCD, and Lorentzian polarization and absorption
(SADABS, max./min. transmission 1.0000/0.7616) corrections
were performed. Of 6048 measured data, 1725 were unique (Rint
= 0.1057) and 1090 observed [I > 2σ(I )]. The structure was
solved by direct methods. The non-hydrogen atoms were refined
anisotropically, the hydrogen atoms were idealized and refined
isotropically using a riding model. Structural refinements were
performed with the full-matrix least-squares method on F 2 for
all data (G. M. Sheldrick, SHELXTL version 5.10, Bruker
AXS, Madison WI, 1997) to give conventional R1 = 0.0453 and
wR2 = 0.0512 for 139 parameters with residual electron extremes
of 0.162 and Ϫ0.172 e ÅϪ3
.
CCDC reference number 189436.
See http://www.rsc.org/suppdata/dt/b2/b206597e/ for crystal-
lographic data in CIF or other electronic format.
26 P.-A. Jaffrès and R. E. Morris, J. Chem. Soc., Dalton Trans., 1998,
2767.
27 C. P. Casey, E. L. Paulsen, E. W. Beuttenmueller, B. R. Proft,
L. M. Petrovich, B. A. Matter and D. R. Powell, J. Am. Chem. Soc.,
1997, 119, 11817.
28 C. P. Casey, E. L. Paulsen, E. W. Beuttenmueller, B. R. Proft,
B. A. Matter and D. R. Powell, J. Am. Chem. Soc., 1999, 121, 63.
29 J. D. Unruh and J. R. Christenson, J. Mol. Catal., 1982, 14, 19.
30 D. J. Peterson, J. Organomet. Chem., 1967, 8, 199.
31 M. Slany, M. Bardaji, A.-M. Caminade, B. Chaudret and J.-P.
Majoral, Inorg. Chem., 1997, 36, 1939.
Acknowledgements
We thank the University of St. Andrews for a studentship
(L. R.). R. E. M. thanks the Royal Society for the provision of
a University Research Fellowship.
32 D. Chantreux, J.-P. Gamet, R. Jacquier and J. Verducci, Tetrahedron,
1984, 40, 3087.
33 E. Arpac and L. Dahlenburg, Z. Naturforsch., Teil B, 1980, 35, 146.
34 A. Buhling, P. C. J. Kamer, P. vanLeeuwen, J. W. Elgersma,
K. Goubitz and J. Fraanje, Organometallics, 1997, 16, 3027.
35 W. Schalk, P. Wirth, G. E. M. Dannenberg and V. Schmied-
Kowarzik, Ger. Pat., 1118981, 1959.
36 K. J. Haxton, L. Ropartz, D. J. Cole-Hamilton and R. E. Morris, to
be submitted.
37 D. Evans, J. A. Osborn and G. Wilkinson, J. Chem. Soc. A., 1968,
3133.
38 D. Evans, G. Yagupsky and G. Wilkinson, J. Chem. Soc. A, 1968,
2660.
39 G. Gregorio, G. Montrasi, M. Tampieri, P. Cavalieri d’Oro,
G. Pagani and A. Andreetta, Chim. Ind. (Milan), 1980, 62, 389.
40 M. E. Broussard, B. Juma, S. G. Train, W.-J. Peng, S. A. Laneman
and G. G. Stanley, Science, 1993, 260, 1784.
References
1 P. W. N. M. van Leeuwen and C. Claver, Rhodium Catalysed
Hydroformylation, Kluwer Academic Publishers, Dordrecht, 2000.
2 D. Astruc and F. Chardac, Chem. Rev., 2001, 101, 2991.
3 G. E. Oosterom, J. N. H. Reek, P. C. J. Kamer and P. W. N. M. van
Leeuwen, Angew. Chem., Int. Ed., 2001, 40, 1828.
4 N. Brinkmann, G. Giebel, G. Lohmer, M. T. Reetz and U. Kragl,
J. Catal., 1999, 183, 163.
5 D. de Groot, E. Eggeling, J. C. de Wilde, H. Kooijman, R. J. van
Haaren, A. W. van der Made, A. L. Spek, D. Vogt, J. N. H. Reek,
P. C. J. Kamer and P. W. N. M. van Leewen, Chem. Commun., 1999,
1623.
6 N. J. Hovestad, E. B. Eggeling, H. J. Heidbüchel, J. T. B. H.
Jastrzebski, U. Kragl, W. Keim, D. Vogt and G. van Koten, Angew.
Chem., Int. Ed., 1999, 38, 1655.
7 C. P. Casey and L. M. Petrovich, J. Am. Chem. Soc., 1995, 117, 6007.
8 M. Kranenburg, Y. E. M. van der Burgt, P. C. J. Kamer, P. W. N. M.
van Leeuwen, K. Goubitz and J. Fraanje, Organometallics, 1995, 14,
3081.
9 A. W. Kleij, R. A. Gossage, J. T. B. H. Jastrzebski, J. Boersma and
G. van Koten, Angew. Chem., Int. Ed., 2000, 39, 176.
10 R. Breinbauer and E. R. Jacobsen, Angew. Chem., Int. Ed., 2000, 39,
3604.
41 R. C. Matthews, D. K. Howell, W.-J. Peng, S. G. Train,
W. D. Treleavan and G. G. Stanley, Angew. Chem., Int. Ed. Engl.,
1996, 35, 2253.
42 Z. Freixa, M. Pereira, A. A. C. C. Pais and J. C. Bayón, J. Chem.
Soc., Dalton Trans., 1999, 3245.
43 I. del Rio, W. G. J. de Lange, P. W. N. M. van Leeuwen and C. Claver,
J. Chem. Soc., Dalton Trans., 2001, 1293.
44 L. A. van der Veen, M. D. K. Boele, F. R. Bregman, P. C. S. Kamer,
P. W. N. M. van Leeuwen, F. Goubitz, J. Fraanje, H. Schenk and
C. Bo, J. Am. Chem. Soc., 1998, 120, 11616.
11 C. Francavilla, M. D. Drake, F. V. Bright and M. R. Detty, J. Am.
Chem. Soc., 2001, 123, 57.
12 V. Maraval, R. Laurent, A.-M. Caminade and J.-P. Majoral,
Organometallics, 2000, 19, 4025.
13 P. Bhyrappa, J. K. Young, J. S. Moore and K. S. Suslick, J. Mol.
Catal. A: Chem., 1996, 113, 109.
45 Materials Studio Suite, Accelrys Inc., San Diego, 2000.
46 J. Rankin, A. C. Benyei, A. D. Poole and D. J. Cole-Hamilton,
J. Chem. Soc., Dalton Trans., 1999, 3771.
47 D. J. Peterson and H. R. Hays, J. Org. Chem., 1965, 30, 1939.
4334
J. Chem. Soc., Dalton Trans., 2002, 4323–4334