Investigation of rhodium catalyzed hydroformylation of ethylene in supercritical carbon dioxide by in situ FTIR spectroscopy

Article abstract of DOI:10.1016/S0040-4020(02)00246-6

The reactions of RhH(CO)L₃ (L=P(3,5-(CF₃)₂C₆H₃)₃) with CO, H₂, C₂H₄ and mixtures of these in supercritical carbon dioxide (scCO₂) were investigat

Full text of DOI:10.1016/S0040-4020(02)00246-6

TETRAHEDRON
Pergamon
Tetrahedron 58 #2002) 3929±3941
Investigation of rhodium catalyzed hydroformylation of ethylene
in supercritical carbon dioxide by in situ FTIR spectroscopy
Shaker Haji and Can Erkeyp
Environmental Engineering Program, Department of Chemical Engineering, University of Connecticut, Storrs, CT 06269-3222, USA
Received 12 November 2001; accepted 5 January 2002
AbstractÐThe reactions of RhH#CO)L3 #LˆP#3,5-#CF3)2C6H3)3) with CO, H2, C2H4 and mixtures of these in supercritical carbon dioxide
#scCO2) were investigated using high-pressure FTIR spectroscopy. The results were compared to the behavior of the conventional catalyst,
RhH#CO)#PPh3)3, in organic solvents. RhH#CO)L3 does not dissociate in scCO2 and it is converted to RhH#CO)2L2 and to [Rh#CO)2L2]2 in
the presence of CO and mainly to RhH#CO)L2 in the presence of an equimolar mixture of CO and H2. In the presence of CO and C2H4, the
peaks observed in the acyl region and the terminal metal carbonyl region indicate the formation of three different acylrhodium complexes
which are Rh#CO)L2#COEt), Rh#CO)2L2#COEt), and Rh#CO)3L#COEt). Similar species were also observed during the hydroformylation
reaction. The ®rst ever detection of the presence of Rh#CO)L2#COEt) under hydroformylation conditions provides direct evidence for the
mechanism originally proposed by Wilkinson and co-workers. The carbonyl stretching frequencies of all of the rhodium±carbonyl species
are shifted to higher wavenumbers due to a reduction of electron density at the metal center by the CF3 groups. q 2002 Elsevier Science Ltd.
All rights reserved.
1. Introduction
solution, which is in¯uenced by the nature of the solvent.
Using scCO2 may shift such equilibria in the desired
direction.
As a reaction solvent in homogeneous catalysis, super-
critical carbon dioxide #scCO2) offers many advantages
over conventional organic solvents.1,2
5. scCO2 is non-¯ammable, non-toxic, environmentally
acceptable, cheap, readily available in large quantities
and has a low critical temperature and a moderate critical
pressure.
6. The strong dependency of the solubility of solutes in
scCO2 to temperature and pressure in the vicinity of
the critical point may be exploited in development of
ef®cient catalyst separation/recovery and product puri®-
cation schemes.
1. Many gases exhibit higher solubilities in scCO2 than in
organic solvents.3 Thus, based simply on increased
concentration of reactants, faster reactions are expected
to occur within the scCO2 phase provided that the
reaction order with respect to the gaseous substrate is
positive.
2. Conducting the reaction in a single phase can eliminate
the problems when mass transfer controls the reaction
rate across the gas±liquid interface.
3. The design, scale-up and operation of reactors operating
in single phase are much simpler than multiphase
reactors.
4. The unusual solvent properties of scCO2 may lead to
increased rates or selectivities due to solvent effects in
homogeneous catalysis. Transition metal complexes in
catalytic cycles participate only in a few types of
reactions such as ligand exchange, insertion, oxidative
addition and reductive elimination. The overall rate of
the reaction and in some cases the selectivity depend on
the equilibria between the reactive intermediates in
Although scCO2 has very attractive properties as a reaction
medium in homogeneous catalysis, the low solubility of
conventional homogeneous catalysts in scCO2 has hindered
developments in this area.4 One way to increase solubility in
scCO2 is to utilize CO2-philic moieties such as ¯uoroether,
¯uoroalkyl, ¯uoroacrylate, siloxane, or phosphazene. In the
pioneering example on developing chelating agents for
scCO2 extraction of heavy metals from aqueous solutions,
¯uorination of the ethyl groups of the diethyldithiocarba-
mate #DDC) ligand was found to enhance the solubilities of
Cu#DDC)2, Ni#DDC)2, and Co#DDC)3 in scCO2 by three
orders of magnitude.5 This discovery has been one of the
key developments in scCO2 research. Subsequently, a wide
variety of reagents functionalized with CO2-philic groups
were developed for scCO2 applications.6,7
Keywords: in situ FTIR; hydroformylation; rhodium; supercritical carbon
dioxide; ¯uorinated phosphines; homogeneous catalysis.
p
During the past 3 years, signi®cant advances have also
been made in development of synthetic methods for
Corresponding author. Tel.: 11-860-486-4601; fax: 11-860-486-2959;
e-mail: cerkey@engr.uconn.edu
0040±4020/02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved.
PII: S0040-4020#02)00246-6

3930
S. Haji, C. Erkey / Tetrahedron 58 ,2002) 3929±3941
Figure 1. Spectra of 1: #a) in scCO2 at 508C and 115 bar #3 mM), and #b) in solid form.

S. Haji, C. Erkey / Tetrahedron 58 ,2002) 3929±3941
3931
homogeneous catalysts, which exhibit high solubilities in
scCO2. These started with the pioneering work of Kainz
and Leitner who prepared triarylphosphine ligands with
per¯uoroalkyl tails.8 The rhodium complexes of these
ligands were found to be suf®ciently soluble in scCO2 and
could catalyze the hydroformylation of ole®ns. Along
the same time period, an interesting coincidence and an
important development was the invention of ¯uorous
biphasic systems #FBS) by Horvath and Rabei9 as a means
to tackle the problems associated with catalyst recovery/
recycle in homogeneous catalysis. The success of this
approach depends on favorable partitioning of ¯uorous
homogeneous catalysts into the ¯uorous phase. Subse-
quently, quite a few studies appeared in the literature on
synthesis, characterization and reactive properties of
¯uorinated ligands, primarily phosphines.10±14 Our group
synthesized and characterized several ¯uoroalkyl and
¯uoroalkoxy substituted tertiary aryl phosphine ligands
{[3,5-#CF3)2C6H3]3P, #4-CF3C6H4)3P, #3-CF3C6H4)3P,
#4-CF3OC6H4)3P, [4-F#CF2)4#CH2)3C6H4]3P]} and rhodium
complexes of some of these ligands #HRhCOL3) as hydro-
formylation catalysts.15,16 Organometallic complexes of
similar CO2-philic phosphine ligands were found to
ef®ciently catalyze other types of reactions such as enantio-
selective hydrogenation,17 hydrovinylation18 and hydro-
formylation19 in scCO2.
mediates in scCO2 solutions with ligands substituted with
¯uorous groups. In this article, we report the results of our
investigations of reactions of hydridocarbonyltris#3,5-bis-
#tri¯uoromethyl)phenylphosphine)rhodium #I), RhH#CO)-
[P#3,5-#CF3)2C6H3)3]3 #1), with CO, H2, and ethylene in
dense CO2 using a high-temperature high-pressure trans-
mission FTIR probe. Ethylene was selected as the substrate
to simplify interpretation of the spectra. The results were
compared to the behavior of the conventional catalyst 3 in
organic solvents.
2. Results and discussion
The concentration of 1 ranged from 2 to 3 mM in all experi-
ments to achieve a satisfactory signal-to-noise ratio.
2.1. The nature of 1 in carbon dioxide
350 mg #3 mM) of HRh#CO)L3 #1) was placed in the vessel.
The vessel ®rst was heated to 508C and then was pressurized
with CO2 up to 115 bar bringing the solvent to supercritical
conditions. The solution became yellow immediately with
complete dissolution of 1. The spectra were taken every
5 min for 30 min. There were no changes in the successive
spectra and the spectra after 30 min is shown in Fig. 1a. The
bands at 2041 and 1964 cm21 were assigned to the metal
hydride #nRhH) and the carbonyl #nCO) in 1, respectively.
The spectra of 1 in solid form is shown in Fig. 1b
#nRhHˆ2039 cm21 and nCOˆ1952 cm21). The n#CO) is
a sensitive indicator of electron density at the metal center,
yielding a relative measure of the amount of p-backbonding
from an occupied metal d-orbital to the empty pp-orbital of
the carbonyl. Comparison of nCO of 1 with that of the
conventional catalyst 3 #1952 vs 1923 cm21, solid spectra)21
reveals that metal±carbonyl bond is weaker in 1 compared
to 3 which is due to a reduction of electron density at the
metal center caused by the electron withdrawing CF3
groups. The ®rst possibility is that electron density reduc-
tion is caused by a strong electron back donation to the
¯uorinated ligand 2 if it is s-donor/p-acceptor. The second
possibility is that the electron density reduction is due to less
electron donation from 2 compared to triphenylphosphine to
the metal center. This would be true if 2 is pure s-donor.
An important consideration in incorporation of ¯uorous
groups into conventional ligands is the effect of electron
withdrawing nature of ¯urous groups on activity and selec-
tivity. For example, in hydroformylation in scCO2, the
activity of the catalyst formed in situ from Rh#CO)2#ACAC)
and [3,5-#CF3)2C6H3]3P #2) as ligand, was higher by
a
factor of
5
than that when the ligand was
[4-F#CF2)4#CH2)3C6H4]3P.20 Likewise, per¯uoroalkyl sub-
stituted BINAPHOS resulted in higher regioselectivities
compared to the unsubstituted ligand in enantioselective
hydroformylation in organic solvents.19 Our primary objec-
tive in this study is to probe the effects of ¯uorous groups on
the nature and distribution of intermediates in scCO2 by
in situ FTIR spectroscopy for a reaction that is reasonably
well understood, namely hydroformylation of ole®ns by
rhodium/phosphine complexes. In situ FTIR spectroscopy
is a powerful tool well suited particularly for such investi-
gations since the intermediates can be identi®ed and
monitored at catalytic concentrations under reaction
conditions.
The similarity of the spectra of 1 in solid form and in solu-
tion indicates that 1 does not dissociate in scCO2. The slight
shift in nCO in solution indicates the effect of the solvent
on the electronic structure of 1. The NMR spectra of 1 in
acetone-d6 shows that 1 does not dissociate in acetone also.
Hydroformylation of ole®ns catalyzed by hydridocarbo-
nyltris#triphenylphosphine)rhodium #I), RhH#CO)#PPh3)3
#3), has been extensively studied by a wide variety of
techniques including FTIR spectroscopy by Wilkinson and
his co-workers21±24 and by Morris and Tinker.25 In the late
eighties, development of mid-range infrared optical ®bers
enabled the measurements to be carried out in situ at high
temperatures and pressures.26 The studies by Moser et al.
using Cylindrical Internal Re¯ectance FTIR #CIR/FTIR) at
extended temperatures and pressures improved our under-
standing of the underlying mechanism of the hydroformyl-
ation reaction.27,28 Even though, in situ FTIR spectroscopy
has been used for probing homogeneous chemical reactions
in supercritical ¯uids,29 no studies in the literature has been
reported on spectroscopic identi®cation of reactive inter-
1
Both H NMR and 31P{1H} NMR showed no evidence of
any band or broad band near the frequencies of the pure
ligand 2. This behavior is markedly different from the
behavior of the conventional catalyst 3, which was found
to undergo ligand dissociation in solvents such as benzene,
dichloromethane, and cyclohexane. Evans and co-workers21
suggested the following equilibria based on molecular
weights measured by osmometric techniques:
RhHꢀCO†ꢀPPh3† 2OPPh3 RhHꢀCO†ꢀPPh3† 2OPPh3 RhHꢀCO†ꢀPPh3†
3
2
1 PPh3
1 PPh3
ꢀ1†
They also reported an appreciable difference in the metal

3932
S. Haji, C. Erkey / Tetrahedron 58 ,2002) 3929±3941
Figure 2. #a) Spectra of 1 #2.73 mM) in scCO2 at 508C and 115 bar with CO #165 mM), and #b) after the addition of H2.

S. Haji, C. Erkey / Tetrahedron 58 ,2002) 3929±3941
3933
Figure 3. Spectra of 1 #2.73 mM) in scCO2 at 508C and 115 bar with 1:1 mixture of CO/H2 #150 mM each).
hydride stretching frequency in IR spectra of 3 in solution
and in solid form which is not the case observed with 1.
Such behavior is indicative of a stronger metal±ligand bond
in 1 than in the conventional catalyst due to tri¯uoromethyl
groups.
more, the smaller electron density on the metal results
in more electron donation from H2 and a stronger metal
hydride bond. The solvent could also have some effect on
these shifts. The IR spectrum in Fig. 2a also indicates
the presence of an appreciable amount of the dimer,
[RhH#CO)2L2]2 #6); two bands of this species are clearly
seen at 1827 and 1799 cm21 and were assigned to the
bridging carbonyl groups. The analogous dimer reported
previously21,27 are reported to have four bands, two in the
carbonyl-bridging region and the other two in the terminal
carbonyl region. In this case, the terminal carbonyl could
not be observed since the three strong bands of RhH#CO)2L2
#5) might be obscuring them.
2.2. Reaction of 1 with carbon monoxide in carbon
dioxide
300 mg #2.73 mM) of 1 was placed in the vessel. After
the vessel was heated to 508C and ¯ushed with CO, it
was pressurized with CO to 4.5 bar #165 mM) and then
pressurized with CO2 up to 115 bar. The mixture became
a single phase in a few minutes with complete dissolution of
1. While ®lling the reactor with CO2, the solution color
rapidly changed from yellow to dark orange to ®nally
dark red. The IR spectrum is shown in Fig. 2a. One of the
new formed complexes is expected to be the hidrydodi-
carbonylmetal complex, RhH#CO)2L2 #5), for two reasons;
the appearance of three bands and the appearance of two
of those bands at the terminal M±CO region. The metal
hydride frequency appeared at 2076 cm21 while the two
carbonyl-stretching frequencies were at 2056 and
2034 cm21. T hen#CO) of 5 are higher than the n#CO) of
The results indicate that in the presence of CO, 1 undergoes
ligand dissociation in solution based on the following
equilibria:
2 L
1 CO
RhHꢀCO†L3 O RhHꢀCO†L2 O RhHꢀCO†2L2
1 L
2 CO
ꢀ2†
2 H2
O ‰RhꢀCO†2L2Š2
1 H2
and the equilibria in the ®rst two reactions in Eq. #2) are
shifted way to the right in the presence of excess CO.
RhH#CO)2#PPh3)2 #1980 and 1939 cm21 21
due to a reduc-
)
tion of electron density at the metal center. Such an increase
in n#CO) with decreasing basicity of the ligand was also
observed by Moser et al. for RhH#CO)2#P#p-Z-C6H4)3)2
where Z is N#CH3)2, OCH3, H, F, Cl, or CF3.27 When Z
was CF3, the frequencies were 2050, 1997, and 1953
cm21. Therefore, the existence of a second CF3 group results
in a very signi®cant shift to a higher wavenumber. Further-
When hydrogen was added to the solution, the spectrum in
Fig. 2b showed the emergence of new bands and weakening
of others. Two of those new bands were assigned
to RhH#CO)L2 #4) #nRhHˆ2066 cm21 and nCOˆ2027
cm21). The amounts of both the dicarbonyl complex 5 and
the dimer 6 were reduced appreciably, which provides

3934
S. Haji, C. Erkey / Tetrahedron 58 ,2002) 3929±3941
Figure 4. The spectra of 1 #2.2 mM) in liquid CO2 at 1.58C and 40 bar in different CO/C2H4 mole ratios #those spectra were not subtracted from the solvent±
carbon monoxide solution, the peaks at 1944 and 2070 cm21 are solvent peaks): #a) with CO/C2H4ˆ1 #140 mM each); #b) with CO/C2H4ˆ0.35 #50 mM CO
and 140 mM ethylene); and #c) with CO/C2H4ˆ0.035 #10 mM CO and 285 mM ethylene).

S. Haji, C. Erkey / Tetrahedron 58 ,2002) 3929±3941
3935
Figure 4. %continued)
further proof of the existence of the equilibrium in Eq. #2).
The lowest band assigned to 5 #2034 cm21) could not be
detected because it was probably obscured by the strong
complex 5 as can be seen upon close inspection of Fig. 3. In
case of the conventional catalyst the analog of complex 4
was observed by Evans and co-worker21 to absorb at 2000
and 1920 cm21 in dichloromethane, however, that species
was observed when 3 was dissolved in a solvent. Moser et
al.27 found no evidence of the analog of 4 under
hydroformylation conditions with the conventional catalyst
system. In studies of hydroformylation of 1-hexene by RhH-
#CO)#PPh3)3 #3) using an in situ high-pressure 31P{1H}
NMR probe,30 the analog of 4 could not be detected. The
existence of 4 in scCO2 shows that the equilibrium compo-
sition for the reactions given in Eq. #2) are strongly affected
by the CF3 groups and also perhaps by the solvent. Since the
color of the solution in this experiment was dark yellow, the
red color solution in the previous experiment could have
been caused by either the dicarbonyl complex 5 or the
dimer 6. At the end of the experiment, the reaction mixture
was cooled, depressurized, and NMR spectra were taken for
the solid. 1H NMR and 31P{1H} NMR spectra showed peaks
for the free ligand, in case of 31P{1H} NMR the peak was
broad. In addition, 1H NMR and 31P{1H} NMR showed that
most of 4 reverts back to 1 after the removal of CO/H2 and
the solvent.
band of
4
#2027 cm21). Three other shoulders at
2012, 1996, and 1973 cm21 and another weak peak at
1867 cm21 could not be assigned to any complex. The
color of the solution remained dark red throughout the
experiment.
2.3. Reaction of 1 with carbon monoxide/hydrogen in
carbon dioxide
This experiment is similar to the one above, except that the
hydrogen was added at the same time with CO at a 1:1 ratio.
The reaction was carried out at 508C with 300 mg of 1
#2.73 mM), 8 bar of 1:1 CO/H2 mixture #150 mM each)
and the rest CO2 to make the total pressure of the system
115 bar. The yellow powder 1 dissolved immediately. The
CO/H2 mixture dissolved after a few minutes resulting in a
single-phase dark yellow solution. The spectra were taken
every 10 min for four times. There were no changes in the
successive spectra and the spectra after 30 min is shown in
Fig. 3. Two distinguished bands are seen clearly at 2065 and
2027 cm21, which were assigned to the metal hydride
#nRhH) and the carbonyl #nCO) of 4, respectively. Since
the bands of 1 were not detected, the equilibrium for the ®rst
reaction in Eq. #2) is shifted away to the right. Although the
catalyst undergoes ligand dissociation in the presence of
CO/H2, the presence of hydrogen inhibited CO coordination
to the dissociated catalyst 4. The very weak bands observed
at 1827 and 1803 cm21 were assigned to the dimer 6 which
should be in equilibrium with a small amount of dicarbonyl
2.4. Reaction of 1 with carbon monoxide and ethylene in
carbon dioxide
240 mg #2.2 mM) of 1 was charged into the reactor. At
1.58C, the reactor was pressurized with ethylene up to
3.15 bar #140 mM) then with CO up to 6.3 bar #140 mM)
and ®nally with CO2 up to 42 bar #CO/Rhˆ64, CO/C2H4ˆ
1:1). The solution was initially two phases with a yellow

3936
S. Haji, C. Erkey / Tetrahedron 58 ,2002) 3929±3941
Figure 5. #a) Spectra of the hydroformylation of ethylene #63 mM) in CO2 #148C, 65 bar) with 1:1 mixture of CO/H2 #240 mM each) catalyzed by 1 #2.73 mM)
#peaks at 1944 and 2070 cm21 are solvent peaks); and #b) The spectra at the end of the hydroformylation reaction at 508C and 200 bar.

S. Haji, C. Erkey / Tetrahedron 58 ,2002) 3929±3941
3937
catalyst-rich bottom phase. The IR spectra of the bottom
phase were acquired at this temperature every 2.5 min for
15 min. Subsequently, the vessel was heated to 508C. All the
spectra had similar peaks, but the intensities decreased
slightly with increasing temperature. While heating, the
system turned into one phase at 248C. With further heating
to 508C, the pressure increased up to 162 bar. The solution
color stayed the same throughout the entire period of
heating. The spectrum at 1.58C and 42 bar is shown in
Fig. 4a which shows that 1 is converted to different species.
Three peaks appeared in the acyl frequency range and
another seven peaks in the terminal metal carbonyl range.
The effect of decreasing CO concentration is shown in Fig.
4b #CO/C2H4ˆ0.35, 50 mM of CO and 140 mM of ethyl-
ene) and Fig. 4c #CO/C2H4ˆ0.035, 10 mM of CO and
285 mM of ethylene). Comparing Fig. 4a±c, it was
concluded that since the intensities of the three peaks in
the acyl frequency range changed with CO/C2H4 mole
ratio, they were for three different acylrhodium species.
When the ratio of CO/C2H4 charged to the reactor was
solvent peaks. It is also worth mentioning that when only
ethylene is charged to the reactor, no peaks were observed in
the 1620±1700 cm21 region.
2.5. Hydroformylation reaction of ethylene with 1 in
carbon dioxide
The preliminary experiments conducted at 508C indicated
that the ethylene hydroformylation reaction was very fast to
follow the intermediates. Therefore, the reaction was carried
out at 148C and 65 bar in liquid CO2 with 300 mg of 1
#2.73 mM), 11.5 bar of 1:1 CO/H2 mixture #240 mM each)
and 1.5 bar ethylene #63 mM). The reaction mixture was
initially two phases with a yellow catalyst-rich bottom
phase. As usual, 1 dissolved immediately. IR spectra of
the homogeneous bottom phase were acquired every
2.5 min for 10 min. The consecutive spectra showed no
changes in the reaction mixture species and the spectra at
the end of 10 min is shown in Fig. 5a #the spectra is not
subtracted from solvent±CO solution, peaks at 1947 and
2071 cm21 are due to the solvent). The peak at 1741 cm21
was assigned to the carbonyl group of propionaldehyde,
based on the spectrum of propionaldehyde in CO2 at the
same conditions, indicating small conversion. Weak bands
of ethylene also were observed #for example peak around
3100 cm21 and a hump at around 1899 cm21), indicating
some unconverted ethylene. As in the previous set of experi-
ments, all of the peaks in the 1975±2086 cm21 region were
visible #1975, 1999, 2009, 2014, 2033, 2051, and
2086 cm21), as well as the peaks assigned to the acyl groups
of 7, 8, and 9 #1645, 1672, and 1695 cm21). The dimer 6 also
appeared in very low concentration, however, the band at
1799 cm21 was shifted to 1795 cm21. Then the heating was
started while the data acquisition continued. While heating,
the intensities of propionaldehyde peaks #1741, 2989, 2951,
2906, 2821, and 2724 cm21) increased, especially when the
solution turned into one phase at 278C and 81 bar. At the
same time the acylrhodium complexes' peaks were decreas-
ing in intensities slowly. Changes were also noticed in the
1900±2100 region. The color started to change to dark
orange and then to dark red at the end of the reaction. The
reaction stopped when all the ethylene was consumed. The
intensity of the propionaldehyde peaks stopped increasing
and leveled off and the three acylrhodium peaks dis-
appeared. The spectra at the end of the reaction at 508C
and 200 bar is shown in Fig. 5b. The following species
reduced, the intensity of the ®rst acyl peak #1645 cm21
)
increased along with the peak at 1975 cm21, while that of
the third acyl peak #1695 cm21) decreased along with the
peak at 2086 cm21 and both peaks at 2008 and 2033 cm21
disappeared. The intensity of the middle acyl peak
#1673 cm21) changed slightly. This behavior indicates that
the three species are mono-, di-, and tricarbonyl acyl-
rhodium complexes. The bands at 1645 and 1975 cm21
were assigned to the carbonyl in the acyl group and the
terminal metal carbonyl, respectively, in the monocarbonyl
acylrhodium intermediate, Rh#CO)L2#COEt) #7). The band
at 1673 cm21 and the two bands at 2014 and 2051 cm21
were assigned to the carbonyl in the acyl group and the
two terminal metal carbonyls, respectively, in the dicarbon-
yl acylrhodium intermediate, Rh#CO)2L2#COEt) #8). The
band at 1695 cm21 and the three bands at 2008, 2033, and
2086 cm21 were assigned to the carbonyl in the acyl group
and the three terminal metal carbonyls, respectively, in the
tricarbonyl acylrhodium intermediate, Rh#CO)3L#COEt)
#9). The presence of CO more likely converted 1 to 4,
which further reacted with ethylene to form many different
species. A plausible pathway is given below:
1 C2H4
2 L
RhHꢀCO†L3 O RhHꢀCO†L2 O RhHꢀCO†L2ꢀC2H4†
1 L
2 C2H4
1 CO
O RhꢀCO†L2ꢀEt† O RhꢀCO†2L2ꢀEt†
2 CO
were observed: propionaldehyde with
1741 cm21
a
, the dimer 6 with peaks at 1798 and
peak at
ꢀ3†
1 CO
O RhꢀCO†L2ꢀCOEt† O RhꢀCO†2L2ꢀCOEt†
1828 cm21, the species, RhH#CO)L2 #4), with peaks at
2 CO
2026 and 2064 cm21, and ®nally the hydridodicarbonyl
complex 5 with two of its peaks at 2075 and 2056 cm21
.
1 COO= 2 L RhꢀCO†3LꢀCOEt†
A weak band was observed at 1993 cm21 but could not be
related to any species. After the completion of the hydro-
formylation reaction and under the presence of extra CO/H2,
the major species in the reaction mixture are 4 and 5. Subse-
quently, the reaction mixture was cooled, depressurized and
the recovered solid residue was analyzed by NMR. The
31P{1H} NMR showed the presence of 1 as a major species,
which indicates that during the removal of CO/H2, 5 was
converted to 4 which reassociated with the free ligand 2 to
form 1 in accordance with the equilibria given in Eq. #2).
Two other doublets with very small intensities also appeared
2 CO= 1 L
Among the three acyl species, analogues of 7 and 9 were
not detected before in conventional hydroformylation
studies. The band at 2001 cm21 seen in Fig. 4a, was noticed
to appear in wide range of frequencies in different experi-
ments with different CO/C2H4 ratios, from 1997 to 2002
cm21. Since the intensity of the peak was not changing
with CO/C2H4 ratio, it was not assigned to any acylrhodium
species. The broad band observed at 1899 cm21 is one of the
bands of ethylene. The peaks at 1944 and 2070 cm21 are

3938
S. Haji, C. Erkey / Tetrahedron 58 ,2002) 3929±3941
Figure 6. Summary of the structure, IR bands, and the reactions of the compounds detected.
on the 31P{1H} NMR which might be phosphine±metal
species.
ethylene in supercritical carbon dioxide with RhH#CO)L3
#LˆP#3,5-#CF3)2C6H3)3). Catalyst concentration for obtain-
ing a satisfactory signal-to-noise ratio was found to be
around 2 mM. Marked differences were found between the
behavior of RhH#CO)L3 in supercritical carbon dioxide and
the RhH#CO)#PPh3)3 in organic solvents. RhH#CO)L3 does
not dissociate in scCO2 and it is converted to RhH#CO)2L2
and to [Rh#CO)2L2]2 in the presence of CO and mainly to
RhH#CO)L2 in the presence of an equimolar mixture of CO
3. Conclusions
High pressure in situ FTIR spectroscopy was found to be a
powerful tool for investigating the effects of ligand modi®-
cation on the nature of intermediates in hydroformylation of

S. Haji, C. Erkey / Tetrahedron 58 ,2002) 3929±3941
3939
Figure 7. Schematic diagram of the experimental setup.
and H2. In the presence of CO and C2H4, the peaks observed
in the acyl region and the terminal metal carbonyl region
indicate the formation of three different acylrhodium
complexes which are Rh#CO)L2#COEt), Rh#CO)2L2#COEt),
and Rh#CO)3L#COEt). Similar species were also observed
during the hydroformylation reaction. A summary of
the reactions is shown in Fig. 6 within the framework of
the mechanism originally proposed by Wilkinson and
co-workers. The IR bands of the observed species are also
shown in Fig. 6. The carbonyl stretching frequencies of
all of the rhodium±carbonyl species are shifted to higher
wavenumbers due to a reduction of electron density at the
metal center by the CF3 groups.
4. Experimental
4.1. Experimental apparatus
A schematic diagram of the experimental apparatus is
shown in Fig. 7. The set-up consists of a custom manu-
factured, 51.5 mL stainless steel reactor ®tted with two
sapphire windows #100 ID, Sapphire Engineering, Inc.),
poly-ether±ether±ketone o-rings #Valco Instruments, Inc.),
a T-type thermocouple assembly #Omega Engineering,
DP41-TC-MDSS), a pressure transducer #Omega Engi-
neering, PX300-7.5KGV), a vent line, and a rupture disk
assembly #Autoclave Engineers). A high-pressure±high
temperature FTIR sampling head is screwed through the
reactor body to reach inside the reactor. The reactor rests
on a magnetic stir plate and is heated to appropriate
temperatures by a circulating heater #Fisher Scienti®c) via
a machined internal heating coil. The reactor is pressurized
with CO2 from a syringe pump #ISCO, 260D) equipped with
a cooling jacket to the desired reaction pressure.
The turnover frequency #TOF) obtained with 1 in scCO2
is about ®ve times higher than the TOF obtained with
RhH#CO)#PPh3)3 in conventional organic solvents at
comparable conditions and also about ®ve times higher
than the TOF obtained with RhH#CO)[#4-F#CF2)4-
#CH2)3C6H4)3P]3 in scCO2.20 Since the electronic properties
of RhH#CO)#PPh3)3 and RhH#CO)[#4-F#CF2)4#CH2)3-
C6H4)3P]3 are similar, the low basicity of the ligand 2 is
responsible for the higher activity rather than the effect of
the solvent. Based on the spectroscopic measurements
presented in this manuscript, the high activity is most likely
due to a shift in the equilibrium distribution of the cata-
lytically active intermediates. The species 7 is expected to
undergo oxidative addition by hydrogen and the concen-
tration of 7 in solution should be directly proportional to
the activity. Since the analog of 7 was never detected under
hydroformylation conditions in organic solvents with the
conventional catalyst, the higher activity observed is most
likely related to the higher concentration of 7 in solution.
Furthermore, the decrease in activity with an increase in CO
concentration is also related to the equilibria between mono-,
di-, and tricarbonyl acyl intermediates. The equilibria are
shifted towards the catalytically inactive di- and tricarbonyl
acyl rhodium intermediates with increasing CO concentra-
tion. Moreover, increasing the CO/H2 ratio over 1.0 will
promote the formation of the inactive dimer species 6,
resulting in a decrease in activity.
4.2. Spectra collection
Spectroscopic data are obtained using a Bruker Vector 22
FTIR spectrometer with a Remspec mid-IR ®ber optic
system comprising a bundle of 19 optical ®bers, which
transmit in the mid-IR, 5000±900 cm21. Seven of these
®bers are attached to a signal launch module bolted to the
collimated external beam port of the spectrometer; 12 ®bers
are attached to an external, liquid-nitrogen cooled MCT
detector provided with specialized optics to optimize
capture of the mid-IR signal from the end of the ®ber
bundle. The combined bundle of 19 ®bers is ®tted with a
high-pressure sampling head. A zinc sul®de optical transfer
crystal inside the head transmits the signal into a
double-pass, variable-path transmission cell made from
stainless steel and returns the signal to the ®ber bundle.
The sampling head is rated for use at pressures up to
200 bar. IR spectra of solids were measured with the same
spectrometer but with internal DTGS detector. The solid
sample was grounded into ®ne particles which was then

3940
S. Haji, C. Erkey / Tetrahedron 58 ,2002) 3929±3941
Figure 8. Structure of the ligand and catalyst.
mechanically pressed into a thin wafer. A transparent ®ller,
KBr, was used to give the wafer suf®cient mechanical stabi-
lity and to dilute the solid sample down to 5±10% by
weight.
3,5-bis#tri¯uoromethyl)phenylphosphine #2), was prepared
using a procedure similar to that described by Bhattacharyya
et al.31 Six grams of 3,5-bis #tri¯uoromethyl)bromobenzene
and 50 cm3 of diethyl ether were put in a three-necked
round-bottom ¯ask. An addition funnel containing 25 cm3
of diethyl ether was connected to the ¯ask. The ¯ask was
placed in dry ice±acetone bath #around 2788C) and ¯ushed
with nitrogen. 1.31 g of n-butyllithium #1.6 M solution in
hexanes) was injected to the addition funnel and the solution
was added dropwise to the ¯ask over 1 h and stirred for a
further 1 h. A solution of 1.03 g of phosphorus trichloride
#10% excess) in 25 ml of diethyl ether was added dropwise,
through the addition funnel, to the reaction mixture at
2788C over a further hour. The reaction mixture was then
allowed to warm slowly to room temperature with con-
tinuous stirring over 12 h. The mixture was hydrolyzed
with 25 ml of 10% ammonium chloride aqueous solution.
The organic phase was collected and washed twice with
water #around 2£25 ml). The organic phase was puri®ed
by vacuum distillation. The condensate was collected as a
4.3. General procedures
For each experiment, a certain amount of catalyst was
placed in the reactor and the vessel was sealed. After
acquiring a background spectrum of the empty reactor
with optical ®bers attached, the vessel was heated or cooled
to the desired temperature. Before the system was charged
to the desired pressures of the gaseous reactants #CO, H2,
C2H4), the reactor was ¯ushed several times with the ®rst
added gas. After the system was charged with the gaseous
reactants, carbon dioxide was added to the required
pressure. Then the IR spectra were acquired as required.
Spectral scans were taken at 4 cm21 resolution and signal
averaged over 96 or 128 scans, depending on application.
No smoothing algorithms were applied to improve visual
quality of any spectra presented. To obtain high quality
spectra, the carbon dioxide #solvent) spectra or the spectra
of a mixture of carbon dioxide and 1:1 mixture of CO/H2
or only CO at operating conditions and concentrations
were subtracted from the spectra of the reaction mixture
whenever possible.
1
white crystalline product with a yield of 75%. H NMR in
acetone-d6 d 8.17 #s, 1H), d 8.21 #d, J#PH)ˆ6.8 Hz, 2H).
31P{1H} NMR in acetone-d6 d 20.44 #s).
0.093 g #0.36 mmol) of dicarbonylacetylacetonato rhodium
#I), #CO)2Rh#acac), and 1.21 g #1.81 mmol) of 2, were
placed in the high-pressure reactor. After the temperature
was raised to 508C, the reactor was pressurized with a 1:1
mixture of CO/H2 to 8 bar. Carbon dioxide was pumped in
until a pressure of 140 bar was reached. After 45 min, the
reactor was cooled down and the CO2/CO/H2 mixture was
slowly vented. The products were collected and washed
with chloroform to yield a yellow powder of hydridocarbo-
nyltris#3,5-bis#tri¯uoromethyl)phenylphosphine)rhodium#I),
HRh#CO)[P#3,5-#CF3)2C6H3)3]3 #1), #0.44 g, 57%). 1H NMR
in acetone-d6 d 29.58 #q, J#RhH)ˆ13.32 Hz, 1H), d 8.01 #s,
18H), d 8.19 #s, 9H). 31P{1H} NMR in acetone-d6 d 47.00
#d, J#PRh)ˆ157.1 Hz). The structures of the ligand 2 and the
catalyst 1 are shown in Fig. 8.
4.4. NMR spectra
The 1H NMR and 31P NMR spectra were acquired at
268C on a Bruker DRX-400 spectrometer #400.13 and
161.98 MHz, respectively). Chemical shifts are reported
in ppm #d) relative to tetramethylsilane #1H) or 85%
H3PO4 #31P).
4.5. Ligand and catalyst preparation
All the chemicals were purchased from Aldrich and used
without further puri®cation. First, the ¯uorinated ligand,

S. Haji, C. Erkey / Tetrahedron 58 ,2002) 3929±3941
3941
13. Hamilton, D. J.; Payne, M. J. Chem. Commun. 1997, 1127±
1130.
Acknowledgements
14. Fawcett, J.; Hope, E. G.; Kemmitt, R. D. W. P.; Russell, D. R.;
Stuart, A. M. J. Chem. Soc., Dalton Trans. 1998, 3751±3763.
15. Davis, T.; Erkey, C. Ind. Engng Chem. Res. 2000, 39, 3671±
3678.
The authors gratefully acknowledge ®nancial support from
the National Science Foundation, Grant# CTS-0080823.
The authors also thank the School of Engineering at the
University of Connecticut for a generous grant to purchase
the FTIR unit.
16. Palo, D. R.; Erkey, C. Ind. Engng Chem. Res. 1999, 38, 3786±
3792.
17. Kainz, S.; Brinkmann, A.; Leitner, W.; Pfaltz, A. J. Am. Chem.
Soc. 1999, 121, 6421±6429.
References
18. Wegner, A.; Leitner, W. Chem. Commun. 1999, 1583±1584.
19. Francio, G.; Leitner, W. Chem. Commun. 1999, 1663±1664.
20. Palo, D. R.; Erkey, C. Organometallic 2000, 19, 81±86.
21. Evans, D.; Wilkinson, G. J. Chem. Soc. ,A) 1968, 2660±2665.
22. Evans, D.; Osborn, J. A.; Wilkinson, G. J. Chem. Soc. ,A)
1968, 3133±3142.
1. Erkey, C. In Supercritical Fluids Technology in Materials
Science and Engineering; Sun, Y.-P., Ed.; Marcel Dekker:
New York, 2002; pp 189±226.
2. Jessop, P.; Ikariya, T.; Noyori, R. Chem. Rev. 1999, 99, 475±
493.
23. Yagupsky, G.; Brown, C. K.; Wilkinson, G. J. Chem. Soc. ,A)
1970, 1392±1401.
3. Tsang, C.; Street, W. Chem. Engng Sci. 1981, 36, 993±999.
4. Palo, D. R.; Erkey, C. J. Chem. Engng Data 1998, 43, 47±49.
5. Laintz, K.; Wai, C.; Yonker, C.; Smith, R. J. Supercrit. Fluids
1991, 4, 194±198.
24. Brown, C. K.; Wilkinson, G. J. Chem. Soc. ,A) 1970, 2753±
2764.
25. Morris, D. E.; Tinker, H. B. Chemtech 1972, 554±559.
26. Moser, W. R.; Berard, J. R.; Melling, P. J.; Burger, R. J. Appl.
Spectrosc. 1992, 46, 1105±1112.
6. DeSimone, J.; Maury, E.; Menceloglu, Y.; Combes, J.;
McClain, J.; Romack, T. Science 1994, 65, 356±359.
7. Yazdi, A. V.; Beckman, E. Ind. Engng Chem. Res. 1996, 35,
3644±3652.
27. Moser, W. R.; Papile, C. J.; Brannon, D. A.; Duwell, R. A.;
Weininger, S. J. J. Mol. Catal. 1987, 41, 271±292.
28. Moser, W. R.; Papile, C. J.; Wieninger, S. J. J. Mol. Catal.
1987, 41, 293±302.
8. Kainz, S.; Koch, D.; Baumann, W.; Leitner, W. Angew.
Chem., Int. Ed. Engl. 1997, 36, 1628±1629.
9. Horvath, I. T.; Rabai, J. Science 1994, 266, 72±75.
10. Horvath, I.; Kiss, G.; Cook, R. A.; Bond, J.; Stevens, P.; Rabai,
J.; Mozeleski, E. J. Am. Chem. Soc. 1998, 120, 3133±3143.
11. Juliette, J. J. J.; Horvath, I. T.; Gladysz, J. Angew. Chem., Int.
Ed. Engl. 1997, 36, 1610±1612.
29. Buback, M. Angew. Chem., Int. Ed. Engl. 1991, 30, 641±653.
30. Bianchini, C.; Lee, H. M.; Meli, A.; Vizza, F. Organometallics
2000, 19, 839±843.
31. Bhattacharyya, P.; Gudmunsen, D.; Hop, E. G.; Kemmitt,
R. D. W.; Paige, D. R.; Alison, M. J. Chem. Soc., Perkin
Trans. 1 1997, 3609±3612.
12. Sinou, D.; Pozzi, G.; Hope, E. G.; Stuart, A. M. Tetrahedron
Lett. 1999, 40, 849±852.