Full kinetic description of 1-octene hydroformylation in a supercritical medium

Article abstract of DOI:10.1016/j.molcata.2011.06.012

The kinetics of the hydroformylation of 1-octene in a supercritical carbon dioxide medium, catalyzed by a tris(3,5-bis[trifluoromethyl]phenyl)phosphine- modified rhodium catalyst, have been investigated. The influence of the concentration of carbon dioxid

Full text of DOI:10.1016/j.molcata.2011.06.012

Journal of Molecular Catalysis A: Chemical 346 (2011) 1–11
Contents lists available at ScienceDirect
Journal of Molecular Catalysis A: Chemical
journal homepage: www.elsevier.com/locate/molcata
Editor’s Choice paper
Full kinetic description of 1-octene hydroformylation in a supercritical medium
Ard C.J. Koekena,∗, Leo J.P. van den Broekea,1, Berth-Jan Deelmanb,2, Jos T.F. Keurentjesa
a Process Development Group, Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven, The Netherlands
b Arkema Vlissingen B. V., P.O. Box 70, 4380 AB Vlissingen, The Netherlands
a r t i c l e i n f o
a b s t r a c t
Article history:
The kinetics of the hydroformylation of 1-octene in a supercritical carbon dioxide medium, catalyzed
by a tris(3,5-bis[trifluoromethyl]phenyl)phosphine-modified rhodium catalyst, have been investigated.
The influence of the concentration of carbon dioxide, reactants, catalyst precursors, and the reaction
temperature has been determined. A kinetic model was developed, which describes the concentration-
time profiles of the reactants, the linear and branched aldehydes, and the internal alkenes. Using the
kinetic model activation energies for hydroformylation of 1-octene to nonanal and 2-methyloctanal were
determined. Throughout the concentration ranges studied an approximate first order dependence of
the hydroformylation rate on the hydrogen and catalyst concentration was found which indicated that
oxidative addition of hydrogen was the rate limiting step. The increase in reaction rate and regioselectivity
with an increase in ligand concentration is a striking feature of the catalyst investigated here. At higher
concentrations the reaction rate was found to have a strong negative order dependence on the carbon
monoxide concentration. The reaction rate had a positive order in 1-octene at a concentration lower
than 0.5 mol L−1 while saturation kinetics were observed at a higher concentration. The results were
explained by invoking the contribution of both monophosphine and diphosphine rhodium species to the
hydroformylation catalysis.
Received 22 March 2011
Received in revised form 28 June 2011
Accepted 29 June 2011
Available online 7 July 2011
Keywords:
Hydroformylation
Phosphorus ligands
Rhodium
Supercritical fluids
Carbon dioxide
© 2011 Elsevier B.V. All rights reserved.
1. Introduction
because long-chain alkenes are too sparingly soluble in water to
obtain acceptable space-time yields [6].
Hydroformylation of alkenes using homogeneous catalysts is
applied on a commercial scale to make aldehydes that serve as
starting materials for the production of detergents, plasticizers,
and solvents [1–4]. Consequently, controlling the kinetics of hydro-
formylation reactions is a field of high current interest [4]. The
applied catalysts are complexes of rhodium or cobalt, commonly
with phosphines or phosphites as modifying ligands. In order to
reuse the catalyst, separation steps like distillation or extraction
are required, which could, in principle, have a detrimental effect
on catalyst activity and selectivity. The Ruhrchemie/Rhône-Poulenc
(RCH/RP) hydroformylation process is a well-known example
where, by means of an aqueous phase which preferentially dis-
solves the catalyst, the reaction and the separation step of the
homogeneous rhodium catalyst are integrated [5]. However, this
approach is limited to the hydroformylation of short-chain alkenes,
Supercritical fluids have received considerable attention as
alternative reaction media for the hydroformylation of long-chain
alkenes such as 1-octene [7]. Supercritical carbon dioxide (scCO2)
is of particular interest since it has accessible critical properties,
is nonflammable, is easily separated from the reaction product by
depressurization, and has a low toxicity. Furthermore, a single-
phase reaction system can be created with a high diffusivity of
the dissolved species and a high solubility of permanent gases like
carbon monoxide and hydrogen [8].
One of the restrictions of applying scCO2 as a solvent is the lim-
ited solubility of common homogeneous catalysts in this medium.
For hydroformylation, carbonylation and hydrogenation catalysts
this drawback has been primarily overcome by modifying ligands
with certain functional groups [9–31] or by using phosphines of low
molecular weight [32,33]. Recently, it was demonstrated that even
rhodium catalysts modified with the well-known triphenylphos-
phine or triphenyl phosphite ligand were applicable with high
efficiency in CO2-rich supercritical medium when high densities
were applied [34]. An alternative approach to using a single phase
supercritical medium is the application of a CO2 expanded organic
solvent. CO2 expanded solvents appear to allow for improved
hydroformylation selectivity in some cases [35,36]. Attaching per-
fluoroalkyl groups on the ligands of the catalyst appears to be the
most effective means to ensure a sufficient catalyst solubility in a
∗
Corresponding author. Present address: Inorganic Chemistry and Catalysis,
Utrecht University, Universiteitsweg 99, 3584 CG Utrecht, The Netherlands.
Tel.: +31 622736392.
E-mail addresses: a.c.j.koeken@uu.nl, ardkoeken@hotmail.com (A.C.J. Koeken),
berth-jan.deelman@arkema.com (B.-J. Deelman).
1
Present address: TNO Science and Industry, Separation Technology,
Schoemakerstraat 97, P.O. Box 6012, 2600 JA Delft, The Netherlands.
2
Fax: +31 113 612984.
1381-1169/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.molcata.2011.06.012

2
A.C.J. Koeken et al. / Journal of Molecular Catalysis A: Chemical 346 (2011) 1–11
involved in the reaction, n-octane (Aldrich, >99%), 2-octene (ABCR,
mixture of E and Z, 98%), and nonanal (Fluka, >95%) used for the
GC-analysis were used as received.
2.2. Hydroformylation in carbon dioxide
The details of the high pressure batch reactor setup are described
in Ref. [39]. A typical hydroformylation experiment was started
by charging the desired amounts of [Rh(CO)2acac] and the phos-
phine ligand into the empty reactor and subsequently closing the
reactor. Fig. 1 shows an example of typical conditions in terms of
temperature (a) and pressure (b) during an experiment.
The reactor volume was carefully filled with argon and subse-
quently evacuated for three times. Next, the stirring was switched
on with a stirring rate of 700 rpm and the desired amounts of car-
bon monoxide and hydrogen gas were charged to the reactor at
room temperature (see Fig. 1b: at t = −1.5 h up to about 7.4 MPa
in this case). The reactor content was heated to a temperature of
50 ◦C (Fig. 1a: at t = −1.4 h), and consecutively CO2 was charged into
the reactor at a constant flow typically up to a total pressure such
that about 14.5 mol L−1 CO2 was present (Fig. 1b: at t = −1.2 h). In
the case when 1 mol L−1 of CO, 1 mol L−1 of H2, and 14.5 mol L−1
of CO2 was applied this corresponded to about 26 MPa total pres-
sure at 50 ◦C. When the CO and H2 concentration were varied, the
total reactor pressure required to achieve a CO2 concentration of
14.5 mol L−1 at 50 ◦C was estimated using the Peng–Robinson equa-
tion of state and the binary interaction coefficients as previously
reported in Ref. [39]. A period of at least 0.5 h at reaction tem-
perature was taken for the in situ formation of the active catalyst
complexes from the [Rh(CO)2acac] and L1 (Fig. 1: t = −0.5–0 h). The
reaction was started by the addition of 1a, which was done by
opening the valve between the pump and the reactor. Fast pres-
sure equalization occurred and consecutively the desired volume
Scheme 1. Reaction scheme for the hydroformylation of 1-octene (1a), with the two
main products nonanal (2a) and 2-methyloctanal (2b). The side products are (E,Z)-
2-octene (1b, 1c), (E,Z)-3-octene (1d, 1e), (E,Z)-4-octene (1f, 1g), 2-ethylheptanal
(2c), 2-propylhexanal (2d), and n-octane (3).
CO2-rich reaction medium. The ‘tunability’ of the solvent properties
of scCO2 by relatively small changes in temperature and pressure
allows the precipitation and reuse of the perfluoroalkyl-substituted
catalysts [9,10]. Furthermore, catalyst activity and selectivity can
be enhanced by attachment of perfluoroalkyl groups on phosphine
ligands used in hydroformylation [37].
Although there is significant current interest in using
scCO2 as solvent for hydroformylation, detailed kinetic stud-
ies in single phase supercritical media are scarce. There is
a study by Erkey and co-workers which deals with tris(3,5-
bis[trifluoromethyl]phenyl)phosphine (L1) modified Rh [38]. In
that study relatively low temperatures (40–50 ◦C) and a relatively
low and constant initial pressure were used. Turnover frequencies
(TOF) in the range of 1000–2000 molaldehyde mol−1 h−1 and a negli-
gible isomerization of 1-alkene were observed. RThhe kinetic model
developed in that study did not include the regioselectivity of the
reaction nor did it include isomerization of the 1-alkene.
Using the same system as studied by Davis and Erkey [38] we
recently found for scCO2 that application of a higher temperature
(i.e. 70 ◦C instead of 50 ◦C) and a higher pressure (i.e. 40–50 MPa
instead of 28 MPa) resulted in considerably higher, and therefore
industrially more attractive TOF values [13,14]. Here, an extensive
description of the reaction kinetics of 1-octene (1a) hydroformy-
lation covering a wider range of conditions is presented, which
describes the overall reaction rate as well as the chemo- and
regioselectivity as a function of the concentration of reactants and
catalyst precursors (Scheme 1). The relation between the observed
kinetics and the catalytic cycle is discussed and the results are
compared to other kinetic studies of rhodium-catalyzed hydro-
formylation of linear 1-alkenes.
2. Experimental
2.1. Materials
Carbon dioxide, carbon monoxide, and hydrogen, grade 5.0,
4.7, and 5.0, respectively, were obtained from Hoekloos (The
Netherlands). Prior to use CO2 was passed over a Messer Oxisorb
filter to remove oxygen and moisture. 1-Octene, 1a, obtained
from Aldrich, was passed over activated alumina, dried with pre-
treated molsieves 3A (Aldrich, 4–8 mesh), and stored under argon.
The rhodium precursor, rhodium(I) dicarbonyl acetylacetonate,
[Rh(CO)2acac], was obtained in the form of dark green crystals from
Fluka. Ligand L1, tris(3,5-bis[trifluoromethyl]phenyl)phosphine, is
a white to light yellow solid and was supplied by Arkema (Vlissin-
gen, The Netherlands). All catalyst precursors were stored under
argon. The solvent toluene (Merck, analytical grade), the inter-
nal standard n-decane (Aldrich, >99% purity) and the substances
Fig. 1. (a) Typical temperature – time history during a hydroformylation experi-
ment; (b) corresponding pressure – time history.

A.C.J. Koeken et al. / Journal of Molecular Catalysis A: Chemical 346 (2011) 1–11
3
of 1a was pumped into the reactor, which generally did not take
more than 30 s. While pumping 1a into the reactor, the medium
temperature increased 1–3 ◦C as a result of a fast pressure increase
and the start of the exothermic reaction (Fig. 1: t ≈ 0 h). The tem-
perature stabilized within 15 min. During the remainder of reaction
the reactor temperature was maintained within a deviation of less
than 1 ◦C from the desired temperature.
of a line fitted through the conversion or yield data points up to a
conversion where there was a linear trend (typically up to a con-
version of 60%). A distinction is made between the linear aldehyde
product, 2a, and total amount of aldehydes, 2a–2d, abbreviated as
‘ald’. So, R1a,0, Rald,0, R2a,0, and R1b–g,0 are the rate of conversion of
1a, the rate of formation of aldehydes, the rate of formation of 2a,
the rate of isomerization of 1a, respectively.
Samples were withdrawn from the high pressure mixture at reg-
ular intervals into a calibrated volume of 0.182 mL. The ‘spikes’ in
pressure and temperature at regular time intervals are a result of
sampling. At typical reaction conditions the change in pressure as
a result of taking a sample was in between 0.1 and 0.2 MPa. The
content of the sample volume was carefully bubbled through a
vial with a solution of n-decane in toluene and afterwards rinsed
with additional toluene solution to collect residual 1a and its reac-
tion products quantitatively. Subsequently, the sample volume was
dried by alternately applying an argon flow and vacuum. The time
for taking a sample and preparing for a next one was in the order of
10 min. Generally, a minimum reaction time of 3 h was observed,
after which the reaction mixture was rapidly cooled, the gases were
vented and the remaining liquids consisting of reaction products
and catalyst were collected.
The TOF based on the formation rate of aldehydes was calculated
as follows:
Rald,0
3600 s
1 h
TOFald
=
×
(5)
nRh
where nRh is the amount of Rh in mol. The (cumulative) n:iso
ratio was obtained by dividing the concentration of linear aldehyde
product by the sum of the concentrations of the branched aldehyde
products:
C2a
n : iso =
(6)
C2b + C2c + C2d
The initial differential n:iso ratio, n:iso0, was estimated with the
following equation:
R2a,0
Rald,0 − R2a,0
n : iso0
=
(7)
To ensure that the reactor was cleaned properly, blank reaction
runs were performed regularly. The concentration of catalyst pre-
cursors were chosen such that catalytic complexes would dissolve
completely for the conditions applied here.
The initial differential selectivity for 2a was calculated with the
following equation:
R2a,0
R1a,0
S2a,0
=
(8)
2.3. Analysis and calibration
The samples were analyzed off-line using a Fisons Instruments
GC-FID equipped with a Restek Rtx-5 column (fused silica, length
30 m, internal diameter 0.53 mm) with helium as the carrier gas.
Calibration was done for 1a, E-2-octene and Z-2-octene (1b and 1c),
n-octane (3) and nonanal (2a), the response factors for the other
octene and aldehyde isomers were taken to be equal to those of 1a
and 2a, respectively.
2.5. Kinetic model
The concentration profiles of carbon monoxide and hydrogen
were determined from their initial concentration and the stoi-
chiometry of the reactions. The reaction network used for the
kinetic model (Scheme 2) was based on the reactions described in
Scheme 1. The main internal octene isomers are E-2-octene (1b) and
Z-2-octene (1c). The other octene isomers (1d–1g) were observed
to some extent in the GC analysis but in very small amounts. There-
fore, isomers 1b, 1c and the other octene isomers were lumped
together as 1bc. Similarly, the aldehyde isomers 2b, 2c, and 2d were
lumped together as 2bcd.
Besides hydroformylation, isomerization of 1a and hydrogena-
tion of 1a and its isomers are taken into account. The model allows
for a description of the n:iso ratio, because the formation of linear
(2a) and branched aldehydes (2b, 2c, 2d) are considered separately.
Catalyst deactivation was assumed to be negligible. Consequently,
the concentration of rhodium and ligand were considered to be
constant during the reaction. For a batch reaction in a fixed reactor
volume the mass balances for the reactants and products are given
as Eqs. (9)–(16):
2.4. Reaction parameters
To obtain normalized concentration profiles for 1a and its reac-
tion products, each concentration obtained by GC analysis, CGC,i
,
was divided by the sum of all obtained concentrations and multi-
plied by the concentration based on the total amount of 1a, n1a in
mol, injected and the reactor volume, Vreactor in L:
CGC,i
n1a
ꢀ
Ci =
×
(1)
Vreactor
iCGC,i
for i = 1a–1g, 2a–2d and 3.
The outcome of the hydroformylation was expressed in one
of the following parameters. The definitions used were based on
Westerterp et al. [40]. The conversion, X, was given by:
C1a,0 − C1a
dCCO
= −rhf,1 − rhf,2 − rhf,3
(9)
X =
× 100%
(2)
dt
dCH
dt
C1a,0
The subscript 0 indicates the normalized concentration at t = 0 h.
The overall selectivity, Si, towards a product i was defined as:
2
= −rhf,1 − rhf,2 − rhf,3 − rhg,1 − rhg,2
(10)
Ci
Si =
× 100%
(3)
C1a,0 − C1a
where Ci is the normalized concentration of a product i, with
i = 1b–1g, 2a–2d or 3.
The overall yield, Yi, for a product i was then:
Ci
Yi =
× 100%
(4)
C1a,0
The initial overall rate of reaction R0 (units mol s−1), were estimated
by multiplying the initial amount of 1a in mol, n1a,0, with the slope
Scheme 2. Reaction network. The rate expressions rhf,1, rhf,2, rhf,3, riso,f, riso,b, rhg,1
and rhg,2 are incorporated in the mass balances.
,

4
A.C.J. Koeken et al. / Journal of Molecular Catalysis A: Chemical 346 (2011) 1–11
dC1a
dt
= −rhf,1 − rhf,2 − rhg,1 − riso,f + riso,b
(11)
(12)
(13)
(14)
(15)
(16)
dC1bcd
dt
= −rhf,3 − rhg,2 + riso,f − riso,b
dC2a
dt
= rhf,1
dC2bcd
dt
= rhf,2 + rhf,3
dC3
dt
= rhg,1 + rhg,2
dCL1
dCRh
dt
=
= 0
dt
For all computations Matlab version 7.7.0.471 (R2008b) was
used. Using the built-in function script ‘lsqnonlin.m’ the so-called
objective function was minimized [41]. Vector F is the input for
‘lsqnonlin.m’ and is defined as follows:
F(p) = Data − Model(p)
(17)
The objective function is the square of F. ‘Data’ represents a matrix
containing the normalized concentration data obtained at the
respective sample times of all the experiments. ‘Model’ is a matrix,
which contains the calculated concentration values for CO to L1
of all experiments. The model concentration values were evalu-
ated by solving the mass balances containing the rate equations at
the respective experimental sample times. The mass balances were
numerically integrated using the built-in ODE solver ‘ode113.m’.
The rate equations contained the adjustable parameters repre-
sented by the vector p. The optimal values for p were calculated
using ‘lsqnonlin.m’. For the case presented here ‘lsqnonlin.m’ uses
a large scale trust-region reflective Newton method to calculate the
optimal set of parameters associated with the minimum value of
the square of the function defined in Eq. (17) [41]. To obtain an esti-
mate of the confidence intervals for the optimized values of p the
built-in Matlab function ‘nlparci.m’ was used.
Fig. 2. (a) Concentrations (calculated as explained in Section 2) of the main reac-
tants and products and (b) concentrations of isomers as a function of time for a
hydroformylation experiment with ligand L1. Experimental conditions: T = 70 ◦C,
CRh = 2.6 × 10−4 mol L−1, CL1 = 1.0 × 10−3 mol L−1, CCO = 15 mol L−1. Note that the
2
initial concentrations of 1a, CO and H2 at t = 0 h are based on the amounts of reac-
tants charged to the reactor and not determined by analysis of a sample. Fig. 1 gives
the temperature and pressure as a function of time for this experiment.
the n:iso ratio decreased slightly at a high conversion of 1a (reac-
tion time > 0.25 h). Hydrogenation of the product aldehydes into the
corresponding alcohols was not observed. Some hydrogenation of
1a was observed, but isomerization of 1a was the dominant side
reaction.
3. Results and discussion
3.1. General description of hydroformylation experiments
The reactions were carried out at temperatures ranging from 40
to 80 ◦C and pressures ranging from 24 to 51 MPa. A temperature of
70 ◦C was chosen to determine reaction kinetics in detail. At 70 ◦C
the main reaction, the direct formation of 2a and 2b from 1a, CO,
and H2 was essentially complete in 1 h with linear over branched
aldehyde ratios (n:iso) ranging between 2.3 and 3.7. Generally, the
sum of the concentrations of 1a and the reaction products deter-
mined by sampling from either the top or bottom of the reactor
was the same, indicating that the reaction took place under single
phase conditions (see Supplementary Information section A.1).
The concentrations of reactants and products as a function of
time for a typical hydroformylation experiment conducted in CO2
with L1 as the ligand are depicted in Fig. 2. The corresponding reac-
tion conditions are given in Fig. 1. The concentrations of CO and H2
after the start of the reaction were derived from the stoichiometry
of the reactions and the measured concentrations of the reaction
products. After 1 h of reaction the concentration of 2a was almost
constant. The catalyst derived from L1 gave rise to a significant
amount of octene isomers (1b–g), resulting in a maximum concen-
tration of 0.034 mol L−1 octene isomers after approximately 0.3 h
of reaction (Fig. 2b). It can be seen that after almost complete con-
version of 1a the concentration of 2a remained constant. As a result
of the hydroformylation of internal octene isomers (mainly 1b and
1c) the concentrations of 2b, 2c, and 2d continuously increased,
which is most clearly shown for 2c and 2d in Fig. 2b. As a result,
3.2. Influence of the CO2 amount
Three different amounts of CO2 were applied, namely 1.2, 1.4,
and 1.6 mol in a fixed reaction volume of 0.108 L, for the hydro-
formylation of 55 mmol of 1a. In Table 1 conditions and results
are shown for the corresponding experiments. Pmax is the pressure
reached when the alkene is injected at t = 0 h. In the case that 1.2 mol
of CO2 was used, a two-phase reaction mixture was observed dur-
ing the reaction. Significantly higher total concentrations, derived
from the sum of the concentrations of 1a and reaction products,
were determined by GC-analysis of samples taken from the bot-
tom part of the reactor [42]. Both the use of 1.4 and 1.6 mol of CO2
resulted in a single phase mixture throughout the course of the
reaction.
Using 1.2 mol of CO2 resulted in a significantly higher n:iso ratio
(3.8 after 3 h) and a reaction rate about 10% higher compared to
entries 2 and 3. The result obtained using 1.4 mol of CO2 (entry 2,
Table 1) is rather similar to the result obtained when 1.6 mol of
CO2 was used (entry 3, Table 1). The regioselectivity expressed in
n:iso3h and S2a,3h are about 3.3–3.4 and 77–78%, respectively. The
initial reaction rates found when using either 1.4 or 1.6 mol of CO2
are also very similar. The total drop in pressure as result of reaction
and sampling (ꢀP in Table 1) is of the same order of magnitude

A.C.J. Koeken et al. / Journal of Molecular Catalysis A: Chemical 346 (2011) 1–11
5
Table 1
Results of the hydroformylation at different pressures.a
c
Entry
nCO (mol)
Pmax (MPa)
ꢀPb (MPa)
R1a,0 (10−6 mol s−1
)
Rald,0 (10−6 mol s−1
)
TOFald (103 molald molRh h−1
)
S2a,3h (%)
n:iso0
n:iso3h
2
1d
2
3
1.2
1.4
1.6
24.0
30.5
40.2
7.1
9.5
12
60
54
54
56
51
50
7.6
6.5
6.6
78
77
76
4.2
3.8
3.6
3.8
3.4
3.3
a
General conditions: Vreactor = 0.1076 L, stirrer speed = 700 rpm, T = 70 ◦C, nCO = 108 mmol, nH = 108 mmol, n1a = 55 mmol, nRh = 27 ␮mol, nL1 = 1.35 mmol.
2
b
c
Total change in pressure as a result of reaction and sampling.
The n:iso ratio calculated with the initial rates, see Section 2.
Two-phase system.
d
as the difference in initial pressure for these experiments. It can
kinetic model which could fit all the concentration data. The kinetic
model can describe the concentrations of reactants and products as
a function of time using mass balances derived for a fixed volume
batch reactor:
thus be concluded that the change in CO2 concentration as well as
the absolute pressure drop has little effect on the kinetics of the
reaction when a CO2 amount in the range of 1.4–1.6 mol is used.
Further experiments to study the kinetics were carried out with
1.6 mol of CO2.
ꢁ
dCi
dt
=
ꢁijrj
(18)
3.3. Description of the reaction rates and selectivity with a kinetic
model
with Ci being the concentration of reactant or product i, rj the reac-
tion rate for a certain reaction step j, and ꢁij the stoichiometry of
step j with respect to reactant or product i. Applying this kinetic
model we tested several empirical rate equations (rj) in order to
get the best description of the experimental data using the reaction
network described in Scheme 2. Using this kinetic model the hydro-
formylation mechanism can then be discussed in more detail. For
each reactant generally three different initial concentration values
were chosen. Each experiment described in Table 2 corresponded to
about6–8datapointsperreactantorproductandthiswassufficient
to obtain an accurate kinetic model.
At a temperature of 70 ◦C and with a fixed amount of 1.6 mol
of CO2 the effect of varying the (initial) concentration of a reactant
or catalyst precursor was investigated. In Table 2 the conditions
and selected results on the initial rate and selectivity of the hydro-
formylation of 1a in CO2 rich mixtures are shown.
Generally, the initial differential n:iso ratio (n:iso0) was higher
than the n:iso ratio obtained after 3 h reaction (n:iso3h, see
Supplementary Information). This was a result of the hydroformy-
lation of internal octenes, mainly 2-octene, at a high 1a conversion.
In the initial stage of the reaction hydroformylation of internal
alkenes was not yet significant. It can be seen in Table 2 that in all
cases the initial rate of conversion of 1a (R1a,0) was higher than the
formation rate of aldehydes (Rald,0). Most of the 1a that was not con-
verted into aldehydes was isomerized to internal octenes (R1b–g,0).
The L1 modified Rh catalyst was highly active in hydroformylation
For the hydroformylation steps rhf,1 and rhf,2 (Scheme 2) empir-
ical rate equations of the form closely related to those proposed
by Chaudhari and co-workers were used [43,44]. These empirical
rate equations for rhf,1 and rhf,2 show resemblance to the rate equa-
tion derived from the catalytic cycle assuming oxidative addition of
hydrogen to the acyl-rhodium intermediate as the rate determin-
ing step [45]. An increase in the formation rate of 2a and the small
increase in isomerization activity of 1a (Table 2, R1b–g,0) with an
increase in ligand concentration was observed. Therefore, a term
that describes the weak dependence of the hydroformylation rate
(to form 2a) and the isomerization rate on the ligand concentra-
tion was added. Hydroformylation of internal octenes (rhf,3) was
with TOF values ranging from 3300 to 9300 molald molRh h−1
.
For each experiment in Table 2 concentrations of reactants and
products were determined as a function of time. Consequently, each
batch experiment gave considerable information with regard to the
reaction kinetics. In order to obtain more insight into the underlying
kinetic parameters of the individual reactions, we chose to use a
assumed to be first order in octene and CO. Isomerization (riso,f
)
Table 2
Overview of conditions and main results of the experiments.a The corresponding concentration profiles were the input for determining the optimal parameter values for the
kinetic model and are given in the supporting information.
b
c
c
c
c
d
e
Entry
pmax (MPa)
nCO (mmol)
nH (mmol)
n1a (mmol)
nRh (␮mol)
nL1 (␮mol)
R1a,0
Rald,0
R2a,0
R1b–g,0
TOFald
n:iso0 (–)
S2a,0 (%)
2
1f
2
3
4
5
6
7
8
9
10
11f
12f
13
14
15
40.6
39.0
44.5
38.2
44.0
34.5
52.3
40.2
39.5
40.2
40.7
40.2
49.2
50.2
50.8
107
81.5
162
109
109
109
109
108
109
109
105
108
109
108
108
107
108
111
86.5
166
110
110
111
109
110
108
108
108
108
108
55.2
54.4
55.0
55.0
54.7
12.1
109
55.4
54.8
55.1
55.3
54.2
105
27.3
27.7
29.3
26.6
26.6
27.3
26.6
13.8
52.5
13.8
27.2
27.2
53.2
53.8
28.1
110
107
110
107
107
107
107
55
214
104
273
1359
218
2709
1344
37
48
29
29
79
30
44
20
93
21
46
54
34
44
27
26
67
27
40
18
79
19
42
50
99
25
33
19
19
49
19
29
12
58
13
32
39
73
3.0
4.1
2.3
3.5
11
3.4
4.1
2.3
14
2.2
3.2
4.1
5.9
14
4.5
5.7
3.3
3.5
9.1
3.5
5.4
4.6
5.4
4.8
5.6
6.6
6.7
8.9
9.3
2.7
2.9
2.3
2.6
2.6
2.5
2.6
2.4
2.8
2.6
3.0
3.6
2.8
3.7
3.4
67
68
64
64
63
63
65
62
62
64
70
73
70
71
61
105
147
92
105
105
132
73
105
56
19
a
General conditions: T = 70 ◦C, Vreactor = 0.1076 L, nCO = 1.6 mol, stirrer speed = 700 rpm. Note that the amounts can be converted in concentration by dividing with the
2
reactor volume.
b
Maximum reactor pressure reached upon injection of 1a.
10−6 mol s−1
.
c
d
e
f
103 molald molRh h−1
.
Initial differential selectivity for 2a, see Section 2.
Average results of duplicate experiments.

6
A.C.J. Koeken et al. / Journal of Molecular Catalysis A: Chemical 346 (2011) 1–11
was taken as first order in 1a and rhodium. For triphenylphosphine
and diphosphine modified Rh catalysts a lower CO pressure results
in a higher proportion of isomerization and hydrogenation [45,46].
Therefore, a negative order ε in carbon monoxide was taken into
account for the isomerization reaction. Hydrogenation rates were
assumed to be first order in alkene, hydrogen, and rhodium. The
rate equations are given as Eqs. (19)–(24), which together with the
mass balances, Eqs. (9)–(16) (see Section 2.5), make up the kinetic
model:
all 16 parameters to minimize the objective function (Eq. (17), Sec-
tion 2). When all 16 parameters were included in the optimization
the lowest value for RSS was found. In case 2, values for parameters
˛, ˇ, ꢂ, ı, and ε were used based on the values used in case 1 but
rounded off to the closest half integer value. In this case the opti-
mized parameter values differ somewhat from those obtained in
case 1. However, the confidence intervals of the optimized values
in case 2 have considerable overlap with those of case 1. Also, the
RSS value for case 2 is only slightly higher than for case 1. In both
cases the values for khg,1 and khg,2 have a small significance, since
the corresponding confidence intervals are quite broad.
For case 1 in Table 3 ꢂ has a value of 2, which is the max-
imum value ꢂ can have according to the optimization routine
(Table A7 in Supplementary Information). In the case where ꢂ = 5
(Supplementary Information section A.5, Table A6, case 6) the con-
fidence intervals for both K1a and ꢂ were rather broad. Therefore,
a value for ꢂ higher than 2 was not expected to be meaningful,
because it would not result in a significantly enhanced kinetic
model.
khf,1CCOCˇ C1aCRıh
(1 + KCO,1CCO)˛(1 + K1aC1a
KL1,1CL1
H
2
rhf,1
=
(19)
(20)
ꢂ
ꢂ
(1 + KL1,1CL1
)
)
)
khf,2CCOCˇ C1aCRıh
H
2
rhf,2
=
(1 + KCO,2CCO)˛(1 + K1aC1a
rhf,3 = khf,3CCOCˇ C1bcCı
(21)
(22)
H
Rh
2
kiso,fC1aCRh
KL1,2CL1
riso,f
=
CCεO
(1 + KL1,2CL1
)
A model (case 3) where Eqs. (19) and (20) were replaced by
versions in which the term describing saturation in 1a is absent and
˛ is directly associated with the carbon monoxide concentration
(Eqs. (25) and (26)) was also tested. This model is similar to the one
proposed by Davis and Erkey [38] except for the ligand saturation
term that was kept in Eq. (25).
rhg,1 = khg,1CH C1aCRh
(23)
(24)
2
rhg,2 = khg,2CH C1bcCRh
2
where khf,1, khf,2, khf,3, kiso,f, khg,1, and khg,2 are the reaction rate
constants for the respective hydroformylation, isomerization, and
hydrogenation steps. C indicates the concentration of a reactant or
catalyst precursor. ˛, ˇ, ꢂ, ı, ε, KCO,1, KCO,2, K1a, KL1,1, KL1,2 are the
empirical kinetic parameters.
khf,1CCOCˇ C1ꢂaCı
KL1,1CL1
H
Rh
2
rhf,1
=
(25)
(26)
(1 + KCO,1C˛
)
(1 + KL1,1CL1
)
CO
khf,2CCOCˇ C1ꢂaCı
At 70 ◦C the Rh catalyst modified with L1 resulted in some iso-
merization of 1a into internal octenes. The formation of internal
octenes is thermodynamically favored relative to 1a, and the equi-
librium concentration of 1a will be very small [47]. Therefore, we
H
Rh
2
rhf,2
=
(1 + KCO,2C˛
)
CO
The data up to a conversion of 90% (39 samples corresponding
to 273 concentration values) were taken into account. Convergence
could not be reached when the optimization was tried with all
the data (121 samples corresponding to 847 concentration values).
Also, the optimization calculations took longer when Eqs. (25) and
(26) (on average 6 h) in the optimization were used instead of Eqs.
(19) and (20) (on average 0.5 h). Presumably, the computation of
have chosen to neglect the reverse process, i.e. riso,b = 0 mol s−1
.
In Table 3 the optimization results and the total residual sum
of square errors (RSS) for four different cases are listed. For cases 1
and 2 all data were used as input for the calculation. For cases 3 and
4 the data up to a conversion of 90% were used as input for the cal-
culation. Case 1 is the set of results which were obtained by varying
Table 3
Optimization results with 95% confidence intervals.
Parametera
Case 1b
Case 2b
Case 3c (rate equations (25) and (26))
Case 4c
d
d
d
khf,1 (L1+ˇ+ı mol−1−ˇ−ı s−1
khf,2 (L1+ˇ+ı mol−1−ˇ−ı s−1
)
)
)
754 393
94 42
2.17 0.57
6.2 2.5
949 166
111 27
2.7 0.5
6.04 0.52
4.03 0.53
0.97 0.04
2.15 0.13
4.7 2.4
2.5f
53 25g
8.7 5.6h
1.7 2.5i
31 15j
1319 310
95 53
4.7 3.2
7.1 0.8
3.9 1.1
0.95 0.04
2.11 0.13
2.2 1.1
2.5f
khf,3 (L1+ˇ+ı mol−1−ˇ−ı s−1
d
KCO,1 (L mol−1
KCO,2 (L mol−1
)
)
d
4.3 1.5
19 14j
K1a (L mol−1
)
0.95 0.68
2.06 0.13
4.3 1.9
2.50 0.19
1.02 0.05
2.0 1.0
0.97 0.02
0.88 0.15
0.34 0.05
0.035 0.022
0.28 0.11
0.1446
–
KL1,1 (103 L mol−1
)
)
2.22 0.15
4.6 4.6
2.07 0.07
0.91 0.06
0.45 0.02
0.97 0.02
1f
0.280 0.054
0.017 0.059
0.6 1.4
0.1121
KL1,2 (103 L mol−1
˛(–)
ˇ(–)
ꢂ (–)
ı (–)
ε (–)
1f
1f
2f
2f
1f
1f
1f
1f
kiso,f (L1−ε mol−1+ε s−1
)
0.31 0.02
0.036 0.022
0.29 0.11
0.1458
0.37 0.06
khg,1 (L2 mol−2 s−1
khg,2 (L2 mol−2 s−1
)
)
0
1
0.05
1
RSSe (mol2 L−2
)
0.1010
a
The parameters khf,1 to khg,2 are part of the vector p in the optimization function F.
b
c
d
e
f
Calculation using all data up to 100% conversion (121 samples corresponding to 847 concentration values).
Calculation using data up to 90% conversion (39 samples corresponding to 273 concentration values).
The units of khf,1, khf,2, khf,3, KCO,1, and KCO,2 are dependent on the values of ˛ (for case 3), ˇ,ꢂ (for case 3) and ı.
The residual sum of squares.
These parameters were at a fixed value and were not part of the optimization calculations.
g
h
i
Unit: L1+ˇ+ꢂ+ı mol−1−ˇ−ꢂ−ı s−1
Unit: L1+ˇ+ꢂ+ı mol−1−ˇ−ꢂ−ı s−1
.
.
Unit: L1+ˇ+ı mol−1−ˇ−ı s−1
.
Unit: L˛ mol−˛
.
j

A.C.J. Koeken et al. / Journal of Molecular Catalysis A: Chemical 346 (2011) 1–11
7
The values obtained for ˛, ˇ, ꢂ, and ı, correspond to some extent
to the values reported by Davis and Erkey who have found values of
1.61 0.23 for ˛, 0.83 0.05 for ˇ, 0.40 0.04 for ꢂ, and 0.94 0.08
for the order in catalyst at a ligand to Rh ratio of 3 to 1 [38]. The opti-
mization was also performed using rate equations (19)–(24) with
fixed values for ˛, ˇ, ꢂ, ı, and ε with the data up to a conversion of
90% (case 4, Table 3). The residual sum of square errors was con-
siderably smaller than obtained with rate equations (25) and (26).
Apparently, the kinetic model based on Eqs. (19) and (20) (cases
1, 2 and 4) gives a better description of the data than the kinetic
model with the rate equations proposed by Erkey and co-workers
(case 3).
In Fig. 3a the calculated concentration data with the optimized
parameters of case 1 in Table 3 are compared to the experimental
concentration data. Fig. 3a gives no indication of systematic errors
and, therefore, it can be concluded that the model gives a good rep-
resentation of the experimental data. In Fig. 3b a parity plot is used
to compare the experimentally determined initial reaction rates
(R1a,0, Rald,0, R2a,0 in Table 2) to the initial rates predicted with the
model using the same ‘linearized’ approach as used for the exper-
imental data. The initial rates calculated with the kinetic model
agree well with the experimentally determined rates. In the supple-
mentary information further comparisons between experimental
and predicted data are provided for cases 3 and 4 in Table 3.
In Fig. 4 concentration-time profiles are given for the experi-
ments in which the initial carbon monoxide concentration has been
varied. The lines represent the results obtained with the model.
Fig. 4 shows that the prediction based on the kinetic model with the
optimized values in Table 3 (case 1) gives a satisfactory description
of the concentration data of 1a, 2a, branched aldehydes and inter-
nal octenes. The model gives a good description of the influence of
carbon monoxide on the regioselectivity as well. This can be seen
in Fig. 4b and c, because the concentrations of 2a and 2b–2d as a
function of time are represented correctly.
Fig. 3. (a) Comparison of the experimental concentrations and the predicted con-
centrations obtained with the parameters of case 1 in Table 3. (b) Comparison of the
experimental rates and the predicted rates obtained with the parameters of case 1
in Table 3 using the same ‘linearized’ approach as used for the experimental data.
the differential equations becomes problematic with Eqs. (25) and
(26) when the concentrations of 1a and carbon monoxide approach
zero. In Table 3 results are given for this optimization using the
rate equations (25) and (26) together with rate equations (21)–(24)
(case 3).
3.4. Influence of temperature and the apparent activation energy
In Fig. 5 the aldehyde yield as a function of time is given for dif-
ferent reaction temperatures using a ligand amount of 0.11 mmol.
Fig. 4. Comparison of the experimentally obtained normalized concentration profile of 1a (a), 2a (b), 2b–2d (c), and 1b–1g (d) indicated by markers and the predicted model
concentration profiles indicated by the lines: [CO]0 = 0.76 mol L−1 – · – · – ·, [CO]0 = 1 mol L−1 – – –, [CO]0 = 1.5 mol L−1 ——.

8
A.C.J. Koeken et al. / Journal of Molecular Catalysis A: Chemical 346 (2011) 1–11
tive activation energies. The data point corresponding to 80 ◦C, at
3.4 × 10−4 mol J−1, appears to deviate from the linear trend when
considering the values of ln(khf,1) and ln(khf,2) at 40–70 ◦C. The devi-
ation of this data point is most likely due to the relatively low
sampling frequency in combination with the fast reaction, which
results in an underestimation of the initial reaction rate. In addi-
tion, the initial rate of isomerization to internal octenes is higher at
80 ◦C than at 70 ◦C and the formed internal octenes might inhibit the
hydroformylation of 1a (see Supplementary Information, Table A3).
The estimated activation energies, with 95% confidence intervals,
are 74 11 kJ mol−1 for the formation of 2a and 76 22 kJ mol−1 for
the formation of 2b, in the temperature range of 40–70 ◦C. These
values are in the range of values reported Rh-catalyzed hydro-
formylation of ethene, propene, and 1a [38,48,49]. The results on
khf,3, kiso,f, khg,1, and khg,2, in particular, at the low reaction tem-
peratures (≤60 ◦C) could not be determined accurately enough to
determine activation energies for the corresponding reaction steps.
This is evidently a result of the low isomerization and hydrogena-
tion activity at a temperature below 70 ◦C. In the Supplementary
Fig. 5. Comparison of the experimentally determined aldehyde yield (markers) and
the aldehyde yield predicted with the kinetic model (lines).
The initial rate of aldehyde formation increased with a factor of
2–2.5 with every 10 ◦C increase in reaction temperature in the
range of 40–70 ◦C. With a value between 2.7 and 2.9 the n:iso0
did not vary significantly with the temperature. However, the
initial rate of isomerization R1b–g,0, increased with temperature
(Table A3 in the supplementary information).
Information the optimized values for khf,1, khf,2, khf,3, kiso,f, khg,1
and khg,2 are given.
,
3.5. The catalytic cycle and comparison with literature
The solid lines in Fig. 5 are predictions based on the kinetic
In Scheme 3 the generally accepted mechanism of hydro-
formylation is depicted. It is based on the reaction mechanisms
suggested by Wilkinson and co-workers for the starting compound
HRhCO(PPh3)3 [50], and by Heck based on the starting compound
HCo(CO)4 [51]. The so-called dissociative pathway with sequence
C2–C4–C5–C6–C7–C8, or C2a–C4a–C5a–C6a–C7a–C8a, has been
generally accepted as the main mechanism for the hydroformy-
lation of alkenes. Starting from C1 or C3, the first step is the
dissociation of PR3 or CO, respectively, to form the reactive inter-
mediate C2. An alkene coordinates then to C2 resulting in complex
C4. Through migratory insertion of the alkene into the Rh–H bond
complex C5 is formed. Subsequently, CO coordinates to C5, which
results in the n-alkyl C6. A second migratory insertion but now of
CO into the Rh-alkyl bond then can take place to form C7 followed
by oxidative addition of dihydrogen to form C8 and reductive elim-
ination to form aldehyde product and the intermediate C2. This is
still a simplified scheme. For a more detailed discussion the reader
is referred to Refs. [3,4,45] and the references cited therein.
Besides the formation of linear aldehydes from the linear 1-
alkenes, a certain amount of branched aldehydes is formed. This
can be explained by 2,1-insertion of the alkene in C4 to form
the branched alkyl species C11 [52,53]. C11 can result in the for-
mation of 2b through CO insertion, oxidative addition of H2 and
reductive elimination of the branched aldehyde product similar to
the sequence C5–C6–C7–C8–C2. ␤-Hydrogen elimination from C11
results in 2-octene that can undergo further isomerization (through
reinsertion/␤-hydrogen elimination steps) or hydroformylation. In
general, the HRhCO(PPh3)3 system has a reasonably high selectivity
for the n-aldehyde which can be explained by the high preference
for 1,2-insertion of the 1-alkene into the Rh–H bond combined with
a low isomerization activity.
model, calculated with the optimized parameters khf,1, khf,2, khf,3
kiso,f, khg,1, and khg,2, and with fixed values for KCO,1, KCO,2, K1a
,
,
KL1,1, KL1,2 ˛, ˇ, ꢂ, ı and ε, as given in Table 3 (case 1). In princi-
ple, all 16 kinetic parameters presented for case 1 in Table 3 can be
temperature dependent. To obtain the best estimate for the temper-
ature dependency of all these parameters, reaction kinetics should
be measured at several temperatures. Since we do not have the
details of the reaction kinetics at different temperatures, we have
assumed that the reaction orders (˛, ˇ, ꢂ, ı, and ε) and the param-
eters KCO,1, KCO,2, K1a, KL1,1, KL1,2 are temperature independent. A
similar approach was taken by Davis and Erkey [38]. By doing so it
is possible to obtain the temperature dependence of khf,1 and khf,2
.
Subsequently, estimates for the apparent activation energies Eact,1
and Eact,2 can be made by means of the Arrhenius equation, Eq. (27):
/RT
khf,i = k0,i e−E
,
i = 1 for R2a,0
,
i = 2 for R2b,0
(27)
act,i
with k0,i being the pre-exponential factor, T the temperature, and
R the gas constant. In Fig. 6 the natural logarithm of khf,1 and khf,2
are given as a function of (R × T)−1. The data point corresponding
to 80 ◦C was not taken into account to determine the respec-
A number of empirical models for the reaction rate of the hydro-
formylation with Rh-catalysts have been proposed. Van Leeuwen
and co-workers discussed two types of simplified rate equations
(Eqs. (28) and (29)) as a starting point for the modeling of the
reaction kinetics [3,54,55]:
p1CalkeneCRh
RI =
(28)
(29)
p2 + CLigand
Fig. 6. Apparent activation energy for the formation of the linear (2a) and branched
aldehydes (2b–2d). The data point at 80 ◦C has not been used in the estimation of
the apparent activation energy. Every data point has error bars, but in some cases
these are not visible as a result of a relatively small error.
p3CH CRh
p4 +2CCO
RII
=

A.C.J. Koeken et al. / Journal of Molecular Catalysis A: Chemical 346 (2011) 1–11
9
Scheme 3. The catalytic cycles of the hydroformylation of 1-octene, 1a, to nonanal, 2a. R3P is the monodentate phosphine ligand L1 with R = –C6H3–3,5-(CF3)2. The Rh species
with R1 = –PR3 are designated with the code outside of the brackets. The Rh species with R1 = –CO are designated with the code inside the brackets.
where p1 to p4 are constants. The generally accepted rate equation
is RI (Eq. (28)), for which alkene coordination or the insertion of
the alkene into the Rh–H bond can be considered to be the rate-
determining step, i.e. the sequence C2–C4. Type I kinetics (Eq. (28))
is associated with Rh-catalysts modified with triphenylphosphine
[48,54]. Type II kinetics (Eq. (29)), which is considered to be less
common, has been observed for Rh-catalysts coordinated with CO
[56,57] and bulky phosphites, like tris(2-t-butyl-4-methyl-phenyl)
phosphite [54,58] but has also been observed for HRh(CO)(PPh3)3
[44]. For type II kinetics the rate-determining step is considered
to be the oxidative addition of dihydrogen, i.e. the step C7–C8.
Although CO is needed for step C5–C6 it is usually not rate limiting.
In practice the reaction kinetics of hydroformylation are often
more complicated than Eqs. (28) and (29) suggest. In several papers
a more extensive rate equation that also incorporates the saturation
Fig. 7. Reaction order as
tant or catalyst precursor based on the kinetic parameter values of
a
function of normalized amount of reac-
in alkene has been derived for the case where C7–C8 is the rate
determining step [44,59]:
case
1
in Table 3. Reference conditions: CCO = 1 mol L−1
,
CH = 1 mol L−1
,
2
C1a = 0.5 mol L−1
,
CRh = 2.5 × 10−4 mol L−1
,
and CL1 = 1 × 10−3 mol L−1
;
maxi-
v1CCOCH CalkeneCcatalyst
= 1.5 mol L−1
, ,
C1a,max = 1.1 mol L−1
1 + v2CCO + v3CCOCalkene2 + v4C2 Calkene + v5C3 Calkene
mum amounts: CCO,max = 1.5 mol L−1
,
CH
R =
(30)
,max
2
CRh,max = 5.5 × 10−4 mol L−1, CL1,max = 2.5 × 10−2 mol L−1
.
CO
CO
where v1 to v5 are the constants in which the rate coefficients and
equilibrium constants of the reaction steps in the catalytic cycle
are lumped together. The empirical rate equation we have found
resembles this theoretically derived rate equation. In addition to
a first order in dihydrogen, it also includes the saturation kinetics
in CO, alkene and ligand. A possible higher order inhibition by CO
has been explained by the formation of di- and tricarbonyl species
such as C1, C1a, C9, and C9a that are all outside of the catalytic
cycle. The saturation or even inhibition in alkene at higher alkene
concentrations is generally explained by the formation of alkene
complexes of rhodium alkyl species that are also outside of the
catalytic cycle (not shown in Scheme 3) [60].
the rate equations. The reaction orders in CO, 1a, and L1 (ꢃi), for
the primary reactions, the formation of 2a and 2b from 1a, were
evaluated using the following equation:
d ln(rhf,1 + rhf,2
d ln Ci
)
ꢃi =
(31)
where Ci is the concentration of component i (CO, H2, 1a, Rh, and
L1). Eq. (31) represents the slope of the curve obtained by plot-
ting the logarithm of the rate (rhf,1 + rhf,2, Eqs. (19) and (20)) as a
function of the logarithm of reactant concentration. This is the dif-
ferential method to evaluate reaction orders [61]. See Section A.5
To further support the discussion, the reaction orders in carbon
monoxide, hydrogen, 1a, rhodium, and ligand are plotted as a func-
tion of their respective normalized concentrations in Fig. 7. When
considering the rate equations for rhf,1 and rhf,2 it is clear that ˇ
and ı directly represent the orders in H2 and Rh, respectively. For
CO, 1a, and L1 the reaction order cannot be readily discerned from
in the Supplementary Information for formulas defining ꢃCO, ꢃ1a
and ꢃL1
The reaction order in 1a changes from 1 to zero and even
becomes slightly negative when C1a increases. The order in CO,
which initially is close to zero, develops into a strong nega-
tive order (ca. −1.25) at higher CO concentrations. The orders in
,
.

10
A.C.J. Koeken et al. / Journal of Molecular Catalysis A: Chemical 346 (2011) 1–11
hydrogen and rhodium are approximately one in the range of con-
centrations investigated. Saturation kinetics are observed in ligand
concentration, going from an order of 0.33 at a ligand concentra-
tion of 1 × 10−3 mol L−1 to essentially zero at a concentration of
phine complexes that lack the steric crowding required for a more
selective alkene insertion into the Rh–H bond.
A pronounced change in reaction rate and selectivity was found
when the CO2 amount was decreased to 1.2 mol (entry 3, Table 1).
A similar observation was made in the hydroformylation of 1-
hexene in the presence of a Rh-catalyst by Cole-Hamilton and
co-workers when they reached biphasic conditions [33]. It is plau-
sible that the higher reaction rate and regioselectivity observed
using 1.2 mol of CO2 were the result of the two-phase reaction con-
ditions. In the two-phase reaction system a higher local Rh and
L1 concentration in the lower 1a- and product-rich phase can be
expected.
Although the unique ligand properties of L1 seem to be responsi-
ble for the effects observed, without comparative studies with other
solvents it is not possible to assess the exact solvent effect of carbon
dioxide on the hydroformylation kinetics. Earlier studies indicated
that results obtained in scCO2 could be compared to those obtained
in hexane or toluene but in this case much lower and constant par-
tial pressures (0.5 MPa) of CO and H2 were used [13] resulting in
significantly lower concentrations of these reagents (0.065 mol L−1
CO and 0.033 mol L−1 H2 in hexane [67,68]). The fact that similar
TOF values were found in this study as for the hydroformylation in
hexane in a previous study [13] may be due to the opposite effects
of H2 and CO pressure on the rate of hydroformylation.
2.5 × 10−2 mol L−1
.
All in all, in particular at higher 1a and CO concentration the
kinetics of the Rh/L1 system show resemblance with the type II
kinetics (Eq. (29)). The positive effect of hydrogen concentration
could in principle also be caused by the equilibrium between C1
and C10 [44,62]. However, this is not likely, because the catalyst is
formed in the presence of 1 mol L−1 hydrogen and 1 mol L−1 carbon
monoxide. These concentrations of hydrogen and carbon monox-
ide are considerably higher than commonly applied in experiments
using organic solvents. In earlier spectroscopic studies under lower
H2 and CO concentrations than applied here it was already found
that HRh(L1)3CO is converted predominantly into HRh(L1)2CO and
that dimeric species such as C10 and C10a are not formed in sig-
nificant amounts [63]. Also in the case of formation of dimeric Rh
species one would not expect a clean first order in hydrogen as is
found here. Type II kinetics is mostly associated with the use of a
rhodium catalyst modified with electron-withdrawing and steri-
cally hindered ligands like bulky phosphites. It is known that the
application of electron-withdrawing ligands decelerates the oxida-
tive addition of hydrogen [3]. L1 contains six electron-withdrawing
trifluoromethyl groups, so, L1 is considerably less basic and more
sterically hindered than triphenylphosphine [64,65]. Therefore, it
is plausible for L1 that oxidative addition of hydrogen is the rate
determining step under conditions where the olefin concentration
is high.
For a rhodium catalyst modified with triphenylphosphine an
increase in concentration of species C3, which has three coordi-
nated phosphine ligands, is the main cause of a decrease in reaction
rate when the triphenylphosphine concentration is increased
[1,3,48]. The observation for L1 that the reaction rates increase with
an increase in ligand concentration, i.e. an increase in L1:Rh ratio, is
in sharp contrast with what is observed for triphenylphosphine–Rh
systems. The origin of the observed saturation kinetics rather than
the expected inhibition in L1 might be that C3 is not kinetically rele-
vant under the reaction conditions [63]. The latter may be explained
by the low basicity of L1 although steric consequences of the bis-
meta substitution, in terms of cone angle for example, may play a
role too [64,65]. The absence of C3 even at high ligand concentration
was also reported by Moser et al. for rhodium catalysts modified
with para-substituted triarylphosphines [66].
4. Conclusion
The high activity of the Rh/L1 catalytic system for the hydro-
formylation of 1-octene in scCO2 at 40–70 ◦C is a result of the unique
electronic and steric properties of L1 which cause diphosphine and
very likely also monophosphine Rh species to contribute to the
catalysis. The current work shows that, despite of the rather exotic
but environmentally advantageous solvent system used (scCO2),
the kinetics can still be discussed within the classical framework
originally proposed by Wilkinson and co-workers for the system
HRh(CO)(PPh3)3 [62].
The kinetic model derived here is complementary and more
broadly applicable than previously reported models. The rate of
formation of linear and branched aldehydes, inhibition by CO, iso-
merization, the effect of alkene and ligand were all specifically
incorporated in one kinetic model. It could be shown that oxidative
addition of hydrogen is the rate limiting step. The observed kinet-
ics and the corresponding mathematical model presented here are
expected to be of relevance for other rhodium catalysts modified
with electron-withdrawing phosphine ligands.
It can be assumed that under the reaction conditions employed
in this study a part of the catalysis involves rhodium species
with only one L1 ligand, and that hydroformylation through the
‘monophosphine cycle’ (C2a–C4a–C5a–C6a–C7a–C8a) plays a role.
This is implied by the clear enhancement of selectivity with an
increase in L1 concentration. An increase in the L1 concentration
would shift the equilibria to such an extent that the cycle starting
with C2 becomes dominant over the one starting with C2a. This
explanation is in line with the lower basicity of ligand L1 com-
pared to PPh3 and with spectroscopic studies recently obtained
by Haji and Erkey et al. who observed Rh(acyl)(CO)3(L1), a species
corresponding to C9a [63].
Acknowledgements
This work was supported by an EET grant (EETK 01115) from
SenterNovem (agency of the Dutch Ministry of Economic Affairs).
The contributions of Stefan de Bakker and Paul van Bavel to the
experimental work are gratefully acknowledged.
Appendix A. Supplementary data
The differential selectivity values based on the initial rate mea-
surement for 2a lie in between 61% and 73% (Table 2) and are
moderate [34]. Nevertheless, the n:iso ratios observed for the Rh/L1
catalyst are comparable to those for the PPh3-modified rhodium
catalyst. The lower differential selectivity for 2a is clearly due to the
higher proportion of internal alkenes formed that are only partially
hydroformylated to branched aldehydes [14,34]. The high isomer-
ization activity of the L1-modified Rh system is most probably also
related to the electron-withdrawing ligand system employed [3]
and could be another indication of the involvement of monophos-
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.molcata.2011.06.012.
References
[1] M. Beller, B. Cornils, C.D. Frohning, C.W. Kohlpaintner, J. Mol. Catal. A 104 (1995)
17–85.
[2] P.W.N.M. van Leeuwen, Homogeneous Catalysis, Understanding the Art
Kluwer Academic Publishers, Dordrecht, 2004, pp. 125–138.
,
[3] P.W.N.M. van Leeuwen, Homogeneous Catalysis, Understanding the Art
Kluwer Academic Publishers, Dordrecht, 2004, pp. 139–174.
,

A.C.J. Koeken et al. / Journal of Molecular Catalysis A: Chemical 346 (2011) 1–11
11
[4] M. Beller, J. Seayad, A. Tillack, H. Jiao, Angew. Chem. Int. Ed. 43 (2004)
3368–3398.
[5] C.W. Kohlpaintner, R.W. Fischer, B. Cornils, Appl. Catal. A 221 (2001) 219–225.
[6] B. Cornils, in: B. Cornils, W.A. Herrmann, I.T. Horváth, W. Leitner, S. Mecking,
H. Olivier-Bourbigou, D. Vogt (Eds.), Multiphase Homogeneous Catalysis, vol.
1, Wiley-VCH, 2005, pp. 27–39;
H. Klein, R. Jackstell, M. Beller, Chem. Commun. 17 (2005) 2283–2285.
[7] (a) S. Bektesevic, A.M. Kleman, A.E. Marteel-Parrish, M.A. Abraham, J. Supercrit.
Fluids 38 (2006) 232–241;
[37] For example: J.D. Unruh, J.R. Christenson, J. Mol. Catal. 14 (1982) 19–34;
H. Klein, R. Jackstell, K.D. Wiese, C. Borgmann, M. Beller, Angew. Chem. Int. Ed.
40 (2001) 3408–3411.
[38] T. Davis, C. Erkey, Ind. Eng. Chem. Res. 39 (2000) 3671–3678.
[39] A.C.J. Koeken, S.J.M. de Bakker, H.M. Costerus, L.J.P. van den Broeke, B.-J. Deel-
man, J.T.F. Keurentjes, J. Supercrit. Fluids 46 (2008) 47–56.
[40] K.R. Westerterp, W.P.M. van Swaaij, A.A.C.M. Beenackers, Chemical Reactor
Design and Operation , 2nd ed., John Wiley & Sons, Chichester, 1984.
[41] The Matworks website: http://www.mathworks.com (last accessed 17.01.11).
[42] Concentrations which are more than 20% higher than the concentration based
on the amount of 1a injected and the volume of the reactor. See also the
Supplementary Information.
[43] R.M. Deshpande, R.V. Chaudhari, Ind. Eng. Chem. Res. 27 (1988) 1996–2002.
[44] S.S. Divekar, R.M. Deshpande, R.V. Chaudhari, Catal. Lett. 21 (1993)
191–200.
[45] C.D. Frohning, C.W. Kohlpaintner, H.-W. Bohnen, in: B. Cornils, W.A. Herrmann
(Eds.), Applied Homogeneous Catalysis with Organometallic Compounds, vol.
1, 2nd ed., Wiley-VCH, Weinheim, 2002, pp. 31–103.
[46] R.P.J. Bronger, J.P. Bermon, J. Herwig, P.C.J. Kamer, P.W.N.M. van Leeuwen, Adv.
Synth. Catal. 346 (2004) 789–799.
[47] T.C. Morrill, C.A. D’Souza, Organometallics 22 (2003) 1626–1629.
[48] P. Cavalieri d’Oro, L. Raimondi, G. Pagani, G. Montrasi, G. Gregório, A. Andreetta,
Chim. Ind. 62 (1980) 572–579.
[49] G. Kiss, E.J. Mozeleski, K.C. Nadler, E. VanDriessche, C. DeRoover, J. Mol. Catal.
A 138 (1999) 155–176.
[50] D. Evans, J.A. Osborn, G. Wilkinson, J. Chem. Soc. A: Inorg. Phys. Theor. 12 (1968)
3133–3142.
(b) P.G. Jessop, J. Supercrit. Fluids 38 (2006) 211–231.
[8] P.G. Jessop, T. Ikariya, R. Noyori, Science 269 (1995) 1065–1069.
[9] S. Kainz, D. Koch, W. Baumann, W. Leitner, Angew. Chem. Int. Ed. Engl. 36 (1997)
1628–1630.
[10] D.R. Palo, C. Erkey, Ind. Eng. Chem. Res. 37 (1998) 4203–4206.
[11] (a) L.J.P. van den Broeke, E.L.V. Goetheer, A.W. Verkerk, E. de Wolf, B.J. Deelman,
G. van Koten, J.T.F. Keurentjes, Angew. Chem. Int. Ed. 40 (2001) 4473–4474;
(b) E.L.V. Goetheer, A.W. Verkerk, L.J.P. van den Broeke, E. de Wolf, B.-J. Deelman,
G. van Koten, J.T.F. Keurentjes, J. Catal 219 (2003) 126–133.
[12] A.M.B. Osuna, W.P. Chen, E.G. Hope, R.D.W. Kemmitt, D.R. Paige, A.M. Stuart,
J.L. Xiao, L.J. Xu, J. Chem. Soc. Dalton Trans. 22 (2000) 4052–4055.
[13] A.C.J. Koeken, M.C.A. van Vliet, L.J.P. van den Broeke, B.J. Deelman, J.T.F. Keuren-
tjes, Adv. Synth. Catal. 348 (2006) 1553–1559.
[14] A.C.J. Koeken, M.C.A. van Vliet, L.J.P. van den Broeke, B.J. Deelman, J.T.F. Keuren-
tjes, Adv. Synth. Catal. 350 (2008) 179–188.
[15] D. Koch, W. Leitner, J. Am. Chem. Soc. 120 (1998) 13398–13404.
[16] D.R. Palo, C. Erkey, Organometallics 19 (2000) 81–86.
[17] K.-D. Wagner, N. Dahmen, E. Dinjus, J. Chem. Eng. Data 45 (2000) 672–677.
[18] Y. Shimoyama, M. Sonoda, K. Miyazaki, H. Higashi, Y. Iwai, Y. Arai, J. Supercrit.
Fluids 44 (2008) 266–272.
[19] C.T. Estorach, A. Orejón, A.M. Masdeu-Bultó, Green Chem. 10 (2008) 545–552.
[20] D. Bonafoux, Z.H. Hua, B.H. Wang, I. Ojima, J. Fluorine Chem. 112 (2001)
101–108.
[21] S. Kainz, A. Brinkmann, W. Leitner, A. Pfaltz, J. Am. Chem. Soc. 121 (1999)
6421–6429.
[22] Z.K. Lopez-Castillo, R. Flores, I. Kani, J.P. Fackler, A. Akgerman, Ind. Eng. Chem.
Res. 42 (2003) 3893–3899.
[23] G. Franciò, K. Wittmann, W. Leitner, J. Organomet. Chem. 621 (2001) 130–142.
[24] W. Leitner, Nature 405 (2000) 129–130.
[25] Y.L. Hu, W.P. Chen, L.J. Xu, J.L. Xiao, Organometallics 20 (2001) 3206–3208.
[26] C.D. Ablan, D. Sheppard, E.J. Beckman, M.M. Olmstead, P.G. Jessop, Green Chem.
7 (2005) 590–594.
[51] R.F. Heck, Acc. Chem. Res. 2 (1969) 10–16.
[52] W.A. Herrmann, M. Prinz, in: B. Cornils, W.A. Herrmann (Eds.), Applied Homo-
geneous Catalysis with Organometallic Compounds, vol. 3, 2nd ed., Wiley-VCH,
Weinheim, 2002, pp. 1119–1130.
[53] L.A. Oro, D. Carmona, in: J.G. de Vries, C.J. Elsevier (Eds.), The Handbook of
Homogeneous Hydrogenation, Wiley-VCH, Weinheim, 2007, pp. 3–30.
[54] A. van Rooy, J.N.H. de Bruijn, K.F. Roobeek, P.C.J. Kamer, P.W.N.M. van Leeuwen,
J. Organomet. Chem. 507 (1996) 69–73.
[55] S.C. van der Slot, P.C.J. Kamer, P.W.N.M. van Leeuwen, J.A. Iggo, B.T. Heaton,
Organometallics 20 (2001) 430–441.
[56] G.W. Liu, R. Volken, M. Garland, Organometallics 18 (1999) 3429–3436.
[57] J.H. Feng, M. Garland, Organometallics 18 (1999) 417–427.
[58] A. van Rooy, E.N. Orij, P.C.J. Kamer, P.W.N.M. van Leeuwen, Organometallics 14
(1995) 34–43.
[27] A. Galia, A. Cipollina, G. Filardo, O. Scialdone, M. Ferreira, E. Monflier, J. Supercrit.
Fluids 46 (2008) 63–70.
[59] M. Rosales, A. Gonzalez, Y. Guerrero, I. Pacheco, R.A. Sanchez-Delgado, J. Mol.
Catal. A 270 (2007) 241–249.
[28] M.F. Sellin, D.J. Cole-Hamilton, J. Chem. Soc. Dalton Trans. 11 (2000) 1681–1683.
[29] J.J.M. de Pater, C.E.P. Maljaars, E. de Wolf, M. Lutz, A.L. Spek, B.-J. Deelman, C.J.
Elsevier, G. van Koten, Organometallics 24 (2005) 5299–5310.
[30] J.J.M. de Pater, D.S. Tromp, D.M. Tooke, A.L. Spek, B.-J. Deelman, G. van Koten,
C.J. Elsevier, Organometallics 24 (2005) 6411–6419.
[60] R.M. Deshpande, S.S. Divekar, B.M. Bhanage, R.V. Chaudhari, J. Mol. Catal. 77
(1992), L13–L17.
[61] O. Levenspiel, Chemical Reaction Engineering , 3rd ed., John Wiley & Sons,
Chichester, 1999, pp. 38–82.
[62] C.K. Brown, G. Wilkinson, J. Chem. Soc. A: Inorg. Phys. Theor. 17 (1970)
2753–2764.
[63] S. Haji, C. Erkey, Tetrahedron 58 (2002) 3929–3941.
[64] C.A. Tolman, Chem. Rev. 77 (1977) 313–348.
[65] K.K. Banger, A.K. Brisdon, C.J. Herbert, H. Ali Ghaba, I.S. Tidmarsh, J. Fluorine
Chem. 130 (2009) 1117–1129.
[66] W.R. Moser, C.J. Papile, D.A. Brannon, R.A. Duwell, S.J. Weininger, J. Mol. Catal.
41 (1987) 271–292.
[31] J.J.M. de Pater, B.-J. Deelman, C.J. Elsevier, G. van Koten, Adv. Synth. Catal. 348
(2006) 1447–1458, and references cited therein.
[32] I. Bach, D.J. Cole-Hamilton, Chem. Commun. 14 (1998) 1463–1464.
[33] M.F. Sellin, I. Bach, J.M. Webster, F. Montilla, V. Rosa, T. Aviles, M. Poliakoff, D.J.
Cole-Hamilton, J. Chem. Soc. Dalton Trans. 24 (2002) 4569–4576.
[34] A.C.J. Koeken, N.E. Benes, L.J.P. van den Broeke, J.T.F. Keurentjes, Adv. Synth.
Catal. 351 (2009) 1442–1450.
[35] P.G. Jessop, B. Subramaniam, Chem. Rev. 107 (2007) 2666–2694.
[36] M.J. Muldoon, Dalton Trans. 39 (2010) 337–348.
[67] W. Gao, R.L. Robinson Jr., K.A.M. Gasem, J. Chem. Eng. Data 46 (2001) 609–612.
[68] R. Koelliker, H. Thies, J. Chem. Eng. Data 38 (1993) 437–440.