Kinetic explanation for the temperature dependence of the regioselectivity in the hydroformylation of neohexene

Article abstract of DOI:10.1002/cctc.201300818

The kinetics of Rh-catalyzed neohexene hydroformylation were investigated with the bulky monodentate ligand tris(2,4-di-tert-butylphenyl)phosphite. The hydrogenolysis of the Rh-acyl intermediate was identified as the rate-limiting step for both the linear and the branched aldehydes. Rate equations for both aldehydes were derived and kinetic parameters were estimated. Increased aldehyde linearity at higher temperatures, frequently observed in hydroformylation, was elucidated by deuterioformylation experiments. These showed that at 100 °C the formation of linear Rh-alkyl was more reversible than the formation of the branched derivative. The ratio of linear to branched Rh-acyl species was determined by in situ high-pressure IR spectroscopy experiments, which allowed the difference in the activation energies for the hydrogenolysis steps towards the aldehyde isomers to be quantified. The hydrogenolysis of Rh-acyl was found to be the step that caused the greatest temperature effect on the regioselectivity. Kinetics arouses curiosity: Studying the kinetics of neohexene hydroformylation catalyzed by a bulky-phosphite-modified Rh catalyst brings up the question: "What is the reason behind the temperature dependence of regioselectivity?" We answer this question by using mechanistic tools such as deuterium labeling and in situ IR spectroscopy. The hydrogenolysis step of the catalytic cycle seems to be the main step that is responsible. Copyright

Full text of DOI:10.1002/cctc.201300818

CHEMCATCHEM
FULL PAPERS
DOI: 10.1002/cctc.201300818
Kinetic Explanation for the Temperature Dependence of
the Regioselectivity in the Hydroformylation of Neohexene
[
a]
[a]
[b]
[b]
Sabriye Gꢀven, Marko M. L. Nieuwenhuizen, Bart Hamers, Robert Franke,
[
b]
[b]
[a, c]
Markus Priske, Marc Becker, and Dieter Vogt*
The kinetics of Rh-catalyzed neohexene hydroformylation were
investigated with the bulky monodentate ligand tris(2,4-di-tert-
butylphenyl)phosphite. The hydrogenolysis of the Rh–acyl in-
termediate was identified as the rate-limiting step for both the
linear and the branched aldehydes. Rate equations for both al-
dehydes were derived and kinetic parameters were estimated.
Increased aldehyde linearity at higher temperatures, frequently
observed in hydroformylation, was elucidated by deuteriofor-
mylation experiments. These showed that at 1008C the forma-
tion of linear Rh–alkyl was more reversible than the formation
of the branched derivative. The ratio of linear to branched Rh–
acyl species was determined by in situ high-pressure IR spec-
troscopy experiments, which allowed the difference in the acti-
vation energies for the hydrogenolysis steps towards the alde-
hyde isomers to be quantified. The hydrogenolysis of Rh–acyl
was found to be the step that caused the greatest tempera-
ture effect on the regioselectivity.
Introduction
The hydroformylation of alkenes, discovered in 1938
by Otto Roelen, has become one of the most impor-
tant homogenously catalyzed reactions in industry
[
1]
and has been studied extensively. Nowadays, rhodi-
um is most commonly applied as the catalyst metal,
and a great variety of phosphorus ligands are avail-
able to tune the catalyst for activity and selectivity.
Knowledge about the influence of their steric and
electronic properties usually allows good prediction
of the expected activity and selectivity of the catalyst
system. Bulky monodentate p-accepting ligands give
rise to extremely active but less regioselective cata-
lysts, whereas chelating ligands possessing a large
bite angle allow for very high selectivity towards the
[
2]
linear aldehyde product. However, knowledge of
the catalyst itself is usually not enough for the pre-
diction of reactivity; different elementary steps in the
mechanism can become rate limiting for different
Scheme 1. Mechanism of Rh-catalyzed hydroformylation when using bulky monophos-
phites.
[
3]
substrates, even for isomers, depending on the position of the
[3,4]
double bond in the molecule.
Regioselectivity in Rh-catalyzed hydroformylation is a widely
[
a] S. Gꢀven, M. M. L. Nieuwenhuizen, Prof. Dr. D. Vogt
Faculteit Scheikundige Technologie
Technische Universiteit Eindhoven
P.O. Box 513, 5600 MB Eindhoven (The Netherlands)
Fax: (+44)131-650-6453
[5–8]
[9]
studied subject,
owing to its industrial relevance, and the
linear aldehyde is typically the desired product in large-scale
[1]
applications. In the literature, the formation of Rh–alkyl in the
reaction mechanism (Scheme 1) is generally referred to as the
step that determines the regioselectivity, especially at low tem-
peratures, for which the formation of Rh–alkyl is not reversi-
E-mail: d.vogt@ed.ac.uk
[
b] Dr. B. Hamers, Prof. Dr. R. Franke, Dr. M. Priske, Dr. M. Becker
Evonik Oxeno GmbH
Paul-Baumann-Str. 1, 45772 Marl (Germany)
[5]
ble. At higher temperatures, the higher reversibility of the
[
c] Prof. Dr. D. Vogt
School of Chemistry
University of Edinburgh
branched Rh–alkyl formation step has been shown to favor the
[2,10,11]
linear product.
However, a detailed in situ kinetic study of
the rate-limiting Rh–acyl hydrogenolysis step revealed a differ-
ence in energetics for the linear and branched isomers for sty-
West Mains Road Edinburgh EH9 3JJ Scotland (United Kingdom)
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cctc.201300818.
[6]
rene hydroformylation. Finally, it is well reported that hydro-
ꢁ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 603 – 610 603

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
formylation regioselectivity is governed by a combination of
different steps in the mechanism and is dependent on reaction
Results and Discussion
[
7,12]
conditions, catalytic system, and substrate.
Kinetics
For the design and operation of a jet-loop reactor with inte-
grated membrane filtration for continuous homogeneously
catalyzed hydroformylation, we needed accurate kinetic data
for a number of model substrates, such as neohexene, and for
Neohexene hydroformylation produces two isomeric alde-
hydes, 4,4-dimethylpentanal and 2,3,3-trimethylbutanal, as
shown in Scheme 2, which will be referred to as the linear (l)
and branched (b) aldehydes, respectively.
[
13]
a wider range of operating conditions. While examining the
kinetics of neohexene hydroformylation, we came across some
inconsistencies that could not be explained by the common in-
terpretation. Aroused by curiosity, we decided to perform an
in-depth study.
Bulky phosphites are reported to result in monocoordinated
[15–17]
active Rh species.
In this study, we used a high molar
excess of the ligand (L/Rh varying from 15 to 30) to ensure the
complete coordination of all Rh centers and to avoid observing
mixed kinetics of ligated and unmodified Rh. A mechanistic
study performed by van Leeuwen et al. reports that the IR
spectra of the Rh–H species contained three carbonyl bands if
an excess of L/Rh=15 was applied, which points to the pres-
The kinetics of neohexene hydroformylation (Scheme 2)
were previously investigated for the tris(2-tert-butyl-4-methyl-
[
14]
phenyl)phosphite-modified rhodium catalyst.
Van Leeuwen
et al. found a first-order dependence of the reaction rate on
alkene concentration and hence proposed alkene coordination
[16]
ence of monoligated species. The hydroformylation mecha-
nism proposed for the bulky phosphite monoligated Rh cata-
[3]
lyst is given in Scheme 1. According to this mechanism and
[15]
the findings in the literature, we derived the following rate
r) equations [Equations (1) and (2)] for linear and branched al-
(
dehydes by assuming that the hydrogenolysis step was rate
limiting for both the linear and branched aldehydes (detailed
derivation of the equations is given in the Supporting Informa-
tion).
Scheme 2. Neohexene hydroformylation.
to be the rate-limiting step. Under the reaction conditions
studied, they did not observe any branched product and con-
cluded that the steric bulk of the substrate, which hampers its
coordination, also prevents its coordination in the branched
fashion.
k K K K ½Rhꢀ½Alkeneꢀ½H ꢀ
FA AB BC CF
2
rl ¼ 1
ð1Þ
ð2Þ
0 0
CF FG
þ K K ðK K þ K K Þ½Alkeneꢀ½COꢀ
AB BC
CF FG
0
0
k K K K ½Rhꢀ½Alkeneꢀ½H ꢀ
FA AB BC CF
2
Under the same conditions used for tris(2-tert-butyl-4-meth-
rb ¼
0 0
CF FG
1
þ K K ðK K þ K K Þ½Alkeneꢀ½COꢀ
AB BC
CF FG
[
14]
ylphenyl)phosphite, Selent et al. observed the formation of
5
% branched aldehyde in the product by using the bulky
In the equations, k stands for the rate constant and K stands
for the equilibrium constant of the step specified in the sub-
script (see Scheme 1 for reaction steps). It is possible to simpli-
fy these equations by lumping equilibrium constants together
as given in Equations (3) and (4).
[15]
tris(2,4-di-tert-butylphenyl)phosphite rhodium catalyst.
For-
mation of both aldehydes followed pseudo-first-order kinetics,
with an initial constant reaction rate followed by a first-order
alkene dependence at higher conversions. In situ IR spectros-
copy studies by the same authors confirmed the reformation
of the Rh–hydride species together with the Rh–acyl species as
observable intermediates during the reaction. According to
this study, hydrogenolysis of the Rh–acyl species is the rate-
limiting step over the entire conversion range. The authors ob-
served an increase in the amount of branched aldehyde pro-
duced with decreasing temperature and proposed that this
might be due to the difference in activation energies of hydro-
genolysis steps of the isomeric aldehydes.
k ½Rhꢀ½Alkeneꢀ½H ꢀ
l
2
rl ¼ 1
ð3Þ
ð4Þ
þ K½Alkeneꢀ½COꢀ
k ½Rhꢀ½Alkeneꢀ½H ꢀ
b
2
rb ¼
1
þ K½Alkeneꢀ½COꢀ
Parameters are k and k are Arrhenius-type rate constants
l
b
with frequency factors A and A for the linear and branched al-
l
b
dehydes, respectively, and with activation energies DEAl and
Aiming to elucidate the reasons behind the temperature de-
pendence of the regioselectivity, we investigated the kinetics
of neohexene hydroformylation for the tris(2,4-di-tert-butylphe-
nyl)phosphite rhodium catalyst and performed deuterioformy-
lation reactions and in situ IR spectroscopy measurements to
quantify the contributions of different elementary steps in the
mechanism to this behavior.
DE .
Ab
Two reactions were performed: both at 808C, 20 bar CO,
one at 20 bar H and the other at 40 bar H (1 bar=100 kPa);
2
2
to check the assumption that hydrogenolysis is the rate limit-
ing step for both aldehydes. If any one of the aldehydes had
a different rate-limiting step, the linear to branched aldehyde
ratio (l/b ratio) would change. Both reactions gave the same
linear/branched aldehyde ratio of 18.5. Moreover, monitoring
the reaction profile by sampling during the reaction gave a sim-
ilar profile for both aldehydes, as shown in Figure 1.
During this reaction, the l/b ratio was constant at approxi-
mately 23.2ꢁ0.6. This behavior was confirmed at both 80 and
ꢁ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 603 – 610 604

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
linear and branched aldehydes given in Equations (5)–(7). We
assumed that the concentrations of CO and H were constant
2
for these semibatch reactions, as we made sure to operate at
non-mass-transfer limited conditions and we kept compensat-
ing both gases by keeping the reaction pressure constant. (De-
tailed explanation of how gas concentrations in solution were
calculated is given in the Supporting Information.) To estimate
these equations, the optimization routine uses updated guess-
es of the rate equation parameters mentioned for Equations (3)
and (4) for each iteration by starting from user-supplied initial
guesses. The optimizer compares the experimental concentra-
tion values to the concentration values estimated at times de-
termined as explained before by solving the differential equa-
tions and keeps changing the parameter values until the differ-
ence between the estimated and experimental data converges
to a minimum.
Figure 1. Sampling semibatch deuterioformylation reaction showing reac-
tion profiles for both the linear and branched aldehydes and the total alde-
hyde profile obtained by the gas uptake curve. Reaction conditions: [neo-
ꢂ4
hexene]=2.7m, [Rh]=1.0ꢂ10 m, L/Rh=27, CO=20 bar, D
T=1008C.
2
=20 bar,
d½neohexeneꢀ
ð5Þ
ð6Þ
ð7Þ
¼
ꢂr ꢂ rb
l
dt
d½linear aldehydeꢀ
¼
rl
508C, as given in the Supporting Information (Table S1). Both
dt
the experiment with double the hydrogen pressure and the
sampling experiments confirmed that the hydrogenolysis step
was the common rate-limiting step for both aldehydes.
d½branched aldehydeꢀ
¼ rb
dt
Semibatch hydroformylation experiments were performed to
determine the parameters in the rate equation—frequency fac-
tors, activation energies, and the lumped equilibrium constant
K. Reaction variables appearing in the rate equation (Rh and
Optimal parameter values are presented in Table 1 with their
95% confidence intervals. Figure 2 compares experimental
concentration data from the kinetic experiments and data esti-
mated by using the parameter values presented in Table 1. Es-
alkene concentrations, H and CO pressures, and reaction tem-
2
perature) were varied, as given in the Supporting Information.
The ligand concentration was also varied for some of the ex-
periments to confirm that L/Rh=15 was already a large
enough excess of ligand and that increasing the L/Rh ratio
would not affect the regioselectivity (Figure S1 shows l/b vs. L/
Table 1. Parameter values for rate equations of linear and branched alde-
hydes [Equations (3) and (4)].
Parameter
Value
ꢂ2
ꢂ1
17
11
A
l
[m min
]
(1.45ꢁ0.66)ꢂ10
85.54ꢁ1.86
Rh). H and CO concentrations were kept constant by using
ꢂ1
2
DEAl [kJmol
]
ꢂ2
ꢂ1
a mass flow controller that fed syngas (CO/H =1:1) to keep
Ab [m min
]
(5.83ꢁ0.90)ꢂ10
56.87ꢁ0.31
2
ꢂ1
the reaction pressure and the preadjusted CO/H ratio con-
DEAb [kJmol
]
2
ꢂ2
K [m
]
40.00ꢁ1.05
stant. This approach enabled us to register the gas uptake
values over time, which was easily correlated to alkene conver-
sion, that is, total aldehyde concentration (as given in
Figure 1), as the maximum amount of alkene hydrogenation
observed in the study was 1.4% and the formation of alcohol
or heavies, such as aldol condensation products, was not ob-
served.
Concentration–time data for kinetic evaluation was obtained
by using the gas uptake curves. Given that l/b was shown to
be constant over the reaction range, the respective concentra-
tions of the linear and branched aldehydes were calculated by
using the l/b value determined at the end of the reaction time.
Concentration values at 10, 30, 50, and 70% neohexene con-
version were used together with the time data determined by
using the “lookup” function of Microsoft Excel for the gas
uptake data of each experiment (given in the Supporting Infor-
mation). Nonlinear regression analysis of the kinetic data was
performed by using a MATLAB routine developed according to
[
18]
the algorithm reported by Koeken et al. This routine numeri-
cally integrates the batch design equations for neohexene and
Figure 2. Parity plot comparing estimated and experimental concentration
data from the semibatch experiments.
ꢁ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 603 – 610 605

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
timated data are in good agreement with the experimental
data.
constant for the formation of the linear aldehyde should have
[15]
a stronger temperature dependence.
For the sake of simplicity, we have lumped together the
equilibrium and rate constants in Equations (1) and (2) and ex-
pressed them in the form of Arrhenius dependency. Writing
them in their original forms as in Equation (9) will help to
show the role of these two probable sources of temperature
effects on the regioselectivity.
Regioselectivity
Given that the l/b ratio was shown to be constant during the
reaction, it was possible to model the regioselectivity as given
in Equation (8) by using the parameter values given in Table 1.
ꢀ
ꢁ
ꢀ
ꢁ
ꢀ
ꢁ
ꢂDEAl
ꢂ
ꢃ
A
¼ A
FA exp
ꢂDEFA=
RT
exp
exp
ꢂDGCF=RT
Al exp
=
l
b
rl
rb
RT
Al
Ab
ꢂDE þ DE
r
l
k
FA
K
K
CF
Al
Ab
¼
¼
ꢀ
ꢁ ¼ exp
ð8Þ
¼ k
0
0
ꢀ
ꢁ
ꢀ
ꢁ
ð9Þ
ꢂDE0
ꢂDG0
ꢂDEAb=RT
RT
r
b
FA
CF
0
Ab exp
FA
exp
FA
=
RT
CF
=
RT
Equation (8) suggests that the regioselectivity is a function
Lumped activation energies DEAl and DEAb, which represent
of temperature only. Figure 3 shows the percentage of linear
aldehyde in the product depending on the reaction tempera-
the whole cycle for linear and branched aldehydes, differ by
ꢂ1
28.7 kJmol , as shown in Equation (8). This difference is a com-
bination of the differences in activation the energies (DEFA and
DE’ ) for the hydrogenolysis step and the Gibbs free energies
FA
of the equilibrium steps (DGCF and DG’ ) between the p com-
CF
plex (species C and C’) and Rh–acyl (species F and F’). To quan-
tify these effects, deuterioformylation and in situ IR spectrosco-
py experiments were performed.
Deuterioformylation experiments
Hydroformylation experiments performed under D2 pressure
instead of H give direct information on the reversibility of the
2
[10,20]
Rh–alkyl formation step (C to D).
The use of the Rh catalyst
under D pressure will label the alkene reversed from branched
2
Rh–alkyl at C1, whereas it will label the alkene reversed from
linear Rh–alkyl at C2, as shown in Scheme 3.
Figure 3. Percentage of linear aldehyde in the product as a function of tem-
perature; values were obtained from kinetic hydroformylation experiments
and estimated by Equation (8). Data points at 323 and 373 K belong to deu-
terioformylation experiments.
ture, obtained from the semibatch hydroformylation reactions
performed for the kinetics and estimated by using Equation (8).
There is a clear increase in the percentage of linear aldehyde
in the product with increasing temperature, especially evident
in the lower temperature region. The model predicts the
higher temperature region, for which more experimental data
was available, better than the lower temperature region. Spe-
cifically, the data points for 508C, measured for deuterioformy-
lation experiments, are quite higher than the predicted values.
Further experiments in the lower temperature region are re-
quired to clear this point.
Scheme 3. Deuterium labeling of alkene molecules reversed from Rh–alkyl
formation under deuterioformylation conditions.
Three deuterioformylation reactions were performed with
This behavior can be the result of a shift in the equilibrium
for Rh–alkyl formation with increasing temperature that favors
D /CO (20:20 bar) at 50, 80, and 1008C to investigate the effect
of temperature on Rh–alkyl reversibility. A fourth experiment
2
[
19]
the formation of linear Rh–alkyl. Another reason can be a dif-
ference in the activation energies of the hydrogenolysis steps.
If the activation energy for the formation of the linear alde-
hyde was higher than that for the branched isomer, the rate
was performed at 508C and 20:10 bar D /CO to check if CO
2
pressure had any effect on the reversibility of the Rh–alkyl spe-
cies. There might be a different effect on the equilibrium fol-
lowing linear Rh–alkyl formation (D to E) compared to the
[21]
branched one (D’ to E’). If one of these equilibria is more
ꢁ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 603 – 610 606

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
sensitive to the CO concentration, this would shift the rate-lim-
iting step to CO insertion into the Rh–alkyl species (D to E) at
lower CO concentrations. However, care should be taken to
prevent CO depletion, that is, not to work under gas–liquid
mass-transfer control. Deuterium incorporations into the C1
The temperature sensitivity of an equilibrium arises from the
entropy change associated with this step. Given DG=
DHꢂTDS, the entropy term will affect the equilibrium more at
higher temperatures. Rh–alkyl formation should have a nega-
tive entropy change, as one molecule is formed from two mol-
ecules and a negative change in entropy will favor the reverse
reaction, even more so at increased temperature. Thus, the in-
crease in reversed linear Rh–alkyl with temperature suggests
that the negative change in entropy for linear Rh–alkyl forma-
tion is higher than the negative change in entropy for
branched Rh–alkyl formation and consequently that linear Rh–
alkyl has lower entropy. Reducing the CO pressure did not
have an effect on the reversibility at 508C. This is an expected
result and confirms that hydrogenolysis is the common rate-
limiting step.
(terminal) and C2 (internal) positions were determined by anal-
ysis of samples taken during the course of the reactions by
2
H NMR spectroscopy. As seen from Figure 4, the fraction of
High-pressure IR experiments
Although deuterioformylation experiments give a clear idea of
the reversibility for the linear and branched Rh–alkyl species,
they do not help to quantify the activation energy difference
in the hydrogenolysis step. Only an analysis of the respective
amounts of the Rh–acyl species would give direct information
Figure 4. Deuterium incorporation into the C1 (terminal) and C2 (internal)
[6]
on that, as explained in the derivation of Equations (10) and
positions of neohexene during deuterioformylation: [neohexene]=2.7m,
(
11).
ꢂ4
[
Rh]=1.0ꢂ10 m, L/Rh=27, CO=20 bar unless otherwise stated,
2
D =20 bar.
dCl aldehyde
¼
k ½Fꢀ½H ꢀ
ð10Þ
ð11Þ
FA
2
dt
deuterium atoms incorporated with respect to the number of
aldehyde molecules produced increased with increasing tem-
perature for both positions.
dCl aldehyde k ½Gꢀ½H ꢀ
FA
2
¼
dt
K ½COꢀ
FG
An important observation is that at 1008C the number of D
atoms on C2 was higher than that on C1. Hence, the equilibri-
um for linear Rh–alkyl formation is more temperature sensitive
than the equilibrium for branched Rh–alkyl formation. This be-
havior is easily observed in Figure 5, for which the fraction of
D atoms in the C1 and C2 positions averaged over the entire
conversion range is plotted.
According to the justified assumption that l/b is constant
during the reaction, it is possible to write Equation (12), and by
comparing Equation (12) to Equation (9), Equation (13) can be
written as given.
ꢀ
ꢁ
ꢂDEFA
0
0
AFA exp
=
K ½Gꢀ
dCl aldehyde
dCb aldehyde
l
b
k K ½Gꢀ
RT FG
FA FG
0
¼
¼
¼
ꢀ
ꢁ
0
ꢂDE0
k K ½G ꢀ
FA FG
0
0
A
exp
FA
=
K ½G ꢀ
RT FG
FA
ꢀ
ꢁ
ꢀ
ꢁ
ꢀ
ꢁ
ð12Þ
0
0
ꢂDEFA
DS ꢂDSFG
DH ꢂDHFG
FG
FG
RT
0
AFA exp
exp
exp
l½G ꢀ
RT
R
¼
ꢀ
ꢁ
0
b½Gꢀ
0
ꢂDEFA
A
exp
FA
RT
ꢀ
ꢁ
ꢀ
ꢁ
ꢀ
ꢀ
ꢁ
ꢂDGCF
ꢂDGCF
ꢂDGFG
KFG exp
=
exp
exp
=
exp
exp
=
½
Gꢀ
RT
RT
RT
¼
ꢀ
ꢁ ¼
ꢀ
ꢁ
ꢁ
0
ꢂDG0
ꢂDG0
ꢂDH þ DH0
ꢂDG0
½
G ꢀ
0
K
exp
CF
=
CF
=
FG
=
FG
RT
RT
RT
ꢂ
ꢃ
ꢂ
ꢃ
0
DSCG ꢂ DS
CG
CG
CG
¼
exp
exp
R
RT
ð13Þ
The ratio of the pre-exponential factors A /A’ enables us
to calculate the difference in activation entropy changes, as
given in Equation (14).
FA
FA
Figure 5. Average deuterium incorporation into the C1 (terminal) and C2 (in-
ternal) positions of neohexene during deuterioformylation for conditions
given in Figure 3.
ꢁ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 603 – 610 607

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
ꢂ
ꢃ
0
A
A
DSFA ꢂ DS
FA
FA
¼
exp
ð14Þ
Table 2. Ratio of linear to branched Rh–acyl species and reaction condi-
tions for the IR spectroscopy experiments.
0
R
FA
[
a]
[a]
T [K]
H
2
CO
[G]/[G’]
The ratio of linear to branched Rh–acyl [G]/[G’] also provides
2
3
328
343
98
13
4.85
3.87
6.59
3.75
35.15
34.21
29.75
31.00
1.04
1.12
1.50
1.68
direct information on the Gibbs free-energy differences of the
steps between species C and G, as given in Equation (13).
In situ IR spectroscopy experiments were performed to quanti-
fy the respective amounts of linear and branched Rh–acyl spe-
cies. Figure 6 shows the spectra recorded in the first 10 min of
the reaction performed at room temperature.
[
a] In units of bar at 258C (1 bar=100 kPa).
of hydrogen used in these experiments (Table 2) enabled us to
take a number of spectra before the Rh–acyl band was con-
ꢂ1
cealed under the aldehyde band at 1734 cm . In this way, we
were able to obtain the spectra of the acyl C=O bond of the
Rh–acyl species as given in Figure 7, with desired species G
ꢂ1
and G’ overlapping at approximately 1690 cm .
Figure 6. IR spectra recorded in the first 10 min of neohexene hydroformyla-
ꢂ4
tion, [neohexene]=0.54m, [Rh]=3.14ꢂ10 m, L/Rh=26, T=258C.
Negative bands appearing after the addition of neohexene
ꢂ1
(
n˜ =1824.4 cm ) belong to hydride species A, characterized by
Figure 7. IR spectra recorded 8 min after neohexene addition for the 408C IR
spectroscopy experiment after neohexene subtraction, showing the decon-
volution of the Rh–acyl species.
ꢂ1
bands at 2013, 2043, and 2092 cm . These values are in close
agreement with the ones reported in the literature ( n˜ =2013,
041, and 2091 cm ). They are negative because the back-
CO
ꢂ1 [15]
2
ground was measured after complete conversion of the cata-
lyst into the Rh–hydride species, and after substrate addition
some of this species turned into Rh–acyl. The Rh–acyl species
After deconvolution of this band, linear and branched coor-
dinatively saturated Rh–acyl species G and G’ were character-
ꢂ1
ized by the bands at 1693 and 1689.5 cm , respectively, with
(
i.e., G, G’) is characterized by three bands in the terminal CO
the values given in Table 2. The deconvolutions were per-
ꢂ1
[24]
region: n˜ =1994.6, 2019.8, and 2078 cm , which appeared im-
formed by CasaXPS, and the final values were obtained by
mediately after the addition of the substrate, also in agree-
averaging the acyl ratios calculated from a number of spectra
for each temperature. (More data on deconvolutions are avail-
able in the Supporting Information)
[
15]
ꢂ1
ment with the reported values ( n˜ CO =1995, 2017, 2079 cm ).
The bands belonging to the acyl C=O bond should appear
[
22,23]
in the organic carbonyl region
at approximately 1690–
Rearranging Equation (13), the natural logarithm of the Rh–
acyl ratios ln([G]/[G’]) describes a line as given in by Equa-
tion (15).
ꢂ1
1
700 cm . These enabled us to determine the ratio of linear
and branched Rh–acyl. This band was recently reported to
ꢂ1
[17]
appear at 1690 cm for this catalyst–substrate system. How-
ꢂ
ꢃ
ꢂꢂDH
ꢃ
ꢂ
ꢃ
0
ꢂ1
ꢂ1
½
Gꢀ
þ DH0
DSCG ꢂ DS
ever, the aldehyde (1734 cm ) and neohexene (1642 cm )
bands overlap in this region, and it was not possible to ob-
serve the relatively small C=O bands, especially at higher con-
versions. Therefore, the spectra of pure neohexene were sub-
tracted from the recorded reaction spectra. The neohexene
CG
CG
CG
ln
¼
þ
ð15Þ
0
½
G ꢀ
RT
R
Linear regression with data given in Table 2 allowed
ꢂ1
DH ꢂDH’ to be calculated as 9.7 kJmol and a difference
CG
CG
ꢂ1
ꢂ1 ꢂ1
signal at 1824.4 cm was used to determine the proper
of changes in entropy DS ꢂDS’ as 32.5 Jmol
K
(a figure
CG
CG
amount of neohexene to be subtracted; spectra of different
amounts of neohexene were subtracted and the one that elim-
inated the band at this position was chosen. The low pressure
showing the linear fit is given in Figure S2). It would be justi-
fied to assume that the total change in free energy from spe-
cies C to F is negative, because the coordination of an alkene
ꢁ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 603 – 610 608

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
is commonly thought to be a high-energy change step, as evi-
denced by the energy profiles of hydroformylation for a range
Conclusions
[
8,21,22,25]
of substrates.
The change in free energy of the equilib-
Detailed analysis of the kinetics and mechanism of Rh-cata-
lyzed neohexene hydroformylation was performed for the
tris(2,4-di-tert-butylphenyl)phosphite ligand. The rate-limiting
step for both the linear and branched aldehydes was shown to
be the hydrogenolysis of the Rh–acyl species. Rate equations
were derived, and equation parameters were estimated within
very narrow confidence intervals by using nonlinear regression.
These estimates were used to predict the regioselectivity,
which was in good agreement with the experimental data.
Deuterioformylation and in situ IR spectroscopy experiments
were performed to reveal the reasons behind the increase in
linear aldehyde selectivity with increasing temperature.
rium between F and G should also be negative. We also know
that the difference in Gibbs free-energy changes is negative
(
DG ꢂDG’ <0), from which we can deduce jDG j > j
CG
CG
CG
DG’ j. So, a higher negative value of DG favors the forward
CG
reaction (C to G) of the linear path, even more so with increas-
ing temperature, because the entropy difference DS ꢂDS’ is
CG
CG
positive and increases the difference in Gibbs free-energy
changes with increasing temperature; as should be evident
from the expression given in Equation (16).
DGCG ꢂ DG0 ¼ DHCG ꢂ DH0 ꢂ TðDSCG ꢂ DS0CGÞ
ð16Þ
CG
CG
The deuterioformylation experiments showed that the for-
mation of both the linear and branched Rh–alkyl species was
reversible and that the reverse reaction was favored at higher
temperatures. This finding suggested that the entropy change
involved in Rh–alkyl formation was negative. At 1008C, linear
Rh–alkyl formation was more reversible than branched Rh–
alkyl formation, which implies the negative entropy change for
linear Rh–alkyl formation is larger than that for branched Rh–
alkyl formation: that is, the equilibrium for linear Rh–alkyl for-
mation (C to D) is more temperature sensitive than the equilib-
rium for branched Rh–alkyl (C to D’). High-pressure IR spectros-
copy experiments performed to investigate the linear and
branched Rh–acyl species (G and G’) showed an excess
amount of the linear Rh–acyl species relative to that of the
branched species. Moreover, the ratio of linear to branched
Rh–acyl increased with increasing temperature. These findings
suggest that the total forward reaction of non-common equi-
librium steps of the linear and branched catalytic cycles (C to
G and C to G’) are favored for the linear aldehyde over the
branched aldehyde.
It is now possible to obtain the values of the expressions
given in Equations (17) and (18) by using the data in Table 2
for linear regression of Equation (12). (Figure showing the
linear fit is given in Figure S3.)
ðDEFA ꢂ DH Þ ꢂ ðDE0 ꢂ DH0FGÞ
ð17Þ
FG
FA
DSFA ꢂ DS0 ꢂ DSFG þ DS0
ð18Þ
FA
FG
ꢂ1
Thus, (DE +DH )ꢂ(DE’ +DH’ )=DH ꢂDH’ =19 kJmol
FA
GF
FA
GF
GA
GA
ꢂ1
and DS ꢂDS’ +DS ꢂDS’ =DS ꢂDS’ =71 Jmol K. For
FA
FA
GF
GF
GA
GA
the formation of the activated state of the hydrogenolysis
[
26]
step, the entropy change should be positive. Garland et al.
#
ꢂ1 ꢂ1
reported an activation entropy of DS =(121ꢁ14) Jmol
K
for the hydrogenolysis step in neohexene hydroformylation
catalyzed by an unmodified Rh catalyst. A positive value for
the activation entropy suggests that the transition state is
highly unstable. Degrees of freedom are “liberated” in going
from the ground state to the transition state, which, in turn, in-
crease the rate of the reaction. The entropy change involved in
going from G to F or from G’ to F’ should also be positive, and
consequently, both DSGA and DS’GA are positive.
Furthermore, the reverse step of the equilibrium from G to F
and subsequent hydrogenolysis of linear acyl F was found to
have a larger enthalpy change than the branched equivalent
(G’ to F’ and hydrogenolysis of F’). The contribution of this
part of the cycle to the temperature dependence of regioselec-
tivity was approximately twice that from species C to G (and C
to G’). Finally, the larger positive entropy changes for both
mentioned parts of the linear aldehyde cycle make the linear
aldehyde the favored isomer under typical hydroformylation
conditions.
So, hydrogenolysis of five-coordinated linear Rh–acyl G
under typical hydroformylation conditions is faster than that of
branched isomer G’ although it has a higher activation energy
ꢂ1
(
DH ꢂDH’ =19 kJmol ) because it has a larger positive ac-
GA
GA
ꢂ1 ꢂ1
tivation entropy (DS ꢂDS’ =71 Jmol K ).
GA
GA
In summary, the part of the reaction cycle from species C to
G has a higher enthalpy change than the equivalent part in
ꢂ1
Experimental Section
the branched cycle, DH ꢂDH’ =9.7 kJmol . The hydroge-
CG
CG
nolysis of linear acyl G creates an enthalpy change that is
Semibatch reactions for kinetics were performed in custom-built
stainless steel autoclaves (100 mL) equipped with a mechanical
gas-impeller stirrer. H and CO concentrations were kept constant
2
ꢂ1
1
9 kJmol more than the change involved in the hydrogenol-
ꢂ1
ysis of branched acyl G’, DH ꢂDH’ =19 kJmol . The sum of
GA
GA
by using a mass flow controller that fed syngas (Linde Gas CO/H =
these two figures determines the temperature sensitivity of
2
ꢂ1
1:1, ꢃ99.998 vol%) to keep the reaction pressure and the pread-
the l/b ratio; this adds up to 28.7 kJmol , which was found as
justed CO/H ratio constant. Reaction temperature was monitored
2
the difference between the lumped activation energies for
during the reaction and was constant at the set temperature with
a maximum of 28C overshoot in the first minutes. A total reaction
volume of 10 mL (neohexene and toluene) was used for the kinetic
experiments and the maximum amount of neohexene used was
5.5 mL. A stirring rate of 1200 rpm was set for all kinetic experi-
linear and branched aldehydes, DE ꢂDE , as given in Equa-
Al
Ab
tion (8). The difference in the parts of the reaction cycle from
G to A and from G’ to A’ has a higher contribution to the tem-
perature dependence of regioselectivity.
ꢁ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 603 – 610 609

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
ments. Conversion was measured by gas chromatography (GC) by
using an Agilent DB1 column (30 mꢂ0.32 mm i.d.) at the end of
the reaction time to confirm the values obtained from gas uptake
Acknowledgements
The authors are grateful for the financial support of the Europe-
an Community’s Seventh Framework Programme (FP7/2007–
(
GU) data. The conversion values were generally in good agree-
ment, with x /x =1ꢁ0.05. If x /x <1, the difference can be at-
GC GU
GC GU
2013) under grant agreement No. 228867. We also thank Mr. T.
tributed to leaking gas or an overshoot in the measurement of the
amount of neohexene by using a graduated syringe. For x /x
Staring for technical assistance.
>
GC GU
1
, the initial amount of neohexene can also be lower as a result of
experimental error or loss of neohexene while charging the volatile
neohexene to the autoclave under Ar atmosphere. For the sake of
consistency, the amount of neohexene charged into the autoclave
was recalculated according to the GU value at full conversion for
all cases. All solutions were prepared under an atmosphere of
argon by using Schlenk techniques. The substrate solution contain-
ing neohexene (Sigma–Aldrich, ꢃ97%) and decane (Sigma–Aldrich,
ꢃ99%) as an internal standard was filtered over neutral alumina
Keywords: deuterioformylation
spectroscopy · kinetics · phosphite
·
hydroformylation
·
IR
[
[
[
1] P. W. N. M. van Leeuwen, C. Claver, Rhodium Catalyzed Hydroformyla-
tion, Springer, 2002.
2] S. C. van der Slot, J. Duran, J. Luten, P. C. J. Kamer, P. W. N. M. van Leeu-
wen, Organometallics 2002, 21, 3873–3883.
3] A. van Rooy, E. N. Orij, P. C. J. Kamer, P. W. N. M. van Leeuwen, Organo-
metallics 1995, 14, 34–43.
and degassed. A stock solution of the catalyst [Rh(acac)(CO) ,
2
acac=acetylacetonate, obtained as a generous loan from OMG]
and ligand [tris(2,4-di-tert-butylphenyl)phosphite, Sigma–Aldrich,
9
8%] was prepared in dry and degassed toluene (purchased from
Biosolve and purified over custom made alumina columns).
[
[
4] B. Breit, W. Seiche, Synthesis 2001, 2001, 0001–0036.
5] G. Alagona, C. Ghio, R. Lazzaroni, R. Settambolo, Organometallics 2001,
The substrate solution was brought into an addition funnel con-
nected to the autoclave by a tube that kept the funnel at the
same pressure as the autoclave and a valve that enabled con-
trolled charge of the substrate solution into the autoclave. The cat-
alyst solution was charged into the autoclave, and it was pre-
formed under reaction conditions for 1 h, after which the reaction
was started by adding the substrate solution. Complete conversion
of the catalyst precursor into the Rh–H species was confirmed to
occur in 40 min at 508C, which implies the preformation was com-
plete for all the experiments in this study. (IR spectra of the prefor-
mation is given in Figure S4.)
2
0, 5394–5404.
[
[
[
6] J. Feng, M. Garland, Organometallics 1999, 18, 417–427.
7] A. L. Watkins, C. R. Landis, J. Am. Chem. Soc. 2010, 132, 10306–10317.
8] R. Lazzaroni, R. Settambolo, G. Alagona, C. Ghio, Coord. Chem. Rev.
2010, 254, 696–706.
[9] K.-D. Wiese, D. Obst in Catal. Carbonylation React. (Ed.: M. Beller),
Springer, Berlin, 2006, pp. 1–33.
[
[
[
[
[
10] R. Lazzaroni, G. Uccello-Barretta, M. Benetti, Organometallics 1989, 8,
2
323–2327.
11] R. Lazzaroni, R. Settambolo, G. Uccello-Barretta, A. Caiazzo, S. Scamuzzi,
J. Mol. Catal. A 1999, 143, 123–130.
12] C. P. Casey, S. C. Martins, M. A. Fagan, J. Am. Chem. Soc. 2004, 126,
5
585–5592.
Deuterioformylation reactions were performed with the same type
of autoclave as mentioned above equipped with a sampling unit
that enabled sampling during the reaction. The total reaction
volume was 30 mL and the neohexene volume was 15 mL.
13] S. Gꢀven, B. Hamers, R. Franke, M. Priske, M. Becker, D. Vogt, Cat. Sci.
Technol. 2014, DOI: 10.1039/C3CY00676J..
14] A. Van Rooy, J. N. H. de Bruijn, K. F. Roobeek, P. C. J. Kamer, P. W. N. Van
Leeuwen, J. Organomet. Chem. 1996, 507, 69–73.
2
[15] C. Kubis, R. Ludwig, M. Sawall, K. Neymeyr, A. Bçrner, K.-D. Wiese, D.
H NMR spectroscopy was performed with a Varian Unity Inova 500
1
2
Hess, R. Franke, D. Selent, ChemCatChem 2010, 2, 287–295.
spectrometer operating at 499.8 MHz for H (76.72 MHz for H),
equipped with a 5 mm AutoX DB-PFG probe. Prior to measure-
ment, the 908 pulse width was calibrated by using a solution of
[
16] T. Jongsma, G. Challa, P. W. N. van Leeuwen, J. Organomet. Chem. 1991,
21, 121–128.
4
[
17] C. Kubis, D. Selent, M. Sawall, R. Ludwig, K. Neymeyr, W. Baumann, R.
C D in CHCl . Spectra were recorded by using a 1535 Hz spectral
6
6
3
Franke, A. Bçrner, Chem. Eur. J. 2012, 18, 8780–8794.
width by employing 908 pulses and a relaxation delay of 20 s. Data
was processed by using VNMRJ 2.2d software and was filtered by
using 1 Hz line broadening. Integrals were normalized to the alde-
hyde signal at 10.1 ppm
[18] A. C. J. Koeken, L. J. P. van den Broeke, B.-J. Deelman, J. T. F. Keurentjes,
J. Mol. Catal. A 2011, 346, 1–11.
[19] R. Lazzaroni, P. Pertici, S. Bertozzi, G. Fabrizi, J. Mol. Catal. 1990, 58, 75–
85.
[
20] C. P. Casey, L. M. Petrovich, J. Am. Chem. Soc. 1995, 117, 6007–6014.
In situ high-pressure IR spectroscopy experiments were performed
in a custom-built stainless steel autoclave (50 mL) with an integrat-
ed flow cell equipped with a ZnS windows and a mechanical gas-
impeller stirrer (800 rpm for all FTIR measurements). The spectra
were recorded with a Shimadzu FTIR 8300 spectrometer in the ab-
sorbance mode. A solution of the catalyst in hexane was pre-
formed under an atmosphere of syngas, and the reaction was
started after complete hydride formation by adding the substrate,
which was kept in a dropping funnel. Background measurements
were performed just before the addition of the substrate. FTIR
spectra of pure neohexene were later subtracted from the meas-
urements in proper amounts. Catalyst and substrate concentrations
[21] C. Ghio, R. Lazzaroni, G. Alagona, Eur. J. Inorg. Chem. 2009, 98–103.
[22] G. Liu, R. Volken, M. Garland, Organometallics 1999, 18, 3429–3436.
[23] P. C. J. Kamer, A. van Rooy, G. C. Schoemaker, P. W. N. M. van Leeuwen,
Coord. Chem. Rev. 2004, 248, 2409–2424.
[
[
[
[
24] http://www.casaxps.com/.
25] G. Alagona, R. Lazzaroni, C. Ghio, J. Mol. Model. 2011, 17, 2275–2284.
26] M. Garland, P. Pino, Organometallics 1991, 10, 1693–1704.
27] J. H. Hildebrand, Solubility of Non-electrolytes, Reinhold Pub. Corp.,
1
936.
[
28] U. J. Jꢃuregui-Haza, E. J. Pardillo-Fontdevila, A. M. Wilhelm, H. Delmas,
Lat. Am. Appl. Res. 2004, 34, 71–74.
[29] L. C. Yen, J. J. McKetta, AIChE J. 1962, 8, 501–507.
ꢂ4
were kept constant at [neohexene]=0.54m, [Rh]=3.14ꢂ10 m,
Received: September 25, 2013
and L/Rh=26 for all experiments (total reaction volume 10 mL).
Published online on January 8, 2014
ꢁ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2014, 6, 603 – 610 610