Exploring between the extremes: Conversion-dependent kinetics of phosphite-modified hydroformylation catalysis

ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Chem. Eur. J. 0000, 00, 0 – 0

ÝÝ

These are not the final page numbers!

Conversion-Dependent Kinetics of Phosphite-Modified Hydroformylation Catalysis

FULL PAPER

[

19]

nents consisting of the 16- and

8-electron hydrido and acyl

rhodium complexes, respective-

GC.

The reaction proceeds highly chemoselective and

1

with the same kinetics for both aldehyde products, 4,4-dime-

thylpentanal and 2,3,3-trimethylbutanal, with the latter iso-

aldehyde representing a constant 0.099 molar fraction over

the whole conversion range. Therefore, in Figure 1a the sum

[15a,d]

ly, at a quasi-equilibrium.

Consequently, all elementary

steps preceding acyl complex

formation are considered to

belong to one reaction step.

Scheme 2. Michaelis–Menten-

type catalysis.

We shall show that the quasi-

stationarity of intermediates is not given for the Rh/mono-

phosphite catalytic system under investigation. But with

[

Cat] ![S] the Bodenstein approximation holds for nearly

0 0

the entire conversion range and the experimental data can

be analyzed by the Michaelis–Menten approach.

Hydroformylation of 3,3-dimethyl-1-butene in n-hexane:

3

,3-Dimethyl-1-butene as a substrate was converted with

a rhodium/tri(2,4-di-tert-butylphenyl)phosphite (TDTBPP or

L) catalyst at [Rh]=0.3 mm, T=308C, p_C_O=1 MPa, p_H₂

MPa, see Scheme 3. As a change to our earlier study in cy-

=

1

Scheme 3. Hydroformylation of 3,3-dimethyl-1-butene with a rhodium/

tri(2,4-di-tert-butylphenyl)phosphite catalyst.

clohexane, n-hexane was used as a solvent because the back-

ground subtraction required during the preprocessing of the

FTIR spectroscopic data could be performed in a much

more reliable manner. The olefin selected is not able to un-

dergo double-bond isomerization, which would otherwise

alter the kinetic behavior.

Figure 1. Hydroformylation of 3,3-dimethyl-1-butene. a) Concentration

versus time plot of the sum of aldehydes (n+iso, black curve). Experi-

mental data obtained from FTIR spectroscopy at the reference wave-

ꢁ1

number n˜ =1765.1 cm to avoid nonlinearities of the absorption at the

aldehyde carbonyl band maximum, compared to the regression curve

(gray curve) from the numerically integrated Michaelis–Menten equa-

tion. b) Transient phase kinetics of acyl complex formation;

Before the substrate was added, [(acac)Rh(CO) ] (acac=

2

ꢁ

1

[

HRh(CO)

3

L] (main component): n˜ (CO)=2015, 2043, 2093 cm ,

acetylacetonate anion) in the presence of twenty equivalents

of TDTBPP was reacted to the pentacoordinate hydrido

ꢁ1

2

L

2

] (minor component) n˜ (CO)=2028, 2068 cm , acyl com-

plex: n˜ (CO)=1995, 2019, 2072, and 2079 cm . Conditions: [Rh]=0.3,

[TDTBPP]=6 mm; [olefin] =0.9m; T=308C; p =1, p_H₂=1 MPa; sol-

ꢁ1

complex [HRh(CO) L] under conditions intended for catal-

3

0

CO

vent: n-hexane.

ysis. Complete conversion to the hydride required 16 h at

3

08C. The final IR spectrum shows the main component

ꢁ

1

[HRh(CO) L] with n˜ (CO)=2015, 2043, and 2093 cm and

3

contributions of a minor compound at n˜ (CO)=2028 and

of aldehydes is used to illustrate the dynamics of the prod-

uct formation. These experimental concentration versus

time data of the aldehyde(s) were used for a nonlinear re-

ꢁ

1

2

068 cm , attributed to the bisphosphite hydrido complex

[5c,d]

[

HRh(CO) L ].

After addition of the olefin a slow hy-

2

droformylation reaction started without induction time,

taking >15 h for full conversion.

The concentration profile of the aldehyde product, as de-

termined by IR spectroscopy, did fit to that obtained by

gression to compute the possible maximum rate V =1.52ꢃ

sat

ꢁ

3

ꢁ3

ꢁ1

10 moldm min

(for [CatS]=[Cat]₀) and K =

m

ꢁ

3

0.173 moldm by numerical integration of the Michaelis–

Menten equation [see Eq. (4)].

Chem. Eur. J. 2012, 00, 0 – 0

ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

&

3

&

ÞÞ

These are not the final page numbers!

D. Selent et al.

d½Pꢀ

dt

V ½Sꢀ

V ð½Sꢀ ꢁ ½PꢀÞ

plementation of solvent background subtraction into the

spectra-decomposition routine allowed for a more precise

reconstruction of the concentration profiles of the organo-

metallic intermediates.

sat

0

V ¼

¼

ð4Þ

K þ ½Sꢀ K þ ð½Sꢀ ꢁ ½PꢀÞ

m

0

To further support the result of the regression procedure,

we simulated the Michaelis–Menten mechanism by solving

the set of ordinary differential Equations (5a)–(5d).

Figure 2a shows a spectrum stack covering the full con-

version range. The first spectrum is obtained after catalyst

preformation and represents an equilibrium mixture of the

hydrido complexes [HRh(CO) L] and [HRh(CO) L ] in

d½Sꢀ

¼

ꢁk ½Sꢀ½Catꢀ þ k_ꢁ₁½CatSꢀ

ð5aÞ

ð5bÞ

3

2

1

dt

[5c,d]

a 13.3:1 ratio.

After the addition of the olefin the acyl

L] is rapidly populated to the

d½Catꢀ

dt

complex [RC(O)Rh(CO)

3

obs

¼

ꢁk ½Sꢀ½Catꢀ þ k_ꢁ₁½CatSꢀ þ k₂½CatSꢀ

1

dominant intermediate. A certain fraction of the hydrido

d½CatSꢀ

dt

obs

¼

k₁½Sꢀ½Catꢀ ꢁ k_ꢁ₁½CatSꢀ ꢁ k₂½CatSꢀ

ð5cÞ

ð5dÞ

d½Pꢀ

obs

¼

k₂½CatSꢀ

dt

obs

For that purpose, the known value of k₂and the esti-

mated values of k and k obtained from a variation of the

1

ꢁ1

hydrogen pressure, which will be described below, have

been used. Thus, evidence is given that the kinetic model

implementing a steady-state approximation is also valid

[20]

when the entire conversion range is taken into account.

The initial rate V₀was calculated to 1.27ꢃ

ꢁ

3

ꢁ3

ꢁ1

10

moldm min according to Equation (6a).

V ½Sꢀ

sat

0

V ¼

ð6aÞ

ð6bÞ

0

K þ ½Sꢀ

m

0

V₀

½CatSꢀ

¼

V_s_a_t½Catꢀ þ ½CatSꢀ

As can be derived from Equation (6b), a 0.84 molar frac-

tion of the catalyst resides in the intermediate acyl complex

(

CatS) at the beginning of the reaction. This does point to

the fact, that the system does not operate under full satura-

tion conditions at any time. An interesting result arised

from a rapid scan measurement following the evolution of

the reaction directly after olefin addition, see Figure 1b. The

formation of the acyl complex with n˜ (CO)=1995, 2019,

ꢁ

1

2

072, and 2079 cm is fast, with most of the hydrido com-

plex conversion taking place within a time span of five to

seven seconds, after some delay, which probably is due to

mixing phenomena. Unfortunately, the respective dynamics

could not be followed precisely. The application of

a stopped-flow methodology would be appropriate. Howev-

er, it becomes clear that the formation of the acyl intermedi-

ate is much faster than the hydroformylation. The transient

phase kinetics is therefore negligible.

Figure 2. Hydroformylation of 3,3-dimethyl-1-butene. a) FTIR spectra

after raw spectra treatment, see text for details. First spectrum: mixture

ꢁ1

of [HRh(CO)

(

3

L] ( n˜ (CO)=2015, 2043, 2093 cm ) and [HRh(CO)

2

L

2

]

ꢁ1

n˜ (CO)=2028, 2068 cm , the low frequency band is obscured here) as

obtained after catalyst preformation. Next spectrum, 6.3 min after addi-

tion of the olefin: The acyl complex [RC(O)Rh(CO) L] ( n˜ (CO)=1995,

2019, and 2079 cm , indicated by asterisk) is dominant. For the origin of

the band observed at n˜ =2072 cm , see further results and discussion

below. Background: n-hexane+conversion-dependent amount of sub-

Improved spectroscopic analysis and concentration profiles

of organometallic intermediates: In a preceding paper we

showed that the hydrido complex (Cat) does coexist with

the acyl complex (CatS) over a wide conversion range. How-

ever, it has not been detected at a maximum substrate con-

3

ꢁ1

ꢁ

1

strate. b) Spectra of the acyl and hydrido complexes obtained by the ap-

[11]

centration. By changing the solvent from cyclohexane to

n-hexane and advanced background treatment, now better

resolved FTIR spectra were obtained. Furthermore, the im-

[21]

plication of the PCD software on the spectra sequence.

Conditions:

[

Rh]=0.3, [TDTBPP]=6 mm; [olefin]

0

=0.9m; T=308C; pCO =1 MPa,

H

p ₂=1 MPa; solvent: n-hexane.

&

4

&

ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Chem. Eur. J. 0000, 00, 0 – 0

ÝÝ

These are not the final page numbers!

Conversion-Dependent Kinetics of Phosphite-Modified Hydroformylation Catalysis

FULL PAPER

complex remains unreacted, which is in agreement with the

evaluation of the product concentration profile by the kinet-

ic model. With an increasing olefin consumption, the intensi-

ties of the bands for the acyl complex decrease, whereas

those for the hydrido complex increase, finally resulting in

a spectrum almost identical to that prior to the catalytic re-

action. In Figure 2b the spectra of the hydride and the acyl

complexes are given. Due to the very low rhodium concen-

tration and the treatment of the conversion-dependent back-

ground required, artifacts were generated. See for example,

the low frequency shoulder of the band located at n˜ =

ꢁ

1

2

043 cm in the spectrum of the hydrido complex. Howev-

er, this did not hinder further analysis.

We want to describe briefly how the spectra shown in

Figure 2 and the concentration profiles of the organometal-

lic species as depicted in Figure 3a were obtained. The

pure-component decomposition (PCD) software was ap-

[

21]

plied.

took also into account the conversion dependency of the

The raw spectra required data preprocessing that

[22]

background.

Thus, it was possible to remove negative

ꢁ

1

components, especially from a band at n˜ =1990 cm of the

olefin, and to correct a strong drift of the zero line. The re-

sulting series of spectra in the spectral interval of 1960–

ꢁ

1

2

120 cm is that shown in Figure 2a.

The concentration profiles of the acyl complex and the

hydrido complex were obtained from the single-run spectra

decomposition under the restriction of a constant rhodium

concentration. Additionally, the set of ordinary differential

equations for the Michaelis–Menten mechanism [Eq. (5a)–

(

5d)], was included as a regulative constraint. The kinetic

obs

constants k , k , and k

2

are usually determined in a way

1

ꢁ1

that the concentration profiles are fitted to such system of

differential equations. As the spectral factorization by

means of PCD is computed first this can be called an a pos-

teriori kinetic study. In contrast to this we used the error be-

tween a preliminary guess of the concentration profiles and

their least-squares-fit to the system of ordinary differential

equations as a further regularization of the PCD reconstruc-

tion functional. This approach can be considered as an

a priori kinetic analysis in which the structure of the differ-

ential equation regulates the result of the spectral factoriza-

tion by PCD. Such kinetic regularization is known from the

Figure 3. Hydroformylation of 3,3-dimethyl-1-butene. a) Concentration

profiles of olefin and aldehydes as obtained conventionally from FTIR

spectroscopy, and profiles of the organometallic components from PCD.

The concentration range of the rhodium complexes is given by the right

hand axis. Least-squares-fit from the Michaelis–Menten kinetic model

(

black curves: data, gray curves: regression). b) Comparison of the rate

profile obtained from numerical differentiation of the aldehyde concen-

tration profile (from the FTIR measurement, representing the 15% to

full conversion range) with the numerical product of the rate constant

obs

k

2

with the acyl complex (CatS) concentration profile, and linear de-

[23]

pendence between the rate of aldehyde formation and the acyl complex

concentration. Conditions: [Rh]=0.3, [TDTBPP]=6 mm; [olefin]

.9m; T=308C; pCO =1, p ₂=1 MPa; solvent: n-hexane.

literature.

Numerical experiments that compare the re-

0

=

sults of an a-posteriori and an a-priori kinetic analysis clear-

ly show an improved spectral reconstruction and a better ki-

0

H

[24]

netic modeling by means of the latter approach. Figure 3a

shows the results, namely the concentration profiles of the

organometallic species together with those of the organic

educt and the product, the latter as obtained from conven-

tional FTIR analysis. Each of these profiles compares well

to the corresponding least-squares-fit based on a Michaelis–

Menten kinetic model. We further assumed that the concen-

tration of the acyl complex reaches zero at full olefin con-

version. The regression procedure did only allow for the de-

that the formation of the acyl complex is fast with respect to

mixing, thus preventing the recording of reliable spectra

within approximately 2 min after olefin addition. The math-

ematical background is that the Hessian matrix of the recon-

struction error functional is nearly singular in the k and k

1

ꢁ1

directions. However, the K_mvalue, which is composed from

the three rate constants, is in very good agreement with that

from fitting the numerically integrated Michaelis–Menten

equation to the profile of the aldehyde concentration (see

Table 1). It is obvious that the investigated system with two

observable organic and two catalytic components can be de-

obs

termination of a distinct value of k₂, whereas the values

for k and k for the pre-equilibrium remained numerically

1

ꢁ1

unstable. This uncertainty is probably caused by the fact

Chem. Eur. J. 2012, 00, 0 – 0

ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

&

5

&

ÞÞ

These are not the final page numbers!

D. Selent et al.

Table 1. Kinetic constants obtained from different regression methods.

unchanged. For the calculation of the molar concentrations

of hydrogen in n-hexane at 308C we used literature data.

[

b]

[26]

System of ordinary

differential equations,

PCD![S], [Cat],

MM-equation,

FTIR+GC![P]

[

a]

It can be seen from Figure 4 that there is indeed a clear

obs

linear correlation between the values of k₂and the hydro-

AHCTUNGTRENNUNG[ CatS], [P]

obs

ꢁ1

k

V

K

2

[min

]

5.00

1.51ꢃ10

0.179

5.05

1.52ꢃ10

0.173

ꢁ

3

ꢁ1

ꢁ3

sat [moldm min

]

ꢁ

3

m

[moldm

]

[

a] Integration/regression results from fitting the numerically integrated

system of ordinary differential equations of the Michaelis–Menten mech-

anism [Eqs. (5a)–(5d)] to the four observable concentration profiles, see

Figure 3a. [b] Integration/regression results from fitting the numerically

integrated Michaelis–Menten equation [Eq. (4)] to the product concen-

tration profile, see Figure 1a.

scribed by a Michaelis–Menten-type kinetics over the full

conversion range.

Furthermore, Equation (5d) defines that the rate of prod-

uct formation is linearly correlated with the concentration

of the catalyst substrate complex. The expected result,

therefore, is getting identical plots when comparing the rate

profile of aldehyde formation with the profile obtained from

obs

the numerical product of the rate constant k₂and the con-

Figure 4. Hydroformylation of 3,3-dimethyl-1-butene. Influence of the hy-

centration of the intermediate acyl complex or alternatively

from the numerical product of V_s_a_tand the molar fraction of

the same acyl complex. The plots given in Figure 3b confirm

such a linear correlation, which has been found also for un-

obs

drogen concentration on k

2

. Conditions: [Rh]=0.3, [TDTBPP]=6 mm;

[

olefin] =0.9m; T=308C; pCO =1 MPa; solvent: n-hexane. For data on

0

[26]

hydrogen concentration in solution see Table 2.

ments are indicated by half-filled circles.

Repetition experi-

[3c,6a]

modified hydroformylation catalysis.

By this relation it

is obvious that the rate of product formation is limited by

the hydrogenolytic step of the mechanism over the entire

conversion range, which is also the rate-controlling step at

gen concentration in solution; respective data are given in

Table 2. The plot of ln ACHTUNGTRNENUG( k ) versus ln([H ]) verifies the

2 2

obs

high substrate concentrations, where [S]@K . In the region

linear relationship, a partial reaction order with respect to

the hydrogen concentration of 1.05 is obtained.

m

of the reaction where [S]!K , the free catalyst (Cat) is the

m

predominant complex and the rate constant is a composed

pseudo-first-order one. Therefore, in principle no step can

be assigned as the rate-controlling step.

Table 2. Head space hydrogen pressures applied, effective hydrogen con-

centration in solution as calculated from data given in reference [26], and

[13,25]

obs

constants k

p ₂[MPa]

.20

.50

2

.

Influence of the hydrogen concentration on the rate and on

the concentration profiles

ꢁ3

obs

ꢁ1 [b]

obs

ꢁ1 [c]

H

Conc. H

2

[mol dm

]

k

2

[min

]

k

2

[min ]

0

0.011

0.027

0.054

0.053

0.080

0.107

0.133

0.163

0.209

0.208

0.76

2.32

5.05

5.07

7.09

10.25

11.80

15.63

17.89

17.91

0.74

2.08

5.00

5.02

7.21

10.31

12.81

16.27

18.98

19.18

Characterization of the pre-equilibrium: The good agree-

ment between the Michaelis–Menten kinetic model applied

here and the experimentally observed dependency of the

rate of product formation on substrate concentration

prompted us to study the influence of the concentration of

the co-substrate hydrogen in more detail. Equations (1) and

1.01

[

a]

1.00

1.49

1.98

2.55

2.99

3

.81

[

a]

(

2) indicate that the hydrogen concentration will influence

[

a] Repetition experiments. [b] Values from Figure 4. [c] Values from

the rate characteristically and also the product concentration

obs

Figure 5.

profile. The pseudo constant k₂is, besides K , a directly

m

accessible physical quantity obtained by the integration/re-

gression procedure. Furthermore, changing the hydrogen

concentration will set a new balance between the concentra-

tions of Cat and CatS at any distinct point of olefin conver-

sion, thus also altering the concentration profiles for these

intermediates.

obs

The values of k₂were obtained from V_s_a_tafter fitting

the numerically integrated Michaelis–Menten equation to

[27]

3

ꢁ1

the aldehyde profiles. A value of 89.1 dm mol min for

k₂is directly derived from the slope. Two additional experi-

ments performed at 1 and 3.8 MPa hydrogen pressure, re-

spectively, verify the reproducibility of the results.

For all experiments a linear relationship is also found for

different hydrogen concentrations when the rate of product

A set of experiments was performed with different hydro-

gen partial pressures in the range between 0.2 and 3.8 MPa

H and a constant p =1 MPa, with all other parameters

2

CO

&

6

&

ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Chem. Eur. J. 0000, 00, 0 – 0

ÝÝ

These are not the final page numbers!

Conversion-Dependent Kinetics of Phosphite-Modified Hydroformylation Catalysis

FULL PAPER

To check this point we calculated the initial rates by using

Equation (6a) with V_s_a_tand K_mobtained from our experi-

ments and arbitrarily chosen initial substrate concentrations.

With these initial rates the observable partial reaction order

with respect to the hydrogen concentration was obtained

from the slope of a respective double logarithmic plot ln(V₀)

versus ln([H ]). It can be seen from Figure 6 that the reac-

2

Figure 5. Correlation of the product formation rate with the concentra-

tion of the intermediate acyl complex at different hydrogen head space

partial pressures for the hydroformylation of 3,3-dimethyl-1-butene.

Rates were calculated by numerical differentiation of the aldehyde con-

centration curve after smoothing. Conditions: [Rh]=0.3, [TDTBPP]=

6

0

mm; [olefin] =0.9m; T=308C; pCO =1 MPa; solvent: n-hexane.

formation is plotted against the concentration of the inter-

mediate acyl complex obtained from PCD (see Figure 5).

Head space pressures of hydrogen are given, because they

correlate linearly with the solution concentrations of H₂

within the pressure range applied (see Table 2). The slopes

Figure 6. Partial reaction order with respect to the hydrogen concentra-

tion for different initial substrate concentrations for the hydroformylation

of 3,3-dimethyl-1-butene. The calculation is based on the experimentally

determined constants V_s_a_tand K_mand Equation (6a).

obs

of those straight lines represent the respective k₂values,

which are listed in Table 2. By linear regression of the data

obs

set k₂values versus hydrogen concentration, the constant

tion under study is of nearly first order with respect to the

hydrogen concentration at the starting point of the reactions

with high olefin concentrations applied, but subsequently

decreases with olefin consumption. Such broken order has

3

ꢁ1

k was calculated to 93.3 dm mol min which is in good

2

3

ꢁ1

agreement with the former value of 89.1 dm mol min .

The observed first-order dependency of the product for-

mation rate on the hydrogen concentration is strictly only to

[6f]

indeed been found for other hydroformylation catalysts.

obs

be expected when one analyses the values of k₂versus the

From the definition of K_min Equation (2c), an informa-

tion concerning the establishment of the pre-equilibrium

can be obtained by variation of the hydrogen concentration.

hydrogen concentrations or when the pre-equilibrium is es-

tablished. Then, a first-order dependence is observed also

by using initial rates, independent from the catalyst satura-

tion. We will show below that the pre-equilibrium, which is

part of the Michaelis–Menten-type mechanism depicted in

Scheme 2, is not established and therefore the first case is

relevant for our reaction.

The first-order dependency illustrated in Figure 4 corre-

sponds to the limiting situation where [S]@K_mand all cata-

lytic material is in the form of the acyl complex and the rate

In the case of an established pre-equilibrium with k

@

1

ꢁ

[13]

k [H ], Equation (7a) is valid.

2

k

^k1

1

K

ꢁ1

K_m

¼

ð7aÞ

ð7bÞ

k

^k1

1

ꢁ

1

K_m

¼

þ _k k ½H ꢀ

2

1

obs

is determined by k₂. The K constant, as given with Equa-

m

Under these conditions, the constant K_mshould be inde-

pendent from the hydrogen concentration. On the other

hand, when the pre-equilibrium is not established a linear

dependence of K on [H ] is expected. The constants k and

tions (2) and (3), which do also contain the hydrogen con-

centration, then is negligible, thus the substrate concentra-

tion is eliminated from the equation. However when one

analyses the dependency of the initial rates from the hydro-

gen partial pressure for the case where [S] is in the order of

magnitude of K_mthe latter is not negligible and the sub-

strate concentration cannot be eliminated. This results in

a broken order with respect to the hydrogen concentration,

which will show a shift, depending on the (initial) substrate

concentration.

m

2

1

k

can then be estimated from a plot of K versus k [H ]=

m 2 2

ꢁ

1

obs

k₂[Eq. (7b)] (see Figure 7 and Table 3).

obs

There is a significant dependence of K_mon k₂. There-

fore, the pre-equilibrium is not established. The relation

seems to be linear, as expected, with a deviation for the first

data point that originates from the experiment at 0.2 MPa of

Chem. Eur. J. 2012, 00, 0 – 0

ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

&

7

&

ÞÞ

These are not the final page numbers!

D. Selent et al.

range as described above. Figure 8 shows two selected sets

of profiles, which were obtained at 0.2 and 3.81 MPa, respec-

[28]

tively, of hydrogen partial pressure. For comparison and

verification, each individual profile is accompanied by the

result of a simultaneous integration/regression of the set of

ordinary differential Equations (5a)–(5d).

obs

Figure 7. Plot of the Michaelis–Menten constant K

hydroformylation of 3,3-dimethyl-1-butene at different hydrogen partial

pressures. Conditions: [Rh]=0.3, [TDTBPP]=6 mm; [olefin] =0.9m; T=

08C; pCO =1 MPa; solvent: n-hexane. Open circles represent repetition

m

versus k

2

, for the

0

3

experiments.

m

Table 3. Dependency of the Michaelis constant K on the hydrogen par-

tial pressure.

ꢁ

3

p

H₂[MPa]

m

K [mol dm ]

0.20

0.50

1.01

1.00

1.49

1.98

2.55

2.99

3.81

0.03

0.11

0.17

0.21

0.26

0.28

0.34

0.36

[

a]

[

a]

[

a] Repetition experiments.

hydrogen. At such low head space pressures of hydrogen,

the determination of K is more difficult because the corre-

Figure 8. Influence of the hydrogen pressure (p_H₂= 0.20 (a) or 3.81 MPa

(b)). Concentration profiles and results of simultaneous integration/re-

gression for the educt, products (aldehyde sum), and observable rhodium

complexes (right-hand axis) during the hydroformylation of 3,3-dimethyl-

m

sponding small amounts of hydrogen present in solution can

lead to high molar fractions of the acyl complex also at low

substrate concentrations.

1

-butene. Conditions: [Rh]=0.3, [TDTBPP]=6 mm; [olefin]

0

=0.9m; T=

Constants obtained by taking all data points into account

3

08C; pCO =1 MPa; solvent: n-hexane (black curves: data, gray curves:

3

ꢁ1

were k =57.1 dm mol min and k =3.8 min . It is evi-

regression).

1

ꢁ1

obs

dent that k ₁is of similar magnitude, compared to k₂(see

ꢁ

Table 2). Therefore, a derivation of the rate equations based

on the quasi-equilibrium approximation is not valid for this

reaction system but a steady-state approach is suitable!

The results described above show, that the hydrogen con-

centration has a deciding influence on the behavior of the

catalytic system under study. An impressive tool for illus-

The profile set obtained for 0.2 MPa hydrogen pressure

shows the acyl rhodium complex to dominate over the hy-

drido complex until >90% of conversion, with a nearly zero

reaction order with respect to the olefin observed. This reac-

tion took approximately 4500 min for completion. In con-

trast, the acyl complex is quickly depopulated at 3.81 MPa

trating the consequences of the variation of [H ] are the

2

concentration profiles, which include pure organic compo-

nents but also the observable organometallic intermediates.

Such profile sets have been determined for all experiments

of the variation of the hydrogen partial pressure series. Each

profile was reconstructed for the entire olefin conversion

of H and a much faster aldehyde production takes place.

2

The pressure-dependent initial concentrations of the acyl

complex clearly indicate also that the pre-equilibrium is not

established.

&

8

&

ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Chem. Eur. J. 0000, 00, 0 – 0

ÝÝ

These are not the final page numbers!

Conversion-Dependent Kinetics of Phosphite-Modified Hydroformylation Catalysis

FULL PAPER

Further investigations on the intermediates

The hydrido complexes: After catalyst preformation two hy-

drides, [HRh(CO) L] and [HRh(CO) L ], are present as po-

3

2

tentially catalytically active species with the latter being the

minor component. Because it was not possible to determine

the exact molar fraction of that complex from the FTIR

spectra taken at p_C_O=p_H₂=1 MPa, we performed a systemat-

ic variation of the CO partial pressure at a constant hydro-

gen partial pressure of 1 MPa and the same Rh/P ratio of 20

as has been applied for the catalytic batches. At a very low

ꢁ3

CO partial pressure of approximately 10 MPa, only the vi-

brational bands of pure [HRh(CO) L ] are observable (see

2

Figure 9).

We used the ratio of the peak areas of this hydrido com-

ꢁ

1

ꢁ1

plex band at n˜ =2068 cm and the band at n˜ =1740 cm of

an internal reference (see the Experimental Section for fur-

ther quantification). Figure 9 illustrates how the fractions of

the complex depend on the CO partial pressure. There is

some variance in the data due to the low rhodium concen-

tration of 0.3 mm and the need for background subtraction.

As expected from the corresponding mass action law, the

fraction of the bis ACHTUNGTRENNUNGl i ACHTUNGTRENNUNGg and hydrido complex shows an inverse

relation to the partial pressure of carbon monoxide. By

least-squares fitting an equilibrium constant K=1.7 has

been determined for the substitution reaction depicted in

Scheme 4. From these data a significant mole fraction of

0.07 was calculated for the standard carbon monoxide pres-

sure of 1 MPa.

We assume that [HRh(CO) L ] is part of the pseudocom-

2

ponent consisting of quasi-equilibrated hydrido complexes,

and that any ligand exchange involving TDTBPP and CO

required to form the intermediate acyl rhodium complex is

fast.

Figure 9. a) Experimental FTIR spectrum of the pure bis ACHUTNRGENNUGl i ACHUTNGRENUNgG and hydrido

ꢁ

1

ꢁ3

complex, n˜ (CO)=2028 and 2068 cm at about pCO =10 /p ₂

upper trace). Spectrum of [HRh(CO) L] at the experimental standard

partial pressures, =1 MPa (lower trace). b) Fraction of

[HRh(CO) ] in dependence on the partial pressure of CO. Experi-

mental data according to the internal reference and fit according to

HRh(CO) ]/[Rh] =K[L]/(K[L]+[CO]). All measurements performed

in n-hexane, with [Rh]=0.3 and [TDTBPP]=6 mm.

H

=1 MPa

(

3

p

CO =p ₂

H

[

29]

Scheme 4. Equilibrium between the hydrido rhodium complexes formed

with tri(2,4-di-tert-butylphenyl)phosphite.

2 2

L

[

2

L

2

0

Earlier work did not consistently address the question of

the structure of the rhodium hydrido carbonyl complexes

formed with TDTBPP. This prompted us to perform DFT

calculations on molecules containing the actual phosphite

ligand. Because spectroscopy will not always allow directly

for a proper structural assignment of hardly isolable inter-

mediates, a comparison of experimental FTIR spectra with

such obtained from DFT calculations of a model compound

clearly be seen for e- ACHTUNGTERNNN[UG HRh(CO) L], where the spectrum

3

ꢁ

1

from DFT is red shifted by approximately 50 cm , but the

band pattern obtained compares well to the experimental

one.

For [HRh(CO) L ], a bisequatorial phosphite arrange-

2

ment can be deduced. The calculations predict a band at n˜ =

[30]

ꢁ1

can help with identification.

Geometry optimization and frequency calculations were

carried out for both isomers with an axially (a) and an equa-

1950 cm originating from a stretching vibration of the Rhꢁ

H bond, which, probably due to low intensity and solvent ef-

fects, has not been detected during the experiment.

torially (e) coordinated phosphite in [HRh(CO) L], as well

3

as for the e,e and e,a isomers of [HRh(CO) L ]. The calcu-

The acyl complex: The acyl rhodium complex formed in the

intermediate step of carbon monoxide insertion is not only

important for supporting the mechanistic rationale proposed

for the hydroformylation reaction. This complex represents

2

lated frequencies shown in Figure 10 were not scaled, there-

fore shifts in the wavenumbers between the calculated and

the experimental vibrational spectra naturally exist. This can

Chem. Eur. J. 2012, 00, 0 – 0

ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

&

9

&

ÞÞ

These are not the final page numbers!

D. Selent et al.

Figure 11. Hydroformylation of 3,3-dimethyl-1-butene at 308C in n-

hexane at pCO =4.65 and p ₂

0; [olefin] =0.9m. FTIR spectra of the carbonyl region taken 10 min

after olefin addition to the solution of preformed catalyst. Ac=acyl com-

plex [C 13C(O)Rh(CO) (TDTBPP)] (for an explanation of the band

pattern see text); aldehyde=4,4-dimethylpentanal and 2,3,3-trimethylbu-

H

=0.2 MPa; [Rh]=1.24 mm; TDTBPP/Rh=

2

0

6

H

3

ACHTUNGTRENNUNG

ꢁ

1

tanal. The drop of the base line at n˜ =1825 cm originates from the sub-

traction of olefin from the background consisting of n-hexane and 3,3-di-

methyl-1-butene.

Figure 10. Comparison of experimental (two lower traces) spectra and

spectra from DFT calculations of the rhodium hydride complexes formed

with tri(2,4-di-tert-butylphenyl)phosphite. For optimized geometries see

the Supporting Information.

batch methodology should further enhance the reliability of

such spectroscopic work. We partly followed this approach

by monitoring the spectra with a variety of carbon monox-

ide pressures ranging from 0.2 to 4.7 MPa, and also by

changing the TDTBPP/Rh ratio. To reduce background cor-

rection problems arising from extended olefin consumption,

[Rh] was set back to 0.3 mm. This also slowed down all reac-

tions enough to perform the IR spectroscopic measurements

within a 5% initial conversion range. The results are shown

in Figure 12. When compared to Figure 11, no change within

the spectral region of terminal carbonyls is observed. The

the only component of the catalytic cycle that is detectable

after olefin activation and before reformation of Cat. It

therefore justifies a demand that is essential for the kinetic

formalism applied here. Spectroscopic evidence for the for-

mation of an acyl complex formed during the hydroformyla-

tion of 1-octene has been provided for a Rh/tri(2-tert-butyl-

4

-methylphenyl)phosphite catalyst. The authors succeeded

ꢁ

1

ꢁ1

in detecting a band at n˜ =1690 cm of the rhodium-bound

band in question at n˜ =2072 cm does resist the pressure

acyl moiety by rapid scan technique before it was overlaid

with the band of the strongly absorbing aldehyde group.

changes, and also a reduced concentration of the monophos-

phite has no effect. We conclude, that the spectrum record-

[10d]

In our experiments, we could not observe a band at simi-

lar wavenumbers probably because of the low rhodium con-

centration of 0.3 mm applied. New measurements were

therefore done at [Rh]=1.24 mm. To further stabilize the

acyl complex, we adjusted the partial pressures to p_C_O=4.65

and p_H₂=0.2 MPa. Interestingly, the preformation of the hy-

drido complexes is now accompanied by a side reaction.

ed is indeed that of [C H C(O)Rh(CO) AHCTUNGTRNNEU(G TDTBPP)] and

6 13 3

that this acyl complex, in contrast to the results obtained

for the hydrido complexes, is not in equilibrium to a

detectable amount with the bisphosphite derivative

[C H C(O)Rh(CO) ACHTGNUTRNEUN(G TDTBPP) ].

2 2

6

13

One of the questions remaining is that of the structure of

[C H C(O)Rh(CO) (TDTBPP)]. An equatorial coordina-

AHCTUNGTRENNUNG

6

13

3

ꢁ

1

Bands for bridging CO groups at n˜ =1814 and 1835 cm

tion of the structurally comparable tri(2-tert-butyl-4-methyl-

phenyl)phosphite is discussed in the literature, with the axial

positions of the trigonal bipyramidal (tbp)-structured mole-

cule accommodated by the acyl fragment and a carbon mon-

ꢁ

1

and further bands at n˜ =2032 and 2049 cm are observed,

0

which we assign to dimeric Rh complexes. After the addi-

tion of the olefin, the hydrido complex instantly is converted

to the acyl complex, whereas the dimer disappears in

a slower reaction that takes 10 min. It also does react to the

acyl complex, without any further intermediate observable.

The modified reaction conditions allowed for the observa-

tion of an acyl carbonyl band, which is located at n˜ =

[10d]

oxide ligand, respectively.

a similar spectral pattern in the terminal carbonyl region as

is observed for the hydrido complex e-[HRh(CO) L]. For

For such geometry, we expect

AHCTUNGTRENNUNG

3

the TDTBPP ligand used in this study, such similarity is not

given. The spectrum of the corresponding acyl complex

compares well to that of matrix-isolated acyl complexes of

ꢁ

1

690 cm . This band gained intensity simultaneously with

ꢁ1

the known bands at n˜ =1995, 2019, 2072, and 2079 cm (see

rhodium and more stable acyl derivatives of cobalt, contain-

ꢁ

1

[6d,31]

Figure 11). The band at n˜ =2072 cm has not been observed

earlier with the same reaction studied in cyclohexane sol-

vent. Thus, the change of the methodology applied for mea-

surement and data processing allowed for better spectro-

scopic identification. Consequently, the application of a semi-

ing triphenylphosphine.

Thus, a structure with an axially

coordinated phosphite is probably given. We performed

DFT calculations as well as high-pressure (HP) NMR spec-

troscopic investigations to verify this hypothesis. For the cal-

culations the actual acyl moiety (CH ) CCH CH C(O), and

3

2

&

10

&

ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Chem. Eur. J. 0000, 00, 0 – 0

ÝÝ

These are not the final page numbers!

Conversion-Dependent Kinetics of Phosphite-Modified Hydroformylation Catalysis

FULL PAPER

Figure 13. Experimentally observed FTIR spectrum of the acyl complex

13C(O)Rh(CO) (TDTBPP)] and the respective spectra obtained

[

C

6

H

3

ACHTUNGTRENNUNG

from DFT calculations for the phosphite ligand coordinated in an axial

or equatorial position. For optimized geometries see the Supporting In-

formation.

ric vibrational modes as present for a C₃_vpoint group is

here cancelled and one frequency is shifted to the red of

ꢁ

1

about 16 cm . Both vibrational modes still have similar in-

tensities. However, the calculations do not explain the addi-

tional splitting of the symmetrical stretching mode absorbing

ꢁ

1

weakly at n˜ =2075 cm .

NMR spectroscopy has been used for the characterization

of isolable rhodium acyl complexes as well as of sufficiently

populated intermediates formed during catalysis with high

1

3

rhodium loadings applied. C NMR shifts for the carbonyl

2

carbon atom range from d=199.3 ppm ( J

in pentamethylcyclopentadienyl rhodium(III) b-trimethylsi-

lylpropionyl intermediates of intermolecular olefin hydroa-

ACHTUNGTRENNU(GN C,Rh)=63.6 Hz),

AHCTUNGTRENNUNG

ꢁ

1

cylation to d=235.4 ppm ( n˜ (CO)=1675(s), 1750 cm (m))

in dichloro and chelate complexes. An acetyl complex of

Rh bearing a tetrapod NP ligand could be prepared from

the corresponding hydride acetyl Rh complex by reduction

[33]

I

Figure 12. Hydroformylation of 3,3-dimethyl-1-butene at 308C in n-

hexane, [Rh]=0.3 mm, [olefin] =0.9m. FTIR spectra of the acyl complex

3

III

0

taken within a 5% olefin conversion range at a) different carbon monox-

ide partial pressures and b) different ratios of TDTBPP/Rh. Background

correction involves olefin and n-hexane.

1

with NaBH . For this complex, a H NMR resonance at d=

4

2

.03 ppm (q, J

A

H

U

G

R

N

U

G

ꢁ

1 [34]

methyl group, together with an IR band at n˜ =1575 cm .

Because of the low metal concentrations usually applied in

rhodium-catalyzed hydroformylation, advantage is taken

from labeling with CO for the direct detection of the qua-

ternary acyl carbon atom by C NMR spectroscopy. Thus,

a structurally intact TDTBPP ligand were used. The result is

shown in Figure 13, which compares the experimentally ob-

tained FTIR spectrum of the acyl complex with those ob-

tained from DFT calculations for both isomers. There is

a good agreement of the spectral pattern between the ex-

periment and the calculated spectrum representing a mole-

cule with the phosphite coordinated in an axial position.

Frequency analysis revealed, that both intense bands at n˜ =

1

3

1

3

1

3

acyl intermediates Rh C(O)

ACHTUNGENTRNUNG( CH ) Ph were unambiguously

2 2

detected in toluene at d=234–235 ppm with rhodium con-

[6e]

centrations of 21 mm applied. The signal of the Rh-bound

13

heptanoyl carbon became observable after CO versus CO

exchange and slowing down the rate of intra- and intermo-

lecular exchange processes in a combined IR and NMR

ꢁ

1

1995 and 2019 cm represent asymmetric stretching vibra-

[6f]

tions. The splitting of these terminal n(CO) bands can be at-

tributed to the carbonyl group of the acyl. Weinholdꢄs NBO

analysis shows that the oxygen lone pairs of the three CO li-

gands can donate charge into the corresponding RhꢁC(O)

spectroscopic study of 1-hexene hydroformylation. For hy-

droformylation catalysts modified by bulky monophosphites,

acyl rhodium complexes so far have been investigated inten-

sively only by FTIR spectroscopy.

[32]

anti bonds to different extent. This leads to a lengthening

of the RhꢁC(O) bond of about 0.025 ꢅ and a shortening of

the corresponding CꢁO bond (0.005 ꢅ) for the ligand that is

For our NMR spectroscopic study of the Rh/TDTBPP/

CO/H /3,3-dimethyl-1-butene system we used a modified

2

10 mm sapphire NMR tube allowing for precise control of

head space pressure and ensuring gas saturation of the

neighbored to the carbonyl group of the acyl. This results in

weaker coupling for one of the asymmetric vibrational

modes of the CO ligands. The degeneracy of both asymmet-

[35]

sample for the entire measurement.

To enhance the

chance of detection of rhodium acyl complexes, the concen-

Chem. Eur. J. 2012, 00, 0 – 0

ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

&

11

&

ÞÞ

These are not the final page numbers!

D. Selent et al.

tration of the rhodium precursor [(acac)Rh(CO) ] was set to

ACHTUNGTRENNUNG

2

.5 mm, representing an eight-fold concentration compared

cause the time-dependent concentration profiles for organic

1

to catalysis, in [D ]n-hexane. Partial pressures of H and of

as well as metal-organic components as seen in the

H

14

2

3

1

CO, 0.25 and 3.75 MPa, respectively, were adjusted to slow

down the hydrogenolytic aldehyde formation. The ratio Rh/

TDTBPP/3,3-dimethylbut-1-ene was 1:20:360 with the olefin

added directly to the mixture of catalyst precursor and

monophosphite at room temperature. The first P NMR

spectrum recorded under an argon atmosphere after dissolv-

ing all components, besides the signal of free ligand at d=

and P NMR measurements do fit qualitatively to the re-

sults of our in situ FTIR study we conclude that the new

proton signal observed at d=2.69 ppm belongs to

[37]

[C H C(O)Rh(CO) ACHTGNUTERNN(UGN TDTBPP)].

3

6

13

3

1

Conclusion

1

30.6 ppm, exhibited the sharp doublet of [(acac)Rh(CO)L]

at d=118.4 ppm (J(P,Rh)=293 Hz). Immediately after the

A

C

H

T

U

N

G

T

R

E

N

U

N

G

There are still a couple of interesting facets of olefin hydro-

formylation worth being studied. This also holds for the rho-

dium-catalyzed reaction, despite the fact that it is well estab-

lished in both, industry and academia. Our work presented

here does exemplify that in situ spectroscopy allows not

only for the identification of organometallic intermediates,

but also for a time-resolved determination of individual con-

centrations of components if the respective methodological

prerequisites are met. The kinetic approach we applied

takes benefit from the fact that data are available for the

entire olefin conversion range. Rates at catalyst saturation

became accessible by such kinetic analysis as well as further

details as are the concentration of the catalyst substrate

syngas was added, the signal of the rhodium complex started

to broaden, now indicating an equilibrium occurring be-

tween free and coordinated phosphite. After 2 h at room

temperature, the rhodium-precursor-to-catalyst transforma-

tion and the beginning of the hydroformylation catalysis is

1

3

observed. In the carbonyl region of the C NMR spectrum

peaks for dissolved CO (d=184.8 ppm) and acetylacetone

(

d=190.4 ppm), together with distinct amounts of 4,4-dime-

thylpentanal (d=198.3 ppm) and 2,3,3-trimethylbutanal (d=

01.9 ppm, 9:1 ratio of aldehydes) are assigned. Within the

subsequent hours a new broad phosphorus signal appeared

at d=134.1 ppm (d, J(P,Rh)=137 Hz), which shifted mar-

ginally and became sharper upon cooling to 241 K (d=

2

AHCTUNGTRENNUNG

obs

complex, and the value of the constant k₂. It was shown

1

5

35.5 ppm (d, J

% of signal intensity, pointing to the coordination of one

A

C

H

T

U

N

G

T

R

E

N

U

N

G

(P,Rh)=125.4 Hz)) and permanently held

with the help of K_mthat the pre-equilibrium is not estab-

lished, and therefore kinetic derivations based on the

steady-state approximation are suitable for this reaction

system. The rate-controlling step is dependent on the sub-

strate concentration, but is always limited by hydrogenolysis.

We conclude that the hydroformylation mechanism of the

transition-metal catalytic system investigated in this paper is

consistent with Michaelis–Menten-type kinetics. A further

interesting result is the difference in the coordination site,

equatorial versus axial, which is preferred by the phosphite

ligand upon coordination at the saturated hydrido and acyl

complexes, respectively. DFT calculations on these com-

plexes did not reproduce all features of the experimental IR

spectra in detail but do strongly support the initial assump-

tions of the respective assignments. Based on the results

from DFT calculations and as further analyzed with the

NBO theory, the origin of an additionally occurring asym-

metric stretching vibration of the carbon monoxide ligands

of the acyl complex is not a result of steric hindrance but of

distinct orbital interactions. There are certainly more of

such “hidden stories” to discover, which, complementary to

literature data, will lead to a better understanding of this

highly interesting type of catalyst.

phosphite to rhodium in this intermediate. The rhodium hy-

dride complexes remained below the detection limits during

a longer period of time, whereas large amounts of olefin

were converted to aldehyde (47% within 16 h at 301 K).

However, we were not able to detect signals for Rh-bound

carbonyls or the acyl group by measuring long term as well

as INEPT and HMBC C– H NMR spectra within the tem-

perature range of 301 and 241 K, even at stopped syngas

flow to further reduce the probability of acyl complex hy-

1

3

1

drogenolysis by dissolved H (d=4.55 ppm). In the H NMR

2

spectra, the intense signals of the olefin caused severe over-

laps. However, the appearance of a new broad triplet in the

proton spectrum at d=2.69 ppm, assignable to rhodium acyl

a-methylene protons, was synchronized with the observa-

tion of the phosphorus signal at d=134.1 ppm and correlat-

ed to a less intense carbon resonance at d=38.8 ppm. These

signals began to disappear after the olefin conversion

reached 90% and signals, which compare to that of the

[36]

known hydride complexes [HRh(CO) ACHTUNGTERNN(UNG TDTBPP)] (d=

3

136.7 ppm (d,

JACHTUNGTRENNUNG

A

C

H

T

U

N

G

T

R

E

N

U

N

G

(TDTBPP) ] (d=134.3 ppm (d, JACHTGNUTRENNUNG

2

[5d]

detectable in a 1:1 ratio.

Interestingly, even at these high olefin conversions the

acyl intermediate could be repopulated to be the predomi-

nant one simply by stopping the syngas flow in the NMR

cell, thus slowing down hydrogenolysis and reformation of

the rhodium hydride. Full olefin conversion within the

NMR sample was achieved within four days with 91% selec-

tivity in 4,4-dimethylpentanal. After depressurizing the

NMR cell and purging the solution with argon, [HRh(CO)₂-

Experimental Section

Materials: 3,3-Dimethyl-1-butene (>99% (GC), Sigma–Aldrich, 95%)

was distilled over sodium and stored under argon. n-Hexane (Sigma–Al-

drich, >99%) was distilled over Sicapent (Merck) and stored under

argon. Dodecane (Sigma–Aldrich, >99%), which was used as internal

GC-standard, was dried by storage over Sicapent for one week and dis-

tilled after separation of the drying agent and stored under argon. Fur-

&

12

&

ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Chem. Eur. J. 0000, 00, 0 – 0

ÝÝ

These are not the final page numbers!

Conversion-Dependent Kinetics of Phosphite-Modified Hydroformylation Catalysis

FULL PAPER

ther chemicals used in this study: acetylacetonato dicarbonyl rhodium(I)

39.46% Rh, Umicore), tri(2,4-di-tert-butylphenyl)phosphite (TDTBPP

Sigma–Aldrich, 98%). Gases used in this study: synthesis gas (CO/H

:1, from carbon monoxide 99.997% and hydrogen 99.999%, Linde),

carbon monoxide (99.997%, Linde), hydrogen (99.9993%, Linde), and

argon (99.999%, Linde).

was injected, and the FTIR measurements and the sampling procedure

were started.

(

ꢁ

1

2

=

FTIR spectra were recorded between n˜ =3950 and 700 cm with a spec-

ꢁ1

1

tral resolution of 2 cm . Per FTIR spectrum ten scans were collected

(double-sided, forward–backward) with a mirror speed set to 40 kHz. In-

tervals between two measurements were 37 s (for the first 120 min) and

after that 67 s.

As

a

phosphite stabilizer bis(2,2,6,6-tetramethyl-4-piperidyl)-sebacate

(

Tinuvin 770 DF, 100%, Ciba) was added in equimolar amounts to the

Rapid scan measurements: For these measurements the Bruker software

[

11]

ꢁ1

ligand.

tool OPUS-Chrom was used. The spectral resolution was set to 4 cm

.

One scan per spectrum was recorded (single-sided) with a delay time of

.244 s between spectra at a mirror speed of 40 kHz.

As an internal reference for the investigation of the equilibrium between

the hydrido complexes, the carbonyl band at n˜ =1740 cm of bis(2,2,6,6-

tetramethyl-4-piperidyl)-sebacate has been used.

Devices and procedures: The HP FTIR apparatus for performing the hy-

droformylation reactions consisted of a 200 mL stainless steel autoclave

with gas entrainment impeller and an oil bath thermostat (premex reac-

tor AG, Leimen, Germany) equipped with a heatable transmission flow-

through IR cell (Dr. Bastian Feinwerktechnik GmbH, Wuppertal, Ger-

many) and an automated sampling device (ASD) for taking GC samples

during the reaction (amplius GmbH, Rostock, Germany) (see Figure 14).

0

ꢁ

1

A 7890 A GC System from Agilent Technologies with a Petrocol DH 150

column (Supelco, Inc.) was used for GC analyses.

NMR spectra were obtained on a Bruker Avance 400 spectrometer by

using a high-pressure gas flow cell connected to devices allowing for con-

[

35]

tinuous gas supply and the control of gas flow and pressure.

Computational Details

Geometry optimization and frequency calculations were performed by

[

38]

using the Gaussian 03 and Gaussian 09 program packages. For all cal-

culations done we used the PBE exchange density functional, the PBE

gradient-corrected correlation density functional and the DGDZVP basis

[

39]

set. No imaginary frequencies were found, which indicates that the op-

timized structures are at least local minimum structures on the potential

energy surface.

Acknowledgements

We thank Evonik Industries AG for financial support. For helpful discus-

sions and suggestions we especially thank Prof. Detlef Heller.

[

1] a) Rhodium-Catalyzed Hydroformylation (Eds.: P. W. N. M. van

Leeuwen, C. Claver), Kluwer, Dordrecht, 2000; b) K.-D. Wiese, D.

Obst in Topics in Organometallic Chemistry, Vol. 18 (Ed.: M.

Beller), Springer, Heidelberg, 2006.

Figure 14. Scheme of the experimental HP FTIR apparatus equipped

with an automated sampling device.

[

2] Mechanisms in Homogeneous Catalysis (Ed.: B. Heaton), Wiley-

VCH, Weinheim, 2005.

Circulation of the reaction solution through the IR cell and back to the

autoclave was realized by a micro gear pump (mzr-7255, HNP Mikrosys-

teme GmbH, Parchim, Germany). A 6-port valve (Knauer GmbH,

Berlin, Germany) with a 6.5 mL loop was used for olefin injection. A

Bruker Tensor 27 FTIR spectrometer with a MCT-A detector was used.

ZnS (Korth Kristalle GmbH, Kiel, Germany) was used as window mate-

rial. The mean optical path length was 0.48 mm resulting from a wedged

[

3] a) M. Garland in Mechanisms in Homogeneous Catalysis (Ed.: B.

Heaton), Wiley-VCH, Weinheim, 2005 , pp. 151–193; b) C. Li, E.

Widjaja, M. Garland, J. Catal. 2003 , 213 , 126–134; c) J. Feng, M.

Garland, Organometallics 1999 , 18 , 417–427 ; d) C. Li, E. Widjaja,

M. Garland, J. Am. Chem. Soc. 2003 , 125 , 5540–5548 ; e) C. Li, L.

Chen, M. Garland, J. Am. Chem. Soc. 2007 , 129 , 13327–13334.

4] a) W. Chew, E. Widjaja, M. Garland, Organometallics 2002 , 21 ,

[

2

spacer. Pressurization facilities were installed for synthesis gas (CO/H =

Chem. 2003 , 75 , 4499–4507.

982–1990 ; b) E. Widjaja, C. Li, M. Garland, Organometallics 2002 ,

1 , 1991–1997 ; c) E. Widjaja, C. Li, W. Chew, M. Garland, Anal.

1

:1) as well as hydrogen and carbon monoxide. For regulation of the con-

2

sumed gas (CO/H =1:1) and keeping the pressure constant during the

reaction a pressure controller (Brooks Instrument, Hatfield, PA, USA)

was used. No mass transport limitation occurs with such experimental

5] a) D. Evans, J. A. Osborn, G. Wilkinson, J. Chem. Soc. A 1968 ,

3

133–3142; b) C. K. Brown, G. Wilkinson, J. Chem. Soc. A 1970 ,

753–2764 ; c) T. Jongsma, G. Challa, P. W. N. M. van Leeuwen, J.

[11]

setup under the conditions applied.

Solutions in n-hexane were prepared from the solids by using a syringe

and Schlenk techniques. All components except the olefin were trans-

ferred into the autoclave. The stirrer speed was set to 1500 rpm, the

speed of the micro-gear pump (displacement volume: 48 mL) was set to

Organomet. Chem. 1991 , 421 , 121–128; d) R. Crous, M. Datt, D.

Foster, L. Bennie, C. Steenkamp, J. Huyser, L. Kirsten, G. Steyl, A.

Roodt, J. Chem. Soc. Dalton Trans. 2005 , 1108–1116.

[6] a) M. Garland, P. Pino, Organometallics 1991 , 10 , 1693–1704; b) C.

Fyhr, M. Garland, Organometallics 1993 , 12 , 1753–1764 ; c) G. Liu,

R. Volken, M. Garland, Organometallics 1999 , 18 , 3429–3436 ; d) J.

Zhang, M. Poliakoff, M. W. George, Organometallics 2003 , 22 ,

1612–1618; e) J. M. Brown, A. G. Kent, J. Chem. Soc. Perkin Trans.

2

333 rpm. After heating the solution to 308C the overhead pressure was

about 0.15 MPa (argon, solvent vapor), the system was pressurized with

MPa carbon monoxide and the respective hydrogen partial pressure.

After preformation of the hydrido complexes was complete, the olefin

1

Chem. Eur. J. 2012, 00, 0 – 0

ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

&

13

&

ÞÞ

These are not the final page numbers!

D. Selent et al.

Dalton Trans. 2009 , 2496–2509; b) C. Godard, S. B. Duckett, S.

1987 , 1597–1607; f) S. C. van der Slot, P. C. J. Kamer, P. W. N. M.

van Leeuwen, J. A. Iggo, B. T. Heaton, Organometallics 2001 , 20 ,

30–441.

trum of the background including 3,3-dimethyl-1-butene. The zero

lines of the spectra were corrected. By computing the difference be-

tween these two spectra we obtained an approximation of the olefin

spectrum. Furthermore the concentration profile of the olefin

(during the hydroformylation process) was accessible from a flank

of the C=C valence vibration band at n˜ =1646.8 cm . The computed

olefin spectrum was multiplied with the olefin concentration and

then subtracted from the zero-line-corrected series of spectra ob-

tained from the hydroformylation reaction solution.

4

[

7] a) C. Godard, S. B. Duckett, S. Polas, R. Tooze, A. C. Whitwood,

Polas, R. Tooze, A. C. Whitwood, J. Am. Chem. Soc. 2005 , 127 ,

ꢁ

1

4

994–4995 ; c) L. Damoense, M. Datt, M. Green, C. Steenkamp,

Coord. Chem. Rev. 2004 , 248 , 2393–2407; d) C. Godard, S. B. Duck-

ett, C. Henry, S. Polas, R. Tooze, A. C. Whitwood, Chem. Commun.

Chemom. Intell. Lab. Syst. 1998 , 44 , 77–98.

004 , 1826–1827.

[23] a) A. de Juan, M. Maeder, M. Martꢉnez, R. Tauler, Chemom. Intell.

Lab. Syst. 2000 , 54 , 123–141 ; b) H. Haario, V.-M. Taavitsainen,

[24] For the details on the mathematical algorithm and the optimization

procedure see: M. Sawall, A. Bçrner, C. Kubis, R. Ludwig, D.

Selent, K. Neymeyr, unpublished results.

8] a) A. B. Permin, R. Eisenberg, J. Am. Chem. Soc. 2002 , 124 , 12406–

Chem. Rev. 2000 , 100 , 439–493.

7

2407; b) P. P. Deutsch, R. Eisenberg, Organometallics 1990 , 9 , 709–

18; c) L. Versluis, T. Ziegler, Organometallics 1990 , 9 , 2985–2992;

d) M. Solꢆ, T. Ziegler, Organometallics 1996, 15, 2611–2618; e) T.

Matsubara, N. Koga, Y. Ding, D. G. Musaev, K. Morokuma, Organo-

metallics 1997 , 16 , 1065–1078 ; f) M. Torrent, M. Solꢆ, G. Frenking,

9] Unsaturated complexes have been observed under non-catalytic

conditions only. a) D. Evans, G. Yagupsky, G. Wilkinson, J. Chem.

[25] K. J. Laidler, J. Chem. Educ. 1988 , 65 , 250–254.

[26] E. Brunner, J. Chem. Eng. Data 1985 , 30 , 269–273. Comment: solu-

bility data for hydrogen in n-hexane were available for T=298.15,

323.15, and 373.15 K. We calculated the solubility for our reaction

temperature of 303.15 K as follows: for each temperature mole frac-

tions for respective partial pressures of hydrogen were obtained

from linear regression. By ln(x)=A/T+B, mole fractions for

303.15 K were calculated for each partial pressure.

[27] See the Supporting Information (SI-D), for a presentation of the

graphs obtained from the fitting process for each hydrogen concen-

tration applied.

Soc.

A 1968 , 2660–2665; b) R. V. Kastrup, J. S. Merola, A. A.

Oswald, Adv. Chem. Ser. 1982 , No. 196, Chapter 3; c) M. A. Free-

man, D. A. Young, Inorg. Chem. 1986 , 25 , 1556–1560. For an 16-

electron acyl complex of iridium(I) see: G. Yagupsky, C. K. Brown,

G. Wilkinson, J. Chem. Soc. A 1970 , 1392–1401.

[

10] a) P. C. J. Kamer, J. N. H. Reek, P. W. N. M. van Leeuwen in Mecha-

nism in Homogeneous Catalysis (Ed.: B. Heaton), Wiley-VCH,

Weinheim, 2005, pp. 231–269; b) E. Zuidema, L. Escorihuela, T. Ei-

chelsheim, J. J. Carbꢇ, C. Bo, P. C. J. Kamer, P. W. N. M. van Leeu-

wen, Chem. Eur. J. 2008 , 14 , 1843–1853; c) A. van Rooy, P. C. J.

Kamer, P. W. N. M. van Leeuwen, K. Goubitz, J. Fraanje, N. Veld-

man, A. L. Spek, Organometallics 1996 , 15 , 835–847; d) P. C. J.

Kamer, A. van Rooy, G. C. Schoemaker, P. W. N. M. van Leeuwen,

Coord. Chem. Rev. 2004 , 248 , 2409–2424; e) A. van Rooy, E. N.

Orij, P. C. J. Kamer, F. van den Aardweg, P. W. N. M. van Leeuwen,

J. Chem. Soc. Chem. Commun. 1991 , 1096–1097; f) A. van Rooy,

E. N. Orij, P. C. J. Kamer, P. W. N. M. van Leeuwen, Organometallics

[28] For the full set of concentration profiles as well as a table containing

obs

the values of k

2

m

and K , which were determined for different hy-

drogen pressures, see the Supporting Information (SI-E).

[29] R. Koelliker, H. Thies, J. Chem. Eng. Data 1993 , 38 , 437–440. We

calculated the solubilities for our reaction temperature at 303.15 K

the same way as in reference [23] from the data for T=293.15,

323.15, and 373.15 K.

[30] a) A. D. Allian, Y. Wang, M. Saeys, G. M. Kuramshina, M. Garland,

Vibrational Spectroscopy 2006 , 41 , 101–111 ; b) A. D. Allian, M.

Tjahjono, M. Garland, Organometallics 2006 , 25 , 2182–2188; c) A.

Christiansen, C. Li, M. Garland, D. Selent, R. Ludwig, R. Franke,

A. Bçrner, ChemCatChem 2010 , 2 , 1278–1285; d) A. Christiansen,

C. Li, M. Garland, D. Selent, R. Ludwig, A. Spannenberg, W. Bau-

mann, R. Franke, A. Bçrner, Eur. J. Org. Chem. 2010 , 2733–2741;

e) A. Boddien, B. Loges, F. Gꢁrtner, C. Torborg, K. Fumino, H.

Junge, R. Ludwig, M. Beller, J. Am. Chem. Soc. 2010 , 132 , 8924–

8934; f) F. Gꢁrtner, A. Boddien, E. Barsch, K. Fumino, S. Losse, H.

Junge, D. Hollmann, A. Brꢀckner, R. Ludwig, M. Beller, Chem.

Eur. J. 2011 , 17 , 6425–6436; g) D. Selent, R. Franke, C. Kubis, A.

Spannenberg, W. Baumann, B. Kreidler, A. Bçrner, Organometallics

2011 , 30 , 4509–4514.

995 , 14 , 34–43; g) A. van Rooy, J. N. H. de Bruijn, K. F. Roobek,

P. C. J. Kamer, P. W. N. M. van Leeuwen, J. Organomet. Chem. 1996 ,

07 , 69–73.

5

[

11] C. Kubis, R. Ludwig, M. Sawall, K. Neymeyr, A. Bçrner, K.-D.

Wiese, D. Hess, R. Franke, D. Selent, ChemCatChem 2010 , 2 , 287–

pp. 490 –508; c) D. G. Blackmond, Angew. Chem. 2005 , 117 , 4374–

95.

12] A. A. Dabbawala, H. C. Bajaj, R. V. Jasra, J. Mol. Catal. A 2009,

02, 97–106.

13] H.-J. Drexler, A. Preetz, T. Schmidt, D. Heller in The Handbook of

Homogeneous Hydrogenation (Eds.: J. G. de Vries, C. J. Elsevier),

Wiley-VCH, Weinheim, 2007, pp. 257–293.

[

3

[

14] C. Bergounhou, D. Neibecker, R. Mathieu, J. Mol. Catal. A 2004,

[31] a) P. C. Ford, S. Massick, Coord. Chem. Rev. 2002 , 226 , 39–49;

b) S. M. Massick, J. G. Rabor, S. Elbers, J. Marhenke, S. Bernhard,

J. R. Schoonover, P. C. Ford, Inorg. Chem. 2000 , 39 , 3098–3106; c) J.

Somlyai-Haꢈsz, F. Haꢈz, V. Galamb, A. Benedetti, C. Zucci, G.

Pꢈlyi, T. Krꢀmmling, B. Happ, T. Bartik, J. Organomet. Chem. 1991,

419, 205–217; d) I. Kovꢈcs, F. Ungvꢈry, Coord. Chem. Rev. 1997,

161, 1–32.

2

20, 167–182.

[

15] a) F. G. Helfferich, Kinetics of Homogeneous Multistep Reactions in

Comprehensive Chemical Kinetics, Vol. 38, 2nd ed. (Ed.: N. J. B.

Green), Elsevier, Amsterdam, 2004; b) J. H. Espenson in Encyclope-

dia of Catalysis, Vol. 4 (Ed.: I. T. Horvꢈth), Wiley, New York, 2003 ,

4

393; Angew. Chem. Int. Ed. 2005 , 44 , 4302–4320; d) D. Murzin, T.

[32] a) NBO Version 3.1, E. D. Glendening, A. E. Reed, J. E. Carpenter,

F. Weinhold; b) A. E. Reed, L. A. Curtiss, F. Weinhold, Chem. Rev.

1988 , 88 , 899–926; c) F. Weinhold, C. R. Landis, Valency and Bond-

ing, Cambridge University Press, Cambridge, 2005. See the Support-

ing Information, SI-G, for a detailed description of the analysis.

[33] a) A. H. Roy, C. P. Lenges, M. Brookhart, J. Am. Chem. Soc. 2007 ,

129 , 2082–2093; b) G. L. Moxham, H. Randell-Sly, S. K. Brayshaw,

A. S. Weller, M. Willis, Chem. Eur. J. 2008 , 14 , 8383–8397; c) J. A.

Miller, J. A. Nelson, Organometallics 1991 , 10 , 2958–2961; d) M. A.

Garralda, R. Hernꢈndez, L. Ibarlucea, E. Pinilla, M. R. Torres, M.

Zarandona, Organometallics 2007 , 26 , 1031–1038.

Salmi, Catalytic Kinetics, Elsevier, Amsterdam, 2005 .

[

16] J. W. Rathke, R. J. Klingler, T. R. Krause, Organometallics 1991 , 10 ,

350–1355.

[

17] J. A. Christiansen, Adv. Catal. 1953 , 5 , 311–353.

18] E. J. King, C. Altman, J. Phys. Chem. 1956 , 60 , 1375–1378.

19] See the Supporting Information, SI-B, for plots of the GC data.

20] a) See the Supporting Information, SI-C, for results of the simula-

tion of the Michaelis–Menten mechanism and the following integra-

tion/regression procedure; b) K. J. Laidler, Can. J. Chem. 1955 , 33 ,

metallics 1991 , 10 , 820–823.

614–1624.

[

21] K. Neymeyr, M. Sawall, D. Hess, J. Chemom. 2010, 24, 67–74.

22] Otherwise a successful factorization appeared to be impossible. To

get rid of these perturbing signals we used a separate spectrum of

the background at the reaction conditions, and also a separate spec-

[34] C. Bianchini, A. Meli, M. Peruzzini, F. Vizza, F. Bachechi, Organo-

[35] D. Selent, W. Baumann, A. Bçrner (LIKAT e.V.), DE 10333143,

17.07., 2003.

&

14

&

ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Chem. Eur. J. 0000, 00, 0 – 0

ÝÝ

These are not the final page numbers!

Conversion-Dependent Kinetics of Phosphite-Modified Hydroformylation Catalysis

FULL PAPER

[

36] In [CH

3

CH

2

C(O)Ir

A

C

H

T

U

N

G

T

R

E

N

U

N

G

(dppe)(CO)] (dppe=1,2-bis(diphenylphosphino)-

[39] a) J. P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 1996 , 77 ,

3865–3868 ; b) J. P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett.

1997 , 78 , 1396 ; c) N. Godbout, D. R. Salahub, J. Andzelm, E.

Wimmer, Can. J. Chem. 1992 , 70 , 560–571; d) C. Sosa, J. Andzelm,

B. C. Elkin, E. Wimmer, K. D. Dobbs, D. A. Dixon, J. Phys. Chem.

1992 , 96 , 6630–6636.

ethane), the propionyl methylene protons resonate at d=2.83 ppm

ꢁ

1

(

n˜ =1635 cm ), see reference [8b].

[

37] See the Supporting Information, SI-G, for a selection of spectra.

38] a) Gaussian 03, Revision D.01, M. J. Frisch et al. Gaussian Inc.,

Wallingford CT, 2004; b) Gaussian 09, Revision A.1, M. J. Frisch

et al. ; Gaussian Inc., Wallingford CT, 2009. For the complete list of

authors see the Supporting Information, SI-H.

Received: February 23, 2012

Published online: && &&, 2012

Chem. Eur. J. 2012, 00, 0 – 0

ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

&

15

&

ÞÞ

These are not the final page numbers!

D. Selent et al.

Catalysis

C. Kubis, D. Selent,* M. Sawall,

R. Ludwig, K. Neymeyr, W. Baumann,

R. Franke, A. Bçrner ........ &&&& — &&&&

Exploring Between the Extremes:

Conversion-Dependent Kinetics of

Phosphite-Modified Hydroformylation

Catalysis

Catalyst at work: Organic as well as

organometallic components have been

followed by FTIR spectroscopy during

the rhodium-catalyzed hydroformyla-

tion. All concentration profiles

the acyl complex intermediate is not

established. Spectroscopic and results

from DFT calculations show that the

coordination mode of the phosphite in

the acyl complex is different from that

in the corresponding hydride (see

figure).

obtained fit to Michaelis–Menten-type

kinetics. The pre-equilibrium towards

&

16

&