Mechanistic investigation of the ruthenium-N-heterocyclic-carbene-catalyzed amidation of amines with alcohols

Article abstract of DOI:10.1002/chem.201202400

The mechanism of the ruthenium-N-heterocyclic-carbene-catalyzed formation of amides from alcohols and amines was investigated by experimental techniques (Hammett studies, kinetic isotope effects) and by a computational study with dispersion-corrected density functional theory (DFT/M06). The Hammett study indicated that a small positive charge builds-up at the benzylic position in the transition state of the turnover-limiting step. The kinetic isotope effect was determined to be 2.29(±0.15), which suggests that the breakage of the C-H bond is not the rate-limiting step, but that it is one of several slow steps in the catalytic cycle. Rapid scrambling of hydrogen and deuterium at the α position of the alcohol was observed with deuterium-labeled substrates, which implies that the catalytically active species is a ruthenium dihydride. The experimental results were supported by the characterization of a plausible catalytic cycle by using DFT/M06. Both cis-dihydride and trans-dihydride intermediates were considered, but when the theoretical turnover frequencies (TOFs) were derived directly from the calculated DFT/M06 energies, we found that only the trans-dihydride pathway was in agreement with the experimentally determined TOFs. On the right pathway: Experimental and theoretical investigations indicate that the Ru-catalyzed amidation of amines with alcohols (see scheme) proceeds by alcohol coordination, β-hydride elimination, nucleophilic attack, and β-hydride elimination to the amide. Copyright

Full text of DOI:10.1002/chem.201202400

FULL PAPER
DOI: 10.1002/chem.201202400
Mechanistic Investigation of the Ruthenium–N-Heterocyclic-Carbene-
Catalyzed Amidation of Amines with Alcohols
Ilya S. Makarov, Peter Fristrup,* and Robert Madsen*[a]
Abstract: The mechanism of the ruthe-
nium–N-heterocyclic-carbene-catalyzed
formation of amides from alcohols and
amines was investigated by experimen-
tal techniques (Hammett studies, kinet-
ic isotope effects) and by a computa-
tional study with dispersion-corrected
density functional theory (DFT/M06).
The Hammett study indicated that a
small positive charge builds-up at the
benzylic position in the transition state
of the turnover-limiting step. The kinet-
ic isotope effect was determined to be
2.29
(ꢀ0.15), which suggests that the
catalytically active species is a rutheni-
um dihydride. The experimental results
were supported by the characterization
of a plausible catalytic cycle by using
DFT/M06. Both cis-dihydride and
trans-dihydride intermediates were
considered, but when the theoretical
turnover frequencies (TOFs) were de-
rived directly from the calculated DFT/
M06 energies, we found that only the
trans-dihydride pathway was in agree-
ment with the experimentally deter-
mined TOFs.
ꢁ
breakage of the C H bond is not the
rate-limiting step, but that it is one of
several slow steps in the catalytic cycle.
Rapid scrambling of hydrogen and
deuterium at the a position of the alco-
hol was observed with deuterium-la-
beled substrates, which implies that the
Keywords: amides · density func-
tional calculations · isotope effect ·
reaction mechanisms · ruthenium
Introduction
procedures, it is important to gain a better understanding of
the underlying reaction mechanisms. So far, the umpolung
approach with a-bromonitroalkanes and the decarboxylative
pathway with a-ketoacids have been studied with 18O-la-
beled substrates,[7] whilst the first step in the dehydrogena-
tive reaction with ruthenium pincer complexes has been in-
vestigated by low-temperature NMR spectroscopy.[8] In ad-
dition, computational studies have been performed on the
ruthenium-catalyzed transformations with pincer and dia-
mine–diphosphine complexes.[9]
The amidation reaction between primary alcohols and
amines catalyzed by ruthenium–N-heterocyclic-carbene
complexes was first reported by our group in 2008;[5b] since
then, this transformation has been further investigated with
regard to catalyst precursors and substrate scope.[10] This re-
action is most-effectively cata-
The (carbox)amide group constitutes one of the most-signifi-
cant functional groups in organic chemistry. The synthesis of
amides is usually performed from carboxylic acids and
amines by using a coupling reagent or by prior conversion
of the acid into an activated derivative. Recently, however,
several new and fundamentally different approaches to
amide synthesis have been developed.[1] These emerging
methods include the umpolung synthesis from a-bromoni-
troalkanes,[2] the decarboxylative condensation of a-ketoa-
cids and hydroxylamines,[3] and the metal-catalyzed coupling
of primary alcohols and amines.[4–6] This latter reaction can
be carried out by aerobic oxidation with heterogeneous gold
catalysts[4] or by dehydrogenation with various homogeneous
and heterogeneous catalysts.[5,6] The dehydrogenative syn-
thesis of amides from alcohols and amines can be performed
both in the presence or absence of a hydrogen acceptor,
such as a ketone or an alkene.[5,6] The most-attractive proce-
dures would avoid the need for hydrogen scavengers alto-
gether, which can be achieved with homogeneous ruthenium
pincers,[5a] carbenes,[5b,c] and diamine–diphosphine[5d] com-
lyzed by complex 1 (Figure 1)
in the presence of PCy3 and
KOtBu. No stoichiometric addi-
tives are necessary and the ami-
dation reaction produces hydro-
gen gas as the only byproduct.
[5e]
plexes, as well as with heterogeneous Ag/Al2O3 and Au/
The reaction is believed to pro-
ceed through dehydrogenation
of the alcohol into the corre-
hydrotalcite catalysts.[5f] To further develop these amidation
Figure 1. Structure of com-
pound 1.
sponding aldehyde, which stays
coordinated to the ruthenium
[a] I. S. Makarov, Dr. P. Fristrup, Prof. Dr. R. Madsen
Department of Chemistry, Building 201
Technical University of Denmark, 2800 Kgs. Lyngby (Denmark)
Fax : (+45)4593-3968
catalyst (Scheme 1). Then, nucleophilic attack by the amine
forms the hemiaminal, which is dehydrogenated to afford
the amide. The fact that the intermediate aldehyde remains
coordinated to the ruthenium center is an important obser-
vation that has been confirmed experimentally by a cross-
E-mail: pf@kemi.dtu.dk
rm@kemi.dtu.dk
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/chem.201202400.
Chem. Eur. J. 2012, 18, 15683 – 15692
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
15683

high yield with both electron-donating and electron-with-
drawing groups in the para position (Table 1).
Because the amidation reaction occurs with the negligible
formation of byproducts, the course of the competition reac-
tions could be followed by measuring the disappearance of
the alcohols by using GC. Assuming that all of the sub-
strates react according to the same mechanism and that the
reaction is first order in the alcohol, their relative reactivi-
Scheme 1. Dehydrogenative synthesis of amides from alcohols and
amines.
over experiment.[10e] However,
the mechanism of the amida-
tion reaction has not previously
been subjected to a more-thor-
ough examination.
Herein, we report a com-
bined experimental and theo-
retical mechanistic investigation
of the formation of amides
from alcohols and amines cata-
lyzed by ruthenium–N-hetero-
cyclic-carbene complex 1.
Scheme 2. Competition experiments for the amidation of benzyl alcohol versus that of para-substituted benzyl
alcohols 2b–2g.
Table 1. Amidation of para-substituted benzyl alcohols 2a–2g.
Results and Discussion
For the experimental study, the catalyst system was modified
slightly to obtain more-accurate and reproducible results.
Tricyclohexylphosphine is easily oxidized by air and com-
mercial samples contain various amounts of impurities that
are difficult to remove. To solve this problem, Netherton
and Fu replaced trialkylphosphines with their corresponding
HBF4 salts in several metal-catalyzed reactions.[11] These
salts are air-stable and the phosphine can be released into
the reaction with a Brønsted base. Moreover, because a
base is already required for the amidation reaction, we de-
cided to replace PCy3 with PCy3·HBF4 and increase the
amount of KOtBu accordingly. This modification gave
more-consistent results and, therefore, the experimental
mechanistic study was performed on the following system:
Entry
Compound
X
Yield [%][a]
1
2
3
4
5
6
7
2a
2b
2c
2d
2e
2 f
2g
H
CF3
F
90
70
80
94
100
88
Me
OMe
SMe
NMe2
100
[a] Yield of isolated product after 24 h.
ties (kX/kH) can be obtained as the slope of the line when
ln(c0/c) for one para-substituted benzyl alcohol is plotted
against the same values for benzyl alcohol. This plot result-
ed in good linear correlations for all six para-substituted
benzyl alcohols (Figure 2), which confirmed the assumption
that the amidation reaction was first order in the alcohol. In
each case, the correlation coefficient was ꢂ0.99 and the
benzyl alcohols with electron-donating para substituents re-
acted faster than alcohols with electron-withdrawing groups.
The slopes could then be used to construct the Hammett
plot based on s values from the literature[13] (Figure 3). The
best correlation was achieved with s+ values, which afforded
a straight line with a small negative slope (1=ꢁ0.15). This
result indicates that a small positive charge is build-up at
the benzylic position in the transition state of the turnover-
[RuCl2ACHTUNGTRENNUNG(IiPr)ACHTUNGTRENNUNG(p-cymene)] (5 mol%), PCy3·HBF4 (5 mol%),
and KOtBu (15 mol%) in refluxing toluene.
Hammett study: We have previously used Hammett studies
with para-substituted benzaldehydes to analyze the turn-
over-limiting step in several metal-mediated reactions.[12]
This method makes it possible to determine the change in
charge at the benzylic position between the starting material
and the transition state. Subsequently, the development of
charge can be simulated in silico, based on a proposed cata-
lytic cycle, and can be used to discriminate between differ-
ent mechanistic scenarios. In this case, the Hammett study
will be comprised of a series of competition reactions be-
tween benzyl alcohol and various para-substituted benzyl al-
cohols with hexylamine as the amine component
(Scheme 2). First, the amidation reactions of hexylamine
with different para-substituted benzyl alcohols were carried
out and the reaction was found to proceed cleanly and in
C
limiting step. The correlation with Creary’s s values was
poor and, therefore, a radical intermediate is not involved in
the amidation reaction. The oxidation of an alcohol into an
amide is likely to proceed through two consecutive b-hy-
dride eliminations, which result in the formation of an alde-
hyde and an amide, respectively. Either of these elimination
steps are good candidates for the turnover-limiting step be-
15684
www.chemeurj.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2012, 18, 15683 – 15692

ꢁ
Ru NHC-Catalyzed Amidation of Alcohols and Amines
FULL PAPER
Figure 2. Kinetic data for the amidation of para-substituted benzyl alco-
hols 2b–2g in competition with compound 2a (“0” denotes initial con-
centration, X is the concentration of compounds 2b–2g, and H is the
concentration of compound 2a).
Figure 3. Hammett plot for the amidation of alcohols 2a–2g.
cause, in both cases, a partial positive charge is formed at
the benzylic position in the transition state.
Reaction order: To gain more information about the role of
each component, we sought to determine the kinetic order
in these constituents. First, the order in the amine was exam-
ined by varying the concentration of hexylamine from 0.2–
0.5m whilst keeping the concentration of benzyl alcohol
(0.5m) and the ruthenium catalyst (25 mm) constant. The ini-
tial rate of the reaction was determined for each concentra-
tion of amine and the rates were plotted against the amine
concentrations to give a straight line, thus indicating a first-
order dependence in the amine (Figure 4a).
The reaction order in phosphine was examined by varying
the concentration of PCy3 from 0–5 mm whilst keeping the
concentrations of benzyl alcohol and hexylamine constant
(0.2m). A non-linear dependence was observed and an
order of 0.5 in phosphine was derived from a double-loga-
rithmic plot (Figure 4b). This reaction order will be further
addressed in the computational study and used to pinpoint
the role of phosphine in the reaction mechanism.
Figure 4. a) Plot of camine versus rinit; b) plot of lncPCy versus lnrinit; c) Plot
3
of lnc1 versus lnr’init
.
made it impossible to obtain reproducible results. At higher
loadings (>6%), the initial rate measurements were also ac-
companied by significant uncertainties. After some experi-
mentation, we found that more-reproducible results could
be obtained if complex 1 was treated with PCy3·HBF4 and
KOtBu in refluxing toluene for 45 min before the alcohol
and the amine were added. With this modification, the con-
centration of compound 1 was varied from 5–37 mm, whilst
the concentrations of benzyl alcohol and hexylamine were
kept constant (0.5m). These data resulted in a straight line,
as shown in Figure 4c, thus indicating a first-order depend-
ence on the ruthenium catalyst. The experiments in Fig-
ure 4a were repeated under these slightly modified condi-
tions and the same linear dependence was observed, which
illustrates that the overall kinetics of the reaction have not
been altered. The minimum amount of ruthenium catalyst
for complete conversion under these conditions was about
1%, where the complete amidation of hexylamine with 2-
Attempts to establish the reaction order in the ruthenium
catalyst were initially met with difficulties. At low catalyst
loadings (ꢃ2% or 10 mm), the amidation either proceeded
with a long initiation period or did not proceed at all, which
Chem. Eur. J. 2012, 18, 15683 – 15692
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
15685

I. S. Makarov, P. Fristrup, R. Madsen
phenylethanol was observed in 24 h. This result should be
compared to 0.1% loading with a ruthenium pincer com-
plex[5a] and to 4–5% loading with ruthenium–triazolylidi-
ne[5c] and diamine–diphosphine[5d] complexes.
As a result, the exchangeable protons are most likely the
ꢁ
two a protons on the primary alcohol, the O H proton, and
ꢁ
the two N H protons in the primary amine. In a second ex-
periment, non-deuterated 2-phenylethanol was reacted with
BnND2 and the deuterium content in the starting alcohol
was monitored again as the reaction progressed (Figure 5,
entry 2). This experiment showed that two hydrogen atoms
in the alcohol were exchanged with deuterium and that the
reaction must occur with the a protons because exchange of
Deuterium-labeled substrates: Initially, we planned to deter-
mine the kinetic isotope effect (KIE) by using competition
experiments in a similar way to the Hammett study. [a,a-
D2]Benzyl alcohol would be allowed to compete with benzyl
alcohol in a reaction with hexylamine and the KIE would be
calculated by measuring the disappearance of the two alco-
hols, which would be separable by GC. However, we quickly
discovered that this simple experiment was not feasible be-
cause a rapid scrambling of deuterium and hydrogen occur-
red at the a position of the alcohol under the amidation con-
ditions. Accordingly, we decided to study this scrambling in
more detail and, to exclude a possible side-reaction with
benzylic radicals, these experiments were performed with 2-
phenylethanol (3) as the alcohol substrate.
ꢁ
the O H proton cannot be measured by GCMS. The scram-
bling occurred in such a fashion that an equilibrium was
slowly reached at which the hydrogen/deuterium ratio for
the two a protons was 3:2. This ratio is the same as that be-
tween the exchangeable hydrogen and deuterium atoms in
the starting materials and confirms that all five protons take
part in the scrambling. This result was verified by repeating
the experiment with [D2]-a,a-2-phenylethanol and non-deu-
terated benzylamine (Figure 5, entry 3). Again, equilibrium
was reached at which the hydrogen/deuterium ratio for the
two a protons in the alcohol was 3:2. However, in this case,
the exchange occurred much more rapidly and was observed
even before the amide had started to form. This result
means that there is a reversible step at the beginning of the
reaction, which most-likely involves a b-hydride elimination
and a migratory insertion. More significantly, the scrambling
implies that a ruthenium–dihydride species is involved in
the catalytic cycle.
First, the source of the atom scrambling was determined
(Table 2). The reaction between non-deuterated 2-phenyle-
thanol and benzylamine was performed in [D8]toluene and
Table 2. Experiments to determine the positions of the scrambled atoms.
This result suggests that the two chloride ligands in com-
plex 1 are not present in the catalytically active species,
which, instead, is likely to be a ruthenium–dihydride species.
To gain further support for this rationale, the two chloride
atoms in complex 1 were replaced with a different halogen
group. It has previously been shown that diiodide complexes
Entry
R1
R2
Solvent
1
2
3
H (3)
H (3)
D ([D2]-3)
H (4)
D ([D2]-4)
H (5)
[D8]toluene
toluene
toluene
[RuI2ACHTNUTRGENNUG(NHC)ACHTUNGTREN(NGUN p-cymene)] (NHC=IMe, IiPr, and ICy) will
the relative amounts of non-deuterated, mono-deuterated,
and di-deuterated alcohols were monitored by GCMS. This
experiment showed no change in the deuterium content of
the alcohol during the reaction, which demonstrates that no
exchange with the solvent occurs (Figure 5, entry 1).
also catalyze the amidation reaction,[10e,h] although the yields
vary, possibly owing to the lower stabilities of these com-
plexes. However, when we measured the initial rate with
complex [RuI2ACTHNUTRGENNG(U IiPr)ACHUTNTGREN(NUGN p-cymene)] under the standard condi-
tions with benzyl alcohol and hexylamine, we obtained a
value of 3.09 mmminꢁ1. This value is essentially the same as
with complex 1 (3.37 mmminꢁ1), which strongly suggests that
the halides are not bound to the ruthenium center in the
catalytic cycle.
Kinetic isotope effect: With the knowledge of the deuterium
scrambling and the atoms that are involved in this process
in hand, we designed an experiment to determine the KIE
of the overall reaction. First, the hydrogen atoms at the ex-
changeable positions in both the alcohol and the amine
were replaced with deuterium and the initial rates for both
the deuterated and non-deuterated substrates were then
measured in two separate experiments. For convenience,
commercially available and fully deuterated [D10]-1-butanol
was selected as the alcohol, whilst BnND2 was chosen as the
amine part. The initial rate with these deuterated substrates
was 6.44
ACHTUNGTRENNUNG
Figure 5. Experiments with deuterium-labeled substrates.
ed 1-butanol and benzyl amine was 14.77
N
15686
www.chemeurj.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2012, 18, 15683 – 15692

ꢁ
Ru NHC-Catalyzed Amidation of Alcohols and Amines
FULL PAPER
These values gave an experimental KIE of 2.29
G
ed phosphine. Therefore, these spectroscopic data provide
further support for the formation of mono- and dihydride–
ruthenium species during the reaction. In addition, the
NMR spectroscopy experiments have demonstrated that
complexes both with and without phosphine are present in
the reaction mixture.
ꢁ
which suggests that the breakage of the C H bond is not
the rate-limiting step, but instead is one of several slow
steps in the catalytic cycle (see below).
NMR spectroscopy: The amidation reaction was also ana-
lyzed by NMR spectroscopy to identify possible intermedi-
ates during the transformation. First, we studied whether p-
cymene stayed coordinated to the ruthenium center
throughout the catalytic cycle. The reaction between com-
pounds 3 and 4 was performed in [D8]toluene at 1108C with
Computational study: To increase our understanding of the
reaction pathway, our investigation was extended by per-
forming a computational study, in line with earlier work.[12]
A simplified system was chosen in which ethylamine and
benzyl alcohol were used as reactants and 1,3-diisopropyli-
midazol-2-ylidene (IiPr) and PCy3 were coordinated to the
ruthenium atom. All of these calculations were performed
by using the M06 functional, which includes non-bonding in-
teractions (not the case with DFT/B3LYP). In all of these
calculations, the total energy (DGtot) was represented by a
combination of gas-phase energy (Escf), solution-phase
energy (Esolv), and Gibbs free energy (DG), as shown in
[Eq. (1)]. This approach was first suggested by Wertz[14] and
has later been applied in several studies of transition-metal-
catalyzed reactions.[12a,b,15] This procedure avoids the time-
consuming and error-prone calculation of numerical fre-
quencies in the solution phase.
15 mol% of compound
1 and with PPh3 instead of
PCy3·HBF4 to avoid the presence of additional signals in the
aliphatic region of the spectrum. Samples were removed
from the reaction mixture and analyzed at ambient tempera-
ture. We found that, after only 3 min, 85% of p-cymene was
in the solution in its unbound form and, after 10 min, p-
cymene had completely decoordinated from the ruthenium
atom.
Then, the reaction between 2-phenylethanol and benzyla-
mine was monitored in [D8]toluene with the NMR probe
temperature set at 708C. A rather high catalyst loading was
employed in this experiment with 40% of compound 1,
40% of PCy3·HBF4, and 120% of KOtBu. During the reac-
tion, several clusters of signals were detected in the hydride
region of the spectrum. After 3 h, this cluster included low-
intensity signals at d=ꢁ7.44 and ꢁ7.54 ppm, very low-inten-
sity signals in the range d=ꢁ10.66 to ꢁ11.13 ppm, high-in-
DGtot ¼ DGꢁEscfþEsolv
ð1Þ
First, we were interested in identifying the ligands that
could be coordinated to the ruthenium center during the
catalytic cycle. The precursor complex (1) is an 18-electron
ruthenium(II) species, which loses p-cymene during the ini-
tiation step. Another possible ligand is the amine moiety,
which is present in stoichiometric amounts. The DFT calcu-
lations show that the coordination of one molecule of amine
is very favorable, with a decrease in DGtot from ꢁ31 to
ꢁ107 kJmolꢁ1, depending on the other ligands on the ruthe-
nium atom. This result strongly implies that an amine is
bound to the metal center throughout the reaction. Howev-
er, the coordination of a second molecule of amine at the
apical position of the complex is less favorable than the co-
ordination of a phosphine at this position (DGtot increases
from 6 to 40 kJmolꢁ1, depending on the other ligands on
ruthenium atom).
A detailed study of the initiation of the reaction is
beyond the scope of this investigation. However, for similar
ruthenium(II)–dichloride complexes, it has been established
that, in the presence of alcohols, the chloride ligands can be
replaced with alkoxide and hydride groups.[16] Thus, because
the experimental study indicates that both chloride anions
are replaced by other ligands, we decided to use 16-electron
complex 5, in which a hydride and an alkoxide ligand are
coordinated to the ruthenium atom, as a starting point.
Our calculations show that complex 5 adopts a distorted
octahedral orientation in which the two bulky ligands (IiPr
and phosphine) are in the apical positions and the amine,
alkoxide, and hydride groups occupy the equatorial posi-
tions (Scheme 3). The alkoxide group must have an adjacent
tensity doublets from d=ꢁ17.41 to ꢁ17.89 ppm (J
ꢄ20 Hz), as well as high-intensity doublet at d=
ꢁ18.04 ppm (J(H,H)=7.1 Hz). These observations clearly
a
ACHTUNGTRENNUNG
reveal that several hydride species are formed during the
amidation reaction.[10h] Moreover, the doublet at d=
ꢁ18.04 ppm shows that there is a dihydride species that
does not contain a phosphine group. The doublets from d=
ꢁ17.41 to ꢁ17.89 ppm and their coupling constants suggest
the presence of ruthenium–hydride complexes in which one
phosphine group is coordinated cis to the hydride atom.
Over time, the intensity of the signals decreased and some
of them disappeared.
To study the reaction under conditions that were more
similar to the actual setup, the amidation reaction was re-
peated in refluxing [D8]toluene with a catalyst loading of
20%. After 30 min, a sample was withdrawn and analyzed
1
by H and 31P NMR spectroscopy at room temperature. In
1
the H NMR spectrum, several additional signals as well as
the signals given above were observed in the hydride region,
including
ꢁ15.04 ppm (J
a
singlet at d=ꢁ9.70 ppm,
a
doublet at
ACHTUNGTRENNUNG
ꢁ17.8 ppm. The presence of these signals also suggests the
formation of several complexes in which the phosphine is cis
to the hydride atom, as well as a complex without phos-
phine. The 31P NMR spectrum reveals a group of signals in
the range d=46–51 ppm, a low-intensity signal at d=
57.2 ppm, and a high-intensity signal at d=10 ppm. This
latter signal is from free PCy3, whereas the others may be
assigned to ruthenium intermediates that contain coordinat-
Chem. Eur. J. 2012, 18, 15683 – 15692
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
15687

I. S. Makarov, P. Fristrup, R. Madsen
Scheme 3. Proposed catalytic cycle (a: cis-dihydride pathway; b: trans-dihydride pathway).
empty site for b-hydride elimi-
nation to occur. As a conse-
quence, the amine and hydride
groups can be positioned in two
different ways and, thus, form
two isomers (5a and 5b).
Either of these isomers can
serve as an entry point into the
catalytic cycle.
Interestingly, these two iso-
mers, compounds 5a and 5b,
have almost the same energy
(DDGtotACHTUNGTRENNUNG
(5aꢁ5b)=5.8 kJmolꢁ1),
which means that they can both
be formed at the beginning of
the reaction. To distinguish be-
tween these two possible path-
ways for the reaction, the entire
catalytic cycles starting from
either isomer 5a or isomer 5b
were calculated. We expected
that, by direct comparison be-
Figure 6. Energy profiles for the two possible catalytic cycles (TS1a–TS5a: X=H, Y=EtNH2; TS1b–TS5b:
X=EtNH2, Y=H).
tween their experimental results and their calculated energy
profiles, it would be possible to reach a conclusion about the
orientation of the ligands. From the experiments with the
deuterated substrates, it is known that deuterium scrambling
takes place before the formation of the amide and that the
rate of exchange is much higher than the rate of the amide-
forming reaction. Most likely, the deuterium exchange
occurs in the first b-hydride-elimination step when a dihy-
dride species (6) is formed. The calculations show that the
formation of species 6a is exothermic, whereas the forma-
tion of species 6b is endothermic (Figure 6). In addition, the
activation energy for the reverse reaction is lower in the
(DEa(TS2bꢁTS1b)=48.9 kJmolꢁ1,
DEa(TS2aꢁTS1a)=
18.5 kJmolꢁ1). These facts suggest that the observed b-hy-
dride elimination is more plausible in the case of isomer 6b
than with isomer 6a. Consequently, the noted equilibrium
between deuterated and non-deuterated substrates will be
determined by the b pathway in the first b-hydride-elimina-
tion step.
To gain additional support for the b route, the two path-
ways were compared quantitatively by calculating the turn-
over frequencies (TOFs) with the energetic span model. The
concept of energetic span was introduced by Amatore and
Jutand[17] and further developed by Kozuch and Shaik.[18]
This model replaces the classical Curtin–Hammett princi-
ple of the rate-limiting step in the catalytic cycle with the
rate-limiting states, that is, the TOF-determining transition
case of species 6b (DDEaACHTUNGTRENNUNG
the difference in energy between the two transition states
(6bꢁ6a)=ꢁ30.4 kJmolꢁ1), whilst
(TS2
and
TS1)
is
lower
with
species
6a
15688
www.chemeurj.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2012, 18, 15683 – 15692

ꢁ
Ru NHC-Catalyzed Amidation of Alcohols and Amines
FULL PAPER
state (TDTS) and the TOF-determining intermediate (TDI).
The AUTOF program[19] was used to calculate the TOFs
from the computationally obtained energy states. These cal-
culations show that, for the a pathway, TS4a is the TDTS
and species 6a is the TDI, whereas, for the b pathway, TS2b
is the TDTS and species 5b is the TDI. In other words, the
rate of the reaction is defined by the difference in energy
(energetic span) between species 6a and TS4a (DE=
189.1 kJmolꢁ1) for the a pathway and between species 5b
and TS2b for the b pathway (DE=119.0 kJmolꢁ1). As a
result of the large differences in energetic span between the
two pathways, the calculated TOF for the catalytic cycle
that proceeds through the a pathway is 1.04ꢁ10ꢁ8 hꢁ1,
whereas the TOF for the catalytic cycle that proceeds
through the b pathway is a factor of ꢁ106 higher (7.38ꢁ
10ꢁ1 hꢁ1). This large difference between the calculated
TOFs, along with the data on deuterium scrambling, clearly
indicates that the orientation of the ligands in the b pathway
is consistent with the experimental results, whereas that in
the a pathway is not. Moreover, the calculated TOF is close
to the average experimental value of 8.00ꢁ10ꢁ1 hꢁ1, which
lends further support to the conclusion that the b cycle is
the dominant product-forming pathway in this reaction. The
trans orientation of the hydrides in this route makes the spe-
cies less stable and, consequently, more reactive.
an empty site and undergo b-hydride elimination. Notably,
when molecular hydrogen has dissociated from species 12b
to form species 5b, the alkoxide ligand changes its coordina-
tion from h1 to h3 by engaging in an agostic interaction be-
ꢁ
tween Ru and the a-C H bond of the alkoxide. This interac-
tion has also previously been observed in computational
studies of the b-hydride elimination with alkoxides.[20] As
ꢁ
shown in Figure 7, the Ru H bond length changes from
For the b pathway, the ability to dissociate a molecule of
phosphine was examined to explain the spectroscopic obser-
vations (see above). Calculations were performed for all of
the intermediates (5b–12b) and revealed that one inter-
mediate (6b) was more stable without a coordinated phos-
phine (Table 3).
Table 3. Energy of PCy3 dissociation from intermediates 5b–12b.
Compound
5b
6b
7b
8b
9b
10b 11b 12b
DGdissoc [kJmolꢁ1
]
41.2 ꢁ6.4 29.4 61.1 63.7 16.0 43.8 15.7
However, the barrier for the addition of an amine to this
species (without a coordinated phosphine) is 21.7 kJmolꢁ1
higher than that to species 6b, which, in total, makes the
pathway without phosphine 15.3 kJmolꢁ1 less favorable. Be-
cause species 6b is located between two rate-limiting states,
the concentration of this intermediate will have a strong in-
fluence on the overall rate of the reaction. This result corre-
lates with the experimental observations, where the addition
of phosphine shifts the equilibrium towards species 6b and
increases the overall rate, but the order of the reaction with
respect to phosphine is less than one because some phos-
phine dissociates off. In addition, the doublet in the
1H NMR spectrum at d=ꢁ18.04 ppm can be assigned to the
dihydride species that is formed by PCy3 dissociation from
species 6b.
Figure 7. Calculated structures of compounds 12b (top) and 5b (bottom).
3.317 ꢂ (12b) to 2.137 ꢂ (5b). A similar agostic interaction
is observed in the transformation from species 8b into spe-
cies 9b. This result indicates that, during the catalytic cycle,
ruthenium is electron poor and can be stabilized by receiv-
ing electron density from its ligands. This conclusion is also
supported by experimental observations because amides are
formed faster and in higher yields when electron-rich phos-
phines and NHCs, as well as benzyl alcohols, with electron-
donating substituents in the para position are used (see
above).
Having determined the orientation of the ligands, the
next step was to establish the geometrical details of the spe-
cies that are involved in the catalytic cycle. All of the com-
pounds in the catalytic cycle are 18-electron complexes,
except for two intermediates, 5b and 9b, which both have
After the first b-hydride elimination, complex 6b is
formed. In this intermediate, the aldehyde acts as a h2
Chem. Eur. J. 2012, 18, 15683 – 15692
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
15689

I. S. Makarov, P. Fristrup, R. Madsen
ꢁ
ligand by binding to the ruthenium center through the p sy-
stem of the C=O bond (Scheme 4). However, the geometry
of the carbonyl carbon atom is close to that of an sp3 hybri-
out-of-plane angle for the C H bond: 1698, dC O =1.235 ꢂ),
which is then attacked by the amine to form species 7b
(Scheme 4).
=
The energetic parameters for the transformations 5b!6b
and 6b!7b determine the overall rate of the reaction.
These
steps
have
approximately
equal
energy
(Ea(5bꢁTS1b)=68.8 kJmolꢁ1, Ea(6bꢁTS2b) = 68.7 kJmolꢁ1)
and, consequently, both steps should contribute equally to
the limitations of the reaction. In the Hammett study, these
two steps have opposite influence on the rate (and, hence,
the 1 value) of the reaction. In the first step, electron-donat-
ing groups facilitate the b-hydride elimination, whereas, in
the second step, they have the opposing effect in the nucleo-
philic addition to the aldehyde.
To model the Hammett study, the energy difference be-
tween the TDI and the TDTS was calculated for several
benzyl alcohols with the following para substituents: NMe2,
OMe, SMe, Me, F, Cl, CF3. Their relative reactivities were
calculated by using Equation (2).
Scheme 4. Equilibrium between the isomers of compound 6b.
ꢁ
dized carbon atom (out-of-plane angle for the C H bond:
1368, dC O =1.309 ꢂ), which implies that complex 6b is
=
more-correctly represented by a three-membered oxaruthi-
nacycle (Figure 8). In the subsequent addition of the amine,
complex 6b rearranges into the aldehyde h1 isomer (6b’;
kX dEH ꢁ dE
X ; dE ¼ EðTDTSÞ ꢁ EðTDIÞ
ð2Þ
lg
¼
2:303RT
kH
Different calculated energies were used in [Eq. (2)], but
only the gas-phase energies and solution-phase energies
showed good correlation with the s+ values (Figure 9).
Figure 9. Hammett plot with calculated gas-phase and solvation energies
(red points were excluded from the linear regression).
As can be seen from Figure 9, two substituents did not
follow the overall good correlation; H and Me are both
non-polar substituents and we ascribe the difference be-
tween these and the remaining polar substituents to inaccur-
acies in the solvation model. When examining the actual 1
values that were calculated by using the gas-phase energies
(1gasphase =ꢁ0.68) and the solution-phase energies (1solv
=
ꢁ0.73), it is clear that the calculated slopes are somewhat
higher than the experimental value (1exp =ꢁ0.15). However,
the calculated reactivity follows the same trend as the exper-
Figure 8. Calculated structures of compounds 6b (top) and 6b’ (bottom).
15690
www.chemeurj.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2012, 18, 15683 – 15692

ꢁ
Ru NHC-Catalyzed Amidation of Alcohols and Amines
FULL PAPER
imental data, that is, that substrates with electron-donating
substituents are more reactive than those with electron-with-
drawing substituents.
but when the theoretical turnover frequencies (TOFs) were
derived directly from the calculated DFT/M06 energies, we
found that only the trans-dihydride pathway (Scheme 3,
cycle b) was in agreement with the experimentally deter-
mined TOFs. The overall good agreement between the ex-
perimental and theoretical data illustrates that modern theo-
retical methods have matured to a point where their results
can be used to predict and rationalize experimental observa-
tions. This result opens up the possibility for a number of
new applications, such as in silico ligand screening, which is
currently underway in our laboratory.
The KIE for the reaction was also calculated by using the
energetic-span model. In that case, the catalytic cycle with
deuterated analogues had a dE value of 142.3 kJmolꢁ1,
which resulted in a TOF of 1.4·10ꢁ2 hꢁ1. Thus, the KIE value
was calculated as KIE=TOFH/TOFD =3.78, which was sig-
nificantly higher than that observed experimentally (KIE=
2.29). There may be several reasons for this discrepancy:
First, benzyl alcohol was used in the computational model
system whereas [D10]-1-butanol was utilized in the experi-
ment. This result indicates that the nature of the substituent
on the a carbon atom of the alcohol does have a profound
influence on the observed KIE value in the amidation reac-
tion. Another way to calculate the KIE value is by using
Equation (3).
Computational Details
All calculations were performed with Jaguar[21] (version 7.8, release 109)
by using the M06[22] or B3LYP functionals[23] in combination with the
LACVP* basis set.[24] During our initial investigations, it became clear
that the B3LYP functional did not adequately describe the non-bonded
interactions that were responsible for discriminating between the possible
reaction pathways; this deficiency of the B3LYP functional is well-known
and the problem has been addressed previously by either appending a
classic dispersion term[25] or by using a functional that incorporates terms
for kinetic energy density.[26] Among the most-successful of these latter
approaches are the M0x family of functionals reported by Truhlar and
co-workers; herein, we chose the M06 functional, which has been opti-
mized with a particular focus on organometallic systems and has been
used successfully in an earlier study.[27]
DGHa ꢁDGDa
kH
kD
RT
ð3Þ
KIE ¼
¼ eꢁ
; DGa ¼ DGtotðTS2bÞ ꢁ DGtotð6bÞ
The KIE value for the system with EtOH was calculated
to be 3.27, which is very close to the value obtained by
using the energetic span model. This result shows that either
of these two models may be used for the KIE calculations.
However, the KIE value from these calculations is still
higher than the experimental value (2.29); the reason for
this difference may partially lie in the experimental determi-
nation of the KIE. Commercially available [D10]-1-butanol
was used for the KIE experiments and it contained a small
amount of C4D9OH. The amidation reaction was performed
with 5% of PCy3·HBF4, which, after deprotonation with
KOtBu, would exchange hydrogen and deuterium with
[D10]-1-butanol. Presumably, 5–10% of C4D9OH was present
in the experiment with [D10]-1-butanol and this isotopic im-
purity would lower the value of the experimentally deter-
mined KIE.
All of the structures were optimized in the gas phase and the single-point
solvation energy was calculated for the optimized structures by using a
standard Poisson–Boltzmann solver with suitable parameters for benzene
as the solvent (dielectric constant: e=2.284, probe radius: r=2.600 ꢂ).
Gibbs free energies were obtained from the vibrational-frequency calcu-
lations for the gas-phase geometries at 298 K and 383 K. All of the transi-
tion states were characterized by the presence of one negative vibrational
frequency.
Acknowledgements
We thank the Danish Council for Independent Research–Technology and
Production Sciences (DFF) for financial support.
Conclusion
[1] a) V. R. Pattabiraman, J. W. Bode, Nature 2011, 480, 471–479;
b) C. L. Allen, J. M. J. Williams, Chem. Soc. Rev. 2011, 40, 3405–
3415; c) C. Chen, S. H. Hong, Org. Biomol. Chem. 2011, 9, 20–26.
[2] B. Shen, D. M. Makley, J. N. Johnston, Nature 2010, 465, 1027–1032.
[3] J. W. Bode, R. M. Fox, K. D. Baucom, Angew. Chem. 2006, 118,
1270–1274; Angew. Chem. Int. Ed. 2006, 45, 1248–1252.
[4] a) J.-F. Soulꢃ, H. Miyamura, S. Kobayashi, J. Am. Chem. Soc. 2011,
133, 18550–18553; b) Y. Wang, D. Zhu, L. Tang, S. Wang, Z. Wang,
Angew. Chem. 2011, 123, 9079–9083; Angew. Chem. Int. Ed. 2011,
50, 8917–8921.
[5] a) C. Gunanathan, Y. Ben-David, D. Milstein, Science 2007, 317,
790–792; b) L. U. Nordstrøm, H. Vogt, R. Madsen, J. Am. Chem.
Soc. 2008, 130, 17672–17673; c) A. Prades, E. Peris, M. Albrecht,
Organometallics 2011, 30, 1162–1167; d) N. D. Schley, G. E. Dober-
einer, R. H. Crabtree, Organometallics 2011, 30, 4174–4179; e) K.
Shimizu, K. Ohshima, A. Satsuma, Chem. Eur. J. 2009, 15, 9977–
9980; f) J. Zhu, Y. Zhang, F. Shi, Y. Deng, Tetrahedron Lett. 2012,
53, 3178–3180.
The mechanism of the Ru–NHC-catalyzed amidation reac-
tion between alcohols and amines was investigated by using
experimental and theoretical methods. A Hammett study in-
dicates that a small positive charge is built-up at the benzyl-
ic position in the transition state of the turnover-limiting
step. The small value indicates that the rate of the reaction
is not dominated by a single elementary reaction; instead, it
is likely that two steps with opposite electronic character
both influence the reaction rate.
The kinetic isotope effect was experimentally determined
ꢁ
to be 2.29
N
bond is not the rate-limiting step, but that it is one of several
slow steps in the catalytic cycle. These experimental results
were further supported by the characterization of a plausible
catalytic cycle by using DFT/M06 calculations. Both cis-di-
hydride and trans-dihydride intermediates were considered,
[6] a) T. Zweifel, J.-V. Naubron, H. Grꢄtzmacher, Angew. Chem. 2009,
121, 567–571; Angew. Chem. Int. Ed. 2009, 48, 559–563; b) A. J. A.
Chem. Eur. J. 2012, 18, 15683 – 15692
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
15691

I. S. Makarov, P. Fristrup, R. Madsen
Watson, A. C. Maxwell, J. M. J. Williams, Org. Lett. 2009, 11, 2667–
2670.
Creary, M. E. Mehrsheikh-Mohammadi, S. McDonald, J. Org.
Chem. 1987, 52, 3254–3263.
[7] a) J. P. Shackleford, B. Shen, J. N. Johnston, Proc. Natl. Acad. Sci.
USA 2012, 109, 44–46; b) I. Pusterla, J. W. Bode, Angew. Chem.
2012, 124, 528–531; Angew. Chem. Int. Ed. 2012, 51, 513–516.
[8] M. Montag, J. Zhang, D. Milstein, J. Am. Chem. Soc. 2012, 134,
10325–10328.
[14] D. H. Wertz, J. Am. Chem. Soc. 1980, 102, 5316–5322.
[15] J. K. C. Lau, D. V. Deubel, J. Chem. Theory Comput. 2006, 2, 103–
106.
[16] a) E. Solari, S. Gauthier, R. Scopelliti, K. Severin, Organometallics
2009, 28, 4519–4526; b) A. Aranyos, G. Csjernyik, K. J. Szabꢅ, J.-E.
Bꢆckvall, Chem. Commun. 1999, 351–352.
[9] a) G. Zeng, S. Li, Inorg. Chem. 2011, 50, 10572–10580; b) H. Li, X.
Wang, F. Huang, G. Lu, J. Jiang, Z. Wang, Organometallics 2011, 30,
5233–5247; c) A. Nova, D. Balcells, N. D. Schley, G. E. Dobereiner,
R. H. Crabtree, O. Eisenstein, Organometallics 2010, 29, 6548–6558.
[10] a) A. Maggi, R. Madsen, Organometallics 2012, 31, 451–455; b) C.
Chen, Y. Zhang, S. H. Hong, J. Org. Chem. 2011, 76, 10005–10010;
c) A. Sølvhøj, R. Madsen, Organometallics 2011, 30, 6044–6048;
d) J. Zhang, M. Senthilkumar, S. C. Ghosh, S. H. Hong, Angew.
Chem. 2010, 122, 6535–6539; Angew. Chem. Int. Ed. 2010, 49, 6391–
6395; e) J. H. Dam, G. Osztrovszky, L. U. Nordstrøm, R. Madsen,
Chem. Eur. J. 2010, 16, 6820–6827; f) S. C. Ghosh, S. H. Hong, Eur.
J. Org. Chem. 2010, 4266–4270; g) S. Muthaiah, S. C. Ghosh, J.-E.
Jee, C. Chen, J. Zhang, S. H. Hong, J. Org. Chem. 2010, 75, 3002–
3006; h) Y. Zhang, C. Chen, S. C. Ghosh, Y. Li, S. H. Hong, Organo-
metallics 2010, 29, 1374–1378; j) S. C. Ghosh, S. Muthaiah, Y.
Zhang, X. Xu, S. H. Hong, Adv. Synth. Catal. 2009, 351, 2643–2649.
[11] M. R. Netherton, G. C. Fu, Org. Lett. 2001, 3, 4295–4298.
[12] a) P. Fristrup, M. Tursky, R. Madsen, Org. Biomol. Chem. 2012, 10,
2569–2577; b) P. Fristrup, M. Kreis, A. Palmelund, P.-O. Norrby, R.
Madsen, J. Am. Chem. Soc. 2008, 130, 5206–5215; c) J. H. Dam, P.
Fristrup, R. Madsen, J. Org. Chem. 2008, 73, 3228–3235; d) P. Fristr-
up, L. B. Johansen, C. H. Christensen, Chem. Commun. 2008, 2750–
2752.
[17] C. Amatore, A. Jutand, J. Organomet. Chem. 1999, 576, 254–278.
[18] S. Kozuch, S. Shaik, Acc. Chem. Res. 2011, 44, 101–110.
[19] a) A. Uhe, S. Kozuch, S. Shaik, J. Comput. Chem. 2011, 32, 978–
985; b) S. Kozuch, S. Shaik, J. Phys. Chem. A 2008, 112, 6032–6041;
c) S. Kozuch, S. Shaik, J. Am. Chem. Soc. 2006, 128, 3355–3365.
[20] N. Sieffert, M. Bꢄhl, J. Am. Chem. Soc. 2010, 132, 8056–8070.
[21] Jaguar, Version 7.8, Schrodinger, LLC, New York, NY, 2010.
[22] Y. Zhao, D. G. Truhlar, Theor. Chem. Acc. 2008, 120, 215–241.
[23] a) A. D. Becke, J. Chem. Phys. 1993, 98, 5648–5652; b) C. Lee, W.
Yang, R. G. Parr, Phys. Rev. B 1988, 37, 785–789.
[24] P. J. Hay, W. R. Wadt, J. Chem. Phys. 1985, 82, 270–283.
[25] a) S. Grimme, J. Comput. Chem. 2004, 25, 1463–1473; b) M. Elstner,
P. Hobza, T. Frauenheim, S. Suhai, E. Kaxiras, J. Chem. Phys. 2001,
114, 5149–5155.
[26] a) B. B. Averkiev, Y. Zhao, D. G. Truhlar, J. Mol. Catal. A 2010, 324,
80–88; b) Y. Zhao, D. G. Truhlar, Acc. Chem. Res. 2008, 41, 157–
167.
[27] X. Bantreil, G. Prestat, A. Moreno, D. Madec, P. Fristrup, P.-O.
Norrby, P. S. Pregosin, G. Poli, Chem. Eur. J. 2011, 17, 2885–2896.
[13] a) C. Hansch, H. Gao, Chem. Rev. 1997, 97, 2995–3059; b) C.
Hansch, A. Leo, R. W. Taft, Chem. Rev. 1991, 91, 165–195; c) X.
Received: July 5, 2012
Published online: October 15, 2012
15692
www.chemeurj.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2012, 18, 15683 – 15692