Palladium(0)-catalyzed trimerization of arylisocyanates into 1,3,5-triarylisocyanurates in the presence of diimines: A nonintuitive mechanism

Article abstract of DOI:10.1021/ja068291k

We show here that palladium(0) (dibenzylideneacetone) complexes bearing 1,10-phenanthroline constitute efficient catalysts for the cyclotrimerization of aromatic isocyanates. For the first time, the mechanism of this reaction has been investigated experimentally and theoretically with group 10 catalysts. This investigation provides a very consistent picture of the catalytic cycle. Notably, we establish that the reaction does not proceed by stepwise cycloadditions or ring insertions involving metallacyclic intermediates, as might have been anticipated. Rather, in our proposal, the initial steps of the mechanism resemble the chain-growth process operative during the anionic polymerization of isocyanates and feature charge-separated intermediates. These steps are then followed by ring closure on the metal center of the last intermediate formed to yield a seven-membered metallacycle that reductively eliminates the cyclotrimer and re-forms the active species. In addition, we conclusively show that the (known) palladacycles that could be isolated during the experimental investigations are not catalytic intermediates but result from catalyst deactivation. Thus, with Pd(0) diimine catalysts, the actual trimerization mechanism appears to be a blend between the two types of mechanisms proposed thus far for the oligomerization of heterocumulenes with very different catalysts. In conclusion, this work contributes to a better understanding of the reactivity of arylisocyanates in the vicinity of late group 10 metal centers in low oxidation state and sheds some light on the detrimental self-poisoning processes observed during the reductive carbonylation of nitroaromatic substrates catalyzed by related catalysts in non-nucleophilic media.

Full text of DOI:10.1021/ja068291k

Published on Web 05/18/2007
Palladium(0)-Catalyzed Trimerization of Arylisocyanates into
1,3,5-Triarylisocyanurates in the Presence of Diimines: A
Nonintuitive Mechanism
Fre´de´ric Paul,*,† Solenne Moulin,† Olivier Piechaczyk,‡ Pascal Le Floch,*,‡ and
John A. Osborn§,
Contribution from the Laboratoire “Sciences Chimiques de Rennes”, UniVersite´ de Rennes 1,
Campus de Beaulieu, CNRS, 35042 Rennes Cedex, France, Laboratoire He´te´roe´le´ments et
Coordination, Ecole Polytechnique, CNRS, 91128 Palaiseau Cedex, France, and Laboratoire de
Chimie des Me´taux de Transition et de Catalyse, UniVersite´ de Strasbourg, Institut Le Bel,
CNRS, 4 rue Blaise Pascal, F-67070 Strasbourg Cedex, France
Received November 23, 2006; E-mail: frederic.paul@univ-rennes1.fr; lefloch@poly.polytechnique.fr
Abstract: We show here that palladium(0) (dibenzylideneacetone) complexes bearing 1,10-phenanthroline
constitute efficient catalysts for the cyclotrimerization of aromatic isocyanates. For the first time, the
mechanism of this reaction has been investigated experimentally and theoretically with group 10 catalysts.
This investigation provides a very consistent picture of the catalytic cycle. Notably, we establish that the
reaction does not proceed by stepwise cycloadditions or ring insertions involving metallacyclic intermediates,
as might have been anticipated. Rather, in our proposal, the initial steps of the mechanism resemble the
chain-growth process operative during the anionic polymerization of isocyanates and feature charge-
separated intermediates. These steps are then followed by ring closure on the metal center of the last
intermediate formed to yield a seven-membered metallacycle that reductively eliminates the cyclotrimer
and re-forms the active species. In addition, we conclusively show that the (known) palladacycles that
could be isolated during the experimental investigations are not catalytic intermediates but result from catalyst
deactivation. Thus, with Pd(0) diimine catalysts, the actual trimerization mechanism appears to be a blend
between the two types of mechanisms proposed thus far for the oligomerization of heterocumulenes with
very different catalysts. In conclusion, this work contributes to a better understanding of the reactivity of
arylisocyanates in the vicinity of late group 10 metal centers in low oxidation state and sheds some light
on the detrimental self-poisoning processes observed during the reductive carbonylation of nitroaromatic
substrates catalyzed by related catalysts in non-nucleophilic media.
have focused on cyclotrimerization reactions of isococyanates,7,8
especially with group 10 metal catalysts.9 These reactions
Introduction
The reactivity of isocyanates in the coordination sphere of
various group 10 metal centers was a field only partially
explored by the early 1990s,1 but some renewed interest for
their reactivity has been shown more recently.2 From an atom-
economy perspective, such reactions are quite attractive since
they can provide a very valuable access to various types of
heterocycles,3 as amply demonstrated not only by the seminal
work of Hoberg with nickel complexes,4,5 but also by other
researchers later on.6 Among these investigations, much fewer
provide, however, an easy access to 1,3,7-isocyanurates (3a-
e; Chart 1), which constitute important activators and additives
for processing numerous urethane-based polymers and
coatings.7b,8,10 We have shown previously that such isocyanate
oligomerization reactions are likely at the origin of self-
(4) For catalytic reactions, see, for instance: (a) Hoberg, H. J. Organomet.
Chem. 1988, 358, 507-517. (b) Hoberg, H.; Guhl, D. J. Organomet. Chem.
1989, 375, 245-257. (c) Hoberg, H.; Ba¨rhausen, D.; Mynott, R.; Schroth,
G. J. Organomet. Chem. 1991, 410, 117-126. (d) Hoberg, H.; Nohlen, M.
J. Organomet. Chem. 1991, 412, 225-236.
(5) For stoichiometric reactions, see, for instance: Hoberg, H.; Guhl, D. J.
Organomet. Chem. 1989, 378, 279-292.
(6) See, for instance: (a) Hsieh, J.-C.; Cheng, C.-H. Chem. Commun. 2005,
4554-4556. (b) Zhou, H.-B.; Alper, H. J. Org. Chem. 2003, 68, 3439-
3445.
† Universite´ de Rennes 1.
‡ Ecole Polytechnique.
§ Universite´ de Strasbourg.
Deceased.
(1) (a) Braunstein, P.; Nobel, D. Chem. ReV. 1989, 89, 1427-1445. (b) Cenini,
S.; La Monica, G. Inorg. Chim. Acta 1976, 18, 279-293.
(2) (a) Guiducci, A. E.; Boyd, C. L.; Mountford, P. Organometallics 2006,
25, 1167-1187. (b) Mindiola, D. J.; Hillhouse, G. L. J. Am. Chem. Soc.
2001, 123, 4623-4624. (c) Dunn, S. C.; Hazari, N.; Cowley, A. R.; Green,
J. C.; Mountford, P. Organometallics 2006, 25, 1755-1770. (d) Wang,
H.; Chan, H.-S.; Okuda, J.; Xie, Z. Organometallics 2005, 24, 3118-3124.
(e) Owen, G. R.; Vilar, R.; White, A. J. P.; Williams, D. J. Organometallics
2003, 22, 4511-4521.
(7) (a) Tang, J.-S.; Verkade, J. G. Angew. Chem., Int. Ed. Engl. 1993, 32,
896-898. (b) Tang, J.-S.; Verkade, J. G. J. Org. Chem. 1994, 59, 4931-
4938.
(8) (a) Foley, S. R.; Yap, G. P.; Richeson, D. S. Organometallics 1999, 18,
4700-4705. (b) Foley, S. R.; Zhou, Y.; Yap, G. P. A.; Richeson, D. S.
Inorg. Chem. 2000, 39, 924-929 and references therein.
(9) Zhitinkina, A. K.; Shibanova, N. A.; Tarakhanov, O. G. Russ. Chem. ReV.
1985, 54, 1104-1125.
(10) Puzstai, Z.; Vlad, G.; Bodor, A.; Horvath, I. T.; Laas, H. J.; Halpaap, R.;
Richter, F. U. Angew. Chem., Int. Ed. 2006, 45, 107-110.
(3) Noak, R.; Schwetlick, K. A. Z. Chem. 1987, 27, 77-89.
9
7294
J. AM. CHEM. SOC. 2007, 129, 7294-7304
10.1021/ja068291k CCC: $37.00 © 2007 American Chemical Society

Pd(0)-Catalyzed Trimerization of Arylisocyanates
A R T I C L E S
Chart 1. Structures of Selected Compounds
Scheme 2. Reaction of Phenylisocyanate (5a) with Complex 4a
common organic solvents. Indeed, their precipitation in the
reaction medium drives the various equilibria toward their
quantitative formation. In accordance with the equilibrium
situation, we have verified that the starting Pd(dba)3 complex
could be regenerated in solution when 4a was exposed to a 10-
fold excess of dba.
The pale orange air-sensitive Pd(0) complexes 4a-d were
obtained in good yields and were characterized by FABMS
spectrometry. The low solubility of the complexes associated
with the lability of the dba ligand which slowly generates
palladium black in solution complicates any subsequent puri-
fication. As a result, only poor elemental analyses were obtained
for these compounds. The strong fluxionality of the dba ligand
Scheme 1. Synthesis of Complexes 4a-d
1
poisoning during the catalytic carbonylation of various nitroaro-
matic substrates when these transformations are conducted in
non-nucleophilic/non-protic solvents.11 Given that such carbo-
nylation reactions have aroused an enormous industrial interest
as potential means to replace the traditional phosgene-based
processes in the production of isocyanates, a better understand-
ing of the detrimental side processes was highly desirable,
especially with (Pd/o-phen)-based catalytical systems (o-phen
) 1,10-phenanthroline), which constitute the most active
catalytic systems identified to date for this reaction.12,13 With
such catalysts, self-poisoning was often accompanied by the
formation of palladacycles such as 1 or 2 (Chart 1).11
In the hope of getting additional clues on the way these
palladacycles can form, we have studied the reaction of PhNCO
with various Pd(0)-diimine precursors. At first, the reactivity
of various (N-N)Pd(dba) complexes (4a-d) isolated from the
Pd(dba)3 precursor and the corresponding diimine was examined
(Chart 1). Several among these complexes proved to constitute
quite efficient catalysts for the trimerization reaction. The scope
of this reaction was subsequently investigated experimentally
and by density functional theory (DFT). Calculations carried
out on this catalytic transformation allowed for the determination
of a precise mechanism in full accordance with the experimental
results.
is revealed by H NMR. This fluxionality renders the diimine
ligand symmetric at the NMR time scale and impedes observa-
tion of olefinic dba protons coordinated to the Pd(0) center at
ambient temperature, but a one-to-one ratio between dba and
phenanthroline ligands can be inferred from the 1H NMR signals
of complex 4a at low temperatures (see Supporting Information).
Infrared spectroscopy also clearly indicates that the dba ligand
is monocoordinated in a π-fashion in the solid state, as suggested
by the slight νCO shift to lower wavenumbers observed relative
to free dba (ca. 15 cm-1).14
Reaction of 4a-d with Aromatic Isocyanates. In line with
the fluxionality observed by NMR, several studies have shown
that the dba ligand was quite labile in 4a-d and that these Pd-
(0) complexes could act as convenient sources of the (N-N)-
Pd0 fragment.14-16 However, in our initial experiments, the
inherent low solubility of 4a-d severely restricted the available
concentration of (N-N)Pd0 precursor in solution and proved
detrimental for obtaining clean stoichiometric reactions with
aromatic isocyanates, due to the competitive decomposition of
4a-d into palladium metal. To overcome this problem, we
decided to use the isocyanate reactant in large excess (i.e.,
directly as solvent) to speed up the reaction kinetics.
Thus, when 4a is suspended in neat PhNCO (5a), a slightly
exothermic process takes place and large amounts of off-white
product precipitate in the medium, rendering it viscous and
completely heterogeneous after a few hours (Scheme 2).
Analysis of the precipitate indicated that the latter was mostly
composed of 1,3,5-triphenylisocyanurate (3a), the cyclic trimer
of isocyanate, which evidently results from the catalytic
cyclotrimerization of phenylisocyanate 5a (40-45%) after 30
h (Table 1, entry 1). Traces of 1,3-diphenylene uretedione (6a),
the cyclic dimer, were also characterized (0.6%), but no
additional product could be identified apart from unreacted
PhNCO (5a). In this reaction, the isolated yield of trimer 3a
corresponds to a catalytic turnover of ca. 120 cycles. Control
experiments revealed that, under otherwise similar conditions,
neither free dba, free o-phen, Pd adsorbed on charcoal (Pd/C),
Results
Synthesis of (N-N)Pd(dba) Complexes. The desired Pd(0)
precursors 4a-d were isolated from the known Pd(dba)3‚C6H6
complex and various diimine ligands, following a procedure
used by Ishii et al. in 1974 for the isolation of (bipy)Pd(dba)
and more recently by Stahl and co-workers for the isolation of
(bc)Pd(dba) (bc ) bathocuproin) (Scheme 1).14,15 Notably, the
successful isolation of 4a-d rests on their poor solubility in
(11) (a) Paul, F.; Fischer, J.; Ochsenbein, P.; Osborn, J. A. C. R. Acad. Sci.
2002, 5, 267-287. (b) Paul, F.; Fischer, J.; Ochsenbein, P.; Osborn, J. A.
Angew. Chem., Int. Ed. Engl. 1993, 32, 1638-1640.
(12) Paul, F. Coord. Chem. ReV. 2000, 203, 269-323.
(13) Cenini, S.; Ragaini, F. ReductiVe Carbonylation of Organic Nitro Com-
pounds; Kluwer Academic Publishers: Dordrecht, The Netherlands, 1997.
(14) Ukai, T.; Kawazura, H.; Ishii, Y.; Bonnet, J. J.; Ibers, J. A. J. Organomet.
Chem. 1974, 65, 253-266.
(16) (a) Amatore, C.; Broeker, G.; Jutand, A.; Khalil, F. J. Am. Chem. Soc.
1997, 119, 5176-5185. (b) Amatore, C.; Jutand, A. Coord. Chem. ReV.
1998, 178-180, 511-528. (c) Takahashi, Y.; Ito, T.; Sakai, S.; Ishii, Y. J.
Chem. Soc., Chem. Commun. 1970, 1065-1066.
(15) Stahl, S. S.; Thorman, J. L.; Nelson, R. C.; Kozec, M. A. J. Am. Chem.
Soc. 2001, 123, 7188-7189.
9
J. AM. CHEM. SOC. VOL. 129, NO. 23, 2007 7295

A R T I C L E S
Paul et al.
Table 1. Catalytic Trimerization of Phenylisocyanate by 4a
cosolvent
isolated yield
a
entry
[4a]0
[PhNCO]/[4a]
(ratio vs PhNCO)
t (h)
T (°
C)
of 3a
1
2
3
4
5
6
7
8
9
1.0 × 10-2
2.4 × 10-2
1.2 × 10-2
1.2 × 10-2
1.2 × 10-2
1.2 × 10-2
1.2 × 10-2
1.2 × 10-2
1.2 × 10-2
2.2 × 10-2
2.2 × 10-3
2.4 × 10-2
0.8 × 10-2
0.8 × 10-2
0.8 × 10-2
0.8 × 10-2
0.8 × 10-2
0.8 × 10-2
800
147
350
350
350
350
350
350
350
40
/
30.0
30.0
120.0
120.0
120.0
120.0
14.0
0.5
0.2
7.0
1.0
3.0
16.0
0.3
0.3
0.3
0.3
25
25
25
25
25
25
25
25
140
140
80
0
-10
25
45
15
46
3
CH2Cl2 (2:1)
CH2Cl2 (1:1)
C6H6 (1:1)
(CH3)2CO (1:1)
PhNO2 (1:1)
PhNO2 (1:1)
PhNO2 (1:1)
PhNO2 (1:1)
PhNO2 (50:1)
C6H6 (50:1)
CH2Cl2 (2:1)
PhNO2 (2:1)
PhNO2 (2:1)
PhNO2 (2:1)
PhNO2 (2:1)
PhNO2 (2:1)
PhNO2 (2:1)
65
80
99
80
10b
0d
10c
11c
12
13
14
15g
16h
17i
18j
20
0e
147
350
350
350
350
350
350
10f
67
25
22
16
20
7
25
25
25
25
0.3
a Initial concentration of 4a in moles per liter. b Estimated after leaving the reaction medium 30 additional hours at 25 °C. c Five equivalents of o-phen
also present in the medium. d Isolation of palladacycle 2 (95%: see Supporting Information). e Isolation of palladacycle 2 (28%) along with palladium black.
f Estimated after leaving the reaction medium at 25 °C. g In the presence of a 10-fold excess of o-phen. h In the presence of a 5-fold excess of dba. i In the
presence of a 5-fold excess of 1,9-dihydroanthracene. j In the presence of a 5-fold excess of duroquinone.
nor Pd(dba)3 alone produced any visible reaction. Thus, the
catalytic transformation must be imparted to 4a. Accordingly,
with such a hypothesis, when Pd(dba)3 and excess o-phen are
simultaneously introduced in the medium, or when another
potential source of “(o-phen)Pd0” such as (o-phen)Pd(PhNO)n
is introduced in the medium,17 the exothermic formation of
isocyanurate is again observed. Thus, the “(o-phen)Pd0” frag-
ment appears as a constitutive structural unit of the active
species.
other hand, cooling of the reaction medium also results in a
decrease of the catalytic activity, as qualitatively evidenced by
the absence of visible precipitation after 3 h at 0 °C (entries 12
and 13). Clearly, the trimerization reaction appears to be favored
at ambient temperature with 4a.
Finally, the effect of various additives on the isolated yield
of trimer 3a was investigated under conditions for which the
reaction is far from completion (Table 1, entries 14-18). First,
the presence of excess of the catalysts ligands such as o-phen
or dba was checked (entries 15 and 16). Thus, while the yield
seems quite unaffected by excess o-phen, a 5-fold excess of
dba seems to slightly hinder the reaction (compare entries 15
and 16 with entry 14). We also verified that additives such as
1,9-dihydroanthraquinone or duroquinone (Table 1, entries 17
and 18), well known to trap radicals, do not poison the reaction.
This suggests that the mechanism does not proceeds via a radical
pathway.
The reaction observed in neat PhNCO is often incomplete, a
fact that can be attributed to the heterogeneous nature of the
reaction medium, rendering stirring inefficient after some time.
To determine if the conversion of isocyanate (5a) in 1,3,5-
triphenylisocyanurate (3a) could be improved, the use of a
cosolvent was tested. With dichloromethane, no better conver-
sions were obtained (Table 1, entries 2 and 3). Typically, using
a 1:1 (volumic) mixture of PhNCO (5a) with dichloromethane
delayed precipitation of the trimer (3a), but no higher yields
were obtained even after 120 h (Table 1, entry 3). Further
dilution of the phenylisocyanate substrate 5a led to even lower
conversions into 3a, even when higher catalyst loadings are used
(Table 1, entry 2). The medium apparently stays homogeneous,
but the formation of metallic palladium is visible after several
hours. We then realized that the nature of the cosolvent used
could exert a strong influence on the catalyst turnover, with
more polar cosolvents favoring higher conversions (Table 1,
compare entries 3-6). Thus, when nitrobenzene was used as
cosolvent, we observed a quite quantitative conversion of the
starting isocyanate into isocyanurate after 14 h (Table 1, entry
7). Actually, monitoring of the reaction progress over time
reveals that the conversion of the phenylisocyanate (5a) is quite
complete already after 1 h in this solvent (Table 1, entry 8).
The temperature also appears to strongly influence the
catalytic transformation (Table 1, entries 9-13). As can be seen
from entries 9-11 in Table 1, heating of the reaction medium
has a detrimental effect, since the isolated yield of 3a drops to
10% after ca. 10 min of heating at 140 °C (entry 9). On the
The fate of the palladium complex 4a was examined next.
The low yield in 1,3,5-triphenylisocyanurate (3a) in run 2 (Table
1) when dichloromethane was used as cosolvent allowed for
selective precipitation of metal-containing species. Thus, the
known palladacycle 1 (ca. 30-50% of starting 4a) was isolated
at the end of the reaction along with metallic palladium.11,12
When the reaction was run at lower temperature in the same
cosolvent (Table 1, entry 12), the formation of metallic Pd was
not observed and 1 was again dominantly isolated (ca. 80% of
4a) along with some palladacycle 2 (ca. 20% of 4a) and trimer
3a (10% of starting PhNCO). Yet, when 4a is heated above
140 °C in nitrobenzene in the presence of a much slighter excess
of phenylisocyanate (5a) and free o-phen in excess, the quasi-
quantitative conversion of 4a into 2 is observed along with traces
of metallic palladium (Table 1, entry 10 and Scheme 3). Notably,
the palladacycle 1 could never be isolated under these condi-
tions.
We had previously established that both 1 and 2 were
unreactive against phenylisocyanate (5a) at ambient tempera-
ture.11 Thus, these complexes can by no means constitute
intermediates in this transformation and correspond rather to
(17) Paul, F. Ph.D. Thesis, Strasbourg 1, 1993.
9
7296 J. AM. CHEM. SOC. VOL. 129, NO. 23, 2007

Pd(0)-Catalyzed Trimerization of Arylisocyanates
A R T I C L E S
Scheme 3. Reaction of Phenylisocyanate (5a) with Complex 4a
under Heating
Table 3. Catalytic Trimerization of Phenylisocyanate by Various
(N-N)Pd(dba) Precursors at 25 °C in ArNCO/PhNO2 (1:2)a
isolated yield
entry
N
−N
t (min)
of 3a
1
2
3
4
5
6
o-phen
dmphen
dmphen
tmphen
bipy
15
15
1050
15
15
995
17
0
0
18
9
57
Table 2. Catalytic Trimerization of Various Arylisocyanates by 4a
at 25 °C in ArNCO/PhNO2 (1:2) Mixturesa
bipy
isolated
entry
Ar
[ArNCO]/[4a]
t (min)
yieldb
a The initial concentration of the various (N-N)Pd(dba) complexes 4a-d
is 0.8 × 10-2 mol L-1 ([PhNCO]/[(N-N)Pd(dba)] ) 350).
1
2
3
4
5
6
7
Ph (5a)
350
810
350
310
310
278
278
15
120
15
15
1010
15
17
0
49
0
12
3
8
2,6-(iPr)2(C6H3) (5b)c
Chart 2. Relative Energies (kcal/mol) of Methylisocyanate
Oligomers (n ) 2, 3, 4) with Regard to Methylisocyanate
4-F(C6H4) (5c)d
4-Me(C6H4) (5d)
4-Me(C6H4) (5d)
4-MeO(C6H4) (5e)
4-MeO(C6H4) (5e)
920
a Initial concentration of 4a: 0.8 × 10-2 mol L-1
.
b Of corresponding
1,3,5-triarylisocyanurate. c Initial concentration of 4a: 2.9 × 10-2 mol L-1
.
d Initial concentration of 4a: 0.4 × 10-2 mol L-1
.
be evidenced and only decomposition in metallic Pd is observed
after several days. We believe that the steric strain introduced
on the ortho positions of the dmphen ligand by the methyl
groups is responsible for that lack of reactivity.
Scheme 4. Reaction of 2,6-Di(iso-propyl)phenylisocyanate (5b)
with Complex 4a
DFT Modeling of the Catalytic Transformation. DFT
calculations were next carried out to determine the complete
catalytic cycle and rationalize the experimental observations.
All these calculations were carried out within the framework
of DFT using the Gaussian 03W set of programs.42-46 The
structure of the 14 VE metallic fragment [(o-phen)Pd] was kept
unmodified, but the phenylisocyanate was replaced by meth-
ylisocyanate to save computation time (see Supporting Informa-
tion for further details). To take into account the role played
by the polarity of the medium, single point calculations were
realized using the polarized continuum model (PCM) method
using dichloromethane as solvent. The choice of dichlo-
romethane is not arbitrary and results from the fact that this
solvent was used as cosolvent several times. Furthermore, the
dielectric constant of CH2Cl2 (ꢀ ) 9.1) is close to that of
phenylisocyanate (ꢀ ) 8.9), which was used as solvent in the
initial experiments.
deactivated forms of the catalyst. We have now checked that
neither 1 nor 2 can be obtained from insertion of an (o-phen)-
Pd0 fragment in 3a and subsequent rearrangement. Indeed, when
4a is refluxed with an excess of 3a (i.e., 15 equiv), in the
presence of an excess of free o-phen, the reaction leads mostly
to decomposition of the compound 4a in metallic Pd. Therefore,
both 1 and 2 do not result from trapping of the (o-phen)Pd0
fragment by the 1,3,5-triphenylisocyanurate (3a) catalytically
generated during the catalysis.18
The reaction of 4a with other arylisocyanates was also
attempted (Table 2). Thus, when 4a is reacted with a sterically
crowded isocyanate such as the 2,6-di(iso-propyl)phenylisocy-
anate (5b), two palladacycles can be isolated instead of the
corresponding 1,3,5-triarylisocyanurate (3b) (Scheme 4). The
former (7) is a known four-membered metallacycle11 and is
essentially present as a side product (5%), while the other is a
new five-membered palladacycle (8), analogous to 1, which
constitutes the dominant form (70%) of the Pd(0) precursor
introduced.
A first important point is the relative stability of the cyclic
oligomers (R-NCO)n that may form during the process (n ) 2,
3, 4). As expected, the trimer was found to be more stable than
the dimer and the tetramer (Chart 2).
In the initial step, it was assumed that coordination of the
isocyanate on the [(o-phen)Pd]0 fragment (noted I hereafter)
preferentially takes place at the CdN double bond. This
preference can indeed be expected when one examines the MO
scheme of the methylisocyanate. The HOMO and the HOMO-1
were found to be more developed on the CdN double bond,
while the HOMO-2 and HOMO-3 are mostly located on the
π-system of the CdO double bond (see Supporting Information).
Furthermore, the LUMO and the LUMO+1 were also found to
be more developed on the nitrogen atom. Note also that the
same MO ordering is found for phenylisocyanate. This hypoth-
esis was further confirmed by calculations of the possible 16
VE η2-complexes IIa (coordination of methylisocyanate to
fragment I through the CdN double bond) and IIb (coordination
of methylisocyanate to fragment I through the CdO double
bond) whose formation were found to be exothermic. As shown
Now, with less crowded arylisocyanates such as 5c or 5e,
the catalytic reaction takes place and is significantly influenced
by the para substituent (Table 2). Apparently, the reaction
proceeds faster with electron-withdrawing groups.
Finally, we have examined the reaction of phenylisocyanate
(5a) with the other Pd(0) precursors 4b-d. Except with the
dmphen complex (4b), the catalytic trimerization takes place
to a comparable extent (Table 3). With 4b, no trimerization can
(18) Note that 2 and the known corresponding four-membered palladacycle could
be isolated from 4a and 6a under similar conditions (see Supporting
Information). However, this does not constitute a proof of reactivity of 4a
with the cyclodimer.19
(19) It is known that uretediones equilibrate with the corresponding isocyanates
upon heating. It constitutes often the favored product under ambient
conditions, but the kinetics of this reaction are very slow.20
(20) Tiger, R. P.; Sarynina, L. I.; Entelis, S. G. Russ. Chem. ReV. 1985, 41,
774-785.
9
J. AM. CHEM. SOC. VOL. 129, NO. 23, 2007 7297

A R T I C L E S
Paul et al.
Scheme 5. Coordination of Methylisocyanate to Complex
[(o-phen)Pd] I To Form IIa and IIb; NBO Charges Are Given in e
Units
Table 4. Geometrical Data for C-O and C-N Bonds in Free
Methylisocyanate and in Complexes IIa and IIba
bond
CH3NCO
IIa
IIb
C-O
C-N
1.18 (1.83)
1.20 (1.95)
1.22 (1.65)
1.28 (1.56)
1.26 (1.46)
1.26 (1.71)
a Distances are given in angstroms. Respective Wyberg bond indexes
are given in parenthesis.
in Scheme 5, complex IIa was found to be 4.3 kcal/mol more
stable than complex IIb. In both compounds, the CdO and Cd
N have lost their double bond character and a significant
lengthening of bond distances is noted with respect to free
methylisocyanate. This is also confirmed by calculation of
Wyberg bond indexes (WBI), as summarized in Table 4. Also,
NBO calculations indicate that the nucleophilic character of the
nitrogen atom in IIa and IIb is slightly enhanced upon
coordination, its charge increasing from qNBON ) -0.56 e in
methylisocyanate to -0.68 e in IIa and -0.66 e in IIb. This
observation will prove important to rationalize the subsequent
steps of the catalytic cycle. Note also that the positive charge
on the palladium atom increases in IIa and IIb in comparison
to that in I (Scheme 5).
Figure 1. Views of the transition states TSIIa-IIb (a) connecting complexes
IIa and IIb and TSIIb-III (b) connecting the two structures IIb and III as
given by DFT. Most significant bond distances (Å) and bond angles (deg):
TSIIa-IIb: Pd-N1, 2.298; Pd-N2, 2. 2.275; Pd-C1, 2. 1.953; Pd-O1,
2.636; Pd-N3, 2.511; C1-O1, 1.226; C1-N3, 1.270; C2-N3, 1.455; N1-
Pd-N2, 73.8. TSIIb-III: Pd-N1, 2.105; Pd-N2, 2.228; Pd-C1, 2.269;
Pd-O1, 2.053; C1-O1, 1.274; C1-N3, 1.270; C2-N3, 1.458; C3-N3,
2.215; C3-O2, 1.192; C3-N4, 1.232; C4-N4, 1.452; N1-Pd-N2, 76.8;
C1-N3-C2, 116.9; O2-C3-N4, 161.1.
A transition state TSIIa-IIb connecting these two minima was
found (∆EqIIa-IIb ) 20.6 kcal/mol; Scheme 5; Figure 1a). The
negative frequency characterizing this transition state corre-
sponds to a rotation around the palladium carbon bond axis.
These results where the two coordination modes lie close in
energy resemble those reported for ketene on diphosphane Ni-
(0) and Pt(0) fragments.21
local minimum on the potential energy surface (PES) and results
from the attack of the uncoordinated nitrogen atom of IIb on a
free molecule of isocyanate. The optimized geometry for the
transition state (TSIIb-III) connecting IIb to III is shown in
Figure 1b. This coupling reaction was found to be slightly
exothermic (∆EIIb-III ) -4.4 kcal/mol) and requires a relatively
weak activation energy of ∆EqIIb-III ) 14.2 kcal/mol (Scheme
6). It is noteworthy that the structure of complex III features a
small interaction between the R-hydrogen atom of the o-phen
ligand and the oxygen atom of the second molecule. The
electronic structure of the second isocyanate is significantly
disturbed by the attack of the nitrogen atom, the NBO charge
of the nitrogen atom is increased (qNBON ) -0.69 e III
compared to qNBON ) -0.66 e in IIb or -0.56 e in free
methylsiocyanate), and the CdO and CdN bonds have partially
lost their double bond character (C-O, 1.25 with WBI ) 1.40;
C-N, 1.29 with WBI ) 1.58). The latter might be sketched by
a zwitterionic valence bond mesomer (IIIz) which certainly
possesses a non-negligible weight in the bonding description
of III.
Having shown the preference for CdN coordination, we next
examined the introduction of a second molecule of methyliso-
cyanate. In a first series of calculations, we tried to coordinate
two molecules of methylisocyanate to fragment I to see whether
an 18 VE bisisocyanate complex such as [(o-phen)Pd-
(MeNCO)2] could be obtained. Whatever the starting structure
proposed, all attempts resulted in the re-formation of either
complex IIa or IIb. This result prompted us to investigate a
second sphere mechanism in which one molecule of methyl-
isocyanate would react with the nucleophilic nitrogen atom. We
logically focused our attention on complex IIb in which the
nitrogen atom is not coordinated to the palladium center. Indeed,
one may expect than coordination of nitrogen to Pd in IIa makes
this complex less reactive toward nucleophilic attack on
methylisocyanate. The reactivity of complex IIa will be
discussed later. The new complex III was characterized as a
We next inquired about the formation of the five-membered
metallacycle 1 which was isolated. Formation of this complex
(corresponding to IV in our calculations) from complex III was
found to be strongly exothermic (∆E ) -42.6 kcal/mol) and
only involves a weak activation barrier (∆Eq
) 7.6 kcal/
(21) (a) Hofmann, P.; Perez-Moya, L. A.; Steigelmann, O.; Reidel, J. Organo-
metallics 1992, 11, 1167-1176. (b) Wright, C. A.; Thorn, M.; McGill, J.
W.; Sutterer, A.; Hinze, S. H.; Prince, R. B.; Gong, J. K. J. Am. Chem.
Soc. 1996, 118, 10305-10306.
III-IV
mol). An isomer (V) of complex IV featuring a Pd-O bond in
place of the Pd-N bond of IV was also characterized as a local
9
7298 J. AM. CHEM. SOC. VOL. 129, NO. 23, 2007

Pd(0)-Catalyzed Trimerization of Arylisocyanates
A R T I C L E S
Scheme 6. Energetic Pathway for the Transformation of IIb into III; NBO Charges Are Given in e Units
Scheme 7. Formation of Complexes IV and V from Complex III
minimum on the PES. Its formation also requires a weak
activation energy starting from complex III (∆Eq
) 9.0
yields complex IX in which one molecule of trimer is
coordinated to the 14 VE fragment I through one C-N bond
(Scheme 9). This elimination process is slightly endothermic
III-IV
kcal/mol) but appears to be less exothermic (∆E ) -32.3 kcal/
mol) than that of IV. Both IV and V are connected by a
transition state (TSIV-V; ∆Eq
) 49.0 kcal/mol) which
IV-V
involves a rotation around the C3-N3 bond (Scheme 7). Given
the energetic closeness of the two energy barriers, DFT predicts
that the formation of both complexes can occur during the
process. A view of the transition state TSIII-V is presented in
Figure 2a. Importantly, DFT confirms that complex IV
behaves as a thermodynamic sink, in line with experimental
observations.
Complex V turns out to be the only possible reactive
intermediate to explain the formation of the cyclic trimer. As
with III, complex V features one exocyclic nitrogen atom that
is prone to react with a third molecule of methylisocyanate
(qNBON ) -0.62 e). As shown in Scheme 8, complex VII which
results from the nucleophilic attack of V onto MeNCO was
found to be a local minimum onto the PES and its formation is
only slightly endothermic (∆E ) 4.5 kcal/mol).
As with the transformation between IIb and III, complex V
is connected to VII through TSV-VII (Figure 2b) whose
formation requires an activation energy of ∆EqV-VII ) 16.4 kcal/
mol. Again the chain growth elementary process operative here
is not a cycloaddition or an insertion but rather a nucleophilic
attack of the corresponding terminal nitrogen atoms on the
electrophilic carbonyl of the free methylisocyanate. Also, the
new intermediate VII might be sketched by a zwitterionic
valence bond structure (VIIz) which certainly possesses a non-
negligible weight in its bonding description. This compound
can then rearrange into the metallacyclic structure VIII in an
exothermic fashion (∆E ) -13.8 kcal/mol) via a nucleophilic
attack of the external nitrogen atom on the palladium center
Figure 2. Views of the transition states TSIII-V (a) connecting the two
structures III and V and TSV-VII (b) connecting the two structures V and
VII as given by DFT calculations. Most significant bond distances (Å) and
angles (deg): TSIII-V: Pd-N1, 2.229; Pd-N2, 2.309; Pd-C1, 1.891; Pd-
O1, 2.319; Pd-O2, 2.647; C1-O1, 1.235; C1-N3, 1.317; C2-N3, 1.458;
C3-N3, 1.477; C3-O2, 1.268; C3-N4, 1.289; C4-N4, 1.445; N1-Pd-
N2, 73.5; O1-C1-N3, 132.3. TSV-VII: Pd-N1, 2.185; Pd-N2, 2.075;
Pd-C1, 1.962; Pd-O2, 2.000; C1-O1, 1.122; C1-N3, 1.380; C2-N3,
1.458; C3-N3, 1.401; C3-O2, 1.301; C3-N4, 1.298; C4-N4, 1.449; C5-
N4, 2.021; C5-O3, 1.196; C5-N5, 1.254; C6-N5, 1.458; N1-Pd-N2, 81.7.
with an activation barrier of ∆Eq
) 23.0 kcal/mol (see
VII-VIII
Supporting Information).
The last step of the process involves the formation of the
cyclotrimer from complex VIII. This coupling process involves
two distinct steps. The first is the reductive elimination that
9
J. AM. CHEM. SOC. VOL. 129, NO. 23, 2007 7299

A R T I C L E S
Paul et al.
Scheme 8. Energetic Pathway Explaining the Formation of Complex VIII from Complex V; NBO Charges Are Given in e Units
Scheme 9. Reductive Elimination Process Leading to 1,3,5-Trimethylisocyanurate and to Complex I Formation
Scheme 10. Elimination of CO from Complex VIII
(∆EVIII-IX ) 6.4 kcal/mol) and requires a quite weak
activation barrier (∆Eq
) 10.6 kcal/mol; Figure 3). It is
VIII-IX
therefore likely to occur under ambient conditions. The decom-
plexation of the cyclotrimer, which allows the re-formation of
the active species I, is also slightly endothermic (∆EIX-I ) 5.9
kcal/mol).
Another experimental observation concerns the formation of
metallacycle 2 upon thermolysis of [(o-phen)Pd(dba)] with
excess PhNCO. A likely explanation for this experimental fact
is to consider that metallacycle VIII undergoes a retro-CO
insertion of the carbonyl group bonded to palladium. Actually,
this process, which requires a rather large activation energy (50.1
kcal/mol; see Supporting Information for TSVIII-X), was found
to be endothermic (∆E ) 24.9 kcal/mol with regard to VIII),
in line with the required heating of the reaction medium to
isolate 2 (Scheme 10).
Scheme 11. Energetic Pathway Relative to the Formation of
Complexes IV and V by Insertion of One Molecule of
Methylisocyanate into Complex IIa
We then turned our attention again to the first step. As
previously discussed, two η2 complexes IIa and IIb can be
formed, but the reactivity of complex IIa, which involves
coordination of methylisocyanate through the CdN bond, had
not been considered thus far. Two different attacks on this
intermediate differing by the orientation of the incoming
molecule of methylisocyanate were considered. In one case, the
methyl group is directed toward the o-phen ligand, whereas in
the second case the CdO groups of the incoming isocyanate
point toward the ligand. These attacks lead, respectively, to
complexes IV and V, which have been discussed earlier. The
two corresponding transition states TSIIa-IV and TSIIa-V from
IIa were computed (see Supporting Information). They also
present weak activation barriers (19.2 kcal/mol for the formation
of IV and 17.7 kcal/mol for the formation of V; Scheme 11).
Figure 3. View of the transition state TSVIII-IX connecting the two
structures VIII and IX as given by DFT calculations. Most significant bond
distances (Å) and angles (deg): Pd-N1, 2.282; Pd-N2, 2.125; Pd-C1,
1.971; Pd-O2, 2.253; Pd-N5, 2.076; C1-O1, 1.222; C1-N3, 1.419; C1-
N5, 1.923; C2-N3, 1.459; C3-N3, 1.392; C3-O2, 1.225; C3-N4, 1.392;
C4-N4, 1.459; C5-N4, 1.418; C5-O3, 1.225; C5-N5, 1.362; C6-N5,
1.450; N1-Pd-N2, 78.4.
9
7300 J. AM. CHEM. SOC. VOL. 129, NO. 23, 2007

Pd(0)-Catalyzed Trimerization of Arylisocyanates
A R T I C L E S
Scheme 12. Complete Energetic Pathway of the Catalytic Cycle Proposed; Energies Are Given in Kilocalories per Mole
Scheme 13. Proposed Catalytic Cycle; Scheme Features the
Different Complexes and Intermediates Calculated
Obviously, an alternative entry in the catalytic cycle from IIa
may be provided via V. This also shows that IIa will lead to
the side formation of IV to a comparable extent in the catalytic
process.
On the basis of these data and taking into account the level
of theory considered, it is difficult to conclude which one is
favored under the reaction conditions as both energetic
pathways are comparable. However, it is important to keep in
mind that steric factors have not been really taken into
account since methylisocyanate was used in calculations
instead of phenylisocyanate. One may propose that the
presence of a phenyl group probably disfavors the formation
of complex IV since in the transition state this phenyl group
would strongly interact with the o-phen ligand. This reasoning
emphasizes the formation of complex V as the first important
intermediate in the catalytic cycle. The most significant
information from these calculations is the fact that both
complexes IIa and IIb can enter the cycle. The complete
energetic pathway describing this trimerization process is
presented in Scheme 12.
from the starting isocyanates.19 A view of the whole catalytic
cycle is shown in Scheme 13. Expectedly, the trimerization
process is energetically downhill, and the various TS values
computed have a lower energy than the initial state, in line with
the spontaneous occurrence of this process under ambient
conditions. It is also obvious that the most stable compound
corresponds to IV, in line with the selective isolation of
palladacycle 1, after the death of the catalytic system, and to
VIII, the seven-membered palladacycle (never experimentally
observed) formed just before reductive elimination of the
isocyanurate takes place.
Finally, a last point that remained to be examined was the
possible palladium-promoted formation of uretediones, the
isocyanate cyclodimer, which was also isolated once in the
course of the previous investigations (6a), albeit in very low
yields. One simple way to form this byproduct would be by
reductive elimination from the intermediate IV. This mechanistic
step, which would simultaneously regenerate the active species
I, is akin to that previously computed from seven-membered
palladacycle VIII. Likewise, we found that an intermediate
complex (VI) might form during this process, which features
an interaction between the formed four-membered ring and the
fragment I (see Supporting Information). However, now the
reductive elimination is much more strongly disfavored from a
thermodynamic point of view, since the formation of I and of
Discussion
Mechanistic Proposal. In accordance with the working
hypothesis previously made,11a,12 we have evidenced now that
4a can react with phenylisocyanate (5a) to form the known
palladacycles 1 and 2, the latter complex being obtained after
subsequent decarbonylation. In addition, we have shown that
the Pd(0) precursors 4a-c catalyze the trimerization of various
aromatic isocyanates into the corresponding 1,3,5-triphenyliso-
cyanurates. Group 10 transition metal complexes are quite scarce
for such a reaction, and the good activity reported for 4a and
4c is remarkable in this connection.8,9 The present catalytic
system bears a strong analogy with the (bipy)(Et)2Ni (9) or
(bipy)2Ni (10) trimerization catalysts reported in 1975 by
Kashiwagi et al.22,23 A better turnover rate is presently obtained
with 4a.25 However, in contrast with the first of these catalysts,
the dimer is highly endothermic (∆Eq
) 64.4 kcal/mol).
IV-I
Notably, the intermediate VI is also found to be very high in
energy (∆EIV-VI ) 44.1 kcal/mol and ∆Eq
) 45.4 kcal/
IV-VI
mol), a fact that would be in line with the likely occurrence of
the reverse reaction.
Overall, these results rationalize the inertness stated for
palladacycle 1, since neither the formation of the 14 VE complex
I nor that of the isomeric palladacycle V is expected to occur
from IV under ambient conditions. Thus, uretediones are
certainly not catalytically generated as side products in this
reaction. Most likely, when isolated, these products originate
9
J. AM. CHEM. SOC. VOL. 129, NO. 23, 2007 7301

A R T I C L E S
Paul et al.
Scheme 14. Mechanisms Previously Envisioned for the Cyclotrimerization of Isocyanates
no linear polymer has ever been detected, even after running
the reaction at lower temperatures, conditions that are supposed
to favor the linear polymer.26,27 Notably, with the present Pd-
(0) catalysts, the precipitation of the cyclotrimer during catalysis
increases the viscosity of the reaction medium and certainly
limits their efficiency. Several attempts to overcome this
limitation by using a cosolvent resulted in decreased conversions
and pointed out the importance of having a large isocyanate
concentration present in the reaction medium.7b,8a This seems
to be a general feature of related systems, since many outstand-
ingcatalystsrecentlyreportedwerealsousedinneatisocyanate.7a,8b,28
Along with the heterogeneous reaction media, the incomplete
solubilization of the complexes 4a-d at the start of the reaction,
especially when cosolvents are used, constitutes a severe
limitation to obtain meaningful kinetic data on these systems.
Nevertheless, we have shown that the simultaneous presence
of a Pd(0) source and a given diimine ligand is needed for the
catalysis to proceed, and most possibly it is the complex 4a
itself that initiates the catalysis. Indeed, Stahl and co-workers
have shown that related (N-N)Pd(dba) precursors could
exchange the dba ligand for an olefin via a tetrahedral-shaped
18 VE intermediate.29 A similar intermediate with isocyanate
instead of the olefin would presently constitute a convenient
entry to the catalytic cycle and lead to a π-complex (of type
IIa or IIb) analogous to known compounds.30
intermediates could be disregarded,31 since the catalysis proceeds
in the presence of radical traps such as 9,10-dihydroanthracene
or duroquinone (Table 1). Notably quinones, such as duro-
quinone, are known to form stable and insoluble adducts with
4c.14,15,32 This can explain the decrease in the yield observed in
the presence of the latter additive. Moreover, one of the most
important findings is that a “metallacyclic” mechanism, involv-
ing successive concerted oxidative coupling reactions, as shown
in Scheme 14 and that resembles these often invoked for the
cyclotrimerization of alkynes on metal centers,33 is presently
not operative. Indeed, we have previously clearly demonstrated
experimentally that the palladacycle 1 was not an intermediate
in this transformation, even under forcing conditions.11a The
DFT calculations confirm that this type of compound (IV)
constitutes a thermodynamic sink for the catalytic system. Quite
clearly, this palladacycle constitutes a deactivated form of the
catalyst. Note that similar complexes were previously isolated
by other researchers from M(0) precursors (M ) Ni, Pd, Pt)
and ArNCO.22,34
Another mechanism often invoked in such a trimerization
reaction is the so-called “zwitterionic” mechanism (Scheme
14b). This type of mechanism was initially proposed by Sashoua
et al.35 for the anionic polymerization of isocyanates.10,20,26,36
It was then often extended to catalytic cyclotrimerization
Regarding the mechanism of the catalytic transformation, we
have experimentally shown that a mechanism involving radical
(30) (a) Hoberg, H.; Korff, J. J. Organomet. Chem. 1978, 150, C20-C22. (b)
Vasapollo, G.; Nobile, C. F.; Sacco, A. Inorg. Chim. Acta 1984, 83, 125-
128.
(31) For an example of radicalar coupling involving alkynes, see: Yang, J.;
Verkade, J. G. J. Am. Chem. Soc. 1998, 120, 6834-6835.
(32) Popp, B. V.; Stahl, S. S. J. Am. Chem. Soc. 2006, 128, 2804-2805.
(33) (a) Canovese, L.; Visentin, F.; Chessa, G.; Uguagliati, P.; Levi, C.; Dolmella,
A. Organometallics 2005, 24, 5537-5548. (b) Mu¨ller, C.; Lachicotte, R.
J.; Jones, W. D. Organometallics 2002, 21, 1118-1123. (c) Grotjahn, D.
B. In ComprehensiVe Organometallic Chemistry II; Hegedus, L. S., Ed.;
Pergamon: Oxford, 1995; Vol. 12, pp 741-770. (d) Parshall, G. W.; Ittel,
S. D. Homogeneous Catalysis; Wiley: New York, 1992; pp 208-213. (e)
Dieck, H.; Munz, C.; Mu¨ller, C. J. Organomet. Chem. 1990, 384, 243-
255. (f) Otsuka, S.; Nakamura, A. AdV. Organomet. Chem. 1976, 14, 245-
281.
(34) (a) Hoberg, H.; Radine, K.; Milcherheit, A. J. Organomet. Chem.
1985, 280, C60-C62. (b) Hoberg, H.; Oster, B. W.; Kru¨ger, C.; Tsay, Y.
H. J. Organomet. Chem. 1983, 252, 365-373. (c) Beck, W.; Rieber, W.;
Cenini, S.; Porta, F.; La Monica, G. J. Chem. Soc., Dalton Trans. 1974,
298-304.
(35) Shashoua, V. E.; Sweeny, W.; Tietz, R. F. J. Am. Chem. Soc. 1960, 82,
866-873.
(22) Kashiwagi, T.; Hidai, M.; Uchida, Y.; Misono, A. J. Polym. Sci., Part B:
Polym. Phys. 1975, 8, 173-175.
(23) It has been demonstrated that (bipy)(Et)2Ni (10) reductively eliminates
n-butane to generate the transient “(bipy)Ni(0)” species upon heating at
ambient temperature.24
(24) Uchino, M.; Asagi, K.; Yamamoto, A.; Ikeda, S. J. Organomet. Chem.
1975, 84, 93-103.
(25) For (bipy)(Et)2Ni (10), turnover
) 10, and for (bipy)2Ni (9),
turnover ) 22 in 3a after 8 h. Note that 2,2′-bipyridine alone was also
inactive.
(26) Parodi, F. In ComprehensiVe Polymer Science; Allen, G., Bevington, J.
C., Eds.; Pergamon Press: Oxford, 1989; Vol. 5, pp 387-412.
(27) Bur, A. J.; Fetters, L. J. Chem. ReV. 1976, 76, 727-746.
(28) Nambu, Y.; Endo, T. J. Org. Chem. 1993, 58, 1932-1934.
(29) (a) Popp, B. V.; Thorman, J. L.; Morales, C. M.; Landis, C. R.; Stahl, S.
S. J. Am. Chem. Soc. 2004, 126, 14832-14842. (b) Stahl, S. S.; Thorman,
J. L.; de Silva, N.; Guzei, I. A.; Clark, R. W. J. Am. Chem. Soc. 2003,
125, 12-13.
9
7302 J. AM. CHEM. SOC. VOL. 129, NO. 23, 2007

Pd(0)-Catalyzed Trimerization of Arylisocyanates
A R T I C L E S
reactions when nucleophilic organic catalysts were used,7 but
firm evidence for zwitterionic intermediates was only gathered
recently with phosphane catalysts,10 never with transition metal
catalysts.20 Our present computational results would actually
be more in accordance with such a mechanism, at least
concerning the initial steps. Note that although seldom encoun-
tered or isolated with transition metals,37 zwitterionic intermedi-
ates were nevertheless sometimes invoked in related processes.6b,38
Presently, DFT reveals that some intermediates, such as V for
instance, cannot be envisioned as true zwitterions, since the
charge is strongly delocalized. However, the important point
according to the calculations is that reaction with incoming
isocyanate molecules takes place on the pendant terminal
nitrogen atom, with stepwise formation of bonds, in each case.
Thus, except for the last steps, the actual mechanism differs
markedly from the “metallacyclic” mechanism previously
considered (Scheme 14a). In fact, when starting from type II
intermediates, each incremental step is more accurately envi-
sioned as a nucleophilic attack from the terminal nitrogen of
the palladium complex, rather than as an insertion of the
incoming isocyanate in the growing chain or as a metallacyclic
rearrangement (Scheme 13). This readily explains the sensitivity
to steric strain evidenced when diimine ligands bearing methyl
groups in positions 2 and 9 (Table 3) or isocyanate substrates
bearing isopropyl groups in positions 2 and 6 (Table 2) were
used. In addition, such a mechanism would perfectly be in line
with the previous finding of Stahl and co-workers, who
evidenced that 16 VE (N-N)Pd(L)0 complexes were quite
nucleophilic.29a,39 Also, the increased rate observed for
isocyanates bearing electron-withdrawing groups as well as the
higher yields of isocyanurate obtained in polar solvents are
consistent with such a nucleophilic polymerization mecha-
nism.10,26,41
oligomerization process. Calculations indicate that the ring
closure becomes favored over further chain extension (Scheme
13), the intermediate VII being much more prone to cyclize
into VIII than to react further with free isocyanate. Also, with
VII, no evidence for a transition state resulting from ring closure
on the R-carbonyl could be obtained from our calculations.
Apparently, with this intermediate, ring closure at the metal is
energetically more favored over typical processes occurring
following the zwitterionic mechanism. Then, depending on the
temperature of the medium, the metallacyclic intermediate VIII
can either regenerate the active species or eliminate carbon
monoxide to form type X palladacycles, such as 2, also inactive
for the catalysis. All these steps are precisely these that would
terminate a metallacyclic oligomerization process (Scheme 14a).
Thus, the mechanism that we propose for these Pd(0) diimine
catalysts is actually a blend of the “zwitterionic” and “metal-
lacyclic” mechanisms. In addition, we also found that the
formation of the very stable five-membered intermediate IV
(corresponding to 1) can occur competitively with that of VIII
at earlier stages of the catalysis. This explains the slow
deactivation of the catalyst under ambient conditions.
Conclusions
This work reveals that diimine-Pd(0) complexes constitute
novel and efficient catalytic precursors for cyclotrimerizing
aromatic isocyanates under ambient conditions without the
occurrence of competing linear polymerization. The experi-
mental and theoretical (DFT) mechanistic investigations con-
ducted during this study demonstrate that the mechanism of this
transformation does not proceed by successive cycloadditions
on metallacyclic intermediates followed by the final reductive
elimination of the isocyanurate, as might have been inferred
by analogy with alkyne oligomerization reactions on group 10
metals, nor does the catalysis proceed by a typical zwitterionic
process, which would lead to side formation of linear isocyanate
polymers. Rather, our findings suggest that, at least in a formal
sense, the trimerization mechanism on Pd(0) diimine complexes
starts like an anionic polymerization mechanism, where the Pd-
(0) center plays the role of a nucleophilic initiator and gives
rise to zwitterion-like intermediates. However, these charge-
separated intermediates, upon lengthening, are prone to cyclize
at the metal center and give rise to several metallacyclic
intermediates of different stabilities. Eventually, a seven-
membered metallacycle (VIII) is formed that generates the
cyclotrimer and re-forms the active species upon reductive
elimination. Notably, we have also evidenced here that, accord-
ingly to our starting hypothesis, the known palladacycles 1 and
2 can be isolated from phenylisocyanate and the (o-phen)Pd-
(dba) precursor. Actually, these palladacycles are inactive for
the trimerization reaction and constitute deactivated forms of
the catalyst. Their formation is readily explained in light of the
present investigation.
In contrast to previous examples of related zwitterionic
trimerizations (Scheme 14b), this transformation is not a “living”
process, due to the absence of formation of linear isocyanate
polymers. Indeed, a clear termination step takes place for the
(36) (a) Ahn, J.-H.; Shin, Y.-D.; Nath, G. Y.; Park, S.-Y.; Rahman, M. S.; Samal,
S.; Lee, J.-S. J. Am. Chem. Soc. 2005, 127, 4132-4133. (b) Ahn, J.-H.;
Lee, J.-S. Macromolecules 2001, 34, 5070. (c) Shin, Y.-D.; Kim, S.-Y.;
Ahn, J.-H.; Lee, J.-S. Macromolecules 2001, 34, 2408-2410. (d) Lee, J.-
S.; Ryu, S.-W. Macromolecules 1999, 32, 2085-2087.
(37) (a) De Renzi, A.; Di Blasio, B.; Panunzi, A.; Pedone, C.; Vitagliano, A. J.
Chem. Soc., Dalton Trans. 1978, 1392-1397. (b) Baba, A.; Oshiro, Y.;
Agawa, T. J. Organomet. Chem. 1975, 87, 247-256. (c) Hasegawa, S.;
Itoh, K.; Ishii, Y. Inorg. Chem. 1974, 13, 2675-2679.
(38) (a) Scriban, C.; Glueck, D. S.; Zakharov, L. N.; Kassel, W. S.; DiPasquale,
A. G.; Golen, J. A.; Rheingold, A. L. J. Am. Chem. Soc. 2006, 25, 5757-
5767. (b) Andrews, M. A.; Gould, G. L.; Voss, E. J. Inorg. Chem. 1996,
35, 5740-5742. (c) Adams, R. D.; Falloon, S. B. Organometallics 1995,
14, 1748-1755. (d) Adams, R. D.; Falloon, S. B. J. Am. Chem. Soc. 1994,
116, 10540-10547.
(39) The (o-phen)Pd0 fragment is isolobal to a singlet carbene fragment, as
outlined by several researchers.40
(40) (a) Hoffmann, R. Angew. Chem., Int. Ed. Engl. 1982, 21, 711-724. (b)
Yoshida, T.; Tatsumi, K.; Otsuka, S. Pure Appl. Chem. 1980, 52, 713-
727. (c) Otsuka, S. J. Organomet. Chem. 1980, 200, 191-205.
(41) Given the low solubility of the Pd(0) catalyst precursors (4a-d), the latter
effect might also be connected with solubility issues.
(42) Frisch, M. J.; et al. Gaussian 03, revision B.04; Gaussian, Inc.: Pittsburgh,
PA, 2003.
(43) (a) Becke, A. D. J. J. Phys. Chem. 1993, 98, 5648-5662. (b) Perdew, J.
Acknowledgment. This article is dedicated to John A. Osborn
(1939-2000). The CNRS, the Ecole Polytechnique, and the
IDRIS (for computer time, project No. 51616) are thanked for
supporting this work. F.P. also acknowledges “Rhone-Poulenc
Recherches” for financial support when this study was initiated
under the supervison of J.A.O. and S. Sinbandhit for NMR
assistance.
P.; Wang, Y. Phys. ReV. B 1992, 45, 13244-13249.
(44) Hay, P. J.; Wadt, W. R. J. Chem. Phys. 1985, 82, 299-310.
(45) Ehlers, A. W.; Bohme, M.; Dapprich, S.; Gobbi, A.; Hollwarth, A.; Jonas,
V.; Kohler, K. F.; Stegmann, R.; Veldkamp, A.; Frenking, G. Chem. Phys.
Lett. 1993, 208, 111-114.
(46) (a) Barone, V.; Improta, R.; Rega, N. Theor. Chem. Acc. 2004, 111, 237-
245. (b) Cossi, M.; Scalmani, G.; Rega, N.; Barone, V. J. Chem. Phys.
2002, 117, 43-54. (c) Cossi, M.; Barone, V.; Cammi, R.; Tomasi, J. Chem.
Phys. Lett. 1996, 255, 327-335. (d) Miertus, S.; Scrocco, E.; Tomasi, J.
Chem. Phys. 1981, 55, 117-129.
9
J. AM. CHEM. SOC. VOL. 129, NO. 23, 2007 7303

A R T I C L E S
Paul et al.
Supporting Information Available: (i) Complete experimen-
tal section, (ii) VT H NMR spectra of 4a, (iii) complete ref
42, (iv) structures of selected transition states, (v) energetic
pathway explaining the formation of the isocyanate dimer by
reductive elimination, and (vi) computed Cartesian coordinates,
thermochemistry, PCM energies, and lower frequencies of all
theoretical structures. This material is available free of charge
via the Internet at http://pubs.acs.org.
1
JA068291K
9
7304 J. AM. CHEM. SOC. VOL. 129, NO. 23, 2007