Unexpected four-membered over six-membered ring formation during the synthesis of azaheterocyclic phosphonates: Experimental and theoretical evaluation

Article abstract of DOI:10.1021/ja0584119

The cyclization of functionalized aminophosphonates is studied on both experimental and theoretical grounds. In a recently described route to phosphono-β-lactams [Stevens C. V.; Vekemans, W.; Moonen, K.; Rammeloo, T. Tetrahedron Lett. 2003, 44, 1619], it

Full text of DOI:10.1021/ja0584119

Published on Web 06/13/2006
Unexpected Four-Membered over Six-Membered Ring
Formation during the Synthesis of Azaheterocyclic
Phosphonates: Experimental and Theoretical Evaluation
Veronique Van Speybroeck,*,† Kristof Moonen,‡ Karen Hemelsoet,†
Christian V. Stevens,*,‡ and Michel Waroquier†
Contribution from the Center for Molecular Modeling, Ghent UniVersity, Proeftuinstraat 86,
B-9000 Ghent, Belgium, and Research Group SynBioC, Department of Organic Chemistry,
Faculty of Bioscience Engineering, Ghent UniVersity, Coupure links 653,
B-9000 Ghent, Belgium
Received December 12, 2005; Revised Manuscript Received April 26, 2006;
E-mail: veronique.vanspeybroeck@ugent.be; chris.stevens@ugent.be
Abstract: The cyclization of functionalized aminophosphonates is studied on both experimental and
theoretical grounds. In a recently described route to phosphono-â-lactams [Stevens C. V.; Vekemans, W.;
Moonen, K.; Rammeloo, T. Tetrahedron Lett. 2003, 44, 1619], it was found that starting from an ambident
allylic anion only four-membered rings were formed without any trace of six-membered lactams. New anion
trapping experiments revealed that the γ-anion is highly reactive in intermolecular reactions. Ab initio
calculations predict higher reaction barriers for the γ-anion due to restricted rotation about the C-N bond
and due to highly strained transition states during ring closure. The sodium or lithium counterion, explicit
dimethyl ether solvent molecules, and bulk solvent effects were properly taken into account at various
levels of theory.
1. Introduction
lactams without any trace of the corresponding six-membered
lactams. In a quite similar reaction where the anion was
stabilized by a carbonyl group instead of a phosphonate group,
the same experimental preference was observed toward the
formation of the four-membered ring. However no explanation
for this peculiar reactivity was given.9,10 The higher reactivity
of the R-canonical anion compared to the γ-canonical anion
could be believed to be important for the observed reactivity;
however, steric factors may also play a role since exclusive
γ-alkylation was observed using â,γ-unsaturated R-silyloxy-
phosphonates.11
Intramolecular ring-closure competition between a four-
membered and a six-membered ring is generally accepted to
lead to the six-membered ring.2 Several ring-closure studies have
been performed on the formation of lactones from ω-bromo-
alkanoate ions to evaluate the influence of the ring size.3,4 When
the exo-tet situation is considered during ring closure,5 the six-
membered ring preference can be explained by a smaller Baeyer
strain (angle strain) as well as a smaller Pitzer strain (due to
eclipsed conformations of the hydrogens). The possibility to
form a four- or a six-membered ring exists when an R-alkenyl
anion can ring close in an exo-tet mode either directly or through
its mesomeric canonical form. However, a surprising preference
for the formation of a four-membered ring was found during
the synthesis of azaheterocyclic phosphonates.6-8 When treated
with a base, N-chloroacetyl-1-aminoalkenyl phosphonates de-
rived from cinnamaldehyde exclusively lead to phosphono-â-
The main goal of the present paper is to unravel the origin
of this remarkable selectivity toward four-membered ring
formation. This is done both on experimental and theoretical
grounds. In addition to the previously reported synthesis of
4-phosphono-â-lactams new experiments were set up to trap
the intermediate anions prior to cyclization to obtain information
about the factors controlling the stability of the mesomeric
resonance contributors. Further microscopic insight into stabi-
lization and charge distribution is obtained by means of ab initio
theoretical calculations.
† Center for Molecular Modeling.
‡ Research Group SynBioC.
(1) Deleted in proof.
(2) Smith, M. B.; March, J. Kinetic Requirements for reactions. AdVanced
Organic Chemistry, 5th ed.; J. Wiley and Sons: New York, 2001; pp 278-
282.
The theoretical study of organometallic species as encountered
here is challenging as the effects of the metallic counterion and
solvent all contribute to the final reaction preference. To evaluate
(3) Galli, C.; Illuminati, G.; Mandolini, L.; Tamborra, P. J. Am. Chem. Soc.
1977, 99, 2591-2597.
(4) Mandolini L. J. Am. Chem. Soc. 1978, 100, 550-554.
(5) Baldwin, J. E. J. Chem. Soc., Chem. Commun. 1976, 734-736.
(6) Stevens, C. V.; Vekemans, W.; Moonen, K.; Rammeloo, T. Tetrahedron
Lett. 2003, 44, 1619-1622.
(9) Bossio, R.; Marcos, C. F.; Marcaccini, S.; Pepino R. Tetrahedron Lett.
1997, 38, 2519-2520.
(7) Moonen, K.; Laureyn, I.; Stevens, C. V. Chem. ReV. 2004, 104, 6177-
(10) Marcaccini, S.; Pepino, R.; Pozo M. C. Tetrahedron Lett. 2001, 42, 2727-
2728.
6215.
(8) Moonen, K.; Stevens, C. V. Synthesis 2005, 3603-3612.
(11) Hata, T.; Nakajima, M.; Sekine, M. Tetrahedron Lett. 1979, 2047-2050.
9
8468
J. AM. CHEM. SOC. 2006, 128, 8468-8478
10.1021/ja0584119 CCC: $33.50 © 2006 American Chemical Society

Cyclization of Functionalized Aminophosphonates
A R T I C L E S
the contribution of various parameters on the geometries,
energies, and the reaction outcome, the reaction route was
modeled in different environments: in the gas phase, with
transmetalation with sodium or lithium as the counterion, and
finally with inclusion of solvent interactions. The theoretical
results are correlated with the experimental data.
From a theoretical point of view the stabilization of allylic
anions by phosphorus substituents has attracted considerable
interest.12-19 The review of Katritzky20 on the regioselectivity
of reactions of heteroatom-stabilized allylic anions with elec-
trophiles presents both experimental and theoretical consider-
ations on this item. Unsymmetrically substituted allyl anions
can react with electrophiles both intra- or intermolecularly at
two sites, making these species of considerable importance. An
early theoretical work on the stabilization of allylic anions by
heteroatom-containing subgroups was presented by Denmark
and Cramer where neutral, anionic, and lithiated P-allyl and
P-methylphosphonic diamides were discussed.16
proved to be the key structural feature of this class of antibiotics.
Furthermore, increasing resistance is a major threat in hospitals
and in the community29 and is mainly caused by the emergence
and spread of â-lactamases, which are â-lactam-deactivating
enzymes evolutionarily related to penicillin-binding proteins
(PBPs).30-33 Therefore, current research is focusing on new
functionalized â-lactams having â-lactamase-inhibiting activity
or combining antibacterial activity with enhanced stability
toward â-lactamases.34,35 In addition to these well-known
antibacterial properties, â-lactams have also found application
in inhibiting serine proteases. Members of this major class of
enzymes are involved in numerous physiological processes
including protein turnover, digestion, blood coagulation and
wound healing, fertilization, cell differentiation and growth, cell
signaling, the immune response, and apoptosis.36,37
2. Experimental and Calculation Methods
1
General. H NMR spectra were recorded at 300 MHz with CDCl3
as solvent and tetramethylsilane (TMS) as internal standard. 13C NMR
spectra were recorded at 75 MHz and 31P NMR spectra at 121 MHz.
MS spectra were measured using electron spray ionization (4000 V).
THF was dried and distilled over sodium (benzophenone ketyl control).
The absolute value of the coupling constants (J) in Hz and assignments
Recently, Pratt and co-workers performed several advanced
calculations on the aggregation states of lithium carbenoids in
the gas phase and in ethereal solvent.21 From a methodological
point of view, this work deserves attention as it incorporates
the effects of the metallic cation which is further embedded in
solution. It was explicitly shown that charge-separated species
may be dramatically stabilized by solvent interactions. Pratt,
Vaˇn Nguˇyˆen, and Ramachandran have additionally studied the
performance of various electronic structure methods for the
accurate reproduction of barrier heights for lithium enolates.22
Recently, in a theoretical study performed by Ando, the origin
of π-facial stereoselectivity in the alkylation of enolates was
investigated.23 In this study, the effects of the counterion, explicit
coordination of solvent molecules with the lithium ion, and bulk
solvent effects were found to be important for the reaction
mechanism and energetics. Bearing this in mind, it is necessary
to evaluate the influence of each of these effects on the reaction
outcome for the present study.
1
of H and 13C peaks were determined using COSY, HSQC, HMBC,
and DEPT experiments.
Typical Procedure for the Synthesis of Phosphono-â-lactams 4.
A NaH suspension [0.24 g (6 mmol, 1.2 equiv)] in mineral oil is washed
three times with petroleum ether to remove the oil, and 15 mL of dry
THF is added. Then, 5 mmol of the corresponding N-chloroacetyl amino
alkenylphosphonate 1 in 5 mL of THF is added dropwise, and the
reaction mixture is refluxed for 2 or 3 h protected from moisture with
a CaCl2 tube. After cooling, the mixture is poured into 25 mL of water
and extracted with 20 mL of diethyl ether. The remaining water phase
is then washed two times with 10 mL of diethyl ether. The combined
organic phases are dried with MgSO4. The â-lactams are obtained in
good purity after filtration and evaporation of the solvent. Further
purification can be performed using column chromatography.
Ab Initio Molecular Orbital Calculations. All ab initio calculations
were carried out with the GAUSSIAN 03 software package.38 Density-
functional theory (DFT) methods39 have been shown to be more efficient
than wave function based procedures such as highly correlated post-
Hartree-Fock methods due to their excellent cost-to-performance ratio.
Despite the importance of phosphorus compounds in chemistry, only
limited levels of theory studies were conducted on such elements. The
work of Leyssens and Peeters addresses the accuracy of various post-
The main motivation for studying the remarkable selectivity
of the ring-closure reaction lies in the pharmaceutical importance
of the â-lactam ring as part of several important antibiotics,
including penicillin.24 Also some electron-deficient monocyclic
â-lactams appeared to be very active, including the commercially
available azthreonam,25-28 in which the â-lactam heterocycle
(12) Schleyer, P. V.; Clark, T.; Kos, A. J.; Spitznagel, G. W.; Rohde, C.; Arad,
D.; Houk, K. N.; Rondan, N. G. J. Am. Chem. Soc. 1984, 106, 6468-
6475.
(13) Streitwieser, A., Jr.; Rajca, A.; McDowell, R. S.; Glaser, R. J. Am. Chem.
Soc. 1987, 109, 4184-4188.
(27) Gordon, E. M.; Ondetti, M. A.; Pluscec, J.; Cimarusti, C. M.; Bonner, D.
P.; Sykes, R. B. J. Am. Chem. Soc. 1982, 104, 6053-6060.
(28) Brogden, R. N.; Heel, R. C. Drugs 1986, 31, 96-130.
(29) Neu, H. C. Science 1992, 257, 1064-1073.
(14) Denmark, S. E.; Dorow, R. L. J. Am. Chem. Soc. 1990, 112, 864-866.
(15) Denmark, S. E.; Dorow, R. L. J. Org. Chem. 1990, 55, 5926-5928.
(16) Denmark, S. E., Cramer, C. J. J. Org. Chem. 1990, 55, 1806-1813.
(17) Kranz, M.; Denmark, S. E. J. Org. Chem. 1995, 60, 5867-5877.
(18) Koch, R.; Anders, E. J. Org. Chem. 1995, 60, 5861-5866.
(19) (a) Leyssens, T.; Peeters, D. J. Mol. Struct. (THEOCHEM) 2004, 673, 79-
86. (b) Leyssens, T.; Peeters, D. J. Mol. Struct. (THEOCHEM) 2004, 686,
71-82.
(30) Bulychev, A.; Massova, I.; Miyashita, K.; Mobashery, S. J. Am. Chem.
Soc. 1997, 119, 7619-7625.
(31) Golemi, D.; Maveyraud, L.; Vakulenko, S.; Tranier, S.; Ishiwata, A.; Kotra,
L. P.; Samama, J.-P.; Mobashery, S. J. Am. Chem. Soc. 2000, 122, 6132-
6133.
(32) Meroueh, S. O.; Minasov, G.; Lee, W.; Shoichet, B. K.; Mobashery, S. J.
Am. Chem. Soc. 2003, 125, 9612-9618.
(20) Katritzky, A. R.; Piffl, M.; Lang, H. Y.; Anders, E. Chem. ReV. 1999, 99,
665-722.
(33) Fisher, J. F.; Meroueh, S. O.; Mobashery, S. Chem. ReV. 2005 105, 395-
424.
(21) Pratt, L. M.; Streitwieser, A. J. Org. Chem. 2003, 68, 2830-2838.
(22) Pratt, L. M.; Vaˇn Nguˇyˆen, N.; Ramachandran, B. J. Org. Chem. 2005, 70,
4279-4283.
(34) Sandanayaka, V. P.; Prashad, A. S. Curr. Med. Chem. 2002, 9, 1145-
1165.
(35) Hammond, M. L. J. Antimicrob. Chemother. 2004, 53, Suppl. S2, ii7-ii9.
(36) Bonneau, P. R.; Hasani, F.; Plouffe, C.; Malenfant, E.; LaPlante, S. R.;
Guse, I.; Ogilvie, W. W.; Plante, R.; Davidson, W. C.; Hopkins, J. L.;
Morelock, M. M.; Cordingley, M. G.; Deziel, R. J. Am. Chem. Soc. 1999,
121, 2965-2973.
(23) Ando, K. J. Am. Chem. Soc. 2005, 127, 3964-3972.
(24) Du¨rckheimer, W.; Blumbach, J.; Lattrell, R.; Scheunemann, K. H. Angew.
Chem. 1985, 24, 180-202.
(25) Imada, A.; Kitano, K.; Kintaka, K.; Muroi, M.; Asai, M. Nature 1981,
289, 590-591.
(37) Powers, J. C.; Asgian, J. L.; Ekici, O. D.; James, K. E. Chem. ReV. 2002,
102, 4639-4750.
(38) Frisch, M. J.; et al. Gaussian 03; Gaussian, Inc.: Pittsburgh, PA, 2003.
(39) Example of a reference work: Parr, R. G.; Yang, W. Density-Functional
Theory of Atoms and Molecules; Oxford University Press: New York, 1989.
(26) Sykes, R. B.; Cimarusti, C. M.; Bonner, D. P.; Bush, K.; Floyd, D. M.;
Georgopapadakou, N. H.; Koster, W. H.; Liu, W. C.; Parker, W. L.;
Principe, P. A.; Rathnum, M. L.; Slusarchyk, W. A.; Trejo, W. H.; Wells,
J. S. Nature 1981, 291, 489-491.
9
J. AM. CHEM. SOC. VOL. 128, NO. 26, 2006 8469

A R T I C L E S
Van Speybroeck et al.
Scheme 1. Synthesis of 4-Phosphono-â-lactamsa
a Reagents and conditions: (i) 1.1 equiv NaH, THF, ∆, 3 h.
Hartree-Fock and DFT methods on the geometries and energies of
phosphorus-containing compounds.19 The systems studied in that work
are relatively small, and the use of highly correlated electronic structure
methods such as MP2, MP4, QCISD, CCSD(T) was feasible. Their
results show that the geometries resulting from DFT(B3LYP), MP2,
and QCISD computations using the 6-31++g(d,p) are usually very
close to each other. Another study by Pratt et al. revealed the structures
of fluoromethyllithium and chloromethyllithium carbenoids in the gas
phase and in ethereal solvent,40 and it was found that the optimized
geometries from the mPW1PW91/6-31+g(d) and QCISD/6-31+g(d)
level agree to a large extent.
electron density than neutral atoms, it is essential to add diffuse
functions in the basis set.45 The following basis sets were used:
6-31+g(d), 6-311 g(d,p), 6-311++g(d,p). The first one represents a
double-ú basis, whereas the others are triple-ú basis sets. They further
differ in the number of diffuse and/or polarization functions added.
The vibrational frequencies of the optimized structures were
calculated at the same level of theory at which the geometries were
optimized. The transition states were verified to have only one
imaginary frequency and hence correspond to a first-order saddle point
on the potential energy surface.
3. Experimental Results
The systems of interest to the current work are too large to perform
geometry optimizations at the post-Hartree-Fock level. DFT methods
are therefore more suitable, and in the light of the previous discussion
the B3LYP and mPW1PW91 exchange correlation functionals were
selected to establish their influence on the geometry. The B3LYP
functional is the most popular version of a hybrid DFT method
combining the Becke exchange functional with the LYP correlation
functional and which also contains 20% of exact Hartree-Fock
exchange.41 According to various references, the B3LYP methods
provide excellent low-cost performance for geometry optimizations.42
This is further confirmed by specific studies performed by Leyssens
and Peeters on the properties of the phosphonate functional group.19
The mPW1PW91 functional is also a hybrid functional containing 25%
of exact exchange. The latter uses the modified Perdew-Wang
exchange functional that has improved the long-range behavior as
proposed by Adamo and Barone.43 DFT models may substantially
underestimate the activation energies for SN2 substitution reactions.
More information can be found in a study performed by Kormos and
Cramer44 where the relatively good performance of the mPW1PW91
functional is also shown. Another interesting paper within this respect
is the study recently performed by Pratt and co-workers on the
performance of various electronic structure methods for the evaluation
of barrier heights for gas-phase reactions of lithium enolates.22 They
showed that when DFT methods succeeded in locating a transition-
state structure, the geometries of the forming and breaking bond were
in reasonable agreement with MP2 calculations. Also single-point MP2
energies at B3LYP geometries provide a reasonable estimate of the
lithium enolate activation barriers in the gas phase. However, since
the systems under consideration are quite large, such a strategy is
computationally not feasible.
The synthesis of the phosphono-â-lactams under investigation
was performed, starting from the appropriate N-chloroacetyl-
1-amino alkenylphosphonates, 1, which were obtained according
to a literature procedure.6 Upon treatment with a strong base,
typically sodium hydride, a phosphorus-stabilized anion is
formed, which can be represented by two canonical resonance
contributors 2 and 3 (Scheme 1). The ring closure, performed
in 3 h at 66 °C, showed that only one product was formed with
a characteristic chemical shift between 24.01 and 24.75 ppm
(31P NMR), while the chemical shift for a vinylic phosphonate
is expected at significantly higher field (10-20 ppm). The same
products were obtained at room temperature using LiHMDS as
a base in diethyl ether as a solvent.
The proposed â-lactam structure was confirmed upon further
spectroscopic investigation. The very high infrared absorption
of the carbonyl (>1750 cm-1) is typical for highly strained rings.
Furthermore, the ring CH2(3) appears as a second-order spin
system in the 1H NMR spectrum, involving a geminal coupling
constant of 14.6-15.3 Hz and also quite large 31P couplings
(5.5-5.8 Hz), indicating the near presence of the phosphorus
atom. According to the integral of the signals in the region of
6-7 ppm in the proton spectrum, two alkenyl protons are present
in the molecule. A E-coupling of 16.2 Hz was found for all
five products, next to smaller 31P couplings. All aforementioned
phosphorus couplings disappeared when the proton spectrum
was run with selective 31P decoupling. All 13C peaks could be
attributed to the appropriate carbon in the azetidinones using
2D techniques (HSQC and HMBC) together with DEPT spectra.
The quaternary carbon atom bearing the phosphonate group is
expressed as a doublet (J ) 166.7-168.5 Hz) in the 13C
spectrum, with a chemical shift clearly within the aliphatic
Another issue of importance is the basis set used for the calculations.
As the present study involves anions, which have a more spread-out
(40) Pratt, L. M.; Ramachandran, B.; Xidos, J.; Cramer, C. J.; Truhlar, D. G. J.
Org. Chem. 2002, 67, 7607-7612.
(41) Becke, A. D. J. Chem. Phys. 1993, 98, 5648-5652.
(42) Coote, M. L. J. Phys. Chem. A 2004, 108, 3865-3872.
(43) Adamo, C.; Barone, V. J. Chem. Phys. 1998, 108, 664-675.
(44) Kormos, B. L.; Cramer, C. J. J. Phys. Org. Chem. 2002, 15, 712-720.
(45) Lynch, B. J.; Zhao, Y.; Truhlar, D. G. J. Phys. Chem. A 2003, 107, 1384-
1388.
9
8470 J. AM. CHEM. SOC. VOL. 128, NO. 26, 2006

Cyclization of Functionalized Aminophosphonates
A R T I C L E S
region (58.06-59.82 ppm). From this carbon, a clear HMBC
coupling was observed with the CH2(3) of the four-membered
ring.
same ratio. The reaction was then repeated under the same
conditions, however, using methyl iodide as an intermediate soft
electrophile. This might provide a better model for the intramo-
lecular alkylation. In this case, the γ-methylated product 11 was
formed as a mixture of (E,Z)-isomers (δ 31P ) 16.3 and 16.2).
Purification using column chromatography or crystallization
failed. The structure was confirmed, however, by comparing
the 13C, 31P, and DEPT spectra from the crude reaction mixture
with vinyl phosphonate 10 and by determining its mass spectrum
using LC-MS. No other phosphorus-containing compounds
were detected using 31P NMR.
The results obtained from the intermolecular reactions with
allylic anion 9 are in contradiction with the exclusive â-lactam
formation in the intramolecular reaction. To elucidate these
peculiarities, ab initio molecular orbital calculations were
performed as presented in the following section, giving micro-
scopic insight into the reaction preference.
An overview of the regioselectivity of intermolecular reactions
between heteroatom-stabilized allyl anions and electrophiles has
been presented by Katritzky et al.20 Even though an intramo-
lecular reaction is considered in our system, some important
directing factors could be identified such as nature of the
counterion, reaction conditions, the nature of the electrophile,
steric characteristics of the anion and the electrophile, coordina-
tion between the electrophile and the substrate prior to bond
formation, etc. Lithium and sodium have been evaluated as
counterions giving the same selectivity. The reaction conditions,
on the other hand, have an influence on the reaction rate but
not on the selectivity. Formation of the anion occurs fast and
can be followed visually by the evolution of hydrogen gas from
the reaction mixture upon addition of sodium hydride at room
temperature. After the H2-gas evolution ceases, the reaction
needs a further 3 h of reflux in THF (bp 66 °C), while the
reaction did not complete during an overnight reflux period in
diethyl ether (bp 35 °C). The high number of functionalities in
the substrate calls for a more profound investigation of other
determining factors. Furthermore, the intramolecular reaction
under investigation involves additional complications compared
to the intermolecular reactions, such as anion geometry and
substrate conformation.
4. Ab Initio Results
Free Anions. The energetically most favored structure of the
free carbanion is visualized in Figure 1 (labeled as 2a in Scheme
1). This structure was found by applying various internal
rotations about single bonds and selecting stepwise the most
stable conformation along the rotational potential. For the
discussion of the geometry some of the carbon atoms are labeled
as shown in Figure 2 (Anion I). The most important geometrical
parameters are listed in Table 1.
A phosphonate group is known to greatly stabilize carbanions
in the R-position through electrostatic interactions.15,20 This
might explain the lack of reactivity in the γ-position. To study
the selectivity of the reaction, an experiment was set up to trap
the allyl anion with common electrophiles, such as a proton (or
a deuteron) or methyl iodide. For this purpose, N-acetyl-1-amino
alkenylphosphonate, 8, was synthesized using a modified
procedure. Imine 6 was phosphonylated regioselectively using
dimethyl phosphite in methanol (97% yield),46 and the resulting
aminophosphonate 7 was then acetylated using acetyl chloride
and DMAP/pyridine (63% overall yield). The resulting N-acetyl-
1-amino alkenylphosphonate, 8, is sterically and electronically
very similar to the N-chloroacetyl substrates but is not prone to
intramolecular reactions because of the lack of a leaving group.
Initial attempts to prepare the anion with LiHMDS and to trap
it with methyl iodide resulted in complex mixtures. When the
anion was quenched with water, an unseparable mixture of
starting material and two new products was found in low yield
after extraction. The generation of several side products was
probably induced by the acidic acetyl protons leading to
unselective deprotonation by LiHMDS. However, when sodium
hydride was used instead, deprotonation proceeded selectively,
and a clean mixture of starting material 8 and its isomer 10
was found after quenching with water. Furthermore, both
isomers could be separated using column chromatography,
which allowed the unambiguous determination of the structure
of the resulting isomer with the use of selected NMR techniques.
The ratio of both isomers was rather surprisingly in favor of
the γ-protonated isomer 10, suggesting it as an important
contributor to the canonical resonance system of the allyl anion.
When deuterium oxide was used to quench the allyl anion, the
corresponding monodeuterated products were obtained in the
One particular internal rotation deserves special attention, i.e.,
about the C1-C2 bond, as it gives rise to either the E or Z
geometry of the γ-anionic form (R and γ refer to the phosphorus
atom). The E geometry cannot lead to six-membered rings. The
latter conformer was found to be highly unstable (14.8 kJ/mol),
indicating that the Z geometry is preferred which can either close
at the γ- or R-position.
Within the context of a further cyclization toward four- or
six-membered ring formation, the C5-C1 and C5-C3 distances
are relevant, as they represent the distances from the R- and
γ-carbon centers of the allylic unit to the chlorinated carbon
atoms. They amount to 2.78 and 3.65 Å, respectively, and
indicate that the most stable conformer of the anion has a
characteristic geometry that is suitable for four-membered ring
formation. Also, the internal rotation about the C1-N bond, will
play a crucial role in the reaction mechanism, as it determines
the relative position of the chlorinated carbon atom (C5) toward
the R- or γ-carbon center (respectively, C1 or C3). In the
energetically most favored conformation the dihedral angle
C2C1NC4 amounts to -82.19°, and C5 points away from the
carbon atom at the γ-position of the phosphorus (cf. Figure 1).
To unravel the mechanism of the further ring closure, it is
instructive to study the polarization of charge of the carbanion.
Therefore, we calculated partial atomic charges according to
the ChelpG scheme47 on the optimized geometries. The latter
method uses electrostatic derived charges and has proven its
accuracy for calculating atomic populations.48,49 The results are
given in Table 2. The phosphorus atom has a highly positive
charge (0.88), whereas the negative charge is delocalized over
the allyl unit with a slight polarization toward the γ-carbon atom
(47) Breneman, C. M.; Wiberg, K. B. J. Comput. Chem. 1990, 11, 361-373.
(48) Sigfridsson, E.; Ryde, U. J. Comput. Chem. 1998, 19, 377-395.
(49) Martin F.; Zipse, H. J. Comput. Chem. 2005, 26, 97-105.
(46) Van Meenen, E.; Moonen, K.; Acke, D.; Stevens, C. V. ArkiVoc 2006,
31-35.
9
J. AM. CHEM. SOC. VOL. 128, NO. 26, 2006 8471

A R T I C L E S
Van Speybroeck et al.
Figure 1. Optimized conformers of the anions in different molecular environments: 2a in the gas phase, 2a + Na including one sodium ion, 2a + Na +
1DME and 2a + Na + 2DME including one sodium ion and one or two DME molecules. All optimizations are performed at the B3LYP/6-31+g(d) level.
patterns (cf. Figure 2). It turns out that the net amount of
negative charge on the allyl unit systematically increases by
removing the phosphonate group and/or the nitrogen centered
group (Anion III, Anion IV, Anion V). In the presence of the
phosphonate group (Anion IV) a highly polarized carbon-
phosphorus bond is obtained with a large positive charge (1.12)
on the P atom. Previous results are in accordance with the
findings of Cramer and Denmark.16
Protonation Energies. The reactivity of the R- and γ-carbon
atoms (R and γ are referred with respect to the phosphonate
group) toward electrophilic attack can be estimated by calculat-
ing the protonation energies at both carbon positions (see
Scheme 2):
∆Ex ) E(Anion + Hx) - E(Anion) - E(H+)
(1)
with x ) R or γ. In Scheme 2, the R- and γ-protonated forms
are related to 8 and 10, respectively. The results are shown
schematically in Figure 3. For the Anions I and II having both
the phosphonate group and the nitrogen-containing subgroup,
the protonated molecule at the γ-position is more stable by 23
and 37 kJ/mol, respectively. When removing either the phos-
phonate or nitrogen centered subgroup, the stability sequence
inverts. These findings are in accordance with the experiments
described in the previous section: the γ-carbon has a larger
proton affinity than the R-carbon. However, these thermody-
Figure 2. Allyl stabilized carbanions.
(C3). Also, the oxygen of the PdO bond carries a large negative
charge and is the basic reason for the positively charged
phosphorus atom. The origin of this charge distribution is studied
more in detail by calculating the energetically most favored
structure of some other free anions with different substitution
9
8472 J. AM. CHEM. SOC. VOL. 128, NO. 26, 2006

Cyclization of Functionalized Aminophosphonates
A R T I C L E S
Table 2. Atomic Charges Calculated According to the ChelpG
Scheme at the B3LYP/6-31+g(d) Level of Theory for Various
Anions
Table 1. Most Important Geometrical Characteristics of Reactants
and Transition States at the B3LYP/6-31+g(d) levela
Gas Phase Structures
anion I (2a)
anion II
anion III
anion IV
anion V
Anion I (2a)
TS4
TS6
C5
Cl
C1
C2
C3
N
0.21
-0.29
-0.39
0.10
-0.47
0.20
0.13
-0.29
-0.32
-0.15
-0.20
0.15
0.30
-0.31
-0.43
0.11
-0.55
0.23
C5-C1
2.7762
3.6454
1.8131
0.34
2.1233
3.7046
2.4117
9.27
2.8210
2.5544
2.3520
C5-C3
-0.63
0.24
-0.59
-0.75
0.28
-0.66
Cl-C5
C3C2C1N
C2C1NC4
C1NC4C5
-1.32
-36.02
-15.24
-82.19
-1.11
-116.51
10.28
P
0.88
0.89
1.12
O
-0.65
-0.67
-0.70
Transmetalated Structures
Anion I
+
Na
TS4
+
Na
TS6
+
Na
anion I
Na
+
anion I
Li
+
anion I
Na
+
anion I
+
+
1DME
Na
+2DME
C5-C1
2.8176
3.6139
1.7986
1.9825
3.6417
2.4820
6.59
2.7987
2.1833
2.4213
C5-C3
C5
Cl
C1
C2
C3
N
P
O
Na/Li
O1
O2
-0.05
-0.20
-0.34
0.04
-0.51
-0.04
0.81
0.03
-0.21
-0.42
0.01
-0.40
0.02
0.96
-0.69
0.54
0.03
0.06
Cl-C5
-0.22
-0.34
0.09
-0.47
-0.01
0.80
-0.70
0.55
-0.36
-
-0.22
-0.30
0.10
-0.49
-0.02
0.83
-0.61
0.49
-0.20
-0.19
C3C2C1N
C2C1NC4
C1NC4C5
Na-C1
Na-C3
Na-C6
Na-C7
-4.41
-2.25
-80.36
-117.36
-31.36
-16.72
2.86
11.28
2.7097
2.9223
2.7938
2.7873
3.6904
2.8038
3.5203
2.8747
2.9833
2.7610
2.7168
3.0384
-0.64
0.59
Anion I
+
Li
TS4
+
Li
TS6 + Li
C5-C1
C5-C3
Cl-C5
2.8327
3.6866
1.7986
1.9674
2.8895
2.4902
5.93
2.7920
2.4085
2.4432
Scheme 2 a
C3C2C1N
C2C1NC4
C1NC4C5
Li-C1
-2.08
-1.64
-83.80
-120.25
-31.42
-16.28
2.53
11.08
2.3744
2.5211
2.5803
2.5595
3.7625
2.4360
Li-C3
3.6033
2.4987
2.7311
2.4183
2.3279
2.7642
Li-C6
Li-C7
Transmetalated Structures Coordinated
with Dimethylether Molecules
Anion I
+
Na+1DME
TS4
+
Na
+
1DME
TS6 + Na + 1DME
C5-C1
2.8142
3.6081
1.7994
-5.12
79.53
1.9995
3.6616
2.4685
5.23
2.7905
2.2261
2.4100
C5-C3
Cl-C5
C3C2C1N
C2C1NC4
C1NC4C5
Na-C1
Na-C3
Na-C6
Na-C7
Na-O1
O1-C6
O1-C8
-2.68
-116.69
-31.66
-17.73
2.47
10.87
2.7773
3.5396
3.0571
3.2443
2.2795
5.0397
4.4539
3.0057
2.6789
2.8753
2.9603
3.8839
2.2779
5.1811
4.3069
3.8638
2.7726
3.5234
2.2758
4.7951
5.0642
a All distances are reported in Å, angles and dihedrals in degrees. O1 is
the oxygen of the DME molecule.
namic findings cannot explain the exclusive reaction toward
four-membered rings. Most probably the transition state toward
six-membered ring formation is characterized by energetically
unfavorable effects, which inhibit the ring closure at the γ-site.
Transmetalation. Although the free gas-phase anions provide
useful information, the counterion may contribute to the factors
that affect the regioselectivity of heteroatom-stabilized allylic
anions. Therefore, the free anion with the presence of a sodium
counterion was optimized. Various stable structures were
obtained when starting from different initial positions of the
sodium cation. The most stable configuration is visualized in
Figure 1 (labeled as 2a+Na). The sodium counterion prefers a
position between the allylic subsystem and the benzyl group
a Reagents and conditions: (i) 2 equiv HPO(OMe)2, MeOH, ∆, 3 h; (ii)
1.5 equiv AcCl, 0.2 equiv DMAP, 2 equiv pyridine, THF, rt (1 h), ∆ (1 h);
(iii) 1) 1.1 equiv NaH, -78 °C, THF; 2) -78 °C f rt (2h); (iv) H2O, 2 h
rt; (v) 3 equiv MeI, 2 h, rt.
(cf. Table 1: Na-C1 ) 2.71 Å, Na-C6 ) 2.87 Å, Na-C7 )
2.98 Å). The sodium atom makes an additional contact with
the oxygen of the PdO bond (Na-O ) 2.27 Å). The lithiated
anions have similar geometrical characteristics (cf. Table 1: Li-
C1 ) 2.37 Å, Li-C6 ) 2.50 Å, Li-C7 ) 2.73 Å). The atomic
charges are not drastically altered by incorporating the coun-
terion (cf. Table 3). For further cyclization, it is also important
to analyze the effect of the counterion with respect to the
9
J. AM. CHEM. SOC. VOL. 128, NO. 26, 2006 8473

A R T I C L E S
Van Speybroeck et al.
Table 3. Reaction Barriers for Four- and Six-Membered Cyclization (∆Eq(4), ∆Eq(6), ∆Gq(4), ∆Gq(6)) Expressed in kJ/mol at Various Levels
of Theorya
gas phase results
level of theory
∆
Eq(4)
∆
Eq(6)
∆(
∆
Eq)
∆
Gq(4)
∆
Gq(6)
∆(∆
Gq)
B3LYP/6-311g**//B3LYP/6-311g**
mPW1PW91/6-31+g(d)//B3LYP/6-311g**
BP86/6-31+g(d)//B3LYP/6-311g**
B3LYP/6-31+g(d)//B3LYP/6-31+g(d)
B3LYP/6-311++g(d)//B3LYP/6-311++g(d,p)
mPW1PW91/6-31+g(d)//MPW1PW91/6-311g**
75.72
91.43
70.54
81.05
82.43
90.78
98.12
111.69
86.32
101.83
103.12
110.45
22.40
20.25
15.78
20.78
20.69
19.67
70.28
96.99
26.71
75.61
85.34
100.70
109.32
25.09
23.98
without inclusion
of bulk solvent effects
with inclusion
of bulk solvent effects
B3LYP/6-31+g(d)//B3LYP/6-31
+g(d)
∆
Eq(4)
∆
Eq(6)
∆(∆
Eq)
∆
Gq(4)
∆
Gq(6)
∆(∆
Gq)
transmetalated structures (Na)
transmetalated structures (Li)
transmetalated structures (Na) + 1DME molecule
transmetalated structures (Na) + 2DME molecule
157.14
167.66
149.69
145.06
160.54
177.65
140.75
140.39
3.40
9.99
-8.94
-4.66
95.31
102.17
88.00
121.45
147.31
103.31
85.62
26.14
45.13
15.31
26.84
58.78
a ∆(∆Eq) is the difference between the reaction barrier for six- and four-membered ring formation, i.e. ∆Eq(4) - ∆Eq(6). In the notation Level1//Level2,
Level1 is the electronic level of theory used for the energetics, whereas Level2 is the level of theory used for the geometry optimization.
solvation energy (CSE) is estimated by coordinating the sodium
ion to one ore more ether oxygens. Dimethyl ether (DME) was
chosen as the coordinating solvent instead of THF. DME has
about the same basicity as THF but is significantly smaller for
computations. An important question concerns the degree of
coordination. For solvent-separated ion pairs, four-coordinated
lithium cations have been recognized in NMR studies.52,53
However, for contact ion pairs as encountered here, the degree
of coordination may be dependent on the nature of the coun-
terion, and the coordination may be expected to be smaller due
its electrostatic effect.54 The degree of coordination for the
system under consideration is estimated by coordinating the
sodium cation with one, two, three, or four DME molecules
Figure 3. Protonation energies (∆Ex) for various anions (B3LYP/
6-31+g(d) level).
and calculating the CSE energies. These values are shown in
Figure 1. Coordination of the first two ether molecules is highly
exothermic (CSE amounts to -53.8 and -88.7 kJ/mol). As the
coordination with the second DME is still highly exothermic,
an additional DME molecule was placed close to the ion pair.
Coordination of the third DME molecule causes only a slight
additional stabilization (-98.9 kJ/mol versus -88.7 kJ/mol) and
is located at a large distance of 4.74 Å from to the sodium ion
(compared with Na-O distances of 2.17 and 2.18 Å for the
first two DME molecules). The CSE amounts to -102.7 kJ/
mol for four DME molecules, indicating that the coordination
effectiveness is largely reduced from the third DME molecule
on.
chemically active area of the molecule. The C5-C1 and C5-C3
distances are not considerably altered. This conclusion also holds
for the C2C1NC4 dihedral angle which determines the position
of the C5-C1 unit to the allylic system. This proves that the
counterion exerts limited influence on the structural parameters
of the system under consideration.
Effect of Solvation. The computations discussed thus far refer
to unsolvated compounds. However, THF was used as solvent
in our experiments, and solvation may be expected to be
important in these ether-like solvents. The solvation originates
from two contributions: coordination of ether oxygens to the
sodium cation (coordination solvation) and the electrostatic
effect of the solvent dielectric medium (dielectric solvation).
The latter effect is more easily to deal with from a computational
point of view as it is usually approximated by enclosing the
molecule in a cavity within a continuous dielectric medium.
Recently, an article by Kelly, Cramer, and Truhlar appeared
on the accurate determination of solvation free energies, which
may be interesting for a more detailed reading on this subject.50
The dielectric solvation energies (DSE) were obtained using a
CPCM (also known as COSMO) option in Gaussian 03 on the
unsolvated B3LYP/6-31+g(d) structures.51 The coordination
The DSE is obtained as the sum of two terms, an electrostatic
term, which has a negative sign and which is derived from the
interaction of the solute charges with the solvent, and a
nonelectrostatic term, that is generally positive and consisting
of the free energy needed for the formation of the cavity in the
continuum, a dispersion term and a repulsion term. The value
for the DSE on the monomeric ion pair formed by the anion
and the sodium ion amounts to 12.22 kJ/mol and is non-
negligible. These values are in line with the numbers reported
by Pratt and Streitwieser.21 The geometrical parameters with
(52) Reich, H. J.; Green, D. P.; Medina, M. A.; Goldenberg, W. S.; Gudmunds-
son, B. O.; Dykstra, R. R.; Philips, N. H. J. Am. Chem. Soc. 1998, 120,
7201-7210.
(53) Carlier, P. R.; Lo, C. W.-S. J. Am. Chem. Soc. 2000, 122, 12819-12823.
(54) Schade, C.; Schleyer, P. V. R. AdV. Organomet. Chem. 1988, 27, 169-
170.
(50) Kelly, C. P.; Cramer, C. J.; Truhlar, D. G. J. Chem. Theory Comput. 2005,
1, 1133-1152.
(51) (a) Cossi, M.; Rega, N.; Scalmani, G.; Barone, V. J. Comput. Chem. 2003,
24, 669-681. (b) Barone, V.; Cossi, M. J. Phys. Chem. A 1998, 102, 1995-
2001.
9
8474 J. AM. CHEM. SOC. VOL. 128, NO. 26, 2006

Cyclization of Functionalized Aminophosphonates
A R T I C L E S
Figure 4. Optimized conformers of the transition states in different molecular environments: TS4 and TS6 transition states for four- and six-membered
lactam formation in the gas phase; TS4+Na and TS6+Na including one sodium ion; TS4+Na+1DME and TS6+Na+1DME including one sodium ion and
one DME molecule. All optimizations are performed at the B3LYP/6-31+g(d) level.
inclusion of solvent molecules, confirm once more that the
chemical active area for cyclization remains nearly the same.
R-carbon atom, i.e. the reactive center for four-membered ring
formation. In the case of six-membered ring formation large
distortions are needed from the original geometry of the anion
to adapt the SN2-like transition structure (cf. Table 1): the
dihedral angle C2C1NC4 varies from -82.2 to -36.0°. To get
an idea about energetic variations associated with these distor-
tions, the rotational potential in terms of this geometrical variable
was calculated. This was done by a stepwise variation of the
C2C1NC4 torsional angle and optimizing all other degrees of
freedom. To reduce the computational cost, the B3LYP/
3-21+g(d) level of theory was adapted. The results are shown
Transition-State Structures. To unravel the experimentally
obtained reaction preference toward four-membered lactams,
calculations are performed to determine the reaction barriers.
The transition states toward four- and six-membered ring
formation in the gas phase are visualized in Figure 4, and some
relevant geometrical parameters are given in Table 1.
Both transition states resemble an intramolecular SN2-like
reaction characterized by an umbrella-like inversion at C5. In
the reacting anion the C5 carbon atom is oriented toward the
9
J. AM. CHEM. SOC. VOL. 128, NO. 26, 2006 8475

A R T I C L E S
Van Speybroeck et al.
and six-membered ring formation (i.e. ∆(∆Eq) ) ∆Eq(6) -
∆Eq(4), where ∆Eq(4) and ∆Eq(6) are the barriers for four- and
six-membered cyclization) show much less variation (16-22
kJ/mol, depending on the electronic level of theory used),
predicting a large preference for four-membered ring formation
at all levels of theory. The reason for this preference must be
traced back to the high amount of energy needed to reach a
conformation suitable for six-membered ring formation (cf.
Figure 5).
At second instance, the influence of the cation and solvent
effects on the reaction barriers is discussed. Inclusion of the
sodium cation has a dramatic influence on the reaction barriers.
Both barriers increase with approximately 60-80 kJ/mol at all
levels of theory. Similar findings were found by Ando.23
Moreover, there is not a clear preference anymore for four- or
six-membered ring formation (∆(∆Eq) ) 3.4 kJ/mol). However,
since the reactions were performed in THF, solvation of the
counterion is expected to be important. At first, bulk solvent
effects are taken into account using the CPCM model, lowering
the reaction barriers to about 95 and 121 kJ/mol (at the B3LYP/
6-31+g(d)//B3LYP/6-31+g(d) level) for four- and six-mem-
bered ring formation, respectively. Furthermore, the preference
for cyclization at the R-position is correctly predicted by
including bulk solvent effects (∆(∆Gq) ) 26.1 kJ/mol) and thus
emphasizing the necessity to treat solvent effects in a proper
way.
Figure 5. Part of the rotational potential in terms of the C2C1NC4 dihedral
angle. The precursors for four- and six-membered ring formation (Pr4 and
Pr6) are indicated by a gray circle.
in Figure 5 for the appropriate range of the dihedral angle.
Starting from the most stable favorable conformation (φ )
-82.9°) the torsional angle reaches values of -116 and -36°
in the four- or six-membered transition states, respectively (cf.
Table 1). Due to the strongly asymmetric shape of the rotational
potential around the minimum, the associated energy to induce
such distortions amounts to approximately 15 and 38 kJ/mol.
Therefore, a large amount of energy is needed to reach the
transition state toward the six-membered lactams. The structures
of the transition states with incorporation of the counterions
are also visualized in Figure 4. The chemically active part for
the ring-closure reaction is not altered substantially; only the
phenyl groups slightly relax in the presence of the sodium cation.
Furthermore, the distances of the forming C-C bond are shorter
(∼0.15 Å) at all levels of theory. This is in accordance with
the results of Ando.23
Reaction Barriers. In the further course of the discussion,
both the bare reaction barriers ∆Eq (i.e. the electronic energy
differences between transition state and corresponding reactant)
and the corresponding free energies (∆Gq) including zero-point,
thermal, and entropic contributions are evaluated. When PCM
computations are used, the energies have the status of free
energies, since they implicitly take into account the thermal and
entropic contributions of the solvent.55 The values are given in
Table 3 at various levels of theory.
First, the influence of the level of theory on the energetic
results is discussed. It is generally known that B3LYP geom-
etries are quite accurate19,42,56 and that variations on the reaction
barriers are small by adopting another level for the geometry
optimization. This is confirmed by our calculations: mPW1PW91/
6-31+g(d)//B3LYP/6-311 g** and mPW1PW91/6-31+g(d)//
mPW1PW91/6-31+g(d) predict barriers for four-membered
cyclization of 91.43 and 90.78 kJ/mol, respectively. The
functional form has a much larger influence on the reaction
barriers leading to respective values between 70 to 91 kJ/mol.
The selection of the most suitable level of theory is jeopardized
by the lack of direct experimental data on the reaction barriers.
However, the differences between the reaction barriers for four-
Finally explicit solvent molecules were taken into account
as explained previously. The barriers were calculated for one
and two DME molecules. Oncemore, inclusion of bulk solvent
effects is required to predict the correct reaction preference. The
difference between the activation free energy for four- and six-
membered ring formation (∆(∆Gq)) amounts to 15.31 and 26.84
kJ/mol in the case of adding one or two DME molecules,
respectively. The value of ∆(∆Gq) may be expected to alter
only slightly with inclusion of more DME molecules, consis-
tently with the result found for the CSE values. All theoretical
calculations are now consistent, explaining the four-membered
ring preference. The values of ∆(∆Gq) amount to 25.09, 26.14,
and 26.84 kJ/mol at the B3LYP/6-31+g(d) level for the gas-
phase computations, the calculations with inclusion of the
sodium counterion and the inclusion of the sodium and two
DME molecules, respectively.
Summarizing, for this particular reaction a correct qualitative
picture is predicted using the gas-phase computations. Solvent
effects do alter the absolute values of the reaction barriers but
do not affect the global R/γ selectivity. This might be anticipated
as the geometries of the chemical active part of the system are
not significantly altered by inclusion of the molecular environ-
ment. However, based on this single application, these conclu-
sions may not be generalized.
Origin of the Four-Membered Ring Preference. Finally,
suggestions are validated to unravel the true origin of the four-
membered ring formation preference. As already pointed out,
the rotation about the C-N bond is largely hindered. Moreover
the associated potential is asymmetric, i.e a rotation toward the
γ-carbon atom is disfavored (cfr. Figure 5). To investigate more
in depth these energetically unfavorable effects, the reaction
barriers for a number of allyl-stabilized anions differing in the
substituents X, Y, and Z attached at various positions of the
anion (cf. Figure 6) were calculated.
(55) Tomasi, J.; Persico, M. Chem. ReV. 1994, 94, 2027-2094.
(56) Gomez-Balderas, R.; Coote, M. L.; Henry, D. J.; Radom, L. J. Phys. Chem.
A 2004, 108, 2874-2883.
9
8476 J. AM. CHEM. SOC. VOL. 128, NO. 26, 2006

Cyclization of Functionalized Aminophosphonates
A R T I C L E S
Figure 7. Iso-surfaces (value 0.005) of the Fukui function (f-(r)) for
electrophilic attack in the most stable conformer of the anion, which is the
precursor for four-membered ring formation.
Figure 6. Difference between four- and six-membered activation free
energies for a variety of allyl stabilized anion. The calculations are performed
at the B3LYP/6-31+g(d)//B3LYP/6-31+g(d) level of theory. For Anions
I, Ic, Id and Ie also two DME molecules and bulk solvent effects were
included.
softness values calculated using the natural population analysis
(NPA) and CHELPG scheme are given in Table S1 of the
Supporting Information. The R-carbon atom is the softest center,
whereas the γ-carbon atom is somewhat harder; thus, the
condensed local indices indeed predict that cyclization prefer-
entially occurs at the R-position. This can further be illustrated
by visualizing hf(r) of the reactive anion (Figure 7). The largest
value for the Fukui function is indeed found at the R-carbon
atom. The specific reaction is an example of a reaction which
is frontier-orbital controlled, and the Fukui function is an
appropriate indicator to describe the site selectivity.58,59 The
specific selectivity encountered here has been seen in various
other systems such as HCHO, NCS, and malonaldehyde anion.60
In the anion trapping experiment with a proton, which is
known to be very hard, the reaction may be expected to occur
preferentially at the hardest center, which is the γ-carbon atom
in our case. However, since the alkylation with methyl iodide
which is a soft electrophile, also occurs at the γ-position, it can
be concluded that the site-selectivity is primarily dictated by
geometric constraints, whereas the frontier-orbital interactions
are of minor importance.
When substituting the bulky phosphonate group by a tertiary
butyl group which is approximately as voluminous, the four-
membered ring is still preferred by 26.13 kJ/mol (Anion Id).
This indicates that the true nature of the four-membered ring
formation preference is not related to the specific properties of
the phosphonate group. Nevertheless, the phosphonate group
plays a crucial role in the anion stability, allowing selective
deprotonation of the substrates. Also Bossio et al. and Marcac-
cini et al. found preference for four-membered ring formation
in the case where the anion was stabilized by a carbonyl
group.9,10 Even when substantially reducing the size of the X
and Y substituent, four-membered rings are always preferentially
formed. This indicates that the size of the X and Y group are
not the primary cause for the largely hindered rotation about
the C-N bond. The anion can exist in at least two conformers
which are schematically shown in Figure 6 and which differ in
the relative orientation of the groups 1 and 2 which are attached
to the nitrogen-carbon bond. Our calculations point out that
Conformer a, which is ready for six-membered ring formation,
is highly unstable due to the large amount of energy needed to
bring the groups 1 and 2 closer toward each other.
Furthermore, it was investigated whether the site selectivity
seen in the reactive anion under study can further be rationalized
by inspecting appropriate DFT-based reactivity indicators.
Therefore, a variety of global and local indices, such as the
global hardness (η), local softness (s(r)), and Fukui function
(f(r)), were applied to the most stable conformer. For more
explanation on their definitions we refer to the textbook of Parr
and Yang39 and a review by Geerlings et al.57 The reactions
studied here concern an ambident nucleophile, where the soft
and hard centers are located in the same system. Moreover, the
reactions are intramolecular SN2 cyclization reactions, and the
Fukui function for electrophilic attack (hf(r)) is believed to
indicate the most favorable site. The condensed Fukui and
Summarizing, the four-membered ring preference is due to a
largely hindered internal rotation around the C-N bond of the
1-aminoalkenyl-phosphonate, which prevents the anion in
reaching the conformer suitable for six-membered ring forma-
tion.
5. Conclusions
In this study a peculiar ring closure of functionalized
aminophosphonates toward four-membered phosphono-â-lac-
tams has been unraveled both on experimental and theoretical
bases. Starting from an ambident allylic anion, no trace of six-
membered lactams was found, and only the highly strained four-
membered rings are formed through intramolecular alkylation.
Nevertheless, new anion trapping experiments indicate that the
(58) Chattaraj, P. K. J. Phys. Chem. A 2001, 105, 511-513.
(59) Fukui, F. Theory of Orientation and Stereoselection; Springer-Verlag:
Berlin, 1973; p 134; Science (Washington, D.C.) 1982, 218, 747.
(60) Pearson, R. G. Chemical Hardness: Applications from Molecules to Solids;
Wiley-VCH Verlag GMBH: Weinheim, 1997.
(57) Geerlings, P.; De Proft, F.; Langenaeker, W. Chem. ReV. 2003, 103, 1793.
9
J. AM. CHEM. SOC. VOL. 128, NO. 26, 2006 8477

A R T I C L E S
Van Speybroeck et al.
γ-position is more reactive in intermolecular reactions. This was
confirmed by the calculation of theoretical protonation energies,
which ascribe the stabilization of this isomer to the presence of
the phosphonate- and nitrogen-containing subgroup. However,
it was found that the remarkable selectivity is typically
associated with intramolecular reactions and was primarily due
to a restricted rotation around the C-N bond of the 1-amino-
alkenyl-phosphonate studied. Therefore, the transition state for
six-membered ring formation is energetically disfavored. The
typical SN2-like transition state is geometrically strained for six-
membered ring formation and consequently substantially (ap-
proximately 25 kJ/mol) more activated. The system of interest
is challenging as it contains solvated sodium or lithium ions.
However, the R/γ-selectivity is not notably influenced by
accounting for solvent effects.
Acknowledgment. This work is supported by the Fund for
Scientific Research-Flanders (FWO) and the Research Council
of Ghent University.
Supporting Information Available: Complete ref 38, reactiv-
ity indices for the reactive anion, structural characterization of
all compounds, Cartesian coordinates of all optimized reactants,
transition structures. This material is available free of charge
via the Internet at http://pubs.acs.org.
JA0584119
9
8478 J. AM. CHEM. SOC. VOL. 128, NO. 26, 2006