A combined theoretical and experimental study of efficient and fast titanocene-catalyzed 3-exo cyclizations

Article abstract of DOI:10.1021/ja050268w

The mechanism of titanocene mediated 3-exo cyclizations was investigated by a combined theoretical and experimental study. A gradient corrected density functional theory (DFT) method has been scaled against titanocene dichloride, the parent butenyl radica

Full text of DOI:10.1021/ja050268w

Published on Web 04/20/2005
A Combined Theoretical and Experimental Study of Efficient
and Fast Titanocene-Catalyzed 3-exo Cyclizations
Joachim Friedrich,*,† Michael Dolg,† Andreas Gansa¨uer,*,‡
Daniel Geich-Gimbel,‡ and Thorsten Lauterbach‡
Contribution from the Institut fu¨r Theoretische Chemie der UniVersita¨t zu Ko¨ln, Greinstrasse 4,
50939 Ko¨ln, Germany, and Kekule´-Institut fu¨r Organische Chemie und Biochemie der
UniVersita¨t Bonn, Gerhard Domagk Strasse 1, 53121 Bonn, Germany
Received January 14, 2005; E-mail: joachim_friedrich@gmx.de; andreas.gansaeuer@uni-bonn.de
Abstract: The mechanism of titanocene mediated 3-exo cyclizations was investigated by a combined
theoretical and experimental study. A gradient corrected density functional theory (DFT) method has been
scaled against titanocene dichloride, the parent butenyl radical, and in bond dissociation energy (BDE)
calculations. The BP86 method using density fitting, and a basis set of triple-ú quality emerged as a highly
reliable tool for studying titanocene mediated radical reactions. The computational results revealed important
kinetic and thermodynamic features of cyclopropane formation. Surprisingly, the â-titanoxy radicals, the
first intermediates of our investigations, were demonstrated to possess essentially the same thermodynamic
stabilization as the corresponding alkyl radicals by comparison of the calculated BDEs. In contrast to
suggestions for samarium mediated reactions, the cyclization was shown to be thermodynamically favorable
in agreement with earlier kinetic studies. It was established that stereoselectivity of the cyclization is governed
by the stability of the intermediates and thus the trans disubstituted products are formed preferentially.
The observed ratios of products are in good to excellent agreement with the DFT results. By a combination
of computational and experimental results, it was also shown that for the completion of the overall
cyclopropane formation the efficiency of the trapping of the cyclopropylcarbinyl radicals is decisive.
Introduction
Over the past decades, radical chemistry has been increasingly
used in the synthesis of complex molecules.1 Limitations also
exist, however. One of the relatively unexplored but potentially
very useful fields is constituted by the synthesis of small rings.2
Difficulties encountered in the preparation of three- and four-
membered rings are due to the strain of these compounds. Thus,
ring opening proceeds usually much faster than ring closure.
Strategies in classical free radical chemistry circumventing this
problem have emerged recently. They include incorporation of
some of the products’ strain into the substrates, fast and
irreversible fragmentation reactions following the initial cy-
clization, and stabilization of the cyclopropylcarbinyl radical
through conjugation. However, selective trapping to yield the
cyclopropane remains difficult and in sterically unbiased cases
Figure 1. Typical kinetics of the 3-exo cyclization.
reactions in the presence of Bu3SnH predominantly yielded the
ring opened products.
The kinetics and thermodynamics of the 3-exo cyclization,
on the other hand, have been studied thoroughly.3 As a
consequence, the rate constants and activation energies have
been determined with high accuracy. The ring opening of the
parent system shown in Figure 1 is among the fastest radical
reactions known (k ≈ 108 s-1). Nevertheless, the 3-exo cycliza-
tion itself is also a relatively fast reaction (k ≈ 104 s-1) that is
kinetically attractive if the reverse reaction can be retarded or
the cyclopropyl carbinyl radical trapped selectively.3a,c
† Institut fu¨r Theoretische Chemie der Universita¨t zu Ko¨ln.
‡ Kekule´-Institut fu¨r Organische Chemie und Biochemie der Universita¨t
Bonn.
Because of the accurate and reliable kinetic studies, the
depicted equilibrium has also attracted considerable interest in
(1) For leading references see: (a) Radicals in Organic Synthesis; Renaud, P.,
Sibi, M. P., Eds.; Wiley-VCH: Weinheim, 2001. (b) McCarroll, A. J.;
Walton, J. C. Angew. Chem. 2001, 113, 2282-2307; Angew. Chem., Int.
Ed. 2001, 40, 2225-2250. (c) Curran, D. P.; Porter, N. A.; Giese, B.
Stereochemistry of Radical Reactions; VCH: Weinheim, 1996. (d) Fossey,
J.; Lefort, D.; Sorba, J. Free Radicals In Organic Synthesis; Wiley:
Chichester, 1995.
(2) For a recent review, see: Srikrishna, A. In Radicals in Organic Synthesis;
Renaud, P., Sibi, M. P., Eds.; Wiley-VCH: Weinheim, 2001; Vol. 2, p
151. For selected examples, see: (a) Srikrishna, A.; Sharma, G. V. R.;
Hemamalini, P. Chem. Commun. 1990, 1681-1683. (b) Malacria, M. Chem.
ReV. 1996, 96, 289-306. (c) Journet, M.; Malacria, M. J. Org. Chem. 1994,
59, 718-719. (d) Devin, P.; Fensterbank, L.; Malacria, M. J. Org. Chem.
1998, 63, 6764-6765.
(3) Rate constants of the cyclopropylcarbinyl radical ring opening: (a) Bowry,
V. W.; Lusztyk, J.; Ingold, K. U. Chem. Commun. 1990, 923-925. (b)
Bowry, V. W.; Lusztyk, J.; Ingold, K. U. J. Am. Chem. Soc. 1991, 113,
5687-5698. (c) Beckwith, A. L. J.; Bowry, V. W. J. Am. Chem. Soc. 1994,
116, 2710-2716. (d) Newcomb, M.; Horner, J. H.; Emanuel, C. J. J. Am.
Chem. Soc. 1997, 119, 7147-7148. (e) Newcomb, M.; Choi, S. Y.; Horner,
J. H. J. Org. Chem. 1999, 64, 1225-1231. Rate constant of ring closure:
(f) Newcomb, M.; Glenn, A. G.; Williams, W. G. J. Org. Chem. 1989, 54,
2675-2681. (g) Effio, A.; Griller, D.; Ingold, K. U.; Beckwith, A. L. J.;
Serelis, A. K. J. Am. Chem. Soc. 1980, 102, 1734-1736. Equilibrium
constants: (h) Halgren, T. H.; Roberts, J. D.; Horner, J. H.; Martinez, F.
N.; Tronche, C.; Newcomb, M. J. Am. Chem. Soc. 2000, 122, 2988-2994.
9
10.1021/ja050268w CCC: $30.25 © 2005 American Chemical Society
J. AM. CHEM. SOC. 2005, 127, 7071-7077
7071

A R T I C L E S
Friedrich et al.
the field of theoretical chemistry.4 A recent study by Radom
assessed the quality of high level computational methods for
the determination of activation energies and reaction enthalpies
of the parent butenyl radical.4b More complex systems were not
investigated, and thus issues of stereoselectivity cannot be
addressed.
In the field of metal mediated radical chemistry, only very
few examples of 3-exo cyclizations have been observed.5 As
early as 1966, Barton described a single example of a chro-
mium(II) mediated radical addition to an enone in a steroid
derivative.5a Another isolated report by Lange concerned a low
yielding samarium diiodide mediated fragmentation cyclization
sequence involving an acrylate.5b Similar but general and high
yielding methodology was explored by Guibe´ in the same year
in samarium diiodide mediated cyclizations of δ-halo-R,â-
unsaturated esters.5c A radical mechanism was postulated, and
surprisingly it was proposed that opening of the intermediate
ester substituted cyclopropylcarbinyl radical was thermodynami-
cally favored. A 3-exo cyclization of similar iodides mediated
by zinc dust has been described very recently.5d
of this reaction has been reported independently by Ferna´ndez-
Mateos.8j
Despite their preparative usefulness, the overall mechanistic
picture of these reactions, including our own, has remained
unclear, and the factors influencing chemo- and stereoselectivity
are still unknown. The essential issue of catalyst control of
selectivity could therefore not be rationally addressed so far.
Of course, computational chemistry has become a unique and
highly useful tool for studying reaction mechanisms with the
emergence of powerful computers and sophisticated ab initio
methods. However, high level wave function-based ab initio
methods are still restricted to relatively small molecules. The
titanocene containing intermediates and transition structures are
too complex to be studied by these procedures. We therefore
decided to use density functional theory (DFT) first principle
methods in this study. To establish their accuracy for the
problems discussed here, they were scaled against the pertinent
cases of titanocene dichloride, the parent butenyl radical
cyclization, and the homolytic bond dissociation energies of the
intermediates studied here. These calibration studies revealed
that the DFT methods are highly reliable tools for our purposes.
The computational studies delivered the structures and
energies of all relevant transition structures and intermediates
composing a complete mechanistic picture. In particular, it was
established in contradiction to the proposal by Guibe´5c and in
agreement with a study by Beckwith3c that the cyclization is
thermodynamically favored. In combination with carefully
designed experiments, it was further demonstrated that the
equilibration of the cyclopropylcarbinyl radical is faster than
its reduction by a second equivalent of Cp2TiCl in the carbonyl
substituted cases. Thus, the overall course of the reaction is
determined by a unique combination of thermodynamic and
kinetic factors.
We have reported the first 3-exo cyclization6 catalytic7 in
titanocene(III) reagents based on the stoichiometric reductive
epoxide opening introduced by Nugent and RajanBabu.8 The
efficiency of the 3-exo cyclization is thought to be based on
selective reductive trapping of cyclopropylcarbinyl radicals.
Most recently, a single example of the stoichiometric version
(4) For excellent theoretical treatments of the opening of the cyclopropylcarbinyl
radical, see: (a) Martinez, F. N.; Schlegel, H. B.; Newcomb, M. J. Org.
Chem. 1996, 61, 8547-8550. (b) Smith, D. M.; Nicolaides, A.; Golding,
B. T.; Radom L. J. Am. Chem. Soc. 1998, 120, 10223-10333. (c) Coosky,
A. L.; King, H. F.; Richardson, W. H. J. Org. Chem. 2003, 68, 9441-
9452.
(5) (a) Barton, D. H. R.; Basu, N. K.; Hesse, R. H.; Morehouse, F. H.; Pechet,
M. J. M. J. Am. Chem. Soc. 1966, 88, 3016-302. (b) Lange, G. L.; Merica,
A. Tetrahedron Lett. 1999, 40, 7897-7900. (c) David, H.; Afonso, C.;
Bonin, M.; Doisneau, G.; Guillerez, M.-G.; Guibe´, F. Tetrahedron Lett.
1999, 40, 8557-8561. (d) Sakuma, D.; Togo, H. Synlett 2004, 2501-2504.
(6) Gansa¨uer, A.; Lauterbach, T.; Geich-Gimbel, D. Chem.-Eur. J. 2004, 10,
4983-4990.
Results and Discussion
(7) For reviews, see: (a) Gansa¨uer, A.; Narayan, S. AdV. Synth. Catal. 2002,
344, 465-475. (b) Gansa¨uer, A.; Lauterbach, T.; Narayan, S. Angew. Chem.
2003, 115, 5714-5731; Angew. Chem., Int. Ed. 2003, 42, 5556-5573.
For pertinent original catalytic contributions, see: (c) Gansa¨uer, A.;
Pierobon, M.; Bluhm, H. Angew. Chem. 1998, 110, 107-109; Angew.
Chem., Int. Ed. 1998, 37, 101-103. (d) Gansa¨uer, A.; Bluhm, H.; Pierobon,
M. J. Am. Chem. Soc. 1998, 120, 12849-12859. (e) Gansa¨uer, A.;
Lauterbach, T.; Bluhm, H.; Noltemeyer, M. Angew. Chem. 1999, 111,
3112-3114; Angew. Chem., Int. Ed. 1999, 38, 2909-2910. (f) Gansa¨uer,
A.; Pierobon, M.; Bluhm, H. Synthesis 2001, 2500-2520. (g) Gansa¨uer,
A.; Pierobon, M.; Bluhm, H. Angew. Chem. 2002, 114, 3341-3343; Angew.
Chem., Int. Ed. 2002, 41, 3206-3208. (h) Gansa¨uer, A.; Bluhm, H.; Rinker,
B.; Narayan, S.; Schick, M.; Lauterbach, T.; Pierobon, M. Chem.-Eur. J.
2003, 9, 531-542. (i) Barrero, A. F.; Rosales, A.; Cuerva, J. M.; Oltra, J.
E. Org. Lett. 2003, 5, 1935-1938. (j) Gansa¨uer, A.; Rinker, B.; Pierobon,
M.; Grimme, S.; Gerenkamp, M.; Mu¨ck-Lichtenfeld, C. Angew. Chem.
2003, 115, 3815-3818; Angew. Chem., Int. Ed. 2003, 42, 3687-3690. (k)
Justicia, J.; Rosales, A.; Bun˜uel, E.; Oller-Lo´pez, J. L.; Valdivia, M.;
Ha¨ıdour, A.; Oltra, J. E.; Barrero, A. F.; Ca´rdenas, D. J.; Cuerva, J. M.
Chem.-Eur. J. 2004, 10, 1778-1788. (l) Gansa¨uer, A.; Rinker, B.; Ndene-
Schiffer, N.; Pierobon, M.; Grimme, S.; Gerenkamp, M.; Mu¨ck-Lichtenfeld,
C. Eur. J. Org. Chem. 2004, 2337-2351. (m) Justicia, J.; Oltra, J. E.;
Cuerva, J. M. J. Org. Chem. 2004, 69, 5803-5806.
(8) For pertinent stoichiometric contributions, see: (a) Nugent, W. A.;
RajanBabu, T. V. J. Am. Chem. Soc. 1988, 110, 8561-8562. (b) RajanBabu,
T. V.; Nugent, W. A. J. Am. Chem. Soc. 1989, 111, 4525-4527. (c)
RajanBabu, T. V.; Nugent, W. A.; Beattie, M. S. J. Am. Chem. Soc. 1990,
112, 6408-6409. (d) RajanBabu, T. V.; Nugent, W. A. J. Am. Chem. Soc.
1994, 116, 986-997. (e) Ferna´ndez-Mateos, A.; Martin de la Nava, E.;
Pascual Coca, G.; Ramos Silva, A.; Rubio Gonza´lez, R. Org. Lett. 1999,
1, 607-609. (f) Hardouin, C.; Doris, E.; Rousseau, B.; Mioskowski, C.
Org. Lett. 2002, 4, 1151-1153. (g) Barrero, A. F.; Oltra, J. E.; Cuerva, J.
M.; Rosales, A. J. Org. Chem. 2002, 67, 2566-2571. (h) Anaya, J.;
Ferna´ndez-Mateos, A.; Grande, M.; Martia´n˜ez, J.; Ruano, G.; Rubio
Gonza´lez, R. Tetrahedron 2003, 59, 241-248. (i) Ferna´ndez-Mateos, A.;
Mateos Buro´n, L.; Rabanedo Clemente, R.; Ramos Silva, A. I.; Rubio
Gonza´lez, R. Synlett 2004, 1011-1014. (j) Ferna´ndez-Mateos, A.; Mateos
Buro´n, L.; Mart´ın de la Nava, E. M.; Rabanedo Clemente, R.; Rubio
Gonza´lez, R.; Sanz Gonza´lez, F. Synlett 2004, 2553-2557.
Computational Studies. The geometry optimizations were
carried out within the framework of density functional theory
(DFT) with the BP86/TZVP method (Becke-Perdew gradient
corrected exchange and correlation density functional9 com-
bined with a polarized split-valence basis set of triple-ú quality)
using the RI-approximation (resolution of identity) within the
TURBOMOLE10 program package (if not otherwise noted). For
the DFT energy calculations of the optimized structures, the
more extended TZVPP basis set was applied. The ab initio cal-
culations of this work were performed with the GAUSSIAN,11
MOLPRO,12 and TURBOMOLE10 program packages at the RI-
BP86/TZVP optimized structures (if not otherwise noted). The
zero point energy (ZPE) was obtained from numerical force
constants based on RI-BP86/TZVP-calculations (if not otherwise
noted). The error introduced by this approximation was checked
and in the case of TiCp2Cl2 found to be smaller than 0.3 kcal/
mol, which is probably negligible as compared to the density
functional and basis set errors. The accuracy of the single point
approximation was checked for the acid system of Tables 4 and
(9) (a) Becke, A. D. Phys. ReV. A 1988, 38, 3098-3100. (b) Perdew, J. P.
Phys. ReV. B 1986, 33, 8822-8824.
(10) Ahlrichs, R.; Ba¨r, M.; Baron, H.-P.; Bauernschmitt, R.; Bo¨cker, S.; Ehrig,
M.; Eichkorn, K.; Elliott, S.; Furche, F.; Haase, F.; Ha¨ser, M.; Horn, H.;
Huber, C.; Huniar, U.; Ko¨lmel, C.; Kollwitz, M.; Ochsenfeld, C.; O¨ hm,
H.; Scha¨fer, A.; Schneider, U.; Treutler, O.; von Arnim, M.; Weigand, F.;
Weis, P.; Weiss, H. Turbomole 5; Institut fu¨r Physikal. Chemie, Universita¨t
Karlsruhe, 2002.
9
7072 J. AM. CHEM. SOC. VOL. 127, NO. 19, 2005

Titanocene-Catalyzed 3-exo Cyclizations
A R T I C L E S
Scheme 1. 3-exo Cyclization of the Butenyl Radical 2
Table 1. Stability of â-Titanoxy Radicals and Alkyl Radicals (BDE
at 298.15 K in kcal/mol)
system/method
exp.1a
exp.14
B3LYP15
RI-BP86a,b
H-C2H5
98
95
92
101.1
97.8
96.6
100.0
99.4
95.3
92.2
106.0
96.6
92.2
(CH3)2CH-H
(CH3)3C-H
H-cyclopropyl
Cp2ClTiOCH2CHCH3-H
Cp2ClTiOCH2C(CH3)2-H
106
106.3
a This work, RI-BP86/TZVPP//RI-BP86/TZVP including the ZPE ob-
tained from numerical RI-BP86/TZVP frequencies, the hydrogen energies
were obtained from Gaussian 03 calculations (see Table 1). b Includes a
thermal correction to the BDE of 1.5 kcal/mol (2.5RT).
Table 2. Geometry Parameters of the 3-exo Cyclization of 2
Calculated with Different Methods (Numbering from Scheme 1;
Bond Lengths in angstroms, Angles in deg)
5 to be 0.03-0.08 kcal/mol. Therefore, the error introduced by
the approximation is negligible, and the basis set error is reduced
to a minimum.
system
method
C1
−
C2
C2-C3
C3
−
C4
C1
−
C3
C1C2C3C4
1
QCISD/6-31G*
B3-LYP/6-31G*
1.484 1.494 1.388 1.915
1.482 1.496 1.385 1.916
Thermodynamic Considerations of Radical Stability:
Calculation of Bond Dissociation Energies (BDE). To un-
derstand the effects of the Cp2TiCl group on the thermodynamic
stability of radicals, we calculated the BDEs of the correspond-
ing titanocene alkoxides. This is of essential importance for
understanding the reactivity of the â-titanoxy radicals because
in reversible reactions the equilibria and in irreversible reactions
the activation energies will be affected. To guarantee a correct
quantitative description, we decided to calibrate the DFT
methods against the experimentally determined values for typical
hydrocarbons, for example, CH3-H, (CH3)2CH-H, and
(CH3)3C-H.
The Accuracy of the RI-BP86 Method for the Calculation
of BDEs. To check the accuracy of our DFT-calculations, we
compared the DFT-BDE of methane with experimental and high
level ab initio values. The RI-BP86/TZVPP//RI-BP86/TZVP
BDE of 104.8 kcal/mol agrees very well with the experimental
BDE of 1051a kcal/mol and the CCSD(T)/cc-pVQZ//QCISD/
cc-pVTZ value of 104.0 kcal/mol (coupled cluster singles and
doubles with perturbative triples correction energy using Dun-
nings13 correlation consistent basis set of polarized quadruple-ú
quality at the quadratic configuration interaction singles and
doubles optimized geometry with Dunnings13 correlation con-
sistent basis set of polarized triple-ú quality). More details of
these benchmark calculations are given in the Supporting
Information.
QCISD/cc-pVDZ 1.493 1.502 1.398 1.917
RI-BP86/TZVPa
QCISD/6-31G*
B3-LYP/6-31G*
1.484 1.498 1.385 1.920
1.500 1.510 1.336
1.497 1.510 1.333 2.509
2
3
-119.0b
119.4b
QCISD/cc-pVDZ 1.508 1.520 1.346
116.7b
RI-BP86/TZVPa
QCISD/6-31G*
B3-LYP/6-31G*
1.495 1.517 1.337 2.485 -114.5b
1.498 1.524 1.466 1.524
1.496 1.534 1.457 1.534
QCISD/cc-pVDZ 1.509 1.534 1.473 1.534
RI-BP86/TZVPa
1.496 1.538 1.452 1.542
a Results of this work. b The variation of the angles can be explained by
the symmetry. Two symmetry equivalent rotamers of the molecule exist.
Additionally, the dihedral potential is very flat. The RI-BP86/TZVP energy
varies by less than 0.1 kcal/mol for the C1C2C3C4-angle in the range from
110° to 120°.
radicals is shown on the basis of their BDEs. The RI-DFT values
are in excellent agreement with the known experimental data.
Surprisingly, the inductive effect of the alkoxide is negligible
thermodynamically. The computational results clearly demon-
strate that no significant stabilization or destabilization of the
radical center in â-titanoxy radicals is exerted by the ClCp2TiO-
group. Thus, as compared to alkyl radicals, no electronic effects
have to be considered when employing the â-titanoxy radicals
in retrosynthetic planning or in synthetic applications. Concern-
ing the control of stereoselectivity of their reactions, the
â-titanoxy radicals offer excellent prospects for a reagent
controlled course due to the high steric bulk of the titanocene
moiety. This feature will be relevant to both transition and
product structures and must be carefully incorporated into any
planning.
The Accuracy of the DFT Calculations for the 3-exo
Cyclization. With these encouraging results in hand, we next
checked the accuracy of DFT calculations for the 3-exo cycli-
zation of the parent but-3-enyl radical 2 shown in Scheme 1.
The cyclization constitutes an ideal reaction for scaling of
DFT-methods for three reasons. First, the system has been
intensely studied experimentally. Second, it is small enough to
be calculated with high accuracy even by the most advanced
ab initio methods. Therefore, a number of computational results
are available for comparison and identification of errors. Finally
and most importantly, the cyclization proceeds through inter-
mediates that, as already established by our BDE study, are
thermodynamically closely related to those of our titanocene
mediated reaction.
Calculation of BDEs for â-Titanoxy Radicals. In Table 1,
the relative stability of four alkyl radicals and two â-titanoxy
(11) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M.
A.; Cheeseman, J. R.; Montgomery, J. A., Jr.; Vreven, T.; Kudin, K. N.;
Burant, J. C.; Millam, J. M.; Iyengar, S. S.; Tomasi, J.; Barone, V.;
Mennucci, B.; Cossi, M.; Scalmani, G.; Rega, N.; Petersson, G. A.;
Nakatsuji, H.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa,
J.; Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Klene, M.;
Li, X.; Knox, J. E.; Hratchian, H. P.; Cross, J. B.; Adamo, C.; Jaramillo,
J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.; Austin, A. J.; Cammi, R.;
Pomelli, C.; Ochterski, J. W.; Ayala, P. Y.; Morokuma, K.; Voth, G. A.;
Salvador, P.; Dannenberg, J. J.; Zakrzewski, V. G.; Dapprich, S.; Daniels,
A. D.; Strain, M. C.; Farkas, O.; Malick, D. K.; Rabuck, A. D.;
Raghavachari, K.; Foresman, J. B.; Ortiz, J. V.; Cui, Q.; Baboul, A. G.;
Clifford, S.; Cioslowski, J.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz,
P.; Komaromi, I.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M.-A.;
Peng, C. Y.; Nanayakkara, A.; Challacombe, M.; Gill, P. M. W.; Johnson,
B.; Chen, W.; Wong, M. W.; Gonzalez, C.; Pople, J. A. Gaussian 03;
Gaussian, Inc.: Pittsburgh, PA, 2003.
(12) Werner, H. J.; Knowles, P. J.; Schu¨tz, M.; Lindh, R.; Celani, P.; Korona,
T.; Rauhut, G.; Manby, F. R.; Amos, R. D.; Bernhardsson, A.; Berning,
A.; Cooper, D. L.; Deegan, M. J. O.; Dobbyn, A. J.; Eckert, F.; Hampel,
C.; Hetzer, G.; Lloyd, A. W.; McNicholas, S. J.; Meyer, W.; Mura, M. E.;
Nicklaâ, A.; Palmieri, P.; Pitzer, R.; Schumann, U.; Stoll, H.; Stone, A. J.;
Tarroni, R.; Thorsteinsson, T. Molpro 2002.1; Institut fu¨r Theoretische
Chemie, Universita¨t Stuttgart, 2002.
In Table 2, the accuracy of selected geometry parameters as
compared to results from QCISD ab initio calculations with cc-
(13) Dunning, T. H., Jr. J. Chem. Phys. 1989, 90, 1007-1023.
9
J. AM. CHEM. SOC. VOL. 127, NO. 19, 2005 7073

A R T I C L E S
Friedrich et al.
Table 3. Selected Reaction and Activation Energies for the Ring
Opening Reaction of the Cyclopropylcarbinyl Radical from Ref 4
and the Present Work (Reaction Energies and Activation Barriers
in kcal/mol at 0 K without ZPE; the Experimental Values Are
Corrected to 0 K, and the ZPE Was Subtracted by Smith;4b the
Calculations with RI-Approximation Were Performed with
TURBOMOLE 5.610
)
method
geometry
energy
∆
E#
∆E
B3LYP/6-31G*4b B3LYP/6-31G*
8.4 -2.3
B3LYP/6-31G*4b B3LYP/6-311+G(3df,2p) 7.3 -3.4
B3LYP/6-31G*4b UMP2/6-311+G**
QCISD/cc-pVDZ4c QCISD/cc-pVDZ
QCISD/cc-pVDZ4c CCSD(T)cc-pVTZ
15.9 -0.1
10.4 -3.5
9.7a -3.6a
6.7 -1.4
6.9 -0.7
16.5 +1.3
12.4 +0.9
10.3 -3.5
12.0 -2.8
12.0 -2.7
9.6 -2.7
RI-BP86/TZVP
RI-BP86/TZVP
RI-BP86/TZVP
RI-BP86/TZVP
RI-BP86/TZVP
RI-BP86/TZVP
RI-BP86/TZVP
RI-BP86/TZVP
exp.4b
RI-BP86/TZVP
RI-BP86/TZVPP
RI-MP2/TZVPP
RI-CC2/TZVPP
QCISD/cc-pVDZb
CCSD/cc-pVTZc
CCSD/cc-pVQZc
CCSD(T)/cc-pVTZc
7.46 -1.7, -1.9, -2.8,
-3.7,d -4.6
a The most accurate ab initio calculation available so far. b Performed
with Gaussian 03.11 c Performed with MOLPRO 2002.1.12 d Temperature
and ZPE corrected experimental value of Halgren.3h For reasons of
consistency, we used the B3LYP-based corrections of Smith.4b
pVDZ4c and the 6-31G*4a basis sets of polarized valence
double-ú quality are shown. Additionally results of the standard
DFT B3LYP/6-31G*4b approach are included for comparison.
The RI-BP86/TZVP approach is found to generate structural
parameters in excellent agreement with the computationally
lavish wave function-based methods. Only in the case of the
C
1-C3-bond are the deviations larger than 0.02 Å. The reason
Figure 2. Reactions investigated in this computational study.
for this exception can be found in the extremely flat potential
between the two atoms in the course of the 3-exo cyclization
(vide infra).
activation energy. Considering the uncertainties associated with
the experimental reaction energies, the DFT approximation must
be considered as very satisfactory for our purposes.
Table 3 lists a summary of the theoretical results concerning
the activation and reaction energies from refs 4 and of the DFT
and ab initio calculations of the present work. The most reliable
theoretical values obtained so far by the CCSD(T)/cc-pVTZ//
QCISD/cc-pVDZ method are chosen as a benchmark. As
compared to experimental energy barrier of 7.5 kcal/mol, the
benchmark value of 9.7 kcal/mol overestimates the barrier by
2.2 kcal/mol. The RI-BP86/TZVPP//RI-BP86/TZVP approach
yields an activation barrier of 6.9 kcal/mol, which is in excellent
agreement with the experimental value.
These results clearly indicate that the RI-approximation in
combination with DFT constitutes a powerful tool for the
computational investigation of the titanocene mediated 3-exo
cyclizations. This is due to the excellent agreement between
the calculated reaction energy, activation energy, and geometries
of intermediates and transition structures with those of other
more expensive computational approaches as well as the
experimental values. We note that approaches applying com-
putationally less demanding ab initio treatments such as MP2
or CC2 perform distinctly worse than the pure DFT approach
and are therefore not recommended.
Computational Results of [Cp2TiCl] Mediated 3-exo
Cyclizations. We decided to study the 3-exo cyclizations of
various radicals shown in Figure 2, to gain a comprehensive
understanding of the potential energy hypersurface (PES), the
geometries of intermediates, and the transition structures.
Moreover, these computational results were compared to closely
related experimental studies to obtain a clear picture of the
overall reaction including the final radical reduction that is also
relevant for related cyclizations.
Because of the substantial variation (-1.7 to -4.6 kcal/mol)
and the large errors of the experimentally determined reaction
energies, the situation is less clear-cut in this case. The ab initio
benchmark calculations yield a calculated reaction energy of
-3.6 kcal/mol close to the upper limit of the measured values.
The DFT approaches RI-BP86/TZVPP/RI-BP86/TZVP yield
energies (-1.4 and -0.7 kcal/mol) that are close to the lower
limit.
To check the effect of the single point approximation of the
ab initio calculations at the DFT structures on activation barriers
and reaction energies, QCISD/cc-pVDZ//RI-BP86/TZVP and
CCSD(T)/cc-pVTZ//RI-BP86/TZVP calculations were per-
formed and compared to the ab initio results of Coosky.4c The
QCISD case yields results nearly identical to the full ab initio
values. In the CCSD(T) case, the reaction energy differs by only
1 kcal/mol, whereas again agreement is obtained for the
Thermodynamic Considerations. The reaction energies are
summarized in Table 4. The addition of radical stabilizing
groups renders the cyclization slightly exothermic.
This is in agreement with two kinetic studies of free radicals
by Ingold (stabilization by a phenyl group)3a and Beckwith
9
7074 J. AM. CHEM. SOC. VOL. 127, NO. 19, 2005

Titanocene-Catalyzed 3-exo Cyclizations
A R T I C L E S
Table 5. RI-BP86/TZVPP//RI-BP86/TZVP Activation Energies
(kcal/mol) for the Reactions of 4, 7, 10, and 13 Including ZPE
(Barriers for the Opening of 6, 9, 12, and 15 in Brackets)
Table 4. RI-BP86/TZVPP//RI-BP86/TZVP Reaction Energies for
the Reactions of 4, 7, 10, and 13 in kcal/mol Including the ZPE
entry
substrate
product
∆E (trans, cis)
entry
substrate
product
E
act trans, cis
1
2
3
4
5
6
7
4a
4b
7a
7b
7c
10
13
6a
6b
9a
9b
9c
12
15
-8.1, -6.8
-9.3, -8.2
-6.0, -4.9
-5.9, -4.7
-3.8, -3.1
-3.5, -2.1
-1.7, -1.4
1
4a
6a
1.3, 1.2
(9.4, 8.1)
1.0, 0.9
2
3
4
5
6
7
4b
7a
7b
7c
10
13
6b
9a
9b
9c
12
15
(10.3, 9.1)
1.9, 0.7
(7.8, 5.6)
0.8, 0.9
(6.7, 5.6)
3.0, 2.6
(6.9, 5.7)
3.3, 3.5
(6.8, 5.6)
2.2, 3.0
(stabilization by an ester group)3c that suggest a thermodynami-
cally favored cyclization but do not provide reaction energies.
Our results also indicate that Guibe´’s mechanistic proposal
implying an endothermic ring closure in the samarium mediated
cyclization of δ-halo-R,â-unsaturated esters is incorrect.5c
An interesting observation was made for the ester substituted
radical lacking the geminal dimethyl group (entry 6). In this
case, the exothermicity of the reaction was reduced by about 3
kcal/mol for both cis and trans cases. It should be noted that
for the substrate radicals a stability difference of only 0.7 kcal/
mol in favor of 7b was observed. The stabilization in the case
of 9b must therefore be due to an effect of the cyclopropane,
most likely the increased electronegativity of the ring carbons.16
Thus, methyl substitution will result in additional stabilization
of 10 versus 7 as indeed observed in the larger reaction energy.
Because the reaction energy is only slightly negative, we also
investigated the effect of stabilizing the substrate radical by an
additional methyl group for the ester substituted olefin (entry
7). It was found that the exothermicity of cyclization is reduced
to just about 1.4 (cis) and 1.7 (trans) kcal/mol as compared to
entry 4 [4.7 (cis) and 5.9 (trans) kcal/mol]. This result is again
in good agreement with the stability difference of secondary
and tertiary radicals according to their BDEs (ca. 4 kcal/mol).
The computational results suggest another experimental
means for probing the mechanism of the reaction. In case of a
reversible 3-exo cyclization, the more stable trans isomer is
predicted to be formed, except for 15 where the additional
methyl substituent should result in an unselective cyclization.
Discussion of the Energies of Activation. The activation
energies shown in Table 5 are among the lowest in radical
chemistry and suggest that our 3-exo cyclizations are extremely
fast.
(3.8, 4.4)
calculated structures of 7b and 10. Thus, the effect of the two
methyl groups discussed above is already stabilizing transition
structure 8b and results in an acceleration of the cyclization,
too.
The barriers of ring opening are relatively small, and thus
ring opening is fast. Within the series of substrates where
substitution of both radicals involved in the cyclization differs
only in the carbonyl substituent, that is, 7a, 7b, 7c, and 10, the
barriers for ring opening are very similar within the respective
cis and trans series. Because the activation energies are of the
same magnitude as for the cyclopropylcarbinyl radical (3) (trans
series) or slightly lower (cis series), ring opening of 9a, 9b, 9c,
and 12 will have a rate constant of more than 7 × 107 s-1, the
value measured for 3.
A readily reversible 3-exo cyclization will occur in the
absence of efficient trapping reagents. The differences in
activation energies for the formation of the cis and trans
disubstituted cyclopropanes are too small to be discussed
quantitatively. Qualitatively, it is clear, however, that these
differences will be smaller than the corresponding ∆E as the
forming bonds between C1 and C3 are much longer than in the
products.
Summary of the Computational Results. The computational
results demonstrate that the 3-exo cyclizations investigated here
are very fast and thermodynamically favored reactions. The
thermodynamic stabilization as compared to the parent butenyl
radical (2) is due to a delocalization of the enoyl radical formed,
as shown above in Figure 3.
However, ring opening is also predicted to be swift and
comparable to the parent butenyl radical. Thus, in the absence
of efficient trapping reagents, the stereoselectivity of the overall
process is expected to be under thermodynamic control favoring
the trans disubstituted cyclopropylcarbinyl radicals. To dem-
onstrate the effects of the carbonyl groups on both reaction and
activation energy, the PES of the reactions of 2 and 7b are
compared in Figure 4.
In agreement with chemical intuition for radicals 7a-7c, the
barrier is highest for 7c that possesses the largest mesomeric
stabilization of the carbonyl group. The course of the cyclization
and evolution of spin density is shown in Figure 3 for 7b.
With radical 10 lacking the geminal methyl groups, the
activation energy is higher by ca. 2.5 kcal/mol as compared to
7b. As no substantial destabilization of the substrate (0.7 kcal/
mol) is observed, no reactive rotamer effect seems to be
operating.18 Also, no angle compressions were observed in the
Inspection of PES demonstrates that the transition structure
and product radical from 7b are stabilized by about the same
energy, presumably by mesomeric effects of the ester group,
as compared to 2.
The question of the speed of reduction of the cyclopropyl-
carbinyl radicals is too complex to be studied even for the DFT
(14) (a) Lide, D. R., Ed. CRC Handboock of Chemistry and Physics, 83rd ed.;
CRC Press: Boca Raton, FL, 2003. (b) McMillen, D. F.; Golden, D. M.
Annu. ReV. Phys. Chem. 1982, 33, 493-532. (c) Berkowitz, J.; Ellison, G.
B.; Gutman, D. J. Phys. Chem. 1994, 98, 2744-2765. (d) Seetula, J. A.;
Slagle, I. R. J. Chem Soc., Faraday Trans. 1997, 93, 1709-1719.
(15) Bartberger, M. D.; Dolbier, W. R.; Lusztyk, J.; Ingold, K. U. Tetrahedron
1997, 53, 9857-9880.
(16) Wiberg, K. B.; Bader, R. F.; Lau, C. D. H. J. Am. Chem. Soc. 1987, 109,
1001-1012.
(17) (a) Portmann, S.; Lu¨thi, H. P. Chimia 2000, 54, 766-770. (b) Flu¨kiger,
P.; Lu¨thi, H.; Portmann, S.; Weber, J. MOLEKEL 4.2; Swiss Center for
Scientific Computing: Manno, 2000-2002.
(18) For two reviews, see: (a) Jung, M. E. Synlett 1999, 843-846. (b) Jung,
M. E.; Piizzi, G. Chem. ReV. 2005, 105, in press.
9
J. AM. CHEM. SOC. VOL. 127, NO. 19, 2005 7075

A R T I C L E S
Friedrich et al.
Figure 3. Evolution of spin density during the cyclization of 7b and mesomeric stabilization of 9b. The spin density was calculated and visualized with
MOLEKEL.17 The isovalue of the surface was set to 0.025.
Table 6. Titanocene-Catalyzed Cyclopropane Formation (for
Details of Reaction Conditions and Compound Characterization,
See Supporting Information; for Reasons of Clarity, Only the Major
Isomer Is Shown)
Figure 4. Comparison of the RI-BP86/TZVP PES of the reactions of 7b
and 2 (without ZPE).
methods and is therefore addressed experimentally as described
in the following sections.
Experimental Results. We examined epoxyolefins 16, 18,
20, 22, 24, and 26, as substrates for comparison with the
“computational” substrates. We have established before that
changing a geminal dimethyl group by a cyclohexyl group and
a methyl ester by the computationally lavish tert-butyl ester does
not alter the outcome of such reactions substantially while
volatile intermediates were avoided.
The results summarized in Table 6 demonstrate that the
carbonyl substituted epoxides 16, 18, 24, and 26 give the desired
cyclopropanes in reasonable to high yields. The observed
diastereoselectivities are in good to excellent agreement with
the values expected for a reversible cyclization (Table 4). This
implies that trapping of the cyclopropylcarbinyl radical by a
second equivalent of Cp2TiCl is slower than ring opening. A
comparison of the experimental product ratios with the calcu-
lated ratios based on ∆E is possible because an advantageous
cancellation of thermodynamic corrections in the cis and trans
isomers can be expected. It is also reasonable to assume a
In the case of the phenyl and vinyl substituted epoxides 20
and 22, none of the desired cyclopropane was observed in the
crude reaction mixture or could be isolated. Instead, complex
mixtures of polymeric materials were obtained. Here, trapping
similarly advantageous cancellation of the preexponential factors
of the Arrhenius equation for the cis and trans reaction
pathways, because the intermediates and transition structures
involved are quite similar.
9
7076 J. AM. CHEM. SOC. VOL. 127, NO. 19, 2005

Titanocene-Catalyzed 3-exo Cyclizations
A R T I C L E S
Figure 6. Effects of olefin geometry and concentration on the diastereo-
selectivity of the 3-exo cyclization.
Figure 5. RI-BP86/TZVP optimized structures of cyclopropylcarbinyl
radicals 6a and 9b.
Thus, all experimental and computational evidence is in
support of a reversible 3-exo cyclization and a slower and
irreversible trapping of the cyclopropylcarbinyl radicals. If this
trapping is too slow, no cyclopropanes can be isolated and
polymerization occurs instead as observed for 20 and 22.
of the cyclopropylcarbinyl radicals is too slow to be prepara-
tively useful.
The structures of the cyclopropylcarbinyl radicals 6a and 9b
shown in Figure 5 give an indication why trapping is fast in
the enol radical case and slow in the benzyl radical case.
In the former case, the approach of the second equivalent of
Cp2TiCl is sterically difficult as the radical bearing carbon is
“buried” within 6a. Trapping of 6a to yield a benzylic
organotitanium compound must be considered as more difficult
and thus slower as in 9b where the more exothermic formation
of an enolate takes place additionally.
Considering the relative magnitude of these thermodynamic
and kinetic aspects, we suggest that thermodynamic consider-
ations are not necessarily decisive. This is because from the
termination of titanocene-catalyzed 5-exo cyclizations it is
known that alkyl radicals are reduced faster than benzylic
radicals even though the less stable organometallic intermediates
are formed.
In the case of the reduction of vinyl substituted 6b where
steric hindrance will be less relevant, the thermodynamic factors
may be more important, however.
While the observed diastereoselectivity constitutes a strong
hint for a reversible cyclization, we sought further evidence to
support our hypothesis. In case of a reversible reaction,
diastereoselectivity should be independent of the substrate’s
olefin geometry, diastereomeric composition at the epoxide, and
its concentration in the reaction medium.
It is well documented for irreversible radical cyclizations,
for example, in the 5-exo case,19 that diastereoselectivity is
strongly dependent on olefin geometry in the preparation of
bicyclic systems. We therefore examined epoxides 28 as
substrates as demonstrated in Figure 6.
The results amply demonstrate that stereoselectivity of the
cyclization is independent of substrate concentration and of
olefin geometry. Bicyclic product 29 was obtained in essentially
the same diastereomeric ratio in all three reactions. This is
readily explained by a trapping of the enol radical that is slower
than ring opening.
Conclusion
In summary, we have described a complete mechanistic
scheme of titanocene mediated 3-exo cyclization of vinyl
epoxides containing carbonyl, phenyl, and vinyl groups and
determined the structures of relevant intermediates and transition
structures by DFT calculations. The RI-BP86/TZVP method
used was scaled against the parent butenyl radical, the BDE of
pertinent alkyl radicals, and Cp2TiCl. It emerged as a highly
reliable tool for describing and predicting metal mediated and
free radical reactions.
The titanocene mediated 3-exo cyclizations were investigated
by a combination of experimental and computational studies.
They turned out to be exothermic and reversible in contradiction
to an earlier mechanistic proposal.5c The energies of activation
for ring closure are among the lowest known for radical
reactions. The crucial step for a successful completion of the
reaction is constituted by a swift trapping of the cyclopropyl-
carbinyl radicals that is possible only for carbonyl substituted
electrophilic radicals irrespective of the exothermicity of the
cyclization.
Because our study provides a complete picture of the complex
mechanism of the titanocene mediated 3-exo cyclization, the
results and methods applied also should be of distinct relevance
for the successful design and understanding of other unusual
radical cyclizations.
Acknowledgment. We are indebted to the Deutsche For-
schungsgemeinschaft (SFB 624 “Template-Vom Design
chemischer Schablonen zur Reaktionssteuerung”) and the
Fonds der Chemischen Industrie (Dozentenstipendium and
Sachbeihilfen to A.G.) for generous financial support.
Supporting Information Available: The RI-BP86/TZVP
optimized structures and experimental details. This material is
available free of charge via the Internet at http://pubs.acs.org.
(19) (a) Enholm, E. J.; Trivellas, A. Tetrahedron Lett. 1989, 30, 1063-1066.
(b) Enholm, E. J.; Trivellas, A. J. Am. Chem. Soc. 1989, 111, 6463-6465.
JA050268W
9
J. AM. CHEM. SOC. VOL. 127, NO. 19, 2005 7077