Homolysis of weak Ti-O bonds: Experimental and theoretical studies of titanium oxygen bonds derived from stable nitroxyl radicals

Article abstract of DOI:10.1021/ja044512f

Titanium-oxygen bonds derived from stable nitroxyl radicals are remarkably weak and can be homolyzed at 60 °C. The strength of these bonds depends sensitively on the ancillary ligation at titanium. Direct measurements of the rate of Ti-O bond homolysis in Ti-TEMPO complexes Cp₂TiCl(TEMPO) (3) and Cp₂TiCl(4-MeO-TEMPO) (4) (TEMPO = 2,2,6,6-tetramethylpiperidine- N-oxyl, 4-MeO-TEMPO = 2,2,6,6-tetramethyl-4-methoxypiperidine-N-oxyl) were conducted by nitroxyl radical exchange experiments. Eyring plots gave the activation parameters, ΔH? = 27(±1) kcal/mol, ΔS? = 6.9(±2.3) eu for 3 and ΔH? = 28(±1) kcal/mol, ΔS? = 9.0(±3.0) eu for 4, consistent with a process involving the homolysis of a weak Ti-O bond to generate the transient Cp₂Ti(III)Cl and the nitroxyl radical. Thermolysis of the titanocene TEMPO complexes in the presence of epoxides leads to the Cp ₂Ti(III)Cl-mediated ring-opening of the epoxide followed by trapping by the nitroxyl radical. The X-ray crystal structure of the Ti-TEMPO derivative, Cp₂-TiCl(4-MeO-TEMPO) (4), is reported. DFT (B3LYP/6-31G*) calculations and experimental studies reveal that the strength of the Ti-O bond decreases dramatically with the number of cyclopentadienyl groups on titanium. The calculated Ti-O bond strength of the monocyclopentadienyl complex 2 is 43 kcal/mol, whereas that of the biscyclopentadienyl complex 3 is 17 kcal/mol, a difference of 26 kcal/mol. These studies reveal that the strength of these Ti-O bonds can be tuned over an interesting and experimentally accessible temperature range by appropriate ligation on titanium.

Full text of DOI:10.1021/ja044512f

Published on Web 02/25/2005
Homolysis of Weak Ti-O Bonds: Experimental and
Theoretical Studies of Titanium Oxygen Bonds Derived from
Stable Nitroxyl Radicals
Kuo-Wei Huang,† Joseph H. Han,‡ Adam P. Cole,† Charles B. Musgrave,‡,§ and
Robert M. Waymouth*,†,‡
Contribution from the Departments of Chemistry, Chemical Engineering, and
Materials Science and Engineering, Stanford UniVersity, Stanford, California 94305
Received September 10, 2004; E-mail: waymouth@stanford.edu
Abstract: Titanium-oxygen bonds derived from stable nitroxyl radicals are remarkably weak and can be
homolyzed at 60 °C. The strength of these bonds depends sensitively on the ancillary ligation at titanium.
Direct measurements of the rate of Ti-O bond homolysis in Ti-TEMPO complexes Cp2TiCl(TEMPO) (3)
and Cp2TiCl(4-MeO-TEMPO) (4) (TEMPO ) 2,2,6,6-tetramethylpiperidine-N-oxyl, 4-MeO-TEMPO ) 2,2,6,6-
tetramethyl-4-methoxypiperidine-N-oxyl) were conducted by nitroxyl radical exchange experiments. Eyring
plots gave the activation parameters, ∆Hq ) 27((1) kcal/mol, ∆Sq ) 6.9((2.3) eu for 3 and ∆Hq ) 28((1)
kcal/mol, ∆Sq ) 9.0((3.0) eu for 4, consistent with a process involving the homolysis of a weak Ti-O
bond to generate the transient Cp2Ti(III)Cl and the nitroxyl radical. Thermolysis of the titanocene TEMPO
complexes in the presence of epoxides leads to the Cp2Ti(III)Cl-mediated ring-opening of the epoxide
followed by trapping by the nitroxyl radical. The X-ray crystal structure of the Ti-TEMPO derivative, Cp2-
TiCl(4-MeO-TEMPO) (4), is reported. DFT (B3LYP/6-31G*) calculations and experimental studies reveal
that the strength of the Ti-O bond decreases dramatically with the number of cyclopentadienyl groups on
titanium. The calculated Ti-O bond strength of the monocyclopentadienyl complex 2 is 43 kcal/mol, whereas
that of the biscyclopentadienyl complex 3 is 17 kcal/mol, a difference of 26 kcal/mol. These studies reveal
that the strength of these Ti-O bonds can be tuned over an interesting and experimentally accessible
temperature range by appropriate ligation on titanium.
Introduction
bonds derived from stable radicals, in analogy to approaches
utilized in living free-radical polymerization.12-14 This approach
requires metal-ligand bonds of the appropriate energy to be
homolyzed thermally, and for this reason, we were attracted to
the stable nitroxyl radical TEMPO (2,2,6,6-tetramethylpiperi-
dine-1-oxyl).
At the onset of our studies, little was known of the
coordination chemistry or bonding energetics of stable nitroxyl
radicals to early metals.15-17 We have shown that the coordina-
tion chemistry of hydroxylaminato ligands (i.e., those derived
from nitroxyl radicals or hydroxylamines) depends sensitively
on the ligation environment of titanium.18-21 In the absence of
other sterically demanding ligands, the hydroxyl-derived ligand
Transition-metal radicals have a rich reaction chemistry that
is often distinct from their diamagnetic congeners.1 Odd-electron
organometallic complexes derived from low-valent transition
metals have proven to be useful for a variety of organic
transformations;2-5 those from Ti(III), in particular, have
engendered interest for their utility in regioselective epoxide
opening and pinacol couplings.5-10 As part of our interest in
the chemistry of Ti(III) polymerization catalysts,11 we initiated
studies to develop a general strategy for generating Ti(III)
complexes from the homolysis of weak metal-ligand covalent
† Department of Chemistry.
‡ Department of Chemical Engineering.
§ Department of Materials Science and Engineering.
(1) Astruc, D. Electron Transfer Processes in Transition-Metal Chemistry;
Wiley-VCH: New York, 1995.
(12) Benoit, D.; Harth, E.; Fox, P.; Waymouth, R. M.; Hawker, C. J.
Macromolecules 2000, 33, 363-370.
(13) Marque, S.; Le Mercier, C.; Tordo, P.; Fischer, H. Macromolecules 2000,
33, 4403-4410.
(2) Skrydstrup, T. Angew. Chem., Int. Ed. Engl. 1997, 36, 345-347.
(3) Molander, G. A.; Harris, C. R. Tetrahedron 1998, 54, 3321-3354.
(4) Gansauer, A.; Lauterbach, T.; Bluhm, H.; Noltemeyer, M. Angew. Chem.,
Int. Ed. 1999, 38, 2909-2910.
(14) Hawker, C. J.; Bosman, A. W.; Harth, E. Chem. ReV. 2001, 101, 3661-
3688.
(15) Golubev, V. A.; Voronina, G. N.; Chernaya, L. I.; Dyachkovskii, F. S.;
Matkovskii, P. E. Zh. Obsh. Khim. 1977, 47, 1825-1832.
(16) Brindley, P. B.; Scotton, M. J. J. Organomet. Chem. 1981, 222, 89-96.
(17) Evans, W. J.; Perotti, J. M.; Doedens, R. J.; Ziller, J. W. Chem. Commun.
2001, 2326-2327.
(5) Spencer, R. P.; Schwartz, J. Tetrahedron 2000, 56, 2103-2112.
(6) Rajanbabu, T. V.; Nugent, W. A. J. Am. Chem. Soc. 1994, 116, 986-997.
(7) Gansauer, A.; Bluhm, H. Chem. ReV. 2000, 100, 2771-2788.
(8) Gansauer, A.; Pierobon, M.; Bluhm, H. Synthesis 2001, 2500-2520.
(9) Gansauer, A.; Bluhm, H.; Rinker, B.; Narayan, S.; Schick, M.; Lauterbach,
T.; Pierobon, M. Chem.sEur. J. 2003, 9, 531-542.
(18) Huang, K.-W.; Waymouth, R. M. J. Am. Chem. Soc. 2002, 124, 8200-
8201.
(10) Gansauer, A.; Rinker, B. In Titanium and Zirconium in Organic Synthesis;
Marek, I., Ed.; Wiley-VCH: Weinheim, Germany, 2002; pp 435-450.
(11) Mahanthappa, M. K.; Waymouth, R. M. J. Am. Chem. Soc. 2001, 123,
12093-12094.
(19) Mahanthappa, M. K.; Huang, K.-W.; Cole, A. P.; Waymouth, R. M. Chem.
Commun. 2002, 502-503.
(20) Mahanthappa, M. K.; Cole, A. P.; Waymouth, R. M. Organometallics 2004,
23, 836-845.
9
10.1021/ja044512f CCC: $30.25 © 2005 American Chemical Society
J. AM. CHEM. SOC. 2005, 127, 3807-3816
3807

A R T I C L E S
Huang et al.
To investigate the structural effects of biscyclopentadienyl
ligation on the TEMPO ligand, an X-ray structural study of 4
was carried out (3 did not yield suitable crystals for an X-ray
structure). Crystallization of 4 from pentane afforded thin brown
plates suitable for X-ray crystallographic analysis. The structure
of 4 reveals that 4-MeO-TEMPO coordinates to the titanium in
an η1 geometry (Figure 4). The Ti-O bond distance of 1.824
Å and the N-O bond distance of 1.406 Å are indicative of a
reduced TEMPO ligand, as observed for 2.19 Comparison of
the bonding geometry of 4 relative to that of the monocyclo-
pentadienyl compound 2 reveals several significant differences,
notably, a longer Ti-O bond (1.824 Å for 4 versus 1.752 Å
for 2) and a smaller Ti-O-N bond angle (145.6° for 4 versus
155.7° for 2).19 The longer Ti-O bonds for the biscyclopen-
tadienyl derivative 4 (and 3, vide infra) is consistent with
trends observed for mono- and biscyclopentadienyl titanium
alkoxides.23-26 This lengthening also correlates with our ex-
perimental observations that the Ti-O bond of 3 is more readily
homolyzed than that of 2. The Ti-O-N angle of 4 is similar
to that observed in other titanocene alkoxides and sulfides.23,27,28
The smaller Ti-O-N angle of 4 relative to that of 2 may be
indicative of reduced pπ to dπ donation from the oxygen to the
titanium center upon substitution by the second cyclopentadienyl
ligand,23-25 although as pointed out by Parkin and Rothwell,28,29
M-O-R angles are not predictive indicators of the degree of
π-donation in early metal alkoxides.
To provide further evidence for the homolytic Ti-O bond
cleavage of 3 and to provide an experimental estimate of the
Ti-O bond strengths of the biscyclopentadienyl complexes 3-5,
we investigated the kinetics of Ti-O bond homolysis by nitroxyl
exchange experiments. Previous studies on the thermolysis of
3 in the presence of carbon tetrachloride (Figure 2) revealed
that the coupling of nitroxyl radical with Cp2TiCl (8) was much
faster than chlorine atom abstraction from CCl4,30 which
precluded a direct measurement of the rate of Ti-O bond
homolysis.18 Among other possible trapping reagents for
Ti(III) (such as CBr4), we found that nitroxyl radicals underwent
clean exchange reactions that could be readily monitored by
1H NMR (Figure 5).
Figure 1. Homolysis of Ti-O bonds derived from stable radicals.
Figure 2. Thermolysis of titanium TEMPO complexes in CCl4.
TEMPO adopts an η2-coordination geometry for Cl3Ti(TEMPO)
1 in the solid state, but in solution, the TEMPO ligand is labile
and undergoes an η1-η2 interconversion.21 The presence of
cyclopentadienyl ligands enforces an η1-coordination geometry
on the TEMPO ligand for CpTiCl2(TEMPO) 2.19
The thermal stability of the cyclopentadienyl titanium TEMPO
complexes CpTiCl2(TEMPO) 2 and Cp2TiCl(TEMPO) 3 de-
pends sensitively on the number of cyclopentadienyl groups on
titanium; complex 2 was stable for days at 100 °C in CCl4,
whereas the TEMPO ligand is exchanged for chloride upon
thermolysis of 3 in CCl4 (Figure 2).18
Mechanistic studies provided evidence for the homolytic
cleavage of a weak Ti-O bond of 3.18 The facile homolysis of
the Ti-O bond of 3 is indicative of a remarkably weak Ti-O
bond, consistent with our hypothesis that Ti-O bonds derived
from stable radicals should be much weaker than those derived
from alkoxides (approximately 90 kcal/mol).22 What we had
not anticipated was the dramatic influence of the cyclopenta-
dienyl ligands on the strength of these Ti-O bonds; the fact
that 2 is thermally stable at 100 °C for extended periods suggests
that the strength of these Ti-O bonds might be tunable over
an interesting and experimentally accessible temperature range
by appropriate manipulation of the ligand environment at
titanium. In this paper, we report experimental studies on the
homolysis of weak Ti-O bonds derived from stable radicals
combined with theoretical studies to assess how the strength of
these bonds can be tuned by appropriate choice of the ligand
environment or nature of the nitroxyl radical.
The thermolysis of 3 in the presence of excess 4-MeO-
1
TEMPO (10) was monitored by H NMR.31 Initial rates were
monitored as the system approached equilibrium at high
conversion. Plots of ln([3]0/[3]t) versus time were linear with a
zero intercept; a plot of the pseudo-first-order rate constant kobs
versus [4-MeO-TEMPO] () [10]0) at 60 °C in d6-benzene
indicated that the rates were zero order in nitroxyl 10. Monitor-
(23) Kalirai, B. S.; Foulon, J. D.; Hamor, T. A.; Jones, C. J.; Beer, P. D.; Fricker,
S. P. Polyhedron 1991, 10, 1847-1856.
(24) Thorn, M. G.; Vilardo, J. S.; Lee, J.; Hanna, B.; Fanwick, P. E.; Rothwell,
I. P. Organometallics 2000, 19, 5636-5642.
Results and Discussion
(25) Nomura, K.; Naga, N.; Miki, M.; Yanagi, K. Macromolecules 1998, 31,
7588-7597.
To investigate the role of ligand effects on the bonding
geometry and energetics of this family of compounds, we
prepared a series of titanium hydroxylaminato compounds 1-5
(Figure 3). The synthesis was carried out by reduction of a
suitable Ti(IV) halide precursor, followed by trapping of the
Ti(III) intermediate with the nitroxyl radical.18,19
(26) Huffman, J. C.; Moloy, K. G.; Marsella, J. A.; Caulton, K. G. J. Am. Chem.
Soc. 1980, 102, 3009-3014.
(27) Calhorda, M. J.; Carrondo, M. A. A. F. D. T.; Dias, A. R.; Frazao, C. F.;
Hursthouse, M. B.; Simoes, J. A. M.; Teixeira, C. Inorg. Chem. 1988, 27,
2513-2518.
(28) Howard, W. A.; Trnka, T. M.; Parkin, G. Inorg. Chem. 1995, 34, 5900-
5909.
(29) Coffindaffer, T. W.; Rothwell, I. P.; Huffman, J. C. Inorg. Chem. 1983,
22, 2906-2910.
(21) Mahanthappa, M. K.; Cole, A. P.; Waymouth, R. M. Organometallics 2004,
(30) Tenhaeff, S. C.; Covert, K. J.; Castellani, M. P.; Grunkemeier, J.; Kunz,
C.; Weakley, T. J. R.; Koenig, T.; Tyler, D. R. Organometallics 1993, 12,
5000-5004.
23, 1405-1410.
(22) Calhorda, M. J.; Carrondo, M. A. A. F. D. C. T.; Dias, A. R.; Domingos,
A. M. T. S.; Simoes, J. A. M.; Teixeira, C. Organometallics 1986, 5, 660-
(31) Budzelaar, P. H. M. gNMR, version 4.1; Adept Scientific, Herts,
(www.adeptscience.com); Cherwell Scientific Publishing: 1999.
667.
9
3808 J. AM. CHEM. SOC. VOL. 127, NO. 11, 2005

Homolysis of Weak Ti−O Bonds
A R T I C L E S
Figure 3. Titanium hydroxylaminato complexes.
Table 1. Value of kobs for the Thermolysis of 3 and 4 in the
Presence of Nitroxyl Radicalsa
1
entry
[10]0 (M)
[9]0 (M)
T (
°C)
k )
obs (10-4 s-
1b
0.100
0.150
0.200
0.250
0.200
0.200
0.200
0.200
0.200
60
60
60
60
40
45
50
55
65
40
50
60
65
9.3 ( 1.0
9.3 ( 0.5
9.3 ( 0.7
10 ( 1
0.63 ( 0.11
1.4 ( 0.2
2.9 ( 0.8
4.7 ( 0.4
17 ( 2
0.32 ( 0.05
1.2 ( 0.2
4.2 ( 0.5
9.7 ( 1.3
2b
3b
4b
5b
6b
7b
8b
9b
10c
11c
12c
13c
0.200
0.200
0.200
0.200
Figure 4. ORTEP drawing of 4. Selected bond distances (Å) and angles
(deg): Ti1-O1, 1.824(2); Ti1-Cl1, 2.5346(12); N1-O1, 1.406(3); Cp1-
Ti, 2.019; Cp2-Ti, 2.028; Cl1-Ti1-O1, 98.00(8); Ti1-O1-N1,
145.64(19); O1-N1-C11, 109.1(2); O1-N1-C15, 103.8(2); C11-N1-
C15, 120.4(3); Cp1-Ti-Cp2, 127.0.
a [3]0 ) 1.0 × 10-2 M, [4]0 ) 1.0 × 10-2 M. b Plotting ln([3]0/[3]t) vs
time. c Plotting ln([4]0/[4]t) vs time.
kBT) versus 1/T (k ) rate constant; h ) Planck constant; kB )
Boltzmann constant; T in Kelvin). The kinetics and activation
parameters, ∆Hq ) 27((1) kcal/mol, ∆Sq ) 6.9((2.3) eu (∆Gq
) 24.2 kcal/mol at 60 °C) for 3 and ∆Hq ) 28((1) kcal/mol,
∆Sq ) 9.0((3.0) eu (∆Gq ) 24.7 kcal/mol at 60 °C) for 4, are
consistent with those of a process involving the homolysis of a
weak Ti-O bond and are comparable to those obtained for the
homolysis of cobalt-carbon bonds within a similar temperature
range.32,33,35,39 Accounting for possible contributions from cage
effects37,38,40-42 places an upper-limit of the Ti-O bond
dissociation energy of 3 of approximately 25 kcal/mol, which
constitutes the weakest Ti-O bond known for neutral com-
plexes.22,43,44 The slower rate of homolysis of 4 relative to that
of 3 indicates that the activation energies of Ti-O homolysis
reactions depend on the nature of the nitroxyl radical, as
observed for the homolysis of organic alkoxyamines.13,45
Figure 5. Nitroxyl radical exchange.
ing the initial rates under conditions where k-2[10]0 . k-1[9]
yields an estimate of k1, the rate constant for Ti-O bond
homolysis (eq 1).
k-2[10]
d[3]
dt
rate ) -
)
k1[3]
(1)
(
)
k-1[9] + k-2[10]
Equilibrium measurements with known concentrations of 3,
9, 4, and 10 yielded an equilibrium constant for nitroxyl radical
exchange, K ) ([4][9]/[3][10]) ) 2.0 ( 0.2 (Table 2). These
The rate constants of Ti-O homolysis of 3 and 4 in
d6-benzene at different temperatures (40-65 °C) were measured
(Table 1). The thermolysis of 4 was carried out under similar
conditions with TEMPO as the nitroxyl radical. These experi-
ments revealed that kobs for the homolysis of 4 was ap-
proximately one-half that of 3. The enthalpies of activation for
bond homolysis reactions are related to the bond dissociation
energies32-38 and were obtained from Eyring plots of ln(kh/
(38) Koenig, T.; Hay, B. P.; Finke, R. G. Polyhedron 1988, 7, 1499-1516.
(39) Experimental considerations limited the temperature range (15 °C) over
which the activation parameters could be measured, and thus some error
can be expected in the Eyring analysis.
(40) Visanath, D. S.; Natarajan, G. Data Book on the Viscosity of Liquids;
Hemisphere Publishing Corp.: New York, 1989.
(41) Braden, D. A.; Parrack, E. E.; Tyler, D. R. Coord. Chem. ReV. 2001, 211,
279-294.
(42) Male, J. L.; Lindfors, B. E.; Covert, K. J.; Tyler, D. R. J. Am. Chem. Soc.
1998, 120, 13176-13186.
(32) Halpern, J. Polyhedron 1988, 7, 1483-1490 and references therein.
(33) Halpern, J. Bull. Chem. Soc. Jpn. 1988, 61, 13-15.
(34) Daikh, B. E.; Finke, R. G. J. Am. Chem. Soc. 1992, 114, 2938-2943.
(35) Brown, K. L.; Zhou, L. Inorg. Chem. 1996, 35, 5032-5039.
(36) Hay, B. P.; Finke, R. G. Polyhedron 1988, 7, 1469-1481 and references
therein.
(43) Schmittel, M.; Soellner, R. Angew. Chem., Int. Ed. Engl. 1996, 35, 2107-
2109.
(44) Schmittel, M.; Burghart, A.; Werner, H.; Laubender, M.; Soellner, R. J.
Org. Chem. 1999, 64, 3077-3085.
(45) Skene, W. G.; Belt, S. T.; Connolly, T. J.; Hahn, P.; Scaiano, J. C.
Macromolecules 1998, 31, 9103-9105.
(37) Koenig, T.; Finke, R. G. J. Am. Chem. Soc. 1988, 110, 2657-2658.
9
J. AM. CHEM. SOC. VOL. 127, NO. 11, 2005 3809

A R T I C L E S
Huang et al.
Table 2. Equilibrium at 60 °Ca
styrene oxide yielded clean first-order kinetics out to four half-
lives. The rate constant measured from these experiments at 60
°C in C6D6 toluene (kobs ) 10.2((0.2) × 10-4 s-1) is in very
good agreement with the rate of Ti-O bond homolysis measured
from the initial rates in the nitroxyl exchange experiments (Table
1).50 These studies reveal that under these conditions, the rate
of trapping of the CpTi(III)Cl fragment 8 by styrene oxide is
faster than that of the nitroxyl radical 9, a consequence of the
lower concentration of 9 engendered by its rapid reaction with
the benzyl radical 16 (Figure 7).
entry
[4]0 (M)
[10]0 (M)
[9]0 (M)
[3]/[4]
K
1
2
3
1.0 × 10-2
8.0 × 10-3
2.4 × 10-2
9.24 × 10-2
1.18 × 10-1
6.56 × 10-2
1.85 × 10-1
1.07 × 10-1
2.08 × 10-1
1.1
0.52
1.7
1.7 ( 0.3
2.3 ( 0.3
1.9 ( 0.3
a Initial concentration of [3] ) 0. The equilibrium was reached after
150 min at 60 °C.
experiments indicate that the Ti-O bond energy of 4 is greater
than that of 3 by approximately 0.5 kcal/mol (assuming ∆S° ≈
0), comparable to the bond energy difference of C-O bonds
between N-benzoxyl-2,2,6,6-tetramethylpiperidine and N-
benzoxyl-4-hydroxyl-2,2,6,6-tetramethylpiperidine (30.8 versus
31.4 kcal/mol, respectively).45
The influence of the nature of the cyclopentadienyl ligand
was investigated by replacing one Cp ligand of 3 with a Cp*
(pentamethylcyclopentadienyl) group.46 The rate of thermolysis
of Cp*CpTiCl(TEMPO) (5) in the presence of 4-MeO-TEMPO
(3.2((0.3) × 10-3 s-1 at 50 °C) was found to be 10 times
greater than that of 2. This result provides a further indication
that the strength of these Ti-O bonds can be sensitively tuned
through the choice of appropriate ancillary ligands.
These studies provide clear evidence that the homolysis of
weak bonds derived from stable radicals provides a novel
strategy for the generation of odd-electron Ti(III) organome-
tallics. As Nugent and Gansauer have shown that Cp2Ti(III)Cl
is a powerful reagent for the ring-opening of epoxides,4-10 we
investigated the reactivity of 3 with epoxides (Figure 6).
Thermolysis of 3 in the presence of styrene oxide at 100 °C for
10 min in toluene yielded the titanocene alkoxyamine 12 in
quantitative yield by 1H NMR (72% isolated yield). Similarly,
thermolysis of 3 in the presence of indene oxide 13 at 100 °C
for 5 min yielded a single diastereomer of the titanocene
alkoxyamine 14, which could be hydrolyzed to the alkoxyamine
15 (94% isolated yield). The high diastereoselectivity of the
latter process is intriguing and implies that the biscyclopenta-
dienyl titanium fragment constrains the approach of the nitroxyl
radical, yielding a highly diastereoselective coupling of the
nitroxyl and benzyl radical fragments. Thus, ring-opening of
epoxides by the titanocene TEMPO complex 3 provides a
stereoselective strategy for the ring-opening of benzylic epoxides
with the introduction of an alkoxyamine at the benzylic position
which can be used for further elaboration. Aryl diols, such as
1,2-indanediols, are useful intermediates for agrochemical and
pharmaceutical applications47 and can be obtained from the
enantioselective epoxidation of indene.48,49 Further studies on
the scope and limitations of these radical epoxide opening
reactions are in progress.
DFT Calculations. The results of the nitroxide exchange and
the epoxide ring-opening experiments provide clear evidence
that homolysis of Ti-O bonds derived from stable nitroxyl
radicals provides a novel strategy for generating Ti(III) inter-
mediates. These studies reveal that these Ti-O bonds are
considerably weaker than those of titanium alkoxides, at least
for the biscyclopentadienyl derivatives 3 and 4. However, the
observation that the Ti-O bonds of monocyclopentadienyl
complexes 2 are not readily homolyzed at 100 °C reveals that
the strengths of these weak Ti-O bonds are sensitive to ancillary
ligation at titanium. To investigate the influence of ancillary
ligands on the bonding energetics of these Ti-O bonds, we
carried out a series of DFT calculations using the Gaussian 98
program51 on complexes 1-4, 6, and 7. To calibrate our method,
we first surveyed a variety of DFT methods and basis sets and
compared the optimized geometries of the calculated structures
of 1 and 2 with X-ray crystal structures. On the basis of these
studies, we found that the B3LYP52,53 gradient-corrected
exchange hybrid DFT method combined with a 6-31G(d)54-57
basis set resulted in a good compromise of accuracy and
computational expense. This hybrid functional has been shown
to have relatively good accuracy at moderate computational cost
and has been used previously to investigate titanium-oxygen
bonds.57-59 This level of theory has been shown to underestimate
the experimentally measured bond dissociation energies of
organic alkoxyamines,60 by approximately 7-8 kcal/mol;
nevertheless, as we were most interested in the effects on
ancillary ligands on the trends in the Ti-O bond energies, the
B3LYP/6-31G* level of theory was judged adequate to provide
an estimate of these trends.
Structural Simulations of Cl3Ti(TEMPO) (1) and CpTiCl2-
(TEMPO) (2). Three initial binding configurations of TEMPO
(50) See Supporting Information for the derivation of the rate law and figures
giving a summary of the kinetic data.
(51) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M.
A.; Cheeseman, J. R.; Zakrzewski, V. G.; Montgomery, J. A., Jr.; Stratmann,
R. E.; Burant, J. C.; Dapprich, S.; Millam, J. M.; Daniels, A. D.; Kudin,
K. N.; Strain, M. C.; Farkas, O.; Tomasi, J.; Barone, V.; Cossi, M.; Cammi,
R.; Mennucci, B.; Pomelli, C.; Adamo, C.; Clifford, S.; Ochterski, J.;
Petersson, G. A.; Ayala, P. Y.; Cui, Q.; Morokuma, K.; Rega, N.; Salvador,
P.; Dannenberg, J. J.; Malick, D. K.; Rabuck, A. D.; Raghavachari, K.;
Foresman, J. B.; Cioslowski, J.; Ortiz, J. V.; Baboul, A. G.; Stefanov, B.
B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.; Gomperts, R.; 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.; Andres, J. L.; Gonzalez, C.; Head-Gordon, M.; Replogle, E. S.; Pople,
J. A., Gaussian 98, ReV. A.11.2; Gaussian, Inc.: Pittsburgh, PA, 2001.
(52) Becke, A. J. Chem. Phys. 1993, 98, 5648-5652.
The facile trapping of the benzylic radical upon ring-opening
of styrene oxide by TEMPO provided a further experimental
protocol to measure the rate of Ti-O bond homolysis. The
kinetics of the ring-opening of styrene oxide were measured at
60 °C in d6-benzene under pseudo-first-order conditions with
excess styrene oxide. Thermolysis of 3 in the presence of excess
(53) Lee, C.; Yang, W.; Parr, R. Phys. ReV. B 1988, 37, 785-789.
(54) Ditchfield, R.; Hehre, W.; Pople, J. J. Chem. Phys. 1971, 54, 724-728.
(55) Hehre, W.; Ditchfield, R.; Pople, J. J. Chem. Phys. 1972, 56, 2257-2261.
(56) Harihara, P.; Pople, J. Theor. Chim. Acta 1973, 28, 213-222.
(57) Dobado, J. A.; Molina, J. M.; Uggla, R.; Sundberg, M. R. Inorg. Chem.
2000, 39, 2831-2836.
(58) Bergstrom, R.; Lunell, S.; Eriksson, L. Int. J. Quantum Chem. 1996, 59,
427-443.
(59) Flisak, Z.; Szczegot, K. J. Mol. Catal. 2003, 206, 429-434.
(60) Marsal, P.; Roche, M.; Tordo, P.; Claire, P. D. J. Phys. Chem. A 1999,
103, 2899-2905.
(46) Rogers, R. D.; Benning, M. M.; Kurihara, L. K.; Moriarty, K. J.; Rausch,
M. D. J. Organomet. Chem. 1985, 293, 51-60.
(47) Page, P. C. B.; Carnell, A. J.; Mckenzie, M. J. Synlett 1998, 774-776.
(48) Hughes, D. L.; Smith, G. B.; Liu, J.; Dezeny, G. C.; Senanayake, C. H.;
Larsen, R. D.; Verhoeven, T. R.; Reider, P. J. J. Org. Chem. 1997, 62,
2222-2229.
(49) Jacobsen, E. N.; Zhang, W.; Muci, A. R.; Ecker, J. R.; Deng, L. J. Am.
Chem. Soc. 1991, 113, 7063-7064.
9
3810 J. AM. CHEM. SOC. VOL. 127, NO. 11, 2005

Homolysis of Weak Ti−O Bonds
A R T I C L E S
Figure 6. Epoxide opening with titanium TEMPO complex 3.
conversion is facile at room temperature.21 For the η1 structure
of 1c, the large Ti-O-N angle (164°) is similar to that found
in (ArO)2TiCl261 and consistent with the contribution of pπ-dπ
electron donation from oxygen to titanium (vide infra).28,57
However, the loss of the N-Ti coordination increases the energy
of 1c compared to 1a and 1b.
Geometry optimizations of metallocene 2 yielded geometrical
parameters in good agreement with those from the crystal
structure of 2.19 The mean absolute difference of the five
Ti-C, two Ti-Cl, two C-N, one Ti-O, and one N-O bond
distances is 0.02 Å, providing further evidence that the B3LYP/
6-31G* level of theory is appropriate to capture the key
geometrical characteristics of these complexes (Table 3).
Comparison of the structures of Cl3Ti(TEMPO) 1 versus
CpTiCl2(TEMPO) 2 reveals that the η2-coordination of the
TEMPO ligand results in a lengthening of the Ti-O and N-O
bonds of 1a relative to that in 2. For the η1-coordinated TEMPO
in 1c, these distances are more closely analogous to those of 2.
These results reveal that the Ti-O and Ti-Cl bonds of 1c are
slightly shorter than those of 2 (1.723 versus 1.753 Å for Ti-O
bonds, 2.219 versus 2.279 Å for Ti-Cl bonds), implicating that
the addition of a Cp ligand in 2 leads to lengthening of the
Ti-Cl and Ti-O bonds and a decrease in the Ti-O-N bond
angle (167.7 for 1c versus 155.7° for 2).26,28
Structural Simulation of Cp2TiCl(TEMPO) (3) and the
Effect of the Additional Cp. As our experimental results had
indicated that the Ti-O bonds for the biscyclopentadienyl
complexes 3 and 4 were weaker than that of the monocyclo-
pentadienyl complex 2,18 we were particularly interested in the
influence of additional cyclopentadienyl ligands on the bonding
geometries and energetics of Ti-O bonds. Geometry optimiza-
tions of 3 (Figure 9) reveal a N-O bond length of 1.426 Å and
a Ti-O bond length of 1.842 Å, consistent with a covalent
Ti-O bond to a reduced TEMPO ligand. Single-point energy
calculations at the B3LYP/6-311G**//B3LYP/6-31G* level of
theory show that the spin densities of Ti, O, and N atoms are
all zero, consistent with this interpretation. Analysis of the
bonding parameters reveals that the addition of the second
cyclopentadienyl ligand has a significant influence on the Ti-O
Figure 7. Kinetics of epoxide opening with titanium TEMPO complex 3.
Figure 8. Calculated geometries for Cl3Ti(TEMPO) 1.
for complex 1 were calculated in order to locate global and local
minima during geometry optimizations. For these calculations,
a chair conformation for the piperidine ring was initially
assumed and subsequently confirmed for the optimized geom-
etries. For the η2 geometries, the oxygen was placed at either
the axial (1a) or the equatorial position (1b) of the TiCl3
fragment and optimized to a locally stable geometry (Figure
8). The B3LYP/6-31G* geometry optimizations indicated that
1b and 1c are less stable than 1a by 2 kcal/mol after ZPE
correction. The identification of 1a as the global minimum is
consistent with the crystallographically determined structure for
1.19 The slightly higher energy of 1b is likely a consequence of
the steric repulsion between the axial Cl atom and the two axial
methyl groups of the piperidine ring as both 1a and 1b have
similar core Ti-O-N bonding geometries. The DFT calcula-
tions reveal that the η1 geometry in 1c is only moderately higher
in energy than the η2 geometry in 1a, which is consistent with
solution measurements that indicate that the η1-η2 inter-
(61) Waratuke, S. A.; Thorn, M. G.; Fanwick, P. E.; Rothwell, A. P.; Rothwell,
I. P. J. Am. Chem. Soc. 1999, 121, 9111-9119.
9
J. AM. CHEM. SOC. VOL. 127, NO. 11, 2005 3811

A R T I C L E S
Huang et al.
Table 3. Selected Experimental and Calculated Bonds Lengths in 1-4a
geometrical
parameter
geometrical
parameter
geometrical
parameter
exp 1
calcd 1a
calcd 1c
exp 2
calcd 2
exp 4
calcd 4
calcd 3
Ti-O
1.838
2.112
2.258
2.218
1.433
1.527
79.4
1.842
2.151
2.272
2.220
1.416
1.535
81.5
1.723
-
Ti-O
1.753
1.758
Ti-O
1.824
1.845
1.842
Ti-N
Ti-N
Ti-N
Ti-Cla
Ti-Cle
N-O
N-C
Ti-O-N
2.215
2.219
1.392
1.502
167.7
Ti-Cl
Ti-Cp
N-O
N-C
Ti-O-N
2.279
2.015
1.412
1.502
155.7
2.279
2.056
1.403
1.505
154.3
Ti-Cl
Ti-Cpa
N-O
N-C
Ti-O-N
2.534
2.028
1.406
1.536
145.6
2.400
2.106
1.420
1.500
144.5
2.404
2.107
1.426
1.506
144.8
a Bond distances are reported in angstroms and angles in degrees.
energies of the Ti-O bonds. Unless otherwise specified, all
reported bond energies have been zero-point corrected and
represent the energy at 0 K. In the cases of 3 and 4, thermal
corrections were also applied to obtain the enthalpy of the bond
homolysis at 333.15 K in order to compare to experimental
results. At the B3LYP/6-31G* level of theory, the calculated
Ti-O bond dissociation energies were observed to depend
sensitively on the ancillary ligation at Ti. For the Cl3Ti(TEMPO)
complexes 1a and 1c, the Ti-O bond is calculated to be 56
and 54 kcal/mol, respectively. This calculated bond energy is
substantially weaker than that of typical Ti-O bonds (90 kcal/
mol),22 as we had anticipated from the higher stability of the
nitroxyl radical relative to typical alkoxy radicals. What we had
not anticipated was the dramatic influence of the cyclopenta-
dienyl ligands on the strength of these Ti-O bonds. Replace-
ment of one of the chlorides with a cyclopentadienyl ligand to
give compound 2 results in a weakening of the Ti-O bonds by
approximately 11-13 kcal/mol (Table 4).
A more dramatic and unanticipated effect was observed on
the addition of a second cyclopentadienyl ligand, as in 3. The
calculated bond energy of 2 is 43 kcal/mol, while those of 3
and 4 are both 17 kcal/mol, a difference of 26 kcal/mol!
Comparison of the calculated (gas phase) bond energies,
including thermal corrections to 333.15 K, with those derived
from the nitroxyl exchange experiments for 3 and 4 (Table 4)
reveals that the B3LYP/6-31G* level of theory leads to an
underestimation of the bond energies if the experimental value
is taken to be 25 kcal/mol. This 7-8 kcal/mol underestimation
is similar to that observed from calculated and experimental
estimates of the C-O bond strengths of alkoxyamines at this
level of theory.60,63
The significantly lower calculated Ti-O bond energy of 3
relative to that of 2 or 1 provides a rationale for our experimental
observations that thermolysis of the biscyclopentadienyl complex
3 results in the homolytic cleavage of the Ti-O bond at 60 °C,
whereas complex 2 is stable to homolytic cleavage at 100 °C
for at least 24 h.18 As we were unable to measure the homolytic
cleavage of the Ti-O bond in 2, we have no experimental
estimate for the strength of this bond; nevertheless, the fact that
2 is stable at 100 °C is consistent with the calculated difference
of 26 kcal/mol.
Analysis of the structural parameters of 2 and 3 reveals that
the lower Ti-O bond energies correlate to the longer Ti-O
bonds of the geometry-optimized or crystallographically char-
acterized complexes. The much lower bond energy of 3 relative
to that of 2 or 1 might be attributed to both steric and electronic
effects. Analysis of the nonbonded contacts between the
Figure 9. B3LYP/6-31G* optimized bond lengths (Å) and angles (deg)
for 3: Ti-Cp1, 2.214; Ti-Cp2, 2.107; Ti-O, 1.842; Ti-Cl, 2.404; N-O,
1.426; N-C11, 1.506; Cp1-Ti-Cp2, 128.8; Ti-O-N, 144.8; Cl-Ti-O,
97.0; O-N-C11, 107.5; O-N-C10, 109.1; C11-N-C15, 118.8.
and Ti-Cl bond lengths.26 Comparison of the B3LYP/6-31G*
optimized structure of 3 to that determined crystallographically
for 2 reveals a significantly longer Ti-O bond distance of 3
relative to that of 2 (1.842 Å for 3 versus 1.753 Å for 2) and a
smaller Ti-O-N bond angle (144.8° in 3 versus 155.7° in 2),
as evidenced as well by the crystal structures of 4 versus 2
(Table 3).
Analysis of the molecular orbitals corresponding to Ti-O
bonding reveals a combination of σ- and π-bonding for the
Ti-O bonds. Presented in Figure 10 are the molecular orbitals
corresponding to the Ti-O bonding interactions in 1c, 2, and 3
using an isodensity surface with a value of 0.03. For 1c and 2,
a Ti-O σ-bond and two Ti-O π-bonds are clearly evident;
similar multiple bonding between Ti and O was described for
Cl3TiOCH3.57 In contrast, for 3, the molecular orbitals reveal a
σ-bond (HOMO-39) and a π-bond (HOMO-11) in the Cl-Ti-O
plane, but as π-bonding in the Cp-M-Cp plane is dominated
by the strongly π-donating cyclopentadienyl ligands, the Ti-O
π-interaction is diminished, analogous to the situation described
for titanocene dithiolene complexes.62 Thus, analysis of the
molecular orbitals suggests that there is less Ti-O π-bonding
in the biscyclopentadienyl complex 2 relative to that in 3 as a
consequence of the strong π-donor properties of the cyclopen-
tadienyl ligand.
Calculated Ti-O Bond Energies. To assess the influence
of the ancillary ligand environments on the strength of the Ti-O
bonds, we carried out calculations on the bond dissociation
(62) Flemmig, B.; Strauch, P.; Reinhold, J. Organometallics 2003, 22, 1196-
(63) Once more experiments become available, a higher level of theory might
be warranted to obtain better agreement between experiment and theory.
1202.
9
3812 J. AM. CHEM. SOC. VOL. 127, NO. 11, 2005

Homolysis of Weak Ti−O Bonds
A R T I C L E S
Figure 10. Isodensity plots of selected orbitals of 1c, 2, and 3 with a value of 0.03.
Table 4. Calculated Ti-O Bond Energies
1a
1c
2
3
4
6
7
calcd Ti-O
1.842
1.723
1.758
1.842
1.845
1.862
1.769
bond distance (Å)
Ti-O bond energy
(kcal/mol) at 0 K (333.15 K)
56
54
43
17 (17)
17 (17)
30
50
hydrogen atom of the TEMPO methyl groups and those of the
cyclopentadienyl ligands in 2 (2.780 Å) and 3 (2.229 Å) suggests
that the sterically congested nature of 3 might contribute to a
destabilization of the Ti-O bond. To assess the magnitude of
these steric effects as well as to investigate the influence of the
nature of the nitroxyl ligand on the Ti-O bond strength, we
calculated the Ti-O bond energy of a less sterically hindered
analogue 6, where the TEMPO ligand is replaced by the
1-piperidinyloxyl ligand (1-PO). The calculated Ti-O bond
energy for 6 is 30 kcal/mol, 13 kcal/mol stronger than the Ti-O
bond energy of 3. This calculation reveals that the strength of
Ti-O bonds derived from nitroxyl radicals also depends on the
nature of nitroxyl ligand, as we observed experimentally from
comparisons of 3 and 4 and as anticipated from analogous
studies of the C-O bond of organic hydroxylamines.13,45
Comparison of the O-H bond strength of N-hydroxypiperidine
(77.0 kcal/mol)64 versus that of N-hydroxyltetramethylpiperidine
(69.6 kcal/mol)65,66 reveals a difference of 7.4 kcal/mol in the
O-H bond strengths. If the strength of the Ti-O bond correlates
to the strength of the O-H bond of the hydroxylamines (not
necessarily a good assumption), then we can get a crude estimate
of the steric contribution to the destabilization of the Ti-O bond
energy of 3 of approximately 6 kcal/mol (that is, the Ti-O bond
of 3 is approximately 6 kcal/mol weaker than we might predict
on the basis of the relative strengths of the O-H bond of the
corresponding hydroxylamines).
Another estimate of the steric destabilization of 3 can be
calculated from the vertical bond (or snap) dissociation energies
(vBDE) of 3 and 6. This calculation involves a single-point
energy calculation for the Cp2TiCl and nitroxyl fragments upon
cleavage of the Ti-O bond at the optimized geometry of the
bound complexes. The geometries of the Cp2TiCl and nitroxyl
fragments are then optimized to get an estimate of the relaxation
energies of the two fragments following the bond homolysis.
To the extent that a component of the relaxation energy of the
Cp2TiCl fragment following bond homolysis represents the steric
distortion of the Cp2TiCl component in the bound complexes 3
and 6, the differences in the relaxation energies provide an
estimate of the steric effects on the Ti-O bond strengths.
These calculations reveal that the vBDEs for 3 and 6 are 52
and 60 kcal/mol, respectively. The relaxation energies of the
Cp2TiCl fragment derived from 3 and 6 are 18 and 14 kcal/
mol, respectively, whereas the relaxation energies for the
nitroxyl fragments are similar at 15 kcal/mol. This calculation
reveals that the relaxation energy of the Cp2TiCl fragment
(64) Bordwell, F. G.; Liu, W. Z. J. Am. Chem. Soc. 1996, 118, 8777-8781.
(65) Bordwell, F. G.; Liu, W. Z. J. Am. Chem. Soc. 1996, 118, 10819-10823.
(66) Mahoney, L. R.; Mendenha, G. D.; Ingold, K. U. J. Am. Chem. Soc. 1973,
95, 8610-8614.
9
J. AM. CHEM. SOC. VOL. 127, NO. 11, 2005 3813

A R T I C L E S
Huang et al.
derived from 3 is approximately 4 kcal/mol greater than that of
6. This value is in reasonable agreement with the magnitude of
the steric effect estimated from the difference in O-H bonds
strengths.
1,3-di-tert-butylcyclopentadienyl) and indicated that the value
for Cp/2TiI2 was comparable to that of TiI4 (56(2) kcal/mol).67
The lower Ti-I bond energy of Cptt TiI2 was attributed to steric
2
effects. These data would suggest that the addition of Cp*
ligands does not lead to a significant decrease in the Ti-I bond
strengths, which is at variance with our observations for Ti-
ONR2 bonds. Nevertheless, it is known that Ti-Cl bond lengths
increase upon substitution of chloride ancillary ligands with Cp
ligands (av Ti-Cl: TiCl4 ) 2.163 Å,69,70 CpTiCl3 ) 2.22 Å,71
Cp2TiCl2 ) 2.364 Å72), and similar trends have been observed
for Ti-OR bonds.23-26 Recent results from Ti and Cl K-edge
X-ray absorption spectroscopy and DFT calculations reveal that
the covalency of Ti-Cl bonds decreases in the series TiCl4,
CpTiCl3, Cp2TiCl2, largely as a consequence of the strongly
donating cyclopentadienyl ligand.73 To the extent that longer
bond lengths and lower covalencies correlate to weaker bonds,
these trends would suggest that the influence of Cp ligands on
the bonding of Ti-X bonds may not be unique to the nitroxyl-
derived ligands, but representative of a more general trend in
early transition-metal complexes. Further studies are warranted
to test this hypothesis. Moreover, these results imply that the
assumption of the transferability of Ti-X bond strengths among
different ligand environments, which has proven to be useful
in organometallic thermochemistry, should be re-examined,
particularly for cyclopentadienyl ligands.22,67,68,74,75
A similar set of calculations involving the 1-PO analogue of
2, CpTiCl2(1-PO) (7), does not show the same steric effect. The
ZPE-corrected bond dissociation energy increases from 43 kcal/
mol in 2 to 50 kcal/mol in 7. This difference of 7 kcal/mol is
smaller than the change of 13 kcal/mol observed in the bis-Cp
structures and is more closely in line with the differences in
the O-H bond energies (7.4 kcal/mol). The vBDEs of 2 and 7
are 67 and 73 kcal/mol, respectively. The relaxation energies
for the CpTiCl2 fragment derived from 2 and 7 are both 11
kcal/mol, suggesting that there is very little steric effect
contributing to the difference in Ti-O bond energies for the
monocyclopentadienyl complexes 2 and 7.
Comparison of these estimates reveals that steric effects are
only partially responsible for extraordinarily weak Ti-O bond
energy in 3 and cannot account for the 26 kcal/mol difference
in the bond energies calculated for 2 and 3. Moreover,
comparison the Ti-O bond energies of the less sterically
hindered analogues 6 and 7 shows a similarly large decrease in
the Ti-O bond energy upon the addition of a second Cp ligand
(the calculated bond energy of 6 is 20 kcal/mol weaker than
that of 7). This leads us to conclude that steric interactions
between the cyclopentadienyl ligands and the TEMPO ligand
are likely not the dominant effect that weakens the Ti-O bond
in the biscyclopentadienyl complexes 3-6, but rather the
remarkable weakening of the Ti-O bond is due to an electronic
effect of adding the strongly donating cyclopentadienyl ligand.
One measure of this electronic effect can be obtained from the
coefficients of the component atomic orbitals on the molecular
bonding orbitals for the Ti-O bonds. For example, the
coefficient of the oxygen pz orbital (corresponding to the Ti-O
σ-bond) changes from 0.154 to 0.136 to 0.100 for 1c, 2, and 3,
respectively.
Conclusion
In conclusion, Ti-O bonds derived from stable nitroxyl
radicals are considerably weaker than those of titanium alkox-
ides, and in appropriate ligation environments, they can be
homolyzed at slightly elevated temperatures to generate Ti(III)
complexes and nitroxyl radicals. The titanium(III) intermediates
are rapidly trapped by nitroxyl radicals, but can competitively
ring-open benzylic epoxides in high yield.
The strength of these Ti-O bonds is extremely sensitive to
the ancillary ligation of titanium. Calculations at the B3LYP/
6-31G* level of theory reveal that the addition of strongly
donating cyclopentadienyl ligands results in a weakening of the
Ti-O bond by up to 26 kcal/mol, providing a means of
modulating the Ti-O bond energy over an interesting and
experimentally accessible temperature range by appropriate
manipulation of the ligand environment at titanium. The strong
sensitivity of these Ti-O bonds to ancillary ligand environments
implies that the influence of ancillary ligands, particularly
cyclopentadienyl ligands, on Ti-X bond strengths is greater
than previously anticipated.
The results of our experimental and theoretical studies indicate
that the bond strengths of Ti-O bonds derived from the stable
nitroxyl radical TEMPO are exquisitely sensitive to ancillary
ligation environment. Calculations reveal that these Ti-O bonds
weaken significantly upon substitution of chloride ancillary
ligands with Cp ligands (Table 4, Cl3Ti(TEMPO), 1c ) 54 kcal/
mol; CpCl2Ti(TEMPO), 2 ) 43 kcal/mol; and Cp2ClTi-
(TEMPO), 3 ) 17 kcal/mol). These effects could be due to a
destabilization of the Ti-O bonds by ancillary ligation or a
preferential stabilization of the Ti(III) products by the cyclo-
pentadienyl ligands. Both of these effects are likely to play a
role, but our view is that the ability of cyclopentadienyl ligands
to stabilize the highly coordinatively and electronically unsatur-
ated Ti(III) species is likely a significant factor leading to the
lower Ti-O bond energies. Are these observations specific to
Ti-O bonds derived from nitroxyl radicals or representative
of general bonding trends in early transition-metal systems? If
the Cp effect is primarily a consequence of stabilizing the
Ti(III) products, then these observations should be general to a
variety of Ti-X bonds, but the evidence from the literature is
not definitive on this question, due to the paucity of data on
M-X bond strengths for a variety of ligation environ-
Experimental Section
Computational Details. All of the quantum mechanical computa-
tions have been performed with the Gaussian 98 package51 using the
all-electron 6-31G(d) basis set54-56 for all atoms at B3LYP level of
theory.52 The exchange functional is a linear combination of Slater’s
(67) King, W. A.; Di Bella, S.; Gulino, A.; Lanza, G.; Fragala, I. L.; Stern, C.
L.; Marks, T. J. J. Am. Chem. Soc. 1999, 121, 355-366.
(68) Simoes, J. A. M.; Beachamp, J. L. Chem. ReV. 1990, 90, 629-688.
(69) Morino, Y.; Uehara, H. J. Chem. Phys. 1966, 45, 4543-4550.
(70) Troyanov, S. I.; Snigireva, E. M. Russ. J. Inorg. Chem. 2000, 45, 580-
585.
(71) Engelhardt, L. M.; Papasergio, R. I.; Raston, C. L.; White, A. H.
Organometallics 1984, 3, 18-20.
(72) Clearfield, A.; Warner, D. K.; Saldarriagamolina, C. H.; Ropal, R.; Bernal,
I. Can. J. Chem. ReV. Can. Chim. 1975, 53, 1622-1629.
(73) Debeer, G. S.; Brant, P.; Solomon, E. I. J. Am. Chem. Soc. 2005, 127,
667-674.
ments.22,67,68 Marks determined the Ti-I bond strengths for Cptt -
2
TiI2 (40.6(5) kcal/mol) and Cp/2TiI2 (52.3(6) kcal/mol) (Cptt )
9
3814 J. AM. CHEM. SOC. VOL. 127, NO. 11, 2005

Homolysis of Weak Ti−O Bonds
A R T I C L E S
local spin density, Becke’s 1988 nonlocal, and exact (Hartree-Fock)
exchange terms. The correlation functional is a combination of Lee-
Yang-Parr and VWN local spin density correlation functionals.53
Geometry optimizations of all the structures were performed using the
default convergence criteria without any geometric constraints. Fre-
quency calculations were performed to confirm that geometries were
at the minimum of an energy well. The unscaled frequencies obtained
from these calculations were also used to determine the zero-point
energies (ZPE) as well as thermal corrections when applicable. The
thermal corrections included translation, rotational, vibrational, and
electronic contributions to the partition function. All internal rotations
were treated using a harmonic oscillator approximation.
General Considerations. Nitrogen gloveboxes and standard Schlenck
techniques were applied in handling all oxygen and moisture-sensitive
compounds. Pentane and toluene were purchased from Aldrich and
purified through towers containing alumina and Q5 prior to use.
Tetrahydrofuran and d6-benzene were vacuum transferred from purple
sodium/benzophenone solutions. Styrene oxide obtained from Aldrich
was dried over CaH2 and distilled in vacuo prior to use. Cp2TiCl2 was
obtained from Strem Chemicals, Inc. and used without further purifica-
tions. TEMPO and 4-MeO-TEMPO were purchased from Aldrich and
sublimed in vacuo prior to use. Zinc dust was purchased from Aldrich,
dried in vacuo overnight, and stored under N2 prior to use. Cp*CpTiCl246
and indene oxide76 were synthesized according to the literature
procedures.
in 1.0 mL of d6-benzene solution were added various amounts of
4-MeO-TEMPO (18.6-46.5 mg) (Table 1). These reactions were
monitored by H NMR on a Varian Mercury 400 at 40-65 °C. The
1
ratio of ([2]t/[2]0) was determined by the integration of the Cp proton
signal of Cp2TiCl(TEMPO)/the sum of the integration of the Cp proton
signals in both Cp2TiCl(TEMPO) and Cp2TiCl(4-MeO-TEMPO) using
gNMR31 analysis, assuming a 10% error for NMR integrations.
Thermolysis of Cp2TiCl(MeO-TEMPO) (4) in the Presence of
TEMPO (9). Cp2TiCl(4-MeO-TEMPO) (4.0 mg) ([4]0 ) 1.0 × 10-2
M), TEMPO (31.2 mg), and a few crystals of hexamethylbenzene (used
as the internal standard) were dissolved in 1.0 mL of d6-benzene
solution. These reactions were monitored by 1H NMR on a Varian
Mercury 400 at 40-65 °C.
Thermolysis of Cp*CpTiCl(TEMPO) (5) in the Presence of
4-MeO-TEMPO (10). Cp*CpTiCl(TEMPO) (4.4 mg) ([5]0 ) 1.0 ×
10-2 M), TEMPO (37.2 mg), and a few crystals of hexamethylbenzene
(used as the internal standard) were dissolved in 1.0 mL of d6-benzene
solution. These reactions were monitored by 1H NMR on a Varian
Mercury 400 at 25-50 °C.
Estimation of the Cage Effect.
kh
kd
[Ti-TEMPO] y\ \
kz [TEMPO‚‚‚TiIII] yk z [TEMPO•][•TiIII]
c D
Koenig and Finke have discussed the importance of radical cage effects
in detail.37 If the cage effect is important, k1 (kobs) and k-1 should be
referred to as (kdkh/(kc + kD)) and (kckD/(kc + kD)), respectively, where
the fractional cage efficiency Fc ) kc/(kc + kD). The bond dissociation
energy37
1H NMR spectra were recorded at 400 MHz on a Varian Mercury
400. 13C NMR spectra were recorded at 100 MHz on the same
instrument. The temperatures of the probe for variable temperature
studies were calibrated by ethylene glycol. 1H and 13C chemical shifts
were referenced relative to tetramethylsilane by internal residual solvent
peaks. Elemental analyses were carried out at either Desert Analytics
Laboratory (Arizona, USA) or Atlantic Microlab, Inc. (Georgia, USA).
High-resolution mass spectra (HRMS) were performed on a Micromass
Q-Tof hybrid quadrupole-time-of-flight LC-MS in the Vincent Coates
Foundation Mass Spectrometry Laboratory at Stanford University.
Synthesis of Cp2TiCl(4-MeO-TEMPO) (4) and Cp*CpTiCl-
(TEMPO) (5). Complex 4: To a mixture of Cp2TiCl2 (747 mg, 3 mmol)
and Zn (451 mg) was added 30 mL of THF. This mixture was stirred
for 30 min. This green mixture was then filtered to remove excess Zn.
A solution of 558 mg of 4-MeO-TEMPO (3 mmol) in 10 mL of THF
was cannulated into the filtrate. After stirring for 5 h, THF was removed
in vacuo. The residue was extracted with 40 mL of toluene. The
resulting solution was filtered to remove ZnCl2. Toluene was removed
in vacuo and the residue extracted with 100 mL of pentane. The
resulting brown solution was filtered. Brown-red thin plates were
obtained after recrystallization from the pentane filtrate (666 mg,
BDE ≈ ∆Hqobs (soln) + (1 - Fc)[ ∆Hqd - ∆Hq ] - ∆Hq-1(soln)
c
where ∆Hq is the enthalpy of escape from the cage, ∆Hq the radical
d
c
recombination enthalpy, ∆Hq-1(soln) the enthalpy of formation of the
cage, and Fc is defined above. If we assume that Fc ) 1, that solvation
of the various transition states are similar (δ (solvation) ) 0), and that
∆Hq-1(soln) ≈ 1.9 kcal/mol (the activation enthalpy for a diffusion-
controlled reaction in benzene),40 then we can calculate an upper limit
of the Ti-O BDE of 3 ≈ ∆Hq (soln) - ∆Hq-1(soln) ) 27((1) -
obs
1.9 ) 25((1) kcal/mol, or approximately 25 kcal/mol.
Synthesis of 12: Cp2TiCl(TEMPO) (370 mg, 1 mmol) and styrene
oxide (240 mg, 2 mmol) were dissolved in 10 mL of toluene in a
reaction tube. The reaction mixture was heated at 125 °C for 10 min
and cooled to room temperature. Then, 15 mL of pentane was added
into the mixture. The product was recrystallized twice at -50 °C to
1
give a pale yellow solid (345 mg, 72%). H NMR (C6D6): δ 0.86 (s,
3H), 1.11 (s, 3H), 1.29 (s, 3H), 1.50 (s, 3H), 1.20-1.50 (bm, 6H),
4.71 (dd, J ) 7.2, 11.6 Hz, 1H), 4.81 (dd, J ) 4.4, 7.2 Hz, 1H), 5.02
(dd, J ) 4.4, 11.6 Hz, 1H), 5.73 (s, 5H), 5.84 (s, 5H), 7.11 (t, J ) 7.6
Hz, 1H), 7.21 (t, J ) 7.6 Hz, 1H), 7.34 (t, J ) 7.6 Hz, 1H). 13C NMR
(C6D6): δ 143.3, 128.3, 127.9, 127.3, 116.3, 87.5, 84.9, 60.1, 40.7,
34.5, 20.6, 17.5. Anal. Calcd: C, 66.19; H, 7.41; N, 2.86. Found: C,
66.13; H, 7.37; N, 2.72.
1
55.5%). H NMR (C6D6, 60 °C): δ 1.01 (s, 6H), 1.28 (s, 6H), 1.54
(m, 2H), 1.83 (m, 2H), 3.13 (s, 3H), 3.26 (m, 1H), 6.07 (s, 10H). 13C
NMR (C6D6, 60 °C): δ 117.1, 71.5, 63.0, 55.4, 46.1, 32.5 (b), 22.7
(b). Anal. Calcd: C, 60.09; H, 7.56; N, 3.50. Found: C, 60.27; H,
7.82; N, 3.65.
Complex 5: This was prepared from Cp*CpTiCl2 and TEMPO using
the same procedure as for 4 but on a smaller scale (1 mmol). Brown-
red needles were obtained after recrystallization twice from the pentane
Reaction of (3) with Indene Oxide (13): Cp2TiCl(TEMPO) (20.7
mg, 0.056 mmol) and indene oxide (7.4 mg, 0.056 mmol) were
dissolved in 0.7 mL of d6-benzene in a J-Young tube. The reaction
1
(140 mg, 31.9%). H NMR (C6D6, 60 °C): δ 1.07 (s, 6H), 1.28 (s,
3H), 1.36-1.62 (m, 6H), 1.15 (s, 3H), 1.85 (s, 15H), 6.15 (s, 5H). 13
C
1
was heated at 100 °C for 5 min. From the quantitative yield by H
NMR (C6D6, 60 °C): δ 125.4, 118.1, 64.3, 61.2, 43.3, 42.1, 33.7, 32.0,
22.7, 22.1, 17.1, 13.1. Anal. Calcd: C, 65.53; H, 8.71; N, 3.18.
Found: C, 65.45; H, 8.81; N, 3.04.
Thermolysis of Cp2TiCl(TEMPO) (2) in the Presence of 4-MeO-
TEMPO (10). To 3.7 mg of Cp2TiCl(TEMPO) ([2]0 ) 1.0 × 10-2 M)
and a few crystals of hexamethylbenzene (used as the internal standard)
NMR, only one product was observed, >99% trans selectivity as
confirmed by ROESY. 1H NMR (C6D6): δ 1.00 (s, 3H), 1.20 (s, 6H),
1.39 (s, 3H), 1.13-1.52 (bm, 6H), 2.94 (dd, J ) 3.0, 16.8 Hz, 1H),
3.26 (dd, J ) 6.0, 16.8 Hz, 1H), 5.27 (d, J ) 3.0 Hz, 1H), 5.50 (ddd,
J ) 3.0, 3.0, 6.0 Hz, 1H), 5.94 (s, 5H), 5.98 (s, 5H), 7.08-7.13 (m,
3H), 7.67 (dd, J ) 1.6, 7.2 Hz, 1H). 13C{1H} NMR (C6D6): δ 143.1,
141.6, 128.8, 127.6, 126.2, 125.2, 116.6, 116.1, 96.1, 92.8, 60.5, 60.1,
40.7, 40.5, 39.7, 34.4, 33.9, 20.7 (br), 17.6.
(74) Schock, L. E.; Marks, T. J. J. Am. Chem. Soc. 1988, 110, 7701-7715.
(75) Marks, T. J. Bonding Energetics in Organometallic Compounds; American
Chemical Society: Washington D.C., 1990; Vol. 428.
(76) Imuta, M.; Ziffer, H. J. Org. Chem. 1979, 44, 1351-1352.
Isolation of (15): The d6-benzene solution of 14 was diluted with 2
mL of diethyl ether. This solution was mixed with 2 mL of 0.5 M HCl
9
J. AM. CHEM. SOC. VOL. 127, NO. 11, 2005 3815

A R T I C L E S
Huang et al.
(aq). The organic layer was separated. The aqueous layer was extracted
with 3 × 1 mL of ether. The organic layers were combined, washed
with brine (2 × 2 mL), and dried over Na2SO4. Then the solution was
eluted through a silica gel column. Solvent was removed in vacuo to
anisotropically. Hydrogen atoms were located by difference Fourier
synthesis and were constrained to idealized geometries in a riding
(AFIX) refinement. A single free rotation angle about the C-CH3 bond
vector was also refined in the case of methyl groups. Scattering factors
were taken from the International Tables for Crystallography.80 All
calculations were performed using the CrystalStructure81,82 crystal-
lographic software package, except for refinement, which was per-
formed using SHELXL-97.78
1
give 15 as a pale yellow oil (15.2 mg, 94%). H NMR (CDCl3): δ
1.13 (s, 3H), 1.29 (s, 6H), 1.30 (s, 3H), 1.32-1.56 (m, 6H), 2.76 (dd,
J ) 8.4, 14.8 Hz, 1H), 3.16 (dd, J ) 7.2, 14.8 Hz, 1H), 3.45 (br, 1H),
4.71 (dt, J ) 8.4, 7.2 Hz, 1H), 5.35 (d, J ) 7.2 Hz, 1H), 7.16-7.23
(m, 3H), 7.37 (m, 1H). 13C{1H} NMR (CDCl3): δ 140.2, 138.4, 127.9,
126.6, 124.8, 124.1, 90.2, 79.9, 61.6, 59.7, 40.4, 40.2, 36.9, 34.7, 33.7,
20.4 (2C), 17.2. HRMS Calcd for C18H28NO2 (M + 1 for [M + H]+):
290.2120. Found: 290.2112.
Thermolysis of Cp2TiCl(TEMPO) (3) in the Presence of an
Excess Amount of Styrene Oxide (11): Cp2TiCl(TEMPO) (4.5 mg,
0.012 mmol), styrene oxide (29.3 mg, 0.244 mmol), and a few crystals
of hexamethylbenzene (used as the internal standard) were dissolved
in 1.0 mL of d6-benzene. The reaction was monitored by 1H NMR on
a Varian Mercury 400 at 60 °C. The ratio of ([3]t/[3]0) was determined
by the integration of the Cp proton signal of 3/the sum of the integration
of the Cp proton signals in both 3 and 12, assuming a 10% error for
NMR integrations.
X-ray Data Collection and Reduction. Full details are contained
in the Supporting Information. A colorless rhombic crystal of Cp2TiCl-
(MeO-TEMPO) (4) having approximate dimensions of 0.44 × 0.25 ×
0.04 mm3 was mounted on a quartz fiber using Paratone N hydrocarbon
oil. All measurements were made on a Bruker-Siemens SMART77 CCD
area detector with graphite monochromated Mo KR radiation (λ )
0.71073 Å). Data were integrated by the program SAINT with box
parameters of 1.0 × 1.0 × 0.6°. Of the 9114 reflections which were
collected, 3308 were unique (Rint ) 0.070). No decay correction was
applied. The structure was solved by direct methods78 and expanded
using Fourier technique.79 The non-hydrogen atoms were refined
Acknowledgment. We acknowledge the financial support
from the National Science Foundation (NSF-CHE 0097839).
We are indebted to Wei-Chen Chen at Caltech, Abhishek Dey
at Stanford, and Yuan-Chung Cheng and Kan-Lian Hu at MIT
for useful discussions. K.-W.H. acknowledges a Regina Casper
Stanford Graduate Fellowship. J.H.H. acknowledges an NSF
Graduate Fellowship. This paper is dedicated to the memory
of Mr. Chao-Chun Huang and Mrs. Yu-Mei Lin-Liu; and to
Professor Henry Taube on the occasion of his 90th birthday.
Supporting Information Available: Cartesian coordinates for
the calculated structures of 1a-c, 2-4, 6, and 7 at B3LYP/
6-31G*, synthesis, characterization, and X-ray crystallographic
data of 4, figures and tables giving a summary of kinetic data.
This material is available free of charge via the Internet at
http://pubs.acs.org.
JA044512F
(79) Beurskens, P. T.; Admiraal, G.; Beurskens, G.; Bosman, W. P.; de Gelder,
R.; Israel, R.; Smits, J. M. M. DIRDIF99; The DIRDIF-99 Program System,
Technical Report of the Crystallography Laboratory, University of
Nijmegen: The Netherlands, 1999.
(80) International Tables for Crystallography, Vol. C.; Kluwer Academic
Publishers: Dordrecht, The Netherlands, 1992; Tables 4.2.6.8 and 6.1.1.4.
(81) CrystalStructure 2.00: Crystal Structure Analysis Package, Rigaku and
MSC, 2001.
(82) Watkin, D. J.; Prout, C. K.; Carruthers, J. R.; Betteridge, P. W. CRYSTALS
Issue 10; Chemical Crystallography Laboratory: Oxford, U.K.
(77) Smart, Area-Detector Software Package; S. I. A., Inc.: Madison, WI, 1995.
(78) Sheldrick, G. M. SHELXS97: A Program for Crystal Structure Refinement;
University of Go¨ttingen: Go¨ttingen, Germany, 1997.
9
3816 J. AM. CHEM. SOC. VOL. 127, NO. 11, 2005