9,10-Dehydroanthracene: p-Benzyne-type biradicals abstract hydrogen unusually slowly

Article abstract of DOI:10.1021/ja960181y

The 9,10-dehydroanthracene biradical, a model for the p-benzyne-type biradicals implicated in DNA cleavage by the enediyne antitumor antibiotics, was prepared by photodissociation of a propellane in solution. Trapping products characteristic of biradicals, e.g. anthracene-d₂, are found. The rates of hydrogen abstraction by the biradical from acetonitrile and isopropyl alcohol are measured directly by flash photolysis/transient absorption spectroscopy, giving second-order rate constants of k(MeCN,abstr) = (1.1 ± 0.2) x 10³ M^-1 s^-1 and k(i-PrOH,abstr) = (6.5 ± 0.6) x 10³ M^-1 s^-1 at room temperature, which are 100-200 times lower than the corresponding rate constants for phenyl or 9-anthryl radical. A second decay route for the biradical is found, and assigned, based on thermochemical, kinetic, and trapping arguments to a retro-Bergman reaction that converts the 9,10-dehydroanthracene biradical into the ring-opened 3,4-benzocyclodeca-3,7,9-triene-1,5-diyne. Although the retro-Bergman reaction is relatively fast, k ~ 4 x 10⁵ s^-1 at room temperature, it is competitive with hydrogen abstraction by the biradical only because the hydrogen abstraction is slower than expected. Through-bond coupling in the 1,4-biradical is discussed as a rationalization for the 100- to 200-fold reduction in the abstraction rate.

Full text of DOI:10.1021/ja960181y

4896
J. Am. Chem. Soc. 1996, 118, 4896-4903
9,10-Dehydroanthracene: p-Benzyne-Type Biradicals Abstract
Hydrogen Unusually Slowly
Marc J. Schottelius and Peter Chen*
Contribution from the Laboratorium fu¨r Organische Chemie der Eidgeno¨ssischen of
Technischen Hochschule, Zu¨rich, Switzerland
ReceiVed January 19, 1996X
Abstract: The 9,10-dehydroanthracene biradical, a model for the p-benzyne-type biradicals implicated in DNA cleavage
by the enediyne antitumor antibiotics, was prepared by photodissociation of a propellane in solution. Trapping
products characteristic of biradicals, e.g. anthracene-d2, are found. The rates of hydrogen abstraction by the biradical
from acetonitrile and isopropyl alcohol are measured directly by flash photolysis/transient absorption spectroscopy,
giving second-order rate constants of kMeCN,abstr ) (1.1 ( 0.2) × 103 M-1 s-1 and ki-PrOH,abstr ) (6.5 ( 0.6) × 103
M-1 s-1 at room temperature, which are 100-200 times lower than the corresponding rate constants for phenyl or
9-anthryl radical. A second decay route for the biradical is found, and assigned, based on thermochemical, kinetic,
and trapping arguments, to a retro-Bergman reaction that converts the 9,10-dehydroanthracene biradical into the
ring-opened 3,4-benzocyclodeca-3,7,9-triene-1,5-diyne. Although the retro-Bergman reaction is relatively fast, k ≈
4 × 105 s-1 at room temperature, it is competitive with hydrogen abstraction by the biradical only because the
hydrogen abstraction is slower than expected. Through-bond coupling in the 1,4-biradical is discussed as a
rationalization for the 100- to 200-fold reduction in the abstraction rate.
Introduction
limiting in most of the mechanistic studies, there have been no
kinetic data pertaining to the subsequent chemistry of the
biradical. In particular, the rates of hydrogen atom abstraction
are not known, and have been commonly assumed to be the
same as those for the comparably substituted phenyl radicals.
The hydrogen abstraction rate gains added importance in light
of the proposed reversibility of the Bergman cyclization, which
would necessarily occur in kinetic competition with ring-opening
of the biradical. Motivated by the desire to prepare a p-benzyne-
like biradical by a fast photodissociation, so that subsequent
biradical kinetics could be observed, we have explored the
chemistry of 2. The route 2 f 1 has the additional attractive
features of (i) a good UV chromophore in 2 for photodissocia-
tion, (ii) easy synthesis from readily available starting materials,
(iii) facile modification with either σ or π substituents at sites
sterically remote from the reactive centers, and (iv) a good UV
chromophore in 1 for transient absorption spectroscopy. There
are two previous reports pertaining to 1. Chapman et al.8
reported the photochemical double decarbonylation of an-
thraquinodiketene, as a precursor to 1, from which an an-
thracene-like absorption spectrum (in a low-temperature glass)
and hydrogen and chlorine atom abstractions by the biradical
were seen. Neither rate data nor a mass balance were reported
in that work. Darby et al.9 reported anthracene in an attempted
preparation of 3,4-benzocyclodeca-3,7,9-triene-1,5-diyne (3),
presumably via the intermediacy of 1, as indicated by deuterium
incorporation at the 9,10 positions. As in the previous work,
no mass balance or rates were reported.
The chemistry of 1,4-biradicals structurally related to p-
benzyne has recently become the target of renewed activity,
following the realization that these biradicals are the likely
intermediates in the biological activity of the enediyne group
of antitumor antibiotics1 exemplified by calicheamicin, dyne-
micin, and esperamicin. In prior studies, the substituted
p-benzyne biradicals were generated by pericyclic closure of
cyclic enediynes,2 subsequent to a chemical triggering event,
to afford a trappable species. The process of biradical formation
has been thoroughly explored, with both strain energy3 and
spatial proximity4 of one terminus of the enediyne system to
the other as indices to which cyclization rate can be correlated.
Nevertheless, it has been commented upon that neither criterion
alone provides a satisfactory explanation of all of the extent
data.5 In further studies, benzannelation of the enediyne has
been reported variously to have either no effect5 or a marked
decelerating effect6 on the Bergman cyclization. Moreover,
experimental evidence for a kinetically-significant reversibility
of the Bergman cyclization in at least some cyclic systems has
recently emerged.6,7
Another issue important to the modification and rational
design of biradical-based therapeutic agents has remained wholly
unexplored though, despite the intense activity in this area.
Because thermal cyclization of the enediyne has itself been rate
X Abstract published in AdVance ACS Abstracts, May 1, 1996.
(1) Nicolaou, K. C.; Dai, W. M. Angew. Chem., Int. Ed. Engl. 1991, 30,
1387.
We report here the successful determination of absolute rates,
both for the hydrogen abstraction reaction of 1 from acetonitrile
and isopropyl alcohol, and for a competing kinetic process,
which we attribute to a retro-Bergman reaction in solution. We
find that, in agreement with qualitative valence bond consid-
erations and ab initio calculations10 at the CASPT2N level,
hydrogen abstraction by 1 is up to 100-200 times slower than
(8) Chapman, O. L.; Chang, C. C.; Kolc, J. J. Am. Chem. Soc. 1976, 98,
5703.
(9) Darby, N.; Kim, C. U.; Salau¨n, J. A.; Shelton, K. W.; Takada, S.;
Masamune, S. J. Chem. Soc., Chem. Commun. 1971, 1516.
(2) Bergman, R. G. Acc. Chem. Res. 1973, 6, 25.
(3) Snyder, J. P. J. Am. Chem. Soc. 1990, 112, 5367.
(4) Nicolaou, K. C.; Ogawa, Y.; Zuccarello, G.; Kataoka, H. J. Am. Chem.
Soc. 1988, 110, 7247. Nicolaou, K. C.; Zucarello, G.; Ogawa, Y.;
Schweiger, E. J.; Kumazawa, T. J. Am. Chem. Soc. 1988, 110, 4866.
(5) Grissom, J. W.; Calkins, T. L.; McMillen, H. A.; Jiang, Y. J. Org.
Chem. 1994, 59, 5833.
(6) Semmelhack, M. F.; Neu, T.; Foubelo, F. Tetrahedron Lett. 1992,
33, 3277. Semmelhack, M. F.; Neu, T.; Foubelo, F. J. Org. Chem. 1994,
59, 5038.
(7) Yoshida, K.; Minami, Y.; Otani, T.; Tada, Y.; Hirama, M. Tetrahe-
dron Lett. 1994, 35, 5253. Iida, K.; Hirama, M. J. Am. Chem. Soc. 1995,
117, 8875.
S0002-7863(96)00181-3 CCC: $12.00 © 1996 American Chemical Society

9,10-Dehydroanthracene
J. Am. Chem. Soc., Vol. 118, No. 20, 1996 4897
amounts of the analytes. The derived sensitivity factors were used to
convert integrated peak areas into true product ratios. For the
9-methoxyanthracene and 9-ethoxyanthracene versus anthracene or
2-methylanthracene, the GC/MS peak areas were calibrated against 1H
NMR integrations. Because of the trace, <0.5%, contamination of 2b
by anthracene photodimer, a small correction is included in the reported
product ratios. Ordinarily, the effect would be negligible for such a
small impurity, but the relative quantum yields for photodissociation
of anthracene photodimer (φ ) 0.55)12 versus 2b (φ ) 0.01) mean
that roughly 10% of the detected anthracene from samples of 2b derives
from the impurity. While the anthracene/2-methylanthracene ratio over
several GC/MS injections of a single sample is reproducible to within
1%, the run-to-run variability of the ratio in identically treated samples
is about 10%.
Scheme 1
that for the comparable monoradical (9-anthryl or phenyl in this
case). Furthermore, we find that the competing process,
assigned as ring-opening of 1 to 3,4-benzocyclodeca-3,7,9-
triene-1,5-diyne (3), is also fast, with a rate of k ≈ 4 × 105 s-1
at room temperature. A nucleophile-induced transannular
cyclization of strained enediynes is suggested as the origin for
the ionic trapping products previously attributed to dipolar
biradicals.
Flash photolysis experiments were performed at room temperature
with an Applied Photophysics (Leatherhead, Surrey, U.K.) laser kinetic
spectrometer, adapted to operate with a Spectra-Physics GCR-3G Nd3+
-
YAG laser. Transient absorption spectra with lifetimes in the range
from 10 ns to >500 µs, with OD changes of at least 5 × 10-4, can be
observed cleanly following laser photolysis at 266 nm (10 mJ/pulse).
The degassed solution, typically 80 µM concentration, was flowed
through a well-stirred quartz cuvette at a rate sufficient so that all
products from one laser shot are flushed away before the next shot.
Removal of the product arenes from each shot prior to the next was
found to be crucial in that both the anthracenes and their metastable
triplets have broad, intense UV/vis absorption spectra that mask the
spectral regions in which we needed to work. The role of diffusion
out of the observation zone for transient decays up to the millisecond
time scale could be excluded by tracking the temporal profile of the
251-nm anthracene band (ꢀ ) 220 000) following a single laser shot.
The single-shot conversion could also be determined from the absorption
at 251 nm to be 0.4% of the material in the irradiated volume. Transient
absorption spectra, taken pointwise, were signal averaged over 20 shots
at each wavelength to improve the signal-to-noise ratio. Each decay
rate is the average of approximately 15 repeated measurements.
Uncertainty bounds (1 σ) of about 10% were found for the decay rates,
measured on a given run. Repeated runs to determine run-to-run
reproducibility gave a comparable uncertainty bound.
Experimental Section
The synthetic route to 2 is modified from the procedure by
Applequist and Searle.11 Anthracene (for 2a) or 2-methylanthracene
(for 2b) and 9,10-dichloroanthracene were co-irradiated (Hanovia
450-W medium pressure Hg lamp/uranyl glass filter) in degassed
CH2Cl2 solution to afford the 4πs + 4πs adduct in moderate yield (63%
for anthracene and 50% for 2-methylanthracene after column chroma-
tography on silica gel). Reduction of the anthracene-dichloroan-
thracene adduct with triphenylmethyl sodium and subsequent recrys-
tallization of the product from superheated ether yields analytically pure
2a in 50% isolated yield as a stable, crystalline solid. 2a: mp 338-
342 °C dec; 1H NMR (500 MHz, CDCl3, 25 °C, TMS) δ 4.96 (s, 2H),
6.89 (m, 8H), 7.02 (dd, 4H), 7.14 (dd, 4H); MS (70 eV EI), m/z 354
(100) M+; UV/vis (MeCN) λmax (ꢀ) 283 (3818), 274 (5127), 269 (5194),
215 (60000). The analogous reduction of the 2-methylanthracene-
dichloroanthracene adduct proceeds in much lower yield, with the
consequence that the 2b does not crystallize from the reaction mixture
without an intermediate purification by column chromatography on
silica gel. Subsequent recrystallization from xylene affords 2b in 10%
yield with >99.5% purity, as judged by NMR and GC/MS. Curiously,
the only detectable impurity is the known anthracene photodimer, which
was not present in the mixture prior to chromatography. 2b: mp 253-
255 °C dec; 1H NMR (500 MHz, CDCl3, 25 °C, TMS) δ 2.15 (s, 3H),
4.92 (s, 1H), 4.93 (s, 1H), 6.71 (m, 1H), 6.91 (m, 6H), 6.99 (s, 1H),
7.04 (m, 5H), 7.14 (m, 2H); MS (70 eV EI), m/z 368 (100) M+; UV/
vis (MeCN) λmax (ꢀ) 280 (4200), 273 (5810), 215 (49600). Anal. Calcd
for C29H20: C (94.53); H (5.47). Found: C (94.30); H (5.58).
Bulk irradiation was done on degassed, 100 µM solutions of 2 at
266 nm (Spectra-Physics GCR-3G Nd3+-YAG, 6-ns pulse width, 20
Hz unfocused, 100 mW average power) in a vigorously stirred, long-
path cell set so that all of the incident UV laser light is absorbed.
Workup consisted of only filtration of the photolysate through a short
plug of silica gel to remove insoluble material resulting from the
photoreaction, followed by careful evaporation of solvent under reduced
pressure. Laser power measurements for quantum yield determination
were made with a Scientech (Arapahoe, CO) Model 390107 calibrated
laser pyrometer. The frequency-quadrupled YAG laser was used
instead of a more conventional light source because of the particular
constraints posed by 2. There exists a narrow spectral window around
270 nm in the absorption spectrum of anthracene or 2-methylanthracene
in which the arene itself would not be an internal filter. Fortunately,
2 has an absorption maximum near this window; at 266 nm, ꢀ for 2a
and anthracene are 4560 and ∼200, respectively, so electronic excitation
of anthracene or 2-methylanthracene (especially at low conversions of
2 to products) can be ignored. Fluorescence quantum yields were
measured against naphthalene as a standard in a Perkin-Elmer LS-50B
luminescence spectrometer and are uncorrected. GC/MS measurements
were performed in EI mode on a Fisons MD800 instrument equipped
with a DB-5 capillary column. Relative detection sensitivity for 2,
anthracene, and 2-methylanthracene in the GC/MS was determined
directly by injection of reference solutions containing equimolar
All solvents were spectroscopic grade and were checked by UV/vis
to ensure that there was no absorbing species at 266 nm.
Ab initio calculations were performed on an IBM RS/6000 Model
590 workstation using the Gaussian 92 and MOLCAS-3 suites of
programs.13 Given the biradical character of the species involved, we
judged it essential to use multideterminantal methods to obtain
chemically reliable results.14 Geometries were optimized at CASSCF-
(2×2), (4×4), and (6×6) levels with a relatively small basis set, 3-21G.
The energies were subsequently recomputed, at the 3-21G geometries,
with a much larger basis set, 6-31G*, and perturbative inclusion of
dynamic correlation effects. These latter CASPT2N/6-31G* calcula-
tions were done with the same (2×2), (4×4), and (6×6) active spaces.
The (2×2) active space included only the two nonbonding orbitals of
1. Adding one pair of σ and σ* orbitals involved in the through-bond
coupling15 produced the (4×4) active space. Inclusion of two σ and
σ* pairs yielded the (6×6) active space. A full description of the ab
initio calculations on 1 and its isomers will be given in a subsequent
(12) Yamamoto, S.; Grellman, K.-H.; Weller, A. Chem. Phys. Lett. 1980,
70, 241.
(13) Gaussian 92, Revision A; M. J. Frisch, G. W. Trucks, M. Head-
Gordon, P. M. W. Gill, M. W. Wong, J. B. Foresman, B. G. Johnson, H.
B. Schlegel, M. A. Robb, E. S. Replogle, R. Gomperst, J. L. Andres, K.
Raghavachari, J. S. Binkley, C. Gonzales, R. L. Martin, D. J. Fox, D. J.
DeFrees, J. Baker, J. J. P. Stewart, J. A. Pople; Gaussian, Inc.: Pittsburgh,
PA, 1992. MOLCAS-3; K. Andersson, M. P. Fu¨lscher, G. Karlstro¨m, R.
Lindh, P. A. Malmqvist, J. Olson, B. O. Roos, A. J. Sadlej, M. R. A.
Blomberg, P. E. M. Siegbahn, V. Kello¨, J. Noga, M. Urban, P. O. Widmark,
University of Lund, Sweden, 1994.
(14) Hrovat, D. A.; Morokuma, K.; Borden, W. T. J. Am. Chem. Soc.
1994, 116, 1072. Olivucci, M.; Bernardi, F.; Celani, P.; Ragazos, I.; Robb,
M. A. J. Am. Chem. Soc. 1994, 116, 1077.
(15) A recent review of experimental and theoretical work on through-
bond coupling may be found in: Paddon-Row, M. N.; Jordan, K. D. In
Modern Models of Bonding and Delocalization; Liebman, J. F., Greenberg,
A., Eds.; VCH Publishers: New York, 1988; Chapter 3.
(10) Logan, C. F.; Chen, P. J. Am. Chem. Soc. 1996, 118, 2113.
(11) Applequist, D. E.; Searle, R. J. Am. Chem. Soc. 1964, 86, 1389.

4898 J. Am. Chem. Soc., Vol. 118, No. 20, 1996
Schottelius and Chen
Figure 2. Representative total ion GC/MS trace for the photolysate
of 2b in isopropyl alcohol. The four peaks are identified (with retention
times in minutes) as anthracene (13.5), 2-methylanthracene (14.8),
9-isopropoxyanthracene (16.3), and 2b (22.3) by their 70-eV EI mass
spectra.
Figure 1. UV/vis absorption spectra for 2a in acetonitrile. The
extinction coefficient of 2a at 266 nm is 4560. See text for ꢀ at other
wavelengths. Successive spectra, taken at 30-s intervals during 266-
nm photolysis, are overlaid, showing the growth of the anthracene
spectrum. The anthracene band systems at 251 (ꢀ ) 220 000) and 375
nm (ꢀ ) 7400) would absorb incident broadband UV, necessitating
the use of monochromatic laser photolysis at 266 nm.
10 mL of solvent, run to 25% conversion, meant that there were
approximately 0.1 mg of products. The irradiation of 2a in
ethanol was repeated on a 50-fold larger scale to produce enough
9-ethoxyanthracene so that the 1H NMR spectra could be taken
to confirm that there was indeed only the ether and none of the
isomeric hydroxyethylanthracene. Irradiation of 2a in CH3OD
yielded, in addition to anthracene, an adduct with the mass
spectral fragmentation pattern consistent with 10-methoxyan-
thracene-9-d. The location of the deuterium in the product was
confirmed by 2H NMR to be exclusively the 9-position. These
experiments are important in that they connect the mass spectra18
to structures unambiguously determined by NMR. The trapping
products for all of the other experiments in alcohols could then
be identified with confidence by GC/MS alone. A representative
total ion GC/MS trace for the products of bulk photolysis is
shown in Figure 2 to illustrate the cleanliness of the photore-
action. The results of the trapping experiments are summarized
in Table 1. One should note that, in the photolyses of 2b, the
combined amounts of the anthracene and 9-alkoxyanthracene
products do not add up to the amount of 2-methylanthracene,
even for the photolysis in isopropyl alcohol. In that case, with
the best hydrogen-donating solvent19 used in this study, ap-
proximately 65% of the reactive intermediates produced in the
photodissociation were captured as isolable products.
publication, along with details of the potential surface for interconver-
sion between isomers.
Results
In the typical bulk photolysis experiment, irradiation of a
degassed 100 µM solution of 2a in acetonitrile at 266 nm affords
after 30 min 25% conversion of 2a to anthracene, as judged by
UV/vis spectra taken at regular intervals. The quantum yield
of anthracene production is relatively low: φ ) 0.02 and 0.01
from 2a and 2b, respectively. There is also a surprisingly large
fluorescence quantum yield of 0.13 in xylene-like emission
peaked at 297 nm (structureless with a fwhm of 26 nm) from
2a. The absorption spectrum of 2a prior to photolysis, and at
30-s intervals during photolysis, is shown in Figure 1. The
integrated UV absorption cross section of 2a together with the
fluorescence quantum yield yields an excited state lifetime16
for S1 of ∼1 ns. Given the typical arene intersystem crossing
rates17 of kST ∼ 106-107 s-1, it is unlikely that intersystem
crossing from the initially-produced S1 state occurs to a
significant extent. The photolyses in acetonitrile produced, in
addition to the arenes, an uncharacterizable polymeric material
which imparted a visible turbidity and yellow tint to the solution.
Photolysis of either 2a or 2b in aliphatic alcohols was much
In the flash photolysis experiment, two transients were
observed following photolysis of 2a at 266 nm. The transient
absorption spectra are shown in Figures 3 and 4. The same
two spectra were observed in both degassed CH3CN and
isopropyl alcohol, indicating that solvent-derived species were
not involved. The decays for both transients were first order20
and give lifetimes in CH3CN of τ ) 2.6 µs and 1.0 ms. In
1
cleaner with no visible turbidity or color change. By both H
NMR and GC/MS analysis, the only products (other than
unreacted 2) after workup at low-to-moderate conversions in
the reaction in alcohols were substituted anthracenes. Photore-
duction of the anthracene products by the alcohols was evident
only at high conversions. In any of the solvents, anthraquinone
was formed in large amounts if the solution was not carefully
degassed, so care was taken to exclude O2 from the reaction
mixtures. Identification of the products in the irradiations in
methanol, ethanol, and isopropyl alcohol was done principally
by GC/MS. This was necessitated by the experiments in
deuterated solvents, for which a 100 µM concentration of 2 in
(18) EI mass spectra with fragmentation patterns are also reported in
the literature for 9-methoxy- and 9-ethoxyanthracene: Gillis, R. G.; Long,
G. J.; Moritz, A. G.; Occolowitz, J. L. Org. Mass. Spectrosc. 1968, 1, 527.
Klasinc, L.; Komadinovic, S.; Stefanovic, D.; Guesten, H. Croat. Chem.
Acta 1976, 48, 109.
(19) The choice of hydrogen donors is limited by the need to photolyze
at 266 nm. For example, although 1,4-cyclohexadiene is transparent at the
photolysis wavelength, it is unavoidably contaminated with 1,3-cyclohexa-
diene and benzene, both of which absorb strongly enough at 266 nm to
cause problems when they are present at the ∼1% level in the solvent.
Benzylic donors such as cumene also cannot be used because of their ππ*
transitions in the near-UV.
(20) For the slowest decay, τ ) 1.0 ms, the decay could be followed out
for <2 lifetimes. Although the decay is first-order as far as we can
determine, we do not exclude the possibility that instrumental limitations
prevent us from seeing nonexponential behavior in the longest-lived
transient. Error bounds were determined on the decay rates rather than on
the lifetimes, and are reported with those rate measurements.
(16) Strickler, S. J.; Berg, R. A. J. Chem. Phys. 1962, 37, 814.
(17) Turro, N. J. Modern Molecular Photochemistry; Benjamin/Cum-
mings: Menlo Park, CA, 1978; p 111. Given the similarity of both the
absorption and dispersed fluorescence spectra of 2 to those of o-xylene,
the comparison should be a safe one.

9,10-Dehydroanthracene
J. Am. Chem. Soc., Vol. 118, No. 20, 1996 4899
Table 1. Product Ratios in the Photolyses of 2a and 2b in acetonitrile and aliphatic alcohols, As Determined by Calibrated GC/MS
Integrationsa
trapping agent (solvent)
2a
2b
CH3CN
CD3CN
CH3CN/H2O
CH3OH
CH3OD
An + polymer
An + polymer
An + polymer
An + 2-MeAn + polymer [1:18]
An-d2 + 2-MeAn + polymer [1:150]
An + MeOAn [3.2:1]
An + 2-MeAn + MeOAn [1:12:6.7]
An + MeOAn + MeOAn-d [8.8:1.7:1]
CD3OD
An-d2 + 2-MeAn + MeOAn-d4 [1:23:6.9]
An + 2-MeAn + EtOAn[1:7.3:5.2]
An + 2-MeAn + i-PrOAn [1:5.7:2.8]
An-d2 + 2-MeAn + i-PrOAn-d8 [1:10:3.0]
CH3CH2OH
(CH3)2CHOH
(CD3)2CDOD
An + EtOAn[2.3:1]
An + i-PrOAn [2.4:1]
a The products and abbrevations are as follows: anthracene (An), 2-methylanthracene (2-MeAn), anthracene-9,10-d2 (An-d2), 9-methoxyanthracene
(MeOAn), 10-methoxyanthracene-9-d (MeOAn-d), 10-perdeuteriomethoxyanthracene-9-d (MeOAn-d4), 9-ethoxyanthracene (EtOAn), 9-isoprop-
oxyanthracene (i-PrOAn), 10-perdeuterioisopropoxyanthracene-9-d (i-PrOAn-d8). Run-to-run reproducibility of the observed ratios is approximately
10%.
ns limit of the instrument, followed by a slow increase over
time. The rate of the increase is difficult to determine accurately
because it is a small increment on top of a large offset, but
single exponential fits yielded rate constants for the appearance
of anthracene after the initial 10 ns of approximately 2 × 104
and 5 × 104 s-1 in acetonitrile and isopropyl alcohol, respec-
tively. The reliability of these latter two values is difficult to
assess, but they are certainly good to within a factor of 2.
Relative energies for the three isomers, 1, 3, and 4, along
with the key bond lengths and complete geometry specifications
in Cartesian coordinates are given in Table 2. The computa-
tional results are preliminary in the sense that only the three
isomers are reported here. A complete study is underway to
compute the potential surface on which the three isomers would
interconvert. Nevertheless, for the purpose of facilitating
interpretation of the present experimental results, the relative
energies of 1, 3, and 4 are sufficient.
Figure 3. Transient absorption spectrum for the fast transient from
photolysis of 2a in isopropyl alcohol. The spectrum was taken 0.2 µs
after the laser shot.
Discussion
The trapping studies, in particular the observation of an-
thracene-d2 from 2b in perdeuterated methanol and isopropyl
alcohol, clearly support the intermediacy of the 9,10-dehydroan-
thracene biradical 1 in the photochemical reactions of 2 as a
qualitative conclusion. A more quantitative look at both the
trapping and flash photolysis data produces answers to three
important questions: (i) Can the rate of hydrogen abstraction
by 1 be extracted from the transient absorption? (ii) One
expected to see 1 as a transient in the flash photolysis
experiment. What is the second transient? It cannot be
9-anthryl radical, given the known rate of reaction (see below)
for that radical. (iii) What is the origin of the 9-alkoxyan-
thracene products that accompany anthracene formation in the
photolyses of 2 in aliphatic alcohols? One would not consider
them to be typical radical trapping products. The three questions
Figure 4. Transient absorption spectrum for the slow transient from
photolysis of 2a in isopropyl alcohol. The spectrum was taken 80 µs
after the laser shot to exclude any contribution from the fast transient.
will be addressed in sequence, followed by a rationalization of
the observed behavior on simple theoretical grounds.
As an answer to the first question, we identify the fast
isopropyl alcohol, the corresponding lifetimes were shorter, with
τ ) 2.1 and 500 µs for the “fast” and “slow” transients,
respectively. The spectrum of the slow transient is much weaker
than that of the fast transient, presumably due to a lower
extinction coefficient, and is partially overlapped by the 375
nm band system of anthracene produced by the same laser shot.
Accordingly, the lifetimes were taken at the high-energy side
of the band, where anthracene absorption is negligible. The
risetime of the fast transient is not detectable on the fastest time
scales available to us, and that for the slow transient at short
times is obscured by the partially overlapping absorption of the
fast transient. The temporal profile at 251 nm shows a large
jump in optical density that cannot be resolved within the 10-
transient in the flash photolysis experiment with the 9,10-
dehydroanthracene biradical 1, which means that the fast
transient decay can be related to the hydrogen abstraction rate
by 1 from acetonitrile or isopropyl alcohol. One starts with
the trapping results for irradiations of 2b in acetonitrile versus
isopropyl alcohol in Table 1. Anthracene clearly derives from
hydrogen abstraction from the solvent, and if there were no other
kinetically significant eVent, it would be formed in an 1:1 ratio
with 2-methylanthracene. Interestingly, the ratio of anthracene
to 2-methylanthracene differs markedly from 1:1, going from
1:18 to 1:5.7 as one goes from acetonitrile to isopropyl alcohol.
One writes the simple kinetic scheme below, ignoring for the
moment the actual identity of the competing process. Using

4900 J. Am. Chem. Soc., Vol. 118, No. 20, 1996
Schottelius and Chen
Table 2. Results of the ab Initio Calculations for 1, 3, and 4a
1
3
4
CAS(6×6)/6-31G*
-534.72002
0 kcal/mol
-536.49502
0 kcal/mol
CAS(6×6)/6-31G*
-534.72937
-5.9 kcal/mol
-536.48897
+3.8 kcal/mol
CAS(6×6)/6-31G*
-534.61089
+68.6 kcal/mol
-536.41588
CASPT2N/6-31G*
CASPT2N/6-31G*
CASPT2N/6-31G*
+49.7 kcal/mol
C9-C10
C9-C11
C11-C12
2.70 Å
1.37 Å
1.47 Å
C1-C2
C2-C3
C3-C4
C2-C5
C1-C6
1.21 Å
1.41 Å
1.43 Å
2.73 Å
3.00 Å
C9-C10
C9-C11
C11-C12
1.60 Å
1.40 Å
1.47 Å
C
C
C
C
C
C
C
C
C
C
C
C
C
C
H
H
H
H
H
H
H
H
0.00000
0.71410
-0.71410
1.40835
-1.40835
0.73547
-0.73547
1.35168
-1.35168
-0.73547
0.73547
-1.40835
1.40835
-0.71410
0.71410
1.23151
-1.23151
-2.47962
-2.47962
-2.47962
2.47982
3.63837
3.63837
2.48331
2.48331
1.22206
1.22206
0.00000
0.00000
-1.22206
-1.22206
-2.48331
-2.48331
-3.63837
-3.63837
4.57690
4.57690
2.48343
2.48343
C
C
C
C
C
C
C
C
C
C
C
C
C
C
H
H
H
H
H
H
H
H
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.69589
-0.69589
1.39215
-1.39215
0.71416
-0.71416
1.36380
-1.36380
-1.50185
1.50185
-1.66854
1.66854
-0.72660
0.72660
1.22871
-1.22871
2.46312
3.66474
3.66474
2.47787
2.47787
1.26055
1.26055
0.00399
0.00399
-1.19638
-1.19638
-2.59428
-2.59428
-3.54863
-3.54863
4.59438
4.59438
2.47524
2.47524
C
C
C
C
C
C
C
C
C
C
C
C
C
C
H
H
H
H
H
H
H
H
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
3.74724
-3.74724
3.74724
-3.74724
2.61404
-2.61404
2.61404
-2.61404
1.39558
-1.39558
1.39558
-1.39558
0.00000
0.00000
4.70170
-4.70170
4.70170
-4.70170
2.63581
-2.63581
2.63581
0.72412
0.72412
-0.72412
-0.72412
1.46375
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
1.46375
-1.46375
-1.46375
0.73367
0.73367
-0.73367
-0.73367
0.80000
-0.80000
1.21273
1.21273
-1.21273
-1.21273
2.53442
2.53442
-2.53442
-2.53442
-2.46312
-2.69073
2.69073
-1.11961
1.11961
-2.48343
-2.48343
-4.57690
-4.57690
-2.92738
-2.92738
-4.54937
-4.54937
-1.23151
1.23151
-2.63581
A Total energies (hartress) and relative energies (kcal/mol referenced to 1) are given for both CASSCF/6-31G* and CASPT2N/6-31G* calculations
with the same (6×6) active space. Selected bond lengths and complete Cartesian coordinates (in Å) are given for the geometries optimized at the
CASSCF/3-21G level of theory. The geometry optimizations for 1 and 3 were done with a (6×6) active space, while that for 4 was done with a
(2×2) active space. The geometry for 4 corresponds to a minimum in the CASPT2N/6-31G* energy, computed at a series of CAS(2×2)/3-21G
points describing the 1 f 4 conversion in steps of 0.1 Å in the C9-C10 distance (all other parameters were allowed to relax). The latter contrivance
was needed because, although 4 lies in a shallow well according to CASPT2N energies, there is no minimum at the CASSCF level of theory.
Scheme 2
molarity of neat acetonitrile, 19 M, and isopropyl alcohol, 13
M, the second-order rate constants for hydrogen abstraction by
1 are kMeCN,abstr ) (1.1 ( 0.2) × 103 M-1 s-1 and ki-PrOH,abstr
)
(6.5 ( 0.6) × 103 M-1 s-1. For the competing kinetic process,
which has up-to-now been simply designated as “other”, one
obtains absolute rates in acetonitrile and isopropyl alcohol by
subtracting the pseudo-first-order abstraction rates from the total
rates. For the reaction in the two solvents, one thus arrives at
kMeCN,other ) (3.6 ( 0.4) × 105 s-1 and ki-PrOH,other ) (4.0 (
0.5) × 105 s-1, indicating that, whatever the process is, it has
no pronounced solvent dependence. Examining the product
ratios for CH3CN versus CD3CN and (CH3)2CHOH versus
(CD3)2CDOD, one sees that the kinetic isotope effect on
hydrogen (deuterium) abstraction is much larger for acetonitrile
than for isopropyl alcohol. The variation of kH/kD follows the
expected trend for hydrogen transfer reactions as a function of
the position of the transition state. The kinetic isotope effect
is larger for a more symmetrical transition state,22 which, in
this case, means that as the C-H bond strength of the donor
approaches that of the arene C-H bond, the isotope effect
should increase. The position of the transition state, as measured
by the relative rates of H-abstraction, going from acetonitrile
to isopropyl alcohol, moves in the proper direction.23 The
deuterium kinetic isotope effect also gives the reason for our
the 2-methylanthracene as an internal reference, and taking the
hydrogen abstraction reaction from solvent to be pseudo first
order, the hydrogen abstraction rates can be derived by a
combination of the total decay rate and the product ratios. For
CH3CN and (CH3)2CHOH as solvent and hydrogen donor, the
lifetimes of the fast transient, τ ) 2.6 and 2.1 µs, respectively,
correspond to total decay rates of kMeCN,fast ) (3.8 ( 0.4) ×
105 s-1 and ki-PrOH,fast ) (4.8 ( 0.5) × 105 s-1. Assuming that
none of the 2-methylanthracene is removed, the ratio of
anthracene to 2-methylanthracene represents that fraction of the
total decay rate corresponding to hydrogen abstraction. For
CH3CN and (CH3)2CHOH, one therefore obtains (2.1 ( 0.5)
× 104 and (8.4 ( 0.8) × 104 s-1 as pseudo-first-order rates of
hydrogen abstraction in the two solvents. The match of these
two pseudo-first-order rates, each calculated from a decay of
the fast transient combined with a product ratio, to the rate of
appearance of anthracene absorption in the same solVent21
confirms that the fast transient is the biradical 1. Using the
(21) As can be inferred from the identification of the slow transient (see
later text), its longer lifetime and much lower ꢀ at 251 nm mean that only
that fraction of 1 which decays by hydrogen abstraction should contribute
to the growth of anthracene absorbance on the microsecond time scale.
(22) Bell, R. P.; Cox, B. G. J. Chem. Soc. B 1971, 783.

9,10-Dehydroanthracene
J. Am. Chem. Soc., Vol. 118, No. 20, 1996 4901
Scheme 3
CAS(6×6)/6-31G* level of theory, 4 lies 68.5 kcal/mol above
1. At the CASPT2N/6-31G* level, the difference falls to 49.7
kcal. It is evident that 4 is much higher in energy than 1, in
contrast to what one infers about the slow transient. It is
furthermore thermochemically inaccessible from 2 by photolysis
at 266 nm, based on an estimate for ∆Hf[2a] starting from the
∆Hf[anthracene] and ∆H for cleavage of anthracene photo-
dimer.27 The remaining candidate for the slow transient, 3,4-
benzocyclodeca-3,7,9-triene-1,5-diyne 3, is predicted by the
CAS(6×6)/6-31G* and CASPT2N/6-31G* calculations to lie
5.9 kcal/mol below and 3.8 kcal/mol above the biradical 1. At
this level of theory, one can only say (conservatively) that ab
initio calculations place 1 and 3 close in energy. The result is
qualitatively reasonable given that the aromaticity of 3 (which
is a [10]annulene) would at least partially offset the strain
incurred upon inclusion of the acetylenic units within a 10-
membered ring. As precedent, a retro-Berman reaction of
p-benzyne-type biradicals, in competition with hydrogen ab-
straction, has been recently reported in a model study7 of the
C-1027 chromophore. The experimental evidence, in that case,
for kinetically significant reversibility in the Bergman cyclization
is unequivocal for the enediyne moiety included in a 9-mem-
bered ring. Subsequent reexamination of the literature finds
comparable instances in some 10-membered rings as well.6
Accordingly, it is completely reasonable that 1 could go to 3 in
competition with hydrogen abstraction. From the recent
redetermination28 of the Arrhenius parameters for the parent
retro-Bergman reaction, p-benzyne going to the enediyne, we
have log A ) 14.1 and Ea ) 20.7 kcal/mol for the unelaborated
biradical. One needs to postulate a reduction of the Ea to
approximately 12-13 kcal/mol to bring the retro-Bergman
reaction rate of 1 into consistency with kMeCN,other as determined
in an earlier section. The 8 kcal/mol lowering of the transition
state for retro-Bergman reaction would imply, by microscopic
reversibility (using an estimated 2-3 kcal/mol energy difference
between 1 and 3 and the same Arrhenius preexponential factors),
that 3, if it were to be prepared independently, would have a
lifetime at room temperature of a few milliseconds with respect
to cyclization, making it unisolable under ordinary conditions.
Darby et al.,9 in their attempt to produce 3 by a double
elimination reaction, recovered only polymeric material and
anthracene derivatives, just as the rate estimates imply. With
the combination of results from ab initio calculation and
precedent in other enediyne/radical systems, we make the
assignment of the slow transient to 3,4-benzocyclodeca-3,7,9-
triene-1,5-diyne (3).
initial trouble in detecting anthracene-d2 in the photolysis of
2a in CD3CN. For acetonitrile, kH/kD ∼ 7.5 and contains a
primary plus two secondary deuterium isotope effects for a
relatively strong C-H(D) bond. Given that hydrogen abstrac-
tion from CH3CN already barely competes with the “other”
kinetic process, deuteration of the solvent would slow the
abstraction enough so that anthracene-d2 would be difficult to
detect in the presence of a much larger amount of anthracene
produced concurrently in the experiment. One further piece of
information can be extracted from the spectrum of the fast
transient. By comparison to the anthracene absorption at 251
nm, the molar extinction coefficient of 1 at λmax ) 295 nm can
be estimated as ꢀ ∼ 100 000.
To answer the second question posed by the results, we
propose that the slow transient is the ring-opened structure 3,
3,4-benzocyclodeca-3,7,9-triene-1,5-diyne. A priori, there are
three other obvious candidates, the ring-closed butalene24 isomer
4, the 9-anthryl radical, and some addition product25 of 1 and
2 (leading to polymer). We exclude the addition product by a
kinetic argument. The concentration of 2 is never above 100
µM, and is often lower. A bimolecular addition reaction with
a typical Ea of a few kilocalories per mole would proceed with
a pseudo-first-order rate of ∼ 103 s-1, which is far too slow to
be the process converting 1 to the slow transient. We can also
exclude 9-anthryl radical as a candidate for the slow transient
on kinetic grounds. The pseudo-first-order lifetime of 9-anthryl
radical in neat acetonitrile has been reported in an electrochemi-
cal study26 to be only 50 ns, which is much too short to be
compatible with the 1.0-ms lifetime of the slow transient in
acetonitrile. Because there is no evidence in the flash photolysis
experiment that the slow transient reverts to the fast transient
up to the millisecond time scale, one infers that the slow
transient is a species derived from 1 in an exothermic reaction.
With the conditions of the experiment, i.e. room temperature
and micro-to-millisecond time scales, the exothermicity of the
1-to-slow transient reaction need not be more than 2-3 kcal/
mol for any reversibility to become kinetically unimportant. Of
the remaining two candidates for the slow transient, 3 and 4,
we can now exclude 4 on thermochemical grounds. At the
Given that we believe the slow transient to be 3,4-benzocy-
clodeca-3,7,9-triene-1,5-diyne (3), the 9-alkoxyanthracene prod-
ucts from photolysis of 2a or 2b in aliphatic alcohols must be
derived from 3. The 2-fold reduction of the slow transient’s
lifetime, from 1.0 ms to 500 µs, going from acetonitrile to
isopropyl alcohol, is presumably the kinetic signature of the
reaction of 3 with ROH. Trapping of an alcohol on oxygen
instead of carbon suggests that the reaction is ionic as opposed
to radical-based. Addition of an alcohol across the in-plane
π-bond of a strained cycloalkyne is not too surprising. It has
been shown experimentally that, while cis-bending of an alkyne
is accompanied by almost no change in the in-plane π-HOMO
energy (as measured by the ionization potential in photoelectron
spectroscopy29), a marked lowering of the π*-LUMO occurs
(as measured by the electron affinity in electron transmission
(23) If one uses the para C-H bond in phenyl radical as a model, the
C-H bond strength in 9-anthryl radical at the 10-position would be 108
kcal/mol, using the ∆Hf,298[p-benzyne] ) 138 kcal/mol as reported
independently in two recent works: Wenthold, P. G.; Squires, R. R. J. Am.
Chem. Soc. 1994, 116, 6401; and Reference 28. Because the arene C-H
bond is stronger than any of the solvent C-H bonds, one would therefore
expect the relative rates for abstraction from acetonitrile and isopropyl
alcohol to be a measure of the symmetry of the transition state.
(24) Dewar, M. J. S.; Lee, W.-K. J. Am. Chem. Soc. 1974, 96, 5569.
Breslow, R.; Napierski, J.; Clarke, T. C. J. Am. Chem. Soc. 1975, 97, 6275.
Breslow, R.; Khanna, P. L. Tetrahedron Lett. 1977, 3429.
(25) We do not consider addition products of 1 to solvent, e.g. acetonitrile,
because the absorption spectrum of the slow transient is unchanged going
from acetonitrile to isopropyl alcohol. The next most abundant species in
solution to which a radical-like species could add would be unreacted
propellane 2.
(27) The required ∆H is assumed to be the same as that measured for
the photodimer of 9-cyanoanthracene of -20 kcal/mol: Guarini, G.; Sarti-
Pantoni, P. Mol. Cryst. Liq. Cryst. 1970, 6, 423. Dontai, D.; Guarini, G.;
Sarti-Pantoni, P. Mol. Cryst. Liq. Cryst. 1972, 17, 187.
(26) M’Halla, F.; Pinson, J.; Save´ant, J. M. J. Am. Chem. Soc. 1980,
102, 4120.
(28) Roth, W. R.; Hopf, H.; Horn, C. Chem. Ber. 1994, 127, 1765.

4902 J. Am. Chem. Soc., Vol. 118, No. 20, 1996
Schottelius and Chen
Scheme 4
and 9-anthryl radicals is kinetically unimportant. Accordingly,
the lowering of the hydrogen abstraction rate for 1 by two orders
of magnitude does not derive from steric problems, which leaves
electronic factors as the most likely cause. In a recent
publication,10 we have investigated the hydrogen abstraction
reactions of p-benzyne and phenyl radical from methanol by
ab initio theory at the CASPT2N/6-31G** level. It was found
that the rate for the biradical was depressed by a factor of
approximately 14 due to an increased activation energy relative
to the reference radical. An extra energetic increment in Ea,
corresponding to most of the singlet-triplet splitting in the
biradical, served to slow radical-like reactions by the singlet
ground-state biradical. Qualitatively, the energetic stabilization
of the singlet biradical by through-bond coupling must be repaid
as one “uncouples” (to use valence-bond language) the two
nominally nonbonding electrons in p-benzyne to perform an
abstraction reaction. The present results for the 9,10-dehy-
droanthracene biradical constitute the first clear experimental
verification of both the qualitative and computational prediction
that singlet biradicals are inherently poor hydrogen abstraction
agents. This generalization has not been commonly considered
in discussions of the action of the enediyne antitumor antibiotics.
For example, with such a slow abstraction rate, competing
reactions, like the retro-Bergman reaction in the current example,
become important. More interesting is the notion that the
biradical might live long enough to show selectivity in its
chemistry, either by way of differential rates of reaction with
different hydrogens or even possibly by way of equilibration
between different binding sites or orientations for the biradical.
It is important to realize that in p-benzyne itself, and the 9,10-
dehydroanthracene biradical, the two radical sites are identical
by symmetry, so that the only way in which the orbitals are
split, and a singlet ground state induced, is by either through-
space or through-bond coupling. In the natural products, the
p-benzyne biradicals are asymmetrically substituted, which, by
breaking the degeneracy of the orbitals, should give a larger
singlet-triplet splitting. The larger singlet-triplet splitting
would lead, in turn, to an even larger rate depression for
abstraction reactions. Structural changes which affect the
singlet-triplet splitting by increasing or decreasing through-
bond and/or through-space coupling, e.g. inductive electron
withdrawal on the σ-framework or steric compression of the
radical-to-radical distance, should allow a rational modulation
of hydrogen abstraction rates by remote substitution. The
anthracene framework of 1 lends itself well to a systematic
exploration of substituent effects on hydrogen abstraction rates.
These studies are underway.
spectroscopy30). The same effect appears in computational
studies on strained cycloalkynes.31 Accordingly, strained cy-
cloalkynes ought to be susceptible to attack by nucleophiles.
In the case of 3, nucleophilic attack can induce a transannular
cyclization, for which there is ample precedent32 in the reduction
of macrocyclic polyynes. Furthermore, two recent reports33,34
on macrocycles related to the enediyne series of compounds
find transannular cyclization products in which it could be
claimed that aromatization is triggered by nucleophilic addition
to a strained sp-hybridized site in the ring. Neutral water, a
stronger acid and a weaker nucleophile than the alcohols, does
not trap 3. (It would have presumably given anthrone).
Products from attack at what becomes the 9-position would be
favored if an initial nucleophilic attack were to be reversible.
One can speculate that the observation of ionic products in
neocarzinostatin and related systems34 arise not from “dipolar
biradicals”, i.e. zwitterionic cyclized structures, but rather from
nucleophile-induced transannular reactions. Because of the
symmetry of both 1 and 3, there can be no permanent charge
separation in the isolated molecules. Although neocarzinostatin
itself lacks that symmetry, it would be entirely reasonable that
it could react in the same way. It is important to note here that
1,4 σ,π-biradicals such as the R,3-dehydrotoluene biradical
studied by Myers et al.,35 in contrast to 1,4 σ,σ-biradicals,
probably do react via a dipolar structure36 made possible by
the much greater electronic difference between the two radical
sites.
Having suggested answers to the three questions posed at
the onset of this section, one should offer a rationale for why
the hydrogen abstraction rate for a p-benzyne-type biradical
should, in fact, be so slow. In the present study, we have
measured the second-order rate constants kMeCN,abstr ) (1.1 (
0.2) × 103 M-1 s-1 and ki-PrOH,abstr ) (6.5 ( 0.6) × 103 M-1
s-1. By way of comparison, the analogous second-order rate
constants for phenyl radical37 are (1.04 ( 0.18) × 105 and (1.38
( 0.07) × 106 M-1 s-1 for acetonitrile and isopropyl alcohol.
The one report of an absolute rate constant for hydrogen
abstraction by 9-anthryl radical,26 which one might argue is more
sterically hindered than phenyl, lists ∼106 M-1 s-1 for abstrac-
tion from CH3CN; clearly, any steric difference between phenyl
It remains only to comment upon some of the prior work
that pertained to this biradical and related systems. Two prior
reports, mentioned in the Introduction, claimed to have also
found evidence for the 9,10-dehydroanthracene biradical as a
reactive intermediate. Chapman et al.8 photolyzed 9,10-
anthraquinodiketene in low-temperature glasses and matrices,
observed radical trapping products upon warming to room
temperature, and most importantly, reported an UV/vis absorp-
tion spectrum that they assigned to the biradical. Both our
group38 and another group working independently39 have been
unable to produce a trappable biradical from either anthraquin-
odiketene or p-quinodiketene40 as photochemical precursors to
(29) Schmidt, H.; Schweig, A.; Krebs, A. Tetrahedron Lett. 1974, 1471.
Bieri, G.; Heilbronner, E.; Kloster-Jensen, E.; Schmelzer, A.; Wirz, J. HelV.
Chim. Acta 1974, 57, 1265. Carlier, P.; DuBois, J. E.; Masclet, P.; Mouvier,
G. J. Electron Spectrosc. 1975, 7, 55.
(30) Krebs, A.; Ru¨ger, W.; Ng, L.; Jordan, K. D. Bull. Soc. Chim. Belg.
1982, 91, 363. Ng, L.; Jordan, K. D.; Krebs, A.; Ru¨ger, W. J. Am. Chem.
Soc. 1982, 104, 7414.
(31) Several computational studies are reviewed by: Krebs, A.; Wilke,
J. Top. Curr. Chem. 1983, 109, 189.
(32) Bradshaw, J. D.; Solooki, D.; Tessier, C. A.; Youngs, W. J. J. Am.
Chem. Soc. 1994, 116, 3177. Gleiter, R.; Ritter, J.; Irngartinger, H.;
Lichtentha¨ler, J. Tetrahedron Lett. 1991, 32, 2883 and references therein.
(33) Toshima, K.; Ohta, K.; Ohtake, T.; Tatsuta, K. Tetrahedron Lett.
1991, 32, 391.
(34) Sugiyama, H.; Yamashita, K.; Fujiwara, T.; Saito, I. Tetrahedron
1994, 50, 1311.
(35) Myers, A. G.; Dragovich, P. S.; Kuo, E. Y. J. Am. Chem. Soc. 1992,
114, 9369.
(36) Concurrent biradical and ionic trapping products from R,3-dehy-
drotoluene, independently prepared by the photolysis of the (m-diazoni-
umphenyl)acetate inner salt support the observation of dual reactivity in
ref 35. Logan, C. F. Ph.D. Thesis, Harvard University, 1996.
(37) The most complete and consistent set of absolute rate constants for
phenyl radical come from: Scaiano, J. C.; Stewart, L. C. J. Am. Chem. Soc.
1983, 105, 3609. There are several other determinations as well which
give similar numbers: Janzen, E. G.; Nutter, D. E., Jr.; Evans, C. A. J.
Phys. Chem. 1975, 79, 1983. Kryger, R. G.; Lorand, J. P.; Stevens, N. R.;
Herron, N. R. J. Am. Chem. Soc. 1977, 99, 7589.
(38) Kim, Y.-M. Diplomarbeit, ETH Zu¨rich, 1995.
(39) Sander, W. Private communication.

9,10-Dehydroanthracene
J. Am. Chem. Soc., Vol. 118, No. 20, 1996 4903
the 1,4-biradicals. The absorption spectrum in the original
report, assigned in that work to 1, does not match the spectrum
for the fast transient in the present report. It more closely
resembles the spectrum of the slow transient, but the quality of
the data really does not permit a definitive statement. The report
by Darby et al.9 of trapping products from 1 in the attempted
preparation of 3 is, as discussed above, completely consistent
with the present experimental study. It would be interesting to
attempt to measure the rates coming from the direction 3 f 1
rather than 1 f 3 but that is beyond the scope of this
investigation. One final piece of prior work deserves comment.
Johnson and Bergman41 prepared a propellane, going by a longer
synthetic route involving the Ag+-catalyzed rearrangement of
bicyclopropenyls to Dewar benzenes, which would have given
anthracene and p-benzyne. Upon thermolysis, though, their
propellane underwent a complicated skeletal rearrangement to
benzfluoranthene, which was attributed to an initial Cope
rearrangement involving one of the isolated double bonds of
the propellane. Upon photolysis, evidence was found for
prismane formation by intramolecular 2πs + 2πs cycloaddition
of the two isolated double bonds in the propellane. The
difference between the propellane precursors 2 and the closely
related compound in Johnson and Bergman’s work occurs
because, in the dibenzo compounds such as 2, aromaticity in
the fused benzenoid moieties would be sacrificed in going
through either the Cope rearrangement or the intramolecular
cycloaddition, thus disfavoring the unproductive reactions.
Conclusion
We report the first experimental determination of absolute
rates of hydrogen abstraction by biradicals related to the
p-benzyne-based DNA-cleaving agents. In comparison to the
analogous rates by phenyl or 9-anthryl radical, the 9,10-
dehydroanthracene biradical abstracts hydrogens more slowly
by two orders of magnitude. The slow hydrogen abstraction
reaction allows other chemical processes to compete, namely a
kinetically significant reversibility of the Bergman reaction, for
which qualitative evidence has recently appeared. Rationaliza-
tion of the rate depression in biradicals in terms of through-
bond coupling suggests structural changes which may allow the
synthetic chemist to tune the reactivity of biradical-based DNA-
cleavage agents.
Acknowledgment. The authors would like to thank Prof.
Dr. W. H. Koppenol (ETH Zu¨rich) for the use of the solution-
phase flash photolysis spectrometer. Helpful suggestions from
Prof. W. v. E. Doering (Harvard) are also acknowledged. This
work was supported financially by the ETH Zu¨rich and the
Kontaktgruppe fu¨r Forschungsfragen (Basel).
(40) Marquardt, R.; Sander, W.; Laue, T.; Hopf, H. Liebigs Ann. 1995,
1643.
(41) Johnson, G. C.; Bergman, R. G. Tetrahedron Lett. 1979, 2093.
JA960181Y