Tuning the singlet-triplet energy gap in a non-Kekule series by designed structural variation. The singlet states of N-substituted-3,4-dimethylenepyrrole biradicals

Article abstract of DOI:10.1021/ja963113k

Semiempirical quantum chemical calculations (AM1/CI and PM3/CI) confirm the qualitative perturbational prediction that electron-withdrawing groups on the ring nitrogen of a 3,4-dimethylenepyrrole should diminish the energy separation of the singlet and triplet states to near zero. Syntheses of a series of precursors of such biradicals have been developed. Study of the chemistry and spectroscopy of the biradicals has revealed persistent singlet states for the cases where the substituent is methyl, isobutyryl, and pivaloyl. In the cases of N-arenesulfonyl-3,4-dimethylenepyrroles, both a singlet and a triplet form can be observed as persistent species. In this paper, the properties of the singlets in this series are described. Although energy transfer from the excited triplet state of the sensitizer xanthone to the diazene precursor of N-p-toluenesulfonyl-3,4-dimethylenepyrrole is observed by nanosecond time-resolved spectroscopy, the chemical behavior of the biradical intermediate is the same as that observed in the direct photolysis or thermolysis of the diazene. The reactive form of the biradical under these conditions appears to be the singlet.

Full text of DOI:10.1021/ja963113k

1406
J. Am. Chem. Soc. 1997, 119, 1406-1415
Tuning the Singlet-Triplet Energy Gap in a Non-Kekule´ Series
by Designed Structural Variation. The Singlet States of
N-Substituted-3,4-dimethylenepyrrole Biradicals
Linda C. Bush,1a,c Richard B. Heath,1a Xu Wu Feng,1a Patricia A. Wang,1a
Ljiljana Maksimovic,1a Anne In Song,1a Wen-Sheng Chung,1a Alain B. Berinstain,1b
J. C. Scaiano,*,1b and Jerome A. Berson*,1a
Contribution from the Department of Chemistry, Yale UniVersity,
New HaVen, Connecticut 06520-8107, and the Ottawa-Carleton Chemistry Institute,
UniVersity of Ottawa Campus, Ottawa, Canada K1N 6N5
ReceiVed September 4, 1996X
Abstract: Semiempirical quantum chemical calculations (AM1/CI and PM3/CI) confirm the qualitative perturbational
prediction that electron-withdrawing groups on the ring nitrogen of a 3,4-dimethylenepyrrole should diminish the
energy separation of the singlet and triplet states to near zero. Syntheses of a series of precursors of such biradicals
have been developed. Study of the chemistry and spectroscopy of the biradicals has revealed persistent singlet
states for the cases where the substituent is methyl, isobutyryl, and pivaloyl. In the cases of N-arenesulfonyl-3,4-
dimethylenepyrroles, both a singlet and a triplet form can be observed as persistent species. In this paper, the properties
of the singlets in this series are described. Although energy transfer from the excited triplet state of the sensitizer
xanthone to the diazene precursor of N-p-toluenesulfonyl-3,4-dimethylenepyrrole is observed by nanosecond time-
resolved spectroscopy, the chemical behavior of the biradical intermediate is the same as that observed in the direct
photolysis or thermolysis of the diazene. The reactive form of the biradical under these conditions appears to be the
singlet.
Experimental Design. Tuning the Multiplet Gap in
N-Substituted-3,4-dimethylenepyrrole Biradicals. Prediction
of the spacings of the lowest energy multiplets of π-conjugated
non-Kekule´ molecules is a stringent test of theory. Although
in most cases these separations are difficult to measure, the effect
of structure on the qualitative trends of their direction and
magnitude should be estimable by theoretical calculation. We
are particularly attracted to the use of computationally designed
systematic structural variations as means of tuning the spacings
and hence providing experimental tests of such estimates. The
necessary elements of this approach are a parent system with
near-degenerate multiplet states, a variable structural feature that
can be used to perturb their separation, and a theoretical means
for selecting potentially fruitful structural alterations. Since the
singlet-triplet gap in such species can be made quite small,
the tuning process ultimately might offer the power of choice
over the multiplicity of the ground state.
Figure 1. Perturbative interaction of filled heteroatomic p-π orbital
with the NBMOs of TME.4
The present paper describes the preparation and characteriza-
tion of the singlet states of several such biradicals, and an
accompanying paper11b reports results of the search for the
corresponding triplet states.
Our present computational approach was implicit in theoreti-
cal considerations by Du, Hrovat, and Borden,4 which were
further developed in later corollaries.3,5,6 Figure 1, a diagram-
The disjoint hydrocarbon biradical tetramethyleneethane
(TME, 1) provides a useful starting structure for such a process.
Although the size of the energy separation between the lowest
triplet and singlet states of TME is not yet known from
experiment, theory2 predicts a very small gap (ET - ES ∼ -1
to +2 kcal/mol). A heteroatom bridge connecting two termini
of the TME molecule, as in previous work3 from this laboratory
on 3,4-dimethylenefuran and 3,4-dimethylenethiophene, intro-
duces a variable source of electronic perturbation.
(3) (a) Berson, J. A. Mol. Cryst. Liq. Cryst. 1989, 176, 1 and references
cited therein. (b) Stone, K. J.; Greenberg; M. M.; Blackstock, S. C.; Berson,
J. A. J. Am. Chem. Soc. 1989, 111, 3659. (c) Greenberg, M. M.; Blackstock,
S. C.; Stone, K. J.; Berson, J. A. J. Am. Chem. Soc. 1989, 111, 3671. (d)
Greenberg; M. M.; Blackstock, S. C.; Berson, J. A. Tetrahedron Lett. 1987,
28, 4263. (e) Haider, K. W.; Clites, J. A.; Berson, J. A. Tetrahedron Lett.
1991, 32, 5305.
(4) Du, P.; Hrovat, D. A.; Borden, W. T. J. Am. Chem. Soc. 1986, 108,
8086.
(5) Lahti, P. M.; Ichimura, A. S.; Berson, J. A. J. Org. Chem. 1989, 54,
958.
X Abstract published in AdVance ACS Abstracts, February 1, 1997.
(1) (a) Yale University. (b) Ottawa-Carleton Chemistry Institute. (c)
Current address: Department of Chemistry, Salisbury State University,
Salisbury, MD 21801.
(2) (a) Borden, W. T.; Davidson, E. R. J. Am. Chem. Soc. 1977, 99,
4587. (b) Du, P.; Borden, W. T. J. Am. Chem. Soc. 1987, 109, 930. (c)
Nachtigall, P.; Jordan, K. D. J. Am. Chem. Soc. 1992, 114, 4743. (d)
Nachtigall, P.; Jordan, K. D. J. Am. Chem. Soc. 1993, 115, 270.
(6) Blackstock, S. C., Unpublished results, Yale University, 1990.
S0002-7863(96)03113-7 CCC: $14.00 © 1997 American Chemical Society

Singlet-Triplet Energy Gap in a Non-Kekule´ Series
J. Am. Chem. Soc., Vol. 119, No. 6, 1997 1407
matic display of the concept, shows the effect of a heteroatom
on the frontier orbitals of the disjoint hydrocarbon tetrameth-
yleneethane (TME, 1). Note that the symmetric component of
the TME frontier orbitals is of the proper symmetry to mix with
the pz orbital of the heteroatom X. Thus, in the 3,4-dimethyl-
eneheterocycle, e.g., 2a,b and 3a-h, the relative energies of
the original TME NBMOs (ψA, ψS) are perturbed by an amount
that, to first order, will be inversely proportional to the energy
separation between one of them and the lone-pair pz orbital.
An oversimplified but graphic way to think about the
perturbation and its consequences starts with the recognition4
that the orbital (ψS) that is of the proper symmetry to interact
with the pz orbital is initially lower than ψA, which is not
perturbed by pz. Therefore, the perturbation of ψS by pz, if weak,
could bring ψS and ψA closer together than they are in TME
itself, thereby tending to favor a triplet ground state, but if
strong, could drive ΨS enough higher than ψA to favor a singlet.
Both of the previous examples (2a,b) of heterocyclic TME
derivatives have singlet ground states whose chemistry is
consistent with an S-above-A ordering of the frontier orbitals,3
which suggests that the perturbations by O and S in the furan
and thiophene cases is of the strong type.
Figure 2. Calculated vertical and optimized triplet-singlet energy
separation (relative to that of TME set as zero) as a function of the
LUMO π-electron coefficient at N for a series of N-substituted-3,4-
dimethylenepyrroles and other biradicals. The values were obtained
with AM1, OPEN (4,4)CI. The points 1-3 represent compounds as
follows: (1) TME, 1; (2) DCP, 6; (3) 3,4-dimethylenepyrrolium ion,
7. Points 4-10 represent N-substituted-3,4-dimethylenepyrroles with
substituents as follows: (4) p-nitrobenzoyl; (5) p-benzoylbenzoyl; (6)
benzoyl; (7) acetyl; (8) cyano; (9) methyl; (10) hydrogen. The curve is
merely a smooth connection of the points.
In the case of the pyrrole TME derivative 3a, one would
expect (and calculations4-6 confirm) that the perturbation would
be even stronger than in the furan and thiophene systems.
However, the functional nature of the NH group in the pyrrole
offers a special opportunity to manipulate the electronegativity
of the heteroatom by replacing H with electron-releasing or
electron-withdrawing groups (ERG or EWG) and hence to tune
the singlet-triplet gap by variations of the molecular structure.
EWGs, for example, would lower the energy of the lone pair
pz orbital in Figure 1 and thereby decrease the perturbation of
the symmetric component of the TME NBMOs. A sufficiently
EW substituent should result in a triplet-singlet gap similar to
those in hydrocarbon TMEs. Three examples of such hydro-
carbons exist: TME (1) itself,7 as well as two derivatives, 2,3-
dimethylenecyclohexa-1,3-diene (4)8 and 5,5-dimethyl-2,3-
dimethylenecyclopenta-1,3-diene (5).9 Experimentally, all three
gap or even reverse the preference and make the triplet the
ground state. A goal of the present work is to find examples
of this inversion of spin multiplet energies and thereby
demonstrate the predicted perturbative effect.
AM1/CI Calculations. Semiempirical calculations with
configuration interaction (CI)3c,5,10 can provide quantitative
estimates for the perturbative effect of the N-substituents on
the separation of the multiplet energies. We now11 have applied
the AM1/CI method12 to this problem in the cases of the pyrrole
series shown in Figure 2. In order to avoid the interaction of
configurations containing σ-MOs, which usually led to spin
contamination of the states and impeded standardization of the
CI level across the series, we chose as CI space all configura-
tions in which four valence electrons freely occupied four
orbitals around the nonbonding level (two above and two
below). This reference state, OPEN(4,4), provided a reasonable
level of CI while keeping the interaction restricted to π-MOs.
The effect of geometry was studied by performing two sets
of calculations of the multiplet separation (ET - ES), one for
the “vertical” separation of the states when both the singlet and
triplet are assumed to have the geometry optimized for the spin
of these compounds have been shown to have thermally
accessible and persistent triplet states. Since in the parent NH
case 3a,3-6 the singlet is predicted to lie only slightly below
the triplet, strong EWGs should be able to narrow the multiplet
(10) Lahti, P. M.; Rossi, A. R.; Berson, J. A. J. Am. Chem. Soc. 1985,
107, 2273.
(11) (a) Preliminary communication: Bush, L. C.; Heath, R. B.; Berson,
J. A. J. Am. Chem. Soc. 1993, 115, 9830. (b) Bush, L. C.; Maksimovic, L.;
Feng, X. W.; Lu, H. S. M.; Berson, J. A. Ibid. 1997, 119, 1416
(accompanying paper). (c) Lu, H. S. M.; Berson, J. A. Ibid. 1996, 118,
265. (d) Lu, H. S. M.; Berson, J. A. Ibid. 1997, 119, 1428 (accompanying
paper).
(12) (a) Stewart, J. P. P. J. Comput. Chem. 1989, 10, 209. (b) Stewart,
J. P. P. J. Comput. Chem. 1989, 10, 221. (c) Dewar, M. J. S.; Zoebisch, E.
G.; Healy, E. F.; Stewart, J. J. P. J. Am. Chem. Soc. 1985, 107, 3092.
(7) (a) Dowd, P. J. Am. Chem. Soc. 1970, 92, 1066. (b) Dowd, P.; Chang,
W.; Paik, Y. H. J. Am. Chem. Soc. 1986, 108, 7416.
(8) Dowd, P.; Chang, W.; Paik, Y. H. J. Am. Chem. Soc. 1987, 109,
5284.
(9) Roth, W. R.; Kowalczik, U.; Maier, G.; Reisenauer, H. P.; Sustmann,
R.; Mu¨ller, W. Angew. Chem., Int. Ed. Engl. 1987, 26, 1285.

1408 J. Am. Chem. Soc., Vol. 119, No. 6, 1997
Bush et al.
very close to that calculated (point 2, -1.0 kcal/mol) for the
hydrocarbon DCP 6. The effect of geometry optimization for
the most EWG-substituted cases (points 4-6, Figure 2) is
mainly to lower the (ET - ES)rel values.
Both of the curves of Figure 2 decline to nearly flat plateaus
in the region of (ET - ES)rel ∼ 0 to -1 kcal/mol. Thus, in the
context of the current work, the main message to be read from
Figure 2 is that an N-alkyl group (cf. methyl, point 9) should
result in a singlet 3,4-dimethylenepyrrole, and an acetyl sub-
stituent (point 7) may be too weakly electron-withdrawing to
overcome the preference for a singlet, but 3,4-dimethylenepyr-
roles with EWG N-substituents as strong as p-nitrobenzoyl
(point 4) or stronger should produce an accessible triplet state.
As tests of these predictions, we have synthesized and studied,
with varying degrees of scrutiny, a number of N-substituted-
3,4-dimethylenepyrroles, including the methyl and benzyl
derivatives 3b,c as examples of cases in which a singlet state
is clearly predicted, the isobutyryl and pivaloyl derivatives 3d,e
as examples of moderately EWG cases, and the p-toluenesulfo-
nyl and p-bromobenzenesulfonyl derivatives 3f,g as examples
bearing strong EWGs. We chose the arenesulfonyl derivatives
instead of the p-nitrobenzoyl derivative because of anticipated
experimental problems in handling the latter hydrolytically
active amide. Because the AM1/CI program was not param-
eterized for sulfonyl functionality, we could not confirm
computationally the expected location of the arenesulfonyl points
in the plateau region of Figure 2. Our working hypothesis that
arenesulfonyl substituents would be sufficiently strong EWGs
was based on the assumption that the EW power of the
substituent might be estimated roughly by comparison of the
acidities of the corresponding acids, which show arenesulfonic
acids (pKa -7) to be much stronger even than p-nitrobenzoic
acid (pKa 3.4).
Figure 3. Structure of N-benzenesulfonyl-3,4-dimethylenepyrrole
biradical as calculated by PM3/CI. Values of bond lengths and angles
are given in the text.
state of lower energy, and one for the states with geometries
individually optimized. In all of the pyrroles, the nitrogen and
its ligand bonds were found to be planar within a few degrees.
Whether the actual nitrogen configuration is slightly pyrami-
dalized cannot be answered with confidence from the AM1/CI
calculations, since the results seemed to depend upon the level
of CI used. In all the calculations, the CH2 groups were assumed
to be coplanar with the heterocyclic ring. Justification of this
assumption is given elsewhere.11c
Figure 2 provides gratifying computational confirmation of
the qualitative perturbational ideas. The LUMO coefficient on
nitrogen is taken as a measure of the mixing of the nitrogen
lone pair orbital with the TME NBMOs (see Figure 1). The
shapes of the curves of Figure 2 are in accord with the
expectation that this mixing should decrease as the N-substituent
becomes more electron-withdrawing and that, ultimately, the
multiplet separation should approach those of the parent
hydrocarbons TME (1) and 2,3-dimethylenecyclopenta-1,3-diene
(DCP, 6). As a model for a 3,4-dimethylenepyrrole with the
TME NBMO-N: interaction fully switched off, we used the
hypothetical N-protonated pyrrolium ion species 7.
Molecular Geometries by PM3/CI Calculations. We were
able to calculate molecular geometries for the benzenesulfonyl
derivative by means of another semiempirical algorithm, PM3/
CI,12 which is parameterized for this group. As calculated by
PM3/CI, both the singlet and triplet states of the N-benzene-
sulfonylpyrrole biradical prefer the conformation shown in
Figure 3. The phenyl and pyrrole rings lie in planes related by
the N-S-phenyl ipso C angle of ∼100°. This conformation is
close to that determined13 a by single crystal X-ray analysis for
N-benzenesulfonylpyrrole, which also shows the rings in nearly
perpendicular planes, with the N-S-phenyl ipso C angle of
∼105°. In the biradical, the plane defined by these three atoms
bisects the O-S-O angle, placing the oxygens on one side of
the plane defined by the nitrogen and its attached carbons. The
latter structural feature matches that calculated13b by ab initio
methods for N-methylmethanesulfonamide, in which the oxy-
gens lie on one side of the H-N-methyl C plane. It should
be noted that although the PM3/CI calculations on N-benzene-
sulfonyl-3,4-dimethylenepyrrole again show little or no pyra-
midalization at nitrogen, the ab initio calculation13b of N-meth-
ylmethanesulfonamide predicts that a pyramidalized config-
uration for nitrogen is preferred in that case.14
The absolute value of the multiplet gap (ET - ES) calculated
for TME by the AM1/CI procedure is 7 kcal/mol, which is
substantially larger than that obtained by ab initio methods. As
we have noted above, triplets have been observed for TME and
two derivatives, so that it is likely that the actual gap is small.
For the purpose of scaling, we assume that the predictions of
the gap by ab initio methods are more accurate than the
semiempirical ones, and we arbitrarily assign the gap for TME
as 0.0 kcal/mol (see Figure 2) by reference to the small
calculated2 ab initio value. The absolute values of (ET - ES)
calculated by AM1/CI for the other molecules then are
diminished by 7 kcal/mol to give (ET - ES)rel, thus scaling them
to the TME ∼ 0 standard. This step cannot be defended
rigorously, since there is no guarantee that the departure of the
AM1/CI calculated value from the true value is the same in all
cases. However, some confidence in the validity of the
procedure is provided by the reasonably close match of the
scaled AM1/CI value of (ET - ES)rel for the NH case 3a, 8.4
kcal/mol, with the MRSD CI ab initio absolute (ET - ES) value,
6.8 kcal/mol.4 Nevertheless, although the resulting curves of
Figure 2 are anchored by reference to ab initio values at both
ends, they should be interpreted as indicative of trends rather
than as quantitative predictions of the multiplet gaps.
The experimental studies now have demonstrated the exist-
ence of persistent singlet states for the N-methyl-, N-isobutyryl-,
and N-pivaloyl-3,4-dimethylenepyrroles and of persistent triplet
states for the N-tosyl and N-brosyl compounds. The latter two
findings bring welcome confirmation of the validity of the
experimental design, but an unexpected complication has
Figure 2 (point 3) shows that in the pyrrolium ion 7, the
N-protonation indeed effectively has turned off the TME
NBMO-N lone-pair interaction, as would be expected intuitively.
(ET - ES)rel is about -1.5 kcal/mol for this biradical, a value
(13) (a) Beddoes, R. L.; Dalton, L.; Joule, J. A.; Mills, O. S.; Street, J.
D.; Watt, C. I. F. J. Chem. Soc., Perkin Trans. 2 1986, 787. (b) Bindal, R.
D.; Golab, J. T.; Katzenellenbogen, J. A. J. Am. Chem. Soc. 1990, 112,
7861.

Singlet-Triplet Energy Gap in a Non-Kekule´ Series
J. Am. Chem. Soc., Vol. 119, No. 6, 1997 1409
Scheme 1a
Scheme 2a
a Methods: (1) KOt-Bu, 18-Crown-6, THF, 63%; (2) TsCl,
Bu4NHSO4, KOH, CH2Cl2, 91%; (3) DIBAL, CH2Cl2, 77-91%; (4)
Ph3PCl2, CH2Cl2, 78%; (5) N2H2(CO2t-Bu)2, NaH, DMSO, 87%;
alternatively, 14f can be obtained in lower yield (67%) but more
conveniently by using KOt-Bu in THF as the base.
a Methods: (1) KOH, MeOH-THF, 70-80%; (2) N-acyl- or
N-arenesulfonylimidazole, Cs2CO3, 90%; (3) HCl (g), Et2O, 85-100%;
(4) DMAD.
The tert-butyl ester groups of 14f or of the other N-substituted
analogues can be removed with hydrogen chloride in diethyl
ether (Scheme 2) to give the hydrazines 16d-f, which can be
converted to the N-isobutyryl (17d), N-pivaloyl (17e), N-tosyl
(17f), N-brosyl (17g), or potentially other diazenes by the
usual3b,c dimethyl azodicarboxylate (DMAD) dehydrogenation
procedure.
emerged in that we also have found strong evidence of the
existence of persistent singlet states of the two arenesulfonyl
derivatives.
The present paper describes the properties of these singlet
biradicals. Experimental documentation is given in three
sections: synthesis of the biradical precursors and generation
of the biradicals; preparative reaction chemistry of the biradicals
and its implications for assignments of the spin multiplicity of
the reactive transients; and spectroscopic characterization of the
biradicals. An accompanying paper11b deals with the triplet
states.
Synthesis of N-Substituted-3,4-dimethylenepyrroles. Our
first samples of the desired diazene precursors 17a of N-sub-
stituted-3,4-dimethylenepyrroles were obtained by rather im-
practical routes described elsewhere.11a,17b Currently we prefer
the synthetic approach shown in Scheme 1. It is based upon
the ready availability of 3,4-dicarboethoxypyrrole 10 from the
p-tosylmethyl isocyanide-fumaric ester addition15-17 and in-
corporates as a crucial modification the use of the N-tosyl group
in the intermediate 13 to suppress the heterolytic reactivity of
the chloromethyl groups. Specific methods (Schemes 1 and 2)
for the conversion of 10 via 13 to the key intermediate
N-tosylpyrrolodihydropyridazine biscarbamate 14f and for the
replacement of the tosyl group of 14f by other N-substituents
permit its application to a variety of derivatives.
Chemistry of the N-Substituted-3,4-dimethylenepyrrole
Biradicals from Diazenes 17. Thermal and Direct Photo-
chemical Generation and Capture. The chemical interception
experiments now to be described were intended to answer two
questions: (1) Does the deazetation of a diazene 17 lead to a
capturable transient 3,4-dimethylenepyrrole biradical intermedi-
ate of structure 3? (2) If so, is this same intermediate the carrier
of the visible color and UV-vis spectrum observed spectro-
scopically by the methods described below? For question 1,
we applied preparative and kinetic procedures used3 earlier in
the cases of the analogous biradicals 3,4-dimethylenefuran 2a
and 3,4-dimethylenethiophene 2b. Question 2 also can be
addressed through an established method.3,17,19,20 Relative
reactivities of a series of trapping olefins, as determined by
preparative competition experiments, constitute a characteristic
fingerprint for the reactive intermediate. An independent
fingerprint is based upon the reactivities determined by direct
time-resolved spectroscopic observation of the quenching of a
transient chromophore. If the two fingerprints match, the
evidence becomes persuasive that the carrier of the spectroscopic
signal truly has the structure of the preparatively significant
intermediate.
Although much of our detailed knowledge of these biradicals
comes from photochemical studies (see below), we also have
generated what probably are the same species by thermal
deazetation of the corresponding diazenes. The thermal stability
(and hence the ease of manipulation) of these precursors is a
sensitive function of the nature of the N-substituent. For
example, the N-methyl diazene 17b is unstable above -60 °C,
but the N-tosyl diazene 17f is reasonably stable at 0 °C. As
monitored by 1H NMR spectroscopy, tosyl diazene 17f decom-
Detosylation of 14f is effected smoothly by direct saponifica-
tion with KOH in methanol-THF. The key to the latter process
is the resistance of the tert-butyl ester functions of the
biscarbamate 14f to alkaline hydrolysis, which permits the
selective removal of the tosyl group.18 Introduction of other
substituents on the nitrogen then can be achieved by acylation
with an N-acylimidazole in the presence of cesium carbonate.
(14) The PM3/CI calculations of (ET - ES)rel gave a plot very similar in
shape to that obtained with AM1-CI. Again, EW substituents caused a steep
diminution of the multiplet energy gap, and the curve reached a plateau of
small-to-zero gap. However, the restriction to π-orbitals in the CI space
was difficult to achieve in this set of calculations, and the real significance
of the (ET - ES)rel results is dubious. For example, a closer inspection reveals
that although the shape of the PM3/CI and AM1-CI curves are similar, the
actual order of the points is disturbingly different. Thus, although we had
expected arenesulfonyl to be more effective than acetyl in diminishing the
preference for the singlet, the PM3-CI calculations reversed this order,
predicting that acetyl would be more effective than benzenesulfonyl.
Fortunately, our experimental results eventually showed that arenesulfonyl
was better for our purpose than aliphatic acyl, although the reasons for this
turned out to be more intricate than the simple EWG considerations that
had guided our earlier thinking.
(15) Arnold, D. P.; Nitschinski, L. J.; Kennard, C. H. L.; Smith, G. Aust.
J. Chem. 1991, 44, 323.
(16) Cf.: Van Leusen, A. M.; Sidering, H.; Hougenboom, B. E.; Van
Leusen, D. Tetrahedron Lett. 1972, 5337.
(17) (a) Heath, R. B.; Bush, L. C.; Feng, X. W.; Berson, J. A.; Scaiano,
J. C.; Berinstain, A. B. J. Phys. Chem. 1994, 97, 13355. (b) Bush, L. C.
Ph. D. Dissertation, Yale University, New Haven, CT, 1994, pp 102-106.
(18) (a) The replacement of N-arenesulfonyl groups with hydrogen by
photolysis of sulfonamides in the presence of NaBH4 has been reported.18b-f
We also have applied this reaction to the cases of the tosyl derivative 14f
and the corresponding benzenesulfonyl derivative. (b) Abad, A.; Mellier,
D.; Pe`te, J. P.; Portella, C. Tetrahedron Lett. 1971, 4555. (c) Mellier, D.;
Pe`te, J. P.; Portella, C. Tetrahedron Lett. 1971, 4559. (d) Umezawa, B.;
Hoshino, C.; Sawaki, S. Chem. Pharm. Bull. 1969, 17, 1115, 1120; 1970,
18, 182. (e) Art, J. F.; Kestemont, J. P.; Soumillion, J. Ph. Tetrahedron
Lett. 1991, 32, 1425. (f) Horspool, W. M. In The Chemistry of Sulphonic
Acids, Esters and their DeriVatiVes; Patai, S, Rappoport, Z., Eds.; John
Wiley: New York, 1991; Chapter 13. (g) Pillai, V. N. R. Org. Photochem.
1987, 9, 225.
(19) (a) Scaiano, J. C.; Wintgens, V.; Bedell, A.; Berson, J. A. J. Am.
Chem. Soc. 1988, 110, 4050.
(20) Scaiano, J. C.; Wintgens, V.; Haider, K.; Berson, J. A. J. Am. Chem.
Soc. 1989, 111, 8732.

1410 J. Am. Chem. Soc., Vol. 119, No. 6, 1997
Bush et al.
Scheme 3
solutions of the N-tosyl diazene 17f containing appropriately
substituted alkenes, such as dimethyl maleate, dimethyl fuma-
rate, maleonitrile, fumaronitrile, acrylonitrile, and the three
stereoisomeric 2,4-hexadienes, were warmed to the diazene’s
decomposition temperature (<25 °C), the N-tosyl biradical 3f
could be captured to give fused (24f) and bridged (25f) 1:1
adducts (Table 1). In these cases, the reaction was stereospecific
for syn addition (cis-alkenes gave only cis adducts, and trans-
alkenes gave only trans adducts). The regiochemistry of the
addition varied with the trapping agent. With fumaronitrile
(FN), the fused trans adduct predominated over the bridged trans
one by a ratio of 92:8, whereas with maleonitrile, the cis endo
bridged product was favored over the cis fused one by a ratio
of 57:43. Note that the regiospecificity (fused/bridged product
ratio) of capture by FN is higher with the N-tosyl biradical 3f
(92:8) than with the N-pivaloyl biradical 3e (see below: 64:36
in MTHF, 45:55 in CH2Cl2).
poses in a first-order reaction in CDCl3 with a half-life of 14.5
min at 5 °C. The N-pivaloyl compound 17e decomposes at 5
°C with a half-life of about 50 min.
The thermal reactions with the nitrile trapping agents
proceeded in good yield and gave little or no dimeric material
20f or 21f. With the sluggishly reactive (see below) cis,trans-
2,4-hexadiene as trapping agent, the presence of these dimers
of the biradical was evident, but even so, it was possible to
isolate a 47% yield of adduct. Again, the reaction appeared to
be highly stereospecific. The formation of only one adduct from
cis,trans-2,4-hexadiene (Table 1) suggests that there also may
be a regiospecific preference for addition of the biradical at a
trans rather than a cis double bond of the diene.
In the absence of trapping agents, the biradicals formed in
the decompositions of the diazenes dimerize (see Scheme 3).
In the case of the N-pivaloyl (17e) and N-tosyl (17f) derivatives,
for example, we have identified21 the symmetrical (20) and
1
unsymmetrical (21) dimers by H NMR spectroscopic com-
parison with the corresponding dimers3 in the furan and
thiophene series. Careful examination of the NMR spectra of
solutions of the diazenes during thermal decomposition shows
the growth of not only these dimers but also of small amounts
of other products: the cyclic peroxide 23 and/or its rearrange-
ment product hydroxyaldehyde 22; hydrazone 19 formed by
tautomerization; and the adduct 18 of the biradical with any
inadvertent slight excess of dimethyl azodicarboxylate used in
the oxidation step generating the diazene from the hydrazine
17 (see Scheme 2). These reactions closely parallel those
observed previously3 in the furan and thiophene series.
When any of these diazenes is allowed to decompose
thermally in the presence of an olefinic trapping agent or
dioxygen, 1:1 adducts with the corresponding biradical are
formed in high yield. Previous work3 has shown that the
thermal decompositions of the furan and thiophene diazenes give
similar adducts. The mechanism of the cycloaddition in the
latter cases has been established by kinetic means. The reaction
involves two successive steps, namely deazetation to give the
biradical reactive intermediates 2a or 2b, followed by capture
of the biradical. This pathway requires that the overall rate of
disappearance of the diazene precursor be kinetically first-order
in diazene and zero-order in trapping agent, in agreement with
the experimental facts.3 By analogy, one would expect the same
mechanism to prevail in the pyrrole cases studied here. We
now have demonstrated this for the cases of the N-pivaloyl and
N-tosyl diazenes 17e,f, for which we observe pseudo-first-order
thermal decomposition rate constants, monitored at 10 °C by
low-temperature NMR spectrometry, that are essentially inde-
pendent of the concentration of acrylonitrile, maleonitrile, or
dioxygen in the reaction mixture.
When the decomposition of the diazene 17f in the presence
of maleonitrile, acrylonitrile, or cis,cis-2,4-hexadiene was ef-
fected by photolysis (λ ) 355-360 or 370 nm) of solutions at
temperatures below the threshhold for thermal decomposition,
the products were nearly the same as those observed in the
thermal decomposition (Table 1). Control experiments showed
that the products were stable under the conditions of photolysis.
The structure of the adducts, the strict stereospecificity of
the cycloadditions, and other evidence to be described, point to
the singlet N-tosyl biradical 3f as the reactive entity in the
thermal and direct photochemical deazetations of the diazene
17f.
Capture of the N-Pivaloyl Biradical 3e Thermally and
Photochemically Generated from the Diazene 17e. The
cycloadditions of the thermally generated biradical 3e also are
completely stereospecifically syn, but both fused and bridged
regioisomers are formed in each case (Table 1). Direct
photolysis (370 nm) of the N-pivaloyl diazene 17e gave the same
stereospecific products as the thermal reactions (shown in Table
1) but in different proportions of regioisomers. The apparent
anomaly was resolved when it was found that the adducts
undergo secondary photochemistry at unequal rates upon
continued irradiation. This photoreaction, whose nature we do
not yet know, precludes the use of the fused:bridged product
ratio as a fingerprint for a particular reactive intermediate in
the photodeazetation of N-pivaloyl diazene 17e. Nevertheless,
we are confident that the singlet biradical 3e is the common
reactive species in the thermal and direct photochemical
deazetation of this diazene.
Structures of the Adducts. The adducts obtained from
alkene cycloadditions to the thermally generated N-pivaloyl and
N-tosyl biradicals are shown in Table 1, which shows also the
corresponding results for the trapping reactions (described
below) of maleonitrile (MN) with the biradical 3f generated by
direct photolysis of the diazene 17f. The structures and
configurations were established by 1H and 13C NMR spectros-
copy and high-resolution mass spectroscopy. Details are given
in the Supporting Information.
Relative Reactivities of Alkenes Toward Singlet Biradicals
3e,f by Competition Experiments. Comparison with Reac-
tivities by Absolute Rate Measurements. Using methods
similar to those previously described,3b we have measured the
relative rates of reaction of the thermally or photochemically
generated singlet biradicals N-pivaloyl- (3e) and N-tosyl-3,4-
dimethylenepyrrole (3f) in solvent MTHF (or CH2Cl2) at 25
°C with a series of alkenes: dimethyl maleate (ME), acrylonitrile
(AN), dimethyl fumarate (FE), fumaronitrile (FN), maleonitrile
Capture of the N-Tosyl Biradical 3f Thermally and
Photochemically Generated from the Diazene 17f. When
(21) See the Supporting Information.

Singlet-Triplet Energy Gap in a Non-Kekule´ Series
J. Am. Chem. Soc., Vol. 119, No. 6, 1997 1411
Table 1. Distribution of Products (%)a,h Obtained from Alkenes and Biradicals 3e,f Generated Thermally or Photochemically (hν) from
Diazenes 17e,f
alkeneb
ssssf
f: R ) Ts
FN
MN
trans (92c)
trans (8c)
cis (43 ( 3c,f)
cis endo (57 ( 3c,f)
cis exo (∼0c)
cis endo (51c)
cis exo (∼0c)
endo or exo (20c)
endo or exo (18c)
∼0c,e,g
MN(hν)
cis (49c)
AN
AN (hν)
(80c)
(82c)
cis,cis-2,4-hexadiened
cis,cis-2,4-hexadiene (hν)
cis,trans-2,4-hexadiene
cis,trans-2,4-hexadiene (hν)
trans,trans-2,4-hexadiene
trans,trans-2,4-hexadiene (hν)
FE
cis (92c), trans (8c)
cis (92c), trans (8c)
cis (∼100c)
cis (∼100c)
trans (∼100c)
trans (∼100c)
trans (70)
∼0c,e,g
∼0c
∼0c
∼0c
∼0c
e: R ) t-BuCO
trans (30)
ME
cis (82)
cis endo or exo (9)
cis exo or endo (9)
trans (36,a 55c)
cis endo or exo(56c)
cis exo or endo (∼0e)
endo or exo (24a)
exo or endo (15a)
FN
MN
trans (64,a 45c)
cis (52,a 44c)
AN
(61,a 41c)
endo + exo (59c,g
)
a Solvent is 2-methyltetrafuran (MTHF) unless otherwise indicated. b FE ) dimethyl fumarate, X ) X′ ) CO2Me. ME ) dimethyl maleate, X
) X′ ) CO2Me. FN ) fumaronitrile, X ) X′ ) CN. MN ) maleonitrile, X) X′ ) CN. AN ) acrylonitrile, X ) CN, X′ ) H. From cis,cis-
and cis,trans-2,4-hexadiene, X ) cis-propenyl, X′ ) Me; from trans,trans-2,4-hexadiene, X ) trans-propenyl, X′ ) Me. c Solvent CH2Cl2. d Initial
composition 97% cis,cis-, 3% cis,trans-. e A trace peak of a third product of retention time close to that of the bridged product is observed. f Average
of two separate runs. g Analysis by NMR. h Analysis by HPLC unless otherwise indicated.
Table 2, column 1, shows the reactivity sequence for the
N-pivaloyl-3,4-dimethylene biradical 3e in MTHF solvent. A
comparison value, not shown in Table 2, is the FN/AN ratio of
400 in CH2Cl2 for the capture of the 3e intermediate generated
by 370 nm photolysis of 17e. As was previously reported,17
the relative rates for the N-tosyl (3f) biradical correlate linearly
with those of the 3,4-dimethylenethiophene biradical 2b3b in a
log-log relationship, with a slope of 1.24 (r ) 0.990). A similar
relationship is now seen in the case of the relative reactivities
of the N-pivaloyl system 3e and 2b (Figure 4), where the slope
is 1.07 (r ) 0.998).
Laser flash photolysis17a (355 nm) of the N-tosyl diazene 17f
in CH2Cl2 solution generates a species of λmax 593 nm, which
also can be observed in the blue matrix-immobilized ESR-silent
preparations described below. As has been reported elsewhere,17a
monitoring the 593 nm absorption of samples containing alkene
trapping agents in fluid media allows one to obtain absolute
Figure 4. Linear regression analysis of log relative rate of capture of
and hence relative rate constants (Table 2, column 3) for the
capture of the transient, which we believe to be the singlet
the N-pivaloyl-3,4-dimethylenepyrrole and 3,4-dimethylenethiophene3b,d
biradicals by alkenes. The alkene points are dimethyl maleate (ME),
N-tosyl-3,4-dimethylenepyrrole biradical 3f. The absolute rate
acrylonitrile (AN), dimethyl fumarate (FE), fumaronitrile (FN), and
constants can be derived from the relative values in Table 2
maleonitrile (MN).
and the FN value17a of 3.2 × 107 M-1 s-1. The relative rate
constants are to be compared with the values from competition
experiments on CH2Cl2 solutions of 17f and pairs of alkenes
photolyzed at 370 nm and monitored by high-performance liquid
chromatography (HPLC) of the products. In the case of the
pairwise competition FN:FE, any unreacted diazene 3f after
photolysis was quenched with maleic anhydride. The latter
results are given in Table 2, column 2.
(MN), and N-methylmaleimide (MM). The results are given
in Table 2 (columns 1-4). Table 2 also lists relative reactivities3
for the singlet 3,4-dimethylenefuran (2a) and 3,4-dimethyl-
enethiophene (2b) biradicals and for a singlet state of the TMM
biradical 2-isopropylidenecyclopentane-1,3-diyl, 2622 (columns
5-7) . We estimate that the experimental error in each of the
relative reactivities in the present work is about 15%.

1412 J. Am. Chem. Soc., Vol. 119, No. 6, 1997
Bush et al.
Table 2. Relative Rates of Reaction of Singlet Biradicals with Alkenesa
biradical
alkenef
MA
MM
MN
FN
FE
3eb,h
3fd,i
3fd,e,g
51300g
3fd,h,j
2ac,g
2bc,g
7000
26c,h
>27400
2930
235
14100
4230
2240
746
487
300
140
100
1.0
0.17
180
160
67
3.8
1.0
343k
183
1593
513
2240g
507e,g
1.0g
761
149
1.0
AN
ME
1.0
0.233
1.0
cis,cis-HD
∼0.002
a Compounds 3e,f: present work; compounds 2a-b and 26: text refs 3 and 22, respectively. b Solvent 2-methyltetrahydrofuran. c Solvent acetonitrile.
d Solvent CH2Cl2. e Diethyl fumarate used as trapping alkene instead of dimethyl fumarate. f MA ) maleic anhydride; MM ) N-methylmaleimide;
MN ) maleonitrile; FN ) fumaronitrile; FE ) dimethyl fumarate; AN ) acrylonitrile; ME ) dimethyl maleate; cis,cis-HD ) cis,cis-2,4-hexadiene.
g Transient generated by laser flash photolysis at 355 nm and rate determined by nanosecond time-resolved spectroscopy. h Thermally initiated
preparative competition experiment. i Photochemically initiated (370 nm) preparative competition experiment. j Competition ratios in column 4
determined by direct comparison of MM vs FN, FN vs MN, FN vs FE, and AN vs cis,cis-HD. FN vs AN is assumed to be the same as in column
3. k Not shown in the table is the value 400 for the FN/AN reactivity ratio toward 3e generated by 370 nm photolysis in CH2Cl2.
The relative rate value for 3f reaction with FN vs. FE was
independently confirmed by a competition for the thermally
generated reactive intermediate. In this case, the unreacted
diazene 17f was quenched with N-methylmaleimide. The
observed rate ratio FN:FE of 3.0 (Table 2, column 4) was in
good agreement with the value 3.3 obtained in the previous
series of experiments (Table 2, column 2) in which the transient
was generated by solution-phase direct photolysis of the diazene
17f.
Placing the data points of columns 2 and 3 of Table 2 on a
common basis shows that the relative values of the second-
order rate constants directly determined (column 2) for three
members of the series were MA 101 ( 20, FN 4.4 ( 1, and FE
1.0, a series to be compared with the competitively determined
values for the same alkenes of > 54 ( 8, 3.3 ( 1, and 1.0
(Table 2, column 3). The competition factor for maleic
anhydride (Table 2, column 2) is a minimum value, since the
adduct is unstable upon storage in the reaction mixture and
quickly declines in concentration. Within the limitations
imposed by this circumstance, no significant differences appear
between the relative reactivities determined by the two methods.
The agreement supports the proposal that the reactive entity
responsible for the preparative formation of the cycloadducts
of Table 1 from the thermolysis and 370 nm photolysis of
N-tosyl diazene 17f is the same singlet species as the spectro-
scopically observed carrier of the 593 nm absorption in the laser
flash photolysis17a and matrix photolysis (see below) experi-
ments.
(1) in the flash photolysis studies17a of the series in which the
diyl (rather than the alkene) structure is varied, the rates correlate
with the diyl HOMO energy, not with the diazene HOMO
energy; (2) we find no UV-vis or NMR evidence of charge-
transfer complex formation in cooled solutions of diazene-
alkene mixtures; (3) in the thermal cycloadditions, the hypo-
thetical preassociation mechanism would predict second-order
kinetics, whereas the observed kinetics always are first-order
in diazene and zero-order in alkene. We conclude that FMO
influence is being exerted in the diyl-alkene cycloadditon
transition state itself.
As we have previously shown, the variations in the rates of
these reactions are controlled largely by changes in the entropy
of activation, not in the enthalpy of activation. How the
variations in diyl HOMO-alkene LUMO energy differences
are translated into variations in activation entropies has been
explained by a surface-crossing model.20
Attempts to Effect Chemical Interception of a Triplet
State of the N-Tosyl-3,4-dimethylenepyrrole Biradical 3f. In
the thermolysis or direct photolysis of diazenes 17e,f, the species
trapped in the reactions described so far appear to be well-
defined singlets. There remains the question whether the
corresponding triplet species can be trapped under other
conditions. Based upon precedent in previous work from this
laboratory,22a-e which demonstrated the generation and capture
of both the singlet and triplet states of a trimethylenemethane
(TMM) derivative, 2-isopropylidenecyclopentane-1,3-diyl 26,
we applied three tests: the first was the dilution effect on the
product composition in cycloadditive capture of the biradical
when the transient is formed in its singlet state by pyrolysis or
direct photolysis;22c the second was the effect on the distribution
of products when the transient was generated by triplet
photosensitization;22d the third was the CIDNP effect in the 1H
NMR spectrum of freshly generated dimers of biradical 3f.22e,f
A fourth test generated the triplet biradical 3f in matrix-
immobilized admixture with a trapping agent and then liberated
the partners for reaction by thawing the matrix.
Although the reactivity of cis,cis-2,4-hexadiene is too low
for accurate measurement by the competition method, an
approximate determination (Table 2, column 4) shows it to be
∼0.2% as reactive as acrylonitrile. Thus, the diylophilic
reactivities of alkenes toward the N-tosyl biradical 3f cover a
range of at least 2.5 × 106.
Origin of the Reactivity Sequence. We favor a frontier MO
(FMO) rationalization of the reactivity sequence observed here
and in other cycloaddition reactions of singlet 3,4-dimeth-
yleneheterocycles,3,17a,19,20 in which diyl HOMO-diylophile
LUMO interaction dominates. One might imagine a subtly
different interpretation in which the diazene rapidly forms a
charge-transfer complex with the alkene, which then slowly
decomposes by deazetation of the diazene moiety. Capture of
the newly formed biradical by its solvent-caged olefin partner
then completes the reaction. In this mechanism, FMO control
would be exerted over the equilibrium constant for diazene-
alkene complex formation, rather than in the diyl addition
transition state.
Dilution Effects. In the TMM system (26), the energy of
the triplet lies below that of the singlet. Hence, initial generation
of the singlet can be followed by competition between inter-
molecular capture of the singlet by alkenes and essentially
(22) Summarized: (a) Berson, J. A. Acc. Chem. Res. 1978, 11, 446 and
references cited therein. (b) Berson, J. A. In Diradicals; Borden, W. T.,
Ed.; Wiley: New York, 1982; Chapter 4. (c) Duncan, C. D.; Corwin, L.
C.; Davis, J. H.; Berson, J. A. J. Am. Chem. Soc. 1980, 102, 2350. (d)
Berson, J. A.; Duncan, C. D.; O’Connell, G. C.; Platz, M. S. J. Am. Chem.
Soc. 1976, 98, 2358. (e) Bushby, R. J.; McBride, J. M.; Tremelling, M.;
Berson, J. A. J. Am. Chem. Soc. 1971, 93, 1544. (f) Closs, G. L. J. Am.
Chem. Soc. 1971, 93, 1545
Several lines of evidence argue against this interpretation:

Singlet-Triplet Energy Gap in a Non-Kekule´ Series
J. Am. Chem. Soc., Vol. 119, No. 6, 1997 1413
irreversible intersystem crossing to the triplet, which in turn
can be captured. The singlet and triplet forms of that TMM
display differing chemistry: the singlet reacts both regiospe-
cifically and stereospecifically, giving largely fused product by
syn addition, whereas the triplet reacts with little specificity,
giving both bridged and fused products by both syn and anti
addition. Increasing dilution of the reaction mixture decreases
the rate of capture of the initially formed singlet, thereby
relatively favoring intersystem crossing and shifting the product
composition in favor of triplet-derived adducts.
In the present work, we examined the effect of changing the
concentration of maleonitrile on the products of its capture of
N-tosyl biradical 3f generated by direct photolysis of diazene
17f at -40 °C, a temperature at which the diazene is thermally
stable. Any unreacted 17f was scavenged with maleic anhydride
or N-methylmaleimide, and the products were analyzed by
HPLC. For maleonitrile concentrations of 3.17, 6.24, 28.5, and
80 mM the only products observed by HPLC analysis were the
cis adducts, fused cis, and bridged cis-endo. The trans fused
adduct could not be detected in these reaction mixtures. With
1.6 M maleonitrile, the fused adduct mixture consisted of 97.2%
cis and 2.8% trans. The small amount of trans adduct at high
concentration presumably came from the ∼0.5% of fumaronitrile
in our maleonitrile samples. We conclude that no dilution effect
occurs in the reactions of the tosyl biradical 3f. Several
hypotheses to explain this result may readily be imagined, but
the main conclusion in the context of the objectives here is that
the triplet state of 3f has not been trapped under these conditions.
Moreover, in this case, the dilution test fails to answer whether
the triplet even has been formed.23
Photosensitization Experiments. Triplet sensitizers with
triplet energies in the range 62-74 kcal/mol24a should be
appropriate for the sensitization of a diazene, whose triplet
energy is likely to be in the range 55-65 kcal/mol.24b
Preliminary tests showed that several such sensitizers indeed
facilitated the photodeazetation of the N-tosyl diazene 17f and
that xanthone was the most efficient of the sensitizers tried. A
more direct demonstration of triplet energy transfer was provided
by a nanosecond time-resolved study of the quenching of the
xanthone triplet absorption, λmax 630 nm, by the N-tosyl diazene
17f. As Figure 5 shows, the pseudo-first-order rate constant
for decay of the xanthone triplet depends linearly (r ) 0.987)
on the concentration of diazene 17f. The seond-order quenching
rate constant derived from the slope of the linear correlation of
Figure 5 is 1.5 × 109 M-1 s-1. This rate is near the diffusion-
controlled limit and is in accord with the Dexter mechanism25
of energy transfer. Although new UV-vis maxima that might
be associated with a transient triplet state of biradical 3f could
not be observed under the laser flash conditions, the quenching
experiments suffice to demonstrate the occurrence of energy
transfer to diazene 17f. To our knowledge, this is the first such
observation with a precursor of a tetramethyleneethane deriva-
tive.
Figure 5. Quenching of the xanthone triplet absorption by N-tosyl
diazene 10f in CH2Cl2 solution at -50 °C.
competing photosensitized stereomutation of the alkenic trapping
agents would impede our attempts to demonstrate this by
stereochemical means.
A similar problem arose in the studies of the TMM 26,22d
where the stereochemical criterion for triplet formation in the
photosensitization experiments also had to be abandoned because
of competing stereomutation of the alkene trapping agents.
However, it was possible there to demonstrate the interception
of triplet 26 from the regiochemistry of xanthone-photosensitized
cycloadditions. These gave ratios of fused and bridged products
close to those observed in the "infinitely" dilute thermal
reactions, which had been identified as indicative of triplet
chemistry. Attempts to apply this strategy to the case of the
N-tosyl-3,4-dimethylenepyrrole biradical 3f failed, however,
because the regiochemistry of the cycloadditon to alkenes was
independent of whether the reactive intermediates were gener-
ated from diazene 17f by thermolysis, direct photolysis, or
xanthone-sensitized photolysis (see below).
In an attempt to minimize the alkene stereomutation problem,
we used alkenes that are relatively inefficient triplet quenchers
as trapping agents: the cis,cis-, cis,trans-, and trans,trans-2,4-
hexadienes (see Table 1). Indeed, we found that these sub-
stances interconvert only slowly under the conditions of the
xanthone photosensitization experiments. However, the yields
of cycloadducts formed were too low to permit accurate
determinations of products.
Better yields of adducts were obtained with maleonitrile as
the trapping agent, but this alkene was known24a to be an
efficient triplet quencher, and in fact we found that stereomu-
tation was so severe that no conclusions on the trapping
stereochemistry could be reached.
However, photolyses of diazene 17f to Very low conVersion
in the presence of very pure MN under xanthone-photosensitized
conditions gave recovered alkene with a ratio MN/FN of 99:1,
thus preserving the alkene stereochemistry to a degree sufficient
to detect even a small amount of non-stereospecific cycload-
diton. Any remaining diazene was quenched thermally with
maleic anhydride. Aside from the maleic anhydride adduct, the
only products were the MN adducts, cis-endo bridged and cis
fused in the ratio of 56:44. These values agree with those
obtained (Table 1) in the thermal and direct photolytic experi-
ments (57:43 and 51:49, respectively). Significantly, none of
the corresponding trans adducts could be detected.
Capture of the Transient Produced by Photosensitization.
Transfer of triplet energy to the diazene 17f presumably creates
the (localized) lowest triplet of the azo chromophore. If this
state deazetates to the triplet state of biradical 3f, one might
hope to capture the latter as characteristic cycloadducts with
appropriate trapping agents. However, we anticipated that the
(23) It is not clear from this, however, whether the exclusively singlet-
derived products are formed because the 3f biradical spin states reach an
equilibrium strongly favoring the singlet 13f before capture, or because the
singlet reacts much faster than the triplet. For this reason, this experiment
cannot be used to assign the spin of the ground state.
(24) (a) Murov, S. L. Handbook of Photochemistry; Marcel Dekker: New
York, 1973; p 3. (b) Engel, P. S.; Steel, C. Acc. Chem. Res. 1973, 6, 275.
(25) Dexter, D. L. J. Chem. Phys. 1953, 21, 836.
The results strongly suggest that even when produced by
photosensitization, biradical 3f ultimately reacts by stereospecific
syn addition with pure MN to give the same regioisomeric

1414 J. Am. Chem. Soc., Vol. 119, No. 6, 1997
Bush et al.
Table 3. Optical Absorption Maxima of Singlet
N-Substituted-3,4-dimethylenepyrrole Biradicals (3)a,b
mixture obtained in the reactions of singlet 3f (Table 1). Thus,
if triplet 3f is formed under these conditions, either its chemistry
is exactly the same as that of the singlet, or more probably, its
rate of trapping by MN is too slow to compete with intersystem
crossing to the singlet, which then is essentially the exclusiVe
biradical reaction partner in the alkene cycloadditon.23
CIDNP Effects. Fast thermolytic generation of the triplet
TMM 26 in fluid solution from the diazene precursor led to
formula
R
λ
max, nm
ꢀ, M-1 cm-1
3
3b
3h
CH3
COCH3
600
650
536
656
665
(580?)
544
590
343
a
a
a
a
the formation of dimers of 26 with numerous strongly emissive
1H NMR resonances.22e This observation was interpreted22f as
direct evidence of the involvement of at least one triplet of 26
in the dimerization. In the thermal decomposition of 17f, the
precursor of the biradical 3f, similar attempts did not produce
discernible emissions or enhanced absorptions in the dimers’
1H NMR spectrum. This result is not surprising, since the
ordering of the multiplet energies in 26 is clearly T-below-S
(by about 13-15 kcal/mol).22a-e Formation of 326 thus follows
hard on the heels of thermal generation of 126 and leads readily
to magnetically polarized dimer, whereas 33f may lie uphill from
thermally generated 13f in solution, and dimers are formed from
the singlet without polarization.
3d
3e
COi-Pr
COt-Bu
∼4900
∼3500
∼3900
∼5000
a
a
a
3f
Ts
Bs
300(?)
595
3g
a Not measured. b In MTHF glass at 77 K.
approaches might have had the best chance for success, since
at least they generate the elusive triplet first. The other two
methods, thermally initiated CIDNP and dilution, both generate
the singlet first. It will be self-evident that interception of the
triplet under the latter conditions is inherently difficult.
Generation of Triplet Biradical 3f in the Presence of
Trapping Agent in Matrix-Immobilzed Form Followed by
Annealing. A frozen solution of diazene 17f and excess
fumaronitrile in CD2Cl2 was irradiated at 77 K with monochro-
matic 265 nm light, conditions that produce the characteristic
ESR spectrum assigned11b to the triplet biradical 33f. The
temperature was raised to 196 K, warm enough to melt the glass
but well below the thermal decomposition point of the diazene.
The sample was re-glassed at 77 K, re-irradiated, and re-thawed
for a total of 75 1-min irradiations. After the sample had thawed
to 230 K, the NMR spectrum (taken at -40 °C) showed that
little or none of the diazene had been consumed (<2%), and
none of the trapping products could be discerned. Apparently,
under these conditions, only a very small amount of the diazene
is converted to 33f, which is in accord with the results of a spin-
count experiment.11b Similar low conversions are observed in
fluid solutions with monochromatic 265 nm light.
Generation and Characterization of the Kinetically Stable
Singlet States of 3,4-Dimethylenepyrrole Biradicals in Low-
Temperature Matrices. In either fluid media or solvent
glasses, each of the diazene precursors 17b,d,e,f,g,h showed
absorbances near 370 nm as expected for the azo n,π*
chromophore. Irradiation of glassy 2-methyltetrahydrofuran
(MTHF) samples of any of these at 370 nm at 77 K generated
a blue color and a UV-vis absorption spectrum dominated by
a broad band centered in the region 590-665 nm (see Table
3), which we assign to the corresponding biradicals 3. As
already has been described,17a a 590 nm absorption can be
detected by nanosecond time-resolved spectroscopy of flash-
photolyzed fluid solutions of the precursor diazene 17f. We
propose that the singlet species 3f is the common carrier of the
spectra in the solution and matrix phases, and we think it is
likely that the corresponding singlets 3b,d,e,g,h are responsible
for the blue color and long-wavelength absorption in the other
instances.
A maleonitrile trapping experiment was carried out in MTHF
at -40 °C with NiSO4-filtered light (230-325 nm). Note that
at 5 K, these irradiation conditions produce a much stronger
ESR spectrum than is obtained with monochromator-filtered 265
nm light.11b Treatment with N-methylmaleimide to quench
residual diazene and conventional work-up still showed no
detectable trace of maleonitrile or fumaronitrile trapping product
(HPLC analysis). A separate control experiment showed that
authentic trapping products 24 and 25 (X ) CN), were readily
decomposed by 265 nm irradiation, each adduct’s concentration
being diminished by 41% (NMR analysis relative to dimethyl
hydrazodicarboxylate internal standard) after 12 min photolysis
in solution at -30 °C. (As we already have noted, under our
conditions, monochromatic 370 nm light is without effect on
the products 24 and 25 X ) X′ ) CN). We believe that the
major reasons for the failure to observe the trapping products
of the triplet 33f in the matrix experiments with MN at 265 nm
are the low photoconversions 17f f 33f and the photochemical
instability of the trapping products under the irradiation condi-
tions. We ultimately found11b that the cycloadducts of the 3f
manifold with dimethyl azodicarboxylate (DMAD) are relatively
photostable and can be isolated. This has important implications
in another context,11b but the DMAD diylophile of course cannot
be used as a stereochemical probe.
The extinction coefficients (ꢀ) of the matrix-immobilized
transients 3f,g were difficult to determine accurately because
of secondary photobleaching of the new chromophore, which
precluded attempts to achieve complete photoconversion of the
precursor 17f (or 17g) to the biradical 3f (or 3g) (see below).
A rough value of ꢀ ∼ 5000 M-1cm-1 for 3f was obtained from
low-conversion experiments in which the concentration of the
biradical was estimated from the decrease in intensity of the
370 nm azo absorption of the precursor 17f, combined with the
assumption that the photoproduct from 17f was entirely the
biradical 3f. The estimate probably is close to the true extinction
coefficient, since it gives the value 2kt ) 2.8 × 1010 M-1 s-1
when used in conjunction with the nanosecond laser flash
photolysis measurements17a to calculate the rate constant for
dimerization of 3f in fluid CH2Cl2 solution at 220-230 K. This
value is close to the diffusion-controlled rate 1.5 × 1010 M-1
s-1 calculated from theory. Also, it matches the rates deter-
mined experimentally17a,19,20 for other singlet biradical dimer-
izations, including the 3,4-dimethylenethiophene singlet, for
which the extinction coefficient could be measured3c directly
on matrix-immobilized samples. A similar value of ꢀ, 4900
Μ-1 cm-1 , is observed for the N-pivaloyl derivative 3e.
Thus, four unsuccessful attempts have been made here to
demonstrate the independent capture chemistry of the triplet 33f.
Of the four methods, the photosensitization and matrix-thawing

Singlet-Triplet Energy Gap in a Non-Kekule´ Series
J. Am. Chem. Soc., Vol. 119, No. 6, 1997 1415
spectroscopy. The relative reactivities of UV-vis-active tran-
sients, as measured in this way, match those measured by in
preparative competition experiments. This concordance, com-
bined with the structures of the adducts with alkenes, demon-
strates that the transient blue species are indeed singlet 3,4-
dimethylenepyrroles.26,27
Experimental Section. General synthetic and spectroscopic
procedures are described in the Supporting Information.
Syntheses of Schemes 1 and 2. See Supporting Information.
Preparation and Characterization of the Products Listed
in Table 1. See Supporting Information.
In addition to the 590 nm band, the tosyl biradical 3f showed
an absorption maximum at 343 nm and possibly at 300 nm.
Although diazene absorbances in the same general regions made
these short-wavelength bands difficult to detect in most samples,
the presence of absorptions there authentic to 3f was suggested
by the fact that the 340 and 300 nm regions were active in its
photobleaching.
With the 1000 W Hg-Xe monochromator system, secondary
photobleaching seems to have its maximum near 400 nm,
although the absorption spectrum of the blue samples does not
show a maximum there. Samples obtained by fully photo-
bleaching 3f showed no UV-vis absorption beyond 350 nm.
The complex NMR spectrum of the product(s) of photo-
bleaching can be recorded at 183 K in solution but rapidly
disappears at 243 K. A firm assignment of structure is not yet
available.17b
We also have observed that thermal decomposition (270-
300 K) of diazene 17f in a co-solid with adamantane, followed
by cooling of the sample to 77 K, gave a visibly blue
preparation. A similar blue color often was seen during the
preparation of solid samples of the hydrazine 16f (Scheme 2).
Presumably, the solid acts as a support for a small amount of
biradical 3f formed by adventitious oxidation of the hydrazine
and deazetation of the resulting diazene. The identity of the
oxidant is not clearly established. Ordinary molecular dioxygen
would seem to be a likely candidate, but the blue color appeared
even in samples that we had made extra efforts to protect from
air. The blue preparations remained so at room temperature in
the dark under vacuum for about 1h, and for weeks in a dark
freezer at 190 K. Such observations recall those previously
reported3c for the thiophene biradical 2b. They indicate that
despite its extraordinarily fast dimerization in fluid media, the
singlet biradical 3f, when isolated and immobilized, is a
kinetically stable species, even at room temperature. We
postpone to the accompanying paper11b a discussion of the
question whether the singlet is actually the ground state of
N-tosyl-3,4-dimethylenepyrrole 3f, but we note here that blue
matrix-immobilized preparations of singlet 3f showed no ESR
signals of a triplet species even after storage for as long as a
month at 77 K.
Conclusions. The chemistry of the transients formed upon
thermal, direct photochemical, or photosensitized deazetation
of the precursors of a series of N-substituted-3,4-dimethyl-
enepyrroles strongly suggests that the reactive entities captured
in cycloadditions to alkenes are singlet biradicals. These ESR-
silent, deeply colored singlet species can be detected directly
in the matrix immobilized state by their persistent UV-vis
absorption. These absorptions also appear in the solution phase
during nanosecond laser flash photolysis, and their kinetic
behavior can be monitored by nanosecond time-resolved
Kinetics of Thermal Deazetation of N-Toluenesulfonyl
Diazene 17f and N-Pivaloyl Diazene 17e Monitored by NMR.
(Details are given in the Supporting Information). The kinetics
were carried out by procedures similar to those described
elsewhere.3b At 283 K, the rates for 17f in CDCl3 were strictly
first-order (r2 ) 0.996-0.999). The rate constants (×104 s) kd
were independent of the concentration [MN] of maleonitrile:
[MN] ) 0.0 M, kd ) 8.6 (8.0 measured by a second investiga-
tor); [MN] ) 0.026 M, kd ) 9.2; [MN] ) .059 M, kd ) 7.2
(measured on a different day); [MN] ) 0.1 M, kd ) 9.5.
Similarly, for 17e in CDCl3 at 277.9 K, for [MN] ) 0.0 M,
[O2] ) 0.0 M. kd ) 2.4; for [O2] ) saturated in CDCl3, 2.0,
2.3; for [MN] ) 0.354 M, kd ) 2.5. See Supporting Informa-
tion.
Quantum Chemical Calculations. The AM1/CI results are
given in the Supporting Information to our preliminary
communication11a and ref 17b. These results are given again
together with the PM3/CI results in the Supporting Information
to the present paper.
Acknowledgment. We are grateful for support of this work
by grants from the National Science Foundation and the
American Chemical Society’s Arthur C. Cope Scholar Award
program and by a Dox Fellowship (to L.C.B.). The participation
of M. Walker in the early stages of these studies is gratefully
acknowledged.
Supporting Information Available: Synthetic details of
Schemes 1 and 2, preparations and characterizations of trapping
products listed in Table 1 and Scheme 3, descriptions of trapping
studies and analyses of product ratios, results of AM1/CI and
PM3/CI calculations, and kinetics of thermal deazetation of
diazenes 17e,f (72 pages). See any current masthead page for
ordering and Internet access instructions.
JA963113K
(26) (a) Our so far unsuccessful attempts to observe a solid state 13C
NMR signal for singlet 3f by methods used in other systems27 are described
elsewhere: (b) Reference 17b, pp 99-106.
(27) (a) Zilm, K. W.; Merrill, R. A.; Greenberg, M. M.; Berson, J. A. J.
Am. Chem. Soc. 1987, 109, 1567. (b) Zilm, K. W.; Merrill, R. A.; Webb,
G. G.; Greenberg, M. M.; Berson, J. A. J. Am. Chem. Soc. 1989, 111, 1523.
(c) Greenberg, M. M.; Blackstock, S. C.; Berson, J. A.; Merrill, R. A.;
Duchamp, J. C.; Zilm, K. W. J. Am. Chem. Soc. 1991, 113, 2318. (d)
Reynolds, J. H.; Berson, J. A.; Kumashiro, K. K.; Duchamp, J. C.; Zilm,
K. W.; Rubello, A.; Vogel, P. J. Am. Chem. Soc. 1992, 114, 763. (e)
Reynolds, J. H.; Berson, J. A.; Kumashiro, K. K.; Duchamp, J. C.; Zilm,
K. W.; Scaiano, J. C.; Berinstain, A. B.; Rubello, A.; Vogel, P. J. Am.
Chem. Soc. 1993, 115, 8073.