Electron-transfer-induced tautomerization in methylindanones: Electronic control of the tunneling rate for enolization

Article abstract of DOI:10.1021/ja003708m

The radical cations generated from 4-methyl- and 4,7-dimethylindanone, as well as their deuterated isotopomers, isolated in Argon matrices, were found to undergo enolization to the corresponding enol radical cations at rates that differ by orders of magnitude. It is shown by quantum chemical calculations that the effect of the remote methyl group in the 4-position is of purely electronic nature in that it stabilizes the unreactive π-radical relative to the reactive σ-radical state of the 7-methylindanone radical cation. The observed kinetic behavior of the two compounds can be reproduced satisfactorily on the basis of calculated heigth and width of the thermal barrier for enolization, using the Bell model for quantum mechanical tunneling. High-level calculations on the methylacrolein radical cation show that barriers for enolization in radical cations are overestimated by B3LYP/6-31G*.

Full text of DOI:10.1021/ja003708m

J. Am. Chem. Soc. 2001, 123, 2377-2387
2377
Electron-Transfer-Induced Tautomerization in Methylindanones:
Electronic Control of the Tunneling Rate for Enolization
Pawel Bednarek,† Z. Zhu,†,‡ Thomas Bally,*,† Tomasz Filipiak,§ Andrzej Marcinek,§ and
Jerzy Gebicki*,§
Contribution from the Institute of Physical Chemistry, UniVersity of Fribourg, Perolles,
CH-1700 Fribourg, Switzerland, and Institute of Applied Radiation Chemistry, Technical UniVersity,
90-924 Lodz, Poland
ReceiVed October 17, 2000. ReVised Manuscript ReceiVed December 27, 2000
Abstract: The radical cations generated from 4-methyl- and 4,7-dimethylindanone, as well as their deuterated
isotopomers, isolated in Argon matrices, were found to undergo enolization to the corresponding enol radical
cations at rates that differ by orders of magnitude. It is shown by quantum chemical calculations that the
effect of the remote methyl group in the 4-position is of purely electronic nature in that it stabilizes the unreactive
π-radical relative to the reactive σ-radical state of the 7-methylindanone radical cation. The observed kinetic
behavior of the two compounds can be reproduced satisfactorily on the basis of calculated heigth and width
of the thermal barrier for enolization, using the Bell model for quantum mechanical tunneling. High-level
calculations on the methylacrolein radical cation show that barriers for enolization in radical cations are
overestimated by B3LYP/6-31G*.
1. Introduction
Scheme 1
Keto-enol tautomerization is a very well understood reaction
in organic chemistry.1 Apart from some notable exceptions,2
enols are usually less stable3 and exist only at very low
concentration under equilibrium conditions.4 However, the
situation is different in the corresponding radical cations where
gas-phase experiments have shown that enol radical cations are
usually more stable than their keto tautomers.5 This is due to
the fact that enol radical cations profit from allylic resonance
stabilization that is not available to ketones. Moreover, in the
case of aromatic carbonyl compounds, enolization results in the
formation of o-quinoid alcohols that are much easier to ionize
than their benzenoid precursors (see Scheme 1).
We have shown that certain classes of radical cations, carrying
a wide range of H-donor and H-acceptor groups, undergo
spontaneous tautomerization when generated in low-temperature
matrices.6,7 In our earlier studies we had focused mainly on
whether enolization does occur or not, but we had already noted
that some of these reactions must involve quantum mechanical
tunneling.8,9 In our recent work on the occurrence of electron-
transfer induced tautomerization in radical cations of NADH
analogues10,11 we found that the rate of H-transfer depends
strongly on the distance and the orientation of the H-donor and
the H-acceptor group, and we encountered indeed a borderline
case where hydrogen atom tunneling can be observed to occur
over hours at 12 K.
One way to vary the distance over which the hydrogen atom
must travel in the course of the enolization is by changing the
ring size of the cyclic ketone attached to the tolyl moiety which
serves as the H-donor. Since enolization was found to occur
spontaneously in 5,8-dimethyltetralone (DMT)9 we thought that
† University of Fribourg.
‡ Present address: Agouron Pharmaceuticals, Alanex Division, San
Diego, CA.
§ Technical University of Lodz.
(1) The Chemistry of Enols; Rappoport, Z., Ed.; Wiley: Chichester, 1990.
(2) Hart, H.; Rappoport, Z.; Biali, S. E. Isolable and relatively stable
simple enols. In The Chemistry of Enols; Rappoport, Z., Ed.; Wiley:
Chichester, 1990; p 481.
(3) Guthrie, J. P. Themodynamics of Enols. In The Chemistry of Enols;
Rappoport, Z., Ed.; Wiley: Chichester, 1990; p 75.
(4) Toullec, J. Keto-Enol Equilibrium Constants. In The Chemistry of
Enols; Rappoport, Z., Ed.; Wiley: Chichester, 1990; p 323.
(5) Turecek, F. The Chemistry of Ionized Enols in the Gas Phase. In
The Chemistry of Enols; Rappoport, Z., Ed.; Wiley: Chichester, 1990; p
95.
(9) Marcinek, A.; Michalak, J.; Rogowski, J.; Tang, W.; Bally, T.;
Gebicki, J. J. Chem. Soc., Perkin Trans. 2 1992, 1353.
(6) Gebicki, J. Pure Appl. Chem. 1995, 67, 55.
(10) Marcinek, A.; Adamus, J.; Huzben, K.; Gebicki, J.; Bartczak, T. J.;
Bednarek, P.; Bally, T. J. Am. Chem. Soc. 2000, 122, 437.
(11) Marcinek, A.; Rogowski, J.; Adamus, J.; Gebicki, J.; Bednarek, P.;
Bally, T. J. Phys. Chem. A 2000, 104, 718.
(7) Gebicki, J.; Bally, T. Acc. Chem. Res. 1997, 30, 477.
(8) Gebicki, J.; Marcinek, A.; Michalak, J.; Rogowski, J.; Bally, T.; Tang,
W. J. Mol. Struct. 1992, 275, 249.
10.1021/ja003708m CCC: $20.00 © 2001 American Chemical Society
Published on Web 02/17/2001

2378 J. Am. Chem. Soc., Vol. 123, No. 10, 2001
Bednarek et al.
by going to the corresponding five-membered ring compound,
4,7-dimethylindanone (DMI), we could slow tautomerization
sufficiently to monitor the decay of the parent keto radical
cation.
As we will show, this strategy succeeded. However, more
importantly, we found that removal of the distant 4-methyl group
to give 7-methylindanone (MI) has a profound influence on
the kinetics of the enolization: MI•+ behaves analogous to
DMT•+ rather than DMI•+, in that it decays spontaneously at
12 K in Ar, irrespective of whether it is formed radiolytically
from the neutral ketone or by photolysis of the enol radical
cation. In the present paper we will put this surprising
observation into proper perspective and thus demonstrate for
the first time that, next to the more obvious geometric factors,
purely electronic factors can be instrumental in determining the
kinetics of tautomerizations in radical cations. Analogous
observations have been made in connection with intra- and
intermolecular photoreductions involving neutral species many
years ago,12 but to our best knowledge, never in photoenoliza-
tions involving compounds of the present type.
Figure 1. Electronic absorption spectra obtained after ionization of
DMI in Ar at 1 K (a) and after subsequent irradiation at >640 (b) and
490 nm (c). Difference spectra b-a and c-b highlight the spectral
changes during the last two steps. Difference spectrum d was obtained
after keeping a sample after >640 nm irradiation during 14 h at 12 K
in the dark. Gray bars represent CASPT2 excitation energies and
transition moments for DME•+(cf. Table 6).
2. Experimental Results
Figure 1 shows the spectra obtained after ionization of DMI
in argon at 12 K (a), after subsequent irradiation at >640 nm
(b), and after final irradiation at 490 nm (c). The transformations
induced by the two irradiations are represented also in the
corresponding difference spectra (b-a and c-b) which show
clearly the bleaching and reappearance of a species with sharp
absorptions at 600-720 nm. This absorption is typical for the
chromophore of the o-quinoid radical cations that are formed
by tautomerization from radical cations of benzene derivatives
substituted vicinally by a H-donor and a H-acceptor group (cf.
Scheme 1).7 In the present case, the expected product is the
enol radical cation of DMI which we will abbreviate DME•+.
That DME•+ is indeed the product obtained on ionization of
DMI is also confirmed by the IR spectra in Figure 2, which
represent the difference spectra corresponding to (b-a) and (c-
b) in Figure 1. They show clearly the decrease of the O-H
stretching vibration at 3514 cm-1 on >640 nm irradiation, along
with other bands in the mid-IR region that match fairly well
with those predicted by the results of a B3LYP frequency
calculation shown at the bottom of Figure 2. The main IR
manifestation of the bleaching of DME•+ is the appearance of
a band at 1747 cm-1 which corresponds to the CdO stretching
vibration of DMI•+, shifted by 27 cm-1 to higher energies from
that of neutral DMI.
Figure 2. Difference IR spectra for the interconversion of DMI•+
(peaks pointing up in the solid line) and DME•+ (peaks pointing down
in the solid line). The bottom spectrum represents the results of a
B3LYP/6-31G* frequency calculation on DME•+ (all frequencies scaled
by 0.97).
Interestingly, the same transformation that was observed on
490 nm irradiation, i.e., the re-formation of DME•+ from
DMI•+, is also witnessed after letting a sample obtained after
complete bleaching at >640 nm stand overnight at 12 K
(spectrum d in Figure 1). This indicates that enolization of
(12) See, e.g.: Wagner, P. J. Acc. Chem. Res. 1971, 4, 168.

Tautomerization in Methylindanones
J. Am. Chem. Soc., Vol. 123, No. 10, 2001 2379
Figure 3. Change in the area of the OH stretching band of DME•+
with time, starting with a sample of ionizend DMI after bleaching
DME•+ by >640 nm irradiation. Similar profiles were obtained at
intermediate temperatures and up to 30 K.
Figure 5. Electronic absorption spectra obtained after ionization of
MI (a) and MI-d3, respectively, (b) in Ar at 12 K. (c and d) The spectra
•+
after bleaching of ME-d3 at >590 nm and its re-formation at 450
nm, respectively. Black bars represent CASPT2 excitation energies and
transition moments of MI•+ (c) and ME•+ (d).
contrast to the dimethyl derivative, ME•+ resisted all attempts
to convert it back to the parent keto cation, MI•+, in that the
enol cation proved to be entirely photostable at all wavelengths
above 350 nm (on prolonged irradiation in the UV some
decomposition occurs). That this is not due to an inherent change
in the photochemistry on going from DME•+ to ME•+ is shown
by the results obtained with MI-d3. Although ionization of this
compound results also in partial enolization (spectrum b), this
can be reverted photochemically (spectrum c), and the resulting
keto cation, MI-d3•+, is entirely stable, unless it is exposed to
450 nm irradiation, whereupon it reverts back to the enol radical
cation, ME-d3•+ (spectrum d). The above findings are confirmed
by IR spectra corresponding to those shown in Figure 2.
An intriguing aspect of the spectra shown in Figure 5 is the
difference in shape and the shift between the first absorption
bands of ME•+ and its deuterated derivative (the shift in the
first peak is 325 cm-1!). We have no explanation for this
phenomenon which was, however, observed in different series
of experiments with compounds whose identity had been
scrupulously ascertained.
Thus, the experiments described above show unambiguously
that the presence or absence of a methyl group in what appears
to be a pure “spectator” position has a profound influence on
the rate of enolization in the radical cation of 4-methylindanone.
Its presence slows the reaction from less than minutes at 12 K
(i.e. the time it takes to observe the spectrum of the enol radical
cation after bleaching) to several hours at 12-30 K. As we will
show, the explanation of this surprising phenomenon resides in
the photoelectron spectra of the two compounds (Figure 6)
Figure 4. Electronic absorption spectrum obtained after ionization of
DMI-d3 in Ar at 12 K (a) and after subsequent irradiation at >590 nm
(b). Black bars represent CASPT2 excitation energies and transition
moments of DMI•+ (cf. Table 5).
DME•+ occurs also thermally at 12 K over hours. Following
this reaction by plotting the area under the O-H stretching band
in the IR spectrum against time shows that the rate of enolization
does not depend on the temperature in the range accessible in
Ar matrices (12-30 K) within the error of determination of
the band area (the data for 12 and 25 K are shown in Figure 3).
This is a strong indication that quantum mechanical tunneling
is involved.
This suspicion is confirmed by the experiments with the
derivative where the methyl groups are deuterated (DMI-d6),13
shown in Figure 4. In contrast to DMI, no enolization is
observed on ionization, and when the enol is formed photo-
chemically by irradiation at >490 nm (spectrum b in Figure
4), it remains stable over many days at 12 K as well as on
warming to 30 K. Thus, the facile enolization that was observed
for DMI•+ is completely suppressed on deutration of the methyl
groups. This proves that enolization occurs by tunneling.
Figure 5 shows the results of similar experiments with
7-methylindanone (MI) and its deuterated derivative, MI-d3.14
As in the case of DMI, ionization results in the formation of
the corresponding enol radical cation, ME•+. However, in
(14) In actual fact, the compound contains an additional deuterium atom
in the CH2 group adjacent to the benzene ring (see Synthesis section).
However, this deuterium atom does not affect the spectroscopic or kinetic
features of interest in the present study. Therefore, and to avoid confusion,
we will designate this compound MI-d3 anyway.
(13) For reasons of economy, perdeuterio p-xylene was employed in the
synthesis. Thus, the two remaining aromatic positions in DMI are also
deuterated, but this is not expected to affect the spectroscopic or kinetic
data.

2380 J. Am. Chem. Soc., Vol. 123, No. 10, 2001
Bednarek et al.
Table 1. Relative Energies (in kcal/mol corrected for zero-point
vibrational energy (ZPVE) differences at equilibrium geometries
relative to the most stable state) for Different States of I•+, MI•+
and DMI•+ at Different Geometries. Energies
,
method f
geometryb V
B3LYP/6-31G*
CASPT2/ANOa
12A′′ 12A′ 22A′′
3.5 (0)
5.6 11.1 13.6
(0)d
18.7 18.5
11.0 11.3 15.9
12A′′
12A′
I•+
neutral
9.0
4.3
(0)c
9.6
4.2
(0)f
9.3
(0)
8.1
14.0
6.6e
2.8
DMI•+
neutral
2A′′ (π-radical)
2A′ (σ-radical)
neutral
MI•+
1.3
(0)g
9.0
1.1
9.7 15.3
5.4 10.3
6.7
2A′′ (π-radical)
2A′ (σ-radical)
8.7
1.4h
a For active spaces, see footnotes to Tables 4 and 5. b Optimized at
the B3LYP/6-31G* level. c Total energy: -501.345984 hartrees;
ZPVE)124.1 kcal/mol. d Total energy: -499.738527 hartrees. e ZPVE
) 124.5 kcal/mol. f Total energy: -462.0163499 hartrees; ZPVE )
107.5 kcal/mol. g Total energy: -460.585542 hartrees. h ZPVE )112.2
kcal/mol.
either method to predict absolute ionization energies, but only
the energies of excited states relatiVe to the ground states of
radical cations, as given in Table 1. Below we will use
designations (µ)-1 to denote radical ion states attained be
electron ejection from MOs ψµ of the neutral precursor.
Indanone: All our calculations agree in predicting that the
ground state of the indanone radical cation, I•+, at the geometry
of neutral I (indicated by an open bar in Figure 6) arises by
ionzation from the nO orbital that extends into the C-C σ-bonds
adjacent to the carbonyl group. The first excited state of the
radical cation corresponds to ionization from the “bis-ethylenic”
π MO of the benzene moiety, (π3)-1 (dashed bar in Figure 6),
whereas the next one is formed by electron loss from the “bis-
allylic” benzene MO, (π4)-1 (filled bar).19 Following a “hole”
in the PE spectrum, the next band is due to ejection of an
electron from the πCdO MO (combined to some extent with CH2
pseudo-π MOs and the totally symmetric benzene π MO).
The above sequence of the (nO)-1 and the (π3)-1 states of
Figure 6. Photoelectron spectra of indanone (a), MI (b), and DMI
(c). The bars are schematic representations of the CASPT2 results listed
in Table 1. MOs at the top are from HF-3-21G calculations on neutral
indanone.
which will be discussed in the following section, based on
quantum chemical calculations.
I
•+ is confirmed by B3LYP calculations. The two methods differ
3. Calculations and Discussion
somewhat in their prediction of the gap between these two states
(0.15 eV by CASPT2, 0.39 eV by B3LYP). If we assume that
the ordering of states given by the two methods is indeed correct,
then the PE spectrum indicates that the gap is actually around
0.25 eV. Thus, CASPT2 seems to underestimate sligthly the
energy difference between states attained through ionization
from nO and from π MOs, whereas B3LYP overestimates this
difference by a similar amount.
4,7-Dimethylindanone: According to our calculations, the
state ordering is changed on attaching two methyl groups in
the 4- and 7-positions of indanone. The ground state of DMI•+
corresponds to loss of an electron from the π4 MO which has
thus been destabilized by almost 1 eV from its position in
indanone, owing to the inductive effect and hyperconjugation
with the CH3 groups.20 According to our calculations, the peak
at 8.98 eV should be assigned to the (nO)-1 state. Note that it
has moved only 0.13 eV from its position in parent I•+, in
agreement with the fact that the nO MO does not extend much
into the benzene ring. The splitting between the lowest π- and
3.1. The Photoelectron Spectra of Indanones. To our best
knowledge the photoelectron (PE) spectra of indanones have
not been published to date, and since these spectra form a very
good starting point for the ensuing discussion we will begin by
their assignment. The PE spectra of related carbonyl-substituted
benzenes, such as benzaldehyde or acetophenone, have been
published in the 1970s.15-17 Unfortunately these spectra do not
lend themselves to assignment on the basis of Koopmans’
theorem, because the orbital energy of the n lone pair (nO) is
invariably predicted much too low relative to those of the π
MOs by ab initio Hartree-Fock calculations. Although this can
be rectified by resorting to semiempirical methods, different
such procedures gave a different ordering of states in benzal-
dehyde15 such that no unambiguous assignments of the PE
spectra of Ph-COR derivatives have been provided to date. Thus
we decided to address this question by the CASSCF/CASPT2
method which has recently proven to be quite reliable in
assigning electronic spectra of the Ph-COR type radical
cations.18 In addition we will estimate the important difference
between their lowest energy π- and σ-radical states by the
B3LYP density functional method. Note that we will not use
(19) The numbering of the π MOs shown in Figure 6 is not in line with
their ordering in indanone, but with that in MI and DMI.
(20) This shift is about 20% larger than that on going from benzene to
p-xylene [Lias, S. G. In NIST Chemistry WebBook; NIST Standard Reference
Database No. 69; Mallard, W. G., Linstrom, P. J., Eds.; National Institute
of Standards and Technology, Gaithersburg MD, 20899, February 2000
(http://webbook.nist.gov)]. This is probably due to the fact that the π4 MO
is lowered by 0.3 eV relative to its position in benzene and can therefore
undergo more hyperconjugative interaction with the CH3 pseudo-π MOs
which leads to a stronger destabilization of the π4 MO.
(15) Rabalais, J. W.; Colton, R. J. J. Electron Spectrosc. 1972, 1, 83.
(16) Kobayashi, T.; Nagakura, S. Bull. Chem. Soc. Jpn. 1974, 47, 2563.
(17) Egdell, R.; Green, J. D.; Rao, C. N. R. Chem. Phys. Lett. 1975, 33,
300.
(18) Huben, K.; Zhu, Z.; Bally, T.; Gebicki, J. J. Am. Chem. Soc. 1997,
119, 2825.

Tautomerization in Methylindanones
J. Am. Chem. Soc., Vol. 123, No. 10, 2001 2381
σ-radical states (0.42 eV from the PE spectrum) is again
underestimated, both by CASPT2 (0.24 eV) and B3LYP (0.16
eV). The next excited state of DMI•+ is the (π3)-1 state, 0.35
eV above the ground state according to CASPT2 (0.59 eV from
the PE spectrum).
7-Methylindanone: The PE spectrum of MI•+ is the most
difficult one to assign because the first three ionization events
are lumped together in a single, barely structured PE band.
However, in view of the above assignments this is not surprising,
because a linear interpolation between the ionization energies
of I and DMI leads to predicted values of 9.05 eV for both the
(nO)-1 and the (π4)-1 states and 9.25 eV for the (π3)-1 state of
MI•+, in good accord with the observed PE spectrum. Indeed,
theory agrees with the finding that the (nO)-1 and the (π1)-1
states are nearly degenerate in MI•+. Both CASPT2 and B3LYP
slightly favor the σ-radical over the π-radical ground state, but
in view of the above errors in the predicted gaps between these
two states in I•+ and DMI•+, this prediction should be taken
with a grain of salt.
In conclusion, the PE spectra of the three compounds show
that there exist substantial differences in the relative energies
of the π- and the σ-radical states of the radical cations. We
will demonstrate below that the different reactivity with regard
to enolization observed for MI•+ and DMI•+can be traced back
to differences in these σ/π gaps.
Figure 7. Schematic representation of the mechanism of photoeno-
lization (top) of â-alkylenones and of the corresponding reaction in
the corresponding radical cations (bottom). Note that π-radical states
of â-alkylenones are unreactive toward enolization.
3.2. The Mechanism of Enolization in â-Alkylenones:
Model Calculations. Photoenolization in â-alkylenones is
thought to occur by abstraction of a H atom from the alkyl group
by the singly occupied pσ NBMO of the carbonyl moiety in its
triplet n f π* excited state, followed by a twisting of the CH2
group in the incipient diradical to give the ground-state enol
(after intersystem crossing which occurs during this last event).21
By analogy, one may picture enolization in the corresponding
radical cations (where it is exothermic, see Introduction) to
follow the same mechanism, with the exception that no change
in the spin state is required on the way from reactant to product
(cf. Figure 7).
This picture is confirmed by calculations on the simplest such
system, 3-methylacrolein (MA), whose radical cation has a
σ-radical ground state. The reaction profile for enolization of
MA•+ to its enol form, MAE•+, involves indeed a transition
state, TS1, where the hydrogen atom is roughly halfway between
the donor C- and the acceptor O-atoms (rC‚‚‚H ) 1.35 Å, rO‚‚‚H
) 1.23 Å), but remains in the plane of the molecule. On the far
side of TS1 the reaction proceeds to another transition state,
TS2, which turns out to be that for rotation of the CH2 group
in the product, the Z-butadien-1-ol radical cation (cf. Figure
8). According to IRC calculations this rotation, and hence the
loss of Cs symmetry, only sets in after the H has been almost
fully transferred to the O-atom. Thus, enolization proceeds
diabatically on the 2A′ surface of â-alkylenone radical cations.
To examine the validity of the B3LYP/6-31G* method for
calculating the activation parameters of such reactions, we first
reoptimized the geometries of the three stationary points with
larger basis sets at the B3LYP level, and finally also with the
QCISD/6-31G* method. For MA•+ and MAE•+ the bond
lengths vary by less than 0.01 Å between the different methods
so that, with regard to these equilibrium geometries, the B3LYP/
6-31G* method appears to be appropriate. For TS1 some of
the differences were found to be more important: at the QCISD
level the hydrogen atom is closer to the carbon (rC‚‚‚H ) 1.26
Figure 8. Schematic reaction profile for the enolization of methyla-
crolein radical cation (MA•+) as computed by B3LYP/6-31G*. Note
that the entire H atom transfer takes place in the A′ (“σ- radical”)
2
state.
Å) than to the oxygen atom (rO‚‚‚H ) 1.33 Å),22 in contrast to
B3LYP where the opposite is true (see above; the other bond
lengths are again within 0.01 Å for both methods). This
difference in the geometry of TS1 changes ∆E0q by -0.88 kcal/
(22) For reasons of computational economy, we had to impose Cs
symmetry in the QCISD optimizations as well as in the B3LYP structures
used in the coupled-cluster calculations. In the case of TS1 this resulted in
a stationary point with two imaginary normal modes. One of them
corresponds to a slight twisting of the CH2 group, but the associated
frequency is only -110 cm-1. Reoptimization in C1 resulted in a decerase
in the B3LYP/6-31G* energy of 0.14 kcal/mol, and in a concomitant
increase of the ZPVE of 0.15 kcal/mol. Since the two effects cancel almost
q
(21) See, e.g.: Wagner, P. J.; Chen, C. P. J. Am. Chem. Soc. 1976, 98,
239.
exactly, the error in ∆E0 introduced by this approximation should be
minimal.

2382 J. Am. Chem. Soc., Vol. 123, No. 10, 2001
Bednarek et al.
q
Table 2. 0 K Exothermicities, ∆E0, and Activation Barriers, ∆E0 ,
for Enolization of MA•+ to ME•+ via TS1 at Different Levels of
Theory
method f
basis set V ∆E0 ∆ZPVE ∆E0 ∆ZPVE ∆E0 ∆E0 ∆E0 ∆E0
6-31G* 11.41 2.30 5.85 2.13 12.92 7.73 12.60 5.72
B3LYPa
CCSDb
CCSD(T)b
q
q
q
cc-pVDZ 14.26 2.71 3.35 2.21 14.26 4.79 14.19 3.00
6-311G** 15.05 2.78 4.13 2.13 15.09 4.91 15.31 2.73
cc-pVTZ 16.01 2.76 3.70 2.19 17.16 3.61 17.57 1.13
a All species fully optimized and corrected for ∆ZPVE (as listed in
the adjacent columns) with the corresponding basis set. b At the B3LYP/
6-31G* geometries in Cs symmetry,22 including ZPVE differences of
2.76 (MA•+/TS1) and 2.19 kcal/mol (MA•+/ME•+), respectively,
obtained at the same level.
mol at the B3LYP level. We conclude from this that taking
B3LYP geometries for the transition states for such tautomer-
izations introduces an error of (1 kcal/mol in the calculated
q
∆E0 .
In a second step we carried out single-point calculations by
the coupled cluster method, using again different basis sets.22
The corresponding results, listed in Table 2, indicate that the
barrier for enolization of MA•+ appears to be considerably
overestimated by B3LYP/6-31G*. Interestingly, this does not
become evident when basis sets of DZ-quality are employed,
and/or if triple exciations are excluded from the CC treatment.
However, at the CCSD(T)/cc-pVTZ level, the classical barrier
drops to almost 1 kcal/mol after subtraction of the ZPVE
difference between MA•+ and TS1.23 If one employs the QCISD
geometries for MA•+ and TS1, the barrier increases by 0.9 kcal/
mol at the CCSD(T)/cc-pVTZ level, in line with the above
estimate of the error introduced by the choice of the B3LYP
geometry for TS1.
From the above findings we conclude that enolization
proceeds diabatically in the σ-radical state of â-alkylenone
radical cations to the transition state for CH2 rotation in the
resulting enol radical cations, but that the B3LYP/6-31G*
method overestimates barriers for such enolizations by about 5
kcal/mol. The disruption of the aromatic π-system on enolization
of methylindanone radical cations from their σ-radical states
will lead to some changes in the thermochemistry and, hence,
also in the activation parameters for this reaction, but it is not
expected that the general mechanism is affected.
More importantly, however, the question arises if and how
enolization will proceed if the ground state of an enone is of
π-radical nature, in which case the oxygen pσ NBMO is doubly
occupied and hence poorly predisposed to function as an
acceptor orbital for the migrating hydrogen atom (see Figure
7). Will the partially vacant pπ-AO of the oxygen atom function
as an acceptor, thus necessitating a nonplanar transition state
for enolization, or will it be more economical to promote the
system to the reactive σ-radical state to effect hydrogen
migration? Before looking into this question, we need to assess
the nature of the ground states of the methylindanone radical
cations at their equilibrium geometries.
Figure 9. Energy profile for the enolization of MI•+ and DMI•+
according to B3LYP/6-31G* calculations. All energies are relative to
the π-radical ground state of the two methylindanone radical cations.
For a discussion see the text.
the plane containing the heavy atoms, geometries can be
obtained with the economical B3LYP method which has proven
above to be quite reliable for providing equilibrium geometries.
The results of this exercise are depicted schematically in Figure
9 while the relative energies are listed in Table 1 (absolute
energies and geometries of all stationary points are available in
the Supporting Information).
From Figure 9 it becomes apparent that the behavior of MI•+
and DMI•+ on geometry optimization of the 12A′ and the 12A′′
states is very similar: in both cases, relaxation of the π- and
σ-radical states from the neutral geometry leads to stabilizations
of about 4.3 and 1.5 kcal/mol, respectively, while the “opposite”
state is destabilized by 5.9 and 5.2 kcal/mol, respectively. This
is in line with the observation that the geometries of the two
states are very similar in both cases (cf. Supporting Information).
Thus, the methyl group in the 4-position has no discernible
influence on the changes in structure and energy on relaxation
from the neutral geometry.
However, although MI•+ and DMI•+ have different ground
states at their neutral geometries (cf. 3.1) the π-radical state is
predicted to be more stable after relaxation in both cases, albeit
not by much in the case of MI•+. This is in line with the
observation of optical spectra for these cations (DMI•+, Figure
4a; MI•+, Figure 5c) which are in good agreement with CASPT2
predictions for species with π-radical ground states, whereas
the excited-state structure of the corresponding σ-radical cations
would be incompatible with these spectra (see Section 3.4). Thus
we conclude that on ionization in Ar the π-radical states of MI•+
and DMI•+ are formed and that enolization must occur from
these states.
3.3. Energies of π- and σ-Radical States of Methylin-
danones. To this end we carried out geometry optimizations
of both the π- and the σ-radical states of MI•+ and DMI•+.
Since the two states differ by their symmetry with regard to
Thus we proceeded to evaluate the activation parameters for
enolization of the two methylindanone radical cations by
searching for transition states analogous to TS1 and TS2 in
Figure 8. The results are listed in Table 3 which shows that, if
one takes the energies of the relaxed σ-radical states of MI•+
and DMI•+ as a reference, the 0 K exothermicity ∆E0 for
(23) Unfortunately we found it impossible to employ higher members
of Dunning’s series of correlation-consistent basis sets, which would have
permitted an extrapolation of the CCSD(T) results to the complete basis
set (CBS) limit. However, in view of the almost 2 kcal/mol decrease on
going from the double- to the triple-ú basis set, it is to be expected that the
barrier for enolization of MA•+ will disappear almost completely at the
CBS limit.

Tautomerization in Methylindanones
J. Am. Chem. Soc., Vol. 123, No. 10, 2001 2383
Table 3. Energetics of the Enolization of MI•+, and DMI•+
Evaluated at the B3LYP/6-31G* Level (all energies are relative to
the σ-radical state (2A′) at its equilibrium geometry)
∆E
(0)a
-6.2
+12.3
+3.6
-10.1
(0)c
-0.5
+11.4
+3.6
-8.9
∆ZPVE
∆E + ∆ZPVE
DMI•+ (2A′)
DMI•+ (2A′′)
TS1
(0)b
-0.4
-2.6
(0
(0)
-6.6
+9.7
+3.6
-8.4
(0)
-1.4
+8.4
+2.9
-8.4
TS2
DME•+
MI•+ (2A′)
MI•+ (2A′′)
TS1
1.7
(0)d
-0.9e
-3.0
-0.7
+0.5
TS2
ME•+
a Total energy ) -501.33610 hartrees. b ZPVE ) 124.1 kcal/mol.
c Total energy ) -462.01571 hartrees. d ZPVE ) 107.5 kcal/mol
(PBE1PBE value: 107.2 kcal/mol). e Due to a problem with the
vibrational calculation with the B3LYP functional we resorted the
PBE1PBE functional for the calculation of the ZPVE (108.1 kcal/mol).
Figure 10. Plot of log(k) ) log(kA) ‚ log(Q) vs Ea for hydrogen or
deuterium atom transfer by quantum mechanical tunneling trough a
parabolic barrier of 1.25 Å width at 12 K (where the contribution of kA
is negligible for Ea > 3 kcal/mol) according to the Bell model (see
text). Vertical dashed lines delineate the ranges of Ea that are compatible
with experimental observations for MI•+ and DMI•+, respectively. The
black bars at the bottom represent the barriers calculated by B3LYP/
6-31G*.
enolization is the same in both cases (-8.4 kcal/mol). In line
with qualitative expectations (disruption of the aromatic π-sys-
tem in the σ-radical state on enolization) ∆E0 is smaller by about
4 kcal/mol than in MA•+. Consequently, the activation energies
∆E0 are a few kcal/mol higher than in MA•+, but they also
q
simple model26 assumes a barrier of parabolic shape that hass
at the vibrational level at which a particle of mass m tunnelss
a width b and a heigth Ea. Under these premises, the correction
factor Q(T) that must be applied to the classical Arrhenius rate
constant kA(T) ) A exp(-Ea/RT) takes the simple form
differ by only 0.9 kcal/mol (1.3 kcal/mol after ZPVE correction)
between MI•+ and DMI•+.
However, if we evaluate the exothermicities and the activation
barriers for enolization relative to the π-radical ground states
of MI•+ (∆E0 ) -6.0 kcal/mol, ∆E0 ) 9.8 kcal/mol) and
q
eR
â - a
DMI•+ (∆E0 ) -1.8 kcal/mol, ∆E0 ) 16.3 kcal/mol) then
q
Q(T) )
(â‚e-R - R‚e-â
)
q
differences of 4.2 kcal/mol in ∆E0 and 6.5 kcal/mol in ∆E0
turn up. The latter difference is largely due to the higher energy
that is required to promote the keto radical cation to the reactive
σ-radical state in the case of DMI•+.24 Thus, the effect of the
methyl group in the 4-position is not on the energetics of the
enolization per se as it proceeds from the reactive σ-radical state,
but on the energy of the unreactive π-radical ground state which
is stabilized by 5.2 kcal/mol in DMI•+ as compared to MI•+.
An exhaustive search for alternative transition states that would
allow a hydrogen atom transfer directly from the π-radical state
did not lead to any stationary points, hence we conclude that
enolization must proceed through the σ-radical state.
It is interesting to note that, with ∆E0 ) -1.8 kcal/mol,
DMI•+ is a borderline case with regard to spontaneous eno-
lizaton. Attachment of more electron releasing substituents to
the benzene ring of indanone or extension of the π-system will
stabilize the π-radical state to the point where enolization
becomes endothermic. As a matter of fact, the methylfluorenone
analogue of MI gives a spectrum on ionization in Ar that is
very similar to that of parent fluorenone, with no indication
that any enol radical cation has been formed.25
where
π2
h
Ea
R )
and â ) b 2mE
x
a
RT
This model predicts correctly that the overall rate becomes
nearly independent of temperature at the cryogenic temperatures
that prevail in our experiments, and that the kinetic isotope effect
becomes very large for barriers of the kind that we are concerned
with in the present case.27 According to our calculations, the
hydrogen atom must travel over a distance of b ≈ 1.25 Å in
the methylindanone radical cations that are at the focus of our
study.28 In Figure 10 we plot the rate constant log(k) ) log(kA)
+ log(Q) for T ) 12 K and b ) 1.25 Å. Thereby we used A(12
K) ) 1.83 × 1010 s-1 which we derived from ∆Sq ) -30
J‚mol-1‚K-1 calculated by B3LYP/6-31G* for MI•+ and the
corresponding transition state TS1 for its enolization.
From our experiments we know that in the case of MI•+ the
rate must be in the s-1 to min-1 range, whereas the lifetime of
•+
MI-d3 must be in the dozens of hours (log(k) < 10-5 s-1).
3.4. Tunneling Kinetics. The observed temperature inde-
pendence of the enolization of DMI•+ (cf. Figure 3) and the
fact that deutration of the methyl group all but shuts down this
process at the temperatures attainable in Ar matrix experiments
indicates very strongly that quantum mechanical tunneling is
involved. We had previously been able to model the kinetics
of enolizations quite successfully with the Bell model.6 This
On the other hand, the rate of enolization of DMI•+ is in the
h-1 range (log(k) ) 10-4-10-5 s-1).
The above observations define ranges for Ea of the enolization
in the methylindanone radical cations which are indicated by
vertical dashed lines in Figure 10. It is interesting to note that
the difference in Ea for enolization of MI•+ and DMI•+ is in
same range as the difference in the classical barriers computed
(24) If we take both states at their relaxed geometries this difference
amounts to 5.2 kcal/mol. However, the lowest point on the conical
intersection subspace between the π- and the σ-radical states, through which
the reaction must proceed if it follows a classical mechanism, will lie at a
higher energy.
(25) Zhu, Z. Ph.D. Thesis No. 1152, University of Fribourg, Switzerland,
1997.
(26) Caldin, E. F. Chem. ReV. 1969, 69, 135.
(27) Bednarek, P. Ph.D. Thesis No. 1301, University of Fribourg,
Switzerland, 2000.
(28) The distance b is 1.28 Å if the reaction occurs from the π-radical
state, 1.22 Å if it occurs from the π-radical state. The dependence of the
tunneling rates on b is not very pronounced, therefore we chose the average
value for our simulations.

2384 J. Am. Chem. Soc., Vol. 123, No. 10, 2001
Bednarek et al.
Table 4. Excited States of DMI•+
.
TD-PBE0a
TD-B3LYPa
CASPTb
exptlc
∆E [eV]
main excitation
CASSCFb,f
∆E [eV]
fd,e
∆E [eV]
fd,e
∆E [eV]
fd
0.32
0.98
2.77
<10-4
<10-4
0.0047
0.09
0.90
2.63
<10-4
0.0004
0.0350
0.81
0.80
2.69
1 × 10-6
87%
83%
57%
9%
nO f π4
0.0002
π3 f π4
π2 f π4
π3 f π6
π4 f π5
π1 f π4
π2 f π4
π4 f π5
π3 f π6
π1 f π4
0.0313
2.58
9%
3.50
3.92
0.0024
0.0021
3.36
3.91
0.0227
0.0121
3.84
4.04
0.0297
0.160
>4.1
>4.1
29%
17%
27%
15%
13%
a Calculated with the 6-31G* basis set. Note that both TD-DFT methods predict additional transitions37 that are, however, too weak to be observed
or lie outside the observation range. b Calculated with the ANO-L basis set, active space: (15,12)/[1σ,11π]. c See the spectrum of DMI-d3 in
•+
Figure 4a. d Oscillator strength for electronic transition. e The absolute values of the oscillator strengths from the TD-DFT output of Gaussian 98
cannot be compared to those from the CASPT2 calculations.58 f See MOs in Figure 11.
Table 5. Excited States of MI•+
TD-PBE0a
TD-B3LYPa
∆E [eV]
CASPTb
exptlc
∆E [eV]
main excitation
CASSCFb,f
∆E [eV]
fd,e
fd,e
∆E [eV]
fd
0.10
0.82
2.62
<10-4
<10-4
0.0036
-0.14
0.74
2.48
<10-4
<10-4
0.0283
0.42
0.65
2.73
4 × 10-8
6 × 10-6
0.0328
-
88%
nO f π4
π3 f π4
π2 f π4
π3 f π6
π2 f π4
π1 f π4
π4 f π5
-
81%
54%
12%
37%
34%
71%
2.58
3.41
0.0016
3.29
0.0195
3.61
0.0029
>4.1
>4.1
4.62
0.0068
4.53
0.0392
4.30
0.1447
a Calculated with the 6-31G* basis set. Note that both TD-DFT methods predict additional transitions to those listed above37 that are, however,
too weak to be observed or lie outside the observation range with the 6-31G* basis set. b With the ANO-L basis set, active space: (15,12)/[3σ,9π].
•+
c See the spectrum of MI-d3 in Figure 5c. d Oscillator strength for electronic transition. e The absolute values of the oscillator strengths from the
TD-DFT output of Gaussian 98 cannot be compared to those from the CASPT2 calculations58 f See MOs in Figure 11.
by B3LYP/6-31G* (solid bars at the bottom). However, these
barriers are too low in both cases, which is not surprising in
view of our above coupled cluster calculations on MA•+ which
indicate that B3LYP/6-31G* overestimates barriers for such
enolizations by a few kcal/mol. If one corrects for this, then
the calculated barriers fall within the ranges predicted by the
Bell model. But then a more sophisticated treatment of tunneling,
paired with a more reliable estimate of the thermal barrier, would
probably be needed to arrive at an entirely consistent picture
of the kinetics of the enolization of methylindanone radical
cations.
DME•+ and ME•+, observed in our experiments. To this end
we carried out CASSCF/CASPT2 calculations which have
proven to yield rather reliable predictions of excited-state
energies and electric dipole transition moments in previous
applications to radical cation spectra.18,30-35 In addition we did
some calculations with the recently introduced methods based
on time-dependent density functional response theory (TD-
DFT).36
The results of these calculations which were all carried out
at the B3LYP equilibrium geometries of the radical cations are
listed in Tables 4-7 and indicated graphically by gray bars in
Figures 1, 4, and 5. From these data we gather that the CASPT2
predictions are generally in good agreement with experiment,
in line with previous findings for similar cases.10,18 The most
conspicuous deviation is found for the third excited state of the
enol radical cations (cf. Figures 1c and 5d), but even this
deviation is within the expected limits of accuracy of this model
((0.2 eV). The agreement of the predictions by the much more
One question that remains to be answered is why, in view of
•+
the long lifetime of MI-d3 generated photochemically from
the correspondig enol radical cation (Figure 5c), a substantial
amount of this enol radical cation is present after ionizaton, in
contrast to DMI-d3•+, where no enolization is observed on
ionization (Figure 4a). The reason for this is that on charge
transfer from Ar (which is the predominant ionization mecha-
nism of embedded substrates on X irradiation of Ar matrices29)
the incipient substrate radical cations are imparted with a
substantial amount of excess energy that they cannot easily
dissipate in solid Ar. This excess energy is therefore available
to drive thermal rearrangements, such as the present tautomer-
izations. The extent to which such reactions occur depends on
the heigth of the barriers to be crossed. Apparently, this barrier
is low enough in MI•+ but too high in DMI•+, in accord with
the calculated 6.5 kcal/mol difference.
3.5. Electronic Structure of the Observed Radical Cations.
Above we had claimed that the observed spectra of DMI•+
(Figure 4a) and MI•+ (Figure 5c) are only compatible with a
π-radical ground-state structure of the two species. We wish to
substantiate this claim by electronic structure calculations which
will also encompass the corresponding enol radical cations,
(29) Bally, T. In Radical Ionic Systems; Lund, A., Shiotani, M., Eds.;
Kluwer: Dordrecht, The Netherlands, 1991.
(30) Fu¨lscher, M. P.; Matzinger, S.; Bally, T. Chem. Phys. Lett. 1995,
236, 167.
(31) Zhu, Z.; Bally, T.; Wirz, J.; Fu¨lscher, M. J. Chem. Soc., Perkin
Trans. 2 1998, 1083.
(32) Bally, T.; Carra, C.; Fu¨lscher, M. P.; Zhu, Z. J. Chem. Soc., Perkin
Trans. 2 1998, 1759.
(33) Bally, T.; Carra, C.; Matzinger, S.; Truttmann, L.; Gerson, F.;
Schmidlin, R.; Platz, M. S.; Admasu, A. J. Am. Chem. Soc. 1999, 121,
7011.
(34) Marcinek, A.; Adamus, J.; Huben, K.; Gebicki, J.; Bartczak, T. J.;
Bednarek, P.; Bally, T. J. Am. Chem. Soc. 2000, 122, 437.
(35) Bally, T.; Bernhard, S.; Matzinger, S.; Roulin, J.-L.; Sastry, G. N.;
Truttmann, L.; Zhu, Z.; Marcinek, A.; Adamus, J.; Kaminski, R.; Gebicki,
J.; Williams, F.; Chen, G.-F.; Fu¨lscher, M. R. Chem. Eur. J. 2000, 6, 858.
(36) Casida, M. E. In Recent AdVances in Density Functional Methods,
Part I; Chong, D. P., Ed.; World Scientific: Singapore, 1995; p 155.

Tautomerization in Methylindanones
J. Am. Chem. Soc., Vol. 123, No. 10, 2001 2385
Table 6. Excited States of DME•+
TD-PBE1PBE
TD-B3LYP
∆E [eV]
CASPTb
∆E [eV]
exptlc
∆E [eV]
main excitation
CASSCFb,d
∆E [eV]
f
f
f
2.23
0.0020
2.16
0.0206
7 × 10-4
1.87
1.74
52%
-26%
66%
42%
+12%
37%
π4 f π5
π3 f π4
π2 f π4
π3 f π4
π4 f π5
π4 f π6
π2 f π5
π1 f π4
2.90
3.15
0.0052
0.0128
2.86
3.11
0.0431
0.1142
0.0500
0.0328
2.58
2.73
2.53
2.95
3.75
0.0026
3.71
0.0094
0.1447
4.30
-11%
-11%
a With the 6-31G* basis set. b With the ANO-L basis set, active space: (15,12)/[1σ,11π]. c See the spectrum of DME•+ in Figure 1c. d See MOs
in Figure 11.
Table 7. Excited States of ME•+
TD-PBE1PBE
∆E [eV]
TD-B3LYP
∆E [eV]
CASPTb
∆E [eV]
exptlc
∆E [eV]
main excitation
CASSCFb,d
f
f
f
2.35
0.0024
2.26
0.0241
1.93
0.0024
1.88e
52%
-24%
63%
38%
+20%
31%
π4 f π5
π3 f π4
π2 f π4
π3 f π4
π4 f π5
π4 f π6
π2 f π5
π1 f π4
2.89
3.26
0.0034
0.0117
2.84
3.25
0.0361
0.0969
2.61
2.78
0.0383
0.1213
2.58
3.00
3.82
0.0056
3.76
0.0283
3.75
0.0310
3.64
-12%
-11%
a With the 6-31G* basis set. b With the ANO-L basis set, active space: (15,12)/[1σ,11π]. c See the spectra in Figure 5a,d. d See MOs in Figure
11. e Average between vertical excitation energies in the first band of ME•+ and ME-d3
.
•+
vs 2.0 kcal/mol, respectively). Since the two electronic states
in question are described well by single configurations this
divergent behavior of the two flavors of DFT theory is puzzling.
Apparently TD-DFT is poorly suited to describe relative energies
of π-radical and σ-radical states in radical ions, and perhaps
also in neutral compounds (n f π* vs π f π* excitations).
Apart from the above low-energy nO f π4 and π3 f π4
excitations, the electronic structure of the methylindanone radical
cations is unsurprising.37 As is often found in radical cations,
the lowest few excited states are described by electron promo-
tions from doubly occupied MOs to the singly occupied MO
(SOMO). Excitations into virtual orbitals only come into play in
the UV range (see Tables 4 and 5). In contrast, the quinoid
chromophore of the enol radical cations gives rise to a pair of
excited states arising from linear combination of the SOMO-1
f SOMO (π3 f π4) and SOMO f LUMO excitations, as is typical
for polyenic radical cations.30 These excitations lead to a weak
red (negative linear combination) and a strong blue absorption
(positive linear combination) in the spectrum. Between the two
lies the second excited state formed by π2 f π4 excitation.
According to CASPT2 the sharp peak at ca. 340 nm in MI•+
(Figure 5) is due to π2 f π6 excitation (cf. MOs on the right
side of Figure 11).
Figure 11. Moplot representation of the molecular orbitals involved
in the excitations of DMI•+ and DME•+ listed in Tables 4 and 6. The
corresponding orbitals for MI•+ and ME•+ (cf. Tables 5 and 7) look
very similar.
economical TD-DFT methods with the observed bands is
generally also satisfactory, although not as good on the average
as those obtained at the CASPT2 level. However, substantial
discrepancies are found between the TD-DFT and the CASPT2
predictions for the first excited state of MI•+ and DMI•+, which
arises by promotion of an electron from the nO to the π4 MO
(or by electron ejection from the nO MO in the neutral
precursors). This state is predicted much too low, especially by
TD-B3LYP which even places it 3.2 kcal/mol below the ground
state at its equilibrium geometry! This is all the more surprising
as the same state was predicted to lie 8.7 kcal/mol aboVe the
π-radical ground state by separate B3LYP calculations on these
two states, in excellent accord with CASPT2 which predicted
this gap to be 9.7 kcal/mol (see Table 1). The same discrepancy
between B3LYP and its TD variant is found for DMI•+ (14.0
Finally, we calculated also the electronic spectrum that would
be expected if MI•+ had a σ-radical ground state, a possibility
that we had to admit in view of the results shown in Figure 9.
However, according to TD-B3LYP only a very weak transition
is expected in the range of the intense absorption of ionized
MI, whereas a rather strong one is predicted to occur around
340 nm where the observed spectrum shows no bands. In
addition, the IR spectrum of MI-d3•+ (not shown) shows a strong
(37) In addition to the transitions that are listed in Tables 4 and 5, the
TD-DFT methods predict a number of nO f π* transitions that carry,
however, very small oscillator strengths. Hence they do not manifest
themselves in the experimental spectra and they were therefore not calculated
by CASPT2 and are not listed in the tables. Also, TD-DFT predicts more
π f π* transitions than CASPT2 in the UV region but these are also not
listed because they lie outside the present observation range. The full results
of the TD-DFT calculations are available in the Supporting Information..

2386 J. Am. Chem. Soc., Vol. 123, No. 10, 2001
Bednarek et al.
none according to one of the procedures described by Khalaf et al.42
band at 1749 cm-1, in accord with the B3LYP prediction for
A
2
the A′′ state (1785 cm-1 after scaling by 0.97), but in stark
solution of â-chloropropionyl chloride (0.105 mol) and p-xylene (0.1
mol) in 13 mL of CS2 was added slowly to a mixture of 16 g of AlCl3
and 65 mL of CS2. After the mixture was stirred for 3 h at room
temperature, the CS2 was removed under vacuum and 125 mL of
concentrated H2SO4 was added to the remaining oily residue. The
mixture was heated to 90 °C for 45 min, then cooled and poured onto
ice. The resulting solid was extracted with diethyl ether and benzene.
The extracts were washed with sodium carbonate solution and water
and dried over sodium sulfate, and the solvents were removed under
vacuum. The crude product was crystallized from methanol (68%, mp
77 °C, lit.42 mp 76-77 °C).
2
disagreement with the prediction for the A′ state where the
CdO stretching mode is shifted to 1478 cm-1 (after scaling by
0.97). Thus we conclude that MI ends up predominantly in the
π-radical minimum on ionization, although vertical electron loss
favors the σ-radical state according to B3LYP (and, marginally,
also to CASPT2).
4. Conclusions
The radical cation formed on ionization of 4,7-dimethylin-
danone in Ar at 12 K is stable for hours but undergoes slow
tautomerization at a rate that is independent of temperature up
to 30 K. The resulting enol radical cation can be reverted
photochemically to the keto form. In contrast, the 7-methylin-
danone radical cation does not persist under these conditions
but decays spontaneously to the corresponding enol radical
cation which furthermore appears to be photostable. Deuteration
of the methyl groups completely suppresses enolization in both
methylindanone radical cations, such that ionized 7-methylin-
danone can now be observed leisurly, too.
The surprising effect of the 4-methyl group, which is distant
from the site of tautomerization, can be traced back to an
electronic factor, i.e., the stabilization of the unreactive π-radical
relative to the reactive σ-radical state of the methylindanone
radical cation. B3LYP calculations of the reaction profile show
that the difference in activation energies can be largely ac-
counted for by the promotion from the π-radical to the σ-radical
state that is nearly thermoneutral in the 7-methylindanone but
requires about 6.5 kcal/mol in the 4,7-dimethylindanone case.
This example represents the first demonstration of state selectiv-
ity in radical ion rearrangements. The above observations and
simulations of the observed kinetics on the basis of the Bell
model for hydrogen atom tunneling show that this is the
dominant mechanism for enolization, in agreement with previous
findings on related systems.8,9
It has been known for a while that aromatic ketones whose
lowest lying triplet states are of π f π* nature are notoriously
less reactive in photoreduction (H-abstraction) than those where
they are of n f π* nature.38,39 The cause of this difference in
reactivity is the same as that which appears to be operative in
the present case, i.e., the presence or absence of a singly
occupied nO orbital in one or the other electronic state of the
ketone (or its radical cation). However, the exact mechanism
by which this difference in reactivity, which increases smoothly
as the energy difference between the two states changes, comes
about is still a matter of debate (thermal equilibration or vibronic
coupling between with the two states).12,40
To our best knowledge, the question of state selectivity has
never been addressed in photoenolizations involving o-alkyl
aromatic ketones, which are, however, known to have close-
lying n f π* and π f π* triplet states.41 It would come as no
surprise if substituents wold have an equally dramatic effect
on the rate of enolization as has been found in the present case
of radical cations.
(b) 4,7-Bis(trideuteriomethyl)indanone (DMI-d6)13 was prepared
analogous to DMI starting from perdeuterated p-xylene.
(c) 7-Methylindanone was made by cyclization of 3-methylhydro-
cinnamic acid with polyphosphoric acid, a procedure that has been
employed successfully for the synthesis of a range of substituted
indanones.43 Typically, a mixture of 1 g of 3-methylhydrocinnamic acid
in 7 mL of polyphosphoric acid was heated to 80 °C for 3 h with
occasional stirring. The reaction mixture was poured into water and
extracted with diethyl ether. The mixture of 5-methyl- and 7-meth-
ylindanone that was obtained in this way was separated by column
chromatography on silica 60F with methylene chloride to give MI in
about 38% yield (mp 52-54 °C, lit.44 mp 54.7-55.2 °C; 1H NMR, IR,
and UV spectra in accord with literature reports44,45).
(d) 3-Methylhydrocinnamic acid was obtained by Perkin condensa-
tion46 of 3-methylbenzaldehyde with acetic anydride in the presence
of potassium acetate and subsequent reduction of the cinnamic acid
with sodium amalgam.47 3-Methylbenzaldehyde (5 mL, 7.7 g; Fluka)
and 4 g of AcOK were dissovled in 9 mL of Ac2O and heated to 155
°C in a dry flask for 6 h. The resulting reaction mixture was poured
onto 100 mL of water and alkalified with sodium carbonate. Diethyl
ether extracts were washed with water and dilute HCl whereupon the
crude cinnamic acid was filtered and recrystallized from EtOH (yield
6.5 g, 84%, mp 114 °C). For the reduction, a solution of 1 g of
3-methylcinnamic acid in 15 mL of 1 M NaOH was added during 15
min to 16 g of sodium amalgam and kept at 50 °C for two more hours
after the addition was terminated. After decanting the mercury the
product was washed with water and the aqueous phases were acidified
with 50% HCl. The crude product was then exctracted with diethyl
ether. After drying and evaporation, 740 mg of a clear pale yellow oil
was obtained which spontaneously crystallized on standing (mp 38-
42 °C, lit.47 mp 42-43 °C).
(e) 7-(Trideuteriomethyl)indanone (MI-d3) was prepared analogous
to MI starting from 3-(trideuteriomethyl)benzaldehyde that was obtained
by reductive condensation of 3-bromobenzaldehyde with CD3I following
the procedure described by Chapman et al.48 To avoid deuterium
exchange in the methyl group during the cinnamic acid reduction, we
substituted NaOH by NaOD, which led of course to incorporation of
two deuterium atoms in the propionic acid chain. However, during the
subsequent cyclization with polyphosphoric acid, these deuterium atoms
were partially lost again. Mass spectra indicated that the final product
still contained one additional deuterium atom that according to the 1H
NMR spectrum is located â to the carbonyl group. However, as the
presence of this additional deuterium atom was not expected to change
the spectroscopic properties of interest in the present study, no efforts
were made to arrive at a sample of isotopically pure MI-d3 (which,
strictly speaking, we should therefore call MI-d4).
5.2. Matrix Isolation and Spectroscopy. Crystals of the compounds
were placed in a U-shaped tube immersed into a water bath and
connected to the inlet system of a closed-cycle cryostat. While the bath
5. Experimental Section
(42) Khalaf, A. A.; Abdel-Nahab, A. M. A.; El-Khawaga, A. M.; El-
Zahry, M. F. Bull Soc. Chim. Fr. Part II 1984, 285.
(43) Koo, J. J. Am. Chem. Soc. 1953, 75, 1891.
(44) House, H. O.; Rasmusson, G. H. J. Org. Chem. 1963, 28, 31.
(45) Buys, T. S. V.; Cerfontain, H.; Geenevasen, J. A. J.; Stunnenberg,
F. Recl. TraV. Chim. Pays-Bas 1986, 105, 188.
(46) Johnson, J. R. In Organic Reactions; Wiley & Sons: New York,
1942; Vol. 1, p 210.
5.1. Syntheses: (a) 4,7-Dimethylindanone (DMI) was made by H2-
SO4/AlCl3 catalyzed cyclization of 2′,5′-dimethyl-3-chloropropiophe-
(38) Pitts, J. N.; Johnson, H. W.; Kuwana, T. J. Phys. Chem. 1962, 66,
2456.
(39) Wagner, P. J.; Kemppainen, A. E. J. Am. Chem. Soc. 1968, 90,
5898.
(40) Wagner, P. J. In Rearrangements in Ground and Excited States; de
Mayo, P., Ed.; Academic Press: New York, 1980; Vol. 3, p 381.
(41) Lamola, A. A. J. Chem. Phys. 1967, 47, 4810.
(47) Dippy, J. F. J.; Page, J. E. J. Chem. Soc. 1938, 1938, 357.
(48) Chapman, O. L.; Johnson, J. W.; McMahon, R. J.; West, P. R. J.
Am. Chem. Soc. 1978, 110, 508.

Tautomerization in Methylindanones
J. Am. Chem. Soc., Vol. 123, No. 10, 2001 2387
was kept between 0 and -10 °C (depending on the total surface of the
crystals in the U-tube) a mixture of argon and CH2Cl2 (1000:1) was
flowing through the tube at a rate of ≈1 mmol/h and swept the
compounds onto a CsI window held at 19 K. There the mixture
accumulated to form a matrix containing a sufficient quantity of the
compound within 2 h.
After taking reference spectra, the samples were exposed to 90 min
of X-irradiation as described previously.29 Photolyses were effected
with a 1 kW Ar plasma discharge lamp through appropriate low-pass
cutoff or interference filters. Electronic absorption (EA) spectra were
taken between 200 and 1200 nm with a Perkin-Elmer Lambda 19
instrument, whereas IR spectra were obtained on a Bomem DA3
interferometer (1 cm-1 resolution) equipped with an MCT detector
(500-4000 cm-1).
Gas-phase photoelectron spectra were measured on a modified
Perkin-Elmer PE 16 instrument operated in pre-retardation (and hence
constant resolution) mode.49 Calibration was effected with a Xe-Ar
mixtured and the spectral resolution was about 20 meV.
5.3. Quantum Chemical Calculations. The geometries of all species
were optimized by the B3LYP density functional method50,51 as
implemented in the Gaussian 98 programs,52,53 using the 6-31G* basis
set. Relative energies and vibrational spectra were calculated at the
same level. Full sets of Cartesian coordinates and absolute energies
(including thermal corrections where available) are given in the
Supporting Information.
In addition we resorted to a recently introduced molecular density
functional method based on time-dependent response (TDR) theory,36
according to which the poles and the residues of the frequency-
dependent polarizability are evaluated, where the former correspond
to vertical excitation energies and the latter to oscillator strengths.
Analogous to the Hartree-Fock variant of TDR (in the popular Tamm-
Dancoff or Random-Phase approximation), excited states are then
described in terms of CI vectors containing single excitations from a
ground state wave function (CIS), which makes for a very transparent
interpretation of the results. However, it should be noted that the
treatment of electron correlation in TD density functional response
theory (TD-DFT) goes far beyond that in a HF-based CIS method which
makes the results generally more accurate.
We used the variant of TD-DFT described recently by Stratmann et
al.57 and implemented in the Gaussian 98 program.52,58 Next to the
B3LYP functional we employed also a recently introduced hybrid
variant of the nonempirical Perdew-Burke-Ernzerhof (PBE)
functional59-61 that was shown to be well suited for the evaluation of
excitation energies of radicals.62 The results obtained with the two
functionals differ somewhat, but one is not consistently in better
agreement with experiment than the other.
The MOs in Figures 6 and 11 were plotted with the Moplot
program,63 which gives a schematic representation of the MO’s nodal
structure within a ZDO-type approximation.64
Acknowledgment. This paper is dedicated to Prof. Edwin
Haselbach (University of Fribourg) on occasion of his 60th
birthday, in gratitude for his continuing advice, for many
exciting discussions, and for his friendship. This work was
supported by the Swiss National Science Foundation (Project
No. 2000-061560.00) and by grants from the Polish State
Committee for Scientific Research (No. 3/T09A/122/18). We
are very indebted to Prof. Michael Allan and Anne-Christelle
Sergenton (University of Fribourg) for measuring the photo-
electron spectra shown in Figure 6.
Excited-state calculations were carried out at the B3LYP/6-31G*
geometries of the radical cations (except where indicated otherwise)
by the CASSCF/CASPT2 procedure54 with the MOLCAS program55
using the [C,N,O]3s2p1d/[H]2s ANO basis set.56 The active spaces were
chosen such as to obtain a satisfactory description of all excited states
of interest at the CASPT2 level. The resulting active spaces are
described in the footnotes to Tables 4-7. To ensure orthogonality, the
CASSCF wave functions were averaged over all excited states of the
same symmetry. Transition moments were calculated on the basis of
these wave functions, using CASPT2 energy differences in the
denominator.
Supporting Information Available: Energies and Cartesian
coordinates from the B3LYP calculations for all stationary points
mentioned in this study (minima and transition states) as well
as full results from TD-B3LYP calculations (ASCII). This
material is available free of charge via the Internet at
http://pubs.acs.org.
(49) Dressler, R.; Neuhaus, L.; Allan, M. J. Electron Spectrosc. Relat.
Phenom. 1983, 31, 181.
(50) Becke, A. D. J. Chem. Phys. 1993, 98, 5648.
(51) Lee, C.; Yang, W.; Parr, R. G. Phys. ReV. B 1988, 37, 785.
(52) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb,
M. A.; Cheeseman, J. R.; Zakrzewski, V. G.; Montgomery, J. A.; Stratmann,
R. E.; Burant, J. C.; Dapprich, S.; Millam, J. M.; Daniels, A. D.; Kudin, K.
N.; Strain, M. C.; Farkas, O.; Tomasi, J.; Barone, V.; Cossi, M.; Cammi,
R.; Mennucci, B.; Pommelli, C.; Adamo, C.; Clifford, S.; Ochterski, J.;
Petersson, G. A.; Ayala, P. Y.; Cui, Q.; Morokuma, K.; Malick, D. K.;
Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Cioslowski, J.; Ortiz, J.
V.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.;
Gomperts, R.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.; Peng,
C. Y.; Nanayakkara, A.; Challacombe, M.; Gill, P. M. W.; Johnson, B. G.;
Chen, W.; Wong, M. W.; Andres, J. L.; Gonzales, C.; Head-Gordon, M.;
Repogle, E. S.; Pople, J. A. Gaussian 98; Rev. A1 ed.; Gaussian, Inc.:
Pittsburgh, PA, 1998.
JA003708M
(57) Stratmann, R. E.; Scuseria, G. E.; Frisch, M. J. J. Chem. Phys. 1998,
109, 8218.
(58) The transition moments that appear in the TD-DFT output of
Gaussian should probably be taken with a grain of salt, because their absolute
values vary by orders of magnitude depending on the number of excited
states (NStates)nn) that are calculated.
(59) Ernzerhof, M.; Scuseria, G. E. J. Chem. Phys. 1999, 110, 5029.
(60) Adamo, C.; Barone, V. J. Chem. Phys. 1999, 110, 6158.
(61) The same functional was introduced simultaneously under two
different names (PBE1PB59 and PBE060). It is invoked by the PBE1PBE
keyword in Gaussian 98, but we list the results under the simpler heading
of PBE0 in Tables 4-7.
(62) Adamo, C.; Barone, V. Chem. Phys. Lett. 1999, 314, 152.
(63) Bally, T.; Albrecht, B.; Matzinger, S.; Sastry, M. G. MOPLOT, a
program for displaying results of LCAO-MO calculations, available on
request from Thomas.Bally@unifr.ch.; Version 3.2 University of Fribourg,
1997.
(53) Johnson, B. G.; Gill, P. M. W.; Pople, J. A. J. Chem. Phys. 1993,
98, 5612.
(54) Andersson, K.; Roos, B. O. In Modern Electronic Structure Theory;
World Scientific Publ. Co.: Singapore, 1995; Part 1, Vol. 2, p 55.
(55) Andersson, K.; Blomberg, M. R. A.; Fu¨lscher, M. P.; Kello¨, V.;
Lindh, R.; Malmqvist, P.-Å.; Noga, J.; Olson, J.; Roos, B. O.; Sadlej, A.;
Siegbahn, P. E. M.; Urban, M.; Widmark, P.-O. MOLCAS; Version 4;
University of Lund: Lund, Sweden, 1998.
(56) Pierloot, K.; Dumez, B.; Widmark, P.-O.; Roos, B. O. Theor. Chim.
Acta 1995, 90, 87.
(64) Haselbach, E.; Schmelzer, A. HelV. Chim. Acta 1979, 54, 1299.