Stepwise versus direct long-range charge separation in molecular triads

Article abstract of DOI:10.1021/ja983716r

Trifunctional electron donor - donor - acceptor molecules are described in which photoinduced charge separation, D₂ - D₁ - A* → D₂ - D₁⁺ - A^- , is followed by a charge migration step D

Full text of DOI:10.1021/ja983716r

J. Am. Chem. Soc. 2000, 122, 3721-3730
3721
Stepwise versus Direct Long-Range Charge Separation in Molecular
Triads
R. J. Willemse,† J. J. Piet,‡ J. M. Warman,‡ F. Hartl,† J. W. Verhoeven,† and
A. M. Brouwer*,†
Contribution from the Institute of Molecular Chemistry, UniVersity of Amsterdam,
Nieuwe Achtergracht 129, 1018 WS Amsterdam, The Netherlands, and IRI, Delft UniVersity of Technology,
Mekelweg 15, 2629 JB Delft, The Netherlands
ReceiVed October 22, 1998. ReVised Manuscript ReceiVed January 27, 2000
Abstract: Trifunctional electron donor-donor-acceptor molecules are described in which photoinduced charge
+
separation, D2-D1-A* f D2-D1+-A-, is followed by a charge migration step D2-D1+-A- (CS1) f D2
-
D1-A- (CS2), leading to a relatively long-lived charge-separated state. The rate of the charge migration process
could be determined in a range of solvents of low polarity. In benzene and dioxane, reversibility of the process
allowed the determination of the free energy difference between CS1 and CS2. The relative energy of the CS2
state is much lower than expected from simple electrostatic models. An increase of the charge migration rate
was found with increasing solvent polarity within a series of alkyl ethers or alkyl acetates. However, an apparent
preferential stabilization of the CS1 state in acetates relative to ethers leads to discontinuities in the
solvatochromic shift behavior of the CT fluorescence from the CS1 state, and in the increase of the charge
migration rate as a function of dielectric constant. In a reference compound lacking the intermediate redox
unit, direct long-range charge separation yielding a D2+-bridge-A- charge-separated state can occur, but the
yield is significantly lower than in the triads.
Introduction
The overall efficiency in multistep charge separation se-
quences depends critically on the competition between forward
charge separation steps and charge recombination processes. In
the “photosynthetic model systems” described in the literature,
large variations are found in the overall quantum yields and
lifetimes of the ultimate charge-separated state. What we find
particularly striking is the large variation in the rates of
recombination of the charge-separated states, remarkably rapid
recombination being found in some cases even with very
substantial distances between the charged sites.2,9,12,16
Many of the model systems studied require relatively polar
solvents to achieve charge separation, which has the drawback
that they usually do not work in immobilized media (in which
solvent reorganization is highly restricted) or at (very) low
temperatures. Moreover, the use of a polar solvent introduces a
large solvent reorganization term which tends to enhance the
rate of charge recombination considerably.9 Thus, being able
to separate charges in a medium of low polarity is an attractive
The process of photoinduced charge separation via a sequence
of short-range electron transfer steps as it occurs in natural
photosynthesis has inspired many chemists to design molecular
systems which may accomplish the same efficient and long-
lived charge separation. Some groups have used building blocks
that resemble the chromophores of Nature,1-3 and more or less
attempt to mimic the photosynthetic apparatus,4 while others
have implemented multistep charge separation schemes in very
differently constructed organic,5-14 inorganic,15-17 and supramo-
lecular systems.18-20
* Author to whom correspondence should be addressed. Fax +31 20
5255670. E-mail: fred@org.chem.uva.nl.
† University of Amsterdam.
‡ Delft University of Technology.
(1) Gust, D.; Moore, T. A.; Moore, A. L. Acc. Chem. Res. 1993, 26,
198.
(2) Osuka, A.; Marumo, S.; Mataga, N.; Taniguchi, S.; Okada, T.;
Yamazaki, I.; Nishimura, Y.; Ohno, T.; Nozaki, K. J. Am. Chem. Soc. 1996,
118, 155.
(3) Osuka, A.; Mataga, N.; Okada, T. Pure Appl. Chem. 1997, 69, 797.
(4) Wasielewski, M. R.; Gaines, G. L.; Wiederrecht, G. P.; Svec, W.
A.; Niemczyk, M. P. J. Am. Chem. Soc. 1993, 115, 10442.
(5) Mes, G. F.; van Ramesdonk, H. J.; Verhoeven, J. W. J. Am. Chem.
Soc. 1984, 106, 1335.
(6) Brouwer, A. M.; Mout, R. D.; Maassen van den Brink, P. H.; van
Ramesdonk, H. J.; Verhoeven, J. W.; Warman, J. M.; Jonker, S. A. Chem.
Phys. Lett. 1991, 180, 556.
(12) Roest, M. R.; Verhoeven, J. W.; Schuddeboom, W.; Warman, J.
M.; Lawson, J. M.; Paddon-Row, M. N. J. Am. Chem. Soc. 1996, 118,
1762.
(13) Greenfield, S. R.; Svec, W. A.; Gosztola, D.; Wasielewski, M. R.
J. Am. Chem. Soc. 1996, 118, 6767.
(14) Paddon-Row, M. N. Acc. Chem. Res. 1994, 27, 18.
(15) Sauvage, J. P.; Collin, J. P.; Chambron, J. C.; Guillerez, S.; Coudret,
C.; Balzani, V.; Barigelletti, F.; De Cola, L.; Flamigni, L. Chem. ReV. 1994,
4, 993.
(7) Wasielewski, M. R. Chem. ReV. 1992, 92, 435.
(8) Brouwer, A. M.; Eijckelhoff, C.; Willemse, R. J.; Verhoeven, J. W.;
Schuddeboom, W.; Warman, J. M. J. Am. Chem. Soc. 1993, 115, 2988.
(9) Van Dijk, S. I.; Wiering, P. G.; van Staveren, R.; van Ramesdonk,
H. J.; Brouwer, A. M.; Verhoeven, J. W. Chem. Phys. Lett. 1993, 214,
502.
(10) Van Dijk, S. I.; Wiering, P. G.; Groen, C. P.; Brouwer, A. M.;
Verhoeven, J. W.; Schuddeboom, W.; Warman, J. M. J. Chem. Soc.,
Faraday Trans. 1995, 91, 2107.
(16) Harriman, A.; Odobel, F.; Sauvage, J. P. J. Am. Chem. Soc. 1995,
117, 9461.
(17) Treadway, J. A.; Chen, P. Y.; Rutherford, T. J.; Keene, F. R.; Meyer,
T. J. J. Phys. Chem. A 1997, 101, 6824.
(18) Yonemoto, E. H.; Kim, Y. I.; Schmehl, R. H.; Wallin, J. O.;
Shoulders, B. A.; Richardson, B. R.; Haw, J. F.; Mallouk, T. E. J. Am.
Chem. Soc. 1994, 116, 10557.
(19) Keller, S. W.; Johnson, S. A.; Brigham, E. S.; Yonemoto, E. H.;
Mallouk, T. E. J. Am. Chem. Soc. 1995, 117, 12879.
(20) Feldheim, D. L.; Grabar, K. C.; Natan, M. J.; Mallouk, T. E. J.
Am. Chem. Soc. 1996, 118, 7640.
(11) Verhoeven, J. W.; Wegewijs, B.; Hermant, R. M.; Willemse, R. J.;
Brouwer, A. M. J. Photochem. Photobiol. A 1996, 95, 3.
10.1021/ja983716r CCC: $19.00 © 2000 American Chemical Society
Published on Web 04/04/2000

3722 J. Am. Chem. Soc., Vol. 122, No. 15, 2000
Willemse et al.
goal. At the same time it is a challenge to achieve this in an
economical way, without wasting too much energy. An essential
problem is therefore the estimation of the energy levels of
charge-separated states in such media. The usual approach21 is
to start from the redox potentials of the electroactive groups,
which are measured in a polar solvent, and to apply a correction
for the smaller solvation energy in the nonpolar solvent using
a crude dielectric continuum model, typically the simple Born
model. Furthermore, the electrostatic interaction energy accord-
ing to Coulomb’s law is taken into account. The total electro-
static energy correction is typically quite large, and subject to
considerable uncertainty. It is one of the purposes of the present
study to obtain accurate information on the relative energies of
charge-separated states, which may serve to test more detailed
molecular approaches to the estimation of electrostatic solvation
energies which will undoubtedly emerge soon.
In the present study we describe a few trichromophoric
electron donor-donor-acceptor (D2-D1-A) molecules (triads)
which are able to undergo stepwise charge separation in solvents
of low polarity.
D2-D1-A* (LE) f D2-D1+•-A-•
(CS1)
(CS2)
D2-D1+•-A-• (CS1) f D2+•-D1-A-•
The molecules are designed such that the charge separation from
the locally excited (LE) state occurs very rapidly, and with
nearly 100% efficiency. The second step is essentially a thermal
electron transfer process, driven by the difference in oxidation
potentials of the donors, but hampered by the electrostatic attrac-
tion between the charges on D1 and A. This step has to compete
with other decay processes of the CS1 state, notably charge re-
combination to the ground state. The first charge migration step
is of crucial importance in multistep charge separation schemes,
because subsequent steps tend to be easier as the change in the
Coulomb energy and the rates of competing recombination
processes decrease with the distance between the redox sites.
Figure 1. Structure of TC3 as determined by single-crystal X-ray
diffraction. Hydrogen atoms are omitted for clarity. Newman projections
indicate dihedral angles along some relevant bonds.
A special property of the molecules studied here is that the
CS1 state is strongly fluorescent. This allows us to accurately
determine the rates of the process CS1 f CS2 using stationary
and time-resolved fluorescence spectroscopy. The triads studied
are depicted below, together with some relevant model com-
pounds which contain only one or two of the chromophores. In
all four trichromophoric compounds (triads) a vinylcyanonaph-
thalene is the electron-accepting moiety (A) and a saturated
amine acts as the first donor (D1). The triads have differently
substituted anilines as their second electron donor (D2). The
chromophores are linked in a linear arrangement via two
saturated rings. Compound BC3, in which the aliphatic amino
nitrogen of TC3 is replaced by a CH group, is a bichromophoric
D2-bridge-A system designed to test the direct long-range
interaction between the terminal chromophores.
It will be shown that in compounds TC3 and TC4 in nonpolar
solvents two sequential steps of electron transfer occur upon
excitation of the acceptor chromophore. Furthermore, it will be
demonstrated that the rate of the second step of electron transfer
is markedly dependent on solvent polarity. In general, the rate
of the conversion of the CS1 state to the CS2 state tends to be
faster upon increasing the dielectric constant of the medium.
However, the rate of the charge migration step is relatively
slower in acetates than in alkyl ethers (specific solvent effect).
Moreover, we will show that the stepwise electron transfer as
occurring in TC3 is more efficient than the direct long-range
electron transfer which takes place in BC3.
Results and Discussion
Ground State Structure. Compound TC3 is a representative
of the trichromophoric donor-donor-acceptor compound series
TC1-TC2-TC3-TC4. Figure 1 shows the structure of TC3
as determined by single-crystal X-ray diffraction as well as the
Newman projections along some relevant bonds. Bond distances
and bond angles of the non-hydrogen atoms of TC3 are given
in the Supporting Information.
The molecule adopts a stretched conformation with both
piperidine rings in a chair conformation. Although the structure
of the triads allows some rotational freedom around the long
(21) Weller, A. Z. Phys. Chem. Neue Folge 1982, 133, 93.

Charge Separation in Molecular Triads
J. Am. Chem. Soc., Vol. 122, No. 15, 2000 3723
Table 1. Optical Absorption Data of the Triads TCn in
Acetonitrile
compd
λ
max (nm) {ꢀmax (M-1 cm-1)}
TC1
TC2
TC3
TC4
236 {45 790} 314 {13 410}
236 {46 500} 315 {13 220}
236 {46 920} 314 {13 990}
236 {45 350} 314 {15 650}
Figure 3. Fluorescence spectra of dyad BC0 and triads TCn in di-
n-propyl ether; excitation at 308 nm. The spectra are scaled in such a
way that the maximum intensity corresponds with the fluorescence
quantum yield.
of the solvatochromic shift of the emission maximum (Lippert-
Mataga relation):24-26
2µ2
F3hc
νjCT ) νj0CT
-
∆f
(1)
Figure 2. Representative UV absorption spectra (cyclohexane): model
donor D3, model acceptor A0, and triad TC3.
In eq 1 νjCT is the wavenumber of the emission maximum of
the CT band in a given solvent, νj0CT denotes the wavenumber
of the fluorescence in the gas phase, and F is the effective radius
of a cavity which the molecule occupies in the solvent. The
solvent polarity function ∆f is
axis of the molecules, this does not alter the distances between
the chromophores. The distance between the first donor (N2)
and the second donor (N1) is 4.3 Å; the edge to edge distance
between the two terminal chromophores (N1-C14) is 7.2 Å.
From the experimentally determined excited state dipole mo-
ments (see below), which reflect the actual charge distributions
in the charge-separated states, we can calculate the correspond-
ing distances between two unit point charges. This gives
effective donor-acceptor distances R(D1-A) ≈ 5.2 Å and
R(D2-A) ≈ 9.7 Å, in good agreement with the molecular
structure.
(n2 - 1)
(ꢀs - 1)
(2ꢀs + 1)
∆f )
-
(2)
(4n2 + 2)
Hence from the slope of a plot of νjCT vs the solvent polarity
parameter ∆f the magnitude of the excited state dipole moment
can be estimated if the cavity radius is known.
Optical Absorption Spectra. The UV absorption spectra
(Table 1, Figure 2) of the triads are dominated by a strong band
at ca. 236 nm and a less intense absorption at ca. 314 nm. These
bands can be attributed to transitions of the acceptor chro-
mophore. In addition, a shoulder occurs at ca. 256 nm due to
the aniline chromophore. The long-wavelength band of the
aniline chromophores is hidden under the much stronger
acceptor transition at 314 nm. Although low-energy charge
transfer states must exist in the dyads and triads,11 transitions
to these states are not observed in the UV absorption spectra,
probably due to a small transition probability.
Characterization of the Emissive CS1 State. Excitation of
the triads at wavelengths between 308 and 320 nm leads to
population of the locally excited state of the acceptor chro-
mophore. The corresponding fluorescence (370 nm),6 however,
is completely absent, and a characteristic charge transfer (CT)
emission is observed instead.22,23 To illustrate this, the fluores-
cence spectra of the bichromophoric reference compound BC0
and the TCn compounds in di-n-propyl ether are depicted in
Figure 3. In Table 2 the maxima and the quantum yields of the
CT fluorescence of the TCn systems and BC0 in a range of
solvents are listed.
Values for the slope and the intercept of the regression lines
in the Lippert-Mataga plots shown in Figure 4 are given in
Table 2. Although the emission maxima of the TCn compounds
are systematically at slightly shorter wavelengths than those of
the bichromophoric reference system BC0, all yield essentially
identical values of the solvatochromic slope 2µ2/hcF3 ≈ 30 000
cm-1. It is evident that the observed emission in all cases
originates from the product of the first step of electron
transfer: the D2D1+A- (CS1) state. An estimate6 of the cavity
radius F ) 5.5 Å leads to an estimate of the excited state dipole
moment of ca. 22 D for this state. It is important to note that
although dielectric continuum theory allows a useful charac-
terization of the solvatochromic shift, a clear deviation from
the ideal behavior is apparent in Figure 4. In going from diethyl
ether to n-pentyl acetate, which have very similar ∆f values, a
red shift of 50 nm occurs. This specific solvent effect also affects
the charge migration process, as discussed below.
Since the photophysical properties of compound TC2 as a
function of solvent and temperature have been extensively
described elsewhere,27-29 we shall now focus especially on the
photophysical behavior of TC1, TC3, and TC4.
(24) Lippert, E. Z. Naturforsch. A 1955, 10, 541.
(25) Mataga, N.; Kaifu, Y.; Koizumi, M. Bull. Chem. Soc. Jpn. 1956,
29, 465.
(26) Beens, H.; Knibbe, H.; Weller, A. J. Chem. Phys. 1967, 47, 1183.
(27) Willemse, R. J.; Verhoeven, J. W.; Brouwer, A. M. J. Phys. Chem.
1995, 99, 5753.
The magnitude of the dipole moment of the emissive state
formed upon photoexcitation can be estimated from an analysis
(22) Hermant, R. M.; Bakker, N. A. C.; Scherer, T.; Krijnen, B.;
Verhoeven, J. W. J. Am. Chem. Soc. 1990, 112, 1214.
(23) Scherer, T.; Hielkema, W.; Krijnen, B.; Hermant, R. M.; Eijckelhoff,
C.; Kerkhof, F.; Ng, A. K. F.; Verleg, R.; van der Tol, E. B.; Brouwer, A.
M.; Verhoeven, J. W. Recl. TraV. Chim. Pays-Bas 1993, 112, 535.
(28) Willemse, R. J. Ph.D. Thesis, University of Amsterdam, 1997.
(29) Willemse, R. J.; Verhoeven, J. W.; Brouwer, A. M. Manuscript in
preparation.

3724 J. Am. Chem. Soc., Vol. 122, No. 15, 2000
Willemse et al.
Table 2. Fluorescence Maxima λmax (nm) and Fluorescence Quantum Yields φf (in Square Brackets) of BC0 and TCn Compounds in Various
Solvents at Room Temperaturea
solvent (∆f)
BC0
TC1
TC2
TC3
TC4
n-hexane (0.092)
419 [0.03]
420 [0.03]
468 [0.19]
475 [0.23]
481 [0.22]
496 [0.21]
540 [0.06]
547 [0.05]
560 [0.01]
554 [0.02]
412 [0.01]
413 [0.02]
458 [0.11]
466 [0.14]
472 [0.16]
483 [0.20]
530 [0.10]
535 [0.06]
558 [0.02]
554 [0.03]
411 [0.01]
413 [0.02]
460 [0.11]
466 [0.12]
472 [0.13]
486 [0.13]
532 [0.06]
539 [0.04]
555 [0.01]
553 [0.01]
412 [0.02]
413 [0.02]
460 [0.03]
469 [0.03]
470 [0.02]
485 [0.01]
531 [0.01]
535 [0.01]
556 [0.01]
553 [<0.01]
410 [0.01]
411 [0.02]
461 [0.03]
466 [0.02]
468 [0.03]
484 [0.01]
530 [0.01]
535 [<0.01]
b
cyclohexane (0.100)
di-n-butyl ether (0.194)
di-n-propyl ether (0.213)
diisopropyl ether (0.237)
diethyl ether (0.251)
n-pentyl acetate (0.259)
n-butyl acetate (0.267)
ethyl acetate (0.292)
tetrahydrofuran (0.308)
b
benzene
dioxane
481 [0.31]
517 [0.18]
470 [0.19]
504 [0.22]
472 [0.21]
503 [0.23]
471 [0.07]
506 [0.04]
470 [0.05]
507 [0.02]
intercept νjCT (0) (cm-1
)
26900 ( 600
29300 ( 2700
27500 ( 700
30500 ( 3000
27500 ( 700
30600 ( 2900
27400 ( 700
30300 ( 2900
27300 ( 800
29200 ( 3900
slope 2µ2/hcF3(cm-1
)
a λexc ) 308 or 314 nm. The last two entries are the values of the intercept νjCT (0) (cm-1) and the slope 2µ2/hcF3 (cm-1) together with the
standard deviations as obtained by linear correlation of ∆f and νjCT according to eq 1. b No CT fluorescence observed.
resolved microwave conductivity (TRMC)31 and time-resolved
optical absorption measurements were carried out, as described
below.
On the basis of the experimentally determined free energy
difference between the CS1 and CS2 states in TC2 as a function
of solvent polarity27 we have argued that in TC3 near-
degeneracy of these states should be expected in solvents having
a dielectric constant of about 2. Therefore, we carefully
reinvestigated the fluorescence decays of TC3 and TC4 in
benzene and dioxane, and indeed a weak long-lived component
could be detected in these decays. This observation is attributed
to feedback from the CS2 state to the CS1 state, which is slightly
higher in energy.
Figure 4. Lippert-Mataga plots (eq 1) for TC1 (0) and TC2 (]).
Data points for TC3 and TC4 coincide with those shown.
Radiative and Nonradiative Decay of the CS1 State in
TC1, TC3, and TC4. In Table 3 the fluorescence decay times
of BC0 and the triads in a series of solvents are listed together
with values of the radiative (kf) and nonradiative (kd) decay rates.
For TC1 the fluorescence quantum yields and decay times of
the CS1 state are similar to those of the dyad BC0. As has been
Analysis of the biexponential decay traces, using the reference
compound TC1 to estimate k1, allows us to determine the rates
noted before for donor-acceptor systems with similar molecular
of the processes depopulating the CS manifold from CS1 (k1)
structures,22 the nonradiative decay rates are relatively high in
and from CS2 (k2) as well as the rates of interconversion k12
and k21. From the equilibrium constant is obtained the free
energy difference (∆G12 ) -kT ln(k12/k21)). The results are
given in Table 4.
Time-Resolved Microwave Conductivity (TRMC) Mea-
surements. In order to confirm the occurrence of the second
nonpolar solvents, then decrease to reach a minimum (in this
case in diethyl ether), and increase again in more polar solvents.
The latter is the usual energy-gap-law behavior. The rapid decay
in nonpolar solvents, in which the energy of the CS state is
high, is related to the existence of an additional decay channel,
as discussed elsewhere.28,30 In the case of compounds TC3 and
step of electron transfer in compound TC3, TRMC measure-
TC4 the quantum yields and decay times of the fluorescence
ments were performed on compounds BC0, TC1, BC3, and
are strongly reduced compared to the values obtained for BC0
TC3. The formation of the highly dipolar CS2 state should be
and TC1. This increased nonradiative decay of the CS1 state
reflected in a large change of the microwave conductivity ∆σ
for TC3 and TC4, which is illustrated in Figure 5, is attributed
to the occurrence of the second step of electron transfer.
of the solution. The TRMC measurements were carried out and
analyzed as fully described elsewhere.31,32 The amplitude of the
Not unexpectedly, we do not observe clear fluorescence from
observed TRMC signal was analyzed to yield the conductivity
the fully charge separated (CS2) state. Due to the small
2
parameter µCS φCS/θ, in which µCS represents the excited state
electronic coupling between D2 and A the radiative transition
probability must be much smaller than for the CS1 state. For
the dyad BC3 a weak CT emission can observed, but in TC3
the fluorescence from the CS1 state is strong enough that CS2
dipole moment, φCS is the quantum yield of formation of the
dipolar species, and θ is the rotational relaxation time. From
the conductivity parameter we can calculate the excited state
dipole moment if φCS and θ are known.
Figure 6 shows the TRMC traces for BC0, TC1, TC3, and
BC3 in benzene and dioxane. The signal intensities in dioxane
emission, if any, is hidden in the long-wavelength tail of the
CS1 band.28 In order to confirm the occurrence of the second
step of electron transfer and to characterize the CS2 state, time-
(31) de Haas, M. P.; Warman, J. M. Chem. Phys. 1982, 73, 35.
(30) Willemse, R. J.; Theodory, D.; Mout, R. D.; Verhoeven, J. W.;
Brouwer, A. M. Manuscript in preparation.
(32) Schuddeboom, W. Ph.D. Thesis, Delft University of Technology,
1994.

Charge Separation in Molecular Triads
J. Am. Chem. Soc., Vol. 122, No. 15, 2000 3725
Table 3. Fluorescence Lifetimes (τf) (ns) and Radiative (kf) and Nonradiative (kd) Decay Rates (107 s-1) of Dyad BC0 and Triads TCn in
Various Solvents at Room Temperaturea
BC0
τf
0.31 6.16 5.0 11 0.19 2.84 6.7 29 0.07 0.91d 7.7 102
0.18 6.71 2.7 12 0.22 6.17 3.6 13 0.04 1.56d 2.6
62
TC1
TC3
TC4
τf
0.05 0.79d 6.3 120
solvent (ꢀs)
φf
kf
kd
φf
τf
kf
kd
φf
τf
kf
kd
φf
kf
kd
benzene (2.28)
dioxane (2.21)
0.02 1.26d 1.6
78
di-n-butyl ether (3.06)
0.19 4.02 4.7 20 0.11 1.92 5.7 46 0.03 0.58
5.2 167
6.5 211
5.3 258
2.6 254
0.03
b
di-n-propyl ether (3.39) 0.23 4.20 5.5 18 0.14 3.13 4.5 27 0.03 0.46
diisopropyl ether (3.88) 0.22 6.38 3.4 12 0.16 3.79 4.2 22 0.02 0.38
diethyl ether (4.24)
0.02 0.40
0.02 0.33
0.01 0.35
5.0 245
6.1 297
2.9 283
0.21 7.40 2.8 11 0.20 5.10 3.9 16 0.01 0.39
n-pentyl acetate (4.75)
n-butyl acetate (5.01)
ethyl acetate (6.00)
0.06 2.98 2.0 32 0.10 3.56 2.8 25 0.01 0.92
0.05 2.16 2.3 44 0.06 2.94 2.0 32 0.01 0.76
0.01 1.21 0.8 82 0.02 1.72 1.2 57 0.01 0.55
1.1 108
1.3 130 <0.01 0.75
1.8 180
0.01 0.96
1.0 103
1.3 132
c
a Excitation at wavelengths between 315 and 320 nm. b Not measured. c No CT fluorescence observed. d Biexponential fluorescence decays are
observed in benzene and dioxane for TC3 and TC4; only the fastest component which has by far the highest amplitude is given.
Figure 5. Nonradiative decay rates of BC0 and TCn compounds in a
series of solvents arranged according to their dielectric constant.
Table 4. Rate Constants (108 s-1) Characterizing the Decay
Processes in TC3 and TC4 in Benzene and Dioxane from
Time-Resolved Fluorescence Measurements
λdet
(nm) k1
∆G12
compd
solvent (ꢀs)
k2 k12 k21 Keq (eV)
TC3 benzene (2.28) 470 3.5 0.7 7.0 0.5 13 -0.07
TC3 dioxane (2.21) 500 1.6 0.4 4.7 0.1 40 -0.09
TC4 benzene (2.28) 470 3.5 0.8 8.8 0.4 23 -0.08
TC4 dioxane (2.21) 500 1.6 0.8 6.2 0.2 34 -0.09
were scaled to correct for the difference in viscosity between
benzene and dioxane, which influences the rotational relaxation
time θ. For all four compounds large dipolar transient signals
were observed upon flash photolysis. The data could be fitted
with a single dipolar intermediate and a small underlying slowly
decaying signal close to the noise level. For TC3 a small
additional short-lifetime component was included in the final
fitting procedure to take into account the presence of the CS1
state. The values of the lifetimes of the dipolar intermediates
and the corresponding values of the conductivity parameter
Figure 6. Time-resolved microwave conductivity transients ∆σ
observed on flash photolysis of benzene and 1,4-dioxane solutions of
the compounds BC0, TC1, TC3, and BC3. The full lines are
convolution fits to the experimental data (dotted lines). The transients
in dioxane are scaled 2.25× to correct for the difference in viscosity
between benzene and dioxane.
2
µCS φCS/θ are compiled in Table 5. For BC0 and TC1 the dipole
the value found for BC0. If the second step of electron transfer
had occurred, a significantly larger value would have been
expected. The fact that the magnitude of the TRMC transient
for TC1 is in fact considerably smaller than that for BC0 as
shown in Figure 6 is a result of the much longer rotation time
for the former molecule.
In the case of compound TC3 the observed signal is seen to
be much larger than for TC1 even though the rotation time for
the former should be even longer. Even based on a maximum
quantum yield of unity the dipole moment of the species
responsible for the TRMC signal of TC3 would be estimated
to be ca. 40 D, i.e., much larger than for BC0 or TC1. It can
therefore be concluded that for TC3 the second electron transfer
moments in Table 5 were calculated using estimates of the
dipolar relaxation times and based on a quantum yield of unity
for the formation of the CS1 state. For TC3 quantum yields of
formation of the CS2 state of 0.68 and 0.75 in benzene and
dioxane, respectively, as derived from the quenching of the
CS1 fluorescence, were used to calculate the dipole moments
in Table 5.
From the TRMC transient observed for BC0 was determined
a dipole moment of 23 D. This value is in good agreement with
the value of ca. 22 D derived from the solvatochromic shift.
The TRMC measurements on TC1 yielded a dipole moment
on photoexcitation of approximately 26 D, which is similar to

3726 J. Am. Chem. Soc., Vol. 122, No. 15, 2000
Willemse et al.
2
Table 5. Lifetime τCS, Conductivity Parameter µCS φCS/θ,
Rotational Relaxation Time θ, Quantum Yield φCS, and the
Calculated Dipole Moment µCS as Obtained from the TRMC
Measurements Using Benzene (BZ) and Dioxane (DX) as Solvents
2
µCS φCS/θ
τ
CS (ns)
(D2/ps)
θa (ps)
φCS
µ
CS (D)
compd BZ DX BZ DX BZ DX
BZ
DX BZ DX
BC0
TC1
TC3 14 23
BC3 19 26
7
3
7
1.91 0.85 277 619
1
1
1
1
23 23
27 26
6.5 1.59 0.66 458 1022
2.44 1.49 545 1216 0.68b 0.75b 44 49
1.14 0.55 542 1209 0.32c 0.28c 44 49
a Reference 32. b Derived from fluorescence results according to φ12
) 1 - (τTC3/τTC1). c Based on the assumption that the dipole moments
of the fully charge separated states of BC3 and TC3 are equal.
Figure 7. UV-vis absorption spectra of A0 (dashed lines) and its
radical anion (full line, produced by electrochemical reduction at room
temperature) in butyronitrile. The spectrum of the radical cation of D3
(dotted line; cosensitization experiment, CH3CN) is included, scaled
step undoubtedly occurs, leading to the highly dipolar CS2 state.
When the values for the quantum yields of the formation of the
CS2 state in benzene and dioxane are used in the analysis of
the TRMC transients, dipole moments of the CS2 state of 44
and 49 D are found, respectively. The increase in dipole moment
compared with compounds BC0 and TC1 corresponds to an
extra charge separation distance of ca. 4 Å, which is close to
the distance between the nitrogen atoms of the two donor groups
as determined in the X-ray crystal structure of TC3. From the
TRMC signals the lifetimes of the CS2 state in benzene and
dioxane are found to be 14 and 23 ns, respectively, considerably
longer than those of the fluorescent intermediate CS1 state.
In the case of compound BC3 the observation of a large
dipolar transient indicates the formation of the long-range
charge-separated state (D2+-bridge-A-), although the observed
signal is seen to be much smaller for BC3 than for TC3.
Assuming now that the conformations and the dipole moments
of the fully charge separated states of TC3 and BC3 are the
same, the quantum yields for charge separation in BC3 can be
estimated from the results on TC3. The values obtained for φCS
in BC3 are listed in Table 5.
to ꢀmax ) 6000 M-1 cm-1 6
.
Figure 8. Transient absorption spectra of TC3 in dioxane. The first
spectrum shown was taken at the maximum of the exciting laser pulse,
the subsequent ones with time increments of 3 ns. Inset: decay of
transient absorption integrated between 490 and 560 nm.
Characterization of the CS2 State by Time-Resolved
Optical Absorption Measurements. Spectra of the Radical
Ions of the Separate Chromophores. It is reasonable to expect
the optical absorption spectra of the ion-pair-like CS states to
resemble a superposition of the spectra of the constituent radical
ions. The model acceptor compound A0 is reversibly reduced
in acetonitrile at E1/2 ) -1.75 V vs SCE (T ) 293 K, V g 100
mV/s). Therefore, the UV-vis absorption spectrum of the
radical anion could be conveniently recorded after electrochemi-
cal reduction of A0 in an OTTLE cell33 (see Figure 7). The
radical anion of A0 displays two absorption bands in the visible
spectrum: a strong absorption band at 446 nm (ꢀ ) 15 500 M-1
cm-1) with a shoulder at 507 nm (ꢀ ) 4800 M-1 cm-1) and a
weak absorption at 735 nm (ꢀ ) 2500 M-1 cm-1). The latter
absorption is in a wavelength region where no spectral overlap
unfortunately, were found to be unstable at room temperature
on the time scale of the spectroelectrochemical measurements
(10-100 s). Their spectra could, however, be readily observed
when the radical cations were generated as transient species by
using a cosensitization technique34,35 with nanosecond time-
resolved detection. The spectrum obtained for D3+• (a broad
band with maxima at 480 and 505 nm, see Figure 7) is in
agreement with the spectra reported for the radical cations of
N-(4-methoxyphenylpiperidine) (λmax 505 nm (ꢀ ) 6000 M-1
cm-1))6 and N,N-dimethyl-4-methoxyaniline (λmax 470 and 490
nm).36 This absorption feature was also found in the UV-vis
spectrum of D3+• electrogenerated in butyronitrile at -80 °C,
but additional bands were also present in this spectrum, possibly
due to secondary products.
Nanosecond Time-Resolved Optical Absorption Measure-
ments. Having determined the radical ion reference spectra, we
could further characterize the CS2 state in TC3 by time-resolved
optical absorption measurements. Figure 8 shows the transient
absorption spectra of TC3 in dioxane as an example. Very
similar spectra were obtained in benzene or diethyl ether. At
short times after the exciting pulse a strong band at ca. 460 nm
3
exists between the absorptions of A-•, D+•, and A* in the
donor-acceptor systems studied. Hence, monitoring the tran-
sient absorption in this wavelength region can in principle give
accurate information about the kinetics of the charge-separated
state, even though the radical anion does not absorb very
strongly.
Oxidation of the model donor compound D3, studied by
means of cyclic voltammetry in acetonitrile, is a chemically
reversible one-electron process at E1/2 ) 0.61 V vs SCE at room
temperature and V g 100 mV/s. The methylenedioxy derivative
D5 is a slightly better electron donor, with E1/2 ) 0.57 V under
the same conditions. The radical cations of D3 and D5,
(34) Majima, T.; Pac, C.; Nakasone, A.; Sakurai, H. J. Am. Chem. Soc.
1981, 103, 4499.
(35) Mattes, S. L.; Farid, S. Acc. Chem. Res. 1982, 15, 80.
(36) Hester, R. E.; Williams, K. P. J. J. Chem. Soc., Perkin Trans. 2
1982, 559.
(33) Krejcik, M.; Danek, M.; Hartl, F. J. Electroanal. Chem. Interfacial
Electrochem. 1991, 317, 179.

Charge Separation in Molecular Triads
J. Am. Chem. Soc., Vol. 122, No. 15, 2000 3727
state at higher solvent polarity. Moreover, the charge recom-
bination in the CS1 state becomes slower upon increasing the
dielectric constant (see Table 3), which further contributes to
the increased efficiency of the second charge transfer step.
Remarkably, in the acetates the formation of the CS2 state is
less efficient. The reason is 2-fold: the nonradiative charge
recombination in the CS1 state is faster in acetates than in alkyl
ethers (see Table 3), while the rate of the second step of electron
transfer is slower in acetates compared to alkyl ethers (see Figure
9). As noted above, the changes in the position of the emission
maxima also indicated that the solvation of the CS1 state in
BC0 and the TCn compounds is relatively better in acetates
than in ethers. The fact that the rate of the conversion of the
CS1 state into the CS2 state is relatively slower in acetates
indicates that this especially favored solvation applies only to
the CS1 state and not (or to a lesser extent) to the CS2 state.
Figure 9. Rates of the second electron transfer step, k12, as calculated
from eq 3 (left axis) and quantum yield of formation of the CS2 state
as calculated from eq 4 (right axis) as a function of the dielectric
constant.
The bulk dielectric properties are clearly not sufficient to fully
describe the solvent effects observed. Details of the microscopic
solvation must be studied to understand this phenomenon.
Preliminary Monte Carlo simulations indicate that it is possible
for the alkyl acetate solvents, which have their dipolar group
on one end of the molecule, to line up with the CS1 dipole in
such a way that both ends of the solvent dipole interact
simultaneously with D+ and A-. The symmetric alkyl ethers,
which have their polar group embedded in the middle of the
molecule, cannot do this. For the CS2 state, the distance between
and a weak band at ca. 730 nm are observed. The former
corresponds to a superposition of bands of the 1-(4-methoxy-
phenyl)piperidine radical cation (D3+•) and the acceptor radical
anion (A0-•); the long-wavelength band is characteristic of the
acceptor radical anion. Unfortunately, the time resolution of our
equipment is insufficient to observe the conversion of CS1 to
CS2 (time scale < 2 ns). This should lead to a clear increase of
the absorption around 500 nm, because saturated amine radical
cations such as D1+• do not absorb strongly in the visible.37
At longer times the CS species have decayed, and two bands
(355, 470 nm) of the local triplet state of the acceptor
chromophore28,30 remain. Thus, the charge-separated states decay
to a local triplet state as well as to the ground state. The latter
pathway is dominant (>85%). The bands attributed to the CS2
state decay with time constants similar to those of the TRMC
signals. A corresponding band of an aniline radical cation
fragment is not observed in the case of TC1 (spectra not shown).
The highly fluorescent CS1 state (with a lifetime of ca. 6 ns)
decays to a local triplet of the acceptor as well as to the ground
state. These results provide further confirmation that the second
electron transfer step does occur in TC3, but not in TC1.
Efficiency of the Second Electron Transfer Step. From the
increased nonradiative decay rate of the CS1 state in TC3 and
TC4 compared to TC1, the rate of the second electron transfer
step, k12, and the quantum yield of formation of the CS2 state,
φCS2, can be calculated using eqs 3 and 4 under the assumption
of irreversibility of the second step of electron transfer:
+
D2 and A- is so large that a single acetate molecule cannot
interact effectively with both sites, and as a result the solvation
is like that of a pair of ions.
Long-Range Electron Transfer vs Stepwise Electron
Transfer. The trichromophoric compounds were designed in
order to accomplish a charge separation over a large distance
via a stepwise electron transfer mechanism. In principle, direct
long-range electron transfer between the terminal chromophores
could occur. In compound BC3, in which the aliphatic amino
nitrogen is replaced by a CH group, the electron transfer must
occur as a single-step process. According to the TRMC results,
the yield of the charge-separated state is ca. 30% in benzene
and dioxane (Table 5). In compound TC3, on the other hand,
two steps of electron transfer occur sequentially upon excitation,
leading to a yield of ca. 70% of the CS2 state in the same
solvents. The reduced quantum yield of emission of the CS1
state of TC3 and TC4 relative to TC1 could have been caused
by direct long-range electron transfer D2-D1-A* f D2+-D1-
A-, but the fact that the lifetime is decreased proportionally
shows that there is indeed an additional decay path of CS1,
which is ca. 100% populated as it is in BC0. From the fact that
for compound BC3 residual local fluorescence is observed, we
conclude that the direct long-range electron transfer between
the terminal chromophores takes place on the same time scale
as the local decay processes of the initially excited chromophore
(ca. 6 × 1010 s-1).28,30 For TC3, on the other hand, no residual
local fluorescence is observed, indicating that the first step of
electron transfer (LE f CS1) is much faster than direct long-
range electron transfer between the terminal chromophores (LE
f CS2). From these results it can be concluded that in nonpolar
solvents the stepwise electron transfer as occurring in TC3 is
ca. three times more efficient than the long-range electron
transfer which takes place in BC3.
1
1
k12
)
-
(3)
(4)
τTCn τTC1
τTCn
φCS2 ) 1 -
τTC1
Figure 9 depicts the rate of the second electron transfer step
and the quantum yield of the CS2 state as a function of the
dielectric constant.
As can be seen, the stepwise charge separation in compounds
TC3 and TC4 is realized with an efficiency of about 70% in
di-n-butyl ether, increasing to more than 90% in more polar
alkyl ethers. This increased efficiency is caused by the fact that
the second electron transfer step becomes faster as a result of
the increased stabilization of the CS2 state relative to the CS1
Driving Force of Charge Migration. The estimation of the
energies of polar excited states in solvents of low polarity is a
long-standing problem in the field. The systems studied in this
work have allowed us to obtain quantitative information on the
interconversion rates and the relative energies of two different
(37) Shida, T. Electronic absorption spectra of radical ions; Elsevier:
Amsterdam, 1988.

3728 J. Am. Chem. Soc., Vol. 122, No. 15, 2000
Willemse et al.
CS states in the same molecule, which was an important goal
of the present study.27 A charge-separated species with a
sufficiently large donor-acceptor distance, such as the CS2 state
in our triads, is quite well described by the commonly used
model of an intramolecular radical ion pair. The observation
that the transient absorption spectrum of TC3 closely resembles
assume that rD ) 4.5 Å and rD ) 3.5 Å, the value for the
2
1
expected slope is 0.72 eV. The aniline donor may indeed be
somewhat larger in size than the saturated amine. We tested
this hypothesis by means of quantum-chemical calculations of
the continuum solvation energy of radical cations of cyclohexyl-
piperidine (D1 model) and N-phenylpiperidine (D2 model). Using
the self-consistent isodensity polarized continuum method38,39
with a dielectric constant of 78.5 we found solvation energies
of 48.7 and 48.3 kcal/mol, respectively, at the HF/6-31G(d)
level, and 47.4 and 47.1 kcal/mol, respectively, using density
functional theory (BLYP/6-31G(d)).40 The corresponding Born
ion radius would be ca. 3.7 Å for both radical cations. Although
there may still be considerable uncertainty as to the reliability
of the computational model used, the result clearly indicates
that the difference in solvation energies of D1 and D2 radical
cations is likely to be small.
+
the superposition of the spectra of the D2 and A- model
systems supports the idea that the radical ion pair picture is
realistic even in solvents of quite low polarity such as benzene
or dioxane. For such an ion pair, in which D+ and A- are
considered to be individually and independently solvated, the
energy ∆GCS2 (relative to the neutral ground state) may be
estimated using the redox potentials, measured in a polar solvent
with dielectric constant ꢀpolar, the Coulomb interaction energy
between unit charges at the distance D2-A, and a correction
for the ion solvation energy based on the Born model:
Another factor to consider is that the assumption of full charge
separation for the CS1 state may not be correct. It is well-known
that mixing occurs of the CS1 charge-separated configuration
with a locally excited state of the acceptor chromophore.41
Although this may reduce the dipole moment and thus the
electrostatic price to be paid in the transition CS1 f CS2, it
should lead to electronic stabilization of CS1 especially in
nonpolar media, which makes the conversion to CS2 more
difficult.
The simple ion-pair model may be appropriate for the CS2
state, but for the CS1 state a description in terms of a dipole
solvation model is probably more suitable. In fact this is what
we did above when we analyzed the solvatochromic shift data
using the Lippert-Mataga method, which is based on the
Kirkwood-Onsager42 model of solvation. The energy of the
CS1 state (∆GCS1) can be estimated as shown in eq 7. The factor
14.4
∆GCS2 ) EoDx - ErAed
-
+
2
ꢀsR
D2A
1
1
1
1
ꢀpolar
14.4
+
-
(eV) (5)
2riAon
s
ion
ꢀ
(
)(
)
2rD
2
The application of this model to CS1 and CS2, neglecting
possible differences in the Born ion radii of the donors, leads
to a simple expression for the free energy change of the charge
migration step as a function of the dielectric constant ꢀs:
∆G12 ) ∆GCS2 - ∆GCS1 ) EoDx - EoDx
-
1
2
14.4
ꢀs
1
1
-
(eV) (6a)
(
RD A RD A
)
2
1
On the basis of the experimentally determined dipole moments
and the molecular structure we can estimate the distances as
RD A ≈ 9.7 Å and RD A ≈ 5.2 Å. This leads to a predicted slope
µ2
(ꢀs - 1)
3 (2ꢀs + 1)
∆GCS1 ) ∆G0CS1
-
(7)
F
2
1
of a plot of ∆G12 versus 1/ꢀs of ca. 1.3 eV (eq 6b):
µ2/F3 is known from the analysis of the solvatochromic shift of
the emission maxima of the triads to be ca. 15 000 cm-1 (1.86
eV) (see Table 2). We assume that the energy of the CS2 state
(∆GCS2) can be appropriately described by the Born approach,
eq 5. Combination of eqs 5 and 7 (using an average Born ion
radius for D2 and A) gives an expression for the difference in
energy (∆G12) between both charge-separated states as a
function of ꢀs:
1.3
ꢀs
∆G12 ) EoDx - EoDx
+
1
(eV)
(6b)
2
As we have previously shown,27 the experimental data for TC2
in a series of ethers give
0.6
ꢀs
∆G12 ) EoDx - EoDx
+
1
(eV)
(6c)
2
14.4 1
1
µ2
(ꢀs - 1)
Equation 6b predicts ∆G12 ≈ +0.25 eV for TC3 (Eox - EDox
≈
∆G12 ) constant +
-
+
(8)
D2
3 (2ꢀs + 1)
av
1
ꢀs
R
(
)
-0.32 V)27 in benzene and dioxane, which is clearly incorrect,
as we now experimentally find ∆G12 ) -0.07 and -0.09 eV,
respectively, in these solvents. This result, on the other hand,
is completely consistent with the empirical relation (6c) found
for TC2, which also correctly predicts that in dibutyl ether and
more polar solvents ∆G12 is sufficiently negative (<-0.12 eV)
that the charge shift process is irreversible.
rion
D2A
F
Two different functions of the dielectric constant ꢀs appear in
eq 8, but they are almost linearly related: over the polarity range
of interest (ꢀ ) 2-10) the linear relationship (ꢀs - 1)/(2ꢀs + 1)
) 0.478 - 0.5678(1/ꢀs) holds within 3%. Thus, we arrive at eq
(38) Foresman, J. B.; Keith, T. A.; Wiberg, K. B.; Snoonia, J.; Frisch,
M. J. J. Phys. Chem. 1996, 100, 16098.
Thus, our experimental results show that formation of the
very polar CS2 state from the less polar CS1 state in a nonpolar
solvent is much more facile than expected from the simple
model which led to eq 6b. The major reason for the discrepancy
is probably that the ion-pair description of the CS1 state is
invalid because the solvation spheres of the donor and acceptor
ions overlap considerably. Thus, counting both the ion solvation
terms and the Coulombic interaction energy to their full extent
leads to an underestimation of the energy of the CS1 state.
A nonnegligible difference in the Born ion radii of D1 and
D2 could also contribute to the discrepancy. If we, for example,
(39) Miertus, S.; Scrocco, E.; Tomasi, J. Chem. Phys. 1981, 55, 117.
(40) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Gill, P. M. W.;
Johnson, B. G.; Robb, M. A.; Cheeseman, J. R.; Keith, T.; Petersson, G.
A.; Montgomery, J. A.; Raghavachari, K.; Al-Laham, M. A.; Zakrzewski,
V. G.; Ortiz, J. V.; Foresman, J. B.; Cioslowski, J.; Stefanov, B. B.;
Nanayakkara, A.; Challacombe, M.; Peng, C. Y.; Ayala, P. Y.; Chen, W.;
Wong, M. W.; Andres, J. L.; Replogle, E. S.; Gomperts, R.; Martin, R. L.;
Fox, D. J.; Binkley, J. S.; Defrees, D. J.; Baker, J.; Stewart, J. J. P.; Head-
Gordon, M.; Gonzalez, C.; Pople, J. A. Gaussian 94; Gaussian, Inc.:
Pittsburgh, PA, 1995.
(41) Bixon, M.; Jortner, J.; Verhoeven, J. W. J. Am. Chem. Soc. 1994,
116, 7349.
(42) Onsager, L. J. Am. Chem. Soc. 1936, 58, 1486.

Charge Separation in Molecular Triads
J. Am. Chem. Soc., Vol. 122, No. 15, 2000 3729
Concluding Remarks
9. Using the numerical values of RD A ) 9.7 Å and µ2/F3 )
2
We have shown that in the triad systems TC3 and TC4 two
∆G12
)
steps of electron transfer can occur upon excitation of the
1
1
µ2
1
ꢀs
+
acceptor chromophore. The first charge-separated D2-D1
-
constant + 14.4
-
- 0.57
(9)
A- (CS1) state can be detected by means of time-resolved
( )
rion
D2A
F3
av
{
}
R
(
)
+
fluorescence spectroscopy, and the fully charge separated D2
-
D1-A- (CS2) state, which is formed from CS1 via a charge
shift, by time-resolved microwave conductivity and transient
optical absorption measurements. Furthermore, we have dem-
onstrated that the rate of the charge shift is markedly dependent
on solvent polarity. In general, the rate of the conversion of the
CS1 state to the CS2 state tends to be faster upon increasing
the dielectric constant of the medium. However, the rate of the
charge migration step is relatively slower in acetates than in
alkyl ethers (specific solvent effect). Moreover, we showed that
the stepwise electron transfer as occurring in TC3 is more
efficient than the direct long-range electron transfer which takes
place in BC3.
1.86 eV and the experimental dependence of ∆G12 on 1/ꢀ (eq
6c), we find that an average ion radius of D2 and A of 4.6 Å
fits the data. This seems a rather large value, compared to the
range of values found in the literature for related organic donors
and acceptors10,13,21,43 and the value estimated from the quantum-
chemical calculation above. Thus, while eq 8 can successfully
account for the experimental observations, it is based on a rather
arbitrary mix of models and requires an assumption of an ionic
radius that is rather large. Obviously, there is a need for a
consistent model that describes both states on an equal footing.
The current use of solvation models in the electron transfer
field is in general very simplistic. We believe that in principle
modern computational chemistry can offer much improvement,44
but we note the following key problems.
(1) Solute electronic structure and electrostatic potential: We
(and others) have used a dipole or two point charges in idealized
cavities. The use of a multisite model representing the charge
distribution, a realistic cavity shape, or separate cavities for the
slow and fast solvent responses45,46 will be an improvement.
Unfortunately, it is currently not straightforward to derive the
charge distribution of charge-separated excited states, including
the effect that the medium will have on the charge distribution.
In principle, the quantum chemical framework is available for
computing the excited state structures and wave functions.47,48
The inclusion of medium effects via dielectric continuum theory
is also possible.49 Practical implementations at ab initio levels,
however, are limited to relatively small systems, while the
accuracy of lower-level semiempirical quantum chemical meth-
ods remains to be established.
The electrostatic continuum models commonly used for
estimating the energies of CS states fail to predict correctly the
relative energies of CS1 and CS2 in low-polarity solvents:
contrary to the simple model prediction (∆G12 ≈ +0.25 eV),
the driving forces found in benzene (∆G12 ≈ -0.07 eV) and
dioxane (∆G12 ≈ -0.09 eV) are sufficient for the charge
migration to proceed.
Experimental Section
Descriptions of the synthetic methods used and details of the
syntheses and characterizations are available in the Supporting Informa-
tion.
Optical Absorption Spectroscopy. Electronic absorption spectra
were recorded on a Hewlett-Packard 8451 A or 8453 diode array
spectrophotometer or on a Cary 3 (Varian). Molar absorption coef-
ficients were determined using concentrations of 10-4-10-5 M.
Commercially available spectrograde solvents were used.
Steady State Fluorescence Measurements. Fluorescence spectra
were recorded on a Spex Fluorolog II emission spectrometer using an
RCA-C31034 GaAs photomultiplier as detector. Fluorescence spectra
were corrected for the wavelength dependent response of the detection
system. Fluorescence quantum yields were determined relative to a
reference solution (9,10-diphenylanthracene in cyclohexane (φ )
0.90),51 quinine bisulfate in 1 N sulfuric acid (φ ) 0.546),51 or
triphenylene in cyclohexane (φ ) 0.08)52) and corrected for the
refractive index of the solvent. The samples were made with an
absorbance between 0.1 and 0.2 (1 cm path length) at the excitation
wavelength and were deoxygenated by purging with argon for 10-15
min. Commercially available spectrograde solvents were used (Merck,
Uvasol), except for di-n-butyl ether (Merck, better than 99%, washed
three times with H2SO4). When the purity of the solvent was found to
be insufficient, the solvent was purified by standard procedures.53 All
alkyl ethers were distilled from CaH2 or LiAlH4 prior to use. All acetates
were washed with a saturated Na2CO3 solution and distilled from CaH2.
Time-Resolved Fluorescence Measurements. Fluorescence decay
curves were measured by means of picosecond time-correlated single
photon counting (SPC). The experimental setup has been fully described
elsewhere.10 A mode-locked argon ion laser (Coherent 486 AS Mode
Locker and Coherent Innova 200 laser) was used to pump synchro-
nously a DCM dye laser (Coherent model 700). The output frequency
was doubled with a BBO crystal resulting in 310-320 nm pulses. A
Hamamatsu microchannel plate photomultiplier (R3809) was used as
detector. The response function (fwhm ca. 20 ps) was obtained by
(2) Solvent properties: It is clear from the present study that
the dielectric continuum description of a solvent can be grossly
inadequate. On the other hand, it has the advantage that it is
easy to apply. The use of simulations based on microscopic
models of solvents has become almost routinely accessible to
many chemists and could in principle become a standard tool
in the electron transfer field.50 For accurate results with solvents
of low polarity, solvent polarizability must be included. At this
moment, there are no established and widely available proce-
dures for doing this.
Accurate description of the solute and the solvent being
already difficult, a truly detailed model treating both at the same
level is not likely to be within reach. On the other hand, it is
desirable that a practical “semiempirical” approach is developed
which does not require heavy computation. Accurate experi-
mental data are required for this purpose, and we hope that our
present work has made a useful contribution to this.
(43) van Dijk, S. I.; Groen, C. P.; Hartl, F.; Brouwer, A. M.; Verhoeven,
J. W. J. Am. Chem. Soc. 1996, 118, 8425.
(44) Cramer, C. J.; Truhlar, D. G. Chem. ReV. 1999, 99, 2161.
(45) Basilevsky, M. V.; Rostov, I. V.; Newton, M. D. Chem. Phys. 1998,
232, 189.
(46) Newton, M. D.; Basilevsky, M. V.; Rostov, I. V. Chem. Phys. 1998,
232, 201.
(47) Mercha´n, M.; Serrano-Andre´s, L.; Gonzalez-Luque, R.; Roos, B.
O.; Rubio, M. TheochemsJ. Mol. Struct. 1999, 463, 201.
(48) Christiansen, O.; Gauss, J.; Stanton, J. F.; Jørgensen, P. J. Chem.
Phys. 1999, 111, 525.
(51) Scaiano, J. C. Handbook of Organic Photochemistry; CRC Press:
Florida, 1989.
(52) Berlman, I. B. Handbook of Fluorescence Spectra of Aromatic
Molecules, 2nd ed.; Academic Press: New York, 1971.
(53) Perrin, D. D.; Armarego, W. L. F.; Perrin, D. R. Purification of
Laboratory Chemicals, 2nd ed.; Pergamon: Oxford, 1980.
(49) Barone, V.; Cossi, M.; Tomasi, J. J. Comput. Chem. 1998, 19, 404.
(50) Ungar, L. W.; Newton, M. D.; Voth, G. A. J. Phys. Chem. B 1999,
103, 7367.

3730 J. Am. Chem. Soc., Vol. 122, No. 15, 2000
Willemse et al.
monitoring at the Raman band of a cell filled with demineralized water.
The cells were painted black with camera varnish on two adjacent sides
to avoid reflection at the quartz/air boundary. The decay traces were
deconvoluted with a computer program54,55 using nonlinear iterative
least-squares fitting.
Cyclic Voltammetry. Cyclic voltammetry was performed using a
BANK-Potentioscan model POS 73 with a Hewlett-Packard 7090A
measurement/plotting system. Scan rates were 100-500 mV/s. Deoxy-
genated acetonitrile solutions were used (dried and stored on 4 Å
molecular sieves) containing ca. 10-3 M of the compound studied and
0.1 M tetraethylammonium tetrafluoroborate (TEAB) as supporting
electrolyte. A platinum disk working electrode and a platinum gauze
auxiliary electrode were used, in combination with a saturated calomel
reference electrode (SCE) connected to the sample solution via a 3 M
KCl salt bridge.
Spectroelectrochemical Measurements. The UV-vis spectroelec-
trochemical measurements were carried out using a Perkin-Elmer
Lambda 5 UV-vis spectrophotometer and an OTTLE cell33 equipped
with a Pt-minigrid working electrode (32 wires/cm) and quartz windows.
The electrode potential was controlled during electrolysis by a PA4
potentiostat (EKOM, Czech Republic). Nitrogen-saturated butyronitrile
solutions were used of the compound under study (10-3 M) with
tetrabutylammonium hexafluorophosphate (TBAH) (3 × 10-1 M) as
supporting electrolyte.
Transient Absorption Spectroscopy. Transient absorption spectra
were measured with an EG&G OMA-III gated intensified diode array
detector. The excitation source was a Lumonics XeCl excimer laser
(308 nm, pulse width ∼ 7 ns). The pulse energy was attenuated to
1-10 mJ hitting the sample cell in a window 10 mm wide and 4 mm
high. The probe light (right angle geometry) was a pulsed 450 W high-
pressure xenon lamp, but the time window for observation was
determined by gating the intensifier of the diode array. Timing of the
pulses was accomplished using a digital delay generator (EG&G 9650).
Spectra were averaged over up to 20 pulses for each delay to improve
the signal-to-noise ratio. Commercially available spectrograde solvents
were used. If the purity of the solvent was found to be insufficient, the
solvent was purified by standard procedures.53 Concentrations were
adjusted to give A1cm ≈ 1.0 at 308 nm, and the samples were
deoxygenated by purging with argon for 15 min.
The cosensitization experiments were carried out using a Q-switched
Nd:YAG laser (355 nm, pulse width ca. 8 ns) as the excitation source.
The radical cations were generated using 1,4-dicyanonaphthalene (DCN)
as the photoexcitable acceptor and biphenyl (E1/2ox ) 1.96 V vs SCE)
as the cosensitizer. The concentrations were 10 mM DCN, 200 mM
biphenyl, and <1 mM donor. The samples were made in acetonitrile
and were not deoxygenated. The optical absorption at 355 nm was 1.0
cm-1, entirely due to DCN.
Time-Resolved Microwave-Conductivity Measurements (TRMC).
The TRMC measurements were carried out and analyzed as fully
described elsewhere.31,32 The excitation source was a Lumonics XeCl
excimer laser (308 nm, pulse width ca. 7 ns, integrated intensity of
approximately 10 mJ/cm2 per pulse). Any transient change occurring
after flash photolysis in the microwave conductivity (dielectric loss)
was monitored as a change in microwave power reflected by the
resonant cavity containing the solution of interest using a Tektronix
7912 transient digitizer. Concentrations were adjusted to give A1cm
1.0 at 308 nm, and the samples were deoxygenated by bubbling with
CO2 for 15 min.
≈
Acknowledgment. The authors are grateful to H. J. van
Ramesdonk and Ing. D. Bebelaar for their invaluable assistance
with the photophysical experiments, to K. Goubitz and J. Fraanje
for the determination of the crystal structure of TC3, and to G.
W. H. Wurpel and E. Rienks for performing the Monte Carlo
simulations on the specific solvation effects. This research was
sponsored by the “Stichting Nationale Computer Faciliteiten”
(National Computing Facilities Foundation, NCF) for the use
of supercomputer facilities and supported (in part) by The
Netherlands Foundation for Chemical Research (SON), with
financial support from the Nederlandse Organisatie voor Weten-
schappelijk Onderzoek (Netherlands Organization for Scientific
Research, NWO).
Supporting Information Available: Descriptions of the
synthetic methods used, details of the syntheses, and charac-
terizations including the crystal structure determination (PDF).
This material is available free of charge via the Internet at
http://pubs.acs.org.
(54) Spectrum Analyser: Kunst, A. G. M., University of Amsterdam,
1992.
(55) Fld Fit: Gedeck, P.; Ja¨ger, W., University of Erlangen-Nu¨rnberg,
Germany, 1994.
JA983716R