Multistate photo-induced relaxation and photoisomerization ability of fumaramide threads: A computational and experimental study

Article abstract of DOI:10.1021/ja802531j

Fumaric and maleic amides are the photoactive units of an important and widely investigated class of photocontrollable rotaxanes as they trigger ring shuttling via a cis-trans photoisomerization. Here, ultrafast decay and photoinduced isomerization in iso

Full text of DOI:10.1021/ja802531j

Published on Web 12/15/2008
Multistate Photo-Induced Relaxation and Photoisomerization
Ability of Fumaramide Threads: A Computational and
Experimental Study
Piero Altoe`,† Natalia Haraszkiewicz,‡,§ Francesco G. Gatti,|, Piet G. Wiering,‡
Ce´line Frochot,‡,∫ Albert M. Brouwer,‡ Grzegorz Balkowski,‡,¶ Daniel Shaw,‡
Sander Woutersen,‡ Wybren Jan Buma,‡ Francesco Zerbetto,† Giorgio Orlandi,†
David A. Leigh,| and Marco Garavelli*,†
Dipartimento di Chimica “G.Ciamician”, UniVersita` di Bologna, Via Selmi 2, 40126 Bologna,
Italy, Van‘t Hoff Institute for Molecular Sciences, UniVersity of Amsterdam, Nieuwe
Achtergracht 166, 1018 WV Amsterdam, The Netherlands, and School of Chemistry, UniVersity
of Edinburgh, The King’s Buildings, West Mains Road, Edinburgh, EH9 3JJ, United Kingdom
Received April 7, 2008; E-mail: marco.garavelli@unibo.it
Abstract: Fumaric and maleic amides are the photoactive units of an important and widely investigated
class of photocontrollable rotaxanes as they trigger ring shuttling via a cis-trans photoisomerization. Here,
ultrafast decay and photoinduced isomerization in isolated fumaramide and solvated nitrogen-substituted
fumaramides (that are employed as threads in those rotaxanes) have been investigated by means of
CASPT2//CASSCF computational and time-resolved spectroscopic techniques, respectively. A complex
multistate network of competitive deactivation channels, involving both internal conversion and intersystem
crossing (ISC) processes, has been detected and characterized that accounts for the picosecond decay
and photochemical/photophysical properties observed in the singlet as well as triplet (photosensitized)
photochemistry of fumaramides threads. Interestingly, singlet photochemistry appears to follow a non-
Kasha rule model, where nonequilibrium dynamical factors control the outcome of the photochemical
process: accessible high energy portions of extended crossing seams turn out to drive the deactivation
process and ground-state recovery. Concurrently, extended singlet/triplet degenerate regions of twisted
molecular structures with significant spin-orbit-coupling values account for ultrafast (picosecond time scale)
ISC processes that lead to higher photoisomerization efficiencies. This model discloses the principles behind
the intrinsic photochemical reactivity of fumaramide and its control.
light absorption,6,9-13 which are an inherent characteristic
of many catenanes and rotaxanes.14,15
1. Introduction
Control of large amplitude submolecular movements is a
prerequisite for the development of artificial devices that
function through motion at the molecular level. Large
amplitude internal translation can be induced by different
types of stimuli: acid base reaction,1-3 redox reaction4-8 and
Fumaramide motifs template the assembly of benzylic amide
macrocycles (called rings) around them (called threads) to form
(5) Altieri, A.; Gatti, F. G.; Kay, E. R.; Leigh, D. A.; Martel, D.; Paolucci,
F.; Slawin, A. M. Z.; Wong, J. K. Y. J. Am. Chem. Soc. 2003, 125,
8644.
† Universita` di Bologna.
(6) Brouwer, A. M.; Frochot, C.; Gatti, F.; Leigh, D. A.; Mottier, L.;
Paolucci, F.; Roffia, S.; Wurpel, G. W. H. Science 2001, 291, 2124.
(7) Fioravanti, G.; Haraszkiewicz, N.; Kay, E. R.; Mendoza, S. M.; Bruno,
C.; Marcaccio, M.; Wiering, P. G.; Paolucci, F.; Rudolf, P.; Brouwer,
A. M.; Leigh, D. A. J. Am. Chem. Soc. 2008, 130, 2593.
(8) Jagesar, D. C.; Hartl, F.; Buma, W. J.; Brouwer, A. M. Chem.-Eur.
J. 2008, 14, 1935.
‡ University of Amsterdam.
§ Present address: Biotechnology Department, Synthon BV Microweg
22, 6503 GN Nijmegen, The Netherlands.
| University of Edinburgh.
Present address: Dipartimento di Chimica, Materiali e Ingegneria
Chimica ”Giulio Natta”, Politecnico di Milano, Via Mancinelli 7, 20131
Milano, Italy.
(9) Murakami, H.; Kawabuchi, A.; Kotoo, K.; Kunitake, M.; Nakashima,
N. J. Am. Chem. Soc. 1997, 119, 7605.
∫ Present address: DCPR, Nancy-University, CNRS, 1 rue Grandville
BP 20451 F-54001, Nancy, France.
(10) Stanier, C. A.; Alderman, S. J.; Claridge, T. D. W.; Anderson, H. L.
Angew. Chem., Int. Ed. 2002, 41, 1769.
¶ Present address: Institute of Inorganic Technology and Mineral Fertil-
izers, Wrocław University of Technology, ul. Smoluchowskiego 25, 50-
372 Wrocław, Poland.
(11) Altieri, A.; Bottari, G.; Dehez, F.; Leigh, D. A.; Wong, J. K. Y.;
Zerbetto, F. Angew. Chem., Int. Ed. 2003, 42, 2296.
(12) Balzani, V.; Clemente-Leon, M.; Credi, A.; Ferrer, B.; Venturi, M.;
Flood, A. H.; Stoddart, J. F. Proc. Natl. Acad. Sci. U.S.A. 2006, 103,
1178.
(1) Bissell, R. A.; Cordova, E.; Kaifer, A. E.; Stoddart, J. F. Nature 1994,
369, 133.
(2) Badjic, J. D.; Balzani, V.; Credi, A.; Silvi, S.; Stoddart, J. F. Science
2004, 303, 1845.
´
(13) Leigh, D. A.; Morales, M. A. F.; Pe´rez, E. M.; Wong, J. K. Y.; Slawin,
(3) Keaveney, C. M.; Leigh, D. A. Angew. Chem., Int. Ed. 2004, 43, 1222.
(4) Tseng, H.-R.; Vignon, S. A.; Stoddart, J. F. Angew. Chem., Int. Ed.
2003, 42, 1491.
A. M. Z.; Carmichael, A. J.; Haddleton, D. M.; Brouwer, A. M.; Buma,
W. J.; Wurpel, G. W. H.; Leo´n, S.; Zerbetto, F. Angew. Chem., Int.
Ed. 2005, 44, 3062.
9
104 J. AM. CHEM. SOC. 2009, 131, 104–117
10.1021/ja802531j CCC: $40.75
2009 American Chemical Society

Fumaramide Threads
A R T I C L E S
Scheme 1a
To access the details of this photoinduced process, a
computational study of the photochemical isomerization and
multistate relaxation channels of the bare photoactive unit (i.e.,
the unsubstituted fumaramide E-1) has been performed in Vacuo,
jointly with photochemical and spectroscopic measurements
(including time-resolved subpicosecond time scale experiments)
on N,N′- and N,N,N′N′-substituted fumaramides/maleamides
(2-6) in dichloromethane/methanol or acetonitrile solution.
Specifically, a high level CASPT2//CASSCF state-of-the-art
quantum mechanical description17 of the chemical system has
been used to map the relevant photochemical channels and
deactivation routes of isolated E-1, which has been used as a
model system for the photochemical reactivity of fumaramides
threads (E-4, E-5 and E-6). Then, recorded time-resolved
observations have been compared/combined to these computa-
tional results to provide a full rationale of fumaramide photo-
chemical activity. Eventually, the purpose is to unveil the factors
that control the photoinduced dynamics of this chromophore,
so that our ability to understand, predict and control its behavior
may be increased. It is also apparent that this information may
turn out to be useful for a more efficient use of these systems
in fumaramide-based rotaxanes (7,8).18
This work adds to the in depth studied field of alkene
photoisomerizations: despite the existing extended (both ex-
perimental and theoretical) literature (spanning from neutral to
polar conjugated organic molecules such as carotenes, retinals,
stilbenes/azobenzenes/diarylethenes and their analogs, coumaric
acid derivatives, etc) that has been widely reviewed both in the
past and more recent times (see refs 19-23), to the best of our
knowledge, the photochemistry of fumaric amides has never
been investigated computationally before. Fumaric esters were
studied in the seminal paper by Hammond,24 while acroleine
photochemistry (a rough precursor of the investigated systems
that, although lacking the amide bond, has similar nfπ*/πfπ*
excited states) has also been documented (see ref 25). The
fumaramide photochemistry will be shown to be unique and
drastically different (and far more complex) than the one
reported for apparently similar conjugated molecules, such as
hexatrienes, acroleines, etc. In this respect, the work is theoreti-
cally valuable by itself as it delivers a novel information on a
system that is a prototypical photoactive molecule with promis-
ing (nanotechnological) applications.
a Investigated fumaramide/maleamide compounds and their photoinduced
isomerization (top). The structural formula of a very simple fumaramide
based rotaxane (middle) and the atoms numbering used in the text are also
displayed (bottom).
rotaxanes in high yields16 (see Scheme 1). The fumaramide unit
(that is also referred to as a station of the thread) also provides
an opportunity to enforce a geometrical change in the thread
after rotaxane formation. In fact, isomerization of the olefin from
E to Z, triggered photochemically, disrupts the near-ideal
hydrogen bonding structure between macrocycle and thread and
changes the internal dynamics governed by those interactions
(such as the shuttling of the ring along the thread). The
photoisomerization interconverts fumaramide (trans) and malea-
mide (cis) isomers of the olefinic unit. The trans olefin holds
its two strongly hydrogen bond accepting amide carbonyl groups
in a preorganized close-to-ideal spatial arrangement for interac-
tion with the amide groups of the macrocycle. Photoisomer-
ization reduces the number of possible intercomponent hydrogen
bonds at the olefin station from four to two and the macrocycle
can change position to another station, if present, that is
essentially indefinitely stable until a further stimulus is applied
to reisomerize the maleamide unit back to fumaramide. The
switching process is reversible: subjecting the maleamide
rotaxanes to heat, or a suitable catalyst, or a second photon of
different wavelength, results in reisomerization to the more
thermally stable trans-olefin isomers, with accompanying rein-
statement of the strong hydrogen-bonding network. As long as
photons of a suitable wavelength are supplied, the process will
continue indefinitely, apart from losses due to side reactions
and fatigue effects.
2. Computational and Experimental Details
2.1. Computational Section. Fumaramide E-1 (a small molecule
of 14 atoms that can be studied at a high ab initio level) has been
(17) Andersson, K.; Malmqvist, P. A.; Roos, B. O.; Sadlej, A. J.; Wolinski,
K. J. Phys. Chem. 1990, 94, 5483.
(18) Gatti, F. G.; Leon, S.; Wong, J. K. Y.; Bottari, G.; Altieri, A.; Morales,
M. A. F.; Teat, S. J.; Frochot, C.; Leigh, D. A.; Brouwer, A. M.;
Zerbetto, F. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 10.
(19) Altoe′, P.; Bernardi, F.; Conti, I.; Garavelli, M.; Negri, F.; Orlandi,
G. Theor. Chem. Acc. 2007, 17, 1041.
(20) Gilbert, A.; Baggott, J. Essentials of Molecular Photochemistry;
Blackwell Science: Oxford, 1991.
(21) Horspool, W. M.; Song, P.-S. CRC Handbook of Organic Photochem-
istry and Photobiology; CRC Press: Boca Raton, FL, 1995.
(22) Klessinger, M.; Michl, J. Excited States and Photochemistry of Organic
Molecules; VCH Publishers: New York, 1994.
(23) Michl, J.; Bonacic-Koutecky, V. Electronic Aspects of Organic
Photochemistry; Wiley & Sons, Inc.: New York, 1990.
(24) Hammond, G. S.; Saltiel, J.; Lamola, A. A.; Turro, N. J.; Bradshaw,
J. S.; Cowan, D. O.; Counsell, R. C.; Vogt, V.; Dalton, C. J. Am.
Chem. Soc. 1964, 86, 3197.
(14) Balzani, V.; Credi, A.; Raymo, F. M.; Stoddart, J. F. Angew. Chem.,
Int. Ed. 2000, 39, 3348.
(15) Barbara, P. F. Acc. Chem. Res. 2001, 34, 409.
(16) Gatti, F. G.; Leigh, D. A.; Nepogodiev, S. A.; Slawin, A. M. Z.; Teat,
S. J.; Wong, J. K. Y. J. Am. Chem. Soc. 2001, 123, 5983.
(25) Reguero, M.; Olivucci, M.; Bernardi, F.; Robb, M. A. J. Am. Chem.
Soc. 1994, 116, 2103.
9
J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009 105

A R T I C L E S
Altoe` et al.
Table 1. Molar Absorption Coefficients at the Absorption Maxima
and at 254 nm (Irradiation Wavelength) of the Compounds Studied
Experimentally
Compound
λmax (nm)
ꢀ (λmax) (M-1cm-1
)
ꢀ (λ254nm) (M-1cm-1
)
E-3
E-4
265
250 (shoulder)
-
-
13500a
13200a
14200b
7000a
7600b
8600b
7500b
E-5
264
7700a
Z-4
Z-5
250 (shoulder)
250 (shoulder)
8700a
7000a
a acetonitrile. b dichloromethane.
(i.e., CASSCF geometry optimizations plus single point CASPT2
energy corrections on top of these structures) is referred to as a
CASPT2//CASSCF approach. All these computations have been
performed in Vacuo.
2.2. Experimental Section. Synthesis. The synthesis of threads
were performed as described previously.18 Suitable threads were
prepared in a single step from fumaryl chloride (E-4, E-5) or maleic
acid (Z-4) and a bulky primary (4) or secondary amine (5).
N-methylated thread Z-5 was conveniently obtained from E-5 upon
sensitized isomerization by 350 nm irradiation in the presence of
benzophenone as a sensitizer, followed by separation from the
benzophenone, and small amounts of photobyproducts, on a silicagel
column using CHCl3/EtOAc as the eluent. E-3 was prepared as
32
described by Montaudo et al.
Spectroscopy. Steady state absorption spectra were recorded on
a Cary 3 (Varian) spectrometer. The details of the setup employed
for the femtosecond transient absorption experiments have been
reported before.33 Briefly, a Hurricane (Spectra Physics) laser/
amplifier system generated a pulse train (130 fs fwhm, 800 nm
center wavelength, 1 kHz repetition rate) that was separated into
two parts. One part was employed to pump an Optical Parametric
Amplifier, which generated pump pulses at 270 nm, the other part
was focused on a calcium fluoride crystal to generate a white light
continuum from 350 to 800 nm used for the probe pulse. The
polarizations of the two beams were set at the magic angle. The
probe beam was coupled into a 400 µm optical fiber after passing
the sample and was detected with a CCD spectrometer (Ocean
Optics, PC 2000). A chopper (Rofin Ltd., 83 Hz) placed in the
excitation beam provided I and I0 depending on the status of the
chopper (open or closed). The total instrumental response was about
200 fs (fwhm). The excitation power was kept as low as ∼5 µJ per
pulse using a pump spot diameter of about 1 mm. Apart from using
a relatively low excitation intensity, thermal effects and photodeg-
radation were also taken care of by stirring the solution. Photo-
degradation was not substantial as shown by comparing absorption
spectra taken before and after the transient absorption experiment.
All experiments were performed at ambient temperature.
Figure 1. Fumaramide active space MOs.
selected as a model of the fumaramide thread E-4 (see Scheme 1).
A complete active space-self-consistent field (CASSCF)/6-31G*
level of theory has been adopted for all geometry optimizations
and minimum energy paths (MEP) computations, using the tools
available in the Gaussian98 suite of programs.26 For this purpose,
a full active space of 14 electrons in 10 orbitals (i.e., all the p
orbitals/electrons of the molecule plus the two p lone pairs on the
oxygen atoms, see Figure 1) has been used. A state average
procedure has also been employed, equally weighting the investi-
gated root (for which the gradient is computed) and all the lower
lying states.
MEPs have been computed by following the prescriptions
described in refs 27 and 28. Briefly, to locate the relaxation channels
(i.e., the MEPs) departing from the lower tip of a conical intersection
point or the FC region, an initial direction of relaxation (IRD) (as
close as possible to the conical intersection/FC point) has been
located, and then a standard fully unconstrained MEP computation
following that IRD has been computed in mass weighted coordinates.
To improve the energy profiles along the computed MEPs, single
point multiconfigurational second-order perturbation theory com-
putations have been carried out to account for correlation effects,
using the complete active space second-order perturbation theory
(CASPT2) method17 implemented in MOLCAS 5.29 To minimize
the influence of weakly interacting intruder states the so-called
imaginary level-shift technique was employed30 (an imaginary level-
shift value of 0.1 has been used throughout all the calculations,
which was selected by a calibration procedure and found to produce
stable results). A 5 roots state average procedure and an ano-s
(10s6p3d/3s2p1d)31 basis set have been employed. This strategy
Details of the photochemical experiments are given in the
Supporting Information.
3. Results and Discussion
The interplay between the potential energy surfaces (PESs)
of the photochemically relevant electronic states is rather
complicated. Two different and alternative interpretative models
(thereafter called as the static and dynamics view) will be
presented, discussed and compared with the experimental data
documented here. All the (experimental and computational)
results have been collected in Tables 1-4 and are presented in
Figures 2XI-13, including a schematic (cartoon like) representa-
(26) Frisch, M. J.; et al. Gaussian 98, revision A.6; Gaussian Inc.: Pittsburgh
PA, 1998.
(27) Bearpark, M. J.; Robb, M. A.; Schlegel, H. B. Chem. Phys. Lett. 1994,
223, 269.
(28) Celani, P.; Robb, M. A.; Garavelli, M.; Bernardi, F.; Olivucci, M.
Chem. Phys. Lett. 1995, 243, 1.
(29) Andersson, K.; et al. MOLCAS, Version 5.4; Lund University: Sweden,
2002.
(32) Montaudo, G.; Maravigna, P.; Caccamese, S.; Librando, V. J. Org.
Chem. 1974, 39, 2806.
(33) Balkowski, G.; Szemik-Hojniak, A.; vanStokkum, I. H. M.; Zhang,
H.; Buma, W. J. J. Phys. Chem. A 2005, 109, 3535.
(30) Forsberg, N.; Malmqvist, P.-A. Chem. Phys. Lett. 1997, 274, 196.
(31) Widmark, P.-O.; Persson, B. J.; Roos, B. O. Theor. Chem. Acc. 1991,
79, 419.
9
106 J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009

Fumaramide Threads
A R T I C L E S
characterized by a single nfπ* electronic transition between
the orbitals 3f8 and 4f8 (see Figure 1), respectively; while
the next two higher energy ones (S3(Bu) and S4(Ag)) are
characterized by single πfπ* excitations. In particular, S3(Bu)
is the bright state (oscillator strength ∼0.1), which absorbs at
239 nm (see Table 2) and is mainly described by a single
excitation between the orbitals 7 and 8 (that are located
respectively on the CdC central bond (C1-C2) and the adjacent
(C1-C3, C2-C4) bonds, see Figure 1. Note that this value is
close to the experimental maxima recorded for the UV/vis
spectrum of the threads E-4 and E-5 in CH2Cl2 (250 and 264
nm respectively, see Table 1 and Figure 2). The red shift
recorded for the thread absorption maximum (as compared to
the computed value of the parent unsubstituted fumaramide
compound E-1) can be attributed to the hyperconjugative effect
of the alkyl substituents. In fact, similar computations on E-2
(251 nm, see Table 2) and E-3 (277 nm, see Table 2), that can
be regarded as model compounds for E-4 and E-5, do nicely
reproduce the experimental values for E-4 and E-5, respectively.
Even considering the broad maximum region of the recorded
solution spectra (see Figure 2), the agreement between computa-
tions (that have been performed in isolated conditions) and
experiments is remarkably good. As our computations in vacuo
reproduce the spectroscopic data recorded in solution, we think
that solvent effects may be considered negligible in this specific
case and that the isolated system models the photochemical
behavior in solution.
Figure 2. UV absorption spectra in dichloromethane.
3.2. E/Z Photochemical Isomerization. Direct Irradiation
Experiments. E and Z-isomers are separable and stable in their
ground states. Thus, it is straightforward to determine by
conventional analytical methods the ratio of the two isomers at
the photostationary state as well as the isomerization quantum
yields. Since the absorption spectra of E and Z-isomer pairs of
threads are quite similar, they are not suitable for accurately
determining the concentration changes following irradiation. For
the same reason they cannot be used to calculate the ratio of
isomers at the photostationary state, which is, in contrast,
possible in the case of many stilbene analogs or azobenzene
analogs described in the literature.9,35-37 Instead, 1H NMR (the
chemical shifts of protons at the double bonds or protons from
NH groups are different in case of E and Z-isomers) or the more
sensitive HPLC method were applied to determine the isomer
ratios.
Photostationary States. Photoisomerization upon 254 or 300
nm irradiation is reversible and after a period of exposure,
typically 30 min. to 1 h, depending on conditions, a photosta-
tionary E-Z equilibrium is attained, where the ratio of both
isomers does not change after further prolongation of the
irradiation. The results of these irradiation experiments are
summarized in Table 3. Since UV absorbance of Z-isomers is
lower than that of E-isomers, it is not surprising that the
photostationary states consist of only 38% and 16% of the initial
E-isomers for the pairs of E-4/Z-4, and E-5/Z-5, respectively.
Quite similar values were obtained by analyzing the mixture of
isomers with 1H NMR and HPLC. The ratio of E and Z-isomers
at the photostationary state depends on the excitation wave-
lengths but the differences are not large.
Figure 3. Transient absorption spectra of E-4 upon excitation using 100
fs pulses of 275 nm and 1 µJ pulse energy. The spectral data depicted
represent the excited-state history at times of 0.7 ps, (red) 1 ps, (violet) 1.3
ps, (green) 2.0 ps (blue) and 20 ps (black) after photoexcitation.
tion of the potential energy surfaces involved in the two (static
and dynamics) models (Figures 7 and 10, respectively).
3.1. Absorption Spectra: The Singlet Manifold. Experi-
ments. The spectroscopic data for threads 4 and 5 are gathered
in Table 1. The spectra (Figure 2) are characterized by a strong
absorption band at ca. 250-260 nm, which appears as a shoulder
on a stronger band at shorter wavelength. Based on a comparison
with other substituted alkenes we can attribute the absorption
bands to πfπ * transitions.34 Such transitions are known to
possess high intensity. The molar absorption coefficient at the
absorption maximum for the simple ethene chromophore is in
the range of 10000 M-1cm-1, but it depends strongly on
substituents. The molar absorption coefficient is higher for E-4,
ca. 14000 M-1cm-1, but it is lower for E-5 (ca. 7000 M-1cm-1).
The absorption maximum for E-5, on the other hand, occurs at
a longer wavelength. The substituents on the nitrogen atoms
have a substantial influence on the transitions of the chro-
mophore because they affect the relative energies of the p-basis
orbitals contributing to the π -MOs. The molar absorption
coefficients for Z-isomers are generally lower than those of the
corresponding E-isomers because the chromophore is less
extended. For the case of Z-5, the chromophore is nonplanar,
which further contributes to a higher vertical excitation energy
and smaller absorption coefficient. None of the compounds
investigated shows detectable fluorescence. Therefore it is not
possible to properly estimate the 0-0 excitation energy.
Photoisomerization Quantum Yields. Quantum yields of
photoisomerizations of 4 and 5 in acetonitrile were measured
Calculations. The first two excited singlet states (S1(Bg) and
S2(Au)) for fumaramide E-1 are almost degenerate and are
(35) Bortolus, P.; Monti, S. J. Phys. Chem. 1979, 83, 648.
(36) Birnbaum, P. P.; Style, D. W. G. Trans. Faraday Soc. 1954, 50, 1192.
(37) Siampiringue, N.; Guyot, G.; Monti, S.; Bortolus, P. J. Photochem.
1987, 37, 185.
(34) Hesse, M.; Meier, H.; Zeeh, B. Spectroscopic Methods in Organic
Chemistry, 2nd ed.; Georg Thieme Verlag: Stuttgart, Germany, 2008.
9
J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009 107

A R T I C L E S
Altoe` et al.
Figure 4. Left: The decay profile for E-4 monitored at 375 nm over a 60 ps time scale, with in the insert the decay for short pump-probe delays.
The solvent response (dashed/dotted line in the insert) was experimentally determined from the excitation of a sample containing exclusively the
solvent. The blue line is a biexponential fit with lifetimes of 1.7 ( 0.2 and 19 ( 1 ps obtained after deconvolution of the decay with the solvent
response. Right: The decay profile for E-4 monitored at 650 nm with in the insert the decay for short pump-probe delays and the solvent response.
The blue line derives from a monoexponential fit with a lifetime of 1.7 ps.
Figure 5. Deactivation paths (MEPs) for fumaramide E-1 from the S3(ππ*) Franck-Condon point to the minimum on S1(nπ*). All energy values are
calculated at the CASPT2/ano-s level (filled symbols) or at the CASSCF/ano-s level (empty symbols), which have been scaled to match the energies computed
at the CASPT2/ano-s level.
by monitoring the conversion starting from each of the pure
isomers using HPLC. The total conversion was kept small, so
that corrections for the reverse reaction (excitation of the
photoproduct) could be ignored. The concentrations of initial
E and Z-isomers were ca. 1 mM, the photon flux was determined
using the photocyclization of 1,3-cycloheptadiene (CHD) as
actinometer.38 For details see Supporting Information. Results
are reported in Table 3. The isomerization quantum yields and
the photostationary state composition are related according to
eq 1:
The efficiencies of the isomerization process of E-4 f Z-4
and Z-4 f E-4 are almost the same (ca. 4.4%). The photoi-
somerization process is somewhat less efficient for methylated
threads E-5 and Z-5 than for E-4 and Z-4. The quantum yield
for the Z-5 f E-5 process is even smaller than that for E-5 f
Z-5 (0.28% and 1.5%, respectively).
In general, the photoisomerization quantum yields are rather
low. This means that the relaxation to the ground-state proceeds
via nonproductive funnel, or that the relaxation on the ground-
state surface is biased in favor of the reactant rather than the
product.39 The recorded quantum yields for EfZ and ZfE
photoisomerization of the thread are collected in Table 3.
Time-Resolved Experiments. Figure 3 displays chirp-uncor-
rected transient absorption spectra of E-4 dissolved in dichlo-
romethane for various pump-probe delay times. For short delay
times, the absorption spectrum is dominated by two absorption
bands. The stronger one of these bands is the one in the
350-450 nm region and appears to have a maximum below
φEfZ εZ [Z]PPS
)
×
(1)
φZfE εE [E]PPS
in which [Z]PPS/[E]PPS is the ratio of Z-isomer to E-isomer at
the photostationary state and ꢀZ and ꢀE are the molar absorption
coefficients. This allows a consistency check, results of which
are included in Table 3.
(38) Numao, N.; Hamada, T.; Yonemitsu, O. Tetrahedron Lett. 1977, 18,
1661.
(39) Merchan, M.; Serrano-Andres, L. J. Am. Chem. Soc. 2003, 125, 8108.
9
108 J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009

Fumaramide Threads
A R T I C L E S
Figure 6. S1 minimum energy path between the S1 minimum (Min S1) and the 90° twisted conical intersection (CI S1/S0 90°) computed for fumaramide
E-1. Triangles and diamonds display the S1 and S0 energy profiles, respectively. All values are calculated at the CASSCF/ano-s level (empty symbols),
which have been scaled to match the energies computed at the CASPT2/ano-s level (filled symbols). At CASPT2 level the conical intersection is anticipated,
i.e. it is closer to the TS.
Figure 7. “Static interpretation” of fumaramide photochemistry.
350 nm, that is, outside the region covered by our white-light
window. The other band in the 600-800 nm region is
considerably weaker, but still distinctly present. For longer delay
times, the latter band disappears and only a band with a
maximum at ∼375 nm remains in the 350-500 nm region.
Figure 4 shows decay traces taken at 375 and 650 nm. After
deconvolution with the solvent response, the trace at 375 nm
can be fit well with a biexponential decay with time constants
of 1.7 ( 0.2 ps and 19 ( 1 ps. At 650 nm the solvent response
precludes an accurate analysis of the trace below 1 ps, but for
longer delay times the decay can be satisfactorily fitted with a
monoexponential decay with a time constant of 1.7 ( 0.2 ps.
Similar analyses have been performed for a range of wave-
lengths, and lead to the conclusion that the Decay Associated
Difference Spectrum (DADS) of the fast component peaks at
the aforementioned 350-450 and 600-800 nm bands, while
that of the slow component is maximum around ∼375 nm and
diminishes continuously for longer probe wavelengths.
The transient absorption spectra of the other systems,
unfortunately, were of lower quality. However, analogously to
the results on E-4, a rapid decay is observed for all molecules
thatswhen fitted to a monoexponential decaysleads to decay
times of 1.1 and 1.3 ps for E-5 and E-6, respectively. It would
thus appear that the primary excited-state dynamics are es-
sentially the same in all cases.
The calculations and experimental absorption spectra leave
no doubt that the initial excitation populates the S3(ππ*) state.
Unfortunately, the solvent response at in particular the longer
probe wavelengths do not allow us to observe processes faster
than 300 fs, but we can safely assume that internal conversion
9
J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009 109

A R T I C L E S
Altoe` et al.
Figure 8. S1/S0 crossing seam computed for fumaramide E-1. Each point is obtained by constrained (at a fixed central bond dihedral angle) conical intersection
optimization at the CASSCF level. Energies are re-evaluated at the CASPT2/6-31G* level: filled triangles and diamonds display the S0 and S1 CASPT2
corrected energy profiles, respectively. The blue line indicates the state diabatically connected to FC S0, while the red line indicates the state diabatically
connected to FC S1. It is apparent here that the two states have already swapped at the CASPT2 level: as reported in Figure 6, the CASPT2 crossing region
is displaced (i.e., anticipated) respect to the CASSCF one (dashed lines and empty symbols).
Figure 9. High energy S3/S2, S2/S1 and S1/S0 conical intersections (CI S3/S2, High CI S2/S1 and CI S1/S0 30°, respectively) and their distance in mass
weighted atomic units (a.u.). It is apparent that these points are isoenergetics and that CI S3/S2 and High CI S2/S1 are quite close each other.
of S3 occurs faster than the first observable decay on the order
of 0.5-1.7 ps. We therefore associate this decay component
with the decay of another electronically excited state, most likely
the S1 state. For E-4 a second decay component is observed.
The absence of the 600-800 nm band at longer delay times
indicates a change in the electronic character of the absorbing
state. The steady-state absorption spectrum of E-4 shows a
lowest absorption band at 250 nm, and negligible absorption
for wavelengths larger than 300 nm. This observation makes it
unlikely that the ∼375 nm band observed for pump-probe
delays larger than 10 ps is associated with absorption from a
vibrationally hot ground state, and we therefore tentatively
assign it to excited-state absorption from the T1 state. Further
support for such an assignment might be obtained by recording
the transient absorption spectrum of T1 generated, for example,
by triplet sensitizers, but such experiments have not been
9
110 J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009

Fumaramide Threads
A R T I C L E S
Figure 10. “Dynamic interpretation” of fumaramide photochemistry.
and the amide bonds. This issue and the implications of isomers
equilibria on the recorded experimental data are discussed in
depth in section S1 of the Supporting Information.
Computational Analysis of Singlet Photochemistry: Static
vs Dynamic Model. The results of CASPT2//CASSCF/
6-31G* MEP mapping from the FC point of the first bright
(S3) ππ* singlet state of E-1 are summarized in Figures 5–7,
while Tables 2 and 4 collect the relative energies and the
structural data of all the relevant points discussed in the text,
respectively. A first model for the photochemistry of the
fumaramide unit has been formulated that is based only on
the minimum energy points found (along the MEPs) on the
potential energy surfaces of the photochemically relevant
states. This mechanistic picture (which is schematically
displayed and summarized in Figure 7) will be referred to in
this paper as the static model.
In contrast to that picture, also the high energy regions of
the photochemically relevant crossing seams have been
investigated (see Figures 8-9). These regions are close to
high energy portions of the computed MEPs and may be
accessed dynamically if sufficient vibrational energy is
available. They turn out to be important features for a realistic
description of fumaramide photochemistry and a full under-
standing of the experimental results. The present analysis
contributes to draw an alternative mechanistic picture that
will be referred to in this paper as the dynamic model (and
is schematically depicted in Figure 10).
Figure 11. Cartoon representation for the static (short white arrows) and
dynamic (long yellow arrows) models of fumaramide deactivation. The S1/
S0 intersection space is represented by dashed lines.
(A) The Static Model. As pointed out in the subsection 3.1
above, direct photoexcitation promotes fumaramide to the
spectroscopically allowed πfπ* S3(Bu) state. Relaxation
from the Franck-Condon (FC) region of this state drives
the system to a planar minimum (Min S3) where the C2h
symmetry is preserved. A crossing point with the dark πfπ*
S4(Ag) state (i.e., a CI S4/S3 conical intersection) exists very
close to the minimum region. The energy of this CI (2.7 kcal/
mol above the minimum) has to be considered an upper limit
because this crossing has been found through a linear
Figure 12. Conversion to the Z-isomers of 4 and 5 in acetonitrile after
prolonged irradiation as a function of triplet sensitizer energy.
successful yet. Anyway, the computations documented below
provide a further basis for this interpretation.
Finally, it is worth noting that the fumaric (trans) and maleic
(cis) amides used in this work can adopt several low-energy
conformations via rotations of the single bonds (CdC-CdO)
9
J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009 111

A R T I C L E S
Altoe` et al.
Figure 13. B3LYP/6-31G* calculations on 2 (a) and 3 (b) illustrating the different ground-state structures, associated vertical and adiabatic (0-0) triplet
T1 (ππ*) energies (energies values are relative to the E isomer ground state). The twisted minima on T1 (Min T1) and their relative energies have also been
reported.
interpolation between the S3 minimum and the CI S3/S2 point
(see below). After the conical intersection, the energy of S3
(now the dark πfπ* state, diabatically connected to the FC
S4(Ag) state) decreases until a new crossing is reached with
the nfπ* S2 state (CI S3/S2, see Figure 5). Interestingly,
this motion (from FC to the CI S3/S2) only involves bond-
stretching modes that preserve the planarity of the backbone,
i.e. no double bond isomerization coordinates have yet been
involved. Following decay through the first CI S3/S2 crossing,
it is conceivable that most of the molecules will be collected
on the minimum of the nfπ* S2 state (Min S2), by relaxing
on this surface (see Figure 5).
Again, the molecular deformations driving the system into
this point do not involve C-C twisting motions but the
rearrangement of the skeletal bond distances only. In particular,
it turns out that by following the S2 MEP from the CI S3/S2
crossing to Min S2 a crossing point with the S1 state is
intercepted (CI S2/S1 planar), which is almost degenerate and
geometrically very close to Min S2.
Thus, a sudden and fully efficient depopulation to S1 is
expected through this point. Then, following the S2fS1 hop,
relaxation takes place on the lowest excited S1 (nfπ*) state
and, finally, the planar minimum on this surface (Min S1) is
populated (see Figure 5).
In conclusion, according to this model, the final excited-state
outcome of the initial multistate deactivation process is to
populate the planar minimum (Min S1) on the lowest singlet
excited (nfπ*) state S1(Bg), which thus acts as the collecting
point of the photoexcited molecular population.
this Kasha rule-based model, it is through a planarity
conserving motion (FCfCI S3/S2fCI S2/S1 planarfMin S1)
that the system reaches the S1 minimum starting from the
FC region on a higher energy state. Since this process
involves a barrierless route, it is predicted to occur in an
ultrafast (subpicosecond) time scale.
The S1/S0 deactivation funnel (CI S1/S0 90°) can be
accessed from Min S1 by overcoming a barrier on S1 (TS)
of 11 kcal/mol (see Figure 6): this radiationless decay and
photoproduct formation route is unlikely to be very efficient,
due to the high energy barrier involved (see also the
discussion below). Thus, two other possible mechanisms for
deactivation and ground-state recovery from Min S1 may
become competitive, in principle: i) a radiative path (i.e.,
fluorescence, which has been computed to occur at 547 nm),
and ii) a radiationless channel involving an inter system
crossing (ISC) between S1 and T1 (which are almost degener-
ate at Min S1) and, following relaxation on T1, a back ISC
to S0 (a detailed description of triplet photochemistry will
be provided below in a dedicated subsection).
In conclusion, the following computational evidence may be
summarized for the static (i.e., Kasha rule-based) model (see
Figure 7):
i) The photoexcited population (or at least most of it) is
collected on Min S1 through a barrierless and ultrafast (sub
ps) relaxation process that preserves the planarity of the
system backbone.
ii) A barrier of 11 kcal/mol is involved on going from
Min S1 to CI S1/S0 90°: that is, for this channel to be
populated, energy leaking (i.e., vibrational energy redistribu-
tion) must take place to let the energy flow from the initially
populated stretching modes to the torsional (reactive) modes.
This process takes time (ps time scale at least). Additionally,
the kinetic energy is redistributed not only to the reactive
(central CdC bond) torsional mode, but it is dissipated on
several other (unreactive) modes, which means that not all
It is worth noting that this model is a Kasha-rule based
picture of the process that, substantially, assigns to the lowest
energy excited-state alone the leading role in dictating and
controlling the photochemical behavior of the chromophore.
Remarkably, the whole multistate relaxation path, from FC
to Min S1, is characterized by skeletal stretching deformations
only, while no torsions are involved. That is, according to
9
112 J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009

Fumaramide Threads
A R T I C L E S
Table 2. CASSCF(14,10)/ano-s and CASPT2/ano-s Energies of Fumaramide (E-1), Dimethyl-Fumaramide (E-2) and
Tetramethyl-Fumaramide (E-3). All the States are Ordered by CASSCF Energy
Geometry/
symmetry
E-1 ∆E (kcal/mol)
E-1∆E (kcal/mol)
E-2 ∆E (kcal/mol)
E-3 ∆E (kcal/mol)
State
S0
CASSCFa
CASPT2b (∆E TD-DFT in parenthesis)
CASPT2b
CASPT2b
FC
Ag
Bg
Au
Bu
Ag
0.00
114.91
118.03
158.14
159.63
82.87
0.00
95.75 (90.39)
100.01 (94.55)
119.55 (117.38)
120.98 (117.61)
80.23
0.00
97.91
0.00
95.21
99.12
S1 (nπ*)
S2 (nπ*)
S3 (ππ*)
S4 (ππ*)
T1 (ππ*)
T2 (nπ*)
T3 (nπ*)
T4 (ππ*)
T5 (ππ*)
S0
102.19
113.95
113.08
81.41
103.70
103.17
80.29
109.78
112.41
128.57
131.61
34.86
93.84
97.72
96.05
95.69
102.19
113.95
113.08
98.87
110.37
110.64
27.22
105.80
102.36
Min S1
S1 (nπ*)
S2 (nπ*)
S3 (ππ*)
S4 (ππ*)
T1 (ππ*)
T2 (nπ*)
T3 (ππ*)
T4 (nπ*)
T5 (ππ*)
S0
79.84
79.52
128.54
134.55
150.64
75.56
113.52
111.73
154.17
76.8
80.13
77.43
126.85
128.72
154.90
77.72
108.98
109.82
127.87
73.02
TS
S1
88.64
90.42
S2
146.43
148.86
156.16
82.30
138.73
122.90
153.76
73.98
S3
S4
CI S1/S0 90°
CI S1/S0 30°
CI S2/S1planar
High CI S2/S1
Min S3
S0
S1
85.24
81.18
S2
152.38
154.66
165.07
106.97
110.67
173.37
207.69
223.01
23.56
129.17
142.43
150.38
97.89
S3
S4
S0
S1
99.88
S2
163.11
160.97
181.49
16.54
S3
S4
S0
S1
98.04
83.24
89.07
S2
101.64
133.48
139.19
10.25
S3
108.46
107.71
12.01
S4
S0
S1
103.15
103.85
150.50
153.49
18.08
98.59
99.48
S2
S3
132.56
129.57
7.26
S4
S0
S1
99.24
74.60
81.37
S2
103.46
136.53
139.30
18.24
S3
104.03
106.85
11.06
S4
CI S4/S3
S0
S1
97.24
82.87
89.46
S2
104.02
133.30
138.13
24.97
S3
106.18
107.73
23.02
S4
CI S3/S2
S0
S1
92.82
82.70
S2
117.93
127.45
152.50
28.01
98.54
S3
99.59
S4
119.79
20.81
Min S2
S0
S1
95.34
98.95
138.65
141.22
55.75
81.35
S2
85.11
S3
115.58
116.77
55.24
S4
Min T1
S0
S1 (nπ*)
S2 (nπ*)
S3 (ππ*)
S4 (ππ*)
T1 (ππ*)
T2 (nπ*)
T3 (nπ*)
T4 (ππ*)
T5 (ππ*)
S0
S1 (nπ*)
S2 (nπ*)
S3 (ππ*)
S4 (ππ*)
T1 (ππ*)
T2 (nπ*)
T3 (nπ*)
T4 (ππ*)
T5 (ππ*)
122.15
123.19
153.19
154.77
55.06
121.53
122.08
152.79
154.30
55.26
122.71
123.93
152.52
154.88
55.00
121.22
121.65
152.99
154.56
109.84
112.90
133.64
135.23
56.02
110.85
111.99
133.78
136.28
56.47
109.18
110.95
119.37
120.66
56.04
112.20
113.22
135.39
137.97
TS
S0
a Fumaramide ground-state CASSCF/ano-s energy: -413.7670356036 au. b Fumaramide ground-state CASPT2/ano-s energy: -414.9076535320
au. Dimethyl-fumaramide ground state CASPT2/ano-s energy: -493.2672445046 a.u. Tetramethyl-fumaramide ground state CASPT2/ano-s energy:
-571.6186677367 a.u.
the energy gets available for the photoisomerization process.
This implies that this high-barrier channel must not be an
efficient radiationless (i.e., fluorescence quenching) deactiva-
tion route, and that S1 depopulation may involve, in principle,
other competitive deactivation mechanisms (such as emission
or ISC, see the next two points).
iii) A radiative path, involving fluorescence at around 547
nm from Min-S1, could in principle become competitive at these
9
J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009 113

A R T I C L E S
Altoe` et al.
Table 3. Quantum Yields and Ratios of Quantum Yields for
Photoisomerization of Compounds 4 and 5 in Acetonitrile
so that reactant back formation on the ground-state surface
becomes the favored process. Furthermore, the absence of
fluorescence, the ps excited-state lifetime and the different
singlet and triplet quantum yields observed, reveal that a direct
efficient radiationless decay channel from S1 to S0 exists (that
is, this route does not involve the triplet state, i.e. no ISC occurs).
Starting from these premises, we have investigated the intersec-
tion space between S0 and S1, performing series of constrained
conical intersection optimizations at different values of the
central CdC bond twisting angle. Thanks to this procedure, we
have identified an extended S1/S0 crossing seam which spans
structures of increasing energy from 90° (i.e., the fully twisted
and lowest energy CI S1/S0 90°) to 150° (CI S1/S0 30°,
possessing an almost planar backbone), see Figure 8.
Analogously, an extended crossing seam exists also between
S2 and S1. In particular, we have identified another crossing
between S2 and S1 (High CI S2/S1) that has a planar structure,
yet is much higher in energy then CI S2/S1 planar. Remarkably,
this crossing has the same energy as CI S3/S2 and it is close to
this (2.85 mass weighted a.u. distance only, see Figure 9). This
means that the S2/S1 crossing space spans a region of increasing
energy (and substantially planar structures) that approach the
high energy part of the S2 MEP. It is thus conceivable that high
energy regions of the S2/S1 crossing seam (such as the planar
High CI S2/S1) becomes accessible from CI S3/S2 (i.e., from
the beginning of the S2 MEP) for dynamics reasons: even normal
(thermal) vibrations may easily drive the system into the S2/S1
crossing space from the top of the S2 MEP. Hence, the system
does not necessarily need to fully follow the S2 MEP till to the
S2 minimum to hit the S2/S1 crossing seam and hop to S1, since
this process (i.e., the radiationless decay to S1) can occur much
earlier than predicted previously by the static (Kasha rule-based)
model. If that occurred, as it appears from these analysis, then
the system could jump to S1 at a much higher energy than CI
S2/S1 planar, thus accessing novel high energy regions of the
S1 PES such as the high energy (almost planar) structures of
the S1/S0 intersection space previously shown (see Figures 8
and 9).
In conclusion, it is conceivable that the system decays to high
energy regions of S1 via S2/S1 crossing points nearby the planar
and high energy High S2/S1 CI found close to the CI S3/S2,
and that from here it hits the S1/S0 crossing seam (and, therefore,
it is funneled to S0) much before reaching the twisted CI
minimum (CI S1/S0 90°): i.e. deactivation to S0 may occur via
high energy almost planar (or only slightly twisted) S1/S0 CI
structures. If this were the case, besides opening a very efficient
radiationless decay channel to S0 that quenches the fluorescence,
a very small photoisomerization quantum yield would be
expected, since reactant back formation (following decay from
almost planar molecular structures) would be the most favored
event. A schematic representation of this behavior is reported
and summarized in Figure 10.
This “dynamic interpretation” of fumaramide deactivation can
be schematically represented by using two bidimensional
intersecting surfaces as models for the S1 and S0 PES (see Figure
11). The x-axis represents the relevant stretching coordinate and
the y-axis the rotation of the central CdC bond. It is apparent
that an extended crossing seam exists between these two surfaces
even for small torsional angles, although at higher energies.
Thus, if the system may hit these regions, it can hop down to
S0 before reaching the minimum on S1. Under this hypothesis,
it comes clear why no emission is observed. Moreover, the low
photoisomerization quantum yield observed is a consequence
Isomer
pair
Direct
irradiation
Sensitizeda
b
b
OEfZ
OZfE
O
EfZ/OZfE
)
OEfZ
OZfE
OEfZ/OZfE)
4
5
0.044 0.044
0.015 0.0028 5.27c, 5.58d, 5.36e 0.20 0.0008 37f, 250e
1.05c, 1.15d, 0.99e 0.17 0.20
1.07f, 1.18e
a Using benzophenone as the sensitizer. b At 254 nm. c From the
photostationary state composition upon 300 nm irradiation (Equation 1).
d From the photostationary state composition upon 254 nm irradiation
(Equation 1). e Determined directly from the quantum yields. f From the
photostationary state composition.
conditions, but considering the low oscillator strength of the
transition, it is unlikely to have a high radiative rate and this
path will not play an important role.
iv) Also a competitive (fluorescence quenching) S1fT1 ISC may
not be ruled out at Min S1, since singlet and triplet are almost
degenerate here and their spin-orbit-coupling (SOC) is significant
(see the subsection below for a detailed discussion on triplet
photochemistry).
Experiments support our hypothesis for an unfavored/inef-
ficient photoisomerization process, the singlet photoisomeriza-
tion quantum yield recorded for the threads 4 and 5 being rather
small (see Table 3). On the other hand, emission is not observed
and a very short (ps time scale) excited-state lifetime has been
recorded for 3-6, revealing that an efficient and ultrafast
radiationless decay channel exists that quenches fluorescence
and quickly triggers excited-state decay and reactant back
formation. Although apparently rather surprising for a spin-
multiplicity change event such as ISC (which generally implies
longer timescales to occur), this process seems to be the unique
way to account for the experimental data, provided it is ultrafast
and very efficient. If this were the case, singlet and photosen-
sitized (i.e., triplet) photochemistry should lead to the same
outcome. That is, the recorded singlet photoisomerization
quantum yields for the threads should coincide with the triplet
values, since singlet photochemistry eventually turns out to
become a triplet process. It is anticipated here that this is not
the case: experimental results reported in Table 3 shows that
the photoisomerization quantum yields recorded for the singlet
and the triplet are very different.
Thus, while it can be concluded (according to the Kasha rule-
based view) that Min S1 collects most of the excited population
and that deactivation to S0 must involve either emission or ISC,
unfortunately neither of these mechanisms is supported by the
experiments. This reasoning leads to the conclusion that
fumaramide is not a Kasha rule-based system, i.e. its photo-
chemistry cannot be accounted for dealing with the lowest
energy points of each electronic state only. For a more correct
and realistic interpretation of the experimental results nonstatic
effects need to be accounted for, i.e. it is mandatory to explore
also higher energy regions along (or close to) the MEPs and
the crossing seams, which may turn out to become important
for dynamics and will be the focus of the dynamic model
presented below.
(B) The Dynamic Model. The inconsistencies found between
the experiments and the Kasha rule based-model reported in
the previous section, motivated us to explore higher energy
regions of the photochemically relevant potential energy surfaces
in order to identify other feasible deactivation mechanisms. The
low isomerization quantum yields observed in the singlet
photochemistry suggest that deactivation must involve funnel
regions where the CdC central bond is not yet fully twisted,
9
114 J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009

Fumaramide Threads
A R T I C L E S
Table 4. Fumaramide Bond Distances (Å)
Geometries
Min S2
Bond
FC
Min S1
TS
CI S1/S0 90°
CI S1/S0 30°
CI S2/S1 planar
CI S2/S1 High
CI S3/S2
Min T1
Min S3
C1-C2
C1-C3
C2-C4
O1-C3
O2-C4
N1-C3
N2-C4
1.340
1.492
1.492
1.207
1.207
1.359
1.359
1.409
1.359
1.471
1.368
1.209
1.390
1.377
1.482
1.330
1.473
1.365
1.209
1.388
1.379
1.474
1.349
1.478
1.324
1.209
1.384
1.377
1.460
1.376
1.507
1.352
1.198
1.378
1.349
1.418
1.396
1.396
1.284
1.284
1.362
1.362
1.427
1.403
1.397
1.288
1.284
1.364
1.372
1.461
1.418
1.416
1.294
1.271
1.349
1.375
1.458
1.370
1.458
1.346
1.215
1.339
1.368
1.466
1.471
1.470
1.210
1.209
1.372
1.378
1.430
1.413
1.413
1.273
1.273
1.350
1.350
of the substantially planar molecular geometries (i.e., small
rotation of the central CdC bond) of the S1/S0 crossings that
trigger the decay to the ground-state and promote reactant back
formation.
mophore, but the sterically more crowded Z-isomer is twisted
at the C-C single bonds. Interestingly, this situation does not
apply to the case of the threads 4, in which the photostationary
state has roughly equal amounts of both isomers. This is due to
the fact that Z-4 adopts a planar structure, with an internal
hydrogen bond. As a result, the triplet energies of E-4 and Z-4
are not very different, and selective energy transfer is not
possible. The situation is illustrated in Figure 13 (that reports
results from B3LYP/6-31G* calculations), using compounds
2 and 3 as models for 4 and 5, respectively.
Computational Analysis of the Isomerization in the Triplet
T1 State: S1-T1 and T1-S0 Inter-System Crossings. Our calcula-
tions (at both CASPT2 and B3LYP level) predict a higher (ca.
80 kcal/mol, see Table 2 and Figure 13) vertical (FC) S0-T1
energy gap than the recorded (67 ( 2 kcal/mol) experimental
value. It is reasonable to associate the latter value to the energy
difference between the planar ground-state minimum and the
relaxed planar structure on the triplet T1 surface (called 0-0planar
thereafter). Very remarkably, the computed singlet-triplet
0-0planar energies for E-2 and E-3 (65.9 and 61.9 kcal/mol
respectively, see Figure 13) match fairly well with the experi-
mental value recorded for the threads (note that compounds 2
and 3 are used as models for threads 4 and 5, respectively).
Since our computations in vacuo closely reproduce the spec-
troscopic data recorded in solution, solvent effects do not play
a major role in this specific case and the isolated system should
represent an accurate model of the photochemical behavior in
solution.
Fairly stated, a computational validation of this model would
require a systematic dynamical investigation by performing
trajectories computations (anyway, a simple experiment is
suggested - see section 3.4 below - that may be alternatively
used to validate this mechanistic scheme). Although this is out
of the scope of this work (CASPT2 dynamics is not affordable
even for such a small system), it is worth noting that this non-
Kasha rule behavior does characterize the photochemistry of
other organic chromophores, such as azobenzene40,41 and its
derivatives. For instance, cyclohexenylphenyldiazene (a simple
surrogate of the azobenzene photochromic unit) has been
recently investigated, that show a very similar (Kasha rule
breaking) photophysical scheme.40 This would represent the
second example where such a mechanistic model applies.
Finally, it must be noted that no side products have been
detected in these simulations, which accounts for the effective-
ness of the photoactive fumaramide unit as its photoisomeriza-
tion can be repeated indefinitely without losses due to side
reactions. In particular, excited-state relaxation channels involv-
ing amide bond rotations (which may be a potential source of
side reactions and may produce different isomers as side
photoproducts if different nitrogen substituents are present) turn
out to be unfavored processes (see end section S1 of the
Supporting Information for a detailed discussion).
E/Z Photoisomerization by Triplet Sensitization. In order
to explore the reactivity on the lowest triplet state T1 (ππ*),
the E/Z photoreaction of threads 4 and 5 has been studied by
using a series of triplet photosensitizers, and the isomerization
quantum yields with benzophenone have been recorded and
collected in Table 3. Figure 12 shows the extent of conversion
to the Z-isomer after prolonged irradiation of the E-isomers of
4 and 5 in the presence of sensitizers of different triplet energies.
For very low sensitizer energy, conversion does not occur, but
when the triplet energies are high enough, photostationary states
are obtained. The experimental energy gap recorded in the thread
4 between S0 and T1 is 67 ( 2 kcal/mol (the vertical triplet
energies are probably higher, see also the discussion below),
while the onset of conversion occurs at 59 kcal/mol for 5, and
at 65 kcal/mol for 4. For a range of triplet energies, E-5 can be
almost quantitatively converted to its Z-isomer. This is because
the vertical triplet energy of Z-5 is sufficiently higher that the
sensitizer can only transfer its energy to the E-isomer, which is
thus effectively depleted. A similar situation is found for many
conjugated alkenes24 when the E-isomer has a planar chro-
The T1 wave function is characterized by a single-electron
occupation of the HOMO (π) and LUMO (π*) orbitals on the
central double bond (orbitals 7 and 8, see Figure 1). As a
consequence, a strong weakening of the central CdC double
bond is expected in this state and, consistently, geometry
optimization on T1 leads to a 90° twisted central double bond
minimum (Min T1, see Figure 14). This is a recurring property
of many olefinic systems (e.g., ethylenes, stilbenes, azobenzenes,
etc)40,41 in their lowest triplet (ππ*) state.
Interestingly, besides the usual sensitization process, T1 can
be populated by intersystem crossing (ISC) from the S1
minimum (Min S1). In fact, S1, T2 (nπ*) and T1 (ππ*) are
substantially degenerate at this point (the three states are within
2.5 kcal/mol, see Figure 14 and Table 2). Additionally, the
computed spin-orbit-coupling (SOC) between T1(ππ*) and
S1(nπ*) is quite large (28 cm-1) at Min S1, suggesting that S1
< 199.T1 ISC can be very efficient and fast (ps time scale,
see also the discussion below). Furthermore, constrained
geometry optimizations on T1 at different twisting angles of the
central CdC bond show that a wide region around Min T1
(between 70° and 100° of twisting) exists (Figure 14) where
the PES of T1 and S0 are substantially degenerate. Although
T1-S0 SOC values are rather small along the twisting mode in
this region (ca. 0.02 cm-1, see Figure 14), an out of plane mode
(40) Conti, I.; Marchioni, F.; Credi, A.; Orlandi, G.; Rosini, G.; Garavelli,
M. J. Am. Chem. Soc. 2007, 129, 3198.
(41) Cembran, A.; Bernardi, F.; Garavelli, M.; Gagliardi, L.; Orlandi, G.
J. Am. Chem. Soc. 2004, 126, 3234.
9
J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009 115

A R T I C L E S
Altoe` et al.
Figure 14. T1 and S0 energy profiles along the central bond twisting coordinate for the region surrounding the triplet twisted minimum (Min T1): constrained
CASSCF/6-31G* geometry optimizations followed by single point CASPT2/ano-s corrections are used for this purpose. The energies of S1, T2 and T1 at
the Min S1 geometry are reported on the left up side. Computed SOC values are also reported in italic. The vibrationally activated T1∼>S0 ISC in the Min
T1 region is also displayed: pyramidalization at the carbonyl carbon C3 (C4) is the orthogonal vibrational mode that triggers T1∼>S0 ISC by enhancing
T1-S0 SOC values.
of 730 cm-1 exists that involves the asymmetric pyramidaliza-
tion of the carbonyl carbons C3 (C4) and leads to large T1-S0
SOC values: according to our computations, zero point energy
oscillations for this mode already lead to SOC values up to ca.
30 cm-1 (see Figure 14). Thus, it is apparent that another
efficient and ultrafast (ps time scale) ISC from T1 to S0 that is
activated vibrationally, can easily occur in the twisted region
around Min T1. It is suggested here that the slower (19 ps) decay
component observed in fumaramides singlet photochemistry (see
the discussion above about time-resolved spectroscopic data)
is related to this triplet T1 channel and can be assigned to its
population (S1 ∼> T1) and/or decay (T1 ∼> S0), see Figure
10. Although these ISC processes may appear surprisingly fast
(spin-multiplicity change events generally implies longer ti-
mescales to occur), a similar behavior has been recognized in
other olefinic chromophores (e.g., ethylenes, azobenzenes,
etc),40,41 that, as mentioned above, display a very similar T1/S0
topology with a very efficient and ultrafast (ca. 10 ps time scale)
T1 ∼> S0 ISC (see also the discussion in the following
subsection). This argument nicely supports the previous assign-
ment for T1 being responsible of the 19 ps decay component
observed in E-4, as it well matches with the time scale of these
ISC processes. Additionally, the very short T1 lifetime implies
a difficult spectroscopic detection of this state and well accounts
for the lack of spectroscopic observations of T1 in fumaramides.
Interpretation of Photoisomerization Quantum Yields in
the Triplet State. The experimental quantum yield observed for
the photosensitized isomerization of the threads is about 1 order
of magnitude higher than the one recorded in the singlet (see
Table 1). Remarkably, this higher value is in agreement with
our results that show that the triplet T1 state crosses (and is
degenerate with) the ground-state in the twisted region and that
the path to this twisted region is barrierless (Figure 14). In fact,
these findings support the idea that the decay of T1 can occur
more efficiently via the torsion mechanism with a rate depending
on the strength of the relevant spin-orbit-couplings similarly
to what reported for other olefinic chromphores such as stilbenes
and azobenzenes.40,41 Specifically, in our previous works on
azobenzene41 and cyclohexenylphenyldiazene40 (which, as
mentioned above, present a similar T1/S0 topology with degen-
erate S0 and T1 states at the twisted structures), we have
demonstrated that SOC values of ca. 20 cm-1 (which are of the
same magnitude as the ones computed here for fumaramide,
see the discussion above and Figure 14) are large enough to
ensure that the rate of the T1 ∼> S0 ISC process is of the order
of 1011 s-1. This rate indicates that the system can decay from
T1 to the ground-state in a ca. 10 ps time scale and that torsion
provides indeed a very efficient mechanism for the T1 decay
and isomerization. Thus, a situation similar to that found in
azobenzene41 and cyclohexenylphenyldiazene40 is suggested
here for fumaramide, which should lead to a very short (i.e., ps
time scale) T1 lifetime in agreement with the difficult spectro-
scopic detection of this state.
In conclusion, a more efficient isomerization of the central
double bond (than in the singlet state) is expected through this
route. Indeed, this is what recorded photoisomerization quantum
yields reveal. Finally, the existence of an extended T1/S0
degenerate region that starts from ca. 70° to ca. 100° of central
CdC bond twisting (see Figure 14) well explains why the
observed triplet quantum yields are less than 0.5: as the
degenerate region is first encountered on the reactant side (i.e.,
before 90° fully twisted structure), an ISC leading mainly to
reactant back formation is always expected, as the experiments
reveal indeed.
3.3. Dynamics vs Static Control of the Photochemical
Process. It would be highly rewarding if the dynamics vs the
static model could be tested and proved experimentally. The
key point for such an experiment would be the possibility to
pump directly to the S1 state (which is symmetry forbidden)
with a small and controlled amount of (excess) vibrational
energy such as hitting a vibronic level that can borrow intensity
by Herzberg-Teller mechanism. This would populate selec-
9
116 J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009

Fumaramide Threads
A R T I C L E S
tively the S1 minimum region, and leave no energy available
for activation of the radiationless decay channel to the twisted
CI (CI S1/S0 90°). Under such conditions, which would bias a
Kasha rule-controlled system, emission and/or ISC could in
principle become competitive, as already discussed in subsection
3.2 above. If this were the case, emission might be observed.
Alternatively, if ISC occurs and is a competitive and efficient
process (as we indeed expect to be, see subsection 3.3 above),
singlet photochemistry will turn out to become a triplet process
and the same photoisomerization quantum yield observed in
photosensitized experiments should be recorded also for singlet
photochemistry. This would represent a novel and intriguing
aspect of fumaramide-based photocontrolled molecular devices,
which may be suitably exploited to tailor their photochemical
properties and reactivity as a function of the technological
applications pursued. However, impeded by the small absorption
cross section of the S1 state, such experiments have as yet not
been successful.
which provide a precise and reliable picture of its photochem-
istry. Furthermore, it is also the base of an important family of
photoresponsive nanodevices such as the rotaxane shown in
Scheme 1.16,18 In this respect, it is worth noticing that fuma-
ramide photochemistry does not involve efficient side reactions,
so that it can be operated continuously under illumination with
very little loss, which makes it a very effective photoactive unit
in nanodevices.
Our study also suggests a way to operatively achieve a static
or dynamics control of the photochemistry of this system, thus
switching on between two ‘modes’ that lead to different
photochemical outcomes (i.e., different photoisomerization
quantum yields, excited-state lifetimes, etc). From a computa-
tional standpoint, full quantum-mechanic nonadiabatic molecular
dynamics computations will have to be used to quantitatively
investigate this issue and access the importance of nonequilib-
rium (i.e., dynamics) effects along the relaxation process. This
will be to topic of future investigations. Nevertheless, the static
description of the PESs shown here provides a rationale for
fumaramide photochemistry, both in the singlet and triplet
manifold, that accounts for the observations recorded in the
isolated threads.
4. Conclusions
A computational investigation of fumaramide photochemistry
(as a model of fumaramide threads) jointly to photochemical
and time-resolved spectroscopic measurements on fumaramide/
maleamide compounds revealed a complex multistate photo-
physics of this photoactive unit, opening several intriguing
scenarios. We have identified a pattern of possible deactivation
mechanisms that may involve, in turn, internal conversion,
intersystem crossing or emission. These are all based on the
analysis of the minimum energy paths computed on the PESs
of the photochemically relevant excited states (which are
supposed to drive the relaxation of the excited molecules) and
the crossing seams between them (which are thought to trigger
radiationless decay). For the fumaramide unit in vacuo, it has
been shown that a static (i.e., Kasha rule based) interpretative
model (where Min S1 acts as the collecting point of the
photoexcited population) does not account for all the experi-
mental evidence, while a dynamic picture (involving decay to
S1 and S0 at higher energy regions of the crossing seams) may
easily explain all the observations recorded for fumaramide
compounds in solution. This provides a solid computational
evidence for fumaramide not being a Kasha system (i.e., Kasha
rule is violated).
In conclusion, a rationale for the spectroscopic properties,
photoisomerization quantum yields and lifetimes observed in
the direct (i.e., singlet) irradiation experiments has been
provided, and a model for the sensitized (i.e., triplet) photo-
chemical reactivity as been drawn that accounts for the different
behavior (e.g., different isomerization quantum yields) observed
in the threads.
Acknowledgment. The support by funds from Bologna Uni-
versity (Progetti Strategici d’Ateneo 2005: Progetto CompRenDe)
is gratefully acknowledged. We thank CINECA and INSTM for
granted calculation time. Part of this work has been supported by
the European Union project STAG (STRP 033355), and The
Netherlands Organization for the Advancement of Research (NWO).
Dr. Bert Bakker synthesized N,N,N′,N′-tetramethylfumaramide.
Supporting Information Available: Conformational analysis
(section S1), electronic absorption spectroscopy (section S2),
irradiation experiments (section S3), complete references26 and29
(section S4), Cartesian coordinates of all the optimized structures
discussed in the text (section S5), one table for the triplet
sensitization experiments (Table S1) and two figures (Figure
S1 and S2). Complete refs 26 and 29. This material is available
free of charge via the Internet at http://pubs.acs.org.
This is not the first example of Kasha rule failure: azobenzene
and related systems (such as the phenylazocyclohexene) behave
the same, as it has been previously shown by both experimental
and theoretical works.40,41 Nevertheless, fumaramide provides
a very interesting case because it is a small molecule that can
be studied using very accurate high level ab initio methods,
JA802531J
9
J. AM. CHEM. SOC. VOL. 131, NO. 1, 2009 117