Hula-twist cis–trans isomerization: The role of internal forces and the origin of regioselectivity

Journal of Photochemistry and Photobiology A: Chemistry 237 (2012) 53–63

Contents lists available at SciVerse ScienceDirect

Journal of Photochemistry and Photobiology A: Chemistry journal homepage: www.elsevier.com/locate/jphotochem

Invited feature article

Hula-twist cis–trans isomerization: The role of internal forces and the origin of regioselectivity Werner Fuß ∗ Max-Planck-Institut für Quantenoptik, D-85748 Garching, Germany

a r t i c l e

i n f o

Article history: Received 29 November 2011 Received in revised form 23 January 2012 Accepted 28 January 2012 Available online 9 April 2012 Keywords: Polyenes Double-bond isomerization Potentials Conical intersections Branching Momentum effects

a b s t r a c t In photochemical Hula-twist isomerization of conjugated polyenes, a double bond and an adjacent single bond twist concertedly by 180◦ . It is here considered to be driven by passing through the last conical intersection (CI), between potentials of the dark covalent S1 (2A) and the ground S0 (1A) states. If several such CIs are available, that is, if isomerization of different bonds is possible, regioselectivity can be caused by a sterically induced pre-twist in S0 , so that the pre-twisted group isomerizes: this deformation is amplified in the Franck–Condon region of the spectroscopic (1B) state; on entering the 2A surface with this geometry, the nearest S1 /S0 CI will be chosen, if the reaction is ultrafast. If the reaction is slower, that is, if there is a barrier before each CI, a local pre-twist reduces one of them and thus also selects the site of isomerization. Another pre-twist can be caused in the initially excited (ionic) 1B state by electrostatic effects, thus also giving rise to regioselectivity. The explanations only consider potentials of the molecule, in the majority of cases with no external forces. Also other observations are summarized that support the idea that Hula twist can work without influence of the environment. A variant of Hula twist is also proposed and used, in which the torsion of the two adjacent bonds is not disrotatory as usual but conrotatory. © 2012 Elsevier B.V. All rights reserved.

1. Introduction Hula-twist (HT) isomerization is a mechanism for photochemical cis–trans interconversion of polyenes, in which only one CH group rotates by 180◦ , whereas the groups attached to it reorient in the original plane. An alternative description says that these two groups (or the CC double bond and an adjacent single bond) rotate in disrotatory direction versus the CH group (Scheme 1). Originally postulated in 1985 by Liu et al. for cases, where volume saving seems important [1,2], it was experimentally identified only in 1998 [3] in photoisomerization of previtamin D. In the mean time Liu et al. have found plenty of additional examples, which they have repeatedly reviewed [4–8]. As nearly all examples were carried out in a matrix or under other conditions of mobility limitation, Liu ascribes this mechanism to a medium effect, pointing to the minimal volume need of Hula twist as compared to one-bond flip (OBF). On the other hand, quantum chemistry points to the slope through an S1 /S0 conical intersection (CI) (i.e., the crossing between the potentials of the dark covalent 2A state with that of the ground state), which could drive the molecule towards HT also without interaction with a medium, as is pointed out in [3,9,10].

∗ Tel.: +49 89 3201807. E-mail address: [email protected] 1010-6030/$ – see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jphotochem.2012.01.019

Part of the present work (Section 2) is devoted to summarize the evidence that HT is due to internal driving forces and is the standard path for cis–trans isomerization of nonpolar polyenes. The evidence is based primarily on experimental observations and their comparison with quantum chemical calculations; the latter alone have not yet given a clear-cut decision whether and when HT is preferred over OBF. Sections 3–5 then give explanations, again based on internal forces (pre-twist for steric or electrostatic reasons), for regioselectivities observed in the literature, that were not (or only partially) explained before. It should be noted here (see, e.g. [11]) that normally (with possible exceptions in polar solvents) in all conjugated polyenes (but not in strongly polar double-bond systems such as protonated Schiff bases of unsaturated aldehydes) a two-electron excited totally symmetric (2A: 2Ag in C2h and 2A1 in C2v ) “dark” state at suitable geometry reaches to below the HOMO → LUMO excited (1B: 1Bu in C2h and 1B2 in C2v ) “spectroscopic” state. The latter is optically strongly coupled with the ground state and is usually first populated. It is depleted in all polyenes within 50–250 fs around the 1B/2A CI (survey in [11]) to the dark state, that has only a very weak optical transition probability to the ground state. From there, the population can flow (with time constants that vary by orders of magnitudes [11]) through the 2A/1A CI to the ground state. This is the S1 /S0 CI mentioned above that is also responsible for the photochemical transformation. Valence bond theory describes the 1B

54

W. Fuß / Journal of Photochemistry and Photobiology A: Chemistry 237 (2012) 53–63

Scheme 1. Two alternative ways to describe Hula twist isomerization (i.e., its disrotatory variant, see below) of a conjugated polyene. The broken line not only indicates the rotation axis but also a partial bond formed in the CI. HT and OBF differ in the conformer produced; if the molecules are sufficiently asymmetric to distinguish the conformers (and if the experiment can identify them before rotamerization, e.g. by sufficient time resolution or by freezing the conformer isomerization), there is no ambiguity in the decision, which pathway has been taken.

state as ionic (superposition of zwitterionic structures) and the 2A state as covalent. The former is sensitive to electrostatic effects and its energy may be lowered by (polar or polarizable) solvents [12], whereas the latter is much less so. However, with strong distortion of the molecule, the two states can mix, so that the covalent state can borrow ionic character at such a geometry (see, e.g. [13]). For convenience, the notation 1B and 2A is here kept even when mixing plays a role. Regioselectivity in Hula-twist isomerization was reported by Liu’s group in substituted diphenylbutadienes, in cinnamic esters [14,15] and in naphthyl-phenyl-ethenes [16]. In all these cases, rotation of the CH group closer to the phenyl ring (HT-1) is preferred over that in ␤ position to it (HT-2). Whereas electronic or mass effects were tentatively assumed in [14] for explanation, I invoke the sterically induced pre-twist of the phenyl group (Section 3). A different kind of regioselectivity, caused by electrostatic effects, was found by the groups of Squillacote, Liu and Singh [17–21] and discussed in terms of a zwitterionic state, an allylcation methylene-anion intermediate. Further such examples are discussed on pp. 208–211 of the review [22]. This either means that the (ionic) spectroscopic state (1B) takes over the control, or that the dark state (2A) is not purely covalent but can assume some ionic character. The latter property in fact resulted from the recent quantum chemical calculation of butadiene excited states [13]. On the other hand, this calculation favored OBF over HT, which seems in conflict with the mentioned experimental results, as discussed in Section 2. It is here suggested (Section 4) that in the case of ultrafast reactions, zwitterionic polarization in the 1B state can cause (pre)distortion that in turn guides the molecule on the 2A surface to the (HT-type) conical intersection that is connected with the observed regioselectivity. Section 5 then argues that in slower reactions the effect of pre-twist still exists, but electrostatic effects are not easy to foresee. The decision, into which CI (i.e., into which part of the intersection space) the molecule enters, depends on fine differences of slopes or/and barriers, which are still hard to reliably find out by quantum chemical calculations alone. Section 2 therefore recalls and compiles also experimental evidence. Little noticed theoretical results and some plausibility considerations (such as a correlation diagram) suggest also the existence of another form of HT, with conrotatory (instead of the usual disrotatory) motion of molecular moieties, that is then invoked in Sections 3–5 to explain the pre-twist induced regioselective HT. 2. Hula-twist versus one-bond flip in nonpolar polyenes The traditional view of photochemical cis–trans isomerization of singlet-excited olefins is that of a one-bond flip (OBF) mechanism, in which one double bond rotates by 180◦ . This is established for mono-olefins, where the ␲␲* excited state and the doubly excited dark state have a deep minimum at a twist angle of 90◦ (ignoring other coordinates such as pyramidalization of one carbon that can further lower the energy but do not leave permanent

signatures in the products) (see, e.g. [23]). Also in butadiene there is a driving force – although weaker – towards twist in both the 1B and 2A state [13] (although on the 2A surface probably not in OBF direction, see below), and in trienes it is probably further attenuated. Longer polyenes such as octatetraene are planar in their S1 minimum, as known from spectroscopic evidence [12,24]. In spite of such systematic variations, it is popular to assume OBF for all conjugated systems (e.g., in the recent edition of a photochemistry textbook [25]). Evidence that the simple OBF mechanism cannot be the only one or not even the predominant one, came independently from (a) computation of photochemical reaction paths and (b) the conformation of the primary products of the photoinduced cis–trans isomerization of certain polyenes. 2.1. Computational results In a series of quantum chemical calculations, the groups of Olivucci, Robb and Garavelli found that in conjugated hydrocarbons (polyenes of different length, styrenes and stilbenes) the minimum energy CI between the lowest singlet excited (2A) and the ground state (1A) involve concerted twisting about a double bond and an adjacent single bond [26–35], that is a motion as in HT; a review emphasizing conformational aspects [36] and another one focusing on the discussion of Hula twist [10] is available. Due to the common geometrical structure, a triangle that was interpreted as a three-electron three-center bond, the CI was initially called kinktype [33] and later also HT-type [10]. The twist angle of the two bonds is less than 90◦ (about 60◦ ), and it was pointed out in [10] that on relaxing from the CI on the lower cone the molecule can either complete the HT to 180◦ or turn one of the two bonds back (resulting in “attempted (aborted) HT” leading either to an OBF product or another rotamer) or revert both twists (resulting in internal conversion): as there are no deep valleys on the lower cone [10], it exhibits practically no directional force, and the actual path will be sensitive to momentum effects or influence of an environment. Momentum can be accumulated from the entrance cone to the CI, and the environment can change the slope in specific directions. The calculations were done by CASSCF (complete active space self-consistent field) with all ␲ electrons and all ␲ orbitals as active space. In a recent work [13], Levine and Martínez pointed out that this method does not take electron correlation into account in a balanced way for both the dark (2A) and the spectroscopic (1B) states. They therefore reduced the active space in butadiene to three ␲ orbitals (CASSCF(4/3)), which is better in this respect. In fact, the results compared well with those of CASSCF(4/4) with PT2 correction for correlation. According to these calculations [13], which also followed the dynamics of the molecule, butadiene twists a terminal CH2 group already on the 1B state by 90◦ and with this structure reaches on the 2A surface an OBF-type CI, where it passes to S0 . This twist also imparts some zwitterionic character to the dark state by mixing the two states. An HT-type CI was also found (called there “transoid CI”) but located at higher energy. For systems polar enough that the (ionic) 1B-like state is below the (covalent) 2A-like state, also the Olivucci group found that an OBF-type CI is more easily accessible than an HT-type [10], and both types compete in intermediate cases [10]. Hence zwitterionic electronic structures seem to favor OBF type distortions, whereas covalent states initiate motion towards HT-type isomerization. However, there is also a recent calculation on previtamin D (a steroid triene) on the basis of time-dependent density-functional theory that favors an HT pathway from the spectroscopic state [37]. Furthermore, an accessible HT-type CI (though not at lowest energy) was also found by CASSCF for a protonated Schiff base of pentadienal [38]. It is interesting that HT-type S1 /S0 CIs were recently also found for different systems: symmetric monomethine cyanines [39], using a minimalistic model with

W. Fuß / Journal of Photochemistry and Photobiology A: Chemistry 237 (2012) 53–63

parameter-dependent closed-form solution. These systems are prototypes not only of many dyes but also of the chromophore of green fluorescent protein. It seems that calculations alone cannot yet decide, if and under which conditions an HT-type path is chosen. A comparison with experimental observations would therefore be desirable. This work focuses on nonpolar polyenes, because most evidence is available from this class of compounds. An alternative CI leading to HT products (but at higher energy than the kink-type CI) was suggested by CASSCF calculations of Wilsey et al. [40,41]: it involves partial H migration and an inversion at C(2) , the carbon that in the path via the kink-type CI rotates out of plane. However, it seems that the experimental observations are easier to explain by the kink-type CI (Sections 2.2 and 2.3). 2.2. Experimental observations The many cases of Hula twist observed in a matrix or with other constraints, reviewed by Liu [4–8], were already mentioned. Liu assumed HT to be a path of higher free energy of activation (in particular an entropy term) than OBF in the free molecule and interpreted the preference for HT in a matrix to a volume constraint imposed by it; it had to be assumed that the higher activation energy is selectively reduced by the matrix. In fact, similar phenomena have been observed. Thus, free E,E-octatetraene molecules (in a supersonic jet) isomerize to the Z,E form with an activation energy of 2100 cm−1 (0.26 eV) [29,42], whereas in n-octane solution at 4.2 K a single bond rotates to form the 2-s-cis conformer [24], a reaction interpreted in [5,6] as HT-2 isomerization. From the low temperature, one can estimate a barrier of ≤50 cm−1 (preexponential factor 1013 s−1 assumed, S1 lifetime 200 ns [24]), whereas isomerization of a central C C bond takes 870 cm−1 of activation energy under the same conditions [24]. Obviously the matrix has generally lowered the energetic barriers (probably an effect of the polarizability of the condensed phase [10]), but due to restricted free volume less efficiently for isomerization of the central bonds. Another instructive example is 2,3-dimethylbutadiene-1,4-d2 in an Ar matrix, in which Squillacote and Semple observed trans → cis isomerization of a terminal CHD group [43] instead of an HT-2 reaction observed in unsubstituted butadiene (see [5]). Obviously, due to the methyl substituents, the latter is more inhibited by the matrix than the simple OBF-type rotation of the terminal groups. (Instead of OBF, Liu in [4–7] offers HT-1, involving rotation of the C C and a C H bond. But the electronic interpretation of HT needs p orbitals at all three centers to take part. In Section 2.3, the example is interpreted as aborted HT-2.) Hence a matrix can in fact change relative energies of activation and thus redirect the reaction; according to Section 2.1, a redirection can also take place on the lower cone of a CI, where only (change of) slopes are involved and the concept of activation energy and entropy is not applicable. However, if in free molecules HT is the lower energy path as suggested by the Olivucci calculations, the matrix may have served in many cases only to conserve the primary conformer, as suggested in [3,9,10]. (The HT and OBF products differ only in their conformations.) It would therefore be desirable to detect the conformer products of free molecules, before they can rotate their single bonds. Such a time-resolved experiment has in fact been reported [44]: the initial orientation of the product trans-stilbene relatively to that of the reactant cis-stilbene was easy to explain by HT but not by OBF. In the same work [44], the time for departure from the dark state (or perpendicular minimum) of gaseous stilbene was compared with that for a derivative (“stiff stilbene”), in which single-bond rotation was blocked. As the time for the stiff stilbene was much longer, it was concluded that single-bond rotation (as in HT) facilitates isomerization; stilbene is, however, a limiting case, as the two bonds rotate asynchronously though concertedly [44].

55

HT in free molecules was probably also detected in another example: when E-hexatriene was irradiated in a cold molecular beam and then deposited with Ar on a cold substrate, the 2-s-cis conformer was observed as a product [45], which was interpreted as HT-2 reaction in [8]. Another possible case of HT of free molecules may be indicated by the isomerization times of different cinnamic esters in solution (Section 3). These examples demonstrate that HT can be completed in molecules also without any constraints. By contrast, for OBF in free molecules of conjugated polyenes so far no evidence is known, except in cases with blocked rotation of the single bonds. Liu et al. [8] also point to the viscosity dependence of the cis-stilbene isomerization rate: the observed deviations from prediction by the Kramers theory can be interpreted by assuming that at low viscosity the molecule follows the volume-demanding OBF path but switches to the volume-saving HT mechanism in viscous environment [8]. (A “stiff” derivative of stilbene with suppressed rotation of single bonds in fact obeys Kramers’ theory very well [8].) This seems like a support of the idea that the unconstrained molecules perform OBF. On the other hand, the time-resolved experiment with the free stilbene molecule suggested that an HT CI is reached via a path gradually turning from a C C twist to a direction involving the HT-typical torsion of both the C C and adjacent C C bonds [44]. With increasing viscosity, inhibiting OBF-type motion, this change of direction versus HT could just happen earlier [10]. Ref. [10] also points to the very minor viscosity dependence in diphenylpolyenes, where HT is expected to be more synchronous than in stilbene; this observation is hence consistent with HT at all viscosities. A number of other observations support a path through an HTtype CI in free molecules without excluding an OBF end product via aborted HT:

(a) The sum of the quantum yields for cis → trans and trans → cis isomerization of polyenes is in general  1 (see e.g. Table 1 in [22]); if forward and backward reaction pass through the same intermediate and if other reactions are negligible or can be corrected for, one expects 1 for this sum. Hence, as already pointed out in [10], this small sum is in contradiction to a common intermediate such as an OBF-type CI with 90◦ twist, but is easy to understand by the two different HT-type CIs (cisoid and transoid), which are not in common for forward and backward reaction. Each of them is at only one third (or less in longer polyenes) on the way from the respective reactants to completion of the isomerization. It was also explained in this way, why in longer polyenes the isomerization yield decreases [10]. Even details can be understood: steric strain can increase the dihedral angle in the HT-type CI of Z,E-cycloocta-1,3-diene, so that it can merge with that of the Z,Z isomer; in fact, in this case the sum of quantum yields for forward and backward reactions is near 1 [46]. The argument is not weakened by special cases, where even an OBF-type CI does not guarantee a CI in common for forward and backward reaction: (a) if the molecules are sufficiently asymmetric, torsion by ±90◦ leads to energetically nonequivalent structures (diastereomers), so that also the CIs do not coincide, such as found in the example of [47]. A minor effect on the quantum yield sum can be expected. (b) In the same example, which is a zwitterion, torsion reorients the large dipole moment. (The change from reactant to product amounts to 30 Debye [47]!) The solvent cannot reorient in the ultrashort time of passage through the CI, so that it will tend to restore the reactant from this branching region. In fact, the quantum yield sum was only 0.42 [47]. The delayed solvent response implies that passage to the ground state surface does not happen through

56

W. Fuß / Journal of Photochemistry and Photobiology A: Chemistry 237 (2012) 53–63

the same point of the intersection space (that includes solvent coordinates). (b) The (cisoid) HT CI mediating cis–trans isomerization for butadienes is geometrically very similar to the CI mediating electrocyclic ring closure to cyclobutenes (as well as the backward reaction) [36,48,49]. It was shown by time-resolved spectroscopy that from cyclobutenes not only the latter is reached but also the former [50]. In this way the observed violations of the Woodward–Hoffmann rules in cyclobutene ring opening could be explained in all details [50], even the selective ring opening in anti-Woodward–Hoffmann sense on excitation of Rydberg states or the opposite specificity in some ring-fused derivatives [51]. No similar thing seems conceivable with OBF. (c) Internal conversion of diphenylbutadiene and an analogous compound with the diene integrated in six-membered rings has practically the same efficiency. As already pointed out in [10], this is easy to understand with an HT CI (leading to aborted HT in the ring-fused compound) that requires only minor torsions, whereas with an OBF CI there would be a drastic difference. 2.3. Conclusions from the experimental observations The experimental evidence for general preference of an HTtype CI by nonpolar polyenes demands an improvement of the Levine–Martínez calculations for butadiene [13], that suggested the opposite propensity. (That some slopes are too flat or barriers too high in [13] can also be inferred from the calculated time of >200 fs for departure from the dark state, that is much longer than the measured [52] one of <20 fs.) In the following it is assumed that in butadiene and other nonpolar or mildly polar dienes the molecule crosses over from 1B to 2A long before the double-bond torsion reaches 90◦ (consistent with [13]) and that the 2A polarity borrowed from 1B in this geometry is smaller than calculated for 90◦ in [13]. It should be mentioned that the 1B and 2A states in dienes including diphenylbutadienes (probably also in stilbenes) are energetically not far apart in the minimum region; the sequence is geometry and solvent dependent. One could thus also envisage a direct decay from the ionic state, with OBF isomerization from an allyl methylene geometry, similar as calculated in [13]. On the other hand, Olivucci et al. found that at this geometry the ground state lies too far apart to allow a fast decay [53]. As the experiments (e.g., the small isomerization quantum yields) point to passage through an HT CI, the possibility of direct 1B → S0 relaxation is ignored here. The preference for the HT CI in the free molecules also indicates that the entrance of this channel involves a smaller (free-energy) barrier than that for an OBF CI. (The path in some cases is even barrierless, as indicated by the ultrashort dark-state lifetimes in butadiene [52,54] and other dienes such as those in [50,55].) That is, Liu’s assumption that in free molecules HT is a higher-energy path does not apply to its beginning. But as pointed out above and in [10], there is no energetic preference for HT after leaving the HT CI, and on the lower cone the potential can be influenced by environmental constraints. In fact, Squillacote reports on an example (2,3-dimethylbutadiene-1,4-d2 ) of OBF in a matrix [43], as already mentioned in Section 2.2; it probably represents attempted (aborted) HT with passage through the HT CI and a change of direction thereafter, induced by the restricted free volume. (In this example, OBF is less volume demanding than HT.) But in contrast to OBF, completion of HT isomerization seems not to need any environmental constraint, as demonstrated by the examples of stilbene [44] and hexatriene [45] in the gas phase (Section 2.2); in the absence of other forces, below the CI the molecules seem to be guided by a momentum acquired above the CI. The hesitation to accept HT as the standard mechanism for isomerization of nonpolar polyenes seems to be largely due to a

misinterpretation of the principle of nonequilibration of excited rotamers (NEER) of Jacobs and Havinga [56]. Such doubts were expressed in [57], for instance. This very successful principle [56–59] forbids conformer equilibration in S1 but does not rule out C C single bond rotation by 180◦ on the way from S1 to S0 , as already pointed out in the original work [56] and later with examples of conformer isomerizations by Brouwer and Jacobs [60]. (Further examples are compiled in chapter 4B of [22].) The principle just implies that the conformers have a separate photochemistry. In harmony with this is, for example, the Hula-twist isomerization observed with previtamin D, in which the cZc conformer gave the tEc tachysterol whereas the tZc conformer produced cEc tachysterol1 [3]. Similarly the examples reported in the reviews [4–8] show a separate photochemistry for different conformers (where investigated) and are thus no exception to the NEER principle, as already pointed out in [9,10,36]. A single-step rotation by 180◦ (such as in HT) does not imply equilibration. The misunderstanding was caused by the rationalization of the NEER principle that points to the increase of double-bond character of the original single bonds in the excited state, slowing down their rotation. But for example in polyenes with planar excited state (e.g. octatetraene) diabatic isomerization (which in HT involves concerted twist of both, a double and a single bond, and in OBF only a double bond) does not take place in the excited states (in contrast to adiabatic isomerization, which is much rarer) but only on leaving from it down to the ground state; along this path there is a strong change of the electronic properties and the character of the bonds involved. 2.4. Correlation diagrams and formation of three-membered rings, dis- and conrotatory HT In view of the 90◦ twist in the excited state of monoolefins, the quantum chemical prediction of concerted rotation of two adjacent CC bonds came as a surprise. A simple correlation diagram can provide further insight. The numerical prediction of a kink in the CI of a polyene chain with decreased CCC angle (near 90◦ ) was interpreted in [10] by an attractive 1–3 interaction initially in planar geometry, which then further strengthens (without causing additional strain in the backbone) by disrotatory twist (Scheme 1) of the ␲ orbitals attached to the C(1) and C(3) groups. If the approaching of C(1) and C(3) were continued, a derivative of a cyclopropyl radical would be formed. The orbital correlation diagram (Scheme 2a) shows that this path has energetically a chance from an excited state of butadiene only if ␲4 (LUMO + 1) is involved on the path. Probably this is a also reason (though perhaps not the most important one), why a CASSCF(4/3) calculation (excluding the ␲4 orbital) as that in [13] disfavors the HT CI. Remarkably, in correlation diagrams for Woodward–Hoffmann- (WH-) allowed photochemical processes the LUMO is usually assumed more important that LUMO + 1. With the LUMO, formation of a 1–3 bond requires a conrotatory twist: connect a lower lobe with an upper one in butadiene such as in Scheme 2b. It is here suggested that such a conrotatory form of HT isomerization also exists, although it probably only happens,

1 The conclusions of [3] were questioned in [57], because measured spectra, details of the evaluation procedure and the conversions were not given in [3]. These details can be found in the diploma thesis [61], which is also available from the library [email protected] or from the present author. Two-photon effects (also mentioned in [57]) were avoided by a very low energy density (25 ␮J cm−2 , that is 0.1% of the saturation energy density) of the laser pulses. The conclusion on HT profited from the fact that the spectra of the two tachysterol conformers were known, and the crucial spectrum of the stable conformer was already clearly visible at a dehydrocholesterol conversion of 15%. Also at higher conversions the measured spectra could be simulated by linear combination of the three components with coefficients resulting from the derived kinetics and residual spectra everywhere below 1% of the maximum.

W. Fuß / Journal of Photochemistry and Photobiology A: Chemistry 237 (2012) 53–63

(a)

2.

2

0°

60°

57

(b)

.

.

.

73° conrotation of groups 1 and 3

rotation of CH(2)

Scheme 2. Orbital correlation of butadiene with those of cyclopropyl-methylene diradical, related to dis- and conrotatory Hula twist. In (a), the transformation is made by HT-type rotation of the CH(2) group around a 1–3 axis (disrotatory twist of groups 1 and 3, see Scheme 1) allowing two p orbitals to interact with their lobes (bow in the formula). Bonding and antibonding 1–3 interactions are indicated by the bows on the right-hand side. The angle 60◦ indicates the approximate position of the CI, whereas 73◦ is the complement of the dihedral angle between the ring plane of cyclopropane and a CH bond. A butadiene with a trans (cis) substituent at position 1 would correlate with a cyclopropane species with this substituent (at C(1) ) in trans (cis) to the CH2 radical group. In (b), the three-membered ring is generated by conrotatory twist of the groups 1 and 3. (Anti)bonding interactions are indicated by the curved lines.

if the molecules are pre-twisted in this direction (see below and Section 3). This mechanism is also volume saving, as the same type of intermediate structures are formed as in the normal (disrotatory) variant, with similar motions of the attached groups. Even the final products (conformers) are the same. This is easy to see: the result is not different, if during torsion of a C C bond the attached C C bond rotates by +180◦ or −180◦ (because the difference would be 360◦ ). Also the quantum–chemical calculations predict that both paths to the S1 /S0 intersection space, dis- and conrotatory, are accessible, although the latter only with a minor activation energy [26,36,62,63]; the other calculations consider the disrotatory case only. It is also predicted that the conrotatory path is preferred, if there is a pre-twist in this direction such as in butadiene substituted in positions 2 and 3 with two bulky groups [63], consistent with the effect of pre-distortion discussed in Sections 3 and 5. The barrier in the conrotatory case is easy to interpret: whereas in the disrotatory path an attractive 1–3 interaction can develop already in planar geometry, the opposite case needs first some (conrotatory) torsion of two partial ␲ bonds before the corresponding two orbital lobes can overlap. As mentioned, there are two separate HT-type CIs, one “cisoid” the other “transoid”. They just differ by the configuration (cis or trans) of two substituents (the CH2 radical resulting from one terminal group of the diene and a rest assumed at the other end of the diene) at the long (weak) bond in the triangular structure, as shown in Scheme 3. This scheme also summarizes, which of the two CIs can be reached by dis- or conrotatory motion from the four diene isomers with cis or trans configuration at one double and the single bond. As said above, according to the calculations the conrotatory paths climb over a barrier before the entrance into the CI funnel (unless the compounds are pre-distorted in this direction). It was predicted that from the s-trans reactants E–Z isomerization and formation of bicyclo[1.1.0]butane take place from the same CI (see below), and from the s-cis conformers in addition ring closure to cyclobutene. However, the latter prediction together with Scheme 3 (slightly extended: put a substituent also at the other double bond) would cause a conflict with the Woodward–Hoffmann rules (which are established for the ring closure): in the scheme, a d-cis and a d-trans isomer result in the same (HT) CI structure, which would form the same substituted

cyclobutene, if the HT CI is also responsible for electrocyclization. But in a time-resolved experiment it was found that the paths for E–Z isomerization and electrocyclic reaction are already separated by a barrier, before reaching the entrance cones of the CIs [50]. That is, the ring-closure CI and the HT CI are in different regions of the intersection space; and from the s-trans conformers the electrocyclic path is probably even not competitive with Z–E isomerization. The barrier separating the Z–E and electrocyclic paths can perhaps be rationalized by the different orbitals involved: ring closure uses disrotatory motion at the positions 1 and 4, which is supported by the LUMO (␲3 in Scheme 2), whereas the standard HT path twists the groups 1 and 3 in disrotatory sense, which is favored by the LUMO + 1. The CI in Scheme 2a and b is located at an intermediate geometry with the CH(2) group rotated by about 60◦ and a weak long 1–3 bond. For completion of the HT isomerization (on the lower cone of the CI), the CH group rotation must reach 180◦ , which requires that the 1–3 bond is again loosened. From the same CI also the methylene-cyclopropyl diradicals of Scheme 2 can be formed [34,36,49,53,62,63], which probably precede the bicyclo[1.1.0]butane products observed in small quantum yields from dienes [22]; similar diradical intermediates are assumed also in the formation of bicyclo[3.1.0]hexanes from

d-ciss-cis d-transs-trans

d-ciss-cis d-transs-trans

dis

.

con

.CH2

dis

con

d-transs-cis d-ciss-trans

cisoid

con con

.

dis

.CH . 2

dis

d-transs-cis d-ciss-trans

transoid

Scheme 3. Dis- and conrotatory paths from the various isomers of a substituted butadiene to the cisoid and transoid CI. In the triangular structure, two H atoms are indicated for clarity by short bonds.

58

W. Fuß / Journal of Photochemistry and Photobiology A: Chemistry 237 (2012) 53–63

trienes [64,65]. The HT products from the dis- and conrotatory paths cannot be distinguished. But information can be retrieved from the stereochemistry of the bicyclic products. In fact, a conrotatory stereoselectivity of the groups in positions in 1 and 3 has been observed (but not universally [66]) in the photochemical formation of bicyclo[1.1.0]butanes from s-trans dienes and bicyclo[3.1.0]hexenes from s-trans trienes (see the reviews [22,64]). The conrotatory 1–3 bond formation (leading through an HT CI) hence exists. Note by the way that this interpretation of the stereochemistry of the bicyclic products does not require a zwitterionic intermediate, that was suggested in [67] for ethylidene–cyclooctene to explain the conrotatory stereochemistry; this reference is often quoted as the first experimental evidence for such intermediates. It is interesting that (at least one conformer of) trans-ethylidene–cycloctene is slightly distorted in the diene part to conrotatory direction, according to molecular mechanics. It would be desirable to check for pre-distortion more systematically and on a higher computational level with this example and the others, where the stereochemistry of the bicyclic products has been investigated (see the reviews [22,64]). cis-Stilbene is clearly pre-twisted in conrotatatory direction. One can therefore assume that it also follows this path towards HT. Evidence, taken in favor of the standard (disrotatory) form of HT in the time-resolved experiment [44], would in the same way be valid for the other form of HT.

3. Regioselective HT isomerization by pre-twist Liu’s group reported a regioselective Hula-twist isomerization (HT-1 versus HT-2) in substituted diphenyl-Z,E-butadienes (1), in which rotation of the CH group closer to a phenyl ring (CH(1) ) is by far preferred over that in ␤ position to it (CH(2) ) [14]. Similarly, also two derivatives methylated in the butadiene part showed HT-1 predominating over HT-2 [68], and the isomerization of unsubstituted diphenylbutadiene was interpreted in the same way [14]. Regioselective HT was also reported for several unsymmetric diarylethenes, in which again the rotating group was the benzylic CH [8]. Electronic or mass effects [14] seem not satisfactory for explanation. It is important to note that HT isomerizations are ultrafast in the shorter-chain polyenes (e.g. butadiene [52], cyclic dienes [55], hexatriene [69], Z-stilbene [70]; survey of lifetimes of alltrans polyenes with different chain length in [11]). If the critical steps in such reactions takes less than about 1 ps, the molecule has no time to explore all energetically accessible regions of the excited-state surfaces. (Section 5 will consider, whether this limit also applies to the compounds 1 and 2, and what happens if the processes are slower.) It was suggested in [51] that in such cases the choice between energetically similar paths can be taken by a pre-distortion in the ground state. The principle of least motion (accordancy principle) and the choice between electrocyclic ring opening (cyclohexadienes → hexatrienes) versus ring closure (cyclohexadienes → bicyclobutenes) were interpreted in this way, as well as other effects including environmental constraints or wavelength effects. Fig. 1a explains the principle: if a molecule is pre-distorted (for example, for steric reasons or by environmental constraints) in a direction that is also photochemically or Franck–Condon active, vertical excitation can lead to one of two similar down-slopes. Thus the pre-distortion is amplified, so that the wave packet can be selectively accelerated to one of two conical intersections that would energetically both be accessible. In the given case the initial acceleration (OBF type on the spectroscopic surface!) is towards C(1) C(2) torsion, which together with the phenyl–C(1) pre-twist points to the

(a)

(b) C2 C3 twist HT2 (con)

C1=C2 twist phenyl C1 twist HT1 (con)

4 =C 3 C

C1=C2 twist

ist tw

C1=C2 twist

Fig. 1. (a) Acceleration in the FC region of the excited state towards the direction of ground-state pre-distortion in a substituted diene such as 1: A steep down-slope is met in the excited state at a large torsion (C1 C2) but a nearly horizontal one with small twist (C3 C4). This selective amplification effect can be expected, if the coordinate of pre-distortion is antisymmetric (hence has slope 0 at vanishing excursion) and is Franck–Condon or photochemically active and the process is ultrafast, so that conservation of momentum (direction) is important. (b) On C C twisting (horizontal arrows) in the excited state, the resulting displacement is closer to HT-1 than to HT-2 (i.e. to the entrances of the corresponding CIs), because the initial phenyl–C torsion is much larger than that of the other single bond. In the example, the HT motion is conrotatory, because the initial twist is so.

HT-1 CI (Fig. 1b). The pre-twist of the other adjacent single bond (Scheme 4) is too small to give HT-2 a chance. HT-1 was in fact observed exclusively [14] (Scheme 5). Because in the examples the C C and C C bonds are twisted in the same sense, it is a conrotatory motion towards the HT CI. Similarly HT-1 was found in cis-cinnamic esters (2) and related compounds (OBF was excluded by analogy) [15]. Calculations in [15] showed that the phenyl ring is twisted out of the alkene plane by 21◦ (similar as that given in Scheme 4). One may thus

2 1

HT-1

2

1

OBF 2 HT-2

1

1

1 2

F 1 2 COOR

HT-1

F COOR

2 Scheme 4. Pre-twist induced regioselectivity in HT photoisomerization of Z,Ediphenylbutadienes (1) and Z-cinnamates (2). Molecular mechanics with MM+ force field predicts the following torsions in 1: phenyl C(1) 15◦ , C(1) C(2) 12◦ , C(2) C(3) 2◦ , the other C C and phenyl–C bonds are practically not twisted; in 2: phenyl C(1) 16◦ , C(1) C(2) 9◦ , C(2) C(3) 1.4◦ . With the AMBER force field, the torsions would be about twice larger.

W. Fuß / Journal of Photochemistry and Photobiology A: Chemistry 237 (2012) 53–63

FH2C

C D

3

E,Ereactant FC (1Bu)

CH3

4

Scheme 5. Regioselectivity controlled by polarity: E–Z isomerization predominates at the double bond that has less substituents or has electronegative groups.

consider isomerization also of these compounds as examples of control by pre-distortion, provided that the reaction is ultrafast. A time-resolved study seems to be available only on trans isomers of cinnamic esters and related compounds [71]. In solution this work [71] finds a sub-ps time and a subsequent relaxation within about 3–5 ps; the former is ascribed to relaxation from the spectroscopic state to S1 and the latter to trans → cis isomerization from S1 [71]. While this would be too slow to apply our model, the reaction starting from the cis isomer is expected to be faster by an order of magnitude or more due to a smaller or vanishing barrier, as in stilbene and many other examples (see Section 5). It seems worth to add a side observation in this time-resolved experiment: the authors of [71] were astonished, why the phenyl ester of cinnamic acid was not slower (it was even slightly faster) than the methyl ester or the acid, considering the former’s larger volume demand and larger moment of inertia on rotation around the C C bond. The authors only considered OBF. With Hula twist the mass of the ester group would play no role. Hence this observation might be another support that HT applies also to molecules in solution. Two refinements over Fig. 1 should still be introduced, although they do not change the principle: (a) amplification of the distortion takes place in the spectroscopic state, e.g., in or near the Franck–Condon region. But the molecule can then enter with the distorted geometry to the dark state, and the entrance location and momentum of the wave packet can favor reaching the final CI, if all this happens in ultrafast time (see Section 5), so that the molecule has no time to explore all the S1 surface. In this case, the choice for the final CI is made on a preceding electronic surface. (b) The distortion amplification may happen with delay, only after the Franck–Condon region. Thus in the given case, CC bond length changes probably precede any twist (see e.g. the calculation on butadiene [13].). According to [10,11], also planar (angular) distortions precede Hula twist. 4. Control of Hula twist by electrostatic effects Effects of polar substituents on regioselectivity of cis–trans isomerization are well known (for reviews see chapter 4a of [22] and pp. 132–133 of [59]). Already the pioneering example, 1,3pentadiene-1d (3), was interpreted by the Squillacote group as mediated by an allyl-cation methylene-anion structure of the excited state, in which the anionic part performs the rotation [18]. The allyl methylene model is now in common use, although the polarization had to be assumed with different directions [22,59]. The mechanism (OBF or HT) was not examined in the early cases and in some recent ones [19–21]. In other examples HT was either assumed (in 1-fluoro-2,4-hexadiene (4) in solution [17]) or made probable in cinnamic esters (2) [15]. (The latter were considered above as a case of control by pre-twist of the phenyl group; but the polarity considerations below would lead to the same result, HT-1.) Above, it was argued that HT requires a path via a covalent dark (2A) state, which has only minor polarity (e.g. by mixing with the 1B state in a distorted molecule), if at all. So, how can polar effects take over the control? The clue is distortion in the initially excited spectroscopic (1B) state, which has ionic character. If electron donating or accepting substituents favor formation of an allyl-methylene zwitterion, partial (OBF-like) twist of the double bond at the

CC stretch

H

59

seam 1Bu 2Ag HT1 HT2 Z,Eproduct

HT1, HT3 HT2, HT4

HT3 HT4 E,Zproduct

<90°

<90°

C1-C2 twist

C3-C4 twist

Fig. 2. The 1B and 2A surfaces (view from top) versus CC bond length alternation and two CC twist coordinates, (conically) intersecting at the seam. The gradient difference vector of the CI is dominated by a CC stretch, which is also Franck–Condon (FC) active; the nonadiabatic coupling vector is mainly a planar CCC bending [11]. The wave packet is initially accelerated along the FC active coordinate and can later be deflected to one of two side valleys involving some C C twist (of OBF type, but with <90◦ twist) and some zwitterionic polarization. This deflection is the pre-twist that leads the wave packet – after entering to the 2A surface (that does not have to support any polarization) – to the 2A/1A CI associated with isomerization of one or the other double bond. Due to a tendency for momentum conservation over ultrashort times, straight paths (on the ridge between the valleys) are also possible; they exhibit no regioselectivity.

methylene part can cause the pre-distortion that then on relaxing to the dark covalent state can lead to the HT-type CI leading to Hula twist near the acceptor substituent. However, for this suggestion to work one has to assume that also departure from the dark state is ultrafast (see Section 5). Using the example of 1-fluoro-2,4-hexadiene, Squillacote already formulated such a mechanism [17]: regioselectivity in this case “is determined . . . by the relative efficiencies of the entryways leading to two conical intersections” between the dark state and S0 . The present work supplements: “provided that all steps in the process are fast enough (shorter than ≈1 ps) that there is no time for the molecule to explore all energetically accessible regions of excitedstate surfaces”. It also points to the similarity with pre-distortion for steric reasons (Section 3), which seems to be widespread also in other photochemical rearrangements [51]. Squillacote further explains [17] that the excited-state lifetime is too short for addition of a nucleophilic solvent such as methanol and even for reorientation of a polar solvent. The latter explains unexpected solvent effects. Similar considerations apply to alkenes [72]. It remains only unexplained, why the more polar 4 showed a smaller regioselectivity than 3. An interpretation is possible on the basis of a more detailed relaxation mechanism (Fig. 2): the wave packets initially travel down from the Franck–Condon (FC) region of the spectroscopic (1Bu ) surface along the FC active coordinates (mainly CC bond length alternation, see e.g. [11,13]). Some will continue on this way and in this planar geometry cross over to the dark (2Ag ) surface, with some excursion along the nonadiabatic coupling vector, a planar CCC bend [11]. Another part of them can fall into one of two side valleys involving C C twist, which gradually develop while advancing on the FC coordinate. Whereas the pretwist on these two paths favors the isomerizations as indicated, the

60

W. Fuß / Journal of Photochemistry and Photobiology A: Chemistry 237 (2012) 53–63

undeflected wave packet still has the choice between all S1 /S0 CIs (of HT type, as assumed). Hence the product ratio not only depends on the choice between the two valleys but also on the fraction choosing the central path due to a tendency for momentum conservation. The central path lowers the regioselectivity. It is then suggestive to assume that the CHD rotation in 3 is faster and more efficiently accelerated to a side valley than that of the heavier CHCH2 F group of 4. In this case, the regioselectivity difference between 3 and 4 is not only controlled by electronic effects (shape and depth of the left side valley in Fig. 2, which would suggest a higher selectivity for 4, in contrast to observation) but also by the moment of inertia of the rotating group (following an OBF path on the 1 B surface). If so, one must also admit that within one compound the choice between the two valleys is also controlled by both effects. (An early suggestion of control by the substituent mass alone is disproved by the example of 4 [17].) Mass effects can generally be expected in steps of a process, which are ultrafast, so that momentum conservation plays a role. It was assumed for this mechanism that the polarity is not too strong. Otherwise a pure OBF-type CI may become more easily accessible. This can happen either, if (with suitable distortion such as a strong C C twist) the dark state can strongly mix with the spectroscopic state and thus acquire zwitterionic character (as in the calculation [13]), or if even the spectroscopic state (that prefers OBF) is lowered to below the dark state such as in the protonated Schiff base (PSB) of retinal [35]. With the criteria for lifetime and little strength of polarization, we can probably apply the mechanism of this section to dienes and trienes with alkyl, fluoroalkyl or fluoro substituents. If the conjugated chain is lengthened (to octatetraene or by attaching one or two phenyl or carboxyl groups to dienes or trienes), the lifetime of the dark state becomes too long in the all-trans configuration. However, if there is at least one (cis) double bond with substituents that introduce enough pre-twist (by steric hindrance) to eliminate a torsional barrier on the 2A surface (compare the example of cis- and trans-stilbene), the lifetime can again come to the range ≤1 ps. In Section 3, the diphenyl-Z,E-butadienes and Z-cinnamates (Scheme 4) were considered to be of this kind.

5. Slower processes: control by activation energies in S1 If the departure from S1 takes several picoseconds or even longer, due to the presence of barriers, in solution vibrational relaxation can already take place and alter any path, which will then no longer be controlled by momentum conservation and anything that happens before reaching S1 . In such cases, the competition between different paths, hence the regioselectivity for instance, is controlled by the relative activation energies on the S1 surface. The lifetime of the (dark) S1 state of all-trans polyenes and derivatives is a (nonmonotonic) function of the energetic S1 –S0 gap (compilations can be found in [11,73,74]). The latter decreases on extending the conjugation length. Already the all-trans isomers of octatetraene, diphenyl-butadiene or -hexatriene have much too long lifetimes. Only much beyond carotene the excited polyenes decay again in shorter than 1 ps; but polyenes of such a length do not exhibit photochemical trans–cis isomerization [11]. The long lifetime is caused by a barrier before the entrance into the CI funnel. Phenyl substituents in 1-cis orientation introduce some steric hindrance and thus a pre-twist not only of the phenyl-C single bond but (to a lesser extent) also of the double bond, as known from the example of stilbene. The corresponding isomerization barrier is thereby reduced; in cis-stilbene it nearly vanishes, so that the lifetime becomes ≤1 ps (trans-stilbene about 70 ps in solution) (see e.g. [75]). Fig. 3 explains the principle: without steric

+ steric

S1

only

+ steric

S0

only

steric 0

torsion

Fig. 3. Adding torsional steric hindering (“steric”) to potentials that would result from ␲ interaction alone displaces the S0 minimum (pre-twist) and can cancel a torsional barrier in the spectroscopic and/or the dark state (“S1 ”).

hindering (“␲ only” in the figure) the polyene is assumed in the figure to be planar in S0 (torsion 0) and to have a torsional barrier for leaving the spectroscopic and/or the dark state (“S1 ”). Adding a steric repulsion (lower broken line, “steric”) to these potentials causes a pre-twist in S0 and can cause disappearance of the torsional barrier in S1 . The mechanism works, if the barrier would be expected early, at torsional angles where the steric hindering still plays a role. In diaryl-ethenes, one expects a barrier near the Franck–Condon region; it results from avoided crossing of the primary excited state that mainly affects the aromatic rings, and a higher, ␲-antibonding state coming down on increasing torsion [23]. In the dark state of longer polyenes such as octatetraene there are minor barriers before the entrance to the CI funnel (Section 3), which at least for the HT CI is also at small torsional angles. Reduction of an activation energy implies an acceleration of the process; while this seems evident, steric hindering is normally not associated with increase of speed. Above (Section 3) we supposed (with support by semiempirical calculations, Scheme 4) that the pre-twist is also sufficient in the Z,E-diphenylbutadienes [14] and for the Z-cinnamates [15] of Liu et al., so that the lifetimes are short enough and the momentum mechanism works. If it should turn out, against expectation, that the lifetime is clearly >1 ps, the regioselectivity would be controlled by activation energies. The pre-twist of the phenyl-C and adjacent C C bonds lowers the barrier for isomerization at this location, i.e. hence for the entrance to an HT-1 CI. Hence also in case of control by activation energy instead of momentum, a pre-distortion determines the preferred path. This can probably be generalized to nonpolar or mildly polar polyenes, where the covalent S1 (2A) state is separate from, and not much influenced by, the ionic 1B state. It is the pre-twisted group that rotates in the photochemical process, continuing in the direction of ground-state pre-distortion. By contrast, the influence of any polarity or electrostatic effect to barriers on S1 is much less obvious, because it is only indirect: it primarily affects the more polarizable (ionic) 1B state, which in distorted geometry can mix with the dark state and lend to it some polarizability and affect it energetically. The mixing depends also on the energetic distance, which in turn varies along different coordinates; in some compounds (e.g. diphenylbutadienes) the two states have comparable energies in extended regions of the coordinate space. Furthermore, with stronger polarity the mechanism

W. Fuß / Journal of Photochemistry and Photobiology A: Chemistry 237 (2012) 53–63

can change such that a path via an OBF-type CI is chosen instead of the usual HT-type. Any predictions would therefore require more detailed considerations or direct calculations. On the other hand, there seem to be also some clear-cut cases such as the fluorinated diphenyl-E,E-butadienes of Jin Liu and coworkers [19,20] or the donor–acceptor substituted diphenyl-butadienes of Singh and Mahalaxmi [21]. In these cases the polarity was apparently sufficient to lower the ionic 1B surface to below the 2A potential and cause a pure OBF. This is indicated by the quantum yield: from an OBF CI with its 90◦ twist one can expect yields around 50%, whereas with HT they are at best around 10% [11]. In fact, the mentioned diphenyl-butadienes isomerize with about 50% efficiency [21].

6. Conclusion The most important branching region in photochemical reactions is the last (S1 /S0 ) conical intersection, where often the molecule can not only decide between a way back to the reactant (i.e. internal conversion) and to a product but sometimes also several products [76]. Slopes of the cones can be influenced by substituents, so that quantum yields or product ratios may change; an example was suggested by Olivucci and coworkers: the unusually large quantum yield of 1-cyano-bicyclo[1.1.0]butane from 2-cyano-butadiene, explained by stabilization of the radical (partial) structure in the (HT-type) CI [53]. Also the environment can change the relative slopes, for instance, by hindering motion towards OBF but less so towards HT (Section 2.1 and [10]). An opposite example (deuterated dimethylbutadiene in an Ar matrix), where the environment seems to enforce an OBF path, was presented in Section 2.2 and interpreted as aborted HT (where the wave packet is redirected on the lower cone of the HT CI). But near the CI the wave packet is not only guided by local slopes but also by a momentum acquired on the path before, in particular in the entrance cone, so that valleys in the upper cone are more important than in the lower one (Section 2.1 and [10,53]). The momentum accumulated on approaching an HT-type CI favors, after departure from it, continuation of this direction towards HT isomerization, although on the lower cone there is no energetic preference for this direction. Often there are several CIs (that is, minima of the intersection space), for example, those belonging to isomerization of the different double bonds of a polyene. Regioselectivity happens by a choice between the CIs, hence before entering into one of them (Sections 4 and 5 and [17]). Before such funnels, there can be barriers, in particular for trans–cis isomerizations. Obviously in such cases the different activation energies determine the regioselectivity. Barriers towards isomerization at one site can be lowered by a pre-twist of the corresponding bonds, caused by steric hindering in particular in cis-isomers (Section 5); regioselectivity is hence in favor of the hindered site. External forces can obviously act in the same way [51]. By contrast, it is less easy to overlook the influence of polarity to barrier heights. This is, because it primarily affects the (ionic) 1B state and only by mixing with it also the (covalent) 2A (S1 ) state, with the mixing depending on parameters such as the energy difference and distortion coordinates (Section 5). The control by barrier heights is consistent with the usual picture that the energy landscape of S1 determines the fate of the photochemical reaction. But as pointed out here, this is not the full story: the CIs beyond the barriers are also branching regions. Furthermore, if the S1 lifetime is ultrashort (<1 ps), due to absence of a barrier in some exit channels, such as in short-chain polyenes, a momentum acquired in a preceding excited state (e.g., the spectroscopic 1B state) can dominate the choice between reaction paths (e.g., regioselectivity).

61

This is, because such short times are not sufficient for the molecule to explore the full energetic landscape of S1 , and it is more important towards which exit channel the wave packet is sent by the momentum (Sections 3 and 4). It is also worth noting that 1B → 2A relaxation is fast (50–250 fs [11]) in conjugated polyenes; on transition the molecule gets just an additional momentum towards the nonadiabatic coupling vector, a planar CCC bend of bu symmetry that according to [11] gives the molecule an (additional) push toward the entrance cone of a HT CI. A slight pre-twist of one double bond can be amplified on excitation to the spectroscopic state 1B and, and together with a (strong) pre-twist of one adjacent single bond, direct the molecule towards a regioselective HT direction; completion of the reaction is initiated only when the wave packet arrives on the S1 surface at the entrance of the corresponding HT CI (Section 3). Regioselective pre-twist of a C C bond already in 1B can also be expected (based on the allylmethylene model or the calculation [13]) by electrostatic effects; the wave packet is sent thereby to near the entrance of the corresponding HT and OBF CIs (Section 4). The former will be preferred, because in the short polyenes of Section 4 HT is barrierless, whereas an activation energy is required for OBF, as claimed in Section 2.3. Section 2 summarizes theoretical calculations and considerations and experimental evidence (part of it indirect) that the path to a HT CI is usually preferred over that to an OBF CI, controlled by internal forces. The fact, that also regioselective Hula twist such as HT-1 versus HT-2 can be explained purely by considering potentials of the isolated molecule, also supports this idea. Nevertheless, external forces or constraints (steric interactions) can also interfere: (1) they can reduce or delete a barrier in a specific direction due to a pre-twist (Fig. 3). This is closest to what Liu had in mind, saying that the medium can alter activation energies. (2) They can alter the slope in a specific direction down from the CI. If the change is strong enough to overcome momentum effects, the path can be redirected to give an OBF product or another conformer on leaving from an HT CI (attempted HT, Section 2.1), or the original (HT) path can be consolidated. This mechanism was emphasized in [10]. (3) A pre-twist can be amplified in the Franck–Condon region and, if also the subsequent steps are ultrafast, the molecule can be accelerated to a specific CI (Section 3 and [51]). The variants 1 and 3 can also accelerate the process, a phenomenon not usually associated with steric hindering. But the influence of external and internal forces have in common that they both work by inducing a pre-distortion, which in turn can act in the three variants just mentioned. References [1] R.S.H. Liu, A.E. Asato, Photochemistry of polyenes. 22. The primary process of vision and the structure of bathorhodopsin – a mechanism of photoisomerization of polyenes, Proceedings of the National Academy of Sciences of the United States of America 82 (1985) 259–263. [2] R.S.H. Liu, D.T. Browne, A bioorganic view of the chemistry of vision: HT-n and BP-m,n mechanisms of confined, anchored polyenes, Accounts of Chemical Research 19 (1986) 42–48. [3] A.M. Müller, S. Lochbrunner, W.E. Schmid, W. Fuß, Low-temperature photochemistry of previtamin D: a Hula-twist isomerization of a triene, Angewandte Chemie International Edition 37 (1998) 505–507. [4] R.S.H. Liu, G.S. Hammond, The case of medium-dependent dual mechanisms for photoisomerization: one-bond-flip and Hula-twist, Proceedings of the National Academy of Sciences of the United States of America 97 (2000) 11153–11158. [5] R.S.H. Liu, Photoisomerization by Hula twist: a fundamental supramolecular photochemical reaction, Accounts of Chemical Research 34 (2001) 555–562. [6] R.S.H. Liu, G.S. Hammond, Examples of Hula twist in photochemical cis–trans isomerization, Chemistry – A European Journal 7 (2001) 4536–4544. [7] R.S.H. Liu, in: W.M. Horspool, F. Lenci (Eds.), CRC Handbook of Organic Photochemistry and Photobiology, vol. 2, CRC Press, Boca Raton, 2004, p. 1-11. [8] R.S.H. Liu, L.Y. Yang, Y.P. Zhao, A. Kawanabe, H. Kandori, in: A.G. Griesbeck, M. Oelgemöller, F. Ghetti (Eds.), CRC Handbook of Organic Photochemistry and Photobiology, 3rd edition, CRC Press, Boca Raton, 2012. [9] W. Fuß, S. Lochbrunner, A.M. Müller, T. Schikarski, W.E. Schmid, S.A. Trushin, Pathway approach to ultrafast photochemistry: potential surfaces, conical intersections and isomerizations of small polyenes, Chemical Physics 232 (1998) 161–174.

62

W. Fuß / Journal of Photochemistry and Photobiology A: Chemistry 237 (2012) 53–63

[10] D. Sampedro Ruiz, A. Cembran, M. Garavelli, M. Olivucci, W. Fuß, Structure of the conical intersections driving the cis–trans photoisomerization of conjugated molecules, Photochemistry and Photobiology 76 (2002) 622–633. [11] W. Fuß, Y. Haas, S. Zilberg, Twin states and conical intersections in linear polyenes, Chemical Physics 259 (2000) 273–295. [12] B.S. Hudson, B.E. Kohler, K. Schulten, in: E.C. Lim (Ed.), Excited States, vol. 6, Academic Press, New York, 1982, pp. 1–99. [13] B.G. Levine, T.J. Martínez, Ab initio multiple spawning dynamics of excited butadienes: role of charge transfer, Journal of Physical Chemistry A 113 (2009) 12815–12824. [14] L.Y. Yang, R.S.H. Liu, K.J. Boarman, N.L. Wendt, J. Liu, New aspects of diphenylbutadiene photochemistry. Regiospecific Hula-twist photoisomerization, Journal of the American Chemical Society 127 (2005) 2404–2405. [15] S. Schieffer, J. Pescatore, R. Ulsh, R.S.H. Liu, Regioselective Hula-twist photoisomerization of cinnamate esters in organic glass, Chemical Communications (2004) 2680–2681. [16] L.Y. Yang, R.S.H. Liu, Mechanism of photoisomerization of 1-naphthyl-2phenylethylenes in organic glasses, Photochemistry and Photobiology 83 (2007) 1436–1440. [17] M. Squillacote, J. Wang, J.W. Chen, Electrostatic control of the regioselectivity in the photoisomerization of trans,trans-1-fluoro-2,4-hexadiene: evidence for competing conical intersections, Journal of the American Chemical Society 126 (2004) 1940–1941. [18] M. Squillacote, T. Semple, Polarization in the excited state of 1,3-pentadiene: experimental evidence for an allyl cation-methylene anion species, Journal of the American Chemical Society 109 (1987) 892–894. [19] J. Liu, K.J. Boarman, N.L. Wendt, L.M. Cardenas, Regiospecific photoisomerization of fluorinated (E,E)-1,4-diphenyl-1,3-butadienes, Tetrahedron Letters 46 (2005) 4953–4956. [20] J. Liu, E.L. Suits, K.J. Boarman, Regioselectivity of E/Z photoisomerization of fluorinated cisoid (1E,3E)-1,4-diphenylbutadienes via direct irradiation, Tetrahedron Letters 44 (2003) 8103–8107. [21] A.K. Singh, G.R. Mahalaxmi, Excited state properties of ␣,␻-diphenylpolyenes: photophysical and photochemical studies of donor–acceptor diarylbutadienes, Photochemistry and Photobiology 71 (2000) 387–396. [22] B.H.O. Cook, W.J. Leigh, in: Z. Rappoport (Ed.), The Chemistry of Dienes and Polyenes, vol. 2, Wiley, New York, 2000, pp. 197–255. [23] M. Klessinger, J. Michl, Excited States and Photochemistry of Organic Molecules, VCH, New York, 1995. [24] B.E. Kohler, Octatetraene photoisomerization, Chemical Reviews 93 (1993) 41–54. [25] N.J. Turro, V. Ramamurthy, J.C. Scaiano, Modern Molecular Photochemistry of Organic Molecules, University Science Books, Sausalito, CA, USA, 2010. [26] P. Celani, F. Bernardi, M. Olivucci, M.A. Robb, Excited-state reaction pathways for s-cis buta-1,3-diene, Journal of Chemical Physics 102 (1995) 5733–5742. [27] M. Garavelli, P. Celani, F. Bernardi, M.A. Robb, M. Olivucci, Force fields for ultrafast photochemistry: the S2 (1 Bu ) to S1 (2 Ag ) to S0 (1 Ag ) reaction path for alltrans hexa-1,3,5-triene, Journal of the American Chemical Society 119 (1997) 11487–11494. [28] M. Olivucci, F. Bernardi, P. Celani, I. Ragazos, M.A. Robb, Excited-state cis–trans isomerization of cis-hexatriene. A CAS-SCF computational study, Journal of the American Chemical Society 116 (1994) 1077–1085. [29] M. Garavelli, P. Celani, N. Yamamoto, F. Bernardi, M.A. Robb, M. Olivucci, The structure of the nonadiabatic photochemical trans–cis isomerization channel in all-trans octatetraene, Journal of the American Chemical Society 118 (1996) 11656–11657. [30] M. Garavelli, B.R. Smith, M.J. Bearpark, F. Bernardi, M. Olivucci, M.A. Robb, Relaxation paths and dynamics of photoexcited polyene chains: evidence for creation and annihilation of neutral soliton pairs, Journal of the American Chemical Society 122 (2000) 5568–5581. [31] M.J. Bearpark, M. Olivucci, S. Wilsey, F. Bernardi, M.A. Robb, An MC-SCF study of styrene singlet-state photoisomerization, Journal of the American Chemical Society 117 (1995) 6944–6953. [32] M.J. Bearpark, F. Bernardi, S. Clifford, M. Olivucci, M.A. Robb, T. Vreven, Cooperating rings in cis-stilbene lead to an S0 /S1 conical intersection, Journal of Physical Chemistry A 101 (1997) 3841–3845. [33] P. Celani, M. Garavelli, S. Ottani, F. Bernardi, M.A. Robb, M. Olivucci, Molecular trigger for radiationless deactivation of photoexcited conjugated hydrocarbons, Journal of the American Chemical Society 117 (1995) 11584–11585. [34] F. Sicilia, L. Blancafort, M.J. Bearpark, M.A. Robb, Quadratic description of conical intersections: characterization of critical points on the extended seam, Journal of Physical Chemistry A 111 (2007) 2182–2192. [35] A. Migani, M. Olivucci, in: W. Domcke, D.R. Yarkony, H. Köppel (Eds.), Conical Intersections: Electronic Structure, Dynamics and Spectroscopy, vol. 15, World Scientific, Singapore, 2004, pp. 271–320. [36] M. Olivucci, M.A. Robb, F. Bernardi, in: J. Waluk (Ed.), Conformational Analysis of Molecules in Excited States, Wiley-VCH, New York, 2000, pp. 297–366. [37] E. Tapavicza, A. Meyer, F. Furche, Unravelling the details of vitamin D photosynthesis by non-adiabatic molecular dynamics simulations, Physical Chemistry Chemical Physics 12 (2011) 20986–20998. [38] M. Sumita, K. Saito, Theoretical study on hula-twist motion of penta-2,4dieniminium on the S1 surface under isolated condition by the complete active space self-consistent field theory, Chemical Physics Letters 424 (2006) 374–378.

[39] S. Olsen, R.H. McKenzie, Conical intersections, charge localization, and photoisomerization pathway selection in a minimal model of a degenerate monomethine dye, Journal of Chemical Physics 131 (2009) 234306. [40] S. Wilsey, K.N. Houk, H/vinyl conical intersections for dienes: a mechanism for the photochemical Hula-twist, Photochemistry and Photobiology 76 (2002) 616–621. [41] J.E. Norton, K.N. Houk, H/vinyl conical intersections of hexatrienes related to the hula-twist photoisomerization, Molecular Physics 104 (2006) 993–1008. [42] H. Petek, A.J. Bell, Y.S. Choi, K. Yoshihara, B.A. Tounge, R.L. Christensen, The 2 1 Ag state of trans,trans-1,3,5,7-octatetraene in free-jet expansions, Journal of Chemical Physics 98 (1993) 3777–3794. [43] M. Squillacote, T.C. Semple, Photochemistry of s-cis acyclic 1,3-dienes, Journal of the American Chemical Society 112 (1990) 5546–5551. [44] W. Fuß, C. Kosmidis, W.E. Schmid, S.A. Trushin, The photochemical cis–trans isomerization of free stilbene molecules follows a Hula-twist pathway, Angewandte Chemie International Edition 43 (2004) 4178–4182. [45] Y. Furukawa, H. Takeuchi, I. Harada, M. Tasumi, Matrix-isolation infrared and ultraviolet spectroscopic studies of less stable conformers of 1,3,5-hexatriene, Journal of Molecular Structure 100 (1983) 341. [46] J. Saltiel, J.L. Charlton, in: P. de Mayo (Ed.), Rearrangements in Ground and Excited States, vol. 3, Academic Press, 1980, pp. 25–89. [47] A. Melloni, R. Rossi Paccani, D. Donati, V. Zanirato, A. Sinicropi, M.L. Parisi, E. Martin, M. Ryazantsev, W.J. Ding, L.M. Frutos, R. Basosi, S. Fusi, L. Latterini, N. Ferré, M. Olivucci, Modeling, preparation, and characterization of a dipole moment switch driven by Z/E photoisomerization, Journal of the American Chemical Society 132 (2010) 9310–9319. [48] F. Bernardi, M. Olivucci, I.N. Ragazos, M.A. Robb, Origin of the nonstereospecificity in the ring opening of alkyl-substituted cyclobutenes, Journal of the American Chemical Society 114 (1992) 2752–2754. [49] M. Garavelli, F. Bernardi, M. Olivucci, M.J. Bearpark, S. Klein, M.A. Robb, Product distribution in the photolysis of s-cis butadiene: a dynamics simulation, Journal of Physical Chemistry 105 (2001) 11496–11504. [50] W. Fuß, W.E. Schmid, S.A. Trushin, P.S. Billone, W.J. Leigh, Forward and backward pericyclic photochemical reactions have intermediates in common, yet cyclobutenes break the rules, ChemPhysChem 8 (2007) 592–598. [51] W. Fuß, in: T.E. Simos, G. Maroulis (Eds.), Proc. of the International Conference on Computational Methods in Science and Engineering, Corfu, Greece, American Institute of Physics Conf. Proc., vol. 963 2A, 2007, pp. 627–630. [52] W. Fuß, W.E. Schmid, S.A. Trushin, Ultrafast electronic relaxation of s-trans butadiene, Chemical Physics Letters 342 (2001) 91–98. [53] M. Olivucci, F. Bernardi, S. Ottani, M.A. Robb, Substituent effects in buta1,3-diene photochemistry: a CAS-SCF study of 2,3-dimethylbutadiene and 2-cyanobutadiene excited state reaction paths, Journal of the American Chemical Society 116 (1994) 2034–2048. [54] F. Assenmacher, M. Gutmann, G. Hohlneicher, V. Stert, W. Radloff, Ultrafast dynamics of the 1 1Bu-state of 1,3-butadiene after excitation at 204 nm, Physical Chemistry Chemical Physics 3 (2001) 2981–2982. [55] W. Fuß, S. Panja, W.E. Schmid, S.A. Trushin, Competing ultrafast cis–trans isomerisation and ring closure of cyclohepta-1,3-diene and cyclo-octa-1,3-diene, Molecular Physics 104 (2006) 1133–1143. [56] H.J.C. Jacobs, E. Havinga, Photochemistry of vitamin D and its isomers and of simple trienes, Advances in Photochemistry 11 (1979) 305–373. [57] J. Saltiel, L. Cires, A.M. Turek, in: W.M. Horspool, F. Lenci (Eds.), CRC Handbook of Organic Photochemistry and Photobiology, vol. 2, CRC Press, Boca Raton, USA, 2004, pp. 27/1–27/27. [58] W.H. Laarhoven, H.J.C. Jacobs, in: W.M. Horspool, P.-S. Song (Eds.), CRC Handbook of Organic Photochemistry and Photobiology, CRC Press, Boca Raton, 1995, pp. 143–154. [59] W.H. Laarhoven, in: A. Padwa (Ed.), Organic Photochemistry, vol. 9, Marcel Dekker, New York, 1987, pp. 129–224. [60] A.M. Brouwer, H.J.C. Jacobs, Photochemistry of 2,5-dimethyl-1,3,5-hexatrienes in argon matrices. Formation of isomers and rotamers, Recueil des Travaux Chimiques des Pays-Bas 114 (1995) 449–458. [61] A.M. Müller, Diplomarbeit, Technische Universität München, 1997. [62] M. Olivucci, I.N. Ragazos, F. Bernardi, M.A. Robb, A conical intersection mechanism for the photochemistry of butadiene. A MC-SCF study, Journal of the American Chemical Society 115 (1993) 3710–3721. [63] M. Garavelli, B. Frabboni, M. Fato, P. Celani, F. Bernardi, M.A. Robb, M. Olivucci, Photochemistry of highly alkylated dienes: computational evidence for a concerted formation of bicyclobutane, Journal of the American Chemical Society 121 (1999) 1537–1545. [64] W.G. Dauben, E.L. McInnis, D.M. Michno, in: P. de Mayo (Ed.), Rearrangements in Ground and Excited States, vol. 3, Academic Press, Inc., 1980, pp. 91–129. [65] M. Garavelli, P. Celani, M. Fato, M.J. Bearpark, B.R. Smith, M. Olivucci, M.A. Robb, Relaxation path from a conical intersection: the mechanism of product formation in the cyclohexadiene/hexatriene photochemical interconversion, Journal of Physical Chemistry A 101 (1997) 2023–2032. [66] K. Tanaka, K. Fukui, A stereoselective rule for the cross-bicyclization in linear conjugated polyenes, Bulletin of the Chemical Society of Japan 51 (1978) 2209–2214. [67] W.G. Dauben, J.S. Ritscher, Photochemistry of ethylidenecyclooctenes. Mechanism of bicyclobutane formation, Journal of the American Chemical Society 92 (1970) 2926–2927. [68] L.Y. Yang, R.S.H. Liu, Steric effects in Hula-twist photoisomerization. 1,4Dimethyl- and 2,3-dimethyl-1,4-diphenylbutadienes, Journal of the American Chemical Society 127 (2005) 9378–9379.

W. Fuß / Journal of Photochemistry and Photobiology A: Chemistry 237 (2012) 53–63 [69] W. Fuß, T. Schikarski, W.E. Schmid, S. Trushin, K.L. Kompa, P. Hering, Spectroscopy and ultrafast dynamics of the 2A1 state of Z-hexatriene in gas phase, Journal of Chemical Physics 106 (1997) 2205–2211. [70] W. Fuß, C. Kosmidis, W.E. Schmid, S.A. Trushin, The lifetime of the perpendicular minimum of cis-stilbene observed by dissociative intense-laser field ionization, Chemical Physics Letters 385 (2004) 423–430. [71] J.L. Pérez Lustres, V.M. Farztdinov, S.A. Kovalenko, Femtosecond S3–S1 conversion and structural reorganisation of trans-3-phenylprop-2-enaldehyde and derivatives in solution, ChemPhysChem 6 (2005) 1590–1599. [72] K. Kosma, S.A. Trushin, W. Fuß, W.E. Schmid, Ultrafast dynamics and coherent oscillations in ethylene and ethylene-d4 excited at 162 nm, Journal of Physical Chemistry A 112 (2008) 7514–7529.

63

[73] P.O. Andersson, S.M. Bachilo, R.-L. Chen, T. Gillbro, Solvent and temperature effects on dual fluorescence in a series of carotenes. Energy gap dependence of the internal conversion rate, Journal of Physical Chemistry 99 (1995) 16199–16209. [74] P.O. Andersson, T.M. Gillbro, Photophysics and dynamics of the lowest excited singlet state in long substituted polyenes with implications to the very longchain limit, Journal of Chemical Physics 103 (1995) 2509–2519. [75] S.A. Kovalenko, A.L. Dobryakov, I. Ioffe, N.P. Ernsting, Evidence for the phantom state in photoinduced cis–trans isomerization of stilbene, Chemical Physics Letters 493 (2010) 55–258. [76] F. Bernardi, M. Olivucci, M.A. Robb, Potential energy surface crossings in organic photochemistry, Chemical Society Reviews 25 (1996) 321–328.