Photoinduced Changes in Aromaticity Facilitate Electrocyclization of Dithienylbenzene Switches.

The concepts of excited-state aromaticity and antiaromaticity have in recent years with increasing frequency been invoked to rationalize the photochemistry of cyclic conjugated organic compounds, with the long-term goal of using these concepts to improve the reactivities of such compounds toward different photochemical transformations. In this regard, it is of particular interest to assess how the presence of a benzene motif affects photochemical reactivity, as benzene is well-known to completely change its aromatic character in its lowest excited states. Here, we investigate how a benzene motif influences the photoinduced electrocyclization of dithienylethenes, a major class of molecular switches. Specifically, we report on the synthesis of a dithienylbenzene switch where the typical nonaromatic, ethene-like motif bridging the two thienyl units is replaced by a benzene motif, and show that this compound undergoes electrocyclization upon irradiation with UV-light. Furthermore, through a detailed quantum chemical analysis, we demonstrate that the electrocyclization is driven jointly and synergistically by the loss of aromaticity in this motif from the formation of a reactive, antiaromatic excited state during the initial photoexcitation, and by the subsequent relief of this antiaromaticity as the reaction progresses from the Franck-Condon region. Overall, we conclude that photoinduced changes in aromaticity facilitate the electrocyclization of dithienylbenzene switches.

Introduction

The concept of aromaticity is widely used in organic chemistry to predict and rationalize structures and reactivities of cyclic conjugated molecules residing in their electronic ground state. It was first applied to electronic excited states by Dewar[1] and Zimmerman[2] in the 1960s to explain the formation of four-membered rings in photochemical electrocyclization reactions. These authors proposed that such reactions proceed through aromatic transition structures, which are also known to mediate thermal electrocyclization,[1−3] sigmatropic rearrangement,[3,4] nonpericyclic ring-closure,[3,5] and carbonyl-olefin metathesis[6] reactions. Subsequently, Baird[7] used semiempirical molecular orbital theory calculations to formulate rules for aromaticity and antiaromaticity in the lowest triplet excited state (T1) of cyclic conjugated hydrocarbons. These rules, which are the reverse of Huckel’s rules for the ground state, assert that 4n + 2 and 4n π-electron systems are respectively antiaromatic and aromatic in T1. As argued by Aihara in 1978,[8] it follows from Baird’s rules that reactions of benzene that are energetically hindered in the ground state due to a loss of aromaticity, are favored in T1 due to a loss of antiaromaticity. However, unlike consequences of Huckel’s rules, this implication of Baird’s rules was not put to immediate test.

Much later, in 2008,[9,10] Karadakov demonstrated using multiconfigurational quantum chemical calculations that Baird’s rules apply also to the lowest singlet excited state (S1) of cyclobutadiene,[9] benzene,[9] and cyclooctatetraene.[10] This triggered a renewed interest in excited-state aromaticity (ESA), both in terms of obtaining spectroscopic evidence for the concept[11−14] and applying it to the design of novel synthetic protocols.[15−17] Furthermore, this trend led to the discovery of ways of tuning ESA through steric and electronic substituent effects[18] and to the development of strategies for using ESA to modulate double-bond photoisomerization,[19−22] proton-transfer,[21,23−25] electrocyclization,[26] conformational-planarization,[27−29] and photodissociation[30] reactions.

In a recent review[21] outlining how ESA may improve photochemical reactivity, two broad strategies were discussed, as illustrated in Scheme . The first involves utilizing a gain of antiaromaticity (or a loss of aromaticity) from the initial electronic excitation,[23] and the second involves exploiting a gain of aromaticity during the subsequent geometric relaxation from the vertically excited Franck–Condon (FC) point.[20,22] Interestingly, however, a recent investigation[26] noted that photoinduced changes in aromaticity need not always influence photochemical reactivity in a positive direction. Specifically, through transient absorption spectroscopy and time-dependent density functional theory (TD-DFT) studies of dihydroazulene-vinylheptafulvene photoswitches operated through electrocyclization reactions, it was found that the loss of aromaticity in a benzene motif from the initial electronic excitation hampers the photochemical reactivity by increasing an already existing barrier along the excited-state potential energy surface (PES).[26] In this light, it is worthwhile to investigate whether a similar conclusion is applicable also to other types of photoswitches utilizing electrocyclization reactions and featuring a benzene motif.

Scheme 1

(a) Facilitating an Excited-State Proton Transfer Reaction through a Gain of Anti-Aromaticity in the Excited State[23] and (b) Facilitating a Z/E Photoisomerization Reaction through a Gain of Aromaticity in the Excited State[20,22]

To this end, we here consider dithienylethene switches, which are one of the most extensively studied class of electrocyclization-based photoswitches,[31−39] with applications in photopharmacology,[40,41] molecular electronics,[34,38,42,43] and in the design of complex functional materials.[44−46] Specifically, we report on the synthesis and the experimental and computational characterization of a dithienylbenzene switch where the typical nonaromatic, ethene-like motif[31−39] bridging the two thienyl units, is replaced by an aromatic benzene motif. The corresponding switch in its isomeric open (1a) and closed (1b) forms, and in a closed-oxidized form (1c), are shown in Scheme .

Scheme 2

Chemical Structures of the Open (1a), Closed (1b), and Closed-Oxidized (1c) Forms of the Dithienylbenzene Switch Studied in This Work

Photoswitches like 1a with an aromatic benzene π-linker have a potential key advantage over those with a nonaromatic linker in that the energy stored in the ring-closed isomer 1b is generally larger, because of the transformation of the benzene motif in 1a into a nonaromatic cyclohexadiene motif in 1b. This makes these switches interesting candidates as molecular solar thermal energy (MOST) storage systems.[47−50] However, a recent study[51] on dithienylethene switches with pyrazine-based aromatic π-linkers predicted the photocyclization of such switches to be hindered by the energetic expense associated with the ensuing loss of ground-state aromaticity. Although reached systematically through absorption spectroscopy and crystallographic measurements, and with support from DFT computations,[51] this prediction seems to us somewhat simplistic. In fact, it relies essentially on the assumption that the ground-state aromaticity of the π-linker remains unchanged in the excited state. However, following the foregoing discussion of ESA based on Baird’s rules[7] and their extensions,[9,10,52] this assumption is not valid for a switch featuring a benzene π-linker. Furthermore, it is well-known that the ground-state aromaticity of the thienyl units does not impede the photocyclization of dithienylethene switches.[34,37,38] Therefore, it seems natural to expect the same to hold true for the ground-state aromaticity of the π-linker. It should also be mentioned that Kitagawa and co-workers[53] recently reported spectroscopic evidence for photocyclization of a fluorinated derivative of 1a, in the form of a large absorption red-shift upon UV irradiation. As pointed out by these authors,[53] the nondetection of the ring-closed isomer in previous studies of such photoswitches is not necessarily an indication that photocyclization does not occur but is rather a consequence of the fact that formation of the ring-closed isomer is immediately followed by fast thermal cycloreversion (that restores the ground-state aromaticity of the π-linker).

In this work, the potential problem with detecting the ring-closed isomer 1b is bypassed by allowing, as shown in Scheme , for its subsequent oxidation to the more stable product 1c, which is isolated and characterized by means of UV–vis absorption and NMR spectroscopy. This oxidation is facilitated by the presence of hydrogens instead of alkyl groups at the ring-closing C11 and C13 positions. In particular, this work presents detailed computational evidence supporting the idea that photoinduced changes in the aromaticity of the benzene motif of dithienylbenzene switch 1a exerts a distinctly positive influence on the photocyclization reactivity of this compound.

Results and Discussion

Experimental UV–vis Absorption and 1H NMR Spectra

The UV–vis absorption spectra of 1a shown in Figure display high-energy bands and do not exhibit any solvent dependence, which is further corroborated by spectra recorded in additional solvents given in Figure S1 of the Supporting Information (SI). Upon irradiation of 1a in MeCN with UV-light of 254 nm, no apparent color change was observed. However, as illustrated in Figure , new absorption bands appeared in the near-UV region. Prolonged irradiation of the sample produced clear isosbestic points at 229, 271, and 279 nm, indicating the formation of a single, well-defined product. This photochemical transformation was found irreversible and, as can be inferred from Figures S2–S4 of the SI, quite insensitive to the nature of the solvent.

Figure 1

UV–vis absorption spectra of 1a in different solvents.

Figure 2

Irradiation of 1a in MeCN solution at 254 nm.

The gradual change of the initial UV–vis absorption spectrum of 1a in Figure upon irradiation at 254 nm suggests that the high-energy UV-light induces a 6-electron electrocyclization of this compound. The resulting ring-closed product 1b is expected to undergo either of two reactions: thermal cycloreversion back to the parent species (1a), which is favored by the regeneration of the three aromatic rings, or oxidation by atmospheric O2 to yield a highly stable, aromatic 18 π-electron compound (1c) and H2O. In order to gain structural insight into which of these two reactions is more likely, the photochemical transformation was followed by 1H NMR spectroscopy. The results are presented in Figure and in Figures S6–S8 of the SI. As can be seen from Figures and S6, upon irradiation of a solution of 1a in CD3CN, a new set of peaks appears in the 1H NMR spectrum, upfield to the signals before irradiation, while the intensities of the original signals are reduced. Gratifyingly, based on an analysis of the new peaks (see Figure S7), it was found that they correspond to 1c. Furthermore, consistent with this interpretation, water is the only observed byproduct (see Figure S8). Although 1c has previously been prepared by alternative methods,[54] its photochemical generation has not been reported before.

Figure 3

Irradiation of 1a in CD3CN at 254 nm, followed by 1H NMR spectroscopy. (a) 1H NMR spectrum before irradiation. (b) 1H NMR spectrum after 90 min.

Besides synthesizing 1a and generating 1c photochemically, we also synthesized a tetramethylated derivative of 1a bearing methyl substituents instead of hydrogens both at the “outer” C9 and C15 positions and the ring-closing C11 and C13 positions. For this compound, which has not been prepared earlier and is denoted Me-1a in Scheme , no apparent change of its absorption spectrum (see Figure S5 of the SI) was observed upon irradiation with UV-light. This may indicate that the ring-closed species, being relatively insusceptible to oxidation by the presence of the C11 and C13 methyl groups, undergoes a fast thermal cycloreversion that precludes its spectroscopic detection.[53] We will return to this issue in section .

Scheme 3

Chemical Structures of Additional Dithienylbenzene Switches Studied in This Work

Thermal Electrocyclization and Oxidation

The first task for the computational part of this work is to clarify how probable or improbable the thermal electrocyclization of 1a is. Such an assessment will provide a frame of reference for the subsequent modeling of the photoinduced electrocyclization process. Before this assessment, however, it should be noted that previous studies on diarylethene switches have observed two possible conformations for the open isomer: a parallel conformation with the aryl rings in mirror symmetry, and an anti-parallel conformation with the aryl rings in C2 symmetry.[33,34,37] Analogously, the open isomer 1a also exhibits parallel and anti-parallel conformations, shown in Figure and hereafter denoted 1a-p and 1a-ap, respectively. As can be seen in Table , each of the four DFT methods employed for the calculations, which were performed using an SMD[55] continuum solvation model representation of a MeCN solvent, yields free energies of 1a-p and 1a-ap that are very similar. Furthermore (see Table S1 of the SI), the predicted free-energy barriers for the thermal interconversion between these conformations are small, only 11–25 kJ mol–1, which suggest very fast interconversion processes. Overall, these findings are consistent with previously reported NMR data on dithienylethene switches indicating that the two conformations are equally populated under ambient conditions.[33,34,38] Moreover, these studies showed that the parallel conformation does not undergo photocyclization.[33,34] For this reason, only the 1a-ap conformation was considered in the present modeling of the thermal electrocyclization and photocyclization reactions of 1a.

Figure 4

Conformations of 1a, 1b, and 1c considered in the calculations of this work.

Table 1

Calculated Relative Free Energies for the 1a/1b/1c System (kJ mol–1)a

methodb	1a-p	1a-ap	1b	1c+H2
B3LYP	70.6	72.1	183.3	0
M06-2X	61.6	64.0	186.6	0
ωB97X-D	53.2	55.0	177.9	0
CAM-B3LYP	55.4	56.8	177.2	0

Energies given relative to the total free energy of the 1c+H2 system.

Geometry optimizations carried out using the cc-pVDZ basis set and singlepoint calculations carried out using the cc-pVTZ basis set, employing in both cases an SMD description of a MeCN solvent.

While the free-energy barriers for the 1a-p ↔ 1a-ap interconversion are small, Table shows that each of the four DFT methods predicts a barrier for the 1a-ap → 1b thermal electrocyclization of such magnitude (183–197 kJ mol–1) that the reaction cannot be induced thermally under reasonable conditions. This is in accordance with the reaction being a Woodward–Hoffmann forbidden reaction. It is also notable that the electrocyclization transition structure (henceforth denoted TS) calculated by each method has substantial diradical character, which is another key characteristic of Woodward–Hoffmann forbidden reactions[56−59] and is revealed by the nonzero (∼0.7) expectation value of the spin-squared operator and the associated spin densities given in Table S2 of the SI.

Table 2

Calculated Free-Energy Barriers for the 1a-ap → 1b Thermal Electrocyclization and the 1b → 1a-ap Thermal Cycloreversion of the Parent Switch 1a and Its Methylated Derivatives Me-1a, Me-1a, and Me-1a* (kJ mol–1)a

switch	method	electrocyclization	cycloreversion
1a	B3LYP	182.6	70.5
1a	M06-2X	197.0	76.0
1a	ωB97X-D	190.0	66.6
1a	CAM-B3LYP	190.3	68.8
Me4-1a	ωB97X-D	210.0	72.1
Me2-1a	ωB97X-D	184.7	77.6
Me2-1a*	ωB97X-D	216.9	60.1

All calculations were carried out using the aug-cc-pVTZ basis set and an SMD description of a MeCN solvent.

As for the calculated relative free energies of 1a-ap and 1b, these predict that the overall electrocyclization process is endergonic by 111–123 kJ mol–1 (see Table ), which is readily explained by the loss of aromaticity in the π-linker. Should the electrocyclization be inducible by sunlight, this large endergonicity would be an attractive feature of systems of this kind for potential MOST applications.[47−50] Such applications would also require that the free-energy barriers of 67–76 kJ mol–1 (see Table ) that the calculations ascribe to the 1b → 1a-ap thermal cycloreversion can be increased, so as to ensure that the absorbed solar energy can be stored for a long time. In order to briefly explore this possibility, cycloreversion barriers were also calculated for three methylated derivatives of the parent switch (see Scheme ), including the aforementioned Me-1a compound and two dimethylated compounds differing with respect to whether the methyls occur at the “outer” C9 and C15 positions (Me-1a) or the ring-closing C11 and C13 positions (Me-1a*). Incorporated into Table , the results of these calculations suggest that the C9 and C15 methyls increase the barrier by 11 kJ mol–1, whereas the C11 and C13 methyls decrease it by 7 kJ mol–1. Accordingly, in Me-1a, the two effects add up to a slight increase of the barrier relative to the parent switch. Interestingly, this indicates that the absence noted in section of a change in the experimental absorption spectrum of Me-1a upon irradiation with UV-light, is not related to the influence of the C11 and C13 methyls on the cycloreversion, but rather to their ability to prevent oxidation of the ring-closed species.[53]

Turning thus to the oxidation of 1b by atmospheric O2, which yields 1c and H2O, the calculated free energies in Table S3 of the SI suggest that this reaction proceeds with an astoundingly large exergonicity (ΔG2) of 400–420 kJ mol–1. Thereby, the experimental detection of 1c described in section can be rationalized by the fact that the driving force for oxidation is much larger than the driving force for thermal cycloreversion (111–123 kJ mol–1). Moreover, the driving force for oxidation can be partitioned into two components: the first (ΔG3) being associated with the formation of H2O from H2 and O2, and the second (ΔG2 – ΔG3) reflecting the gain in aromaticity upon formation of the central six-membered ring that fuses the two thienyls in 1c (see Scheme ). As can be seen from Table S3, these two components amount to 217–233 and 177–187 kJ mol–1, respectively.

Calculated UV–vis Absorption Spectra

The second task for the computational part of this work is to assess how well the TD-DFT methodology adopted for the modeling of the photocyclization of 1a reproduces the experimental UV–vis absorption spectrum of this compound in MeCN solution. To this end, the absorption of 1a-p and 1a-ap was calculated with the same four DFT methods employed for the modeling of the thermal processes discussed in section . Again, an SMD[55] description of the MeCN solvent was used, in combination with the cc-pVTZ basis set. For each species, vertical excitation energies from the ground state (S0) to the ten lowest singlet excited states and the associated oscillator strengths were calculated to derive, through convolution with Gaussian functions with the full width at half-maximum set to 0.5 eV, the UV–vis absorption spectra shown in Figure a. In addition, “total” spectra of 1a-p and 1a-ap were derived by assuming the individual contributions of 1a-p and 1a-ap to be quantifiable in terms of their Boltzmann weights. The corresponding spectra are given in Figure b, which also includes the experimental spectrum of 1a for comparison.

Figure 5

(a) UV–vis absorption spectra of 1a-p and 1a-ap in MeCN solution calculated with different DFT methods. (b) Boltzmann-weighted UV–vis absorption spectra of 1a-p and 1a-ap in MeCN solution calculated with different DFT methods and compared with the experimental spectrum of 1a.

As can be inferred from the excitation energies and oscillator strengths summarized in Table S4 of the SI, the high-energy band in the spectra of both 1a-p and 1a-ap in Figure a calculated with the M06-2X, ωB97X-D and CAM-B3LYP methods is dominated by the S0 → S3 transition, which occurs at ∼5.5 eV/∼225 nm. Furthermore, the absorption shoulder at ∼250 nm predicted by these methods can be assigned to the S0 → S1 ππ* transition, which is located at 4.84–5.09 eV/244–256 nm (at the B3LYP level, this transition results in a clearly identifiable lower-energy band). From Figure S13 of the SI, it can be seen that the frontier π-molecular orbitals involved in the S0 → S1 transition are primarily localized on the benzene π-linker, with negligible contributions from the atoms of the two thienyls. This suggests that it is indeed this moiety, and not the thienyls, whose aromaticity may influence the photocyclization of 1a.

Turning to the Boltzmann-weighted total spectra in Figure b, especially M06-2X, ωB97X-D and CAM-B3LYP achieve good overall agreement with the experimental spectrum of 1a. Furthermore, given that the 1H NMR analysis in section attributed the appearance in Figure of new near-UV absorption bands upon irradiation of 1a at 254 nm to the formation of 1c, it is also gratifying that the lower-energy band in the calculated spectra of 1c (see Figure S14b of the SI) is indeed red-shifted relative to those in the calculated spectra of 1a-p and 1a-ap (see Figure a). Overall, then, we believe that the results in Figures and S14b lend credence to the use of TD-DFT for the modeling of the photocyclization of 1a, which we turn to next. Since 1a-ap, but not 1a-p, undergoes photocyclization,[33,34] all calculations pertaining to compound 1a described henceforth were carried out using strictly the 1a-ap species, which is hereafter referred to as simply 1a.

Photocyclization

The results from the modeling of the 1a → 1b photocyclization in the S1 state through static TD-DFT calculations with the ωB97X-D method are summarized in Figure , which for comparison also includes results relevant for the symmetry-forbidden thermal electrocyclization reaction. For example, as was already noted in Table , it can be seen that the thermal reaction is prohibited by a very large barrier. Contrarily, following light absorption and population of the vertically excited S1 FC point of 1a, Figure a shows that photocyclization along the C11–C13 (see Scheme ) reaction coordinate proceeds in a completely barrierless fashion, whereby the S1 and S0 states approach degeneracy (see Figure b) in the region of a presumed S1/S0 conical intersection (CI) seam. The excited-state evolution toward this seam, which affords a decay channel to the S0 state, is driven by a large decrease in S1 energy relative to the FC point at C11–C13 = 3.29 Å. For example, at C11–C13 = 1.80 Å, the decrease in S1 energy amounts to 1.70 eV. In order to corroborate this result in light of potential problems in the treatment of static correlation effects by TD-DFT,[60] singlepoint calculations carried out with the complete active space second-order perturbation theory (CASPT2) method[61,62] yield a value of very similar magnitude (1.74 eV, see Table S6 of the SI). Furthermore, based on the observation that 1b is produced by a ωB97X-D S0 geometry optimization started in the vicinity of the presumed S1/S0 CI seam at C11–C13 = 1.80 Å, it also seems plausible that 1b is indeed the S0 photoproduct of the photocyclization of 1a.

Figure 6

(a) Calculated S1 photocyclization path of 1a and relative energies of 1a, 1b, and the TS geometry along the thermal S0 electrocyclization path. (b) Calculated S1 photocyclization path of 1a and S0 energies at the corresponding geometries.

It should be noted that even though the calculations summarized in Figure suggest that the 1a → 1b photocyclization is a favorable reaction, the underlying procedure to perform S1 geometry optimizations along a predefined C11–C13 reaction coordinate neither accounts for possible competition from other reactions, nor provides an estimate of the time scale for the overall photocyclization process. In order to address these issues, we also carried out nonadiabatic molecular dynamics (NAMD) simulations[63−67] with Tully’s fewest switches algorithm[68] at the level of B3LYP for the DFT and TD-DFT parts. These simulations, which describe the simultaneous evolution of the electronic and nuclear degrees of freedom of the photoswitch, were started from the S1 FC point of 1a and were run for maximally 200 fs with ten different initial nuclear velocities generated randomly from a Maxwell–Boltzmann distribution at 300 K. The key results are summarized in Figure .

Figure 7

Changes in S1 and S0 energies (a) and C11–C13 bond length (b) during the ten NAMD trajectories run for 1a.

As can be seen from Figure a, there is an apparent tendency for the excited-state dynamics to allow the S1 and S0 states to approach degeneracy, and to do so very quickly (within ∼100 fs in many trajectories). Furthermore, it is clear from the rapid decrease in C11–C13 distance observed in all ten trajectories (see Figure b) that a key component of the dynamics is indeed ring-closing along the C11–C13 coordinate. In fact, within 200 fs, four out of ten trajectories oscillate around a C11–C13 distance of 1.5 Å. Since all ten trajectories are evolving in the S0 state at that stage (see Figure S15 of the SI), this means that 40% of the trajectories have completed the reaction and formed the 1b photoproduct within 200 fs. Thus, the 1a → 1b photocyclization appears to be very fast and not much impeded by competing processes. Loosely, this 200 fs time scale is consistent with findings reported in time-resolved spectroscopic studies of ring-closing in diarylethene switches.[69−72] Finally, on a technical note, complementary NAMD simulations reported in Figures S16 and S17 of the SI performed with initial nuclear velocities generated at 500 K instead of 300 K, give results similar to those just described.

Aromaticity Indices

Having obtained computational evidence that the 1a → 1b photocyclization is an ultrafast and efficient reaction, we now set out to investigate if this finding can be rationalized in terms of photoinduced changes in aromaticity of the benzene π-linker. To this end, aromaticity indices of four different types—geometric, electronic, magnetic, and energetic—were calculated for this moiety. Starting with the geometric harmonic oscillator model of aromaticity (HOMA) index,[73,74] which probes the deviation of carbon–carbon bond lengths from an ideal aromatic reference value for benzene, aromatic and nonaromatic systems usually exhibit HOMA values close to 1 and 0, respectively.[73−75] Presenting the HOMA values of 1a and 1b in the S0 state calculated from the corresponding ωB97X-D geometries, as well as the HOMA values along the S1 photocyclization path of 1a calculated based on the corresponding TD-ωB97X-D geometries, Figure a unsurprisingly shows that 1a (0.96) and 1b (0.10) are typical S0 aromatic and nonaromatic systems, respectively. However, more interestingly, as the photoexcited system evolves from the S1 FC point of 1a, the HOMA values are much reduced (from 0.96 to 0.46–0.65), which suggests that there is a loss of aromaticity in the S1 state relative to the S0 state that might contribute to the photocyclization reactivity of 1a.

Figure 8

Changes in HOMA (a) and SA values (b) for the benzene π-linker along the S1 photocyclization path of 1a. Also shown are the HOMA and SA values for 1a, 1b, and the TS geometry along the thermal S0 electrocyclization path.

Continuing with the electronic Shannon aromaticity (SA) index,[76,77] which is formulated in the framework of Bader’s theory of atoms in molecules[78] and measures the variation in electron density at bond critical points, the analysis of these results in Figure b (obtained from S0 and S1 electron densities calculated with ωB97X-D and TD-ωB97X-D, respectively) supports the same conclusion as the analysis of the HOMA values in Figure a. Specifically, noting that aromatic systems typically show smaller SA values than their nonaromatic analogues because of their more uniform distribution of electron density,[76,77] the SA values are increased from 0.0001 for 1a in the S0 state to 0.0014–0.0021 along the S1 photocyclization path.

Despite the agreement between the HOMA and SA values, it is important to point out that these indices do not distinguish between nonaromaticity and antiaromaticity. Moreover, as can be inferred from Figure , they are not able to probe changes in aromaticity during the photoexcitation from the S0 state of 1a to the S1 FC point. In order to circumvent these limitations, the aromaticity of the benzene π-linker in the S0 and S1 states of 1a was assessed magnetically through calculation of nucleus-independent chemical shift (NICS) indices[79,80] by means of a NICS-scan procedure.[81,82] These indices probe ring currents induced by circulating π-electrons, and were calculated from wave functions obtained with the complete active space self-consistent field (CASSCF) method[83] and using gauge-including atomic orbitals.

The results of this investigation are summarized in Figure . We see that the S0 state of 1a shows negative NICS values between −11 and −22 ppm, indicating diatropic ring currents or aromaticity.[79] At the S1 FC point, on the other hand, the NICS values are positive (23–70 ppm), suggesting paratropic ring currents or antiaromaticity.[79] In the CI region along the S1 photocyclization path, in turn, the NICS values are close to zero, reflecting the absence of ring currents or nonaromaticity.[79] Accordingly, the overall picture that emerges from Figure regarding changes in aromaticity during the course of the photocyclization of 1a, is that the π-linker goes from being aromatic in the S0 state to antiaromatic at the S1 FC point, and then turns nonaromatic in the CI region. Thereby, it appears that the photocyclization is driven by the loss of aromaticity and the concurrent formation of a reactive, antiaromatic excited state from the initial photoexcitation, and by the subsequent relief of this antiaromaticity as the reaction progresses from the FC point. In other words, the reaction is facilitated by photoinduced changes in aromaticity.

Figure 9

NICS-scan values for the benzene π-linker in the S0 state and at the S1 FC point of 1a. These values are denoted 1a (S0) and 1a (S1 FC), respectively. Also shown are the NICSzz-scan values for a structure in the CI region (at C11–C13 = 1.80 Å) along the S1 photocyclization path of 1a, for the TS geometry along the thermal S0 electrocyclization path of 1a, and for the S0 state of 1b. These values are denoted CI (S1), TS (S0), and 1b (S0), respectively.

In order to quantify in energetic terms the loss of aromaticity during photoexcitation and the loss of antiaromaticity during the excited-state evolution, two novel aromaticity indices were introduced and calculated for a methylated (at the benzene motif) derivative of 1a and its nonaromatic exocyclic methylene isomer. Denoted 2 and 2-ref, respectively, these compounds are shown in Scheme . The indices in question, which are reminiscent of the isomerization stabilization energy (ISE) aromaticity index,[84] quantify the difference ΔVEE = VEE(2) – VEE(2-ref) in the vertical S0 → S1 excitation energies of 2 and 2-ref, and the difference ΔFCRE = FCRE(2) – FCRE(2-ref) in the S1 FC relaxation energies of 2 and 2-ref (the S1 FC relaxation energy is the decrease in S1 energy from the FC point to the minimum on the S1 PES). Thereby, if there is a distinct loss of aromaticity during the photoexcitation of 1a and similarly a pronounced loss of antiaromaticity during its subsequent excited-state evolution, one expects these indices to be positive and large. Thus, it is encouraging to note from calculations presented in Table S7 of the SI that this is indeed the case, with ΔVEE and ΔFCRE amounting to 1.11 and 0.71 eV, respectively. This finding is consistent with a previous study on cyclopentadienes and siloles demonstrating that aromatic compounds have relatively larger vertical excitation energies than their nonaromatic or antiaromatic analogues.[85]

Finally, we note that it is of interest for future studies to explore how functionalization of the benzene motif with different electronic substituents might influence the photoinduced changes in aromaticity. In this regard, the attachment of strongly electron-withdrawing groups to benzene has previously been found to reverse the character of its T1 state from antiaromatic to aromatic.[18] As a brief investigation of this issue, ΔVEE and ΔFCRE values were also calculated for a derivative of 2 (denoted 3 in Scheme ) featuring an electron-withdrawing nitro group at the benzene motif. Interestingly, the ΔVEE and ΔFCRE values of 0.82 and 0.36 eV for this compound given in Table S7 are noticeably smaller than those of 1.11 and 0.71 eV calculated for 2. Hence, the nitro group appears to inhibit the photoinduced changes in aromaticity.

Conclusions

We have synthesized dithienylbenzene 1a and reported 1H NMR evidence that this compound undergoes electrocyclization upon irradiation with UV-light of 254 nm. Furthermore, through a detailed computational analysis, we have found that the electrocyclization process is facilitated by photoinduced changes in the aromaticity of the benzene motif of 1a. More specifically, there is a loss of aromaticity during the initial photoexcitation that produces a reactive, antiaromatic excited state (NICS values change from distinctly negative to distinctly positive), and subsequently a relief of this antiaromaticity as the photoexcited system evolves from the FC point toward the CI seam (NICS values change from distinctly positive to ∼ zero). Through NAMD simulations, this evolution is predicted to be very fast (∼100 fs) and to be followed by excited-state decay and complete electrocyclization within 200 fs in many (40%) of the simulated trajectories. Altogether, the results of this work show that the concept of ESA holds substantial promise as a tool for designing potent electrocyclization-based photoswitches within the framework of dithienylethenes. In future research, we plan to investigate whether this concept is fruitfully applicable also to the design of photoswitches based on other reactions and frameworks.

Experimental and Computational Details

Synthesis

Dithienylbenzene 1a was synthesized in a single step via a 2-fold Suzuki-Miyaura cross-coupling of 1,2-dibromobenzene and 3-thienylboronic acid in excellent yield (see Scheme ). Full details are given in the SI.

Scheme 4

Synthesis of 1a

UV–vis Absorption and 1H NMR Spectroscopy

Full details are given in the SI.

Quantum Chemical Calculations

Thermal and photochemical reactions were modeled using DFT and TD-DFT,[86−94] respectively. As for the former reactions, S0 stationary points corresponding to 1a, 1b, 1c, and the electrocyclization transition structure (TS) connecting 1a and 1b were located by performing geometry optimizations using the B3LYP,[95,96] M06-2X,[97,98] ωB97X-D,[99] and CAM-B3LYP[100] hybrid density functionals in combination with the cc-pVDZ basis set (in some cases also with the larger aug-cc-pVTZ basis set). For each structure optimized at a specific level of theory, frequency calculations were carried out at that level to obtain Gibbs free energies at room temperature and to ensure that the structure in question is either a minimum with real vibrational frequencies only (1a, 1b, and 1c), or a first-order saddle point with one imaginary vibrational frequency (TS). In order to refine the electronic contribution to the free energies, the geometries obtained with a specific functional were subjected to singlepoint calculations with that functional and the larger triple-ξ cc-pVTZ basis set. All calculations were performed using an SMD[55] continuum solvation model representation of a MeCN solvent.

As for the TD-DFT modeling of the 1a → 1b photocyclization, in turn, this was done within the Tamm-Dancoff approximation (TDA)[101−104] to TD-DFT using exclusively the ωB97X-D functional. This choice was made in part because ωB97X-D is a range-separated hybrid functional, which has proven a useful feature for the description of photoisomerization reactions,[105,106] and in part because ωB97X-D includes empirical 1/R6 atom–atom dispersion corrections[99,107,108] suitable for capturing weak intermolecular interactions between the two thienyl units. Starting from the ωB97X-D/cc-pVDZ S0 geometry of 1a, whose C11–C13 distance is 3.29 Å, the S1 photocyclization path was computed by first locating the vertically excited S1 FC point through a singlepoint calculation with the cc-pVTZ basis set and the SMD (MeCN) model. Subsequently, additional points along the path were obtained by performing a series of constrained S1 geometry optimizations (again with the cc-pVDZ basis set) in which the C11–C13 distance was varied between 3.20 and 1.80 Å in steps of 0.10 Å. In each optimization, all other geometric degrees of freedom than the C11–C13 distance were allowed to relax. For computational expedience, the optimizations were done in the gas phase, but were followed by singlepoint calculations (again with the cc-pVTZ basis set) utilizing the SMD (MeCN) model. Finally, as a complement to the TD-DFT modeling, singlepoint calculations along the S1 photocyclization path were also carried out with the CASPT2 method,[61,62] as further detailed in the SI.

NAMD Simulations

The NAMD simulations were done with the implementation of Tully’s fewest switches algorithm[68] in TURBOMOLE[64,109] at the B3LYP/cc-pVDZ level of theory for the DFT and TD-DFT parts. The reason for using B3LYP, and not ωB97X-D, is that the latter method is not implemented in TURBOMOLE. The simulations were started from the S1 FC point of 1a and were run for maximally 200 fs with two sets of ten different initial nuclear velocities generated randomly from a Maxwell–Boltzmann distribution at 300 and 500 K, respectively. The exact criteria used to enforce a hop between the S1 and S0 states in the simulations are given in the SI.

Aromaticity Indices and Software Used

Full details on the calculations of HOMA,[73,74] SA,[76,77] and NICS[79−82] indices for the benzene π-linker are given in the SI. Briefly, HOMA values were calculated based on S0 and S1 geometries optimized at the ωB97X-D/cc-pVDZ and TD-ωB97X-D/cc-pVDZ levels of theory, respectively. SA values, in turn, were derived from S0 and S1 electron densities computed at the ωB97X-D/cc-pVTZ and TD-ωB97X-D/cc-pVTZ levels of theory, respectively. NICSZZ values,[80] finally, were calculated following a NICS-scan procedure[81,82] using S0 and S1 wave functions obtained at the CASSCF/cc-pVDZ level of theory. More specifically, NICS values were calculated along the axis passing through the geometric center of the π-linker at distances above the center ranging from 1.50 to 2.50 Å in steps of 0.10 Å. This procedure is designed to avoid the undesirable arbitrariness associated with NICS-index calculations that only consider one single point in space.[81,82,110]

The calculations were done with the Gaussian 16,[111] OpenMolcas 18.09,[112] TURBOMOLE 7.4,[109,113] Dalton2016.2,[114,115] and Multiwfn[116] programs. Gaussian 16 was used for DFT and TD-DFT calculations. OpenMolcas 18.09 was used for CASSCF and CASPT2 calculations. TURBOMOLE 7.4 was used for NAMD simulations. Dalton2016.2 was used for NICS-index calculations. Multiwfn was used for HOMA- and SA-index calculations.