Is Excited State Aromaticity a Driving Force for Planarization of Dibenzannelated 8  -Electron Heterocycles?

: Compounds with dibenzannelated heterocycles with eight  -electrons are found in a range of applications. It was argued by Shukla and Wan [ J. Am. Chem. Soc. 1993 , 115 , 2990] that two such compounds, dibenz[ b,f ]oxepine and dibenz[ b,f ]thiepin, adopt planar structures in their lowest singlet excited states due to “attainment of a cyclically conjugated system of 8  electrons in the central ring”. Herein we report on a quantum chemical investigation of the aromatic character in the first excited singlet and triplet states (S 1 and T 1 ) of dibenzannelated seven-and six-membered heterocycles with one, two or three heteroatoms in the 8  -electron ring. The S 1 and T 1 states could have  * or n  * character, and we find that compounds with one or two heteroatoms in the central ring have  * states as their S 1 and T 1 states. These states are to a significant degree influenced by excited state aromaticity, and their optimal structures are planar or nearly planar. Among the heteroatoms, nitrogen provides for the strongest excited state aromaticity whereas oxygen provides for the weakest, following the established trend of

the electronic ground state. Yet, dibenzannelated seven-membered ring compounds with N=N bonds have nonaromatic n* states with strongly puckered structures as their S1 and T1 states.

Introduction
A change in aromaticity is one of the most powerful driving forces to control and modulate reactivity, structure and other molecular properties. [1] In the ground state (S0), Hückel's rule tells that aromaticity is associated with fully conjugated cycles with 4n + 2 π-electrons, [2] and it has had a profound impact on our understanding of various chemical reactions and molecular properties. A change in aromaticity in the lowest electronically excited states, as given by Baird's 4n rule, [3,4] can similarly be a driving force for photoreactivity and changes in excited state properties. [5,6] Baird's rule was formulated for the lowest * triplet state (T1), yet, it has been found through computations that it often extends to the lowest singlet excited state (S1) of small annulenes. [7,8] Thus, annulenes with 4n -electrons can be aromatic in their T1 and S1 states.
The focus herein is on the excited state aromaticity of dibenzannelated molecules with central 8π-electron rings. Compounds of this type are found in a wide range of applications, for example in antipsychotic drugs such as quetiapine and chlorpromazine (Figure 1), and in photofunctional molecular materials for usage as viscosity probes and photoresponsive liquid crystals. [9,10] Dibenzo[b,f] [1,4]oxazepin is a very strong lachrymatory agent known as CR gas, and dibenzodioxin is the core in some of the most environmentally hazardous polychlorinated chemicals known. Finally, oxepin units situated at the edges of graphene nanosheets as cyclic ethers, effectively benzannelated oxepins, have been proposed as the cause for exciton selftrapping observed in graphene quantum dots and carbon dots. [11]  Cyclic 8-electron molecules in their S0 states normally adopt non-planar structures that are non-aromatic rather than antiaromatic. Cyclooctatetraene (COT) in S0 adopts a tubshaped geometry, avoiding the angle strain at the planar D4h symmetric structure, [12] and also azepines, oxepines, and thiepines adopt puckered conformations. [13][14][15][16][17][18][19] The resonance energies of the planar structures of azepine and oxepines obtained through extended Hückel MO theory and early Hartree-Fock computations suggested antiaromatic character, [20,21] later supported by NICS calculations. [14,22] However, the dibenz[b,f]annelated derivatives were found to have positive resonance energies associated with some aromatic character as their benzene rings keep their Hückel aromaticity in line with Glidewell-Lloyd's extension of Clar's rule. [23,24] Still, dibenzo[b,f]oxepin, similarly to oxepin, adopts a saddle-shaped structure in the S0 state. [4] Also dibenzannelated 8π-electron six-membered ring compounds such as phenothiazine and phenoxepine adopt nonplanar conformations in S0. [25] Yet, Baird's rule can lead to aromatic stabilization and planarization of many of these molecules in their lowest excited states. Quantum chemical calculations tell that COT in the S1 and T1 states exhibits planar D8h structures and magnetic properties typical of high degree of aromaticity. [7,8,[26][27][28] Several experimental observations related to large structural changes in the excited states, when compared to the S0 state, have been reported for COT and a number of COT derivatives. [9,29,30] With regard to dibenzannelated heterocycles, dibenz[b,f]oxepin displays a large Stokes' shift and well-defined vibrational fine structure in the fluorescence spectrum, evidences that a change from a V-shaped to a planar conformation occurs in the S1 state ( Figure 2), [31] and similar findings were made for dibenz[b,f]thiepin. Interestingly, dibenz[b,f]oxepin shows an increased photostability when compared to its 10,11dihydrogenated analogue, [31] a feature that could be connected to the gain in S1 state aromaticity of a cyclic system with 4n -electrons. More recently, it was possible to obtain an experimental assessment of excited state aromatic stabilization in a chiral COT derivative. [29] The racemization enthalpy determined through time-resolved CD spectroscopy of the isolated enantiomer revealed its excited state aromatic stabilization to be 21 -22 kcal/mol in both the T1 and S1 states. Yet, there are also limitations to the excited state aromaticity concept. The presence of an 8 cyclooctatetraene ring in the center of acene dimers is not a guarantee for planarization in the S1 state, as it depends on the acene length. [6] Also, is there a similarity in the degree of aromaticity between azepines, oxepines and thiepines in the T1 and S1 states as there is between pyrrole, furan, and thiophene in the S0 state? Finally, with more C atoms exchanged to heteroatoms there will be an increase in the number of n* states, and they may become the S1 and T1 states. Herein we report on a computational study in which we probe if gain of excited state aromaticity, as given by Baird's rule, is a general driver for planarization in the lowest excited states of dibenzannelated 8-electron heterocycles. The compounds can tentatively be described as aromatic chameleon compounds (Figure 4), [32,33] that is, compounds that can adapt their electronic structures so as to comply with the different aromaticity rules in different 6 electronic states: Hückel's rule in S0 with two -sextets and Baird's rule in T1 and S1 with a octet, -duodectet or -hexadectet. Besides providing information on the scope and limitations of the excited state aromaticity concept to tricyclic molecules with overall 16 -electrons, the study also provides insights of the structure-property relationship of molecules that can be of interest in the design of new drugs and photoactived materials. Molecules with aromatic S1 and T1 states may display higher photostability than molecules with n* states as their lowest excited states. [31] Such findings could be of relevance for targeted design of compounds with improved photostability, or the opposite, increased photoreactivity.  showing that it can act as an "aromatic chameleon" compound.

Computational Methods
Geometry optimizations for the S0, S1, and T1 states were carried out using the PBE0 [34] functional and the 6-311+G(d,p) [35] basis set. For selected structures, the effect of dispersion corrections was analyzed by including the GD3 version of Grimme's dispersion model. [36] However, no significant changes were observed in the geometries of S0, S1, or T1 (see Figures   S6 to S10 in the Supporting Information). Time-dependent DFT (TD-DFT) was used for the S1 state optimizations while the unrestricted version of DFT was used for T1 state optimizations. The PBE0 functional was chosen based on benchmark papers for heterocyclic rings. [37][38][39] Stationary points with no imaginary frequencies were confirmed through frequency calculations. The stability of the wave function and spin contamination for the triplet state were checked. Vertical absorption and emission energies were computed at B97XD/6-311+G(d,p)//PBE0/6-311+G(d,p) level, having a good agreement with experimental results [40] (see Table S2 in the Supporting Information). CASSCF/6-31G(d) including all the -orbitals in the active space were carried out at TD-PBE0 geometries to evaluate the configurational weights and verify the character of the S1 state. For molecules containing seven and six members in the central ring the active space used was 16 electrons in 15 orbitals (CASSCF (16,15) and 16 electrons in 14 orbitals (CASSCF (16,14), respectively. For all the molecules investigated, the S1 state has a single-configurational character, and with regard to the triplet states all molecules with * T1 states have <S 2 > below 2.0014. CASSCF energies were corrected using CASPT2/6-31G(d) (using an imaginary shift of 0.1 and a standard IPEA value of 0.25). The CASSCF wave function was also used for computing MCI indices at the S1 state of a few molecules (see below). [41] For all the compounds, TD-DFT emission energies are in good agreement with CASPT2 emission energies and a comparison of the molecular orbitals and states show that TD-DFT is appropriate for the molecules studied. All DFT and TD-DFT calculations were performed with Gaussian 16 revision B.01 [42] while OpenMolcas version 18.09 was used for CASSCF and CASPT2 calculations. [43] The harmonic oscillator model of aromaticity (HOMA) index, [44,45] the multicenter index (MCI), [46] the aromatic fluctuation index (FLU), [47] the anisotropy of the induced current density (ACID) plots, [48,49] and the nucleus-independent chemical shifts (NICS)-XY scans [50,51] were used to quantify the aromatic character of the different systems. The correlation between the aromaticity indexes has been challenged; [52] here we chose to report several indexes for completeness. All five indices of aromaticity were computed at the (U)B3LYP/6-311+G(d,p) level at the PBE0/6-311+G(d,p) geometries.
The HOMA is defined as: where n is the number of bonds considered, α is an empirical constant (for C-C, C-N, C-O, C-S, and N-N bonds α = 257.7, 93.5, 157.4, 94.1, and 94.09, respectively), [53] Ropt is an optimal bond value (1.388, 1.334, 1.265, 1.667, and 1.309 Å for C-C, C-N, C-O, C-S, and N-N bonds, respectively) and Ri stands for a running bond length.
FLU was computed using delocalization indices, δ(A,B), with the expression: where A0  AN and the string = { 1 , 2 , … , } contains the ordered elements according to the connectivity of the N atoms in a ring or in a chosen circuit. V(A) is defined as: and  is a simple function to make sure that the first term is always greater than or equal to 1.
The delocalization indices of Eq. 1 were calculated using the overlaps between occupied molecular orbitals in the atomic basins generated by AIMAll program. [54] The The MCI is an electronic index obtained from Iring values as follows: where ( ) stands for a permutation operator which interchanges the atomic labels 1 , 2 ...
to generate up to the N! permutations of the elements in the string , and the Iring index is defined as: where ( ) is the overlap of natural orbitals i and j in the atom defined in the framework of the QTAIM, [55] and ni are their occupancies. FLU and MCI indices were obtained with the ESI program. [56] NICS-XY scans were performed using the Aroma package. [51] These were computed at 1.7Å above the plane of the molecules using the σ-only model to retrieve the πcontributions.

Results and discussion
The symmetric dibenzannelated compounds 1a -1c with central seven-membered 8-electron cycles are analyzed first. These compounds have only one heteroatom, and two of them have earlier been studied experimentally by Shukla and Wan. [31] We then consider compounds 1d -1g that have either two or three heteroatoms in the central 8-electron cycle. These compounds allow us to probe the limitations of the excited state aromaticity concept because gradually more n* states can now compete with the lowest and potentially Baird-aromatic * states for being the T1 and S1 states. At the end we briefly analyze the dibenzannelated six-membered heterocycles with eight -electron cores (2a -2d) which all have two heteroatoms in the central ring. In three of these, n* states can be the T1 and S1 states. Moreover, the T1 state geometry of each individual compound strongly resembles the S1 state geometry, indicating that the two states, apart from the difference in multiplicity, are of the same character. Also, when regarding the CC bond lengths of the three compounds in the T1 and S1 states one can only see minute variations between the three compounds ( Figure 6). For the S1 state, CC bond lengths in the perimeters are found in the ranges 1.369 -1.425 Å (1a), 1.368 -1.427 Å (1b), and 1.367 -1.430 Å (1c), respectively, and the same applies to the T1 state. The strong resemblance in geometries suggest that the S1 and T1 states in 1b and 1c are of * character, just like in in 1a where there is no competing n* excited state. level. TD-PBE0 was used for the S1 states and UPBE0 for the T1 states. Figure 6: Geometries of 1a -1c in their T1 and S1 states at PBE0/6-311+G(d,p) level. TD-PBE0 was used for the S1 states and UPBE0 for the T1 states. Bond distances are given in Å.
Indeed, an analysis of the orbitals and the electron configurations reveal the similarity between the S1 and T1 states; they are all of singly-excited (HOMO to LUMO) * character    correlation, as was previously observed by Solà and coworkers. [8,57] When comparing the three compounds, the MCI values for the T1 and S1 states are larger for the azepine ring in 1a than for the oxepine ring in 1b, indicating that the local aromaticity of the former is larger. The MCI values for the thiepine ring in 1c in T1 and S1 are intermediate. Interestingly, this is the same order in degree of aromaticity as found in the S0 states of pyrrole, furan, and thiophene, and should reflect the relative energy and size of the 2p or 3p lone-pair orbital of the X atom leading to differences in -orbital overlap with adjacent C atoms. It can also be noted that the aromatic character of the benzene rings in 1a -1c when the three compounds are excited decrease in the aromatic characters according to MCI. As the MCI values in the T1 and S1 states for the benzene rings are similar this further reveals the resemblance between the two states.    for 1a, 1b, and 1c, respectively, suggesting that the nitrogen-containing heterocycle is more aromatic than the others, in agreement with the MCI, FLU, and HOMA results. Yet, thiepine 1c has a slightly less negative value than oxepine 1b. By also calculating the NICS-XY scan ( Figure S13) for the C2v symmetric structure of 1c (a structure with one imaginary frequency) we can attribute the slightly lowered excited state aromaticity of 1c to the minute puckering around the S atom. The NICS plots also reveal a global aromaticity for these three molecules, similarly to dibenzocyclooctatetraene reported by Ayub et al., [33] although the central 8-electron ring makes a substantial contribution to the triplet state aromatic character.
ACID plots using the total contributions ( and ) indicate the predominance of global aromaticity in 1a -1c ( Figure 10A). Yet, the opposite pattern is revealed in the -only ACID because the global currents are weakened and the local ones in the B rings are enhanced ( Figure   10b). Thus, by removing the -contribution, which is not of relevance for the aromaticity in these compounds, one can clearly see the diatropic ring in the central 8π-electron system. The -ACID model indicates that the global aromaticity is larger for 1b and 1c, in agreement with what is found based on the HOMA values for the perimeter.  To conclude, 1a -1c are all aromatic in their T1 and S1 states. Whether the dominating aromatic cycle is the central 8-electron heterocycle or the perimeter varies to some extent between the compounds and between aromaticity index used. Plots of spin density, NICS-XY scans, HOMA and FLU indices favor the presence of a global 16-electron circuit, in a similar way that biphenylene shows a global 12-electron circuit in its T1 state. [33] Yet, the 8-electron circuit becomes dominating when regarding the -only contribution to the ACID, in line with the interpretation given by Shukla and Wan. [31] The three compounds can obviously be labelled as aromatic chameleon compounds, i.e., they can adapt their electron distribution to as to comply with the different aromaticity rules in different electronic states.

Compounds 1d -1g:
Now what is the case of the dibenzannelated compounds with several heteroatoms in the central 8-electron 7MR? Here, the non-annelated C=C double bond was changed to an N=C bond (1d -1f) or an N=N bond (1g), providing more -type lone-pairs from which excitations may occur. The first three in the set are also unsymmetric in the sense that the two benzene rings are non-equivalent.
The geometries of these four compounds in their T1 and S1 states reflect the differences between compounds with * versus those with n* states as the lowest excited states.
Compounds 1d and 1e are planar in both T1 and S1 states while 1f is slightly puckered in S1 (like 1c, both with X = S) and more distinctly so in T1 (for optimized geometries see Figure   S2) However, 1g is even more strongly puckered in both excited states. Now, with regard to the aromaticity in the excited states of 1d -1f we can note the same pattern as for the 1a -1c compounds. Starting with HOMA, 1d is the most aromatic in both the T1 and S1 states. In general, the HOMA values for all circuits of 1d, 1e, and 1f, respectively, closely resemble the corresponding HOMA values of 1a, 1b and 1c, respectively.
This applies to both the T1 and S1 states, and also 1d -1f are slightly more aromatic in their S1 states than in their T1 states when based on HOMA. When considering the electron density based indices (FLU and MCI) we can again see a strong resemblance within each of the pairs 1a/1d, 1b/1e, and 1c/1f. Finally, with regard to NICS-XY it was only run for 1d and 1f, yet these compounds in their T1 states again show a close resemblance with 1a and 1b in the T1 states, and this is also in line with the ACID plots.  yet, instead has a slight puckering at the S atom so that the two benzene rings are equivalent.
The NICS-XY scans of 2a -2c in the T1 states shows strong resemblance between the three compounds as the negative NICS values, corresponding to diatropic (aromatic) ringcurrents, are markedly localized to the B ring ( Figure 12). The benzene rings in 2b and 2c even have NICS values that correspond to weak antiaromatic character. This picture is reinforced further through the -only contributions of the ACID plot because a strong diatropic ringcurrent is visible in the ring B while in rings A and A' one can instead observe local paratropic ring-currents that merge with the diatropic ring-current of ring B in the two CC bonds between the rings (see Figure S11 for the -only ACID of 2a in the T1 state). This picture is starkly different from that of compounds 1a -1g. When considering the MCI values in the S1 and T1 states we observe that the values for the central ring are larger while the values for the two outer benzene rings are smaller than for the previous molecules in their T1 states.  and D2h symmetric molecule. [58] For the much smaller dioxin molecule we find that the structure (planar vs. nonplanar) varies extensively with method with methods including dynamic electron correlation (CASPT2 and the DFT methods) leading to a nonsymmetric structure. Whether dibenzodioxin is planar in its lowest excited states, or not, seems to require a dedicated effort focused on this molecule.

Conclusions
Many dibenzannelated 8-electron heterocyclic molecules are influenced by aromaticity both in their S0 states and in their lowest excited states (S1 and T1). In the S0 states, the two benzene rings are strongly Hückel-aromatic while in the S1 and T1 states the central 8-electron cycle and/or the 16-electron perimeter can be Baird-aromatic. The requirement for the latter is that the S1 and T1 state have * instead of n* character, and according to quantum chemical calculations this is the situation when the central cycle has either one or two heteroatoms. The compounds that are aromatic in their S1 and T1 states also adopt planar or nearly planar excited state structures.
Yet, molecules with S1 and T1 states having n* character, as found in the dibenzannelated compound with N=N bonds and in dibenzodioxin, are strongly puckered in these states and they are nonaromatic. On the other hand, compounds with isolated N atoms in the heterocycles, either as -NR-or as -N=CH-, provide for the most strongly excited state aromatic compounds, hence resembling the situation in the S0 state with pyrrole being more aromatic than furan and thiophene. A further difference exists between compounds with central seven-membered versus those with six-membered 8-electron heterocycles as the former tend to have the 16-electron perimeter as the strongest aromatic circuit while the latter ones have the central ring as the most aromatic circuit.
On the experimental side, the planarization that is observed in the excited state results in large Stokes' shifted emissions as observed by Shukla and Wan. [31] Such large Stokes' shift should indeed be general for many dibenzannelated 8-electron heterocycles.