Synthesis of yellow and red fluorescent 1,3a,6a-triazapentalenes and the theoretical investigation of their optical properties

To expand the function of the fluorescent 1,3a,6a-triazapentalenes as labelling reagents, their fluorescence wavelength was extended to the red color region.


Introduction
Fluorescent organic molecules are an important class of compounds in modern science and technology, and are widely used as biological imaging probes, sensors, lasers, and in lightemitting devices. 1 Thus, the development of useful uorescent organic molecules is crucial for the advancement of many industries, and has been a subject of intensive research. 2 In particular, small uorescent organic molecules have attracted great attention in the eld of chemical biology, because the visualization of biologically active small compounds by introducing uorophores is one of the most useful ways for studying their mechanism. 3 However, several key improvements are needed for the commonly used uorescent molecules. The most highly uorescent molecules possess a relatively large molecular size depending on the target bioactive compounds, and the uorescence-labelled molecules sometimes lose their activity as a result of the structural modications. Furthermore, oen the methods used to synthesize them do not allow for the design of systems whose luminescence properties span a wide range of wavelengths. As a potential uorescent chromophore to overcome the above problems, we have recently discovered that a 1,3a,6a-triazapentalene skeleton without an additional fused ring system is a compact and highly uorescent chromophore. 4,5 In contrast, benzotriazapentalene as an aryl-fused ring system exhibits almost no uorescence (F F < 0.001), 6 and the various related analogues of the aryl-fused 1,3a,6a-triazapentalenes 7 have not been reported to have noteworthy uorescence properties. The limited synthesis of 1,3a,6a-triazapentalenes without an aryl-fused system 8 might be the main reason that they have been previously unrecognized as excellent uorescent chromophores until our nding.
The construction of the 1,3a,6a-triazapentalene skeleton without an aryl-fused ring system was recently established in our laboratory, and the 1,3a,6a-triazapentalenes were readily prepared by the click-cyclization-aromatization cascade reaction of various alkynes with the azide 2 possessing two triates at the C2 and C3 positions (Scheme 1). 4 The click reaction of azide 2 with the alkynes produced triazole A, which underwent cyclization to give triazolium ion B. In the presence of triethylamine, the intermediate B was subsequently converted to triazapentalene 1 by a sequential reaction of E2 elimination and deprotonation (Scheme 1). This cascade reaction was conrmed to be applicable to a wide range of alkynes, and the easy access to the various 1,3a,6a-triazapentalenes was enabled. Furthermore, the 5,5-dimethoxy analog of B was found to be stable enough for isolation, and a strong base was necessary for the elimination of the methoxy group to give 5-methoxy-1,3a,6atriazapentalenes. This method was applicable to the one-pot synthesis of the various 2,5-disubstituted-1,3a,6a-triazapentalenes. 9 Although the 1,3a,6a-triazapentalenes are composed of a zwitter ion, the polarities and the electrical charges are neutralised due to the resonance stabilization of the aromatic compounds and so they are easily manipulated.
The 1,3a,6a-triazapentalenes exhibit not only intense uorescence but also various interesting uorescence properties such as an extremely large Stokes shi (Stokes shi exceeding 100 nm) 10 and large positive uorescence solvatochromism. More interestingly, the 1,3a,6a-triazapentalenes as uorescent chromophores provide an innovative uorescence system that can be tuned both in terms of the uorescence wavelength and the quantum yield by varying the 2-and 5-substituents, respectively. 4,9 For example, the uorescence of the 1,3a,6a-triazapentalenes shied to longer wavelengths due to the inductive effect of the 2-substituents. In fact, the uorescence maxima of the 2-phenyl-1,3a,6a-triazapentalene derivatives exhibited a noteworthy correlation with the Hammett s p value of the substituent on the benzene ring, as shown in Fig. 1. In contrast, the introduction of an electron donating substituent at the C5 position had little effect on the uorescence wavelength, although the enhancement of the push-pull effect on the 10 pelectron system was expected. Meanwhile, the uorescence quantum yields (F F ) were dramatically changed. In fact, the introduction of a methoxy group at the C5 position of 2-(4-cyanophenyl)-1,3a,6a-triazapentalene caused a substantial increase in F F (from 0.15 to 0.57) without having any effect on the uorescence wavelength .
Recently, emission-and/or quantum yield-tunable uorophores have received a great deal of attention as the core skeleton of uorescent probes. 11 The 1,3a,6a-triazapentalene system also provides a novel uorescent molecule that enables the same uorescent chromophore to exhibit various uorescence colors and quantum yields. However, the detailed mechanisms of the above interesting uorescence properties have not been elucidated.
To actually develop the 1,3a,6a-triazapentalenes as uorescent labelling reagents, several goals had to be met: (i) to expand the uorescence wavelength of the triazapentalenes to the red color region, (ii) to conrm that the uorescence of triazapentalene from the inside of cells is observable, (iii) to introduce binding sites, such as a succinimide ester and a maleimide moiety, as labelling reagents, and (iv) to obtain the theoretical explanation of the uorescence properties of 1,3a,6a-triazapentalene. The uorescent labels exhibiting longer emission wavelengths, such as those emitting yellow, orange, and red light, might be more suitable for the living cells and tissues due to the reduction of the light irradiation damage and the potential access to deeper tissue. However, the existing uorescent organic molecules emitting red light have several common problems, including a large molecular size and a small Stokes shi. 11,12 On the other hand, 1,3a,6a-triazapentalene is a compact uorescent chromophore exhibiting a large Stokes shi, and its uorescence wavelength can be tuned based on the inductive effect of C2-substituents. Although the uorescence wavelengths of the 1,3a,6a-triazapentalene derivatives previously reported in a preliminary communication are below the 556 nm (lime green) uorescence wavelength of 2-(4-nitrophenyl)-1,3a,6a-triazapentalene, additional introductions of electron-withdrawing groups on the benzene ring are expected to induce additional and longer wavelength shis. Thus, we became intrigued by the synthesis of 1,3a,6a-triazapentalenes Scheme 1 Single step synthesis of the 1,3a,6a-triazapentalenes (1). possessing additional electron-withdrawing groups in order to investigate the possibility of 1,3a,6a-triazapentalenes emitting yellow, orange, and red light. Herein, we describe the synthesis of 2-phenyl-1,3a,6a-triazapentalene derivatives possessing both electron-withdrawing groups and binding sites on the benzene ring, the observation of their uorescence inside cells, and the computational efforts made to provide a theoretical explanation of the uorescence properties of the 1,3a,6a-triazapentalenes (Fig. 2).
Results and discussion on the synthesis and the fluorescence properties A cyano group was chosen as the electron-withdrawing group due to its small size and excellent stability under UV irradiation. Thus, a suitable position for the introduction of the cyano group to the benzene ring was rst investigated. Treatments of 2 with the phenyl acetylene derivatives possessing a cyano group at the para 3b, meta 3c, and ortho position 3d in the presence of the CuI$ligand complex and triethylamine gave the desired triazapentalenes 1b, 1c, and 1d with yields of 77%, 87%, and 93%, respectively (Table 1).
In comparison with the para-substituent 1b, the introduction of a cyano group at the meta position (1c) induced the undesired shorter wavelength shi, although the F F value was increased to 0.24 (Table 1). 13 In contrast, the ortho-cyano analog 1d exhibited a slightly longer-wavelength shi, and the F F value was also increased (Table 1). 13 Therefore, we found that the ortho position is more suitable for the introduction of the cyano group as an additional electron-withdrawing group for the expansion of the uorescence wavelength to the yellow and red color regions. Thus, we rst tried to synthesize methyl 3-cyano-4-ethynylbenzoate 3e as an alkyne fragment. Commercially available 5-bromo-2-iodobenzonitrile 4 was converted into ethynylbenzonitrile 5 by the Sonogashira coupling reaction with tert-butyldimethylsilylacetylene. 14 The treatment of 5 with 5 mol% of Pd(PPh 3 ) 4 in methanol under a CO atmosphere produced the methyl ester 6 in quantitative yield. Finally, the removal of the TBS group gave the desired alkyne fragment 3e (Scheme 2). Next, we tried to synthesize the dicyano analog, which was expected to induce a further wavelength shi. 3,5-Dicyano-4-iodo-benzoate 7 as a starting material was obtained from the commercially available p-toluidine in 5 steps according to the procedure of Professor Gübel. 15 The Sonogashira coupling reaction of 7 with various acetylenes was initially difficult, and yielded mainly the deiodinated reductive product. 16 Aer various investigations, we found that the reaction with TBS-acetylene under the conditions of 10 mol% of Pd 2 (dba) 3 $CHCl 3 , 20 mol% of trifurylphosphine, 20 mol% of copper(I) iodide, and triethylamine in DMF at 50 C produced the desired coupling product. Finally, the subsequent treatment with TBAF and acetic acid gave the alkyne fragment 3f.  Next, the cascade reaction leading to the production of the 1,3a,6a-triazapentalenes was applied to the prepared alkynes 3e and 3f. The treatment of 3e with 1.2 equiv. of 2 in the presence of 5 mol% of the CuI$ligand complex and triethylamine produced the desired 1,3a,6a-triazapentalene 1e with a yield of 71%. The similar click reaction of 3f also proceeded smoothly to give 1f with a yield of 72%. Furthermore, the comparative analog 1g, which did not possess cyano groups, was also synthesized with a yield of 73% from methyl 4-ethynylbenzoate (3g). Having prepared the desired 1,3a,6a-triazapentalenes 1e, 1f, and 1g, their uorescence properties were examined ( Table  2). Since these three compounds were only slightly soluble in water due to the lipophilicity of the benzene ring, their uorescence spectra were measured in deaerated dichloromethane. The standard analog 1g exhibited a high uorescence quantum yield (F F ¼ 0.44) 13 and green emission (l max em ¼ 510 mm) as predicted from the Hammett s p value of the methyl ester on the benzene ring. As we expected, the mono-cyano analog 1e showed a noteworthy longer-wavelength shi of the uorescence maximum from 510 nm of 1g to 572 nm, and 1e emitted yellow light. Although the uorescence quantum yield (F F ) of 1e was slightly decreased to 0.34, 17 this value was still within the range required for an effective uorescent labelling reagent. Furthermore, the uorescence maximum of the di-cyano analog 1f shied to a still longer-wavelength region (632 nm), and 1f exhibited red uorescence. Therefore, the introductions of the cyano groups were found to induce an approximately 60 nm longer shi of the uorescence maximum in each case, and the development of yellow and red uorescent 1,3a,6a-triazapentalenes was accomplished.
It was especially noteworthy that these long-wavelength uorescent molecules exhibited large Stokes shis, such as the 152 nm shi of 1e and the 166 nm shi of 1f, despite there having been few prior examples of the long-wavelength ('550 nm) organic uorophores exhibiting such large (mega) Stokes shis, 18 since such shis were useful for suppressing the action of background uorescence in the various uorescence analyses. In addition, the molecular sizes of 1e and 1f were considerably smaller in comparison with the conventional yellow and red uorescent molecules. Therefore, the 1,3a,6atriazapentalenes might be practical uorescent chromophores for use as molecular probes to cover the entire region of visible wavelengths, although the further shi toward longer wavelengths is still needed. Furthermore, 1g and 1e exhibited uorescence emission in the solid state with a uorescence maximum similar to that observed in the solution of dichloromethane, whereas the uorescence of 1f in the solid state was not detected.
On the other hand, the extinction coefficient (3) of 1g at 376 nm was 1230 dm 3 mol À1 cm À1 , and this value still needed to be increased for a more bright uorescent reagent. Although the 3 value at 287 nm was 13 800 dm 3 mol À1 cm À1 , at a practical level, this region (ultraviolet) is not a suitable excitation light for imaging probes. Similarly, the extinction coefficients (3) of 1e and 1f in a visible light region were also not high at 630 dm 3 mol À1 cm À1 (420 nm) and 1580 dm 3 mol À1 cm À1 (466 nm), respectively. Therefore, the improvement of the extinction coefficient (3) was the next challenge for the development of more useful bright uorescent labels. So far, we have already found that the introduction of a substituent at the C4 position dramatically increases the 3 value. For example, 4-phenyl analogs of 1g showed a substantial increase in the 3 value from 1230 dm 3 mol À1 cm À1 (376 nm) for 1g to 22 600 dm 3 mol À1 cm À1 (345 nm) with comparable F F values. The 4-phenyl analog of 1e also exhibited a practical 3 value of 4560 dm 3 mol À1 cm À1 (432 nm) and 38 000 dm 3 mol À1 cm À1 (336 nm), although the F F value was decreased to 0.07. 19 Further investigation of 4substituents for the design of practical uorescent labelling reagents is currently underway in our laboratory. Furthermore, the uorescence solvatochromism of 1e was examined. The uorescence spectra of 1e in several solvents are shown in Fig. 3. Basically, the uorescence of 1e shied to the longer wavelength with the Stokes shi being increased by an increase in the solvent polarity from benzene (546 nm) to acetone (645 nm). On the other hand, its uorescence in methanol shied inversely to the shorter wavelength (l max em ¼ 463 nm). Furthermore, since 1e was only slightly soluble in water, its uorescence in water was also measured. The uorescence shied to 476 nm similarly to the uorescence shi observed in methanol. The uorescence quantum yield (F F ) in water was substantially decreased to a value of 0.013. Therefore, the 1,3a,6a-triazapentalenes are expected to change their uorescence wavelength and intensity according to the hydrophobic environment in the cells.
Next, we investigated the applicability of the long-wavelength uorescent 1,3a,6a-triazapentalenes as uorescent probes in a biological system. Since the di-cyano analog 1f was not very stable under UV irradiation and its F F was lower (F F ¼ 0.096) than that of 1e (F F ¼ 0.34), 17 the mono-cyano analog 1e was adopted for this purpose. Thus, HeLa cells were treated with a solution of 1e (10 mM in 0.02% DMSO) 20 and monitored in the 572-642 nm wavelength region. As shown in Fig. 4, the uorescent staining of HeLa cells was successfully observed without washing the cellular medium. The living HeLa cells were clearly visualized as observed using a uorescence microscope, whereas the interiors of the control cells, which were treated with DMSO, were not stained. Since the active uptake of the uorescent 1e by living cells and the uorescence solvatochromism of 1e enhance the uorescence contrast between the cells and the background, it was not necessary to x the cells. Furthermore, a cytotoxic effect on the cells was not iden-tied over the observation period, suggesting that the triazapentalene is suitable for connecting to small biofunctional molecules as a uorescent label. This is the rst experimental evidence that the 1,3a,6a-triazapentalene is applicable to the life sciences eld as a uorescent reagent. Further detailed investigations into the localization and quantitative analysis of 1e inside cells are currently underway in our laboratory.
The actual uorescence observation of 1e inside cells encouraged us to develop 1e as a uorescent labelling reagent. Thus, the conversion of the methyl ester moiety into the Nhydroxysuccinimide ester as a binding site was attempted. The treatment of 1e with 1.2 equiv. of lithium hydroxide afforded carboxylic acid 8, which was directly used for the next condensation reaction. However, although the condensation reaction proceeded smoothly, the removal of the urea analogs generated from the condensing reagent was not straightforward due to the instability of the succinimide moiety of 9. Finally, polymersupported DCC was adopted as a useful condensing reagent to remove the urea by ltration, and the subsequent recrystallization gave the puried 9 in a 60% two-step yield. Having prepared the uorescent labelling reagent 9, the introduction of 9 into amino acids was examined (Scheme 3). The treatment of 9 with glycine ethyl ester in DMF produced labelled glycine 10 in 95% yield. The uorescence-labelled 10 exhibited yellow emission (l max em ¼ 567 nm) with a high quantum yield (F F ¼ 0.37) 17 in  deaerated dichloromethane. Furthermore, the introduction of the tri-peptide Gly-Pro-Leu was also examined, and the labelled tri-peptide 11 was obtained in 82% yield. The uorescence observation of 11 showed a uorescence maximum at 567 nm and an acceptable uorescence quantum yield (F F ¼ 0.24) 17 in deaerated dichloromethane. Therefore, the development of the 1,3a,6a-triazapentalene as a compact uorescent labelling reagent emitting yellow-red light was achieved. Furthermore, although the labelled glycine 10 and tri-peptide 11 were dissolved well in an organic solvent, 21 their uorescence properties in water were also measured. Since the emission maxima of 10 and 11 shied to shorter wavelengths with similar absorption maxima, the Stokes shis became small in water as in the case of 1e. The uorescence quantum yields (F F ) were also reduced to 0.019 (10) and 0.077 (11). 17 These changes in the uorescence properties according to the polarity of the environment might make the 1,3a,6a-triazapentalene useful as a uorescent probe in vivo measurements.
Theoretical investigation of the optical properties of the 1,3a,6a-triazapentalenes In our preliminary communication, we rst reported that the 1,3a,6a-triazapentalene skeleton without an additional fused ring system is a compact and highly uorescent chromophore. However, the detailed mechanisms of the uorescence have not yet been elucidated. In this work, quantum chemical calculations were performed to investigate the optical properties of the 1,3a,6a-triazapentalenes. Most of the theoretical calculations for the optical properties of dye molecules utilize the time-dependent density functional theory (TD-DFT), but in this work the high-level wavefunction-based approach using the complete active space second-order perturbation theory (CASPT2) method are also employed to provide a more reliable description of the excitation energies. The following synthetic 1,3a,6a-triazapentalenes were examined as the model substrates in this investigation: unsubstituted 1,3a,6a-triazapentalene 1a as a basic structure, 2-(4-cyano)phenyl derivative 1b as a standard analog described in the previous communication, and synthetic 1g, 1e, and 1f as described in this article.

Computational details
The equilibrium geometry in the electronic ground state (S 0 ) is determined by the density functional theory (DFT) calculations using the B3LYP functionals, while the geometry optimization in the lowest pp* excited state S 1 (pp*) is performed by the time-dependent DFT (TD-DFT) calculations employing the coulomb attenuated B3LYP (CAM-B3LYP) functionals. 22 The C s symmetry constraint is imposed for 1a, 1b, 1g, and 1e, while no constraint is applied for 1f because the twisted structure is more stable due to the steric hindrance. The choice to employ the CAM-B3LYP functionals is due to the signicant charge-transfer character involved in excitation to the S 1 state. The 6-31 + G(d,p) basis set is used in the DFT calculations and the equilibrium geometries are determined both in the gas phase and in dichloromethane. The solvent effects are taken into account by the polarizable continuum solvation model (PCM), 23 where the radii are taken from the universal force eld. 24 Aer the geometry optimization, the vertical excitation and uorescence energies are calculated at the S 0 and S 1 equilibrium structures (denoted as (S 0 ) min and (S 1 ) min ), respectively, by the TD-DFT(CAM-B3LYP) method. In PCM calculations, the linearresponse method with a non-equilibrium solvation is employed to obtain the vertical excitation energies at (S 0 ) min , while the equilibrium solvation is adapted for the calculation of the excitation energies during the S 1 geometry optimization. The excitation energy is also rened at the DFT-optimized geometries by the CASPT2 (ref. 25) method in order to obtain more reliable excitation energies. A level shi with a value of 0.3 is applied for the CASPT2 calculations. 26 The notation of CASPT2 (m,n) is occasionally used, in which case the active space for a reference state-averaged complete active space selfconsistent eld (SA-CASSCF) wavefunction is composed of m electrons and n orbitals (SA-CASSCF (m,n)). The augmented correlation-consistent polarized double-zeta basis set (denoted as aug-cc-pVDZ) is employed in the CASPT2 calculations. For obtaining the oscillator strengths, the vertical excitation energies calculated by CASPT2 and the transition dipole moments calculated by SA-CASSCF are used.
For 1a, the active space for the reference SA-CASSCF wavefunction is comprised of six p orbitals (four p orbitals are doubly-occupied and two are unoccupied in the closed-shell conguration), and it is therefore denoted as SA-CASSCF (8,6). 1a possesses ten p orbitals and the lowest and highest p orbitals are excluded from the active space. This is justied by the larger active space calculation, which includes all p orbitals (which corresponds to SA-CASSCF (10,8), and the active orbitals at (S 0 ) min are shown in the ESI as Fig. S1 †), where only a difference of $0.01 eV is observed in the S 1 vertical excitation energies. The active space for the other chromophores is composed of twelve electrons distributed in ten p orbitals (SA-CASSCF(12,10)), and the active orbitals of 1b at (S 0 ) min are shown in Fig. S2. † As seen in the gure, the active space of the SA-CASSCF(12,10) wavefunction includes orbitals that correspond to the active orbitals of SA-CASSCF (8,6) in 1a. For all chromophores, the S 0 and S 1 states are averaged with equal weights in the SA-CASSCF calculations, except where otherwise noted.
The DFT and TD-DFT calculations are performed using the Gaussian09 program package 27 while the CASPT2 calculations are carried out using the MOLPRO2010.1 program package. 28 Results and discussion on the optical properties We begin by investigating the character of the excited states of 1a and 1b at (S 0 ) min in the gas phase, followed by the results and discussion on the optical properties of the other chromophores in the gas phase and in dichloromethane.

Simple 1,3a,6a-triazapentalene (1a)
The S 0 and S 1 equilibrium structures of 1a in the gas phase are shown in Fig. 5, along with the bond lengths and the atomic numbering (note that this numbering is different from the previous sections and is only used in the theoretical section). The signicant changes in geometry upon photo-excitation involve the bond elongation of N3-C6 (1.370 / 1.411Å) and N1-N2 (1.344 / 1.376Å).
The vertical excitation energies to the low-energy-lying pp* states are shown in Table 3, where in the CASPT2 calculation the S 0 and lowest three pp* states are averaged with equal weights in the reference SA-CASSCF(8,6) wavefunction. It is noted that, although a couple of np* states are found between these pp* states in the TD-DFT calculations, it is conrmed that the lowest-energy singlet excited-state is characterized by the pp* excitation, and therefore only the pp* states are examined in this investigation. The lowest pp* excited-state, S 1 (pp*), is viewed as the HOMO-LUMO transition (see the natural orbitals in Fig. S1 (ESI) †) and the CASPT2 excitation energy of 4.33 eV (286 nm) is in good agreement with the experimental value of 4.31 eV (288 nm), even though the experimental measurements are performed in dichloromethane. The second pp* excited state is characterized by the HOMO / LUMO+1 transition, and it lies close to the rst pp* state in the CASPT2 calculation. The natural charges of the S 0 and S 1 states at (S 0 ) min and their differences are shown in Fig. S3. † 2. 2-(4-Cyano)phenyl-1,3a,6a-triazapentalene (1b) The S 0 and S 1 equilibrium structures of 1b in the gas phase are shown in Fig. 6, along with the bond lengths and atomic numbering. The transition to the S 1 state involves the bond elongation of C7-C8 (1.394 / 1.444Å) and shortening of the central C8 -C13 bond (1.469 / 1.421Å).
The vertical excitation energies to the low-energy-lying pp* states are shown in Table 4. In the CASPT2 calculation, the S 0 and the lowest three pp* states are averaged with equal weights in the reference SA-CASSCF(12,10) wavefunction. The vertical excitation energies to the S 1 (pp*) state are 3.85 (322 nm) and 3.25 eV (381 nm) for the TD-DFT and CASPT2 calculations, respectively, and the CASPT2 excitation energy is in remarkably good agreement with the experimental value of 3.25 eV (381 nm) (note again that the experimental measurements are performed in dichloromethane). Excitation to the S 1 state is characterized by the HOMO / LUMO transition (see Fig. 7 and also Fig. S2 †), and as expected from the shape of the two relevant orbitals, the S 1 transition involves charge transfer from the 1,3a,6a-triazapentalene skeleton to the substituted phenyl ring. This is clearly seen from the large dipole moment in the S 1 state (19.47 debye) compared to that of the S 0 state (7.11 debye) at (S 0 ) min (see Table S1 †). The charge-transfer character of S 1 is also clear from the natural charges, where the sums of the natural charges in the 1,3a,6a-triazapentalene skeleton (atoms from N1 to H12) are 0.022 and 0.542 in the S 0 and S 1 states, respectively (see also Table S1 †). Since the S 1 state exhibits a charge-transfer character, it may be possible to observe the twisted intramolecular charge transfer (TICT) state involving the rotation of the phenyl ring around the central C8-C13 bond. In order to check this, we performed frequency analysis at (S 1 ) min and conrmed that the planar geometry is the minimum energy structure in the S 1 state.
As seen in Table 4, the second and third pp* states can be described as a mixing of two congurations, HOMO / LUMO+1 (9p / 2p*) and HOMOÀ1 / LUMO (8p / 1p*). It is noted that the HOMO / LUMO+1 (9p / 2p*) transition corresponds to the S 0 -S 1 excitation of 1a, while the S 0 -S 1 transition of 1b corresponds to the excitation to the second pp* state of 1a (see the natural orbitals given in Fig. S1 and S2 †). Therefore, the electronic character of the S 1 state is different between 1a and 1b.
3. Green uorescence (1g), yellow uorescence (1e), and red uorescence (1f) derivatives and comparison with the experimental results The optimized structures of 1f in the S 0 and S 1 states are shown in Fig. 8, where the dihedral angle of d(C7-C8-C13-C15) representing twisting of the phenyl ring around the central C8-C13 atoms is 40.6 degrees at (S 0 ) min , and slightly decreases to 29.7 degrees at (S 1 ) min . The other chromophores (1g and 1e) maintain the planar geometry, and the Cartesian coordinates of the optimized structures are given in the ESI. † Excitation to the S 1 state involves the HOMO / LUMO transition, and all chromophores (1g, 1e, and 1f) exhibit the same charge-transfer character. The vertical excitation and uorescence energies are summarized in Table 5 and Table 6, respectively. We note here that in this table a slight discrepancy is found in the S 1 (pp*) vertical excitation energies of CASPT2 for 1a and 1b with respect to the values shown in Table 3 and 4, since in Table 5 only the S 0 and S 1 states are averaged with equal weights in the reference SA-CASSCF wavefunction.
The CASPT2 calculations are performed only in the gas phase, and therefore we estimate the excitation energies in dichloromethane using the solvatochromic shis of TD-DFT calculations (the estimated values are shown in parenthesis). Fig. 9 shows the comparison of absorption and uorescence wavelengths between the theoretical calculations and experimental results. Although the calculated uorescence wavelengths are shorter than the experimental values, the gure Table 3 The vertical excitation energies (DE in eV and nm) and oscillator strengths (f in a.u.) of 1a for the low-lying pp* states at (S 0 ) min a     clearly demonstrates a good correlation between the two values. The overestimation of the uorescence energies may be attributed to the insufficient treatment of the solvent environments, because excitation involves a signicant charge-transfer character. The explicit treatment of the solvent molecules in the framework of the QM/MM approach or the state-specic approach 29,30 would be appropriate for a more quantitative description of the uorescence energies.
In the ESI (Tables S1 and S2 †), the sums of the natural charges in the 1,3a,6a-triazapentalene skeleton and the dipole moments in the S 0 and S 1 states at (S 0 ) min and (S 1 ) min are given for all chromophores. It is noteworthy that there is a clear correlation between the wavelengths and the natural charges (also the dipole moments) in the S 1 state, where a larger charge separation induces longer absorption and uorescence wavelengths. It is also noted that the absorption and uorescence wavelengths are longer when measured in dichloromethane than when measured in the gas phase because the chargetransfer state is more stabilized in polar solvents.
Finally, we comment that the CASPT2 method is more reliable than the TD-DFT approach, but the computational cost is much more expensive. As seen in the present work, the TD-DFT method predicts slightly higher excitation energies than those by CASPT2, but the correlation with experimental results is surprisingly good. Therefore, for chromophores of a larger size, where the computational costs of CASPT2 calculations are prohibitive, the TD-DFT method can be reliably used to predict the optical properties of 1,3a,6a-triazapentalenes.

Conclusions
The uorescence wavelengths of 1,3a,6a-triazapentalenes were extended to the red color region. Based on the noteworthy correlation of the uorescence wavelength with the inductive effect of the 2-substituent, electron decient 2-(2-cyano-4-methoxycarbonylphenyl)-1,3a,6a-triazapentalene and 2-(2, 6-dicyano-4-methoxycarbonylphenyl)-1,3a,6a-triazapentalene were synthesized. They exhibited yellow and red uorescence and a large Stokes shi respectively, and the 1,3a,6a-triazapentalene system enabled the same uorescent chromophore to cover the entire region of visible wavelengths. The potential applications of the 1,3a,6a-triazapentalene system as uorescent probes in the elds of the life sciences were Table 5 The vertical excitation energies (DE in eV and nm) and oscillator strengths (f in a.u.) for the S 1 state calculated by TD-DFT and CASPT2 at (S 0 ) min

DE (eV)
DE (   investigated, and the 1,3a,6a-triazapentalene system was clearly proven to be useful as a uorescent reagent for living cells. The N-hydroxysuccinimide ester derivative of yellow uorescent 1,3a,6a-triazapentalene as a compact labelling reagent was conrmed to be able to readily label the amino group. Finally, quantum chemical calculations were performed to investigate the optical properties of the 1,3a,6a-triazapentalenes. These calculations revealed that excitation involves signicant chargetransfer from the 1,3a,6a-triazapentalene skeleton to the 2substitutent. The calculated absorption and uorescence wavelengths showed a good correlation with the experimental ones, which allows us to design substituents that exhibit the desired optical properties.