Spin–Vibronic Control of Intersystem Crossing in Iodine-Substituted Heptamethine Cyanines

Spin–orbit coupling between electronic states of different multiplicity can be strongly coupled to molecular vibrations, and this interaction is becoming recognized as an important mechanism for controlling the course of photochemical reactions. Here, we show that the involvement of spin–vibronic coupling is essential for understanding the photophysics and photochemistry of heptamethine cyanines (Cy7), bearing iodine as a heavy atom in the C3′ position of the chain and/or a 3H-indolium core, as potential triplet sensitizers and singlet oxygen producers in methanol and aqueous solutions. The sensitization efficiency was found to be an order of magnitude higher for the chain-substituted than the 3H-indolium core-substituted derivatives. Our ab initio calculations demonstrate that while all optimal structures of Cy7 are characterized by negligible spin–orbit coupling (tenths of cm–1) with no dependence on the position of the substituent, molecular vibrations lead to its significant increase (tens of cm–1 for the chain-substituted cyanines), which allowed us to interpret the observed position dependence.


■ INTRODUCTION
Heptamethine cyanines (Cy7) are small organic chromophores with strong absorption and emission in the near-infrared (NIR) region that are used in diverse biological applications, such as fluorescence probes, 1 pH 2 and metal cation 3−5 sensors, and DNA 6 and protein 7 markers, or for tumor visualization 8 and photocaging. 9,10 The best-known Cy7 derivative, indocyanine green (ICG), is a fluorescent probe approved by the Federal Drug Administration (FDA) 11 for various clinical applications. 12 Some Cy7 derivatives, bearing heavy halogen atoms�usually on the indole core�have been reported to act as photosensitizers. 13−15 Heavy atoms enhance intersystem crossing (ISC) due to strong spin−orbit coupling between the singlet and triplet states (heavy-atom effect, HAE). 16,17 The ISC enhancement can also be achieved by introducing a covalently-linked radical species, such as 2,2,6,6-tetramethyl-1piperidinyloxyl (TEMPO), 18 J-aggregation, 19 or charge transfer. 20,21 Electron-donating substituents in the C4′ positions or electron-withdrawing groups in the C3′ position of the Cy7 chain were also shown to improve ISC, but the effect is relatively small. 22 Triplet-excited Cy7s can be used as singlet oxygen generators in organic synthesis, 23 wastewater treatment, 24 or photodynamic therapy. 25 Some of us have recently introduced a new approach to the synthesis of Cy7 derivatives substituted in the C3′−C5′ positions of the chain under mild conditions using a Zincke salt ring-opening reaction. 26 This allowed us to investigate how the chain substitution modulates their photophysical properties, such as quantum yields of singlet oxygen formation, photodecomposition, and emission. 22 It was demonstrated that the C3′-iodine substitution significantly increases the quantum yield of singlet oxygen production (Φ Δ ). 22 The cyanine dyes have also attracted the attention of theory; the electronic structure of cyanine dyes and their excited states have been modeled by many research groups for over three decades. Models ranging from the simplest yet surprisingly accurate particle-in-a-box model 27 to current electronic structure models (typically at the density functional theory (DFT) level) have been used to explain the substituent and strong vibronic effects responsible for the characteristic asymmetric shape of the electronic spectra of cyanines. Despite the effort, the quantitative modeling of these spectra is still a challenge, 28−36 as standard DFT methods fail to reproduce the exact position of the absorption bands. The reason for this unsuspected failure has been addressed by various research groups. 29,36−39 Interestingly, the electronic states of interest show no significant multi-reference character, the overlap of the involved frontier orbitals is large, and the charge-transfer character is not present. Yet, the accurate peak position is not accessible via widely used DFT methods. From a qualitative point of view, inaccurate DFT energies are related to a particularly strong reorganization of the electron density upon excitation. 29,36,38 ISC in excited cyanine dyes has been studied to a much lesser extent, and the treatment was typically insufficient to understand the photophysics.
In this work, we synthesized several iodine-substituted Cy7 derivatives 1 (Figure 1) to study the effects of the substituent positions on ISC and singlet oxygen production in methanol and aqueous solutions to evaluate their potential application as photosensitizers. To interpret the experimentally observed results, we used quantum chemical methods focusing on the interaction between spin−orbit coupling and vibrational motion. The spin−vibronic coupling proved to be essential for a correct understanding of the ISC phenomena. 40 ■ RESULTS AND DISCUSSION Synthesis. Iodine atom-substituted heptamethine cyanines 1 were synthesized via ring-opening of the corresponding Zincke salts according to the reported procedure ( Figure 1, Scheme 1). 26 This methodology allowed us to prepare compounds 1a−f containing iodine atoms in both the heptamethine (C3′ position) and 3H-indolium moieties. 3H-Indole derivatives 2a−c were synthesized from the corresponding hydrazines via the Fischer indole synthesis, and substituted indolium compounds 3a−d were obtained by alkylation of the nitrogen atom. The other precursors, N-2,4-dinitrophenylpyridinium (Zincke) salts 5a,b, were prepared by the reaction of pyridine or 3-iodopyridine with 2,4-dinitrophenyl tosylate 4. Finally, the synthesis of derivatives 1a−f was carried out via ring-opening of an electron-deficient pyridinium core using 4bromoaniline and the subsequent condensation with the corresponding indolium heterocycle. For 1d,e, 4-bromoaniline as a nucleophile provided cyanines only in low yields; higher chemical yields (86 and 54%, respectively) were obtained using 4-aminobenzonitrile.
In general, Cy7 derivatives are known to exhibit low fluorescence quantum yields (Φ f ), 22,48,49 especially in water, 50 although they still exhibit high molecular brightness (Φ f ε max ) thanks to the large ε max values. Iodo substitution led to the expected suppression of fluorescence by an order of magnitude thanks to enhanced ISC in both solvents (see later). For example, Φ f ∼ 0.08 found for 1b and 1d in methanol is lower than that of unsubstituted Cy7 1a (Φ f = 0.24 47 ) by a factor of 2.5. Fluorescence was further suppressed in PBS (Φ f = 0.04− 0.07), the values which are comparable to that of 1a in water (Φ f = 0.06). 47 Characterization of the Excited States. To understand the photophysics of the studied dyes, we first had to characterize the relevant electronic states. The electronic absorption spectra of 1a−d were calculated at the timedependent density functional theory (TDDFT) level (full TDDFT and with the Tamm−Dancoff approximation, TDA) with the CAM-B3LYP functional. In all cases, the vertical electronic absorption is dominated by the transition between the highest occupied (HOMO) and lowest unoccupied (LUMO) molecular orbitals to give the S 1 state, which exhibits the highest oscillator strength values, as also found for differently substituted cyanine dyes before. 22 Both HOMO and LUMO are delocalized over the molecule. The vertical excitation energies are collected in Table 2; the corresponding molecular orbitals (for 1d) are depicted in Figure 3. The first two triplet states have a mixed HOMO−1 → LUMO and    HOMO → LUMO + 1 character. The experimentally observed vibronic shoulder of cyanine dyes ( Figure 2) has already been interpreted: the 0−0 transition dominates the transition. 22,29,32,37−39, 51 We also compared the values calculated with and without TDA ( Table 2). The values for singlet states obtained within TDA are systematically higher because cyanines are systems where de-excitation for a linear response of the density matrix is important. As pointed out by many authors before, 29,32,35,39 the first π → π* singlet excitation energies are systematically overestimated by TDDFT methods (compare the data in Table 2 and the experimental data in Table 1). The agreement can be improved by techniques proposed in many previous works. 28−36 However, in this work, we do not focus on quantitative modeling of the absorption spectra but on understanding ISC and the related rates and the dependence of ISC on vibrations.
In contrast to the singlet states, TDA provides higher excitation energies for the triplet states that are supposedly more accurate. 52 The TDA values of the triplet excitation energy for the T 1 state are also closer to the ΔSCF values. The S/T energy gap was studied in detail by Zhekova and coworkers,39 who showed that the gap is systematically exaggerated by TDDFT methods independently of the DFT functional used. Table 2 shows that the S 1 /T 1 energy gap in molecules 1a−d is approximately 1 eV. Considering only vertical excitation, we suggest two possible ISC channels: (i) a nonradiative transition from bright S 1 to T 1 by spin−orbit coupling of S 1 to higher vibrational levels of T 1 , or (ii) spin− orbit coupling to the higher triplet state T 2 followed by a rapid internal conversion from T 2 to T 1 . In both scenarios, the most important factors are the spin−orbit coupling (SOC) matrix elements (SOCMEs) and the energy gap between the given pair of states.
We can obtain more information about the plausible mechanism by analyzing the S 1 state. The optimized geometries are collected in Table S3, and their overlap with the corresponding ground state structures is provided in Figure  S46. As can be inferred from the structures, the geometry changes are relatively small in all cases; the changes are most profound in 1a and 1c. The TDDFT excitation energies at the S 1 minimum are listed in Table S4. The T 2 state energy is systematically lower than the S 1 energy for the S 1 minimum structure. Based on the energies of the states, we assume that after excitation to S 1 , the most relevant ISC channel involves SOC to the higher triplet state T 2 , followed by a rapid internal conversion from T 2 to T 1 . The states primarily relevant for the ISC are the S 1 and T 2 states.
Production of 1 O 2 and Photochemical Stability. Triplet-excited cyanines sensitize oxygen ( 3 O 2 ) to give singlet oxygen ( 1 O 2 ) and other reactive oxygen species. 53 The quantum yield of 1 O 2 production (Φ Δ ) is related to an ISC efficiency (Φ isc ); the heavy atom substitution of a dye usually improves Φ isc but also shortens the triplet lifetime, 54 causing reduced quenching by O 2 . The iodine chain substitution in 1b led to an increase in Φ Δ in methanol (0.12) by more than an  The Journal of Organic Chemistry pubs.acs.org/joc Article order of magnitude compared to Φ Δ = 0.0095 found for unsubstituted derivative 1a (Table 3). On the other hand, only a marginal increase (Φ Δ = 0.018) was observed for 1c with the I-substituted indolium core. A higher efficiency found for doubly I-substituted 1d indicates the additivity of HEA (Φ Δ = 0.13), as also observed before. 55 Water-solubilizing groups in 1e and 1f did not affect the Φ Δ values significantly. The effect of an iodide counterion on the 1 O 2 formation has been reported to be negligible. 22, 56 We did not observe any significant decrease in the Φ Δ values for water-soluble derivatives 1e and 1f in PBS (pH 7.4, c = 10 mM, I = 100 mM) solutions (Table 3). Indeed, polar Cy7 substituents have been shown by Burgess and coworkers to affect Φ Δ only marginally. 57 Their work reported that a Cy7 derivative bearing two iodine atoms in the indolium groups and one meso (C4′) chlorine atom exhibits Φ Δ in the range of 0.59−0.79, which are significantly higher values than the maximum Φ Δ found for C4′/indolium-substituted derivatives studied in our work. Those Φ Δ values were calculated using ICG as a reference sensitizer with Φ Δ (PBS) = 0.077. However, this number is an order of magnitude higher than that measured by Pandey and coworkers (Φ Δ (ICG, methanol) = 0.008). 58 In such a case, the Φ Δ values for the reported meso (C4′) chlorine Cy7 derivatives 55 would be similar to those reported in this work (Table 3), although we cannot directly compare Φ Δ values obtained in methanol and PBS. Because of this controversy, we chose methylene blue as a reference sensitizer 59,60 for our study.
Cyanine dyes are known to be chemically degraded (photobleaching) by regioselective oxidative fragmentation with singlet oxygen 61 or via an electron transfer mechanism. 62,63 The extent of photodecomposition depends, among others, on the type of solvent and the length of the cyanine chain. Cy7 derivatives tend to undergo photobleaching more efficiently than the shorter Cy5 or Cy3 analogues. 64 To evaluate the photostability of Cy7 derivatives, quantum yields of decomposition (i.e., disappearance; Φ dec ) were measured under different experimental conditions ( Table 3). The Φ Δ values were two to three orders of magnitude smaller than those in methanol and water; compound 1f was the most reactive in methanol but relatively persistent in water. It has already been reported that cyanine polar groups increase their photostability. 65 The results in Table 3 demonstrate that Cy7, in general, are relatively resistant to the presence of 1 O 2 .
HRMS analyses of the irradiated mixtures of 1b and 1c in methanol suggested that the major degradation process detectable by HRMS is monodeiodination (Figures S50 and S51; a partial loss of iodine was also found to occur in the HRMS analyses of 1c). Photoinduced homolysis of the sp 2 carbon−halogen bond has been known since the 1960s. 66 Homolysis requires the energy of the productive excited state to be greater than the corresponding bond dissociation energy (BDE). 16 The C−I BDE of ∼65 kcal mol −167 is too high for the Cy7 triplet energies (∼29 kcal mol −1 ; Table 2), but the S 1 energy of ∼58 kcal mol −1 ( Table 2) can be sufficiently high. Therefore, in addition to photobleaching related to the reaction of the heptamethine chain with 1 O 2 resulting in a complex chain degradation (see above), deiodination occurs from either a singlet excited state or via an alternative mechanism other than direct homolysis. 66 Because photodegradation of I-containing cyanines 1 is much less efficient than sensitization (Table 3), we did not study these processes further.
Bimolecular rate constants of photooxygenation of Cy7 with 1 O 2 (k r ), produced by rose bengal oxygen sensitization, were measured in aqueous solutions to evaluate the specific reactivities of cyanines (Table 3). DMSO was added to the solutions to increase the solubilities of Cy7 derivatives and the sensitizer ( Figure S3c; see above). The relatively high rate constants cannot be directly compared because the DMSO content affects the 1 O 2 lifetime; 68 however, we can conclude that improved solvation of the delocalized Cy7 cation in water enhances the nucleophilicity of cyanine toward the addition of 1 O 2 , as also previously reported. 22 ISC and Spin−Vibronic Effects. We show below that the experimentally observed 1 O 2 production rates (Table 3) are not consistent with the calculated values of the SOCs in the optimal geometry. However, the molecular vibrations accessible even within the zero-point motion enhance the averaged value of the SOCs by an order of magnitude (leading to an increase in the ISC rate by two orders of magnitude) due to orbital mixing. This effect is called spin−vibronic coupling. In the following paragraphs, we explain this phenomenon in detail.
We can start with Fermi's golden rule, expressing the rate of population transfer k as where Ψ i and Ψ f are the wave functions of the initial and final states, Ĥi f is the Hamiltonian describing the coupling, and E i and E f are energies of the initial and final states, respectively. For ISC, the Hamiltonian ĤS O describes SOC. If the SOC is independent of vibrational motion, the equation simplifies to where |⟨υ fk | υ ia ⟩| 2 is the measure of the vibrational state overlap. Within this approach, we can calculate the SOCMEs as a qualitative estimate of the ISC rate constants and propose that SOC is driven by the electronic character of the states as suggested by El-Sayed rules (the coupling is effective if the spin change is accompanied by a change in angular momentum). 69,70 Therefore, we calculated the SOCMEs for 1a−d in their optimal ground state structures (Table 4) and the S 1 optimized structures (Table S5). The calculated values are small (maximum tenths of cm −1 ) and almost the same for all studied Cy7 derivatives in both the ground state and the S 1 state minima. Another important factor in eq 2 is the energy gap between the states, which is approximately the same for all studied molecules (Table 2 and Table S4). Apparently, the calculated data for the structures in the ground and S 1 state minimum geometries, but neither SOCMEs nor energy gap, can explain an order of magnitude increase in the quantum yield of singlet oxygen production observed for 1b and 1d. However, as Albrecht suggested in his work, 71 there are other mechanisms for mixing the triplet and singlet states, such as vibrational SOC, in which the size of SOCMEs depends on the motion along a particular vibrational coordinate q i . 40 The spin−orbit interaction, including the vibrational SOCs, can be calculated, for instance, within the framework of the first-order perturbation theory as where ĤS O is the SOC operator and Ψ S and Ψ T are the wave functions of the singlet and triplet states. Along these lines, we calculated the SOCMEs for all vibrational coordinates in the ground-state geometries (see Figure 4 for T 2 and S 1 and Figure  S47 for T 1 and S 1 ). Figure 4 shows the changes of SOCMEs along all 186 vibrational modes for 1a−1d; the point dq = 0 corresponds to the optimized structure. The SOCMEs are very small (<1 cm −1 ) for the optimal structures 1a−1d. The motion along several vibrational coordinates is associated with a substantial increase in the SOCMEs for 1a−1d. However, the increase is an order of magnitude higher only for 1b and 1d, that is, for molecules bearing the iodine substituent attached to the heptamethine chain. The out-of-plane vibrations ω 73 for 1b (950.23 cm −1 ) and ω 80 for 1d (941.93 cm −1 ; Figure 5) show the most significant SOCMEs values (at the CAM-B3LYP/ def2-SVP level). Note that the displacements in dq on the order of several units are accessible within the zero-point energy motion, irrespective of the vibrational frequency. A significant increase in SOCMEs connected to the out-of-plane  The Journal of Organic Chemistry pubs.acs.org/joc Article vibrational modes was also reported for porphyrins 72,73 or psoralene. 74 In the optimal geometry, the iodine orbitals are perpendicular to the molecular plane and do not contribute to the frontier molecular orbitals (Figure 3). Once the system is distorted out of the plane, the iodine orbitals mix in the frontiers orbitals ( Figure S48). The selection rules for SOC are lifted because of the possible mixing of the orbitals involved in the n,π* and π,π* states; as a result, the SOCMEs ⟨Ψ S1 | ĤS O | Ψ T2 ⟩ adopt non-vanishing values. The atomic orbital contributions can be quantitatively evaluated in terms of Loẅdin population analysis, where the population of a molecular orbital is projected into the minimal basis set of atomic orbitals of individual atoms. According to the analysis, the iodine atomic orbitals in the 3H-indolium moieties do not contribute to the frontier orbitals, whereas in the C3′ position, their contribution is up to a few percent (4% for HOMO), which affects ISC along the vibrational coordinates.
The change in the TDDFT with TDA energies of the S 1 , T 1 , and T 2 states and energy differences along vibrational modes are shown in Figure S49. The figure demonstrates that the energy varies along the vibrational coordinates and that T 2 and S 1 states exhibit many crossing points. This supports the claim that these states are involved in ISC. On the contrary, the crossing between T 1 and S 1 states is probably less efficient owing to the substantial energy gap for all but one vibrational coordinate (in the case of 1b and 1d).
Based on the theoretical data, we assume that an order of magnitude increase in the quantum yield of 1 O 2 production (Φ Δ ) observed for 1b and 1d can be interpreted in terms of increased efficiency of ISC thanks to the vibronic SOC mechanism between S 1 and T 2 states. The SOCMEs for 1b and 1d show an order of magnitude increase along the vibrational coordinates compared to 1a and 1c, which can lead to two orders of magnitude faster population transfer. Therefore, the vibronic SOC can be viewed as a powerful mechanism for electronic population transfer from the optically bright state to the triplet manifold.

■ CONCLUSIONS
In this work, the singlet oxygen formation efficiency, affinity to singlet oxygen, and photostability of several Cy7 derivatives bearing iodine atom substituents in the C3′ position of the chain and/or the 3H-indolium core in methanol and aqueous solutions were determined to evaluate the magnitude of the heavy-atom effect of iodine substitution. The singlet-oxygen production was found to be an order of magnitude more efficient for the chain substitution than the 3H-indole core substitution.
The observed dependence cannot be explained solely based on electronic structure calculations for the optimized structures of the synthesized compounds. We show that the variation of SOC along the vibrational coordinates must be considered when interpreting ISC in organic molecules. Although theory showing the fundamentals of the spin−vibronic mechanisms of ISC has been known since the 1960s, 40,71 only recent experimental and theoretical advances have proved that this mechanism is important in many systems. We demonstrate that a simple qualitative analysis based on the energy gaps or Condon approximations, which assumes that the SOCMEs between states remain unchanged along vibrational motion, cannot provide a complete understanding of the ISC phenomenon. Consequently, we should always be aware of the interplay between spin, electronic, and nuclear dynamics when describing the excited states of molecules, even those that do not contain a heavy element. It should also be emphasized that the standard way of calculating SOCs, i.e., using an effective charge model with effective core potentials for heavy atoms, provides an unsatisfactory description of the system. The evaluation of spin−vibronic coupling should thus become a standard tool for the theoretical analysis of photochemical reactions. This work not only serves for a better understanding of SOC in substituted cyanine dyes but also can help for their further development as photosensitizers, for example, in photodynamic therapy applications. ■ EXPERIMENTAL SECTION Materials and Methods. Reagents (2,4-dinitrophenol, ptoluenesulfonyl chloride, 3-iodopyridine, 3-methylbutan-2-one, 4iodophenylhydrazine, 4-hydrazinobenzenesulfonic acid, 2,3,3-trimethylindolenine, 1,3-propane sultone, 4-bromoaniline, methyl iodide, 1,3-diphenylisobenzofuran, rose bengal, and methylene blue) and solvents of the highest purity available were used as purchased, or they were purified/dried using standard methods when necessary. 1 H NMR spectra were recorded on 300 or 500 MHz spectrometers; 13 C NMR spectra were obtained on 125 or 75 MHz instruments. 1 H chemical shifts are reported in parts per million (ppm) relative to d 6 -DMSO (δ = 2.50 ppm), CD 3 OD (δ = 3.31 ppm), CDCl 3 (δ = 7.26 ppm), or D 2 O (δ = 4.79 ppm) as internal references. 13 C chemical shifts are reported in ppm with d 6 -DMSO (δ = 77.67 ppm), CDCl 3 (δ = 77.16 ppm), and CD 3 OD (δ = 49.30 ppm) as internal references. Absorption spectra and the molar absorption coefficients were obtained on a UV−vis spectrometer with matched 1.0 cm quartz cells. Molar absorption coefficients were determined from the absorption spectra (the average values were obtained from three independent measurements with solutions of different concentrations). Emission and excitation spectra were normalized and smoothed using standard protocols. Flash column chromatography was performed using silica gel (230−400 mesh). The exact masses of the synthesized compounds were obtained using a triple quadrupole electrospray ionization (ESI) mass spectrometer in a positive or negative mode. Melting points were measured on an automatic melting point apparatus. Synthetic procedures were performed under an ambient atmosphere unless stated otherwise. In specified cases, the structural assignment was made with additional information from gCOSY, gHSQC, and gHMBC experiments.
Fluorescence Measurements. Fluorescence and excitation spectra were measured using a fluorescence spectrometer in 1.0 cm quartz fluorescence cuvettes at 23 ± 1°C. The sample concentrations were adjusted to keep the absorbance below 0.15 at the corresponding excitation wavelength. Each sample was measured three times, and the spectra were averaged. Molar Absorption Coefficients. Stock solutions of 1a−f were prepared from the amounts of 3−6 mg in 10 or 25 mL volumetric flasks. The absorbances at different concentrations (10 −7 −10 −5 M) were measured in 1.0 cm quartz cuvettes. Slopes of the data obtained from the measurements at maximum absorbance wavelengths from three independently prepared solutions were averaged to get molar absorption coefficients using the Beer−Lambert equation A = l c ε(λ), where A is the absorbance, l is an optical length, c is the concentration, and ε(λ) is a molar absorption coefficient. In some cases (especially for the measurements in PBS), DMSO was added to avoid aggregation.
Reaction with 1 where Φ Δ,REF is the quantum yield of singlet oxygen production from a reference, k is the rate of the consumption of the singlet oxygen trap, and I is the amount of absorbed light, where where A λ,t is the absorbance of the sample and I em,λ is the emission intensity of LEDs.
Singlet Oxygen Production Quantum Yields in PBS. All measurements and calculations were performed analogous to those in methanol with the exception of the use of 9,10-anthracenediylbis(methylene)dimalonic acid (ABDA) (c = 1. where I is given by

3-[2-[7-[3,3-Dimethyl-5-sulfo-1-(3-(trimethylammonium)propyl)-3H-indol-1-ium-2-yl]-3-iodo-2,4,6-heptatrien-1-ylidene]-3,3-dimethyl-5-sulfo-1,2-dihydro-3H-indol-1-yl]-propanesulfonate
Bromide (1f). 5b (150 mg, 0.28 mmol) and 4-aminobenzonitrile (98 mg, 0.82 mmol) were dissolved in methanol (2.8 mL), and the mixture was stirred at room temperature for 30 min. Then, 3c (350 mg, 0.82 mmol) and sodium acetate (260 mg, 1.66 mmol) were added and the mixture was stirred at 45°C in an oil bath for an additional 6 h. The reaction mixture was cooled down to room temperature, and the resulting precipitate was filtered off and washed with ethanol (2 × 5 mL), acetone (5 × 10 mL), and diethyl ether (3 × 5 mL). The compound was dissolved in water (10 mL), acetone (30 mL) was added dropwise, and the solution was stored in the fridge (−5°C) for 6 h. The resulting precipitate was filtered off and washed with acetone (5 × 10 mL) and diethyl ether (3 × 5 mL) to give the pure product.  The structures of Cy7 derivatives were optimized in water (used as an archetype of a polar solvent) represented by a dielectric continuum, using the polarizable continuum model (PCM). 87 The electronic structure calculations were performed at the density functional level, using the CAM-B3LYP functional and the def2-TZVP and def2-SVP basis sets for optimization; the structures are shown in Supporting Information. The choice of the functional was motivated by our previous investigation of analogous cyanine molecules 22 and by the fact that the range-separate hybrid functionals have recently been proven to provide an accurate description of the conjugated double bonds. 88 However, the excitation energies of cyanines are generally relatively insensitive to the choice of functional. 37 Vibrational frequencies, SOC matrix elements (SOCMEs), and the TDDFT excitation energies� with and without a TDA 89 �were calculated at the same levels of theory. To calculate the SOCMEs, we used the Breit−Pauli SOC operator with a mean-field approximation with exact two-electron terms as implemented in the ORCA 5.0.1 package. 90−92 To include the scalar relativistic effects, we used the full Douglas−Kroll Hamiltonian, 93 which was a critical step to obtain reliable values of the SOCs. The Hessian with the relativistic Douglas−Kroll Hamiltonian (DKH) and PCM for solvent is available only for the ground electronic state. We assumed that the normal modes of q i are the same for the ground state and the S 1 state because the optimized geometries are very similar. The basis set dependence of SOCMEs is not critical. The tests were performed for 1b and are summarized in Table S6. This approach for the calculations of SOCMEs was shown to lead to errors of ∼5%. 94 The SOCME scans were performed as a function of dimensionless normal coordinates δq i ; the particular molecular distortions were constructed to correspond to variation of the normal coordinates from −l i δq i to +r i δq i , where l i and r i specify the number of steps in positive and negative directions along each normal mode. The relation between the dimensionless normal coordinate and the vector of atomic Cartesian displacements δX is defined in the ORCA code as where L ki is the orthogonal matrix obtained upon numerical diagonalization of the mass-weighted Hessian matrix and M is the vector of atomic masses. The Hessian matrix was obtained for optimized geometries of the ground electronic state at the same level of theory. Note that the optimization of excited-state geometries within the same level of theory with PCM is not implemented yet. The optimization of the S 1 excited-state geometries was performed in a vacuum at the CAM-B3LYP/def2-SVP level, and the geometries were compared to those of the ground state structure at the same level of theory. These geometries were used for calculations of the SOCMEs and TDDFT energies at the CAM-B3LYP/def2-TZVP levels at the S 1 minimum. All electronic structure calculations were performed in ORCA, version 5.0.1. 90−92 ■ ASSOCIATED CONTENT

Data Availability Statement
The data underlying this study are available in the published article and its online supporting material.