Low-Symmetry Phthalocyanines Bearing Carboxy-Groups: Synthesis, Spectroscopic and Quantum-Chemical Characterization

The synthesis and characterization of A3B-type phthalocyanines, ZnPc1–4, bearing bulky 2,6-diisopropylphenoxy-groups or chlorine atoms on isoindoline units “A” and either one or two carboxylic anchors on isoindoline unit “B” are reported. A comparison of molecular modelling with the conventional time dependent—density functional theory (TD-DFT) approach and its simplified sTD-DFT approximation provides further evidence that the latter method accurately reproduces the key trends in the spectral properties, providing colossal savings in computer time for quite large molecules. This demonstrates that it is a valuable tool for guiding the rational design of new phthalocyanines for practical applications.


Introduction
Phthalocyanines (Pcs) are macrocyclic ligands of particular interest because of their high stability, excellent photophysical properties and facile structural modification, which can readily be used to control the properties of corresponding materials and devices [1][2][3]. From this standpoint, low-symmetry Pcs are particularly interesting since combining various functional groups at the ligand periphery enables a rational modulation of their properties such as solubility, compatibility with nanomaterials, and nonlinear optical properties [4]. For example, low-symmetry Pcs bearing one or two carboxylic groups can act as the light-harvesting components of dye-sensitized solar cells (DSSCs) when these sensitizers are grafted on the surface of TiO 2 or ZnO [5]. To control the aggregation and solubility of these Pcs, bulky groups can be introduced varying from relatively small tert-butyl groups [6] to perfluoro-tert-butyl substituents [7] and sterically demanding 2,6disubstituted phenoxy-groups [8][9][10][11]. Moreover, carboxy-substituted Pcs can be conjugated with various nanomaterials, such as gold nanoparticles [12] or carbon nanotubes [13] to provide photoactive hybrid materials.
In addition to the control of aggregation, the substitution pattern should also help to align the frontier orbital energies of the sensitizer and the conduction band of metal oxides, so molecular modelling calculations can be used to guide further synthetic work in a rational manner [14] and improve device performance [15,16]. The accurate prediction of UV-visible absorption (UV-vis) spectra of low-symmetry Pc sensitizers is critical since photon absorption is the primary act in the sequence of physical processes that convert solar energy into electricity. Typically, this task is solved using the time dependent-density functional theory (TD-DFT) method [17][18][19][20], although the relatively high computational

Results
Carboxy-substituted Pcs can be synthesized by a cross-condensation template reaction of phthalonitriles bearing bulky solubilizing groups and those functionalized either with hydroxymethyl or ester groups followed by chromatographic isolation of the low-symmetry target complexes. The CH2OH-substituted complexes can then be oxidized using iodoxybenzoic acid to form an aldehyde followed by oxidation to form a -COOH group by using NaClO2 in the presence of sulfamic acid [6,26,27]. Recently, direct oxidation of Zn[(tBu)3(CH2OH)Pc] to the corresponding carboxy-substituted complex was reported; the anaerobic reaction of the specified complex with KOH catalyzed by ZnO furnished the well-known TT1 dye [28]. Ester-substituted Pcs can be hydrolyzed to form target complexes bearing -COOH groups [8,9]. This method was selected for use in this study (Scheme 1). Synthesis of A3B-Pcs was performed from the previously reported precursors-4,5-bis(2,6-diisopropylphenoxy)-or

Results
Carboxy-substituted Pcs can be synthesized by a cross-condensation template reaction of phthalonitriles bearing bulky solubilizing groups and those functionalized either with hydroxymethyl or ester groups followed by chromatographic isolation of the lowsymmetry target complexes. The CH 2 OH-substituted complexes can then be oxidized using iodoxybenzoic acid to form an aldehyde followed by oxidation to form a -COOH group by using NaClO 2 in the presence of sulfamic acid [6,26,27]. Recently, direct oxidation of Zn[(tBu) 3 (CH 2 OH)Pc] to the corresponding carboxy-substituted complex was reported; the anaerobic reaction of the specified complex with KOH catalyzed by ZnO furnished the well-known TT1 dye [28]. Ester-substituted Pcs can be hydrolyzed to form target complexes bearing -COOH groups [8,9]. This method was selected for use in this study (Scheme 1). photon absorption is the primary act in the sequence of physical processes that convert solar energy into electricity. Typically, this task is solved using the time dependent-density functional theory (TD-DFT) method [17][18][19][20], although the relatively high computational cost can limit its applicability in the context of large molecules with sterically demanding substituents. In this study, the synthesis of a series of zinc phthalocyaninates, ZnPc1-4, with bulky solubilizing diisopropylphenoxy-groups and either one or two carboxylic anchors is reported (Figure 1). Analyses of their UV-vis spectra using the classical TD-DFT approach and its simplified approximation (sTD-DFT) [21,22], demonstrate that the latter approach provides a spectacular speed-up of calculations by orders of magnitude, which is particularly useful in the theoretical treatment of large conjugated molecules such as phthalocyanines [22][23][24][25].

Results
Carboxy-substituted Pcs can be synthesized by a cross-condensation template reaction of phthalonitriles bearing bulky solubilizing groups and those functionalized either with hydroxymethyl or ester groups followed by chromatographic isolation of the low-symmetry target complexes. The CH2OH-substituted complexes can then be oxidized using iodoxybenzoic acid to form an aldehyde followed by oxidation to form a -COOH group by using NaClO2 in the presence of sulfamic acid [6,26,27]. Recently, direct oxidation of Zn[(tBu)3(CH2OH)Pc] to the corresponding carboxy-substituted complex was reported; the anaerobic reaction of the specified complex with KOH catalyzed by ZnO furnished the well-known TT1 dye [28]. Ester-substituted Pcs can be hydrolyzed to form target complexes bearing -COOH groups [8,9]. This method was selected for use in this study (Scheme 1). Synthesis of A 3 B-Pcs was performed from the previously reported precursors-4,5bis(2,6-diisopropylphenoxy)or 4-chloro-5-(2,6-diisopropylphenoxy)-substituted phthalonitriles 1 and 2 for the A ring moieties, and methyl-3,4-dicyanobenzoate 3 or dimethyl-4,5-  2 and DBU afforded mixtures containing mainly the A 4 , A 3 B and A 2 B 2 macrocyclic products. Because of transesterification, methyl groups were replaced by amyl residues, which decreased the overall polarity of the resulting complexes and hampered chromatographic separation of the A 3 B product from the A 4 and A 2 B 2 structures. Hydrolysis of ester bonds in ZnPcAm1-4 resulted in the formation of the acid groups of the target compounds, which can be readily separated, from traces of the relatively nonpolar A 4 and the much more polar A 2 B 2 derivatives. Control over the separation was achieved by thin-layer chromatography (TLC) and matrix-assisted laser desorption ionization-time-of-flight (MALDI-TOF) mass spectrometry. All complexes isolated were characterized by UV-vis and 1 H NMR spectroscopy (See Supplementary Materials).

Synthesis
The UV-vis and magnetic circular dichroism (MCD) spectra of ZnPc1-4 ( Figure 2) were measured in DMF to suppress their aggregation properties so the spectral properties of the monomeric complexes can be compared with those of ZnPc*, a fully symmetric A 4 compound with sterically demanding substituents that was isolated as a byproduct during the synthesis of ZnPc1 and ZnPc2. The introduction of electron-withdrawing carboxygroups and/or chlorine atoms has only a relatively minor effect on the Q-band wavelengths, resulting in a bathochromic shift of up to 7 nm relative to that of the symmetric ZnPc* reference compound.
Molecules 2022, 26, x FOR PEER REVIEW 3 4-chloro-5-(2,6-diisopropylphenoxy)-substituted phthalonitriles 1 and 2 for the A moieties, and methyl-3,4-dicyanobenzoate 3 or dimethyl-4,5-dicyanophthalate 4 on t ring moieties. Condensation of the corresponding pairs of precursors in reflu n-pentanol in the presence of Zn(OAc)2 and DBU afforded mixtures containing ma the A4, A3B and A2B2 macrocyclic products. Because of transesterification, methyl gro were replaced by amyl residues, which decreased the overall polarity of the resul complexes and hampered chromatographic separation of the A3B product from th and A2B2 structures. Hydrolysis of ester bonds in ZnPcAm1-4 resulted in the forma of the acid groups of the target compounds, which can be readily separated, from tr of the relatively nonpolar A4 and the much more polar A2B2 derivatives. Control over separation was achieved by thin-layer chromatography (TLC) and matrix-assisted l desorption ionization-time-of-flight (MALDI-TOF) mass spectrometry. All compl isolated were characterized by UV-vis and 1 H NMR spectroscopy (See Supplemen Materials). The UV-vis and magnetic circular dichroism (MCD) spectra of ZnPc1-4 (Figur were measured in DMF to suppress their aggregation properties so the spectral pro ties of the monomeric complexes can be compared with those of ZnPc*, a fully symm A4 compound with sterically demanding substituents that was isolated as a bypro during the synthesis of ZnPc1 and ZnPc2. The introduction of electron-withdraw carboxy-groups and/or chlorine atoms has only a relatively minor effect on the Q-b wavelengths, resulting in a bathochromic shift of up to 7 nm relative to that of the s metric ZnPc* reference compound.  The effect of introducing the peripheral substituents was analyzed through a comparison of the experimental spectral data with the TD-DFT and sTD-DFT calculations. In contrast to previous studies of substituted phthalocyanines where bulky alkoxy and aryloxy groups were truncated and replaced with methoxy groups to save computational time [22][23][24][25], in the present work, all calculations were performed for molecules with genuine diisopropylphenoxy substituents to further demonstrate the capabilities of the sTD-DFT approach in the context of a desktop computer. The optical properties of porphyrinoid complexes, such as phthalocyanines, can be readily conceptualized through a consideration of the molecular orbitals (MOs) associated with the 16 atom 18 π-electron inner ligand perimeter that have an M L = 0, ±1, ±2, ±3, ±4, ±5, ±6, ±7, 8 sequence in ascending energy terms [29,30]. The highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) have M L = ±4 and ±5 angular nodal patterns, respectively, resulting in Q and B bands with ∆M L = ±1 and ±9 properties at lower and higher energy. Michl [30] introduced an a, s, -a and -s nomenclature for the MOs derived from the four frontier π-MOs of a C 16 H 16 2− parent perimeter (Figure 3), depending on whether there are nodal planes (a/-a) and MO coefficients (s/-s) aligned with the y-axis, which enables the facile comparison of the electronic structures of cyclic polyenes of differing symmetry. In the context of phthalocyanines, the peripheral benzo ring substitution introduces a second frontier π-MO with a 2u symmetry [29,[31][32][33] that complicates the analysis of the electronic structures of phthalocyanines ( Figure 4). The effect of introducing the peripheral substituents was analyzed through a comparison of the experimental spectral data with the TD-DFT and sTD-DFT calculations. In contrast to previous studies of substituted phthalocyanines where bulky alkoxy and aryloxy groups were truncated and replaced with methoxy groups to save computational time [22][23][24][25], in the present work, all calculations were performed for molecules with genuine diisopropylphenoxy substituents to further demonstrate the capabilities of the sTD-DFT approach in the context of a desktop computer.
The optical properties of porphyrinoid complexes, such as phthalocyanines, can be readily conceptualized through a consideration of the molecular orbitals (MOs) associated with the 16 atom 18 π-electron inner ligand perimeter that have an ML = 0, ±1, ±2, ±3, ±4, ±5, ±6, ±7, 8 sequence in ascending energy terms [29,30]. The highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) have ML = ±4 and ±5 angular nodal patterns, respectively, resulting in Q and B bands with ΔML = ±1 and ±9 properties at lower and higher energy. Michl [30] introduced an a, s, -a and -s nomenclature for the MOs derived from the four frontier π-MOs of a C16H16 2− parent perimeter ( Figure 3), depending on whether there are nodal planes (a/-a) and MO coefficients (s/-s) aligned with the y-axis, which enables the facile comparison of the electronic structures of cyclic polyenes of differing symmetry. In the context of phthalocyanines, the peripheral benzo ring substitution introduces a second frontier π-MO with a2u symmetry [29,[31][32][33] that complicates the analysis of the electronic structures of phthalocyanines ( Figure 4).   Spectral band deconvolution studies for a series of axially ligated Zn II Pc complexes by Nyokong and coworkers identified the presence of two intense overlapping Faraday A1 terms in the B band region [34] that were subsequently labelled as the B1 and B2 bands [4,[35][36][37][38][39][40]. These bands can be assigned primarily to the 1a2u→-a/-s and 1b2u→-a/-s one-electron transitions on the basis of the molecular modelling at the B3LYP/6-31G(d) level of theory (Tables 1 and 2). For this reason, the 1a2u MO ( Figure 3) has previously been assumed to be the s MO of ZnPc in the context of Michl's perimeter model [4,22,33,41].
The assignment of the higher energy ππ* transitions of ZnPc1-4 and ZnPc* in the B band region is problematic due to the complicated configurational interaction that is predicted, but broadly similar trends are predicted in the TD-DFT and sTD-DFT calculations ( Figure 5, Tables 1 and 2). An extra intense band is predicted in the 300−400 nm region in the sTD-DFT calculations for ZnPc1-4 and ZnPc* and in the TD-DFT calculations for ZnPc1, ZnPc2 and ZnPc* ( Figure 5) due to the mesomeric effects of the oxygen lone pairs of the peripheral substituents on MOs that are localized primarily on the peripheral benzo rings. A marked destabilization is predicted for these MOs, which includes the 1a2u MO in the context of ZnPc* (Figures 4 and 5, Tables 1 and 2). It is hence reasonable to conclude that the 2a2u→-a/-s one-electron transitions are likely to dominate in the context of the main B band of ZnPc* as is predicted in the TD-DFT and sTD-DFT calculations ( Figure 5, Tables 1 and 2). Spectral band deconvolution studies for a series of axially ligated Zn II Pc complexes by Nyokong and coworkers identified the presence of two intense overlapping Faraday A 1 terms in the B band region [34] that were subsequently labelled as the B 1 and B 2 bands [4,[35][36][37][38][39][40]. These bands can be assigned primarily to the 1a 2u →-a/-s and 1b 2u →-a/-s one-electron transitions on the basis of the molecular modelling at the B3LYP/6-31G(d) level of theory (Tables 1 and 2). For this reason, the 1a 2u MO ( Figure 3) has previously been assumed to be the s MO of ZnPc in the context of Michl's perimeter model [4,22,33,41].
The assignment of the higher energy ππ* transitions of ZnPc1-4 and ZnPc* in the B band region is problematic due to the complicated configurational interaction that is predicted, but broadly similar trends are predicted in the TD-DFT and sTD-DFT calculations ( Figure 5, Tables 1 and 2). An extra intense band is predicted in the 300−400 nm region in the sTD-DFT calculations for ZnPc1-4 and ZnPc* and in the TD-DFT calculations for ZnPc1, ZnPc2 and ZnPc* ( Figure 5) due to the mesomeric effects of the oxygen lone pairs of the peripheral substituents on MOs that are localized primarily on the peripheral benzo rings. A marked destabilization is predicted for these MOs, which includes the 1a 2u MO in the context of ZnPc* (Figures 4 and 5, Tables 1 and 2). It is hence reasonable to conclude that the 2a 2u →-a/-s one-electron transitions are likely to dominate in the context of the main B band of ZnPc* as is predicted in the TD-DFT and sTD-DFT calculations ( Figure 5, Tables 1 and 2).    The predicted energy gaps between the MOs of ZnPc1-4 that are derived from the 1a 1u and 2a 2u MOs of the ZnPc parent complex are significantly smaller than is the case with ZnPc* (Figure 4), since it is only the MOs derived from the 1b 1u and 1e g MOs of ZnPc that are significantly destabilized by the presence of six or three diisopropylphenoxy groups in this context. As a result, multiple bands involving large contributions from the 1a 2u →-a/-s and 2a 2u →-a/-s one-electron transitions are predicted in the B band region in both the TD-DFT and sTD-DFT calculations (Tables 1 and 2). In contrast with the complexity of the calculated spectra of ZnPc1-4 and ZnPc* (Figure 5), only a relatively weak shoulder of absorbance is observed to the red of the main B band envelope in the experimental spectra (Figure 2). In a similar manner, only relatively minor differences are observed in the Faraday A 1 and pseudo-A 1 terms in the B band regions of the MCD spectra of ZnPc* and ZnPc1-4 ( Figure 2). These are likely to be associated with the large orbital angular momentum generated by transitions between the s, -a and -s MOs with M L = ±4 and ±5 angular nodal properties that are largely localized on the inner ligand perimeter [30,38,39]. The MCD spectra in Figure 2 therefore provide direct spectroscopic evidence that the TD-DFT and sTDDFT calculations in Figure 5 do not provide an accurate description of the B band region. In contrast with the relatively consistent Faraday A 1 and pseudo-A 1 term band morphology that is observed in the B band region of the MCD spectra for ZnPc* and ZnPc1-4, substantial differences are predicted in the wavefunctions in the B band region in Tables 1 and 2 and the simulated spectra in Figure 5. Extensive configurational interactions between a large number of ππ* excited states result in significant differences in the contributions from the 1a 2u →-a/-s and 2a 2u →-a/-s one-electron transitions which are highlighted in bold face in Tables 1 and 2. The predicted energy gaps between the MOs of ZnPc1-4 that are derived from the 1a1u and 2a2u MOs of the ZnPc parent complex are significantly smaller than is the case with ZnPc* (Figure 4), since it is only the MOs derived from the 1b1u and 1eg MOs of ZnPc that are significantly destabilized by the presence of six or three diisopropylphenoxy groups in this context. As a result, multiple bands involving large contributions from the 1a2u→-a/-s and 2a2u→-a/-s one-electron transitions are predicted in the B band region in both the TD-DFT and sTD-DFT calculations (Tables 1 and 2). In contrast with the complexity of the calculated spectra of ZnPc1-4 and ZnPc* (Figure 5), only a relatively weak shoulder of absorbance is observed to the red of the main B band envelope in the experimental spectra (Figure 2). In a similar manner, only relatively minor differences are observed in the Faraday A1 and pseudo-A1 terms in the B band regions of the MCD spectra of ZnPc* and ZnPc1-4 ( Figure 2). These are likely to be associated with the large orbital angular momentum generated by transitions between the s, -a and -s MOs with ML = ±4 and ±5 angular nodal properties that are largely localized on the inner ligand perimeter [30,38,39]. The MCD spectra in Figure 2 therefore provide direct spectroscopic evidence that the TD-DFT and sTDDFT calculations in Figure 5 do not provide an accurate description of the B band region. In contrast with the relatively consistent Faraday A1 and pseudo-A1 term band morphology that is observed in the B band region of the MCD spectra for ZnPc* and ZnPc1-4, substantial differences are predicted in the wavefunctions in the B band region in Tables 1 and 2 and the simulated spectra in Figure 5. Extensive configurational interactions between a large number of ππ* excited states result in significant differences in the contributions from the 1a2u→-a/-s and 2a2u→-a/-s one-electron transitions which are highlighted in bold face in Tables 1 and 2. bands are highlighted with large red diamonds, while smaller amber, light blue, black, gray and green diamonds are used for bands arising primarily from transitions between destabilized π-MOs localized on the peripheral benzo rings into the -a/-s MOs, between the a MO into higher energy π* MOs, between σ-MOs associated with the aza-nitrogen lone pairs into the -a/-s MOs, from MOs localized on the sterically demanding -OR substituents into the -a/-s MOs, and between other π MOs of the Pc ligands into the -a/-s MOs, respectively. Simulated spectra are derived from the Chemcraft [42] and Chemissian [43] programs, respectively, with a fixed bandwidth in each case of 2000 cm −1 .

Discussion
The TD-DFT and sTD-DFT calculations ( Figure 5, Tables 1 and 2) generally reproduce trends observed in the experimental spectra in the Q band region (Figures 2 and 5), which are often the most significant from the standpoint of applications such as DSSCs. The simplified approximation has the advantage of a spectacular speed-up in computa- bands are highlighted with large red diamonds, while smaller amber, light blue, black, gray and green diamonds are used for bands arising primarily from transitions between destabilized π-MOs localized on the peripheral benzo rings into the -a/-s MOs, between the a MO into higher energy π* MOs, between σ-MOs associated with the aza-nitrogen lone pairs into the -a/-s MOs, from MOs localized on the sterically demanding -OR substituents into the -a/-s MOs, and between other π MOs of the Pc ligands into the -a/-s MOs, respectively. Simulated spectra are derived from the Chemcraft [42] and Chemissian [43] programs, respectively, with a fixed bandwidth in each case of 2000 cm −1 .

Discussion
The TD-DFT and sTD-DFT calculations ( Figure 5, Tables 1 and 2) generally reproduce trends observed in the experimental spectra in the Q band region (Figures 2 and 5), which are often the most significant from the standpoint of applications such as DSSCs. The simplified approximation has the advantage of a spectacular speed-up in computation time, but the mesomeric interactions of the carboxylic acid moieties are somewhat problematic in the context of the sTD-DFT calculations. The partial replacement of the electron-donating diisopropylphenoxyl substituents of ZnPc* with electron-withdrawing carboxylic acid groups and chlorine atoms (Figure 1) results in a stabilization of the energies of the frontier orbitals of ZnPc1-4 ( Figure 4). Lower molecular symmetry results in a lifting of the degeneracy of the -a and -s MOs. Larger ∆LUMO values (Michl's terminology for the energy splitting of the -a and -s MOs [30]) are predicted for the dicarboxylic acid substituted ZnPc2 and ZnPc4 complexes, since there are large MO coefficients on the peripheral carbons of the B ring moiety of the -a MOs, but not on those of the -s MOs (Figure 3). The electronwithdrawing mesomeric interaction with the carboxylic acid groups is hence expected to stabilize the -a MOs relative to the -s MOs and result in separate xand y-polarized Q 00 bands.
Q band splittings of 29 and 59 nm are predicted for ZnPc2 in the TD-DFT and sTD-DFT calculations (Tables 1 and 2), respectively, while smaller splittings of 23 and 45 nm are predicted for ZnPc4 which has both a chlorine atoms and an isopropylphenoxyl substituent on the A 3 benzo ring moieties. These bands are not resolved in the experimental spectra although significant broadening is observed in the Q 00 bands of ZnPc2 and ZnPc4 relative to that of ZnPc* (Figure 2). It is hence apparent that the extent of the Q band splitting associated with the mesomeric interactions with the carboxylic acid substituents is over-estimated in the sTD-DFT calculations of ZnPc1-4. This results in a significant under-estimation of the energies of the lower energy Q 00 bands (Table 1). In contrast, there is a systematic over-estimation of the Q 00 band energies in the TD-DFT calculations at the B3LYP/6-31G(d) level of theory (Table 2) as has been reported previously for calculations of this type for a wide range of different porphyrins and phthalocyanine-related structures [17,33,41,44,45].
Since the envisaged application for ZnPc1-4 is in DSSCs as photosensitizer dyes coating the TiO 2 photoanode, a preliminary assessment of relevant parameters [46][47][48][49][50] was calculated on the basis of single point DFT calculations (Table 3) using the optimized geometries of the dyes at the B3LYP/SDD level of theory. Since the LUMO energies ( Figure 6) are higher than that of the conduction band (CB) of TiO 2 [46,50], injection of an electron into the TiO 2 photoanode of the DSSC after photoexcitation should be feasible. Favorable open circuit voltage (V oc ) values are predicted to lie in the 1.08−1.41 eV range for ZnPc1-4 (Table 3). Spontaneous Gibbs free energies are predicted for the electron injection (∆G inj ) and dye regeneration (∆G regen ) processes shown in Figure 6 that are required to complete the circuit of the DSSC [46,50]. Favorable light-harvesting efficiency (LHE) values were also derived for the maxima of the Q bands using the oscillator strength (f) values from Table 1. Since these parameters appear to be promising, laboratory studies with DSSCs are already in progress to further assess the suitability of ZnPc1-4 for this application. tion time, but the mesomeric interactions of the carboxylic acid moieties are somewhat problematic in the context of the sTD-DFT calculations. The partial replacement of the electron-donating diisopropylphenoxyl substituents of ZnPc* with electron-withdrawing carboxylic acid groups and chlorine atoms (Figure 1) results in a stabilization of the energies of the frontier orbitals of ZnPc1-4 ( Figure 4). Lower molecular symmetry results in a lifting of the degeneracy of the -a and -s MOs. Larger ΔLUMO values (Michl's terminology for the energy splitting of the -a and -s MOs [30]) are predicted for the dicarboxylic acid substituted ZnPc2 and ZnPc4 complexes, since there are large MO coefficients on the peripheral carbons of the B ring moiety of the -a MOs, but not on those of the -s MOs (Figure 3). The electron-withdrawing mesomeric interaction with the carboxylic acid groups is hence expected to stabilize the -a MOs relative to the -s MOs and result in separate x-and y-polarized Q00 bands. Q band splittings of 29 and 59 nm are predicted for ZnPc2 in the TD-DFT and sTD-DFT calculations (Tables 1 and 2), respectively, while smaller splittings of 23 and 45 nm are predicted for ZnPc4 which has both a chlorine atoms and an isopropylphenoxyl substituent on the A3 benzo ring moieties. These bands are not resolved in the experimental spectra although significant broadening is observed in the Q00 bands of ZnPc2 and ZnPc4 relative to that of ZnPc* (Figure 2). It is hence apparent that the extent of the Q band splitting associated with the mesomeric interactions with the carboxylic acid substituents is over-estimated in the sTD-DFT calculations of ZnPc1-4. This results in a significant under-estimation of the energies of the lower energy Q00 bands (Table 1). In contrast, there is a systematic over-estimation of the Q00 band energies in the TD-DFT calculations at the B3LYP/6-31G(d) level of theory (Table 2) as has been reported previously for calculations of this type for a wide range of different porphyrins and phthalocyanine-related structures [17,33,41,44,45].
Since the envisaged application for ZnPc1-4 is in DSSCs as photosensitizer dyes coating the TiO2 photoanode, a preliminary assessment of relevant parameters [46][47][48][49][50] was calculated on the basis of single point DFT calculations (Table 3) using the optimized geometries of the dyes at the B3LYP/SDD level of theory. Since the LUMO energies ( Figure 6) are higher than that of the conduction band (CB) of TiO2 [46,50], injection of an electron into the TiO2 photoanode of the DSSC after photoexcitation should be feasible. Favorable open circuit voltage (Voc) values are predicted to lie in the 1.08−1.41 eV range for ZnPc1-4 (Table 3). Spontaneous Gibbs free energies are predicted for the electron injection (ΔGinj) and dye regeneration (ΔGregen) processes shown in Figure 6 that are required to complete the circuit of the DSSC [46,50]. Favorable light-harvesting efficiency (LHE) values were also derived for the maxima of the Q bands using the oscillator strength (f) values from Table 1. Since these parameters appear to be promising, laboratory studies with DSSCs are already in progress to further assess the suitability of ZnPc1-4 for this application.    [46,51]. d The open circuit voltage (V oc ) values are calculated using the equation: V oc = |E HOMO (Donor)| − |E LUMO (Acceptor)| − 0.3 [46][47][48][49]51]. |E HOMO (Donor)| and |E LUMO (Acceptor)| are derived from E OX dye and the energy of the conduction band of TiO 2 (E CB ) of −4.00 eV [46,50]. e The Gibbs free energy for injection of an electron from the dye into the TiO 2 photoanode (∆G inj ) was calculated by using the equation: ∆G inj = E OX dye * − E CB [46,51]. f The I/I 3 − redox potential of the electrolyte is assumed to be −4.80 eV so the oxidation potential energy (E OX electrolyte ) is 4.80 eV [46,50]

Methods
MALDI-TOF mass spectra were measured on a Bruker Daltonics Ultraflex mass spectrometer in positive ion mode with 2,5-dihydroxybenzoic acid (DHB) as a matrix. UV-visible absorption (UV-vis) spectra were recorded in CHCl 3 on a Thermo Evolution 210 spectrometer in the 250-900 nm range. Rectangular quartz cuvettes with a 10 nm optical pathlength were used. NMR spectra were measured on a Bruker Avance-III spectrometer at a frequency of 600.13 MHz. Samples were prepared in CDCl 3 (Cambridge Isotope Laboratories, Inc.), and filtered through a layer of alumina before use. Spectra were acquired at ambient temperatures. NMR spectra were referenced to the solvent signal (CHCl 3 , 7.26 ppm). Magnetic circular dichroism (MCD) spectra were recorded with a Chirascan plus spectrometer (Applied Photophysics, UK) equipped with a 1.0 tesla permanent magnet by using both parallel and antiparallel fields.

Computational Details
Geometry optimizations were carried out for unsubstituted ZnPc (ZnPc) as a model complex, and ZnPc* and ZnPc1-4 at the B3LYP/6-31G(d) level of theory by using the Gaussian 09 software package [57]. Conventional TD-DFT calculations were carried out at the CAM-B3LYP/SDD level of theory since the CAM-B3LYP functional contains a longrange correction, while sTD-DFT calculations [21,22] were performed with the ORCA 5.0 package [58] using B3LYP/6-31G(d) optimized geometries. The MO energies and angular nodal patterns of ZnPc, ZnPc* and ZnPc1-4 were calculated at the CAM-B3LYP/6-31G(d) level of theory [59][60][61]. The RIJCOSX approximation with auxiliary basis set def2/J was used to speed-up the sTD-DFT calculations [62,63]. Only the isomers shown for ZnPc1-4 in Figure 4 are analysed in this study, since the calculated spectra of the other possible isomers were found to be broadly similar to those reported.

Conclusions
The rational design of novel phthalocyanine dyes for applications such as DSSCs is complicated by the challenging modelling calculations that are involved in the absence of access to a large computer cluster. This study demonstrates that the simplified sTD-DFT approach can rapidly provide useful information to predict or interpret the spectral trends observed in the Q band region of the UV-vis absorption spectra of π-extended chromophores such as ZnPc1-4 and ZnPc*. However, tt is noteworthy that the extent of the splitting of the Q band into xand y-polarized components is over-estimated in the context of the lower symmetry ZnPc1-4 complexes. It is clear from this series of test calculations for ZnPc complexes with peripheral substituents that introduce large mesomeric and inductive interactions with the π-system of the Pc ligand that the predictions made in the higher energy B band region in both TD-DFT and sTD-DFT calculations need to be treated cautiously. No significant extra insight is likely to be provided by the significantly longer calculation times associated with the conventional TD-DFT approach.