3.1. Molecular Design of the Azo-derivatives:
Azo compounds are usually synthesized with procedures like: azo coupling reaction (coupling of diazonium salts with activated aromatic compounds), Mills reaction (reaction between aromatic nitroso derivatives and anilines) and Wallach reaction (transformation of azoxybenzenes into 4-hydroxy substituted azo-derivatives in acid media) [46
]. They are mostly produced with symmetric structures that have two potential reactive sites that offer the possibility to combine them with other monomers. In this way, polymers capable to react when irradiated by an electromagnetic irradiation can be produced. They are typically sensitive to UV light irradiation and they can have unstable Z
form that can convert back to the more stable E
form within picoseconds.
A preliminary study on the possible substitution pattern for the novel azobenzene derivatives synthesized in this work was performed. The objective was to achieve a well-designed asymmetric structure to:
Produce a molecule with only one reactive site for further modifications and be used as a side-chain group.
Move the excitation wavelength of the E-Z isomerization from the UV region of the absorbance spectrum to the visible light one.
Enhance the stability of the Z form once obtained after irradiation with white light.
To successfully synthesize the first compound AZO1, an azo coupling reaction procedure was chosen with p-Cresol and 4-amino-3-methoxybenzoic acid as precursors. In this specific electrophilic aromatic substitution reaction, a phenol with a methyl group in para
position was selected as the activated arene. The substitution occurs stereotypically in para
to the activating group when a bulky substituent is selected as the attacking species. However, ortho
substitution takes place when the para
position is already occupied by another substituent. The 4-amino-3-methoxybenzoic acid was chosen to give the molecule the desired asymmetry and a reactive site for further modifications. In addition, with the presence of an electron donor group on the aromatic ring, like the methoxy one, we aimed to move the isomerization wavelength to the visible region of the absorption spectrum to trigger the isomerization with white light irradiation [47
]. Lastly, it was reported that the methyl group could enhance the stability of the Z
-isomer prolonging its half-life in the darkness [48
The conditions of the reaction were carefully controlled both in pH and temperature due to the instability of the diazonium salts formed during the process. A mildly acidic or neutral solution is usually needed when an aromatic amine is involved in this reaction: nevertheless, if the solution is too acidic, the concentration of free amine becomes too small and the reaction does not occur. On the other hand, the phenol is usually not sufficiently active for the reaction, but, the more reactive phenoxide ion is a better initiator to generate the target azobenzene, when slightly alkaline conditions are present. Nonetheless, neither amines nor phenols can react in moderately alkaline solution because the diazonium ions are converted into the corresponding diazo hydroxide [49
From the account given by Bandara et al. [50
] regarding a specific set of azobenzene derivatives, the hydroxyl groups can hinder the isomerization by forming hydrogen bonds with the azo group, or other substituent on the opposite ring. We therefore envisaged to apply this same idea for the present work. The AZO1 compound was thus methylated to give AZO2 (Scheme 1
). This modification generates a structure with two electron donor groups in ortho
position on both the phenyl rings, changing the conjugation of the electrons on the –N=N– double bound. This modification was needed given the small isomerization conversion registered for AZO1 E-
isomer when irradiated by white light, as shown in the following sections.
The novel azobenzene compounds were obtained with reasonably good yields and purities. Their structures were characterized by 1H–NMR, 13C–NMR, and UV-Vis spectroscopy.
3.3. UV/Vis Spectra Interpretation:
The UV–visible absorption spectra of the azo-derivatives exhibit two characteristic absorption bands. The first characteristic absorption band is present around 330–320 nm, while the second absorption band is present around 412–325 nm.
Comparing the absorption spectra, it is evident that the methoxy groups attached on both the phenyl rings of AZO2 are blue shifting the λ1
of 26 nm in DMF (Table 2
). This can be explained by the lower conjugation effect that they induce on the –N=N– bond electrons.
To understand the optical properties of the two compounds, an in-silico analysis was carried out with the time dependent density functional (TD-DFT) approach. A DFT geometry optimisation (see computational methods) was performed for the ground state for both structural isomers of AZO1 and AZO2. The simulated electronic vertical transitions of both molecules are displayed on Figure 3
and these will occur in the frontier orbital region mapped out in Figure 4
. The calculated bands (TD-PBE0) compare favourably well to the experimental ones, presenting a redshift in the region of 20–50 nm. The relative intensities are incorrectly described, likely due to solvent interaction effects.
The band maxima corresponding to the electronic excitations are listed in Table 3
and Table 4
for AZO1 and AZO2, respectively. The most intense transition in the visible region can be assigned to the second excited state for both E
-AZO1 and E
-AZO2, each showing oscillator strengths of 0.57 and 0.91 respectively. The conventional assignment of distinctly separate n→ π* and π→ π* transitions is not rigorously valid in the case of E
-AZO1 since there is an admixture of both transitions in the first two bands.
The first and second electronic transitions of E-AZO1 correspond to two maxima calculated to be at 489 and 441 nm and have a majority of π2→ π* character. The other band with significant oscillator strength in this compound is due to a π1→ π* transition occurring at 338 nm.
The spectral features of E
-AZO2 are altogether significantly different in that n→ π* transitions do not admix with the π→ π* type excitations. Since the former transitions are symmetry forbidden, they have a very low oscillator strength while the latter show up as very strong intensity bands. To broaden the study on the optical properties of the two moieties, the samples were dissolved in different solvents, according to the solubility of the different molecules. Figure 5
shows the dipole moments (in Debye) of each solvent compared to the wavelengths of the main peaks (λ1
) of the absorption spectra of the azo-derivatives.
For AZO1 molecule, a broader range of solvents could be tested. The first peculiar aspect is that AZO1 main peak (λ1
) is splitting its signal passing from a more polar solvent like dimethyl sulfoxide to a less polar one like dichloromethane. This can be ascribed to the different mechanisms of excitation followed by the electrons involved in the azo double bond when irradiated by light. In more polar solvents the signal linked to this excitation are overlapped (λ1
), then their wavelength split (λ1
) being better resolved when the polarity of the solvent decreases. The second crucial aspect is highlighted by the hypsochromic shift of λ1
and the bathochromic shift of λ2
of AZO1 molecule in polar solvents. This can be related to the stabilization of the electronic cloud of AZO1, that is favoured by the presence of a less polar solvent like dioxane. This, in fact, can decrease the conjugation effect of the electrons of the azo double bond, shifting lambda to lower wavelengths. AZO2 λ1
is slightly influenced by the solvent polarity, a bathochromic shift is registered passing from a less polar solvent to a more polar one [51
The UV-Vis spectrum was calculated for E
-AZO1 in various implicit solvent dielectric media to determine how the effect that polarity shifts the visible band wavelengths. As can be seen from the values in Supplementary Table S4
, the values of λ1
tend to increase when passing from vacuum to the solvated environment.
Neglecting dispersion interactions between solvent and solute represents the limiting case of solvation in non-polar media. However, the value of λ1 changed only a few for the used selection of solvents, with varying degree of polarity differences. The observed trend for DMSO is not reproduced by the calculation and in fact the value is slightly higher with respect to CH2Cl2 and DMF, but the change is so small that barely any significance can be inferred from it. The failure to account for the experimental trend is a sign that explicit interactions between the solute and the solvent, which are not considered at this level of theory, are essential to explain the subtle band shifts.
3.5. Z–E Thermal Relaxation
The kinetic Z–E
thermal relaxation of AZO1 and AZO2 were investigated at several temperatures in order to determine its thermodynamic parameters. However, in the case of AZO1, thermal relaxation was impossible to study due to the hindered isomerization. A rough estimation of the kinetic constant relative to Z-E
thermal relaxation of AZO1 was attempted from the data obtained in DMF: unfortunately, the obtained value is not reliable because it is falling within the experimental uncertainty of the instrument (for sake of explanation the tentative fitting is shown in the SI with a value R2
= 0.040). The rate of Z–E
thermal relaxation of AZO2 molecule was monitored in the dark at the absorption maximum of its E-
isomer. The thermal Z
isomerization rate, summarized in Table 6
, was determined for AZO2 at different temperatures and solvents. The process was found to be first order in all the studied solvents.
Calculation of the first-order rate constants at various temperatures allowed us to estimate thermodynamic activation parameters such as the activation energy (Ea
), activation enthalpy (ΔH‡
), and activation entropy (ΔS‡
) for thermal Z
reaction using Arrhenius and Eyring equations (equations shown in Appendix A
). We used the linear correlation between the enthalpy and the entropy of activation to determine the mechanism of AZO2 thermal relaxation. In all the solvents, the Arrhenius plots were linear, proving that there is no significant change in the reaction pathway over the examined temperature range. AZO2 compound has a positive enthalpy of activation and negative entropy for all the solvents used (see Table 7
AZO2 in DMF has the highest ΔS‡
absolute value and the lowest ΔH‡
, and therefore exhibits the fastest thermal relaxation [52
]. This is consistent with the calculation that will be shown below. In order to understand the dynamic photochemical and thermal processes involved in the isomerisation of AZO1 and AZO2 a choice of appropriate computational strategy was paramount. The employment of density functional theory is unsuitable in this instance because there is no guarantee, that at every stage in the transformation within the specific excited state hypersurface, the electron wavefunction of the ground state remains singly determinantal. The best methodology is to adopt multi-configurational self-consistent field theory [36
] with a posteriori perturbational corrections for a good description of dynamical electron correlation such as the NEVPT2 approach [40
] (see Computational Methods).
Two conventional mechanisms usually adopted to describe the isomer change involve the either the twisting of the ∡(N=N–C) angle (inversion) or a rotation of the planes that make up the ∡(C-N=N–C) dihedral (rotation). Only the latter was found to exhibit more favourable energies for the photochemical conversion. This contradicts the hydrogen bonding inhibition scenario as presented by Bandara et al. [50
], since rotation of the molecule around the –CNNC– dihedral planes does not necessarily require the termination of the intramolecular hydrogen bond with the adjacent hydroxyl group. An alternative explanation is given below to account for the slower rates of photo-switching.
The alternative twisting scenario may be found in the supporting information section
. More elaborate reaction coordinates combining angular and dihedral changes have also been studied for unsubstituted azobenzene [56
As seen from Table 3
and Table 4
the second excited state (S2
) presents the highest oscillator strength for both AZO1 and AZO2: this will be the most efficient conversion pathway.
A relaxed potential energy surface (PES) scan of the –CNNC– dihedral at the QD-NEVPT2(4,3)//CASSCF (4,3)/def2-TZVP level of theory is presented in Figure 7
and in Figure 8
for AZO1 and AZO2, respectively. Therein the energies of each of the lowest six singlets were monitored throughout this reaction coordinate. The main composition of the states at the beginning of the process (E
-isomer) is qualitatively described with different colours. The curves are qualitatively similar to what was previously reported for azobenzene [57
The photochemical process originates with the promotion of S0
) which can allow the system enough flexibility to distort the dihedral once the π bond strength has a formal bond order of zero. However, due to orbital polarisation there remains a marginally bonding interaction and some thermal energy is still required to break it. This barrier is computed to be 36.0 kJ mol−1
in the case of AZO1 and 27.4 kcal mol−1
for AZO2 and takes place at ∡(CNNC) = 135° and ∡(CNNC) = 143° respectively. One metric to measure the π symmetry bond strength from the electronic density is the natural localized molecular orbital (NLMO) bond order taken from the CASCI(4,3) density matrix. A comparative analysis of the azo bond orders in ground (S0
) and excited (S2
) states is displayed on Table 8
. It may be seen that the calculated π bond order is identical from the outset (S0
). The S0
excitation suppresses the π bond strength differently in AZO1 and AZO2. In AZO1, the π bond remnant is stronger than in AZO2 thereby justifying the additional energetic requirement for the dihedral rotation and rendering AZO1 kinetically more inert in the S2
Disregarding thermal contributions to the transition state energies, i.e., ΔE‡(S2) ≈ ΔG‡(S2) and assuming a high quantum yield for the S2 excitation, the Eyring equation yields a ki(AZO1)/ki(AZO2) ratio of 3.1 × 10−2 using the above calculated values which is reasonably close to experimental one (4.8 × 10−2). Overall, a 9 kJ mol−1 energy difference leads to a decrease in the reaction rate constant close to two orders of magnitude.
Towards more acute angles one hallmark feature [57
] of azobenzene derivatives comes to the fore, which is the conical intersection (CI) allowing S2
to configurationally admix with S0
leading to the cis isomer, in the region 80° < ∡(–CNNC–) < 100° as the lone pairs (n) become degenerate with the π orbitals. The energy gap in this nonadiabatic coupling (
) is 98.3 kJ mol−1
in AZO1 and 83.1 kJ mol−1
in AZO2. From the Landau–Zener relation [58
], it is known that the probability of transition between the two adiabatic surfaces S0
can be given by
is a collection of constants incorporating nuclear velocity, and the difference in reaction coordinate slope in each adiabatic state.
is the squared energy gap between the two states at the avoided crossing. As the values of
are considerable for both molecules (98.3 kJ mol−1
in AZO1 and 83.1 kJ mol−1
in AZO2), the probability of radiationless transition is practically zero. Thus, the S0
transition will require the emission of one photon at 1217 nm (AZO1) and 1437 nm (AZO2) wavelengths (infra-red region).
The potential energy landscapes also afford an overview of the thermal back-isomerization which has the values +74.9 and +74.4 kJ mol−1. These values are in good agreement with the relaxation activation energies determined experimentally even neglecting the necessary thermal corrections (zero point, vibrational, rotational and translational).
In light of the current data, the presence of the –OH group in AZO1 cannot be considered to play a direct role in the rapid back isomerization of AZO1 if one considers this neutral tautomer alone as alluded to in the literature [50
]. In view of the acidic properties of AZO1, all the possible tautomeric zwitterionic isomers were optimized at the PBE0/TZP level and their optical properties assessed (see Supporting Information section
Our results show that a low energy tautomer may play a role in the isomerization process if the following equilibrium occurs (Scheme 2
). This zwitterionic isomer should be more favored in a polar solvent medium [ΔE(DMF) = 3.6 kcal mol−1
Our results show that the presence of the hydroxy group does indeed influence the conversion process in AZO1, but not by creating strong hydrogen bonds with nitrogen but by instead allowing for the existence of diazo-phenolate that has distinctly unique photodynamic properties (Supplementary Figure S6