On the unexpected mechanism of isomerization in tautomerizable azo photoswitches

Abstract

In recent years, there has been no shortage of research achievements in light-responsive materials based on azobenzene photoswitches. Of growing interest is the ability to reversibly tune the competing “dark" thermal cis-trans back-isomerization through protonation effects. The hydroxy-substituted azobenzenes are well-known for their complex pH-dependent behavior, including azo-hydrazone tautomerism. Presently, experimental studies rationalize only qualitatively the marked acceleration in thermal switching upon acquiring the hydrazone tautomer, while the results of theoretical treatments have experienced a persistent cusp problem in calculated potential energy surfaces. Here, using density functional theory, spin-flip, and multireference wavefunction quantum chemical methods, we provide for the first time a comprehensive explanation of thermal switching in the hydrazone tautomer. We show that, through concerted torsion of two dihedral angles, the hydrazone tautomer unexpectedly acquires a maximally puckered transition state, enabling rapid rotation of the entire system. This study demonstrates the exploitative advantages of protonation for tuning thermal isomerization in azobenzene photoswitches.

Azobenzene photoswitches have garnered significant attention for their potential in light-responsive materials, yet challenges remain in understanding their thermal cis-trans isomerization. Here, the authors employ advanced quantum chemical methods to elucidate the rapid thermal switching mechanism in the hydrazone tautomer, revealing puckering of the tautomeric nitrogen as a key factor for accelerating isomerization kinetics.

Introduction

Azobenzene photoswitches are an indispensable tool in the design of light-responsive materials. When irradiated by light, azobenzenes undergo a reversible photoisomerization about the azo bond (N=N) between planar trans and bent cis isomers^1,2. This robust process, accompanied by significant changes in shape, spectra, and dipole moment, is used to photocontrol the properties of materials in a myriad of research areas, including chemical sensors³, biomimetic coatings⁴, photonics⁵, and photopharmacological drugs⁶. The isomerization process is also thermally reversible, typically proceeding in the cis-to-trans direction. The timescale of thermal isomerization varies largely from one photoswitch to another. Indeed, photoswitches have been reported with thermal half-lives from nanoseconds⁷ to years⁸. Yet, the precise engineering of azobenzene-based materials is predicated on exquisite control of the thermal kinetics to fast or slow temporal regimes. Modification of the azobenzene scaffold is a foremost albeit restrictive strategy for controlling the thermal lifetime, leading to the emergence and classification of azobenzene, aminoazobenzene, and psuedo-stilbene-type classes with distinct spectral and thermal kinetics properties^1,2. However, many azobenzene derivatives are also remarkably sensitive to their local environment. As a result, pH has emerged as a simple tool for reversibly tuning thermal kinetics over several orders of magnitude^9–12. Although the photophysical effects of protonation are well documented, the thermal isomerization process is highly contested and, as such, remains an enduring subject of theoretical and experimental investigations. New insights are emerging in how protonation can dramatically alter the behavior of azobenzene photoswitches, which necessitates further exploration to guide improved materials design with protonation as a tuning tool. For example, we found that protonation of an azopyridine photoswitch as a consequence of otherwise routine experimental conditions markedly accelerates thermal isomerization kinetics¹³. Using density functional theory (DFT), we rationalized a six orders of magnitude acceleration by inductive weakening of the azo bond, as well as a mechanism favoring inversion instead of a multistate rotation mechanism preferred by the neutral photoswitch.

Hydroxy-substituted azobenzenes are another class where the local environment, notably pH, can give rise to interesting behavior. Extensive studies have shown that thermal isomerization kinetics are fast (seconds or less) and strongly solvent dependent^14,15, as well as pH dependent, with complex behavior emerging when various ionized species exist at equilibrium^16,17. In addition, azo-hydrazone tautomerism is a prominent feature in this class of photoswitches. Here, tautomerization involves proton transfer from the hydroxyl group to the azo bond, facilitated by protons available from the solvent, dimerization through intermolecular hydrogen bonds^15,18,19, or photochemical excited state intramolecular proton transfer²⁰. Experimental studies support a thermal isomerization mechanism involving a nonphotochromic hydrazone intermediate, an energetically “cheaper” process than direct cis-to-trans isomerization which gives rise to rapid thermal kinetics^21–23. The profound acceleration is often rationalized only qualitatively by weakening of the azo bond in the hydrazone form, enabling facile rotation and recovery of the trans isomer. In addition, theoretical treatments of the isomerization process have encountered singularity problems. Sharp cusps appear in calculated potential energy surfaces (PES), which have been attributed to multi-configurational character of the transition state (TS)²⁴ and umbrella inversion of the pyramidal tautomeric nitrogen¹², or they were not addressed²⁵. To the best of our knowledge, resolution of this singularity problem and a complete, rigorous treatment of thermal isomerization in the hydrazone form remain to be explored.

To gain further insight into protonation effects on hydroxyazobenzenes and thermal isomerization of the hydrazone tautomer, we studied the pseudo-stilbene-type ("push-pull”) photoswitch 5-(8-hydroxyquinolinyl)-phenylazo-4-sulfonate, abbreviated HPAS. The structure of HPAS as well as some of its chemical properties are summarized in Fig. 1. The 4-phenylazo-1-naphthol scaffold is known to exhibit an equilibrium of azo and hydrazone tautomers in solution^21–23. In addition, one of the phenyl rings has been replaced with the metal-coordinating ligand 8-hydroxyquinoline. Thus, possessing both pH responsivity and versatile coordinative abilities towards transition metals, HPAS has been used to devise light-responsive hydrogels²⁶, metal coordination polymers^27,28, and colourimetric metal sensors²⁹. Despite these successful material applications, both experimental and theoretical investigations of the thermal isomerization process are lacking. Beyond HPAS, the azoquinoline scaffold also has far-reaching applications, such as the design of photoswitchable metal organic frameworks³⁰. Therefore, the HPAS photoswitch is not just an ideal theoretical model, but a photoswitch with immediate materials applications.

Fig. 1. Structure and properties of the HPAS photoswitch.

Push-pull substitution pattern and azo-hydrazone tautomeric equilibrium of the model hydroxyazobenzene photoswitch HPAS. The key protonation sites studied in this work are highlighted in blue. The electron-donating group (EDG) and electron-withdrawing group (EWG) are circled in red.

In this work, we investigate theoretically the protonated forms of HPAS and their thermal isomerization mechanisms with high-level multireference DFT and ab initio wavefunction methods. In addition, we provide for the first time a detailed description of thermal isomerization in the hydrazone form, with implications for the strategic design of azobenzene-based materials incorporating hydroxy-substituted azobenzenes and other derivatives exhibiting azo-hydrazone tautomerism.

Results and discussion

Acid–base chemistry and tautomeric equilibria of HPAS

To explore the effect of protonation on the thermal isomerization of HPAS, we first determined the protonated forms of the cis and trans isomers. Starting from the fully deprotonated structure, we considered all possible and relevant protonation sites as shown in Fig. 1. These sites included the azo group (N_azo), quinoline nitrogen (N_8HQ), phenoxide oxygen (O_8HQ), and sulfonate group (SO₃) (Table S3–S10, Supplementary Data 1). We excluded the azo nitrogen nearest the quinoline ring from analysis on the basis of resonance arguments, as the resultant charge separation created from protonation of this site cannot be relaxed through the π system. We carried out these protonation studies in a C-PCM model of water with the M06-2X functional, a theory level which can predict protonation sites and tautomeric equilibria with reasonable accuracy for azobenzenes and related photoswitches compared to more elaborate explicit solvation approaches^31–33.

For both isomers, the phenoxide oxygen, O_8HQ, was the first protonation site, where the free energy was lower than the next most favorable site, N_azo, by at least 10.4 kJ mol⁻¹ and 2.5 kJ mol⁻¹ for any given conformer of the trans and cis isomers, respectively (Table S7). This first protonation forms a hydroxyl group which participates in a stabilizing hydrogen bond with the nearby lone pair on the quinoline nitrogen. By twisting the hydroxyl group out of the hydrogen bond, we estimate an upper-bound strength of the hydrogen bond as 13.0 kJ mol⁻¹ and 11.9 kJ mol⁻¹ for the trans and cis isomers, respectively. In addition, for the cis isomer, there exists a tautomeric equilibrium between azo and hydrazone forms (Fig. 2, ΔG_hyd−azo = 2.0 kJ mol⁻¹ for cis-HPAS^1-), where a proton is transferred from the hydroxyl group to the azo group, significantly weakening the azo double bond. This proton transfer and the inevitable adjustment of electron density surrounding the azo bond have significant effects on the electronic structure of the photoswitch system. Therefore, we sought to rationalize this tautomeric equilibrium by performing a population analysis in the natural bond orbital (NBO) framework³⁴ on the converged wavefunctions of both tautomeric forms. For each tautomer and isomer, we considered only the minimum-energy conformation (Figs. S1–S3).

Fig. 2. HPAS exhibits a tautomeric equilibrium between azo and hydrazone tautomers.

Protonated forms and tautomeric equilibria for the HPAS photoswitch predicted by DFT. Structures are colored according to the approximate pH they are anticipated to appear³⁶. Calculated ΔG ( = G_hyd − G_azo) of tautomerization is given for cis and trans isomers (only cis structures shown for simplicity) of each protonated form, considering a Boltzmann-weighted average of all possible conformations for each tautomer (Figs. S1–S3).

Several lines of evidence emerged explaining the competitiveness of the azo nitrogen protonation site for the cis isomer of HPAS. First, the p character of the tautomeric nitrogen lone pair NBO in the azo form increases from 1.58 in the trans isomer to 1.80 in the cis isomer. As p character of this NBO increases, the lone pair electrons on this nitrogen are held less strongly by the nucleus due to poorer nuclear penetration afforded by p orbitals. As a result, these electrons are more easily shared with an incoming proton, increasing the basicity of this site in the cis isomer. Second, once the tautomeric nitrogen is protonated, negative charge is localized to this site. Consequently, there is a larger charge separation created at the other azo nitrogen and nearby carbon atoms in the phenyl and 8-hydroxyquinoline rings compared to the azo form (Fig. 3a, Fig. S4), where this separation is minimized. Such localization of charge enables favorable electrostatic interactions in polar solvent. We would therefore anticipate the hydrazone form to be less pronounced in nonpolar solvents, which is a well-known factor explaining the preference of either tautomeric form³⁵. Third, in the hydrazone form, the tautomeric nitrogen now accommodates its lone pair of electrons in a nearly unhybridized p orbital. This lone pair is poised to strongly delocalize into neighboring π systems and increase resonance stability (Fig. 3b), which we observed from the NBO occupancy of 1.710, a significant departure from ideal Lewis 2.0 occupation. We additionally observed this effect by performing energetic deletions of all non-Lewis-type NBOs, demonstrating a strong preference for the cis hydrazone form to be stabilized by resonance delocalization effects compared to the cis azo form (Table S1).

Fig. 3. Tautomeric equilibrium is driven by various electronic structure effects.

a Structures of the azo and hydrazone tautomers of cis-HPAS^1- colored by natural charge from [−1, +1]. Charges are explicitly labeled on select atoms, illustrating the localization of charge in the hydrazone tautomer. b Delocalization of the N_azo lone pair into acceptor π^* NBOs in the cis hydrazone tautomer. c Comparison of steric hindrance between the cis and trans isomers of the hydrazone tautomer. The electronic energy of the idealized Lewis structure, E_L, predicted by NBO is lower for the cis isomer by 32.3 kJ mol⁻¹.

Finally, the cis hydrazone form relaxes steric hindrance that otherwise occurs in the trans hydrazone form (Fig. 3c). It is well known that sterically, the cis isomer of azobenzenes is less stable than the trans isomer due to interactions between the bulky ring substituents. The electronic energy of the idealized Lewis representation determined by NBO energetic deletion analysis, E_L, provides a metric to assess steric hindrance. By comparing E_L of the cis and trans isomers of the hydrazone form, we found that the typical steric dichotomy in azobenzenes is reversed (Table S2). Steric hindrance between the tautomeric proton and the nearby 8-hydroxyquinoline ring in the trans isomer is relaxed in the bent cis isomer. This steric effect is likely negated for the class of ortho-hydroxy-substituted azobenzenes, where the tautomeric proton can engage instead in a stabilizing hydrogen bond with a ketone oxygen.

When investigating the second protonation site of HPAS, we considered separate azo and hydrazone protonation “channels”. For both isomers, these channels tended to converge on the quinoline nitrogen, N_8HQ, as the most energetically favorable protonation site. An azo-hydrazone tautomeric equilibrium still exists in both cases, however the equilibrium now favors the hydrazone form instead of the azo form, signified by the change in sign of ΔG_hyd−azo (Fig. 2). According to energetic deletions of non-Lewis-type NBOs for both tautomeric forms, the shift in tautomeric equilibrium is driven principally by steric effects (Table S1). The interaction of the hydroxyl group and protonated quinoline nitrogen introduces a distortion in the 8-hydroxyquinoline ring which is more severe in the azo form than the hydrazone form (Fig. S5). As a result, the strong capacity for resonance delocalization in the hydrazone form now tips the scale of relative stability. In this way, the act of protonating the quinoline nitrogen appears to behave as a “switch” controlling the equilibrium of both tautomers.

Overall, the protonation sequence for the trans isomer agrees well with experimental data, which have identified the same sites with measured acid dissociation constants of 7.94 ± 0.4 and 3.78 ± 0.3 for the hydroxyl group and quinoline nitrogen, respectively³⁶. To our knowledge, no such experimental data is available for the cis isomer. While an azo-hydrazone tautomeric equilibrium has been reported for trans-HPAS around neutral pH conditions³⁶, there is no quantitative data pointing to the preference of one tautomeric form over the other. Quantum chemical calculations on 4-hydroxyazobenzene²⁰ and 4-phenylazo-1-naphthol²³, however, support our finding that the azo tautomer predominates for the neutral pH form of the photoswitch (HPAS^1-). Given that the accuracy of the M06-2X functional for predicting tautomeric equilibria in azobenzene dyes is estimated to be about 0.5 kcal mol⁻¹ from a benchmarking study³⁷, we expect that the results here can only lend support to a qualitative equilibrium of both tautomeric forms, but not a quantitative predominance of one over the other. All raw data for the investigation of protonation sites and tautomeric equilibria are provided in Tables S3–S10 of the Supporting Information.

DFT and ab initio characterization of thermal isomerization in deprotonated HPAS

To establish a frame of reference for thermal isomerization of HPAS, we first studied the fully deprotonated form of the photoswitch, HPAS²⁻ (Fig. 2, right), which is predominant under basic conditions. Canonical mechanisms of azobenzene thermal isomerization include in-plane inversion, out-of-plane rotation, or some hybrid mechanism of the two^1,2. To study these mechanisms more rigorously, we employed a multireference formulation of DFT, spin-flip time-dependent DFT (SF-TDDFT)³⁸, which has been successfully applied to the study of rotation mechanisms in azobenzene and related derivatives^39–41. SF-TDDFT is capable of describing the multiconfigurational character of rotational TSs where π bonding and π^* anti-bonding orbitals become quasi-degenerate. Kohn-Sham DFT will otherwise produce a singularity due to the qualitative incorrectness of a single-determinant wavefunction where only the lowest-energy Kohn-Sham spin orbitals are occupied. We also studied inversion through either azo nitrogen at the same level of theory. We dismissed inversion mechanisms through either azo nitrogen as the calculated barrier heights far exceeding 100 kJ mol⁻¹ are inconsistent with experimental data for typical push-pull-substituted azobenzenes (Fig. S6). Indeed, it is well established that push-pull azobenzenes, such as HPAS²⁻ discussed here, have a propensity for rotation mechanisms due to resonance weakening of the azo double bond. When studying the rotation pathway on the ground state S₀ and first excited T₁ PES, we encountered a failure of SF-TDDFT, shown in Fig. 4. Severe spin contamination (0.5 < 〈S²〉 < 1.5) approaching the TS plagued both S₀ and T₁ surfaces, rendering energetics and identification of states completely unreliable. Further, any information about potential singlet-triplet crossings was lost. In their work on an arylazopyrazole-based photoswitch, Singer et al. reported similar significant spin contamination in the calculated T₁ and S₀ states, which had 〈S²〉 values of 1.44 and 0.78, respectively⁴¹. In contrast to our previous work on azopyridine where SF-TDDFT was reliable in the neighborhood of the TS¹³, HPAS²⁻ is highly asymmetric and, as we will discuss next, displays pronounced open-shell character along the reaction coordinate. We suspect both factors, along with the known spin problems associated with unrestricted Kohn-Sham reference states, contributed to the challenges faced by SF-TDDFT here. Although spin-adapted spin-flip methods have been proposed which largely correct for spin contamination, to our knowledge, the majority of azobenzene theoretical reports employing SF-TDDFT have relied solely on the original formulation of this method^39–41.

Fig. 4. SF-TDDFT method fails to describe thermal isomerization of a push-pull photoswitch.

Top: Relaxed surface scan of HPAS^2- through ∠CNNC rotation on S₀ and T₁ surfaces calculated with the SF-TDDFT method. Energies are reported relative to the cis isomer (Table S11). Bottom: severe spin contamination (0.5 < 〈S²〉 < 1.5) in the neighborhood of the TS (75^∘ < ∠CNNC < 105^∘, gray shaded region).

To rectify the spin-contaminated PES, we investigated the rotation pathway instead with the complete active space self-consistent field (CASSCF) method. The choice of active space is a crucial step in the successful application of CASSCF. We selected a CAS(16,14) active space encompassing key orbitals in the π system of HPAS^2-, including the azo π and π^* orbitals, as well as the azo lone pair orbitals, n₁ and n₂. While the routine active space for azobenzenes is CAS(14,12), where two π orbitals and their π^* counterparts are included on each ring^39,40,42, we found that the extended π system of 8-hydroxyquinoline necessitated an additional pair of π and π^* orbitals on this ring system (Fig. S7, S8).

Energetic single points on SF-TDDFT-optimized geometries yielded a qualitatively correct PES, shown in Fig. 5. Analysis of the S₀ PES revealed that singlet biradical character characteristic of the TS emerges early along the reaction coordinate, for dihedral angles ranging from 60° to 120°. It is in this same region that SF-TDDFT calculations began behaving anomalously. To assess the barrier for thermal isomerization and the relative energy between S₀ and T₁ surfaces, we included dynamical correlation effects with n-electron valence state perturbation theory in its quasi-degenerate formulation (QD-NEVPT2).

Fig. 5. Thermal isomerization of a push-pull photoswitch avoids singlet-triplet crossings.

Energetic single points for ∠CNNC rotation of HPAS^2- on S₀ and T₁ surfaces calculated with SA(2,1)-CAS(16,14)/QD-NEVPT2 on select SF-TDDFT-optimized geometries. Energies are reported relative to the 10^∘ geometry (Table S12). The S₀/T₁ spin-orbit coupling magnitude calculated with the CASSCF wavefunction is plotted for each geometry.

Whereas for azobenzene, there is a defined minimum in T₁ below the S₀ TS^39,40,42, for HPAS²⁻, the S₀ and T₁ surfaces no longer cross. As a result, there is no energetic advantage for a singlet-triplet crossing mechanism in this photoswitch. There is not yet an abundance of experimental data suggesting that singlet-triplet crossing mechanisms always prevail over rotation solely on the S₀ PES for substituted azobenzenes. Based on CASPT2 calculations, Aleotti et al. provided compelling evidence that rotation on S₀ may still be competitive with the multi-state mechanism across different derivatives and experimental conditions⁴³. In our calculations, we observed that the composition of the first excited singlet, S₁, is predominately π→π^* character, rather than n→π^* as for unsubstituted azobenzene. It is well known in the azobenzene literature that push-pull substitution red-shifts the π→π^* transition, until inversion of the π→π^* and n→π^* states². This inversion has been attributed to simultaneous destabilization of the π-bonding orbital on the azo group and stabilization of the corresponding π^*-anti-bonding orbital, such that push-pull derivatives excited to the ππ^* state (S₂) behave like unsubstituted azobenzene excited to the nπ^* state (S₁)⁴⁴.

We observed a similar effect in HPAS²⁻, operated primarily by the strong electron-donating phenoxide group introducing anti-bonding character into the azo π-bonding orbital compared to unsubstituted azobenzene. In the neighborhood of the TS (80° < ∠CNNC < 100°), S₁ mixes strongly with S₀. As a result, the π→π^* character acquired by S₀ would suggest poor coupling with the triplet, which is also of π→π^* character over the entire reaction coordinate. From our initial SF-TDDFT calculations, we observed a similar trend where the weight of the n→π^* transition for the S₁ state tends to decrease in favor of the π→π^* transition approaching the TS. These qualitative observations, combined with the magnitude of spin-orbit coupling computed from the CASSCF wavefunction (Fig. 5), suggest that the relativistic impetus for singlet-triplet crossings is also low.

Thermal isomerization of the hydrazone form through multiple degrees of freedom

If we assume that tautomerization proceeds in water via a solvent-assisted mechanism with a fast pre-equilibrium compared to thermal isomerization kinetics, then the thermal lifetime of the intermediate hydrazone cis isomer will principally determine the kinetics of thermal isomerization at neutral pH. We build off our previous work on the azobenzene food dyes²⁴ by exploring the rotation pathway of the HPAS^1- hydrazone form using the multireference SF-TDDFT method. Initial attempts to locate a TS involved full geometry optimization under constraint of the ∠CNNC azo dihedral angle, which is routine for azobenzene photoswitches. However, this initial TS search produced a PES with a sharp cusp, corresponding to sudden inversion of the tautomeric nitrogen as the geometry relaxed to the trans isomer (Fig. 6a). In a second attempt, we constrained instead the dihedral angle involving the H-N-N=C atoms of the hydrazone linkage, which we term the hydrazone dihedral angle, ∠HNNC. This dihedral, like ∠CNNC, passes through approximately 180° of rotation in switching from the cis to trans isomer. Unfortunately, a cusp once again emerged (Fig. 6b). Analysis of the SF-TDDFT ground state for both relaxed surface scans indicates that rotation of the hydrazone form is well described by the π² electronic configuration on the S₀ surface, ruling out multiconfigurational character as a suspect for the cusp. Furthermore, we attempted to introduce an explicit water molecule to hydrogen-bond the labile tautomeric proton along the reaction coordinate. The sharp cusp in the PES persisted (Fig. S9).

Fig. 6. Thermal isomerization of the HPAS¹⁻ hydrazone tautomer in two torsional degrees of freedom proceeds through a TS with a maximally puckered azo nitrogen.

One-dimensional a ∠CNNC and b ∠HNNC relaxed surface scans of HPAS^1- on S₀ surface calculated with the SF-TDDFT method. Sharp cusps correspond to inversion of the tautomeric nitrogen. Overlaid geometries illustrate the abrupt inversion process. c Multidimensional ∠CNNC and ∠HNNC relaxed surface scan. Overlaid with the PES is the MEP calculated with the NEB-CI algorithm. Empty regions on the plot correspond to unphysical combinations of ∠CNNC and ∠HNNC where geometry optimizations failed. Points along the MEP belonging to each isomerization stage are indicated on top of the plot. d NBO analysis of all images on the MEP. Top: spⁿ hybridization of the tautomeric nitrogen in the N-H single bond of the hydrazone linkage. Bottom: natural localized molecular orbital bond order of the N-N bond in the hydrazone linkage. Torsion stages one and three are shaded gray, and inversion stage two is shaded blue.

Evidently, the inversion of the tautomeric nitrogen is essentially barrier-less, causing the constrained optimization to “cut the corner” of the full multidimensional PES. As a result, we found that a complete description of thermal isomerization of the hydrazone form requires two rotational degrees of freedom: 1) the ∠CNNC dihedral angle and 2) the ∠HNNC dihedral angle. We therefore calculated a multidimensional PES by full geometry relaxation under constraint of both dihedral angles, shown in Fig. 6c; to reduce computational effort, we scaled down to a smaller split-valence def2-SVP/ma-def2-SVP mixed basis set for all optimizations. Spin contamination across all geometries on the S₀ PES was low, 〈S²〉 < 0.2 (Fig. S10). Armed with intimate knowledge of the PES landscape for the hydrazone form, we calculated a minimum energy path (MEP) and first-order saddle point (TS) with the climbing image nudged elastic band (NEB-CI) method using energy-weighted springs⁴⁵. The smooth convergence of NEB-CI was highly dependent on supplying an accurate guess for the TS from the multidimensional scan. In addition, we found that stiff springs (k_low = 0.1 Eh Bohr⁻² and k_high = 1.0 Eh Bohr⁻²) were required to cluster a sufficient number of images in the narrow saddle point region.

Although inversion of the tautomeric nitrogen is understood in some literature reports as an inevitable step of thermal isomerization^12,24,46, our detailed analysis here reveals fully the distinct stages of thermal isomerization, including the inversion process and molecular strategies leveraged by the photoswitch to promote fast relaxation. In the first stage of isomerization, both ∠CNNC and ∠HNNC dihedral angles are exercised to acquire the TS geometry. While umbrella inversion of the tautomeric nitrogen initially suggested a flat sp²-hybridized TS similar to umbrella inversion in ammonia, the tautomeric nitrogen is in fact maximally puckered, as shown by an NBO analysis of all images on the MEP (Fig. 6d). Indeed, the hydrazone form exploits puckering of the tautomeric nitrogen to reduce the bond order of the azo group. At the TS geometry, the bond order from NBO analysis has reached a minimum, dropping below one (Fig. 6d). As a result, this strategy promotes facile rotation of the entire system. In passing, we note that while the M06-2X functional from earlier protonation studies identified an early TS with a puckered tautomeric nitrogen in the cis isomer, the BH&HLYP functional for SF-TDDFT calculations found instead the tautomeric nitrogen to be almost flat. Based on this NBO analysis, we can clearly envision design strategies in which substitution with additional π-donating substituents on the 8-hydroxyquinoline ring can enhance the puckering of the tautomeric nitrogen, thereby reducing the thermal isomerization barrier even further. This design strategy is supported by the experimental work of Garcia-Amorós et al.¹⁴, who found that substitution of multiple OH groups on the same phenyl ring led to a rate enhancement via a cooperative effect. The hydrazone puckering strategy, however, exists in competition with resonance delocalization of the nitrogen lone pair into the neighboring C=N π system, which tends to diminish as the TS is acquired (Fig. S11). Such a tradeoff may be counteracted in the class of ortho-hydroxy-substituted azobenzenes, where the tautomeric proton can participate in a stabilizing hydrogen bond with the ketone oxygen. The mechanism for ortho derivatives remains to be explored in the analysis framework presented here.

In the second stage of thermal isomerization, the tautomeric nitrogen inverts while ∠CNNC remains mostly constant at approximately 130°, followed by a third final stage where both dihedral angles are exercised concurrently again to recover the planar trans isomer. Second-order perturbation analysis of the Fock matrix in the NBO formalism reveals that the hydrazone form may prefer a distinct inversion stage to maximize stabilizing donor–acceptor interactions with the phenyl ring π system as the trans isomer is recovered (Fig. S11). Details of the converged MEP, including an animation, coordinates, and electronic energies, are provided in Supplementary Movie 1, Supplementary Data 2, and Table S13, respectively.

Using the TS geometry located by NEB-CI for a subsequent full TS optimization, we observed an imaginary mode corresponding to torsion of both ∠CNNC and ∠HNNC (Supplementary Movie 2, Table S14), providing further evidence that these two degrees of freedom are jointly required to describe thermal isomerization of the hydrazone tautomer. In addition, we estimate a final activation barrier of 66 kJ mol⁻¹ for thermal isomerization of the hydrazone form. This predicted barrier is significantly lower than for unsubstituted azobenzene, and falls in the range of barriers reported for 4-hydroxyazobenzene^14,15,18 and 4-phenylazo-1-naphthol²¹ scaffolds in polar protic solvents (≈40–70 kJ mol⁻¹). Initially, we were surprised that the barrier for HPAS^2- was lower than HPAS^1- (Fig. 7), despite the former not possessing a hydrazone form. However, a difference of 4 kJ mol⁻¹ is not significant enough on which to base a qualitative conclusion. Presently, there are no thermal kinetics studies on the HPAS photoswitch or similar push-pull hydroxyazobenzenes in basic aqueous solution, which is the subject of future work in our group. In summary, the hydrazone puckering strategy enables HPAS^1- and other hydroxyazobenzenes to achieve thermal isomerization barriers similar to push-pull derivatives, such as HPAS^2- discussed here (Fig. 7).

Fig. 7. The hydrazone thermal barrier is comparable to push-pull systems.

Skeleton PES for the rotation of the HPAS^1- hydrazone tautomer (red) compared to HPAS^2- (blue). Predicted barriers and structures of cis, trans, and TS geometries are shown for each protonated state.

Finally, regarding thermodynamic activation parameters, extensive characterization of hydroxyazobenzenes by Velasco and coworkers^14,15 has highlighted their large and negative activation entropies for thermal isomerization, on the order of −150 J mol⁻¹ K⁻¹. Analysis of the natural charges of HPAS along the hydrazone isomerization MEP revealed a localization of charge to the azo nitrogens and surrounding atoms in the hydrazone linkage upon acquiring the TS geometry (Fig. S12). Alongside the well-described abilities for hydroxy-substituted azobenzenes to participate in hydrogen bonds and dimer formation, this charge localization can stimulate increased organization of solvent molecules in the TS, which may in part drive the large negative activation entropies uniquely exhibited by these derivatives.

Conclusions

In this work, we provide a detailed theoretical investigation of thermal isomerization mechanisms in different protonated states of a model hydroxyazobenzene photoswitch, including, to the best of our knowledge, the first complete treatment of the hydrazone tautomer. On the grounds of spin-orbit coupling and energetic arguments obtained from CASSCF/QD-NEVPT2 calculations, we found that rotation of the fully deprotonated photoswitch proceeds on the S₀ PES, contributing to the growing evidence that singlet-triplet rotation mechanisms do not prevail in all azobenzene photoswitches. Finally, we detailed the mechanism of thermal isomerization in the hydrazone form. By constructing a full multidimensional PES with two strategically chosen dihedral angles, ∠CNNC and ∠HNNC, we obtained a complete picture of the mechanistic landscape and avoided the cusp problem, an unresolved artifact frequently encountered in the literature. We discovered that the hydrazone form utilizes puckering of the tautomeric nitrogen as a tool to weaken the azo bond significantly approaching the TS, thus achieving activation barriers similar to push-pull substituted derivatives ( ≈ 70 kJ mol⁻¹). Overall, this study has implications for the strategic design of hydroxyazobenzenes and myriad other azobenzene photoswitches exhibiting azo-hydrazone tautomerization. Furthermore, we shed light on the intrinsic complexities of thermal isomerization as a function of protonation, an aspect of photoswitch design that is receiving increasing attention towards tunable azobenzene-based materials.

Methods

Computational methods

All quantum chemical calculations were performed with the Orca 5.0 software⁴⁷. Ground state cis and trans geometries and their possible protonated forms were optimized with DFT in a C-PCM model of water⁴⁸ using the M06-2X functional⁴⁹. No dispersion corrections were included for M06-2X given it was originally parametrized and optimized for noncovalent interactions and dispersion-bound complexes⁴⁹. Furthermore, M06-2X was found to cover dispersion interactions well in a study on dispersion-bound complexes⁵⁰. Calculations used a mixed basis set consisting of the minimally augmented basis ma-def2-TZVP^51,52 for oxygen and def2-TZVP basis⁵¹ for all other atoms. Coulomb integrals were evaluated in the Resolution of Identity (RI) approximation⁵³ with a def2/J auxiliary fitting basis set⁵⁴, whereas exchange integrals were evaluated analytically to minimize noise in the gradient during optimizations. All geometries were verified as true local minima by vibrational analysis, which yielded only real frequencies. Select structures, where noted, were characterized in the NBO formalism using the NBO 7.0 software⁵⁵ at the same level of theory. All structures, NBOs, molecular orbitals, and frequencies were visualized with Chemcraft (https://www.chemcraftprog.com).

Thermal isomerization TS searches for rotation and inversion pathways were conducted using the SF-TDDFT method³⁸ in the Tamm-Dancoff approximation⁵⁶ with the Becke “half and half” BH&HLYP functional^57,58, D3BJ dispersion correction (s₆ = 1.000, α₁ = 0.2793, s₈ = 1.0354, α₂ = 4.9615)^59,60, and an identical mixed basis set as ground state calculations. The BH&HLYP functional is a standard choice for SF-TDDFT calculations⁶¹. Continuum solvation effects were included in the linear response C-PCM scheme using equilibrium dielectric conditions in which excited states are treated as completely solvated⁶². The ground state singlet S₀ and first excited triplet T₁ PES were calculated as spin-flip excitations from an unrestricted Kohn-Sham triplet reference state (S = 1, M_s = + 1).

For HPAS^2-, single-point calculations were performed on select SF-TDDFT optimized geometries with the CASSCF method in a C-PCM model of water. An active space of 16 electrons in 14 orbitals was chosen, denoted as CAS(16,14). The active space includes π and π^* orbitals on the azo group, lone pair orbitals n₁ and n₂ on the azo group, two π orbitals and their π^* counterparts on the phenyl ring, and three π orbitals and their π^* counterparts on the 8-hydroxyquinoline ring. Calculations employed the state-averaging approach encompassing two singlets and one triplet with equal weights for each multiplicity block, denoted SA(2,1)-CAS(16,14). To include dynamic correlation effects, additional single points were performed on converged CAS(16,14) wavefunctions with strongly contracted QD-NEVPT2. The cc-pVTZ basis set was used for all CAS(16,14)/QD-NEVPT2 calculations^63,64. CASSCF and QD-NEVPT2 calculations were sped up with the RI approximation using a cc-pVTZ/C auxiliary basis set⁶⁵.

Author contributions

C.H.: conceptualization, theoretical data curation, formal analysis, methodology, validation, investigation, visualization, writing—original draft, review & editing. C.J.B.: writing—review & editing. W.J.P.: conceptualization, methodology, validation, project administration, funding acquisition, resources, supervision, writing—review & editing. O.M.: conceptualization, methodology, validation, project administration, funding acquisition, resources, supervision, writing—review & editing.

Peer review

Peer review information

Communications Chemistry thanks David Picconi and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request. Cartesian coordinates for all structures discussed in the text are provided in Supplementary Data 1 and Supplementary Data 2. Absolute electronic and Gibbs free energies for all structures discussed in the text are provided in the Supporting Information.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42004-026-01952-5.