Phenylimino Indolinone: A Green‐Light‐Responsive T‐Type Photoswitch Exhibiting Negative Photochromism

Abstract Imines are photoaddressable motifs useful in the development of new generations of molecular switches, but their operation with low‐energy photons and control over isomer stability remain challenging. Based on a computational design, we developed phenylimino indolinone (PIO), a green‐light‐addressable T‐type photoswitch showing negative photochromism. The isomerization behavior of this photoactuator of the iminothioindoxyl (ITI) class was studied using time‐resolved spectroscopies on time scales from femtoseconds to the steady state and by quantum‐chemical analyses. The understanding of the isomerization properties and substituent effects governing these photoswitches opens new avenues for the development of novel T‐type visible‐light‐addressable photoactuators based on C=N bonds.


X-Ray analysis
Z-ITI was crystallized from slow diffusion of a mixture of chloroform and heptane. E-PIO was recrystallized and single crystals grown in Et2O. A single-crystal of either E-PIO or Z-ITI was mounted on a cryoloop and placed in the nitrogen stream (100 K) of a Bruker-AXS D8 Venture diffractometer. Data collection and processing was carried out using the Bruker APEX3 software suite. [3] A multi-scan absorption correction was applied, based on the intensities of symmetry-related reflections measured at different angular settings (SADABS). [4] The structure was solved using SHELXT [5] and refinement was performed using SHELXL. [6] The hydrogen atoms were generated by geometrical considerations, constrained by idealized geometries and allowed to ride on their carrier atoms with an isotropic displacement parameter related to the equivalent displacement parameter of their carrier atoms. No A-or B-level alerts were raised by CheckCIF for the fully refined structure. The cif files and videos of the structures are provided as additional files.

Femtosecond Transient Absorption Spectroscopy
The apparatus used for the transient absorption spectroscopy (TAS) measurements has been described in detail before. [7] Briefly, 80 fs pulses centred at 810 nm were produced by an integrated home-made Ti:sapphire oscillator coupled with a regenerative amplifier system (Amplitude Pulsar). The excitation wavelength was set at 500 or 520 nm and excitation power was set at 30-50 nJ for all measurements. Visible pulses were generated by pumping a home-made non-collinear optical parametric amplifier (NOPA) with a portion of the fundamental 810 nm radiation. The pump beam polarization has been set to magic angle with respect to the probe beam by rotating a \2 plate, to exclude rotational contributions. The white light probe pulse was generated by focusing a small portion of the fundamental laser radiation on a 3 mm thick CaF2 window. A portion of the generated white light was sent to the sample through a different path and used as a reference signal. After passing through the sample the white light probe and reference pulses were both directed to a flat field monochromator coupled to a home-made CCD detector. Transient signals were acquired in a time interval spanning up to 500 ps. The sample was contained in a 2 mm quartz cuvette, mounted on a movable holder in order to minimize photodegradation. Measurements were performed at room temperature. Concentrations were adjusted to an absorbance of 0.9 -1.0 OD (for the respective optical path) at the absorption maximum which amounted to about 0.3 -0.5 OD at excitation wavelength. Before and after the measurements, the integrity of the sample was checked on a PerkinElmer LAMBDA 950 spectrophotometer.

Nanosecond Transient Absorption Spectroscopy
Nanosecond transient absorption spectra were recorded with an in-house assembled setup. A different excitation wavelength (480 -530 nm) was used depending on the absorption properties of the sample. The excitation wavelength was generated using a tunable Nd:YAG-laser system (NT342B, Ekspla) comprising the pump laser (NL300) with harmonics generators (SHG, THG) producing 355 nm to pump an optical parametric oscillator (OPO) with SHG connected in a single device. The laser system was operated at a repetition rate of 10 Hz with a pulse length of 5 ns. The probe light running at 20 Hz was generated by a high-stability short arc xenon flash lamp (FX-1160, Excelitas Technologies) using a modified PS302 controller (EG&G). Using a 50/50 beam splitter, the probe light was split equally into a signal beam and a reference beam and focused (bi-convex lens 75mm) on the entrance slit of a spectrograph (SpectraPro-150, Princeton Instruments) with a grating of 150 ln/mm, blaze at 500 nm. The probe beam (A = 1 mm 2 ) was passed through the sample cell and orthogonally overlapped with the excitation beam on a 1 mm × 1 cm area. The excitation energy was recorded by measuring the excitation power at the back of an empty sample holder. In order to correct for fluctuations in the flash lamp spectral intensity, the reference was used to normalize the signal. Both beams were recorded simultaneously using a gated intensified CCD camera (PI-MAX3, Princeton Instruments) which has an adjustable gate of minimal 2.9 ns, normally a gate of 20 ns and software binning is used to improve the dynamic range and signal to noise ratio. Two delay generators (DG535 and DG645, Stanford Research Systems, Inc.) were used to trigger the excitation and to change the delay of the flash lamp together with the gate of the camera during the experiment. The setup was controlled by an in-house written Labview program.

Computational methods
A series of different methods were employed to optimize the structures of the ground state minima and transition states of the differently substituted compounds of the ITI family presented in Fig. 1 of the main text. All the structures were pre-screened using the CREST driver in the xTB 6.3.2 software, [8] using the GFN2-xTB semiempirical level of theory. [9] In this way, the most stable conformers for each structure were picked via the default series of metadynamics and dynamics runs implemented in the driver. The most stable conformers and the transition states connecting them were optimized at the M06-2X/def2-SVP level. [10,11] The nature of the stationary point found was confirmed computing the hessian matrix and inspecting the number of imaginary frequencies found (0 for the minima, 1 for the transition states). Electronic spectra were simulated on the stable and metastable configurations previously optimized, predicting the 30 lowest singlet transitions at the M06-2X/6-311+G(2d,p). The feasibility of the functional to correctly simulate the state ordering in the ITI photoswitches was previously benchmarked. [12] All DFT optimizations were conducted with the Gaussian 16, Rev B.01 software package. [13] Semiempirical calculations and nonadiabatic molecular dynamics simulations were conducted at the OM2/MRCI level of theory, [14,15] as implemented in the MNDO program. [16] The active space in the MRCI calculations included twelve electrons in eleven orbitals (12,11). All the orbitals were of π character or a mixed n and π character (see Figure S62). For the MRCI treatment, three configuration state functions were chosen as references, namely the leading configuration with two singly occupied orbitals (which defines the ROHF formalism) and the two closed-shell configurations derived therefrom (i.e., the singlet configurations with doubly occupied HOMO or LUMO of the closed-shell ground state). The MRCI wavefunction was built by allowing all single and double excitations from these three references (CISD). All the optimizations were conducted including the lowest three singlet states. The geometry optimization of minima employed the BFGS update. The nature of the minima and transition states was checked inspecting the number of imaginary frequencies after the calculation of the force constants at the optimized geometry (i.e., 0 for minima, 1 for transition states). Transition states were located using the eigenvector following algorithm, while the default optimizer was employed for the minima. The conical intersection was optimized using a modified version [17] of the Lagrange-Newton algorithm proposed by Manaa and Yarkony. [18] Nonadiabatic molecular dynamics (NAMD) simulations on the E geometry were performed using the Tully surface-hopping (TSH) method as implemented in the MNDO program, with an analytical evaluation of the nonadiabatic coupling vectors. [19] All simulations were in a canonical ensemble maintained by a Nosé-Hoover thermostat (T=300 K). The sampling of the initial structures and relative initial velocities was obtained via a preliminary ground-state Born-Oppenheimer dynamics run of 20 ps with 0.5 fs time step. The initial structures for the surface-hopping dynamics were chosen from these ground-state trajectories using the filtering procedure implemented in the MNDO program, combining the active space mapping with a threshold reduced to 70%, and the transition probability to the target state, based on a stochastic algorithm. The three lowest singlet states were included in the NAMD runs. The initial state chosen for starting the dynamics was S1. 320 runs of 600 fs with time step of 0.1 fs were evaluated. The default empirical decoherence correction of 0.1 Hartree was used. [20] The S0-S1 conical intersection was also optimized at the SF-BH&HLYP/cc-pVDZ level over three roots, [21] using the penalty function algorithm as implemented in GAMESS-US (Ver: 30 SEP 2020 (R2)). [22] The excited state landscape was obtained as a constrained optimization of the inversion angle and rotation dihedral of the C=N bonds using the TDA(3 states)-ωB97X-D/MIDI! level of theory [23,24] as implemented in the Gaussian 16, Rev B.01 software. The ground state optimizations using the same constraints were done using the GFN2-xTB level from the xTB 6.3.2 package. All these geometries were then corrected at the SF-BH&HLYP/cc-pVDZ level over five states, using GAMESS-US (Ver: 30 SEP 2020 (R2)). Natural charges were computed using NBO 7.0.9 interfaced with Gaussian 16. [25] All xyz coordinates for all the compounds considered are provided as a separate additional file. A video of a productive E-Z NAMD trajectory computed at the OM2/MRCI level is provided as a separate file.

Results and Discussion
Synthesis and characterization Parent Compounds Z-ITI was synthesized according to a literature procedure. [12] E-PIO was synthesized following the synthetic scheme presented in Scheme S1. Scheme S1. Overview of the synthetic route towards E-PIO.

(E)-1-Acetyl-2-(phenylimino)indolin-3-one (PIO)
A crimp top vial was heated under vacuum, cooled to ambient temperature, filled with nitrogen atmosphere, and equipped with compound 3 (22 mg, 0.13 mmol, 1 eq) and nitrosobenzene (14 mg, 0.13 mmol, 1 eq). Dry benzene (2 mL) was added to dissolve the reagents. Then, two drops of DBU were added to the reaction mixture, which turned immediately red. The reaction mixture was stirred for one minute, completion was confirmed by TLC. Subsequently the mixture was diluted with EtOAc (20 mL) and poured onto water (20 mL). The phases were separated, and the organic phase was washed with water (20 mL), NH4Cl (aq, sat., 2 × 20 mL), NaHCO3 (aq, sat., 2 × 20 mL), brine (20 mL), and dried over Na2SO4. The volatiles were removed in vacuo and the pure product was obtained as red crystalline solid (34 mg, 0.13 mmol, quant.). Note: Extensive washing of the organic phase was crucial as PIO is sensitive to high concentrations of acid or base. Incomplete removal of the base lead to decomposition during concentration, while heating the water bath of the rotary evaporator to 50 °C, normal phase (pentane, EtOAc), or reversed phase column (water, MeCN with 0.1% formic acid) is tolerated. Sensitivity to acid is particularly pronounced in the presence of a nucleophile, such as methanol, and lead to hydrolysis of the imine. 1

Nitrosobenzenes
All nitroso benzenes were either freshly prepared or stored in the fridge until usage. Nitrosobenzenes 4-7 were synthesized following a procedure our groups published earlier. [12] Scheme S2. Overview of the synthesis of nitrosobenzenes used in the study. (8) 4-Bromoaniline (1.73 g, 10.1 mmol) and OXONE (6.14 g, 20.0 mmol) were dissolved in DCM (30 mL) and water (40 mL), and vigorously stirred for 3 h at ambient temperature. The process of the reaction was monitored by TLC (SiO2; EtOAc:heptane, 2:6, v/v). Then, the phases were separated, and the aqueous layer was extracted with DCM (2 x 40 mL). The combined organic extracts were dried over Na2SO4 and the volatiles were evaporated in vacuo. The title compound was obtained as yellow powder (862 mg, 4.63 mmol, 46%) and used without further purification. (9) 4-Aminobenzonitrile (1.13 g, 9.56 mmol) and OXONE (6.39 g, 20.8 mmol) were dissolved in DCM (30 mL) and water (40 mL), and vigorously stirred for 3 h at ambient temperature. The process of the reaction was monitored by TLC (SiO2; EtOAc:heptane, 2:6, v/v). Then, the phases were separated, and the aqueous layer was extracted with DCM (2 x 40 mL). The combined organic extracts were dried over Na2SO4 and the volatiles were evaporated in vacuo. The title compound was obtained as yellow powder (1.10 g, 8.33 mmol, 87%) and used without further purification.

Femtosecond Transient Absorption Spectroscopy
We have recorded transient absorption spectra of E-PIO dissolved in toluene upon excitation at 500 nm. The data have been processed using global analysis, applying a linear decay kinetic scheme with three time constants. As can be noticed by observing the EADS in Figure S28,  Analogous to what we observed in case of ITI, [12] we interpret the initial evolution occurring in 1.4 ps as being associated with reaching the conical intersection between the ground and electronically excited state of the molecule, from which it decays to the ground state of both the Z and E isomers. In the following spectral component, the positive band peaked at about 550 nm is associated with absorption from a hot ground state of the E isomer, while the one peaked at 425 nm with the absorption of the hot ground state of the Z isomer. The final spectral component reflects the differential absorption between the bleaching of the E isomer and the ground state absorption of the Z isomer product. Kinetic traces recorded near the maxima of the absorption bands of the two isomers reported in Figure S29 confirm the fast decay of the broad excited state band and the persistence of small, long-living positive and negative contributions. Figure S29. Kinetic traces recorded for PIO in toluene upon excitation at 510 nm.

S31
Transient absorption measurements have been repeated in solvents with different polarities: cyclohexane, dichloromethane, methanol, DMSO, acetonitrile and diethyl ether. The overall excited state behavior of PIO in all solvents is similar to that observed in toluene although in polar solvents the kinetics of the excited state decay are accelerated compared to those in non-polar ones. The evolution of the transient spectra is instead qualitatively similar in all solvents. As a comparison to what is observed in toluene, Figure S30 reports the EADS obtained from a global analysis of PIO dissolved in methanol. In Figure S31 a comparison of the kinetic traces obtained on the maximum of the excited state absorption band in toluene and methanol is reported. The kinetic constants extracted from the global analyses in the various solvents are reported in Table S14.

Nanosecond Transient Absorption Spectroscopy
An estimate of the quantum yield for photoisomerization of the PIO derivatives following excitation with 2.6 mJ laser pulses was obtained using the Lambert-Beer law by calculating the number of excited molecules and the observed bleach upon irradiation corrected for the probability of absorption (see Table S15). Herein it was assumed that the Z-isomer does not absorb in the 510 nm region. The very small absorption at 510 nm observed in the spectra of the Z isomer would lead to an increase of the reported quantum yields.

Low temperature NMR
A sample of E-PIO was dissolved in Et2O-d10 (c = 1 × 10 -3 M, 600 μL) and transferred into an NMR tube. Irradiation was conducted in situ via a fiber optic cable inserted into the NMR tube with a 505 nm LED at -105 °C. In this specific experimental conditions the NMR probe was cooled with liquid nitrogen: the error in the temperature detection is ±10 °C. Moreover, to acquire more points and achieve a better signal-to-noise ratio, the t1 relaxation was cut before its completion (with an additional ±10% error on the integration of the kinetic data). For the above, the kinetic results obtained are only qualitative. These considerations do not hold for the analysis of the distribution of the species at the photostationary state. Upon irradiation, a new set of aromatic peaks appears and more interestingly, the acetyl CH3 group shifts ~ 1.2 ppm upfield. This observation is in line with our theoretical calculations of the geometry of Z-PIO, the most stable conformer of which presents an interaction between the phenyl ring and the CH3 group itself, with an orientation that leads to increased shielding in the NMR signal (for a visualization of the structure, see Figure 2).       Figure S57. Kinetic of the thermal recovery of the doublet signals in the aromatic region of the NMR spectrum (blue from the signal at 8.67 ppm for E-PIO, orange from the signal at 8.25 ppm for Z-PIO), starting from the photostationary state reached after irradiation of E-PIO in Et2O-d10 (1 × 10 -3 M) with a 505 nm LED at -105 °C. The lifetimes that can be inferred from this measurement differ from the one measured using low-temperature-UV-Vis (2300 s vs 290 s, vide infra). We attribute the cause of this discrepancy to the less accurate reading of the temperature for the NMR measurements, that suffers from a larger error than the UV-Vis itself (±10 °C at -105 °C vs ±1 °C of the UV-Vis). Indeed, a five degree difference recorded around -105 °C can cause the lifetime to increase threefold (see Figure  S60). After these considerations and the comparison with the UV-Vis and ns TA data we can confidently state that, within the margin of temperature uncertainty, the same process, viz. the formation of Z-PIO and its thermal relaxation to the stable E form, was observed.
A sample of E-PIO in anhydrous Et2O (c = 3 × 10 -4 M) was prepared and degassed by bubbling N2 through the solution. The sample was irradiated at six temperatures (-110, -105, -100, -95, -90 and -85 °C; the error in the temperature measurement is ±1 °C), using a 505 nm LED, and the thermal relaxation was followed on a UV-Vis spectrophotometer. From the change in absorbance at 443 nm, the rate constants (k) at the different temperatures were determined by fitting a 1 st order rate law. The values were plotted to determine the thermodynamic parameters of isomerization.

Z-to-E isomerisation by irradiation with 455 and 420 nm light
To demonstrate the ability to obtain E-PIO by irradiation of Z-PIO at a different wavelength instead of thermal relaxation, a sample of the switch was first irradiated with a 505 nm LED to a ΔOD (443 nm) of ~ 0.35. Subsequently, the isomerization from E-PIO to Z-PIO was followed in the case of 1) no irradiation, 2) irradiation with a 455 nm LED, and 3) irradiation with a 420 nm LED. Upon irradiation with a 455 or 420 nm LED the original spectrum was not obtained, but a new PSS was reached following faster kinetics than observed under thermal relaxation conditions.  Computational analysis Figure S62. Active space orbitals employed in the OM2/MRCI calculations. Figure S63. Averaged S0 and S1 states population as a function of time.    Figure S66. Heatmap of the energy (in Hartrees) of S1 obtained at the SF-BH&HLYP/cc-PVDZ//TDA-ωB97X-D/MIDI! level of theory (energy diminishes from orange to blue). The scan confirms the presence of an energetic well at ca. 90° of the rotation dihedral (CInt), where the NAMD simulations predict the S1→S0 hoppings to occur. The excited state around the minimum has a diradical character (e.g S 2 =1.09 for angle=120° dihedral=90°), while the ground state is a closed-shell (e.g S 2 =0.27 for angle=120° dihedral=90°). For this reason, a monodeterminantal treatment such as DFT can be used to approximate the excited-state geometries even near the crossing point region. Bottom right: the geometry of the minimal energy point at 90° dihedral and 120° inversion angle. The scan confirms that the minimum energy path that connects E and Z-PIO passes through a region where the inversion dihedral is around 180°, disfavoring the rotation mechanism in the ground state. All the geometries considered are closed-shell (i.e. S 2 < 0.5). The carbon of the CN is pyramidalized in the ground state at geometries with dihedral angles around 90° (see figure in the bottom right corner). Conversely this pyramidalization is lost in the excited state (see Figure S66), confirming the transition from a closed-shell (pyramidalized) to an open-shell (diradical, non-pyramidalized) state.
CInt Z E Figure S68. Comparison of the geometries obtained from the X-Ray of E-PIO (grey) and the corresponding optimized geometries at the M06-2X/def2-SVP level (white). The RMSD is 0.198 (H not considered). The main differences correspond to angle bends, possibly due to the packing in the solid state. Figure S69. Comparison of the geometries obtained from the X-Ray of Z-ITI (grey) and the corresponding optimized geometries at the M06-2X/def2-SVP level (white). The RMSD is 0.053 (H not considered). Figure S70. Comparison of the geometries of the CInt of E-PIO obtained at the SF-BH&HLYP/cc-pVDZ (grey) and the corresponding one optimized at the OM2/MRCI level (white). The RMSD is 0.135. The main difference corresponds to the pyramidalization degree of the C=N bond, slightly more pronounced at the OM2/MRCI level. [b] Energy differences calculated between the E-form and the Z-form. Negative values → E more stable [c] Less stable conformer. Energy differences are always calculated using the most stable conformer.
[d] Natural charge at the imine nitrogen obtained from Natural Population Analysis.
[e] Difference in Natural charges at the imine nitrogen (TS -Z-form).