Direct evidence for hula twist and single-bond rotation photoproducts

Photoisomerization reactions are quintessential processes driving molecular machines and motors, govern smart materials, catalytic processes, and photopharmacology, and lie at the heart of vision, phototaxis, or vitamin production. Despite this plethora of applications fundamental photoisomerization mechanisms are not well understood at present. The famous hula-twist motion—a coupled single and double-bond rotation—was proposed to explain proficient photoswitching in restricted environments but fast thermal follow-up reactions hamper identification of primary photo products. Herein we describe an asymmetric chromophore possessing four geometrically distinct diastereomeric states that do not interconvert thermally and can be crystallized separately. Employing this molecular setup direct and unequivocal evidence for the hula-twist photoreaction and for photoinduced single-bond rotation is obtained. The influences of the surrounding medium and temperature are quantified and used to favor unusual photoreactions. Based on our findings molecular engineers will be able to implement photo control of complex molecular motions more consciously.

wavelength (λ) are reported in nm and the molar absorption coefficients (ε) in L mol -1 cm -1 . Shoulders are declared as sh.
Low temperature UV/vis spectra in EPA glass (ether/isopentane/ethanol 5:5:2) at 90 K were measured on a Varian Cary® 50 spectrophotometer with an Oxford DN 1704 optical cryostat controlled by an Oxford ITC 4 device. Low temperatures were reached by cooling slowly with liquid nitrogen. The spectra were recorded in a quartz cuvette (1 cm). Solvents for spectroscopy were obtained from VWR, Merck and Sigma Aldrich and were dried, degassed and filtrated prior use. For irradiation studies a Mightex FCS-  LED (405 nm) was used as light source. Absorption wavelength (λ) are reported in nm and the molar absorption coefficients (ε) in L·mol -1 ·cm -1 . Irradiations were conducted using LEDs from Roithner Lasertechnik GmbH (305 nm, 365 nm, 405 nm).

Physical and photophysical properties
Thermal atropisomerizations A to B and C to D NMR tubes were charged with 0.8 mg to 2.5 mg of respective isomer A, B, C, or D and 0.7 mL of deuterated solvent. Subsequent heating was carried out in amberized NMR tubes at 82 °C or 100 °C. Kinetics were followed by 1 H NMR measurements at defined time intervals. The equilibrium concentrations of isomers after prolonged heating were obtained from integration of the corresponding signals in the 1 H NMR spectrum.
At 82 °C or 100 °C only the atropisomerizations between A and B and between C and D proceed. The thermal atropisomerizations follow unimolecular first order reactions and proceed towards an equilibrium composition with both atropisomers present, according to Supplementary  From these measurements we could quantify the ground state energy profile of hemithioindigo 1. As we did not observe any thermal double-bond isomerizations over the course of several hours at temperatures >100 °C a lower limit for the energy barriers could be given, which are at least 30 kcal/mol high for each double-bond isomerization. The equilibrium atropisomer compositions at high temperatures deliver the thermodynamic energy differences between the corresponding states (red ∆ values in Supplementary Figure   9) according to the relation of the change of Gibbs free energy and the equilibrium constant -∆G = R⋅T⋅lnK (see also Supplementary Table 1). The theoretical values are in good agreement with the experimentally determined ones.

Photoconversion of A, B, C, and D determined by quantum yield measurements
The photochemical quantum yield of the different photoconversion reactions φ were calculated as the ratio between the numbers of isomerized molecules n(molecules isomerized) and the number of absorbed photons Therefore, the quantum yield measurements conform to initial-slope behavior, where only the initial isomer is photoconverted but not the photoproducts. The photoconversion of D did not conform very well to linear behavior (Supplementary Figure 17). Nevertheless, the initial slope assumption gave good starting points for the comprehensive analysis of all quantum yields φ using a rate matrix as described below.
For hemithioindigo 1 the rate of an individual phototransition e.g. A to B, depends on all other absorbing species present at the same time and is described by the corresponding rate matrix element rA/B (Supplementary Equation 15): Likewise every phototransition from isomer i to isomer j can be written as: We have used the different rate elements ri/j in the rate matrix M1 to simulate our quantum yield measurements. For every incremental irradiation step (∆t) we have used the following expressions  to achieve the corresponding next concentration of the respective isomer: Since every component of ri/j (Supplementary Equation 16) is known from our experiment except for the quantum yields φij the latter can be obtained from the best fit to the rate model M1 (Supplementary Equation   17). To this end we have started the simulation with experimental quantum yield values obtained from the initial slope analyses described above. 10,000 incremental time points were used in the simulations and the quantum yields were adjusted manually until an adequate match of experimental and simulated data were obtained. While the quantum yields of A and B obtained from the simulation were very similar to the ones obtained from the initial slope method (maximum ±2% deviation), the quantum yields of C and D were considerably improved by the simulation. This can clearly be seen by the much better description of the nonlinear photoconversions of C and D using the optimized quantum yields from the fitting simulation (Supplementary Figure 18).

Markov matrix analysis of the photoconversion of A, B, C, and D in different solvents
The The relative percentage of each isomer A, B, C, and D was then plotted against irradiation time for each pure starting isomer. These kinetic data were then simulated using an adjusted Markov matrix accounting for the different conversion probabilities per time increment (1 min in this case).
In general a Markov matrix describes the probabilities for different transitions of different states p(ij) within a given time increment. For a photoreaction the probabilities are directly proportional to the rate ri/j of the transition if all molar absorption coefficients are the same and therefore the kinetics are first order: Additionally we have to include the diagonal elements p(ii) describing the probability that no conversion occurs: Therefore, a Markov matrix can be written for hemithioindigo 1 in which the transition probabilities p(ij) are given:   After irradiation, the EPA glass was warmed to 195 K and cooled down again to 90 K (light blue). As shown in the enlarged UV/vis absorption spectrum inset, no significant photoisomerization could be observed.

Supplementary Tables
Supplementary Table 1