Sunlight-powered kHz rotation of a hemithioindigo-based molecular motor

Photodriven molecular motors are able to convert light energy into directional motion and hold great promise as miniaturized powering units for future nanomachines. In the current state of the art, considerable efforts have still to be made to increase the efficiency of energy transduction and devise systems that allow operation in ambient and non-damaging conditions with high rates of directional motions. The need for ultraviolet light to induce the motion of virtually all available light-driven motors especially hampers the broad applicability of these systems. We describe here a hemithioindigo-based molecular motor, which is powered exclusively by nondestructive visible light (up to 500 nm) and rotates completely directionally with kHz frequency at 20 °C. This is the fastest directional motion of a synthetic system driven by visible light to date permitting materials and biocompatible irradiation conditions to establish similarly high speeds as natural molecular motors.

S ynthetic photo-driven molecular motors 1-5 allow harnessing light energy and converting it into directional motion against the equilibrating force of the 'Brownian storm' 6,7 . They can be used to effectively drive a molecular system away from equilibrium and thus represent the essential component to power future nanomachinery [8][9][10][11] . Since Koumura et al. 1 and Kelly et al. 12 reported the first synthetic molecular motors in 1999, intriguing applications have been put forward [13][14][15] demonstrating unique functions that cannot be established at the molecular scale in any other way 16 . However, to compete with the high efficiency and versatility of natural molecular motors [17][18][19] considerable efforts have still to be made for synthetic systems [20][21][22] . Most light-powered molecular motors require damaging ultraviolet light to perform their task, which is a major drawback for biological or smart materials applications. At present, only a few motor systems are available that undergo unidirectional 360°rotation using visible light, but the speed of their motion is slow, which again impedes applications in (heat) sensitive environments 23,24 . In the following, we report on a novel molecular motor 1, that performs a full (360°) rotation powered by visible light (up to 500 nm, that is, at the maximum wavelength of sunlight) with 495% unidirectionality and at a very fast rate (1 kHz at 20°C). Thus, our new molecular motor system enables for the first time to power fast unidirectional rotation under ambient and non-damaging conditions, representing the next crucial step towards developing highly efficient nanomachines that work at materials and biocompatible conditions.
Our new molecular motor is based on the hemithioindigo (HTI) chromophore 25 , an emerging photoswitch [26][27][28][29][30][31][32][33] that can be operated exclusively by visible light in both switching directions. HTI consists of a thioindigo and a stilbene fragment, which are connected via a central double bond. On irradiation, HTI undergoes efficient and reversible Z/E or E/Z photoisomerization [34][35][36][37] . To confine light-induced rotations around the central double bond to one direction and exclude unwanted back-movements, we implemented additional stereochemical elements to the HTI framework: the sulfur atom was oxidized to the corresponding sulfoxide, introducing a sulfurbased stereocentre (R-or S-configuration), and sterical crowding at the ring-fused stilbene fragment led to helical twisting (P-or M-helicity) around the central double bond. Combined with its Z and E isomeric forms, motor 1 can thus assume four different diastereomeric structures for each configuration of the stereogenic sulfur centre, for example, Z-(S)-(P), Z-(S)-(M), E-(S)-(M) and E-(S)-(P), which can be distinguished by conventional spectroscopic methods. This molecular set-up is related to the Feringa motor system consisting of a sterically overcrowded double bond in combination with carbon-based stereogenic centres 1,38 . In our case, considerably smaller sterical hindrance between the thioindigo and stilbene fragment is needed to warrant unidirectional rotation.

Results
Synthesis. Motor 1 was synthesized in three steps starting from known, easily accessible precursors (Fig. 1). Phenylthioacetic acid 2 (ref. 39) was converted into the corresponding acid chloride, which subsequently underwent cyclization via intramolecular Friedel-Crafts acylation to give benzothiophenone 3. Because of its instability and high tendency to dimerize, benzothiophenone 3 was used in the next synthetic step in crude form. Commercially available 4,7-dimethoxy-1-indanone (4) was methylated twice in the 2-position using sodium hydride and methyl iodide to give indanone 5. Benzothiophenone 3 was subsequently condensed with indanone 5 using boron trifluoride diethyl etherate, after which a final oxidation with hydrogen peroxide furnished sulfoxide motor 1. Crystals suitable for X-ray analysis were obtained for two isomeric forms of racemic motor 1, the Z-(S)-(P)/Z-(R)- Conformational analysis. The (S) and (R) enantiomeric forms of motor 1 are stable (also under irradiation conditions as shown in Supplementary Fig. 14) and could easily be separated using chiral high-performance liquid chromatography (HPLC). For the following spectroscopic studies however, the racemic mixture was analysed for convenience reasons. The Z-(S)-(P)/Z-(R)-(M) isomers are the thermodynamically stable form of motor 1, which can be obtained in 75% yield by heating a toluene-d 8 solution of 1 to 100°C for 8 h, as analysed by nuclear magnetic resonance (NMR) spectroscopy. Under these conditions also 25% of the E-(S)-(P)/E-(R)-(M) isomers are obtained. When a solution of 1 in xylene-d 10 is heated to 130°C for 12 h, 73% of the Z-(S)-(P)/Z-(R)-(M) isomers and 27% of the E-(S)-(P)/E-(R)-(M) isomers are obtained ( Supplementary Fig. 22). This thermal behaviour translates into an energy difference of 0.81 kcal mol À 1 between the Z and E diastereomeric pairs according to the relation of the change of Gibbs free energy and the equilibrium constant-DG ¼ RTln K. First-order kinetic analysis of thermal conversion from the E-(S)-(P)/E-(R)-(M) to the Z-(S)-(P)/Z-(R)-(M) isomers provided an energy barrier of 29.54 kcal mol À 1 for the process ( Supplementary Fig. 21). With such high thermal stability, that is,  isomers at 25°C, the two Z and E diastereomers could conveniently be separated by conventional chromatography and studied individually. Direct assignment of the two separated species to the corresponding diastereomers was straightforward in solution, as NOESY NMR spectra showed specific cross-signals between the thioindigo and the stilbene part of the molecule ( Supplementary Figs 6 and 11). In addition, crystal structure analysis unambiguously confirmed our assignment of the species and their solution spectra. We were able to grow crystals suitable for structure analysis from the second fraction obtained after chromatographic separation, which could be assigned to the Z-(S)-(P)/Z-(R)-(M) isomers (CCDC 1061969). The 1 H NMR and ultraviolet/visible (ultraviolet/vis) spectra of the same crystal batch were recorded, showing only one single species to be present in solution. Thus, we could directly assign the corresponding NMR signals as well as the ultraviolet/vis spectrum to the Z-(S)-(P)/Z-(R)-(M) isomers (see also Supplementary Figs 1-6 and Supplementary Fig. 12, respectively).
In a similar way, the first fraction obtained after chromatographic separation and corresponding NMR and ultraviolet/vis solution spectra could be assigned to the E-(S)-(P)/E-(R)-(M) isomeric form (see Supplementary Figs 7-11 for details). The absorption spectral behaviour of motor 1 (Fig. 2b)-hypsochromic absorption of the Z (up to 470 nm) and bathochromic absorption of the E isomer (up to 505 nm)-is similar to typical HTI photoswitches 40 .
The crystal structures of the Z-(S)-(P) and Z-(R)-(M) isomers (CCDC 1061969) represent the global thermodynamic minimum of motor 1; the structure of the Z-(S)-(P) isomer is shown in Fig. 2a. In this structure, substituents at the double bond are twisted out of planarity, leading to a helical arrangement, in which the methoxy-substituted aromatic ring of the fused stilbene fragment is oriented behind the sulfoxide oxygen atom. On the other side of the double bond, the two methyl substituents of the five-membered aliphatic ring are bisected by the carbonyl oxygen atom of the thioindigo fragment. The helical twist is clearly a result of steric repulsion between the substituents on the stilbene and the thioindigo fragment, which leads to a SC ¼ CC(Ar) dihedral angle of 12.8°. This value is considerably larger than that observed in sterically unrestrained HTIs. The bond length of the central double bond is 1.358 Å and only slightly longer compared to planar HTIs 40 . The mirror image of that structure is reproduced in the crystal structure of the enantiomerically pure Z-(R)-(M) isomer (Fig. 2c, CCDC 1406625).
In solution, less structural information is available but several details could be inferred from NMR analysis. The NOESY NMR spectrum ( Supplementary Fig. 6) allowed not only to confirm the Z configuration of the double bond in solution, but also to obtain relative distances between protons of the thioindigo fragment and the methyl and methoxy groups of the stilbene fragment. On the basis of relative signal intensities, the shortest H-H distance between the thioindigo and the stilbene fragment is found for one methyl group and the aromatic proton in ortho-position to the carbonyl group. On the other side of the double bond the methoxy group located behind the sulfoxide is slightly farther away, but an NOE coupling to the proton in ortho-position of the sulfur atom is still visible. The second methyl group shows no NOE couplings to protons of the thioindigo fragment. This relative order of distances is confirmed exactly by crystal structure analysis ,where the methyl group pointing towards the sulfur oxygen atom is closest to the thioindigo fragment, followed by the aforementioned methoxy group. The second methyl group is located at the third-closest distance from the thioindigo fragment. Both methoxy groups show additional strong NOE couplings to the neighbouring aromatic protons of the stilbene fragment, but no NOE couplings to the aliphatic CH 2 signals are seen. Again this corresponds very well with the conformation seen in the crystal structure, where both methoxy groups point towards the aromatic protons of the stilbene fragment. A Karplus cross-peak analysis of the aliphatic part of the HMBC NMR spectrum (Supplementary Fig. 4) established that the conformation of the five-membered ring of the stilbene fragment in solution is also very similar to that observed in the crystalline state. Two HMBC signals are not observed, which is due to 90.0°dihedral angles between the corresponding proton and carbon atoms. Dihedral angles of 93.6°and 96.8°are indeed found in the crystal structure between the very same atoms. From this qualitative NMR analysis, we conclude that the conformation of Z-(S)-(P)/Z-(R)-(M)-1 in solution is very similar to the one found in the crystal.
Among the metastable forms of motor 1, the E-(S)-(P)/E-(R)-(M) isomers represent the most stable configuration. The crystal structure of the E-(S)-(P) isomer (CCDC 1061970) is shown in Fig. 2b. The geometry is again helical, with a dihedral angle C(carbonyl)C ¼ CC(Ar) of 16.2°. The two methyl groups of the fused five-membered ring of the stilbene fragment are now bisected by the oxygen atom of the sulfoxide group. At the same time, the aryl ring of the stilbene fragment is oriented to the same side as the sulfoxide oxygen atom.
An NMR analysis similar to the one conducted for the Z-   Fig. 23). However, irradiating the solution at À 90°C allowed us to follow the isomerization process of the double bond in greater detail as shown in Fig. 3.  Fig. 32). This gives an energy difference between the two E isomers of at least 1.73 kcal mol À 1 according to the relation of the change of Gibbs free energy and the equilibrium constant -DG ¼ RTln K.
Complementary experiments were performed with a CD 2 Cl 2 solution of pure E-(S)-(P)/E-(R)-(M)-1 at -90°C ( Supplementary  Fig. 33), using again 460 nm light for the irradiation (Fig. 3b, see also Supplementary Fig. 34 for the full spectra). In this case no new signals were observed, but the signals of Z-(S)-(P)/Z-(R)-(M)-1 were found to build up immediately after irradiation. At higher power of irradiation also the E-(S)-(M)/E-(R)-(P) isomers were observed, but only as a photoproduct of the Z-(S)-(P)/Z-(R)-(M) isomers after the latter were already present ( Supplementary  Fig. 35b). An irradiation experiment in diethyl ether-d 10 at -100°C also did not furnish any new signals of a fourth isomeric form ( Supplementary Fig. 36) Fig. 25).  Theoretical description. To gain deeper insight into the function of our new motor system, we performed a theoretical analysis of the ground-state energy profile at the DFT MPW1K level of theory using the 6-31 þ G(d,p) basis set ( Supplementary Fig. 39). The structures of Z-(S)-(P) and E-(S)-(P) isomers were optimized starting from the crystal analytical data of the Z-(S)-(P) isomer. The optimized geometries and the geometries obtained experimentally from crystal structure analysis agree very well (see Supplementary Table 4 Supplementary Figs 40 and 41). All optimized structures were confirmed to be stationary points by frequency analysis. No imaginary modes were found for the minima structures and only one imaginary mode was present for the transition states, proving the latter to be first order saddle points on the hyper-potential energy surface. The calculated ground-state energy profile is shown in Fig. 4, together with the experimentally obtained data. The theoretically found energies are qualitatively in complete agreement with our experimental findings. Moreover, all experimentally quantified energies match those obtained theoretically exceptionally well 41 . The Z-(S)-(P) isomer is found indeed as a global minimum structure in the theoretical description, while the E-(S)-(P) isomer is only 0.8 kcal mol À 1 higher in energy. This finding is perfectly confirmed experimentally by the 75:25 and 73:27 ratio of Z-(S)-(P) and E-(S)-(P) isomers observed in solution after heating to 100°C and 130°C, respectively (corresponding to an energy difference of 0.81 kcal mol À 1 ). The E-(S)-(M) isomer is 3.03 kcal mol À 1 higher in energy compared with the E-(S)-(P) isomer in the theoretical description, which explains why the E-(S)-(M) isomer is completely converted (that is, 495%, given the accuracy of NMR spectroscopy) to the E-(S)-(P) isomer in the thermal step. For this conversion, the theoretical description gives a barrier of 14.62 kcal mol À 1 , a value that again agrees remarkably well with the experimentally measured 13.10 kcal mol À 1 . Interestingly, the calculated energy barrier of the corresponding thermal conversion of Z-(S)-(M) to Z-(S)-(P) is considerably smaller with only 5.54 kcal mol À 1 . Thus, a halflife of 0.75 ms would be found for this process at -90°C-a time range that is not accessible with conventional low-temperature NMR or ultraviolet/vis spectroscopic techniques. The theoretically found energy difference between the Z-(S)-(M) and Z-(S)-(P) isomers is 3.26 kcal mol À 1 , which is sufficiently high to convert the Z-  Separation of enantiomers. To harness unidirectional rotation of motor 1 in applications, the enantiomers have to be separated, which was conveniently achieved using chiral HPLC. The separation of all four diastereomeric forms, Z-(S)-(P), Z-(R)-(M), E-(S)-(P) and E-(R)-(M), was possible in one single run using a CHIRALPAK IC column from Diacel. Circular dichroism spectra of these four isomers were measured ( Supplementary Figs 15-19) and could be assigned to the corresponding absolute configuration of the sulfoxide stereocenter, by comparison with the theoretically obtained circular dichroism spectra calculated at the DFT MPW1K level of theory using the 6-311 þ þ G(d,p) basis set (Supplementary Fig. 20). These assignments were experimentally proven by crystal structure analysis of enantiomerically pure Z-(R)-(M) isomer (CCDC 1406625, Fig. 2c) obtained after semipreparative chiral HPLC separation ( Supplementary Figs 15 and 37).

Discussion
We have developed a novel HTI-based molecular motor 1 that is fuelled by visible light at wavelengths up to 500 nm. To unambiguously prove unidirectionality of the full 360°doublebond rotation, we have established the exact order by which different diastereomeric forms of 1 are interconverted under irradiation conditions. The maximum speed of rotation was obtained from kinetic analysis of the slowest interconversion step, which is the thermal helix inversion of the E isomer. According to our analysis, motor 1 is capable of kHz rotation at ambient temperatures (that is, 20°C) with 495% unidirectionality, which sets a new standard to the performance and sustainability of synthetic molecular motors. With these exceptional properties, motor 1 can be used as a highly efficient molecular power unit for applications in sensitive environments that do not tolerate ultraviolet light or high temperatures. We believe that especially in the fields of material sciences and biology, our new motor system will be of great advantage and we will direct our future efforts towards the implementation of motor 1 into more complex nanomachinery and biological systems.

Methods
General. Reagents and solvents were obtained from Acros, Aldrich, Fluka, Merck or Sigma-Aldrich in the qualities puriss., p.a. or purum and used as received unless stated otherwise. Technical solvents were distilled before use for column chromatography and extraction on a rotary evaporator (Hei-VAP Value). Thin-layer chromatography (TLC) was conducted on Merck Silica 60 F254 TLC plates and visualization conducted with a ultraviolet lamp (254 or 366 nm). Deuterated solvents were obtained from Cambridge Isotope Laboratories and were used without further purification. Thiophenol and 4,7-dimethoxy-1-indanone were purchased at reagent grade from Sigma-Aldrich and were used as received. Boron trifluoride diethyl etherate was purchased from ABCR and freshly distilled before use. Column chromatography was performed with SiO 2 60 (Merck, particle size 0.063-0.200 mm) and distilled technical solvents. HPLC was performed on a Shimadzu HPLC system consisting of an LC-20AP solvent delivery module, a CTO-20A column oven, a SPD-M20A photodiode array ultraviolet/vis detector and a CBM-20A system controller using a semipreparative CHIRALPAK IC column (particle size 5 mm) from Diacel and HPLC grade solvents (2-PrOH and n-heptane) from Sigma-Aldrich and ROTH. 1   Building block syntheses. 2-Phenylthioacetic acid (2) was prepared according to a literature procedure 39 from a commercially available thiophenol. 4,7-Dimethoxy-2,2-dimethyl-1-indanone (5) was prepared from a commercially available 4,7-dimethoxy-1-indanone based on an altered literature procedure 42 .