## Introduction

In spintronics, the use of organic materials as a spin transport material has recently garnered significant attention as they exhibit long spin-relaxation times and long spin-diffusion lengths owing to the weak spin–orbit interaction (SOI) of light elements1,2,3,4,5,6,7. Meanwhile, the weak SOI of organic materials become a drawback when they are used as a spin filter (or spin polarizer), In addition, the exchange interaction in an organic material is extremely weak to induce a ferromagnetic transition above room temperature. A spin-polarized current is, therefore, typically generated by inorganic materials with ferromagnetism or strong SOIs8,9,10,11,12,13. However, the recent finding of spin-selective electron transport through chiral molecules by Naaman et al., i.e., the so-called chirality-induced spin selectivity (CISS) effect, suggests an alternative method of using organic materials as spin filters for spintronics applications14,15,16,17. Although its microscopic mechanism is still under debate, the CISS effect has been validated by numerous experiments such as photoelectron spectroscopy14,18, magnetoresistance (MR) measurements, photoinduced electron transfer measurements19,20,21, spin-polarized conductive atomic force microscopy (AFM)22,23,24, single-molecule scanning tunnel microscopy25, and X-ray circular dichromism measurements26. Moreover, the CISS effect was already demonstrated in several applications such as magnetic memory devices27,28,29, spin-injection devices30, water splitting31,32, and enantiomer separation33,34. However, in all of the cases above, chiral molecules used in the experiments are static molecules such as DNAs14,19,22,33, helicenes18,23, amino acids35,36, and peptides20,21,25,26,28,29,37,38. Hence, the manipulation of spin-polarization (SP) direction by external stimuli has not been realized yet.

In this context, overcrowded alkenes (OCAs) that demonstrate molecular motor functionalities are promising for integration in CISS-based SP solid-state devices because they demonstrate chirality switching by external stimuli. The first unidirectional molecular motor was reported in 1999 by Feringa et al.39. A cis-trans photoisomerization around the central C–C double bond is a basic driving force of the molecular rotation. Owing to the asymmetric stereo chemistry around the double bond, the system undergoes a unidirectional 360° rotation (Fig. 1a). The rotation cycle includes four distinct steps: two photochemical and two thermal steps, resulting in 4 times P/M chirality inversion during the 360° molecular rotation. Because the SP in the CISS effect depends on the chirality of the molecules, namely P (right handed) or M (left handed), the SP direction of electrons that pass through the molecular motors should be switched by light irradiation or thermal treatments40,41,42.

## Rusults

### MR measurements with gold top electrodes

To confirm spin-polarized current generation by an OCA molecule, a solid-state device that sandwiches a thin layer of an OCA molecule with Al2O3 (3 nm)/Ni (50 nm) and a gold (20 nm) electrode (Fig. 2a) was fabricated. The OCA film was prepared by the spin-coating method and a random molecular arrangement is expected (AFM images for the OCA films are shown in Fig. S7). However, we emphasize that finite spin-polarization owing to CISS effects can be expected even for a completely random orientation because the molecular chirality cannot be canceled out. External magnetic fields of upto 1 T were applied perpendicular to the substrates during the measurement, where the magnetization direction of the Ni electrode can be switched between the up and down directions. MR measurements were performed at room temperature (300 K). As the motor molecule, we chose a donor–acceptor substituted-type OCA43 to ensure visible light operation. The following measurements were performed with a 2’R-isomer of the OCA, unless otherwise noted. The OCA was placed in the device as its (2’R-M)-trans (hereinafter, M-trans) isomer or (2’R-P)-cis (hereinafter, P-cis) isomer. The results of spin-valve-like measurements displayed in Fig. 2b show a clear up-spin selectivity (the spin angular momentum is oriented parallel to the direction of electron propagation) for the M-trans device and down-spin selectivity (the spin angular momentum is oriented antiparallel to the direction of electron propagation) for the P-cis device, whose directions are consistent with those reported in previous CISS experiments14,23,27,38. Meanwhile, in the racemic compound, no MR signal was found (Fig. 2c), elucidating that the MR signals had originated from the SP induced by the molecular chirality of the OCA layers. These results clearly indicate the ability of OCA chirality to select the spin orientation. However, the SP could not be inverted by light irradiation, as shown in the same figure. This is attributable to the gold electrode that is macroscopically flexible in its thin (20 nm) form but is extremely rigid at the molecular length scale to allow for molecular motor rotation. Therefore, we decided to employ another type of electrode that exhibits flexibility at the nanoscale.

### MR measurements with PEDOT/PSS top electrodes

Hence, we fabricated a tunnel junction device by replacing gold with a poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT/PSS) (600 nm) electrode (Fig. 3a, b). First, we investigated the CISS effects with MR measurements to reproduce the result of devices with gold electrodes. Subsequently, we investigated the effects of photoirradiation on the CISS effects with this device. The MR curves were recorded after various visible light-irradiation times for a device fabricated with M-cis isomer, as shown in Fig. 3c. In the initial state, a clear antisymmetric MR curve with a negative slope was observed (blue curve). The irradiation time dependence of MR signals at −1 T are plotted in Fig. 3d. The MR signal decreased exponentially as light irradiation proceeded, and finally the sign of the MR signal was inverted to negative, indicating a light-induced spin switching in the spin-polarized current from up-spin selective to down-spin selective through the M-cis-to-P-trans photoisomerization of the OCA-film. A subsequent thermal activation process at 80 °C for the P-trans isomer inverted the signs of MR at −1 T from negative to positive again, as shown in Fig. 3e, implying a thermal-activation-induced spin switching from down-spin selective to up-spin selective through the P-trans-to-M-trans chirality inversion. Similar phenomena were observed in the device fabricated with the M-trans isomer, as shown in Fig. 3f (M-trans-to-P-cis photoisomerization), g (time dependence of MR), and h (P-cis-to-M-cis thermoisomerization). The phenomena with opposite sign were also observed in the device fabricated with 2’S isomers (Supplementary Fig. 3). This series of experiments clearly demonstrated that 4 times spin switching were induced during the 360° molecular rotation of the OCA. When compared with the device with gold electrode, the origin of the successful SP switching by the photoisomerization in the PEDOT/PSS device should be the nanoscale steric freedom in between the PEDOT polymer and inorganic Al2O3/Ni electrodes. Hence, the utilization of at least one flexible electrode is a crucial condition for the device fabrication. This result suggests the importance of flexibility at the molecular scale44 when molecular machines are integrated in solid-state devices.

### Spin-polarized conductive AFM measurements

The absolute MR value of ~2% observed at each state is insignificant compared with other CISS experiments23,35,37,38. This is likely owing to the unpolarized leakage currents through the pinholes in the OCA films because the OCAs are not perfectly packed in the films. For an accurate estimation of the SP rate, we investigated a microscopic experiment, i.e., spin-polarized conductive AFM utilizing ferromagnetic tips. In these measurements, thin-films of OCAs were spin coated on freshly cleaved highly oriented pyrolytic graphite (HOPG) substrates. Current–voltage (IV) measurements were performed in the contact mode with ferromagnetic CoCr tips (coercivity: 300 Oe) magnetized along the up or down direction (Fig. 4a). At least 40 IV traces were recorded at different positions and subsequently averaged for each magnetized direction. The absolute current is always larger under a positive bias irrespective of molecular chirality. This asymmetric feature can be explained by the work function difference between the HOPG substrate and CoCr tip. The work functions of HOPG (in air) and CoCr tip are ~4.85 eV and 4.5 eV, respectively. This implies that the electron injection barrier is much lower at the tip–molecule interface than that at the substrate–molecule interface. Fig. 4b shows the averaged IV traces for the device fabricated with the M-trans isomer measured with the up-direction magnetized CoCr tips before and after visible-light irradiation (the nonaveraged raw data are shown in Supplementary Fig. 6). The current flowing through the M-trans isomer (initial state) was higher compared with that through the P-cis isomer (after photoirradiation). Meanwhile, for CoCr tips that are magnetized in the down direction, the current flowing through the M-trans isomer (initial state) was smaller compared with that through the P-cis isomer (Fig. 4c), indicating light-induced spin switching in the spin-polarized current. Furthermore, we performed experiments for two other case scenarios: SP switching by photoisomerization from M-cis to P-trans measured with CoCr tip magnetized with up and down directions (Fig. 4d, e), separately. In all cases, the observed SP were up-spin selective for M-helicity molecules and down-spin selective for P-helicity molecules. This tendency is consistent with the results of MR measurements above. The SP rate in the M-trans isomer was calculated using the following formula:

$${\mathrm{SP}}\left( \% \right) = \left( {I_{{\mathrm{down}}} - I_{{\mathrm{up}}}} \right)/(I_{{\mathrm{down}}} + I_{{\mathrm{up}}}) \times 100$$
(1)

where Iup and Idown represent the currents when the magnetized direction of the CoCr tip is pointing up and down, respectively. The SP rate estimated at each bias is plotted in Fig. 4f. The SP rate enhanced gradually with increasing bias and finally reached 43% at +1.9 V. Meanwhile, for the photoirradiated states (P-cis isomer), the SP rate was estimated as –44% at +1.9 V. The magnitude of the SP rate in the present experiments is comparable with those in previous CISS experiments. However, the present experiments demonstrated the light-induced inversion of the SP direction in the spin-polarized currents, for the first time.

## Discussion

We now discuss the possible mechanism of the CISS effects in the present system. Hitherto, several theoretical mechanisms for the CISS effects have been proposed45,46,47,48,49. For instance, Matiyahu et al. recently proposed a theoretical model for CISS effects that is a simple tight binding model for electron transport through a single helical molecule, with SOIs on the bonds along the helix48. They introduced quantum interference through additional electron hopping between neighboring chains and allowed for the leakage of electrons from the helix to the environment to break a time-reversal symmetry. Consequently, a net spin polarization of the outgoing electrons was obtained. These results enabled a qualitative understanding of the mechanism for the CISS effects while the significant magnitude of SP rate (e.g., maximum of 43% in the current experiments) arising from small SOIs of organic molecules is still being investigated.

In this study, we performed the calculations for SOIs for the OCAs in one-electron reduced states by the contracted spin–orbit-configuration interaction (SO-CI) method. The electronic states (ϕn; n is the index number) for the anion radical of M-trans were represented by singly excited configurations. The SOI was evaluated by < ϕn |HSOIm > where HSOI is the SOI Hamiltonian, and subsequently diagonalized to obtain the spin–orbit states. This method has been employed for calculating intersystem crossing energies in a photoreactive molecule to yield reasonable values50. For the anion radical state, n and m were chosen within the LUMO to LUMO+25 levels (LUMO = lowest unoccupied molecular orbital for the neutral molecule). We have calculated these transition energies because the OCA molecule should be in a one-electron reduced state during the electron transmission process and should experience interstate transitions that start from a state in an electrode on one side and ends at a state in an electrode in the opposite side. Interestingly, we found that some matrix elements of the nm interaction are extremely large, although most of the interactions are relatively weak (Supplementary Table 1 and Supplementary Fig. 8). Large values exceeding 30 cm−1 are shown for the transition between state 22 (LUMO+21) and state 25 (LUMO+24), for example (Supplementary Fig. 9). In these cases, the transitions contain the π-electron-to-lone-pair or π-electron-to-σ-electron excitation within the same sulfur atom, which should be associated with large orbital angular momentum changes in sulfur atomic orbitals. This result suggests the importance of SOIs within degenerate atomic orbitals in the CISS effect, which is reminiscent of strong SOIs in corrugated graphene and carbon nanotubes51,52. It is notable that the energy difference between up-spin and down-spin becomes twice as large as the SOI energy. Because the room temperature energy is ~200 cm−1, the SOI splitting of 60–70 cm−1 is more than one-third of the room temperature energy, which appears sufficient to explain the large SP observed in our experiments. Although we do not know why or how these strong SOIs affect the whole transmission process, the energy order appears reasonable despite the OCA being composed of only light elements. It is noteworthy that we did not use gold electrodes in the MR measurements (Fig. 3) that excludes the effect of heavy element electrodes that are often reported in literature.

In summary, we have performed MR and spin-polarized conductive AFM measurements that exhibit SP direction switching upon photoirradiation or heat treatment, by utilizing an artificial molecular motor. In these devices, the P/M chirality, which is the origin of SP generation through the CISS effect, is reconfigurable by external stimuli and thus represents a new type of organic spintronics device. During the development of this device, we found that the flexibility of the electrode at the nanoscale is important. The present results are beneficial for the development of next-generation organic photo/thermospintronic devices combined with molecular machines.

## Methods

### Synthesis and characterization of OCAs

OCAs were synthesized according to the reported procedure43. The cis and trans isomers were separated by silica gel column chromatography (n-hexane:CH2Cl2 = 2:1). The optical resolution was performed by a CHIRAL ART Cellulose-SB (SMC) column using n-hexane: methyl tert-butyl ether (65:35). The elution times were 2’S-P-trans: 10.5 min; 2’R-P-trans: 11.5 min; 2’S-M-trans: 12.5 min; 2’R-M-trans: 13.5 min for trans isomer, and 2’S-P-cis: 10.5 min; 2’R-P-cis: 11 min; 2’S-M-cis: 11.5 min; 2’R-M-cis: 13 min for cis isomer. The enantiopurity in each state confirmed by HPLC is >95%. CD spectra were recorded on J-720WI (JASCO) in toluene solution (Supplementary Fig.  1).

### Fabrication process of MR devices

The devices for the MR measurements were fabricated as follows: a Ni layer (50 nm) that acts as a bottom magnetic electrode was fabricated on a cleaned SiO2 substrate by optical lithography and sputtering. Subsequently, the Ni electrode was covered by an insulating Al2O3 layer (3 nm) by the sputtering method. The OCA layer was deposited by spin coating from toluene solution (2500 rpm, 30 s), followed by annealing at 50 °C for 1 h to evaporate the solvent. To prepare the PEDOT/PSS electrodes, we slightly modified a reported method53. A freestanding PEDOT/PSS film obtained in water was laminated once on a Teflon sheet to be dried. Subsequently, the PEDOT/PSS film was peeled off gently from the sheet using a pair of tweezers and laminated on the SiO2/Ni/Al2O3/OCA substrates with the crossbar configuration (Fig. 1). Finally, electrical contacts were established using silver paste.

### MR measurements

Electrical contacts were established in a standard four-probe configuration with a cross-bridge geometry. MR measurements were performed under reduced He pressure in a physical property measurement system (PPMS) (Quantum Design). A constant DC current of 1 μA was applied using the 2636B source meter (Keithley) and the potential difference across the junction was measured using the 2182 A nanovoltmeter (Keithley). A magnetic field of upto 1.0 T was applied perpendicular to the device interface. The temperature of the sample was controlled by the PPMS.

### Spin-polarized conductive AFM measurements

Spin-polarized conductive AFM measurements (I–V measurements) were performed using the JSPM-4210 scanning probe microscope (JEOL). In this experiment, OCA films were deposited on freshly cleaved HOPG substrates by spin coating from toluene solution (2500 rpm, 30 s), followed by annealing at 50 °C for 1 h to evaporate the solvent. The I–V measurements were recorded in the contact mode by applying the bias voltage to the tip (cantilever). CoCr tips (MESP-V2, Bruker) that were magnetized in the up and down magnetic field orientations were used for the I–V measurements. At least 40 I–V traces were recorded and averaged for each chiral state or magnetic field orientation. For each I–V measurement, the tip was placed in a new position.