Dual-radical-based molecular anisotropy and synergy effect of semi-conductivity and valence tautomerization in a photoswitchable coordination polymer

Abstract Organic radicals are widely used as linkers or ligands to synthesize molecular magnetic materials. However, studies regarding the molecular anisotropies of radical-based magnetic materials and their multifunctionalities are rare. Herein, a photoisomerizable diarylethene ligand was used to form {[CoIII(3,5-DTSQ·–)(3,5-DTCat2–)]2(6F-DAE-py2)}·3CH3CN·H2O (o-1·3CH3CN·H2O, 6F-DAE-py2 = 1,2-bis(2-methyl-5-(4-pyridyl)-3-thienyl)perfluorocyclopentene), a valence-tautomeric (VT) coordination polymer. We directly observed dual radicals for a single crystal using high-field/-frequency (∼13.3 T and ∼360 GHz) electron paramagnetic resonance (EPR) spectroscopy along the c-axis, which was further confirmed by angle-dependent Q-band EPR spectroscopy. Moreover, a conductive anomaly close to the VT transition temperature was observed only when probes were attached at the ab plane of the single crystal, indicative of synergy between valence tautomerism and conductivity. Structural anisotropy studies and density functional theory (DFT) calculations revealed that this synergy is due to electron transfer associated with valence tautomerism. This study presents the first example of dual-radical-based molecular anisotropy and charge-transfer-induced conductive anisotropy in a photoswitchable coordination polymer.


INTRODUCTION
Reducing the scale of manufacturing processes using effective, small-sized alternative materials is a recent focus of scientific research, with the miniaturization of pure Si nanostructures according to Moore's law being almost complete [1]. Materials that exhibit different states in response to external stimuli on a single-molecule level are currently being scrutinized, as they represent the ultimate form of miniaturization [2,3]. Therefore, compounds that combine orbitally degenerate metals with organic radical ligands, including spin-crossover (SCO) [4,5] and valence-tautomeric (VT) compounds [6,7], are being increasingly explored as novel bistable magnetic materials.
In several cases, bifunctional VT compounds have been obtained by introducing ligands that exhibit fluorescence [30,31], conductivity [32,33] and photoisomerization [34]. Furthermore, electronic conductivity, a basic property of matter [35], is extensively used by various multifunctional materials [36][37][38]. Lukyanov et al. investigated the effect of metal-ligand electron transfer on the conductivities of two semiquinonate complexes [39]; however, conductivity/magnetism synergy was not analysed. O'Sullivan et al. attempted to synthesize a VT conductive metal polymer [40], while Kanegawa et al. synthesized a novel VT compound by covalently linking a conductive tetrathiafulvalenefunctionalized (TTF-functionalized) phenanthroline ligand, but no relevant conductivity data were provided [33]. Compared with most SCO conductors, the electron-transfer process in the VT complex is intramolecular, indicating a potentially larger transfer rate. In addition, electron transfer occurs simultaneously with the valence tautomerism of the metal center in the complexes. This contributes to the study of the synergistic effect between magnetism and conductivity. Currently, studies on VT conductors are rare in relation to those on SCO conductors, and more conductive VT systems need to be developed. In particular, elucidating synergistic mechanisms within these bifunctional materials is of considerable significance owing to a lack of similar previous studies.

Strategy and chain structures
A pyridyl derivative of a diarylethene (DAE) ligand was used as the linker along with the Co-dioxolene unit, which is a classical VT building block, to form a VT CP with a chain structure. Light may promote the rearrangement of the crystal structure and further influence magnetic properties through modification of the ligand field (isomers differ substantially in shape and conjugation), leading to photocontrol and deep insight into organic radicals on a 1D chain. Here, we synthesized the novel VT o-1·3CH 3 CN·H 2 O CP using an open-form derivative of the DAE ligand. The NMR spectrum and SC-XRD structure of the open-form ligand are shown in Supplementary Fig. S1a and S1b. The 6F-DAEpy 2 ligand isomerizes when irradiated with light and has been widely studied for the construction of photochromic or photoswitchable compounds. David et al. presented two erbium(III)-based coordination systems that showed slow dynamics of magnetization using 6F-DAE-py 2 as bridging ligands [41]. A 1D coordination solid was also synthesized by reaction of the DAE photochromic unit with the dysprosium-based single-molecule magnet [42]. In other cases, the 6F-DAE-py 2 ligand was used as the photoresponsive component in photoswitchable materials [43,44]. However, less attention has been given to the bending structure in space that may lead to specific properties, as depicted in this work. Here, the closed-form polymer (c-1·H 2 O) was formed using the closed-form ligand c-L, which was obtained from a 365-nm UV-light-irradiated solution of o-L.
Notably, o-1·3CH 3 CN·H 2 O and c-1·H 2 O exhibit the same coordination mode, in which each Co(III) center is coordinated to two dioxolene ligands (one in the catecholate state and the other in the semiquinone state) through four oxygen atoms, with two ancillary ligands located in the vertical direction with trans-disposed N atoms of pyridyl groups. The central Co ions are therefore octahedrally configured. Discrete Co-dioxolene units are connected by linear 6F-DAE-py 2 ligands to form 1D chains along the c-axis. However, the 1D chains in o-1·3CH 3  Packing diagrams of o-1·3CH 3 CN·H 2 O and c-1·H 2 O are shown in Fig. 1b and c, with both exhibiting 1D chains. However, these chains differ due to the different configurations of o-L and c-L. o-1·3CH 3 CN·H 2 O crystallizes in the P6 3 /m space group, with the Co ions in distorted octahedral environments. Two types of crystallographically distinct central Co ions, referred to as Co 1 and Co 2 , are observed in these chains due to their zigzag configurations. In o-1·3CH 3 CN·H 2 O, we may observe that the two types of central Co ions spread hexagonally over the ab plane when viewed along the c-axis ( Fig. 2a and b, with Co 1 and Co 2 shown in pink and green, respectively); this corresponds to the hexagonal crystal system of the open-form complex. o-1·3CH 3 CN·H 2 O was subjected to SC-XRD at 100 K. The C−O bond lengths enable the Cat 2-(single bond: 1.344/1.355Å, with the expected value in the range of 1.33−1.39Å) and SQ ·-(double bond: 1.317/1.362Å) states of the two dioxolene ligands coordinated to Co 1 to be identified, whereas Co 2 is located at an imposed inversion center, rendering Cat 2and SQ ·indistinguishable (Supplementary Fig. S1c and S1d). In this case, the electronic states of the dioxolene ligands of Co 2 are averaged and random rather than directional electron transfer is observed between the ligand and metal center. Moreover, the steric configurations of the dioxolene ligands on Co 1 and Co 2 differ. Figure 2c reveals that the semiquinone ligand coordinated to Co 1 (referred to as Radical 1) and the dioxolene ligand coordinated to Co 2 (both referred to as Radical 2) are not coplanar, with a dihedral angle of 68.85 • between them. The crystal structure modes used in the DFT calculations are described in Supplementary Figs S4−S6. In contrast, c-1·H 2 O crystallizes in the monoclinic P2 1 /n space group, with only one type of Co central ion in the chain connected by the less distorted c-L. Packing diagrams of c-1·H 2 O are shown in Supplementary Fig. S7. Additionally, SC-XRD helps to elucidate the oxidation states of the ligands and Co centers in c-1·H 2 O.

Electronic and NMR spectroscopy
Photoconversion between the open-and closedform species (ligands and complexes) due to the effect of DAE ligand photoisomerization was explored. Reversible conversion of o-L and c-L under UV or visible light was confirmed by using UV/vis absorption spectroscopy, with second-timescale responses observed ( Fig. 3a and b). Electronic absorption spectroscopy was then used to qualitatively study the photoconversion of o-1·3CH 3 CN·H 2 O and c-1·H 2 O when irradiated with light. Photocyclization of o-1·3CH 3 CN·H 2 O in dimethylformamide (DMF) is initiated by 365-nm UV light in seconds, with the solution observed to change from colorless to light blue. The reverse reaction occurs when irradiated with visible light (>520 nm, Fig. 3c). Photoisomerization was also studied in the solid state using a KBr pellet (Fig. 3d). Photocyclization occurred in minutes when irradiated with 365-nm UV light and was complete in one hour; the reverse process (from c-1·H 2 O to o-1·3CH 3 CN·H 2 O) was also observed under visible light. Despite the longer response times compared with those observed for pure organic DAE crystals [45] or single-molecule junctions [46], o-1·3CH 3 CN·H 2 O and c-1·H 2 O effectively photoswitch in solution or in the bulk state in a reversible manner.
NMR spectroscopy was used to further qualitatively and semiquantitatively study photocyclization in solution (Supplementary Figs S8−S13). 1 H NMR spectroscopy was used to study the stability of o-1·3CH 3 CN·H 2 O upon heating. 19 F NMR spectroscopy was used to track the photocyclization progress and calculate yields based on 19 F signal intensities. To support the experimental results, we performed DFT calculations based on

EPR spectroscopy
Single crystals of o-1·3CH 3 CN·H 2 O and c-1·H 2 O exhibit significant structural anisotropies involving SQ ·radicals at the Co metal centers. EPR spectroscopy can be used to confirm the presence of SQ ·radicals in o-1·3CH 3 CN·H 2 O and c-1·H 2 O. The X-band (9.1 GHz) EPR spectra of powder samples of the open-and closed-form complexes at 100 K (Fig. 4a) are consistent with the presence of organic radicals (g ≈ 2.0023); broad lines that lack hyperfine features were observed, suggesting that the SQ ·radical is immobilized due to coordination with the LS closed-shell Co(III) center (S = 0). Interactions between the spin magnetic moments of unpaired electrons and metal nuclei were absent in these systems and o-1·3CH 3 CN·H 2 O and c-1·H 2 O exhibited different degrees of line broadening. Subsequently, high-field (∼13.3 T) and -frequency (∼360 GHz) single-crystal EPR spectroscopy was used to significantly enhance the resolution of the Zeeman splitting in o-1·3CH 3 CN·H 2 O, which led to the observation of a remarkable phenomenon. In particular, we directly visualized dual SQ ·radical signals that correspond to two crystallographically independent Co building blocks associated with the zigzag chain structure when the c-axis was aligned with the static field (Fig. 4b). Strong signals centered at 12.86 and 12.91 T with g = 2.000(1) and 1.992(3), respectively, were observed at 50 K. These dual radicals remained during cooling from 50 to 4.2 K. Moreover, only a split peak and no dual radical signals were observed when the static field was applied in other directions ( Supplementary Fig. S19). Such directional observation of dual radicals is rare in VT systems [47,48]. Two crystallographically independent radicals are detected using high-field and high-frequency EPR spectroscopy along the c-axis because this axis corresponds to the extension direction of the 1D chains in o-1·3CH 3 CN·H 2 O. In contrast, no dual SQ ·radical signals were observed for c-1·H 2 O due to its quasi-linear 1D structure.
To conveniently investigate the two directional radicals, we determined the face index of a single We used a single crystal of o-1·3CH 3 CN·H 2 O to collect a full rotational pattern of EPR spectra Natl Sci Rev, 2023, Vol. 10, nwad047 when the magnetic field was perpendicular to the ab plane to further study the two independent radical signals, which directly reflect the orientations of the radicals in the crystal frame. As shown in Fig. 4d, a static field is exerted along the c-axis, i.e. the magnetic field coincides perfectly with the c-axis. The intersection angle of the field and c-axis is defined as θ and was initially 0 • . The single crystal was then rotated to move the c-axis close to the initial position of the ab plane; hence, θ increased gradually. EPR spectra were acquired as the angle was varied from 0 • to 360 • . The X-band EPR spectra of o-1·3CH 3 CN·H 2 O did not show a clear difference between the signals of the two radicals (Supplementary Fig. S21). However, the resolution of the Qband (34.0 GHz) single-crystal EPR spectrum of o-1·3CH 3 CN·H 2 O acquired at 300 K is improved by approximately one order of magnitude, thereby providing higher sensitivity for distinguishing the two independent radicals as the angle was varied from 0 • to 360 • (Fig. 5a). Meanwhile, the angle-dependent behavior of the g 2 value in the Q-band is consistent with these results (Fig. 5b). The g 2 value exhibits a characteristic periodic variation from a minimum of 3.976(0) to a maximum of 4.017(8) as θ in-creases from 0 • to 90 • and then declines to 3.975(6) at θ = 180 • . The angle-dependent EPR data for o-1·3CH 3 CN·H 2 O are likely to be associated with its crystal structure and the two independent radicals. The c-axis corresponds to the extension direction of the 1D chains in o-1·3CH 3 CN·H 2 O. The g z of Co 1 in the octahedral configuration can be identified in angle-dependent EPR spectroscopy incorporating the face index. This is due to the coincidence of g z of Co 1 and the c-axis. Notably, the g 1 value corresponds exactly to the g z value of Radical 1 at θ = 0 • (the field is directed through two conical points of the Co 1 octahedral sphere). The g values of 1.994(0) and 1.993(9) at θ = 0 • and 180 • , respectively, indicate that the g value of 1.992(3) obtained in the high-field and high-frequency spectra is attributed to Radical 1, g 1 . In this case, the g 2 value of Radical 2 is 2.000(1), which is obtained in the high-field and high-frequency spectra. After the c-axis rotates into the initial ab plane (θ = 90 • ), where g 2 realizes its maximum value, the measured g value of the Co 1 center is g xy (Fig. 5c). The field direction coincides with the c-axis again as the single crystal is continuously rotated. One periodic variation is completed as θ is increased from 0 • to 180 • .

Magnetic studies
To evaluate how structural switching affects its physical properties, the magnetic susceptibility of o-1·3CH 3 CN·H 2 O was measured in the range of 5-400 K, the results of which are shown in Fig. 6a. χ m T (where χ m is the molar magnetic susceptibility and T is the temperature) was observed to be almost constant (approximately 0.9 cm 3 mol −1 K) regardless of the temperature below 300 K when o-1·3CH 3 CN·H 2 O was initially heated, consistently with the low-spin electronic structures of the two LS-Co III (3,5-DTCat 2-)(3,5-DTSQ ·-) species. The magnetic susceptibility abruptly increased at approximately 350 K to 6.21 cm 3 mol −1 K at 400 K, which is close to that of two HS-Co II (3,5-DTSQ ·-) 2 units. Magnetic data could not be collected above 400 K due to limitations associated with the maximum operating temperature of our magnetometer. The observed change in magnetic susceptibility suggests that the thermally induced LS-Co III (3,5-DTCat 2-)(3,5-DTSQ ·-) to HS-Co II (3,5-DTSQ ·-) 2 VT transition was almost complete, with T c↑ = 390 K ( Supplementary Fig. S22), indicative of charge transfer from Cat 2to the Co(III) center. o-1·3CH 3 CN·H 2 O displays a gradual incomplete transition from a high-to a low-spin state during backtracking, with T c↓ = 360 K, which is attributable to the loss of solvent molecules at high temperature [49]. This is consistent with the ther-mogravimetric data of o-1·3CH 3 CN·H 2 O, which shows the loss and partial loss of water and acetonitrile molecules (Supplementary Fig. S23). On the other hand, c-1·H 2 O exhibited different behavior ( Supplementary Fig. S24). During initial heating, the magnetic susceptibility of c-1·H 2 O remained in the range consistently with slightly coupled LS-Co III (3,5-DTCat 2-)(3,5-DTSQ ·-) species below 300 K, with a more abrupt spin transition observed at approximately 300 K. A maximum value of 1.99 cm 3 mol −1 K was observed at 325 K, which increased gradually with increasing temperature to 3.11 cm 3 mol −1 K at 400 K. In addition, c-1·H 2 O exhibited a hysteresis loop of nearly 21 K at approximately 300 K (critical temperatures: T c↑ = 311 K, T c↓ = 290 K) upon cooling, which is rare among general VT systems [50][51][52][53][54]. These results suggest that c-1·H 2 O favors cooperativity more than o-1·3CH 3 CN·H 2 O; this higher cooperativity is possibly ascribable to the π -conjugated structure of c- egy adopted in this system to form polymers with switchable physical properties through the introduction of ligands that are photocyclizable. In addition, we used variable-temperature EPR spectroscopy to confirm the VT transition in c-1·H 2 O (Supplementary Fig. S25).

Electrical conductivity studies
The current-voltage (I-V curves) characteristics of o-1·3CH 3 CN·H 2 O and c-1·H 2 O were examined to further explore the multifunctionalities of the VT complexes and the directionalities of the dual radicals ( Supplementary Fig. S26). The I-V curves of o-1·3CH 3 CN·H 2 O and c-1·H 2 O exhibit almost linear relationships, indicative of low but different conducting performance. A two-orders-ofmagnitude higher current was detected using o-1·3CH 3 CN·H 2 O compared with c-1·H 2 O, which exhibits insulating behavior. We then studied the temperature dependence of the electrical conductivity σ in o-1·3CH 3 CN·H 2 O (Fig. 6b). The results of three selected single crystals show crossover at approximately 330 K, where the VT transition occurs, which indicates that semi-conductivity is possibly synchronized with thermally induced valence tautomerism. The average conductivity at 327 K was determined to be 2.1 × 10 -8 S cm −1 , which is characteristic of semiconductors with very small numbers of free carriers (electrons in the conduc-tion bands or holes in the valence bands) compared with the number of atoms. Here, we observed synergy between magnetism and conductivity in o-1·3CH 3 CN·H 2 O. Electrical conductivity increased with increasing temperature below the transition temperature, which is a characteristic semiconductor behavior. However, electrical conductivity was observed to decrease with increasing temperature above 330 K; hence, o-1·3CH 3 CN·H 2 O showed a behavior of decreasing conductivity with increasing temperature, which is ascribable to electron transfer during valence tautomerism. Similar behavior has also been observed in a TTF-chloranil chargetransfer salt, which exhibits a neutral (N) to ionic (I) phase transition [55]. However, such behavior was observed for the first time in a single-crystal VT complex in this study. Notably, while we evaluated the conductivity of the ab plane in a single crystal, different results were obtained regarding the caxis. The resistivity of a single crystal decreased with increasing temperature when two Au wires were placed along the c-axis ( Supplementary Fig. S27), consistently with classic semiconductor behavior. A slight increase was observed between 340 and 360 K, which is likely due to a crack in the crystal caused by the high temperature [56,57]. The conductivity in the ab plane is attributable to charge transfer from the ligand to the Co ion in the VT complex. The Co 1 center in the 1D chain of o-1·3CH 3  transfer from the catecholate ligand to the Co 1 ion (corresponding to Radical 1) occurred during heating, whereas the direction of electron transfer was random for Co 2 (corresponding to Radical 2) because of the averaged dioxolene ligands. Radicals 1 and 2 are distributed in the ab plane; hence, all electron-transfer processes occur in the ab plane, i.e. directional or non-directional electron hopping capable of forming short-range conductive pathways occurs in the ab plane ( Supplementary Fig. 6c). However, no charge transfer between the ligand and the metal was observed along the c-axis, as this corresponds to the direction of the 6F-DAE-py 2 ligand.
To determine the origin of the anomalous semiconductive behavior in o-1·3CH 3 CN·H 2 O, the electrical conductivity was further analysed. Several single crystals were subjected to direct current (DC) electrical resistivity studies in the range of 200-360 K. The low conductivity and strong thermal activation observed below 300 K indicate that charge transport occurred by polaron hopping. In terms of structure, intra/intermolecular packing may provide a likely electronic pathway along the c-axis within o-1·3CH 3 CN·H 2 O, as shown in Supplementary  Fig. S28. The Arrhenius equation was used to fit the linear region of the curve: l og 10 (ρ) = l og 10 (ρ 0 ) + ( k E a / l og 10 e ) · ( 1 / T ). The activation energies of o-1·3CH 3 CN·H 2 O at various temperatures are shown in Supplementary Table S9 along with the corresponding crossover temperatures. The observed increase in conductivity is likely due to a change in activation energy caused by a dynamic charge fluctuation ascribable to charge transfer. According to the Arrhenius plot, it has a lower activation energy (E a ) in the relatively high temperature range (248−273 meV between 283 and 321 K) compared with the higher activation energy in the low temperature range (403−448 meV between 246 and 297 K). As a result, semiconductor behavior was observed before the VT transition.

DFT calculations
The change (conducting anomaly) in conductivity observed in o-1·3CH 3 CN·H 2 O upon valence tautomerism is attributable to a change in the rate of charge-carrier hopping. DFT calculations were used to reveal the synergistic mechanism responsible for the conductivity and magnetism observed in o-1·3CH 3 CN·H 2 O (see Supplementary information for further details). The B3LYP (for isolated systems) and PBE + U (where the U value of the Co d electrons in extended systems is set to 6 eV) functionals were used to describe the different spin states in the low-and high-temperature structures ( Supplementary Fig. S31). The abovementioned DFT methods correctly reproduced the electronic structures of the CoL 2 units (where L represents the doubly charged 3,5-DTCat 2-, singly charged 3,5-DTSQ ·or neutral chelating 3,5-ditert-butyl-o-benzoquinone ligand), which is consistent with the experimental magnetic susceptibility results. The CoL 2 units are in the low-spin LS-Co III (3,5-DTCat 2-)(3,5-DTSQ ·-) state at low temperatures, whereas they are in the high-spin HS-Co II (3,5-DTSQ ·-) FM (3,5-DTSQ ·-) FM state at high temperatures.
The PBE + U band structure of o-1·3CH 3 CN·H 2 O reveals that the system should exhibit high conductivity due to the small band gap (Supplementary Fig. S30); however, this prediction is inconsistent with the experimental results. Pure DFT functionals are known to underestimate band gaps. In principle, the DFT + U formalism can help counteract this [58]; however, while effective on-site coulombic repulsions are only added to the d electrons of the Co atoms, the bands close to the Fermi level mainly comprise the frontier orbitals of the organic ligands. In principle, the over-delocalization problem can be resolved using a hybrid functional (e.g. B3LYP, PBE0) to predict a more accurate band gap; however, this requires calculating the exact exchange term, which entails a high cost under periodic boundary conditions in the standard implementation and is consequently almost infeasible for such a large system.
According to the band structures and calculations of the isolated systems, the bare on-site coulombic repulsion associated with the CoL 2 units is much larger than the bandwidths formed by orbital interactions between CoL 2 units. Therefore, the system is likely to be in a spin-localized state as a Mott insulator. Furthermore, a band structure only reflects conductivity when electrons are added or removed from bands in the vicinity of the Fermi level without electron configuration and lattice relaxation. In addition, the hopping mechanism should dominate conductivity in the case of o-1·3CH 3 CN·H 2 O according to the temperature-dependent electrical conductivity data; hence, the charged CoL 2 units relax to different configurations with lower energies during charge transport.
Based on the molecular structures of the CoL 2 unit at low (100 K) and high (380 K) temperatures extracted from o-1·3CH 3 CN·H 2 O, KS wave functions were calculated using DFT functionals, starting with the listed initial guesses (Supplementary Table S12) and selected using the SCF procedure, the results of which are summarized in Fig. 7. For [CoL 2 ] + , which is positively charged, the system always converges to the LS-Co III (3,5-DTSQ ·-)(3,5-DTSQ ·-) AFM (II + ) (AFM = antiferromagnetic) configuration as the ground state with a closed-shell LS-Co III center and a pair of AFM-coupled semiquinonate radicals. Hence, the spin state of Co should convert from HS-Co II to LS-Co III , even at high temperatures. The two semiquinonate ligands can also be FM-coupled (FM = ferromagnetic), albeit with slightly higher energy. B3LYP KS-DFT calculations reveal that the low-and high-temperature molecular structures differ by 0.0158 and 0.0052 eV, respectively. For negatively charged [CoL 2 ] − , a system with a lowtemperature molecular structure is likely to converge to the closed-shell LS-Co III (3,5-DTCat 2-)(3,5-DTCat 2-) (I − ) configuration. However, the high-temperature molecular structure is divided in a manner that depends on the exact exchange ratio, where the I − configurations and the open-shell HS-Co II (3,5-DTCat 2-)(3,5-DTSQ ·-) FM (V − ) configurations are close in energy (Supplementary Tables S13-S16).
The electron-transfer matrix elements were calculated based on these electronic configurations and the pairwise spatial relationships of the CoL 2 units in the two crystals. According to these calculations, o-1·3CH 3 CN·H 2 O displays 2D semiconducting properties in the ab plane, with the symmetric CoL 2 layers (Co 2 ) playing major conductivity roles, consistently with anisotropic conductivity results ( Fig. 6b and Supplementary Fig. S27). Electrons act as carriers at high temperature; however, hopping events are dominated by hole conduction at low temperatures. In the case of o-1·3CH 3 CN·H 2 O, the carrier density increases with increasing temperature due to thermal excitation, which is common for a semiconductor; consequently, the conductivity increases. The spin states of the CoL 2 units begin to undergo low-to high-spin conversion at approximately 300 K; as a result, a CoL 2 unit may be surrounded by other units with different spin states. The rate of electron transfer then decreases to zero, leading to a decline in conductivity, which explains the peak conductivity observations (Fig.  6b). Therefore, for the electrical conductivity study, the Arrhenius analysis determined the origin of the semiconductive behavior before 320 K, while DFT calculations helped illustrate the mechanism of the unusual change in electrical conductivity during the VT transition process.

CONCLUSION
The photoisomerizable 6F-DAE-py 2 ligand was used to synthesize two VT CPs, namely o-1·3CH 3 CN·H 2 O and c-1·H 2 O, which display VT and photoconversion behavior. Two crystallographically independent SQ ·radicals arranged along the 1D chain were observed in o-1·3CH 3 CN·H 2 O using SC-XRD. Therefore, two clearly distinct radical signals were observed in the high-field/-frequency single-crystal EPR spectra of a VT complex for the first time. The directional variations of these dual radicals, as well as their 180 • periods, were then confirmed by using angle-dependent EPR spectroscopy. Furthermore, o-1·3CH 3 CN·H 2 O exhibited anisotropic conductivity/magnetism synergy. A conductive anomaly close to the VT transition temperature was observed along the ab plane in o-1·3CH 3 CN·H 2 O, while insulator-like behavior was observed along the c-axis. Meanwhile, the synergistic mechanism was explained using DFT calculations. This study is of considerable significance with regard to the anisotropy and synthesis of multifunctional bistable magnetic materials in the molecular magnetism field.

Lead contact
Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Masahiro Yamashita (yama-sita@agnus.chem.tohoku.ac.jp), and Zhao-Yang Li (zhaoyang@nankai.edu.cn).

Materials availability
Complex o-1·3CH 3 CN·H 2 O and c-1·H 2 O can be produced following the procedures outlined below from standard reagents and procedures.

Data and code availability
Crystal data for o-1·3CH 3

Synthesis
All reagents were obtained from commercial sources and used without further treatment. The open/closed-form ligands o-L and c-L were synthesized according to previously reported procedures (Supplementary Scheme S1) [59].
The open-form 6F-DAE-py 2 ligand (o-L) was synthesized as previously described [59]. Based on our experience, we note that separation and purification by column chromatography is very time-consuming; therefore, a gradated mixture of dichloromethane and ethyl acetate (from 5 : 1 to 1 : 1) was used as the eluent to enhance separation efficiency. A single crystal of the closed-form 6F-DAEpy 2 ligand (c-L) was isolated from an acetonitrile solution of o-L by irradiation with UV light (365 nm) for two hours. o-1·3CH 3

X-ray crystallography
o-1·3CH 3 CN·H 2 O and c-1·H 2 O were subjected to single-crystal X-ray diffractometry on a Rigaku Xta-lAB PRO MM007 DW diffractometer with MoK α (λ = 0.71073Å) radiation at 293 and 100 K, respectively. All collected data were integrated and revived using CrystalAlice Pro software. All structures were solved with the Olex2 structure-solution program using direct methods and refined by fullmatrix least-squares against F 2 in the SHELX refinement package. All non-hydrogen atoms were refined anisotropically. Free solvent molecules were removed using the squeeze command. A high beamstop theta(min) limit set and ignorance of the very high angle data that is inappropriate for highly disordered structures account for the relatively high

NMR spectroscopy
NMR spectra were recorded at field strengths of 7.9 T (400 MHz, Bruker Avance II instrument equipped with a broad band fluorine observation (BBFO) probe) and 14.09 T (600 MHz, Bruker Avance III instrument equipped with a QNP Cryoprobe). The temperature-dependent residual nondeuterated methanol signal in deuterated methanol was used to calibrate the temperature [60]. 1 H NMR spectra were referenced against the residual proton signal in DMF-d 7 [61]. The deuterated solvent (DMF-d 7 , Sigma-Aldrich) was dried and degassed using conventional methods prior to use. The NMR samples were prepared and stored in an inert atmosphere in tubes with Teflon plugs (J. Young valves).
NMR DFT calculations were performed using the Gaussian 09 (revision D.01) program. Theoretical studies on VT complexes have shown that the OPBE [62] and B3LYP * [63] functionals give reliable results [64]. More recently, DFT calculations on the VT [Co(3,5-DTSQ ·-)(3,5-DTCat 2-)(4-papy) 2 ] complex have been reported using the OPBE functional [22]. For this reason, the unrestricted B3LYP * functional and the def2svp [65] basis set were used to optimize the geometries of model complexes of the VT 1D chain. A quadratically convergent self-consistent field procedure was applied when convergence could not be achieved using the standard first-order procedure. A 'superfine' integration grid was used and symmetries were not restricted during optimization. In each case, the absence of an imaginary frequency confirms that a local minimum is located on the potential energy surface. Broken-symmetry calculations were performed with the UB3LYP * functional and the def2tzvp [65] basis set using the 'scf = xqc' procedure and 'superfine' integration grids; in these calculations, the stabilities of the wave functions were tested [66] and, where necessary, reoptimized. The xyz coordinates of the optimized structures, energies from single-point calculations and relevant spin-density maps are provided herein.

Magnetization
Magnetization experiments were carried out using a Quantum Design MPMS 5S SQUID magnetometer equipped with a 5-T magnet. Magnetization (M) was measured as the temperature was increased from 5 to 400 K at 5 K·min −1 in a 5000-Oe field (H) in three heating and cooling cycles. The magnetic susceptibility per mole (χ m ) was calculated as χ m = M m/ H.

Electrical conductivity
A single hexagonal crystal of o-1·3CH 3 CN·H 2 O was measured along the ab plane and the c-axis by using the two-probe DC method (Keithley 2400) using silver paste electrodes and soft goldcoated spider silk fibers with a diameter of approximately 25 μm as electrical wires. The proper orientation was established through the face index of the crystal by means of X-ray diffraction. The temperature-dependent conductivity was measured within a sealed cycle cryostat in the temperature range of 150−360 K.

Other characterizations
Powder XRD was performed on a Rigaku chargecoupled device (CCD) diffractometer. Electronic spectroscopy was performed on a Shimadzu UV3600IPLUS ultraviolet/visible/near-infrared spectrophotometer. EPR spectroscopy was performed on a Bruker EMX Plus spectrometer operating in the X-band (9.1 GHz) and Q-band (34.0 GHz) frequencies. High-field (∼13.3 T) and high-frequency (∼360 GHz) single-crystal EPR spectroscopy was performed on a terahertz electron spin resonance apparatus. EPR spectra were acquired at selected temperatures with variations in temperature measured using a helium continuous-flow thermostat. Angle-dependent EPR spectra were obtained by rotating a single crystal of o-1·3CH 3 CN·H 2 O in the plane perpendicular to the ab plane. Electrical conductivity was measured using a helium flow thermostat and temperaturedependent conductivity was determined with a voltage bias of 1.5 V at a ramping rate of 1 K min −1 . A single crystal of o-1·3CH 3 CN·H 2 O fixed on a quartz tube was photographed on a Rigaku XtalAB PRO MM007 DW diffractometer at room temperature. Face indexing was carried out by using the CrysAlisPro 171.40.84a program.