Single-Chain Magnets Based on Octacyanotungstate with the Highest Energy Barriers for Cyanide Compounds

By introducing large counter cations as the spacer, two isolated 3, 3-ladder compounds, (Ph4P)[CoII(3-Mepy)2.7(H2O)0.3WV(CN)8]·0.6H2O (1) and (Ph4As)[CoII(3-Mepy)3WV(CN)8] (2, 3-Mepy = 3-methylpyridine), were synthesized and characterized. Static and dynamic magnetic characterizations reveal that compounds 1 and 2 both behave as the single-chain magnets (SCMs) with very high energy barriers: 252(9) K for 1 and 224(7) K for 2, respectively. These two compounds display the highest relaxation barriers for cyano-bridged SCMs and are preceded only by two cobalt(II)-radical compounds among all SCMs. Meanwhile, a large coercive field of 26.2 kOe (1) and 22.6 kOe (2) were observed at 1.8 K.

Although the SCM-based magnets showing both the long range magnetic ordering and SCM behavior have added new aspects to the SCM researches 34 feasibility of using the Co II ion and the octacyanometallates for the construction of low dimensional magnetic materials of strong magnetic anisotropy. Inspired by aforementioned considerations, we devoted to the preparation of the SCMs using the [M V (CN) 8 ] 3− units and sterically bulky cations, such as the Ph 4 P + and Ph 4 As + cations. Here, we report the successful construction of two cyano-bridged 1D ladder chain compounds, or so-called 3, 3-ladder chain 39

Crystal Structures Description.
Single crystal X-ray diffraction analysis revealed that 1 and 2 are isomorphous and crystallized in the triclinic P1 space group (Supplementary Table S1). Both compounds are made up of anionic cyano-bridged Co-W 3, 3-ladder chains running along the a axis ( Fig. 1), Ph 4 P + (for 1) or Ph 4 As + (for 2) counter cations and lattice water molecules (for 1). Although 1 and 2 are isomorphous, disorder was only found in 1. One of the three 3-Mepy ligands of 1 has a site occupancy fact of only 0.7, while the rest 0.3 is occupied by a water molecule. Thus, each Co II center is in a distorted octahedral N 6 environment with six nitrogen atoms from three [W V (CN) 8 Tables S2 and S3). The continuous shape measures (CShMs) relative to the ideal octahedron calculated using the program SHAPE 2.1 42 was 0.253 (CoN 6 )/0.282 (CoN 5 O) and 0.236 for the Co II centers of 1 and 2, respectively. On the other hand, the polyhedral shape around the W centers does not correspond to ideal symmetry and the [W V (CN) 8 ] 3− unit shows a distorted dodecahedral configuration (TDD) for both 1 and 2 with the CShMs to be 0.56 and 0.66 (Supplementary Table S4). The W-C and C-N bond lengths and the W-C-N bond angles are also consistent with the reported values (Supplementary Tables S2 and S3).
Each [W V (CN) 8 ] 3− unit uses only three cyanides groups in a relatively T-shaped position to connect to three neighboring Co II centers, with the intrachain Co-W distances being 5.32, 5.38, 5.38 Å for 1 and 5.34, 5.39, 5.41 Å for 2, respectively. Bridged by these cyanide groups, a one-dimensional Co-W chain is formed along the a axis. Within the chain, both the Co II and W V centers connect to three neighbors and this kind of chain can be described as a 3, 3-ladder with regularly alternating W V and Co II along the edges 39 . However, as both the Co-N-C and W-C-N bond angles and the torsion angles of the W-C-N-Co connection deviate from 180° (Supplementary Tables S2 and S3), the 3, 3-ladders for both 1 and 2 are slightly distorted. Separated by the bulky Ph 4 P + or Ph 4 As + cations and 3-Mepy ligands ( Supplementary Fig. S2), these chains are well isolated to each other, with the shortest interchain Co···Co, Co···W, and W···W distances being 12.8, 11.1, 12.8 Å for 1 and 13.0, 12.9, 13.0 Å for 2, respectively. These large distances will efficiently prevent any interchain magnetic coupling and lead to the SCM behavior of these compounds, as described below.
Magnetic Properties. Variable-temperature magnetic susceptibility data for compounds 1 and 2 were measured in the temperature range of 1.8-300 K in a direct current (dc) field of 2 kOe. The χ M T vs T plots of 1 and 2 are shown in Fig. 2 and Supplementary Fig. S3. At 300 K, the χ M T values of 3.35 cm 3 K mol −1 (1) and 3.45 cm 3 K mol −1 (2) are much higher than the spin-only value of 2.25 cm 3 K mol −1 for one isolated W V center (S = 1/2) and one isolated high-spin Co II center (S = 3/2) with g = 2.00, indicating the significant orbital contribution of high-spin Co II in an octahedral configuration. Upon cooling, χ M T increases continuously to a maximum value of 73.17 cm 3 mol −1 K (1) and 73.14 cm 3 mol −1 K (2) at 14 K, before dropping quickly at lower temperatures. The increase in χ M T with decreasing temperature indicates the intrachain ferromagnetic coupling between W V and Co II centers, which has been reported in other Co II -W V systems [43][44][45][46][47] . The further sharp decrease below 14 K is attributed to a saturation of the χ M value and/or zero-field splitting (ZFS) effect, which urged us to explore the possibility of magnetic blocking. The field-cooled (FC) and zero-field-cooled (ZFC) magnetization was thus measured at H dc = 10 Oe as shown in Supplementary Figs S4 and S5. The ZFC and FC curves diverge at 10.2 and 8.4 K for compounds 1 and 2, which define the blocking temperatures of these chain compounds.
To verify the dynamics of the magnetization relaxation, ac magnetic measurements were performed on polycrystalline samples of 1 and 2 under a zero dc field with H ac = 1 Oe in the frequency range of 1-1500 Hz (Fig. 3a, Supplementary Figs S6a and S7). Below 12 K, the obvious frequency dependence of both in-phase (χ M ′) and out-of-phase (χ M ′ ′) ac susceptibility was observed as typically observed in SCMs. The Mydosh parameter ϕ = (ΔT/T)/Δ(log f p ) (where f p is the frequency at which a maximum appears in the χ M ′ ′ (f) plot 12,48,49 , was estimated to be ca 0.13, which falls into the category of superparamagnets (either SMMs or SCMs) and considerably greater than that for a spin glass 50,51 . Notably, both the peaks of the χ M ′ and χ M ′ ′ signals are rather broad, indicating the existence of at least two possible relaxation processes. Futhermore, the frequency dependent ac data also show pronounced temperature dependence, from which the semicircular Cole-Cole plots (χ M ′ ′ vs. χ M ′) were obtained ( Fig. 3b and Supplementary Fig. S6b). A generalized Debye mode 52 1 and 2, respectively. These values indicated a relatively wide distribution of relaxation times, which might be caused by the variable distributions in chain length and/or interchain magnetic interactions and/or random defects as found in other cases where α ranges from 0 to 0.7 53 .
For an Ising-like or anisotropic Heisenberg one-dimensional system, the χ M T value in zero applied field is directly proportional to the correlation length ξ. Therefore, the χ M T increases exponentially with lowering temperature, following the equation: 18,20,54 , where C eff is the effective Curie constant, and Δ ξ gives an estimation of the intrachain exchange energy cost to create a domain wall along the chain. Thus, provided the 1D nature of compounds 1 and 2 and the presence of significant anisotropy, plots of ln (χ M ′ T) versus 1/T of both compounds should display a linear region, which is actually observed experimentally. As can be seen in the inset of Fig. 2 and Supplementary Fig. S3, the ln(χ M ′ T) versus T −1 plots measured at H dc = 0 Oe and H ac = 1 Oe at a frequency of 1 Hz feature a linear region in the temperature range of 11.6-30 K for both 1 and 2, yielding Δ ξ /k B = 78(0) K and C eff = 0.83(1) cm 3 K mol −1 for 1, and Δ ξ /k B = 78 (0) K and C eff = 0.88 (1) cm 3 K mol −1 for 2, respectively. Below 11.6 K, ln (χ M ′ T) reaches a maximum and then undergoes a linear decrease with decreasing temperature, which indicates that the correlation length becomes larger than the average distance between two intrinsic defects along the chain, as the interchain interactions are not likely for these well isolated 1D chain 19,55 . From the activation energy Δ ξ , the exchange interaction J in the Ising limit can be estimated through the expression Δ ξ = 4|J|S eff 2,54-56 . Assuming the system at low temperature to be an Ising chain with the effective spin approach S Co, eff = 1/2 and S W = 1/2, the intrachain interaction is calculated as J/k B = 19.5 K for both 1 and 2. As the Δ ξ depends not only on the exchange interaction (J) but also the single-ion anisotropy, the J value calculated from the Ising model is a rough estimation.
The Glauber dynamics predicts that the energy barrier for the spin reversal of an ideal Ising chain should be Δ τ = 2Δ ξ for infinite chains and Δ τ = Δ ξ for finite-size chains, where growth of the correlation length is limited by naturally occurring defects 57,58 . However, for most real SCMs, an anisotropic Heisenberg chain model is more suitable, which takes into account the magnetic anisotropy energy of each magnetic unit 59 . As the correlation length ξ increases exponentially with a decrease in temperature, the overall energy barrier Δ τ is Δ τ = 2Δ ξ + Δ A for the infinite chain at high temperature and Δ τ = Δ ξ + Δ A for the finite-size chain at low temperature, where Δ A represents the intrinsic anisotropic barrier for the individual spin in the absence of magnetic exchange. This is actually the case for 1 and 2. From the Arrhenius plots of relaxation times τ obtained from the frequency-dependent χ M ′ ′ peaks, two thermally activated regions were obviously observed above and below the crossover temperatures of T* = 10.4 and 10.8 K for 1 and 2, respectively. These two regions are corresponding to the infinite-size and finite-size regimes of relaxation, commonly encountered for SCMs. The energy barriers for both of these thermally activated processes were estimated using the Arrhenius law τ = τ 0 exp (Δ τ /k B T), giving the Δ τ1 (τ 01 ) and Δ τ2 (τ 01 ) as 252(9) K (1.5(2) × 10 −13 s) and 169(2) K (4.6(1) × 10 −10 s) for 1 and 224(7) K (6.3(2) × 10 −13 s) and 154(1) K (4.0(1) × 10 −10 s) for 2, respectively (Fig. 4). To ensure the integrity of data, all the peak values of compound 1 (Fig. 4a) were read from the data of ac susceptibility (Fig. 3a). However, as the peak of 11.4 K is too broad to read accurately, the first spot was excluded in the fitting process. Remarkably, the barriers of 252(9) and 224(7) K observed for 1 and 2 are among the highest for single-chain magnets. The radical bridged chain compounds, [Co(hfac) 2  , were reported to display a relaxation barrier of (396 ± 13) K and (398 ± 14) K, respectively. As such, to the best of our knowledge, the energy barriers of compounds 1 and 2 set the new records for cyano-bridged SCMs and are preceded only by these two Co II -radical compounds among all SCMs. Obviously, for 1 and 2 in both the infinite-size and the finite-size regimes, Δ τ1 is larger than 2Δ ξ and Δ τ2 is larger than Δ ξ . From these values, the intrinsic anisotropic barrier Δ A can be estimated either by Δ A = Δ τ1 − 2Δ ξ or Δ A = Δ τ2 − Δ ξ . The obtained values are 96 or 91 K for 1, and 68 or 76 K for 2, respectively. Furthermore, the slow magnetic relaxation of both 1 and 2 were confirmed by the observation of the magnetization hysteresis loops. The loops were firstly measured on the polycrystalline samples at different temperatures and depicted in Fig. 5a and Supplementary Fig. S8a. At 1.8 K, the largest magnetization values at 70 kOe, 2.76 Nμ B (1) and 2.75 Nμ B (2), are significantly lower than the estimated saturation value of 3.2 Nμ B for the effective spin approach S Co,eff = 1/2, g Co = 13/3 60,61 , and g W = 2, S W = 1/2 indicating the significant magnetic anisotropy 34,49,54 . Obvious hysteresis loops can be observed below 5 K, with the coercive field of 11.5 kOe for 1 and 9.9 kOe for 2 at 1.8 K. Fortunately, although the single crystals of 1 and 2 are not big enough for the magnetic measurement on one single crystal, these strip-like single crystals grow parallel and form a bundle of about 1.0 × 1.5 × 6.5 mm ( Supplementary Fig. S9), which is suitable for the anisotropic measurement. By the face index of the crystal (Fig. 6), we can see that the long edge of the crystal strip (and thus the long side of the crystal bundle) is along the a axis, which is parallel to the direction of the 3, 3-ladder. Thus, the anisotropic magnetization measurements were performed with the dc field applied parallel or perpendicular to the direction of the 3, 3-ladder. As can be seen from the loops at 1.8 K (Fig. 5b and Supplementary Fig. S8b), strong magnetic anisotropy of both compounds is reflected incisively and vividly. When the field is applied along the chain direction, the saturated magnetization value is much lower and exhibits a very large hysteresis loop with a coercive field of H c = 26.2 kOe for 1 (22.6 kOe for 2) and a remnant magnetization of M r = 1.28 Nμ B for 1 (1.27 Nμ B for 2) at 1.8 K. With the field being perpendicular to the chain direction however, the compounds are quite easily magnetized. At 1.8 K, M reaches 3.4 Nμ B at 70 kOe (3.9 Nμ B for 2) with H c = 12.5 kOe for 1 (7.9 kOe for 2) and M r = 3.1 Nμ B for 1 (3.6 Nμ B for 2). This behavior is consistent with the ferromagnetic coupling between Co II and W V centers 43,[60][61][62] . The observation of the loops clearly shows the substantial slow magnetic relaxation, indicative of a "magnetic memory" of both compounds 1 and 2.
In summary, two unique cryano-bridged 3, 3-ladder were synthesized from the anisotropic Co II center, the octacyanotungsten, and two bulky counter cations. Because of the intrachain ferromagnetic interaction, strong magnetic anisotropy, as well as the well isolation of individual chains, these 1D compounds behave as single-chain magnets with record high spin reversal barriers (252 (9) and 224(7) K) for the cyano-bridged SCMs. Furthermore, hysteresis loops can be observed for both compounds below 5 K with large coercivities of 26.2 and 22.6 kOe at 1.8 K. Efforts to extend this synthetic strategy towards other SCMs using the 4d/5d cyanometallates and other anisotropic metal centers with higher energy barriers and blocking temperatures are underway.  Methods Materials. All of the reagents and solvents were purchased from commercial sources and used as received.
Caution! Although no problems were encountered in the preparation of the following complexes, in the acidic reaction conditions, suitable precautions should be taken when handling potentially poisonous compounds. It is of the utmost importance that all preparations should be performed and stored in well-ventilated areas.  X-ray Crystallography. Crystallographic data of compounds 1 and 2 were collected on Photon100 CMOS detector and Bruker Smart CCD area-detector diffractometer with Mo-Kα radiation (λ = 0.71073 Å) by using an ω scan mode at 123 K and 296 K, respectively. The diffraction data were treated using SAINT 65 , and all absorption corrections were applied by using SADABS 66 . All non-hydrogen atoms were located by Patterson method 67 using the SHELXS programs of the SHELXTL package and subsequent difference Fourier syntheses. Hydrogen bonded to carbon were determined theoretically and refined with isotropic thermal parameters riding on their parents. H-atoms of water and methanol were first located by difference Fourier E-maps and then treated isotropically as riding. All non-hydrogen atoms were refined by full-matrix least-squares on F 2 . All calculations were performed by SHELXTL-97 68,69 . Measurements. The IR spectra were carried out with a Nexus 870 FT-IR spectrometer using KBr pellets in the range of 400-4000 cm −1 . Elemental analyses of C, H, N were recorded on a PerkinElmer 240 C elemental analyzer. The temperature-dependent magnetic susceptibility, the zero-field cooled magnetization and field-cooled magnetization, the anisotropy measurement at 1.8 K and AC magnetic susceptibility were measured using a Quantum Design MPMS-XL7 superconducting quantum interference device (SQUID) magnetometer. Magnetization hysteresis loop at 1.8-10 K from 7 to − 7 T and back were measured using an eicosane-constrained sample to prevent sample torquing on a Quantum Design VSM SQUID magnetometer. Experimental susceptibilities were corrected for the diamagnetism estimated Pascal′ s tables and for sample holder by previous calibration.