Observation of the single-ion magnet behavior of d8 ions on two-coordinate Co(i)–NHC complexes

The first single-ion magnet (SIM) [Co(IMes)2][BPh4] (IMes: 1,3-dimesitylimidazol-2-ylidene) with d8 electronic configuration has been found in two-coordinate Co(i)–NHC complexes.


Introduction
The great interest in low-coordinate 3d metal complexes has been fuelled not only by their inherent synthetic challenge, as this type of species can be prone to disproportionation and coordination with exogenous ligands, but also by their useful reactivity in mediating small molecule activation and catalysis, 1 and more recently, by their unique magnetic properties that point to the potential of low-coordinate 3d metal complexes as single-ion magnets. 2 Low-coordinate 3d complexes can feature relatively weak and highly anisotropic ligand-elds, within which a metal ion could exhibit large magnetic moments and even slow magnetization relaxation behavior.  2 ] À and [dmp 2 Nin(Co(N(SiMe 3 ) 2 (OEt 2 ))) 2 ] + (dmp 2 Nin ¼ bis(2,6-dimethylphenyl)nindigo radical); 5b as well as the d 9 nickel(I) complex [Ni(6-Mes) 2 ]Br (6-Mes denotes 1,3-dimesityl-3,4,5,6-tetrahydropyrimidin-2-ylide). 8 In addition to these, it is expected that nickel(II) complexes could be very promising candidates for d 8 single-ion magnets as some nick-coordinate d 8 complexes with distinct magnetic properties, among which the slow magnetic relaxation typical for single-ion magnets has been observed for a d 8 transition-metal complex for the rst time. Moreover, theoretical studies disclosed that the dihedral angle between the two NHC planes and the degree of unsaturation of the NHC ligands could dramatically affect the magnetic properties of the low-coordinate cobalt(I)-NHC complexes.

Experimental
All synthetic experiments were performed under an atmosphere of dry dinitrogen with the rigid exclusion of air and moisture using a standard Schlenk line, or a glovebox. All organic solvents were freshly distilled from sodium benzophenone ketyl immediately prior to use. sIMes, 10a IAd, 10b Co(PPh 3 ) 3 Cl, 10c Co(IAd)(PPh 3 )Cl, 10d and [(IMes) 2 Co][BPh 4 ] 10e were prepared according to reported procedures. All other chemicals were purchased from either Strem or J&K Chemical Co. and were used as received unless otherwise noted. 1 H NMR spectra were recorded on a VARIAN 400 MHz or Agilent 400 MHz spectrometer. The chemical shis were reported in units of ppm, referenced to the residual protons of the deuterated solvents for the proton chemical shis. The elemental analysis was performed by the Analytical Laboratory of the Shanghai Institute of Organic Chemistry (CAS). The absorption spectra were recorded with a Shimadzu UV-3600 UV-vis-NIR spectrophotometer.

Preparation of Co(sIMes) 2 Cl
To a THF (15 mL) solution of sIMes (0.611 g, 2.0 mmol), Co(PPh 3 ) 3 Cl (0.881 g, 1.0 mmol) was slowly added at room temperature, during which time the color of the solution turned from pale yellow to deep reddish brown. Aer being stirred overnight and removal of the solvent, the residue was washed with n-hexane (5 mL Â 3) and then extracted with benzene (5 mL). Aer ltration, the benzene solution was recrystallized by vapor diffusion of n-hexane into the solution to afford Co(sIMes) 2 Cl as a red crystalline solid (565 mg, 80%). 1  To a THF (15 mL) solution of Co(sIMes) 2 Cl (0.703 g, 1.0 mmol), Na[BPh 4 ] (0.342 g, 1.0 mmol) was slowly added, during which time the color of the solution turned from deep reddish brown to orange. Aer being stirred for 12 h and removal of the solvent, the residue was washed with ether (5 mL Â 3) and extracted with THF (5 mL). The yellow extraction was ltrated and a small portion of toluene (1 mL) was added. The slow evaporation of THF afforded 2 as an orange crystalline solid (0.691 g, 70%). Absorption spectrum (THF): l max (3) ¼ 412 (2580) nm. The 1 H NMR spectrum of this paramagnetic complex displayed seven characteristic peaks in the range À22.65 to 66.86 ppm. 1  ] (Ar F ¼ 3,5-(CF 3 ) 2 -C 6 H 3 , 0.886 g, 1.0 mmol) was added slowly, during which time the color of the solution turned from reddish brown to brown. The mixture was further stirred at room temperature for 12 h. Aer the removal of the solvent, the residue was washed with n-hexane (5 mL Â 3) and extracted with THF (5 mL). The yellow green extraction was ltrated and a small portion of toluene (1 mL) was added. The slow evaporation of THF afforded 3 as a green crystalline solid (0.795 g, 50%). Absorption spectrum (THF): l max (3) ¼ 315 (1920), 368 (2810) nm. The 1 H NMR spectrum of this paramagnetic complex displayed seven characteristic peaks in the range À22.65 to 66.86 ppm. 1

Magnetic measurements
All the samples were xed by eicosane and paralm to avoid movement during the measurements and were sealed in a glass tube to avoid reaction with moisture and oxygen. Direct current susceptibility and alternative current susceptibility measurements with frequencies ranging from 1 to 997 Hz were performed on a Quantum Design MPMS XL-5 SQUID magnetometer for the polycrystalline samples. All the dc susceptibilities were corrected for the diamagnetic contributions from the sample holder, eicosane and from the molecule, using Pascal's constants. 11

Computational details
Complete active space second-order perturbation theory (CASPT2), considering the effect of the dynamical electronic correlation based on a complete-active-space self-consistent eld (CASSCF) approach, was performed on the cations [Co(IPh) 2 ] 1+ , [Co(sIPh) 2 ] 1+ , and [Co(IAd) 2 ] 1+ to obtain the parameters D and E using a MOLCAS 7.8 program package. 12 The structures of the cations [Co(IPh) 2 ] 1+ and [Co(sIPh) 2 ] 1+ were built based on the structures of the cations in the crystal structures of 1 and 2, respectively, using hydrogen atoms as replacements for all the methyl groups. The structure of the cation [Co(IAd) 2 ] 1+ in the crystal structure of 3 was directly used for calculation without further modication. The basis sets used for all the atoms were atomic natural orbitals from the MOLCAS ANO-RCC library: ANO-RCC-VTZP for the magnetic center ion Co I ; VTZ for close C; and VDZ for distant atoms. These calculations employed the second order Douglas-Kroll-Hess Hamiltonian, where scalar relativistic contractions were taken into account in the basis set. Aer the rst CASSCF calculation, the effect of the dynamical electronic correlation was applied using CASPT2. And then, the spin-orbit coupling was handled separately in the restricted active space state interaction (RASSI-SO) procedure. The active electrons in 10 active spaces include eight 3d electrons, and there are 25 mixed spin-free states (from 10 triplets and from 15 singlets). The coordinates of the structures can be found in the ESI † as an xyz le.  (b) in Scheme 2). 13 These low-coordinate cobalt(I) complexes are stable at room temperature both in the solid state and in solution under a nitrogen atmosphere. The 1 H NMR spectrum of 2 measured in THF-d 8 exhibits four broad peaks corresponding to the resonances of the metal-bound NHC ligands at 66.86, À12.35, À21.19, and À22.65 ppm, and that of 3 shows ve broad peaks at 50.13, 4.69, À6.34, À15.92, and À79.52 ppm. The peak patterns indicate an idealized C 2 symmetry for the cations in solution and free rotation of the adamantyl groups around the N-C bonds. The absorption spectrum of 2 displays one strong charge-transfer band at 412 nm, which is consistent with the charge-transfer band at 413 nm observed in the spectrum of 1, 10e whereas that of 3 appears at 368 nm. In addition to these strong bands, weak absorptions at around 600 nm with an absorption coefficient of ca. 200 mol À1 L cm À1 were noticeable for both complexes (2 and 3), which might correspond to the ligand-eld transitions of the two-coordinate d 8

ions. 1e
Single-crystal X-ray diffraction studies have established the structures of 1-3 as two-coordinate cobalt(I) complexes (Fig. 1). 11 Table 1 summarizes their key structural parameters. In the structures of 1-3, even the shortest Co/Co separations are all longer than 9Å, and no hydrogen-bonding or arene-arene p-interactions are present. Therefore, the intermolecular dipole-dipole interactions, if they exist, could be very small. Similar to 1, 10e the C(carbene)-Co-C(carbene) alignments in 2 and 3 are also linear (178.4(1) and 180 , respectively). The Co-C(carbene) distance in 2 (1.936(2)Å) is identical to that of 1, and is slightly shorter than that of 3 (1.943(3)Å). As compared to their counterpart in the cAAC complex [Co(Et 2 -cAAC) 2 ][BAr F 4 ] (Et 2 -cAAC: 1-(2 0 ,6 0 -diisopropylphenyl)-3,3-diethyl-5,5-dimethylpyrrolidine-2-ylidene) (1.957(2)Å), 14 the Co-C(carbene) bonds in 1-3 are shorter. One of the apparent structural differences among 1-3 is the dihedral angle between the ve-membered NHC planes. The presence of the IAd ligands in 3, which are the most sterically demanding among the three NHC ligands, 15 rendered vertical alignment of the two planes, whereas, smaller dihedral   angles of 39.55 and 35.02 are observed for the structures of 1 and 2, respectively. Another important structural difference is the weak interaction of the cobalt centers with the N-bonded substituents. Apparent secondary metal-NHC interactions between the cobalt center and the adamantyl groups with a shortest Co/C distance of 2.878Å are evidenced in 3. As for the structures of 1 and 2, the shortest Co/C distances involving the ortho methyl groups of the anking mesityls and the cobalt center are 3.838 and 3.651Å, respectively, approaching the sum of the van der Waals radii of Co with C (3.7Å). 16 Some of the two-coordinate iron(I), iron(II), and nickel(II) metal complexes featuring high uniaxial symmetry maintain angular momentum, leading to a larger zero eld splitting. 1c,4-9 In order to examine whether this is the case for the two-coordinate d 8 cobalt(I) complexes or not, we performed static magnetic experiments on the solid samples of 1-3. Magnetic susceptibility measurements conrmed a high magnetic momentum of 3.65 cm 3 mol À1 K (m eff ¼ 5.40 m B ) and 3.26 cm 3 mol À1 K (m eff ¼ 5.10 m B ) at room temperature for 1 and 2, respectively. These values are much higher than the spin-only value of 1 cm 3 mol À1 K for Co(I) of S ¼ 1, implying the contribution of unquenched orbital angular momentum. The susceptibility changes of the two complexes at a low temperature range, however, are different. As depicted in Fig. 2, the c m T value of 1 decreases slowly from 3.65 cm 3 mol À1 K to 2.93 cm 3 mol À1 K when cooling from 300 K to 2 K. For sample 2, a sharp decrease of the c m T value from 1.94 cm 3 mol À1 K to 1.08 cm 3 mol À1 K was observed as the temperature changed from 8 K to 2 K. Compared with the previously reported twocoordinate Co(II) and Co(I) compounds, 17 the room temperature susceptibility of 1 and 2 is slightly larger, which is probably due to the unquenched angular momentum. Compared to that of 1 and 2, the magnetic momentum of 3 is much lower (Fig. 2). The c m T value of 3 at room temperature is 1.94 cm 3 mol À1 K (m eff ¼ 3.94 m B ). A sharp downturn of its c m T value occurs below 50 K, and the c m T value eventually reaches 0.97 cm 3 mol À1 K at 2 K.
The variable-eld variable-temperature magnetization experiments ( Fig. S5-S7 †) conrmed the existence of magnetic anisotropy, which is typical for low coordinate cobalt compounds. 1e,17f,h We attempted to use ANISOFIT 2.0 to quantify the zero-eld splitting parameters D and E from the variable-eld variable-temperature data of 1-3. However, no reasonable ts for the effective spin Hamiltonian: H ¼ DS z 2 + E(S x 2 À S y 2 ) + gm b SB were obtained. This is probably due to the remarkable rst-order orbital angular momentum contribution of the compounds. 18 Consequently, complete active space secondorder perturbation theory (CASPT2) calculations based on the cations [Co(IPh) 2 ] 1+ , [Co(sIPh) 2 ] 1+ , and [Co(IAd) 2 ] 1+ were performed to acquire the D and E values as approximations for those of 1-3. As shown in Table S1, † the calculated energies of the spin-free states were found to be much larger than the spinorbit coupling energies, implying that the description using Russell-Saunders coupling is not necessary. On the other hand, the z component of the orbital angular moment |L z | of [Co(IPh) 2 ] 1+ is 0.179, which is much larger than that of the other two complexes, indicating a small orbital contribution ( . These calculated results imply that 1 might possess relatively stronger magnetic anisotropy compared to 2 and 3, as well as the larger orbital momentum contribution, which is also reected by their variable-temperature static magnetic susceptibilities. Impressed by the large magnetic anisotropy of 1, we further performed dynamic magnetic experiments to probe its magnetic relaxation behavior. In the absence of a dc eld, a temperature-dependent out-of-phase signal c 00 m was observed for 1 while out-of-phase signal peaks were not observed (Fig. S8 †). This can probably be attributed to the very fast magnetic tunneling and the relaxation timescale, which is beyond our instruments. This phenomenon has been found in many mononuclear transition metal compounds. 20 Under an optimized dc eld of 2000 Oe, frequency-dependent and out-ofphase peaks were then observed between 2 K and 10 K (Fig. 3  and S9 †). The frequency-dependent data can be transformed into Cole-Cole plots (Fig. S10 †) and tted using a generalized Debye model, which gives a tted distribution of relaxation time, with a in the range of 0.04 to 0.2 (Table S3 †). Plotting the relaxation time s vs. T À1 gives a distinct curve as shown in Fig. 4. Assuming Orbach process character, tting the relaxation time with the Arrhenius law: s ¼ s 0 exp(U eff /k B T) at high temperature gives a linear t with U eff ¼ 21.3 cm À1 and s 0 ¼ 6.6 Â 10 À6 s. The spin-reversal energy barrier falls in the typical range of the reported barriers of cobalt(II) single-ion magnets, and is also comparable to the reversal barriers of the reported two-coordinate d 7 iron(I) and d 9 nickel(I) complexes with carbene ligation.  features a U eff value of 11.8 cm À1 under a dc eld of 600 Oe. 8 As compared to the two-coordinate iron(II) amide complexes, 4 the U eff values are relatively small. We noted that a crossover occurred around 5 K, which means that Orbach relaxation behaviour can not describe the whole process. And the preexponential factor s 0 is larger than the usual value of 10 À8 s for a typical Orbach process. So other relaxation mechanisms such as Raman processes must also be present. Using the exponential law can give a good t above 4 K, giving n ¼ 4.4(1). These results suggest that an optical acoustic Raman process would be more reasonable. This behavior has also been observed in other mononuclear transition metal single molecule magnets (SMMs) and the still existing curvation might be due to the presence of other processes such as not fully quenched quantum tunneling magnetism (QTM). 4a,21 As mentioned earlier, some nickel(II) complexes have shown large magnetic anisotropies, 9h-i but so far no precedent d 8 complex showing slow magnetic relaxation is known. Among all the reported 3d single-ion magnets, the magnetic relaxation usually occurs in Kramers systems or non-Kramers systems with a negative D value. 2 Complex 1 represents the rst example of a mononuclear d 8 complex with a relatively large positive D value which shows single-ion magnet behavior. Considering the rough structural similarity of 1 with 2 and 3, we speculated that the other two complexes would also exhibit magnetic relaxation behavior, as 2 and 3 possess easy-axis anisotropy. However, further investigations revealed that neither of them show such a property even under applied dc elds (Fig. S11-S14 †).
The different magnetic properties observed for 1-3 raised the question of the origin of the structural distinctions, which, we think, should lie with the specicity of the NHC-metal interactions. The key structural differences around the two-coordinate metal centers in 1-3 are the dihedral angles between the NHC planes, the degree of unsaturation of the NHC ligands (imidazole-2-ylidene in 1 and 3 versus imidazolin-2-ylidene in 2), and the secondary metal-ligand interactions. The difference in the dihedral angles (39.55, 35.02 and 90 for 1-3 respectively) might induce a difference in the Co-C(carbene) p-interactions that could quench orbital angular momentum. 7 To investigate the inuence of the dihedral angle on the magnetic anisotropy, we calculated the D and E values for a series of structures of [Co(IPh) 2 ] 1+ with the dihedral angle a varying from 30 to 90 while keeping the other structural parameters unchanged. As shown in Table 2, the anisotropy parameters show evident dependence on the dihedral angle. When a increases from 40 to 50 , the D value abruptly decreases from 33.4 cm À1 to nearly zero. Upon further increasing the dihedral angle from 50 to 90 , the calculated D values do not show a signicant change.
These results indicate that the large dihedral angle (90 ) of 3 would be one of the causes for the quenched spin-orbit coupling that leads to the small magnetic anisotropy.
The main structural difference of 1 and 2 comes from the degree of unsaturation of the ve-membered NHC rings which might induce different d-p interactions and change the "genuine" molecular symmetry. The importance of the molecular symmetry for the magnetism of transition metal and lanthanide SIMs or SMMs is well-documented. 4, 22 To further clarify the inuence of this factor, we modied the structure of [Co(sIPh) 2 ] 1+ by changing the CH 2 CH 2 backbone into CHCH, keeping the C-C distance unchanged (see atomic coordinates in    [Co(sIPh 0 ) 2 ] 1+ . Furthermore, the reduction of symmetry would induce more transverse interactions as more transverse components occurred (see Table S2, † the non-zero value of L yz ).
These results indicate that the degree of unsaturation of the ve-membered NHC rings can also inuence the magnetic anisotropy of the two-coordinate cobalt(I)-NHC complexes, which should account for the observed difference of the magnetic properties between 1 and 2.
In addition to these factors, the difference in the secondary metal-ligand interactions in 1-3 also caught our attention. The ease with which a NHC complex will incur secondary metalligand interactions should increase with the increasing bulkiness of the NHC ligands (IMes < sIMes < IAd). 15 The short contact distances observed in 3 and the cyclometallation reactions of the cobalt-IMes complexes 10e have indicated the feasibility of these two-coordinate cobalt complexes to form secondary metalligand interactions within the molecule. Recently, Neese and coworkers have predicted that secondary metal-ligand interactions could cause vibronic coupling and decrease the magnetic anisotropy and relaxation time of single-molecule magnets. 4b However, a quantitative account of these metal-ligand interactions needs to include more atoms and orbitals, which is beyond what we can currently handle. Qualitatively, we could infer that, among the three complexes, 1 could be the one incurring the weakest secondary metal-ligand interactions.

Conclusions
In summary, we have found that the high-spin two-coordinate cobalt(I)-NHC complex [Co(IMes) 2 ][BPh 4 ] exhibits a large room temperature magnetic moment and slow magnetic relaxation behavior under an applied dc eld, which represents the rst d 8 single-ion magnet. In contrast to the IMes complex, the analog two-coordinate cobalt(I) complexes with different NHC ligands, [Co(sIMes) 2 ][BPh 4 ] and [Co(IAd) 2 ][BAr F 4 ], did not show such single-ion magnet behavior. A comparison of the molecular structures of these three complexes revealed their similar linear C(carbene)-Co-C(carbene) cores, with different dihedral angles between the NHC planes, a different degree of saturation of the NHC ligands and different separation distances of the cobalt center toward its nearest carbon atoms on the N-bonded substituents. Ab initio investigations on the two former structural factors show that they can largely alter the value of D, which predominates the changing of the Co-C(carbene) p-interactions and the changing of further spin-orbit coupling splitting. The nal structural difference reects the ease with which the cations incur secondary metal-ligand interactions increasing in the order of [Co(IMes) 2 ] + < [Co(sIMes) 2 ] + < [Co(IAd) 2 ] + , which may also be an important factor affecting the magnetic properties of these low-coordinate complexes, probably acting in a direct way, different from vibronic coupling.