Semiquinone radical-bridged M2 (M = Fe, Co, Ni) complexes with strong magnetic exchange giving rise to slow magnetic relaxation

The use of radical bridging ligands to facilitate strong magnetic exchange between paramagnetic metal centers represents a key step toward the realization of single-molecule magnets with high operating temperatures. Moreover, bridging ligands that allow the incorporation of high-anisotropy metal ions are particularly advantageous. Toward these ends, we report the synthesis and detailed characterization of the dinuclear hydroquinone-bridged complexes [(Me6tren)2MII2(C6H4O22−)]2+ (Me6tren = tris(2-dimethylaminoethyl)amine; M = Fe, Co, Ni) and their one-electron-oxidized, semiquinone-bridged analogues [(Me6tren)2MII2(C6H4O2−˙)]3+. Single-crystal X-ray diffraction shows that the Me6tren ligand restrains the metal centers in a trigonal bipyramidal geometry, and coordination of the bridging hydro- or semiquinone ligand results in a parallel alignment of the three-fold axes. We quantify the p-benzosemiquinone–transition metal magnetic exchange coupling for the first time and find that the nickel(ii) complex exhibits a substantial J < −600 cm−1, resulting in a well-isolated S = 3/2 ground state even as high as 300 K. The iron and cobalt complexes feature metal–semiquinone exchange constants of J = −144(1) and −252(2) cm−1, respectively, which are substantially larger in magnitude than those reported for related bis(bidentate) semiquinoid complexes. Finally, the semiquinone-bridged cobalt and nickel complexes exhibit field-induced slow magnetic relaxation, with relaxation barriers of Ueff = 22 and 46 cm−1, respectively. Remarkably, the Orbach relaxation observed for the Ni complex is in stark contrast to the fast processes that dominate relaxation in related mononuclear NiII complexes, thus demonstrating that strong magnetic coupling can engender slow magnetic relaxation.


Introduction
Semiquinone is a ubiquitous free radical in living systems, acting as a radical scavenger 1 and as a facilitator of electrontransfer processes in photosynthesis 2 and aerobic respiration. 3 In these systems, the p-quinone molecule serves as an electron reservoir, shuttling between the three consecutive oxidation states of quinone, semiquinone, and hydroquinone (Scheme 1, top). Although the free semiquinone radical cannot be isolated due to its short lifetime, it can be stabilized by coordination to one or more metal centers. Nevertheless, despite the abundance of quinone in living systems, to our knowledge, there are only six studies reporting crystal structures of p-semiquinone stabilized by transition metal or lanthanide centers. [4][5][6][7][8][9] Of these examples, only one features the p-semiquinone radical anion (p-C 6 H 4 O 2 c À ), as isolated in a structure consisting of two p-stacked radicals coordinated to two Mn II (cyclam) units. 8 In addition to their importance in biological processes, organic radicals have generally garnered much attention as ligands in the design of single-molecule magnets. 10 In particular, radicals can engage in strong, direct magnetic exchange coupling with metal-based spins, giving rise to energetically well-isolated spin ground states that are necessary to realize slow magnetic relaxation at high temperatures. Semiquinonebased radicals are particularly well-suited toward this end, owing to their negative charge, which enables stronger interactions with metal ions relative to neutral radicals, and their chemical versatility. For instance, 2,5-disubstituted semiquinone derivatives, which can act as bis(bidentate) bridging ligands (Scheme 1, bottom), 11,12 have been employed to synthesize radical-bridged dinuclear complexes exhibiting strong magnetic exchange coupling [13][14][15][16] and in some cases single-molecule magnet behavior. [17][18][19][20] In contrast, magnetic coupling involving the bis(monodentate) benzosemiquinone molecule (Scheme 1, top) has not been reported.
The use of bis(monodentate) semiquinone as a bridging ligand for magnetic molecules offers several key potential advantages over the more common bis(bidentate) analogues. 11 First, the presence of only one donor atom per metal ion reduces the overall number of atoms over which the unpaired electron is delocalized, thereby concentrating spin density and promoting stronger exchange. Second, having donor atoms on only two C atoms of the quinone ring leaves the additional four C atoms available for chemical modication. Finally, the occupation of only one metal coordination site by the bridging ligand allows access to metal ions in different coordination geometries. Of particular interest in this regard are trigonal bipyramidal complexes of Ni 2+ , which have been shown to exhibit exceptionally large axial zero-eld splitting (D) owing to the presence of nearly degenerate d x 2 Ày 2 and d xy orbitals. 21,22 For instance, the compound [(Me 6 tren)NiCl](ClO 4 ) (Me 6 tren ¼ tris(2-dimethylaminoethyl)amine) has been shown by high-eld, high-frequency electron paramagnetic resonance spectroscopy to exhibit À180 # D # À120 cm À1 . 22 Despite the potential of [(Me 6 tren)Ni] 2+ as a building unit for singlemolecule magnets, to date no multinuclear molecule featuring this unit has been reported. Moreover, the only example of any complex incorporating multiple trigonal bipyramidal Ni 2+ centers is a dinuclear imidazole-bridged [Ni(m-Im)NiL](ClO 4 ) complex, which features weak antiferromagnetic coupling between Ni 2+ centers. 23 This conspicuous dearth of compounds likely stems from synthetic difficulties associated with the tendency of Ni 2+ to adopt octahedral coordination.
Herein, we report the synthesis and characterization of the semiquinone-bridged complexes [(Me 6 tren) 2 M II 2 (C 6 H 4 O 2 )] 3+ (M ¼ Fe, Co, Ni), obtained via one-electron oxidation of their hydroquinone-bridged precursors. These complexes are the rst to feature paramagnetic, trigonal bipyramidal metal centers bridged through a radical ligand and the rst examples of any structurally characterized dinuclear complexes bridged through a single semiquinone. Strong, direct magnetic exchange mediated by the semiquinone radical results in well-isolated paramagnetic ground states for all complexes, up to 300 K in the case of Ni. Finally, signicant magnetic anisotropy engendered by the trigonal bipyramidal ligand eld of the metal centers leads to slow magnetic relaxation under small applied elds, with moderate relaxation barriers for the Co and Ni complexes.

Synthesis and structures
The semiquinone-bridged complexes were synthesized in three steps starting from the mononuclear precursors [(Me 6 tren)MX] X (X ¼ Cl À or Br À  (3), Scheme 2) in moderate to high yield. Note that the salt metathesis reaction is necessary to generate a THF-soluble precursor, as it was found that coordinating bis(monodentate) semiquinone is displaced by strong donor solvents such as acetonitrile.
Layering Et 2 O onto THF solutions of 1-3 at room temperature afforded large block-shaped crystals of yellow 1$2THF$Et 2 O, green 2$2THF$Et 2 O, and dark brown 3$2THF$Et 2 O, respectively, which were suitable for singlecrystal X-ray diffraction analysis. All three compounds crystallize in the space group P 1. Each structure features two metal centers, both in a distorted trigonal bipyramidal geometry with an axial site ligated by an O atom from a bridging dianionic hydroquinone ( Fig. S1-S3 †). The two metal centers are related through a center of crystallographic inversion symmetry, with M-O-C angles of 124.91(15) , 131.33 (12) , and 132.08 (8) for Fe, Co, and Ni, respectively ( Table 1). Note that 3$2THF$Et 2 O is only the second example of a multinuclear trigonal bipyramidal Ni II complex, 23 and it is the rst example of any molecule incorporating multiple [(Me 6 tren)Ni] 2+ units.
In the equatorial plane of each metal center, the average M-N eq bond length decreases by $0.05Å upon moving from Fe to Scheme (6)) in high yield. Surprisingly, these complexes are remarkably airstable in the solid-state. As monitored by powder X-ray diffraction analysis, 4 shows only slight decomposition aer six months, while 5 and 6 maintain crystallinity for an indenite period of time ( Fig. S4-S6 †). This kinetic stability may be a result of steric protection of the bridging ligand by the [B(C 6 F 5 ) 4 ] À counteranions as THF solutions of 4-6 readily decomposed in air.
Layering Et 2 O onto 1,2-diuorobenzene solutions of 4, 5, and 6 and storage at À25 C afforded plate-shaped crystals of dark green 4$Et 2 O, dark brown 5$Et 2 O, and red 6$Et 2 O, which were suitable for single-crystal X-ray diffraction analysis. The three compounds crystallize in the space group P2 1 /c and are isostructural ( Fig. 1 and S7-S9 †). In each structure, the asymmetric unit contains two distinct half-molecules of the cationic complex, three [B(C 6 F 5 ) 4 ] À counterions, and a molecule of Et 2 O. The bond distances and angles associated with the two halfmolecules differ considerably in each structure, owing to disorder over two positions for one of the half-molecules. For the purpose of structural comparisons in this report, we refer to the half-molecule without disorder.
The metal centers in 4, 5, and 6 retain their trigonal bipyramidal geometry with M-O-C angles of 135.1(2) (Fe), 138.1(4) (Co), and 136.6(4) (Ni), which are slightly greater than those of the unoxidized analogues. Importantly, in contrast to [(cyclam) 2 (Table 1). A close comparison of bond distances in both series of compounds reveals important information related to the oxidation state of the bridging ligand (Table 1). Upon oxidation, the C-O distances decrease from 1.319(3), 1.343(2), and 1.3405(14)Å, for Fe, Co, and Ni, respectively, to 1.291(4), 1.288(7), and 1.296(7)Å. These changes are consistent with oxidation of the bridging ligand from hydroquinone to semiquinone, which involves a net increase in the C-O bond order. The semiquinone C-O distances in 4, 5, and 6 are also consistent with those observed in the structure of [(cyclam) 2 Mn 2 (C 6 H 4 O 2 À c) 2 ] 2+ . 8 The average C-C bond distances appear unchanged within error upon oxidation (Tables S3 and  S4  Across the two series of compounds, the Co congeners feature geometries closest to a perfect trigonal bipyramid, as reected in their s index values of 0.837 (2) and 0.820 (5), where s ¼ 1 for a perfect trigonal bipyramid and s ¼ 0 for a perfect square pyramid. 24 This approximate three-fold symmetry for Co II is consistent with a fully symmetric Table 1 Selected bond distances (Å), angles ( ), and structural index, s, obtained for compounds 1-6 at 100 K Hydroquinone-bridged complexes Semiquinone-bridged complexes a (d xz ,d yz ) 4 (d x 2 Ày 2 ,d xy ) 2 (d z 2 ) 1 electronic conguration. In contrast, the Fe II and Ni II complexes bear an odd number of electrons in the (d xz ,d yz ) or (d x 2 Ày 2,d xy ) sets and therefore experience a Jahn-Teller distortion away from three-fold symmetry, as reected in their lower s indices (Table 1).

Spectroscopy
The oxidation of hydroquinone in 1-3 to semiquinone in 4-6 is concomitant with a blue shi in the C-O stretching frequency, for instance from 1489 to 1500 cm À1 upon oxidation of 1 to 4 (Fig. 2). In the case of the Co and Ni congeners, the C-O stretching frequencies were found to overlap with the vibrational modes of [B(C 6 F 5 ) 4 ] À . We therefore carried out anion exchange reactions for 2, 3, 5, and 6 to access the corresponding PF 6 À salts, denoted as 2 0 , 3 0 , 5 0 , and 6 0 , respectively (see the ESI †). The C-O stretches of 1503 and 1506 cm À1 in 5 0 and 6 0 , respectively, are also blue-shied relative to those for their unoxidized precursors (1487 and 1483 cm À1 for 2 0 and 3 0 , respectively). The C-O vibrational frequencies for all complexes are consistent with values reported previously for hydroquinone and semiquinone coordinated to Mn II . 4,8 In addition, we carried out density functional theory (DFT) calculations on compounds 3 and 6, which closely reproduce the experimental spectra and support the assignment of the C-O stretching frequencies (Fig. S11 †). These calculations also indicate that a second blue-shied absorption band observed at $1340 cm À1 for 4, 5 0 , and 6 0 corresponds to a C-H rocking mode of semiquinone. The 57 Fe Mössbauer spectrum collected for 1 at 5 K features a symmetric Lorentzian doublet with an isomer shi of 1.0320(6) mm s À1 and a quadrupole splitting of 3.198(1) mm s À1 , consistent with a high-spin, ve-coordinate Fe II center (Fig. S14 †). Upon oxidation, the spectrum for 4 at 20 K is best t to two similar symmetric Lorentzian doublets, corresponding to the two distinct molecules in the structure, with isomer shis of 0.996(2) and 1.008(4) mm s À1 and quadrupole splittings of 3.528(8) and 3.23(1) mm s À1 (Fig. S14 †). The minimal decrease in isomer shi observed upon oxidation of 1 to 4 is consistent with a ligand-based oxidation, from hydroquinone to semiquinone, and retention of high-spin Fe II centers.

Static magnetic properties
To probe magnetic coupling in the hydroquinone-bridged compounds 1-3, dc magnetic susceptibility data were collected under an applied eld of 1 T, and corresponding plots of c M T vs. T are shown in the top portion of Fig. 3. At 300 K, c M T ¼ 7.40, 5.18, and 3.36 cm 3 K mol À1 for 1 (M ¼ Fe), 2 (M ¼ Co), and 3 (M ¼ Ni), respectively, values which are consistent with pairs of magnetically non-interacting S ¼ 2 (g ¼ 2.22), 3/2 (g ¼ 2.35), and 1 (g ¼ 2.59) centers. For each complex, the magnitude of c M T decreases with temperature to minimum values of 0.29 (1), 0.22 (2), and 0.10 cm 3 K mol À1 (3) at 2 K, respectively, which behavior is indicative of weak antiferromagnetic superexchange between the metal centers mediated through the diamagnetic hydroquinone ligand. To quantify the strength of this coupling, the susceptibility data and magnetization data were simultaneously t according to the Hamiltonian in eqn (1): Here, the rst, second, and third terms correspond to Zeeman splitting, Heisenberg-Dirac-van Vleck magnetic exchange, and zero-eld splitting, respectively, where S 1 and S 2 are the metalcentered spins, D ion is the zero-eld splitting tensor for individual metal centers, g is the g tensor of individual metal centers, and H is magnetic eld. For each complex, g and D ion for the metal centers were constrained to be equivalent based on the crystallographic inversion symmetry. These ts give exchange constants of J ¼ À0.847(2), À2.595 (5), and À5.94(1) cm À1 for 1, 2, and 3, respectively (Fig. 3, top and Table 2).
The semiquinone-bridged complexes of compounds 4 (M ¼ Fe), 5 (M ¼ Co), and 6 (M ¼ Ni) exhibit starkly different magnetic behavior. In the case of 4 and 5, the c M T curves increase steadily with decreasing temperature to reach This journal is © The Royal Society of Chemistry 2020 Chem. Sci., 2020, 11, 8196-8203 | 8199 maximum values of 9.11 and 6.04 cm 3 K mol À1 at 45 K, respectively, before undergoing subsequent downturns. In the case of 6, c M T decreases only slightly in a linear fashion with temperature from 3.15 cm 3 K mol À1 at 300 K to 3.00 cm 3 K mol À1 at 50 K, characteristic of temperature-independent paramagnetism, before dropping more precipitously below 50 K to a value of 1.68 cm 3 K mol À1 at 2 K. For each complex, this behavior is indicative of strong antiferromagnetic coupling between the semiquinone radical and the two metal centers as well as the presence of signicant zero-eld splitting. To quantify the magnetic coupling, the susceptibility and magnetization data for 4 and 5 were t using the Hamiltonian given by eqn (2): Here, S SQ ¼ 1/2 represents the spin of the semiquinone radical. As above, D ion and g for both metal centers in each complex were constrained to be the same due to the crystallographic inversion symmetry, and g SQ was constrained to be 2.00. These ts afforded exchange constants of J ¼ À144(1) and À252(2) cm À1 for 4 and 5, respectively (Fig. 3, bottom and Table  2). The c M T vs. T data for 6 could not be satisfactorily t to eqn (2). Nevertheless, simulation of the data revealed the presence of very strong antiferromagnetic coupling with an estimated upper bound of J < À600 cm À1 . In line with this result, brokensymmetry DFT calculations carried out using the solid-state structure of 6 predicted an exchange constant of J ¼ À542 cm À1 (see the ESI for details †). Given this large magnitude of J, the S ¼ 3/2 ground state in 6 should be well-isolated from excited states even at 300 K, and indeed the magnetic susceptibility data could be t using a giant spin approximation with S ¼ 3/2 (Fig. 3, bottom). A simultaneous t of the susceptibility and low-temperature magnetization data (Fig. S26 †) for S ¼ 3/2 yielded the following parameters: g k ¼ 2.781(2), g t ¼ 2.357(3), D mol ¼ À21.4(2) cm À1 , E mol ¼ 4.6(2) cm À1 , and c TIP ¼ 6.6(2) Â 10 À4 cm 3 mol À1 (Table 3). Note that 6 represents the rst multinuclear Ni complex with a thermally isolated magnetically coupled ground state at 300 K.
The magnetic data obtained for 4-6 are the rst reported for any p-semiquinone-bridged transition metal complex, and they demonstrate that the semiquinone radical can facilitate extremely strong magnetic coupling via direct exchange through the ligand-based unpaired electron. Indeed, the magnitudes of J found here are much larger than those mediated via most radical ligands, ‡ such as tetrazine (+96 cm À1 with Ni 2+ ), 28 nindigo (À138 cm À1 with Co 2+ ), 29 and phenazine (+149 cm À1 with Ni 2+ ). 30 Further, the coupling observed for 4 and 5 is considerably stronger than that previously reported for Ni <À600 a Obtained by tting or simulating dc susceptibility data for 1-5 and 6, respectively, as described in the text. b Isotropic g value as calculated from averaging the anisotropic g tensor (Table S7). c The axial component of the D ion tensor. Fig. 3 Plots of the molar magnetic susceptibility times temperature (c M T) versus T for 1-3 (top) and 4-6 (bottom) under an applied field of 1 T. Solid lines indicate fits to the data using eqn (1) for 1-3, eqn (2) for 4 and 5, and a giant spin approximation with S ¼ 3/2 for 6, as described in the text. Table 3 Zero-field splitting parameters and corresponding calculated spin reversal barriers (U calc ) compared with experimental spin reversal barriers (U eff ) extracted from dynamic magnetic data for semiquinone-bridged compounds 4 (M ¼ Fe), 5 (M ¼ Co), and 6 (M ¼ Ni)  (4) bis(bidentate) 2,5-dihydroxy-1,4-benzoquinone derivatives with Fe (J ¼ À57 to À65 cm À1 ) 20 and Co (J ¼ À52 cm À1 ). 15 Likely, this prominent difference results predominantly from the delocalization of the unpaired electron over only two O atoms in semiquinone versus four in the bis(bidentate) analogues. In addition, it is possible that the presence of fewer electronegative O atoms increases the energy of the singly occupied molecular orbital of the semiquinone, thus resulting in a better energy match with the metal orbitals. In fact, the coupling constant for 4 is comparable to that determined for the o-semiquinone radical in the mononuclear [PhTt tBu ]Fe(phenSQ) complex (J Fe-L ¼ À127 cm À1 ). 31 Nevertheless, stronger exchange has been observed for semiquinonoid ligands with N-donors (J < À900 cm À1 with Fe 2+ and À440 cm À1 with Co 2+ ), 17,19 as expected for a more diffuse, higher-energy SOMO of N compared to O. 16,32 As such, incorporation of bridging ligands based on p-phenylenediamine radical may provide a route to complexes with even stronger coupling. Below 20 K, the magnetization data for 4, 5, and 6 were t using a giant spin Hamiltonian with values of S ¼ 7/2, 5/2, and 3/2, respectively ( Fig. S24-S26 †), and the corresponding zero-eld splitting parameters are listed in Table 3. With these zero-eld splitting parameters, the program PHI 33 was used to calculate ground to rst excited state splittings (U calc ) of 12.1, 18.4, and 45.6 cm À1 for 4, 5, and 6, respectively (Table 3).

Dynamic magnetic properties
The static magnetization data for the semiquinone-bridged compounds 4-6 indicate magnetic ground states with signicant anisotropy and moderate excited state splittings, suitable for the observation of single-molecule magnet behavior. We thus set out to probe the spin dynamics of each compound via ac magnetic susceptibility measurements. Under a 1000 Oe eld, the Fe II -containing compound 4 exhibits tails in the outof-phase molar magnetic susceptibility ðc 00 M Þ at the low and high frequency limits of the magnetometer (Fig. S27 †). The slow process rapidly diminishes to zero with increasing temperature, which suggests it may be dipolar in origin, 34 and the fast process moves out of the magnetometer range with increasing temperature.
The Co II -containing compound 5 shows slow magnetic relaxation under zero applied eld that is further slowed under dc elds (Fig. S29 †). Under an optimal eld of 2000 Oe, 5 exhibits two peaks in the out-of-phase susceptibility from 1.8-3.0 K, one below 1 Hz that can be ascribed to dipolar interactions 34 and a thermally activated process at higher frequencies (Fig. S31 †). Similarly, under a 1000 Oe eld, the Ni II -containing compound 6 exhibits a single peak in c 00 M from 1.8 to 5.3 K (Fig. 4), and the data feature slight asymmetry that might arise due to the disordered second molecule in the structure (see the ESI †). The ac magnetic susceptibility data for 5 and 6 were t using a generalized Debye model 35 with two relaxation processes (see the ESI for details †). In the case of 6, the relaxation is dominated by a single major component (Table S9 and Fig. S36 †), which is the focus of further discussion below.
Temperature-dependent relaxation times were extracted for 5 and 6 and are plotted versus T in Fig. 5. The Arrhenius plot for each compound exhibits a linear region at high temperatures, which is indicative of spin relaxation through an Orbach process. The temperature-dependent relaxation data were t to eqn (3): where the rst, second, and third terms correspond to Orbach, Raman, and direct processes, respectively. In this expression, s 0 is a pre-exponential factor, U eff is the effective spin relaxation barrier, k B is the Boltzmann constant, T is temperature, C and A are constants for the Raman and direct processes, respectively, n is the Raman exponent (typically 3-9) and H is magnetic eld. 36,37 The data for 5 could be t without the Raman term, yielding U eff ¼ 22.0(1) cm À1 , s 0 ¼ 2.6(1) Â 10 À9 s, and A ¼ 2.67(5) Â 10 4 s À1 T À4 K À1 . The data for 6 were t without inclusion of the direct process term, yielding U eff ¼ 45.9(4) cm À1 , s 0 ¼ 4.0(4) Â 10 À10 s, C ¼ 0.97(3) s À1 K À3 , and n ¼ 3. Note that the suitability of this model for the relaxation dynamics, particularly at higher temperatures, is supported by the s 0 values in the expected range for Orbach relaxation 35 and the similarity between the U eff and U calc values determined from static magnetization measurements for both compounds ( Table 3). The Orbach relaxation observed for 6 is in remarkable contrast to that observed for mononuclear Ni II complexes in similar coordination environments. For example, the trigonal  22 Note that the only previously reported example of a complex with multiple trigonal bipyramidal nickel ions is an imidazole-bridged Ni 2 complex, which features a non-magnetic ground state. 23 Based on these precedents, we postulate that the strong magnetic exchange through the semiquinone bridge in 6 serves to limit fast relaxation via Raman and direct processes and thus enables observation of Orbach relaxation. Our results support a growing number of studies that demonstrate the utility of judiciously selected multinuclear single-molecule magnets featuring radical-bridged, high-anisotropy metal ions for suppressing through-barrier relaxation processes. [17][18][19][20]

Conclusions
The foregoing results show that semiquinone can be employed as a bridging ligand between metal ions to give exceptionally strong metal-radical magnetic coupling and single-molecule magnet behavior in [(Me 6 tren) 2 M 2 (C 6 H 4 O 2 À c)] 3+ (M ¼ Fe, Co, Ni). Magnetic susceptibility data obtained for these complexes reveal the presence of strong metal-semiquinone coupling, with exchange constants of J ¼ À144(1) and À252(2) cm À1 for Fe and Co, respectively, and J < À600 cm À1 for Ni. Importantly, the values of J for Fe and Co are substantially larger in magnitude than those reported for related bis(bidentate) semiquinoid complexes, demonstrating that decreasing the number of donor atoms can increase magnetic coupling strength. Additionally, the enhanced magnetic exchange achieved with semiquinone results in the rst multinuclear Ni complex with a thermally isolated ground state. The presence of eld-induced molecular slow magnetic relaxation for the Co and Ni semiquinonebridged complexes further serves to illustrate the efficacy of coupling anisotropic metal centers to engender single-molecule magnet behavior. Current efforts are underway to incorporate semiquinone bridges into molecules with heavier metal ions and into extended networks.

Conflicts of interest
There are no conicts to declare.