Strong Exchange Coupling in a Trimetallic Radical‐Bridged Cobalt(II)‐Hexaazatrinaphthylene Complex

Abstract Reducing hexaazatrinaphthylene (HAN) with potassium in the presence of 18‐c‐6 produces [{K(18‐c‐6)}HAN], which contains the S=1/2 radical [HAN].−. The [HAN].− radical can be transferred to the cobalt(II) amide [Co{N(SiMe3)2}2], forming [K(18‐c‐6)][(HAN){Co(N′′)2}3]; magnetic measurements on this compound reveal an S=4 spin system with strong cobalt–ligand antiferromagnetic exchange and J≈−290 cm−1 (−2 J formalism). In contrast, the CoII centres in the unreduced analogue [(HAN){Co(N′′)2}3] are weakly coupled (J≈−4.4 cm−1). The finding that [HAN].− can be synthesized as a stable salt and transferred to cobalt introduces potential new routes to magnetic materials based on strongly coupled, triangular HAN building blocks.

Redox-active ligands continue to provide ar ich source of fascinating coordination chemistry. [1] In addition to the fundamental interest in the ability of some ligands to act as electron reservoirs,m etal complexes of redox-active ligands have been developed for ar ange of applications.I mportant examples of ligand non-innocence are found in biological coordination chemistry,with NO,O 2 and dithiolenes featuring prominently. [2] Thec oncepts of ligand non-innocence and metal-ligand cooperativity have been used to design new catalytic reactions, [3] and the electron transport properties of complexes with non-innocent ligands have been harnessed for applications in molecular electronics and redox-flow batteries. [4,5] Thei nfluence of redox-active ligands on moleculebased magnets has also produced striking results,s uch as room-temperature magnetic ordering in [V-(TCNE) x ·y CH 2 Cl 2 ]( x % 2, y % 0.5;T CNE = tetracyanoethylene), [6] and al arge coercive field at 11 Ki nt he lanthanide single-molecule magnet [K (18-c-6)][Tb 2 {N(SiMe 3 ) 4 (m-N 2 )-(THF) 2 ](18-c-6 = 18-crown-6). [7] Hexaazatriphenylene (HAT, Figure 1) is the simplest member of af amily of electron-deficient tris(bidentate) ligands with considerable non-innocent character. [9][10][11][12][13][14] An area in which the potential non-innocence of HATl igands has not yet been exploited is molecular magnetism, principally because stable,s ynthetically useful radical derivatives of HATa re unknown. HAT-type ligands are of interest in this context because they can bind metal ions in atriangular arrangement, which can in principle lead to frustrated or non-collinear spin systems.S tudies of HAT-mediated magnetic exchange are rare and have been limited to the unreduced ligand. [15][16][17][18][19][20] Although the exchange coupling in M 3 HAT complexes with M = Co II ,F e II or Cu II is antiferromagnetic,i t is also very weak (j J j< 2cm À1 for 2 J spin Hamiltonians), hence spin-frustration was not observed.
We were interested to see if strong exchange could be induced in an M 3 HATc omplex by exploiting the noninnocent character of the ligand. Achieving this aim would introduce possibilities for developing new building blocks for the assembly of molecule-based magnets.T oavoid the use of explosive precursors,weused the derivative hexaazatrinaphthylene (HAN). Our aim was to reduce HAN to its corresponding radical anionic form, and then to transfer [HAN]C À to at ransition metal ion. Because HAT-type compounds suffer from poor solubility,w ec hose the twocoordinate cobalt(II) amide [Co(N'') 2 ]( N '' = N(SiMe 3 ) 2 )a s the transition metal synthon, [21] which provides anisotropic d 7 ions and lipophilic trimethylsilyl groups.
Heating potassium, HAN and 18-c-6 in toluene produced [{K(18-c-6)}HAN],[{K(18-c-6)}1], as an analytically pure red powder in 90 %yield (Scheme 1). TheUV/vis/NIR spectrum of [{K(18-c-6)}1]i nT HF consists of as eries of broad, overlapping absorptions at l < 450 nm, which are probably due to transitions from the p-HOMOs to the low-lying p* LUMOs ( Figure S3). Thespectrum also shows awell-resolved transition at 590 nm, which is typical for an organic p-radical and likely corresponds to aS OMO-LUMO p-p*t ransition. [22] TheX -band EPR spectrum of [{K(18-c-6)}1]i nT HF solution at 293 Kf eatures as ingle resonance centred on g = 2.0033, confirming its radical nature,with extensive hyperfine structure (Figures 2a nd S10 in the Supporting Information (SI)). Thes pectrum was simulated using EasySpin with the hyperfine coupling constants in Table S2. Further insight into the molecular and electronic structure of [{K(18-c-6)}1]w as obtained through DFT calculations owing to the reluctance of the compound to form single crystals.T he geometry of [{K(18-c-6)}1]was fully optimized in the C 3 point group using the B3LYP exchange-correlation functional in combination either with 6-311G** or with def2-TZVP triple-z basis sets (see SI for computational details). Forcomparative purposes, the same calculations were also carried out on the pristine [HAN]C À radical anion in the D 3h point group.T he B3LYP/6-311G** (B3LYP/def2-TZVP) structure for [{K(18-c-6)}1] consists of an ear-planar HAN ring system, with the potassium cation positioned perpendicular to the central C 6 unit, with K-C distances to C(1)-C(6) of 3.683-3.684  Table S1). [23] Complexes 2 and 3 have similar molecular structures,with both containing three Co II centres bonded to two nitrogen atoms on the unreduced and reduced HAN ligand, respectively,a nd to the nitrogen atoms of two amide ligands.I n2, the cobalt environments are distorted tetrahedral, featuring relatively long Co À Nb onds to the HAN ligand in the range 2.063(3)-2.118(3) ( average 2.094 ) and relatively short Co À Nb onds to the amide ligands of 1.916(4)-1.   Them agnetic susceptibility of both materials was measured in afield of H = 1Tat T = 2-300 K (Figure 4). For 2,the c M T(T)p roduct declines slowly from 7.19 cm 3 Kmol À1 at 300 Kto0.42 cm 3 Kmol À1 at 2K,and shows aslight inflection at about 30 K. Thei nflection is reproducible across different samples,and is unlikely to be due to torqueing as the samples were restrained in eicosane.T he appearance of the c M T(T) data for 2 suggests weak antiferromagnetic exchange between the cobalt(II) centres.T he high-temperature value of c M T is significantly greater than the value of 5.63 cm 3 Kmol À1 expected for three non-interacting S = 3/2 cobalt(II) ions with g-factors of 2.00, indicating appreciable spin-orbit mixing with orbitally degenerate excited states in 2.T he c M T(T) profile for [K(18-c-6)][3]i smarkedly different to that of 2.
Cooling the sample from 300 Kt o5 0K produces as teady increase in c M T(T)from 9.66 cm 3 Kmol À1 to reach amaximum of 12.19 cm 3 Kmol À1 .T his behaviour arises from strong antiferromagnetic interactions between the radical ligand and the three cobalt(II) centres,e ffectively resulting in af erromagnetic alignment of the metal-based spins.A t lower temperatures,arapid decrease in c M T is observed, reaching av alue of 4.49 cm 3 Kmol À1 at 2K. To quantify the magnetic interactions in 2 and 3,wehave fit the magnetisation and susceptibility data simultaneously with PHI. [24] In order to simplify the Hamiltonians and avoid over-parameterisation, at hree-fold symmetric model was employed, where all three Co II ions are equivalent for both 2 and 3.Owing to the strong magnetic anisotropy,the zero-field splitting (ZFS) of the S = 3/2 Co II ions was taken into account using as ingle axial D parameter, assuming collinear anisotropies for the three sites.I nb oth cases we considered only isotropic exchange between all spin centres.T hus,w e employedĤ 1 ðÞ for 2 [Eq.
(2)],w herê S ! 1 ,Ŝ ! 2 andŜ ! 3 are the S = 3/2 Co II spins,Ŝ ! 4 is the radical S = 1/2, J 1 is the Co-Co exchange, J 2 is the Co-radical exchange, D is the axial Co II ZFS, g Co is the Co II g-factor, g rad is the radical g-factor, m B is the Bohr magneton andH is the magnetic field.
Them agnetic data could be adequately simulated using the parameters given in Table 1, where g rad was fixed at 2.00. While neither simulation is perfect, the model provides good rationalisation of the magnetic data. Them odel can even explain the low-temperature plateau for 2;b elow 40 Kt he thermodynamic population is dominated by as ix-fold neardegenerate ground state,a nd thus c M T becomes linear, until Zeeman depopulation occurs below 5K.Use of rhombic ZFS with an E parameter did not improve the fits.The results show
(2)].T he Co II S = 3/2 ground states have significant ZFS,w ith the D-value of À38(4) cm À1 for 2 being approximately double the value of À15(2) cm À1 for 3.The large ZFS and g > 2f or tetrahedral Co II is consistent with significant mixing of an orbitally degenerate excited state into the ground state through spin-orbit coupling.
In order to verify the magnitude and the sign of D,w e performed complete active space self-consistent field (CASSCF) calculations on 2 and 3 using the experimental atomic coordinates.T he calculations were performed with MOLCAS8 .0 (see SI for details). [25,26] To directly assess the local ZFS of the Co II sites,each structurally inequivalent Co II was investigated independently,w here the remaining Co II sites were substituted with the closed-shell ion Zn II .F or 3, asingly oxidized form without the radical spin was examined using the fixed solid-state structure to examine only the structural contribution to the ZFS.The calculations for 2 gave an average D value of À110(20) cm À1 ,a nE/D value of approximately 0.08 and g av = 2.46, while for 3 they gave an average D value of À57(5) cm À1 ,a nE/D value of approximately 0.12 and g av = 2.41. Thus,the calculations support the experimental result of both an egative D value with small rhombicity,a sw ell as the trend of D 2 % 2D 3 and g > 2. The magnitude of D is overestimated in both cases by af actor of three to four,a nd the g av values are higher than those found experimentally,i ndicating that the calculations overestimate the extent of excited state mixing into the ground state.T he calculations also yield the local orientations of the magnetic anisotropy at each site.F or both 2 and 3,the easy axes of the Co II ions are dictated by the local [N''-Co-N'']p lane and are oriented at angles of 70-808 8 relative to the plane of the HAN ligands ( Figure 5). Accounting for these local orientations using an on-collinear model in PHI did not significantly enhance the quality of the fits.T he different D-values for 2 and 3 are likely to arise as ar esult of the differences in the geometric parameters associated with the individual cobalt-(II) centres.
Complexes of anisotropic 3d metals have stimulated huge levels of activity over the last 25 years owing to their potential single-molecule magnet (SMM) properties. [27,28] One strategy for increasing the anisotropy barrier in an SMM is to employ strong exchange,w hich can lead to greater separation between the magnetic ground state and the excited states. This can be accomplished by using radical ligands. [29][30][31] We were interested to see if complexes 2 and 3 displayed SMM characteristics,h owever measurements of the in-phase and out-of-phase AC magnetic susceptibility as functions of temperature did not reveal any slow relaxation in fields of H dc = 0-1500 Oe.T he absence of SMM behaviour-despite the appreciable anisotropy-can be explained by the nonzero rhombicity of the Co II centres in 2 and 3,w ith the irregular 4-coordinate geometry also contributing.S imilar observations on as trongly coupled, radical-bridged dicobalt species were described recently. [32] In conclusion, [HAN]C À (1)was synthesized as the [K(18c-6)] + salt and shown to be an S = 1/2 radical with spin density distributed across the aromatic system. Adding three equiv- ( 2)h ave similar molecular structures,h owever their magnetic properties are markedly different. In the case of 2,w eak antiferromagnetic exchange between the cobalt centres was identified, with J % À4cm À1 (À2 J formalism). In 3,v ery strong coupling to the [HAN]C À radical ligand was found, with J estimated as À290 cm À1 .T he properties of [HAN]C À provide ab lueprint onto which other metal ions with greater magnetic anisotropy, such as lanthanides,can be incorporated.