Probing Relaxation Dynamics in Five‐Coordinate Dysprosium Single‐Molecule Magnets

Abstract A new family of five‐coordinate lanthanide single‐molecule magnets (Ln SMMs) [Dy(Mes*O)2(THF)2X] (Mes*=2,4,6‐tri‐tert‐butylphenyl; X=Cl, 1; Br, 2; I, 3) is reported with energy barriers to magnetic reversal >1200 K. The five‐coordinate DyIII ions have distorted square pyramidal geometries, with halide anions on the apex, and two Mes*O ligands mutually trans‐ to each other, and the two THF molecules forming the second trans‐ pair. These geometrical features lead to a large magnetic anisotropy in these complexes along the trans‐Mes*O direction. QTM and Raman relaxation times are enhanced by varying the apex halide from Cl to Br to I, or by dilution in a diamagnetic yttrium analogue.

Single-molecule magnets (SMMs) show slow relaxation of magnetisation under certain conditions;t hese have received huge interesti nt he last 25 years due to their potential applications in quantumc omputing, molecular spintronics and ultra-highdensity storage. [1,2] For technological applications, it is important to determine the highest temperature at whichaSMM can retain its magnetisation, [1] and to understand the multiple relaxation mechanisms, which can be involved in their magnetic dynamics. Lanthanide (Ln) SMMs have provided some promising candidatess ince their discovery in 2003. [3,4] The design criteria to synthesise Ln SMMs, [5] with as ingle Dy III centre in a highly axial ligand field environment to generate large magnetic anisotropy and to stabilise the highest m J = AE 15/2 spin state of Dy III as the ground state, has led to Ln SMMs with magnetisation dynamics dominated by an Orbach-like (thermally activated) relaxation process, with very high values of the activation energy (D). [6][7][8][9][10] [7] have similar D values (1815 Ka nd 1760 K, respectively) but very different T B 100s (12 Ki n4 and 53 Ki n5;h ere we use T B 100s to define the temperaturea t which the magnetic relaxation time is 100 si nz erof ield). The reasons behind the significant difference in T B 100s ,d espite having comparable energy barriers, can be attributed to the differences in the other relaxation processes involved, for example,R aman and quantum tunnelling of magnetization (QTM),a nd therefore it is necessary to understand these mechanismsi nd etail to design higher-performing SMMs. [11][12][13][14][15][16] Therefore, systematic studies on as eries of compounds from ag iven family are essential. In particular,aseries of complexes with fine control of only one structuralc haracteristic should be ideal to see its influence on their dynamic properties. To this point, only two large families of Dy-based SMMs with D/k B > 1000 Ka re known: the pentagonal bipyramidal [Dy(Solv) 5 (L 1 )(L 2 )] (Solv = THF,P y; L 1 = OtBu, Cl, Br,P hO;L 2 = OtBu, Cl, Br,P hO) complexes [6] and the metallocene [Dy(Cp R1 )(Cp R2 )] + cations( R 1 = H, Me, tBu, iPr;R 2 = H, Me, tBu, iPr). [7][8][9][10] Here we present af amily of five-coordinate Dy SMMs, [Dy(Mes*O) 2 (THF) 2 X] (Mes* = 2,4,6-tri-tert-butylphenyl) (X = Cl, 1;B r, 2;I ,3), designed such that the Dy coordination sphere can be selectively varied at as ingle position to study its influence on the relaxation dynamics. In this system,t he sterically demanding aryloxide ligand was employed to reduce the coordination number at the Dy centre. Alkoxide and aryloxidebased ligandsh ave been widely used in Ln chemistry, [17] and in synthesisingL nS MMs in recent years. [6,14,[18][19][20] The [Dy(Mes*O) 2 (THF) 2 X] complexesw ere prepared directly by the salt metathesis reactions of two equivalents of NaOMes*w ith the parent halide in THF (Scheme1). Similarly,t he diamagnetic Y(III) analogous compound, 1-Y and a5 %d oped sample 5%Dy@1-Y were synthesised to perform complementary NMR spectroscopy and dilution experiments, respectively.
The single-crystal X-ray analysis of 1-3 and 1-Y (Figure 1a nd Figures S1-S3 in Supporting Information) reveal that 1, 2 and 1-Y crystallise in the P2 1 /c space group, whereas 3 crystallises in C 2 /c (Tables S1-S2). All molecules contain aL n III ion in ar are pentacoordinate distorted square-based pyramidal geometry, having ah alide anion at the apex of the pyramid with two trans-Mes*O ligands, andt wo trans-THF molecules making the square base (Tables S3-S4). The only monomeric five-coordinate Dy SMM knowni nt he literature, [Dy(NHAr) 3 (THF) 2 ]( Ar = C 6 H 3 iPr 2 -2,6), [21] has an energy barrier of 34 K, arising from a trigonal bipyramidal geometry with three anionic anilide donors and two neutralTHF donors.The distortion of the coordination sphere for five-coordinate systemsc an be quantified by the geometricp arameter t 5 = (bÀa)/60, where b and a are the largest ands econd-largest angles in the coordination sphere, respectively.T he t 5 parameter quantifiest he degree of trigonality within the structural continuum between squarebased pyramid (t 5 = 0) and trigonal bipyramid (t 5 = 1). [22] For complexes 1 to 3, t 5 was found to be 0.348, 0.344 and 0.340, respectively,w hich is consistent with as ignificantly distorted square-based pyramid (Table S4). This distortion is mainly due to the four O-donors in the square base, which are not in a single plane.I n1,t he angle at Dy between the O-donors from the THF ligand is 167.3(1)8,a nd the angle between the Odonors from aryloxide is 146.4(1)8.C ontinuous shape measurement calculations also favour as quare-based pyramid (Table S5). [23] The DyÀOb ond distances for 1 to 3 (Table S3) for the anionic oxidesO 1a nd O2 fall in the range 2.116(4) to 2.124(3) , whilst the neutral oxygen donors O3 and O4 show longerD y À Ob onds:f or 1,2 .370(3) and 2.352(3) ;f or 2,2 .346(4) and 2.370(4) ;f or 3,2 .366(3) and 2.366(3) .W hen movingf rom X = Cl to Br to Ii nt he series, the DyÀXb ond distance increases from 2.537(1) to 2.6981(6)t o2 .9540(6) ,w hich is attributed to the increasei nt he ionic radii of the halide present. The intermolecular nearest Dy··Dyd istances in the crystal packing (Figures S4-S12) varies as 7.7, 7.9 and 10.7 for 1, 2 and 3 respectively.C omplex 1-Y is isostructural to 1 (Tables S2-S4)  To investigate the magnetic properties of this family of complexes, dc (directc urrent) and ac (alternating current) susceptibility measurements were performed. The temperature dependence of the dc magnetic susceptibility was performed under an appliedf ield of 0.1 T ( Figure S28). At 270 K, the measured cT values of 13.8, 13.7, 13.5 and 13.7 cm 3 mol À1 K, for 1, 2, 3 and 5%Dy@1-Y (when normalised per mol of Dy complex) respectively,a re close to the expected value (14.17 cm 3 mol À1 K) for af ree Dy III ion. [24] The continuous decreaseo ft he cT product upon cooling to low temperatures suggestss trong crystal-field splitting. The field dependence of magnetisation measured at multiple temperaturesb elow 15 K (between AE 7T )s hows slow magnetisation dynamics and blocking ( FiguresS29-32). The magnetisation at 1.85(1) Ka nd 7Tsaturates at M sat = 5.3, 5.0, 4.6 and 5.3 N A m B for 1, 2, 3 and 5%Dy@1-Y (when normalised per mol of Dy complex), respectively.T he slow dynamics observed by dc measurements were furthers tudied using ac susceptibility with frequenciesu pt o 10 kHz. As shown in Figures S33-40, the in-phase (c')a nd outof-phase( c")c omponentso ft he ac susceptibility are strongly temperature and frequency dependenti nz ero-dc field for all the complexes. The magnetisation relaxation time and its associated distribution were estimated as af unctiono ft he temperature (Figure 3) from the fitting of the experimental c' versus. n and c" versus n data to the generalised Debye model (see Figures S33-40). [25][26][27] In zero dc-field, paramagnetic relaxation [26][27][28] usually involves the three main mechanisms includingR aman, [28,29] thermally activated (Orbach-like) [28,30] and quantum tunnelling (QTM) [28] relaxation pathways, as summarised in the following equations [Eq. (1)and (2)]: Figure 1. View of the [Dy(OMes*) 2 (THF) 2 Cl]moleculars tructure in 1 from its X-ray crystal structure at 100 Kw ith thermal ellipsoidsdrawn at 50 %p robability level (Dy turquoise,C lgreen,Ored, Cgrey). Ha toms are omitted for clarity.
As shown in Figure 3, the above five-parameters model can reproduce almost perfectly the t versus T À1 data for 1, 2 and 3 (see Ta ble 1), which clearly display three temperature domains associated with dominating Orbach (above~55 K), Raman (~55-10K)a nd QTM (below~10 K) processes. It is interesting to note that the blockingt emperature of theses ystems, T B 100s , is not defined for 1, 2 and 3, as the QTM relaxation time is systematically smaller than 100 seconds. For the dilutedc ompound, 5%Dy@1-Y,t he QTM regime is not observed in the availablee xperimentalw indow and thus the relaxation time was modelled considering only Orbach andR aman processes down to 8K.
The first conclusion that can be drawn from these experimental results( Ta ble 1), is that the activation energy of the Orbach process is statistically the same regardless of the halide present at D/k B % 1200 K, and does not change for 1 upon dilution;t his result agrees with our CASSCF-SOc alculations. The lack of change in D/k B is probably because the halide is not on the principal anisotropy axis. The modification of the phonon bath moving from 1 to 5%Dy@1-Y also has little influence on D/k B .T his conclusion is not surprising as the Orbach mechanism is primarily intrinsic to the electronic structure of the SMM, which is not much perturbed by the choice of halide here.
On the other hand, the halide substitution has ac lear impact on the Ramanr elaxationw ith ac haracteristict ime that increases as Cl < Br < I( Figure 3, Table 1). While the C and n parameters are relatively similar along the series, the intrinsic Raman relaxation time for ag iven temperature between 55 and 10 K( for example at 30K which we define as t Raman 30K ) changes by af actor of 8b etween 1 and 3.I nc ontrastt ot he Orbach process, dilution significantly impacts the C and n parameters which decrease( /24) and increase ( 1.3), respectively. This is as triking differencet ot he bis-cyclopentadienyl dysprosium(III) SMMs, [7] or in some pentagonalb ipyramidal Dy SMMs, where there was no difference between the Raman parameters for the pure and doped materials. [6c] This suggests that 1 is far more sensitive to the precise crystal lattice and associated phonon bath.I ti st empting to link this sensitivity to the faster relaxation observed here.
The halide variation also varies the QTM time which increases Cl < Br < I( by af actor 6b etween the Cl and Ia nalogues;   (Figures S34, S36, S38 and S40) and the log-normal distribution as describedi nreference [27].The solid lines are the best fit discussed in the text.  Table 1. D, t 0 , C, n and t QTM parametersg enerated from the fit of the relaxation time-temperature dependence( Figure 3) for 1-3 and 5%Dy@1-Y.T hese parameters are givenw ith their fitting error in parenthesis and their estimated standard deviations (AE)b asedo nt he estimated standard deviations of the relaxation times showninF igure 3.