Dysprosium Single-Molecule Magnets Involving 1,10-Phenantroline-5,6-dione Ligand

: The two mononuclear complexes of the formula [Dy(tta) 3 ( L )] ( 1 ) and [Dy(hfac) 3 ( L )] ( 2 ) (where tta - = 2-thenoytrifluoroacetylacetonate and hfac - = 1,1,1,5,5,5-hexafluoroacetylacetonate) were obtained from the coordination reaction of the Dy(tta) 3 ·2H 2 O or Dy(hfac) 3 ·2H 2 O units with the 1,10-phenantroline-5,6-dione ligand ( L ). Their structures have been determined by X-ray diffraction studies on single crystals, and they revealed a supramolecular assembly of tetramers through σ-π interactions. Both complexes displayed a Single-Molecule Magnet (SMM) behavior without an external applied magnetic field. Magnetic relaxation happened through Orbach, Raman and Quantum Tunneling of the Magnetization (QTM). Wavefunction theory calculations were realized to rationalize the magnetic properties. analyses of the compounds were performed at the Centre Régional de Mesures Physiques de l’Ouest, Rennes. A Quantum Design MPMS-XL SQUID magnetometer was used to measure the dc magnetic susceptibility from solid polycrystalline samples. A 0.2 kOe, 2 kOe and 10 kOe magnetic field was applied respectively for the temperature ranges of 2–20 K, 20–80 K and 80– 300 K. These measurements were all corrected for the diamagnetic contribution, as calculated with


Introduction
The design of lanthanide coordination complexes is crucial for the development of emitting materials [1,2] and molecular magnetism fields [3][4][5]. On the one hand, the specific characteristic of the lanthanide ions, such as intense, long-lived and sharp-line emissions, makes them ideal candidates for a plethora of applications ranging from luminescent-and bio-sensors [6,7], Organic Light-Emitting Diodes (OLEDs) [8], time-resolved fluoro-immunoassays [9] and imaging [10,11]. Nevertheless, the observation of such emissions needs to be sensitized by an indirect process, since the forbidden Laporte f-f transitions [12] have very weak absorption coefficients which lead to an inefficient direct excitation process. The indirect sensitization process is achieved using organic chromophores which possess adequate triplet and/or intra-ligand charge transfer excited states to efficiently perform an energy transfer to the spectroscopic levels of the lanthanide [13,14]. On the other hand, the strong magnetic anisotropy associated with a high magnetic moment makes lanthanide ions suitable for the observation of Single-Molecule Magnet (SMM) behavior [15]. Such magnetic bistability could find interest in potential applications such as high density storage devices [16], quantum computing [17] or spintronics [18]. In order to observe SMM behavior, the crucial parameter is the magnetic anisotropy, which must be uniaxial or planar depending on the electronic distribution of the f electrons, i.e., oblate or prolate [19]. The control of the magnetic anisotropy is driven by the nature and the position of the ligands around the metal center. Thus, recently, the use of cyclopentadienyl-based ligands has allowed for the design of mononuclear lanthanide SMMs with a high blocking temperature [20][21][22][23]. These two research fields highlight the important role of the organic ligands associated to the lanthanide ion.
To go one step forward, more complex ligands with multi-functionality, such as the combination of redox-activity with multi coordination sites, could be used. These kind of ligands can allow the observation of multi-emission or -SMM behavior [24][25][26][27]. In this context, the 1,10-phenantroline-5,6dione (L) appears as a promising multifunctional ligand since it combines two potential coordination sites with electro-activity [28,29], leading to the design of potential multi-properties compounds, such as a combination of spin crossover and valence tautomerism, as has recently been proposed from a computational point of view [30]. Its electro-activity could be exploited to modulate the physical properties as a function of its oxidation state, i.e., to design redox magnetic and photo-emissive switches, as already published with other redox-active ligands [31][32][33]. The L ligand was employed to design both pure 4f [34][35][36][37][38][39] and hetero-bimetallic nd4f [40][41][42] coordination complexes to study their biological activities and photo-physical properties. It is worth noting that the elaboration of hetero-bimetallic compounds implies the reduction of the ligand in its 5,6-dihydroxy-1,10phenantroline form. To the best of our knowledge, a magnetic study was not performed for lanthanide complexes involving the L ligand.

X-ray Structures
[Dy(tta)3(L)] (1). 1 crystallized in the tetragonal space group I4 (N°79) ( Table S1). The asymmetric unit is composed of one complex of [Dy(tta)3(L)], and its Ortep view is depicted in Figure S1. The Dy(tta)3 unit is linked to the 1,10-phenantroline-5,6-dione ligand (L) through the nitrogen bidentate coordination site and to three ttaancillary ligands (Figure 1a). The nature and the arrangement of these ligands led to a N2O6 environment with a slightly distorted triangular dodecahedron symmetry (TDD-8) (the distortion from the ideal symmetry CShMTDD−8 = 0.455 was determined using the SHAPE 2.1 program [43] with CShM = Continuous Shape Measures).  The crystal packing revealed a supramolecular assembly leading to the formation of tetramers through π-σ interactions between one carbonyl group and the central aromatic ring of the phenantroline moiety (O1· · · Cπ distances range from 2.885 to 3.168 Å) (Figure 2a). π-π interactions between the thiophene ring of the ttaligand and the same central aromatic ring of the phenantroline moiety participate to the stabilization of the tetramers. The shortest Dy-Dy intermolecular distance is 9.845 Å.
[Dy(hfac)3(L)] (2). The X-ray structure of 2 is similar to that of 1. It crystallized in the tetragonal I4 space group (N°82) ( Figure S2 and Table S1). The Dy III ion is coordinated to three hfacligands and to L through the nitrogen bischelating coordination site (Figure 1b). The difference of the electron withdrawing between ttaand hfacled to a slight variation of the Dy-Ohfac and Dy-N bond lengths with 2.333(8) and 2.535(8) Å, respectively. The N2O6 coordination sphere with a D2d triangular dodecahedron symmetry is slightly more distorted than the one determined for 1 (CShMTDD−8 = 0.726). The oxidation of the L ligand is retained with C=O bond lengths of 1.210 (14) and 1.198(12) Å. The supramolecular tetramers are stabilized only by the π-σ interaction (O1· · · Cπ distances range from 2.802 to 3.148 Å) (Figure 2b), since the thiophene rings were replaced by CF3 groups. The shortest Dy-Dy intermolecular distance is 9.247 Å.
The two complexes 1 and 2 highlighted crystal packing which completes the different intermolecular arrangements that could be observed in the literature for compounds made from tris(β-diketonate) of lanthanide associated with phenanthroline derivatives. Indeed, non-substituted phenantroline or π-extended phenantroline led to parallel complex arrangements thanks to π-π interactions [44], phenantroline substituted with an electron donor group such as tetrathiafulvalene led to "head-to-tail" parallel dimers of complexes through donor-acceptor and π-π interaction [45], and phenantroline substituted by an electron rich atom with lone pairs led to perpendicular arrangements of complexes like for compounds 1 and 2.

Ab Initio Calculations
State-Averaged-Complete Active Space Self-Consistent Field/Restricted Active Space State Interaction including Spin Orbit coupling (SA-CASSCF/RASSI-SO) calculations were performed for the two mononuclear complexes to extend the understanding and to rationalize the magnetic properties (see computational details). The computed energy splittings are depicted in Tables S2 and  S3. For both systems, the Dy III center presents an almost pure Ising ground state, 92% MJ = | ± 15/2> for 1 and 95% MJ = | ± 15/2> for 2, defined by a g-tensor with a main component gZ = 19.43 for 1 and gZ = 19.56 for 2 and exhibiting negligible transversal components with gX = 0.00, gY = 0.01 for 1 and gX = 0.01, gY = 0.01 for 2. At this point, the Quantum Tunneling of the Magnetization (QTM) is expected weak, and a slow magnetic relaxation of the magnetization at zero field should be observable. One could observe that the weak difference of the uniaxial character of the magnetic anisotropy ground state follows the distortion trend of the coordination sphere. The main component of the ground state g-tensor of the Dy III center for each complex is represented in Figure 1. To give more insights on the orientation of the magnetic axis, the molecular electrostatic potential is investigated using the homemade Calculated Molecular Multipolar Electrostatics (CAMMEL) software which uses previously calculated Local Properties (LOPROP) charges to compute the electrostatic potential on a sphere centered on the lanthanide ion (see computational details). The resulting potential for both molecules is represented in Figure S3 along with the orientation of the ground state anisotropy. For both systems, the main magnetic component appears perpendicular to the plane containing the 1,10phenantroline-5,6-dione ligand, i.e., the most charged direction as expected for an oblate ion [19]. The matrix elements of the transition magnetic moments between the Stark sub-levels for the Kramers ion of each complex have been computed in order to give more insights into the relaxation mechanisms ( Figure 3). At the molecular scale, no major difference between the two compounds was identified except the higher energy gap between the ground and first excited state for 1 (230.4 K) than for 2 (194.3 K), as already observed when changing the hfacancillary ligands with ttaones [45][46][47].

Static Magnetic Measurements
The dc magnetic properties of 1 and 2 were determined by measuring their thermal dependence on the magnetic susceptibility ( Figure 4). The room temperature values are 13.90 cm 3 ·K·mol −1 for compound 1 and 13.89 cm 3 ·K·mol −1 for compound 2. These values are close to the expected one of 14.17 cm 3 ·K·mol −1 for an isolated Dy III ion ( 6 H15/2 ground state multiplet) [48]. Decreasing the temperature, χMT decreases monotonically down to 11.19 cm 3 ·K·mol −1 for 1 and 11.45 cm 3 ·K·mol −1 for 2 due to the thermal depopulation of the MJ states. Below 5 K, the decreasing is more pronounced, probably due to dominant antiferromagnetic intermolecular interactions ( Figure 4). The magnetization at 2 K displayed classical behavior for the isolated highly anisotropic Dy III ion. The simulated curves reproduced the experimental χMT(T) and M(H) measurements well, leading to the conclusions that the nature of the ground state and energy splitting are correctly computed and identified.

Dynamic Magnetic Measurements
The χM' (in-phase) and χM'' (out-of-phase) components of the ac susceptibility were measured for both compounds 1 and 2. For both compounds, a χM'' was detected in the zero magnetic field, as suggested by ab initio calculations, with maxima localized at 310 Hz and 70 Hz at 2 K, respectively for 1 and 2. They highlighted a frequency dependence of the in-phase ( Figure S4) and out-of-phase signal of the magnetization (Figure 5), allowing an analysis of the temperature variations (Tables S4  and S5) of the magnetic susceptibility in the framework of an extended Debye model (see SI) [49,50]. The out-of-phase component intensity of the magnetization is in agreement with the fact that almost 100% of the sample slowly relaxes, as shown by the normalized Argand ( Figures S5 and S6). The Arrhenius plots of the relaxation time depicted in Figure 6 could be fitted in the zero field with a combination of two thermally activated regimes (Orbach and Raman) and one thermally independent regime (QTM): τ −1 = τ0 −1 exp(-Δ/kT) + CT n + τTI −1 (where τ is the relaxation time, Δ is the energy barrier, C is the Raman constant, n is the exponential constant describing the thermal dependence of the Raman and τTI is the relaxation time for the thermal independent regime). The best fits depicted in Figure 6 were obtained using the parameters of Table 1.  The involvement of the Raman process is supported by the discrepancy between the calculated energy barrier (Δ = 230.4 K for 1 and Δ = 194.3 K for 2) and the experimental barrier extracted from the high temperature region of the Arrhenius plot (Δ = 61(5) K for 1 and Δ = 83(5) K for 2) which can be explained by the operating under-barrier relaxation mechanism [51][52][53][54]. It is worth noting that a satisfying fit could only be obtained by fixing the energy gap to the computed value for each compound. The expected n value for Kramers ion should be 9 [48], but the presence of both acoustic and optical phonons could lead to lower values between 2 and 7 [55][56][57]. The anisotropy barrier values extracted for 1 and 2 are very close to those found for their analogues involving the 1,10phenantroline ligand [58,59].
The application of an external magnetic field is efficient in cancelling the QTM. Thus, a field dependence of χM is studied (Figures 7 and S7). For the two compounds, the presence of an external field led to a significant shift of the out-of-phase component to a lower frequency due to the cancelling of the fast relaxation of the magnetization through QTM. 1200 Oe and 1000 Oe have been chosen as optimal values of the applied magnetic field for 1 and 2, respectively. Under these applied fields, 1 and 2 highlighted a frequency dependence of the inphase ( Figure S8) and out-of-phase signals of the magnetization (Figure 8). An analysis of the framework of the extended Debye model allowed an extraction of the temperature dependence of the relaxation time ( Figure 6, Tables S6 and S7). The out-of-phase component intensity of the magnetization and the normalized Argand indicated that almost the whole sample is slowly relaxing (Figures S9 and S10). The relaxation time follows two thermally dependent relaxation processes (Orbach and Raman) ( Table 1). The best fit was obtained by fixing the Orbach process with the energy barrier (Δ) and relaxation time (τ0) determined for the corresponding compound for H = 0 Oe, since such a magnetic relaxation should be field-independent. Replacing the tta -(compound 1) with hfac -(compound 2) led to a slowing down of the magnetic relaxation time, which is the inverse trend to that usually observed for such chemical engineering [45][46][47]. Indeed, from a molecular point of view, the replacement of hfacby ttaled to a decrease of the coordination polyhedron distortion and an increase of the computed energy gap between the ground and first excited states (Tables S2 and S3). Thus, the explanation of the inverse trend might be of intermolecular origin with the presence of a significant antiferromagnetic interaction, as suggested by the quick decreasing of the χMT product at a very low temperature (Figure 4). Even if the σ-π interactions are similar in both compounds, the intermolecular distance between the Dy III ions, as well as the angles between the magnetic anisotropy vectors, change slightly, leading to a variation of the intermolecular magnetic interactions, which could modify the dynamic magnetic properties. Such a hypothesis is supported by the difference of the frequency dependence of χM'' at 2 K under a 200 Oe applied field (Figure 7), which displayed a clear double contribution at frequencies of 2.5 and 400 Hz for 1 but not for 2. Such an observation should be an indication of stronger intermolecular dipolar interactions in 1 than in 2 [60].

Conclusions
In this article, two mononuclear Dy III complexes involving the 1,10-phenantroline-5,6-dione ligand (L) were structurally and magnetically studied. X-ray diffraction on a single crystal revealed a supramolecular assembly of four mononuclear compounds through σ-π interactions for the two [Dy(tta)3(L)] (1) and [Dy(hfac)3(L)] (2) derivatives. Both compounds behaved as a Single-Molecule Magnet in a zero applied magnetic field with the relaxation of the magnetization occurring through Orbach, Raman and QTM mechanisms. The latter could be efficiently cancelled under an optimal applied field, leading to the slowing of the magnetic relaxation. The experimental magnetic properties could be reproduced and rationalized by ab initio calculations. In zero and applied magnetic fields, 2 displayed better performances than 1, which is in contradiction to what is usually observed when changing the ttaancillary ligand with hfac -. Such observations might be attributed to variations of the intermolecular dipolar interaction.

Crystallography
Single crystals of 1 and 2 were mounted on an APEXIII D8 VENTURE Bruker-AXS diffractometer for data collection (MoKα radiation source, λ = 0.71073 Å), from the Centre de Diffractométrie (CDIFX), Université de Rennes 1, France (Table S1). The structure was solved and refined, respectively, with a direct method using the SHELXT program [64] and a full matrix leastsquares method on F 2 using the SHELXL-14/7 program [65]. The complete crystal structure results as a CIF file are deposited as Supporting Information. The CCDC number is 1,991,212 for 1 and 1,991,211 for 2.

Physical Measurements
The elemental analyses of the compounds were performed at the Centre Régional de Mesures Physiques de l'Ouest, Rennes. A Quantum Design MPMS-XL SQUID magnetometer was used to measure the dc magnetic susceptibility from solid polycrystalline samples. A 0.2 kOe, 2 kOe and 10 kOe magnetic field was applied respectively for the temperature ranges of 2-20 K, 20-80 K and 80-300 K. These measurements were all corrected for the diamagnetic contribution, as calculated with Pascal's constants. The ac magnetic susceptibility measurements were performed on the same apparatus.

Computational Details
The atomic positions were extracted from the X-ray diffraction crystal structures of the 1 and 2 compounds. The wavefunction theory calculations (WFT) were carried out with the help of the OpenMolcas software package [66]. In these calculations, the complete active space self-consistent field [67] (CASSCF) approach was used to treat the static correlation effects arising from the partially filled 4f shell of the Dy(III) ion. The second-order Douglas-Kroll-Hess [68][69][70][71] scalar relativistic (SR) Hamiltonian was used to treat the scalar relativistic effects, in combination with the all-electron atomic natural orbital relativistically contracted (ANO-RCC) basis set from the Molcas library [72][73][74]. The basis sets were contracted to the triple-ζ plus polarization (TZP) quality for the Dy and O atoms bonded to the lanthanide (Dy = 25s22p15d11f4g2h/8s7p4d3f2g1h; O = 14s9p5d3f2g/4s3p2d1f), and to the double-ζ quality for H, C, F and S atoms (H = 8s4p3d1f/2s; C = 14s9p5d3f2g/3s2p, F = 14s9p5d3f2g/3s2p, S = 17s12p5d4f2g/4s3p1d). The calculations employed the state-averaged formalism at the SR level by taking into account the 21 sextet, the 224 quartet and the 490 doublet spin states arising from the 9 electrons spanning the seven 4f orbitals (i.e., CAS(9,7)). The spin-orbit coupling (SOC) was then introduced within a state interaction among the basis of calculated SR states using the restricted active space state interaction (RASSI) approach [75]. Herein the SOC matrix is diagonalized using the calculated 21 SR sextet, 224 SR quartet and the 98th lowest SR doublet spin states. The EPR g-factors were calculated according to [76], as implemented in the RASSI module of OpenMolcas, whereas the magnetic susceptibility and magnetization calculations were performed using the Single-Aniso of OpenMolcas, as detailed in [77].
The molecular electrostatic potential is calculated from the ab-initio LOPROP charge analysis [78]: where , , and are respectively the charge, dipole, quadrupole moments and displacement vector of the i-th atom. The resulting molecular electrostatic potential is mapped and represented using the home-made CAMMEL code. The potential is drawn on a sphere defined by the user around the central lanthanide ion, for a given state (ground state in this case). For a clearer representation of the potential, the intensity can be directly related to both the color (red = high potential and blue = low potential) and the height of the irregularities. This program has already been used in previous works to give some hints on the orientation of magnetization axes [79,80].
Supplementary Materials: The following are available online at www.mdpi.com/2312-7481/6/2/19/s1, Figure S1. ORTEP view of 1. Thermal ellipsoids are drawn at 30% probability; Figure S2. ORTEP view of 2. Thermal ellipsoids are drawn at 30% probability; Figure S3. Molecular Electrostatic potential centered on the lanthanide ion for 1 (left) and 2 (right). The orientation of the ground state anisotropy axis is represented as a dark line. Figure S4. Frequency dependence of in-phase (χM') and out-of-phase components (χM") at 5 K for 1 in a zero applied magnetic field. Red lines are the best fitted curves; Figure S5. Frequency dependence of χM' in the temperature range (a) 2-14 K for 1 and (b) 2-20 K for 2 in a zero applied magnetic field; Figure S6. Normalized Argand plot between 2 and 14 K for 1 in a zero applied magnetic field; Figure S7. Normalized Argand plot between 2 and 14 K for 2 in a zero applied magnetic field; Figure S8. Frequency dependence of χM' at 2 K at various external fields for (a) 1 and (b) 2; Figure S9. Frequency dependence of χM' in the temperature range 2-14 K for (a) 1 and for (b) 2 in a 1200 and 1000 Oe applied magnetic field, respectively; Figure S10. Normalized Argand plot between 2 and 14 K for 1 in a 1200 Oe applied magnetic field; Figure S11. Normalized Argand plot between 2 and 14 K for 2 in a 1000 Oe applied magnetic field; Table S1: X-ray crystallographic data for 1 and 2; Table S2: Computed energies, g-tensor and wavefunction composition of the ground state doublet in the effective spin ½ model for 1; Table S3: Computed energies, g-tensor and wavefunction composition of the ground state doublet in the effective spin ½ model for 2; Table S4: Best fitted parameters (χT, χS, τ and α) with the extended Debye model 1 at 0 Oe in the temperature range 2-13 K; Table S5: Best fitted parameters (χT, χS, τ and α) with the extended Debye model 2 at 0 Oe in the temperature range 2-13 K; Table S6: Best fitted parameters (χT, χS, τ and α) with the extended Debye model 1 at 1200 Oe in the temperature range 3.5-12 K; Table S7: Best fitted parameters (χT, χS, τ and α) with the extended Debye model 2 at 1000 Oe in the temperature range 3.5-14 K. Acknowledgments: B.L.G. and V.M. thank the French GENCI/IDRIS-CINES center for high-performance computing resources.

Conflicts of Interest:
The authors declare no conflict of interest. The founding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.

Abbreviations
The following abbreviations are used in this manuscript: