2.1 Crystal structure

Compound 1 crystallizes in the I4/m space group of the tetragonal system (Table S1, Supporting Information). The structure of 1 consists of an anionic 2D network, {Dy^III[Cu(Me₂pma)₂]₂Cl(H₂O)}^2-, and Li(H₂O)₄⁺ countercations, which are intercalated within adjacent square-grid layers of (4⁴·6²) net topology growing in the ab plane (Fig. S1), together with disordered crystallization water molecules (Fig. 1).

Within the anionic oxamato-bridged framework of 1, each Cu^II ion is surrounded by two Dy^III ions, whereas each Dy^III ion is surrounded by four Cu^II ions, thus accounting for the final 2:1 Cu:Dy stoichiometry (Fig. 1a–b). Thus, each square grid is defined by Dy^III corners and Cu^II ions residing in the middle of the square side arranged pretty close to an ideal square grid with a Cu(1)–Dy(1)–Cu(1c) [(c) = y, −x, z] angle of 88.7°. Furthermore, squares are decorated and almost entirely filled by the aromatic rings of the oxamate ligands pointing inward the pores.

Overall, the connectivity of the Cu^II and Dy^III ions in 1, acting as biconnectors and tetraconnectors (Fig. S2), respectively, yields a 2D net similar to those found when n-Bu₄N salt of the copper(II) precursor [Cu^II(Me₂pma)]^2- is reacted with the nitrate salt of the Sr^II and Ba^II ions (Fig. 1c–b) [58]. Each Cu^II ion is tetra-coordinated by two bis(bidentate) oxamate bridging ligands in a square-planar CuN₂O₂ geometry [with average Cu–O and Cu–N bond distances of 1.86(2) and 1.85(5) Å, respectively]. In turn, the Dy^III ions are ten-coordinated being surrounded by four oxamate bridges from different [Cu^II(Me₂pma)]^2- entities, a chlorine, and a water molecule in a distorted bicapped cubic geometry MnO₉Cl (Fig. 2). The two crystallographically distinct Dy-O_oxamate bond lengths of 2.37(1) and 2.44(2) Å [for Dy-O2 and Dy-O3 and their symmetrical equivalents, respectively (Fig. 2a)] are shorter than those of the Dy-O_w [2.59(2) Å for Dy-O4] and Dy-Cl [2.62(3) Å] ones but in agreement with those found in literature for analogue environments [59,60].

Eclipsed anionic heterobimetallic open frameworks of 1 (Figs. S3 and S4) interact with Li(H₂O)₄⁺ cations through multiple hydrogen-bonding O(oxamate)···H-Ow host–guest interactions [O(oxamate)···H-Ow of 3.02(4) Å] (Figs. 1 and 3, S2 and S4) describing a supramolecular motif reminiscent of Li-ion batteries where solvated alkali metal ions are intercalated within anionic graphite-like layers. As shown in Fig. 3, hydrophobic and hydrophilic regions alternate as a result of the disposition of the dimethyl-substituted phenylene spacers, pointing outward or inward of the voids, respectively. The terminal chlorine atoms are pretty confined in the hydrophilic sides, the aforementioned Li⁺ intercalation being further stabilized or likely driven by hydrogen bonds involving Li(H₂O)₄⁺ cations and the coordinated water molecules [Li-Ow···H–O1w of 2.9(4) Å].

Even if it was not possible to find a reasonable model for the disordered water molecules (see Supporting Information), the structural analysis revealed an estimated volume of accessible solvent voids of 1936.1 Å³ that represent up to 29.7% of the total unit cell volume [6190.5 Å³]. This feature would likely account for the embedding cocrystallized water molecules placed within the sheets, in good agreement with the elemental analyses for 1.

2.2 Thermogravimetric analysis and X-ray powder diffraction of 1

The solvent contents of compound 1 were determined by thermogravimetric analysis (TGA) under dry N₂ (Fig. S5). It shows a very fast mass loss from room temperature to around 70 °C, followed by a plateau under further heating up to 200 °C, when decomposition starts. Overall, the mass loss of ca. 18.5% at 150 °C corresponds to ca. 13 molecules per formula unit, and, together with elemental analysis, the final chemical nature of the compound was determined (see Supporting Information).

The powder X-ray diffraction (PXRD) pattern of a polycrystalline sample of 1 (Fig. 4) confirms the isostructurality with the crystal selected for single-crystal X-ray diffraction and pureness of the bulk.

2.3 Magnetic properties

Fig. 5a shows the temperature dependence of the direct current (dc) magnetic susceptibility of 1, in the form of the χ_MT versus T plot (χ_M being the dc molar magnetic susceptibility per Cu₂Dy unit). At room temperature (R. T.), χ_MT value for 1 (14.76 cm³ K mol⁻¹) is similar to that expected for the sum of two square-planar Cu^II ions (χ_MT = 0.80 cm³ mol⁻¹ K with g_Cu = 2.1 and S_Cu = ½) and one Dy^III ion [χ_MT = (Nβ²g_Dy²/3k_B)J(J + 1) = 14.15 cm³ mol⁻¹ K with J =15/2 and g_Dy = 4/3]. Upon cooling, χ_MT for 1 decreases, most likely as a consequence of the first-order angular momentum and crystal field effects of the Dy^III ion [61] (⁶H_15/2 term with S = 5/2 and L = 5), and it attains a minimum at ca. 14 K (inset of Fig. 5a). Below the minimum, χ_MT increases sharply to reach a χvalue of 19.32 cm³ K mol⁻¹ at 2.0 K. Such a sharp increase below 14 K (Fig. 5a) can hardly be explained only through intralayer ferromagnetic interactions. Antiferromagnetic interactions between Cu^II and Dy^III ions could turn 11 into a ferrimagnetic system with also a sharp increase of χ_MT at low temperatures. However, the observed minimum value of χ_MT is too high, exceeding that expected for two Cu^II ions magnetically uncoupled with the ground doublet of the Dy^III ion.

The M versus H plots (M being the molar magnetization per Cu₂Dy unit and H the applied dc magnetic field) at 2.0 K (Fig. 5b) further confirm this ferromagnetic behavior. The isothermal magnetization curve of 1 exhibits a quite fast saturation with a maximum M value of 6.55 Nβ at 5.0 T, which is very similar to that expected, considering a parallel alignment of the spins of two Cu^II (S_Cu = 1/2) and one Dy^III (S_Dy = 5/2) ions, thus suggesting a particularly strong ferromagnetic interaction between neighboring metal ions. However, no hysteresis was observed for this compound at 2.0 K because of its small value, requiring a lower temperature to allow its record. Moreover, within the 2D anionic Cu^II₂Dy^III and 1D neutral Cu^IIDy^III networks, the ferromagnetic nature of the magnetic coupling between the Dy^III and the Cu^II ions through the oxamate bridge was already reported in the past [54,62]. Therefore, it seems obvious to expect that the observed magnetic behavior is a consequence of a ferromagnetic intralayer coupling. Thus, even if χ_MT decreases in the range 300–314 K for 1, owing to the depopulation of the excited Stark sublevels of Dy^III ions, the ferromagnetic interaction within the plane can slightly mask this drop of χ_MT.

Unfortunately, there is no analytical expression capable of simulating the magnetic behavior of this regular 2D network, but it contains two different metal ions, and one of them possesses coupled spin and angular momenta. This last point makes the study of the magnetic properties particularly tricky, even more, when some effects are acting at the same time, that is, when the effect of the magnetic coupling is playing its role during the depopulation of the excited doublets caused by the spin-orbit coupling (SOC) on the Dy^III ion. In other words, some methodologies based on effective unities could help in a relatively simple way when a phenomenon is dominant, but in 1, the magnetic coupling is strong enough so that its effect comes together with those from SOC shown by the Dy^III ion. In such circumstances, the simulation of the magnetic behavior of 1 becomes an impossible task.

Therefore, the experimental data above 50 K, where the influence of the magnetic coupling is not significant yet, were analyzed. The spin Hamiltonian used and summarized in Eq. 1 embodies the SOC of the Dy^III ion and the Zeeman effect of this ion and two uncoupled Cu^II ions. Structural distortions in the coordination sphere of the Dy^III ion appear as an axial zfs parameter (Δ) on its angular momentum. The spin–orbit coupling constant was considered constant and equal to the value of the free ion (λ = −380 cm⁻¹). The ideal spin momentum (S = 5/2) for the Dy^III ion in this model has, therefore, a g-factor equal to 2.0, the value for the free electron (g_e). The low delocalization of the f orbitals in the rare-earth complexes supports the assignation of the unity as value for κ. Having in mind these particularities, with a g-factor of 2.00 for the Cu^II ion (g_Cu), the Δ parameter for the best fit above 50 K takes a value of +66.5 cm⁻¹, which is in agreement with values for other dysprosium(III) complexes reported previously [54]. The agreement factor between experimental and simulated data, defined as F=∑(Pexp−Pcalcd)2/∑Pexp2∂2Ω∂u∂v, with P being the measured physical property, is F = 1.6 x 10⁻⁵. A simulation with these values agrees reasonably well with the experimental curve; however, it is not perfect even if any restraint on the values of λ and κ is removed. Probably, a possible effect of the magnetic coupling occurs even above 50 K, which, if small or moderate, was also observed in a Dy₂Cu₃ network with oxamate as a bridging ligand [54].

CASSCF/NEVPT2 calculations on a DyCu₄, or even on a DyCu model, could procure both the splitting of the ground term promoted of the ligand field on the Dy^III ion and the magnetic coupling between Dy^III and Cu^II ions. This study is prohibitively expensive because of the excessive size of the chosen active space to build the ground and excited states. Besides, the magnetic interactions between 4f and 3d ions involve the empty 5d orbitals of the lanthanoid ions, which requires an enormous increase of the active space [63–65]. Moreover, these factors make the convergence of this kind of calculations difficult. Because there can be no doubt about the ferromagnetic nature between Dy^III and Cu^II ions, choosing a model that allows studying only the splitting of the f orbitals and the ground term promoted of the ligand field on the Dy^III ion is a good option. In this case, from the experimental geometry of 1, the appropriate model (DyZn₄) is built replacing paramagnetic Cu^II ions by diamagnetic Zn^II ones in, keeping the electronic features of the chemical surrounding (Fig. 6). This splitting leads to a few low-lying excited sextets very near to the ground state (Fig. 7 and S6), as demonstrated for the CASSCF/NEVPT2 calculations. Specifically, 10 of these excited states are in the range of 1161 cm⁻¹, and three of them place only 31, 258, and 266 cm⁻¹ above the ground state. This fact agrees with a stronger influence of the SOC on the splitting of the states and the magnetic properties than the ligand-field effect. By the previous discussion about the analysis of the magnetic data, the presence of close low-lying excited states makes the analysis of the simulation of the magnetic behavior of 1 difficult, which becomes impossible when the magnetic couplings throughout the 2D network are included. Next sextet excited states are placed 7550 cm⁻¹ above the ground state. Quartet excited states are placed beyond 24,000 cm⁻¹.

On the other hand, 10 donor atoms organized in a bicapped square prism constitute the coordination sphere of the Dy^III ion. In this coordination polyhedron, four chelating oxamate ligands in “equatorial” positions occupy the eight vertices of the square prism. Considering each oxamate ligand with two coordinate donor atoms as the only coordination point, the coordination would look like an octahedron in a D_4h symmetry and with four oxamate ligands in its equatorial plane, this being the cause of the use of “equatorial” term in the present discussion. Although the Dy-O(oxamate) bond lengths are longer than those in a previous system, the Dy-O(water) bond length is shorter than that of the majority of the reported dysprosium(III) compounds. The weaker ligand field of “axially” coordinated chloride and water ligands than the “equatorial” oxamate groups makes the dysprosium(III) compound exhibit an in-plane anisotropy [66–68], that is, an easy magnetization appears in the “equatorial” plane, which divides the coordination polyhedron into two separate square pyramids. This particular magnetic anisotropy could spread throughout the 2D DyCu₂ network through intralayer ferromagnetic couplings.

Compound 1 shows a paramagnetic-to-ferromagnetic phase transition at T_C = 7.5 K, which is confirmed by both (i) the presence of sharp non–frequency-dependent peaks at 7.5 K in the χ″_M versus T plot (Fig. 8a) and (ii) the divergence of the field-cooled magnetization (FCM) and the zero-field-cooled magnetization (ZFCM) curves at the same temperature (Fig. 8b). In general, a 2D network cannot display a magnetic order, but 1 does. The non–frequency dependence of the χ″_M peaks (Fig. 8) rules out a slow relaxation of the magnetization on the Dy^III ion. On the other hand, the presence of moderate ferromagnetic intralayer interaction cannot explain this magnetic ordering without the inclusion of (i) ferromagnetic interlayer interaction between neighboring planes or (ii) a magnetic anisotropy that 1 behaves as a magnetic XY plane [69]. The first case is based on the most likely dipolar nature of the interactions, but they are often weak and antiferromagnetic. However, in 1, the first-order SOC in the Dy^III ion with an in-plane anisotropy transfers a magnetic anisotropy to the DyCu₂ layer, this being an XY plane where a long-range magnetic ordering is possible without the presence of the interlayer couplings, which seems reasonable from the crystal structure. Yet the behavior of this kind of 2D systems is not simple. They exhibit a particular phase transition of infinite order from a high-temperature disorder phase to a low-temperature quasi-ordered phase. However, they can explain the observed behavior for 1. Interestingly, although a good number of oxamato-based CPs merely with 3d metal ions exhibiting ferromagnetic ordering have been reported, ferromagnetic ordering in CPs involving 4f metals is very rare. Thus, with isolated exceptions, such as polymeric compounds involving interchain interactions [40,45] and a 2D compound reported by Evangelisti et al. [55], in which a ferromagnetic ordering was observed below 2.0 K by specific heat measurements, compound 1 is one of the very first 3d-4f magnets and indeed exhibits the largest T_C reported so far for lanthanide-based CPs.

4.1 Materials

All chemicals were of reagent-grade quality. They were purchased from commercial sources and used as received. The proligand HEt–Me₂pma and the mononuclear copper(II) complex (n-Bu₄N)₂[Cu(Me₂pma)₂] · 2H₂O were prepared as previously reported [25,57]. Preparation of Li₂[Cu(Me₂pma)₂] · 2H₂O is given in detail in the Supporting Information.

4.2 Preparation of [Li^I(OH₂)₄]₂[Dy^IIICu^II₂(Me₂pma)₄ClH₂O] ^. 4H₂O (1)

Well-formed dark green prisms of 1, which were suitable for X-ray diffraction, were obtained by slow evaporation of 10 mL of a water/acetonitrile/methanol solution (1:10:10 volume ratio) containing stoichiometric amounts of Li₂[Cu^II(Me₂pma)₂] ^. 2H₂O (0.10 g, 0.20 mmol) and DyCl₃ (0.027 g, 0.10 mmol) with an excess of LiCl (0.5 mmol). After several days standing on air, green crystals of 1 appeared. They were filtered off and air-dried. Yield: 59%; elemental analysis calcd. (%) for C₄₀H₆₂Cu₂DyLi₂N₄O₂₅Cl (1337.9): C 35.91, H 4.67, N 4.18; found: C 35.78, H 4.59, N 4.21; IR (KBr): ν = 1601 (C]O).

4.3 Single-crystal X-ray diffraction

Crystal data for 1: C₄₀H₆₅₂Cu₂DyLi₂N₄O₂₁Cl, M_r = 1236.76, tetragonal, space group I4/m, a = 15.9307(5), c = 24.3926(9) Å, V = 6190.5(5) ų, T = 90(2) K, λ = 0.71073 Å, Z = 4, ρ_calc = 1.356 g cm⁻³, μ = 1.983 mm⁻¹, 2188 unique reflections [35,182 measured (R_int = 0.0270)] and 1898 observed with I > 2σ(I), 1.53° ≤ θ ≤ 25.24°, R = 0.1319 (0.1380 for all data), wR = 0.3715 (0.3806 for all data) with 114 parameters and 18 restraints, the final Fourier-difference map showed maximum and minimum height peaks of 4.144 and −3.600 e Å⁻³. Crystal structure deposited at Cambridge Crystallographic Database with CCDC 1891550.

Single-crystal of 1 was mounted on glass fibers in a grease drop and very quickly placed on a liquid nitrogen stream cooled at 100 K to avoid the degradation upon dehydration. Diffraction data were collected on a Bruker-Nonius X8APEXII CCD area detector diffractometer using graphite-monochromated Mo Kα radiation (λ = 0.71073 Å). The data were processed through the SAINT [70] reduction and SADABS [71] multi-scan absorption software. Despite the high quality of the full set of data, a quite low θ_max of diffraction (θ = 27°) was obtained (detected as Alerts A in the checkCIF), even if many efforts have been made to extract the best diffraction data from the sample. Furthermore, high residual electron density, which is less than 1 Å from the metal atom, is due to the effect of the dysprosium and even copper ripples. However, the solution and refinement parameters are suitable, compared with analogue MOF structures previously reported; thus, we are convinced that the structure found is consistent [72–75].

The structure was solved by the Patterson method and subsequently completed by Fourier recycling using the SHELXL-2013 software package [76,77]. All nonhydrogen atoms were refined anisotropically except Li⁺ alkali metal and some disordered carbon or oxygen atoms from ligands and water molecules. The hydrogen atoms were set in calculated positions and refined as riding atoms. The high thermal vibration parameters (Alerts A and B in the checkCIFs) displayed for some atoms in the organic ligand are consequence of contributions from different factors, including (a) the flexibility of the framework and consequential disorder, (b) the high residual electron density produced by the methyl groups that are dynamic components of the walls, and (c) the use of some bond length and angle restrains during the refinements or fixed positions of some highly disordered atoms. Furthermore, and as expected for such systems, the lattice water molecules were highly disordered and cannot be satisfactory modeled (at the origin of the first levels Alerts A). Those found from the ΔF map were the coordinated ones refined with restraints and their hydrogen atoms were neither found nor calculated. As a consequence, the contribution to the diffraction pattern from the highly disordered water molecules of crystallization (6 molecules of H₂O located in the voids of the lattice that amount to ca. 30% percentage void volume of the unit cell) was subtracted from the observed data through the SQUEEZE method, implemented in PLATON [78]. The final formulation for each compound is consistent with the residual electron density and volume. The final full-matrix least-squares refinements on F², minimizing the function ∑w(|F_o|-|F_c|)², reached convergence with the values of the discrepancy indices given in Table S1. High R values (levels Alert A in checkCIFs) are, most likely, mainly affected by the ripples and contribution of the highly disordered solvent to the intensities of the low angle reflections. The final geometrical calculations and all the graphical aspects were carried out with CRYSTAL MAKER and WinGX [79–81].

4.4 X-ray powder diffraction

A polycrystalline sample of 1 was introduced into a 0.5 mm borosilicate capillary and then it was mounted and aligned on an Empyrean PANalytical powder diffractometer, equipped with Cu Kα radiation (λ = 1.54056 Å). For each sample, five measurements were collected at room temperature (2θ = 2–40°) and merged in a single diffractogram.

4.5 Magnetic measurements

Variable-temperature direct current (dc) and alternating current (ac) magnetic susceptibility measurements were carried out with a Quantum Design SQUID magnetometer. The susceptibility data were corrected for the diamagnetism of both the constituent atoms and the sample holder.

4.6 Theoretical calculations

To evaluate the splitting of the states promoted by crystal field on the Dy^III ion in 1, the relative energies of the ground and low-lying excited states were computed by calculations based on the second order N–electron valence state perturbation theory (NEVPT2) applied on the wave function, which was previously obtained from complete active space (CAS) calculation. These calculations were performed with the version 4.0 of the ORCA program on a DyZn₄ molecular model built from the experimental geometry of 1 and by replacing paramagnetic Cu^II ions by diamagnetic Zn^II ones, to keep to the utmost, the electronic features of the system [82]. The TZVP basis set proposed by Ahlrichs and the auxiliary TVZ/C Coulomb fitting basis sets were used for all atoms [83,84]. Electronic relativistic effects introduced by the dysprosium atom were considered through the ZORA Hamiltonian [85] and the SARC (segmented all-electron relativistically contracted) version of the TZVP basis set (SARC-ZORA-TZVP) [86–90]. The energies of 21 sextet and 12 quartet states generated from an active space with nine electrons in seven f orbitals were calculated. See Supporting Information for further details.