Synthesis, structure and magnetic properties of tetrakis-[mu]-carboxylato-bis(dodecylnicotinato)dicopper(II) complexes; crystal and molecular structure of the decyl carboxylate derivative

a INQUIMAE, Departamento de Quõ Âmica InorgaÂnica, Analõ Âtica y Quõ Âmica Fõ Âsica, Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires, PabelloÂn II, Ciudad Universitaria, 1428 Buenos Aires, Argentina, Instituto de Ciencias, Universidad Nacional de General Sarmiento, Roca 850, 1663 San Miguel, Buenos Aires, Argentina, Departamento de Fõ Âsica, Facultad de Ciencias Exactas, Universidad Nacional de La Plata and IFLP (CONICET), CC 67, 1900 La Plata, Argentina, and Institut de Physique et Chimie des MateÂriaux de Strasbourg, Groupe des MateÂriaux Organiques, 23 rue du Loess, 67037 Strasbourg CEDEX, France


Introduction
Investigations on dicopper carboxylates exhibiting the lantern structure have been directed for a variety of interests (Mehrotra & Bohra, 1983;Kato & Muto, 1988). The structural and magnetic properties of the acetate derivative in the past were considered to be very important in the understanding of magnetic exchange in dinuclear compounds, relevant topics in inorganic and physical chemistry, and in the development of appropriate models to describe the magnetic coupling between metal centers (Bleaney & Bowers, 1952;Kahn, 1993).
All the copper carboxylate liquid crystals reported thus far are polymeric, the axial position of each Cu 2+ ion being occupied, in the Cu 2 (O 2 CR) 4 series, by an O atom of a neighbor dimer. To break this polymeric structure into distinct dimers we decided to`block' the axial positions of the longchain dicopper carboxylates by forming bis(pyridine derivative) adducts (like in the nicotinamide antifungals). As in the equatorial carboxylate, the axial ligand we used also bears a long chain.
2.2. Synthesis of Cu 2 (O 2 CC n À 1 H 2n À 1 ) 4 (C 5 H 4 NCOO-C 12 H 25 ) 2 compounds Dodecyl nicotinate (2 mmol) was dissolved in 10 ml of boiling n-heptane. The corresponding copper carboxylate (0.5 mmol in 10 ml of boiling n-heptane), synthesized as described by Ibn-Elhaj et al. (1992), was added dropwise. The mixture was re¯uxed for 30 min and then cooled to 255 K overnight. A green solid was recovered by ®ltration at low temperature. The solid was dried in vacuum over silica. The compound with n = 10 crystallized as green platelet crystals, from which single-crystal samples were selected for crystallographic studies. Anal

Physicochemical methods
IR spectroscopy was carried out on a Nicolet FTIR 510P with KBr pellets and UV±vis measurements on a HP8453 spectrometer. 1 H NMR spectra were run on a Bruker AC200 in CDCl 3 . Differential scanning calorimetry (DSC) was performed on a Shimadzu DSC-50 calorimeter. For optical microscopy observations, a Leitz DMRX microscope equipped with a Leitz 1350 heating stage device was used. Elemental analyses were performed on a Carlo Erba CHNS-O EA1108 instrument.
Magnetic measurements between 2 K and room temperature were performed on a Quantum Design MPMS-XL susceptometer in a 5000 G magnetic ®eld. The powdered samples, typically 5±15 mg, were placed in a cellophane bag. Raw magnetic data were corrected for the sample holder contribution and for the sample diamagnetism, which was evaluated using Pascal's constants (Selwood, 1956).
Variable-temperature powder X-ray diffraction (XRD) patterns were recorded using an INEL CPS-120 curved position-sensitive detector and Cu K radiation from an INEL X-ray generator; room-temperature data were collected on Siemens D5000 equipment.
2.4. Single-crystal diffraction data and structure solution and refinement of Cu 2 (O 2 CC 9 H 19 ) 4 (C 5 H 4 NCOOC 12 H 25 ) 2 Crystal data, data collection procedure, structure determination methods and re®nement results are summarized in Table 1. 1 As expected for a complex with extended organic ligands, the samples turned out to diffract poorly. Only about 59% of the X-ray intensity data up to 0.841 A Ê resolution were above two standard deviations of measurement errors.
Several H atoms were detected at approximate locations in a difference Fourier map. They, however, were positioned stereochemically and re®ned with the riding model. During the re®nement, the CH 3 H atoms were allowed to rotate as a rigid group around the CÐCH 3 bond so as to maximize the sum of the electron density at the three calculated hydrogen positions.

Synthesis and characterization
The formation of the expected complexes was ®rst studied in solution. Experiments carried out with Cu 2 (O 2 CC 9 H 19 ) 4 dissolved in toluene showed a progressive conversion of the 678 nm absorption band of the starting complex into a new band centered at ca 702 nm, as the dodecyl nicotinate concentration increased. This shift is consistent with a change in the coordination number around Cu 2+ from four to ®ve. For example, the pyridine adduct band shifts down in energy to 713 nm owing to its strong interaction with copper. Job's method (Reilley & Sawyer, 1961) clearly indicated a 1:2 complex-to-nicotinate ligand ratio, as expected for a bisadduct, but also showed that the stability constant for this Cu 2 (O 2 CR) 4 (nicotinate) 2 compound is not very high. To overcome this we performed the synthesis of the compounds in excess dodecyl nicotinate ligand; their high solubility in non-polar solvents indicated that a low-temperature crystallization process was necessary.
The characterization of the solid product was completed by the crystal structure determination of the decyl derivative by XRD methods (see below) and IR spectroscopy. The main IR bands are, in addition to the # CH 3 , # CH 2 stretching vibrations characteristic of long-chain organic compounds, the # CO 2 vibration of the ester at 1725±1740 cm À1 , the # py symmetric stretching mode of the pyridine ring (1591 cm À1 in the free ligand) shifted by coordination to 1616±1618 cm À1 , as well as the # CO 2 symmetric and asymmetric stretching modes of the bridging carboxylates at 1424±1437 cm À1 (depending on the carboxylate chain length) and 1599±1588 cm À1 , respectively. These values agree with those found for bis-nicotinamide copper carboxylates (Kozlevcar et al., 1996).

Crystallographic structural results and discussion for the n = 10 derivative
Atomic fractional coordinates and equivalent isotropic displacement parameters are given in Table 2. Selected bond distances and angles are listed in Table 3. Fig. 1 is an ORTEP (Johnson, 1965) drawing of the dimeric complex.
The Cu 2 (O 2 C(CH 2 ) 8 CH 3 ) 4 (C 5 H 4 NCO 2 (CH 2 ) 11 CH 3 ) 2 complex is located on a crystallographic inversion center. The pair of Cu II ions in a dimer are bridged through the carboxylate groups of four CH 3 (CH 2 ) 8 CO 2 À ligands. Each Cu atom is in an elongated octahedral environment, coordinated equatorially to four carboxylate O atoms of two independent CH 3 (CH 2 ) 8 CO 2 À moieties [CuÐO bond distances vary from 1.950 (3) to 1.966 (2) A Ê ] and axially by the pyridine nitrogen of a C 5 H 4 NCO 2 (CH 2 ) 11 CH 3 molecule [d(CuÐN) = 2.183 (3) A Ê ] and by the other inversion-related Cu atom located at 2.615 (1) A Ê . The local molecular symmetry is close to D 4h . The nicotinate ring plane is approximately coplanar to one of the CÐOÐCuÐOÐC bonds, but slightly bent by 11 with respect to the CuÐCu direction.   The CuÐCu distance is the second shortest CuÐCu distance found for Cu 2 (O 2 C n H 2n À 1 ) 4 L 2 compounds. This can be related to the moderate basicity of the axial pyridinic ligand (Melnõ Âk et al., 1985;Melnõ Âk, 1982); in fact, the only compound which exhibits a shorter CuÐCu distance is the pyrazine derivative, the least basic axial ligand (pK a = 0.65) for which the bis-adduct has been structurally characterized. The CuÐO distances are also slightly shorter than the corresponding values reported previously in the range 1.96±1.98 A Ê (Melnõ Âk et al., 1985). The observed CuÐO eq /CuÐN ax distances relationship is as expected on the basis of the local electroneutrality principle (Kato & Muto, 1988;Pauling, 1960) for such a weak basic ligand.
One of the two independent alkyl chains of the equatorial carboxylates is completely elongated, in a fully extended zigzag trans arrangement, whereas the other aligns parallel to the ®rst one after a gauche conformation at the C(12)ÐC (13) bond. The alkyl chain of the nicotinate ligand is also in an extended conformation, after turning at the C(37) atom, rotated from the nicotinate plane to align with the carboxylate chains (see Fig. 1). Both the equatorial ligands and the nicotinate alkyl chain are straight and tilt slightly with respect to the plane perpendicular to the NÐCuÐCuÐN axis. Three crystallographycally non-equivalent alkyl chains converge in a plane: that belonging to the elongated carboxylate is positioned between the alkyl chain of the other non-equivalent carboxylate from the same dimer and a third one belonging to a nicotinate of a neighboring molecule. These alkyl chains are completely interdigitated (see Fig. 2), leading to the formation of two different layers along the crystal: one is de®ned by the polar copper carboxylate cores and the second, non-polar one contains the alkyl chains. This type of polar±nonpolar layer arrangement is typically found in mesomorphic systems.
3.3. Phase transitions and thermal behavior of the Cu 2 (O 2 CC n À 1 H 2n À 1 ) 4 (C 5 H 4 NCO 2 C 12 H 25 ) 2 series The compound in which n = 20 exhibits a lamellar phase from room temperature up to 377 K, where it melts to the isotropic liquid. The interlamellar distance, d 100 , is 49.4 A Ê at 298 K, but it sharply changes to 49.1 A Ê at 325 K and then to 52.8 A Ê at 339 K. This contraction±expansion sequence correlates with phase transitions detected in DSC measurements. Table 4 contains the thermodynamic parameters of these transitions. The experimental interlamellar distance is in good agreement with the molecular length in an elongated conformation, estimated as 52 A Ê by extrapolation from the solved structure for the n = 10 derivative. The n = 18 homologue showed a transition to a lamellar phase at 333 K (d 100 = 46.7 A Ê , consistent with an estimated length of 47 A Ê ). During the phase transitions of both compounds some differences in the high diffraction angle region have been observed, revealing changes in the crystalline order at the aliphatic chain packing level. The powder XRD patterns of the n = 12, 14 and 16 derivatives are much more complicated, precluding any straightforward assignment. As mentioned above, the crystal structure of the decyl derivative can also be interpreted as a layered arrangement of polar cores on the ab plane separated by aliphatic chains, but now the interlamellar separation is about half that expected for a segregated bilayer. This is due to the intercalation of the alkyl chains belonging to neighboring dimers related to each other by a unit cell translation along the c axis.
Interestingly, the differences in the supramolecular features between long-chain and short-chain derivatives can be correlated with molecular characteristics. Indeed, the IR spectra showed that the position of the # CO 2 symmetric stretching band depends on the chain length: it appears at 1599 cm À1 for the n = 10, 12, 14 and 16 compounds, and at 1588 cm À1 for the n = 18 and 20 derivatives. The difference between the # CO 2 asymmetric and symmetric stretching mode wavenumbers decreases from 176 cm À1 for n = 10 to 150 cm À1 for n = 20, thus re¯ecting a more symmetric coordination (Deacon & Phillips, 1980) of the equatorial ligands as the chain length increases.
No liquid crystal phases have been found for these compounds, despite the presence of the long alkyl chains and the rich polymorphism they exhibited before melting to the isotropic liquid. All the studied compounds decompose between 493 and 513 K. The thermal stability is higher than those of nicotinamide adducts of copper carboxylates (Kozlevcar et al., 1996). Fig. 3 shows the molar magnetic susceptibility (1) of the Cu 2 (O 2 CC 9 H 19 ) 4 (C 5 H 4 NCOOC 12 H 25 ) 2 dimer measured as a function of temperature. As for the other lantern-structured copper carboxylates, these data correspond to a system with two antiferromagnetically coupled S = 1 2 centers (the increase of 1 at low temperature is due to a very small amount of a paramagnetic impurity, very likely being a monomeric Cu 2+ species). The octadecyl derivative, Cu 2 (O 2 CC 17 H 35 ) 4 (C 5 H 4 N-COOC 12 H 25 ) 2 , exhibits essentially the same 1 versus T variation.

Magnetic behavior
The magnetic behavior of the copper unpaired electrons (S 1 = S 2 = 1/2) in the dimeric complex can be described by the following Hamiltonian (Bleaney & Bowers, 1952) where the ®rst term describes ligand ®eld effects, the second is the exchange coupling between the electrons, and the third and fourth contributions are the spin-orbit and Zeeman interactions. The crystal ®eld effects lead to a mainly metal d(x 2 À y 2 ) ground state orbital with its lobes along the CuÐO bonds, energetically well separated from the electronic excited states. The exchange interaction produces the largest (®rstorder) perturbation to the spin fourfold-degenerated ground state by splitting it into diamagnetic singlet (total spin S = 0) and paramagnetic triplet (S = 1) levels, energy separated in |J|. Perturbation theory up to second-order including exchange spin-orbit and Zeeman interactions shows that the triplet state can be described by the effective spin (S = 1) Hamiltonian where D and g are the ®ne and gyromagnetic tensors, respectively. The ®ne interaction splits the triplet state in the absence of an external magnetic ®eld into a singlet and a doublet. It arises from the combined effect of ligand ®eld, exchange and spin-orbit interactions and vanishes for uncoupled (J = 0) electrons. The symmetric D and g tensors share the same principal (x,y,z) axes, where the spin Hamiltonian adopts the form Single-crystal EPR measurements in the closely related system of Cu 2 (CH 3 COO) 4 (H 2 O) 2 shows that the dimer exhibits approximate axial symmetry and therefore E 9 0, g x 9 g y 9 g c and g z = g || (Bleaney & Bowers, 1952). Assuming the same holds for our system, (3) further simpli®es to H eff DS 2 z g c H x S x H y S y g jj H z S z X 4 The ®ne and Zeeman interactions (a fraction of 1 cm À1 ) are much smaller than the exchange interaction (in the present case, a few hundredths of a cm À1 ). This interaction, in turn, is much less than the energy separation between electronic ground and excited states. This warrants the description of the powder molar magnetic susceptibility by the Van Vleck expression, which to second-order approximation leads to the Bleaney±Bowers equation &X 5 The TIP contribution to 1 corresponds to temperature-independent paramagnetic effects and the last term in (5) accounts for the presence of small amounts of paramagnetic impurities in a proportion &. g 2 is an isotropic average of the squared gtensor, the g p value can be taken to be equal to 2 and the other parameters have their usual meaning. Table 2 Fractional atomic coordinates and equivalent isotropic displacement parameters (A Ê 2 ).
U eq 1a3AE i AE j U ij a i a j a i Xa j X Least-squares ®t of (5) to the susceptibility data leads to the following parameters: ÀJ = 347 (1) cm À1 , g = 2.037 (4) and & = 0.65 (1)%; the value of the TIP is not meaningful because of uncertainties in the diamagnetic correction of such a large molecule. The corresponding 1(T) curve is compared in Fig. 3 with the experimental susceptibility.
The compound shows a strong intramolecular antiferromagnetic interaction, which can be attributed to CuÐCu superexchange through the carboxylate groups. In order to correlate these magnetic results with structural characteristics, we veri®ed that the diffraction pattern of the bulk powdered sample used for magnetic experiments effectively agrees with that calculated from the crystallographic data. In general, the values of the magnetic parameters agree with those already published for Cu 2 (O 2 C n H 2n À 1 ) 4 L 2 compounds, where L coordinates via N atoms. The more relevant parameter, |J|, is near the upper limit of the thus far reported values (308±353 cm À1 ; Kato & Muto, 1988;Kawata et al., 1992;Melnõ Âk et al., 1985), which seems in agreement with the short CuÐO bond distances. The correlation of |J| with the basicity of the ligand has been the subject of some controversy in the literature. Indeed, some authors invoked an increase of the exchange parameter as the axial ligand becomes a stronger electron donor (Melnõ Âk et al., 1985;Melnõ Âk, 1982); contrarily, others claimed (Kato & Muto, 1988) that the AF superexchange increases as the axial ligand basicity decreases, a fact that has been interpreted as a consequence of the strong CuÐ O bonds (shorter distances) present in order to maintain the local electroneutrality of copper(II) ions, which enhance the overlap between magnetic orbitals through the bridge's HOMO, resulting in a stronger AF interaction. Our results seem to support the conclusion drawn by Kato and Muto. However, the magnetic coupling described by J depends simultaneously on several factors and it cannot be assigned only to the ligand basicity. For example, it has been found that there is no correlation between J and the CuÐCu distance (as expected for a bridged-mediated coupling  Transition temperature (K) and enthalpy (in kJ mol À1 within parentheses) for Cu 2 (O 2 CC n À 1 H 2n À 1 ) 4 (C 5 H 4 NCOOC 12 H 25 ) 2 derivatives.
C 1 , C 2 , C 3 refer to crystalline phases; I refers to the isotropic phase.

Conclusions
The use of long-chain nicotinate derivatives as axial ligands was successful in order to break out the polymeric structure of dicopper carboxylates. Even if they did not present liquid crystal phases, there is evidence in the crystal structure of the decylcarboxylate adduct of a microsegregation of the alkyl chains from the remaining, and more polar, part of the molecule. In fact, the compounds with the longest alkyl chains exhibit a lamellar crystalline structure whose interlamellar distance was related to the largest dimension of one molecule. Those results indicate that this system could be slightly modi®ed to induce the appearance of liquid crystal smectic phases.
The structural analysis of the decyl derivative allowed the interpretation of the magnetic behavior of this compound, which exhibits one of the strongest antiferromagnetic interactions among this type of dinuclear copper compounds. The short CuÐO distance, in addition to the coplanarity of the CuÐOÐCÐOÐCu group, warrant a large overlap between the d x 2 Ày 2 magnetic orbital of the Cu 2+ cation and the lone pair of the O atom that leads to a strong superexchange interaction between the metallic atoms through the carboxylate's bridge.