Spin-transition frameworks based on bistetrazole and triazine ligands Quesada Vilar, Manuel

Citation

Quesada Vilar, M. (2007, March 29). Spin-transition frameworks based on bistetrazole and triazine ligands. Retrieved from https://hdl.handle.net/1887/11463

Version: Corrected Publisher’s Version

License: Licence agreement concerning inclusion of doctoral thesis in the Institutional Repository of the University of Leiden

Downloaded from: https://hdl.handle.net/1887/11463

Note: To cite this publication please use the final published version (if applicable).

6

Coordination Dependence of Magnetic

Properties within a Family of Related

Dinuclear [M ^II ₂ ] Complexes of a

Triazine-Based Ligand

Abstract

The coordination chemistry of the polydentate ligand 2,4,6-tris(dipyridin-2-ylamino)- 1,3,5-triazine (dpyatriz) with Fe^II has been explored, leading through variation of the counter- ion and the solvent system to the preparation of three different dinuclear complexes:

[Fe2(dpyatriz)2(H2O)2(CH3CN)2](ClO4)4 (6), [Fe2(dpyatriz)2(H2O)2(CH3OH)2](BF4)4 (7) and [Fe2(dpyatriz)2Cl2](CF3SO3)2 (8). The X-ray structures of these compounds have revealed that, besides the difference in the non-coordinated anion, complex 6 differs from complex 7 only in the nature of the terminal ligands. Bulk magnetisation studies and Mössbauer spectroscopy have shown that such a subtle difference may produce a change to the crystal field on the metal atoms, leading to an important diversity in magnetic behaviour. Thus, complex 6 presents a partial spin–crossover centred at approximately 265 K, whereas com- plex 7 shows two uncoupled high-spin Fe^II centres over most of the studied temperature range, with the effect of very weak antiferromagnetic coupling and/or single-ion zero-field splitting present at low temperature. By contrast, complex 8, in which one of the nitrogen atoms of the triazine ring of dpyatriz is coordinated for the first time, displays (very uncommon) ferromagnetic coupling interactions between both Fe^II ions within the molecule, leading to an S = 4 spin ground-state. A fit of the experimental data led to a value of the coupling constant of J = + 0.23 cm^–1 (using H = –2 J S1 S2 as the convention for the exchange Hamiltonian) and provided an estimation of D = 0.63 cm^–1 for the ground-state axial zero-field splitting parameter. A magneto-structural comparison of complex 8 with the nickel, [Ni2(dpyatriz)2Cl2](CF3SO3)2 (9) and the cobalt, [Co2(dpyatriz)2Cl2](CF3SO3)2 (10) derivatives, clarifies the mechanism of interaction between the iron centres forming the dimers.

Parts of this chapter have appeared in the literature: Quesada, M.; de Hoog, P.; Gamez, P.; Roubeau, O.; Aromì, G.; Donnadieu, B.; Massera, C.; Lutz, M.; Spek, A. L.; Reedijk, J., Eur. J. Inorg. Chem.

2006, 1353-1361.

6.1 Introduction.

The controlled design of molecule-based functional materials represents one of the research lines inspired by nanotechnological applications. The area of coordination chemistry plays a primary role in this respect, as exemplified by the spectacular progress demonstrated in the context of molecular magnetism.¹ The search for molecular magnetic materials with specific functions is very often based on the use of ligands that will give rise to predictable chemical and physical properties. For instance, it is well known that azole-based ligands have the ability to generate the appropriate type of crystal field around the Fe^II ions, so as to prompt the appearance of the phenomenon of spin crossover (SCO) within the metal ion.^{2, 3} Another example is the exploitation of the potential of certain ligand systems to propagate ferromagnetic interactions between paramagnetic metal ions with the aim of generating high- spin molecules or magnetically ordered materials. Among such ligand-frameworks are (end- on) N3– groups^{4, 5} and m-phenylene links (the latter usually facilitating ferromagnetism through a spin-polarisation mechanism).⁶ Even more desirable in the preparation of versatile functional materials are systems capable of exhibiting different properties resulting from introducing subtle changes of environment or chemical nature.

With the aim of preparing novel coordination supramolecular architectures, the use of the new multinucleating ligand 2,4,6-tris(dipyridin-2-ylamino)-1,3,5-triazine (dpyatriz, Scheme 1)⁷ in coordination chemistry reactions involving various metals has been recently started. The versatility of this ligand has been demonstrated by the diversity of some of the coordination compounds obtained. Thus, the reaction with Cu(NO3)2 produces a 1D coordination polymer whose copper centres are magnetically coupled,⁸ whereas a similar reaction involving CuCl2 allows the characterisation of the trinuclear discrete molecule [(CuCl)3(dpyatriz)2][CuCl4]Cl.⁹ The same process, using Zn(NO3)2, leads to the formation of the tetranuclear molecular species [Zn4(dpyatriz)2(NO3)8].¹⁰ In these products, the dpyatriz li- gands have their three dipyridylamine (dpya) chelating units engaged in coordination. By contrast, the analogous reaction with Co^II produces the dinuclear complex [Co2(dpyatriz)2(NO3)2(CH3OH)2](NO3)2 in which each of the multidentate ligands maintains one of their dpya chelating units uncoordinated. Very recently, the use of solvothermal conditions has allowed the preparation of coordination compounds showing significant structural differences compared to the complexes obtained under normal conditions of temperature and pressure. In this manner, coordination polymers of Cu^II, Zn^II and Cd^II have been synthesised and crystallographically characterised.¹¹ Despite the variety of coordination modes and structures reported, none of the complexes described so far shows the coordination of the dpyatriz ligand via one or more of its s-triazine ring N-donors. It is therefore conceivable that the versatility of dpyatriz can be exploited in the search of new supramolecular architectures exhibiting novel magnetic phenomena.

With this objective, the reactivity of this ligand with various sources of Fe^II ions has been explored. The ability of dpyatriz to form dinuclear complexes of magnetically coupled Fe^II ions is illustrated with the compounds [Fe2(dpyatriz)2(H2O)2(CH3CN)2](ClO4)4 (6), [Fe2(dpyatriz)2(H2O)2(CH3OH)2](BF4)4 (7) and [Fe2(dpyatriz)2Cl2](CF3SO3)2 (8), whose single-crystal X-ray structures have been determined. Thus, the nature of the terminal ligands

completing the coordination sites of the octahedron appears to affect either the ligand field around the metal ions or the coordination mode of the multinucleating ligand, resulting in three related complexes with very diverse magnetic properties, namely spin-crossover (6), antiferromagnetic exchange (7) or (very uncommon for Fe^II) ferromagnetic coupling (8). The synthesis of complexes 6, 7 and 8 and their molecular structures are described, and the origin of their very different magnetic behaviour is discussed. The physical properties of compound 8 are compared with those of the isostructural nickel(II) (9) and cobalt(II) (10) derivatives.

6.2 Synthesis

In attempts to take advantage of the versatility of dpyatriz for the preparation of complexes with new magnetic properties, reactions with various sources of Fe^II applying different experimental conditions have been performed.

Reaction of Fe(ClO4)2 with dpyatriz in acetonitrile, in the presence of ascorbic acid (in order to avoid the aerial oxidation Fe^II→Fe^III) led to a light-pink solution, from which pink crystals could be obtained by slow diffusion of Et2O.¹ The compound is formulated as the dinuclear discrete complex [Fe2(dpyatriz)2(H2O)2(CH3CN)2](ClO4)4 (6), as established by X- ray crystallography (see below). The coordinated water molecules originate either from the original hydrated Fe^II salt, or from the non-anhydrous solvent. Thus, the reaction leading to the formation of 6 can be described through a very simple chemical equation (Eq. 6.1).

2 Fe(ClO4)2 + 2 dpyatriz + 2 H2O + 2 CH3CN →

→ [Fe2(dpyatriz)2(H2O)2(CH3CN)2](ClO4)4 (6.1) It has been observed that crystals of 6 crumble upon exposure to air, yielding a yellowish crystalline material. The elemental analysis suggests the replacement of CH3CN molecules by H2O, although no structural proof for the identity of the resulting material has been obtained. Replacement of the terminal ligands, such as CH3CN, by atmospheric water, with preservation of the core topology is a very common process in coordination chemistry.¹² IR spectroscopy confirms the absence of CH3CN in this new adduct, while the remaining spectral features are unchanged.

An analogous reactivity is observed using Fe(BF4)2·6H2Oand CH3OH as a solvent.

The slow evaporation of the solvent from the reaction mixture results in the formation of light green crystals of a new complex, whose formula, i.e. [Fe2(dpyatriz)2(H2O)2(CH3OH)2](BF4)4

(7), has been revealed by single−crystal X−Ray diffraction (see below) and elemental analysis. Thus, equation 6.1 can be applied to describe the formation of 7, replacing CH3CN and ClO4− by CH3OH and BF4−, respectively. Unlike 6 however, compound 7 is found to be air stable. The similarity in the preparation of 6 and 7, as well as their probable related solid- state structure, has prompted us to test their inter-conversion. Indeed, when crystals of 6 are placed in CH3OH, a slightly greenish powder is formed that does not show SCO properties (this solid material does not turn purple when cooled to liquid N2 temperature). Similarly, if

11 see chapter 2 for details on the experimental procedures and recipes for all compounds presented in this chapter.

crystals of 7 are immersed in CH3CN, a slightly purple powder is obtained. Although this equilibrium has not been further investigated (mostly because the crystals crush into powders, rendering their structural characterisation more difficult), it appears that 6 and 7 can be inter- converted via a simple modification of their solvent environment.

Table 6.1. Crystallographic parameters of complexes 6−8

Compound 6 Compound 7 Compound 8 Formula [C74H64Cl4Fe2N28O18] [C74H79B4F16Fe2N24O10] [C84H62Cl2F6Fe2N26O6S2]

Fw 1887.03 1928.59 1892.32

Diffractometer Nonius KappaCCD Nonius KappaCCD Bruker AXS Smart 1000 Wavelength [Å] 0.71073 0.71073 0.71073

Temp [K] 293(2) 150(2) 293(2)

Crystal size [mm] 0.30 × 0.20 × 0.05 0.54 × 0.48 × 0.12 0.36 × 0.30 × 0.24 Crystal colour colourless yellow yellow

Crystal system triclinic monoclinic triclinic Space group P 1 (no. 2) P21/c (no. 14) P 1 (no. 2)

a [Å] 11.417(2) 13.282(2) 13.250(2)

b [Å] 13.470(3) 14.499(2) 13.621(2)

c [Å] 14.100(3) 22.6080(8) 13.682(2)

α [°] 109.80(3) 90 64.419(3)

β [°] 93.60(3) 91.703(5) 70.521(3)

γ [°] 96.50(3) 90 75.965(3)

V [ų] 2015.2(7) 4351.8(9) 2085.6(5)

Z 1 2 1

Dx [g cm^-3] 1.558 1.472 1.507

µ [mm^-1] 0.582 0.438 0.547

refl. measured 12334 68949 23330

(sin θ/λ)max [Å⁻¹] 0.62 0.65 0.66 abs. corr. psi-scans multi-scan multi-scan abs. corr. range 0.87 - 0.97 0.79 - 0.92 0.87 - 1.00

refl. unique 8121 9974 9310

parameters 649 590 547

restraints 278 0 4

extinction 0.0269(19) - -

R1/wR2 [I>2σ(I)] 0.0467 / 0.1149 0.0506 / 0.1278 0.0781 / 0.2426 R1/wR2 [all refl.] 0.0604 / 0.1259 0.0780 / 0.1438 0.1166 / 0.2726

S 1.094 1.031 0.951

ρmin/max [e Å^-3] -0.51 / 0.38 -0.63 / 1.10 -1.09 / 1.57

The addition of stronger coordinating Cl⁻ anions into the system described above, leads to a structurally closely related complex, but with significant differences regarding its magnetic behaviour. Thus, reaction of dpyatriz and Fe(CF3SO3)2·4H2O in benzyl cyanide in the presence of NBu4Cl, at 100 ºC in a sealed pressure tube, yields yellow crystals of the compound [Fe2(dpyatriz)2Cl2](CF3SO3)2 (8). Compound 8 is formed under these conditions within a few days, according to the reaction depicted in equation 6.2.

2 Fe(CF3SO3)2 + 2 dpyatriz + 2 NBu4Cl →

→ [Fe2(dpyatriz)2Cl2](CF3SO3)2 + 2 NBu4CF3SO3 (6.2) The structure of compound 8 has been unveiled by X-Ray crystallography (see below), which shows that the Cl⁻ ions are coordinated to the Fe^II centres. In addition, one N-atom of the triazine ring of the dpyatriz ligand is unprecedently bound to the metal ion, resulting in an important modification of the magnetic properties of the complex. It has been observed that when stoichiometric amounts of the chloride source are used, the ensuing yields are inferior compared with reactions with sub-stoichiometric amounts of this reagent. A possible reason is that increased amounts of NBu4Cl may favour the formation of the dinuclear anion [Fe2OCl6]²⁻ at the expense of complex 8. This anion is very stable and quite ubiquitous in aerobic reactions involving Fe^II and chloride.¹³

6.3 Description of Structures.

Crystallographic data for complexes 6−8 are collected in Table 6.1, whereas selected metric parameters are listed in Table 6.2.

[Fe2(dpyatriz)2(H2O)2(CH3CN)2](ClO4)4 (6).

Complex 6 is a centrosymmetric cationic dinuclear unit (Figure 6.1, left) comprising two octahedral Fe^II ions bridged and chelated by two dpyatriz ligands through two of their dpya moieties. The third dpya unit is free, and therefore available for the subsequent coordination to an additional metallic centre (see Chapter 8). The coordination sphere of each metal is completed, at the basal plane, by a molecule of H2O and one of CH3CN, at cis positions of the slightly distorted FeN5O octahedral chromophore. The Fe−O and Fe−N(CH3CN) distances are 2.140(2) and 2.151(3) Å, respectively, whereas the Fe−N(dpyatriz) distances range from 2.159(2) to 2.177(2) Å.² The Fe···Fe separation is 9.232(2) Å, while the shortest intermolecular Fe···Fe distance is 8.992 Å. Four perchlorate counter anions are located between the dinuclear units, and exhibit disorder of their oxygen atoms. In addition, two lattice acetonitrile molecules are found in that space. Only weak electrostatic and Van der Waals interactions appear to be involved in the crystal packing. The cation of 6 is structurally analogous to the one observed for the equivalent complex¹⁰ with Co^II.

[Fe2(dpyatriz)2(H2O)2(CH3OH)2](BF4)4 (7).

The structure of 7 (Figure 6.1) is closely related to that of 6. It also consists of centrosymmetric Fe^II dinuclear entities held together by two dpyatriz ligands showing the same coordination mode. The only differences are the replacement of perchlorate by tetrafluoroborate and substitution of the CH3CN ligand by CH3OH. This solvent exchange has significant consequences on the magnetic properties of the resulting coordination compound (see Section 6.4). The geometric parameters are comparable to those of 6, and are reported in

Table 6.2. For instance, the Fe−N bond lengths range from 2.152(2) to 2.190(2) Å, whereas the Fe−O distances are 2.1708(18) Å (CH3OH) and 2.0846(18) Å (H2O).² The intra-dimer Fe−Fe vector measures 9.2593(15) Å and the shortest interdimer Fe−Fe distance is 8.3577(14) Å. The BF4− ions are again located between the diiron cations. The free dipyridylamine moieties of the complex cation forms hydrogen bonds with a CH3OH molecule. (O−H···N distances of 3.112(3) and 2.816(3) Å, respectively).

Figure 6.1. ORTEP representation at the 50% probability level of the cation of [Fe2(dpyatriz)2(CH3CN)2(H2O)2](ClO4)4 (6), left and [Fe2(dpyatriz)2(H2O)2(CH3OH)2](BF4)4

(7), right. Hydrogen atoms are left out for clarity. Only unique non-carbon atoms are labelled.

[Fe2(dpyatriz)2Cl2](CF3SO3)2 (8).

The cationic part of complex 8 (Figure 6.2) is based as well on a centrosymmetric dinuclear unit in which two octahedral Fe^II ions are connected by two dpyatriz ligands.

However, in 8 the binding mode of the multidentate ligand is different from that observed in both previous complexes. Indeed, each dpyatriz ligand is coordinated to an iron ion in chelating fashion with three atoms, involving the two nitrogen atoms N1 and N4 from one dpya moiety, and the N-donor N7 from the triazine ring. Two pyridyl N-atoms (N11’ and N12’) from dpya of a second dpyatriz ligand complete the basal plane of the octahedron. In this way, two pyridyl rings in each dpyatriz ligand remain uncoordinated. Contrary to complexes 6 and 7, the octahedral coordination environment around the Fe^II ion is not completed by solvent molecules, but by a terminal Cl⁻ ion. The Fe1−Cl1 and Fe1−N7(triazine) distances are 2.336(2) and 2.220(4) Å, respectively, while the other four Fe- N bonds involving the pyridyl groups vary from 2.205(4) to 2.234(4) Å (see Table 6.2).^{2, 14} To the best of our knowledge, complex 8 represents the first crystallographically characterised compound where the central triazine moiety of the ligand dpyatriz is involved in metal-

bonding interactions. The triazine rings of the two ligands are parallel to each other, although shifted with respect to one other. The ring centroids although largely parallel are separated by a distance of 3.584(1) Å. The intramolecular Fe···Fe [-x, 1-y, 1-z] separation is 8.324(1) Å, while the shortest intermolecular separation between two metals is 9.041(2) Å. The two positive charges of the cationic complex [Fe2(dpyatriz)2Cl2]²⁺ are balanced by two disordered CF3SO3− anions located in the crystal lattice. Two molecules of solvent (benzyl cyanide) are present for each dinuclear complex, which are strongly interacting with the electron-poor triazine rings via their electron-rich nitrile N-atom (the smallest distance of N1(solvent) to the least-square plane passing through the ring being 3.038(9) Å).

Table 6.2. Selected interatomic distances [Å] and angles [º] for [Fe2(dpyatriz)2(H2O)2(CH3CN)2](ClO4)4 (6) (left), [Fe2(dpyatriz)2(H2O)2(CH3OH)2](BF4)4 (7) (middle) and [Fe2(dpyatriz)2Cl2](CF3SO3)2 (8) (right).

Compound 6 Compound 7 Compound 8

Fe1-O1W 2.140(2) Fe1-O2 2.1708(18) Fe1-Cl1 2.336(2) Fe1-N1 2.160(3) Fe1-N1 2.167(2) Fe1-N1 2.229(3) Fe1-N3 2.177(2) Fe1-N6 2.176(2) Fe1-N4 2.205(4) Fe1-N4 2.159(2) Fe1-O1W 2.0846(18) Fe1-N7 2.220(4) Fe1-N6 2.167(2) Fe1-N3 2.190(2) Fe1-N11' 2.234(4) Fe1-N13 2.151(3) Fe1-N4 2.152(2) Fe1-N12' 2.215(4) O1W-Fe1-N1 88.45(10) O2-Fe1-O1W 88.07(8) Cl1-Fe1-N1 94.7(1) O1W-Fe1-N3 90.87(10) O2-Fe1-N3 85.57(7) Cl1-Fe1-N4 97.2(1) O1W-Fe1-N4 171.34(10) O2-Fe1-N4 93.25(7) Cl1-Fe1-N7 169.2(1) O1W-Fe1-N6 93.50(10) O2-Fe1-N1 89.50(8) Cl1-Fe1-N11' 93.1(1) O1W-Fe1-N13 83.19(10) O2-Fe1-N6 176.83(7) Cl1-Fe1-N12' 92.8(1) N1-Fe1-N3 84.96(9) O1W-Fe1-N1 92.42(8) N1-Fe1-N4 93.1(2) N1-Fe1-N4 100.00(9) N1-Fe1-N4 177.02(8) N1-Fe1-N7 76.6(1) N1-Fe1-N6 94.10(9) N3-Fe1-N6 97.54(7) N4-Fe1-N7 77.1(2) N1-Fe1-N13 171.01(10) N3-Fe1-N4 92.02(8) N1-Fe1-N11' 171.9(2) N3-Fe1-N4 91.75(9) N6-Fe1-N4 86.02(8) N1-Fe1-N12' 91.8(2) N3-Fe1-N6 175.51(9) N1-Fe1-N3 87.04(8) N4-Fe1-N11' 88.3(2) N3-Fe1-N13 88.47(9) O1W-Fe1-N3 173.62(8) N4-Fe1-N12' 168.4(2) N4-Fe1-N6 84.08(9) Fe···Fe^a 9.2593 N7-Fe1-N11' 95.9(1) N4-Fe1-N13 88.47(9) N7-Fe1-N12' 93.9(2)

N6-Fe1-N13 89.74(10) N11-Fe1-N12' 85.5(2)

Fe···Fe^a 9.232 Fe···Fe^b 8.324(1)

Symmetry operation a: [1−x,1−y,1−z]; b: [-x, 1-y, 1-z].

6.4 Magnetochemistry.

The temperature dependence of the magnetisation for compounds 6-8 has been studied on polycrystalline samples in the 2-350 K range. These χmT vs T plots are depicted in Figures 6.3 and 6.4, for 6, 7 and 8, respectively, where χm symbolises the molar paramagnetic susceptibility.

Figure 6.2. Labelled ORTEP representation at the 30% probability level of the cation of [Fe2(dpyatriz)2Cl2](CF3SO3)2 (8). Hydrogen atoms are left out for clarity. Only unique non- carbon atoms are labelled.

For compound 6 (Figure 6.3), the χmT product at 340 K amounts to 6.27 cm³ K mol⁻¹, which is in the range expected for two non-interacting high spin (HS, S = 2) Fe^II centres (6 cm³ K mol⁻¹, for g = 2). As the temperature decreases, the χmT value starts immediately to drop gradually down to 2.2 cm³ K mol^-1 in the vicinity of 200 K, from where it remains constant upon further cooling. This is consistent with a process of spin-crossover (SCO) by a fraction of the Fe^II centres present in the material (calculated to be 63%, see below), from HS to low spin (LS, S = 0). At very low temperatures, a further decrease of χmT is observed, resulting either from zero field splitting (ZFS) of the remaining Fe^II HS centres and/or from antiferromagnetic coupling of the remaining HS-HS pairs. Thus, compound 6 exhibits a gradual and incomplete SCO, centred at about 265 K, but does not exhibit a plateau at χmT ≈ 3.0 (50% of Fe^II centres LS), which is often observed for Fe^II dimers undergoing a spin transition.^{15, 16} This behaviour indicates a lack of strong intermolecular interactions, which would stabilise domains of HS-LS pairs.¹⁷

In order to present a quantitative description of the spin-transition process, Mössbauer spectra of 6 (Figure 6.3) have been collected at 293 K (near the transition temperature) and at 77 K. Signals from both, the HS and LS states are detected at both temperatures. From these experiments, the fraction of Fe^II centres in the HS state at 77 K has been calculated to be 37%.

6 at 77 K presents a HS doublet with I.S. = 1.46 mms^-1, Q.S. = 3.45 mms^-1 and relative intensity = 37% and a LS singlet with I.S. = 0.77 mms^-1, Q.S.= 0.26 mms^-1 and relative intensity = 63%.

The temperature dependence of the HS fraction deduced from susceptibility and Mössbauer data has been fitted to an equation resulting from a regular solution model (Eq.

6.3),¹⁸ where γHS is the variable HS fraction, Γ is the interaction parameter,² and the other terms having their usual meaning. It has to be noted that a normalisation of the data has been performed to take into account the residual fraction of 37% present at 77 K, as measured by Mössbauer spectroscopy. Such a high residual fraction, although often ascribed to defects, may arise from a thermal equilibrium between states of the same energy (∆H = 0) but with different degeneracies, and thus different degrees of freedom.

Figure 6.3. (a) Plot of experimental χmT vs T for compound 6 (circles) collected with an applied field of 1000 Oe, and the magnetic behaviour of compound 6a (squares, see text for details). The solid line is a fit to the regular solution theory model (see text for details), expressed as the HS fraction γHS. (b) Mössbauer spectra of 6 at 77 and at 293 K. The dashed lines represent subspectra used in the fitting procedure, while full lines are the overall fits to the experimental data.

Even though it represents a crude approximation, the calculated values are normalised to the fraction of species undergoing spin-crossover.

Ln 1−γ_HS γ_HS

⎛

⎝ ⎜ ⎞

⎠ ⎟ =∆H + Γ 1− 2

(

)

RT −∆S

R (6.3) The best fit provides (Fig. 6.3, solid line) a value of Γ = 2.9(2) kJ mol⁻¹ with ∆H = 11.0(1) kJ mol⁻¹, and ∆S = 40(1) J mol⁻¹K⁻¹. The cooperativity parameter Γ is in agreement with the absence of strong interactions between the Fe^II centres.² The entropy variation deduced is about three times of that expected from the change in spin manifold (Rln5 = 13.4 J mol⁻¹K⁻¹), indicating, as usually observed, the presence of a vibrational contribution to the entropy variation. Both, the transition entropy and enthalpy derived here are in agreement with the calorimetric characterisation of weakly cooperative SCO systems.^19-22 The magnetic

behaviour, mentioned above, is only retained if the compound is kept sealed and/or in contact with acetonitrile. If complex 6 is left exposed to air, the CH3CN molecules are most likely replaced by atmospheric water (see above), producing the material 6a, which remains HS throughout the whole range of temperatures.

The above results show that the polydentate ligand dpyatriz constitutes a new member of the family of polypyridyl ligands capable of producing SCO Fe^II compounds in the presence of appropriate co-ligands (generation of the suitable crystal field around the Fe^II centres).^23-25 The dinuclear system described here is unusual, since most of the examples reported so far are mononuclear complexes. In addition, this diiron unit represents a potential building block for the preparation of extended structures, a very attractive prospect as cases of high-dimensional SCO systems are still scarce.

Figure 6.4. (a) Linear plot of the experimental χmT vs T for compound 7, and its semi- logarithmic form (inset). The applied field was 1000 Oe. The full line corresponds to the best fit to a theoretical expression contemplating ZFS of Fe^II ions (see text for details). (b) Plot of the experimental χmT vs T for compound 8, and its semi-logarithmic form (inset). The applied field was 1000 Oe. The full lines are best fits to the corresponding theoretical expression (see text for details).

In contrast to complex 6, the χmT product of 7 (Figure 6.4) remains practically constant at approximately 8.3 cm³ K mol⁻¹ from room temperature to 50 K, where it starts to decrease to reach a value of 2.65 cm³ K mol⁻¹ at 2 K. This drop can be ascribed either to antiferromagnetic exchange coupling within the dimeric units in 7, and/or to single-ion ZFS of the Fe^II centres (S = 2). The absence of a maximum in the χm vs T plot indicates that the exchange coupling is very weak. In such situations, both effects (if present) are approximately of the same order of magnitude, and cannot therefore be separated. Consequently, the experimental data have been reproduced considering either the exchange coupling (Eq. 6.4 and 6.5)¹⁸ or the axial ZFS of individual Fe^II ions (Eq. 6.6 and 6.7),¹⁸ including, in both cases, a temperature independent paramagnetism (TIP).

(

)

ˆ 2J S S

H =− ⋅ (6.4)

x x x x

x x

x x

B A

m e e e e

e e

e e

T k

g N

20 12 6 2

20 12

6 2

2 2

9 7 5 3 1

60 28

10 2

+ + + +

+ +

= β +

χ

(6.5)

( )

⎢⎣⎡ − +

×

= 1

3 ˆ 1

ˆ 2 D S² S S

H _z

(6.6)

( ) ( ( ) ( ) )

⎥⎦

⎢ ⎤

⎣

⎡

+ +

− +

− +

× +

= ^{−x − x} _−x ^−x _{− x} ^{−x − x}

B m A

e e

x e e x e e

e T

k g N

4

4 4

2 2

2 2 1

3 / 4

/ 1 6 3 2 8

2 3

2 β 1

χ

(6.7)

In these expressions, D represents the single-ion ZFS parameter, J is the exchange coupling constant between Fe^II ions within the dinuclear entities, while x is either J/kBT (Eq.

6.5) or D/kBT (Eq. 6.7). The experimental data are indeed reproduced correctly with both models, yielding on the one hand g = 2.32(1), J/kB = −0.54(1) K (J = −0.37 cm⁻¹) and TIP = 1.6×10⁻³ cm³ K mol^-1, and on the other hand g = 2.41(1) and D/kB = 5.71(6) K (D = 3.97 cm⁻¹), with TIP = 1.17×10⁻³ cm³ K mol⁻¹. The latter fitting expression leads to a better value (χ² of 0.1 compared to 0.5), indicating that, although both effects are likely to be present, the ZFS is probably the dominant one. This value is the sum of the squared errors between the original data and those calculated from the fitting curve (not to be confused with the magnetic susceptibility). In general, a low χ² value reflects a good fitting of the experimental data. The equation used to calculate χ² is

χ²= y_i− x_i σ_i

⎛

⎝ ⎜ ⎞

⎠ ⎟

2

i

∑

The fact that the Fe^II centres in 7 are at best very weakly coupled (or uncoupled) is consistent with the long distance separating them (see Section 6.3). This lack of interaction is reflected by the absence of magnetic super-exchange pathway connecting the two metallic centres. On the other hand, positive values of D are often found in Fe^II HS species.²⁶

The magnetic properties of complex 8 differ dramatically from those of complexes 6 and 7 (Fig. 6.4 (b)). The value of χmT at 300 K is 7.05 cm³ K mol⁻¹, which is close to the one expected for two uncoupled Fe^II HS centres, and remains practically constant down to approximately 50 K. From that temperature downwards, χmT starts to increase upon cooling to reach a maximum of 7.85 cm³ K mol⁻¹ at T = 10 K. Then, the χmT value suddenly drops to 4.74 cm³ K mol⁻¹ at 2 K. This behaviour is consistent with the presence of ferromagnetic exchange between the Fe centres of the dinuclear complex, resulting in a spin ground state of S = 4. The decrease of χmT at lower temperatures is most likely due to ZFS and not to interactions between the complexes, since no intermolecular exchange pathways can be

envisaged from the crystal structure of 8. In order to quantify this behaviour, the experimental data have been fitted using a χm = f(T) expression (Eq. 6.8) derived from the Van Vleck Equation for two exchange Fe^II centres (S = 2), including a term D for the ZFS of the spin ground state (S = 4) of the dinuclear unit.²⁷

χ_m = N_Ag²β² k_BT

6e^17x+ 24e^8x+ 54e^−7x+ 96e^−28x+ 84e^−8y+ 30e^{−14 y}+ 6e^−18y

e^20x+ 2e^17x+ 2e^8x+ 2e^−7x+ 2e^−28x+ 7e^−8y + 5e^{−14 y}+ 3e^−18y (6.8)

In this equation, x = D/3kBT and y = J/kBT. The best fitting of the data have been obtained with g = 2.12(1), J/kB = 0.33(1) K (J = 0.23 cm^-1), and D/kB = 0.91(2) K (D = 0.63 cm^-1).

Ferromagnetic exchange coupling between Fe^II centres is very uncommon.²⁸ Such ferromagnetic interactions may be achieved via spin-polarisation mechanism, where the electronic π cloud of species linking paramagnetic centres is polarised by the spin of the latter.

For example, the alternation in sign of the polarisation found within aromatic rings has allowed their use as linkers to induce predictably ferromagnetic or antiferromagnetic interactions between spin carriers.⁶ Complex 8 is the only one within the series of complexes herein reported, whose crystal structure reveals that the triazine rings are directly bound to the Fe^II. As a result, the weak ferromagnetic exchange between the metallic centres may occur through a feeble π−π interaction between these two polarised aromatic rings, despite the fact that they are not totally overlapping each other (Figure 6.2).

6.5 Solid-State Absorption Spectroscopy

The diffuse reflectance spectra of bulk 6 at 293 K and around 100 K both display two very broad absorption bands centred at 11000 cm⁻¹ and 18200 cm⁻¹ (see Figure 6.5). These bands are typical for HS ⁵T2→⁵E and LS ¹A1→¹T1 transitions, respectively, in six-coordinated Fe^II showing SCO.² The variation in the relative intensities of these bands with temperature is in agreement with the spin-crossover behaviour of 6, detected by magnetic susceptibility measurements. The corresponding ligand field parameter is derived as 10Dq^HS = 11000 cm⁻¹.

The field parameter is calculated as follows: the ligand field strength 10Dq^HS can be determined directly from the ligand field spectrum, using the maximum of the ⁵T2→⁵E transition. A good approximation to the ligand field strength of the low-spin state² can be obtained from the corresponding spectrum using 10Dq^LS=E(¹T1)-E(¹A1)+(E(¹T2)-E(¹T1))/4.

Unfortunately, the transition ¹A1→¹T2 is not clearly observed for 6, probably due to an overlap with the metal-to-ligand charge transfer band.

It is known,² that a SCO behaviour may be expected for HS species having their 10Dq^HS in the range 10500-12000 cm⁻¹, as observed in the present study for 6. The ⁵T2→⁵E transition for complex 7 is shifted towards higher energies, which is in agreement with its HS ground state, and its lack of spin-crossover behaviour.

Figure 6.5. Diffuse reflectance spectra of 6 at room temperature and ca. 100 K.

6.6 The use of other metals.

For comparative reasons, two new complexes isostructural to compound 8 have been synthesised from different metal triflates, i.e. Ni(CF3SO3)2·6H2O (9) and Co(CF3SO3)2·6H2O (10), and dpyatriz, using identical reaction conditions. As shown in Figure 6.6, the complexes exhibit structural similarities with the formation a dinuclear species, bridged by two dpyatriz ligands. It thus appears that the nature of the metal ion used does not strongly affect the structural features of the corresponding coordination compounds. Similarly to 8, two chloride anions complete the octahedral coordination sphere of the metal ions. Each dpyatriz ligand acts as a tris-chelate for one metal ion (two pyridine and one triazine N-atoms) and as a bis- chelate (two pyridine N-atoms) for the other one. The counterion is disordered in all three compounds. The lattice solvent molecules occupy the same position, namely with the nitrile N-atom pointing toward the centroid of the electron-deficient aromatic ring, which suggests that the solvent-triazine interactions are quite strong.

Figure 6.6. ORTEP representations of (a) [Co(dpyatriz)2Cl2](CF3SO3)2 (10) and (b) [Ni(dpyatriz)2Cl2](CF3SO3)2 (9)

The main structural characteristic of complexes 8−10, with respect to compounds 6 and 7, is the unique coordination of the triazine ring to the metal ion. This feature is rarely observed, probably due to the instability of the resulting compounds. In fact, degradation reactions of the triazine ring, via methoxylation, hydroxylation or hydrolysis, have been observed for the tpt (2,4,6-tris(2-pyridyl)-1,3,5-triazine) derivative.²⁹ Although the initial assumption for such reactions has pointed at a metal-induced angular strain as the cause of the instability,³⁰ recent experimental and theoretical work indicate that the main reason is actually the electron withdrawing effect of the coordinated metal ion, which favours a nucleophilic attack of the aromatic ring.^{29, 31, 32}

Figure 6.7. (a) Schematic representation of the triazine ligand coordinated to the metal as observed for compounds 8, 9 and 10. The figure shows the (N-C)i bond distance which shortens upon coordination to the metal ion, owing to the electron withdrawing effect of the metallic centre. Nc stands for coordinated nitrogen atom and M symbolises the metal ion. (b) benzylcyanide-triazine interaction observed for the nickel complex.

One of the structural consequences of the coordination of the triazine is the shortening of the (N-C)i bond distance in the aromatic ring (see Figure 6.7(a)). Accordingly, the electron density on the carbon atoms next to the coordinated N-triazine atom is decreased, giving rise to the observed shortening of the (N-C)i bonds. Thus, all N-C bond lengths of a non- coordinated triazine ring (compounds 6 and 7) are in the same range. However, for compounds 8−10, the (N-C)i bond distances are shorter than the other ones. The N-C bond distances for the triazine ring of the dpyatriz ligand are longer than those of the tpt ligand.²⁹ This difference may be explained by the greater electron withdrawing character of the pyridyl groups of the tpt ligand, compared to the dipyridylamino moieties of the dpyatriz ligand. As a result, tpt-based complexes are more sensitive to a nucleophilic attack, and a number of tpt

degradation reactions have been reported. For the [M(dpyatriz)2Cl2](CF3SO3)2 series, the (N- C)i bond lengths depend on the nature of the metal coordinated by the N-triazine atom. The (N-C)i bond lengths follow the order Fe < Co < Ni (Table 6.3), which would suggest that the iron derivative is the most susceptible to a nucleophilic attack. Surprisingly, the iron atom is the less electronegative of the metal ions aforementioned. Thus, the longest (N-C)i bond distance would have been anticipated for this metal, which is obviously not the case. The small difference between the electronegativities of the three metal ions investigated is such that other important factors like the spatial arrangement of the ligand, or the solvent-π interactions present for the three complexes may also play a crucial role. Indeed, in all three coordination compounds, benzylcyanide solvent molecules are π-stacked with the electron- deficient triazine ring (see Figure 6.7(b)). This type of interactions, as well as anion-π interactions, have been already reported in literature for triazine-based compounds.^{33, 34} There is no correlation between the strength of the interaction (measured via the distance separating the two interacting molecules) and the (N-C)i bond length of the triazine ring, or the ∆(N-C)bi

(i.e. the difference between the longest and the shortest N-C distance in the ring) (Table 6.3).

Table 6.3. Main bond distances of the triazine ring for the different compounds investigated.

(N-C)i is defined in Figure 6.7(a), (N-C)av is the average value of all distances, (N-C)b is the longest distance, and ∆(N-C)bi is the difference between (N-C)b and (N-C)i. N13-Centroid represents the distance between the nitrogen of the solvent molecule and the centroid of the triazine ring.

Compound 6 Compound 8 Compound 9 Compound 10 (N-C)i --- 1.303 (6) 1.330 (4) 1.323 (4) (N-C)av 1.333 1.330 (7) 1.337 (5) 1.337 (4) (N-C)b 1.3366 (29) 1.355 (6) 1.338 (4) 1.348 (3)

∆(N-C)bi 0.0075 0.052 (6) 0.007 (4) 0.024 (4) N13-Centroid --- 3.360 3.440 3.524

The coordination of the triazine ring results in slightly different N-C distances, but these modifications of the bond lengths do not appear to be drastic enough to generate instability of the compound. Actually, the major resulting effect on compound 8, with respect to 6 and 7, is the change of magnetism. Compound 8 exhibits ferromagnetic properties, while compound 6 is a spin-transition material, and compound 7 shows antiferromagnetism.

Surprisingly, compounds 9 and 10, both display antiferromagnetic interactions between the metal centres, with an S = 0 ground state. As shown in Figure 6.8, the χmT vs T plot characterises a magnetic behaviour for a very weakly interacting dinuclear Ni^II species. The χmT value at room temperature is 2.01 cm³mol⁻¹K, which is close to the value expected for 2 Ni^II ions. As the temperature decreases, χmT remains constant until T = 50 K, where a decrease is observed. This feature indicates weak antiferromagnetic interactions or/and zero field splitting of the Ni^II centres. Similarly to compound 7, the experimental data is satisfactorily fitted to both models (see Section 6.4)^{18, 27} giving the following values: g = 1.99, J/kB = −0.546(1) K (J = −0.382(5) cm⁻¹), and TIP = 6×10⁻⁶ cm³ K mol⁻¹, and on the other

hand, g = 1.99 and D/kB = -0.648 K (D = −0.4505 (88) cm⁻¹), with TIP = 6×10⁻⁶ cm³ K mol⁻¹. In the present case, the former fitting is slightly better, but still both interactions must be considered for the decrease in χmT observed at low temperatures.

(6.9)

(6.10)

Similarly to the nickel(II) derivative, the cobalt(II) coordination compound shows an antiferromagnetic behaviour as well between the cobalt centres of the dinuclear unit. The 3.73 cm³mol⁻¹K χmT value at room temperature is close to the one expected for a dicobalt species.

In this case, the decrease of χmT starts at a higher temperature due to the spin-orbital coupling, which is known to be strong for cobalt-based complexes.²⁷

Figure 6.8. (a) χmT vs T plot for [Ni(dpyatriz)2Cl2](CF3SO3)2 (9). The line is the fitting to a dinuclear model of antiferromagnetic interactions. (b) Observed χmT vs T plot for [Co(dpyatriz)2Cl2](CF3SO3)2 (10).

In summary, the ferromagnetic behaviour observed for compound 8 is assumed to originate from the polarisation of the triazine rings as a result of their direct coordination to the metal centres, and to their interactions via π-π stacking. Considering this assumption, both the polarisation of the ring itself together with the π-π interaction are responsible for the different magnetic behaviour observed for 8, apart, obviously, from the magnetic orbitals involved. The ring polarisation reflects the type of interaction encountered; the lack of polarisation results in a non-interacting, or weakly antiferromagnetic system, like compound 7. The strong or weak character of the π-π interactions most likely affects the strength of the magnetic interactions. Assuming that the ring polarisation is reflected in the N–C distances of the triazine ring a plausible explanation for the different magnetic behaviour can be derived.

The comparison between the N-C distances of the triazine rings of compounds 8−10 (to quantify the degree of polarisation) reveals that only the Fe^II derivative exhibits a polarised triazine ring. Both the cobalt and especially the nickel complexes show comparable distance

(

)

ˆ 2J S S

H =− ⋅

( )

⎢⎣⎡ − +

×

= 1

3 ˆ 1

ˆ 2 D S² S S

H _z

values for all N-C bonds constituting the triazine ring, which indicates a non-polarisation of this aromatic ring. The non-polarised ring apparently hinders the communication between the metal centres, preventing the occurrence of ferromagnetic interactions. A through-ligand pathway is therefore the only manner in which the centres may communicate, as observed in the case of compound 7. This interaction mechanism not only correlates with the magnetic behaviours observed for compounds 8−10, but also with those characterizing compounds 6 and 7.

6.7 Conclusion Remarks

The reactivity of the polydentate ligand 2,4,6-tris(dipyridin-2-ylamino)-1,3,5-triazine (dpyatriz) with various Fe^II sources has been investigated for the first time. Small variations in reaction conditions have led to a new family of dinuclear complexes showing a great diversity in their magnetic behaviour. Thus, the simple exchange of solvent, from CH3CN to CH3OH gives rise to a drastic modification of the magnetic properties, from a SCO behaviour for [Fe2(dpyatriz)2(H2O)2(CH3CN)2](ClO4)4 (6) to weak antiferromagnetic coupling for [Fe2(dpyatriz)2(H2O)2(CH3OH)2](BF4)4 (7). This effect is obviously caused by the difference in crystal field produced by the coordination of the distinct solvent molecules, which, in both cases act as terminal ligands to the iron(II) centres. When a coordinating anion is present during the reaction, for instance Cl⁻, the resulting complex [Fe2(dpyatriz)2Cl2](CF3SO3)2 (8) is obtained. The crystal structure of 8 differs significantly from those of the coordination compounds 6 and 7, since the Fe^II ions are coordinated by N-atoms of the triazine ring. This participation of the triazine ring within the coordination environment of the metal most likely leads to a (very rare) ferromagnetic exchange between the irons. Accordingly, the magneto- structural comparison of compounds 8, with the related Ni^II and Co^II species reveals that the polarisation of the triazine ring, arising from its coordination to the metal ion, may explain the peculiar magnetic behaviour of compound 8. However, DFT calculations should be undertaken to establish the details of nature of the magnetic orbitals participating to this super-exchange, and to rationalise the coupling observed. Nevertheless, these results suggest that the fine-tuning of the physical properties of the dinuclear system may be easily achieved through a careful choice of the terminal ligands completing the basal plane of the two iron octahedrally based coordination environments. In addition, the prospect to design and prepare polymeric SCO systems can be envisaged through the use of appropriate dimetallic moieties as a building blocks (see chapter 8).

6.8 References

1. Miller, J. S.; Drillon, M., Magnetism: Molecules to Materials V. Wiley-VCH:

Weinheim, 2005.

2. Gütlich, P.; Hauser, A.; Spiering, H., Angew. Chem. Int. Ed. Engl. 1994, 33, 2024.

3. van Koningsbruggen, P. J., Special classes of iron(II) azole spin crossover compounds.

In Spin Crossover in Transition Metal Compounds I, Topics in Current Chemistry, Gütlich, P.; Goodwin, H. A., Eds. Springer: 2004; Vol. 233, pp 123-149.

4. Boudalis, A. K.; Donnadieu, B.; Nastopoulos, V.; Modesto Clemente-Juan, J.; Mari, A.; Sanakis, Y.; Tuchagues, J.-P.; Perlepes, S. P., Angew. Chem. Int. Edit. 2004, 43, 2266-2270.

5. Karmakar, T. K.; Ghosh, B. K.; Usman, A.; Fun, H.-K.; Rivière, E.; Mallah, T.;

Aromí, G.; Chandra, S. K., Inorg. Chem. 2005, 44, 2391-2399.

6. Lloret, F.; De Munno, G.; Julve, M.; Cane, J.; Ruiz, R.; Caneschi, A., Angew. Chem.

Int. Ed. Engl. 1998, 37, 135-138.

7. de Hoog, P.; Gamez, P.; Driessen, W. L.; Reedijk, J., Tetrahedron Lett. 2002, 43, 6783-6786.

8. Gamez, P.; de Hoog, P.; Roubeau, O.; Lutz, M.; Driessen, W. L.; Spek, A. L.; Reedijk, J., Chem. Commun. 2002, 1488-1489.

9. Demeshko, S.; Dechert, S.; Meyer, F., J. Am. Chem. Soc. 2004, 126, 4508-4509.

10. Gamez, P.; de Hoog, P.; Lutz, M.; Driessen, W. L.; Spek, A. L.; Reedijk, J., Polyhedron 2003, 22, 205-210.

11. Casellas, H.; Massera, C.; Gamez, P.; Manotti-Lanfredi, A. M.; Reedijk, J., Eur. J.

Inorg. Chem. 2005, 2902-2908.

12. Aromí, G.; Roubeau, O.; Helliwell, M.; Teat, S. J.; Winpenny, R. E. P., Dalton Trans.

2003, 3436-3442.

13. Machkour, A.; Mandon, D.; Lachkar, M.; Welter, R., Inorg. Chem. 2004, 43, 1545- 1550.

14. Dunbar, K. R.; Quillevere, A., Angew. Chem. Int. Ed. 1993, 32, 293-295.

15. Gaspar, A. B.; Ksenofontov, V.; Real, J. A.; Gütlich, P., Chem. Phys. Lett. 2003, 373, 385-391.

16. Real, J. A.; Bolvin, H.; Bousseksou, A.; Dworkin, A.; Kahn, O.; Varret, F.;

Zarembowitch, J., J. Am. Chem. Soc. 1992, 114, 4650-4658.

17. Real, J. A.; Gaspar, A. B.; Muñoz, M. C.; Gütlich, P.; Ksenofontov, V.; Spiering, H., Bipyrimidine-bridged dinuclear iron(II) spin crossover compounds. In Spin Crossover in Transition Metal Compounds I, Topics in Current Chemistry, Gütlich, P.; Goodwin, H. A., Eds. Springer: 2004; Vol. 233, pp 167-193.

18. O'Connor, C. J., Prog. Inorg. Chem. 1982, 29, 203-283.

19. Jakobi, R.; Romstedt, H.; Spiering, H.; Gütlich, P., Angew. Chem. Int. Ed. Engl. 1992, 31, 178-180.

20. Nakamoto, T.; Tan, Z. C.; Sorai, M., Inorg. Chem. 2001, 40, 3805.

21. Roubeau, O.; de Vos, M.; Stassen, A. F.; Burriel, R.; Haasnoot, J. G.; Reedijk, J., J.

Phys. Chem. Solids 2003, 64, 1003-1013.

22. Sorai, M.; Maeda, Y.; Oshio, H., J. Phys. Chem. Solids 1990, 51, 941-951.

23. Moliner, N.; Muñoz, M. C.; van Koningsbruggen, P. J.; Real, J. A., Inorg. Chim. Acta 1998, 274, 1-6.

24. Sugiyarto, K. H.; Craig, D. C.; Rae, A. D.; Goodwin, H. A., Aust. J. Chem. 1994, 47, 869-890.

25. Toftlund, H., Coord. Chem. Rev. 1989, 94, 67-108.

26. Carlin, R. L., Magnetochemistry. Springer-Verlag: Berlin, 1985.

27. Kahn, O., Molecular Magnetism. In Molecular Magnetism, VCH: New York, 1993; p 131.

28. Oshio, H.; Hoshino, N.; Ito, T.; Nakano, M., J. Am. Chem. Soc. 2004, 126, 8805-8812.

29. Zhou, X. P.; Li, D.; Zheng, S. L.; Zhang, X. J.; Wu, T., Inorg. Chem. 2006, 45, 7119- 7125.

30. Lerner, E. I.; Lippard, S. J., J. Am. Chem. Soc. 1976, 98, 5397-5398.

31. Paul, P.; Tyagi, B.; Bhadbhade, M. M.; Suresh, E., J. Chem. Soc. Dalton Trans. 1997, 2273-2277.

32. Paul, P.; Tyagi, B.; Bilakhiya, A. K.; Bhadbhade, M. M.; Suresh, E.; Ramachandraiah, G., Inorg. Chem. 1998, 37, 5733-5742.

33. de Hoog, P.; Gamez, P.; Mutikainen, H.; Turpeinen, U.; Reedijk, J., Angew. Chem. Int.

Ed. 2004, 43, 5815-5817.

34. Mooibroek, T. J.; Teat, S. J.; Massera, C.; Gamez, P.; Reedijk, J., Cryst. Growth Des.

2006, 6, 1569-1574.