Synthesis of 4-hydroxy-2,6-di(pyrazol-1-yl)pyridine, and the Spin State Behaviour of Its Iron(ii) Complex Salts

Treatment of 4-hydroxy-2,6-dibromopyridine with four equivalent of deprotonated pyrazole in hot diglyme affords 4-hydroxy-2,6-di(pyrazol-1-yl)pyridine (L) in low yield. The three complex salts [FeL2]X2 (X − = PF6 − , 3) have been prepared, and crystallographically characterised as their methanol solvates. The solvate structures contain complexes that are fully high-spin (1 and 3), or in a mixed high:low spin state population at 150 K (2). Bulk samples of 1 and 2 obtained from methanol/diethyl ether contain a second, minor crystal phase that exhibits an abrupt spin-transition near 200 K. Recrystallisation of 1 and 2 from nitromethane/diethyl ether affords powder samples that are highly enriched in this spin-transition phase.


Synthesis
Following the usual protocol for the synthesis of bpp derivatives [20], treatment of 2,6-dibromo-4hydroxypyridine [21] with 4 equiv.K[pz] in hot diglyme for 3 days afforded L in 11% yield after purification by silica column chromatography (Scheme 2).While the yield of L is low, it was sufficient to explore its coordination chemistry.Despite the excess of pyrazole used, a significant quantity of the monosubstituted by-product 2-bromo-4-hydroxy-6-(pyrazol-1-yl)pyridine (L′, Scheme 2) was also isolated from the reaction, implying that the second substitution step in the reaction is sluggish [22,23].An attempt to obtain improved yields of L by THP-protection of the 2,6-dibromo-4-hydroxy-pyridine hydroxyl function (THP = tetrahydropyran-2-yl [24]) was unsuccessful, since displacement of the protected O-THP group occurred during the pyrazole coupling step.Although we did not obtain its crystal structure, NMR data in (CD3)2SO demonstrate that L adopts the hydroxypyridine tautomer shown in Scheme 2, rather than the alternative pyridone form preferred by terpyOH in condensed phases [25].Evidence includes the OH proton resonance at 11.5 ppm (the NH proton peak from a pyridone tautomer would come near 7 ppm); and, the pyridyl C4 chemical shift of 168.4 ppm (the pyridone tautomer would have a C=O group at this position, resonating near 180 ppm) [25].

Crystallographic Characterisation
Methanol solvate crystals of 1-3 were obtained by slow crystallisation from MeOH/Et2O, and characterised at low temperature (100 or 150 K).Crystals of 2•1.75MeOH•0.25H2Oand 3•2MeOH are isostructural (orthorhombic, Pccn, Z = 8) and, while 1•2MeOH (monoclinic, C2/c, Z = 4) is not isostructural with the other two salts it shows several similarities with them.All the crystals contain the expected six-coordinate [FeL2] 2+ cation, which is C2-symmetric in 1•2MeOH but has no crystallographically-imposed symmetry in the other two salts (Figures 1 and 2).The metric parameters in the structures show that the iron centres in 1•2MeOH and 3•2MeOH are fully high-spin at the temperature of measurement (Table 1).That is most evident in the parameters Σ and Θ, which are calculated from the N-Fe-N bond angles in the complex and are particularly sensitive to the spin state population of the iron centre [26,27].In contrast Σ, Θ and the Fe-N distances in 2•1.75MeOH•0.25H2Oare somewhat lower than in the other two structures, which implies that [FeL2] 2+ contains a mixture of high-spin and low-spin molecules at 150 K.However, in view of the magnetic data described below, it is unclear whether the spin-state population of 2•1.75MeOH•0.25H2Oat 150 K is temperature-independent, or indicative of thermal spin-crossover.Table 1.Selected bond distances (Å) and angular parameters (°) for the crystal structures in this work.α, Σ and Θ are indices characteristic for the spin state of the complex [26,27], while θ and ϕ are measures of the Jahn-Teller distortion sometimes shown by these iron centers in their high-spin state [28].Typical values of these parameters in [Fe(bpp)2] 2+ derivatives are given in ref. [1].The cations in all three structures deviate significantly from the idealised D2d symmetry expected for a six-coordinate complex with this ligand combination.This structural distortion is often observed in high-spin [Fe(bpp)2] 2+ derivatives [3], and is caused by a Jahn-Teller distortion of the orbitally degenerate high-spin iron(II) configuration [28].The distorted structures can be quantified by two more parameters, ϕ and θ, that describe the relative dispositions of the L ligands in the molecule [1,28].The most noteworthy aspect of 1•2MeOH-3•2MeOH is their reduced values of ϕ, the trans-N{pyridyl}-Fe-N{pyridyl} angle, which range between 165.55(15) ≤ ϕ ≤ 169.06(7)° in the three structures.For comparison, a complex with ideal D2d symmetry would show ϕ = 180° [28].We have previously proposed that complexes with ϕ < 172° should remain high-spin at all temperatures [1], which is consistent with the high-spin nature of 1•2MeOH and 3•2MeOH.However the activity of 2•1.75MeOH•0.25H2Otowards spin-crossover is unexpected in that regard; although this structure has a slightly higher value of ϕ than 1 and 3 (Table 1), it is still lower than our proposed threshold value.
In each structure, the hydroxyl groups in the complex form O-H…O hydrogen bonds to methanol solvent molecules, which in turn donate a O-H…X (X = F or O) hydrogen bond to a neighbouring anion.In 1•2MeOH, this leads to discrete {[FeL2][MeOH]2[BF4]2} hydrogen bond assemblies in the lattice (Figure 1).In contrast, in 2•1.75MeOH•0.25H2Oand 3•2MeOH one anion lying on a crystallographic inversion centre accepts hydrogen bonds from two different methanol molecules, leading to {([FeL2][MeOH]2X)2(μ-X)} + (X − = ClO4 − or PF6 − ) assembly structures (Figure 2).In each case, the hydrogen-bond assemblies in the lattice only interact with their neighbours through weak van der Waals contacts.The hydrogen bonded anions and solvent in 2•1.75MeOH•0.25H2Osuffer from disorder, but are crystallographically ordered in isostructural 3•2MeOH.That may also influence the different spin-state properties of those crystals.

Characterisation of the Bulk Materials
The lattice solvent is lost from all the methanol solvates upon exposure to air, yielding solvent-free powders by microanalysis (referred to as 1 m -3 m ).While 3 m is high-spin at all temperatures, magnetic susceptibility data imply that bulk samples of 1 m and 2 m are not phase-pure.In both samples, 15%-25% of the solid exhibits an abrupt spin-transition at 210-220 K, with the remaining material being fully high-spin (1 m ) or low-spin (2 m ; Figure 3).The approximate 1:3 high-spin:low-spin population in 2 m at room temperature, predicted by the magnetic data, is consistent with its colouration which is noticably darker than for 1 m or 3 m .However, the low-spin nature of 2 m at 150 K from these data contrasts with the crystal structure of the parent solvate 2•1.75MeOH•0.25H2Owhich is ca.65% high-spin at that temperature.We conclude that the spin-state behaviour of 3 m and the main component of 1 m is unaffected by desolvation of the parent methanol solvate materials.However, removal of the solvent from 2•1.75MeOH•0.25H2O,to form 2 m , has a strong effect on the properties of the complex.Changes of spin state induced by removal of lattice solvent are well known in the literature [8,29,30].The partial, abrupt spin-transitions shown by bulk samples of 1 m and 2 m are probably associated with a second crystal phase of the complexes.Consistent with that suggestion, recrystallisation of 1 and 2 from nitromethane/diethyl ether afforded new solvent-free powders 1 n and 2 n , which now have the same pale yellow colouration.These exhibit abrupt spin-transitions at the same temperatures as the minor phases in the methanolic samples (Figure 3).The transition in 1 n exhibits T½ = 211 K and is 78% complete at 150 K, suggesting there is still a minor fraction of a high-spin phase in this new sample.
In contrast, the transition in 2 n is centred at T½ = 207 K and proceeds to completeness, implying that sample is phase-pure.Both spin transitions exhibit a narrow 1-2 K thermal hysteresis in the SQUID magnetometer at a scan rate of 2 Kmin −1 , which we have also observed in a number of other compounds from the [Fe(bpp)2] 2+ family [31,32].
These observations were confirmed by DSC measurements (Figure 4), which afforded thermodynamic parameters for the transitions that are similar to those shown by the parent compound [Fe(bpp)2][BF4]2 [28].For 1 n : T½↓ = 209 K, T½↑ = 215 K, ΔH = 22.4 kJmol −1 and ΔS = 106 Jmol -1 K -1 (corrected for the incompleteness of the transition).For 2 n : T½↓ = 205 K, T½↑ = 210 K, ΔH = 21.7 kJmol −1 and ΔS = 105 Jmol −1 K −1 .The wider thermal hysteresis in these data, compared to the magnetic susceptibility measurements, is probably a consequence of the more rapid thermal scan rate of 10 Kmin −1 that was used in the DSC experiments [33].The errors on the measured thermodynamic parameters will also be relatively large, because the transitions lie near the lower temperature limit for our calorimeter (Figure 4).X-ray powder diffraction data at 298 K showed that the spin-transition phases of 1 n and 2 n are isostructural, and unrelated to the methanol solvate structures (Figure 5).Since there is no apparent crystalline contaminant in the powder pattern of 1 n , the residual high-spin fraction of this sample may arise from amorphous material.The isolation of the spin-transition phases from two different solvent mixtures implies they should be solvent-free forms of the complexes.However, that remains to be confirmed since we have been unable to obtain these phases as single crystals.

Instrumentation
Elemental microanalyses were performed by the University of Leeds School of Chemistry microanalytical service. 1 H-NMR spectra employed a Bruker (Coventry, UK) DPX300 spectrometer operating at 300.2 MHz ( 1 H) or 75.5 MHz ( 13 C).Electrospray mass spectra (ESI MS) were obtained on a Waters (Elstree, UK) ZQ4000 spectrometer, from MeCN feed solutions.All mass peaks have the correct isotopic distributions for the proposed assignments.X-ray powder diffraction patterns were measured using a Bruker (Coventry, UK) D2 Phaser diffractometer.DSC measurements used a TA (Elstree, UK) Instruments DSC Q20 calorimeter, heating at a rate of 10 Kmin -1 .Solid state magnetic susceptibility measurements were performed on a Quantum Design (Leatherhead, UK) SQUID or SQUID/VSM magnetometer, in an applied field of 1000 or 5000 G with a scan rate of 2 K•min -1 .A diamagnetic correction for the sample was estimated from Pascal's constants [34], and a previously measured diamagnetic correction for the sample holder was applied.

Synthesis of the Complexes
The same basic method, described below for [FeL2][BF4]2, was used for all the complexes in this work.A solution of L (0.20 g, 0.88 mmol) and Fe[BF4]2•6H2O (0.15 g, 0.44 mmol) in nitromethane (15 cm 3 ) was refluxed until all the solid had dissolved (ca. 2 h).The cooled solution was concentrated in vacuo to ca. 5 cm 3 .Slow diffusion of diethyl ether vapour into the filtered solution afforded yellow microcrystals of the product.The other complex salts were prepared using analogous reactions, with appropriate amounts of Fe[ClO4]2•6H2O or Fe[PF6]2, as required (Fe[PF6]2 was prepared in situ by treating FeCl2•4H2O with 2 equiv.AgPF6.The precipitated AgCl was removed by filtration before addition of the L ligand).Yields ranged from 34%-75%. For

Crystal Structure Determinations
The single crystals were grown by slow diffusion of diethyl ether vapour into methanol solutions of the complexes.Diffraction data for 1•2MeOH and 2•1.75MeOH•0.25H2Owere measured using a Bruker (Coventry, UK) X8 Apex diffractometer, with graphite-monochromated Mo-Kα radiation (λ = 0.71073 Å) generated by a rotating anode.Data for 3•2MeOH were collected with an Agilent (Stockport, UK) Supernova dual-source diffractometer using monochromated Cu-Kα radiation (λ = 1.54184Å).Experimental details of the structures determinations in this study are given in Table 2.All the structures were solved by direct methods (SHELXS97 [35]), and developed by full least-squares refinement on F 2 (SHELXL97 [35]).Crystallographic figures were prepared using XSEED [36].  ) contain the supplementary crystallographic data for this paper.These data can be obtained free of charge from The Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/data_request/cif.

Structure Refinement of 1•2MeOH
This structure was solved in P 1 , then transformed up to C2/c using the ADSYMM routine in PLATON [37].The asymmetric unit contains half a formula unit, with Fe(1) lying on the crystallographic C2 axis 0, y, 1/4.The unique BF4 − ion is disordered, and was refined over three sites with occupancies of 0.5:0.3:0.2.These three partial anions share a common wholly occupied F atom, and were modelled using the refined restraints B-F = 1.39(2) and F...F = 2.27(2) Å.All non-H atoms with occupancy ≥0.5 were refined anisotropically, while H atoms were placed in calculated positions and refined using a riding model.The highest residual Fourier peak of +1.4 eÅ −3 is 0.9 Å from Fe(1).

Structure Refinement of 2•1.75MeOH•0.25H2O
This structure was originally solved in P21/c, then transformed to Pccn using PLATON as before [37].The asymmetric unit contains one complex dication, one crystallographically ordered anion; two disordered half-anions lying on special positions; and two solvent sites.Both disordered half-anions were modelled over two orientations with a 0.3:0.2occupancy ratio.The refined restraints Cl-O = 1.46(2) and O...O = 2.38(2) Å were applied to these partial anion sites.One of the solvent sites contains a crystallographically ordered methanol molecule, but the other was refined as a mixture of methanol (refined occupancy 0.75) and water (0.25).These residues were modelled without restraints.All non-H atoms with occupancy >0.5, plus all the partial Cl atoms, were refined anisotropically.All H atoms in the complex dications and methanol molecules were placed in calculated positions and refined using a riding model.H atoms from the partial water molecule were not located, and could not be included in the final model.They are accounted for in the density and F(000) calculations, however.

Structure Refinement of 3•2MeOH
The solvated PF6 − salt is isostructural with 2•1.75MeOH•0.25H2O,but does not suffer from the same anion or solvent disorder at the temperature of measurement (100 K).No disorder was detected during refinement of this structure and no restraints were applied to the model.A residual Fourier peak of +1.4 eÅ −3 occupies the same position as the fractional water content in the perchlorate structure, but this refined to <10% occupancy and so was not included in the final model.

Conclusions
The synthesis of 4-hydroxy-2,6-di(pyrazol-1-yl)pyridine (L) has been achieved, in two steps from commercially available precursors with a low but usable yield.Compound L is the first ligand from the bpp family with an oxygen substituent at the pyridine ring (Scheme 1).Three salts of [FeL2] 2+ have been prepared, and crystallographically characterised as their methanol solvates.While two of these remain high-spin at low-temperature, the solvated perchlorate salt (2) exhibits a mixed high/low spin state population at 150 K. Notably 2•1.75MeOH•0.25H2Oalso exhibits significant anion and solvent disorder at low temperature, whereas high-spin 3•2MeOH does not.The different spin state properties of the two isostructural crystals may be related to this disorder, which would result in a less rigid lattice environment around the complex molecule in the perchlorate salt.
Bulk samples of 1 and 2 isolated from methanol/diethyl ether, 1 m and 2 m , appear to contain a second, solvent-free contaminant phase that undergoes an abrupt spin-transition near 200 K.Samples enriched in the latter phase, 1 n and 2 n , are obtained by recrystallising from nitromethane/diethyl ether which has allowed the spin-transition to be properly characterised.While a crystal structure of 1 n and 2 n was not achieved, the form of their spin transitions is reminiscent of other complexes from the [Fe(bpp)2] 2+ family, that adopt a particular type of crystal packing (the "terpyridine embrace") [31].
Our future work is aimed towards using L as a precursor to [Fe(bpp)2] 2+ derivatives bearing other oxygen-based functionalities (Scheme 1; R = OR′, O2CR′ etc. where R′ = alkyl or aryl).

Figure 1 .
Figure 1.View of the hydrogen bonded formula unit in 1•2MeOH.Atomic displacement ellipsoids are at the 50% probability level.Only one orientation of the disordered BF4 − ion is shown, and C-bound H atoms have been omitted.Colour code: C, white; H, pale grey; B, pink; F, cyan; Fe, green; O, red.Symmetry code: (i) −x, y, 1/2−z.

Figure 3 .
Figure 3. Variable temperature magnetic susceptibility data for 1 (a), 2 (b) and 3 (c).The black circles are from samples crystallised from nitromethane/diethyl ether (1 n and 2 n ), and the red diamonds are material obtained from methanol/diethyl ether (1 m , 2 m and 3 m ).

Figure 5 .
Figure 5. Experimental X-ray powder diffraction patterns (black) of the spin-crossover active phases 1 n (a) and 2 n (b), and simulations based on their methanol solvate crystal structures (red).The black traces are isostructural with each other, but not with the red traces.

Table 2 .
Experimental details for the crystal structures in this work.