Ligand-directed synthesis of { MnIII 5 } twisted bowties †

have been synthesised by using two Schiff base ligands derived from 3,5-diamino-1,2,4-triazole, following two different preparative routes, either using the pre-formed ligand (for L1) or via a metal-mediated template synthesis (for L2). The {MnIII 5 } structure is unusual, being based on two corner-sharing perpendicular {Mn3} triangles forming a twisted bow-tie. The magnetic studies reveal antiferromagnetic coupling between Mn(III) ions while electrochemical experiments are consistent with a quasi-reversible Mn(III)↔Mn(IV) redox process at the central manganese ion.


Introduction
][8] Mn(III) ions are employed in the design of nanomagnets due to the characteristic magnetic anisotropy (D), [9][10][11][12] whereas Mn(II) ions are used in molecular magnetic refrigerants, considering the relatively large number of unpaired electrons and the typical isotropic octahedral environment. 13,149][20] One of the advantages of the second approach is the degree of control in the assembly of the metal ions through the pre-design of the ligand.Schiff base ligands have been used in this type of strategy, [20][21][22] and in some cases, these ligands can provide additional redox properties to complexes as a result of the presence of iminic groups. 23Importantly, the iminic bonds make the ligand flexible enough to accommodate different coordination environments based on the type and/or valence of the metal ions.Considering these factors, we proposed the use of the 3,5-diamino-1,2,4-triazole derivatives, N-(2-hydroxy-3-methoxybenzylene)-3,5-diamino-1,2,4-triazole (H 2 L1) and N-salycilene-3,5-diamino-1,2,4-triazole (H 2 L2) to direct the synthesis of new polymetallic manganese complexes.Unlike other diamines, 3,5-diamino-1,2,4-triazole derivatives have not been explored in the synthesis, magnetic and electrochemical studies of high-nuclearity complexes.Herein we describe two different synthetic methodologies for the preparation of two new pentanuclear Mn(III) compounds, 2) and report their structural, magnetic and electrochemical properties.

Materials and physical measurements
All reagents and solvents were obtained from commercial suppliers and used without further purification.
Crystallographic data were collected for 1 and 2 at 100 K using Mo-Kα radiation (λ = 0.71073 Å).For 1 a Bruker APEXII CCD diffractometer with an Oxford Cryosystems n-Helix lowtemperature device mounted on a sealed tube generator was used; for 2 a Rigaku AFC12 goniometer equipped with an (HG) Saturn724+ detector mounted on an FR-E+ SuperBright rotating anode generator with HF Varimax optics (100 µm focus). 24oth structures were solved using SHELXT 25 and refined using full-matrix least squares refinement on F 2 using SHELX2014 26 within OLEX2. 27SQUEEZE was used to calculate and account for diffuse solvent in 2. 28 Further details of disorder modelling and hydrogen atom treatment are given in the ESI.† The powder X-ray patterns for 1 and 2 were collected on a PANalytical XPert MPD, with Cu Kα1 radiation at ambient temperature over a range of 5°< 2θ < 50°using a step size of 0.0167°.The calculated pattern was generated from Mercury using the .cif of the crystal structure at 100 K.
The IR spectra were measured using a FTIR-8400S SHIMADZU IR spectrophotometer.The microanalyses and mass spectrometry analyses were performed by the analytical services of the School of Chemistry at the University of Glasgow.The 1 H-NMR spectrum of H 2 L1 was obtained using a Bruker AVI 400 M MHz.Magnetic measurements of complexes 1 and 2 were carried out on phase-pure (Fig. S4 in the ESI †) polycrystalline samples that were lightly powdered and constrained in eicosane, using a Quantum Design MPMS-XL SQUID magnetometer.Data were corrected for the diamagnetic contribution of the sample holder and eicosane by measurements, and for the diamagnetism of the compounds.Electrochemical measurements were performed on a CH Instruments 760D Electrochemical Workstation using CHI Version 10.03 software.Electrochemical experiments for obtaining the cyclic voltammograms (CVs) were conducted at 298 K using a CH Instruments glassy carbon button working electrode (area = 0.071 cm 2 ), BASi Ag/AgNO 3 pseudo reference electrode, and Pt mesh counter electrode in a single compartment cell.Bulk electrolyses were performed in a similar manner, but using a large surface area reticulated glassy carbon electrode (BASi) as the working electrode.All electrode potentials were referenced to the ferrocene/ferrocenium couple by doping in samples of ferrocene to the electrolyte, and all potentials given in this work are reported relative to the ferrocene/ferrocenium couple.All electrochemical experiments were conducted in electrolyte solutions prepared using dry N,N-dimethylformamide (DMF), and were thoroughly degassed with argon.The supporting electrolyte was tetrabutylammonium hexafluorophosphate at a concentration of 0.2 mol dm −3 .Solutions were agitated between acquisition of individual CVs and CVs were corrected for resistance, using the iR compensation function of the potentiostat.CVs were collected from unstirred solutions and bulk electrolyses were undertaken in stirred solutions.

Results and discussion
][31][32][33] Surprisingly, there is a lack of magnetic studies performed on such complexes with other 3d ions (vide supra) such as Mn.Furthermore, the only report of the coordination chemistry of L1 and L2 is with oxovanadium(IV). 34H 2 L1 was prepared from the condensation reaction between 3,5-diamino-1,2,4-triazole and 2-hydroxy-3-methoxy-benzaldehyde (o-vanillin) in methanol.The synthetic route to H 2 L1 (Scheme 1) is simplified compared to that previously reported, 34 as we obtained pure H 2 L1 straight from the condensation without re-crystallisation (checked by 1 H NMR). 34 The reaction between H 2 L1 and Mn(CH and several solvent molecules (four water and four acetonitrile molecules for 1; one acetonitrile and two methanol molecules for 2).Note that for 2 a poorly defined solvent region was additionally treated using SQUEEZE. 28The solvent accessible voids were calculated to be 190 Å 3 containing 42 electrons per complex which can be attributed to 4 molecules of H 2 O, given the results from the elemental analysis.As both anionic complexes are isostructural, the following description is valid for 1 and 2 (see structure in Fig. S1 of the ESI †).The structure of 1 involves five Mn(III) ions arranged within two fused {Mn 3 O} units which share a central Mn(III) ion (Fig. 2).The oxidation state of each manganese centre has been confirmed by bond length and charge balance considerations, and also by bond valence sum calculations (BVS) with values presented in Table S2.† 37 The two corner-sharing {Mn 3 O} units lie in planes almost perpendicular to one another with angles of 86.84°(1) and 86.75°(2) (see Fig. S2 of the ESI †).Each Mn(III) center displays a tetragonally distorted octahedral geometry through the coordination of four doubly deprotonated Schiff base ligands (L1 2− for 1, and L2 2− for 2) and four bidentate acetate auxiliary ligands (see Fig. 2).Two different Mn(III) ions can be distinguished based on the dissimilarities of their first coordination sphere (see Fig. The outer manganese centres (Mn2-Mn5) in 1 and 2 display a characteristic elongated Jahn-Teller (JT) distortion (Fig. 3), as expected for high-spin d 4 Mn(III) ions (d Avg Mn-N = 2.298 Å (1), 2.283 Å (2); d Avg Mn-N = 2.035 Å (1), 2.032 Å (2)).The elongated JT axis around the central manganese (Mn1) however is comparatively less distinct.This is evident on considering the bond lengths around Mn2-Mn5 as internal standards, and comparing these to those of Mn1.The Mn1-N   distances (N1A-Mn1-N1B) are shorter than expected (2.162 Å (1), 2.158 Å ( 2)) and the Mn1-N distances (N1C-Mn1-N1D) are slightly longer than would be expected (2.096 Å (1), 2.074 Å ( 2)) compared to those of the outer Mn(III) ions.This indicates possible disorder of the JT axis between these two positions. 38ote, the Mn1-O1 and Mn-O2 distances are as expected on comparison to the other Mn-O distances not situated along the JT axes of Mn2-Mn5.We have therefore assumed the axis defined by O1-Mn1-O2 is not involved in the disorder.

Magnetic properties
Variable-temperature magnetic susceptibility data were measured in an applied field of 1000 Oe in the range of 290 to 2 K (see Fig. 4).Variable-field magnetisation measurements were also carried out between 0 and 5 T at 2, 4 and 6 K for each complex (Fig. S6 †).At 290 K the experimental χ M T values observed for 1 and 2 (11.5 cm 3 mol −1 K and 11.9 cm 3 mol −1 K respectively) are lower than the theoretical spin-only value for five non-interacting Mn(III) ions (15.0 cm 3 mol −1 K, considering S = 2, and g = 2).Below 290 K, the χ M T value decreases, reaching a minimum value of 0.79 cm 3 mol −1 K for 1 and 0.80 cm 3 mol −1 K for 2 at 2 K, consistent with predominant intramolecular antiferromagnetic interactions.The susceptibility data for 1 and 2 were fitted using the program Phi, 44 considering the magnetic model shown in Fig. 4 (inset) and the spin Hamiltonian shown below (eqn (1)).
Initially zero-field splitting (ZFS) was not included to avoid over-parameterisation.Two different exchange interactions were considered for complexes 1 and 2 (see magnetic model above)that between the central Mn1 and each of the outerlying manganese centres ðJ 1 Þ and the interaction between the Mn(III) centres of each peripheral pair ðJ 2 Þ.6][47] The best fit gives the coupling constants J 1 ¼ À3:10 cm À1 (1), −2.82 cm −1 (2), and J 2 ¼ À8:26 cm À1 (1), −7.43 cm −1 (2) (R = 98% and 95%).Antiferromagnetic interactions are also seen in a mixedvalence {Mn III 4 Mn II } complex with a related structure ðJ 1 ¼ À1:15 cm À1 and J 2 ¼ À2:40 cm À1 Þ. 41 Given the magnitude of the exchange interactions it is clear that axial zero-field splitting (D) should also be included, as this can be of the same order of magnitude as J . 3 However all attempts to include D lead to unreasonable values and/or prevent the fits from converging.
The simple exchange-coupling model which ignores ZFS suggests that the ground state of both 1 and 2 is S = 0 with low-lying excited states: the closest lying S = 1 state within 3 cm −1 and the closest lying S = 2 state within 6 cm −1 (Fig. S5 †).The magnetisation curves (Fig. S6 †) would then result from population of these low-lying S > 0 states.Note that inclusion of single-ion ZFS in the model (vide supra) could modify the obtained J values and hence the ordering of the spin states, which may reveal a non-zero ground state.Additional alternating-current (ac) measurements were performed, with the absence of any frequency-dependent out-ofphase signal (Fig. S7 †) consistent with either an S = 0 ground state or a small spin ground state with a negligible D parameter.Note that the Jahn-Teller axes of the four outer manganese ions are nearly perpendicular to each other (Fig. 3), which would yield a relatively small net axial magnetic anisotropy. 48

Electrochemical studies
Along with the redox properties of the metal ions, Schiff base ligands can potentially impart intriguing redox activity to their  complexes, as a result of the presence of iminic groups. 49The redox behavior of complexes 1 and 2 was therefore probed by cyclic voltammetry (CV).Fig. 5a shows the CV of compound 2 over the range +1 to −2.5 V vs. the ferrocene/ferrocenium couple (the CV of complex 1 over the same range is very similar).A CV of ligand H 2 L1 is provided in the ESI (Fig. S8 †) for comparison.Fig. 5a indicates that at least two essentially irreversible reductive redox processes that cannot be attributed to the free ligand are evident at around −0.9 V and −1.8 V.These reductive waves did not become more reversible when the scan window was restricted.However, in the oxidative portion of the voltammogram, a more reversible wave (centred around +0.25 V) was apparent.Fig. 5b shows a comparison of the CVs for this oxidative process for complexes 1 and 2 over a smaller potential window than in Fig. 5a.Again, the free ligand shows no redox waves in this region (Fig. S8 †).The shape profile was very similar in both cases, and E 1/2 values of 0.21 V and 0.23 V were obtained for complexes 1 and 2 respectively (these values are the same within the ±20 mV error quoted for the type of Ag/AgNO 3 pseudo reference employed).Taken together, and in view of the structural similarities between complexes 1 and 2, these data suggest that the redox process giving rise to this wave the same origin in both complexes.
Bulk electrolysis on complex 1 at +0.47 V indicated that the charge passed during this oxidation process was equal to 95% of that expected for a one-electron process.It is therefore plausible that this redox process corresponds to the one-electron oxidation and re-reduction of the unique, central Mn(III) centre Mn1 in these complexes given the different coordination environment displayed -{MnN 4 O 2 }compared to Mn2-5 -{MnN 2 O 4 }.The kinetics of the redox process giving rise to the wave at +0.21 V for complex 1 were probed by scanrate dependency studies.As is evident from Fig. 6, the return (cathodic) peak currents are always smaller than the oxidative currents, with this effect becoming less apparent as the scan rate is increased (i.e. the wave becomes more reversible the shorter the length of time the compound remains oxidised).This is consistent with an oxidation process followed by an irreversible chemical decomposition.Thus at fast scan rates, there is a greater chance that the complex can be reduced back to its original form before decomposition can occur.Moreover, Fig. 6 shows that even at slow scan-rates, the peak-to-peak separation is well over 100 mV, indicating that electron transfer is rather slow.This could imply that a structural re-arrangement occurs during this redox event and the associated geometrical changes required around a Mn(III) centre upon oxidation to Mn(IV) (and upon re-reduction to Mn(III)) could account for the slow electron transfer observed.It is interesting to note that the crystallographic analysis indicates possible disorder of the Jahn-Teller axis at the central Mn(III), which may make the redox event easier at this site.

Conclusions
We have shown for the first time that the 1,2,4-triazole derivatives H 2 L1 and H 2 L2 can be utilised in the synthesis of polymetallic transition complexes.This leads to two novel pentanuclear Mn(III)-based complexes with an unusual twisted bowtie structure, where the Mn(III) ions are antiferromagnetically coupled.The electrochemical studies reveal a quasi-reversible redox process: a one-electron oxidation Mn(III) → Mn(IV) and re-reduction Mn(IV) → Mn(III) which we suggest to be centred at the central Mn(III) ion in both complexes.Future work will focus on modifying the Schiff base ligands to impart  additional redox states towards redox-switchable molecular magnetic materials.

Fig. 2
Fig. 2 Structure of the anionic complex in 1. C, grey; Mn, lavender; N, blue; O, red; only Mn and N, O atoms involved in the coordination of the Mn centres are labelled.Hydrogen atoms and solvent molecules are omitted for clarity.

Fig. 3
Fig. 3 Detail of the metal core of 1.Only O, N, and metal ions involved in the Jahn-Teller (JT) axes of Mn2-Mn5 are labelled.JT axes are highlighted in lavender.

Fig. 4
Fig. 4 Temperature dependence of χ M T for 1 (shown in white) and 2 (blue).The inset shows the topology of the intramolecular magnetic interactions considered to fit 1 and 2 (see text for details).The solid lines correspond to the fit of the data.

Fig. 5
Fig. 5 (a) CV of a 2 mM solution of complex 2 at a scan rate of 100 mV s −1 over the range +1 to −2.5 V vs. ferrocene/ferrocenium.(b) Comparison between CVs of 2 mM solutions of complexes 1 and 2, both at a scan rate of 100 mV s −1 over a narrower potential window.

Fig. 6
Fig. 6 Comparison between CVs of a 4 mM solution of complex 1 at various scan rates as indicated.