Eva
Zahradníková
a,
Jean-Pascal
Sutter
*b,
Petr
Halaš
a and
Bohuslav
Drahoš
*a
aDepartment of Inorganic Chemistry, Faculty of Science, Palacký University Olomouc, 17. listopadu 12, 771 46 Olomouc, Czech Republic. E-mail: bohuslav.drahos@upol.cz; Fax: +420585 634 954; Tel: +420 585 634 429
bLaboratoire de Chimie de Coordination du CNRS (LCC-CNRS), Université de Toulouse, CNRS, Toulouse, France. E-mail: jean-pascal.sutter@lcc-toulouse.fr
First published on 15th November 2023
Large uniaxial magnetic anisotropy, expressed by a negative value of the axial zero-field splitting parameter D, has been achieved in a series of trigonal prismatic Co(II) complexes with the general formula [Co(L)X]Y, where L = 1,5,13,17,22-pentaazatricyclo[15.2.2.17,11]docosa-7,9,11(22)-triene, X = Cl−(1a,b), Br−(2), N3−(3), NCO−(4), NCS−(5), NCSe−(6), and Y = Cl−(1), Br−(2), NCS−(4), NCSe−(5), ClO4−(3,6). Complexes 1–6 are six-coordinate with the distorted trigonal prismatic geometry imparted by the pentadentate pyridine-/piperazine-based macrocyclic ligand L and by one monovalent coligand X−. Based on magnetic studies, all complexes 1–6 exhibit strong magnetic anisotropy with negative D-values ranging from about −20 to −41 cm−1. This variation in D (i.e. the increase of magnetic anisotropy) parallels the trend obtained by theoretical calculations and the lesser distortion of the coordination sphere with respect to the trigonal prismatic reference geometry. AC magnetic susceptibility investigations revealed field-induced single-molecule magnet behaviour for all complexes except Cl− derivative 1. The series investigated represents a rare example of Co(II) complexes with a robust trigonal prismatic geometry.
For transition metal ions, the characteristics of magnetic anisotropy for a dn electronic configuration are directly related to the geometry of their coordination sphere.4,5 Although remarkable anisotropy has been obtained for low-coordination complexes,5–7 some of the best SMMs exceeding U = 400 cm−1 are extremely unstable8,9 and hardly suitable to be involved in supramolecular assembly reactions. An alternative is provided by seven-coordinate pentagonal bipyramidal complexes of transition metals.5,10,11 In this geometry, Fe(II) and Ni(II) complexes possess negative D-values and have been involved in the preparation of polynuclear molecular nanomagnets.11–16 Co(II), with its large orbital angular momentum, revealed substantial but positive D-values in this coordination geometry.17–19 Hence, the design of structurally robust Co(II) complexes with negative D-value remains a stimulating but tricky target.
Six-coordinate Co(II) complexes could be an interesting alternative, not those with the common octahedral geometry, which generally provide positive D-values, but rather those with a trigonal prismatic geometry. In this environment Co(II) is expected to exhibit negative D-values,4,5 as was confirmed by previously reported examples including complexes of the Schiff base ligand,20 clathrochelates,21–23 or “tripodal-type” ligands24,25 with D-values ranging from −31 cm−1 up to −128 cm−1 (for more details see Table 4 in the Results and discussion section). However, such compounds remain scarce because it is very difficult to reach a trigonal prismatic coordination geometry which generally results from steric restraints coming from the ligand(s) occupying all coordination sites. We observed that a macrocyclic pentaaza pentadentate ligand, L (Fig. 1) led to a trigonal prismatic coordination sphere for transition metals, including Co(II) as illustrated for [Co(L)(CH3CN)](ClO4)2.26 We have further studied this trigonal prismatic system and report herein a series of Co(II) complexes based on this ligand in association with a monodentate halido/pseudohalido coligand. Our objective was to evaluate the effect of different ligand field strengths and structural modifications on the magnetic anisotropy of Co(II). Based on magnetic studies, we show that all compounds are characterized by strong axial anisotropy, with D values between −20 and −41 cm−1, in agreement with values derived from theoretical calculations. Moreover, SMM-like behaviors have been observed for the majority of the complexes.
![]() | ||
Fig. 1 Structural formula of macrocyclic ligand L together with atom numbering used for NMR assignment. |
A one channel just infusion™ linear pump NE 300 (New Era Pump Systems, Inc., Farmingdale, NY, USA) and an HSW Norm-Ject 20 ml syringe were employed for ligand synthesis (infusion rate 1 mL min−1).
1H and 13C NMR spectra were recorded at the laboratory temperature on a 400 MHz Varian NMR spectrometer for high-resolution solution-state NMR (Varian, Palo Alto, CA, USA): 1H 399.95 MHz, (CDCl3, tetramethylsilane) δ = 0.00 ppm, 13C 100.60 MHz, (CDCl3, residual solvent peak) δ = 77.0 ppm. The multiplicity of the signals was indicated as follows: s – singlet, d – doublet, t – triplet, quin – quintet, and m – multiplet. Deuterated solvent CDCl3, containing 0.03% of TMS, from Sigma Aldrich was used as received. The atom numbering scheme used for NMR data interpretation is shown in Fig. 1.
The mass spectra were collected on an LCQ Fleet mass spectrometer (Thermo Scientific, Waltham, MA, USA) equipped with an electrospray ion source and three-dimensional (3D) ion-trap detector in the positive mode.
IR spectra were recorded on a Jasco FT/IR-4700 spectrometer (Jasco, Easton, MD, USA) using the ATR technique on a diamond plate in the spectral range 4000–400 cm−1.
Elemental analysis (C, H, N) was realized on a Flash 2000 CHNO-S analyzer (Thermo Scientific, Waltham, MA, USA).
Magnetic measurements for all the samples were carried out with a Quantum Design MPMS 5S SQUID magnetometer in the temperature range of 2–300 K. The measurements were performed on polycrystalline samples mixed with grease and put in gelatine capsules. The temperature dependences of the magnetization were measured in an applied field of 1 kOe and the isothermal field dependences of the magnetization were collected up to 5 T at temperatures between 2 and 8 K. The molar susceptibility (χM) was corrected for the sample holder, grease and for the diamagnetic contribution of all the atoms using Pascal's tables.27 AC susceptibility data were collected in the frequency range 1–1500 Hz. Assessment of the ZFS parameters has been done considering an S = 3/2 spin for Co(II), and the software PHI28 was used for fitting χMT = f(T) and M = f(H) behaviors.
Co(II) complexes 1, 2, and 5 were prepared in the same way. Equimolar amounts of metal salt (0.330 mmol, i.e. 78 mg of CoCl2·6H2O or 107 mg of CoBr2·6H2O or 70 mg of Co(SCN)2·2H2O) and ligand L (100 mg; 0.330 mmol) were mixed in methanol (4 ml). The complexes were crystallized by vapour diffusion of diethyl ether into the resulting methanolic solution at 7 °C. Complex 1 obtained by the above-mentioned procedure was recrystallized either from methanol/nitromethane (4 ml, 10/1, v/v) or methanol/water/acetone solution (7.5 ml, 2/0.5/5, v/v) by diethyl ether vapour diffusion at 7 °C, yielding single-crystals suitable for X-ray analysis of complex 1a or 1b, respectively (crystallization of 1b was completed within 24 h by standing at 298 K).
Co(II) complexes 3, 4 and 6 were prepared as follows: Co(ClO4)2·6H2O (120 mg; 0.330 mmol) was dissolved in methanol (2 ml) and NaN3 (86 mg; 1.32 mmol) or NaNCO (86 mg; 1.32 mmol) or KNCSe (190 mg; 1.32 mmol) were added. The solution was filtered and the filtrate was added dropwise to a methanolic solution (2 ml) of L (100 mg; 0.330 mmol). The complexes were crystallized as mentioned above.
In order to evaluate magnetic properties, the post-Hartree–Fock method CASSCF(7,5)/NEVPT2 was utilized. The same basis sets were used for geometry optimizations with the addition of def2-TZVP/C as an auxiliary basis set40 and the chain-of-sphere integration method for Hartree–Fock exchange integrals (COSX).41Ab initio ligand field orbitals were calculated on molecules oriented according to the D-tensors.
Subsequently, a series of six complexes with Co(II) halides and pseudohalides was prepared. Complexes 1a, 1b, 2 and 5 were prepared by direct mixing of the Co(II) salt with the ligand in methanol. In the case of complexes 3, 4 and 6, the Co(II) salt had to be prepared in situ by mixing Co(II) perchlorate with NaN3, NaNCO or KNCSe, which was then added to the ligand methanolic solution. Except for complexes 1a and 1b, which required to be recrystallized from different solvent mixtures, highly crystalline solids were obtained for all complexes directly from the reaction medium. The formation of all complexes was subsequently confirmed by mass spectrometry, elemental analysis, infrared spectroscopy, and single crystal X-ray analysis. It is worth mentioning that no inert conditions were required for the preparation of these complexes and they all are perfectly stable in air in their solid form. This is a quite uncommon feature for Co(II) complexes with pentaaza macrocyclic ligands which usually undergo easy and fast oxidation to Co(III).43,44
The measured IR spectra are very similar for all studied complexes. Vibrations corresponding to the macrocycle can be observed in the positions ∼3300 cm−1 (N–H stretching vibrations); ∼1600, ∼1500 and ∼1460 cm−1 (CC and C
N aromatic vibrations). For complexes 3–6, stretching vibrations of coordinated anions at 2064 cm−1 (–N
N
N), 2210 cm−1, (–N
C
O), ∼2080 and ∼2060 cm−1 (N
C
S; –N
C
Se, two signals due to the presence of one coordinated anion and one non-coordinated counter anion) or those of the perchlorate anion at ∼1080 and ∼620 cm−1 were observed as well.
Compound | 1a | 1b | 2 | 3 | 4 | 5 | 6 |
---|---|---|---|---|---|---|---|
a R int = ∑|Fo2 − Fo,mean2|/∑Fo2. b R 1 = ∑(||Fo| − |Fc||)/∑|Fo|; wR2 = [∑wR2(Fo − Fo2)2/∑w(Fo2)2]1/2. | |||||||
Formula | C18H33Cl2CoN5O | C17H35Cl2CoN5O3 | C35H62Br4Co2N10O | C17H29ClCoN8O4 | C18H29ClCoN6O5 | C19H29CoN7S2 | C19H29CoN7Se2 |
M r | 465.32 | 487.33 | 1076.44 | 503.86 | 503.85 | 478.54 | 572.34 |
Temperature (K) | 293(2) | 276(2) | 293(2) | 293(2) | 293(2) | 293(2) | 293(2) |
Wavelength (Å) | 1.54184 | 1.54184 | 1.54184 | 1.54184 | 1.54184 | 1.54184 | 1.54184 |
Crystal system | Orthorhombic | Triclinic | Monoclinic | Orthorhombic | Orthorhombic | Monoclinic | Monoclinic |
Space group | Pbca |
P![]() |
P21/c | Pbca | Pbca | P21/n | P21/n |
a (Å) | 10.7580(2) | 8.51040(10) | 28.5309(4) | 15.3931(2) | 15.2438(2) | 17.2598(4) | 17.3749(2) |
b (Å) | 13.7934(3) | 10.46290(10) | 10.45941(15) | 16.2764(2) | 16.37741(19) | 8.01170(10) | 8.20480(10) |
c (Å) | 31.8783(6) | 12.91000(10) | 14.18701(16) | 17.5144(3) | 17.6591(2) | 17.7444(4) | 17.9050(2) |
α (°) | 90 | 79.8000(10) | 90 | 90 | 90 | 90 | 90 |
β (°) | 90 | 81.4420(10) | 90.7236(12) | 90 | 90 | 111.562(2) | 111.654(2) |
γ (°) | 90 | 88.9580(10) | 90 | 90 | 90 | 90 | 90 |
V, Å3 | 4730.38(16) | 1118.75(2) | 4233.30(10) | 4388.13(11) | 4408.67(10) | 2281.99(8) | 2372.36(6) |
Z | 8 | 2 | 4 | 8 | 8 | 4 | 4 |
D calc, g cm−3 | 1.307 | 1.447 | 1.689 | 1.525 | 1.518 | 1.393 | 1.602 |
μ, mm−1 | 7.896 | 8.439 | 10.887 | 7.618 | 7.591 | 7.761 | 9.330 |
F(000) | 1960 | 514 | 2176 | 2104 | 2104 | 1004 | 1148 |
θ range for data collection (°) | 2.772–67.684 | 3.517–67.684 | 3.098–67.684 | 4.691–67.684 | 4.688–67.684 | 3.055–67.684 | 3.029–67.684 |
Refl. collected | 4291 | 4329 | 7640 | 3991 | 3999 | 4062 | 4287 |
Independent refl. | 3456 | 4141 | 6033 | 3343 | 3261 | 3490 | 3873 |
R(int)a | 0.0533 | 0.0311 | 0.0472 | 0.0435 | 0.0356 | 0.0563 | 0.0381 |
Data/restrains/parameters | 4291/0/245 | 4329/3/266 | 7640/0/471 | 3991/0/281 | 3999/0/280 | 4062/0/262 | 4287/0/0 |
Completeness to θ (%) | 99.7 | 99.9 | 99.1 | 99.8 | 99.7 | 97.6 | 99.6 |
Goodness-of-fit on F2 | 1.070 | 1.052 | 1.147 | 1.042 | 1.051 | 1.054 | 1.023 |
R 1, wR2 (I > 2σ(I))b | 0.0541/0.1579 | 0.0288/0.0720 | 0.0662/0.1658 | 0.0502/0.1459 | 0.0485/0.1360 | 0.0529/0.1333 | 0.0413/0.1075 |
R 1, wR2 (all data)b | 0.0647/0.1661 | 0.0303/0.0728 | 0.0856/0.1764 | 0.0588/0.1531 | 0.0590/0.1439 | 0.0604/0.1433 | 0.0455/0.1102 |
Largest diff. peak and hole/A−3 | 0.647/−0.548 | 0.445/−0.478 | 1.354/−0.723 | 0.882/−0.556 | 0.666/−0.448 | 0.672/−0.731 | 1.026/−0.643 |
CCDC number | 2279916 | 2279918 | 2279920 | 2279922 | 2279923 | 2279924 | 2279925 |
Complexes crystallized in space groups Pbca (1a, 3, 4), P (1b), P2/1n (5, 6), and P2/1c (2). All complex cations shown in Fig. 2 have similar structural features. The Co(II) is surrounded by a macrocycle L coordinated by all five nitrogen donor atoms and by one monovalent coligand. The coordination sphere for all complexes has distorted trigonal prismatic geometry, which was confirmed by continuous shape measures (deviation between the real and ideal polyhedron geometry) calculated by using the program Shape 2.1 (Table S1†).45,46 The distortion of the polyhedron increases in order 1a,b ∼ 2 → 5 → 6 → 4 → 3 and it is slightly larger for complexes with pseudohalides (3–6) than with halides (1, 2). The coordination bond distances for 1–6 are listed in Table 2 and their comparison is displayed in Fig. 3. Selected bond angles are listed in Table S2.†
![]() | ||
Fig. 3 Comparison of the metal–donor atom distances in complexes 1a–6. Two values are given in the case of complex 2 for two crystallographically independent molecules found in the asymmetric unit. |
Complex 1a/1b | Complex 2 | Complex 3 | Complex 4 | Complex 5 | Complex 6 | |
---|---|---|---|---|---|---|
Co–N1 | 2.097(3)/2.1017(14) | 2.109(6) | 2.102(3) | 2.107(3) | 2.096(3) | 2.098(3) |
2.090(6) | ||||||
Co–N2 | 2.306(3)/2.3192(15) | 2.274(6) | 2.337(3) | 2.346(3) | 2.312(3) | 2.307(3) |
2.255(7) | ||||||
Co–N3 | 2.209(3)/2.2047(16) | 2.197(7) | 2.204(3) | 2.209(3) | 2.199(3) | 2.199(3) |
2.223(7) | ||||||
Co–N4 | 2.257(3)/2.2571(15) | 2.254(6) | 2.278(3) | 2.281(3) | 2.262(3) | 2.269(3) |
2.212(7) | ||||||
Co–N5 | 2.241(3)/2.2146(14) | 2.256(6) | 2.210(3) | 2.213(3) | 2.232(2) | 2.234(3) |
2.302(7) | ||||||
Co–Cl1/Br1/N6 | 2.4719(9)/2.5082(5) | 2.6729(15) | 2.096(3) | 2.059(3) | 2.077(3) | 2.082(3) |
2.6607(15) |
It can be noticed that the macrocyclic ligand L is significantly twisted. Its shortest coordination bond distances in all complexes are between the central Co(II) atom and the pyridine nitrogen atom N1 with a narrow variation (2.090–2.107 Å) within the series (Fig. 3). The bond lengths between Co(II) and the aliphatic secondary NH groups (N2 and N5) are the same only in 2, which coordination polyhedron is the closest to the regular trigonal prism. Otherwise, the difference between Co–N2 and Co–N5 distances is about ∼0.08 Å, the first being larger than the second, the largest difference is for complexes 3 and 4 (∼0.13 Å, Fig. 3). For the coordination bonds involving the piperazine N-atoms, there is also significant difference between them (∼0.06 Å), with shorter Co–N3 in comparison with Co–N4 ones, which is rather constant within whole series (Fig. 3). Thus, three Co–N coordination bonds (N1, N3, N5) are shorter compared to the other two (N2, N4), which confirms the flexibility of the macrocycle.26 For complexes 1 and 2, the bond to the halide ligand is rather long (2.472 Å and 2.673/2.661 Å), whereas for complexes 3–6, the bond with the N-donor coligand atom is 2.059–2.096 Å, shorter than the Co–N1 bond distances.
The two halogenido complexes 1a and 2 formed supramolecular 1D chains via N–H⋯X hydrogen bonds including the non-coordinated halogenide counter anion (Fig. S3 and S5†). This counter anion forms two N–H⋯X hydrogen bonds between the secondary amino groups of two macrocyclic units and additionally, forms one hydrogen bond to a co-crystallized CH3OH solvent molecule. The latter hydrogen bond is present in each 1D chain in complex 1a while only for one half of the chains in complex 2. These 1D chains are symmetrically accommodated next to each other along the b or c axis, respectively (Fig. S3 and S5†), without any non-covalent interaction between each other. On the other hand, complex 1b formed supramolecular dimers via N–H⋯Cl hydrogen bonds and these dimers are further connected into supramolecular 1D chains by extensive system of N–H⋯O and O–H⋯Cl hydrogen bonds. These 1D chains are then connected into a supramolecular 3D network by these extensive O–H⋯Cl hydrogen bonds as well as π–π stacking interactions (Fig. S4,†Cg⋯Cg distance is 3.635 Å). The crystal packing of complexes 3 and 4 is very similar to that of previous complexes and again it consists of supramolecular 1D chains via N–H⋯N and N–H⋯O hydrogen bonds between the first secondary amino group of the macrocycle and coordinated azido or cyanato coligand (Fig. S6 and S7†). These 1D chains are separated by perchlorate anions which are connected via N–H⋯O hydrogen bonds to the second secondary amino group of the macrocycle (Fig. S6 and S7†), but they are not isolated because they are connected to each other via π–π stacking interactions (Cg⋯Cg distance is 3.686 Å (complex 3) or 3.611 Å (complex 4)). On the other hand, the crystal packing of complexes 5 and 6 is much more complex. Similar supramolecular 1D chain formation was observed for both complexes (Fig. S8 and S9†), but the geometry of the chains was very different from those in previous complexes. Furthermore Caromatic–H⋯S/Se hydrogen bonds were observed as well as very weak π–π stacking interactions (Cg⋯Cg distance is 5.532 Å for 5 or 5.512 Å for 6) which all together form an extensive supramolecular 3D network (Fig. S8 and S9†).
![]() | ||
Fig. 4 Temperature dependence of the χMT and the isothermal field dependence of the magnetization (inset) for complex 3. The empty circles are experimental data points, and the full lines represent the best fit with parameter set given in Table 3. |
Compound | 1a Cl | 1b Cl | 2 Br | 3 N3 | 4 NCO | 5 NCS | 6 NCSe |
---|---|---|---|---|---|---|---|
Experimental data | |||||||
χ M T (cm3 mol−1 K) | |||||||
300 K | 2.53 | 3.12 | 3.04 | 2.60 | 2.40 | 2.70 | 2.53 |
2 K | 1.72 | 2.49 | 1.80 | 1.69 | 1.63 | 1.70 | 1.50 |
M (μB, 2 K, 5 T) | 1.97 | 2.49 | 2.29 | 2.0 | 1.90 | 2.27 | 2.0 |
ZFS parameters | |||||||
D (cm−1) | −36 ± 2 | −38.7 ± 0.9 | −21.1 ± 0.7 | −35.2 ± 0.3 | −41.2 ± 0.2 | −19.8 ± 0.5 | −22.0 ± 0.4 |
E (cm−1) | — | 11.3 ± 0.5 | 3.5 ± 0.5 | 4.2 ± 0.1 | — | 5.8 ± 0.5 | 4.8 ± 0.2 |
g | 2.48 | 2.85 | 2.61 | 2.44 | 2.38 | 2.55 | 2.32 |
Magnetization relaxation processes parameters (with respect to eqn (1) ) | |||||||
U/kB (K) | — | — | — | — | — | 32.15 ± 0.08 | — |
τ 0 (s) | — | — | — | — | — | 5.44 ± 0.08 × 10−7 | — |
R (K−n s−1) | — | — | 78 ± 365 | 0.051 ± 0.004 | 0.015 ± 0.008 | — | 0. 51 ± 0.02 |
n | — | — | 2.6 ± 2 | 6.68 ± 0.05 | 7.4 ± 0.3 | — | 6.80 ± 0.09 |
D (K−1 s−1) | — | — | 89 ± 730 × 10−6 | 5.0 ± 0.7 × 10−6 | — | 6.7 ± 0.3 × 10−6 | 1.6 ± 0.2 × 10−5 |
τ QTM (s) | — | — | 0.04 ± 4 | — | — | — | — |
ZFS parameters and g values based on CASSCF/NEVPT2 calculations for the first coordination sphere of the complexes | |||||||
D (cm−1) | −50.1 | −48.6 | −63.0/−59.6 | −37.9 | −31.8 | −34.5 | −33.1 |
E/D | 0.055 | 0.066 | 0.053/0.052 | 0.058 | 0.057 | 0.047 | 0.056 |
g x | 2.122 | 2.120 | 2.095/2.102 | 2.141 | 2.151 | 2.141 | 2.140 |
g y | 2.182 | 2.189 | 2.170/2.170 | 2.187 | 2.190 | 2.171 | 2.171 |
g z | 2.728 | 2.712 | 2.849/2.815 | 2.606 | 2.544 | 2.556 | 2.537 |
Calculated g-tensor values of the lowest Kramers doublet with a pseudospin S = ½ | |||||||
g x | 0.342 | 0.402 | 0.326/0.321 | 0.365 | 0.357 | 0.296 | 0.346 |
g y | 0.371 | 0.444 | 0.356/0.349 | 0.392 | 0.381 | 0.313 | 0.372 |
g z | 8.125 | 8.072 | 8.466/8.376 | 7.772 | 7.592 | 7.636 | 7.573 |
Further insight was provided by CASSCF/NEVPT2 calculations to obtain ZFS parameters D and E (Table 3). The energy diagrams with the d-orbital splitting, ligand-field terms and ligand-field multiplets are shown in Fig. 5. The splitting of d-orbitals is far from that of an ideal trigonal prismatic ligand field, and the second and third lowest energy orbitals (sharing 3 electrons in the HS d7 configuration) are no longer degenerate, which is in agreement with rather large distortion of the coordination polyhedron (Table S1†). As a consequence of such lifting of degeneracy, the first-order spin–orbit coupling is not active in 1–6 and the magnetic anisotropy arises only from the second-order spin–orbit coupling, i.e. ZFS effect. The calculations also revealed that the change of the coligand X, i.e. Br− → Cl− → N3− → NCO− → NCS− → NCSe− results in a progressive shift toward higher energy (from about 1000 to 2000 cm−1) of the d-orbital with the 3rd lowest energy (Fig. 5 left), which is directed towards the coligand (see Fig. S16–S23 in the ESI†), thus the separation between the 2nd and 3rd orbitals increases. The above-mentioned trend is in line with the position of the ligands in the spectrochemical series, and follows the increase of the ligand-field strength. As a consequence of such an increase in the energy difference, the splitting of the lowest two LF multiplets decreases (Fig. 5) and thus |D| is reduced in the same order Br− → Cl− → N3− → NCY− (Y = O, S, Se), see below for a detailed discussion. Therefore, the largest negative D values can be anticipated for a coligand with the weakest ligand field, typically X = Br− and Cl− in the reported series. The contributions to D of the excited states gathered in Table S3 in ESI† confirm that in all cases the largest and negative contribution arises from the mixing with the first excited state. Table S3† also gives the composition of the first two energy states showing their multiconfigurational character without any contribution higher than 60%. This means that any deeper satisfactory analysis (assignment of d-orbitals, evaluation of contributions to D tensor) is difficult in this case.
In general, when the occupation of d-orbitals by 7 electrons (for Co(II)) is considered in an ideal trigonal prismatic ligand field, the transition from the ground to the first exited state is provided by the transfer of one electron between the dxy and dx2−y2 orbitals (they are sharing 3 electrons). In this process, the magnetic quantum number ml is not changed, Δml = 0, therefore the contribution to the D-tensor will be large and negative. Thus, in order to better understand the origin of the observed negative D-values and have a comparison with less distorted trigonal prismatic Co(II) systems studied previously, we tried to make an accurate assignment of the d-orbitals obtained by ab initio ligand field theory (AILFT) calculations. Therefore, calculations with the molecule oriented according to the D-tensor were taken into account (the principal axes of the D-tensor for each complex are shown in Fig. S16–S23†) and by this approach, the composition of each AILFT d-orbital was obtained (Tables in Fig. S16–S23†). It was therefore possible to correctly assign the AILFT d-orbitals, which proved impossible when a non-oriented molecule was considered. Thus, the d-orbitals with the 2nd and 3rd lowest energy in Fig. 5 left appeared to have mixed contributions from both dx2−y2 and dxy orbitals. For example, in 1a, these orbitals contribute 0.42 and 0.32 respectively to the 2nd lowest AILFT d-orbital and 0.44 and 0.50 to the 3rd (see Fig. S16†). Their composition depends on the specific coordinated coligand, but in all cases these two orbitals have the strongest contributions to these AILFT d-orbital (see Fig. S16–S23†). As a result, the coupling of the ground state and the first excited state involves an electron transfer between these two orbitals with Δml = 0 providing a negative and large contribution to D-values (Table S3†). This negative contribution of the 1st excited state is decreasing as the energy separation between the 2nd and 3rd d-orbitals becomes larger, following the trend Br− → Cl− → N3− → NCO− → NCS− → NCSe−. We have also analysed the contributions to the D-tensor arising from transitions between each electron configuration for the ground and first exited state (see Table S3†). The transitions with Δml = 0 prevail and this is in accordance with the overall large negative contribution of the first exited state to the D-value, which in turn governs the overall D of the compounds.
In order to further confirm the magnetic axiality of the ground Kramers doublet, the g-tensors values for a pseudospin S = 1/2 were calculated as well (Table 3). The g-values gx, gy ∼ 0 and gz ∼ 8 unambiguously confirmed a highly axial ground Kramers doublet with the Ising-type g-tensor which is in line with the obtained negative D-values and which is comparable20 with other trigonal prismatic Co(II) complexes with gz ∼ 9.24,25
The calculated D values appear to be generally overestimated compared to the experimentally obtained ones, but the variation along the series follows the same trend, with the exception of complex 2 with a smaller D value and complex 4 with a rather higher D value. Non-covalent interactions, such as hydrogen bonds, are known to cause such discrepancies, especially when the donor atom of the coordinated ligand is involved. Such hydrogen bonds are indeed found between anion from the outer coordination sphere and the coordinated NH– group of the ligand. We therefore also performed calculations of ZFS parameters with fragments containing uncoordinated anions, these however gave values even further away from those found experimentally.
It is worth noting that ZFS parameters are calculated using spin Hamiltonian which is meaningful only when no low-lying excited state is populated. To reliably describe a system, the norms of its projected states should be closest to 1. Norm values for our complexes are around 0.9 or higher – this from our experience should not cause such a significant deviation as ours but it certainly is a contributing factor. Lastly, the reference structure was obtained at higher temperatures and slight changes in the distortion of the coordination sphere at lower temperature may have an effect on the experimental D-value. The aforementioned sources of errors are the most likely reason for the discrepancies.
When the D values obtained for 1–6 are compared to those reported for other trigonal prismatic Co(II) complexes (Table 4), they are among the smallest. This appears to correlate with a more distorted trigonal prismatic coordination sphere, and, likely, to the heteroleptic chromophore {CoN5X}. As discussed above, such a situation leads to a less favourable ligand-field splitting pattern of 3d orbitals and a strongly mixed multiconfiguration composition of wave functions of the ground and low-lying LF states. Thus, the energy separation of dxy and dx2−y2 orbitals is higher in comparison with other Co(II) complexes in Table 4, which results in lower magnetic anisotropy.24 Recently, we have observed similar behaviour in a pentagonal bipyramidal system in which a strong distortion of the coordination sphere had a substantial effect on d-orbital splitting and magnetic anisotropy.48 However, as Table 4 shows, there seems to be no decisive effect of geometry distortion on the D values obtained. Some of the most negative values were found for coordination polyhedra that deviate significantly from trigonal prismatic geometry; and vice versa. In fact, the magnetic anisotropy of these complexes is directly conditioned by the energy diagram of the d orbitals. In particular, the energy difference between the dxy and dx2−y2 orbitals is crucial, as it determines whether or not the 1st-order spin–orbit coupling takes place, but also the contribution of ZFS when the degeneracy is lifted. Thus, even when the actual coordination polyhedral deviates from ideal geometry, a rather small energy difference between these orbital results in a large and negative D, a situation that applies to [CoL4]2+ derivatives22 mentioned in Table 4. The ligand-field acting on the Co(II) center is directly involved in the d-orbital splitting but its effect in heteroleptic systems is more difficult to apprehend. So, it should be stressed here that the geometry distortion should be related to the D-values with great care because the geometry itself does not describe all the aspects of the ligand field (metal–ligand interactions governing the electronic structure of the complex), which is the main driving force for the magnetic anisotropy. On the other hand, the theoretical insights discussed above clearly underline the importance of the coligands interacting with the low-lying d-orbitals.
Complex | Deviation for TPR-6 | D-value (cm−1) | Ref. |
---|---|---|---|
a Calculated using the Shape 2.1 program from the cif file. b Values given for the derivative with boron substituent R = F. L1 = piperazine Schiff base, L2 = tris(pyrazoloximate)phenylborate, L3 = tris(methylimidazoloximate)phenylborate, L4 = tris(pyridylhydrazonyl)phosphorylsulfide, and L5 = tris(1-methylimidazolhydrazonyl)phosphorylsulfide. | |||
1a | 4.720 | −36 ± 2 | This work |
1b | 4.772 | −38.7 ± 0.9 | This work |
2 | 4.779/4.715 | −21.1 ± 0.7 | This work |
3 | 5.544 | −35.2 ± 0.3 | This work |
4 | 5.476 | −41.2 ± 0.2 | This work |
5 | 5.307 | −19.8 ± 0.5 | This work |
6 | 5.338 | −22.0 ± 0.4 | This work |
[CoL1] | 1.382a | −31 | 20 |
Clathrochelate 1 | 0.162a,b | −63b | 21 |
[CoL2]Cl | 0.828a | −82 | 22 |
[CoL3](ClO4) | 0.905a | −102.5 | 23 |
[CoL4][CoCl4] | 3.839 | −60.6 | 24 |
[CoL4][ZnCl4] | 3.196 | −87.2 | 24 |
[CoL4](ClO4)2 | 3.002 | −116.6 | 24 |
[CoL4](BF4)2 | 2.759 | −127.6 | 24 |
[CoL5](ClO4)2 | 0.533 | −95.2 | 25 |
[CoL5](BF4)2 | 0.486 | −98.9 | 25 |
The possibility that complexes 1–6 exhibit slow relaxation of their magnetization at low temperatures has been explored by AC susceptibility measurements performed in the absence and with an applied static magnetic field. All of them, except for complex 1, show the appearance of an out of phase component of their susceptibility, χ′′M, in the presence of an applied field (Fig. S10–S15†).
For complex 2, two relaxation dynamics could be evidenced, suggesting different relaxation characteristics for the two complexes present in the crystal lattice. The observation of the second feature was dependent on the strength of the applied field (Fig. S11†) and the best compromise between strong signal and single (major) relaxation seemed to be for HDC = 1.6 kOe. Therefore, AC data were collected with an applied field of 1.6 kOe. Relaxation time, τ, was obtained from χ′′M = f(ν) behaviors (Fig. 6a). The temperature dependence of τ was analysed considering various relaxation mechanisms; in no case it was reproduced by a single model. The Arrhenius law in association with the direct model gave a satisfactory fit but the value for τ0 was much too large (ca. 10−5 s), suggesting that relaxation is not thermally activated. Reasonable results have been obtained considering Raman and direct processes with a QTM contribution (respectively the second, third, and fourth terms in eqn (1)). Best fit parameters are gathered in Table 3. Note that the uncertainty on each value is huge.
τ−1 = τ0−1 exp(−U/kBT) + RTn + DH2T + 1/τQTM | (1) |
For azido-complex 3, the optimal HDC was found to be 2 kOe (Fig. S12†). The frequency dependence of χ′′M obtained in this applied field and the resulting temperature dependence of the relaxation time are plotted Fig. 6b. The variation of τ was best analysed when considering Raman and direct relaxation processes; best-fit parameters are given in Table 3.
For cyanate complex 4, the longest relaxation time at 2 K was observed for HDC = 1.4 kOe (Fig. S13†) which was considered as the optimal field for subsequent investigations. For this compound, the χ′′M = f(ν) behaviors showed two contributions indicative of two distinct dynamics. At low T this gave rise to two well defined maxima, one in the lower frequency range and the second for higher frequencies. Both maxima shift to higher frequencies with increasing temperature, and above 5 K the two contributions merge. The analysis of χ′′M = f(ν) was done by a model comprising two extended Debye expressions yielding τ1 for the signal spanning over the whole frequency domain, and τ2 for the other (Fig. 6c). These two dynamics are tentatively attributed to a slow relaxation of the magnetization of molecular and lattice origin respectively, for τ1 and τ2. The behaviour 1/τ1 = f(T) could be modelled by a Raman expression (eqn (1)).
For complex 5, the field dependence studies showed that the signal for χ′′M becomes stronger with the field but that the maximum of χ′′M = f(ν) is shifted to larger frequencies (Fig. S14†). Since the high-frequency contribution (attributed to QTM) was suppressed for HDC > 1.4 kOe this field was considered the optimal field. The χ′′M = f(ν) behaviour obtained by applying this static field and the deduced temperature dependence of τ are depicted in Fig. 6d. The variation of τ was best modelled when contributions from both an Orbach and a direct process were considered (first and third terms in eqn (1)). The linear variation of ln(τ) = f(1/T) is characteristic of temperature activated relaxation (Fig. S14†).
For related isoselenocyanate complex, 6, the signal for χ′′M became stronger with a field of up to 2 kOe and levelled for larger fields (Fig. S15†). The maximum of χ′′M = f(ν) did not move with the strength of the applied field; therefore HDC = 2 kOe was chosen for the AC susceptibility studies, the plot of χ′′M = f(ν) is given in Fig. 6e. Temperature dependence of the relaxation time could be well modelled when Direct + Raman contributions were considered to take place (Fig. 6e and Table 3).
Interestingly, the pentadentate ligand provides the complexes with a robust structural geometry that allows the sixth coordination site to be substituted without compromising the trigonal prismatic coordination polyhedron and thus preserving easy-axis magnetic anisotropy. This feature makes these complexes very interesting units for the design of polynuclear SMMs using the building block/complex-as-ligand approach.
Footnote |
† Electronic supplementary information (ESI) available. CCDC 2279916, 2279918, 2279920 and 2279922–2279925. For ESI and crystallographic data in CIF or other electronic format see DOI: https://doi.org/10.1039/d3dt02639f |
This journal is © The Royal Society of Chemistry 2023 |