Aza-macrocyclic complexes of the Group 1 cations – synthesis , structures and density functional theory study †

The Group 1 complexes, [M(Me6[18]aneN6)][BAr ] (M = Li–Cs; Me6[18]aneN6 = 1,4,7,10,13,16-hexamethyl1,4,7,10,13,16-hexaazacyclooctadecane; BAr = tetrakis{3,5-bis(trifluoromethyl)-phenyl}borate), are obtained in high yield by reaction of the macrocycle with M[BAr] in anhydrous CH2Cl2 solution, and characterised spectroscopically (H, C{H}, Li, Na, and Cs NMR), by microanalysis and, for M = Li, K, and Rb, by single crystal X-ray analysis. The structures show N6-coordination to the metal ion; the small ionic radius for Li leads to a puckered conformation. In contrast, the K ion fits well into the N6 plane, with the [BAr] anions above and below, leading to two K species in the asymmetric unit (a hexagonal planar [K(Me6[18]aneN6)] + cation and a ‘[K(Me6[18]aneN6)(κ-BAr)2] anion’, with long axial K⋯F interactions). The Rb ion sits above the N6 plane, with two long axial Rb⋯F interactions in one cation and two long, mutually cis Rb⋯F interactions in the other. The unusual sandwich cations, [M(Me3tacn)2] (M = Na, K; distorted octahedral, N6 donor set) and half-sandwich cations [Li(Me3tacn)(thf)] + (distorted tetrahedron, N3O donor set), [Li(Me4cyclen)(OH2)] , and [Na(Me4cyclen)(thf)] + (both distorted square pyramids with N4O donor sets) were also prepared (Me3tacn = 1,4,7-trimethyl-1,4,7-triazacyclononane, Me4cyclen = 1,4,7,10-tetramethyl-1,4,7,10-tetraazacyclododecane). Density functional theory (DFT) calculations, using the BP86 and B3LYP functionals, show that the accessibility of the [M(Me3tacn)2] + sandwich cations depends strongly on the M ionic radius, such that it is sufficiently large to avoid steric clashing between the Me groups of the two rings, and small enough to avoid very acute N–M–N chelate angles. The calculations also show that coordination to the Group 1 cation involves significant donation of electron density from the p-orbitals on the N atoms of the macrocycle, rather than purely electrostatic interactions.


Introduction
The coordination chemistry of Group 1 cations with neutral ligands is dominated by oxygen-donor ligands such as alcohols, ethers, and water, including the ubiquitous crown ethers and cryptands which are frequently used as ligands towards Group 1 cations. 1,2 The corresponding chemistry with the less electronegative neutral polyaza macrocycles has been much less studied. We recently reported the preparation of several unusual complexes based upon coordination of Na + to Me 3 tacn (1,4,7-trimethyl-1,4,7-triazacyclononane) and Me 4 cyclam (1,4,8,11-tetramethyl-1,4,8,11-tetraazacyclotetradecane), including structural characterisation of the sandwich cation salt, [Na(Me 3 tacn) 2 ][BAr F ] (BAr F = tetrakis{3,5-bis(trifluoromethyl)phenyl}borate) and [Na(Me 4 cyclam)(thf )][BAr F ], containing tetradentate coordination of the macrocycle. 3 The Me 3 tacn complexes were initially isolated serendipitously from reactions of Na[BAr F ] with SiCl 4 in the presence of the N-donor macrocycles, towards preparation of potential precursors for the supercritical fluid electrodeposition (SCFED) of semiconductor materials. 4 The Na + preferentially coordinated to the N-donor macrocycle, and we attributed this unusual behaviour to the low lattice energy of Na[BAr F ], coupled with the high solubility of the [BAr F ] − salts in non-competitive, very weakly coordinating solvents, such as CH 2 Cl 2 and toluene.
Neutral diamines (e.g. N,N,N′,N′-tetramethylethylenediamine) are often used as complexation agents to increase the reactivity of alkyl lithiums 5 and to increase the solubility of Experimental All preparations were carried out under a dry dinitrogen atmosphere using standard Schlenk and glove box techniques.
[Li(thf) 4 ][BAr F ] and Na[BAr F ]·2thf were synthesised using a slight modification of the literature procedure. 12 The lithium salt was isolated as [Li(OH 2 ) 4 ][BAr F ], which can be converted to the thf adduct by stirring in thf for 16 h over 4 Å molecular sieves. Filtration and removal of solvents affords the thf adduct. Crude Na[BAr F ] was recrystallised from thf/n-hexane, resulting in the isolation of Na[BAr F ]·2thf. K[BAr F ], Rb[BAr F ], and Cs[BAr F ] were synthesised via cation exchange of Na[BAr F ]·2thf in water at 95°C with excess (5 mol. equiv.) KNO 3 , RbNO 3 , or CsNO 3 , respectively. 13 Me 3 tacn was synthesised by a literature procedure. 14 Me 4 cyclen was obtained by methylation of cyclen (Sigma) using formic acid/formaldehyde. 15 Me 6 [18]aneN 6 (Sigma) was used as received. CH 2 Cl 2 was dried by distillation from CaH 2 , toluene distilled from Na, and n-hexane distilled from Na/K alloy. 1 H and 13 C{ 1 H} NMR spectra were recorded in CD 2 Cl 2 solution at 298 K using a Bruker AV II-400 spectrometer and are referenced to the residual CH 2 Cl 2 resonance. 7 Li, 23 Na, and 133 Cs NMR spectra were obtained on a Bruker AV II-400 spectrometer at 298 K (unless otherwise stated) and referenced to a 0.1 mol dm −3 solution of LiCl, NaCl, or CsNO 3 in D 2 O, respectively. Microanalyses were undertaken at London Metropolitan University.
[Me 4 cyclenH][BAr F ]. K[BAr F ] (250 mg, 0.24 mmol) was suspended in CH 2 Cl 2 (5 mL) and a solution of Me 4 cyclen (70 mg, 0.60 mmol) in CH 2 Cl 2 (2 mL) was added. The reaction was stirred for 4 h then concentrated to ∼5 mL, n-hexane was layered on top and some colourless crystals were obtained. Required

Results and discussion
In view of the limited literature examples of aza-macrocyclic complexes with Group 1 cations, their expected lability in solution, and the lack of diagnostic spectroscopic signatures for these complexes, X-ray crystallographic data provide the key characterisation method. We therefore describe the solid state data first, comparing these with computed structures obtained using DFT, followed by their solution NMR spectroscopic data.
DFT calculations were carried out using the BP86 and B3LYP functionals. In general, these functionals perform equally well in replicating bond distances and angles in the cations investigated (Tables S2, S3, S7, S8 and S12-S23 †). The results from the BP86 functional are discussed in the main manuscript, while those computed with the B3LYP functional are collected in the ESI. †  6 )] + cation contains an N 6 donor set (without retention of thf ) and the macrocycle is in a puckered conformation; the distortion from an ideal octahedron is severe, as seen by two of the 'trans' N-Li-N angles which show a large deviation from 180°, whilst the Li-N bond lengths vary by ca. 0.3 Å. They are longer than the sum of ionic radii for Li (6-coordinate) and N (2.39 Å), 20 but significantly shorter than the sum of van der Waals radii (3.78 Å), 21 consistent with the mismatch between the small Li + cation and the 18-membered ring. The 1 H NMR spectrum is broad at 298 K, indicating dynamic behaviour in solution, while the 7 Li NMR spectrum shows a singlet at −0.30 ppm.

Hexa-aza macrocyclic complexes
The computed minimum energy geometry of the isolated [Li(Me 6 [18]aneN 6 )] + cation (Fig. 2) has C 2 symmetry. The macrocycle is puckered as in the crystal structure, and overall the computed geometric parameters are also in good agreement with the experimental values (Tables S2 and S3 †). Fig. 3 shows the isodensity plots of the frontier molecular orbitals (FMOs) of [Li(Me 6 [18]aneN 6 )] + . The results show that the highest occupied molecular orbital (HOMO) is centred on the six N atoms of the Me 6 [18]aneN 6 ring, and is composed of the N 2p lone pairs with their associated lobes oriented in the direction of the Li + ion. Population analysis shows that the low-lying HOMOs (HOMO−1 to HOMO−5) are also dominated by the N 2p lone pairs, however, an increasing contribution of Li + is observed from HOMO−1 to HOMO−5 (i.e. HOMO−1, HOMO−3, HOMO−4, and HOMO−5 have 3%, 6%, 6%, and 10% contributions from the 2p z , 2p x , 2p y , and 2s orbitals of Li + , respectively). LUMO and LUMO+1 are antibonding molecular orbitals (ABMOs) and correspond predominantly to the 2s and 2p orbitals of the Li + centre.
[Na(Me 6 [18] 6 )] + , while structure 2 (D 3d symmetry) has a geometry similar to its K + and Rb + counterparts (vide infra). Structure 1 is more stable than 2 by 28.2 kJ mol −1 in the gas phase (Tables S12, S13, S18 and S19 †). It is important to note that the presence of the [BAr F ] − anions in the crystal will influ-ence the relative stabilities of structures 1 and 2 in the isolated solids.
[Rb(Me 6 [18]aneN 6 )][BAr F ] also crystallises with two Rb + cations in the asymmetric unit, with the Rb + ion sitting out of the N 6 plane by 0.561(3) Å (Rb1) and 0.382(3) Å (Rb2). For Rb1, this has the effect that one Rb⋯F interaction is much shorter than the other. The Rb1-F12 distance of 3.235(2) Å is much shorter than the sum of van der Waals radii (4.67 Å) ( Fig. 6a), 21 and is also significantly shorter than the corresponding K⋯F distance. The Rb1⋯F12 i distance (4.128(1) Å) is also within the sum of van der Waals radii. Rb2 has two long Rb⋯F distances of 3.951(1) and 4.233(1) Å, both of which are on the same side of the N 6 ring (Fig. 6b).
The BP86 DFT calculations satisfactorily reproduce the experimental X-ray structures of [M(Me 6 [18]aneN 6 )] + (M = K and Rb) with D 3d symmetry (Fig. 7), which contrasts with C 2 symmetry identified for the Li + counterpart. In the computed structures both metal cations lie within the plane of the Me 6 [18]aneN 6 ring, with six equivalent bond distances (K-N =    6 )] + (M = K or Rb, respectively), which are very similar. As for the Li + complex discussed above, the HOMO corresponds mainly to the 2p orbitals of N atoms, and the LUMO (and LUMO+1) correspond predominantly to the ns (np) orbitals of K + and Rb + respectively; the main metal contribution is in the lower HOMO−3 to HOMO−5 orbitals.
The relevant charge densities for the complex cations are shown in Table 1, together with those for Me 6 [18] 6 )] + complex, 5 and 6, are located by the BP86 DFT computations, with structure 5 lower in energy than structure 6 by 43.1 kJ mol −1 (Fig. 9). In 5, the Cs + ion lies out of the N 6

Tri-aza macrocyclic complexes
Based upon our earlier work on Na-Me 3 tacn complexes, 3 we also attempted to extend the chemistry of Me 3 tacn to the other members of Group 1 (Scheme 2).
Using   ] were unsuccessful, its identity is clear from the spectroscopic and analytical data. DFT calculations on the Me 3 tacn complexes with Na + and Li + were undertaken to probe their relative stabilities. The computed bond distances and angles at Na + in [Na(Me 3 tacn) 2 ] + (Fig. 10, Tables S7 and S8 †) agree well with the experimental values reported earlier. 3 For Li + , the calculations show that the Li-N bond in [Li(Me 3 tacn) 2 ] + (S 6 symmetry) is very long (2.460 Å), whereas in the half-sandwich cation [Li(Me 3 tacn)] + the Li-N distance of 2.029 Å (Fig. 10) is more akin to the reported Li-N bond distances. 9,23 This strongly suggests that the much smaller Li + ion can comfortably accommodate only one Me 3 tacn (consistent with the isolation of the [Li(Me 3 tacn)(thf )][BAr F ] half-sandwich), but introducing a second Me 3 tacn is much less favourable, due to steric clashing between the Me groups on the two rings. As noted earlier, 3 the isolation of the bis-Me 3 tacn sandwich cation is unusual; the large ionic radius of Na + apparently permitting its formation (the Ag + analogue is the only other structurally authenticated example). 24 [K(Me 3 tacn) 2 ][BAr F ] was successfully prepared and isolated similarly to the Na + analogue and is isostructural (Fig. 11), forming a centrosymmetric very distorted octahedral cation. This contrasts with the hexagonal planar coordination present in [K(Me 6 [18]aneN 6 )] + , although the bond distances and angles at K + are actually quite similar.
The K-N bonds are 0.2-0.3 Å longer than the corresponding Na-N bonds, while the intra-macrocyclic N-K-N angles are more acute by 8-9°. These differences may reflect a poorer match between the large K + cation and the small nine-membered Me 3 tacn ring, although as we have noted, they are similar to the angles in [K(Me 6 [18]aneN 6 )] + . As shown in Fig. 11, the Me groups of the two rings are well separated.
The calculations on [K(Me 3 tacn) 2 ] + show that the HOMO and HOMO−1 are doubly degenerate (as are HOMO−2 and HOMO−3) (Fig. 13), and they are mainly located on the N atoms of the Me 3 tacn ring, corresponding to N 2p valence orbitals. The LUMO, LUMO+1 and LUMO+2 are mainly localised on K + . The LUMO corresponds to the 4s orbital of K + , while LUMO+1 and LUMO+2 correspond mainly to the 4p x and 4p y orbitals, respectively. The FMOs of the [Na(Me 3 tacn) 2 ] + analogue follow similar trends (Fig. S5 †). Table 3 shows the charge densities on the metal and N atoms of Me 3 tacn, [M(Me 3 tacn)] + and [M(Me 3 tacn) 2 ] + , M = Na and K. The natural charge on K + is lower than the formal charge of +1, showing a significant electron density transfer from the ligands. The charge densities on the N atoms for [M(Me 3 tacn)] + and [M(Me 3 tacn) 2 ] + are more negative than in the 'free' ligand, because electron density is withdrawn from σ C-H and σ C-N orbitals of the ligand.
As in the [M(Me 6 [18]aneN 6 )] + analogues, the charge densities on M show that less electron density is transferred to the metal in the K + case, suggesting a relatively weaker interaction between the K + and the Me 3 tacn ligand(s), and correspondingly longer K-N distances, than in the Na + case.
Attempts to prepare analogous complexes with Rb + and Cs + ions failed; the optimised minimum energy structures of [M(Me 3 tacn) 2 ] + (M = Rb, Cs) are presented in Tables S16, S17, S22 and S23. †    (3), except for the dissociation energy of the [Li(Me 3 tacn) 2 ] + complex which is much lower. This appears to be mainly a consequence of the inability of the small lithium ion to accommodate six-coordination because the presence of the Me substituents on one macrocycle sterically impedes the approach of the second.
It seems clear, therefore, that isolation of the [M(Me 3 tacn) 2 ] + cations is a fine balance between the metal ion being sufficiently large to avoid significant steric clashing between the Me groups of the two rings, but small enough to avoid very acute N-M-N chelate angles.

Tetra-aza macrocyclic complexes
Changing the ligand from Me 3 tacn to Me 4 cyclen has the effect of both increasing the denticity and expanding the macrocyclic binding cavity. Reaction of [Li(OH 2 ) 4 ][BAr F ] with Me 4 cyclen in   The crystal structure shows a five-coordinate square pyramidal (τ = 0.02) 25 cation with the macrocycle tetradentate and one apical water molecule (Fig. 14). The NMe groups all lie on the same side of the N 4 plane as the metal, with the Li cation displaced out of the N 4 plane by 0.758(7) Å. The same product is obtained irrespective of the ratio of Me 4 cyclen : Li[BAr F ] used.
The reaction between Me 4 cyclen and Na[BAr F ]·2thf in CH 2 Cl 2 yielded [Na(Me 4 cyclen)(thf )][BAr F ], with the thf ligand apical (Fig. 15). The τ value of 0.00 confirms an ideal squarebased pyramidal geometry and the sodium cation is out of the N 4 plane by 1.225(2) Å.
Although sandwich complexes of the Group 1 cations with 12-crown-4, e.g. [Na(12-crown-4) 2 ] + , are well known, 1    the cations are expected to be exchanging rapidly in solution, these values can be compared with the Na + -crown ether cations which have chemical shifts to low frequency of 0 ppm, and therefore indicate that the aza macrocycle coordination is retained in CH 2 Cl 2 solution. Solutions of the potassium complexes in CH 2 Cl 2 solution do not show a 39 K resonance at room temperature (295 K) or on cooling to 185 K. Studies of K + -crown ether cations over a range of temperatures and K : crown ether ratios in a range of donor solvents show that often only the 'free' K + resonance is seen. The [K(crown)] + is often not observed most likely due to fast quadrupolar relaxation in the low symmetry environment. 26c,d In [K(Me 6 [18]aneN 6 )] + and [K(Me 3 tacn) 2 ] + it is probable that fast ligand exchange in solution produces low symmetry environments and fast quadrupolar relaxation (the corresponding 1 H NMR spectra show only broad singlets for the macrocyclic CH 2 groups). For similar reasons, 26e no 85 Rb resonance was observed at any temperature between 298 and 185 K.
The Q value for the 133 Cs nucleus is small, hence NMR spectra are readily observed. The VT 133 Cs NMR data for [Cs(Me 6 [18]aneN 6 )][BAr F ] show a singlet at 54.1 ppm (298 K) and this splits into two singlets at 73.3 and 58.7 ppm at 183 K, which we attribute to the slowing of dynamic processes and the presence of two significant stereoisomers ('all up' and alternating 'up-down'). These chemical shifts are also significantly to high frequency of those typically observed in crown ether adducts. 26f

Conclusions
By taking advantage of the lower lattice energies associated with M[BAr F ] precursors, which leads to increased solubility in very weak donor solvents such as CH 2 Cl 2 , the unusual Group 1 cation complexes [M(Me 6 [18]aneN 6 )][BAr F ] can be obtained in good yield for all members from M = Li to Cs. Structural characterisation of several of these (M = Li, K, and Rb) allows comparisons down the Group, as well as with the rarely observed bis-Me 3 tacn sandwich cations, [M(Me 3 tacn) 2 ][BAr F ], isolated for M = Na and K. The combined experimental and DFT study indicates that the isolation of the [M(Me 3 tacn) 2 ] + cations requires a fine balance of, on one hand, the metal ion being sufficiently large to avoid significant steric clashing between the Me groups of the two rings, and on the other, the ion being sufficiently small to avoid extremely acute N-M-N chelate angles. In contrast, the very small Li + ion forms only the half-sandwich [Li(Me 3 tacn)(thf )] + cation. These complex cations show significant structural differences which correlate closely with the trends in ionic radii down Group 1 and the available macrocyclic binding cavity. The [BAr F ] − anions also show quite significant M⋯F interactions, particularly towards the larger metal ions (K + and beyond), despite being large, diffuse ions with delocalised charge.
DFT calculations show very good agreement with the experimentally determined structures and confirm significant donation of electron density from the N atoms of the ligand upon complexation, which is accompanied by transfer of some electron density to N from the σ C-H and σ C-N bonding orbitals. The nature of the FMOs show that contributions from the metal orbitals are only significant in the lower energy valence occupied orbitals, while the HOMO itself is dominated by the N 2p orbitals. The calculations also show that coordination of the aza macrocycle to the Group 1 cation in these complexes involves significant donation of electron density from the p-orbitals on the N atoms, rather than purely electrostatic interactions.