Selective Anion Binding by a Cofacial Binuclear Zinc Complex of a Schiff-Base Pyrrole Macrocycle

Abstract

The synthesis of the new cofacial binuclear zinc complex [Zn₂(L)] of a Schiff-base pyrrole macrocycle is reported. It was discovered that the binuclear microenvironment between the two metals of [Zn₂(L)] is suited for the encapsulation of anions, leading to the formation of [K(THF)₆][Zn₂(μ-Cl)(L)].2THF and [Buⁿ₄N][Zn₂(μ-OH)(L)] which were characterized by X-ray crystallography. Unusually obtuse Zn-X-Zn angles (X=Cl: 150.54(9)° and OH: 157.4(3)°) illustrate the weak character of these interactions and the importance of the cleft pre-organization to stabilize the host. In the absence of added anion, aggregation of [Zn₂(L)] was inferred and investigated by successive dilutions and by the addition of coordinating solvents to [Zn₂(L)] solutions using NMR spectroscopy as well as isothermal microcalorimetry (ITC). On anion addition, evidence for de-aggregation of [Zn₂(L)], combined with the formation of the 1:1 host-guest complex, was observed by NMR spectroscopy and ITC titrations. Furthermore, [Zn₂(L)] binds to Cl⁻ selectively in THF as deduced from the ITC analyses, while other halides induce only de-aggregation. These conclusions were reinforced by DFT calculations, which indicated that the binding energies of OH⁻ and Cl⁻ were significantly greater than for the other halides.

Introduction

The recognition and binding of anions such as halides, phosphates, and sulfate are key features of many biological processes, including those associated with information transmission, whereas a breakdown in the normal regulation of anion transport is implicated in several disease states, including cystic fibrosis.¹ A critical role for anion recognition is either recognized or proposed in a number of industrial processes, such as nuclear waste extraction and remediation,² as well as base and precious metal extraction.³ Special attention has been given to metal-containing anion receptors as the metal centre can play a crucial structural role in organizing anion-binding groups into a geometry appropriate for the recognition of a specific anion.^4-6 Furthermore, electrostatic interactions between the targeted anion and a Lewis acidic metal centre can promote a binding event.^7;8 For example, in cascade complexes, first discovered by Lehn,⁹ a binucleating capsular ligand was used to accommodate two metallic centers within the cavity, typically in a pseudo-pyramidal geometry with a vacant site. These complexes, in which the two metallic centers are held at a well-defined distance from one another, can then recognize an anion of the appropriate size as the result of bridging coordination interaction involving the two metal centers. This principle has been further explored by several groups; in particular, Lu found that dicobalt cryptates could be used to recognize selectively bromide and chloride over fluoride and iodide.¹⁰ Separately, Fabrizzi showed that halides, cyanides, azide and hydroxide were encapsulated by a binuclear copper bis(tren) complex,⁶ while McKee and Nelson discovered that azide and cyanide could be accommodated within a binuclear copper cryptate.¹¹ Other binuclear complexes in which the two metals adopt pyramidal geometries have been synthesized and have been shown to be capable of accommodating anions such as phosphates and their derivatives in bridging modes while in certain cases also acting as sensors.^8;12 While the detection of chloride using Lewis acidic late transition metal hosts remains relatively rare,¹³ zinc porphyrins, species that are well known to bind strong donor ligands,¹⁴ have been shown to bind anions through electrostatic interactions.^5;15

Cofacial or Pacman diporphyrins are a special class of porphyrins that combine the known coordinative properties of porphyrins with precise organization of the two metal binding sites through the use of a rigid, covalent link, so as to provide a binuclear microenvironment that is particularly well suited for small molecule redox chemistry.¹⁶ The unique binuclear environment provided by these and other molecules created through the assembly of two metalloporphyrins, has been exploited in some aspects of host-guest chemistry. For example, using dimeric porphyrins, the recognition of molecules, such as DABCO,^17;18 and other bifunctional nitrogen donors, such as bipyridines, azides, anilines and diamines, has been successfully achieved.^18;19 Systems containing two zinc porphyrins have been designed to bind and signal the absolute configuration of diamines, aminoalcohols, and diols by exciton-coupled circular dichroism.²⁰ Also, increasing interest has been shown in the use of metallodiporphyrins to recognize larger molecules such as fullerenes.²¹ Looking at the wide range of electron rich substrates that can guest between two metalloporphyrins hosts, it is surprising that the anion-recognition ability of these systems remains largely unknown.

We developed recently a new approach to the formation of binuclear Pacman complexes that relies on the ability of the large, binucleating Schiff-base pyrrole macrocycles H₄L° (Chart 1) to fold at aryl ‘hinges’ upon metal coordination and so present a rigid binuclear microenvironment. These complexes were found to resemble structurally Pacman diporphyrins and to also carry out catalytic redox chemistry.²² Furthermore, binuclear copper and cobalt complexes of L° exhibited preferential in-cleft binding of a strong electron donor ligand such as pyridine, albeit with a distorted bonding mode due to the constrained space.²³ The modular nature and straightforward synthesis of H₄L° allowed us to develop new macrocycles including H₄L (Scheme 1) in which the two MN₄ donor environments are separated by two anthracenyl hinges, a feature that allowed formation of the complex [Pd₂(L)] in which the two metal compartments are cofacial with a Pd⋯Pd separation of 5.4 Å.²⁴ In this manuscript we report the synthesis and anion-binding properties of the new binuclear zinc complex [Zn₂(L)] in which the zinc cations act both as structural agents to pre-organize the ZnN₄ compartments into a cofacial binuclear microenvironment and as Lewis acids to accommodate an anion within the molecular cleft.

Chart 1. Binuclear Pacman complexes derived from the Schiff-base pyrrole macrocycle H₄L°.

Scheme 1. The synthesis and anion-binding reactions of the binuclear Pacman complex [Zn₂(L)].

Results and Discussion

Synthesis of [Zn₂(L)]

The reaction between H₄L and ZnEt₂ in hot THF followed by the addition of diethyl ether affords the binuclear zinc complex [Zn₂(L)] in good yield as an orange precipitate (Scheme 1). Even though the ¹H NMR spectrum of [Zn₂(L)] is surprisingly complex and difficult to interpret, no obvious pyrrole NH resonances are seen, which is consistent with double-metallation of the macrocycle having occurred.

This conclusion is supported further by the IR spectrum, in which the N-H stretch is absent (normally observed at 3250 cm^-1 in H₄L) and a decrease in the C=N vibration from 1616 cm^-1 (H₄L) to 1574 cm^-1 upon metallation is seen. The nano-ESI mass spectrum in acetonitrile exhibits peaks with appropriate isotopic patterns at m/z 1004 and 1068 for the water and bis-acetonitrile adducts [Zn₂(L)] (H₂O) and [Zn₂(L)] (CH₃CN)₂ respectively; elemental analysis also supports the expected molecular composition. Unfortunately, and despite numerous attempts, crystals of [Zn₂(L)] suitable for X-ray diffraction could not be grown and so the solid state structure remains undetermined. However, the numerous resonances observed in the ¹H NMR spectrum of [Zn₂(L)] combined with its poor solubility suggests that aggregation is occurring (see below).

Synthesis of the binuclear zinc complex [K(THF)₆][Zn₂(μ-Cl)(L)]

As an alternative synthetic route to [Zn₂(L)], the reaction between K₄L and ZnCl₂ was carried out in THF and was found to generate the cofacial ‘ate’ complex [K][Zn₂(μ-Cl)(L)]. The ¹H NMR spectrum of [K][Zn₂(μ-Cl)(L)] displays the expected number of resonances and the cofacial geometry of this complex was inferred from the non-equivalence of the endo- and exo-meso-ethyl substituents forming two overlapped quartets at 2.10 ppm and two triplets at 0.99 and 0.79 ppm; the non-equivalence of the ethyl groups was reinforced by appropriate resonances in the ¹³C{₁H} NMR spectrum and assignments were confirmed by 2D NMR experiments. Unfortunately, the bulk isolation of the product proved challenging due to the loss of THF solvent leading to apparent elimination of KCl. However, the slow evaporation of a THF solution of [K][Zn₂(μ-Cl)(L)] afforded a small quantity of crystals suitable for X-ray diffraction. The solid state structure of [K(THF)₆][Zn₂(μ-Cl)(L)] 2THF was determined and is shown in Figure 1, with crystal data and selected bond lengths and angles detailed in Tables S1 and S2, respectively.

Figure 1.

The X-ray crystal structure of [K(THF)₆][Zn(μ-Cl)(L)]⋯2THF (displacement ellipsoids are drawn at 50 % probability). For clarity, the [K(THF)6]⁺ cation, THF solvent of crystallization and all hydrogen atoms are omitted.

Consistent with what was inferred from an analysis of the solution NMR data, the complex adopts a double-pillared cofacial structure in the solid state, with a zinc cation coordinated by each of the two N₄-donor pockets of the ligand. The accommodation of the chloride ion within the cleft forces the mouth to open and results in the two N₄-donor planes deviating from co-planarity by 26.6°. In a similar manner to that observed in the solid state structure of the previously reported binuclear palladium complex [Pd₂(L)],²⁴ the two anthracenyl backbones exist in a face-to-face arrangement with the π-surfaces stacked against one another. In this case, the shortest atom separation is 3.51(1) Å and the co-planarity of the aryl planes is high (3° deviation from the mean planes of the two anthracenes). The molecule is also C₂-twisted by an average of 13.8° between the N₄-donor plane and the anthracenyl backbone. The two zinc atoms, Zn1 and Zn2, are five coordinate and adopt pseudo-square pyramidal geometries with the Zn1 and Zn2 cations displaced into the cleft from the basal N₄-plane by 0.58 Å and 0.56 Å respectively. This results in a Zn1⋯Zn2 separation of 4.532(1) Å i.e., ca. 0.9 Å shorter than in the [Pd₂(L)] analogue. The chloride ion is symmetrically bridged between the two metal centers at distances of 2.340(2) Å and 2.346(2) Å from Zn1 and Zn2, respectively, and subtends a Zn1-Cl1-Zn2 angle of 150.54(9)°. The structure also includes an octahedral potassium cation coordinated by six molecules of THF.

The bimetallic complexes of H₄L can be considered as structural, and in some cases, functional double-pillared analogues of cofacial diporphyrins.²⁵ In this case however, the structure of [K(THF)₆][Zn₂(μ-Cl)(L)] represents a unique example. Indeed, all known binuclear zinc diporphyrin complexes bind to the metals in a square planar geometry within the porphyrin plane and to the best of our knowledge, [K(THF)₆][Zn₂(μ-Cl)(L)] is the first X-ray solid state structure of a cofacial binuclear zinc complex with an anion bridging the two metals.²⁶ Furthermore, the presence of a bridging chloride anion is rare in polymetallic zinc chemistry and the Zn1-Cl1-Zn2 angle of [K(THF)₆][Zn₂(μ-Cl)(L)] is unusually obtuse, being among the largest found in the Chemical Structural Database (mean 96.21°, range 71.53 to 179.97°, number of examples: 105).²⁷ These structural data lead us to suggest that the interaction between the Zn cations and the chloride anion is relatively weak and likely a direct consequence of the pre-organization of the binuclear cleft, an inference that is fully consistent with the difficulties encountered in isolating this fragile compound.

Synthesis and structure of [ⁿBu₄N][Zn₂(μ-OH)(L)]

The reaction between [Zn₂(L)] and one equivalent of ⁿBu₄NOH in THF to generated the anion bound complex [ⁿBu₄N][Zn₂(μ-OH)(L)] that was characterized by ¹H NMR spectroscopy only and proved very difficult to isolate in a pure form. This outcome was ascribed to the hydrated and highly hygroscopic character of hydroxide source which, according to the appearance of resonances for the unmetallated H₄L in the ¹H NMR spectrum, acts to decompose the [Zn₂(L)] host. Nevertheless, in an attempt to synthesize [ⁿBu₄N][Zn₂(μ-F)(L)] from [Zn₂(L)] and ⁿBu₄NF in air, crystals of [ⁿBu₄N][Zn₂(μ-OH)(L)] were instead isolated. These proved suitable for X-ray diffraction analysis. The solid state structure was determined (Figure 2) and crystal data are detailed in Table S1, with selected bond lengths and angles in Table S2; refinement of the guest anion as fluoride led to a less satisfactory residual. Furthermore, the reaction between Li₄L and ZnI₂ did not generate the expected iodide-bridged complex [Li][Zn₂(μ-I)(L)] but instead crystals of [Li(THF)₄][Zn₂(μ-OH)(L)] 2THF were isolated from the reaction mixture. The X-ray crystal structure was determined and the zinc-ate core was found to be near identical to that of [ⁿBu₄N][Zn₂(μ-OH)(L)] so will not be discussed further (see Figure S1). However, its elucidation provide support for the notion that the original complex, [Zn₂(L)], binds hydroxide (presumably derived from a small quantity of water of crystallization) over fluoride.

Figure 2.

The X-ray crystal structure of [ⁿBu₄N][Zn₂(OH)(L)] (displacement ellipsoids are drawn at 50 % probability). For clarity, the ⁿBu₄N⁺ counter-ion and all hydrogen atoms except on the oxygen atom O1 are omitted.

The solid state structure of [ⁿBu₄N][Zn₂(μ-OH)(L)] reveals a hydroxyl group bridging the two zinc cations that adopt a geometrical arrangement similar to that seen in [K(THF)][Zn₂(μ-Cl)(L)]. The two N₄-donor planes are found to deviate from co-planarity by 16.8°, and the π-stacking anthracenyl backbones exhibit a shortest-atom separation of 3.322(8) Å with a displacement from co-planarity of 5.5°. The two zinc cations Zn1 and Zn2 are separated by 3.871(1) Å, i.e., ca. 0.5 Å shorter than in the chloride-bridged analogue. The Zn-bridging hydroxide distances are 1.978(4) and 1.969(4) Å for Zn1-O1 and Zn2-O1, respectively. The bridge forms a Zn1-O1-Zn2 angle of 157.4(3)° which, as in the chloride analogue, is unusually obtuse, especially for a hydroxyl group bridged between two zinc cations and is the largest found in literature (mean 111.13°, range 82.81 to 147.19°, number of examples: 137).²⁷ This feature again provides support for the inference that the cofacial structure is rigid and is unable to flex sufficiently to optimize the interaction between the zinc cations and the hydroxide anion. This conclusion is reinforced by the observation that, compared to [K(THF)₆][Zn₂(μ-Cl)(L)], the two zinc cations are displaced even further into the cavity (N4 plane to Zn distance of 0.748 Å and 0.745 Å for Zn1 and Zn2, respectively) and the twist angle is smaller. These features may be a consequence of the steric clash between the meso-ethyl substituents and the fact that the two metals have to move within the cavity in order to obtain a small enough Zn⋯Zn separation to allow anion binding to occur. While all known binuclear zinc cofacial diporphyrins exhibit square planar geometries at the metal, the binuclear lutetium complex [Lu₂(μ-OH)₂(DPA)] (where DPA is a single-anthracenyl-pillared diporphyrin) was shown by X-ray crystallography to have two hydroxyl anions bridging the two metals within the cavity in a manner similar to the ‘ate’ complex [Zn₂(μ-OH)(L)]^-.²⁸ In the former Lu case, however, the increased size of Lu³⁺ (ionic radius 85 pm) compared to Zn²⁺ (74 pm), and the inability of the porphyrin to expand its N₄-donor cavity presumably leads to the observed out-of-plane binding of the Lu metal center and its corresponding propensity to stabilize bridging interactions with the co-bound anions.

Solution speciation and anion binding of [Zn₂(L)]

The isolation of zinc complexes of L that appear pre-disposed towards anion binding, coupled with the poor quality ¹H NMR data acquired for the parent binuclear zinc complex [Zn₂(L)], encouraged us to evaluate possible aggregation and anion binding phenomena. In the first instance, the ability of [Zn₂(L)] to bind chloride in solution was evaluated by ¹H NMR spectroscopy (Figure 3). The addition of an excess of ⁿBu₄NCl to a solution of [Zn₂(L)] caused the ¹H NMR spectrum to simplify considerably, leading immediately to signals that are ascribed to a mixture of two compounds. These signals resolve into a set of resonances consistent with the presence of only a single species over the course of 72 h. The presence of a double set of resonances at 2.91/2.45 and 1.41/1.12 ppm for the meso-ethyl substituents in the ¹H NMR spectrum recorded after 72 h is consistent with the formation of a complex of cofacial geometry, i.e. the ultimate generation of [ⁿBu₄N][Zn₂(μ-Cl)(L)]. The second set of resonances seen in the ¹H NMR spectrum immediately on addition of Cl⁻ is similarly consistent with a binuclear complex of cofacial structure and is likely due to the formation of monomeric [Zn₂(L)], i.e. the addition of Cl⁻ induces de-aggregation.

Figure 3.

The reaction between [Zn₂(L)] (16 mM) and ⁿBu₄NCl in d₅-pyridine monitored over time by ¹H NMR spectroscopy (only the aromatic region shown).

Numerous examples of the self-assembly of zinc porphyrins have been reported and make use of the Lewis acidity of the zinc cation in these macrocyclic complexes, which is compensated by the addition of an extra ligand in the axial position.¹⁴ Considering that the structural properties of cofacial Schiff-base pyrroles are similar to those of cofacial diporphyrins, the propensity of [Zn₂(L)] to accept a strong donor ligand is likely and in the absence of a suitable interaction, the underlying coordination deficiency that drives ligand association is compensated by aggregation. Although the interactions leading to the formation of aggregates and their speciation has not yet been determined, the consistent lack of resolution seen in the ¹H NMR spectra of [Zn₂(L)] and the presence of some signals below 0 ppm are similar to those observed by Balaban and coworkers for the aggregation of zinc porphyrins.²⁹ In order to understand this aggregation process better, a sequential dilution study of [Zn₂(L)] monitored by ¹H NMR spectroscopy was carried out in THF (Figure 4). On successive dilution from 4.8 to 0.1 mM gradual resolution towards a single set of signals is observed. These signals are identical to those seen above on addition of Cl ⁻ and are consistent with the de-aggregation to form monomeric [Zn₂(L)] at lower concentrations. Better resolved signals are also seen in the ¹H NMR spectrum when the dilution experiment is carried out in a good donor solvent, such as pyridine. Such a finding is consistent with the intuitively reasonable expectation that the monomeric species is stabilized by the pyridine donor ligands.

Figure 4.

¹H NMR spectra of [Zn₂(L)] in THF recorded at concentrations ranging from 0.1 mM (top) to 4.8 mM (bottom). The stars indicate the position of the resonances of the monomeric species.

Attempts to evaluate further the speciation of [Zn₂(L)] in solution by UV-visible spectroscopy were hindered by its sensitivity towards protic impurities at the μM level and spectrophotometer saturation at mM concentrations. Furthermore, the addition of excess chloride to a THF solution of [Zn₂(L)] made little change to the UV-visible spectrum, making it difficult to assign spectra to different species in solution (Figure S3); as such, the solution speciation was studied by isothermal microcalorimetry (ITC) and NMR spectroscopy.

Each ITC experiment comprised the sequential addition of a solution of anionic guest to a mM solution of [Zn₂(L)] which acts a host (Figure 5). The de-aggregation phenomenon was studied using ⁿBu₄NCl as an anion source and DMF as a solvent, as while [Zn₂(L)] proved to be soluble enough to carry out ITC experiments in this solvent, no evidence of anion binding was observed. ITC experiments were also carried out at different temperatures and concentrations, and highlighted that more dilute and/or warmer samples required less anion to reach full dissociation, a finding that again supports the proposed aggregation/de-aggregation of [Zn₂(L)]. Furthermore, at lower concentrations, i.e. when full dissociation is achieved, a simple dilution pattern was observed.

Figure 5.

ITC traces corresponding to the titration of [Zn₂(L)] with ⁿBu₄NCl in DMF: (a) 50 °C, 1.12 mM, n = 0.17; (b). 25 °C, 0.67 mM, n = 0.44; (c) 25 °C, 0.34 mM, n = 0.18; (d) 25 °C, 0.13 mM, n = n/a.

Ideally, a solution of the fully de-aggregated [Zn₂(L)] would be required to carry out the anion binding studies. Unfortunately, a balance had to be found between the ability of a solvent to solvate [Zn₂(L)] and its own binding ability, i.e. a strong enough donor solvent is required to dissociate the aggregate but too good a donor binds strongly to the zinc centers of the metal complex, thereby competing with the anion binding event. Furthermore, the highly hygroscopic nature of the host complex required dry conditions to inhibit the formation of the hydroxyl-bridged species. Likewise, the presence of water could weaken significantly the strength of anion–metal cation interactions. Therefore, ¹H NMR anion binding experiments were carried out in THF or pyridine and care was taken to exclude moisture.

On sequential addition of ⁿBu₄NCl the ¹H NMR spectrum of [Zn₂(L)] in THF becomes more resolved and the signals for the chloride complex appear while the resonances for the free host vanish. This is just as is expected for a de-aggregation event followed by anion binding (Figure S2). Unfortunately, the aggregated nature of the host does not allow for accurate integration of the resonances. While this lack of precision precluded carrying out a Job plot analysis, a split of, for example, the signal for the 9H-anthracene proton from 9.65 ppm into 9.80 ppm and 9.16 ppm is seen to occur, as would be expected given this two step chemical transformation (i.e., de-aggregation followed by anion-cation interaction). The same experiment carried out in pyridine-d₅ led to similar observations, in that two sets of well resolved signals were observed corresponding to the monomeric [Zn₂(L)] from dilution and anion-bound [Zn₂(μ-Cl)(L)]⁻. As noted above, equilibration over 72 hours afforded a single set of signals that are consistent with the formation of [ⁿBu₄N][Zn(μ-Cl)(L)] as the final, stable product. While such a conversion is consistent with all available data, the dynamic nature of the transformation renders quantitative measurements difficult using NMR spectroscopy. Therefore, we considered ITC to be a more useful method to investigate the proposed anion binding events.

ITC binding experiments were carried out in DMF and THF and focused on halides, which were studied in the form of their tetrabutylammonium salts (Figure 6). No reliable data were obtained for fluoride, possibly due to its hygroscopic nature and the potential for formation of a hydroxyl adduct. Therefore, fluoride ion was excluded from the study and the investigations focused on chloride, bromide and iodide. In DMF, the dissociation of the aggregate is observed on chloride addition but no evidence of binding was detected, leading us to suggest that DMF competes strongly with the anion at the binding site. The addition of bromide and iodide salts both lead to patterns typical of dilution only, which is consistent with their lower electronegativity and softer character compared to chloride. When a less coordinating solvent such as THF is used, changes consistent with de-aggregation followed by anion binding of chloride are observed sequentially. Furthermore, the underlying data are consistent with a 1:1 binding event and are thus in full accord with the solid state data in which the anion is bound within the cavity of the Pacman-shaped molecule. The host complex [Zn₂(L)] also exhibits selectivity for Cl⁻ over Br⁻ and I⁻ as only de-aggregation phenomena are observed for the latter halides in THF solvent. In all cases, it should be noted that only small quantities of heat were generated during de-aggregation and anion-binding events and so reliable thermodynamic data were difficult to reproduce; typically, the de-aggregation and binding events for Cl⁻ amount to 7.4 and 6.1 kcalmol^-1 respectively. Unfortunately, the poor solubility of [Zn₂(L)] in THF limited the accessible concentration and therefore prevented the determination of more accurate values. The titrations were carried out at different host [H] and guest [G] concentration (at [H]/[G] of 1 mM/20 mM, 1 mM/10 mM and 0.5 mM/10 mM) and these variations did not affect the overall nature of the binding isotherms, thus providing important support for the proposed selective binding of chloride and with a 1:1 host:guest ratio (Table 1).

Figure 6.

ITC traces of the titration of 1.11 mM [Zn₂(L)] by 10.16 mM ⁿBu₄NCl in THF: (a) full trace; (b) fitting profile for de-aggregation phenomenon (n = 0.6); and (c) fitting profile for binding event (n = 1.1)

Table 1.

ITC data in DMF and THF for the titration of a 0.5 mM solution of [Zn₂(L)] by a 10mM solution of anion.

Anion	Solvent	De-aggregation		Binding
		Equivalence	K (M−1)	Equivalence	K (M−1)
Cl−	DMF	0.44	158000 ±1.2×104	not observed	---
Br−	DMF	not observed	---	not observed	---
I−	DMF	not observed	---	not observed	---
Cl−	THF	0.58	271000 ±1.5×104	1.07	32200 ±1×103
Br−	THF	0.73a	40000 ±1.5×104	not observed	---

^a20 mM guest solution

Calculations

DFT calculations were carried out to determine the structure of the binuclear zinc complex [Zn₂(L)] and its propensity to act as a host for OH⁻, Cl⁻, Br⁻, and I⁻ anions. The structure of the binuclear palladium complex [Pd₂(L)] was used as a starting point for geometry optimization of [Zn₂(L)] at the M05-2X/LANL2DZ level of theory. Bulk solvation effects were accounted for via the polarized continuum model (PCM), and the counterpoise method was use to correct for errors in the calculated binding energies of the anion-bound host-guest complexes [Zn₂(X)(L)]⁻ arising from the incompleteness of the basis set. Absolute binding energies for the series of anions were determined and are detailed in Table 2 along with a comparison of geometrical data for experimental and calculated structures.

Table 2.

Comparison of calculated and experimental (italicized) structural parameters and counterpoise-corrected, solvated anion binding energies for [Zn₂(μ-X)(L)]^- where X = OH, Cl, Br, I.

	M⋯M (Å)	M-X-M (o)	M o.o.pa (Å)	Biteb (o)	Twistc (o)	Binding energyd / kcal.mol-1
Zn2(L)	4.98	n/a	0.14	8.4	34.2	n/a
Pd2(L) expe	5.38	n/a	0.06	15.3	29.4	n/a
Zn2(OH)(L)-	3.90	153.4	0.68	15.5	8.5	-49.3 (-31.8; -28.9)
Zn2(OH)(L)- exp	3.87	157.4	0.75	16.8	10.4	n/a
Zn2(Cl)(L)-	4.68	150.5	0.54	28.9	11.2	-9.8 (+6.2; +5.9)
Zn2(Cl)(L)-exp	4.53	150.5	0.56	26.6	15.7	-6.1f
Zn2(Br)(L)-	4.84	138.9	0.54	35.0	9.3	-4.0 (+11.0; +10.4)
Zn2(I)(L)-	5.03	130.1	0.53	40.7	8.8	+1.4 (+16.9; ----)

^ao.o.p = distance of Zn out of N₄ donor plane into the cleft;

^bdihedral angle subtended by the two N₄ donor planes;

^cdihedral angle subtended by the N₄ donor plane and anthracenyl backbone;

^dM05-2X/LANL2DZ level of theory (B3LYP/LANL2DZ and B3LYP/6-31g(d,p) successively in brackets), a negative value indicates anion-binding;

^eref 24;

^fdetermined by ITC

The substitution of Pd by Zn in the calculated structure of [Zn₂(L)] causes significant changes to the gross molecular structure (Table 2). In particular, in order to incorporate the smaller cation the cofacial molecule adopts a significantly greater lateral twist (34.2 ° cf 29.4 ° in [Pd₂(L)]) similar to that seen by us in the structural comparison of the o-phenylene-backbone-based complexes [Pd₂(L°)] and [Ni₂(L°)]. Furthermore, the [Zn₂(L)] complex is more cofacial than its Pd counterpart (bite angles 8.4 vs. 15.3° respectively) which results in a shortening of the Zn⋯Zn separation by 0.4 Å compared to the Pd analogue. Addition of the anion within the molecular cleft of [Zn₂(L)] also causes significant structural changes; no exogenous binding of anion is seen. The calculated structure for [Zn₂(OH)(L)]⁻ bears a close resemblance to the experimentally determined structure of [Bu₄N][Zn₂(OH)(L)] with the Zn cations moving 0.68 Å into the cleft resulting in a Zn⋯Zn separation of 3.90 Å, an obtuse Zn-OH-Zn of 153.4° and a significantly reduced lateral twist angle of 8.5°. This latter angle is underestimated when compared to the experimental data which can be related to the essentially gas phase environment of the calculations compared to the condensed phase of the experimentally determined structure, which will be perturbed by crystal packing interactions. When the guest is chloride, the calculated structure is again similar to the experimental, albeit with an underestimated lateral twist angle. The calculated structures for Br⁻ and I⁻ show that these anions are poorly accommodated by the macrocyclic complex. In particular, the bite angle is increased by 6° for Br⁻ and 12° for I⁻ which results in an increase in the Zn⋯Zn separation of 0.2 and 0.4 Å respectively and a concomitant decrease in the Zn-X-Zn angle by 10-20° as compared to the Cl⁻ and OH⁻ analogues. The lack of vertical expansion between the two ZnN₄ compartments contrasts to that seen in the porphyrinic analogues which are able to extend vertically from 3.5 to 7.8 Å and is likely due to the presence of two anthracenyl pillars between the two compartments rather than the single pillar commonplace in Pacman diporphyrins. Comparison of the guest anion binding energies in the [Zn₂(L)] host shows an important anion-dependency. In particular, the binding energy of OH⁻ is 5 times that of the most strongly-interacting halide, Cl⁻, and accounts for the experimentally-observed propensity for the [Zn₂(L)] host to scavenge hydroxyl from trace quantities of water present in the crystallization media. The calculated Cl⁻ binding energy at 9.8 kcalmol^-1 is in reasonable agreement with that determined experimentally by ITC (6.1 kcalmol^-1). The calculated binding energy of Br⁻, at −4.0 kcalmol^-1, in light of the overestimation of the binding energy of Cl ⁻ and coupled with the inherent error in the accuracy of the calculation method, is likely to correspond to a complex in which the anion is extremely weakly bound (> 1 kcalmol^-1), or is in fact unbound, as the I ⁻ complex is predicted to be. Experimentally, this is supported by the ITC observation that Cl ⁻ forms a 1:1 host:guest complex, whilst the addition of Br ⁻ and I ⁻ result only in dilution.

In the comparison of the three levels of theory, the calculated trends in the anion binding energies agree qualitatively with the trends observed experimentally (Table 2). Significantly stronger binding is observed for the hydroxide ion, than for the halides, which themselves show decreasing binding energies along the series Cl ⁻ > Br ⁻ > I ⁻. Interestingly B3LYP, irrespective of the basis set used, predicts each of the halide-ligand complexes to be unbound, disagreeing with that seen experimentally. It is likely that the underestimation of the binding energies calculated with B3LYP is a direct result of its improper description of dispersive interactions. In contrast, the binding energies calculated with M05-2X/LANL2DZ are in much better agreement with the available experimental measurements.

Conclusions

The anthracenyl-hinged Schiff-base pyrrole macrocycle H₄L proved to be suitable for the complexation of two zinc cations and the resulting complex [Zn₂(L)] was characterized together with two of its “ate” compounds in which a chloride or hydroxide anion bridges the two metals within the binuclear cofacial molecular cleft. The anion-free complex [Zn₂(L)] was found to be an aggregate in solution and its dissociation was observed as a result of dilution, heating, or upon addition of certain anions. Significantly, in an appropriate solvent, [Zn₂(L)] was found to accommodate selectively a chloride anion over the other halides as inferred from ITC experiments, while evidence of favored hydroxyl binding was also obtained. These host-guest interactions were found to be weak, a result that was corroborated by DFT calculations. These latter revealed a distinct preference for hydroxide over chloride in the case of complex [Zn₂(L)], and confirmed the experimentally-observed finding that bromide and iodide give rise to much less thermodynamically-stable host-guest interactions. The binding energy for [Zn₂(μ-Cl)(L)] ⁻ calculated at the M05-2X/LANL2DZ level of theory was in good agreement with the experimental binding energy determined using ITC. Comparison with calculations at different levels of theory highlighted the importance of selecting a functional that includes a realistic description of mid-range correlation when investigating the binding in neutral-anionic non-covalent complexes. It is clear from this work that the preformed, cofacial arrangement of Lewis acidic zinc cations facilitates the weak binding of small anions such as hydroxide and chloride within the molecular cleft; as such, wider investigations are planned to evaluate the effect of this binding event on the spectroscopic properties of these materials. In fact, preliminary results have been obtained that show that [Zn₂(L)] also acts as a receptor for tetrahedral ions such as phosphate and thus may have a broader role to play as an anion recognition and sensing system. The flexible nature of the complex and the ability to respond to various stimuli in different chemical terms, i.e., de-aggregation vs. anion binding lead us to suggest that complex [Zn₂(L)] and its analogues could be employed as a chemically-controlled switching element.

Experimental Section

General

The synthesis of H₄L was carried out as described in the literature,²⁴ while all other chemicals were used as purchased. With the exception of [Bu₄N][Zn₂(μ-OH)(L)], all the zinc complexes were synthesized under nitrogen using Schlenk and glovebox techniques. Dry solvents (THF, DMF, toluene and diethylether) were purified by passage through Vacuum Atmospheres solvent drying towers, pyridine was distilled from potassium and stored over molecular sieves, pyridine-d₅ and benzene-d₆ were dried over potassium, trap-to-trap vacuum distilled, and freeze-pump-thaw degassed three times. ¹H NMR spectra were recorded at 298 K on a Bruker ARX250, DPX360, DMX500 or AVA600 spectrometer operating at 250.13, 360.13, 500.13 and 599.81 MHz respectively; ¹³C NMR spectra were recorded at 298 K on a Bruker ARX250, DPX360 or DMX500 spectrometer operating at 62.90, 90.55 and 125.77 respectively. Dilution experiments and titrations were recorded on a Bruker DMX500 spectrometer. All NMR spectra were referenced to residual protio-solvent resonances. When the solvent is THF, the ¹H NMR experiment was recorded with double-presaturation of THF with a small quantity of benzene-d₆ added as a deuterium lock. IR spectra were recorded on a JASCO FT/IR 460 Plus spectrometer in the range 4000-400 cm^-1. The UV-Vis spectrum of [Zn₂(L)] in THF was recorded on a PerkinElmer Lambda 9 UV/Vis/NIR Spectrophotometer. The mass spectrum of [Zn₂(L)] was recorded on a Q-Tof 2 (Micromass Waters Corporation, Manchester, UK) mass spectrometer using a nano-electrospray ionization source and its elemental analysis was carried out by Mr. Stephen Boyer at the London Metropolitan University. Microcalorimetric titrations were carried out using a MicroCal VP-ITC instrument and were processed with the Origin software provided.

Synthesis of [Zn₂(L)]

A solution of 1.1M diethyl zinc in toluene (3.5 mL, 3.85 mmol) was added dropwise to a solution of H₄L (1g, 1.16 mmol) in THF (10 mL) at -78°C, allowed to reach room temperature and heated to reflux for 16 h. The orange suspension was treated with diethylether (20 mL) and the supernatant decanted by cannula transfer. The resulting orange solid was washed with diethylether (20 mL) and dried under vacuum (0.584 g, 51%).

¹H NMR (pyridine-d₅, 500.13 MHz): Broad overlapping resonances in the range 10-0.5 ppm.

IR (nujol): v 1610 (w), 1574 (C=N), 1546 (C=C) cm^-1.

UV-Vis (THF, 12 μM): λ_max 238 nm (ln ε = 10.59), 266 (10.85), 359 (10.52), 434(10.48).

Nano-ESI-MS: 1068 ([M+82]⁺, <1%, bis-MeCN adduct), 1004 ([M+18]⁺, 2.5%, H₂O adduct), 861 ([M − 2 Zn + 1]⁺, 15%, ligand), 803 ([M− 2 Zn − 58]⁺, 74%, double Et loss on ligand), 429 ([(M− 2 Zn)]²⁺/2, 33%, ligand), 413 ([(M− 2 Zn) - 34]²⁺/2, 33%, double Me loss on ligand, doubly charged species)

Analysis. Found: C, 70.41; H, 4.84; N, 11.23. C₅₈H₄₈N8Zn₂ requires: C,70.52; H, 4.90; N, 11.34 %

Synthesis of [K][Zn₂(μ-Cl)(L)]

THF (10mL) was added to a stirred mixture of H₄L (0.200 g, 0.23 mmol) and KH (0.044g, 1.15mmol, 5eq) at -78°C and allowed to reach room temperature. After 4h, the resulting solution was added dropwise to a solution of ZnCl₂ (0.063g, 0.46 mmol) in THF (1 mL) at room temperature. The red solution turned yellow almost instantly and a solid appeared. The resulting mixture was stirred for 16 h after which the reaction mixture consisted of a thick suspension. Complete filtration was prevented by the very fine solid and NMR data were recorded directly from the liquors.

¹H NMR (THF/benzene-d₆, 500.13 MHz): δ_H 8.91 (s, 2H, ArH), 7.67 (br, 4H, imino), 7.64 (br, 2H, ArH), 7.20 (d, 4H, J = 8.4 Hz, ArH), 6.84 (t, 4H, J = 8.4 Hz, ArH), 6.44 (d, 4H, J = 3.3 Hz, pyrrole CH), 6.33 (d, 4H, J = 6.8 Hz, ArH), 6.08 (d, 4H, J = 3.3 Hz, pyrrole CH), 2.10 (two superimposed q, 8H, J = 7.1 Hz, CH₂), 0.99 (t, 6H, J = 7.2 Hz, CH₃), 0.79 (t, 6H, J = 7.3 Hz, CH₃).

¹³C{¹H} NMR (THF/benzene-d₆ 125.77 MHz): δ_C 157.2 (s, imino), 156.0 (s, quaternary), 151.2 (s, quaternary), 134.9 (s, quaternary), 132.0 (s, quaternary), 127.1 (s, quaternary), 126.4 (s, CH), 125.1 (s, CH), 125.0 (s, CH), 120.3 (s, CH), 117.3 (s, CH), 116.7 (s, CH), 111.4 (s, CH), 47.3 (s, CH₂), 39.2 (s, CH₂), 10.7 (s, CH₃) 9.90 (s, CH₃). Note: resonance due to meso-quanternary carbon is obscured by resonances due to THF.

Synthesis of [ⁿBu₄N][Zn₂(μ-Cl)(L)]

N-tetrabutylammonium chloride (28 mg, 0.1 mmol) was added to a suspension of [Zn₂(L)] (10 mg, 0.01 mmol) in d₅-pyridine and the system was allowed to equilibrate for 64 h during which full dissolution of the sample occurred.

¹H NMR (pyridine-d₅, 500.13 MHz): δ_H 9.79 (s, 2H, ArH), 7.89 (s, 4H, imino), 7.83 (s, 2H, ArH), 7.43 (d, 4H, J = 8.5 Hz, ArH), 7.07 (dd, 4H, J = 7.0 and 8.0 Hz, ArH), 6.75 (d, 4H, J = 3.3 Hz, pyrrole CH), 6.53 (d, 4H, J = 3.3 Hz, pyrrole CH), 6.34 (d, 4H, J = 6.8 Hz, ArH), 3.67 (m, free ⁿBu₄NCl), 2.91 (m, 4H, CH₂), 2.45 (m, 4H, CH₂), 1.88 (m, free ⁿBu₄NCl), 1.41 (t, 6H, J = 7.22 Hz, CH₃), 1.34 (m, free ⁿBu₄NCl), 1.12 (t, 6H, J = 7.53 Hz, CH₃), 0.88 (t, J = 7.3 Hz, free ⁿBu₄NCl). Resonances of the [ⁿBu₄N]⁺ counter cation of [Zn2(μ-Cl)L]^- overlap with the resonances for free ⁿBu₄NCl.

UV-Vis (THF, 12 μM): λ_max 257 (ln ε= 10.54), 351 (9.79), 367 (9.78), 435 (9.70).

Synthesis of [ⁿBu₄N][Zn₂(μ-OH)(L)]

Solid ⁿBu₄NOH·30H₂O (74 mg, 0.09 mmol) was added to a suspension of [Zn₂(L)] (50 mg, 0.05 mmol) in THF (5 mL) at room temperature and the mixture stirred for 48 h during which full dissolution of the solids occurred. The solvent volume was reduced to 0.5 mL under reduced pressure, Et₂O (20 mL) was added and the orange precipitate that formed was isolated, washed with Et₂O (5 mL) and dried under vacuum to yield [ⁿBu₄N][Zn₂(μ-OH)(L)] as an orange solid (37 mg, 57%).

¹H NMR (600 MHz, THF/benzene-d ₆): δ_H 9.28 (s, 2H, ArH), 7.70 (s, 4H, imino), 7.60 (s, 2H, ArH), 7.21 (d, J = 8.4 Hz, 4H, ArH), 6.91 (m, 4H, ArH), 6.40 (d, J = 3.3 Hz, 4H, pyrrolic ArH), 6.32 (d, J = 6.7 Hz, 4H, ArH), 6.14 (d, J = 3.3 Hz, 4H, pyrrolic ArH), 2.81 (br, 15H, ⁿBu₄N), 2.49 (m, 4H, CH₂), 2.25 (m, 4H, CH₂), 1.26 (t, J = 7.2 Hz, 6H, CH₃), 1.09 (br.m, 40H, ⁿBu₄N), 0.89 (t, J = 7.3 Hz, 6H, CH₃), 0.676 (br.m, 52H, ⁿBu₄N). No resonances were observed for the hydroxyl proton and may be obscured by the THF-suppression.

Over time, [ⁿBu₄N][Zn₂(μ-OH)(L)] reacts with water (solvent of crystallization from ⁿBu₄NOH·30H₂O or from the atmosphere) and de-metallation of the complex occurs. This prevented the isolation of a high purity sample of [ⁿBu₄N][Zn₂(μ-OH)(L)] and repeated attempts at obtaining satisfactory elemental analytical data proved unsuccessful.

Crystallographic details

X-Ray diffraction data on single crystals of [K][Zn₂(μ-Cl)(L)] and [ⁿBu₄N][Zn₂(μ-OH)(L)] were collected at 150 K on a Bruker SMART 1000 diffractometer equipped with a CCD area detector using graphite monochromated using Mo Kα radiation (λ= 0.71073 Å). Details of each data collection and refinement are given in Table S1. The structures were solved by direct methods and refined by full-matrix least square refinement on |F|² using SHELXTL-97. All non-hydrogen atoms were refined with anisotropic displacement parameters while hydrogen atoms were placed at calculated positions and included as part of the riding model. The bridging hydroxyl hydrogen atom H001 was located using a 2-site free-variable refinement procedure and refined with fixed thermal parameters. The data for both [K][Zn₂(μ-Cl)(L)] and [ⁿBu₄N][Zn₂(μ-OH)(L)] were weak, presumably due to facile desolvation of the crystal, and so have resulted in relatively high residuals. For [K][Zn₂(μ-Cl)(L)], θ_max was lowered to prevent noise collection and some atoms in disordered molecules of THF could not be modeled satisfactorily. For [ⁿBu₄N][Zn₂(μ-OH)(L)], the relatively high residual electron density was located close to the Zn1 (0.9 Å) and is not at a sensible position for a hydrogen atom on the bridging hydroxide. Other high residual electron density was located close to Zn2.

¹H NMR spectroscopic dilution experiments and titrations

The dilution experiment was carried out using a 5 mM sample of [Zn₂(L)] in dry THF-d₈ with successive dilutions down to 0.1 mM. To ensure a suitable signal-to-noise in the ¹H NMR spectra, the number of scans was increased accordingly. NMR titrations were carried out in THF with double presaturation by the addition of an ca. 1 mM [Zn₂(L)] and ca. 12 mM ⁿBu₄N.Cl solution to an ca. 1 mM [Zn₂(L)] solution, or in pyridine-d₅ by addition of an ca. 16 mM [Zn₂(L)] and ca. 195 mM ⁿBu₄N.Cl solution to an ca. 16 mM [Zn₂(L)] solution (in each case, the highest concentration was limited by the solubility of [Zn₂(L)]).

Microcalorimetric titrations

Titrations were carried out by the automated sequential addition of the salt solution to a [Zn₂(L)] solution at 298K. Extensive efforts were made to exclude moisture from the sample as this could decompose the binuclear zinc complex, weaken the receptor-anion interactions, and result in preferential binding of OH^-. Host solutions were prepared under argon atmosphere in glove box at least 16 hours in advance to allow full dissolution, stored under argon and used within 24 hours of opening. The guest samples were also prepared with dry solvents and used within 12 hours. For each anion investigated, at least three concordant titrations were carried out with different host and guest concentrations (typically 1 or 0.5 mM and 10 or 20 mM respectively).

Calculations

Initial structures of the ligand and ligand-anion complexes were built using ArgusLab,³⁰ based on the crystal structure of the binuclear palladium complex [Pd₂(L)], and used as starting points for unconstrained geometry optimizations. The absence of imaginary frequencies was used to confirm that each calculated structure was a minimum on the molecular potential energy surface. All calculations were carried out using Gaussian09,³¹ running under Linux. Geometry optimisations and binding energy calculations were conducted at several levels of theory. Initial calculations employed Becke's popular three parameter exchange functional³² coupled with the Lee-Yang-Parr³³ correlation functional (B3LYP) in combination with the LANL2DZ basis set of Hay and Wadt,³⁴ which comprises an effective core potential (ECP) plus double zeta basis for heavy atoms, with the all-electron valence double zeta basis set developed by Dunning (D95V) for lighter atoms.³⁵ This basis set was chosen primarily due to the size of the complexes under study and the need to balance accuracy with computational cost, and, secondly, to satisfy the requirement for a balanced basis set spanning the atomic range from hydrogen to iodine. Comparative calculations using Pople's³⁶ 6-31g(d.p) basis set (which includes polarisation functions on all atoms) for the chloride, hydroxide and bromide complexes (it is not available for iodine) were undertaken to investigate the effect of basis set size on the calculated binding energies, which were observed to differ by at most 2 kcalmol^-1 (Table 2). Thus, the less computationally demanding LANL2DZ basis set was favoured. Both B3LYP/LANL2DZ and B3LYP/6-31g(d,p) have been shown predict calculated structures good agreement with those determined experimentally for large organometallic compounds.³⁷ However, the improper inclusion of dispersion interactions within B3LYP is of concern in the study of non-covalently bound complexes, such as those in this work. As a comparison the M05-2X functional of Zhao and Truhlar,³⁸which has been shown to recover significantly more mid-range correlation than B3LYP,³⁹ at a similar computational cost was also tested, again, in combination with the LANL2DZ basis set. Bulk solvent interactions were modelled via the self-consistent reaction field (SCRF) using the polarisable continuum model (PCM),⁴⁰ employing the default parameters for THF within Gaussian 09.

Ligand-anion binding energies were calculated using:

ΔE = E[[Zn₂(μ-X)(L)] + ZPE] - E[[Zn₂(L)] + ZPE] - E[(X)] + BSSE

where E[Zn₂(μ-X)(L)] is the energy of the solvated host:guest complex, E[Zn₂(L)] is the energy of the solvated complex alone, E[X] is the energy of the solvated anionic species, and the ZPE are the appropriate zero-point energy corrections for each component. The counterpoise method of Boys and Bernardi ⁴¹ was employed to correct for basis set superposition errors (BSSE) in the calculated binding energies, arising as a consequence of the incompleteness of the basis set.