Applications of halogen bonding in solution

The σ-hole

The concept of σ-hole has its origin in pure quantum mechanical calculations, as it can be traced back to the observation of a positive region, much like those presented in Fig. 1a in the surface potential of molecules as simple as tetrachloromethane [11]. Building on this observation, extensive research would follow, notably by Legon, Murray, Politzer, Clark and co-workers, which was summarized recently [4, 6, 7]. It is important to realize that the σ-hole exists even in absence of a bond to the halogen, it is a consequence of the anisotropy of the electronic distribution.

In the simplest case, if a halogen is bound to another atom, even more so if this other atom is electron withdrawing, some electronic density will vacate the halogen, resulting in a more pronounced expression of the already present anisotropy. The bonding itself will take place with the half-full p-orbital. Its electronic density participating to the bounding event, the direct opposite side is vacant and depleted in electronic density; this vacant p-orbital is, roughly speaking, the σ-hole. The other p-orbitals remain populated and are the basis for the negative belt that can be observed in Fig. 1a [1].

The consequence of this forced anisotropy localized among the orbitals gives rise to an important difference with hydrogen bonding: Lewis acids can interact with the negative belt giving halogens a dual Lewis acid and base character of the halogen (Fig. 1b).

It should be noted that the concept of σ-hole is not exclusive to halogens as it has been observed systematically with groups IV–VI of the periodic table.

Definition of halogen bonding

The definition of halogen bonding involved several steps. Presumably, halogen bonding was first observed more than 200 years ago in iodine. Indeed, the color change observed when titrating ammonia into iodine is due, formally, to halogen bond formation [12]. More recently, the Nobel prize winner Prof. Hassel observed the complex formed between, for instance, dioxane and bromine [13]. With the boom of X-ray crystallography, halogen bonding was observed in an increasing number of situations. A survey of the crystallographic database revealed many of the common characteristics of this noncovalent interaction [14].

Given the need of a definition, an IUPAC task group was appointed and proposed the following definition: “A halogen bond occurs when there is evidence of a net attractive interaction between an electrophilic region associated with a halogen atom in a molecular entity and a nucleophilic region in another, or the same, molecular entity” [15] which is by necessity wide.

To have a practical understanding of halogen bonding, particularly in the context of solution-phase, it is useful to return to the electrostatic potentials shown in Fig. 1a.

As halogen bonding includes an important electrostatic component, any Lewis base can act as halogen-bond acceptor (Fig. 1c). The extent in which a Lewis base is a good halogen-bond acceptor is not straightforward and relates to the concept of basicity scales [16, 17]; indeed, the affinity of a given Lewis acid, take a hydrogen-bond donor or AlCl₃ for instance, towards a Lewis base like a carbonyl oxygen, is not necessarily the same, as the interaction taking place can have different natures. In the case of halogen bonding, the electrostatic nature shared with hydrogen bonding gives a good indication to this end, but should be handled with care [18].

The directionality of the halogen bond is to be considered as well. Here again, pure electrostatic considerations help to understand this phenomenon, yet not necessarily to explain it completely: the presence of a negative residual region at 90° from the carbon–halogen bond suggests significantly less angular freedom as compared to the all “positive” hydrogen [1, 19]. As the σ-hole is affected naturally by the interaction with a Lewis base, the linearity of the halogen bond is better explained by charge-transfer interactions and lone pair repulsion [20]. The higher directionality of halogen bonding is often referred to and significant pieces of evidence exist for crystals; indeed, the overall angle observed for halogen bonds in the crystal database show that it is geometrically less flexible than hydrogen bonding [21]. In solution-like media, this geometrical preference is difficult to investigate and only a few examples could be related to the higher directionality of halogen compared to hydrogen bonding [22].

Applications of halogen bonding

Since the rediscovery of halogen bonding, it has mainly been applied in the solid state. Extensively exploited in crystal engineering for the past 20 years, it is hardly new for that community anymore. As this field is not covered here, the reader is referred to some recent reviews [10, 21, 23–25]. Applications include surface chemistry, where its directionality was intelligently exploited for the assembly of molecular monolayers [26, 27] and liquid crystals [28] among others [29].

Concerning halogen bonding in solution, two recent reviews have been published recently focusing on the methods used to investigate halogen bonding in solution [30], and the intrinsic thermodynamics and emerging applications [31]. The applications in solution will be discussed in detail in the following.

Scope

This review aims to summarize the applications of halogen bonding in solution-phase that have been reported during the past 10 years. With the exception of the monovalent binding events, i.e., one-to-one binding between a monovalent halogen-bond donor and a Lewis base, progress in the solution-phase is almost comprehensively restraint to the past decade.

The fields of anion receptors, molecular recognition, catalysis and anion transport are covered in a comprehensive way, while the applications in medicinal chemistry are addressed partially with representative examples.

Interlocked structures

The construction of interlocked structures has attracted the chemists’ attention since several years, not only from the purely proof-of-concept point of view, but also because of the numerous potential applications of nano-mechanical structures [36]. Nowadays, a substantial number of reports exist in the fields of nanomachines, nano-scale switches and nanoelectronics [37, 38].

An excellent example of how interlocked structures can be applied is substrate sensing. The unmatched ability of natural systems in substrate recognition is partially due to the existence of perfectly adapted pockets to recognize the substrate of interest. A similar situation can be reached with interlocked structures; especially, structures that use templates for their assembly are particularly well suited as receptors later on [39, 40].

Halogen bonding having similar characteristics to hydrogen bonding, yet being intrinsically more directional, its application could lead to more selective systems. Interlocked structures, pseudorotaxanes, rotaxanes and catananes, assembled partially and exclusively by halogen bonding, have been explored by the Beer group over the past 5 years. In that period of time, not only the possibility to access such constructs via halogen bonding was proven, but it has been used also in anion sensing applications [41, 42].

The natural first step was the pseudorotaxane formation by anion-templated assembly. The use of a halogen-functionalized imidazolium thread component 2 with isophthalamide macrocycle 1 in the presence of chloride anions gave access to the desired pseudorotaxane 3 (Fig. 3). While the use of halogens other than bromide was not reported, the replacement of bromine by hydrogen had an impairing effect on the affinity constants. Additionally, the methylated version, which showed similar affinities, was confirmed by ¹H, ¹H ROESY NMR spectroscopy to adopt an inverted conformation for stearic reasons stressing the strong linear tendency of halogen bonds [43].

Fig. 3

Pseudorotaxane employing halogen bonding reported by Beer and co-workers.

The assembly of the first rotaxane employing halogen bonds followed shortly after. Towards this goal, the 5-iodo-1,2,3-triazolium group was used as halogen-bond donor. Several model compounds (4a–d, Fig. 4) were evaluated for pseudorotaxane formation with the isophthalamide macrocycle 1 by means of NMR titrations. A preference for the bromide anion was observed with a 2-fold increase in affinity as compared to chloride and a 4-fold increase in the case of iodide. The 1,2,3-triazolium variant showed decreased yet detectable affinity, and the use of non-coordinating BF₄^– lead to barely detectable associations with macrocycle 1.

Fig. 4

Rotaxane assembly by anion-halogen bond templation.

The assembly of rotaxane 7 was achieved using 5-iodo-1,2,3-triazolium axle bromide 6 and bisvinyl-functionalized compound 5 with Grubbs’ second-generation catalyst in CH₂Cl₂ (Fig. 4). The formation of the rotaxane was confirmed by ¹H,¹H ROESY NMR spectroscopy, ESI-MS and single crystal X-ray diffraction [44].

Anion exchange with NH₄PF₆ gave access to a bromide anion receptor with affinities in the range of >10⁴ M^–1 in a 1:1 MeOH/CHCl₃ mixture, which were significantly higher than those reported before for the hydrogen bond variant [45]. The affinity of rotaxane 7 in a competitive water media was explored next and affinities >10³ M^–1 in MeOH/CHCl₃/H₂O (45:45:10) were observed for bromide and iodide with a preference for the latter. This selectivity was attributed to steric constraints.

The preparation of a concatenated structure being promising towards an enhancement of the steric, electronic, and lipophilic properties of the potential anion-binding pocket was attempted using the simplest version in the form of a [2]catenane. In this example, the authors employed bromo-4,5-methylimidazolium derivative 8, which mixed in equimolar quantities with the bromide version form a 2:1 complex with the bromide anion. Further reaction under ring-closing metathesis conditions with Grubbs’ second-generation catalyst yielded catenane 9 (Fig. 5). The catenane formation was confirmed by NMR spectroscopy and mass spectrometry [46].

Fig. 5

[2]Catenanes assembled by halogen-bond formation.

The presence of naphthalene units in the catenane precursor, and thus the catenane itself, was shown to report by fluorescence on the complexation of anions. Indeed, removal of the bromide template by anion exchange with NH₄PF₆^– gave a receptor selective for chloride (K_a > 10⁶ M^–1) and bromide (K_a > 10⁵ M^–1) in CH₃CN. The observed selectivity, 10 times better for chloride than bromide and the virtually non-affinity for other anions was attributed to the specificity of the binding pocket.

The assembly of [2]catenane structures was further explored by using pyridine-containing macrocycle 13. This approach was interesting because no anion templation was necessary and the main directing interaction was designed to be halogen bonding. The formation of the pseudorotaxane between macrocycle 13 and thread components 10a–b and 11a–c confirmed the expected increase stability with increasing polarizability (e.g., derivatives from iodine were better than from bromine), a preference to form the pseudorotaxane by halogen bonding over the less favorable hydrogen bonding and an increased stability with pyridinium over triazolium compounds, probably due to more favorable aromatic donor-acceptor interactions [47].

With the most promising candidate, the iodopyridinium derivative 11c, the catenane formation was explored, leading to catenane 14 as confirmed by NMR spectroscopy and mass spectrometry.

Similarly, the formation of pseudorotaxanes and catenanes was explored with the isophthalamide 1 used before for the first pseudorotaxane, leading to catenanes 12a–b. As their assembly is based on the anion-templation strategy, anion-binding affinities were investigated, but only weak associations were observed (e.g., 340 M^–1 for 12b and iodide in 1:1 MeOH/CHCl₃). The preference towards iodide over the smaller halides was in line with that observed for rotaxane 7, albeit for the rotaxane better affinities were observed.

The possibility to use exclusively halogen bonding to assemble interlocked structures with receptor/sensor capabilities was addressed by using the all-new design depicted in Fig. 6. A combination of bis-iodo-1,2,3-triazole 16 macrocycle precursor and bis-iodo-1,2,3-triazolium axle 15 under Stewart-Grubbs’ metathesis conditions in presence of chloride as template gave access to rotaxane 17 as confirmed by NMR spectroscopy and solid-state crystal structures. The rhenium(I) bipyridyl center was included in the macrocycle component to report on anion binding by fluorescence [48].

Fig. 6

Components and resulting [2]catenane assembled purely by halogen bonding.

The binding affinities, as reported by fluorescent quenching in competitive media, i.e., H₂O/CH₃CN up to 50 %, were as high as 10⁵ M^–1 with a higher affinity towards iodide and only weak binding to oxo-anions. While binding affinities decreased steadily with increasing water content, the selectivity between anions was naturally accentuated due to the hydration free energies. The observed binding affinities are the highest observed for halide-sensing interlocked structures, underlining the suitability of halogen bonds in this field.

Anion receptors

When considering the nature of halogen-bond donors, it is obvious that such interaction should give rise to, if any, anion receptors. Indeed, the formation of one-to-one complexes of several halogen-bond donors with Lewis bases has been reported repeatedly. Most notably, the discovery of halogen bonding is often traced back to the titration of iodine in aqueous ammonia in the 19^th century [12].

In this review we will not cover monovalent receptors such as iodine or perfluoroiodobenzene. Particularly, iodine has been exhaustively studied and the formation of complexes in solution and solid state are included in the more than 200 years of research [49]. In the context of halogen bonding, it is worth pointing to the research in the field of the diiodine basicity scale [18]. It establishes a quasi-orthogonality of the diiodine basicity scale with the better known hydrogen-bond, dative-bond or cation basicity scales. This is of paramount importance for the design of supramolecular systems employing halogen bonding in solution and solid state. Similar studies exist for other dihalogens [16, 17].

Monovalent organic halogen-bond donors have been characterized in solution for their affinities towards Lewis bases; examples including iodoalkynes [50], 2-iodo imidazolium salts [51], and perfluorinated haloalkanes and arenes have been reported [52] and recently reviewed [31]. Importantly, computational methods are becoming increasingly reliable for the study of molecular affinities in this context [53].

Ironically, the first multivalent “receptor” using halogen bonding was a cation receptor; halogen bonding was employed to enhance the receptor affinity in presence of a suitable anion. It was reported in 2005 by Resnati, Metrangolo and co-workers [54].

This receptor used a tertiary amine as scaffold: the tris[2-(2-hydroxyethoxy)ethyl]amine was functionalized with a iodoperfluorophenyl or perfluorophenyl units to yield receptors 18a–b (Fig. 7). On itself the tris[2-(2-hydroxyethoxy)ethyl]amine scaffold can complex cations in virtue of its structure that is, not surprisingly, similar to the family of crown and aza-crown ethers. It was hypothesized that upon substitution of the terminal alcohols with strong halogen-bond donors, i.e., a iodoperfluorophenyl unit, the cation binding would be enhanced by the peripheral coordination of halide anions.

Fig. 7

Anion receptors based on perfluorinated iodoarenes by Metrangolo, Resnati, Taylor and co-workers [54–56].

It was observed that cation coordination with cations such as Na⁺, K⁺ and Cs⁺ was indeed taking place, as evidenced by ESI-MS. In addition, crystal structures for the heteroditopic receptor for NaI were also resolved showing the expected Na⁺ coordination by the heteroatoms and halogen-bond formation in the periphery.

The affinity constants were measured by ¹⁹F NMR spectroscopy using the well characterized [18]crown-6 as reference, and the apparent association constant K_a(1) showed a 20-fold increase over K_a(2). This is the first reported example of a multidentate receptor based on halogen bonding and to report on association constants.

Interestingly, a high selectivity for iodide was observed. This was not surprising at that time, as it was incorrectly considered to be the “natural” strength scale in solution for halogen bonding [54]. It has since been established that the halide anion strength scale is similar to the one for hydrogen bonding, which is Cl^– > Br^– > I^– under non-biased conditions in solution [57].

During the following years, there were no reports on receptors using halogen bonding until 2010, when Taylor and co-workers reported a series of compounds (19–22, Fig. 7), among which receptor 22 was the first receptor using exclusively halogen-bond donors to achieve good anion-binding affinity (1.9 × 10⁴ M^–1 in acetone) by a real multidentate coordination. The chosen halogen-bond donor, 2,3,4,5-tetrafluoro-6-iodobenzylester, had the advantage of providing a convergent multidentated binding and, moreover, it allowed following the binding event by ¹⁹F NMR spectroscopy. As usual for receptor studies, anion bindings were determined and a 1:1 binding stoichiometry was in support of the proposed complex formation. The increase in binding affinities by increasing the number of halogen-bond donors and significantly lower associations in absence of the iodine provided conclusive evidence for this breakthrough example [55].

Several new bidentated and tridentated receptors were reported later based on the same halogen-bond donor unit and, whereas no better affinities were observed, it was clearly established that the strong directionality of halogen bonding is a key parameter to consider in the design of supramolecular structures using halogen bonding [58]. Moreover, it was reported that the use of perfluoroiodoalkanes and arenes to achieve binding results in a positive entropic contribution to binding [58, 59].

The possibility of combining hydrogen- and halogen-bond donors in receptors, not unlike other naturally occurring interactions, is attractive and was explored with compounds 23–25 (Fig. 7). The urea unit was used as primary binding interaction and the influence of potential halogen-bond donors in proximity was investigated [56].

Systematic structural changes in receptor 23a–b to assess the linker size and rigidity effect concluded that a short and flexible linker was best due to, on the one hand, the entropic penalty of a longer linker and the larger deviations from the ideal linear geometry of the halogen-bond formed with the phenyl linker, on the other hand.

While the contribution to binding by the halogen-bond formation could be assessed in all cases, the contribution of anion-π interactions had to be investigated for possible interference [60–62]. Receptors 24a–c were used to gain insights into this additional interaction. Affinities towards oxoanions were observed with phenolate anion being the best, with an affinity K_a = 1.9 × 10⁵ M^–1 in CH₃CN for 24a. Careful analysis of the data revealed that halides are recognized significantly better with the halogen-bond donors than with only the anion-π contribution (i.e., comparing 24a–24b).

Finally, receptors 25a–b were investigated; in this case, the halogen-bond geometries were predicted to be less penalized, and a higher contribution to binding (up to 2.0 ± 0.1 kcal·mol^–1) was measured due to halogen bonding. While phenolate remained the preferred guest, the halide anions showed a 30-fold increased affinity when comparing 25a–25b, confirming the major role of halogen bonding [56].

Perfluorohalo arenes and alkanes are by far the most exploited moieties in the halogen-bonding field and, not surprisingly, the first to be employed in solution applications, as well. Alternatives exist and were notably exploited by Beer and co-workers [23, 41, 51].

Taking into account the usefulness of the imidazolium unit as a hydrogen-bond donor and previous reports where halogen-bond formation was observed with 2-iodo- and bromoimidazolium, the possibility to use a 2-haloimidazolium unit to achieve anion binding was explored [63–65]. Imidazoliophanes 26a–b were investigated first: one important difference between imidazoliophanes 26a and 26b is evident by the possibility to isolate syn and anti conformers of 26a; Receptor 26b exists as a mixture of isomer due to the fast exchange at room temperature. For the binding studies, the two conformers had to be differentiated. It became evident from crystal structures and NMR titration experiments that the syn-26a was capable of binding simultaneously with both bromoimidazolium units leading to overall higher affinities [66].

These higher affinities were observed with syn-26a and bromide (889 ± 37 M^–1), followed by iodide, with undetected affinity for chloride, while 26b showed a relatively constant affinity for all three anions (ca. 100 M^–1) in 9:1 CHCl₃/H₂O. This selectivity highlights not only the usefulness of the bromoimidazolium as halogen-bond donor but also demonstrates one of the fundamental differences between hydrogen and halogen bonds: the steric demands which in this case allow for the preorganization of receptor 26a in the syn configuration and account for both the higher affinity and selectivity.

The same binding geometry was later exploited in imidazoliophane receptors 27a–d that incorporate naphthalene in the macrocyclic framework to serve as fluorescent reporter. Due to the larger macrocycle and the reduced steric demands, only iodoimidaliophane 27d was isolated as syn and anti conformers. A significant higher selectivity and affinity was observed with these receptors: out of the five receptors, only 27c and syn-27d would show changes in the fluorescent spectrum upon titration with a range of potential anions. Moreover, only bromide and iodide were suitable guests for these receptors with selectivity for bromide (27c) and iodide (syn-27d) with a ca. 20-fold preference for one anion over the other. With affinities close to 10⁶ M^–1 in competitive 9:1 MeOH/H₂O media, both receptors are exceptional. It is remarkable that 27c can rearrange to its syn-conformer upon anion binding with no significant penalization to the binding affinity. The counterintuitive and opposed selectivity for iodide or bromide is most likely explained by a competition between favorable binding geometries and solvation [67].

Synthetically easily to access, triazoles are promising neutral halogen-bond donors and have been explored with iodotriazole and triazole zinc(II)-metalloporphyrin receptors 28a–b (Fig. 8). A major advantage of such a system is the possibility to monitor binding following the changes of the absorption spectrum, namely the Soret band. Significant bathochromic shifts of the Soret band upon titration in CHCl₃ with tetrabuthylammonium (TBA) halide salt were observed with both receptors 28a–b, which was confirmed additionally by NMR titrations. Interestingly, oxoanions like acetate, hydrogenphosphate and sulfate showed similar shifts with significantly higher, over 100-fold, affinity constants [68].

Fig. 8

Anion receptors reported by Beer and co-workers.

While halide binding was best with the iodotriazole variant (halogen bonding), oxoanions show the opposite trend having better affinities with the triazole version (hydrogen bonding). A preference in binding of tetrahedral over trigonal anions was observed as well.

In other solvents, like acetonitrile and acetone, the affinity constants increased dramatically showing K_as > 10⁶ M^–1 for most anions with the noteworthy exception of iodide. Remarkably, the observed affinities only showed poor correlation with typical solvent parameters, especially in the case of 28b. The authors found a good correlation with the Guttmann acceptor number [69] which is an empirical measure of the Lewis acidity of the solvent. This is important as solvent effects on halogen bonding are still little explored and the available data suggest that significant differences exist as compared to hydrogen bonding [57, 59, 70]. Overall, halide recognition was better with halogen bonding, and oxoanions were best recognized with hydrogen bonding in this example [68].

Although not explicitly included in this section, due to their other functional properties, interlocked structures (e.g., 17) catalysts (e.g., 36) and anion-transporters (e.g., 60) are ipso facto anion receptors with affinities as high as 10⁵ M^–1 even in water competitive media. Importantly, studies exist where functional structures like 43a–b were investigated in detail for their anion affinities largely confirming their binding properties [20]. Interlocked structures are most interesting with regards to anion receptors, because their receptor pockets are reminiscent of natural phosphate and sulfate receptors [42].

Further examples on molecular recognition

The formation of host-guest complexes by multivalent halogen bonding is greatly dominated by the complexation of anions while the complexation of molecules, much like the interlocked structures prior to mechanical fixation, has received less attention. Two examples by Navarro-Vazquez, Rissanen and their respective co-workers describe in detail the formation of host-guest complexes. Additional NMR spectroscopic evidence confirms the existence of these complexes in solution [71, 72].

The work of Erdélyi and co-workers on the symmetry of halogen bonding has provided the community with significant insights that have potential applications in organic chemistry. They used pyridine-based halogen-bond acceptors and notably rigid bis-pyridines to assess the symmetry of the halogen bond formed with iodous and bromous halogen-bond donors.

Studied by a combination of NMR spectroscopy exploiting the method of isotopic perturbation of equilibrium processes [73] and computational methods, it was demonstrated that, in contrast to hydrogen bonding, halogen bonding is symmetric in solution [70, 74, 75].

Characterization of ion transport systems

The characterization of ion transport systems was one of the original bottlenecks in the research of ion transport systems. Indeed, detection of ion transport and the kinetics of such a process required the development of new methods. Those methods not being usually encountered, as compared to NMR titrations or ESI-MS spectroscopy, a brief introduction is given below of the most common ones. For detailed descriptions the reader is directed to the specialized literature [103, 104].

Large unilamellar vesicles

Probably the most common method to assess synthetic ion channels nowadays; it consists of using large unilamellar vesicles (LUVs). For this experiment, LUVs are formed by various methods (dialysis, extrusion, etc.) with defined lipid composition, typically dipalmitoyl phosphatidylcholine or dioleoyl phosphatidylcholine or the naturally occurring mixture “Egg yolk” phosphatidylcholine, and defined content dispersed in an aqueous solution of equal or similar composition.

An ion gradient is generated by the addition of a base, typically NaOH, and upon the addition of a putative ion transporter the evolution of the ionic species of interest can be followed by several methods: ion selective electrodes [96, 105], fluorescence [106], etc.

Fluorogenic LUVs are particularly interesting as they can be used in virtually any laboratory because no special equipment is required. The most versatile fluorophore used in fluorogenic LUVs is 8-hydroxy-1,3,6-pyrenetrisulfonate (HPTS) [107] which is used in the so-called HPTS assay [103]. HPTS is a pH-sensitive fluorophore and measures the efflux or influx of anions or cations by the concomitant change in pH due to proton or OH^- transport (Fig. 13a). The HPTS assay cannot, a priori, discriminate between the different processes and mechanisms that lead to a pH-change, and as such it is the ideal method to be used in initial screening studies. Variations exist that allow the discrimination of symport and antiport transport mechanisms, anion versus cation selectivities, etc. These more specific experiments are described elsewhere [103, 109].

Fig. 13

The most common methods to characterize synthetic ion transport systems. (a) The HPTS assay, a typical experiment with fluorogenic vesicles. (b) Typical fluorescent trace for a single HPTS experiment. (c) Typical dose-response curve to determine the effective molarity needed to obtain 50 % transport activity (EC50) and the Hill coefficient (n). (d) Typical set-up for planar bilayer conductance experiments. (e) Current flow changes in response to a carrier anion transporter. Adapted from reference [108] with permission, © 2012 Nature Publishing Group.

In a typical experiment, Egg yolk phosphatidylcholine LUVs loaded with HPTS (EYPC-LUVs⊃HPTS) are subjected to, first, a NaOH pulse followed by the addition of the putative transporter and finally the equilibration of the system by adding a detergent to calibrate for the maximal emission (Fig. 13b). The fluorescence is recorded at two different excitation wavelengths in order to determine the ratiometric fluorescence kinetic trace which can be directly correlated to ion transport. If the fractional transport value, i.e., the highest relative value before addition of the detergent, is reported as a function of the transporter concentration, the dose-respond curve (DRC, Fig. 13c) of the transporter is obtained. This transporter-specific DRC reports the effective molarity needed to obtain 50 % transport activity (EC₅₀) and the Hill coefficient (n) which can give information regarding the number of molecules required to achieve transport [103]. Other fluorophores can be used in a similar manner to address membrane stability, i.e., carboxyfluorescein [103], or to detect specific anions like chloride with lucigenin [110].

Anion transport with anion-π interactions and halogen bonding

Shortly after the first reports on anion receptors using halogen-bond donors, the possibility to exploit this noncovalent interaction to transport anions across the lipid bilayer membrane was explored. The first anion transport system reported was designed to allow for multitopic binding to anions. The calix[4]arene scaffold was selected in order to take advantage of the coordination of the tetramethylamonium (TMA) cation to the upper rim by cation-π interactions. Perfluoro-m-iodophenyl groups were employed as halogen-bond donors, while the fully fluorinated aromatic version was used as control (60, 63, Fig. 14) [113].

Fig. 14

Calix[4]arenes decorated with halogen-bond donors and anion-π surfaces to achieve anion transport in bilayer membranes; proposed binding of tetramethylammonium (TMA) cation and chloride are shown.

Good anion transport in the HPTS assay with calix 63 as compared to the almost negligible activity of calix 60 was attributed to favorable anion-π interactions for the former and excessive binding of the latter. Both situations have been reported in the literature: anion-π interactions have been used to achieve anion transport in comparable simpler systems [101, 114], and inhibition by excessive binding is known as the goldilocks principle [85].

Weaker halogen-bond donors with calix 61 or a less favorable conformation (62) resulted in significant improvement in the anion transport activities. Further affinity investigations by ¹⁹F NMR spectroscopy confirmed binding by halogen bond formation and the proposed one-to-one stoichiometry for all receptors/transporters except calix 62 with a one-to-two stoichiometry. Additional evidence for the proposed involvement of halogen bonding was implied by the significant anion selectivity of the transport system.

Anion transport with halogen-bond donors

Only moderate transport activities with calix-based transporters and less than favorable binding geometries observed by DFT models suggested that, on the one hand, the calix[4]arene scaffold was detrimental for the activity and, on the other hand, perfluorinated iodoarenes halogen-bond donors were nicely suited for the task in hand.

The possibility to achieve anion transport using exclusively halogen-bond donors in absence of a pre-organizing scaffold was investigated next. Several perfluorinated halogen-bond donors (64–73, Fig. 15a), were tested in the HPTS assay, and surprisingly high activities were observed. Contributions from anion-π type interactions were safely excluded with transporters 64–69 where no π-surface exist. Also were excluded effects originated from perfluorinated materials (74) and pure anion-π interactions (75). Weaker halogen-bond donors (75, 78) were inactive as was the stronger yet too hydrophilic diiodopyridinium 77 [108].

Fig. 15

(a) Halogen-bond donors explored in the context of transmembrane anion transport. (b) Top, proposed structure responsible for anion-translocations; bottom, ball-and-stick model of chloride bound to six perfluoro-1-iodobutane by DFT simulations. Adapted from reference [108] with permission, © 2012 Nature Publishing Group.

Additional experiments performed with variations of the HPTS assay confirmed high anion selectivity, high dependence on membrane viscosity and thus suggested an anion antiport mechanism ruling the transport.

The observed activities with small molecules were astonishing. While anion transport has been reported with simple molecules [115], encounting molecules like trifluoroiodomethane (6), which is a gas at room temperature (b.p. –22 °C) that was bubbled directly through the suspension of EYPC-LUVs⊃HPTS to activate anion transport and represents the smallest organic anion transport system reported, was intriguing and required further confirmation.

Conductance experiments in planar bilayer membranes confirmed the existence of transport by a carrier mechanism (Fig. 13e) and revealed high anion selectivity.

The proposed mechanism of transport implies the formation of an unstable hexa-coordinated complex with the anion in the middle (Fig. 15b), which in virtue of the hydrophobicity of the outer shell shuttles through the lipid bilayer and exchanges the anion for a hydroxide. This is supported by the observed requirement of a minimum of five molecules to achieve transport in the HPTS assay.

Halogen-bond donor channels

The last available contribution by Matile and co-workers was the development of the family of anion transporters depicted in Fig. 16. To maximize anion transport, a formal channel, i.e., a molecule or ensemble of molecules capable to promote anion translocation without moving significantly, was proposed. The p-octaphenyl scaffold has been reported before, using hydrogen bonding to achieve proton transport by an ion-hopping mechanism [116]. It was hypothesized that a similar process could take place with halogen bonding.

Fig. 16

Halogen-bond cascades explored for channel-like anion transport in bilayer membranes; arrows show the proposed mechanism of action.

The transport activity of rigid rods 79–82 was investigated in the HPTS assay and an over 2300-fold increase in transport activity compared to the simple pentafluoroiodobenzene (70) was observed. Moreover, the observed activity per halogen-bond donor increased faster than exponentially with a cooperativity coefficient [117] of 3.37, a cooperativity coefficient that strongly support the proposed anion-hopping mechanism. Indeed, purely multivalent systems like polymeric constructs typically show cooperativity coefficients between 1 and 2 [118].

Additionally, the anion-π analogue, i.e., the fully fluorinated version, showed significantly lower activities confirming the suitability of halogen bonding for anion transport and the generality of the p-octaphenyl scaffold as applied to ion transport. Additional experiments confirmed anion selectivity and membrane stability, as well. The proposed mechanism involves the perpendicular arrangement in the lipid membrane of at least one rigid rod and subsequent hopping of the anion from one halogen-bond donor to the next.

With detectable transport activity at lipid/transporter ratios as high as 20 000:1 for 82, high selectivity, even with incredibly small systems like 69 and having explored most transport mechanism (i.e., sym- and antiport carriers and channels), halogen bonding is confirmed as a valuable and reliable noncovalent interaction to achieve anion transport.

Halogen-bonded gels

Gel formation is a widespread application of supramolecular chemistry. The use of noncovalent chemistry has the additional advantage to give self-healing properties and to reversibly trigger gel formation by external stimuli [131, 132].

The first proof-of-the-concept example where halogen bonding is used to trigger gel formation was reported recently [133]. Formation of two-component gels was possible using compounds 94 and 95 in presence of halogen-bond linker 98 in competitive methanol/water mixtures upon rapid cooling. Similarly, with gelator 96 it was possible to trigger gelation by addition of TBACl or 4,4′-bipyridine (97) in DMSO/water. Equimolar mixtures of 94 and 96 also formed gels in DMSO/water. The microscopic structure of all three gels differed significantly: there was loss of organization, which was highest for the first case, and increasingly thinner fibers for the latter two.

These examples prove the possibility to form gels either by the use of gelators including the halogen-bond forming groups or by introducing partners of gelation.

Halogen bond-assisted electron transfer

A kinetic study of the redox reaction of halogenated methane, namely tetrabromomethane and tribromonitromethane, with electron donor N,N,N′,N′-tetramethyl-p-phenylenediamine (TMPD) has shown that electron transfer in this case proceeds by an inner-sphere electron-transfer mechanism as opposed to the electron transfer of the same acceptor with decamethylferrocene, which occurs by an outer-sphere electron transfer [134].

Such a situation should be considered in systems where electron transfer is important, like photosystems [135].

Assembly of gold nanoparticles on quartz and silicon surfaces

It has been demonstrated that functionalization of gold nanoparticles with halogen-bond donors (99, Fig. 19) leads to aggregation of the nanoparticles in the presence of a linker accepting halogen bonds (101). Aggregation of gold nanoparticles was easily followed by changes in the surface plasmon absorption band: in absence of the ditopic halogen-bond acceptor or with monotopic capper 104 no changes in the absorption spectrum were observed. In a similar way, using perfluorophenyl 100, which lacks the halogen-bond donor, only minor changes were observed. These observations were further confirmed by transmission electron microscopy (TEM) [136].

Fig. 19

Gelators and halogen-bond linkers explored by Resnati, Metrangolo and co-workers; Molecules applied to the functionalization of gold nanoparticles [133, 136, 137].

The same strategy was applied to functionalize silicon and quartz surfaces. Indeed, quartz and silicon functionalized with monolayers of terminal-pyridine organic compounds can interact with gold nanoparticles functionalized as described before. Subsequent growth of the gold nanoparticles layer was accomplished in presence of different linkers, (102–104) showing significant differences when employing linear (102), 2D (103) and 3D (104) linkers, confirming that molecular geometry influences, while not transferring, the aggregation patterns as confirmed mainly by atomic force microscopy (AFM) [137].

The possibility to use halogen bonding in this context suggests the orthogonal application in surface patterning. It also indicates that the formation of well-defined nanoparticles-based assemblies is possible.

Purification of iodoperfluoroalkanes by halogen bonding

Halogen bonding between iodoperfluoroalkanes and chloride has been used to separate the former from hexane. Using strong ion exchange, that is silica functionalized with 3-hydroxy-N,N,N-trimethylpropan-1-aminium chloride, it was possible to separate iodoperfluoroalkanes, mono- and diiodo, methane to dodecane, from hexane in a column like set-up [138].

Further evidence confirming the binding and selective extraction of analytes was provided by UV/Vis, Raman and NMR spectroscopy. This system was applied to the recovery of iodoperfluoroalkanes in soil samples and showed good performance in the selective recovery of diiodo compounds with ng·g^–1 accuracy.