Halogen Bonding in Supramolecular Chemistry - Chemical Reviews

Halogen Bonding in Supramolecular Chemistry - Chemical Reviews...

4 downloads 206 Views 38MB Size

Review pubs.acs.org/CR

Halogen Bonding in Supramolecular Chemistry Lydia C. Gilday, Sean W. Robinson, Timothy A. Barendt, Matthew J. Langton, Benjamin R. Mullaney, and Paul D. Beer* Chemistry Research Laboratory, Department of Chemistry, University of Oxford, Mansﬁeld Road, Oxford OX1 3TA, United Kingdom 4.4.5. XB with Anions in the Solid State 4.5. Solid-State Architectures 4.6. Applications and Functional Materials 4.6.1. Analytical Chemistry, Chemical Resolution, and Recycling 4.6.2. Surface-Bound Polymers and Thin Films 4.6.3. Liquid Crystals 4.6.4. Surface Relief Gratings 4.6.5. Nanoparticles 4.6.6. Interfaces 4.6.7. Materials with Optical Properties 4.6.8. Solid-State Molecular Recognition 4.6.9. Topochemical Reactions and Solid-State Synthesis 4.6.10. XBs with Radicals 4.6.11. Conducting and Magnetic Materials 4.6.12. Halogen-Bonding Supramolecular Gels 4.7. Conclusions and Outlook 5. Halogen Bonding in Biological Systems 5.1. Analysis of the Protein Data Bank: Halogen Bonds to Amino Acids 5.2. Measurement of Biological Halogen Bonds: Assessing the Contribution to Ligand Binding Aﬃnities 5.2.1. Direct Measurement 5.2.2. Indirect Measurement of the Biological Halogen-Bond Strength 5.3. Computational Approaches to Halogen Bond Structure: Energy Relationships in Biological Environments 5.3.1.1. Hybrid QM−MM Computational Approaches 5.3.1.2. Low Computational Cost Methods: Application to High-Throughput Drug Screening 5.4. Halogen Bonding in Medicinal Chemistry: Success Stories and Application to Systematic Drug Discovery 5.4.1. Halogenated Herbicides 5.4.2. Practical Implications for Drug Discovery 5.5. Transmembrane Anion Transport Mediated by Halogen-Bond Donors 5.6. Conclusions and Outlook 6. Halogen Bonding in Solution

CONTENTS 1. Introduction to Halogen Bonding 1.1. Nature of the Halogen Bond 1.2. Scope of the Review 2. Computational and Theoretical Investigations of Halogen Bonding 2.1. Quantum Mechanics Methods 2.2. σ-Hole Model of Halogen Bonding 2.3. Other Contributions to the Nature of Halogen Bonding 2.4. Recent Examples of Computationally Investigated Halogen-Bonded Complexes 2.4.1. XB to Neutral Species 2.4.2. XB to Anions 2.4.3. XB in Protein−Ligand Complexes 2.4.4. Electron-Transfer Processes Aﬀected by XB Interactions 2.5. Classical Force Field Calculations 2.6. Conclusions and Outlook 3. Gas-Phase Studies of Halogen-Bonding Interactions 4. Halogen Bonding in the Solid State 4.1. Introduction to Crystal Engineering and Functional Materials 4.2. Fundamentals 4.3. Halogen-Bonding Hierarchy 4.3.1. Ranking Halogen-Bond Donors 4.3.2. HB/XB Complementarity/Competition 4.3.3. Predicting XBs 4.4. Control of Solid-State Supramolecular Architectures 4.4.1. Polymorphism 4.4.2. Stoichiometry 4.4.3. Tautomeric Control 4.4.4. XBs Involving Metals and Metal-Bound XBs © 2015 American Chemical Society

7119 7119 7120 7120 7120 7120 7122 7123 7123 7126 7127 7127 7127 7129 7130 7131 7131 7132 7134 7134 7136 7138 7138 7138 7139 7140

7142 7144 7145 7145 7147 7148 7149 7150 7150 7151 7152 7153 7155 7155 7156 7156 7158 7158

7161 7161 7162

7163 7163

7163

7164 7166 7166 7167 7168 7168

Special Issue: 2015 Supramolecular Chemistry Received: November 28, 2014 Published: July 13, 2015

7140 7118

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews 6.1. Thermodynamics of Neutral Halogen Bonding in Solution 6.1.1. Inorganic Halogen-Bond Donors 6.1.2. Organic XB Donors 6.1.3. Comparisons between Organic and Inorganic XB Donors 6.1.4. Weaker XB Interactions in Solution 6.2. Halogen-Bond-Templated Self-Assembly 6.3. Anion Recognition and Sensing 6.4. Reactivity and Catalysis 6.5. Conclusions and Outlook 7. Conclusion Author Information Corresponding Author Notes Biographies Acknowledgments Glossary References

Review

solid-state gas matrixes at low temperature showed the presence, albeit small, of lattice eﬀects in these systems.12 Using molecular beam resonance spectroscopy, Klemperer and co-workers demonstrated that HF acts as a Lewis base in “anti-hydrogen-bonded” complexes with dihalogens (HF···ClF and HF···Cl2).13 The term halogen bonding was ﬁrst used by Dumas in 1978 to describe interactions of CCl4, CBr4, SiCl4, and SiBr4 with tetrahydrofuran, tetrahydropyran, pyridine, anisole, and di-n-butyl ether in organic solvents.14 It was not until the mid-1990s, however, that the nature and applications of the halogen bond began to be intensively studied.3 Extensive and systematic work by Legon and coworkers, who used matrix isolation and supersonic expansion of gases to study the rotational spectra of prereactive gas-phase complexes formed between dihalogens and NH3, H2S, CO, C2H2, C2H4, or HCN, drew attention to the similarities between halogen-bonding and better-known hydrogen-bonding interactions.15 The ability of iodo- and bromoperﬂuorocarbons to form noncovalent interactions with neutral and anionic electron donors was reported by Resnati and co-workers, who highlighted the ability of haloperﬂuoroalkanes to function as halogen-bond donor moieties in crystal engineering, and revealed the widespread promising potential of XB interactions in supramolecular chemistry.16 Currently, XB is being exploited for a range of functional applications. In the solid phase such applications include liquid crystals,17 controlling electrical and magnetic properties,18 nonlinear optics,19 the separation of isomers,20 and control reactions in the solid state.21 In the solution phase, while considerably less developed than in the solid state, applications have been reported in catalysis,22 in interactions with metal complexes,23 for anion binding in the solution and solid states24 and in biological systems,25 in drug design,26 and in protein− ligand complexation.25b,27

7168 7169 7169 7171 7171 7172 7174 7180 7184 7184 7184 7184 7184 7184 7185 7186 7186

1. INTRODUCTION TO HALOGEN BONDING Supramolecular chemistry is “the chemistry of the intermolecular bond, covering the structures and functions of the entities formed by association of two or more chemical species”.1 In addition to the coordinate bond, there are a variety of well-established noncovalent intermolecular interactions, such as electrostatic, hydrogen-bonding (HB), aromatic donor−acceptor, cation− and anion−π, and solvatophobic interactions. Halogen bonding (XB) is arguably the least exploited noncovalent interaction, especially in the solution phase, and its promising potential in supramolecular chemistry applications is only now beginning to be realized. A halogen bond (R−X···Y−Z) 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”.2 The ability of halogen atoms to interact with Lewis bases has been known for some time.3 As early as 1814, adduct formation between iodine and ammonia was observed,4 which was identiﬁed as NH3·I2 by Guthrie in 1863.5 Analogous complexes with other dihalogens followed shortly after. 6 Similar phenomena were reported between aromatic hydrocarbons and iodine in 1948,7 and the nature of analogous complexes with ether, thioether, and carbonyl solvents was reported in 1950: the interaction then ascribed to electron donor−acceptor molecular complexes.8 In the 1950s and 1960s, Hassel and coworkers studied crystal structures of molecular halogen and Lewis base complexes, describing the interaction as a “halogen molecular bridge”, and Bent suggested “that the O···Br−Br interaction is energetically comparable to a strong hydrogen bond”.9 In his 1970 Nobel lecture, Hassel summarized his results and highlighted the importance of halogen atoms in selfassembly.10 In the early 1970s, solution studies of organic bases such as amines, amides, alcohols, and thiols revealed that XB can compete with HB: the peak in the infrared spectrum arising from solute−solute interactions was observed to be signiﬁcantly weakened by the addition of a cosolute that could halogen bond, and the hydrogen-bond-breaking ability was shown to increase in the order F < Cl < Br < I.11 Furthermore, infrared spectroscopy of Lewis base−dihalogen complexes isolated in

1.1. Nature of the Halogen Bond

An XB interaction is typiﬁed by an X···B interaction between a halogen-bond donor (X) and an acceptor (B), wherein the internuclear distance is less than the sum of their van der Waals (vdW) radii. The XB interaction is typically colinear with the R−X covalent bond: indeed, in the gas phase, the angle between components is usually at least 175° (although deviations do occur as a result of secondary interactions or crystal packing eﬀects). XB is comparable in strength (energies up to 200 kJ mol−1)16d to the more well-known intermolecular interactions such as HB: the workhorse of the supramolecular toolbox. Similar trends in energetic and geometric properties are found for HB and XB interactions. Halogen atoms are larger than protons, however, so XB interactions can be more sensitive to steric bulk or secondary interactions. Furthermore, the strength of halogen bonds can be ﬁne-tuned by varying the halogen atom and the motif to which it is covalently bound: increasing the electron-withdrawing ability of the substituent leads to increased halogen-bond donor strength. The exact nature of the halogen-bonding interaction is still disputed. Transfer of electron density occurs from the Lewis basic site (electron donor, halogen-bond acceptor) to a Lewis acidic site (electron acceptor, halogen-bond donor), but the bonding contributions of this noncovalent interaction (electrostatic eﬀects, charge transfer, polarization, and dispersion forces) often depend upon on the individual interacting species. 7119

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

1.2. Scope of the Review

complexes. In particular, DFT−SAPT allowed for the treatment of extended complexes (∼35 atoms), and the total interaction energy was decomposed into electrostatic, induction, dispersion, and exchange-repulsion terms.34 Quantum theory of atoms in molecules (QTAIM)35 and natural bond order (NBO) analysis36 have also been used to calculate the distribution of electron density in halogen-bonded complexes, and provided evidence in some cases of chargetransfer-type interactions. The natural orbitals for chemical valence (NOCV)37 combined with the extended-transition-state (ETS)38 energy composition analysis approach has been used by Michalak and co-workers. This method allowed the separation and quantitative assessment of the contributions to deformation density originating from the electron-charge-transfer channels to describe the XB interaction.39 The interacting quantum atoms (IQAs) method40 allows decomposition of the total energy of the system into intraatomic and interatomic contributions, from which interfragment interaction energies can be calculated. Within this approach, classical Coulombic electrostatic and exchangecorrelation contributions can be discriminated. Intermolecular perturbation theory (IMPT)41 was used by Allen and co-workers to determine the nature and geometry of intermolecular interactions between halogens and oxygen or nitrogen.42

XB is emerging as a noncovalent interaction to exploit in modern supramolecular chemistry. In particular, in recent years the ability of polarized halogen atoms to interact with neutral or anionic Lewis bases in the gas, solid, and liquid phases has received an ever-increasing amount of attention. Furthermore, comparisons with control species or hydrogen-bonding analogues are revealing unique characteristics of XB in terms of strength, selectivity, and interaction geometry. Much of the current understanding of XB interactions is based upon a combination of ab initio calculations, gaseous-phase studies, and X-ray crystal structure determination; solution-phase examples of XB interactions are much rarer. This review presents a summary of the recent advances in XB, with a particular focus on supramolecular chemical applications.

2. COMPUTATIONAL AND THEORETICAL INVESTIGATIONS OF HALOGEN BONDING This section gives an overview of how computational methods have been applied to understand the nature of XB interactions, starting with quantum mechanics (QM) and moving on to methods based on classical force ﬁelds. 2.1. Quantum Mechanics Methods

QM is used to study weak intermolecular interactions because the characteristics of the noncovalently bonded system can be diﬃcult to separate from other eﬀects in condensed phases. Information on molecular geometry, energy, vibrations, and bonding interactions can be elucidated, and simple approaches also exist to allow the eﬀects of solvent and neighboring groups to be computationally assessed. For a more comprehensive discussion on various QM methods and their applicability to XB interactions, readers are directed to a review by Karpfen28 (a theoretical characterization of trends in XB, in complexes of the types B···XY and B···X− CF3) and recent papers by Chudzinski and Taylor29 and Hobza and co-workers,30 which also delve into molecular mechanics methods (further discussed in section 2.5). Herein we will highlight the most commonly used methods. XB interactions are often relatively weak, and the interaction potentials are quite soft; consequently, only very high-level QM calculations that take into account electron correlation give reliable results, which are in turn sensitive to the chosen level of electron correlation. Self-consistent-ﬁeld molecular orbital approaches may be used to compute electrostatic surface potentials with optimized Slater-type orbitals with Gaussian functions (STO-3G*) to rationalize certain characteristics of XB interactions, although the calculations on this minimal basis set are diminishing.31 In geometry optimization, the counterpoise (CP) correction to the basis set superposition error (BSSE) often needs to be included. The Møller−Plesset second-order (MP2)32 approach is conventionally used to calculate the intermolecular potential energy surfaces but can overestimate XB interactions compared with the coupled cluster singles and doubles approach including a triples contribution via perturbation theory (CCSD(T)), by which calculations are performed at the complete basis set (CBS) limit. Density functional theory (DFT) methods have also been used to describe halogen-bonded complexes. Decomposition of the total interaction energy into physically well-deﬁned terms has been undertaken by a number of methods.33 Symmetry-adapted perturbation theory (SAPT) analysis has been used for energy decomposition of XB

2.2. σ-Hole Model of Halogen Bonding

Halogen atoms are among the more electronegative elements, and covalently bonded halogens are usually considered to be electron rich. It is somewhat counterintuitive, therefore, that halogen atoms can interact with Lewis bases. Politzer and coworkers’ extensive studies of the electrostatic potential surface of halogen-bonding systems provide one rationale to the phenomenon of halogen bonding.25b,43 A free, neutral, ground-state atom has, on average, a spherically symmetric distribution of positive electrostatic potential everywhere (the nuclear attraction dominates over that of the surrounding electrons).44 It is only when atoms are bonded that surrounding electronic charge is redistributed, being polarized toward the bonding region, which results in regions with more or less electron density on the atom’s equatorial or bond terminus, respectively.45 This anisotropic charge distribution is usually shown by plotting the electrostatic potential VS(r) on the molecular surface, typically taken to be the 0.001 au (e/bohr3) contour of its electron density and which has been shown to resemble the van der Waals surface:46 those for covalently bonded halogen atoms in CF3X are shown in Figure 1.25b,43a,c,47 The electrostatic surface potential VS(r) at any point in space was calculated mathematically, but it may be experimentally determined using diﬀraction techniques.48 Figure 1 shows that when X = F, all halogen surfaces have negative potential as expected for a very electronegative atom. As one of the ﬂuorine atoms is substituted for a diﬀerent

Figure 1. Molecular electrostatic potential, in hartrees, at the 0.001 e/ bohr3 isodensity surface of CF3X (from left to right, X = F, Cl, Br, I). Reprinted with permission from ref 43b. Copyright 2007 Springer. 7120

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

VS,max.51 Politzer and co-workers also studied the interactions between F3M−X or H3M−X (M = C, Si, or Ge; X = Cl, Br, or I) and NCA (A = H, F, Cl, Br, or I). A linear ΔE−VS,max relationship was observed for 18 R−X···NCH complexes (R = F3M and H3M, calculated from the M06-2X/6-311G(d) energy minima at 0 K), as illustrated in Figure 2, which the authors interpreted as support for an electrostatic driving force for XB.50,52

halogen, however, a region of positive potential develops (red), becoming larger as the size of the halogen increases, and correlates with increased halogen polarizability, as illustrated for CF3Br and CF3I molecules. This region of positive potential at the terminus of the C−X bond is known as the σ-hole43b and is surrounded by a belt of negative potential. The radius of X contracts along the extension of the C−X bondan eﬀect sometimes referred to as polar ﬂattening.45f,h Politzer and co-workers have also used NBO population analysis of CF3X to rationalize why σ-holes form in some situations but not in others.43b In its neutral state, a halogen atom has three p-orbitals, and each contains an average of 5/3 electrons. In CF3X one of the X atom’s electrons is involved in a C−X bonding orbital, σCX, and the others form three lone pairs. Two lone pairs were determined to occupy the px- and pyorbitals on X (perpendicular to the C−X bond axis); the third lone pair is in an orbital which is mostly s-character in nature, with a small degree of p-hybridization along the C−X axis. This p-hybridization was calculated to be much greater with F (25%) compared with when X is Cl (12%), Br (8.5%), or I (8%). The contribution of each X to the C−X bond was determined to be primarily a p-orbital; the s-contribution to the C−X bond was only 12% (Cl), 9% (Br), and 9% (I) compared with 25% when X = F. Consequently, the three lone pairs can be considered to occupy s2px2py2-orbitals when X is Cl, Br, or I (similar to an X+ ion with a vacant p-orbital), forming a belt of negative potential around the equator of atom X, and only the outermost region, the σ-hole, retains the positive potential intrinsic to the free neutral atom. When X = F, the ﬂuorine atoms had a much larger share of the bonding electron density owing to their much greater electronegativity, and also there was found to be a signiﬁcantly larger degree of sp-hybridization. Both factors were considered to prevent the existence of σ-holes in CF4. In the σ-hole model XB occurs through the interaction of the σ-hole with electron-rich partners, and its ﬁxed position along the C−X bond axis governs the near linearity of R−X···B interactions. The sides of polarized halogens being negatively charged also allow for the possibility of side-on interaction with positive sites.49 Since this intermolecular interaction is a result of the polarization of the atom’s electronic charge toward the covalent bond, anything that enhances this polarization will increase the magnitude of the σ-hole: increased polarizability of the atom itself or the electron-withdrawing nature of other atoms/groups within the molecule. The most positive potentials VS,max of the σ-holes of halogen atoms in a series of molecules have been determined.50 For a given molecular framework, RX, the value of VS,max increased in the order X = F < Cl < Br < I, which reﬂected the falling electronegativity of halogen atoms as well as the increasing polarizability on descending group VII. The positive potential was also found to increase with the greater electronwithdrawing nature of R. Computed interaction energies ΔE give a measure of the strength of the interaction and have been calculated for some halogen-bonded complexes, and the trends have been found to correlate with the most positive potential of the σ-hole: VS,max. For a given halogen-bond acceptor B, on changing X from F to Cl to Br to I or when R is more electron withdrawing, the VS,max becomes more positive and the interaction energy becomes more negative (a stronger interaction). Riley et al. investigated complex formation between halobenzenes and acetone and demonstrated that the interaction energies correlated well with the respective values

Figure 2. Interaction energies versus halogen σ-hole VS,max values for 18 R−X···NCH complexes. The VS,max values are on the R−X halogens prior to interaction. Reprinted with permission from ref 50. Copyright 2013 Wiley-VCH Verlag GmbH & Co. KGaA.

Using electrostatic-surface-potential maps calculated at the CAM-B3LYP-D3(bj)/def2-QZVP level, Hobza and co-workers reported the size and magnitude of the σ-hole for a range of substituted aromatic XB donor groups (Figure 3) and showed

Figure 3. Five aromatic cores which were substituted by various chemical groups. X represents a halogen atom or hydrogen; a, b, and c are positions substituted with hydrogen or F, CN, NO2, and CH3 groups.53

that, with increasing magnitude of the σ-hole, the stabilization energy between XB donors and hydrogen ﬂuoride increased and the directionality decreased. The electrostatic-energy contribution was determined to be an important contribution to binding energies, and angular variation was attributed to exchange-repulsion and electrostatic terms.53 Lu et al. have calculated ΔE, ΔG, ΔH, and ΔS for complexes formed between perﬂuoroiodobenzene or perﬂuoroiodoethane and Lewis bases NH3, H2O, Cl−, Br−, and I− using the quantum mechanical DFT/B3LYP method.54 Although gas-phase ΔE and ΔH values were calculated to be favorable (negative), ΔS values were large and negative. Large decreases in entropy were observed for some HB processes55 and, indeed, in any gasphase complex formation, since the process is accompanied by a loss of translational and rotational degrees of freedom.56 Unfavorable entropy changes can be countered by strong 7121

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

the electron density is locally concentrated (0) at a given point in space. Furthermore, the Laplacian is related to the kinetic and potential components of the total energy and can indicate the presence of shared or closed-shell interactions61 (∇2ρ) > 0. Sosa and co-workers carried out a systematic study at the B3LYP/6-311++G** level of theory on halogen bonds formed between Lewis bases (NH3, H2O, H2S) and acids (ClF, BrF, IF, BrCl, ICl, IBr).60a A linear relationship between the interaction energy and electron density at the halogen-bond critical point was determined, which was cited as evidence that this topological indicator is a good descriptor of the halogen-bond interaction strength.

interaction energies or lattice/solvent eﬀects to give an overall thermodynamically favorable process; interactions with halide ions gave suﬃciently negative ΔE values (e.g., for C6F5I···Cl−, ΔE = −25.8 kcal mol−1). Some complexes, however, were calculated to be thermodynamically unstable in the gas phase if ΔE and ΔH are only slightly favorable (e.g., for C6F5I···NH3 and C6F5I···OH2, ΔE = −5.4 and −3.1 kcal mol−1 with ΔG = 2.3 and 3.4 kcal mol−1, respectively). Using DFT calculations (PBE functional) and the polarizable continuum model (PCM), for the solvent, Lu et al. also computed solvation eﬀects on the overall stability of these R− X···B (R = C6H5, C6F5, C2F3) complexes in cyclohexane, chloroform, acetone, and water.54,57 In solvent, all halogen bonds remained highly linear, and the order of halogen-bond donor strength (I > Br > Cl) persisted in both the gas and solution phases. Furthermore, the strength of the complexes with neutral bases (NH3, H2O, and acetone) changed little between the gas and solution phases; calculated interaction energies became slightly more positive with increasing solvent polarity (the solvent has a small destabilizing eﬀect on weak halogen bonds), although the intermolecular X···Y distances decreased for C6F5Br···OH2(NH3) and C6F5I···OH2(NH3) dimers. For the remaining neutral complexes, the X···Y distances were calculated to elongate or remain unchanged. With anionic bases such as Cl−, Br−, and I−, however, the interaction energies became signiﬁcantly less negative on going from the gas phase into solution. The contribution of the Lewis base to σ-hole bonding is also signiﬁcant and has been explored by Politzer and co-workers.58 Nitrogen in NH3 was calculated to have a more negative electrostatic potential than in HCN (VS,min = −46 and −33 kcal mol−1, respectively), and interaction energies with R1R2Z (R1 = R2 = F, CN, or CH3; Z = O, Se, or S) were computed at both the B3PW91/6-311G(3df,2p) and the MP2/6-311++G(3df,2p) levels using the B3PW91 geometries and determined to be more negative for NH3 than HCN. The correlation between ΔE and VS has been cited as evidence that XB interactions are driven by electrostatic attraction between the partially positively charged σ-hole on X in a general molecule R−X and the (partial) negative counterpart B. An interesting cooperative eﬀect has been calculated by Politzer et al. for X−CN species (X = F, Cl, Br): these contain both halogen-bond donor and acceptor portions and were shown to form linear chains of XB interactions.59 The authors found that the magnitude of the interaction energy calculated at the B3PW91/6-311G(3df,2p) level for forming two XB interactions (X−CN···X−CN···X−CN) was more than double the value for forming one (−2.9 vs −1.3 kcal mol−1). This was rationalized by the changes in polarization (an intrinsic component of electrostatic interactions) of the entire molecule on the formation of one XB interaction: the electric ﬁelds of the positive σ-hole (R−X) and the negative site (B) induced rearrangements in the electronic densities of both B and R−X, respectively, the extent of which depended on both the polarizabilities and strengths of electric ﬁelds on each component. In X−CN species, these charge density rearrangements served to enhance the donor ability of the noncoordinating nitrogen atom and the acceptor ability of the uninvolved σ-hole for the second interaction. In addition to this electrostatic viewpoint, the σ-hole can also be visualized through density Laplacian maps.60 The Laplacian of electron density (∇2ρ) provides a measure of the local curvature of the electron density (ρ(r)) and indicates whether

2.3. Other Contributions to the Nature of Halogen Bonding

The purely electrostatic nature of XB interactions has been questioned by some research groups, and several computational studies have been reported which suggested that the extent of electrostatic versus charge-transfer character varies with the nature of the XB donor group.45g,51b,62 Weaker XB interactions may closely correlate with values of the maximum electrostatic potential, but extended studies to encompass a broader range of XB complexes show it is important to consider dispersion, charge transfer, and other factors. Coupled cluster (CCSD(T)/aug-cc-pVTZ) calculations performed at the CBS limit on halomethane···formaldehyde and CH3−nFnX···OCH2 (n = 0−3) halogen-bonded dimers aﬀorded binding energies in the range of −1.05 to −3.72 kcal mol−1, and SAPT analysis (vide supra) of these complexes revealed the halogen bonds depended on both electrostatic and dispersion interactions.51b,63 Dimers comprising weaker halogen bonds (e.g., X = Cl, Br) were shown to be principally dispersive in nature, but with some electrostatic contributions. As the size of the halogen atom or the extent of ﬂuorination increased, however, the electrostatic nature of the halogen bond increased while the dispersive contribution to the halogenbonding interaction diminished. The authors suggested that even when the contribution was relatively small, electrostatic eﬀects probably played a signiﬁcant role in determining the geometries of halogen-bonding systems. A recent perspective paper by Riley and Hobza summarized theoretical investigations which have sought to distinguish the relative importance of electrostatic and dispersion interactions.30 Calculated CCSD(T)/aug-cc-pVTZ energies of interactions between bromomethanol molecules revealed that, in general, the halogen-bonded complexes studied depended more strongly on dispersion and less strongly on electrostatics than their hydrogen-bonded analogues, where electrostatics dominated. Exchange repulsion was also calculated to be greater for hydrogen bonding, which was attributed to the reduced intermolecular separation in a hydrogen bond. The solvation penalty was also determined to be diﬀerent from (and lower than) that of a typical hydrogen bond. Tuning of the halogenbonding interaction by systematic ﬂuorination of bromomethane was also investigated, and the contribution from electrostatics signiﬁcantly increased with successive ﬂuorine substitution. Indeed, stronger electrostatic attraction can shift the equilibrium separation of the complex, which gives rise to strong attraction from dispersion and induction eﬀects as well as increased repulsion from exchange-repulsion eﬀects. The relative importance of electrostatic, dispersion, and induction contributions to HB and XB interactions was recently described by Riley and co-workers, who calculated the singlepoint energies at the CCSD(T)/aug-cc-pVTZ level of the 7122

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

substituted organic molecules and anionic bromometallates revealed the importance of the covalent component in XB.75 Charge-transfer interactions for bromine-containing molecules and bromide anions have also been studied experimentally and theoretically (using ab initio MP2 and DFT calculations with dispersion-corrected wB97X-D, B97-1, and M06-2X functionals in combination with the 6-311+G (d,p) basis set). A limited correlation between interaction energies and VS,max was seen, but a much better correlation was obtained with a combination of electrostatic and orbital components, which mirrored the variations in inter- and intramolecular separations.76 As summarized in Figure 4, HOMO/LUMO mixing is related to the energy lowering arising from expanding the

possible intermolecular interactions in bromo- and iodomethanol dimers (geometries were optimized at the counterpoisecorrected MP2/cc-pVTZ level). The authors determined that dispersion interactions play the largest role in stabilizing XB interactions, while the electrostatic energy dominates for the corresponding HB analogues.64 Hobza and co-workers suggested the origin of XB may be diﬀerent in diﬀerent complexes (perturbation SAPT analysis was used to determine the energy contributions) since in most cases the dispersion energy was the signiﬁcant contributor and only in the strongest halogen bonds did the electrostatic energy become dominant.65 Furthermore, they suggested that the electrostatic (polarization) energy was responsible for the structure of a halogen bond and the dispersion energy for its stabilization. As early as 1989, Desiraju and Parthasarathy reported evidence that charge-transfer or donor−acceptor eﬀects contribute to the overall halogen-bonding intermolecular interactions9b,49c,66 (Politzer has rebutted their arguments67). Trends in the charge-transfer halogen-bonded complexes between amines and dihalogens or of the type X−Y···B have been theoretically explored by Karpfen and co-workers, who suggested the properties of such XB interactionsfor example, dipole moment enhancements and intermolecular charge transferparallel those characteristic of HB.28,68 More recently, Chudzinski and Taylor investigated complexes of electron-poor iodo species with nitrogen- or oxygen-containing Lewis bases and reported poor correlation between the free energies of XB, calculated using a variety of methods for geometries optimized at the MP2/6-31+G(d,p) level of theory,29 and those predicted by the pairwise electrostatic interaction model.69 Zou et al. calculated complexes between organic/inorganic halogens and ammonia at the MP2(full)/Aug-cc-PVTZ, MP2(full)/ Lanl2DZ* or CCSD(T)/Aug-cc-PVDZ level of theory and suggested the XB interactions were dominated by a chargetransfer-type contribution.62a,c Using relativistic DFT at zeroth-order regular approximation ZORA-BP86/TZ2P, Wolters and Bickelhaupt have computationally investigated a range of strongly X-bonded trihalides DX···A− and their hydrogen-bonded DH···A− analogues (D, X, A = F, Cl, Br, I).70 The authors presented a physical model of the halogen bond based on quantitative Kohm−Sham molecular orbital (MO) theory, energy decomposition analysis (EDA), and Voronoi deformation density (VDD) analysis of the charge distribution71 from which they determined that both HB and XB interactions have a sizable covalent component arising from HOMO−LUMO interactions. The Kohn−Sham MO approach was also employed by Palusiak to highlight the covalent nature of halogen bonds formed between CH3X and formaldehyde.72 Li and co-workers performed quantum chemical calculations at the MP2/aug-cc-pVTZ level to study the “halogen-hydride” XB interactions of BH3NH3 with dihalogens. QTAIM (MP2(FC)/aug-cc-PVTZ level) and NBO analyses suggested chargetransfer contributions were important in complex formation.73 Rosokha and co-workers have undertaken experimental studies in conjunction with theoretical investigations to probe the nature of XB interactions. Similar Mulliken dependence of the absorption spectra was observed for XB complexes involving CBr4 or CHBr3 and which are more typical chargetransfer complexes.74 Comparing the locations of intermolecular contacts with the electrostatic surface potentials and molecular orbital shapes (calculated at the B3LYP/6311+G(dp) level) for complex formation between bromo-

Figure 4. Simpliﬁed molecular-orbital diagram of the intermolecular complex (right) resulting from the frontier orbital mixing (the HOMO of the electron donor with the LUMO of the electron acceptor), as illustrated for the CBr4/Br− association (left). Reprinted with permission from ref 76. Copyright 2013 Wiley-VCH Verlag GmbH & Co. KGaA.

variational space of interacting species to include unﬁlled orbitals, and the formation of a bonding orbital of the supramolecular complex.76 The stability of the complex depended on the energy diﬀerence and overlap of the orbitals involved; maximum overlap with the R−X bond σ*-LUMO occurred when the electron donor’s HOMO was oriented along the R−X bond, giving an alternative rationale for the strict linearity of XB interactions. Recently, Solimannejad and co-workers performed MP2/ aug-cc-pVTZ calculations to analyze the X···X interactions in crystalline dihalomethane CH2X2 compounds. Their results demonstrated the X···X distances to be shorter than the sum of the van der Waals radii, and with a linear orientation. According to SAPT, the stabilities of such interactions were largely dependent on dispersion eﬀects.77 In summary, the origin of XB interactions depends, in part, on the nature of the interacting species. If XB is accounted for mainly by electrostatic attraction, the thermodynamic and structural characteristics of the complexes are expected to vary in parallel with the maximum potential on the surface of the halogen-bond donors (VS,max). On the other hand, if the bonding is related primarily to charge-transfer (orbital) interactions, the same characteristic should be correlated with the amount of charge transfer (DQCT) or stabilization energy (ECT). The following sections highlight some of the recent advances in computing halogen-bonding interactions and reveal the varying nature of the interaction. 2.4. Recent Examples of Computationally Investigated Halogen-Bonded Complexes

2.4.1. XB to Neutral Species. 2.4.1.1. Dihalogen Halogen-Bond Donors and Lewis Bases, XY···B. Some of 7123

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

the simplest halogen-bonded complexes are those formed between dihalogens XY and ammonia. Indeed, the ﬁrst experimentally characterized halogen bond to be reported was the complex H3N·I2. Many theoretical calculations have since been applied to these types of complexes, including Hartree− Fock (HF), MP2, and various DFT approaches.62a,68,78 Since then, calculations have encompassed a wide range of dihalogen−Lewis base complexes. Through DFT-D3 calculations (B97D/def2-QZVP), Hobza and co-workers determined the theoretical binding energies of the XB found in crystals of bromine- and iodine-containing TTF derivatives with I3− and I2. Among the complexes analyzed, neutral complexes with iodine were found to be the most stable (−13.8 kcal mol−1 for I2···1,3-dithiole-2-thione-4carboxylic acid). In all complexes studied, the dispersion energy contribution is important, and is comparable to the electrostatic energy: only in the I3−···I2 dimer, with a very strong halogen bond, was the electrostatic interaction found to dominate.65a Calculations beyond the MP2 approximation, even for systems composed of a few atoms, are quite rare. CCSD(T) calculations including geometry optimizations were undertaken by Karpfen for ClF and Cl2 interacting with ammonia.68a Hobza and co-workers have used CCSD(T) calculations at the CBS limit on many halogenated complexes as a benchmark for parametrization and validation of lower level QM methods.79 Large stabilization energies for halogen-bonded complexes 1,3dithiole-2-thione-4-carboxylic acid (DTCA)···I2 and 1,4diazabicyclo[2.2.2]octane (DABCO)···I2 were determined by CCSD(T)/CBS and compared with those calculated by other computational approaches. DFT-D (B97-D3/def2-QZVP) overestimated the strength of these interactions, a common problem with charge-transfer-type interactions; the M06-2X density functional coupled with the def2-QZVP basis set, however, was more reliable. DFT−SAPT analysis revealed charge-transfer energy, included in the induction energy, stabilized the halogen bond in these complexes. Alkorta and co-workers have investigated the structures and binding energies of FCl···N and FCl···CNQ (Q = CN, NC, NO2, F, CF3, Cl, Br, H, CCF, CCH, CH3, SiH3, Li, Na) Xbonds and analogous H-bonds using MP2/aug′-cc-pVTZ calculations, and the one- and two-bond spin−spin coupling constants across the complexes were also determined using EOM-CCSD calculations.63c−h The authors classiﬁed the calculated interactions as “traditional”, “chlorine-shared”, and “ion-pair” types of X-bond depending on the relative lengths of the F−Cl and Cl−B bonds in the FCl···B complex. In the complex F−Cl···CN−SiH3, the Cl−C distance is 49% of the sum of the van der Waals radii, the F−Cl bond is lengthened by 0.26 Å, and ΔE = −15.5 kcal mol−1: these features were attributed to the presence of a degree of additional covalent character to the σ-hole interaction. Extending this work, Karadakov and co-workers studied the halogen bonds F−X··· CNQ, where X = Cl, Br, and I, and found diﬀerent types of halogen bonds (traditional, ion-pair, and halogen-shared) are distinguishable when X = Cl but purely traditional XB interactions occur when X = I.80 Politzer and co-workers considered these properties to be atypical of XB interactions and, through investigating the related complexes formed between Cl−Cl and either CNR or SiNR, attributed these observations to strong polarization of the halogen-bondaccepting C or Si atom by the electric ﬁeld of the positive σhole of the F−Cl chlorine atom, which may in turn result in some sharing of electrons by the C or Si atoms.81

Wang and co-workers have performed a number of calculations on iodine dimers (I−I···I−I) as prototypes of general X···X interactions, in addition to SAPT analyses on I2 dimers and benzene dimers.82 Sandwich, parallel-displaced, and T-shaped geometries were computed, and the validity/ reliability of the various computation methods undertaken was summarized. Electrostatics, dispersion, and induction eﬀects were all determined to contribute to the halogen-bond interaction. Alkorta and co-workers have reported that pure and hybrid DFT, and ab initio computational methods, can overestimate interaction energies for XB interactions to anionic XB acceptors by comparison to binding energies and equilibrium distances obtained using the CCSD(T)/aug-cc-p-VTZ level of theory. Furthermore, the authors do not recommend the use of dispersion-corrected methods to properly describe XB interactions.83 A recent study by Alkorta and co-workers used MP2/aug′-ccpVTZ calculations to probe the interactions in the complexes H2FP···ClY. Both single molecules possess σ-holes and lone pairs, and should be able to form either halogen-bonded or pnictogen-bonded complexes depending on which species acts as the electron-pair donor or acceptor. Three diﬀerent types of halogen bonds (traditional, chlorine-shared, and ion-pair) were identiﬁed, as were two pnictogen bonds (Figure 5).84 With

Figure 5. Schematic representation of the halogen-bonded and pnictogen-bonded complexes H2FP···ClY.

electronegative substituents Y = F or CN, X-bonded complexes were identiﬁed (sketches XB-T, XB-S, and XB-IP), whereas when Y = Me or H pnictogen-bonded complexes resulted (sketches ZB-1 and ZB-2). The authors also used NBO analyses to show that charge-transfer contributions stabilized both pnictogen-bonded and halogen-bonded complexes; the dominant XB charge-transfer interaction involved P lone pair donation to the antibonding σ*Cl−Y orbital. 2.4.1.2. Haloﬂuoromethane Halogen-Bond Donors and Lewis Bases, CF3X···B. The strength of an XB interaction has been shown to depend on the size of the σ-hole, which may be enhanced by covalently attaching electron-withdrawing groups to the halogen atom. The ability of perﬂuorohalo-functionalized alkanes and aromatics (vide infra) to function as halogen-bond donors has been computationally investigated. 7124

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

and 3- and 4-substituted pyridine complexes with BrCl and BrF at the B97D/6-311+G(d) level of theory. In addition, the combination of para-substituted bromobenzenes with parasubstituted pyridines has also been studied. Reasonable statistical parameters were obtained (R2 > 0.9) when compared to the interaction energies, and the authors suggested it provides a basis on which to extend investigations to a much larger set of compounds to be studied.89 Meng and co-workers performed MP2 calculations using the aug-cc-pVDZ or MP2/aug-cc-pVDZ-PP (for iodine) basis set, QTAIM, and noncovalent interaction (NCI) index studies on a series of X···N halogen bonds between halogen-bond donors C6F4XY (X = Cl, Br, I; Y = F, CN, NO2, LiNC+, NaNC+) and halogen-bond acceptor 1,2-diaminoethane. The work presented was consistent with typical trends observed for halogen-bond donors: electron-withdrawing groups enhance the size of the σhole, and the maximum positive electrostatic potential VS,max, strengthens the halogen-bonding interaction.90 2.4.1.4. Aromatic π-Systems as Halogen-Bond Acceptors. The XB acceptor motif is often a lone pair on a heteroatom such as nitrogen or oxygen, although XB interactions can also form to π-systems. The stability of XB complexes between dihalogens and aromatics, including pyridine, furan, and thiophene, have been calculated. According to the Legon− Millen rules used to predict the angular geometries of B···HX complexes, for a complex in which B has both a lone pair and πelectrons, the lone pair dominates as σ-type interaction occurs. This was conﬁrmed for complexes of the type pyridine−XY (X, Y = F, Cl, Br), optimized at the MP2/aug-cc-pVDZ level of theory (where the stronger electrostatic potential is located at the pyridine nitrogen atom).91 This is in contrast to similar interactions between the same dihalogens and furan92 or thiophene,93 where the π−halogen-bonded interaction was the most stable. Using NBO analysis, the pyridine−XY interaction has been calculated to have signiﬁcant n(N) → σ*(XY) chargetransfer character with electrostatics determined to contribute between 9% and 28% of the total attractive interaction energies for all complexes studied.91 Wang et al. have performed an SAPT energy decomposition analysis for the σ− and π−XB interactions in thiophene−XY complexes and concluded they were predominantly inductive in nature, in contrast to the electrostatic interpretation put forward by Zeng et al.94 XB interactions between benzene and halocarbon derivatives were determined to be predominantly dispersive, with only a very small charge-transfer contribution, and their strength was found in the narrow range of −1.29 to −3.16 kcal mol−1, comparable with that of analogous CH−π intermolecular forces.95 Halogen-bonded interactions between CF3X and both aromatic and nonaromatic π-systems, such as toluene, ethene, and propene, were investigated by Hauchecorne and coworkers.96 Infrared and Raman spectroscopic studies provided evidence for the formation of 1:1 and 1:2 C−X···π halogen bonds; complexation enthalpies were determined using spectra recorded at diﬀerent temperatures and compared with those measured for the equivalent, hydrogen-bonded complexes. The experimental complexation enthalpies of C−Br···π and C− H···π were similar, and weaker than that of the C−I···π halogen bond, which was supported by enthalpies obtained by ab initio calculations. Even less common are XB interactions between non-carbonbased aromatic π-systems and halogen atoms. Qi et al. studied the interactions between the electron-deﬁcient positive aromatic three-membered ring (BNN)3+ and dihalogen

The attractive XB interactions between halomethanes and rare gases have been evaluated through QM calculations. Potential energy surface minima obtained were shown to correspond to linear and T-shaped geometries of approach, much more anisotropic than had been previously expected.85 The precise structures of noble-gas−halogen complexes are still to be determined; the interaction energies were found to be relatively low ( Br− > Cl− > F−.16b Karadakov and co-workers predicted, and subsequently calculated using a variety of ab initio and DFT methods with the aug-cc-pVTZ and aug-cc-pVTZ-PP basis sets, that the strength of the interaction should increase with the size of the noble gas.86 The distance between X and B (B = Ar, Kr, or Xe) is typically slightly less than the sum of the van der Waals radii, and the binding energies indicate (weakly) attractive interactions dominated by dispersion interactions. The X···B distances decreased and binding energies increased as X became more polarizable/electron-deﬁcient, and the interactions also became stronger as with increasingly polarizable raregas halogen-bond acceptors. In particular, the authors’ calculations for CF3I···Kr agree well with those determined experimentally, which is cited as evidence of the accuracy of their calculations and the validity of the many ab initio and DFT methods analyzed. The formation of hydrogen-, halogen- and/or chalcogenbonding interactions between hypohalous acids YOH (Y = F, Cl, Br, or I) and dimethylchalcogen derivatives X(Me)2 (X = S, Se, or Te) was analyzed by Alkorta and co-workers at the MP2/ aug′-cc-pVTZ computational level.87 The most stable interaction energies were obtained for XB with BrOH and IOH, and NBO analysis demonstrated that interactions resulted from donation of chalcogen lone pairs to the antibonding orbital of the O−Y bond; red shifts were observed in the O−Y stretching modes. DFT−SAPT calculations were also carried out and suggested that the XB interaction was mostly electrostatic in nature with signiﬁcant contributions from dispersion (FOH) or induction (IOH) in certain cases.87 2.4.1.3. Aromatic Halogen-Bond Donors. The use of Xfunctionalized aromatic groups as halogen-bond donors is also an intense area of research, driven in part by the relative ease of incorporating a variety of electron-withdrawing groups which can enhance polarization at the halogen atom. Linear free energy relationships (LFERs) in halogen bonds formed between pyridine or cyanobenzene derivatives and functionalized iodobenzene species have been studied by Frontera and co-workers at the RI-MP2/aug-cc-pVDZ level of theory. Good correlations in the Hammett plots (EBSSE vs σ) were obtained (0.967 < R2 > 0.982) for X-bonded complexes of X-functionalized pyridine and cyanobenzene derivatives. Furthermore, when electron-withdrawing and -donating groups were included, a good correlation with the standard values of the aromatic substituent constant (σ-values) was observed, which is cited as evidence that the inﬂuence of substituents on the binding energies is due to induction eﬀects.88 Alkorta and co-workers have explored the Hammett−Taft LFER using the optimized structures of an impressive number (459) of Xbonded complexes involving substituted iodo- and bromobenzenes complexed with simple Lewis bases (NH, NCH, and CNH), 4-substituted 1-bromobicyclo[2,2,2]octanes with NH, 7125

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

favored lone pair interaction, particularly in the presence of electron-withdrawing groups attached to the carbonyl carbon atom.104 2.4.1.6. Cooperativity of XB Interactions. Del Bene et al. have analyzed the cooperativity of X bonds in the formation of ternary halogen-bonded complexes by varying the nature of one component and studying the systematic eﬀect on the structure, binding energies, and spin−spin coupling constants.105 Recently, Alkorta and co-workers used DFT calculations to explore cooperativity in 4-functionalized pyridine···XCN···XCN triads (X = Cl, Br); the electronic properties of the complexes were studied using the molecular electrostatic potential (MEP) and minimum average local ionization energy, with parameters derived from AIM and NBO methodologies.106 The ternary complexes were calculated to have shorter intermolecular distances than those in the corresponding binary complexes, which was attributed to the cooperative eﬀect of halogen bonds in these systems. In all dyads and triads, the polarization component of the total energy was the most important stabilizing contribution and was calculated to be 44−51% of all the attractive contributions. The ternary complexes were found to be more stable than two binary complexes, and a linear relationship was found between the cooperativity and the total stabilization energy in the ternary complexes.106 Solimannejad and co-workers investigated the complementarity in halogenbonded complexes NCX···NCX···XCH3 (X = Cl, Br), demonstrating the cooperative eﬀects of the X···X interaction on the strength of the X···N bond, which were dominated by electrostatic attractive forces.107 2.4.1.7. Hard−Soft Acid−Base Perspective. Herrebout, De Proft, and co-workers investigated XB interactions from a hard−soft acid−base (HSAB) perspective: hard and soft interactions were distinguished and characterized by atomic charges, electrophilicity, and local softness indices.108 Despite having a weaker contribution than electrostatics, charge-transfer interactions were computed to have a pivotal role in stabilizing complexes formed by carbon-based X-bond donors. The relative importance of the n → σ*-type orbital interaction was larger going from Cl to Br to I (XB donors) and from O to N (XB acceptors). Unexpected C−X bond strengthening was observed for some complexes and rationalized using molecular orbital theory: negative charges near X increase the energy of the X atomic orbital, and the C−X bond becomes more covalent/less polar, overcoming the bond-weakening eﬀect of n → σ*-type charge transfer. 2.4.1.8. XB Involving Metal Complexes. Arunan and coworkers reported theoretical investigations which showed that square pyramidal Fe(CO)5 can function as an X-bond acceptor, forming X-bonded complexes with ClF and ClH. The XB interaction with ClF is signiﬁcantly stronger, which was attributed to the larger size of the σ-hole on the chlorine atom of this molecule. QM calculations revealed signiﬁcant mutual penetration occurs, although the distance between Fe and Cl is very close to the sum of their van der Waals radii in Fe(CO)5···ClH.109 2.4.2. XB to Anions. In addition to lone pairs on heteroatoms or the π-bond, anions can also act as halogenbond acceptors. Triiodide, I3−, is a limiting case of the more general D···X−Y halogen bond, in which there are two equal I−I linkages shorter than in the comparable halogen-bonding systems, but longer than the bond length in a I2 molecule. Triiodide formation is usually due to I− + I2 addition, beginning with I2 polarization

molecules XY employing MP2 calculations based on the 6311+ G(2d) and aug-cc-pVDZ basis sets and found the calculated binding energies followed the expected trends in the dipole moment of the dihalogen, the polarizability of X, and the electron-withdrawing strength of Y.97 All-metal aromatic species such as Al42−, initially described by Li et al.,98 were also able to establish a π−XB interaction with halohydrocarbons (XCH3, XCH2CH3, XCHCH2, and XCCH, with X = Cl, Br, I).99 The structure, properties, and nature of these π−XB interactions were investigated using QM calculations. One particular type of π−XB interaction was strong (ΔE = −21.64 kcal mol−1 with iodoethyne in a side-on geometry, determined at the MP2/aug-cc-pVDZ level). 2.4.1.5. Nonaromatic π-Systems. XB interactions with allylic π-systems have also been computationally investigated. Using ab intio (MP2/-311+G**) and DFT (B3LYP/6311+G**) methods, Lenoir and Chiappe have calculated the formation of a donor−acceptor π−XB complex to be an essential intermediate in the electrophilic bromination of alkenes and alkynes, in reasonable agreement with experimentally determined thermodynamic values.100 Systematic studies of complexes between ethene derivatives C2H4−nFn and dihalogen XF (X = Cl, Br) complexes were carried out by Li and co-workers at the MP2/aug-cc-pVDZ and CCSD(T)/aug-cc-pVDZ levels, who reported a reduction in interaction energies upon F substitution, while the π−XB bond was shorter and stronger when X = Br than when X = Cl.101 Matter et al. used ab intio methods to calculate the interaction energies of halobenzene···benzene dimers, the results of which were in agreement with experimental data and highlighted the importance of the XB interaction in the recognition of the biomolecule AVE-3247 (see later sections).27 Very recently, Zhuo et al. performed theoretical calculations using a number of computation methods with the aug-cc-pVDZ basis set on a series of complexes formed between XCCF derivatives (X = Cl, Br, I, chosen since sp-hybridization and electron-withdrawing F substituents are known to enhance the XB donor strength) and acetylene or its analogues RCCR (R = Na, CuPH3, AuPH3, AgPH3). The stability of the complexes increased with increasing π-donor ability of the acetylene and with the polarizability of the halogen donor atom X. The authors reported a new class of C−X···H halogen bonds where the π-electrons are donated to the FCC−X antibonding orbitals with simultaneous back-donation by the halogen atom lone pair into the RCCR antibonding orbital as revealed by NBO and atoms in molecules (AIM) analyses.102 Alkorta and co-workers have probed the XB interactions formed between F−Cl and PCY (Y = NC, CN, F, H, CCH, CCF, CH3, Li, Na) to investigate the eﬀect of substituent Y on the XB acceptor ability of PCY molecules. Using a combination of MP2/aug′-cc-pVTZ level calculations, SAPT analysis, and computed spin−spin coupling constants, two diﬀerent XB interactions were identiﬁed depending on the nature of the substituent Y: in complexes with F−Cl···P the XB interaction was formed through the P lone pair, while in complexes involving the interaction of F−Cl the XB interaction involved the PC triple bond.103 In a study of host−guest inclusion complexes between cucurbit[6]uril and molecular dibromine and diiodine, X-ray crystallography and QM calculations provide evidence of perpendicular XB interactions between dihalogen and urea residues, which revealed that halogen bonding with the carbonyl π-system can be competitive with the commonly 7126

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 6. Structure of XB acceptor and donor groups investigated by Riley and Hobza in the study of XB interactions in protein−ligand complexes.113

backbone nitrogen with the proper number of hydrogen atoms (obtained by atomic manipulation of the corresponding crystal structures).113 The use of DFT−SAPT is unrealistic for such large, computationally demanding systems, and instead, a comparison of Hartree−Fock (HF) and MP2 results was used to estimate the relative contributions from electrostatic and dispersion forces. 2.4.4. Electron-Transfer Processes Aﬀected by XB Interactions. Experimental rates of electron-transfer (ET) reactions of halogenated molecules in which the halogen atom is sterically hindered are consistent with outer-sphere electron transfer (Marcus theory). If reagents form halogen-bonded complexes, however, rate constants may be many orders of magnitude higher than predicted by the outer-sphere theory.114 Rosokha and co-workers have studied the reactions of brominated electrophiles R−Br (R−Br = CBr3NO2, or CBr4) with well-known halogen-bond acceptors (N,N,N′,N′-tetramethyl-p-phenylenediamine and decamethylferrocene, prototypical electron donors) to establish the role of halogenbonding prereactive species in these ET reactions.115 High reaction rates were well-accommodated if the electronic coupling of R−Br and halogen-bond acceptors (evaluated from the spectral and structural characteristics of their halogenbonded complexes) was taken into account.

and subsequent electron transfer at short internuclear distances. The I−/I3− redox couple is a fundamental component in dyesensitized solar cells (DSSCs) and ionic liquid electrolytes. Mealli and co-workers have developed a DFT-based model of I3− and I42− anions with a focus on the σ-electron density delocalization rather than a simple electrostatic interpretation.110 Esterhuysen and co-workers have performed a theoretical study of the triiodide ion and the I3−···I3− dimer (both in the gas phase and in an implicit continuum solvent model using various levels of theory and basis sets) to understand the driving force behind the formation of [I3−]∞ chains and to study the eﬀect of a uniform electric ﬁeld (as found in solution) on the I3− ion and its interactions.111 The MP2/cc-pVTZ-PPcalculated I3− bond length and I3−···I3− intermolecular distances were consistent with those obtained from the Cambridge Structural Database. Moreover, the inexpensive WFT method (MP2/TZ) yielded values for the I3−···I3− interaction energy closest to those obtained at the CCSD/a-TZ/MP2/TZ level of theory. Calculations in vacuum gave a repulsive interaction and a global energy minimum at inﬁnite separation. In solvent, however, the repulsive electrostatic contribution to the interaction energy decreased, so the total interaction energy became more stabilizing. Furthermore, QM calculations revealed a substantial dependence on the dielectric constant on the ethanol or water solvent medium. Awwadi and co-workers used crystallographic and ab initio calculations to investigate simple R−Y···X− (R−Y···X−; R = methyl, phenyl, acetyl, or pyridyl; Y = F, Cl, Br, or I; X− = F−, Cl−, Br−, or I−) interactions with a focus on the eﬀect of changing the nature of the R groups and the halogen atom. The distance dependence and angular dependence of the halogenbond strength were studied and provided further evidence that XB interactions are characterized by a linear C−Y···X− geometry and a Y···X− separation less than the sum of the van der Waals radii. The strength of halogen···halide XB interactions depended on the halogen atom, the halide anion, the hybridization at the ipso-carbon and the nature of the R group.112 2.4.3. XB in Protein−Ligand Complexes. Hobza and coworkers have investigated the strength and character of XB interactions in protein−ligand complexes. Many of these X bonds were chosen as “best case scenarios” in terms of interaction energies within protein−ligand complexes (Figure 6); i.e., the model complexes composed of a ligand and the protein backbone carbonyl group bound to the neighboring

2.5. Classical Force Field Calculations

Describing XB using molecular mechanics (MM) is diﬃcult, owing to the inability of conventional force ﬁelds to reproduce the anisotropic charge density distribution on the halogen atom. However, the methods based on classical force ﬁelds are of paramount importance in studying the dynamic behavior of large supramolecular systems comprising XB interactions. Indeed, the quantum methods described above were applied in the investigation of XB on small complexes. Ibrahim has modeled a σ-hole explicitly by modifying a force ﬁeld through addition of an extra point of positive charge that represents the σ-hole on a halogen atom116 (historically, the addition of an extra point charge has been used to model the lone pair on a Lewis base117). This positive extra point (PEP; a massless particle) approach, coupled to the general Amber force ﬁeld (GAFF),118 was then used to study the XB interaction between a series of halogenated organic molecules and several Lewis bases. The comparison between the interactions energies obtained by MM and QM calculations at the DFT and MP2 calculations for each complex showed good agreement. The accuracy of the method was also validated by the assessment of the free energies of solvation of 7127

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

halobenzene molecules relative to benzene (estimated by thermodynamic integration).116 Subsequently, the helpfulness of this approach to the study of XB in drug screening and molecular biology was elucidated by assessment of binding interaction energies within a protein−ligand complex via molecular dynamics (MD) simulations using the ﬀ99SB force ﬁeld.119 Extending this work using MM, Ibrahim reported XB to be electrostatic and van der Waals in nature, and the strength of the interaction depended upon the relative sizes of the attractive interactions (between the Lewis base and the positive σ-hole), repulsive interactions (between the Lewis base and the negative charge on the halogen atom), and attractive/repulsive van der Waals interaction.120 An equivalent approach using a pseudoatom instead of an extra point of charge was employed with GAFF to perform MD simulations on complexes of proteins with halogenated ligands that reproduce experimental data.121 The results were also compared with crystallographic data and with hybrid QM−MM calculations. Moreover, the PEP approach, as devised by Ibrahim, has also been extensively employed by Felix and co-workers for the theoretical investigation of halide recognition by macrocycles122 (Figure 7) and interpenetrated (Figure 8) and interlocked structures (Figure 9)123 mediated by XB. These studies comprised QM calculations to determine the position of the explicit σ-hole (ESH) and the atomic charges (vide infra), followed by MD simulations using the GAFF force ﬁeld in diﬀerent solvent media, including pure water, organic solvents, and aqueous organic media. Detailed descriptions of the aniontemplated construction of these supramolecular architectures and their anion-binding properties will be discussed in section 6. The atomic charges used in the force ﬁeld calculations of XB are obtained through the restrained electrostatic potential (RESP) methodology.124 As required by this approach, the charges were calculated at the HF/6-31G* level of theory, a part of iodine which is not included in this basis set. Furthermore, the σ-hole was modeled by placing a positive charge at suitable distance along the C−X bond.124b The halogen atom point charge was replaced by two charges: the ﬁrst representing a σ-hole placed at a ﬁxed distance from the halogen atomic center and the second representing the halogen atom, such that the sum of the σ-hole charge and the halogen atom charge is equal to the usual value obtained by the RESP method. Another approach involved assigning the charge of the ESH and adjusting the partial charges within the rest of the molecule by employing the RESP methodology. A third approach116 required only the distance between the halogen atom and σ-hole to be determined before the calculation of the electrostatic potential of the whole molecule, including the ESH. Hobza and co-workers have incorporated the σ-hole formalism represented by a massless positive point charge into a docking program suite. This was shown to improve the reliability of protein−ligand geometries determined by the docking, especially when more than one XB is established between the ligand and the protein’s active site, which could be useful for future computer-aided drug design.125 Jorgensen and Schyman have implemented a partial positive charge modeling scheme of the σ-hole in the OPLS-AA and OPLS/CM1A force ﬁelds.126 The corresponding modiﬁed versions, OPLS-AAx and OPLS/CM1Ax, were tested with gasphase complexes of halobenzenes and Lewis bases, and showed

Figure 7. (a) DFT-optimized structures of the XB complex between the bromoimidazolium macrocycle and iodide showing the corresponding halogen-bonding interactions as yellow dashed lines together with their distances (Å) and angles (deg) given in italics and Wiberg indexes in parentheses. (b) Representative snapshots of bromide and iodide complexes in CH3OH/H2O (9:1) solution showing both associations surrounded by water molecules. Methanol molecules and the receptor hydrogen atoms were omitted for clarity. Halogen bonds are represented as yellow dashed lines. Reprinted from ref 122. Copyright 2012 American Chemical Society.

reasonable agreement with results from ab initio MP2/aug-ccpVDZ(-PP) calculations. The aforementioned modeling studies clearly show how to overcome a limitation of classical force ﬁelds when studying XB interactions. Indeed, classical force ﬁelds have been developed with the assignment of atomic point charges, which leads to the 7128

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 8. DFT-optimized structure of the XB pseudorotaxane assembled with a chloride template showing the linear C−Br···Cl− halogen bond and the N−H···Cl− hydrogen bonds, represented as black and yellow dashed lines, respectively. Reprinted with permission from ref 123a. Copyright 2010 Wiley-VCH Verlag GmbH & Co. KGaA.

treatment of Cl, Br, and I as “large ﬂuorines” with little diﬀerentiation between the halogen atoms and σ-holes, as denoted by Clark et al. (see Figure 10).127 These authors have attempted to surpass this limitation by calculating QM local properties, such as the molecular electrostatic potential, local ionization energy, electron aﬃnity, or polarizability to produce augmented molecular property ﬁelds for 3D-QSAR (threedimensional quantitative structure−activity relationship) applications. Another approach to overcome the limitations of classical force ﬁelds has recently been developed. Clark and co-workers have devised a neglect of diatomic diﬀerential overlap (NDDO) based polarizable force ﬁeld (the hpCADD force ﬁeld), which directly combines electrostatics (Coulombic and exchange) taken from the NDDO wave function with classical force ﬁeld potentials to obtain more accurate structures and relative energies. This innovative hpCADD force ﬁeld is implemented in the EMPIRE program,128 and may represent a suitable approach for the molecular modeling of large molecular systems relying on XB interactions. 2.6. Conclusions and Outlook

Figure 9. (a) Representative co-conformation of chloride binding by bromoimidazolium-containing catenane through two XB interactions in acetonitrile solution. Reprinted with permission from ref 123b. Copyright 2012 Wiley-VCH Verlag GmbH & Co. KGaA. (b) Representative co-conformation of iodopyridinium-containing catenane in chloroform solution. Reprinted with permission from ref 123c. Copyright 2013 Wiley-VCH Verlag GmbH & Co. KGaA. (c) Representative snapshot showing the iodide binding to iodotriazolecontaining rotaxane through two XB and NH interactions, surrounded by four water molecules. The cyclodextrin stoppers are depicted by a space-ﬁlling representation. In general, solvent molecules and PF6 counterions were omitted for clarity.

XB interactions have been intensively investigated both theoretically and computationally. As the size of the system to be studied or the basis set increases so too does the computational cost of such calculations. As a consequence, the majority of XB computational investigations have been carried out for XB complexes between simple XB donors, such as dihalogens and haloﬂuoromethanes, and simple Lewis bases, and only a handful of those reported study the eﬀect of multiple XB interactions. High-level computational calculations or those on larger systems remain rare. A number of key trends have been identiﬁed, which are supported by experimental evidence discussed in later sections of this review. Extensive studies on the σ-hole model of XB reveal the strength of interaction correlates with the positive potential of the σ-hole, which may be tuned through changing either the chemical environment of the molecule, in a manner similar to that of HB, or (more unusually) the atom itself. Thermodynamic and structural characteristics of XB are shown to vary in parallel with the size of the σ-hole, which is presented by some as irrefutable evidence that the interaction may be explained fully through electrostatic/polarization and dispersion eﬀects. Others have reported XB with signiﬁcant or

dominant charge-transfer contributions; such interactions are determined to follow trends in amount of charge transferred or stabilization energy. The exact nature of the XB interaction, however, remains to be elucidated. Describing XB interactions using molecular mechanics is particularly challenging since conventional force ﬁelds are largely unable to model the anisotropic charge distribution of a polarized halogen atom. Addition of a single-point positive charge is one approach that has enabled MD simulations to be performed on larger host−guest interactions and also modeled in the solution phase. Classical forces ﬁelds in combination with 7129

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

stronger gas-phase σ-hole interactions, such as those formed to anionic XB acceptors,54,81b are thermodynamically stable. Microwave spectroscopy has been extensively used to explore the fundamental nature of the halogen bond. Legon and coworkers performed systematic studies on a series of prereactive B···XY complexes, where B is a Lewis base and XY is a dihalogen: either homonuclear (F2, Cl2, Br2) or heteronuclear (ClF, BrCl).132 In accordance with ab initio calculations, the extent of electric charge redistribution revealed B···XY complexes are all Mulliken outer complexes133 (weak, with no signiﬁcant charger transfer, as opposed to inner complexes of the type [BX]+···Y−, which are stronger and with a signiﬁcant degree of charge transfer), except H3N···BrCl and H3N···ClF, which had a larger amount of electric charge redistribution. As shown in Figure 11a, the amount of electronic charge transferred was small (on the order of a few percent) and decreased as the ionization energy of the Lewis base increased. Figure 10. Isocontour plots of various halobenzenes using B3PW91/ cc-pVTZ-PP on the 0.001 au molecular surface (A) or OPLS-AA 2005 force ﬁeld derived charges (B). Red indicates negative electrostatic potential; blue refers to positive electrostatic potential. Reprinted from ref 127b. Copyright 2012 American Chemical Society.

PEP charges have limited utility, however, and the development of more accurate models is ongoing.

3. GAS-PHASE STUDIES OF HALOGEN-BONDING INTERACTIONS Theoretical studies invariably involve the isolated complex, whereas experimental work is often in condensed phases. Experimental investigations on isolated complexes either in inert gas matrixes at low temperature or in the gas phase at low pressure have been extensively used for direct comparison with theory. In contrast to experimental studies on hydrogen bonds, gasphase experimental data come from rotational, rather than vibrational, spectroscopy. Analysis of the rotational spectra of complexes under study allows rotational constants to be determined, which are related to moments of inertia and, therefore, the distribution of the mass of the complex in space. Hence, these data can be used to determine the separation of B and XY (radial geometry) and their relative orientation in space (angular geometry). The stretching force constant kσ, obtained from the centrifugal distortion constant, gives information on the strength of the interaction. Changes to the halogen nuclear quadrupolar coupling constants χαβ(X) and χαβ(Y) give information on the changes to the electric ﬁeld gradient at X and Y and, hence, in the electric charge distribution of XY. Strong complexes may be observed in equilibrium gas mixtures of two components at low temperature using microwave spectroscopy, but for more weakly bound complexes, supersonic jets or beams of gas must be employed, for example, pulsed-jet, Fourier-transform microwave spectroscopy (full details of the experimental techniques employed are provided by Legon15,129). Gas-phase complex formation is often accompanied by ΔG > 0 despite ΔH being negative,54,130 which is due to the relatively large negative values of TΔS for weakly and moderately bound complexes. Interaction of two molecules in a complex reduces their translational and rotational degrees of freedom,56,131 and the total entropy of the system is reduced. Consequently, only

Figure 11. (a, top) Fraction δi of an electronic charge transferred from B to XY on formation of B···XY plotted against the ionization energy IB of B for the series XY = Cl2, BrCl, and ICl. The solid curves are the functions δi = A exp(−aIB) that best ﬁt the points for each series B··· XY. Data for B···Br2 are nearly coincident with those of B···BrCl and have been excluded for the sake of clarity. Reprinted with permission from ref 134. Copyright 2008 Springer-Verlag. (b, bottom) Fraction δp of an electronic charge transferred from X to Y on formation of complexes B···XY plotted against the intermolecular stretching force constant kσ (a measure of the strength of the halogen bond) for the series XY = Cl2. The solid line represents the least-squares ﬁt of the points. Reprinted with permission from ref 129. Copyright 2010 The Royal Society of Chemistry. 7130

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Furthermore, for a given B, the amount of electric charge transferred from X to Y paralleled the axial dipole polarizabilities of the halogen/inter-halogen molecule,135 which was also reﬂected in the strength of the interaction as measured by the rotational force constant (Figure 11b), both of which were interpreted as evidence of a weak, electrostatic type of interaction involving the permanent, unperturbed electric charge distributions of the two components. These observations are similar to those observed previously for a large number of analogous hydrogen-bonded complexes B···HX,136 which are consistent with a simple electrostatic model of interaction. Angular geometries of a range of B···XY complexes were also determined and found to strongly parallel those of the analogous B···HX hydrogen-bonded complexes, and with a greater tendency to linearity of the B−X−Y bond angle. Legon et al. concluded that the Legon−Millen rules137 used to rationalize angular geometries of traditional B···HX complexes (which are implicitly electrostatic in origin and involve identifying the directions of greatest nucleophilicity in the electron donor B) also apply, after suitable modiﬁcation, to geometries of B···XY complexes. The only exceptions occurred for Lewis bases furan and thiophene, in which there is both an aromatic π-system and a nonbonded electron pair: these were rationalized, however, when the electric charge distributions of the heteroaromatic molecules were considered.15,134 As such, the authors suggested the angular geometry of halogen-bonded complexes was determined largely by electrostatics irrespective of whether the interaction itself is electrostatic in origin. Radial geometries of B···XY and B···HX are also systematically related and show that the van der Waals radius of atom X in the dihalogen XY is shorter along the XY axis than perpendicular to it. More substantial degrees of charge redistribution occur when the Lewis base is trimethylamine. Studies on the rotational spectra of strongly bound Me3N···ClF138 and Me3N···F2139 showed a signiﬁcant extension of the XY bond length and suggest a substantial charge transfer from the halogen molecule to the amine occurs with this complexa result which parallels the large degree of proton transfer observed for Me3N···HX.140 Legon and co-workers have established a directly proportional relationship between Dσ (the intermolecular dissociation energy) and kσ (the intermolecular quadratic stretching force constant) for a large number of halogen-bonded complexes B··· XY and their hydrogen-bonded analogues B···HX (Figure 12). Values of Dσ were determined using ab initio calculations (at the explicitly correlated level of theory CCSD(T)(F12*) with cc-pVDZ-F12-optimized basis functions), and the iσ values were those established from rotational spectra of B···XY and B··· HX.141 Legon and co-workers have recently extended their investigations into gas-phase halogen-bonding interactions from dihalogen molecules to simple halogen-containing molecules RX such as CF3I with a variety of Lewis bases. The data measured for interactions with amines NH3 and NMe3 are consistent with a C3v symmetric top geometry (conﬁrming the presence of a linear N···I−C halogen bond) and ab initio calculations for the same system.142,143 Moreover, the authors consider the molecule ICF3 to provide a bridge between solid-state investigations involving I−C6H4−I or I− (CF2)n−I and B···ICl gas-phase complexes, and consequently compared their results with a survey of crystallographic data for interactions between halogen-functionalized hydrocarbons and

Figure 12. (a, top) Dσ against kσ for complexes of the types B···X2 and B···XY, where B is one of the Lewis bases N2, CO, C2H2, C2H4, H2S, H2O, PH3, and NH3, X2 is one of the nonpolar dihalogen molecules F2, Cl2, and Br2, and XY is one of the polar dihalogens ClF, BrCl, and ICl. The continuous line is the straight line ﬁtted to the points arising from B···X2 complexes only. (b, bottom) Dσ against kσ for hydrogenbonded complexes of the type B···HX, where B is one of the Lewis bases N2, CO, C2H2, C2H4, H2S, H2O, PH3, and NH3 and X is F, Cl, Br, and I. The continuous line is the straight line ﬁtted to the points (including the origin) by linear regression. Reprinted with permission from ref 141. Copyright 2014 The Royal Society of Chemistry.

O-, N-, or S-containing molecules.45g The gas-phase complex CF3I···NMe3 was found to represent accurately the solid-state (CF2)nI···NH2(CH2)n interactions. Complexes B···ICF3142,144 and B···ICl, where B = N2, OC, C2H2, C2H4, H2O, H2S, PH3, or NH3, were compared,129,131,145 which showed that the halogen bond from ICF3 is systematically longer and weaker than that from ICl, while the angular geometry of B··· is isomorphic in both systems. Experimental studies in the gas phase, in the absence of solvent or other competing interactions, provide a direct link to theoretical studies on the existence and nature of the halogen bond. Once established, there exist a diverse range of supramolecular applications to make use of halogen-boding interactions, and these are the focus of the remainder of this review.

4. HALOGEN BONDING IN THE SOLID STATE 4.1. Introduction to Crystal Engineering and Functional Materials

Halogen bonding has been investigated in the solid state for many years and has become a popular and exploited supramolecular synthon in the well-established ﬁeld of crystal 7131

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

ing a monoiodoacetylene XB donor with ﬂuorination in the para-position, 4-ﬂuoro-1-(2′-iodoethynyl)benzene, resulted in the shortest N···I XB to date of 2.622 Å (Figure 13). The

engineering. While historically the study of XB in the solid state has centered around organic molecules,146 studies involving metal-containing species147 and anions148 have recently come to the fore. At the heart of crystal engineering is the challenge to understand the ﬁne balance between strong and weak intermolecular interactions as they work in concert to determine the solid-state assembly. Desiraju introduced the concept of the supramolecular synthon149 in 1995, and it remains the cornerstone of crystal engineering: useful complementary intermolecular interactions can be identiﬁed and used as synthetic tools for the design and fabrication of supramolecular architectures.150 Furthermore, XB is gaining ground in the construction of complex structural frameworks and functional materials. Several reviews have been written which broadly highlight the use of XB in the context of supramolecular chemistry,16d,45b,146,151 while others are more speciﬁcally focused on metal-containing species,147a anion coordination and aniontemplated assembly,148 and alternative motifs for XB.152 Recently, Crystal Growth & Design and CrystEngComm published special issues with an emphasis on crystal engineering and self-assembly, and materials and biomolecules, respectively. Two reviews have been written about XB in crystal engineering: Aakeroy7 discussed XB brieﬂy in the broader context of crystal engineering, and Desiraju153 considered the analogy of XB with HB and the orthogonal use of XB in the design and manufacture of desirable architectures. In this review, this section of XB in the solid state aims to highlight the recent use of XB in crystal engineering and functional materials with a focus on the applications of XB toward these goals.

Figure 13. Eﬀect of ﬂuorination on N···I distances: (a) XB donors cocrystallized with DMAP and (b) cocrystal structure of 4-ﬂuoro-1(2′-iodoethynyl)benzene and DMAP showing short N···I contacts, the shortest being 2.622 Å (CSD Refcode NOGYUF).157

results showed that ﬂuorination caused a slight shortening of the N···I XB distance; however, the hybridization of the ipsocarbon appeared to be a much more signiﬁcant factor. This trend, whereby the sp-hybridization of the substituent modiﬁes the strength of the halogen bond, is generally accepted to follow the order C(sp)−X > C(sp2)−X > C(sp3)−X. X-ray diﬀraction studies have shown that a halogen bond is characterized by D···X−Y distances that are signiﬁcantly shorter than the sum of the van der Waals radii. Closer contacts are generally indicative of stronger halogen bonds. In addition, a concomitant lengthening of the X−Y bond was also observed, as electron density was donated into the σ-hole. Halogen−halogen contacts tend to be classiﬁed as either type I or type II interactions in the solid state (Figure 14). Type I

4.2. Fundamentals

XB is comparable in strength to the more well-known intermolecular interaction HB.16d Furthermore, it imposes similarly stringent constraints on the geometry and direction of the interaction by way of accepting electron density through the σ-hole. The prominence of the σ-hole is related to the polarizability of the halogen atom and dictates the strength of the halogen-bond donor in the order I > Br > Cl ≫ F. Finetuning of the XB interaction is possible by varying the nature of the atom or moiety covalently bound to the given halogen atom: increasing the electron-withdrawing ability of the substituent leads to increased halogen-bond donor strength. The strength of the XB has also been correlated to the degree of charge transfer between the XB donor and acceptor, illustrating the importance of electrostatics in the formation of charge-assisted XB interactions.154 The corollary of this is that the degree of charge transfer can be altered by the choice of acceptor: more basic acceptors generally increase the degree of charge transfer. Moreover, this in turn can aﬀect the electronic properties of the individual molecules involved in XB formation, leading to neutral−ionic conversions in the solid state.154a When compared to I2, employing an organic spacer aﬀords tunability of the XB donor strength and directionality, in addition to the possibility of forming extended interactions through variation in the number of XB donors, which are not possible with the more easily polarized I2 molecule,155 which can display amphoteric behavior.156 Notable examples of this include the use of perﬂuorinated substituents or the incorporation of a positive charge on an attached heteroaromatic substituent. Recently, the eﬀect of ﬂuorination on the N···I distances of organic iodine XB donors with 4-(N,Ndimethylamino)pyridine (DMAP) was investigated.157 Employ-

Figure 14. Schematic of commonly observed halogen-bond geometries, where X = halogen, Y = C, N, O, halogen, and D = Lewis bases that are neutral (N, O, S, Se, ...) or anionic (Cl−, Br−, I−, ...).158a

contacts are generally considered to be dispersive and are associated with crystallographic inversion centers, and are not halogen bonds, while type II contacts are accepted to be attractive and associated with screw axes and glide planes, and are considered to be true halogen bonds.66a,158 XBs with Lewis bases tend to adopt rather strict geometries that deviate only slightly from linearity with respect to D···X− Y.16d In such cases, the interaction arises along the extension of 7132

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 15. Type II F···F and C−F···S interactions shown in (a) the crystal structure of (pentaﬂuorophenyl)-2,2′-bisthiazole (CSD Refcode KETVUC01), (b) the Δρ(r) map with the isosurfaces drawn at ±0.05 e Å−3, and (c) the ∇2ρ(r) (e Å−5) map drawn on the logarithmic scale in the F···F and S···F interaction regions, +ve and −ve values represented by blue and red colors. Reprinted with permission from ref 160. Copyright 2013 The Royal Chemical Society.

The assertion for type I interactions only being considered as dispersive, geometric interactions is supported by the work of Aakeröy and co-workers,161 who investigated a series of Cu(II)−acac complexes comprised of bifunctional β-diketonate ligands functionalized with methyl, chloro, bromo, or iodo substituents. Apart from the methyl-substituted ligand, the three halo-functionalized ligands all displayed short halogen bonds to the exo lone pair of their respective keto oxygen atoms, extending the structure into 1D chains. This preference is justiﬁed since the XB donor will preferentially interact with the most powerful XB acceptor based on simple electrostatic considerations. However, when complexed to Cu(II), all possible XB acceptors were depleted of charge due to metal chelation and the methyl, chloro, and bromo structures were forced to adopt head-to-head close-packed structures exhibiting type I contacts in lieu of suitable competing interactions (Figure 16). The iodo structure, however, cannot accommodate this packing arrangement due to the more prominent σ-hole and instead adopts an inﬁnite zigzag pattern with no short I···I contacts, a result attributed solely to close-packing. Regarding the halogen−Lewis base interaction, typically, stronger halogen bonds are observed for nitrogen-containing compounds such as amine and pyridine derivatives in comparison with oxygen- and sulfur-containing compounds such as ethers, alcohols, and thioethers. Interestingly, N-oxides tend to be preferred, however, over pyridine compounds. Furthermore, anionic species tend to be better halogen-bond acceptors than neutral molecules, aﬀording XB useful applications in anion coordination chemistry.16d Similarly to HBs, XBs can form bifurcated162 assemblies in the solid state, leading to 1D, 2D, and 3D architectures.16d This property can be useful in the construction of functional materials. To demonstrate the reliability and predictability of XB for the assembly of supramolecular architectures, Desiraju

the X−Y bond and along the axis of the lone pair of electrons donated by the Lewis base. Indeed, this strict geometric requirement has allowed for the formation of molecular polygons: very speciﬁc geometric structures in the solid state.159 The type II interaction has been described as the true form of XB as it arises from an attractive nucleophile−electrophile pairing of the electropositive σ-hole of one polarized halogen with the electronegative equatorial belt of another polarized halogen, whereas the type I interaction is geometrically based since it arises from close packing and is observed for all halogens.153 Indeed, ﬂuorine has been shown to participate in type II interactions despite the more prevalent, close-packed type I interaction, and in one case the type II interaction has been studied in depth using experimental charge density analysis.160 Analysis of the crystal structure of pentaﬂuorophenyl-2,2′-bisthiazole revealed short type II F···F interactions, stabilized by a surrogate C−F···S interaction (Figure 15). A supplementary search of the CSD indicated that F···F interactions are ostensibly accompanied by auxiliary interactions such as C−H···F, X−H···F, or C···F, and the independent existence of F···F interactions is rare. Deformation density and Laplacian maps of pentaﬂuorophenyl-2,2′-bisthiazole clearly displayed a region of charge depletion facing a region of charge concentration on both ﬂuorine atoms involved in the XB. Additionally, an XB to sulfur was distinctly observable from the geometrical alignment of the σ-hole of F toward the valence shell charge concentration (VSCC) of S, in spite of no observable polar ﬂattening and only slight polarization of F. Finally, it should also be noted that, in unsymmetrical type II interactions, where the two halogens are diﬀerent, the XB donor is always the more polarizable of the two halogens, reﬂecting the XB donor ability of the halogen. 7133

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

chlorophenol. These results demonstrate that XB synthons are robust and can be integrated into molecular precursors to yield desirable solid-state constructions. However, to enable the supramolecular chemist to use XB as a tool, one needs to know how it works in relation to other well-known supramolecular synthons. 4.3. Halogen-Bonding Hierarchy

To further the understanding of XB in the context of supramolecular chemistry and crystal engineering, recent studies have sought to build on the general hierarchical principles, outlined above, and to present an accurate ranking of competing solid-state interactions.166 This is essential as XB can be used in concert with HB to engineer complex assemblies with precise stoichiometry and connectivity. XBs can be ranked either relative to one another or against well-established synthons such as HB interactions, donor−acceptor interactions, and van der Waals interactions to institute a hierarchy or some general rules toward synthon preference. 4.3.1. Ranking Halogen-Bond Donors. Aakeröy and coworkers166a examined the relative XB donor strength of pairs of iodo- and bromo-substituted aromatic cocrystal formers with a library of suitable XB acceptor molecules. The molecules chosen for this study included 1-(iodoethynyl)-4-iodobenzene (IEIB) and 1-(bromoethynyl)-4-iodobenzene (BEIB) (Figure 18) as each compound contains two distinctly diﬀerent XB

Figure 16. Crystal structures of Cu(II)−acac complexes with XB ligands showing head-to-head arrangements of the complexes with (a) methyl (CSD Refcode BALKOQ), (b) chloro (CSD Refcode BALKUW), (c) bromo (CSD Refcode BALLAD), and (d) iodo (CSD Refcode BALLEH) substituents.161

and Mukherjee163 recently investigated XB in terms of modular design principles using trichlorophenols. The concept of modularity is the construction of a system from several parts that each have well-deﬁned and distinct roles and can easily be replaced. When applied to crystal engineering, this pertains to the transferability of synthons between crystal structures, the parts being the molecular entities, and the functional groups capable of forming synthons that they contain. In their study, the authors discerned interactions in 3,4,5-trichlorophenol which can be correlated to synthons observed in 4chlorophenol and 3,5-dichlorophenol (Figure 17). It was speculated that each of these synthons existed in solution as a precursor to crystallization and they come together via the Cl···Cl halogen bond in the stages before nucleation. Arguably, the structure of 3,4,5-trichlorophenol can be regarded as the sum of the structures of the two simpler parts. An analogous observation was made when 2,3,4-trichlorophenol was compared with 2,3-dichlorophenol. The authors observed moreover that structural modularity could be induced, especially in cocrystals of these compounds. In particular, the 1:3 cocrystal between 4-chlorophenol and 3,4,5-trichlorophenol displayed synthon equivalence with the α-form of 4-

Figure 18. Molecular structures of XB donor molecules studied for ranking by XB donor ability: Csp-bonded (a) (iodoethynyl)iodobenzene and (b) (bromoethynyl)iodobenzene, activated Csp2bonded (c) 1,4-diiodotetraﬂuorobenzene and (d) 1,4-dibromotetraﬂuorobenzene, and nonactivated Csp2-bonded (e) 1,4-diiodobenzene and (f) 1,4-dibromobenzene.166a

donor functionalities, and while the iodo/bromoethynyl moiety is a known XB donor, systematic studies intended to rank it in comparison to other XB donors had not yet been undertaken. Four well-known XB donors were used for comparison.

Figure 17. Structural modularity in trichlorophenols: the HB tetramer in 4-chlorophenol (CSD Refcode CLPHOL13)164 and the XB- and HBstabilized ladder synthon in 3,5-dichlorophenol (CSD Refcode DCLPHE)165 are both present in the crystal structure of 2,3,4-trichlorophenol with additional Cl···Cl XB interactions (CSD Refcode UZOWEM01).163 7134

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

known in HB systems, was shown to play a signiﬁcant role in the connectivity of XBs in the solid state. In another study by Aakeröy et al.,166b a series of symmetric ditopic (iodo and bromo) XB donors were assessed by cocrystallization with (i) monotopic, (ii) symmetric ditopic, and (iii) dissymmetric ditopic XB acceptors (Figure 20). However, in all cases, disorder was observed in the positions of the iodo and bromo XB donors when interacting with the various XB acceptors. Notably, lowering the symmetry of the XB acceptors by using dissymmetric ditopic XB acceptors was unsuccessful in engineering structures displaying halogen-bond speciﬁcity adequate to establish a hierarchy of the XB donors. It was only when a dissymmetric ditopic XB donor was used in conjunction with dissymmetric ditopic XB acceptors that engineering a structure without positional disorder was successful. However, this only became possible after the introduction of an amide group between the iodo and bromo XB donors to induce order through known amide···amide selfcomplementary hydrogen bonds. While this should have been suﬃcient to order the structure without competing for XB acceptor sites, in one case, N−H···acceptor interactions were able to prevent Br···acceptor interactions from forming and leaving the I···acceptor interactions intact (Figure 20b). Finally, in the case where a symmetric ditopic XB acceptor was used, the desired I···acceptor and amide···amide interactions were obtained with Br···Br type I close contacts present. Consequently, it is clear from this study that XB interactions can eﬀectively compete with HB interactions and overcome them when iodine is used as the XB donor. Bromine is a weaker XB donor but can nevertheless be used to introduce useful structure-directing interactions. The above study highlights the diﬃculty of establishing a reliable ranking of XB donors due to substitutional disorder in the positions of dissymmetric XB donor atoms. Similar results were obtained by Katrusiak and co-workers in their studies on in situ pressure-frozen crystals of 1,2-dihalotetraﬂuoroethanes167 and cocrystals of 1,2-dihalotetraﬂuoroethanes with 1,4-dioxane.168 The investigation into isoelectronic and isostructural 1,2-dihalotetraﬂuoroethanes revealed that X···F (X = Br/I) interactions are the dominant cohesive force, while X···X (X = Br/I) interactions are weaker and play a subordinate

With the premise that electrostatics dominate most conventional halogen bonds, it was surmised that a good XB donor would be one that (i) exhibited a more pronounced σ-hole and (ii) formed cocrystals more eﬃciently with suitable XB acceptors. To that end, molecular electrostatic potential surfaces were calculated (using DFT at the B3LYP/6311+G** level) for the six molecules (Figure 19), and they

Figure 19. From left to right and from top to bottom, XB donor molecules under study and ranked in order of decreasing surface electrostatic potential associated with the σ-hole of the best XB donor atom in the molecule. The electrostatic potential surfaces were calculated at 0.002 isovalue. Potential values are in kilojoules per mol. Reprinted with permission from ref 166a. Copyright 2013 Wiley-VCH Verlag GmbH & Co. KGaA.

were allowed to react with 21 XB acceptor molecules via the solvent-drop grinding methodology. The calculations, Fourier Transform infrared spectroscopy, and single-crystal X-ray diﬀraction analysis all indicated that IEIB and 1,4-diiodotetraﬂuorobenzene were the strongest XB donors of almost equal strength, while 1,4-diiodobenzene and 1,4-dibromobenzene were unsurprisingly the weakest XB donors. It was also evident that perﬂuorination of an XB donor can sensitize the halogen bond toward XB acceptors which contain sterically encumbered sites in geometric proximity to the acceptor site. Additionally, diﬀerences in XB donor−acceptor selectivity were noticed when IEIB and BEIB were compared with certain XB acceptors. Moreover, the concept of “best donor−best acceptor”, well-

Figure 20. Crystal structures of dissymmetric compounds in which disorder was overcome in the structures: (a) 2D sheet formed from the combination of N···I and N···Br XB and CO···H−N HB interactions between the dissymmetric iodo/bromo XB donor and dipyridylethene XB acceptor (CSD Refcode DIMTIE), (b) zigzag network formed by the dissymmetric iodo/bromo XB donor and dipyridylethane via N···I XB and N− H···N HB interactions (CSD Refcode DIMTEA), and (c) 2D sheet formed between the dissymmetric iodo/bromo XB donor and 4,4′-bipyridine via N···I and Br···Br XB and CO···H−N HB interactions (CSD Refcode DIMTAW).166b 7135

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 21. Structures of cocrystals showing the charge-assisted HB interactions driving the formation of the layers with auxiliary XBs: (a) 2D layer formed in (4-chlorobenzyl)ammonium 2,4-dichlorobenzoate (CSD Refcode WIVPUM) and (b) 1D ladder formed in (3-chlorobenzyl)ammonium 2,4-dichlorobenzoate (CSD Refcode WIVQAT).171

Figure 22. Comparison of the crystal structures of (a) 1,4-diiodotetraﬂuorobenzene (CSD Refcode MEKWOO) and (b) hydroquinone (CSD Refcode MEKWUU) with dipyridylethane.172

contained hydrogen-bonded inﬁnite one-dimensional stacks of alternating, parallel cations and anions which cross-link via hydrogen bonds to form inﬁnite two-dimensional sheets (Figure 21a). Relatively short Cl···Cl distances were observed between the 4-chloro substituents of a cation and an anion in adjacent layers. In contrast, the meta-substituted structure contained an inﬁnite one-dimensional hydrogen-bonded ladder with short Cl···Cl contacts arising only between cationic units in adjacent ribbons (Figure 21b). The authors determined that the driving force for the crystallization of these compounds was governed primarily by the strong, charge-assisted hydrogenbond interactions, while secondary noncovalent interactions such as halogen bonding played a subordinate role. In a similar study, Metrangolo and Resnati172 showed the ability of XB to dominate over HB in driving the self-assembly processes of several cocrystals under appropriate conditions. Dipyridylethane was shown to cocrystallize with 1,4-diiodotetraﬂuorobenzene and hydroquinone to aﬀord XB and HB cocrystals, respectively (Figure 22). However, when presented with both 1,4-diiodotetraﬂuorobenzene and hydroquinone in solution, dipyridylethane selectively cocrystallized with 1,4diiodotetraﬂuorobenzene, which indicated the dominant role of the ﬂuoro-activated iodo XB donor over the hydroquinone HB donor. Similar results were obtained when tetramethylethylenediamine (TMEDA) was cocrystallized with diiodotetraﬂuoroethane and ethylene glycol. In these cases, HB was shown to be subordinate to XB. To test the ability of XB to compete with HB, an experiment was designed in which a pyridine-substituted benzimidazole was cocrystallized with a ﬂuoro-activated XB donor molecule (X = F, Br, I) containing an additional two potential HB donors in the form of an oxime and an imine C−H donor (Figure

role in the structures. Interestingly, cocrystals may be obtained for the diiodo compound when cocrystallized with 1,4-dioxane but not with the dibromo compound, revealing that I···O interactions are stronger than Br···O interactions. However, substitutional disorder is still observed in the positions of Br and I in the dissymmetric XB donor molecule. More recently, Metrangolo, Resnati, and co-workers169 cocrystallized 1,2-dihalotetraﬂuoroethanes with HMPA using in situ cryocrystallization techniques to investigate the C−X···O (X = Cl, Br, I) synthons further. In the isostructural series of cocrystals, both the XB donor and acceptor tectons behave as bidentate modules with HMPA accepting two XBs. Disorder is also observed in the dissymmetric XB donor molecules, which is attributed to the comparable surface electrostatic potential and steric hindrance of the halogen atoms. While the strength of the synthons was ranked I > Br > Cl, the results showed that even the weaker C−Cl···O synthon is capable of aﬀecting the crystal structure of the HMPA adduct similarly to its bromo and iodo analogues. 4.3.2. HB/XB Complementarity/Competition. While XB can often dominate the crystal structures engineered using weaker noncovalent interactions, it is generally of the same order of strength as HB and can either compete or cooperate with HB in crystal structures.170 To the best of our knowledge, the seminal example of an investigation into the interplay of XB with HB in solid-state organic salts was by Aakeröy et al. in 1999;171 chloro-substituted benzylammonium benzoate cocrystals were structurally characterized. The eﬀect of chloro substitution in the para- and meta-positions was investigated by comparing the structures of (4-chlorobenzyl)ammonium 2,4-dichlorobenzoate and (3-chlorobenzyl)ammonium 2,4dichlorobenzoate. In the ﬁrst case, the crystal structure 7136

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

23a).173 In this way, two acceptor sites were available for three potential donors. Molecular electrostatic potential calculations

not strong enough to compete with the C−H···pyridine HB (Figure 23c). However, the iodo-substituted cocrystal former exhibited the same oxime···imidazole contact but supplemented with a pyr···I contact displacing the imine C−H. Short type II I···I contacts organize these dimers into 2D sheets (Figure 23d). Consequently, a good HB donor (oxime) is likely to be very competitive for a N-heterocyclic moiety even in the presence of a ﬂuoro-activated organoiodine which was able to displace a less conventional HB donor. XB has also been shown to work in concert with HB in the design of inﬁnite chains using 1,4-diiodotetraﬂuorobenzene and bromo-substituted 2-aminopyrazines.174 Indeed, in these structures, HBs and XBs behave orthogonally with the formation of self-complementary amino N−H···Npyrazine HBs and I···N4pyrazine XBs (Figure 24). Despite the inclusion of

Figure 24. Crystal structure of (dibromoamino)pyrazine with 1,4diiodotetraﬂuorobenzene arranged into inﬁnite 1D chains via selfcomplementary hydrogen bonds and N···I halogen bonds (CSD Refcode PAMLUM).174

electron-withdrawing bromo substituents on pyrazine, the HB and XB synthons displayed high ﬁdelity as they are present in all structures. Nonetheless, it was shown that modulation of the electrostatic charge on the XB acceptor through simple covalent modiﬁcations altered the nature of the speciﬁc halogen bond directly responsible for the cocrystal assembly in each case. Moreover, geometric complementarity minimized synthon crossover as HBs favored two-point contacts, particularly in self-complementary dimers, while most XBs preferred singlepoint contacts. This geometric bias is certainly useful in the design of supramolecular architectures.175 However, there are cases where two-point HBs and XBs have been shown to work cooperatively, such as the study of Nangia and co-workers176 on the construction of molecular tapes using a bifurcated iodo··· nitro XB in tandem with a COOH···pyr HB interaction. The X···nitro (X = Cl, Br) synthon has also been observed in halosubstituted nitroimidazole derivatives.177 It is also interesting to note the geometric eﬀects of varying azobipyridine acceptor sites on the formation of XBs with bifunctional donor molecules containing an activated XB donor and an HB donor.178 In these cocrystal structures, the authors showed that HB was the primary driving force when 3,3′azobipyridine was the cocrystal former, resulting in trimeric supermolecules (Figure 25). In contrast, XB and HB were shown to contribute equally to the assembly of inﬁnite chains when 4,4′-azobipyridine was the cocrystal former. Finally, in terms of avoiding synthon crossover, the HB donors −COOH, −OH, and −CN(R)OH were indistinguishable in the presence of XB donors, which indicated that XB cannot compete with these synthons in the solid state. Recently, Bruce and co-workers have demonstrated the complementarity and ability of HB and XB to interact orthogonally in the preparation of several cocrystals with various heterocyclic XB and HB acceptor molecules.179 Their work supports the assertion that the best donor, be it HB or XB, will interact preferentially with the best acceptor; cocrystals of iodotetraﬂuorobenzoic acid with dithiane exhibited the

Figure 23. (a) Molecular structures of the cocrystal formers investigated, (b) crystal structure of the perﬂuorophenyl oxime with pyridylbenzimidazole (CSD Refcode VITKEP) and the 1D chain formed via hydrogen bonds, (c) crystal structure of the bromoperﬂuorophenyl oxime with pyridylbenzimidazole (CSD Refcode VITKIT) and the 1D chain formed via hydrogen bonds, and (d) crystal structure of the iodoperﬂuorophenyl oxime with pyridylbenzimidazole (CSD Refcode VITKOZ) and the discrete tetrameric supramolecule formed via hydrogen bonds and the only instance of halogen bonds in this series.173

predicted that the imidazole moiety would be a better acceptor than the pyridine and, similarly, the oxime would be a better HB donor than the imine C−H. Keeping with the general best donor−best acceptor rule, experimental investigation revealed that the ﬂuorine-substituted (Figure 23b) and brominesubstituted (Figure 23c) cocrystals displayed similar structures in which oxime···imidazole contacts were formed and supplemented by C−H···pyridine HBs, indicating that Br was 7137

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

which cemented the assertion that XB follows the best donor− best acceptor rule analogously with HB. 4.4. Control of Solid-State Supramolecular Architectures

The ability to control solid-state supramolecular architectures through informed design principles is the heart of crystal engineering. However, controlling certain properties of crystals such as polymorphism, stoichiometry, and molecular tautomerism can be challenging. In this section, notable studies addressing these diﬃculties as well as those concerned with metal-bound XBs and anions are discussed. 4.4.1. Polymorphism. Polymorphism is the ability of a given molecular solid to exist in more than one crystalline state and is a contentious area of research, particularly in the production of pharmaceuticals. A curiosity of polymorphs is that, in some cases, after nucleation of a more stable crystal form, previously prepared less stable forms are no longer accessible without the use of seed crystals; this phenomenon is known as a “disappearing polymorph”.183 Needless to say, the possibility of exerting control over the packing or conformational polymorphism via the introduction of strong structuredirecting interactions such as XB is an appealing prospect, especially in the design of functional materials.184 Gonnade and co-workers185 showed for the ﬁrst time that XB can lead to diﬀerent polymorphic modiﬁcations. They examined the halogen bonds formed in the concomitant polymorphic structures of tris-O-(p-halobenzoyl)-myo-inositol 1,3,5-orthoformates and showed how C−Br···Oether−C contacts in the kinetic structure converted to more electrostatically favorable C−Br···OC contacts in the thermodynamic structure upon heating without loss of single crystallinity (Figure 27). These molecular movements appeared to sacriﬁce the supplementary Br···Br contacts as they rearranged themselves into the thermodynamically stable crystal structure.

Figure 25. Cocrystals of 4-iodotetraﬂuorobenzoic acid with (a) 3,3′azobipyridine (CSD Refcode PETQEM) and (b) 4,4′-azobipyridine (CSD Refcode PETPAH).178

formation of strong self-complementary carboxylic acid dimers accompanied by short I···S XBs (Figure 26a). Additionally,

Figure 26. Crystal structures showing (a) I···S halogen bonds from iodotetraﬂuorobenzoic acid to dithiane (CSD Refcode HIZQOY)179b and (b) stabilizing I···O halogen bonds in the formation of a Meissenheimer-like delocalized ring structure, which forms upon deprotonation of the phenol (CSD Refcode BIYFIA).179a

their studies using phenolate cocrystals have shown that, in the absence of suﬃcient HB donors, XB was able to satisfy the coordination of the phenolate oxygen atom and aided the stabilization of a CO double bond and a Meissenheimer-like delocalized ring structure, which formed upon deprotonation of the phenol (Figure 26b). Similar iodo···chalcogen (S, Se) interactions have been observed by Pennington and coworkers180 in their studies on XBs with thioamides and triphenylphosphine selenides. 4.3.3. Predicting XBs. The above studies aﬀord the crystal engineer valuable insight into the interplay between HB and XB in the solid state. However, the ability to predict crystal structures is still a matter of debate, as it has been for a long time,181 in spite of our knowledge of supramolecular synthons. Nonetheless, Aakeröy and co-workers oﬀered a solution to determine whether supramolecular synthons will either compete with or complement one another through the calculation of molecular electrostatic potentials of molecules.182 Since electrostatics contribute greatly toward XB, much like HB, MEP calculations oﬀer a good starting point for the researcher to determine which donors will prefer which acceptors on the basis of a simple electrostatic potentialbased ranking. This was experimentally conﬁrmed by the cocrystallization of various XB donors with suitable acceptors,

Figure 27. Polymorphic structures of the 1,3,5-orthoformate showing diﬀering C−Br···O contacts: (a) form I stabilized by C−Br···O−C interactions and additional Br···Br interactions (CSD Refcode VEBBUA) and (b) form II stabilized by C−Br···OC interactions (CSD Refcode VEBBUA01).185

Similar results were obtained using the racemic disubstituted analogues of the myo-inositol 1,3,5-orthoformate, which crystallized concomitantly in two forms, one of which was repeatedly obtained in greater yield than the other (Figure 28).186 In both forms, O−H···O HBs linked the molecules into chains. However, the chains were linked diﬀerently: in the ﬁrst, lower yielding form, this was via short C−H···X HB contacts, while the second, higher yielding form exhibited C−X···OC (X = Cl, Br) XB contacts. The ﬁrst form also showed a thermal 7138

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 28. Crystal structures of the two polymorphic forms of myoinositol 1,3,5-orthoformate: (a) form I stabilized by C−Br···OC interactions (CSD Refcode MIXNUD) and (b) form II stabilized by Br···Br halogen-bonding interactions (CSD Refcode MIXNUD01).186

Figure 29. Crystal packing diagrams of polymorphs of 1-phenyl-2methyl-4-nitro-5-bromoimidazole at diﬀerent temperatures: (a) 250 K (CSD Refcode AWAKIS01), (b) 120 K (CSD Refcode AWAKIS04), and (c) 100 K (CSD Refcode AWAKIS05). In each of these cases, interactions such as Br···N halogen bonds facilitate the reversibility of the structural transitions.187

crystal-to-crystal transition to the second form. These results suggest that XB can play a preferential role in nucleation and growth of crystals. In addition, weak Br···N and Br···Br XBs have been shown to play a role in the stabilization of the three reversible polymorphic structures of 1-phenyl-2-methyl-4-nitro-5-bromoimidazole (Figure 29).187 The three structures are nearly identical, with only minor changes in the intermolecular interactions which facilitate the reversibility of the structural transitions. The molecules were connected into chains via Br··· N XBs, and the chains, in turn, were connected by type I Br···Br contacts. These results seem to indicate that the weaker XBs involving Cl or Br XB donors allow for greater freedom in the existence of polymorphic crystal structures. 4.4.2. Stoichiometry. In the ﬁeld of crystal engineering, the preparation of higher order molecular assemblies such as binary (two-component) and ternary (three-component) cocrystals is receiving attention as a means to engineer the properties of crystals more subtly.188 In addition, these structures provide vital information regarding the interaction of various supramolecular motifs in the solid state. When attempting to control the stoichiometry of a particular molecular crystal, several factors need to be considered, such as the primary intermolecular interactions as well as the connectivities between the cocrystal formers.189 Aakeröy and co-workers162a presented a series of cocrystals in which attempts were made to obtain particular stoichiometries of cocrystal formers through the use of complementary XB and HB interactions. In the ﬁrst example, a 1:2 stoichiometry was targeted using the ditopic XB donor 1,4-diiodotetraﬂuorobenzene with amide-substituted pyridine (Figure 30). They hypothesized that this would result in I···Npy XBs with selfcomplementary amide···amide HB interactions. The position of pyridyl substitution was also altered (meta vs para) to determine the eﬀect that might have on the overall shape of the extended assembly. Upon crystallization, both strategies

Figure 30. Crystal structures targeting 1:2 stoichiometries: (a) cocrystal of 1,4-diiodotetraﬂuorobenzene and m-amide-substituted pyridine (CSD Refcode TOJBAW) and (b) cocrystal of 1,4diiodotetraﬂuorobenzene and p-methyleneamide-substituted pyridine (CSD Refcode TOJBEA).162a

successfully resulted in 1:2 stoichiometry, with diﬀerences only in the extended connectivity of the trimers: the metasubstituted pyridine formed 2D sheets, while the parasubstituted pyridine formed an inﬁnite ladder through interchain amide HB interactions. Employing more subtle acceptor cocrystallization agents, pyrazine- and benzimidazole-substituted pyridine, with 1,4diiodotetraﬂuorobenzene aﬀorded complicated results (Figure 7139

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

31). The former example formed 2D sheets with an open framework with the expected HB and XB connectivity.

Figure 32. Crystal structure of the attempt to prepare a ternary cocrystal with particular stoichiometry using 1,4-diiodotetraﬂuorobenzene, pyridine-3,5-dicarboxylic acid, and 2-acetamidopyridine (CSD Refcode TOJCEB).162a

delicate balance between intermolecular interactions at the event of nucleation. 4.4.3. Tautomeric Control. It is well-known that properties such as acidity, hydrophobicity, or polarity can vary signiﬁcantly between the particular tautomers of a compound, and while both tautomers exist in solution, as much as 99.5% of tautomeric molecules crystallize preferentially in one tautomeric state, usually the most stable tautomer. Employing crystal engineering techniques through the use of cocrystal formers to inﬂuence molecular tautomerism in the solid state, CruzCabeza and co-workers190 demonstrated that 1-deazapurine could be selectively crystallized as either the more stable 3H or less stable 1H tautomer with the appropriate cocrystal former (Figure 33). Utilizing two diﬀerent XB donors, the more stable 3H tautomer was selectively crystallized since the I···Nimidazole XB did not interfere with the formation of self-complementary hydrogen-bonded dimers of 3H-1-deazapurine. On the other hand, stabilization of the 1H tautomer required a cocrystal former containing two coplanar HB donors capable of interacting with the pyridine and imidazole nitrogen atoms, therefore separated by ∼2.4 Å, and thus able to perturb the 3H tautomer (Figure 33). Consequently, a series of urea derivatives were investigated: six of these indicated new or cocrystal solid forms via powder X-ray diﬀraction, and four of these aﬀorded crystals suitable for single-crystal X-ray diﬀraction analysis. All four of these cocrystal structures selectively stabilized the 1H tautomer in the desired fashion. This work paves the way for the design of new materials that will likely display diﬀerent physical properties due to the inclusion of rare tautomeric forms. 4.4.4. XBs Involving Metals and Metal-Bound XBs. Despite having been around for a while, metal-containing XBs have recently attracted a surge of attention.147a Some of the seminal work in this area by Brammer and co-workers23d,191 examined the potential of metal-bound halogens as hydrogenbond acceptors, and theoretical calculations demonstrated the equatorial belt of negative electrostatic potential develops concomitantly with the formation of a σ-hole as a result of polarization of the halogen atom. Additionally, Awwadi has investigated the role of XB in directing the structures of tetrahalocuprates192 and the analogy between Fe−Cl···Cl−Fe interactions and C−Cl···Cl−C interactions,193 and Aakeröy has investigated the use of XB in constructing 1D chains and 2D sheets through the introduction of XB donors on the coordinated ligands.194

Figure 31. Crystal structures of 1,4-diiodotetraﬂuorobenzene with (a) pyrazine-substituted pyridine forming a 2D sheet with cavities containing molecules of 1,4-diiodotetraﬂuorobenzene (omitted) (CSD Refcode TOJBIE) and (b) benzimidazole-substituted pyridine in which the assembly process is interrupted by the inclusion of methanol (CSD Refcode TOJBOK).162a

However, a free molecule of 1,4-diiodotetraﬂuorobenzene was included in the void space of the framework, resulting in a loss of stoichiometric control. Similarly, in the latter example, a molecule of methanol disrupted the assembly process by forming a hydrogen bond with the benzimidazole moiety, leaving 1,4-diiodotetraﬂuorobenzene to form one XB with the pyridyl moiety; the second iodine atom of 1,4-diiodotetraﬂuorobenzene did not participate in an XB. Similar results were obtained when cocrystal formers containing two pyridyl moieties were used in which one pyridyl group had been activated through the incorporation of an electron-donating substituent. As expected, XBs were preferentially formed with the electron-rich pyridyl moiety but did not aﬀord the desired stoichiometry as a free molecule of 1,4-diiodotetraﬂuorobenzene was included in the structure. These results indicate the diﬃculty in maintaining stoichiometric control. In an attempt to prepare a ternary cocrystal with 1,4diiodotetraﬂuorobenzene, the authors selected pyridine-3,5dicarboxylic acid and 2-acetamidopyridine as the cocrystal formers as they are known to form heteromeric dimers (Figure 32). The expected stoichiometry was 1:2:4 1,4-diiodotetraﬂuorobenzene:diacid:amide with an XB to the pyridine of the diacid. Crystallization aﬀorded the ternary cocrystal; however, the stoichiometry was unexpectedly 2:1:2 as the molecules assembled into a network with the diacid pyridine participating in two XBs to two molecules of 1,4-diiodotetraﬂuorobenzene. The second iodine of 1,4-diiodotetraﬂuorobenzene formed a C−I···OC XB to an acetamidopyridine molecule. These results clearly demonstrate the diﬃculty in controlling the 7140

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 33. Crystal structures showing stable 3H-deazapurine and metastable 1H-deazapurine structures cocrystallized with XB cocrystal formers [(a) 1,4-diiodotetraﬂuorobenzene (CSD Refcode UDEJIY) and (b) 1,2-diiodotetraﬂuorobenzene (CSD Refcode UDEJOE)] and HB cocrystal formers [(c) cyano-substituted diphenylurea derivative (CSD Refcode UDEKEV), (d) nitro-substituted diphenylurea derivative (CSD Refcode UDEJUK), (e) bromo-substituted diphenylurea derivative (CSD Refcode UDEKAR), and (f) diphenylurea (CSD Refcode UDEKIZ)].190

crucial in controlling the solid-state structures when X = I (Figure 34); this was not the case when X = Br. In the former

Nonetheless, XBs between organic donors and metal-bound ligands are more prevalent than metal-bound XBs.195 An investigation by Brammer and co-workers196 of complex formation between isomolecular salts of tetrahedral halometallate anions and rigid halopyridinium cations reﬂected the competitive and cooperative interplay between hydrogenbonding (N−H···X−M) and halogen-bonding (C−X···X−M) interactions and π−π donor−acceptor interactions. It was observed that a change in the halogens, organic and inorganic, resulted in signiﬁcant changes in the strength of HBs and XBs and, concurrently, considerable changes in the crystal structures of these salts. Varying the organic halogen from F to I resulted in an increase in strength of the XB, whereas varying the metalbound halogen from F to I weakened both XB and HB interactions. The associated crystal structures were divided into three types: in type A structures, the organic halogen is F, no XBs are present, and the structures are dominated by HBs, type B structures display type I contacts through the interaction of small organic halogens with large metal-bound halogens while still dominated by HBs, and type C structures contain large organic halogens and small metal-bound halogens engaged in type II interactions which work cooperatively with HBs in directing the solid-state structures. Interestingly, type A and B structures are isostructural within each structure type, while the type C structures encompass a set of structures dependent upon the structure-directing role of the XBs formed. This variance in type C structures may allow them to adopt a variety of polymorphic structures. More recent work by the same group has probed the use of XB in the crystal engineering of cyanometallates197 due to their photophysical properties and the fact that the cyano group can behave as a pseudo-halide, allowing them to form intermolecular interactions analogous to XBs.198 Furthermore, cyanometallate anions are highly susceptible to changes in the coordination environment, resulting in a luminophore capable of both interacting with other components and reporting such interactions through a change in its spectroscopic properties. Crystallizing [Ru(bipy)(CN)4]2− with N-methylhalopyridinium cations resulted in a series of structures of charge-assisted XBs

Figure 34. Crystal structures of halopyridinium−cyanometallate complexes: (a) [N-methyl-3-iodopyridinium]2[Ru(bipy)(CN)4]·0.5(MeCN) showing C−I···NC(Ru) halogen bonds (CSD Refcode QOXFIT) and (b) [N-methyl-3,5-diiodopyridinium]2[Ru(bipy)(CN)4] showing the network propagated by C−I···NC(Ru) halogen bonds (CSD Refcode QOXGAM).197

case, XBs formed analogously with classical type II interactions in which the σ-hole of I interacts attractively with an “externally directed lone pair of electrons” of a metal-bound cyano group. Since cyanometallates are used to construct magnetic and conducting materials, Brammer et al. extended this work with a series of systematic studies designed to investigate the eﬀect of halogen bonding in such systems. Of particular interest are the cyanometallates of Cr and Fe cocrystallized with halopyridinium compounds in which Br/I XB donors were investigated 7141

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

(Figure 35).199 In these systems, two particular geometries of interaction were observed: where the XB donor (i) interacts

laminated structures of alternating anionic and cationic layers, with the anionic tetrahaloaurate 2D sheets propagated by halogen−halogen close contacts. It was also evident that the tetrabromoaurate structures could be obtained from the benzo crown ether and HAuCl4 in HBr solution via in situ halogen exchange, resulting in a structure that was isomorphous and isostructural with the tetrachloroaurate structure. The authors suggested that the complexes are promising candidates for the extraction of gold due to the ease of stoichiometric precipitation of such complexes and their near-quantitative yields. Lastly, some singular studies on the inﬂuence of XB on the structures of uranyl cation [UO2]2+ units are beginning to surface.201 Most recently, Surbella and Cahill 202 have investigated the eﬀect of intermolecular interactions in the stabilization of the anionic uranyl isothiocyanate tecton, [UO2(NCS)4(H2O)]2−. Of particular interest are the XB interactions introduced through the use of halo-substituted pyridinium countercations. In these structures, the chloropyridinium salt displayed an XB to an axial uranyl oxygen, while the pyridinium N−H formed a hydrogen bond with a sulfur atom of a neighboring uranyl isothiocyanate tecton (Figure 37). Structures incorporating the bromopyridinium or iodopyridinium counterions are isomorphous despite diﬀering space group symmetry and, in contrast to the chloro structure, formed XBs with an isothiocyanate sulfur atom and bifurcated HBs with a water molecule and a neighboring uranyl oxygen atom. Their results and the relative strengths of the interactions were rationalized using the HSAB principle: soft XB donors preferred soft XB acceptors, and the converse with hard donors and acceptors is also true. 4.4.5. XB with Anions in the Solid State. In a recent critical review, Metrangolo, Resnati, and co-workers203 discussed XB as a means toward anion coordination and recognition with a focus on oxoanions via a CSD search. The review encompassed both organic and inorganic oxoanions with a variety of XB donor molecules and the use of XB with anions to facilitate solid-state self-assembly. More recently, the same authors discussed the use of anions as XB acceptors: due to the more diﬀuse, and less directional, electron cloud surrounding the anionic species, the prediction and control of the stoichiometry and geometry of any architectures that may be formed is much more diﬃcult, when compared with a lone pair on a heteroatom.169 However, this allows anions to behave as mono-, bi-, and polydentate XB acceptors. Toward that end, anions have been incorporated as nodes with polyiodinated molecules as organic linkers to form extended architectures analogous to well-known metal−organic frameworks (MOFs).204 Fourmigué and co-workers205 have recently reported the use of 1,3,5-triﬂuoro-2,4,6-tris(iodoethynyl)benzene in the formation of anion organic networks (Figure 38). The authors demonstrated the potent XB donor ability of 1,3,5-triﬂuoro-2,4,6-tris(iodoethynyl)benzene through the formation of a DMSO solvate and 1:1 cocrystal structures with various halide salts, which resulted in the assembly of anion organic networks. Of particular interest is the 1,3,5-triﬂuoro-2,4,6-tris(iodoethynyl)benzene−Et3BuN·Br structure, which displayed a rather rare topology of interacting triangular and octahedral nodes, resulting in two interpenetrating frameworks. This was due to the unprecedented octahedral coordination of Br−. Despite the interpenetration of these two networks, there was still void space within the crystal structure which accommodated the countercation and solvent

Figure 35. Crystal structures of halopyridinium−hexacyanometallate complexes: (a) (3-IpyMe)3[Cr(CN)6] showing C−I···N halogen bonds (CSD Refcode YARSOB), (b) (3-IpyMe)3[Fe(CN)6]·2MeCN showing C−I···N halogen bonds (CSD Refcode YARSUH), and (c) (3,5-I2pyMe)3[Co(CN)6] showing C−I···N halogen bonds (CSD Refcode YARVIY).199

with the lone pair of electrons on the cyano N or (ii) interacts with the π-bond of the cyano group. Shorter halogen bonds were observed throughout the series when the interaction formed with the cyano nitrogen lone pair, which suggested that this is the preferred XB acceptor site. In general, a trend of decreasing X···N distances was observed in the order Cr > Fe > Co, which correlated with increasing XB strength. This was explained by the increased metal d-electron count allowing for greater π-back-donation to the cyanide ligands. Consequently, the partial negative charge associated with the cyanide ligand increased, making it a better XB acceptor, which allowed for the formation of stronger XBs. A recent pioneering study with more precious metals described the impact of XB interactions in the stabilization of new benzo crown ether tetrahaloaurate(III) (Cl or Br) complexes (Figure 36).200 All structures obtained exhibited 7142

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 36. Crystal structures of tetrahaloaurate complexes showing intralayer halogen bonds: (a) [(H3O)(B18C6)0.58(4′-Cl-B18C6)0.42][AuCl4] (CSD Refcode QOCVUB), (b) [(H3O)(4′-Br-B18C6)][AuCl4] (CSD Refcode QOCWAI), (c) [(H3O)(4′-Br-B18C6)][AuBr4] (CSD Refcode QOCXIR), and (d) [(H3O)(B18C6)][AuBr4] (CSD Refcode QOCXOX).200

Figure 37. Crystal structures of halopyridinium−uranyl complexes showing: (a) Cl···O halogen bonds between [UO2(NCS)4(H2O)]2− and 4chloropyridinium (CSD Refcode SIWVOL) and (b) Br···SCN halogen bonds between [UO2(NCS)4(H2O)]2− and 4-bromopyridinium (CSD Refcode SIWVUR).202

molecules. This work showed the analogy of anion organic networks with MOFs and the potential applications of anions in the construction of these 3D coordination assemblies. The anion coordination ability of XB has been demonstrated in the solid state by the structures of a tripodal heteroditopic receptor24b with iodide (Figure 39) and mono-, di-, and

Figure 39. Metrangolo and Resnati’s tripodal heteroditopic anion receptor.24b

Figure 38. Crystal structure of Fourmigué’s anion organic network propagated by I···Br− halogen bonds. The octahedral coordination of Br− facilitates the formation of a 2-fold interpenetrated network with 1,3,5-tris(iodoethynyl)-2,4,6-triﬂuorobenzene as the organic spacer ligand (CSD Refcode IFANAG).205

tridentate 2-iodoimidazolium receptors206 with bromide. While examples of chloride, bromide, and iodide as XB acceptors are known,148,207 only one report of ﬂuoride exists.208 Farnham, Dixon, and Calabrese’s study on ﬂuoride showed that discrete ion pairs were formed when pentaﬂuorophenyl iodide was used as the XB donor. However, a polymeric chain could be formed 7143

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

when ditopic diiodooctaﬂuorobutane was used as the XB donor with ﬂuoride bridges. Indeed, in both these instances, ﬂuoride forms bifurcated XBs with the iodo donors. XB is capable of interacting with other anions such as cyanate, azide, and thiocyanate ions, setting a precedent for their use in crystal engineering due to the possibility of multicenter interactions.209 4.5. Solid-State Architectures

A number of groups have prepared remarkable solid-state architectures with XB as the predominant intermolecular interaction responsible for the stability of these structures. In addition to the above examples, chains210 are formed with bis(N-heterocycle)diacetylenes with ditopic iodo donors211 and with bifunctional molecules212 (containing both XB donor and acceptor moieties) that displayed remarkable synthon robustness in the presence of typically structure-destabilizing and bulky substituents. The formation of chainlike structures is useful in the preparation of materials with conducting and magnetic properties as well as noncentrosymmetric crystals capable of second harmonic generation in optical devices. A variant of the 1D chain is a helix which can be formed using XBs as demonstrated by Metrangolo and Resnati et al.162b and van der Boom et al.213 in their studies on halide-templated dibromopyridinium helices and stilbazole-based helices, respectively (Figure 40).

Figure 41. Crystal structure showing the X3 synthon formed via I···I interactions between molecules of 1,3,5-triiodobenzene (CSD Refcode SAQZOY01).214

Figure 42. Crystal structures of β-sheet formation using cooperative XBs and HBs in symmetric bis((4-halophenyl)amido)alkanes: (a) iodo-substituted molecules separated by n = 2, n = 4, or n = 8 methylene spacer units (CSD Refcodes NUJXIA, NUJXOG, and NUJXUM, respectively) and (b) bromo-substituted molecules separated by n = 2, n = 4, or n = 6 methylene spacer units (CSD Refcodes NUJYAT, NUJYEX, and NUJYIB, respectively).215

Figure 40. Crystal structures of XB helices: (a) Metrangolo and Resnati’s helix formed via Br···I− halogen-bonding interactions in Nmethyl-3,5-dibromopyridinium iodide (CSD Refcode BEYQIG)162b and van der Boom’s helix formed via I···I and N···I halogen bonds in the iodo-substituted stilbazole (CSD Refcode ADOLIP).213

of synthon modularity between the cocrystal structures and those of the parent crystals.215 An extension of these 1D chains and 2D sheets is the use of XB in the construction of higher order 3D assemblies: discrete capsules, cages, and networks.216 Aakeröy et al.217 showed that XB can be used to self-assemble two supramolecules, a tetraiodo-functionalized calix[4]arene XB donor and a tetrapyridyl-functionalized cavitand XB acceptor, into a capsule via four XBs (Figure 43a). Interestingly, the use of inorganic iodine and the (4-(benzoyloxy)-2,2,6,6-tetramethylpiperidinyl)1-oxy (BTEMPO) free radical species as starting materials for cocrystallization resulted in the formation of an anionic 3D cage network constructed from iodine and I5− via multiple XBs and templated by (BTEMPO)22+ dimers (Figure 43b).218 The authors proposed a mechanism for this observation conﬁrmed by UV−vis spectroscopy and ESI-MS: an initial XB complex was formed between I2 and BTEMPO radical followed by Milliken inner charge transfer and a subsequent charge

Regarding the formation of 2D sheetlike structures, an interesting XB supramolecular synthon has come to light, namely, the X3 synthon, in which three XBs form a trimeric unit with interactions reminiscent of type II contacts. In the case of trihalomesitylenes, this synthon is propagated in a 2D sheet throughout the self-assembled crystal structure (Figure 41).214 Similarly, due to their biological signiﬁcance, formation of βsheetlike structures is also desirable and can be achieved using a combination of XB and HB as shown by Samai and Biradha in their study of bis((4-halophenyl)amido)alkanes (Figure 42). Their study showed that the iodo-substituted structures were more susceptible to the formation of bifurcated XBs as these help to stabilize neighboring chains. Cocrystallization of these halo-substituted bis(amido)alkanes resulted in the observation 7144

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Traditionally the province of MOFs, porous materials have garnered much attention in recent years due to their molecular sorption properties and potential in catalyzing molecular transformations. In their search for novel porous materials without metals, Rissanen and co-workers have utilized XB as a means toward preparing such structures.221 Recently, the same authors reported a ﬂexible structural framework that was adapted to accommodate solvent guests in nanosized channels via an induced ﬁt mechanism.222 This framework was constructed from hexamethylenetetramine (HMTA) and Niodosuccinimide bonded by (OC−)2N−I···N XBs (Figure 44). Due to the “breathing” nature of the framework, three distinct solvent-dependent structural topologies were observed via single-crystal X-ray diﬀraction analysis. The ﬁrst displayed ovoid channels which include toluene and chloroform molecules, the second exhibited spherical channels containing nitromethane or acetonitrile, and ﬁnally, the last structure formed a square-grid motif with the largest relative channel volume of the three structures containing either dichloromethane or tetrachloromethane. Guest exchange experiments with the third structure indicated that dichloromethane could be exchanged for tetrachloromethane via solvent- or gas-phase single-crystal to single-crystal transformation despite some decay of the crystals due to the exchange. These results provide interesting opportunities for the construction of metal-free organic host structural frameworks. Lastly, due to potential applications in materials chemistry, biomedicine, and nanotechnology, dendrimers have attracted much attention in recent years, and XB has been incorporated into such structures.223 Metrangolo, Resnati, et al. reported the synthesis of DAB-dendr-(NH−C6F4I)22, a poly(propylenimine) dendrimer functionalized with four iodotetraﬂuorophenyl XB donor groups (Figure 45). Crystallization of the dendrimer with bis(4-pyridyl)ethylene resulted in the self-assembly of a 1:2 supramolecular adduct propagated into large 2D square networks with unprecedented 5-fold interpenetration. This work was extended with the self-assembly of polymers up to the millimeter length via XBs.224 This research opens the door for new tools in the design of dendrimers and polymers. 4.6. Applications and Functional Materials

Recent reviews have highlighted how XB can inﬂuence and augment the properties of functional materials.225 This section provides an insight into the potential of XB to determine material properties. 4.6.1. Analytical Chemistry, Chemical Resolution, and Recycling. Halogen bonding has been exploited in the separation of halogen-functionalized perﬂuorocarbon species. Racemic 1,2-dibromohexaﬂuoropropane (Figure 46a) was cocrystallized with enantiopure (−)-sparteine hydrobromide (Figure 46b), forming crystals which exclusively contained the (S)-enantiomer of the perﬂuorocarbon.226 Single-crystal X-ray structural analysis showed that the cocrystal assembly was driven by interactions between the perﬂuorocarbon bromine atoms and the bromide counterions of the sparteine components. In an extension of this resolution method, a strategy was reported for the separation of α,ω-diiodoperﬂuoroalkane mixtures (with n = 2, 4, 6, 8, 10, and 12 perﬂuoromethylene units) (Figure 47a) by cocrystallization with bis(trimethylammonium)alkane diiodide species (with n = 6 methylene units) (Figure 47a).227 The ammonium-appended alkane was found to have a strong size complementarity with

Figure 43. Structures of interesting halogen-bonded supramolecular architectures: (a) Aakerö y’s molecular capsule comprised of a tetraiodo-functionalized calix[4]arene and a tetrapyridyl-functionalized cavitand self-assembled via N···I halogen bonds (CSD Refcode DEGWET),217 (b) Jin’s anionic 3D cage network formed from pentaiodides, iodine, and (BTEMPO)22+ via multiple halogen bonds (CSD Refcode VIBSUW),218 (c) Goldberg’s tetraarylporphyrin, which yields an inherently chiral architecture upon crystallization, stabilized by N···I and I···π halogen bonds (CSD Refcode DIXXEO),219a and (d) Goldberg’s tin-containing tetraarylporphyrin with nicotinic acid stabilized by I···I halogen bonds (CSD Refcode HIRVUB).220

separation reaction. Consequently, the pentaiodide was formed, presumably stabilized by the amphoteric nature of iodine. Goldberg et al.219 have employed XB in the construction of molecular scaﬀolds based on the asymmetric functionalization of tetraarylporphyrin to aﬀord impressive chiral architectures (Figure 43c,d). 7145

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 44. Crystal structure of Rissanen’s ﬂexible XB framework comprised of hexamethylenetetramine (HMTA) and N-iodosuccinimide (NIS) (CSD Refcode YANGEB).222

Figure 47. (a) Molecular structures of the cocrystal formers α,ωdiiodoperﬂuoroalkane and bis(trimethylammonium)alkane diiodide species and (b) single-crystal X-ray structure of diiodoperﬂuorohexane and bis(trimethylammonium)dodecane showing the size complementarity as well as I···I− halogen-bonding interactions (CSD Refcode XOVBOA).227

47b). It was shown that crystals of a dicationic with bis(trimethylammonium)alkane salt (n + 6) would selectively capture the size-matched perﬂuoroalkane (n) in the vapor phase, and heating the crystals resulted in liberation of the perﬂuoroalkane, which regenerated the dicationic salt lattice. Similar results were obtained when iodoperﬂuorooctane was exploited in the catalyst reclamation and recycling of DABCO from the Morita−Baylis−Hillman reaction (see section 6.4).22k XB has recently been applied for the ﬁrst time to the solidphase extraction of perﬂuorinated iodoalkanes (PFIs).228 Employing a strong anion exchange (SAX) sorbent, analysis showed that absorptivities for diiodo-PFIs were stronger than for monoiodo-PFIs due to C−I···Cl− XB interactions with no adsorption observed for perﬂuorobenzene (a control analyte void of XB donors). This work indicated the potential application of XB in the selective extraction of compounds with strong XB ability. Considering that XB donors such as Br or I are heavy atoms, Bhatt and Desiraju229 reported the use of XB as a means to introduce heavy atoms into the structures of chiral molecules via cocrystallization to enable the determination of the absolute

Figure 45. Crystal structures of (a) the XB dendrimer DAB-dendr(NH−C6F4I)22 (CSD Refcode RIWZON) and (b) the same dendrimer cocrystallized with dipyridylethene, forming a pentafold interpenetrated halogen-bonding network via N···I halogen bonds (CSD Refcode RIWZIH).223

Figure 46. Molecular structures of (a) 1,2-dibromohexaﬂuoropropane and (b) (−)-sparteine hydrobromide.226

the perﬂuoroalkane, with interactions between the iodide counteranions and the perﬂuorocarbon iodine atoms (Figure 7146

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 48. Cocrystal structure containing a heavy-atom cocrystal former, 4-iodophenol, for the determination of the absolute conﬁguration of pregnenolone (CSD Refcode WOMGOV).229

Figure 49. Diagram of Fujita’s MOF for the inclusion and subsequent determination of crystal structures of compounds that are diﬃcult to crystallize, chiral, present in trace quantities, and hazardous. Reprinted with permission from ref 230a. Copyright Nature Publishing Group.

use of X-ray crystallography after liquid chromatographic separation of impure samples. This pioneering work allows for the determination of molecular structures that were previously unobtainable through the incorporation of heavy atoms into a host framework. 4.6.2. Surface-Bound Polymers and Thin Films. As a highly directional analogue to HB, XB is an attractive tool for the development of soft functional materials. In particular, the interaction between polarized iodine and electron-rich nitrogen groups, I···N, has been exploited extensively in the construction of ordered surface-bound assemblies. Highly crystalline halogen-bonded thin ﬁlms can be grown on silicon substrates by physical vapor deposition.231 Gottfried and co-workers have used vapor deposition of 3,5,3″,5″tetrabromo-p-terphenyl (TBrTP) in the Cu(111) surfaceassisted fabrication of halogen-bonded, organometallic, and covalent 2D networks. These networks, which were studied

conﬁguration of the enantiomers (Figure 48). Additionally, provided suitable XB acceptors are present on the chiral molecule, XBs may be formed and stabilize the structure by potentially aiding crystallization of molecules that normally would crystallize poorly. Using this methodology, the authors were able to determine the absolute conﬁgurations of pregnenolone, cholesterol, lamivudine, and zidovudine without chemical modiﬁcation. This concept was extended by Fujita and co-workers230 with the ingenious preparation of a ﬂexible host metal−organic framework constructed from ZnI2 and tris(4-pyridyl)-1,3,5triazine ligands capable of including molecular analytes in solution- and gas-phase single-crystal to single-crystal transformations, eliminating the need for crystallization (Figure 49). This framework allowed the trace detection of analytes as well as the determination of the structures of chiral, highly volatile, and explosive compounds. Furthermore, it enabled the tandem 7147

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

the surface; the upper layers of conjugated oligomers, however, diﬀered greatly. The second layer of P3ATs comprised random strands of polymer intersecting the lower layer at random angles, whereas the XB brominated analogue produced a much more ordered second layer in which the polymers exhibit good alignment and periodicity with the ﬁrst layer. Such evidence of a self-templating eﬀect was attributed to S···Br halogen-bond formation.233 Layer-by-layer assembly is a useful method by which multilayer ﬁlms with speciﬁc features can be strategically assembled. Wang and co-workers used the complementarity between halogen-bond-donating iodoperﬂuorophenyl and halogen-bond-accepting pyridyl units to assemble a multilayer ﬁlm in a tetrahydrofuran−chloroform mixture. An aminefunctionalized substrate was immersed in a solution of poly(4(4-iodo-2,3,5,6-tetraﬂuorophenoxy)butyl acrylate) (PIPBA), followed by immersion in a solution of poly(4-vinylpyridine) (PVPy). Cycling these two steps produced a (PIPBA/PVPy)n multilayer ﬁlm of desired thickness through N···I halogenbonding interactions. The construction of the target multilayer ﬁlm was conﬁrmed by UV−vis experiments, which revealed the presence of absorbance bands corresponding to both components, which increased linearly with increasing bilayer number, indicating that the amount of polymer adsorbed in each cycle was constant. Further proof of the formation of a multilayer through XB was provided by XPS analysis, which highlighted the binding energy of N 1s increased by 0.16 eV and the binding energy of I 3d decreased by 0.31 eV compared with that of pure PVPy or PIPBA, respectively. AFM and XRR techniques were also used to analyze the surface morphology and thickness (Figure 52).234 4.6.3. Liquid Crystals. The nature of the halogen bond has stimulated extensive study into its use in liquid crystalline materials. In particular, XB has been eﬀectively exploited by Bruce and co-workers to induce mesomorphism (a state intermediate between liquid and crystalline phases) from nonmesomorphic components. Early work demonstrated that the self-assembled product of (alkyloxy)stilbazole and iodopentaﬂuorobenzene (Figure 53) was a thermotropic mesophase, which transitioned from a nematic phase (molecules arranged in loosely parallel lines) to a smectic A (SmA) phase (molecules ordered in one direction along the layer normal) upon cooling.235 Shortly thereafter, Xu and co-workers reported a polymeric liquid crystal which was stabilized by halogen bonds between

using scanning tunneling microscopy (STM) and X-ray photoelectron spectroscopy in ultrahigh vacuum, contain pores of varying sizes, correlating with the number of enclosed adatoms (an atom that lies on a crystal surface), which are presumably Br atoms. Chemically modifying surfaces is a desirable method by which the physical and chemical properties of a substrate may be altered or controlled. Comblike polymers with perﬂuorocarbon side chain segments have important biomedical and coating engineering applications on account of their surface behavior,232 and their water-repellent nature has been exploited in the ﬁeld of ﬁber processing. An alternative route to traditional covalent synthetic methods for ﬂuorinating surfaces has been demonstrated by Metrangolo, Resnati, and coworkers, who used XB to template the self-assembly of longchain haloperﬂuorocarbons with suitable, XB-accepting hydrocarbon polymers. Halogen-bonded supramolecular assemblies between poly(4-vinylpyridine) and α,ω-diiodoperﬂuoroalkanes were prepared in chloroform (Figure 50). Following solvent

Figure 50. Preparation of the poly(4-vinylpyridine)−haloperﬂuoroalkane system.17a

evaporation, comparison of the infrared and Raman spectra with those of the starting materials revealed the presence of N··· I XB interactions, which was supported by thermogravimetric analysis. It was also shown that these assemblies could aggregate in solution to form a liquid crystalline phase, driven by the XB interaction.17a Charra, Attias, and co-workers have used molecular selfassembly to control the arrangements of organic π-conjugated polymers on surfaces. Depositions of poly(3-alkylthiophene)s (P3ATs) and their brominated analogues on highly oriented pyrolytic graphite (HOPG) were studied using STM (Figure 51). With both polymers, the ﬁrst layer formed a well-ordered two-dimensional polycrystalline arrangement covering most of

Figure 51. Comparison of STM images of (a) P3AT on HOPG at 3.11 × 10−6 M in phenyloctane (IT = 50 pA, VT = −1520 mV, 109 × 109 nm2) and (b) P3AT·Br on HOPG at 7.78 × 10−6 M in phenyloctane (IT = 50 pA, VT = −1440 mV, 109 × 109 nm2). Reprinted from ref 233. Copyright 2011 American Chemical Society. 7148

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 52. Structures of PIPBA and PVPy and an AFM image of a (PIPBA/PVPy)10 ﬁlm on a silicon slide. Reprinted from ref 234. Copyright 2007 American Chemical Society.

halogen bond, which formed to the methoxy oxygen atom. The 1,3-diiodotetraﬂuorobenzene halogen-bond donor (Figure 54d) was shown to form trimeric complexes with (alkyloxy)stilbazole, which produced a chiral nematic phase due to the bent nature of the complex.238 Furthermore, the difunctional halogen-bond donor α,ω-diiodoperﬂuoroalkane (Figure 54e) also formed trimeric complexes with (alkyloxy)stilbazole, producing a monotropic nematic liquid crystalline phase upon cooling from the isotropic phase.239 A comprehensive systematic study was undertaken on the mesomorphic properties of complexes formed between nitrogen bases (alkyloxy)stilbazole (Figure 53a), (R)-citronellyl- or (S)-citronellyl-substituted (alkyloxy)stilbazole (Figure 55a), an alkyl-substituted pyridine (R = C8H12, OC8H12) (Figure 55b), with various substitutional isomers of activated iodo halogenbond donor compounds (Figure 55c).240 For the achiral systems, short and medium chain length complexes gave predominantly nematic phases, with the longest chain lengths forming monotropic SmA phases. The chiral (R)- and (S)citronellyl-substituted complexes were also dominated by chiral nematic phases, with only a few chiral SmA phases observed. Complexes between (alkyloxy)stilbazole and I2 were also recently shown to have high mesophase stability, and unexpected SmC phases (molecules ordered in one direction tilted along the layer normal) were observed in complexes where the alkyloxy chain contains 10 or 12 methylene units.241 Combining the two difunctional components bis(stilbazole) polyethylene glycol as halogen-bond acceptor and iodoperﬂuorobenzene−benzoic acid-functionalized alkane (Figure 56) as halogen-bond donor resulted in polymer formation.242 The unsymmetrical halogen-bond donor resulted in a random polymeric assembly, and consequently, only the less ordered nematic phase was observed. 4.6.4. Surface Relief Gratings. Photoresponsive azobenzene-based liquid crystalline and azopolymer materials have been studied extensively for the formation of surface relief patterns.243 This concept was recently extended to incorporate an iodoperﬂuorobenzene-functionalized azobenzene (Figure 57a) in an XB mesomorphic complex with a pyridyl-appended compound (Figure 57a), which, after photoalignment, eﬃciently formed a surface relief grating (Figure 57).244 A comparison study between analogous HB and XB complexes

Figure 53. Molecular structures of (a) (alkyloxy)stilbazole and (b) iodopentaﬂuorobenzene, used in the early preparation of liquid crystalline materials.235

the bis(diiodotetraﬂuorophenol)alkane halogen-bond donor (Figure 54a) and the bis(stilbazole) halogen-bond acceptor

Figure 54. Molecular structures of (a) the bis(diiodotetraﬂuorophenol)alkane halogen-bond donor, (b) the bis(stilbazole) polyethylene glycol halogen-bond acceptor, (c) 1,4diiodotetraﬂuorobenzene or 4-iodotetraﬂuorophenol, (d) 1,3-diiodotetraﬂuorobenzene, and (e) α,ω-diiodoperﬂuoroalkane, used in the preparation of liquid crystalline materials.17f−i,237b,238

(Figure 54b).17h The same bidentate bis(diiodotetraﬂuorophenol) alkane was also shown to form trimeric halogen-bonded complexes with (alkyloxy)stilbazole (Figure 53a).236 Trimeric halogen-bonded complexes were also reported between (alkyloxy)stilbazole and either 1,4-diiodotetraﬂuorobenzene or 4-iodotetraﬂuorophenol (Figure 54c), with both assemblies demonstrating nematic phases.237 In the solidstate structure of the 4-iodotetraﬂuorophenol complex, the O− H···N hydrogen bond appeared to be favored over the I···O 7149

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

demonstrated that the higher directionality of the halogen bond can enhance the ﬁrst-order diﬀraction eﬃciency, with a substituent-dependent trend of I > OH > Br > H.245 4.6.5. Nanoparticles. van der Boom and co-workers have made use of XB interactions in the solution-phase supramolecular assembly of gold nanoparticles (AuNPs). AuNPs functionalized with (E)-4-(2,3,5,6-tetraﬂuoro-4-iodostyryl)pyridine 1-oxide were synthesized and assembled with a ditopic halogen-bond-accepting linker (BPEB) and analyzed using UV−vis and TEM techniques (Figure 58). Through varying the assembly time and concentration of the BPEB linker, the aggregate level and morphology could be controlled. The templating role of XB interactions was veriﬁed by studying the analogous assembly of either AuNPs with PEB (containing only one halogen-bond acceptor group) or perﬂuoro-functionalized AuNPs with BPEB: in both cases no signiﬁcant clustering of AuNPs was observed by either UV−vis spectroscopy or TEM experiments.246 In an extension of this strategy, AuNPs functionalized with XB donor ligands were self-assembled on surfaces coated with XB acceptors.247 By comparison with their perﬂuoro analogue, van der Boom and co-workers have shown the attachment of AuNPs functionalized with (E)-4-(2,3,5,6-tetraﬂuoro-4iodostyryl)pyridine-1-oxide onto planar functionalized surfaces is mediated by XB interactions (Figure 59). The resultant structure of the surface-bound AuNP assemblies was controlled by the properties of the monolayers, the nature of the XB acceptor cross-linker, and the number of deposition steps.247 4.6.6. Interfaces. van der Boom and co-workers have used force spectroscopy (FS) to probe XB interactions between two complementary organic interfaces immersed in an organic environment. The FS measurements obtained provided direct experimental information about the solvent dependency of the halogen bond. The authors found that polar and competitive media disrupted the interfacial XB, in agreement with DFT calculations. The adhesion between the iodine-terminated monolayer and the alkyl-CN-functionalized AFM tip in hexane and in ethanol was assessed by measuring the total pull-oﬀ force (Fpull‑off) (hexane, 15.1 ± 2.6 nN; ethanol, 0.3 ± 0.2 nN), and the diﬀerence was attributed to the electrostatic nature of XB (Figure 60). No signiﬁcant adhesion was observed with the control monolayer in either hexane or ethanol.248

Figure 55. Molecular structures of (a) (R)- and (S)-citronellylstilbazole and (b) alkyl-substituted pyridine halogen-bond acceptors and (c) para-substituted iodotetraﬂuorobenzene halogen-bond donors.240

Figure 56. Unsymmetrical iodoperﬂuorobenzene−benzoic acidfunctionalized alkane halogen-bond donor.

Figure 57. (a) Azobenzene cocrystal complex studied in the formation of a surface relief grating, (b) atomic force microscopy view of the spin-coated ﬁlm of the halogen-bonded complex after the surface relief grating inscription, and (c) surface modulation depth of the grating. Reprinted with permission from ref 244. Copyright 2012 Wiley-VCH Verlag GmbH & Co. KGaA. 7150

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 58. Two-step process to direct the formation of halogen-bonding AuNP-based clusters. The AuNPs were functionalized with a halogen-bond donor and subsequently exposed to a halogen-bond acceptor (BPEB) to generate large assemblies. Reprinted with permission from ref 246. Copyright 2010 Wiley-VCH Verlag GmbH & Co. KGaA.

positive in both solvents; thus, the NLO tuning eﬀect was attributed to a strong I···O XB interaction between the XB materials and the DMF solvent, which caused an inversion in the direction of μ. Further study into similar XB systems revealed that the solvent dependence on their NLO properties was a consistent eﬀect.253 The vast majority of eﬃcient phosphorescent materials are based on metal complexes,254 and few purely organic phosphorescent materials have been reported.255 Recently, XB interactions were employed in the construction of phosphorescent materials, which emit strongly at wavelengths ranging from blue to red.256 The crystalline materials contained dibrominated aromatic motifs, substituted with aliphatic hydrocarbon chains, and were cocrystallized with an analogous compound in which a bromine atom had been substituted for an aldehyde group (Figure 62). The phosphorescence was purported to arise from a combination of eﬀects: in the crystal, halogen bonds between the bromine and aldehyde motifs promoted singlet-to-triplet conversion through the “heavy atom eﬀect”, and the halogen bond was also believed to prevent nonradiative decay pathways by increasing the quantum yield of the phosphorescence. Furthermore, diluting the aldehyde chromophore in a matrix of a structurally similar compound suppressed self-quenching, and resulted in more eﬃcient emission. Phosphorescent organic materials have also been constructed by cocrystallization of halogen-bond acceptors with iodoper-

Diﬀerent scanning force microscopy (SFM) techniques have also been used in the study of the crystal faces of XB cocrystals of long-chain perﬂuorocarbons. This has provided information regarding the composition of various crystal planes and can present an alternative tool to X-ray diﬀraction techniques for the analysis of highly ﬂuorinated materials.249 In another study, it has been shown that ethers can form an XB bridge between dichlorotriazine-based interfaces.250 4.6.7. Materials with Optical Properties. In the endeavor to develop materials for optoelectronic applications, XB has been used in the self-assembly of organic thin ﬁlms with nonlinear optical (NLO) properties. Generally, in the solid state, noncentrosymmetric space groups are required to obtain such materials, and XB has been shown to be eﬀective in controlling structures to achieve this.251 The self-complementary terminal binding sites of XB tectons (Figure 61) have been shown to induce ordered polymerization in the solid state.252 Study of their NLO behavior using electric ﬁeld induced second harmonic generation (EFISH) experiments revealed a marked solvent dependence for the XB materials: in chloroform, the μβλ values were +124 × 10−48 and +192 × 10−48 esu for iodoperﬂuoro(dimethylamino)benzene-functionalized stilbazole and the iodoperﬂuoro(dimethylamino)benzene-functionalized butadiene analogue, respectively, while in DMF, the μβλ values were negative, at −380 × 10−48 and −465 × 10−48 esu, respectively. In comparison, the HB analogue, protoperﬂuoro(dimethylamino)benzene-functionalized butadiene, remained 7151

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 59. (A) Formation of assemblies consisting of functionalized AuNPs and organic cross-linkers (BPEB) in solution.246 (B) Schematic representation of the stepwise generation of assemblies consisting of functionalized gold nanoparticles (AuNP-1) and organic cross-linkers (BPEB, TPEB, and TPM) on organic monolayers (M1, M2). Reprinted from ref 247. Copyright 2011 American Chemical Society.

ﬂuorobenzene-based components. C−I···π interactions have been exploited in the assembly of phosphorescent cocrystals of 1,4-diiodotetraﬂuorobenzene with naphthalene, phenanthrene,257 ﬂuorene,258 or carbazole,259 which emit strong green, orange, green, and orange-red phosphorescence, respectively. C−Br···π interactions have also been exploited to a similar eﬀect.260 Using XB to promote organic phosphorescence has also recently been demonstrated in a capped γ-amino acid.261 In addition to incorporation in phosphorescent materials, halogen bonds have been employed in the formation of ﬂuorescent materials. In particular, the ﬂuorescence emission of cocrystals containing stilbene-based components could be modiﬁed through the use of XB tectons.262 It has also been shown that ultrasound-assisted crystallization can be used to produce nanocrystals of the XB cocrystals, which display thermosensitive ﬂuorescence.263 The self-assembly of halogenated squaraine dyes was partially driven by dispersive

halogen−arene interactions.264 Solid-state XB interactions between the [Ru(dcbpy)2(SCN)2] dye (dcbpy = 2,2′bipyridyl-4,4′-dicarboxylate) and I2 provided useful information on the mechanism by which the I−/I3− electrolyte regenerated the dye in dye-sensitized solar cells.265 4.6.8. Solid-State Molecular Recognition. It has been observed that cucurbit[6]uril (CB6) forms stable complexes with molecular dibromine and diiodine (Figure 63).266 In the solid state, this host−guest assembly displayed two diﬀerent types of halogen bonds: one “conventional” linear bond between the dihalogen molecule and an included water molecule H2O···I2, and an unusual perpendicular halogen bond between the dihalogen and the carbonyl groups on the CB6 upper rim. An XB analogue of a deep cavity cavitand has also been reported in the solid state, within which a resorcinarene derivative formed an adduct with bromide anions and molecules of bromotrichloromethane.267 7152

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 60. (a) Schematic representation of the force spectroscopy setup using an alkyl-CN-functionalized AFM tip (XB acceptor) and the modiﬁed surfaces M1 (XB donor) and M2 (a control). (b) Molecular structures of compounds 1 and 2. (c) NClplot for the halogen-bonded model system. Reprinted with permission from ref 248. Copyright 2013 The Royal Society of Chemistry.

Figure 61. Tectons studied for NLO behavior: (a) iodoperﬂuoro(dimethylamino)benzene-functionalized stilbazole and (b) haloperﬂuoro(dimethylamino)benzene-functionalized butadiene (X = I or H).252,253 Figure 63. Single-crystal structure of CB6·I2 (CSD Refcode KEQDUH).266

4.6.9. Topochemical Reactions and Solid-State Synthesis. Recently, Biradha and Santra268 reviewed some of the most recent work into the crystal engineering of molecular

Figure 62. (a) Structures of the phosphorescent crystal hosts/aldehydes and photographs of (b) I(Br)·I(CHO), (c) II(Br)·II(CHO), (d) III(Br)· III(CHO), and (e) IV(Br)·IV(CHO) cocrystals. Reprinted with permission from ref 256. Copyright Nature Publishing Group. 7153

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 64. Molecular structures of (a) the tetrakis(iodoperﬂuorobenzene) XB donor, (b) trans-1,2-bis(4-pyridyl)ethylene, (c) tetrakis(4pyridyl)cyclobutane, (d) the iodoperﬂuorobenzene- and (dimethylamino)benzene-containing tecton, and (e) the iodoperﬂuorobenzene- and (dimethylamino)benzene-functionalized cyclobutane.276,277

solids for use in topochemical reactions. Kohlschütter posed the topochemical principle, which states that “reactions in crystals proceed with minimum atomic and molecular movement”.268,269 Such reactions include [2 + 2] photochemical reactions in which oleﬁns are dimerizedby far the most extensively studied dimerizations.270 Where XB crystal engineering principles are concerned, these may be divided into those cases where there is “halogen substitution”, when the XB donor is incorporated into the molecule to undergo photodimerization, and templation, when the cycloaddition molecule contains no XB donors but rather acceptors; in the latter case, an XB donor template may be employed to engineer a structure suitable for photodimerization reactions. The use of halogen substitution for the generation of crystal structures for use in solid-state [2 + 2] photocycloaddition reactions has garnered some attention in the area of topochemical reactions. These structures are generally characterized by a short axis of 4 Å, which is the “maximum” distance required for cycloaddition.270 Biradha and Santra268 made mention of the following contributions, which are repeated again here brieﬂy for completeness of the review. Schmidt and co-workers made notable progress in implementing this rule in the design of several solid-state photocycloaddition reactions of this type.271 Desiraju and co-workers expanded on this work and demonstrated the importance of Cl···Cl and C−H···Cl contacts in the stabilization of these structures272 and developed a set of rules for engineering structures which adopt β-packing.273 Ramamurthy and co-workers further investigated the photochemical dimerization of coumarins in the solid state and showed that molecules were capable of dimerizing at distances much greater than the accepted range of ∼4 Å.274 More recently, crystal structures have been engineered using type I and II contacts to orient the monomers for photochemical cycloaddition reactions.275 This was successfully exploited in the preparation of a cocrystal assembled through XB interactions between tetrakis(iodoperﬂuorobenzene) (Figure 64a) and trans-1,2-bis(4-pyridyl)ethylene (Figure 64b).276 The ethylene groups in the lattice were suitably aligned to enable photocyclization upon irradiation at 300 nm, quantitatively forming tetrakis(4-pyridyl)cyclobutane (Figure 64c). A similar solid-state reaction has also been achieved by the selfcomplementary iodoperﬂuorobenzene- and (dimethylamino)benzene-containing tecton (Figure 64d), which self-assembles via halogen bonds in the crystal, and undergoes a photocyclization reaction to form the heterotetratopic species (Figure 64e).277

Polydiacetylenes have attracted intense interest for their electronic and optical properties.278 Similar to the [2 + 2] photocyclization reaction, there are certain structural requirements in the solid state to enable this polymerization to occur.279 To achieve these conditions, Goroﬀ and co-workers exploited the highly directional nature of XB to form cocrystals of diiodobutadiyne (Figure 65a) in conjunction with a

Figure 65. Molecular structures of (a) diiodobutadiyne, (b) the bisnitrile-functionalized oxalamide compound, (c) poly(diiododiacetylene), and (d) the single-crystal structure of poly(diiododiacetylene) and the structure-directing bisnitrile compound after polymerization (CSD Refcode CEKFOO).21c

structure-directing bisnitrile compound (Figure 65b).21c,280 HB interactions between oxalamide motifs of the bisnitrile compound favored a 1D crystal packing behavior, while halogen bonds between the polarized iodine atoms of diiodobutadiyne and the nitrile groups of the structuredirecting cocrystal former appropriately aligned the acetylene moieties to form poly(diiododiacetylene) (Figure 65c,d). Other bisnitrile-appended halogen-bond acceptors have also been 7154

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

used to form this polymer,281 and pressure-induced polymerization using a bispyridine-functionalized Lewis base has been demonstrated.282 It has also been shown that the iodine substituents of the polymer may be easily removed by heat283 or reaction with a Lewis base,284 which resulted in carbonization of the material and provided a mild route to prepare carbon nanomaterials. Introducing a cocrystallizing XB template into crystal structures has allowed for photochemical cycloaddition reactions to occur between molecules that do not contain any halogen atoms. Rather, the simple molecules containing XB acceptors are aligned by the template via XBs for eﬃcient conversion upon photochemical irradiation.16a,285 The solid-state synthesis of mixed trihalide species has also been reported, in which crystals of the bishalide salts of 1,6bis(trimethylammonium)hexane absorb dihalogen molecules from the gas or solution phase to form the corresponding trihalide.286 Speciﬁcally, this method was used to form [Br3]−, [I3]−, [Br2Cl]−, and [I2Cl]− species within the crystal lattice. It was further shown that heating the bis-I2 adducts could lead to the elimination of one I2 molecule, quantitatively forming the tetrahalide species [I4]2−, [I2Br2]2−, and [I2Cl2]2− within the crystal lattice.287 In other examples of solid-state reactions, Brammer and coworkers have shown two examples288 in which an MOF can be postsynthetically modiﬁed by the absorption of molecules directly from the gas phase into nonporous crystals, which resulted in a very rare change in covalent bonding of the constituent parts (Figure 66). In both cases, Cu−Nhalopyridine

Figure 67. Crystal structures of cocrystals of radicals stabilized by halogen bonds: (a) 1,4-diiodotetraﬂuorobenzene with PhNN (CSD Refcode ZEFKIG) and (b) 4,4′-diiodooctaﬂuorobiphenyl with PhNN (CSD Refcode ZEFKOM).290

(Figure 67b) as cocrystal formers. Single-crystal X-ray analysis revealed the organization of the radical species into 1D chains via C−I···O−N XBs in 100% supramolecular yield: all donors and acceptors interacting. The bond distances in the radical species were typical, indicating a negligible inﬂuence of XB on the electronic properties of the radical. Very weak antiferromagnetic interactions were observed using the 1,4diiodotetraﬂuorobenzene spacer, while these interactions were much stronger when diiodooctaﬂuorobiphenyl was used despite the magnetic isolation that resulted from employing a longer XB donor. This indicates that magnetic interactions are not transmitted through the XB donor molecules. Analysis of the crystal packing eﬀects explains these observations: while both structures displayed 1D XB chains, these structures arranged the radicals into chains of alternating dimers of PhNN. In the 1,4-diiodotetraﬂuorobenzene structure, these dimers were very similar and were related by a center of symmetry; the dimer and interdimer distances were very similar. However, in the diiodooctaﬂuorobiphenyl structure, the dimers were more well-deﬁned; dimers were spaced more closely and the interdimer distance was greater than in the 1,4-diiodotetraﬂuorobenzene structure, implying stronger dimer interactions and weaker interdimer interactions. Consequently, the diiodooctaﬂuorobiphenyl structure can be described as an antiferromagnetic chain, which was conﬁrmed by α and Jrad−rad values. EPR spectroscopy conﬁrmed that magnetic exchange does not propagate through XBs. 4.6.11. Conducting and Magnetic Materials. A more recent example of radical stabilization through the use of XBs was reported by Jin and co-workers.291 Their study examined the packing eﬀects of 1,2- and 1,4-diiodotetraﬂuorobenzene on the structure of the BTEMPO free radical (Figure 68). Two diﬀerent packing structures were observed: 1,2-diiodotetraﬂuorobenzene formed cyclic tetramers, while 1,4-diiodotetraﬂuorobenzene formed 1D zigzag chains in the solid state. In this way, BTEMPO could be arranged to pack in a head-tohead fashion (1,2-diiodotetraﬂuorobenzene) or side-by-side (1,4-diiodotetraﬂuorobenzene). Furthermore, while the solidstate free radical structure and the XB cocrystals all displayed antiferromagnetic coupling, controlling the structure through use of XBs enhanced the magnetic coupling of the cocrystals.

Figure 66. Reaction of green microcrystalline trans-[CuBr2(3-Brpy)2] with HCl to yield brown microcrystalline (3-BrpyH)2[CuX4] (X = Cl and/or Br). Reprinted with permission from ref 288a. Copyright 2010 Wiley-VCH Verlag GmbH & Co. KGaA.

and H−Cl bonds were broken and replaced with Cu−Cl and H−Nhalopyridine bonds. Powder X-ray diﬀraction experiments showed the change in bonding and the reordering of the molecules in the crystal structures of trans-[CuCl2(4-Clpy)2] (4-Clpy = 4-chloropyridine) and (4-ClpyH)2[CuCl4]. In addition, short XB contacts were observed in stabilizing these structures. This work paves the way for greater versatility in the preparation of solid-state materials, particularly those containing metals, through postsynthetic modiﬁcation. 4.6.10. XBs with Radicals. The possibility of forming organic molecular magnets through the use of long-lived radical species is attainable by using XB to engineer crystal structures in which the radical species are ordered and magnetic exchange pathways are established. Indeed, mononitroxyl radicals are able to act as XB acceptors.289 Espallargas and co-workers290 have studied the structures of the bidentate spin donor nitronyl nitroxide (PhNN = 2-phenyl-4,4,5,5-tetramethylimidazolin-1oxyl 3-oxide) radical with bidentate XB donors 1,4-diiodotetraﬂuorobenzene (Figure 67a) and 4,4′-diiodooctaﬂuorobiphenyl 7155

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

been explored in the development of ferromagnetic294 and antiferromagnetic295 materials. 4.6.12. Halogen-Bonding Supramolecular Gels. A gel has “a continuous structure that is permanent on the analytical time scale and is solid like in its rheological behaviour” 296 and usually comprises two (or more) components where one is a dispersion within the other. Hydro- and organogels are dispersions of liquid within a solid, aerogels are dispersions of gas within a solid, an aerosol is a dispersion of liquid in a gas, and a sol is a dispersion of colloidal particles in a solvent. The properties of gels are wide-ranging and easily tunable and depend on the nature or concentration of the gelling component (the gelator). The supramolecular nature of these gels is such that their dynamic self-association can be perturbed by other competing intermolecular interactions, pH, photo, or redox stimuli, and sonication, and a number of research groups have reported the sol−gel transition can be turned oﬀ or turned on by such eﬀects.297 Anions are good hydrogen-bond acceptors, and Steed and co-workers have shown anions were able to compete with urea self-assembly, which suppressed the urea α-tape motif responsible for gelation of bisurea components.297a−d Acetate, for example, has been used to trigger a gel−sol transition of bisurea gelator in acetonitrile (Figure 70).297e In contrast, the bis(pyridylurea) component (Figure 71) did not gel polar organic or organic−aqueous solvent owing to the presence of gel-inhibiting HB between pyridyl and urea groups, which prevented the formation of the inﬁnite urea α-tape motif.299 In the presence of 1,4-diiodoperﬂuorobenzene, gelation of polar organic−water mixtures has also been reported. Gelation was attributed to XB formation between polarized iodine atoms and XB acceptor pyridyl subunits, as shown by the 1:1 single-crystal X-ray structure (Figure 71a). XB disrupts the pyridyl−urea HB and enables the urea NH··· OC HB interaction to form and was thus shown to be responsible for a “turn-on” response.298 A urea-based gelator containing XB donor groups (Figure 72) has also been investigated to study the generality of XB in supramolecular gelation; gelation depends on the relative strength of HB urea−carbonyl (NH···OC) versus XB iodine− carbonyl (I···OC) interactions, which either enable or inhibit gelation. This species was found to be a nongelator in DMSO and DMSO/water mixtures, suggesting urea carbonyl− perﬂuoro iodine halogen bonding occurs in preference to gelinducing urea α-tape hydrogen bonding. Addition of strongly halogen-bond-accepting 4,4′-bipyridine produced a network of ﬁbers and turn-on gelation in DMSO/water medium. Analogous gelation with pyridine was not observed, suggesting cross-linking of adjacent urea tape strands by a bidentate XBaccepting molecule is crucial for robust gel formation. Gelation, although weak, was also observed in a 1:1 mixture of bis(pyridylurea) and bis(iodoperﬂuorobenzene−urea) gelators.298 The strength of XB in polar aqueous media is suﬃcient to favor gelation by bis(pyridylurea) motifs, similarly to metal coordination, highlighting the promising potential of these interactions in supramolecular gel applications.

Figure 68. Crystal structures of nitroxyl radicals stabilized by halogen bonds: (a) zigzag structure formed upon cocrystallization of BTEMPO with 1,4-diiodotetraﬂuorobenzene (CSD Refcode HISZIU) and (b) discrete cyclic tetramers formed upon cocrystallization of BTEMPO with 1,2-diiodotetraﬂuorobenzene (CSD Refcode HISZEQ).291

Quantum calculations on these structures indicated that XB can play a role in the transmission of spin polarization in the crystal lattice. This suggested that an XB may be employed as a “spin coupler” by introducing spin densities on the XB sitesan exciting possibility in the design and assembly of artiﬁcial spin systems. In a development of previous work in the exploitation of XB in tetrathiafulvalene-based molecular conductors,292 Fourmigué and co-workers recently reported a series of cocrystals incorporating diiodo-functionalized tetrathiafulvalene (Figure 69a).293 This tecton was crystallized with various oxygen-based

Figure 69. Molecular structures of XB donor and acceptor components used for magnetic and conducting materials: (a) diiodofunctionalized tetrathiafulvalene, (b) racemic camphor sulfonate, (c) 1,5-naphthalenebis(sulfonate), (d) 2,6-naphthalenebis(sulfonate), and (e) 2,6-anthracenebis(sulfonate).293a

XB acceptors, including perchlorate (ClO4−), racemic camphor sulfonate (Figure 69b), 1,5- and 2,6-naphthalenebis(sulfonate) (Figure 69c,d), and 2,6-anthracenebis(sulfonate) (Figure 69e). Although diamagnetism was observed in the diiodotetrathiafulvalene−ClO 4 and (diiodotetrathiafulvalene) 2 −2,6naphthalenebis(sulfonate) materials, the assemblies of diiodotetrathiafulvalene with racemic camphor sulfonate, 1,5naphthalenebis(sulfonate), and 2,6-anthracenebis(sulfonate) exhibited semiconducting behavior. XB interactions have also

4.7. Conclusions and Outlook

Analysis of solid-state architectures has aﬀorded a wealth of information pertaining to the nature, geometry, control, and interplay of XB with other intermolecular interactions, most notably HB. While the strength of XB donors can be correlated 7156

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 70. (a) H-bonding interactions in the urea α-tape motif and their disruption by anion coordination. Reprinted with permission from ref 298. Copyright Nature Publishing Group. (b) Competing anion binding and gelation equilibria in a gelator. Reprinted with permission from ref 297e. Copyright The Royal Society of Chemistry.

Figure 71. (a) Structures of the gelator and halogen-bond-donating 1,4-diidoperﬂuorobenzene. (b) Fast cooling of 1:1 bis(pyridylurea)−1,4diiodoperﬂuorobenzene in 4:1 methanol/water forms a robust hydrogel. (c) X-ray crystal structure showing the desired gel-forming urea α-tape interaction and the XB cross-links involving the pyridyl groups (CSD Refcode NEJVOP). Reprinted with permission from ref 298. Copyright Nature Publishing Group.

with the halogen atom polarizability, the ranking of XB synthons is a much more challenging task. However, such information is necessary when considering the very delicate balance of factors controlling nucleation and crystal growth and ultimately determining the structure of molecular crystals. XB is shown to be a vital factor in each of these areas, and the availability of structural data through the CSD certainly makes the design of tectons for the construction of extended architectures and their interactions more predictable. Indeed, studies highlighting the control of polymorphism, stoichiometry, and tautomers through the use of neutral and charged organic as well as inorganic XB donors, are invaluable to the researcher. As shown in many examples, the robustness, strength, and stringent geometric constraints of XB synthons implies that their design into complex structural frameworks will allow them to behave reliably and eﬃciently. XB has already been incorporated into structures ranging from the simplest discrete interactions and 1D chains to

extended 3D cages, networks, and porous frameworks. In addition, XB has been applied to the informed design of many functional materials that display enhanced properties over those analogous systems without XB. It is likely that, with many synthetically accessible targets, XB will be incorporated into many more solid-state frameworks, further developing the understanding of XB synthons and their concerted action with other intermolecular interactions. Furthermore, it is expected that the burgeoning use of XB in the design and fabrication of functional materials such as liquid crystals, conducting and magnetic materials, nanoparticles, and optically active materials, in solid-state molecular recognition, and in topochemical synthesis, will allow for the development of improved display, digital storage, and sensing devices. Moreover, the knowledge gained from solid-state research into XB will further inspire its application in the biological and medical arenas as well as the solution state. 7157

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 72. (a) Structures of the bis(perﬂuoroiodobenzene−urea) gelator and halogen-bond-accepting 4,4′-bipyridine. (b) Mixtures (1% by weight) of (a) gelator, (b) gelator/pyridine, and (c) gelator/bipyridine in 3:1 DMSO/water. (c) SEM image of the dried gel gelator/bipy. Reproduced with permission from ref 298. Copyright Nature Publishing Group.

5. HALOGEN BONDING IN BIOLOGICAL SYSTEMS The importance of XB in biological systems has only recently been highlighted.300 The use of halogen atoms, in particular the lighter elements, as pharmaceutically active ligand substituents is widespread in medicinal chemistry: around 50% of molecules applied in high-throughput screening are halogenated, as are around 25% of the market-leading drugs by revenue. Furthermore, an estimated 25% of medicinal chemistry papers and patents involve the addition of halogen atoms at a late stage of the synthesis.301 Until recently, halogens have been perceived by the medicinal chemistry community as merely lipophilic moieties, involved in nondirectional hydrophobic interactions pointing into vacant spaces with the ligand binding site, and not contributing to the overall protein−ligand complex stability. Halogens are typically incorporated during the drug design process to enhance membrane permeability and to delay the catabolic processes that degrade the drug’s half-life and eﬃcacy. It is only in recent years that the stabilizing eﬀect of XB within protein−ligand complexes has been highlighted, and that the ﬁrst systematic studies of the eﬀects of biological XBs on drug eﬃciency and rational design of halogen-bonding drugs have been reported (vide infra). However, the vast majority of biological halogen bonds formed between a bound ligand and an accessible Lewis base in the protein binding site, such as a backbone carbonyl oxygen or side chain moiety, have been identiﬁed by post hoc analysis of crystallographic data deposited within the Protein Data Bank (PDB) (Figure 73).302

Figure 73. XBs in biological systems. XBs formed between halogenbond donors and acceptors within the peptide backbone or side chains are deﬁned by a bond length d (red) that is shorter than the sum of the respective van der Waals radii, and with a C−X···Y bond angle θ close to 180°. XBs formed with π-systems are deﬁned by the distance from the centroid of the ring and bond angle α from the vector normal to the plane of the ring.

interactions, as deﬁned by the aforementioned bond distance parameters d and deﬁning the halogen bond angles θ as larger than 140°. In total, 248 distinct XB interactions were identiﬁed, in which the halogen atom was distributed as 47% Cl, 22% Br, and 31% I. In recent years, the number of structures deposited in the PDB in which there is a bound halogenated ligand has increased to over 2000, and further systematic studies on the occurrence and geometry of biological halogen bonds have been undertaken.26a,300,304 The crystal structure data in the PDB suggest that the strict constraints on XB geometries that have been inferred from small-molecule solid-state crystal structures and computationally modeled gas-phase interactions discussed earlier may be relaxed in biological systems and are much less predictable. Within a protein−ligand binding site, a multitude of interactions (electrostatics, van der Waals, and HB) in addition to XB are involved in ligand binding. As such, the ligand binding mode and hence the geometry of a stabilizing halogen bond results from a balance of a large number of possible ligand conformations, and this is manifested in a broader range of XB geometries observed in protein−ligand complexes than in more simple nonbiological systems. The mean C−X···Y angles (θ) were found by Zhu et al. to be lower than the optimum 180°,

5.1. Analysis of the Protein Data Bank: Halogen Bonds to Amino Acids

A survey of the PDB by Ho and co-workers in 2004 revealed that XB interactions were common in the crystal structures of proteins with bound halogenated inhibitors.25b A total of 66 out of the 226 entries that contained ligands with C−X bonds showed O···X bond distances of less than the sum of the vdW radii [i.e., d(Cl−O) < 3.27 Å, d(Br−O) < 3.37 Å, and d(I−O) < 3.50 Å]. Later examinations of the PDB in 2010 by Zhu and co-workers303 revealed over 1000 structures in which the bound ligands are halogenated, of which 154 exhibited XB 7158

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

perpendicular to it.306 Calculations on a N-ethylpropanamide model system interacting with bromobenzene revealed that an orthogonal XB to the carbonyl oxygen did not aﬀect the strength of the intramolecular HB formed between the amide NH and the carbonyl oxygen (Figure 75a). Importantly, this

reﬂecting the wider range of geometries found in biological systems (Table 1).303 Such deviations are detrimental to the Table 1. Statistical Evaluation of Biological Halogen-Bond Structural Dataa C−X···Y C−Cl···O C−Br···O C−I···O C−Cl···N C−Br···N C−I···N C−Cl···S C−Br···S C−I···S

no. of PDB entries

no. of XB contacts

mean X···Y distance (Å)

mean C−X···Y angle (deg)

139 90 76 37 8 7 17 3 5

3.040 3.132 3.269 3.082 3.121 3.193 3.363 3.008 3.415

158.5 160 161.2 155 158 155.8 153.5 163.3 155.1

C−X···π

85 60 50 27 8 5 11 3 5 no. of PDB entries

no. of XB contacts

mean X···centroid distance (Å)

mean angle to normal α (deg)

C−Cl···π C−Br···π C−I···π

106 29 11

124 53 12

3.854 3.944 4.112

161.5 153.3 155.4

Figure 75. Orthogonal XB and HB to a peptide carbonyl oxygen: (a) model demonstrating energetically independent HB and XB, (b) representation of dual HB−XB stabilization of a ligand−protein complex, and (c) complex of 3,5,3′,5′-tetraiodo-L-thyronine binding within transthyretin (PDB entry 1SN0). Dashed lines denote XB (red) and HB (blue).

a

Adapted with permission from ref 26a. Copyright 2010 PCCP Owner Societies.

strength of the halogen bond: deviations of 25−30% corresponded to around a 50% reduction in halogen-bond strength compared with the optimum geometry of θ = 180°.305 The most common interactions are X···O, accounting for 53% of all observed XBs, of which three-quarters are to backbone carbonyl oxygen atoms, followed by X···π (33%), X···N (9%), and X···S (5%).26a The X-ray crystal structure of the complex formed by a brominated inhibitor with human aldose reductase has been solved at high resolution (0.66 Å) and provided an opportunity to observe with high precision a biological XB (PDB entry 1US0).25c A halogen bond was observed between the ligand bromine atom and the side chain hydroxylic oxygen of Thr113 (d = 2.97 Å, θ = 153°, Figure 74).

suggested that XB can be introduced as a recognition interaction without disrupting the HB-stabilized protein structure. A halogen bond formed between the σ-hole of the ligand halogen atom and a suitable Lewis base (e.g., backbone carbonyl oxygen) may thus be supplemented by a simultaneous HB to the belt of negative electron density on the halogen atom (Figure 75b).307 This concept was observed within a naturally occurring protein−ligand complex, that of thyroid hormone 3,5,3′,5′tetraiodo-L-thyronine binding within transthyretin, an extra cellular transport protein which is involved in the binding and distribution of thyroid hormones and vitamin A (Figure 75c).308 The crystal structure (PDB entry 1SN0) revealed that the ligand formed two XBs with the carbonyl oxygen atoms of neighboring β-sheets (d = 3.29 and 3.35 Å and θ = 162.3° and 169.7°, respectively). In addition, HBs from the peptide backbone amide to the negative belt of electron density of the iodine atom were observed in the crystal structure, orthogonal to the XB interaction (α(NH−O−I) = 88° and 90°). Orthogonal HB−XB interactions were also observed in the structure of human serum albumin (HSA), a transport protein important for drug action, with 2-hydroxy-3,5-diiodobenzoic acid.309 One of the iodine atoms formed an XB with the carbonyl oxygen of Arg257 (d = 3.46 Å, θ = 169.4°), while an orthogonal HB to the same iodine is observed (α = 87°). The second most common XB interaction observed in the crystal structures deposited in the PDB is that of a halogen bond formed perpendicular to a π-system.310 As with C−X···Y (Y = N, O, S) interactions, the mean bond length between the halogen atom and the aromatic ring increased in the order Cl < Br < I. The distribution of angles α between the vector along the halogen−centroid line and the normal to the ring was 25°, while the π···X−C angles fell predominantly between 160° and 180°. Matter et al. studied the Cl···π interactions of chlorinated inhibitors of factor Xa,311 in which the halogen bond forms

Figure 74. High-resolution (0.66 Å) crystal structure of human aldose reductase complexed with a brominated inhibitor (PDB entry 1US0). The Br ···O XB to the hydroxylic oxygen of the Thr113 side chain is shown (red dashed line).

The peptide amide moieties of the protein backbone participated in HB, which stabilized and deﬁned the secondary and tertiary structures of the protein. As a result, the carbonyl oxygen nonbonding electrons are involved in hydrogen-bond formation, so the only electron density available for accepting XBs comes from the CO π-electrons. The X···O XB in such systems, where the carbonyl oxygen additionally formed a simultaneous HB, has been shown by Ho and co-workers to be energetically independent of the HB, and orientated 7159

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

clin-dependent kinase 9 (Cdk9) (Cdk9 inhibition contributes to the anticancer eﬀect of many Cdk inhibitors, and as a consequence its selective inhibition is of great interest).315 DRB binding induces a conformational change in Cdk9, and binds via two Cl···O XBs (Figure 77), with the binding supplemented by an orthogonal NH···X HB to the chlorine atom’s negative belt of electron density (PDB entry 3MY1). In contrast, DRB bound to the structurally similar, yet conformationally less ﬂexible, Cdk2 active site in a diﬀerent orientation and is 282fold weaker. The crystal structure (PDB entry 3MY5) revealed the presence of one Cl···O XB, supplemented by a Cl···π XB interaction. Halogen bonds between ligands and protein side chains are not limited to X···π interactions, but can form with any available Lewis base, including hydroxyl in serine, threonine, and tyrosine, carboxylate in aspartate and glutamate, sulfur in cysteine and methionine, and nitrogen in histamine (although few examples have been reported, as a consequence of competing HB, metal coordination, and protonation).300 The ﬁnal type of halogen bond observed in protein−ligand crystal structures is that of the halogen−water−hydrogen bridge (XWH).316 In analogy to the classical HB water bridge, within an XWH bridge one of the HBs is replaced by an XB, but behaves in an otherwise similar way. Computational analysis has indicated that the XWH is thermodynamically stable, and is enhanced by cooperativity between the XB and HB interactions. For example, in the complex of a chlorobenzene-derived inhibitor with c-Jun N-terminal kinase (JNK), an XWH bridge is observed in the crystal structure (PDB entry 2B1P). An XB is formed between the chlorine atom and the water oxygen atom (Cl···O, d = 3.1 Å, θ = 164.5°), while the water forms HBs to three surrounding amino acid residues (Figure 78a). In the complex of diclofenac (a chlorinated nonsteroidal anti-inﬂammatory drug) with cytochrome P450 (PDB entry 1NR6), a strong XB is formed to the water molecules (Cl−O, d = 2.8 Å, θ = 158.1°), which in turn forms a hydrogen bond to the carboxy group of Glu297 (2.9 Å) (Figure 78b). A similar water bridge is observed in the complex of diclofenac with lactoferrin, in which the water molecule was bridged between the ligand chlorine atom and the carbonyl oxygen of Val591 (Figure 78c).317 As has been demonstrated in a nonbiological context,51a the strength of a halogen bond can be tuned by varying the

with the Tyr228 phenol aromatic ring, and demonstrated that this type of interaction could stabilize ligand−protein complexes by up to 10 kJ mol−1 (Figure 76, PDB entry 2BQW).27

Figure 76. Cl···π halogen bonding contributed to the stability of a chlorobenzene-derived inhibitor complex with factor Xa (PDB entry 2BQW).311

Multiple XB interactions between a ligand and the protein backbone have been observed, which utilized both X···O and X···π interactions, such as in the complex of halothane (a general anesthetic) within ferritin (a protein involved in iron storage).312 Halothane simultaneously forms two halogen bonds within the protein binding site: X···OC (Leu24, d = 3.10 Å, θ = 145°) and X···π (Tyr28, dcentroid = 4.1 Å, α = 155°). The presence of multiple halogen bonds stabilizing protein− ligand complexes has also been reported by Knapp, Bracker, and co-workers in the complex of a 6,7-dichloro-substituted indole ligand with the carbonyl oxygens of the hinge region of the CDC-like kinase 3 (CLK3),313 and in the complex of dichloro-substituted carbolines with PIM1 kinase (PIM = proviral integration site in moloney murine leukemia virus).314 Baumil and co-workers have demonstrated that multiple Cl··· O halogen bonds are responsible for the selectivity of DRB (5,6-dichlorobenzimidazone-1-β-D-ribofuranoside) toward cy-

Figure 77. DRB binding to Cdk9 and Cdk2 kinases. Dashed lines denote XB (red) and HB (blue). (a) Crystal structure of DRB bound to Cdk9. (b) Crystal structure of DRB bound to Cdk2. 7160

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 78. XWH bridges. Dashed lines denote XB (red) and HB (blue). (a) Schematic representation of an XWH: X = XB donor, Y = HB acceptor, Z = HB donor. (b) JNK complexed with a chlorobenzene-derived inhibitor.318 (c) Diclofenac−cytochrome P450 complex.319

Figure 79. Tuning the halogen-bond strength in aldose reductase inhibitors. Left: inhibitors with the corresponding IC50 values. Right: crystal structure of the aldose reductase−NADP-2 complex. The Br···O halogen bond is depicted as a red dashed line.

Figure 80. HB−XB competition in a DNA junction. Rreprinted from ref 321. Copyright 2013 American Chemical Society.

both ﬂuorine substitution and bromine-to-iodine exchange result in an increase in the XB interaction energy (Figure 79). The general trend of increasing XB strength correlates with the observed IC50 values, with changes of up to an order of magnitude observed. Deviations from this trend can be explained by the changes in the solvation/desolvation contributions to the binding aﬃnity.

electron-withdrawing substituents on the halogen-bond donor scaﬀold. Hobza and co-workers have recently systematically studied the binding of bromo- and iodobenzene-derived inhibitors of aldose reductase.320 The ligands bind within the active site of the protein and form an XB between the bromine or iodine atom of the ligand and the backbone carbonyl oxygen of the Thr113 arene core, which modulates the strength of ligand binding, as determined through the measured IC50 values (the concentration of ligand corresponding to half-maximum enzyme inhibition activity). High-resolution crystal structures of the ligands bound within the protein binding pocket demonstrate that all seven ligands bind in the same manner, and that the structural changes upon ﬂuorine substitution are minimal, while computational techniques demonstrate that

5.2. Measurement of Biological Halogen Bonds: Assessing the Contribution to Ligand Binding Aﬃnities

5.2.1. Direct Measurement. There are few ways to directly measure the energies of XBs within biological systems, and only one example has been reported to date. Ho and coworkers have directly correlated the strength of a halogen bond to that of a hydrogen bond within single crystals of a DNA 7161

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

four-stranded Holliday junction.25e,26b The authors demonstrated that such a junction, which can isomerize between two structurally similar and virtually isoenergetic conformers, can act as an assay for the presence of a hydrogen or halogen bond at the junction crossover, by the preference for one of the two isomeric forms. The brominated junction can adopt two conformations (Figure 80): the ﬁrst is stabilized by a cytosine− phosphate hydrogen bond (H isomer), and the second is stabilized by a halogen bond from a brominated uracil (BrUra) which replaces the cytosine residue (the X isomer) and allows the isomeric form of the junction to be identiﬁed by the location of the bromine atom. The isomorphous crystallization of three DNA constructs, in which the stabilizing XB from the bromine competes with hydrogen bonds in both 1:1 and 2:1 XB:HB ratios, allowed the estimation of the relative energy diﬀerences between HBs and XBs in the system by means of the distribution of X and H isomers in the crystal. This revealed that the XB interaction is 2−5 kcal mol−1 stronger than the analogous HB in this particular DNA junction environment. This result was later conﬁrmed by diﬀerential scanning calorimetry used to measure the thermodynamics of melting of the same DNA junction in solution.25e Recently, this work was extended to determining the halogen-bond strengths in the same DNA junction for F, Cl, Br, and I using the same crystallographic assay and calorimetry techniques.321 The authors demonstrated that the trend in XB bond enthalpies followed the predicted trend of F < Cl < Br < I, but entropic contributions resulted in the strongest halogen-bond free energy being observed in the case of X = Br. 5.2.2. Indirect Measurement of the Biological Halogen-Bond Strength. The contribution of halogen bonds to ligand association within a protein binding pocket can be deduced by comparing the binding aﬃnities of halogenated ligands and the analogous nonhalogenated ligands. To date there have only been a handful of systematic studies into the eﬀect of halogen bonding on ligand binding aﬃnities; however, they provide a useful insight into the utility of XB for enhancing drug binding aﬃnity and eﬃcacy. Matthews and co-workers studied the binding modes of a series of halogenated benzene ligands, of types C6F5X (X = H, F, Cl, Br, I) and C6H5X (X = H, I), within the apolar internal cavity formed in bacteriophage T4 lysozyme following the mutation of Leu99 to alanine.322 All the ligand derivatives bind in the same plane, but are shifted by translation along the plane to diﬀerent extents, depending on the size and nature of the halogen substituents. For the ligand C6F5I, the crystal structure revealed the presence of an XB formed between the iodine atom and the sulfur atom of the Met102 side chain (d = 3.0 Å, θ = 166°, Figure 81a). C6H5I also forms a halogen bond to the Met102 sulfur atom (PDB entry 3DN4), albeit longer (d = 3.3 Å), correlating with the reduced halogen-bonding ability relative to the more electron-deﬁcient ﬂuorinated analogue (Figure 81b). The ligand also occupies an alternate orientation in the crystal structure, rotated about 180° to form an XB to the carbonyl oxygen of Leu84 (d = 3.4 Å, θ = 152°). The association constants of benzene and iodobenzene (C6H5I) were determined using isothermal titration calorimetry to be 17 100 and 51 800 M−1, corresponding to a binding enhancement of around 0.5 kcal mol−1. This diﬀerence is likely to be, in part, due to the formation of the halogen bond. It should be noted that the diﬀerence was not related in this case to the relative strength of HB over XB, but rather to that of XB over the van der Waals contacts it replaces.

Figure 81. Halogenated benzene ligands bound within a nonpolar cavity of a T4 lysozyme mutant: (a) C6F5I forming a halogen bond with the S atom of Met102 (PDB entry 3DN3) and (b) two alternate halogen-bonding conformations of C6H5I (PDB entry 3DN4).

Diederich and co-workers investigated the eﬀect of XB on a series of novel inhibitors of human cathepsin L (hCatL), using a combination of crystal structure data, computational methods, and structure−activity relationships (SARs).323 The inhibitors in the study bind to the enzyme hCatL through a covalent bond, formed by reaction of the thiol of Cys25 with the ligand’s nitrile moiety to form a thioimidate, in addition to hydrogen bonds and lipophilic contacts. The discovery of an XB between the chlorine atom of the chlorinated ligand and the backbone O atom of Gly61 led the authors to prepare a series of analogues, in which the aryl ring of the parent compound (where X = H) is substituted with Me, CF3, F, Cl, Br, and I (blue, Figure 82) to probe the eﬀect of the halogen bond on the binding aﬃnity and the IC50 values, which were determined for the series of ligands using a ﬂuorescence assay.

Figure 82. Structure of hCatL inhibitors. IC50 values refer to the inhibitor concentration required for half-maximum activity.

The chloro-substituted ligand displays a 6-fold reduction in IC50, corresponding to an increase in binding aﬃnity on the order of 1 kcal mol−1, whereas substitution with CH3 (which is similar in size to Cl) or CF3 results in an increase of IC50 by a factor of 2. Substitution with F (which does not generally form halogen bonds) resulted in a slight increase, while Br or I results in a decrease in IC50 by a factor of 2 or 4, respectively. The activity of this class of inhibitors is thus correlated in some way with the strength of the XB donor ability of the substituent halogen atom, with the binding aﬃnity following the expected trend F < H ≪ Cl < Br < I. The same trend was also observed for three further ligand series, in which R1 and R2 were modiﬁed with ﬂuorinated substituents. The X-ray crystal structures of the ligands complexed with hCatL provide further evidence for the XB interaction between the ligand and enzyme binding site. Close contacts between the halogen atom and the 7162

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

5.3. Computational Approaches to Halogen Bond Structure: Energy Relationships in Biological Environments

carbonyl oxygen of Gly61, less than the sum of the respective vdW radii, indicate the presence of the halogen bond. The chlorine-substituted compound (Figure 83) shows a near-

5.3.1.1. Hybrid QM−MM Computational Approaches. The most accurate means of modeling halogen bonds computationally is by using high-level QM calculations on small, model systems.43c,305,324 However, for large macromolecules of biological relevance, such calculations are not feasible due to the large molecular size prohibiting detailed electronic structure calculations with a reasonable computational cost. The most widely used strategy to study halogen bonds in biological systems is a hybrid approach, combining both QM and MM.26a,303,325 The active site, involving relatively few atoms, can be modeled using accurate QM calculations, while the remaining parts can be calculated using the computationally less demanding MM methodology. Recently, Ho and co-workers used QM calculations to probe the halogen bonds formed in the DNA junction systems discussed earlier.25e Using the crystal structure to construct the initial molecular structure, the authors were able then to derive a set of potential energy functions, ab initio, to model the biological halogen bonds taking into account both directionality and bonding.326 5.3.1.2. Low Computational Cost Methods: Application to High-Throughput Drug Screening. Drug development costs by major pharma run into the region of $4−10 billion, with successful drugs often taking over 10 years to reach clinical application. The past decade saw an increase in attrition rates, with failure at phase III trials being dominated by a lack of eﬃcacy.327 New approaches for describing and predicting the binding of drug molecules within protein active sites are, therefore, of high importance. A large proportion of molecules in libraries used for high-throughput assays and available approved drug molecules are halogenated, to improve oral absorption, ﬁll hydrophobic cavities in the binding site, facilitate blood−brain barrier crossing, and prolong the drug lifetime. Computational docking studies oﬀer an attractive way of screening potential drug molecules, without having to synthesize numerous analogues, provided knowledge of the target enzyme’s active site structure is known (e.g., crystal structure data). MM computational techniques are commonly used for such screening processes due to the low computational cost compared with that of QM or QM−MM methods. Computational techniques used in high-throughput computational docking studies need to be able to account for the

Figure 83. Structure of the chlorinated inhibitor (Figure 82, where X = Cl) complexed with hCatL (PDB entry 2XU1). The halogen bond (red dashed line) is formed between the ligand chlorine atom and the carbonyl oxygen of the backbone Gly61 residue.

optimum XB interaction (dO···Cl = 3.0 Å, θ = 174°). Across all three halogenated ligands, the halogen-bond distances are signiﬁcantly shorter than the sum of vdW radii: dO···Br = 3.1 Å, θ = 176°; dO···I = 3.1 Å, θ = 175°. Puckering of the pyrrolidine ring yields enough conformational ﬂexibility to optimize the XB geometry for halo-substituted structures at minimum energetic cost. Computational modeling of halobenzene−N-methylacetamide XB interactions derived CO···X bond lengths consistent with those found in the crystal structures and supports the case for the stabilizing halogen bond to the carbonyl oxygen atom. In hCatL, the halogen bond was formed to a carbonyl oxygen buried in a relatively polar pocket in the protein binding site. The authors also explored XB within an apolar cavity in MEK1 kinase, and demonstrated that a similar trend of decreasing IC50 values was observed when substituting F through I.323a The presence of the XB was conﬁrmed in the crystal structure of the iodo-substituted ligand complexed with MEK1 (Figure 84), with the halogen bond formed between the iodine atom and the carbonyl oxygen of Val127. Binding of the ligand was otherwise dominated by vdW contacts, with no other polar contacts observed in the hydrophobic binding pocket.

Figure 84. Inhibitors of MEK1 kinase (left) and the crystal structure of the MEK1−iodo-substituted ligand complex (right), highlighting the halogen bond (red) formed between the iodine atom and the backbone carbonyl oxygen of Val127 (PDB entry 2YJ8). 7163

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 85. Halogen-bonding HIV reverse transcriptase inhibitors (PDB entry 2BE2) (a,b) . An I···O XB was observed in the crystal structure of the complex with the inhibitor in (a) (red dashed line) (right).

Figure 86. Inhibitor binding to hepatitis C virus NS3-NS4A protease. The brominated inhibitor (green) exhibited a Br···O halogen bond with Asp79 (3.0 Å, red dashed line) and induced a protein conformational change compared to the nonbrominated analogue (X = H blue).

workers implemented this PEP methodology in a molecular docking program for the ﬁrst time329 and, by placing the positive charge closer to the halogen atom center,124b were able to derive XB lengths for four pharmaceutically relevant protein−halogenated ligand complexes that closely resembled the obtained crystal structure data.

directional and attractive bonding nature of halogen bonds, formed between the docked ligand and the active site, to accurately predict the binding aﬃnity and thus eﬃcacy of potential drug molecules. Common programs, such as AMBER, OPLS-AA, CHARMM, and GROMOS, use a single partial charge on each atom to represent electrostatic interactions. In the case of halogen atoms, a partial negative charge is used to account for the observed dipole moments, and thus, they exhibit a purely repulsive interaction with heteroatoms bearing electron lone pairs, such as backbone carbonyl oxygens or side chain oxygen or sulfur atoms. Such MM methods need therefore to be adapted to account for the presence of attractive XB interactions. Recently, various approaches toward adapting MM methods to account for XB interactions have been reported, such as in the OPLS-AA126 or AMBER116,121,328 force ﬁelds. Placing a massless pseudoatom, bearing a partial positive charge and possessing no van der Waals energy (a so-called positive extra point (PEP)), has been shown to improve the modeling of XB interaction geometries and interaction energies in the AMBER force ﬁeld (vide supra).116,121,328 Ibrahim demonstrated that this PEP approach, in which the point charge is placed at the halogen atom surface, calculated XB energies that correlated well with the observed binding aﬃnities of halogenated benzimidazole ligand inhibitors of cyclin-dependent protein kinase 2 (CDK2 kinase).116 Recently, Bronowska and co-

5.4. Halogen Bonding in Medicinal Chemistry: Success Stories and Application to Systematic Drug Discovery

Analysis of the PDB reveals hundreds of examples of XB within protein−ligand complexes, and in many cases the presence of the halogen bonds resulted in higher binding aﬃnity and activity of the ligand toward the protein’s active site. In the following section we review some selected recent success stories, in which XB has been used to improve ligand activity toward a particular target protein, and discuss the potential beneﬁts and challenges of utilizing XB routinely in the design of new drug molecules. Arnold and co-workers have demonstrated that halogenation of HIV reverse transcriptase inhibitors (a key enzyme used by HIV to generate new viral particles) resulted in a potent inhibitor (Figure 85a), stabilized by an XB between the iodine atom and the carbonyl oxygen of Tyr188.330 Jorgenson and coworkers have used free energy perturbation calculations to guide the optimization of inhibitors for the same enzyme, which resulted in the identiﬁcation of a 55 pM inhibitor (Figure 85b) and in which a Cl···O XB is implicated in the enhanced binding 7164

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 87. Left: halogenated ligands targeting MDM2. Right: crystal structure of the iodo-substituted ligand within the MDM2 active site, with XB denoted by a red dashed line (PDB entry 1T4E).

Figure 88. Left: halogenated PDE5 inhibitors. Top right: crystal structure of PDE5 with the bromo-substituted ligand, with the Br···O halogen bond depicted by a red dashed line (PDB entry 3SIE). Bottom right: −(log IC50) versus calculated halogen-bond energies.54

carboxamide inhibitor of integrin-α4β1.333 The IC50 data follow the expected trend of F > Cl ≈ H ≈ CH3 > Br > I, which reﬂects the predicted halogen-bond strength. The same trend was also observed in a series of halogen-substituted 1,4benzodiazepine-2,5-dione ligands, designed as anticancer drugs for targeting the p53 pathway, that bind to the p53 binding site of human murine double mutant 2 (MDM2). The crystal structure of the iodo-substituted ligand (Figure 87, X = I) bound to MDM2 (PDB entry 1T4E) reveals the presence of a halogen bond between the iodine atom and the backbone oxygen of Gly72.334 Replacement of the iodine atom with bromine or chlorine resulted in a decrease in binding aﬃnity and 2-fold and 4-fold increases in IC50, respectively.335 Zhu and co-workers have used XB in ligand optimization, designing a series of potent halogenated inhibitors of phosphodiesterase type 5 (PDE5).336 Using a combined QM−MM computational modeling approach, the strength of the XB formed between the ligand halogen atom and the Tyr612 phenol oxygen was calculated and compared to the observed IC50 values (Figure 88). A strong correlation was observed (a 7-fold decrease in IC50 with the iodo ligand compared to the protic ligand), which demonstrated the crucial

and potency compared with those of the nonhalogenated analogues.331 Lemke et al. have developed a potent inhibitor (Figure 86, X = Br) of the hepatitis C virus NS3-NS4A protease that, at the time of writing, is undergoing phase III clinical trials for the treatment of the virus.332 The crystal structure of the ligand bound within NS3-NS4A protease shows the presence of a Br··· O XB formed between the ligand and the backbone carbonyl oxygen of Asp79 (d = 3.0 Å, θ = 169°, Figure 86). Furthermore, both kinetic studies and SARs support the role of the XB in improving the eﬃcacy of the inhibitor molecule; the association constant increased 3.6-fold relative to that of the protic analogue (Figure 86, X = H). The bromine atom was also shown to improve drug absorption, distribution, metabolism, and excretion (ADME) properties, which highlights the additional advantages of incorporating halogen-bond donor elements within drug molecules. The structure of the protic analogue indicates that it binds in a similar manner, but caused a small change in the protein conformation (Figure 86). Halogenation has also been shown to increase the eﬃcacy of a ligand inhibitor through formation of a halogen bond by Carpenter et al., who reported the halogenated benzimidazole 7165

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 89. Left: halogenated herbicides targeting the IspD enzyme and IC50 data for a series of halogenated ligands. Right: crystal structure of the tribromo ligand (R1 = R2 = R3 = Br) bound to IspD. The halogen bond (red dashed line) and coordination to Cd2+ (yellow sphere, blue dashed lines) are depicted.

Figure 90. (a−d) Halogenated leads for targeting mutant p53. (e) Crystal structure of the optimized inhibitor in (d) binding within the enzyme using XB (red dashed line).

to the backbone oxygen of Val239 (d = 3.25 Å, θ = 164°). Furthermore, the crystal structure of a chlorinated inhibitor (R1 = R2 = Cl; R3 = Br) bound in IspD (PDB entry 4NAL) reveals the presence of a Cl···O halogen bond (d = 3.26 Å, θ = 164°). Halogenation of the ligands results in an improvement of the inhibitory eﬀect compared with that of nonhalogenated analogues, which are unable to form a stabilizing XB, and demonstrates the crucial role of the halogen bond in the binding of these ligands. 5.4.2. Practical Implications for Drug Discovery. Over the past few years there has been an increased awareness toward the use of XB as an intermolecular force in the medicinal chemists’ toolkit. Analysis of protein−halogenated ligand crystal structures has revealed that XBs are far more widespread than ﬁrst thought and play an important role in stabilizing such complexes, leading to enhanced drug activity. Recent systematic studies of halogen bonding in medicinal chemistry applications have demonstrated the utility of the intermolecular XB interactions as a means of enhancing drug potency. However, a number of practical issues remain before the use of XB in drug discovery, particularly systematic drug design, becomes widespread. While the heavier halides (Br and I) form stronger XBs than Cl, they represent a small fraction of halogenated drug molecules. This may be a consequence, in part, of the perceived reactivity of C−Br and C−I bonds, which are typically only used in intermediates in preparation for metal-catalyzed cross-coupling reactions. Furthermore, incor-

role of the XB in enhancing the ligand binding aﬃnity and eﬃcacy. The presence of the XB was conﬁrmed by analysis of the crystal structure of PDE5 with the ﬂuoro, chloro, and bromo ligands (Figure 88). Very recently, the authors undertook a more comprehensive study of the strength of the XB interaction in the same series of halogenated ligands.337 Using a combination of thermodynamic and X-ray crystallographic data, and QM calculations, the contribution to the ligand binding aﬃnity from the X···O halogen bond was shown to be −1.6, −3.1, and −5.6 kJ mol−1 for X = Cl, Br, and I, respectively. 5.4.1. Halogenated Herbicides. Very recently, Diederich and co-workers reported a series of halogenated inhibitors for the IspD enzyme, which forms part of the nonmevalonate pathway for isoprenoid biosynthesis and represents a target for novel approaches toward developing herbicide and antiinfectious disease agents.338 Binding of the highly halogenated pseudilin inhibitors, a series of marine natural products, occurs through an allosteric binding mode using both XB and metal coordination. Binding of the halogenated derivatives is enhanced in the assays by the presence of Cd2+ cations. The crystal structure of the tribromo ligand (R1 = R2 = R3 = Br) with IspD (PDB entry 4NAK, Figure 89) demonstrated that the Cd2+ cation is bound in a tetrahedral coordination by the phenolic OH and pyrrole NH of the ligand, and also to both a Gln238 backbone oxygen and a water molecule. The bromine atom (R1) forms a halogen bond 7166

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 91. Matile’s halogen-bonding anion transporters.343,344

identiﬁed the compounds exhibiting the highest aﬃnities (KD values of 104 and 87 μM for the ligands shown in parts b and c, respectively, of Figure 90). The crystal structures of the p53 cancer mutant Y220C complex with the piperidine fragments reveals a stabilizing halogen bond between the ligand iodine atom and the backbone carbonyl oxygen of Leu145. Further optimization reveals the alkyne-functionalized compound (Figure 90d) is the most potent inhibitor (KD = 9.7 μM), and the determination of the crystal structure of the Y220C− ligand complex conﬁrms the presence of the I···O halogen bond (d = 3.0 Å, θ = 173°, Figure 90e). This work demonstrates the clear potential of increasing the number of halogenated molecules within fragment-based screening libraries as a means of identifying new and innovative drug molecules that utilize XB.

porating heavier halides inevitably increases the molecular weight of the compound, which is regarded as being a critical parameter for aﬀecting the drug’s ADME properties. Medicinal chemists typically use a set of guidelines that predict whether a potential pharmaceutical molecule will possess suitable ADME properties when designing druglike molecules. The ﬁrst of these is Lipinski’s rule of ﬁve,339 which is based on the observation that pharmaceutically active molecules are small and lipophilic, and states that poor absorption of a drug is likely when there are more than ﬁve HB donors and 10 HB acceptors, MW > 500, and the calculated log P > 5 (where P is the partition coeﬃcient between octanol and water). The addition of heavier halogen atoms adds signiﬁcant weight to a molecule, but the eﬀect on size, volume, and log P is not as drastic. More systematic studies are needed to determine the relationship between halogen-bond donors and ADME properties as it is clear that commonly used models, such as the rule of ﬁve, do not account well for the beneﬁcial role of XB in drug discovery. The eﬀect of including halogen-bond donors on metabolism and excretion properties also requires further systematic study; indeed, XB has been implicated in the mechanisms of enzymatic dehalogenations and catabolism.300,340 As a consequence of these concerns, the number of molecules containing the heavier halides in standard highthroughput libraries used for fragment-based screening is very low. However, Boeckler and co-workers have created a halogenenriched fragment library, employing an upper limit on the number of heavy atoms ( IBr and I2 > Br2. To complement this work on interaction energies, Erdelyi and coworkers have conducted an investigation into the geometries of XB in solution.349 Employing the technique of isotopic perturbation of equilibrium, they demonstrated that systems comprised of an electropositive bromine or iodine bound between two nitrogen bases, [N−X−N]+, prefer a static, symmetric arrangement in CD2Cl2 solutions (Figure 92).

Figure 92. A bispyridine halide complex used to determine the symmetric geometry of the [N−X−N]+ XB bond in dichloromethane.349

6.1.1. Inorganic Halogen-Bond Donors. The observance of charge-transfer bands in the UV spectra of the halogenbonded complex formed between molecular diiodine and a neutral Lewis base prompted the measurement of numerous thermodynamic equilibrium constants for their association.7b At around the same time, Larsen and Allred also studied these complexes using NMR spectroscopy.350 However, only recently have these data been used by Laurence et al. to construct an I2 basicity scale: pKBI2 = log Ka.351 It was found that Lewis bases containing N, P, Se, and S donor atoms formed stronger interactions with I2 than those with I, O, Br, Cl, and F donor atoms, a consequence of the decrease in basicity across a period and the increase down a group of the periodic table. Among other trends in halogen-bond acceptors distinguished by this scale was that of amine > pyridine > nitrile, demonstrating that the decreasing p-character of the Lewis basic lone pair reduces the strength of the XB interaction. The construction of LFERs of pKBI2 vs Hammett substituent constants (σ) enabled the I2 basicity of a range of substituted Lewis basic halogen-bond acceptors to be determined. The authors also included a comparison of pKBI2 with other Lewis basicity scales. It showed a poor correlation with scales based upon metal cations (Mn+,

Table 2. Association Constants (Ka) for the 1:1 Stoichiometric Complexes Formed between Tetramethylurea and HB Donor 4-(Phenylazo)phenol or the XB Donor Diiodine in a Variety of Solvents353 solvent n-octane carbon tetrachloride chloroform nitromethane methanol

Ka of the HB complex (M−1)

Ka of the XB complex (M−1)

2400 410

12 6

52 5 CF3CF3CF2I), a consequence of the greater electron-accepting ability of the XB donor halogen atom. This method was also used to rank electron donor moieties relative to the nondonor solvent n-pentane with the same halogen-bond donor 1,4diiodotetraﬂuorobenzene (Table 3). Table 3. Selected 19F NMR Chemical Shift Diﬀerences (Δδ−CF−)a of 1,4-Diiodotetraﬂuorobenzene in n-Pentane and the Given Solvent359

Figure 95. One-to-one stoichiometric XB complex formed between quinuclidine and para-substituted (iodoethynyl)benzene derivatives in benzene solution.357

a

solvent

Δδ−CF−

piperidine tetrahydrothiophene furan

3.60 1.81 0.99

Δδ = δn‑pentane − δused solvent. δIC6F4I in n-pentane was −119.44 pm.

Thermodynamic studies of the XB interactions between C6F5I and group 10 metal ﬂuoride complexes in toluene have been performed by Perutz and Brammer (Figure 98).23d 19F

free energies of binding for 1:1 stoichiometric complexes between quinuclidine and halogen-bond donors containing electron-donating Me 2 N and electron-withdrawing NO 2 substituents in C6D6 to be 4.6 and 7.9 kJ mol−1 respectively. More detailed thermodynamic proﬁles of the association showed that the enthalpic gain from halogen-bond formation was partially oﬀset by an unfavorable entropic term. Following their work studying the XB complexes formed between diiodine and Lewis bases in solution, Larsen and Allred examined the complexation of the organic halogen-bond donor triﬂuoroiodomethane with 2,4,6-trimethylpyridine in cyclopentane by 19F NMR spectroscopy (Figure 96).358 Variable-temperature studies determined enthalpic and entropic interaction energies of −21 kJ mol−1 and −53 J mol−1 K−1, respectively. More recently, Herrebout, Moiana, and coworkers determined thermodynamic data for the 1:1 XB

Figure 98. Examples of neutral XB formation to organometallic species in toluene solution.23d,360

variable-temperature NMR titration experiments were used to calculate the enthalpies and entropies of these interactions in toluene: values of ΔH followed the trend Ni < Pd < Pt down the group. These data were found to be in excellent agreement with DFT calculations of the enthalpy of adduct formation. More recently, the same authors have studied halogen-bond formation with another series of organometallic compounds, bis(η5-cyclopentadienyl)metal hydrides.360 The enthalpies and entropies of 1:1 complexation with C6F5I or iodoperﬂuorohex-

Figure 96. XB complex between 2,4,6-trimethylpyridine and triﬂuoroiodomethane in cyclopentane.358 7170

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Figure 99. XB complexation between perﬂuorohexyl iodide and various Lewis bases in CHCl3 solution.361

ane were determined by 1H NMR titrations in toluene and found to be stronger than those with the hydrogen-bond donor indole, which suggested a closer analogy of XB to dihydrogen bonding than to HB within these types of systems (Figure 98). Cabot and Hunter performed quantitative studies of the 1:1 complexes formed between perﬂuorohexyl iodide and tributylphosphine oxide or a variety of Lewis basic amines (Figure 99).361 Utilizing 19F NMR titration data, the experimentally calculated free energies of XB in CCl4, CHCl3, and benzene showed a good correlation with Hunter’s simple electrostatic model of noncovalent interactions, which draws parallels between XB and HB in solution (Table 4). However, the

which electrostatic and charge-transfer interactions inﬂuence XB, and aﬃrmed the ﬁndings of Hunter and Taylor that electrostatics are particularly important for organic donors. This was supported by further work by Laurence, who observed that LFERs between pKBI2 and log Ka for organic donors (ICN and CF3I) were only observed among the same families of bases (hard or soft).351 Since these XB donors have a stronger electrostatic character than the inorganic donor I2, the family independence could only be restored by addition of an electrostatic parameter to the regression. These conclusions were also reinforced by Taylor’s thermodynamic investigations into the XB interactions between inorganic (I2) and organic (C8F17I, C6F5I) donors and a variety of Lewis bases in alkane solvents.29 Correlations between experimental data and the Hunter electrostatic model predictions were found to be poor, with the association constants calculated for the soft, inorganic donor I2 showing the weakest correlation. These were found to be too low as a result of the neglected charge-transfer contributions to these XB interactions, and instead, DFT exchange-correlation functionals provided excellent linear correlations with the experimental thermodynamic data in both the solution and gas phases. These ﬁndings have more recently been supported by the work of Hunter, Brammer, and Perutz, who found that the stability of XB complexes between I2 and neutral Lewis bases were relatively insensitive to the nature of the solvent. Such observations suggest there is a signiﬁcant charge-transfer contribution to halogen-bond-driven complexation.353 6.1.4. Weaker XB Interactions in Solution. Rebek and co-workers have employed encapsulation techniques to directly observe weak XB interactions by NMR spectroscopy that would otherwise be negligible in bulk solvent.364 The formation of XB between perﬂuorohalocarbons and Lewis bases, such as nitrogen-containing heterocycles or CO groups of amides or lactones, may be detected within a capsule of aromatic cavitand “panels” held together by HB that provide a steric barrier to interference from the solvent (mesitylene) molecules. The capsule has the additional beneﬁt of creating a cleft that aligns guests in a favorable arrangement for XB formation. This resulted in selective halogen-bond formation with only pyridine and γ-picoline because the restricted space of the capsule prevented α- and β-picolines from adopting the correct positions for XB with the Lewis base (Figure 100). In a similar fashion, XB has been observed in aqueous solution upon encapsulation of I2 within a cucurbit[6]uril unit partially driven by halogen-bond formation to the ureido CO groups.104 Pascal has used 15N NMR spectroscopy to investigate the weak N···X halogen bond formed between a tridentate XB

Table 4. Experimental and Predicted Association Constants for XB Complexation between Perﬂuorohexyl Iodide and Various Lewis Bases in CHCl3 Solution361 Lewis base

exptl log Ka

predicted log Ka

Bu3PO Et3N quinuclidine

I−) was observed. This ﬁnal trend is consistent with a signiﬁcant electrostatic contribution to the XB. It is noteworthy that Beer and coworkers have achieved a reversal of this XB halide binding trend through the use of interlocked host systems (vide infra) to overcome the intrinsic preferences of these halogen-bond donor groups for more basic halides. Further analysis of the temperature dependence of Ka values between C8F17I and C6F5I with Cl− revealed that anion association was both enthalpically and entropically favorable (Table 6). Huber and co-workers have conducted isothermal calorimetric titrations to investigate the thermodynamic contributions to the charge-assisted XB formation between halides and cationic bis(haloimidazolium) receptors.383 Their data showed that the nature of the weakly coordinating counteranion had little eﬀect on the interaction strength, and although all systems exhibited a strong solvent dependence on XB formation, the authors commented that this was hard to interpret at this point in time. A marked entropic contribution toward the overall free energy of binding by their cationic receptors to halide anions was also reported, which is consistent with the results reported by Taylor (Table 6). In 2005 Resnati, Metrangolo, and co-workers reported the ﬁrst example of a receptor incorporating halogen-bond donor

Figure 104. An anion-templated XB catenane 10·Br·PF6.376

6.3. Anion Recognition and Sensing

Given the attributes of XB are comparative to those of HB in terms of bond strength and directionality, it is a natural progression to employ XB in an analogous manner for molecular recognition investigations in solution. This section begins with an overview of the thermodynamics of anion complexation in solution and then goes on to summarize examples where this interaction has been utilized in the ﬁeld of anion recognition and sensing. The inherent negative charge and nucleophilicity of anionic species make them strong candidates as XB acceptors. In 1997, thermodynamic investigations of the association of neutral I2 with I− to form I3− revealed the XB interaction between the two species to be as strong as 126 kJ mol−1 in the gas phase379 and 17 kJ mol−1 in water,380 similar to values for HB.381 Since then, however, only a limited number of quantitative analyses of the interactions between anions and halogen-bond donors have been carried out. Taylor and co-workers382 have extended the work of Kochi204d on the energetics of XB interactions between halide anions and CBr4 to include anion binding to iodoperﬂuoroalkanes and iodoperﬂuoroarenes in the moderately polar solvents acetone and dichloromethane (Table 5). Their data 7174

DOI: 10.1021/cr500674c Chem. Rev. 2015, 115, 7118−7195

Chemical Reviews

Review

Scheme 4. Synthesis of Catenane 13·BF4, Templated through a Single Charge-Assisted Halogen Bond377

Scheme 5. First Anion-Templated All-XB Rotaxane 16·Cl·BF4378

Table 5. Association Constants (Ka) for XB Formation between Organic Donors and Anions as Their Tetrabutylammonium (TBA) Salts204d,382 +

Table 6. Enthalpies (ΔH) and Entropies (ΔS) of Interaction between TBA·Cl and C6F5I or C6F17I in Acetone382 a XB complex Cl−·TBA+

−1

XB complex·TBA

solvent

Ka (M )

CBr4···Cl− CBr4···Br− CBr4···I− C6F5I···Cl− C6F5I···Br− C6F5I···I− C6F5I···TsO− C6F5I···NO3− C6F5I···HSO4− C6F17I···Cl− C6F17I···Br− C6F17I···I−

CH2Cl2 CH2Cl2 CH2Cl2 acetone acetone acetone acetone acetone acetone acetone acetone acetone

3.0 2.8 3.2 1.5 1.0 44 Br− > I−) over oxoanions (NO3−, HSO4−, TsO−). These same trends in selectivity were also found for Beer’s XB iodotriazolefunctionalized zinc(II) metalloporphyrin receptor (Figure 107).385

Figure 107. Beer’s XB iodotriazole-functionalized zinc(II) metalloporphyrin receptor.385

Beer and co-workers have incorporated XB donor groups into macrocyclic and interlocked host molecular frameworks. The positively charged bidentate bromoimidazoliophane receptor 23·(PF6)2, was observed to bind bromide selectively (Ka = 889 M−1) in the highly competitive aqueous solvent mixture of 9:1 CD3OD/D2O, as determined by 1H NMR spectroscopy (Figure 108, Table 8).386 The eﬀect of the polarized bromine atom substituents in the macrocyclic synisomer is 2-fold. First, they act as strong XB donors, which leads to stronger binding of halides compared with that of the protic analogue 24·(PF6)2 (Ka = 130 M−1) in the same aqueous solvent mixture. Second, the bromine atoms are sterically bulky and create a preorganized host system, which prevents conformational inversion to the anti-conformation and which is demonstrated to display weaker anion binding. This is

Table 7. Association Constants (Ka) for Receptors 19−22 with Anions (TBA Salts) in Acetone at 295 K384 a receptor 19 20 21 21 21 a

anion −

Cl Cl− Cl− Br− I−

Ka (M−1) 70 1800 19000 3800 760

receptor 21 21 21 22

anion −

TsO HSO4− NO3− Cl−

Ka (M−1) 10

Halogen Bonding in Supramolecular Chemistry - Chemical Reviews

Recommend Documents