Intermolecular interactions in molecular crystals: what ’ s in a name?

Structure – property relationships are the key to modern crystal engineering, and for molecular crystals this requires both a thorough understanding of intermolecular interactions, and the subsequent use of this to create solids with desired properties. There has been a rapid increase in publications aimed at furthering this understanding, especially the importance of non-canonical interactions such as halogen, chalcogen, pnicogen, and tetrel bonds. Here we show how all of these interactions – and hydrogen bonds – can be readily understood through their common origin in the redistribution of electron density that results from chemical bonding. This redistribution is directly linked to the molecular electrostatic potential, to qualitative concepts such as electrostatic complementarity, and to the calculation of quantitative intermolecular interaction energies. Visualization of these energies, along with their electrostatic and dispersion components, sheds light on the architecture of molecular crystals, in turn providing a link to actual crystal properties.


Introduction
The detailed analysis of the interactions between molecules and ions in crystals plays an increasingly important role in modern solid state chemistry, and in particular crystal engineering, where the derivation of predictive structure-property relationships is key to a genuine "engineering" of crystals. This, of course, was articulated "In many cases, with the exception of hydrogen bonding, molecular pairings responsible for the largest part of the interaction energy in a crystal show no particular atomatom feature, no easily identiable "bond", not even aromatic stacks, or the like; they stick together by compatibility of minor and diffuse features in the electrostatic potential, that defy recognition and, a fortiori, classication. Only a quantitative calculation of cohesion energies can reveal true crystal structure determinants". School of Molecular Sciences, University of Western Australia, Perth, WA, Australia. E-mail: mark.spackman@ uwa.edu.au some time ago by Desiraju who described the crystal engineering of molecular solids as the "understanding of intermolecular interactions in the context of crystal packing and the utilization of such understanding in the design of new solids with desired physical and chemical properties". 1 Utilization and design require understanding as an essential precursor, and the context of crystal packing in this statement is especially important. Recent years have seen an explosion of publications focusing on the description, characteristics and relative importance of different non-canonical interactions such as lithium, 2 beryllium, 3 triel, 4-6 tetrel/carbon, 7-10 pnicogen/pnictogen, [11][12][13][14] chalcogen, [15][16][17][18][19] halogen 20-28 and aerogen 29,30 "bonds". (The list of "bonds" or "interactions" is not exhaustive, and the cited literature is only representative). These interactions have also been collected under the more general terms of "s-hole" and "p-hole", 31 referring to localised patches of positive electrostatic potential (ESP) in otherwise negative regions of ESP mapped on a molecular surface. 32 But it is arguable whether this activity has enhanced our understanding of the relationship between the structure of molecules (geometric and electronic), the crystal structures they form, and their consequent chemical and physical properties. Are we getting closer to the requisite intimate, and ultimately useful, understanding of why molecules and ions are arranged in crystals as observed, or are we merely cataloguing more and more examples of intermolecular "bonds" or "interactions" while losing sight of their common origins? In many instances where there is a focus on specic interactions, strongerand frequently more importantinteractions are being completely overlooked.
Our approach to understanding crystal packing combines colour mapping on molecular surfaces of properties derived from molecular wavefunctions and the efficient calculation of remarkably reliable intermolecular interaction energies 33,34 along with a graphical representation of their magnitude. 35 Considerable insight can be obtained via the anisotropy of the network of pairwise intermolecular interaction energies. This whole-of-molecule approach, which is blind to specic atom/atom contacts and/or interactions that may be deemed to be important, complements and challenges some current atom-atom approaches. [36][37][38] In this paper we highlight how these non-canonical interactions can be understoodin the same way as hydrogen bond interactionsthrough their common origin in the redistribution of electron density that results from chemical bonding. This redistribution is directly linked to the molecular ESP, commonly used to rationalise "s-hole" and "p-hole" interactions, 22,23,32,39 and hence to qualitative concepts such as electrostatic complementarity (alluded to by Gavezzotti in the epigraph above) and the calculation of meaningful intermolecular interaction energies. Finally, we show how visualization of the relative importance of electrostatic and dispersion contributions to intermolecular energies can reveal information about the architecture of molecular crystals, and possible links to their physical properties.
Visualizing the redistribution of electrons upon bonding aids in understanding the electrostatic potential The simplest way to understand the nature of the ESP of a molecule is through the electron distribution, and in particular the redistribution that occurs on chemical bonding. The crystallographer's model for the electron density of a molecule in a crystalthe independent atom modelcomprises spherically averaged electron densities located at the nuclear positions. It is not commonly recognised (but see, for example, the discussion in ref. 22) that the ESP of a spherical atom is necessarily positive everywhere, decaying to zero at innity, as a result of the incomplete shielding of the nuclear charge by the electron distribution at any nite distance. So regions of negative ESP for a neutral molecule arise from a local increase of electron density, and this necessarily results in a concomitant decrease of electron density elsewhere in the molecule, and a more positive ESP in that region. Fig. 1 to 4 show how, for a small number of molecules, this relationship can be visualized in several ways. These gures display positive (blue) and negative (red) isosurfaces of the deformation electron density, Dr, which correspond to local increases and decreases, respectively, of electron density relative to the sum of spherical atoms. An isosurface of the promolecule electron density, a molecular surface that closely resembles a van der Waals surface, 40 is used to map Dr as well as the molecular ESP. All calculations in this paper are based on B3LYP/6-31G(d,p) electron densities (B3LYP/DGDZVP where Br, I or Xe are involved) using molecular geometries from known crystal structures, with X-H bond lengths normalized to standard neutron diffraction values. Interaction energies are all based on the CE-B3LYP model 33,34 embodied in CrystalExplorer17. 41 Isosurfaces of Dr in these gures illustrate a number of recurring features for atoms in common bonding environments, and by mapping the same property on a molecular surface these facilitate a better appreciation of the origin of features in the maps of the ESP.
The build up of "lone pair" electron density above and behind the oxygen atom in water (Fig. 1) is clearly directly related to the broad region of negative ESP surrounding the oxygen atom. Similar "lone pair" Dr features are evident for the ring N atoms in s-triazine and 2-amino-5-nitropyridine (Fig. 2), all resulting in prominent electronegative regions in the ESP. The build up of electron density in the C-C bond in acetylene results in a band of negative ESP around the waist of the molecule, and a concomitant positive ESP around the hydrogen atoms. In formamide we see characteristic Dr features around the carbonyl O atom, which are repeated in chloroacetic acid and 2-amino-5-nitropyridine (Fig. 2). Also evident in the Dr maps for formamide and 2-amino-5-nitropyridine are small decits of electron density beyond the N atoms (and between the H atoms) in the planar amino groups, which have been labelled "s-holes" and discussed in the context of the pnicogen bond. 42 However, the Dr maps for those two molecules clearly show that the positive ESP characteristic of the -NH 2 group results from a broad band of negative Dr that is largely due to the electron density decits beyond the H atoms.
We include maps of Dr and ESP for XeO 3 in Fig. 1 as that molecule has been claimed to display a "s-hole" feature at the Xe atom, leading to what the authors proposed as "aerogen bonding". 29 But the Dr map actually shows a small build up above the Xe atom in Fig. 1; the Dr feature that results in a broad electropositive ESP region around the Xe atom comes largely from substantial withdrawal of electron density from the Xe-O bonds and their vicinity. We believe that the term "s-hole"if it is meant to convey a localised region of positive ESP associated with The term "s-hole" most commonly occurs in discussions of halogen bonding, where it refers to the localised positive ESP associated with a decit of electron density in the region beyond terminally bonded halogen atoms. This feature is very clearly seen in Dr isosurface maps for chlorine in chloroacetic acid (Fig. 2), iodine in 1,2-diiodo-1,1,2,2-tetrauoroethane ( Fig. 3) and uorine in hexauorobenzene (Fig. 4). But it should also be obvious that this decit does not occur in isolationit is part of a roughly quadrupolar feature directed along the C-X bond in all cases, and is always accompanied by a torus of electron density decit around the C-X bond. In combination these Dr features lead to localised positive ESP features along the extension of the C-X bond, with negative ESP regions surrounding the bond. As is well known, these features are strongest for iodine, less obvious for chlorine, and barely evident in the ESP map for uorine. Fig. 4 contrasts Dr and ESP maps for benzene and hexauorobenzene. It shows how the positive Dr isosurface features around the H atoms and in the C-C bonds in benzene are not evident in the mapping of Dr on the molecular surface, which is dominated by the electron density decit features. This in turn results in an ESP map that shows electropositive H atoms and broad electronegative regions above and below the ring. Although the benzene skeleton is largely unchanged in hex-auorobenzene, the Dr features associated with the uorine atoms, in particular the torus of positive Dr discussed above, result in moderate electronegative ESP around the F atoms, and broad electropositive regions above and below the ring. The complementary nature of the ESP maps for these two molecules is reected in their quadrupole moments, which are opposite in sign and very similar in magnitude, 43 an important consideration in rationalizing the packing of molecules in the 1 : 1 co-crystal, discussed further in the next section.
In Fig. 1 to 4 we have deliberately mapped the ESP over an identical range for all molecules, namely AE0.025 a.u. (AE65.6 kJ mol À1 per unit charge). Thus, dark red regions are where the ESP on the molecular surface is more negative than À0.025 a.u., and dark blue regions more positive than +0.025 a.u.; white regions depict near-zero potential. In this manner it is quite straightforward to broadly compare the electrostatic nature of various molecules on the basis of the ESP maps in these gures. For example, in Fig. 3 we see that 1,2-diiodo-1,1,2,2-tetrauoroethane displays relatively neutral ESP features across the surface except for the electropositive regions at the extensions of the C-I bonds. In contrast, the ESP for 1,4-dicyanobutane shows strong electronegative regions around the cyano groups, with the butane framework strongly electropositive.

Pictures of electrostatic complementarity correlate with electrostatic interaction energies
Some time ago we showed that mapping molecular ESPs onto Hirshfeld surfaces, in the context of a crystal packing diagram, considerably enhances the discussion of close molecular contacts in the crystal, using the concept of "electrostatic complementarity" between touching surface patches in adjacent molecules. 44 Here we elaborate further on this, with Fig. 5 to 9 showing small clusters of molecules extracted from a number of crystals, with the molecular ESP mapped on Hirshfeld surfaces, along with values of the electrostatic interaction energy computed directly from the two molecular charge (electron + nuclear) distributions.

Formamide
A cluster of three centrosymmetric hydrogen-bonded formamide dimers (Fig. 5) illustrates the sort of insight that can be gained through this approach. There is an obvious complementarity between ESP maps for the molecules involved in these pairs, with an almost perfect matching of zero potential (the white lines), and electronegative (red) patches paired with electropositive (blue) patches in adjacent molecules. The computed electrostatic energy for this pair, À76 kJ mol À1 , reects the strong complementarity in this case. Another hydrogen bond interaction shows a similar red-blue complementarity, but over a smaller area, and this correlates with a smaller electrostatic energy of À50 kJ mol À1 . An electrostatic energy of +5 kJ mol À1 results where electropositive regions of adjacent molecules are in contact.

s-Triazine
Intermolecular interactions for s-triazine are much weaker (Fig. 6). The stacking interaction shown on the le of the gure is stabilizing but electrostatically very weak at only À2 kJ mol À1 . But even here there is an exquisite complementarity between the ESP maps for adjacent molecules in this arrangement; each molecule is rotated 60 with respect to those above and below in order to maximise overlap between positive and negative regions. In the perpendicular direction, adjacent molecules also clearly pack in an arrangement that maximises this red-blue complementarity, and here the stabilizing electrostatic energy between adjacent molecules is somewhat greater but still much smaller than seen for formamide.

Benzene:hexauorobenzene co-crystal
The interactions between adjacent molecules in the co-crystal of benzene and hexauorobenzene are depicted in Fig. 7. In this case the stacking arrangement is  offset, presumably to obtain a lower electrostatic energy (À7 kJ mol À1 , compared with only À2 kJ mol À1 for s-triazine). In the perpendicular direction, the electrostatic interactions are weak to negligible, as expected from the much lighter colours of the ESP for these two molecules. Although these electrostatic energies are quite small, negative values coincide with red-blue complementarity in all cases, and values near zero with contacts that are either blue-blue (for benzenebenzene) or red-red (for hexauorobenzene).
Chloroacetic acid:2-amino-5-nitropyridine co-crystal As observed for formamide, the complementary nature of contacts between chloroacetic acid and 2-amino-5-nitropyridine in the co-crystal (Fig. 8) is obvious, and coincides with two stabilizing electrostatic energies of À130 and À38 kJ mol À1 . Very small positive electrostatic energies arise for close contacts between like molecules, where there is no evidence of complementarity. The arrangement of the molecular pair at top le in this gure is that discussed by Sarkar et al. 42 as  1,2-Diiodo-1,1,2,2-tetrauoroethane:1,4-dicyanobutane co-crystal The nal example in this section (Fig. 9) illustrates the strong electrostatic complementarity between molecules in the linear arrangement of 1,2-diiodo-1,1,2,2-tetrauoroethane and 1,4-dicyanobutane in their 1 : 1 co-crystal, attributed to N/I halogen bonding. 45 The electrostatic energy associated with this molecular pair is considerable at À32 kJ mol À1 , and in line with the expectation from the strong red-blue complementarity at the ends of these two molecules. But this is not the only strong interaction evident in this 2D arrangement in the gure.   There is also a very obvious electrostatic complementarity between cyanobutane molecules in adjacent 1D "chains", and once more we see almost perfect matching of zero potential for these two molecules. The electrostatic energy of À20 kJ mol À1 associated with this pairing is also substantial. A still weaker interaction is also evident between iodoperuoroethane and cyanobutane molecules in adjacent 1D "chains".

Energy frameworks reveal the architecture of molecular crystals
Although our original report and introduction of energy frameworks focused on their application to mechanical properties, we noted then that the accurate and efficient computation of interaction energies, coupled with their visualization in the form of an energy framework, represents a powerful tool for quantitatively exploring interaction energies in molecular crystals. 35 Here we apply this visualization tool to the same molecular crystals for which electrostatic interactions were analysed in the previous section. Fig. 10 to 14 show the same three frameworks for each crystal, and the energy scale usedthe width of the cylinders linking nearest neighbour moleculesis the same in all gures, enabling direct comparison of the relative importance of electrostatic (red) and dispersion (green) contributions to the total intermolecular energies (blue). This also facilitates comparison between molecular crystals, shedding light on which interactions are important, and how this differs between crystals.

Formamide
Energy frameworks for formamide (Fig. 10) show immediately that the total interaction energies for molecular pairs are largely due to electrostatics, as the total energy framework closely mirrors that for the electrostatic contribution; the repulsion energy essentially cancels the sum of smaller polarization and dispersion terms, as well as part of the electrostatic. The cyclic pair of molecules involving two N-H/O hydrogen bonds is evident in the electrostatic and total energy pictures (total CE-B3LYP energy ¼ À64 kJ mol À1 ), and these pairs are linked by single N-H/O hydrogen bonds (À31 kJ mol À1 ) to adjacent molecules. The interaction depicted in Fig. 5 with a positive electrostatic energy of +5 kJ mol À1 results in a total energy of +3 kJ mol À1 . It is important to emphasize that although polarization and dispersion terms are relatively small in this case, they are nevertheless essential for the computation of accurate energies. The accuracy and reliability of these CE-B3LYP energies can be assessed by their summation to estimate a lattice energy; the resulting value of À83 kJ mol À1 compares well with the reference lattice energy of À78.7 kJ mol À1 reported by Otero de la Roza and Johnson. 46

s-Triazine
The energy frameworks for s-triazine (Fig. 11) differ greatly from those for formamide. For the two interactions depicted in Fig. 6, where the electrostatic energies are À2 and À9 kJ mol À1 , for stacking and in-plane nearest neighbours, the ordering of dispersion energies is the reverse, at À16 and À8 kJ mol À1 , Fig. 13 2-Amino-5-pyridine:chloroacetic acid co-crystal (TETXUL01): energy frameworks for separate electrostatic (red) and dispersion (green) contributions to the total nearest neighbour pairwise interaction energies (blue). See Fig. 10 for details. respectively. As a result, the two interactions are very close in energy and just over À12 kJ mol À1 . The electrostatic and dispersion frameworks show quite different architectures, most notably for the stacking interaction, and we conclude that the crystal structure of s-triazine represents a subtle balance between electrostatic and dispersion contributions of similar magnitude and importance. The computed CE-B3LYP lattice energy of À55 kJ mol À1 is also close to the reference value of À60.5 kJ mol À1 . 46 Benzene:hexauorobenzene co-crystal As seen in Fig. 7, stacking interactions are also important for the 1 : 1 co-crystal of benzene and hexauorobenzene, and here the electrostatic energy is largest for the stacking motif, and much less for in-plane nearest neighbours. The corresponding energy frameworks (Fig. 12) show clearly the more dominant role of dispersion in this crystal. The CE-B3LYP energy for the stacking interaction is À22 kJ mol À1 (with electrostatic and dispersion terms of À7 and À28 kJ mol À1 , respectively), almost twice that for the most stabilizing in-plane interaction, that between benzene and hexauorobenzene, of À12 kJ mol À1 (with electrostatic and dispersion terms of À3 and À15 kJ mol À1 , respectively). The total energies for the other various in-plane interactions are all in the narrow range of À4 to À6 kJ mol À1 , with the exception of the benzene-benzene closest pair depicted in Fig. 7, which has a total energy of only À1 kJ mol À1 .
Chloroacetic acid:2-amino-5-nitropyridine co-crystal In their charge density study of the co-crystal between chloroacetic acid and 2amino-5-nitropyridine, Sarkar et al. focused their attention on the -NH 2 /Cl close contact, and the possibility of either a pnicogen bond between N and Cl, or the "less probable" bifurcated N-H/Cl hydrogen bonds. 42 But, and importantly, they Based on the maps of Dr and ESP we have already noted that the interaction involving the -NH 2 /Cl close contact is clearly stabilizing, with an electrostatic a Although all close contacts between symmetry-related atoms (e.g., Cln/Cln, n ¼ 1-3, etc.) in these stacking interactions are identical, an additional close contact results from the offset stacking arrangement. Fig. 16 Closest Cl/Cl and Br/Br contacts (magenta dashed lines) between adjacent molecules in hexachlorobenzene (HCLBNZ12) and hexabromobenzene (HBRBEN02). The atom labelling is the same as that in the original CIF files. See Table 1 for interaction energies between molecules A, B and C, as well as those involved in the stacking motif. energy of À38 kJ mol À1 , but dissecting this into pnicogen vs. bifurcated hydrogen bonds is next to impossible. The total CE-B3LYP energy for this interaction is À33 kJ mol À1 . The much stronger interaction anticipated by Sarkar et al. is that involving a cyclic pair of N-H/O]C and O-H/N hydrogen bonds. The electrostatic energy of À130 kJ mol À1 for this pair dominates the electrostatic energy framework in Fig. 13. Polarization and repulsion energies are also large, and the total energy of À69 kJ mol À1 makes this interaction the strongest in this crystal, and this is shown clearly in the total energy framework in Fig. 13. The interaction with the largest dispersion contribution, namely that coming directly out of the page in Fig. 13, is the stacking interaction between two 2-amino-5-nitropyridine molecules, with an electrostatic energy of +1 kJ mol À1 and a dispersion energy of À26 kJ mol À1 , resulting in a total energy of À11 kJ mol À1 .
1,2-Diiodo-1,1,2,2-tetrauoroethane:1,4-dicyanobutane co-crystal Energy framework pictures for this co-crystal (Fig. 14) show that although the N/ I halogen bond has the most negative electrostatic energy (Fig. 9), it is far from the strongestor most importantintermolecular interaction in this crystal. The strongest is that between cyanobutane molecules in adjacent "chains", with an energy of À25 kJ mol À1 , much greater than for the halogen bonded pair, À14 kJ mol À1 . The mix of electrostatic and dispersion contributions is responsible for this, as is the large repulsion energy associated with the close N/I contact. There are also two other notable molecular pairs, with total CE-B3LYP energies of À14 and À18 kJ mol À1 , both between tetrauoroethane and cyanobutane molecules in adjacent "chains", and for which the dispersion contribution exceeds À20 kJ mol À1 in both cases. These results clearly contradict the observation in ref. 45 that the halogen bonds "are the only strong interactions present" in this co-crystal, and they indicate that the original description of the crystal in terms of 1D "chains" is clearly inaccurate. This linear arrangement of alternating tetra-uoroethane and cyanobutane molecules is observable in the total energy framework in Fig. 14, but other interactions between these molecules are at least as strong (and important), and the cyanobutane-cyanobutane interaction is by far the strongest.
Putting it all together: hexachlorobenzene and hexabromobenzene The crystal structures of hexachlorobenzene (HCB) and hexabromobenzene (HBB) have been discussed in the context of the nature of halogen/halogen interactions and their mechanical properties, in particular bending and nanoindentation. 47,48 HCB has since been subjected to detailed analysis of the experimental charge density at 100 K in order to characterize the nature of Cl/Cl interactions in the crystal, especially the so-called halogen trimer (X 3 or Hal 3 ) synthon. 49,50 Most recently, plastic bending in HCB has also been examined in considerable detail by micro-IR and micro-X-ray diffraction. 51 HBB is isostructural with HCB but, unlike HCB, it only exhibits bending at elevated temperatures ($400 K). 48 These studies acknowledge the strongly anisotropic intermolecular interactions that are responsible for the mechanical behaviour of HCB, and although this anisotropy has been quantied for HCB using counterpoise- corrected BLYP-D/def2-TZVP energies for molecular pairs, 52 no comparison has been made with HBB. Here we use CE-B3LYP model energies, and the separate electrostatic and dispersion contributions, to investigate the interaction anisotropy in HCB and HBB using energy framework diagrams (Fig. 15) and a detailed analysis of the various interactions between a molecule and the 14 nearest neighbours in its rst coordination sphere (Table 1 and Fig. 16). We use the original 100 K structures reported by Reddy et al. 48 The energy framework diagrams (Fig. 15) provide immediate insight into the columnar architecture of these two crystal structures, and especially the differences between them. Electrostatic interactions between adjacent molecules are small for HCB (and only there between molecules arranged in a stacking fashion), but much more signicant for HBB. In both crystals dispersion plays a very important role, again more so for HBB. The total energy frameworks (blue) suggest that HCB is best described in terms of columns involving relatively strong interactions between adjacent molecules (À35 kJ mol À1 ; Table 1), with much smaller interaction energies between molecules in adjacent columns (between À4 and À5 kJ mol À1 ). All interactions are stronger in HBB: stacked pairs at À54 kJ mol À1 , and intercolumn pairs between À7 and À10 kJ mol À1 . The essential difference between HCB and HBB is revealed to be the stronger interaction between molecules in adjacent columns in the latter, resulting in a failure to bend at room temperature. Presumably the observed bending at $400 K is facilitated by expansion of the unit cell, reducing the interaction energies as molecules become further apart.
It is tempting to compare the results in Table 1 with energies reported by Brezgunova et al., 49 based on bond critical point (bcp) properties derived from topological analysis of electron distributions. There are many assumptions involved in such an analysis, as well as substantial experimental errorsincluding model dependencewhen derived from a charge density experiment, and one of us has recently demonstrated that results of that kind are unreliable. 53 But, if we assume that it is somehow possible to derive reliable interaction energies from the electron density and its Laplacian at bcps for intermolecular interactions, we are faced with the dilemma highlighted in Fig. 16 (see also data for Cl/Cl contacts in Table 1). In their analysis of the halogen trimer synthon, Brezgunova et al. 49 summed energy estimates based on just the three bcps that are identied immediately with the triangle of short contacts, namely Cl1/Cl2, Cl2/Cl3 and Cl1/Cl3 in Fig. 16 (the atom labelling is not the same as that used by those authors). But it should be obvious that the interaction between molecules A and B involves three of those short Cl/Cl contacts, and hence three such bcps if they exist. Taken together, the halogen trimer energy, would require summing over all six Cl/Cl bcps identied in the earlier study by Bui et al. 50 (see Table 1 in that work).
The present CE-B3LYP energies in Table 1 simplify this inquiry considerably. The energy associated with the halogen trimer synthon in these crystals is the sum of the three intermolecular energies for pairs depicted in Fig. 16, and given in Table 1: À13 kJ mol À1 for HCB and À25 kJ mol À1 for HBB, suggesting this synthon is nearly twice as strong in HBB than in HCB, in contrast to a difference of only 17% estimated by Brezgunova et al. 49 But once more, we question the focus on this particular structural motif, as it ignores another intercolumn interaction with similar energy to those for the other molecular pairs, namely À4 kJ mol À1 for HCB and À9 kJ mol À1 for HBB (Table 1). As we have done for formamide and s-triazine, we can estimate the reliability of the CE-B3LYP energies for HCB and HBB by comparing computed lattice energies with experimental sublimation enthalpies. The CE-B3LYP estimated lattice energies are À69 (HCB) and À115 (HBB) kJ mol À1 , compared with sublimation enthalpies of 88 AE 12 kJ mol À1 (from 12 measurements) for HCB, and a single value of 118 kJ mol À1 for HBB. 54 (We note that the interaction energies reported in ref. 52 for HCB are all $25% greater than the present CE-B3LYP results, but we cannot identify the origin of this difference).

Conclusions
The question asked in the title of this paper has appeared in the title of recent discussions relating to the nomenclature and denition of polymorphs, salts and co-crystals, 55 pseudopolymorphism, 56 and in the context of what isand is nota halogen bond according to its denition. 57,58 Nomenclature, and being precise about what we mean by the terms we use, is of course enormously important in science. For that reason both the hydrogen bond 59 and halogen bond 24 have been the subjects of recent IUPAC recommendations regarding their use and denition (see also the proposal to systematically name noncovalent interactions by the group of the periodic table to which the electrophilic atom belongs 60 ). But it is not at all obvious to us how those recommendations help in the "understanding of intermolecular interactions in the context of crystal packing". We have identied in this work instances where paying attention to particular interactions, because they are expected to be important, or because they have been given a special name, can result in ignoring signicantly more important interactions.
To best understand the nature of intermolecular interactions in the context of crystal packing we believe a more balanced approach is neededone that avoids any bias towards a focus on specic atom/atom interactions, or what appear to be novel interactions. We have presented one such way of achieving this, using the graphical and computational tools embodied in our research toolbox Crys-talExplorer, but it is certainly not the only one. The broad details of non-covalent interactions, including hydrogen bonds, can be largely understood through their common origin in the redistribution of electron density that results from bonding. This redistribution is directly linked to the molecular electrostatic potential, to qualitative concepts such as electrostatic complementarity, and to the efficient calculation of reliable intermolecular interaction energies. Visualization of these energies, along with their electrostatic and dispersion components, sheds light on the architecture of molecular crystals, in turn providing a link to actual crystal properties.