Report

Imaging the halogen bond in self-assembled halogenbenzenes on silver

See allHide authors and affiliations

Science  13 Oct 2017:
Vol. 358, Issue 6360, pp. 206-210
DOI: 10.1126/science.aai8625

Visualizing halogen bonding

Even though halogen atoms are highly electronegative, a noncovalent bond can form between an electron donor and a halogen atom in a covalent bond. Such interactions are facilitated by the formation of electron-depleted regions in the halogen's covalent bond, a situation least likely for fluorine atoms. Han et al. used noncontact scanning tunneling microscopy with submolecular resolution to explore how the size and polarizability of halogens affect complex formation by halogenated benzene molecules adsorbed on a silver surface (see the Perspective by Neaton). With the help of density functional theory, they show how several weak interactions, including van der Waals forces, electrostatic repulsions, and halogen bonds, affect structure.

Science, this issue p. 206; see also p. 167

Abstract

Halogens are among the most electronegative elements, and the variations in size and polarizability of halogens require different descriptions of the intermolecular bonds they form. Here we use the inelastic tunneling probe (itProbe) to acquire real-space imaging of intermolecular-bonding structures in the two-dimensional self-assembly of halogenbenzene molecules on a metal surface. Direct visualization is obtained for the intermolecular attraction and the “windmill” pattern of bonding among the fully halogenated molecules. Our results provide a hitherto missing understanding of the nature of the halogen bond.

The intermolecular halogen bond, hereafter the halogen bond, is an electrostatically driven, noncovalent interaction that has broad applications, including molecular self-assembly (1), supramolecular chemistry (2, 3), crystal engineering (4, 5), and drug design (6, 7). A halogen atom in a molecule is usually negatively charged. However, the anisotropic distribution of its electron density can lead to an electron-depleted σ-hole at the pole on the extension of the C–X bond (where X is a halogen atom) and an electron-rich equatorial belt perpendicular to the bond direction (810). The halogen bond is formed between the σ-hole and nucleophilic sites of adjacent molecules. A triangular bonding structure involving three halogen atoms, named halogen-3 synthon, has been found in varieties of halogenated molecular crystals (11). In the halogen-3 synthon, the σ-hole of one halogen atom points toward the equatorial region of the adjacent halogen atom to form a “windmill” structure, as shown in fig. S1E.

However, the most electronegative and least polarizable fluorine is not expected to develop the σ-hole and participate in halogen bonding (8, 9, 12). Results from noncontact atomic force microscopy (NC-AFM) suggest the presence of C–F···F–mediated halogen bonding in a network of fully fluorosubstituted phenylene ethynylene molecules (13). However, the rotational freedom of individual fluorinated rings is highly restricted because of the rigid chain structure in these molecules. By contrast, the hexafluorobenzene (C6F6) molecule has higher rotational freedom to accommodate a minimum-energy configuration for the halogen bond.

Previous studies have shown that C6F6 molecules self-assemble into ordered layer structures on different low-index surfaces of coinage metals (14, 15). A free electron–like band is formed in a C6F6 island caused by the hybridization of their lowest unoccupied molecular orbitals with superatom character (14, 15). However, the real-space visualization of the ordered two-dimensional (2D) bonding structures of C6F6 self-assembly has remained unreported, so the nature of the intermolecular interactions leading to the self-assembly remains to be understood.

In recent years, submolecular resolution of chemical structure has been achieved by using a CO tip in NC-AFM (1618) and scanning tunneling microscope (STM) (19). The origin of this enhanced contrast has attracted considerable experimental and theoretical treatment (2022). The CO tip in the NC-AFM probes the vertical (out-of-surface plane) gradient of the forces on the CO molecule, which become highly localized over the atoms and bonds as a result of short-range repulsive forces. In the inelastic tunneling probe (itProbe) based on the STM, the lateral vibrations of the CO molecule on the tip serve as a probe of the gradient of lateral forces on the CO molecule, yielding similar contrast at a small tip-sample gap (23).

We spectroscopically imaged C6F6, hexabromobenzene (C6Br6), and 1,3,5-trifluorobenzene (C6H3F3) adsorbed on a Ag(110) surface at 600 mK with the itProbe (24). The contents of the images were analyzed by density functional theory (DFT) calculations with van der Waals (vdW) correction. Similar intermolecular-bonding structures were observed for C6F6 and C6Br6 self-assemblies, which can be understood by generalizing the classic halogen-bonding mechanism. Images of the intermolecular interactions in the C6H3F3 self-assembly validated the nature of the halogen bond and additionally revealed that the H···F hydrogen bond also played a role in forming the structural order. In addition to imaging the molecular structure, the CO tip facilitated the resolution of atoms on the metal surface and enabled the determination of the molecular adsorption site (19, 24). A CO molecule could be transferred from the surface to the tip by scanning with a small tunneling gap (e.g., 1 nA and 1 mV, corresponding to 1-megohm gap resistance). The presence of CO on the tip was confirmed by the detection of characteristic CO vibrations over the background Ag surface from inelastic electron tunneling spectroscopy (IETS) with the STM (25).

The ordered layer structure formed by C6F6 and imaged with the CO tip (Fig. 1A) showed that the lattice of the first molecular layer was not commensurate with the rectangular symmetry of the underlying Ag(110), resulting in domain boundaries indicated by yellow dashed lines. The growth of the second molecular layer initiated at these domain boundaries. A zoomed-in view of the boxed area in Fig. 1A is shown in Fig. 1B with an overlay of a lattice mesh of the underlying silver substrate. Individual molecules were clearly resolved and formed a nearly hexagonal (rhombic) lattice (see table S2). The molecules preferentially adsorbed on the Embedded Image atomic rows of the Ag(110) but were not in perfect registry with the substrate (see the larger scan in fig. S2A). However, the intermolecular orientations, in particular the rotational angles of the individual C6F6 with respect to one another, were still unresolved on the basis of the topographic images.

Fig. 1 Self-assembled C6F6 island and C6Br6 clusters on a Ag(110) surface.

(A) Constant-current topography of C6F6 island and coadsorbed CO molecule on Ag(110) scanned with a CO-terminated tip (120 Å by 120 Å; set point: 0.1 nA, 0.1 V; z range: 2.56 Å). Domain boundaries are indicated by yellow dashed lines. (B) Constant-current zoomed-in topography of the black square area in (A). The gray rectangular grid represents the underlying Ag(110) lattice with surface atoms at the line intersections (15 Å by 15 Å; set point: 0.12 nA, 60 mV; z range: 0.61 Å). (C) Constant-current itProbe imaging at 1.2 mV of the same area as in (B) (160 pixels by 160 pixels; 15 Å by 15 Å; set point: 0.12 nA, 10 mV). (D) Schematic diagram of the itProbe image in (C). Intermolecular-bonding network highlighted by red and orange dashed lines, indicating the F-3 synthon and trans–type I F···F interaction, respectively. (E) Constant-current topography of C6Br6 trimer, dimer, and monomer (inset) scanned with a CO tip (64 Å by 64 Å and 15 Å by 15 Å for inset; set point: 0.1 nA, 10 mV; z range: 1.90 Å). (F to J) (F), (G), and (I) show constant-height itProbe images at 1.5 mV of the monomer, dimer, and trimer shown in (E). Sizes are 96 pixels by 96 pixels and 12 Å by 12 Å for (F), 100 pixels by 100 pixels and 18.75 Å by 18.75 Å for (G), and 128 pixels by 128 pixels and 20 Å by 20 Å for (I). (H) and (J) show schematic diagrams of the C6Br6 dimer and trimer shown in (E); in (J), red dashed lines show Br-3 synthon, and in (H) and (J), orange dashed lines show trans configuration of type-I Br···Br interaction. The rectangular grids in (E), (H), and (J) show the Ag(110) lattice resolved at close scans with a CO tip (set point: 0.5 nA, 10 mV).

The hindered translational vibration of the tip CO was sensitive to its local environment. The short-range repulsive forces between the CO and the underlying molecule red shifted the CO-hindered translational vibration (19, 23). By monitoring the d2I/dV2 intensity (I, current; V, bias voltage) at a chosen bias of 1.2 mV, the locations of the red-shifted CO-hindered translational vibration were imaged by the itProbe, as shown in Fig. 1C, with the corresponding schematic diagram in Fig. 1D. Each covalent bond structure of the individual molecules (C–C and C–F bonds) was clearly resolved. Furthermore, a rich network of intermolecular interactions within the C6F6 island was uncovered (highlighted by the dashed lines between the molecules). Three neighboring C6F6 self-organized to form a triangular windmill structure similar to halogen-3 synthon in the halogen bond (12, 26). Each molecule rotated ~20° from the fluorine head-to-head configuration.

Within each F-3 synthon shown in Fig. 1D, the F atom in one molecule was not aligned with the C–F bond axis of the adjacent molecule; the C–F···F angle was not 180°. This difference can be understood by considering the interaction between a molecule and the substrate. Calculations by DFT (table S1) indicate that, in agreement with the experiment (fig. S4), individual C6F6 favored the short-bridge SB2 adsorption site on a Ag(110) surface (fig. S3), and the two C–F bonds were aligned along the Embedded Image direction. The orientation of each molecule in the self-assembled island in Fig. 1D was almost the same as the isolated monomer and was only rotated ~4° clockwise from the SB2 configuration because of the intermolecular interactions. In addition to the F-3 synthon illustrated with the triangular red dashed lines, the trans–type I F···F interaction (fig. S1B) was also observed in the itProbe images, as indicated by orange dashed lines in Fig. 1D (24).

The large polarizability of Br leads to a Br···Br halogen bond that facilitates the self-assembly of Br-terminated aromatic molecules on inert surfaces (2733). The adsorption of a C6Br6 molecule on Ag(110) serves as a standard model of halogen bonding to compare with the intermolecular-bonding structure in C6F6 islands. The C6Br6 molecules were adsorbed as monomers, dimers, and triangular trimers, as shown in the topographic image of Fig. 1E. The orientation of the C6Br6 molecules and their intermolecular interactions within a cluster were imaged by constant-height itProbe in Fig. 1, F, G, and I. The corresponding schematic diagrams of the molecules with the underlying Ag(110) lattice are shown in Fig. 1, E (inset), H, and J. The molecular adsorption geometry was determined from the superposition over the same area of atomically resolved scan of the substrate surface and the itProbe images of the molecules (fig. S2, C and D). All C6Br6 monomers were in registry with the substrate in the SB2 adsorption site, as shown in the inset of Fig. 1E and from the DFT calculations.

Both molecules in the C6Br6 dimer nearly maintained the SB2 adsorption configuration, except for a slight rotation and displacement toward each other, as shown in Fig. 1H. Notably, the molecules rotated 30° in the trimer and translated from the SB2 to the LB1 adsorption configuration, as illustrated in Fig. 1J. The Br atoms interacted cyclically to form the close-packed Br-3 synthon, similar to the F-3 synthon in C6F6 islands. However, the C–Br bond axis was almost perfectly aligned with the neighboring Br atom because of the relatively stronger intermolecular interaction compared to the interfacial interaction with the substrate.

Comparison of the measured results with DFT calculations revealed the origin of the contrast in the itProbe image. STM-IETS (25) of the CO tip over the selected points marked in Fig. 2A of the C6F6 layer is shown in Fig. 2B. The vibrational energy varied by >2 meV when the CO tip was positioned over different locations of the adsorbed molecules. For example, the vibrational energy shifted from 3.4 meV in the center of the carbon ring to 1.2 meV over the ring. This difference resulted from the interaction of the CO tip with molecules underneath, which can be described by the potential energy surface (PES). The spatial variations of the PES (Fig. 2C) were sensed by the CO molecule and modified the energy of its vibrations, particularly the CO-hindered translational vibration that can be approximated as a lateral harmonic oscillator. The CO vibrational energy could be easily affected by interaction between CO and C6F6. As illustrated in Fig. 2D, the ridges and apexes in the PES are associated with regions of negative curvature that reduced the overall curvature of the CO vibrational potential well and led to a measurable red shift of the vibrational energy. The 2D imaging of the vibrational intensity at a reduced energy [such as 1.2 compared to 2.5 meV over the Ag background (19)] revealed a real-space view of the apexes and ridges of the PES, as demonstrated by the good agreement between the principal surface curvature of the PES in fig. S5A and the itProbe image in Fig. 1C (also reproduced in fig. S5C).

Fig. 2 Origin of the contrast in itProbe imaging.

(A) Constant-current topography of C6F6 island scanned with a CO tip (15 Å by 15 Å; set point: 0.08 nA, 10 mV; z range: 0.79 Å). (B) d2I/dV2 spectra of the CO tip measured over five different locations of the C6F6 island specified in (A). For clarity, all of the spectra are offset vertically by 400 nA/V2 and with labeled peak positions. The intercept of each spectrum with the orange dashed line at 1.2 mV characterizes the intensity variation caused by the shift of CO-hindered translational energy. Bias voltage modulation: 1 mVRMS at 311.11 Hz (RMS, root mean square); set point: 0.09 nA, 10 mV. (C) Calculations of the 2D PES of the Ag-CO tip and underlying C6F6 lattice without the substrate and for the same area as Fig. 1C. The tetrahedral Ag tip is set at constant height above the C6F6 molecular plane, and the CO molecule attached to the Ag tip is free to relax. The tip height is set at 3.0 Å from the oxygen center to the C6F6 molecular plane prior to CO-tip relaxation. The energy range of the PES, the gray palette, is 16.5 meV. (D) Schematic diagram of the mechanism for the itProbe, showing downshift of CO tip–hindered translational vibration over two ridges [negative (–) curvature] and upshift of CO tip–hindered translational vibration over a valley [positive (+) curvature] of the PES. Here, k is the local curvature of the vibrational well and ω is the vibrational frequency.

We further showed that the ridges and apexes in the PES for CO tip over the C6F6 lattice correlate with the locations of intramolecular and intermolecular bonds in the C6F6 lattice. In particular, intermolecular interactions within the island produced additional features in the itProbe images that were absent for isolated molecules. A general feature of bonds is charge accumulation between atoms, which inevitably disturbs the distribution of Coulomb and exchange-correlation potentials in the vacuum nearby. According to the qualitative arguments in the supplementary materials, the smaller surface curvature of the PES correlates with the charge density of C6F6 in the plane that the oxygen end of CO moves in during the scan. This correlation could be seen by comparing the maps of PES curvature with the total charge density and also the charge-density difference for the C6F6 self-assembly in fig. S5, A, D, and E, in a plane 3 Å above the C6F6 molecular plane.

The bright circular border around the C6Br6 monomer in the itProbe image of Fig. 1F reflected the PES profile created by the large vdW radius of the six Br atoms. By contrast, the itProbe images of the dimer and trimer of C6Br6 exhibited distorted circular borders that were flattened toward the intermolecular region caused by the modification of the PES associated with the intermolecular interactions between adjacent C6Br6 molecules. Although the correlation between the sample charge distribution and the apparent bond contrast in NC-AFM or itProbe images continues to be under debate, the itProbe images in the current study determined the intermolecular interactions that were also revealed by DFT calculations.

DFT calculations incorporating vdW correction have been used to clarify the nature of the intermolecular interactions among halogenated benzenes in the self-assembly. For simplicity, we only considered the hexagonal lattice of molecules without the silver substrate. The molecules were allowed to rotate in the relaxation process. The binding energy and rotation angle θ (defined in Fig. 3A) of the optimized C6F6 and C6Br6 lattices are plotted as a function of lattice constant d in Fig. 3, B and C. The calculated results are summarized in table S2 for comparison with the experiment. The binding energy of the C6F6 lattice without vdW correction was 12 meV, which is <10% of the binding energy when the vdW correction is included (128 meV). Thus, the driving force for self-assembly of C6F6 molecules originates from vdW interactions. In addition, the optimized rotation angle (20.3°) at the most stable lattice constant matched well with the experimental value (21.2° ± 1.3°). Notably, the rotation angle of the molecule was nearly independent of the vdW correction. This independence suggests that the relative orientation of the molecules is not driven by vdW interactions, despite its dominant contribution to the binding energy. Instead, the neighboring C6F6 molecules self-arranged into a windmill pattern to minimize the Coulomb repulsion between the F groups (13). Similarly, for the C6Br6 lattice, the bonding energies were also much larger with the inclusion of the vdW correction, but the molecular orientation was similarly independent of the correction (Fig. 3C and table S2).

Fig. 3 DFT calculations of the 2D hexagonal lattices of C6F6 and C6Br6 without substrate.

(A) Schematic diagram of the 2D hexagonal C6F6 lattice with lattice constant d. The rotation angle θ of the molecule is measured from C–F bond direction to the line connecting the centers of two adjacent molecules. (B) Calculated intermolecular-bonding energy per molecule of the C6F6 lattice and final relaxed rotation angle θ as a function of lattice constant d, with and without vdW correction. The angle is set at 10° initially at each lattice constant d prior to relaxation. (C) DFT results for the C6Br6 lattice. For comparison, experimentally measured (Exp) angle and lattice constant are shown by the green diamond with error bars. The black arrows in (B) and (C) indicate the y axes for energy and angle plots.

The windmill F-3 synthon in the C6F6 self-assembly had similar directionality to the C–F···F contacts observed during previous studies of BPEPE-F18 (13) and to the Br-3 synthon of the C6Br6 trimer in this study. Traditionally, the C–F···F contacts are not classified as halogen bonding because the small polarizability of F atoms prevents the development of a σ-hole at the pole of the C–F bond (8, 9, 12). However, the universal pattern of halogen-3 synthon observed in the self-assembly of hexahalogenbenzene suggests a common bonding mechanism. The halogen-3 synthon in the molecular assemblies arises from the sum of the vdW attraction and the electrostatic interaction between terminating halogen atoms. Molecules rotate to optimize the electrostatic interaction, which minimizes the electrostatic repulsion for the F-3 synthon and maximizes the electrostatic attraction for the Br-3 case.

The conclusions derived from itProbe images of the halogen bond in the self-assembly of hexahalogenbenzene (C6F6 and C6Br6) can be further validated from the self-assembly of 1,3,5-trifluorobenzene (C6H3F3) on a Ag(110) surface. The C6H3F3 molecules formed close-packed islands with a rectangular unit cell as shown in Fig. 4A. The itProbe image in Fig. 4B led to the corresponding schematic diagram in Fig. 4E, showing the molecular orientation of each molecule and the intermolecular interactions between molecules. Similar to C6F6, each C6H3F3 molecule adsorbed on top of the Embedded Image atomic row of the Ag(110) substrate (see also fig. S2). All the molecules in a given row along the [001] direction had the same orientation, but molecules in the adjacent rows were rotated by 60° (or equivalently 180°). Thus, the bonding differed in the two halves of the rectangular unit cell, resulting also in the nonequivalent separation between adjacent rows along the [001] direction, as shown in Fig. 4B.

Fig. 4 Self-assembly of 1,3,5-trifluorobenzene (C6H3F3) island on a Ag(110) surface.

(A) Constant-current topography of a C6H3F3 island scanned with a CO tip (56 Å by 56 Å; set point: 0.1 nA, 100 mV; z range: 1.65 Å). The rectangular unit cell is indicated by a yellow dashed box. The molecules are numbered from 1 to 7 for identification in the itProbe images. (B) Constant-current itProbe imaging at 1.5 mV over the green square area marked in (A) (180 pixels by 180 pixels; 16.88 Å by 16.88 Å; set point: 0.1 nA, 17.5 mV). (C) Magnified constant-current itProbe image at 1.75 mV over the intermolecular region defined by molecules 2, 4, and 5. (D) Same itProbe image as (C) except that the intermolecular region is defined by molecules 4, 5, and 7 (200 pixels by 200 pixels; 9.38 Å by 9.38 Å; set point: 0.1 nA, 17.5 mV). (E) Schematic diagram corresponding to the itProbe image of C6H3F3 self-assembly shown in (B). The intermolecular-bonding network is highlighted by blue, orange, and magenta dashed lines, representing the C–H···F interaction and cis- and trans-configurations of type-I F···F interactions, respectively (F in green and H in white).

The molecule labeled 4 in Fig. 4A did not lie in the center of the rectangular unit cell. This alternating pattern was repeated throughout the entire island and matched the molecular layer structure of the 3D C6H3F3 crystal (34), which is stabilized by the bifurcated C–H···F interactions. Indeed, such V-shape bifurcation structures were captured in the zoomed-in itProbe images in Fig. 4, C and D. These structures appeared when one F atom interacted with two H atoms (Fig. 4C) or two F atoms interacted with one H atom (Fig. 4D). In addition to the V-shape C–H···F interactions, trans– and cis–type I halogen interactions were also observed in Fig. 4B and are highlighted with red and yellow dashed lines in Fig. 4E. The similarities between the 2D island of C6H3F3 and its 3D crystal structure indicate the dominance of the intermolecular interactions over the molecule-substrate interactions. The universal occurrence of type-I halogen interactions in the self-assembly of C6F6, C6Br6, and C6H3F3 validates the previous conclusions on the nature of the halogen bond.

Supplementary Materials

www.sciencemag.org/content/358/6360/206/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S6

Tables S1 and S2

References (3541)

References and Notes

  1. See supplementary materials.
Acknowledgments: This work was supported by the Chemical Sciences, Geosciences, and Biosciences Division, Office of Science, U.S. Department of Energy, under grant no. DE-FG02-04ER15595 (Z.H. and P.J.W.) and DE-FG02-06ER15826 (C.X.); the Office of Naval Research under grant no. N00014-16-1-3150 (P.J.W.); the Condensed Matter Physics Program, Division of Materials Research, National Science Foundation, under grant no. DMR-1411338 (G.C., C.C., and X.W.); the Office of Science, U.S. Department of Energy, under grant no. DE-FG02-05ER46237 (X.W.); and the Chinese National Science Foundation under grant no. 11474056 (Y.Z.). Additionally, X.W. received support from the China Scholarship Council and the National Natural Science Foundation of China (grant no. 61171011). Computer simulations were partially supported by the National Energy Research Scientific Computing Center (NERSC). Experiments were carried out by Z.H., G.C., C.C., C.X., P.J.W., and W.H., and calculations were performed by X.W., Y.Z., and R.W. All data are reported in the main text and supplementary materials.
View Abstract

Navigate This Article