Surface-controlled reversal of the selectivity of halogen bonds

Intermolecular halogen bonds are ideally suited for designing new molecular assemblies because of their strong directionality and the possibility of tuning the interactions by using different types of halogens or molecular moieties. Due to these unique properties of the halogen bonds, numerous areas of application have recently been identified and are still emerging. Here, we present an approach for controlling the 2D self-assembly process of organic molecules by adsorption to reactive vs. inert metal surfaces. Therewith, the order of halogen bond strengths that is known from gas phase or liquids can be reversed. Our approach relies on adjusting the molecular charge distribution, i.e., the σ-hole, by molecule-substrate interactions. The polarizability of the halogen and the reactiveness of the metal substrate are serving as control parameters. Our results establish the surface as a control knob for tuning molecular assemblies by reversing the selectivity of bonding sites, which is interesting for future applications.

This ratio exactly matches the theoretical ratio of bonds that is needed for forming Sierpinski triangles (see Figure S1). On Au(111) the larger clusters form braid-like structures in the fcc regions of the herringbone reconstruction (Figure 2b,e), which also significantly influences the observed bond ratios. and bond distance L for each of the three halogen bonds. The effective bonding angle is close to 120° for all three halogen bonds. (d,e) Simulated force map and bonding configuration of the calculated structure (same panel as in Figure 4f). The bonding angles and distances of the calculated structure (panel e) and the experimental AFM images (panel c) are in remarkable agreement. Parameters: (a-c) amplitude = 40 pm, imaging distance = -23 pm with respect to U = 100 mV and I = 100 pA.

Supplementary Figure S4 -Snapping of IparaBrmeta-TP monomers and dimers to substrate lattices
Supplementary Figure S4. Snapping of IparaBrmeta-TP monomers and dimers to the Cu(111) and Au(111) substrate lattices. Given are the counted number of molecules vs. their orientation on the substrate lattice for Cu(111) (a,b) and Au(111) (c,d). The orientation angles have been determined from several STM overview scans. Due to the uncertainty in the angle measurement a range of +-15° has been chosen around  and [-12-1] directions, respectively (blue and pink regions). Molecules within this range are considered as "snapped to the lattice". For Cu(111) approximately 95 % of the molecules are found within this range, irrespectively if monomers (a) or dimers (b) are considered. On Au(111) a weaker snapping effect is observed. Only 70% of the single molecules (c) and only 55 % of the molecules within dimers (d) snap to the Au lattice. (d,e,f) Three constant-height AFM images of the molecule in (a) for three different imaging distances, which reveal the twisting angles of the three phenyl rings qualitatively. These twisting angles are in good agreement with the calculated structure (see g and h and Ref. 1 ). Furthermore, the image contrast agrees well with the simulated AFM images of IparaBrmeta-TP (see Figure S9). The iodine atom is located 43 pm closer to the Cu(111) surface than the bromine atom (see panel g) due to stronger iodine-Cu interactions.  Figure S6 and Ref. 1 ]. The halogens are located close to bridge sites, the phenyl rings are located above fcc sites, and the X-C bond is aligned with the crystallographic  direction. The side views (b,d) reveal that the iodine is approximately 40 pm closer to the Cu(111) surface plane than the bromine. Figure S8. Contrast origin of the AFM images. (a,e) Atomic height color maps, (b,f) charge densities, (c,g) electrostatic potentials at 3 Å imaging distance, and (d,h) simulated force maps at 3.51 Å imaging distance for the Br···Br···Br (a-d) and Br···Br···I (e-h) windmill structures on the Cu(111) surface as calculated by DFT and our HR-AFM method. The atomic height map in panel (e) clearly shows that the iodine atom is downshifted towards the Cu(111) surface, which leads to lower Pauli repulsion between the CO-tip and the iodine atom. The charge densities in (b,f) reveal the characteristic large electron density around the halogens (yellow regions), which is higher for the bromine atoms than for the iodine atoms. Since the CO-tip has a slight negative charge at its apex, a lower repulsive electrostatic force for the iodine is found compared to bromine atoms. Both effects, i.e. the lower electrostatic repulsion and the lower Pauli repulsion lead to a less bright feature for the iodine atoms in the HR-AFM image (see panel h). The oval shape of the charge density at the halogens is difficult to discern from the images in (b,f). However, the electrostatic potential shown in (c,g) clearly reveals negative oval shaped areas at the halogen atoms (red regions). These repulsive regions are in concordance with the "negative belts" around the halogens (see Figure 1a). The surface potential at the caps of the halogens (at the -holes) is less negative (less repulsive). The relaxation of the probe enhances the asymmetry introduced by the electrostatic potential resulting in oval shaped features in the HR-AFM images (d,h) at the halogens. These oval contrast features are a direct fingerprint of the -holes of the halogens.  Figure  S6 and Ref. 1 for BrparaImeta-TP on Cu(111)]. The simulated images clearly reproduce the observed twisting angles of the phenyl rings, the different brightness of iodine and bromine (iodine appears darker due to the smaller I-Cu distance and lower electrostatic repulsion), and the oval features of the halogens (see explanation in Figure S8). Furthermore, as a consequence of the flexibility of the CO-tip the features change with decreasing tip-sample distance. This is also well reproduced by the simulated images. For example, the halogens appear at large distances as oval features. At smaller distances these features turn into sharp lines, which is in remarkable agreement with our experimental scans [see Figure S6d-f for IparaBrmeta-TP on Cu (111)].

Further details of first-principles calculations and AFM theoretical images
To reproduce the experimental systems based on dimers and trimers of BrparaImeta-TP and IparaBrmeta-TP molecules on Cu(111), we use in our DFT simulations smaller size halobenzene molecules (i.e. bromobenzene and iodobenzene). These molecules adsorbed on a Cu(111) slab displaying a triangular arrangement are suitable models to accurately reproduce the experimental observations of BrparaImeta-TP and BrparaImeta-TP/IparaBrmeta-TP trimers. In particular, we study two windmill structures on Cu(111): One formed by three bromobenzene molecules and another constituted by two bromobenzene molecules plus one iodobenzene molecule. In both cases halogen atoms prefer to occupy bridge positions while the benzene rings stay on fcc-hollow positions. We checked that this configuration is more energetically favorable than other highly symmetrical choices, in agreement with previous calculations of single-molecule adsorption 2 .
These structures were fully relaxed following a conjugate gradient algorithm until forces upon atoms were smaller than 0.01 eVÅ -1 while each electronic self-consistent loop was calculated with a precision of 10 -5 eV. The copper substrate was modeled with a rectangular three-layer slab of 21.80 Å × 17.62 Å, 5√3 × 7, 70 atoms per layer). The atoms of the deepest layer were kept fixed in their bulk positions while those of the two uppermost layers were allowed to relax together with all the atoms belonging to the molecules. A vertical vacuum region slightly larger than 23 Å was established between periodical images and a dipole correction along the z-axis was also used. The reciprocal space was sampled with a 3 × 3 × 1 Monkhorst-Pack mesh 3 during the structural minimization. Similar optimization procedures were carried out for the gas-phase systems and the single molecules adsorbed on Cu(111).
We get adsorption energies of 1.13 eV and 1.32 eV for the bromobenzene and iodobenzene respectively. The bromine atom is farther from the surface than the rest of the molecule while iodine is shifted down, see Figures S8 and S9. The formation energy of the Br···Br···Br (Br···Br···I) windmill on the gas phase is -0.199 eV (-0.217 eV), while for the windmill on the Cu(111) we obtain -3.493 eV (-3.654 eV).
We approximated the intermolecular interaction of the windmill on the surface by subtracting the adsorption energy of each molecule to the formation energy of the structure (see Table S1). Furthermore, we calculated two of the contributions involved in the reduction of the intermolecular interaction induced by the adsorption. The energy penalty to move the molecules from its optimal adsorption configuration, E(mol-subs), and the intermolecular energy variation, E rel (mol-mol), due to the formation of the windmill with the atomic positions of the molecules on the surface instead of their optimal configuration on the gas phase. With this energy we would like to quantify both how much intermolecular energy is lost due to shift of the iodine atom out of the molecular plane, and the energy lost due to the adjustment of the intermolecular distance of the windmill on the surface, slightly different from the result on the gas phase. We have calculated them by subtracting the energy of the gas phase windmill in their optimal atomic configuration from the energy of windmill, without the surface, but in the atomic configuration of the adsorbed one. As shown in Table S1 both contributions cannot explain the intermolecular energy change suffered by the molecules upon absorption on Cu(111).

Intermolecular Energies (meV) Br···Br···Br Br···Br···I
Windmill gas phase -199 -217 Windmill on Cu(111) -100 -70 E(mol-subs) 3 6 Erel(mol-mol) 0 7 From the DFT calculations described before, we use our recently developed method to simulate the HR-AFM images. A detailed description of this method can be found in our previous work. [4][5][6] It is based on a total potential ( ), constituted by four different contributions: electrostatic , short range , van der Waals and tilt component . Essentially, the model describes the electrostatic and short-range interactions in terms of two physical inputs: the total charge densities of the tip and the sample ( and ! ), and the electrostatic potential of the sample " ! which are obtained from a previous plane-wave calculation of the system without the tip. These contributions are given by the following expressions: where 6 and 2 are parameters that are obtained by a fitting procedure of force curves for each specific system 5,6 . In our particular case 6 = 1.1 and 2 =26.84 [eV] (Br···Br···Br) or 2 =32.67 [eV] (I···Br···Br). On the other hand, the van der Waals contribution can be directly extracted from the D3 estimation provided in the DFT calculation. Finally, the tilt potential, which accounts for the possible tip's orientations, is given by where 9 : is a spring constant that can be adjusted to improve the matching with the experimental images. In our case we have set 9 : = 0.46 Jm B . Notice that, instead of performing a direct minimization of the CO tilting taking into account the orientation change of the whole CO charge density as done in previous works 4-6 , we have proceed with a more efficient approximation valid for small angles. Similarly to P. Hapala, et al. 7,8 , we minimize the total energy for small displacements in the <> around the initial position of the apex, considering a rigid tip and restricting the movement to the original plane. This approach speeds up the simulations an order of magnitude with small accuracy loss. All AFM images are calculated without including the substrate explicitly because molecules are physisorbed showing minor discrepancies in the charge density and electrostatic potential when the Cu(111) slab is taken into account. We have chosen the simulated force maps to compare with the experimental AFM images instead of the gradient force images as, with the typical amplitudes used on the experiments, non-negligible variations on the gradient with respect to the tip-sample distance are observed at the distances run by the tip in one oscillation cycle.
Synthesis of BrparaImeta-TP (the procedure has been reported before in Ref. 9 ) General: The chemicals were used as purchased from Sigma-Aldrich, Acros Organics, Alfa Aesar and TCI Europe. Anhydrous solvents were purchased from Acros Organics or obtained from a MBRAUN solvent purification system MB-SPS-800. Products were purified on a Büchi Sepacore® flash chromatography system X-50 (medium pressure liquid chromatography (MPLC)) with Sepacore® Flash cartridges (particle size: 40 -63 µm). NMR spectra were measured on a Bruker Avance II 200 MHz, Avance II 400 MHz or Avance III 600 MHz spectrometer. The 1 H (7.26 ppm) respectively 13 C (77.16 ppm) chemical shift of CDCl3 was used as reference. EI-MS spectra were measured on a Finnigan MAT 95 or a GC-MS HP 5890 with a HP 5971 mass detector. Elemental analyses (EA) were measured on a Thermo FlashEA-1112.

Figure S10. Synthesis of 4-Bromo-3''iodo-p-terphenyl.
Reaction was carried out under an atmosphere of nitrogen, but without dried glassware. No dry solvents were used.

Chemicals:
The chemicals were purchased from Sigma-Aldrich, Acros Organics, Alfa Aesar and TCI Europe. Anhydrous solvents were purchased from Acros Organics. Deuterated solvents were purchased from Euriso -Top GmbH. Solids were dried over Sicapent® and under high vacuum if necessary. Technical grade solvents, used during work-up and purification, were distilled prior to use. 4-Bromo-4'-(trimethylsilyl)-biphenyl (SI-1) was synthesized according to literature. 11

NMR:
NMR spectra were measured on a Avance II 400 MHz, Avance III 400 MHz HD or Avance III 600 MHz spectrometer at 25°C. The 1 H (7.26ppm) or 13 C (77.16ppm) chemical shift of internal residual CHCl3 from CDCl3 was used as reference.

MS:
EI-MS spectra were measured on GC-MS HP 5890 with a HP 5971 mass detector. APCI-MS spectra were measured on a Bruker Micro Mikro-TOF Elemental Analysis: Elemental analysis was performed on a CHN-Analysator: Thermo FlashEA -1112 Series. Samples were weighed on a Mettler Toledo UMX-2 balance.