Beyond the Scholl reaction – one-step planarization and edge chlorination of nanographenes by mechanochemistry

The edge chlorination of the benchmark nanographenes triphenylene and hexa-peri-hexabenzocoronene is conducted mechanochemically. This approach overcomes solubility limitations and eliminates the need for elaborate chlorination conditions. Additionally, the planarization of oligophenylenes and their edge-chlorination can be combined in a one-pot approach requiring as little as 60 minutes.


General methods
Ball mill syntheses were carried out in a planetary ball mill (Fritsch Pulverisette 7 premium line) at a rotational speed of 800 rpm. The reactions were performed in a 20 mL milling vessel made of ZrO2 with 10 milling balls (ZrO2, 10 mm in diameter) if not stated otherwise. Additionally, a Retsch EMAX high-energy ball mill with a 50 ml ZrO2 milling vessel and 25 milling balls (ZrO2, 10 mm in diameter) with a frequency of 1200 rpm and 60 minutes was used.
Bulking material: A bulking agent is an inert additive to the reaction. The size of the milling vessel (20 or 50 mL in the present work) requires a minimum amount of reactants in the milling vessel (ball-to-powder ratio), otherwise the milling balls will collide with each other without transferring the collision energy to the reactants, which causes abrasion of the milling material. We used NaCl as an inert bulking material. This method was first described by Konnert et al. 1 Since we used a total mass of 2 g in the 20 ml milling vessel, we added the amount of NaCl that was necessary to reach this mass. As an alternative we could have also increased the amount of reactants to reach the required mass. Trichloroisocyanuric acid (TCI, 95%), Triphenylene (Alfa Aesar, 98%) and Sodium chloride (Güssig, >98% ) were used as received.

Summary of the results with alternative Lewis acids and chlorination methods
Tab. S1. Reaction conditions and yields of the chlorination of hexabenzocoronene (3) using different Lewis acids (S1-S4) and using different chlorination methods (S5-S8 In a typical synthesis, 0.1 g HBC 1 (0.192 mmol) and an excess of 1.9 g metal chloride (e.g. iron(III) chloride (11.75 mmol, 61 eq.)) were transferred into a 20 mL zirconium oxide grinding jar with ten zirconium oxide 10 mm-diameter grinding balls (3.19 g each). The mixture was then milled for 60 min at 800 rpm in a Fritsch Pulverisette 7 premium line planetary ball mill. After the reaction, the grinding jar was opened and the reaction mixture was poured into water. The crude product was consequently washed with water, methanol and ethanol. The soluble fraction was extracted with CHCl3 which was consequently evaporated and the solid was dried at 80°C. Chlorinated HBC 2 was obtained as a dark red solid (107.5 mg, yield: 49 %). MALDI-TOF (Fig. 3) (TCNQ): 937.14, calc. 1142.60.
Other metal chlorides used with the same mass ratio: CuCl2, AlCl3, MoCl5

Method B: ICl, AlCl3
In a typical synthesis, 0.03 g HBC 7 (0.057 mmol), an excess of 0.563 g aluminum chloride (4.133 mmol, 72 eq.), 1.62 g iodine monochloride (0.5 ml, 9.98 mmol, 174 eq.) and 1.4 g of sodium chloride as bulking material were transferred into a 20 mL zirconium oxide grinding jar with ten zirconium oxide 10 mm-diameter grinding balls (3.19 g each). The mixture was then milled for 60 min at 800 rpm in a Fritsch Pulverisette 7 premium line planetary ball mill. After the reaction, the grinding jar was opened and the reaction mixture was poured into water. The crude product was consequently washed with water, methanol and ethanol. The soluble fraction was extracted with CHCl3 which was consequently evaporated and the solid was dried at 80°C. Chlorinated HBC 2 was obtained as a dark red solid (13.7 mg, yield: 21 %). MALDI-TOF (Fig. S5 In a typical synthesis, 0.05 g HBC 1 (0.096 mmol), an excess of 1.855 g aluminum chloride (13.91 mmol, 145 eq.), 0.290 g trichloroisocyanuric acid (1.25 mmol, 13 eq.) and 1 mL of concentrated sulphuric acid were transferred into a 20 mL zirconium oxide grinding jar with ten zirconium oxide 10 mm-diameter grinding balls (3.19 g each). The mixture was then milled for 60 min at 800 rpm in a Fritsch Pulverisette 7 premium line planetary ball mill. After the reaction, the grinding jar was opened and the reaction mixture was poured into water. The crude product was consequently washed with water, methanol and ethanol. The soluble fraction was extracted with CHCl3 which was consequently evaporated and the solid was dried at 80°C. Chlorinated HBC 2 was obtained as a dark red solid (19 mg, yield: 17 %). MALDI-TOF (Fig. S6) (TCNQ): 898.21, calc. 1142.60.

Method D: NCS, FeCl3
In a typical synthesis, 0.03 g HBC 1 (0.057 mmol), an excess of 0.335 g iron(III) chloride (2.067 mmol, 36 eq.), 0.276 g N-chlorosuccinimide (2.067 mmol, 36 eq.) and 1.4 g of sodium chloride as bulking material were transferred into a 20 mL zirconium oxide grinding jar with ten zirconium oxide 10 mm-diameter grinding balls (3.19 g each). The mixture was then milled for 60 min at 800 rpm in a Fritsch Pulverisette 7 premium line planetary ball mill. After the reaction, the grinding jar was opened and the reaction mixture was poured into water. The crude product was consequently washed with water, methanol and ethanol. The soluble fraction was extracted with CHCl3 which was consequently evaporated and the solid was

Method E: Oxone
In a typical synthesis, 0.1 g HBC 1 (0.192 mmol), an excess of 1.413 g Oxone (KHSO5 · 0.5KHSO4 · 0.5K2SO4) (4.596 mmol, 24 eq.) and 0.487 g sodium chloride (8.338 mmol, 44 eq.) were transferred into a 20 mL zirconium oxide grinding jar with ten zirconium oxide 10 mm-diameter grinding balls (3.19 g each). The mixture was then milled for 60 min at 800 rpm in a Fritsch Pulverisette 7 premium line planetary ball mill. After the reaction, the grinding jar was opened and the reaction mixture was poured into water. The crude product was consequently washed with water, methanol and ethanol. The soluble fraction was extracted with CHCl3 which was consequently evaporated and the solid was dried at 80°C. Chlorinated HBC

Chlorination of Triphenylene
Method A: Metal chloride 0.1 g Triphenylene 3 (0.438 mmol) and an excess of 1.9 g iron(III) chloride (11.75 mmol, 27 eq.) were transferred into a 20 mL zirconium oxide grinding jar with ten zirconium oxide 10 mm-diameter grinding balls (3.19 g each). The mixture was then milled for 60 min at 800 rpm in a Fritsch Pulverisette 7 premium line planetary ball mill. After the reaction, the grinding jar was opened and the reaction mixture was poured into water.
The crude product was consequently washed with water, methanol and ethanol. The soluble fraction was extracted with CHCl3 which was consequently evaporated and the solid was dried at 80°C. Chlorinated Triphenylene 4 was obtained as a brown solid (146.6 mg, yield: 52 %). MALDI-TOF (Fig. 2

One-step planarization and edge-chlorination procedure
In a typical synthesis, 0.1 g hexaphenylbenzene 5 (0.187 mmol) and 2.18 g iron(III) chloride (13.44 mmol, 72 eq.) were transferred into a 20 mL zirconium oxide grinding jar with ten zirconium oxide 10 mm-diameter grinding balls (3.19 g each). The mixture was then milled for 60 min at 800 rpm in a Fritsch Pulverisette 7 premium line planetary ball mill. After the reaction, the grinding jar was opened and the reaction mixture was poured into water. The crude product was consequently washed with water, methanol and ethanol. The soluble fraction was extracted with CHCl3 which was consequently evaporated and the solid was dried at 80°C. Chlorinated HBC 2 was obtained as a dark red solid (89 mg, yield: 42%). MALDI-TOF In the case of attempted bromination, 3.98g iron(III) bromide (13.47, 72 eq.) were used.

Computational Study
For the determination of the energetically most stable oligomer structures, initial geometries for all possible oligomers were generated using a simple algorithm that functionalizes an arbitrary number of initial structures, in this case a single triphenylene structure, with an arbitrary number of functional groups, in this case 1-and 2-triphenylenyl, at every possible functionalization site, where every unique structure it generates is added to the pool of structures to be functionalized, until a certain stopping criterion is met. The stopping criterion that was used here, was the generation of the first pentamer, which was then discarded. The generated molecules are tested for uniqueness by registering their respective canonical SMILES as produced by the Open Babel library 4 and comparing with the already registered ones. The recursive nature of the algorithm ensures the generation of all possible oligomers. To increase the chance that the respective global minima are found in the geometry optimizations of the generated initial structures, rather than energetically less preferable conformers, each unique structure is cloned and rotated around the C-C-C-C dihedral angle, where the two central carbon atoms form the new bond between input structure and functional group, and the rotated structure is also added to the structure pool. Due to the nature of the algorithm, further geometries are only derived from the initially found unique structure and not from the rotated clone, which limits the effectiveness of this procedure to some degree but also the total number of initial geometries which need to be optimized. Unique structures (or rotated clones of these) with unreasonably close atom distances such as in the case of overlapping molecule segments are automatically discarded by the algorithm. In total, six dimer structures for three unique dimer SMILES, 87 trimer structures for 46 unique trimer SMILES and 2069 tetramer structures for 1122 unique tetramer SMILES were found by the algorithm.
Facing the large number of geometries to be optimized, a strategy was chosen that maximizes computational efficiency without sacrificing much of the accuracy of the results. Rather than optimizing all initial oligomer structures at the desired level of theory for the final results, they were optimized with the much faster extended tight-binding method GFN2-xTB, developed by Grimme et al. 5 , and were ranked according to their total molecular energies at the optimum geometries predicted by this method. A selection of the most favorable structures according to this ranking then provided the initial geometries for the final DFT calculations, except for the dimer case, where all six structures were optimized using both methods, serving as a small test study to verify the validity of this procedure. The relative molecular energies of the six dimer structures computed via both methods and the corresponding optimized geometries obtained from DFT calculations are displayed in Fig. S15. The geometries obtained from the GFN2-xTB method are reasonably close to the DFT-optimized ones, as seen in Fig. S16. The six optimized dimer geometries can be identified as conformer pairs of each of the three possible constitutional isomers. Regarding the two used methods, GFN2-xTB generally underestimates the energy differences between the six dimer geometries, in other words, it overstabilizes the less preferred geometries compared to DFT. While Dim-1 is predicted by both methods to be the thermodynamically most stable isomer, Dim-1a being slightly preferred over Dim-1b by both methods, there is a discrepancy regarding the stability of Dim-3a, which the GFN2-xTB method predicts to be much more stable than the DFT calculations do, up to the point where it is preferred over both conformers of Dim-2, whereas both methods are in accord about the least preferable geometry of Dim-3b. Taking these observations into account, GFN2-xTB, while its results quantitatively not agreeing with those computed via DFT, can be considered a valid method for pre-optimizing the initial geometries generated by the above described algorithm and pre-filtering the geometries that are to be further optimized using DFT, since it allows for a qualitative stability ranking of the considered structures.
To remedy the fact, that, with increasing structural complexity, the chance of optimizing towards local minima increases, the pool of trimer structures to be optimized was further supplied with the set of trimer structures that is obtained from running the above-mentioned algorithm with Dim-1a as the single starting structure, ignoring whether this results in any duplicate calculations. The twelve lowest-energy trimer geometries according to GFN2-xTB were then further optimized using DFT. Out of the resulting structures, the four best are displayed in Fig. S17. Analogously, for the tetramer structures, direct descendants of Dim-1a and Tri-1, as generated by the above-described algorithm, were added to the pool of structures to be pre-optimized using GFN2-xTB. This pool was further supplied with the set of tetramer structures that are directly generated by functionalizing Dim-1a with all fragments that can be derived from Dim-1a itself by abstracting a single hydrogen atom. Here, the 16 most favorable geometries, as predicted by GFN2-xTB, were reoptimized using DFT, the four best out of the resulting structures being displayed in Fig. S18. For the computation of the energetically most favorable chlorinated isomers of Dim-1a and HBC a modified version of the above-described algorithm was used, where instead of replacing hydrogen atoms by functional groups, these are simply substituted with chlorine atoms. While for HBC, the optimization of all 22284 unique chlorinated isomers (including non-chlorinated HBC) with GFN2-xTB was still computationally feasible, this was not the case for Dim-1a. Here, the isomer geometries were optimized in batches, where, starting with the third degree of chlorination, only the best geometries were used to seed the next generation of structures in the algorithm, starting at 20 and incrementally increasing to 50 at a substitution level of 11. To obtain reasonable energy diagrams, the ranking of the most favorable isomers per degree of chlorination is limited to direct derivatives of the best isomer of the previous degree of chlorination, both for Dim-1a and HBC. Under this constraint, the best isomer structure per degree of chlorination according to GFN2-xTB was then optimized using DFT to yield the energy diagrams shown in Fig. S19 (Dim-1a) and Fig. 5 (HBC). Note, that the relative total energies Δ , with respect to the number of chlorine atoms in the molecule displayed in these figures were computed as ΔE tot,n Cl = E tot,n Cl + n Cl (E HCl -E Cl 2 ) -E tot,0 to account for the substitution of protons with chlorine, where , is the total energy of the best structure containing chlorine atoms and and 2 are the respective total energies of hydrogen chloride and molecular chlorine, obtained from DFT calculations. For a better understanding, the energy steps between the degrees of chlorination are given in Table S2 (Dim-1a) and Table S3 (HBC). Illustrations of all the molecular structures that are represented in the two energy diagrams are displayed in Fig. S20 (Dim-1a) and Fig. S21 (HBC).