The anti-aromatic dianion and aromatic tetraanion of [18]annulene

π-Conjugated macrocycles behave differently from analogous linear chains because their electronic wavefunctions resemble a quantum particle on a ring, leading to aromaticity or anti-aromaticity. [18]Annulene, (CH)18, is the archetypal non-benzenoid aromatic hydrocarbon. Molecules with circuits of 4n + 2 π electrons, such as [18]annulene (n = 4), are aromatic, with enhanced stability and diatropic ring currents (magnetic shielding inside the ring), whereas those with 4n π electrons, such as the dianion of [18]annulene, are expected to be anti-aromatic and exhibit the opposite behaviour. Here we use 1H NMR spectroscopy to re-evaluate the structure of the [18]annulene dianion. We also show that it can be reduced further to an aromatic tetraanion, which has the same shape as the dianion. The crystal structure of the tetraanion lithium salt confirms its geometry and reveals a metallocene-like sandwich, with five Li+ cations intercalated between two [18]annulene tetraanions. We also report a heteroleptic sandwich, with [18]annulene and corannulene tetraanion decks.


General Experimental Methods
Tetrahydrofuran (THF) and hexanes were purchased as dry from Sigma Aldrich and then freshly distilled over sodium with benzophenone as indicator before use.Other solvents and reagents were purchased from the same company and used directly.THF-d8 for NMR purposes was purchased from Sigma Aldrich and distilled over sodium/potassium alloy in the presence of benzophenone under reduced pressure and then stored inside an argon-filled glovebox.Alkali metals were purchased from Sigma Aldrich, washed with hexanes to remove mineral oil and stored inside a glovebox under argon.All reactions and follow-up procedures were performed in custom-made glass systems under an atmosphere of argon.
2.1 Synthesis of hexadehydro [18]annulene S3 Two previously reported procedures [1,2] were combined and modified: Dialkyne S2 (2.0 g, 9.07 mmol) was deprotected first by vigorous stirring at 45 °C in MeOH (50 mL, directly from a bottle) with K2CO3 (3.42 g, 13.1 mmol, 1.44 equiv.)for 30 min in a closed flask under air and then allowed to reach room temperature (20  °C).The solution was decanted to remove solid salts, the salts were washed with MeOH (5 mL) and solutions combined.The resulting solution of the deprotected alkyne in MeOH was added in one portion (important!) to a solution of copper(II) acetate monohydrate (12.0 g, 59.9 mmol, 6.6 equiv.) in pyridine (200 mL, directly from a bottle) and the mixture was stirred vigorously, protected from light at 20 °C for 2 h in an open flask under air.After that time, aqueous 20% H2SO4 solution (100 mL) was added.The solution became warm and it was allowed to cool to room temperature.The solution was filtered on a Büchner funnel to remove precipitated copper salts, then transferred to a separatory funnel, followed by addition of water (200 mL) and saturated brine (200 mL) and shaken.Diethyl ether (120 mL) was added and the mixture was shaken.After that, another portion of diethyl ether (120 mL) was added and the funnel was slowly inverted 15 times, to avoid formation of emulsion.Two layers were visible.If they were not formed, more diethyl ether (100 mL) was added and the funnel was gently inverted a few more times.The bottom, blue aqueous layer was separated from the upper ether brown/black-colored phase containing some black particles.The ether phase was washed with 20% aqueous H2SO4 (5 ´ 100 mL) to remove the pyridine, followed by washing with water (3 ´ 200 mL).The solution was dried over anhydrous Na2SO4 and filtered.
Petroleum ether (20 mL, 40-60 °C fraction) was added to the resulting solution.The mixture was concentrated on a rotary evaporator equipped with dry ice cooling system, using an ice-cooled water bath, to remove diethyl ether, while not allowing it to become dry, otherwise decomposition is observed.Addition of petroleum ether is important to prevent all the solvent from evaporating.The receiving flask was constantly inside a bucket filled with dry ice during the evaporation process and was emptied every few minutes.
Silica gel column chromatography (4:1 petrol/EtOAc) was performed.The first fraction is orange/red and contains some very quickly decomposing compound (black decomposition product upon drying or concentrating), which is probably a cyclic dimer, but all attempts to characterize it were met with failure due to its instability.The second, intensely yellow fraction contains the desired cyclic trimer.
The yellow fraction containing trimer was collected.Benzene (15 mL) was added, hexane and ethyl acetate were removed on a rotary evaporator in the same conditions as described above.Addition of benzene is important because, in the absence of benzene, the cyclic trimer decomposes when it precipitates out of the solution.However, when benzene is added, it forms stable cocrystals [1] that precipitate during evaporation.The obtained orange-colored solution in benzene (ca.15 mL) was transferred to a vial and stored in a freezer.Yield of S3: 120 mg (18%) (because of the rapid decomposition in the solid state, this yield was determined by taking aliquot of benzene solution, weighing it and evaporating to dryness (decomposition to black material), and weighing it again, calculating the overall yield).

Synthesis of [18]annulene 1
The compound was prepared using a modified literature procedure. [1]To a 100 mL round-bottom flask, a solution of hexadehydro [18]annulene (118 mg, 0.53 mmol) in benzene (10 mL) was added, followed by an additional portion of benzene (40 mL) and quinoline (50 μL, 0.42 mmol).Nitrogen was bubbled through the suspension for 10 min to remove most of the oxygen.After that, the Lindlar catalyst (5% Pd/CaCO3 poisoned with Pb, 49.0 mg, Sigma Aldrich) was added and nitrogen was bubbled for 5 min.After that, hydrogen gas from a balloon was bubbled through the solution with vigorous stirring.The progress of the reaction was monitored by TLC.Usually, the volume of two balloons was found to be enough (1 h reaction time).The mixture should not be allowed to react for too long, otherwise the product gets over-reduced.The yellowgreen mixture was concentrated in vacuo and silica gel chromatography was performed (23:1:1 pentane/cyclohexane/benzene).The fraction containing the product has yellow-green color and the progress of the chromatography can be easily monitored by UV-vis spectroscopy, collecting fractions that give characteristic bands.The product 1 was dried, which provided amber solid (38 mg, 31%).Spectroscopic data were consistent with the literature. [1]The product was stored by adding benzene (10 mL) and the solution was kept at -20 °C in a freezer.

Crystallization of [Li
THF (1.0 mL) was added to a customized glass system (Supplementary Figure 1) [4] containing [18]annulene 1 (3.0 mg, 0.012 mmol) and excess of Li metal (10 mg, 1.4 mmol).The reaction mixture was stirred at 25 °C under argon for 1.5 h.The initial yellow color (neutral macrocycle) changed to brownish-green in 15 minutes.It gradually became grass green over the next 30 min.After another 30 min, the color changed to dark green and the mixture was filtered and the green filtrate was layered with anhydrous hexanes (2.0 mL).The ampule was sealed under argon and stored at 5 °C.Dark green plates started appearing after two days.After a week, crystals were taken for X-ray measurements.Yield: 1.6 mg, 20%.

Measurement and refinement details
Data collection on 1 was performed at the Department of Chemistry, University of Oxford using Rigaku Super Nova A diffractometer at 150 K and then reduced using CrysAlisPro.
Data collection for S3 and 1•2•Li8 was performed at the Department of Chemistry, University of Oxford using Rigaku Synergy DW diffractometer equipped with rotating anode source and HyPix-Arc 150° detector at 100(2) K and subsequently reduced using CrysAlisPro.
Crystallization of S3 was performed by vapor diffusion of MeOH into solution in toluene overnight and provided a new polymorph of hexadehydro [18]annulene S3.Measurement was performed ca.16 h after setting up the crystallization as the crystals decompose quickly in the solid state; most of the material turned black but some crystals were still yellow and diffracted well.
Crystals of [18]annulene 1 were obtained as amber plates by vapor diffusion of methanol into toluene solution overnight.][12] The whole molecule exhibits disorder (ratio 84:16) and was modelled using SADI restraints with default strength applied for appropriate 1,2-and 1,3-distances.The anisotropic displacement parameters of the disordered molecules were restrained to have the same Uij components by using the restraint SIMU, with a standard uncertainty of 0.01 Å 2 .Displacement parameters in the direction of the bonds were restrained using RIGU restraints with default strength.
12•Li8 was crystallized as described in Section 2.5.In this structure model, one annulene tetra-anion and part of the other, as well as four THF molecules, were found to be disordered.This disordered molecule was modelled with two orientations with their relative occupancies refined.The geometries of the disordered parts were restrained to be similar.The anisotropic displacement parameters of the disordered molecules were restrained to have the same Uij components, with a standard uncertainty of 0.01 Å 2 .
1•2•Li8 was crystallized as described in Section 2.7.Several crystals from different batches were tested for diffraction and the data were collected using a high flux rotating anode source.The crystals were of poor quality.Data on the best crystal were collected up to 0.9 Å resolution due to lack of diffraction spots beyond.
The crystal was split.Moreover, beam damage was observed during data collection.Multiple data processing was attempted in order to get the best parameters and at the end, the structure was solved from the .hkl4file generated during attempts of treating the data as twinned, with the second component much weaker and with much higher Rint and therefore not included in the refinement.The structure was found to be severely disordered.Both hydrocarbons in the sandwiched complex as well as lithium cations in between them and some of the coordinated THF molecules were treated as disordered and were modelled using SADI restraints with default strength applied for appropriate 1,2-and 1,3-distances (not including lithium cations in the list of these distance restraints).The anisotropic displacement parameters of the disordered parts were restrained to have the same Uij components by using the restraint SIMU, with a standard uncertainty of 0.01 Å 2 .Displacement parameters in the direction of the bonds in the disordered parts were restrained using RIGU restraints with default strength.Solvent masking (Olex2 subroutine) was used to account for electron density from severely disordered interstitial THF molecule, indicating 304 electrons which corresponds to ca. 7.6 molecules per unit cell.Due to poor data quality and severe disorder in the structure we do not discuss this structure in detail (bond lengths etc.).Despite these problems, the obtained model provides useful information, correlates well with NMR data and therefore is fit for purpose and valuable.The HOMA index can be calculated using the following equation: [14][15][16][17][18] HOMA where Ri and Ropt are the i th bond length of the C-C bond in the analyzed ring and the bond length of benzene ring (Ropt = 1.388Å), respectively.n is the number of C-C bonds in the analyzed ring and α = 257.7 Å -2 is a normalization factor that gives HOMA value of 1 for perfect aromatic benzene ring and a HOMA value of 0 for an alternating nonaromatic Kekulé cyclohexatriene ring.
The uncertainty (standard deviation) in the HOMA value (sH) was calculated from the uncertainty of the bond lengths (sR,i) using the equation: Supplementary Table S2.Analysis of HOMA and BLA in lithium salt of [18]annulene tetra-anion 12•Li8.
For each carbon, BLA is calculated as the absolute value of the difference between the two C-C bonds associated with that atom.For this ring, BLA = 0.014 ± 0.010 Å (here the uncertainty is the standard deviations of the bond length differences) Overall average BLA = 0.014 ± 0.010 Å For both rings together, HOMA = 0.841 ± 0.008

Theoretical calculations
7.1.Coulombic repulsion in different conformers of 1 2-was qualitatively evaluated by calculating atomic charges using the Natural Population Analysis [19] in the natural bond orbital (NBO) package supplied with Gaussian16 [20] .Results are shown in Fig. S65.In conformer B, the most strongly negatively charged atoms (-0.43 |e|) are separated by d1B = 7.8 Å, while analogous atoms in A are separated by 6.3 Å.The negative charge is also spread more uniformly across the ring in the B conformer, with a ~10% smaller mean absolute deviation (MADA = 0.099 |e|; MADB = 0.088 |e|).

configuration 6 .
HOMA and BLA CalculationsHOMA (Harmonic Oscillator Model of Aromaticity) is one of the simplest and most widely used indices for describing aromaticity based on molecular geometry.It uses the C-C bond length in benzene as a standard of perfect aromaticity.