The key role of chirality and peripheral substitution in the columnar organization of bowl-shaped subphthalocyanines

The columnar arrangement of bowl-shaped aromatics is a promising strategy for producing high-performing semiconductors. However, the structural factors that dictate the self-assembly of these molecules remain poorly understood. Herein, we show how chirality and peripheral substitution affect the columnar assembly of subphthalocyanines (SubPcs) in solution. Both aspects are found to influence the structure, stability, and formation mechanism of the supramolecular polymer obtained. Whereas enantiopure tri-substituted SubPcs cooperatively polymerize into homochiral head-to-tail arrays, racemic mixtures socially self-sort, leading to heterochiral columnar polymers. In sharp contrast, hexa-substituted SubPcs polymerize following an isodesmic mechanism, producing highly robust columnar systems. As elucidated by molecular dynamics calculations, the conformational flexibility of these SubPcs, as well as the number of peripheral groups able to intermolecularly interact, underlie these significant differences. The results presented herein pave the way for the realistic application of bowl-shaped π-compounds.


Materials and methods
The monitoring of the reactions was carried out by thin layer chromatography (TLC), employing aluminum sheets coated with silica gel type 60 F254 (0.2 mm thick, Merck).The analysis of the TLCs was performed with an UV lamp of 254 and 365 nm.Purification and separation of the synthesized products were performed by normal-phase column chromatography, using silica-gel (230-400 mesh, 0.040-0.063mm, Merck).Eluents along with the relative ratio in the case of solvent mixtures are indicated for each particular case.Nuclear magnetic resonance spectra ( 1 H-, 13 C-, 11 B-, and 19 F-NMR) were recorded on Bruker AV-300, Bruker AV III HD 400 MHz, or Bruker DRX-500 spectrometers.The deuterated solvent employed in each case is indicated within brackets, and its residual peak was used to calibrate the spectra using literature reference δ ppm values. 1 All the NMR spectra were recorded at room temperature.High-resolution mass spectra (HRMS) were obtained at the Interdepartmental Investigation Service of UAM or IOC mass spectrometry lab of University of Würzburg, employing matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) using a Bruker-Ultraflex-III spectrometer, with a Nd:YAG laser operating at 355 nm or an ultrafleXtreme spectrometer, or an ESI TOF using a Bruker Daltonics microTOF focus instrument.The matrixes and internal references employed are indicated for each spectrum.Infrared spectra were recorded in solid state on a Bruker Vector 22 spectrophotometer.Ultraviolet and visible (UV-Vis) spectra were recorded using solvents of spectroscopic grade in the Organic Chemistry Department of UAM employing JASCO-V660 and PerkinElmer Lambda 2 dual beam absorption spectrophotometers.The logarithm of the molar extinction coefficient (ε) is indicated in brackets for each maximum.Likewise, fluorescence measurements were carried out with a JASCO-V8600 spectrofluorometer and a Horiba Jobin Yvon FluoroMax-3 emission spectrometer.NIR emission spectra were recorded on a Horiba Jobin Yvon FluoroLog3 spectrometer using a 450 W xenon lamp and a Symphony InGaAs array detector in combination with an iHR320 imaging spectrometer.CD spectra were recorded with a JASCO V-815 equipment.Resolution of racemic I 3 -SubPc-Cl was carried out by chiral highperformance liquid chromatography (HPLC) using an Agilent 1200 equipment with a semipreparative Daicel Chiralpak IC column (10 mm ø x 20 mm).Chemicals were purchased from commercial suppliers and used without further purification.FTIR spectra in solution were recorded on a Jasco FT-IR4600 spectrometer using a CaF 2 cell with a path length of 0.1 nm.Dry solvents were purchased from commercial suppliers in anhydrous grade or thoroughly dried before use employing standard methods.Solid, hygroscopic reagents were dried in a vacuum oven before use.Atomic force microscopy (AFM) images were recorded with a Ntegra Prima (NT-MDT) instrument at the Instituto Madrileño de Estudios Avanzados de Nanociencia (IMDEA), in tapping mode.S-3

O
In a 25 mL Schlenk flask, equipped with a magnetic stirrer, 8 (17 mg, 0.013 mmol), PdCl 2 (PPh 3 ) 2 (3.7 mg, 0.0053 mmol), CuI (2.0 mg, 0.011 mmol), and N-H (73 mg, 0.095 mmol) were placed under argon atmosphere.Then, 2.6 mL of a mixture of THF/NEt 3 5:1, which was deoxygenated via three freeze-pump-thaw cycles, was added and the resulting mixture was stirred first at 50 ºC for 30 min, and then at room temperature overnight.After that, the crude was dissolved in DCM and passed through a short celite plug.The solvent was removed by vacuum distillation and the resulting dark solid was purified by column chromatography on silica gel using DCM/MeOH 30:1 as eluent and by size exclusion chromatography on Bio-beads using CHCl 3 as eluent.Upon recrystallization from methanol, 4 was obtained as a green solid (44 mg, 0.0084 mmol).Yield: 64 %;

NMR Studies
The aggregation capability of 1-Rac and 2 was further assessed by NMR spectroscopy.To this end, 1 H NMR spectra of both compounds in TCE were recorded upon increasing the temperature (258 K to 368 K).We employed TCE since in MCH at this high concentration (0.5 mM) both compounds exist in a fully aggregated state (NMR silent).As shown in Figure S5.1, both VT NMR spectra reveal a broadening and shifting of the signals corresponding to the SubPc aromatic core, suggesting headto-tail - stacking.

Nucleation-elongation model for cooperative supramolecular polymerization from variable temperature experiments
The model developed by Ten Eikelder, Markvoort, Meijer and co-workers 7 extends the nucleationelongation equilibrium models designed to describe the growth of supramolecular homopolymers to the case of two monomers and aggregate types and can be applied to symmetric or non-symmetric supramolecular copolymerizations.
A cooperative supramolecular mechanism can be divided into a nucleation (nucleus size of 2) and an elongation phase.The values of T e , ΔH°n, ΔH°, and ΔS° can be obtained from a non-linear leastsquare analysis of the experimental melting curves and the equilibrium constants associated to the nucleation (K n ) and elongation (K e ) phases as well as the cooperativity factor (σ) can be calculated using equations 1, 2, and 3, respectively (see Table 1 in the text). (1)

Nucleation-elongation model for cooperative supramolecular polymerizations from good solvent experiments
In this equilibrium model, the monomer addition steps in the nucleation regime are described by an equilibrium nucleation constant K n with a cooperative parameter (σ): 8 (4) The elongation equilibrium constant K e is defined via: (5) where ΔGº´ is the Gibbs free energy gain upon monomer addition, R the gas constant, and T the temperature.According to denaturation models, the Gibbs free energy is assumed to be linearly dependent on the volume fraction of the good solvent (χ THF ): (6) ΔGº represents the Gibbs free energy gain upon monomer addition in the pure solvent (MCH) and the dependence of ΔGº´ on χ THF is described by the m parameter, which characterizes the ability of the good solvent to associate with the monomer thereby destabilizing the supramolecular aggregated species.

Force Field Parametrization
The quantum mechanically derived force field (QMDFF) used to perform the molecular dynamics calculations on SubPcs 1 and 2 was obtained by parametrizing the molecular fragments showed in Figure S7.1, which were calculated at the density functional theory (DFT) B3LYP/cc-PVDZ level, and transferring the parameters to the full structure.

Parametrization of Fragment 1
Fragment 1 provides all the parameters for the SubPc core, the substituent, and the amide group.In addition, the energy profiles of the four flexible dihedrals ( -) are shown in Figure S7.3.These  1  4 dihedrals include the rotation around the ethynyl group ( ) and those involving the amide group ( - ).Specifically, is represented by the C5−C7−C8−C10 dihedral, by C11−C12−NA−CA, by The total standard deviation for rigid coordinates between the DFT Hessian and the FF parametrization accounts to 4.11 × 10 -3 kJ mol -1 .The molecular structure of Fragment 1 was optimized with the FF and the resulting geometrical parameters were compared with those of the DFT minimum-energy geometry, providing a root-mean square deviation of 0.000 Å, 0.006º, and 0.678º for bond length, bond angles, and dihedrals, respectively.The displacements of all atoms between the FF and the DFT minima to only 0.085 Å.The FF results therefore show a very good fitting of the DFT structure, considering the large number of atoms in Fragment 1 (73), and validate the accuracy of the parameterization obtained.For the flexible coordinates shown in Figure 3, the fitting is also in very good agreement with DFT results.
Another six flexible dihedrals ( -) were defined and their energy profile fitted to DFT calculations The total standard deviation for rigid coordinates between the DFT Hessian and the FF parametrization accounts to 7.87 × 10 -3 kJ mol -1 .The molecular structure of Fragment 2 was optimized with the FF and the resulting geometric parameters were compared with those of the DFT minimum-energy geometry, providing a root-mean square deviation of 0.000 Å, 0.033º, and 0.463º for bond lengths, bond angles, and dihedrals, respectively.The displacements of all atoms between the FF and the DFT minima amounts to 0.028 Å, again evidencing a very good fitting to the DFT structure.S-29

Computational Details
Molecular dynamics (MD) simulations in methylcyclohexane (MCH) as solvent were performed with GROMACS 2021.3 using periodic boundary conditions. 13For SubPcs, we used the quantummechanically derived force field (QMDFF) presented in this work with the atom type selection discussed in the previous section.The corresponding FF of the enantiomer of 1 was obtained by simply flipping the sign of the equilibrium dihedral angle for rigid out-of-plane internal coordinates.
To simulate the solvent MCH, the FF was generated with the PolyParGen tool, 14 and a compressibility of 11.49 × 10 -5 bar -1 was used during the simulations.The initial structures used for monomers, dimers, and octamers were centered in a box large enough that the minimum distance between the box boundaries and the closest atom was of at least of 1.0 nm to avoid spurious interactions due to the periodic boundary conditions.Point charges were calculated by following the ESP procedure with Antechamber at the B3LYP/6-31G** level.The standard protocol for MD simulations was as following: i) energy minimization, ii) solvent equilibration around the solute, and iii) production.
For energy minimization, we used a steepest descent algorithm with 0.01 nm step size until all forces were below 1000 kJ•mol -1 •nm -1 .The equilibration of the solvent consisted of two stages of 1 and 5 ns in steps of 1 fs in which the solute was kept frozen: an initial NVT scheme fixing volume and temperature (298 K), and, subsequently, an NPT scheme where pressure (1 bar) and temperature (298 K) were kept constant.Finally, the production run (NPT scheme) consisted of 20 ns calculations in steps of 1 fs.In all cases, we used a V-rescale thermostat with damping constant of 0.1 ps and a Parrinello-Rahman pressure coupling with damping constant of 5 ps.C−H bonds were constrained with a 4th order LINCS algorithm.The cutoff radius for short-range electrostatic and van der Waals interactions was set to 1.6 nm and we used an order 4 Particle Mesh Ewald for long-range electrostatics.

Figure S5. 1 .
Figure S5.1.Changes in the aromatic region of the 1 H NMR spectra of 1-Rac (top) and 2 (bottom) in CDCl 2 CDCl 2 as a function of temperature at 0.5 mM.

Figure S5. 4 .Table S5. 1 .
Figure S5.4.Evolution of the aggregation degree ( versus T) and the corresponding global fit (solid line) to a   nucleation−elongation model at different concentrations of a) 1-Rac, b) 1-M and c) 1-P.

Figure S7. 2 .
Figure S7.2.Atom type selection.Only one substituent of the SubPc core is emphasized, the other two being equivalent by molecular symmetry.In the dodecyloxy groups, the labeling of the hydrogen atoms (not shown for simplicity) is analogous to that used for the carbon atoms they are bonded to, i.e., HT5, HT6, HT7, and HTM respectively denote hydrogens attached to CT5, CT6, CT7, and CTM.

Figure S7. 3 .
Figure S7.3.Flexible dihedrals -defined on Fragment 1 and their energy profiles computed with DFT (circles)  1  4 or fitted (lines).The colored spheres in the structural model represent the atoms involved in each specific dihedral.Note that the energy scale on the y axis is different for each panel.

Figure S7. 4 .
Figure S7.4.Flexible dihedrals -defined on Fragment 2 and their energy profiles computed with DFT (circles)  5  10 or fitted (lines).The colored spheres on the structural model represent the atoms involved in each specific dihedral.Note that the energy scale on the y axis may differ for each panel.

Figure S7. 5 .
Figure S7.5.Structural models for the columnar (left) and partially dissociated (right) structures found for the 1 2 -M dimer along the MD trajectory in MCH.Top: top view.Bottom: side view.The values of the F•••B and NH•••O distances (in Å) are displayed, and the alkyl chains are hidden for better visualization.

Figure S7. 6 .
Figure S7.6.Structural snapshot of the 1 8(2) -Rac central dimer, extracted from the 1 8 octamer MD trajectory (left: top view; right: side view), emphasizing the competition between NH•••O bonds (top substituent), -stacking between phenyl rings (right substituent), and -stacking in addition to a rotation of the amide group (left substituent).The values of the NH•••O distances are displayed, and the alkyl chains are hidden for better visualization.

Figure S7. 7 .
Figure S7.7.Top views of the minimum-energy QMDFF-optimized structures computed in MCH for the columnar structure of homochiral 1 8(2) -M (left) and heterochiral 1 8(2) -Rac (right).The most favorable orientation to form strong NH•••O bonds between amides is compatible with the π-stacking of the SubPc cores and also of the substituents in 1-M but not in 1-Rac, leading to a less stable structure for the latter.

Figure S7. 8 .
Figure S7.8.a) Minimum-energy QMDFF-optimized structure computed in MCH for the columnar regular selfassembly of the homochiral 1 8 -M octamer showing all atoms.Side (b) and top (c) views of the QMDFF-optimized structure of 1 8 -M hiding non-H-bonded hydrogens and with lateral alkyl chains as wires for visualization.d) Side view of a structural snapshot extracted from the 1 8 -M octamer MD trajectory in MCH.

Figure S7. 10 .
Figure S7.10.Side (a) and top (b) views of the minimum-energy QMDFF-optimized structure computed in MCH for the columnar regular self-assembly of the 2 8 octamer hiding non-H-bonded hydrogens and with lateral alkyl chains as wires for visualization.c) Side view of a structural snapshot extracted from the 2 8 octamer MD trajectory in MCH.

Figure S7. 11 .
Figure S7.11.Structural models for the optimized tail-to-tail dimers of 1-M (left) and 2 (right) at FF level.Top: top view.Bottom: side view emphasizing the SubPc moieties.The value of the B•••B distance is displayed.