The Pink Box: Exclusive Homochiral Aromatic Stacking in a Bis-perylene Diimide Macrocycle

This work showcases chiral complementarity in aromatic stacking interactions as an effective tool to optimize the chiroptical and electrochemical properties of perylene diimides (PDIs). PDIs are a notable class of robust dye molecules and their rich photo- and electrochemistry and potential chirality make them ideal organic building blocks for chiral optoelectronic materials. By exploiting the new bay connectivity of twisted PDIs, a dynamic bis-PDI macrocycle (the “Pink Box”) is realized in which homochiral PDI–PDI π–π stacking interactions are switched on exclusively. Using a range of experimental and computational techniques, we uncover three important implications of the macrocycle’s chiral complementarity for PDI optoelectronics. First, the homochiral intramolecular π–π interactions anchor the twisted PDI units, yielding enantiomers with half-lives extended over 400-fold, from minutes to days (in solution) or years (in the solid state). Second, homochiral H-type aggregation affords the macrocycle red-shifted circularly polarized luminescence and one of the highest dissymmetry factors of any small organic molecule in solution (glum = 10–2 at 675 nm). Finally, excellent through-space PDI–PDI π-orbital overlap stabilizes PDI reduced states, akin to covalent functionalization with electron-withdrawing groups.


Acyclic bis-triazole PDI 3b
To a solution of TMS-protected bis-alkyne PDI 2b (85 mg, 118 µmol) in DCM (20 ml) was added K2CO3 (50 mg) in MeOH (10 ml). The mixture was stirred at rt for 3 min, and completion of the reaction was confirmed by TLC. A further 20 mL of DCM was added to the solution. The solution was then washed with 1 M HCl (2 x 30 mL), water (2 x 30 mL) and brine (30 ml). The organic layer was then dried over anhydrous MgSO4 and concentrated to dryness in vacuo to afford the deprotected PDI bis-alkyne which was used immediately without further purification. This PDI bis-alkyne was immediately re-dissolved in dry DCM (25 mL). To this was added 1,4bis(azidomethyl)benzene (220 mg, 1.18 mmol, 10 equiv) and Tris((1-benzyl-4triazolyl)methyl)amine (TBTA) (12 mg, 24 µmol, 0.2 equiv). The solution was then de-gassed with argon. The copper (I) catalyst Cu(CH3CN)4PF6 (9 mg, 24 µmol, 0.2 equiv) was then added and the solution was once again de-gassed with argon. The reaction was stirred at rt for 12 h. The solvent was then removed in vacuo. The resulting residue was purified by silica gel flash column chromatography (1:99 MeOH-DCM) affording the title compound (as the pure 1,7regioisomer) as a purple solid (63 mg, 66 µmol, 57%).

1,7-TMS-protected bis-alkyne PDI 5b.
To a solution of dibromo PDI 4b 3 (200 mg, 290 µmol) in 1:1 dry NEt3-toluene (30mL) under a nitrogen atmosphere was added Pd(PPh3)2Cl2 (12.94 mg, 17 µmol, 0.06 equiv), CuI (6.09 mg, 32 µmol, 0.11 equiv) and trimethylsilylacetylene (142 mg, 207 µL, 1.45 mmol, 5 equiv). The mixture was thoroughly de-gassed with nitrogen and stirred at 60 °C for 48 h. The solvent mixture was then removed in vacuo. The resulting residue was then re-dissolved in DCM (30 mL) and washed with 1M HCl (50 mL) and water (3 x 50 mL); dried over anhydrous MgSO4 and concentrated to dryness in vacuo. The resulting residue was purified by silica gel flash column chromatography (1:1 n-hexane:DCM) affording the title compound as a red solid (210mg, 290 µmol, 100%). MALDI (TOF) mass spectrum for compound 5b. An accurate mass was obtained by calibration to polyethylene glycol chains that were co-spotted with the sample and therefore also observed in the mass spectrum S15 Figure S1: Truncated 1 H NMR spectra of 1a (Toluene-d8, 400 MHz) at various temperatures ranging from 298 K (bottom spectrum) to 373 K (top spectrum), showing how the spectrum is too broad to assign at room temperature (298K) but sharpens as the temperature is increased. S16 b) Solvent dependent 1 H NMR spectroscopy of macrocycle 1a Figure S2: Partial 1 H NMR spectra of macrocycle 1a in different toluene-d8:TCE-d2 solvent mixtures (373K, 400 MHz). The spectra are aligned using an internal reference standard, poly(dimethylsiloxane), added to each sample. Peaks labelled with an asterisk correspond to the minor species (MP diastereomer), while those without an asterisk correspond to the major species (MM,PP enantiomers). The MM,PP:MP ratios are shown for each spectrum.  In toluene-d8 a NOE between the protons Hc and Hd is observed, in contrast to TCE-d2 ( Figure S4).

d) Determination of rate constants of MM/PP↔MP interconversion by 1 H-1 H EXSY NMR spectroscopy
Quantitative 1 H-1 H EXSY NMR spectroscopy can be used to obtain rate constants for the interconversion between different species. 4 This has been used to quantify the exchange kinetics in supramolecular systems where the exchanging species have different energies and populations. 5 An adapted version of this method has been used here.
All 2D-EXSY NMR spectra were recorded on a Bruker AV NEO 400 (400 MHz) NMR spectrometer. Exchange rates were calculated using the program EXSY CALC. 6 To calculate the exchange rates between two species, diagonal and cross-peak intensities for the exchanging NMR resonances are required from two EXSY NMR experiments at different mixing times. For each data point, one EXSY NMR experiment was carried out with a mixing time of 900 ms and another was carried out with a very short mixing time of 5 ms. The same major and minor exchanging proton signals were used for every EXSY experiment. Here, we used Hc and Hc* in the 1 H NMR spectrum of 1a (Figure 4), as the large chemical shift difference between these signals allowed easy and reliable integration (Δδ = 0.5 ppm).

S20
From here, ΔG ‡ in pure TCE-d2 can be calculated at 298 K: ΔG ‡ 1 (298 K) = 95.5 kJ mol -1 for the forwards MM/PP → MP process and ΔG ‡ -1 (298 K) = 94.8 kJ mol -1 for the backwards MP → MM/PP process. Therefore, the barrier in TCE-d2 at 298 K is close to that in dichloromethane at 298 K, as determined by CD spectroscopy in SI Section 5b.

X-ray crystallography
Purple, needle-like crystals of macrocycle 1b, suitable for single crystal X-ray diffraction, were grown from a racemic mixture of 1b dissolved in chloroform, with slow diffusion of methanol (antisolvent).
Single crystal X-ray diffraction experiments were performed at the UK Diamond Light Source I19-1 3-circle diffractometer ( = 0.6889 Å). 8-10 A suitable single crystal was selected and mounted using fomblin film on a micromount. Data were collected on a dectris-CrysAlisProabstract goniometer imported dectris images diffractometer. The crystals were kept at 100(2) K during data collection (single omega sweep). The structures were solved by direct methods using ShelXT 11 and refined with ShelXL using a least squares method. Olex2 software was used as the solution, refinement and analysis program. 12 The crystal diffracted weakly despite the use of synchrotron radiation and numerous attempts at growing better diffracting crystals. The data used in the refinement was truncated to a resolution of 0.84 Å which reduced completeness but improved signal to noise. Overall the data to parameter ratio is 11.5.A completeness of 83% did not support meaningful modelling of most of the disordered solvent.

Chiral HPLC a) Methods
Chiral chromatographic studies were performed using a Phenomenex i-Amylose-1 chiral column on an Agilent 1290 Infinity analytical HPLC instrument. The flow rate was 1 mL/minute and the detection wavelength was 500 nm. The eluents and injection volumes for each chromatogram are specified in the figure captions.
To separate the enantiomers for chiroptical studies, the system was set up to run automatedly and the enantiomers were collected using an automated fraction collector. For this purification the eluent system was 4:1 toluene:n-hexane and the injection volume was 20 µL.
Figure S10: Chiral HPLC chromatogram of compound 1a dissolved in toluene and eluted with 4:1 (v/v) toluene:nhexane. Using a combination of CD spectroscopy (Section 5) and computational modelling (Section 9), peak A is assigned as the MM enantiomer and peak B is assigned as the PP enantiomer.

S26
To measure the rate of racemisation of compound 1a in toluene, a pure fraction of peak A (MM enantiomer) was obtained by running compound 1a through the chiral HPLC column in 80:20 toluene: n-hexane eluent. The pure fraction of peak A was dried and re-dissolved in toluene. The sample was kept in toluene, and aliquots of it were re-injected into the column over time, allowing the growth of peak B to be monitored ( Figure S12). By measuring the ratio of the integrals of peaks A and B an enantiomeric excess can be calculated for a given time . The resulting data can then be fitted to the equation: Where 0 is the enantiomeric excess at = 0 and is the enantiomerisation rate constant. The racemisation rate constant = 2 . 13 Figure S13: Plot and linear fit of ( ⁄ ) against .
The fitting of this data is shown in Figure S13 and results in a rate constant of enantiomerisation k = 8.82 × 10 -3 h -1 = 2.45 × 10 -6 s -1 and hence a racemisation rate constant krac = 4.90 × 10 -6 s -1 for toluene at room temperature. A free energy of activation for the racemisation process ΔG ‡ = 104.98 kJ mol -1 was determined for toluene at room temperature according to the Eyring equation. This is close agreement with the value for ΔG ‡ determined by time-course CD spectroscopy (Section 5).
Additionally, the chromatogram of a sample of peak A (PP enantiomer) that was kept as a dry solid for three months shows very little interconversion has occurred (MM:PP = 95:5 mol%), allowing us to estimate the enantiomer half-life to be t1/2 = years in the solid state ( Figure S14).

Chiroptical studies a) Circular dichroism studies
Circular dichroism (CD) spectra were recorded on a Jasco J-1500 CD spectrophotometer with a wavelength accuracy ± 0.2 nm (250 to 500 nm), ± 0.5 nm (500 to 800 nm) and a CD root mean square noise < 0.007 mdeg (500 nm). A quartz cuvette with 0.5 mm path length was used. The spectra were recorded at a concentration of 10 µM. The enantiomers were assigned by comparison of their CD spectra in toluene (Figure S15) with the computationally calculated spectra of the enantiomers in toluene (Table S13).

b) Kinetics from time-course CD
To determine the racemisation rate constants, an enantiopure sample of 1a MM (10 μM, MM:PP > 99:1 mol%) was dissolved in the solvent being tested (toluene or DCM) and kept in a sealed cuvette at 25 °C. The CD spectrum was recorded at regular time intervals. The decay in intensity of the strongest peaks between λ = 250 -400 nm was monitored over time and the resulting data was fitted to the following equation: Where 0 is the CD signal intensity at = 0 for a given peak, CD t is the CD signal intensity at time for a given peak and is the enantiomerisation rate constant. The racemisation rate constant is determined from = 2 . 13

c) Circularly-polarised luminescence studies
CPL was measured with a home-built (modular) spectrometer. The excitation source was a broad band (200 -1000 nm) laser-driven light source EQ 99 (Elliot Scientific). The excitation wavelength was selected by feeding the broadband light into an Acton SP-2155 monochromator (Princeton Instruments); the collimated light was focused into the sample cell (1 cm quartz cuvette). Sample PL emission was collected perpendicular to the excitation direction with a lens (f = 150 mm). The emission was fed through a photoelastic modulator (PEM) (Hinds Series II/FS42AA) and through a linear sheet polariser (Comar). The light was then focused into a second scanning monochromator (Acton SP-2155) and subsequently on to a photomultiplier tube (PMT) (Hamamatsu H10723 series). The detection of the CPL signal was achieved using the field modulation lock-in technique. The electronic signal from the PMT was fed into a lock-in amplifier (Hinds Instruments Signaloc Model 2100). The reference signal for the lock-in detection was provided by the PEM control unit. The monchromators, PEM control unit and lock-in amplifier were interfaced to a desktop PC and controlled by a customwritten Labview graphic user interface. The lock-in amplifier provided two signals, an AC signal corresponding to (IL-IR) and a DC signal corresponding to (IL + IR) after background subtraction. The emission dissymmetry factor was therefore readily obtained from the experimental data, as 2 AC/DC.
Spectral calibration of the scanning monochromator was performed using a Hg-Ar calibration lamp (Ocean Optics). A correction factor for the wavelength dependence of the detection system was constructed using a calibrated lamp (Ocean Optics). The measured raw data was subsequently corrected using this correction factor. The validation of the CPL detection systems was achieved using light emitting diodes (LEDs) at various emission wavelengths. The LED was mounted in the sample holder and the light from the LED was fed through a broad band polarising filter and /4 plate (Ocean Optics) to generate circularly polarised light. Prior to all measurements, the /4 plate and a LED were used to set the phase of the lock-in amplifier correctly. The emission spectra were recorded with 0.5 nm step size and the slits of the detection monochromator were set to a slit width corresponding to a spectral resolution of 0.25 nm. CPL spectra (as well as total emission spectra) were obtained through an averaging procedure of several scans. The CPL spectra were smoothed using a shape-preserving Savitzky-Golay smoothing (polynomial order 5, window size 9 with reflection at the boundaries) to reduce the influence of noise and enhance visual appearance; all calculations were carried out using raw spectral data. Analysis of smoothed vs raw data was used to help to estimate the uncertainty in the stated glum factors, which was ±10%.  Table S3 shows that macrocycle 1a is at the upper end of the range of glum values (glum = 10 -2 ) for small organic molecules in solution (current highest is 10 -1 ) [14][15][16][17][18] and the highest for discrete PDI emitters (current highest is 10 -3 ). 19 Furthermore, macrocycle 1a exhibits the most red-shifted CPL spectrum (675 nm) of all small organic emitters reported to date. [19][20][21]  Values are for discrete monomers in solution, as for macrocycle 1a.

Photophysics
All steady state electronic absorption and emission spectra were recorded at a concentration of 10 µM (unless otherwise stated) at 298 K. For UV-Vis-NIR spectroscopy a Cary 5000 spectrophotometer was used, with a wavelength accuracy ≤ 0.08 nm and absorbance accuracy ≤ 0.01 Abs. For fluorescence spectroscopy a Jasco FP-8500 was used with emission and excitation wavelength accuracies ± 1.0 nm. The detector base sensitivity is 8500:1. Quartz cuvettes with 1 cm path length were used.

e) Quantum yields
Absolute fluorescence quantum yields were obtained on an Edinburgh Instruments FLS920 steady-state spectrometer fitted with an integrating sphere. All samples were recorded at a 1 µM with a 7 -8 nm excitation slit and 0.1 -0.2 nm emission slit width. Experiments were carried out in solution using 1 cm path length quartz cuvettes with four transparent polished faces.

Quantification of intramolecular H-type aggregation
To gain a deeper insight into the macrocycle-solvent interactions that promote intramolecular H-type aggregation, the UV-vis absorption spectrum of 1a was recorded in a wide range of solvents. The A0-0 / A0-1 ratio was recorded for each solvent. From here a Gibbs free energy of intramolecular H-type aggregation (∆ agg ) was determined for each solvent, following an adapted method used by Würthner and co-workers. 29 In this method, the following assumptions are made: 1. The A0-0 / A0-1 ratio in toluene corresponds to all molecules of macrocycle 1a being in a state of full intramolecular H-type aggregation.
2. The A0-0 / A0-1 ratio in TCE corresponds to all molecules of macrocycle 1a being in a state where there is no H-type aggregation.
These assumptions are validated by the fact the ε0-0 / ε0-1 ratio for macrocycle 1a is minimised in toluene (0.58), while the ratio is maximised in TCE (1. 19) and is the same as that of the monomeric PDI 3a in either solvent. We note that the MP diastereomer is present as a minor species in chlorinated solvents (~10 mol%) and, from density functional theory calculations (Section 9), is not an H-type aggregate.
From these assumptions, the mole fraction of fully unaggregated molecules can be estimated according to the following equation: Where is the ε0-0 / ε0-1 ratio of the fully H-type aggregated macrocycle in toluene, is the A0-0 / A0-1 ratio of the non H-type aggregated macrocycle in TCE and is the observed A0-0 / A0-1 for a given solvent being investigated. 29 From this, an equilibrium constant can be calculated as follows: Where and are the concentrations of H-type aggregated and non H-type aggregated macrocycles respectively. Hence, a Gibbs free energy of intramolecular H-type aggregation (∆ agg ) can be determined for each solvent according to: As toluene and TCE are the reference solvents for fully H-type aggregated and non H-type aggregated species respectively, they cannot be included in the ∆ plots as they represent asymptotes in the model. The ∆ agg of macrocycle 1a for different solvents were plotted against various solvent scales (Figure S29-36). Good correlations are observed against scales that account for solvent polarity (ε, , Kirkwood-Onsager) or solvent polarity and polarizability (π*, Catalán SPP, ), where Pearson's r = 0.8-0.9. 30 There is no correlation with solvent polarizability (α) or hydrogen bonding (β).

S48
The UV-vis absorption spectrum of macrocycle 1a was measured as the solvent composition was gradually changed from pure toluene to 97:3 (v/v) toluene:TCE, while keeping the concentration of 1a constant (Figure S37). By monitoring the A0-0 / A0-1 ratio for each solvent composition (Figure S38), the free energy of intramolecular H-type aggregation (∆ agg ) can be determined for different toluene:TCE solvent mixtures ( Figure S39). The method used here is based on that developed by Moore and Ray to study the solventinduced folding of phenylene ethynylene oligomers, 35 and used by Würthner and co-workers. 29 The equation below used to calculate ∆ agg results in asymptotes at 0 % and 100 % TCE. However, plotting of the ∆ in the transition region against % TCE gives a straight-line relationship that can be extrapolated to obtain an estimate ∆ agg for full intramolecular H-type aggregation in pure toluene (i.e., at 0% TCE) according to the equation: Where ∆ (Tol) is the free energy for full intramolecular H-type agreggation in pure toluene, [TCE] is the concentration of TCE in the solvent mixture, and is the gradient of the plot. From Figure S39, this gives ∆ (Tol) = -11.17 ± 0.11 kJ mol -1 .

Electrochemistry
Electrochemical experiments were carried out in anhydrous degassed solvents (DCM or 1:1 toluene-DCM) using 0.1 M tetrabutylammonium tetrafluoroborate as the supporting electrolyte. The working electrode was a 3 mm glassy carbon electrode (polished with diamond slurry prior to use), a Pt counter electrode, and a Ag/AgCl reference. All cyclic voltammograms (CVs) were referenced to the Fc + /Fc 0 redox couple and recorded at a scan rate of 0.1 Vs -1 . An Autolab Interface 6 potentiostat was used for electrochemical measurements. The inherent instrument error is V: ± 0.2% (± 2 mV).

Density Functional Theory Calculations a) Conformer search
Conformer searches in toluene and dichloromethane for a simplified model of macrocycle 1, in which the imide-based alkyl chains (R groups) are replaced with methyl groups, were performed using the combination of the CREST code, 36 the GFN2-xTB semiempirical tightbinding method 37 and the analytical linearized Poisson-Boltzmann (ALPB) implicit solvation model. 38 The lowest energy conformers found using CREST were subsequently reoptimized by means of density functional theory using either the B97-3c composite scheme 39 or the combination of the PBE density functional, 40 the D4 dispersion correction method 41 and the def2-TZVP basis-set. 42 Solvation effects in the DFT calculations were described using either the COSMO 43 (toluene, dichloromethane) or COSMO-RS 44 (toluene, dichloromethane) implicit solvation models. All DFT calculations, including the time-dependent DFT and NMR calculations discussed below, are performed using Turbomole 7.5. [45][46] An initial conformer search in the gas phase gave the lowest energy conformers A-H below.
Conformations related by the simple rotation of the methyl group(s) at the imide position(s) were omitted. These structures are provided as .xyz files in the 'DFT structures' folder. Figure S42: Lowest energy conformers A-H (gas phase). The homochiral/heterochiral labels refer to the relationship between the two axially chiral PDI units. Conformer F cannot be labelled in this way since the perylene core of the bottom PDI is pseudo planar. MM and PP assignments are included for conformer A, in agreement with the homochiral conformation of the macrocycle observed experimentally in toluene (1a) and the X-ray crystal structure (1b). The MP assignments is given to the heterochiral conformer H.

S53
The relative energies of conformers A-H were obtained in the gas phase and in toluene and chlorinated solvents (Tables S7-S8). The energy landscape in chlorinated solvents predicted by density functional theory differs from that observed experimentally, most likely because our model does not account for intermolecular macrocycle-solvent hydrogen bonds, specifically impacting the energies of conformers C and H, in which intramolecular hydrogen bonds are broken.   Figure S43: An overlay of the DFT predicted conformer A (red) and the same fragment of the X-ray crystal structure (yellow) of macrocycle 1. The methyl group (DFT structure) or alkyl chains (crystal structure) have been omitted to aid with comparison. Figure S44: A top-down view of conformer C shows the PDI units are rotated by 70° relative to one another, which switches π-π interactions OFF. This contrasts with the 20° rotation in the H-type aggregated conformer A (π-π ON) as seen in the DFT structure ( Figure S43) and X-ray crystal structure (Figure 2c).

b) Predicted UV-vis and CD spectra
Vertical excitation and circular dichroism spectra of the DFT optimised conformers were calculated by single point calculations on the B97-3c optimised structures using the combination of the B97x density functional 47 and the def2-TZVPP basis-set. 42

Predicted UV-vis spectra
The lowest energy excitation (1) in the spectrum of C (i.e., π-π OFF, no H-type aggregation) has a greater intensity than the corresponding peak (1) in conformers A and B (i.e., π-π ON, H-type aggregate). Hence, DFT predicts a red-shifted peak for C, relative to A and B. This is consistent with the experimentally observed red shift of λmax of macrocycle 1a on going from toluene to chlorinated solvents.     The predicted CD spectra of each enantiomer of homochiral conformer A (MM and PP) were used to assign the experimental CD spectra of macrocycle 1a.

c) Predicted 1 H NMR spectroscopy chemical shifts
1 H NMR spectroscopy chemical shifts of DFT optimised conformers were predicted using the Gauge-Including Atomic Orbitals method 42 and PBE-D4/def2-TZVP for structures optimised with the same functional and basis-set combination. The PDI protons Ha-c were chosen because they are informative of PDI-PDI π-π stacking interactions.
The calculated 1 H NMR spectrum of conformer A (H-type aggregate) shows a good agreement with the 1 H NMR spectrum measured in toluene-d8 (Table S14), and a poor agreement with the spectrum measured in TCE-d2 (Table S15). Instead, the calculated spectrum for conformer C (no H-type aggregation) shows a good agreement with the spectrum measured in TCE-d2 (Table S16).