Two-step anti-cooperative self-assembly process into defined π-stacked dye oligomers: insights into aggregation-induced enhanced emission

Aggregation-induced emission enhancement (AIEE) phenomena received great popularity during the last decade but in most cases insights into the packing structure – fluorescence properties remained scarce. Here, an almost non-fluorescent merocyanine dye was equipped with large solubilizing substituents, which allowed the investigation of it's aggregation behaviour in unpolar solvents over a large concentration range (10−2 to 10−7 M). In depth analysis of the self-assembly process by concentration-dependent UV/Vis spectroscopy at different temperatures revealed a two-step anti-cooperative aggregation mechanism. In the first step a co-facially stacked dimer is formed driven by dipole–dipole interactions. In a second step these dimers self-assemble to give an oligomer stack consisting of about ten dyes. Concentration- and temperature-dependent UV/Vis spectroscopy provided insight into the thermodynamic parameters and allowed to identify conditions where either the monomer, the dimer or the decamer prevails. The centrosymmetric dimer structure could be proven by 2D NMR spectroscopy. For the larger decamer atomic force microscopy (AFM), diffusion ordered spectroscopy (DOSY) and vapour pressure osmometric (VPO) measurements consistently indicated that it is of small and defined size. Fluorescence, circular dichroism (CD) and circularly polarized luminescence (CPL) spectroscopy provided insights into the photofunctional properties of the dye aggregates. Starting from an essentially non-fluorescent monomer (ΦFl = 0.23%) a strong AIEE effect with excimer-type fluorescence (large Stokes shift, increased fluorescence lifetime) is observed upon formation of the dimer (ΦFl = 2.3%) and decamer (ΦFl = 4.5%) stack. This increase in fluorescence is accompanied for both aggregates by an aggregation-induced CPL enhancement with a strong increase of the glum from ∼0.001 for the dimer up to ∼0.011 for the higher aggregate. Analysis of the radiative and non-radiative decay rates corroborates the interpretation that the AIEE effect originates from a pronounced decrease of the non-radiative rate due to π–π-stacking induced rigidification that outmatches the effect of the reduced radiative rate that originates from the H-type exciton coupling in the co-facially stacked dyes.


Chemicals
Solvents and chemicals were obtained from commercial suppliers and used without further purification, unless otherwise stated. The solvents for UV/Vis absorption, circular dichroism (CD), fluorescence spectroscopy, circular polarized luminescence (CPL), vapor pressure osmometry (VPO) measurements as well as atomic force microscopy (AFM) were of spectroscopic grade and used as received. 1,2,3-Tris(dodecyloxy)benzene 4 was synthesised according to literature known procedure. S1 Chiral amine 2 S2 was a gracious donation by Prof. Dr. Klaus Ditrich from the BASF SE in Ludwigshafen, Germany, and had an enantiomeric excess of 99.3%.

Melting Points
Melting points were determined on a Linkam TP 94 heating stage and are uncorrected.

Elemental Analysis
Elemental analyses were conducted on a CHNS 932 analyser (Leco Instruments GmbH)

NMR-Spectroscopy
All spectra were recorded in deuterated solvents at a Bruker Avance III HD 400 or Bruker Avance III HD 600 spectrometer (Germany) using either a 5 mm BBFO probe or a 5 mm DCH cryo-probe, both equipped with z-gradient and a temperature control unit. Chemical shifts are reported in ppm relative to residual solvent signal. For multiplicities, abbreviations are used as follows: s = singlet, d = doublet, t = triplet, m = multiplet, br = broad.
The nuclear Overhauser effect spectrum ( 1 H 1 H NOESY) of the dimer of 1 was recorded using the noesygpphpp sequence with a mixing time (d8) of 600 ms. The rotating frame Overhauser effect spectrum ( 1 H 1 H ROESY) of the monomer of 1 was recorded using the roesyphpr.2 sequence with a mixing time (p15) of 300 ms.

S4
Diffusion ordered spectroscopy (DOSY) experiments were recorded on a Bruker Avance III HD 600 with a BBFO probe with z gradient and a maximum gradient strength of 50 G cm 1 .
The DOSY spectra were acquired using the dstebpgp3s (without rotation) or ledbpgp2s (with rotation) pulse sequence. For all experiments, the diffusion gradients were linearly incremented in 32 steps from 2 to 98%. The diffusion time (d20) was set to 50 ms. In case of the dimer of 1 the dstebpgp3s spectrum was evaluated which includes convention compensation. For the higher aggregate no difference was observed between the sequences so the ledbpgp2s spectrum was used for evaluation. Diffusion coefficients were determined by monoexponential fit of the attenuation curves for several peak integrals and averaging of the values. To calculate the hydrodynamic radius according to the Stokes-Einstein equation the value for the dynamic viscosity of undeuterated MCH at the respective temperature was used S3 . Data processing was performed with the TopSpin 3.5 pl 7 software.

Mass Spectrometry
High-resolution ESI-TOF mass spectrometry for the characterization of compounds 3 and 1 was performed on a microTOF focus instrument (Bruker Daltonics) in positive mode with methanol (MeOH) or acetonitrile (MeCN) as solvent.

UV/Vis Spectroscopy
UV/Vis absorption spectra were recorded on a JASCO V-670 or V-770 spectrometer with a spectral bandwidth of 2 nm and a scan rate of 200 nm/min. The temperature was controlled either by a Peltier element (JASCO) or with a NCP-706 thermostat. Quartz cells (Hellma Analytics) with the following path length were used: 0.01 mm (c0 = 1.0 × 10 2 to 2.0 × 10 3 M), 0.1 mm (c0 = 1.0 × 10 3 to 1.7 × 10 4 M), 2 mm (c0 = 1.0 × 10 4 to 1.7 × 10 5 M), 10 mm (c0 = 1.2 × 10 5 to 2.2 × 10 6 M), 50 mm (c0 = 1.2 × 10 6 to 5.2 × 10 7 M), 100 mm (c0 = 4.0 × 10 7 to 9.8 × 10 8 M). Samples in MCH were prepared in silanized vials and measured in silanized cuvettes (except for cuvettes with a path length of d < 2 mm). For concentration-dependent studies, stock solutions were subsequently diluted to adjust the desired concentration. Solutions in MCH were allowed to equilibrate at rt overnight after dilution and 5 to 30 min after a temperature change, before starting the measurement. Repeated measurements were performed to check whether equilibration was complete. Apparent extinction coefficients were calculated according to the Lambert-Beer law and are density corrected for the respective temperatures.

CD Spectroscopy
CD was measured on a JASCO J-810 spectropolarimeter with a spectral bandwidth of 1 nm and a scan rate of 200 nm/min, using the same solvents and quartz cells as for the UV/Vis measurements.
Absolute quantum yields of the solid material were determined on a Hamamatsu Absolute PL Quantum Yield Measurement System CC9920-02.

CPL Spectroscopy
CPL spectra were recorded with a customised JASCO CPL-300/J-1500 hybrid spectrometer.
The dimer sample was measured in a conventional fluorescence quartz cell (Hellma Analytics) with 10 mm path length. The excitation and emission bandwidth were set to 36 nm and 11 nm, S6 respectively. The HAT photomultiplier current was 995 V and the spectrum was recorded with a scan speed of 50 nm/min and a D.I.T. of 2 sec. 30 accumulations were measured.
For the higher aggregate sample in MCH a cylindrical cuvette (Hellma Analytics) with a path length of 1 mm was used. The excitation and emission bandwidth were set to 36 nm and 10 nm, respectively. The HAT photomultiplier current was 950 V and the spectrum was recorded with a scan speed of 50 nm/min and a D.I.T. of 2 sec. 3 accumulations were measured.

Atomic Force Microscopy
AFM measurements were performed under ambient conditions using a Bruker Multimode 8 SPM system operating in tapping mode in air. Silica cantilevers (OMCL-AC200TS, Olympus) with a resonance frequency of ~150 kHz and a spring constant of ~10 N m 1 were used. The solution of the sample in MCH was spin-coated onto n-tetradecylphosphonic acid(TPA)modified Si/SiO2 (100 nm)/AlOx (8 nm) substrates.

VPO Measurement
The vapor pressure osmometry measurements were performed on a KNAUER osmometer with a universal temperature measurement unit. For the measurements in MCH the measurement chamber was heated to 318 K and the chamber lid with the syringes to 320 K. Benzil

Computational Details
Geometry-optimization for monomer (M) and dimer (D) of merocyanine 1 was performed at the density functional theory (DFT) level using the B97D3 functional S5 including dispersion correction and the def2-SVP basis S6 as implemented in the Gaussian 16 program package S7 .
The polarizable continuum model (PCM) S8-10 was used with MCH as solvent. Dodecyl chains were replaced by methyl groups to reduce computational effort. Frequency calculations were performed to verify the optimized geometries as stationary minima. The rotational barrier around the C-N bond of the tetralin substituent of the monomer in the gas phase was calculated by performing a relaxed potential energy surface scan for a 180° rotation around this bond in S7 5° steps and calculating the energy difference between the highest and lowest energy structure.
For this the B97D3 functional S5 and the def2-SVP basis set S6 was used.
Due to the large size of the proposed decamer stack (H), the geometry was optimized using the semi-empirical PM7 Hamiltonian S11 within the MOPAC software package S12 , which also accounts for dispersion interaction. Dodecyl chains were replaced by methyl groups to reduce computational effort. Frequency calculations were performed to verify the optimized geometry as a stationary minimum. TDDFT calculations were performed on the geometry-optimized decamer stack (the benzene substituents were replaced by methyl groups after structure optimization to reduce computational effort for TDDFT calculations) using the B97 S13 functional and the def2SVP basis set S6 as implemented in the Gaussian 16 program package S7 .
The polarizable continuum model (PCM) S8-10 was used with MCH as solvent. The oscillator strength for the first 15 states were calculated and the CD spectrum was simulated with the help of GaussView 5 S14 . The half with at half hight was set to 0.18 eV and the spectrum was shifted 0.64 eV to lower energies to fit the experimental maximum. UV/Vis (CH2Cl2, 298 K, c0  6 × 10 −6 M): max ()  549 nm (116000 M 1 cm −1 ). S10

Silanization Procedure of Glassware
For the silanization of one commercial 1 × 1 cm cuvette a mixture of freshly distilled trimethylsilylchloride (4.60 mL, 50.0 mmol) and NaI (7.50 g, 50.0 mmol) in dry acetonitrile (200 mL) was placed in a 250 mL round bottom flask under nitrogen atmosphere. The cuvette was immersed and the mixture heated to 90 °C overnight. After cooling, the cuvette was washed with distilled water, acetone and CH2Cl2. The amount of silanization reagents was adapted for other glassware by taking into account the approximate outer and inner surface of the respective cuvettes or glass vials as well as the inner surface of the reaction flask.

Solvatochromism
When comparing the absorption spectra of the monomer of merocyanine 1 in different solvents with increasing polarity (Fig. S1), a pronounced negative solvatochromism can be observed.
This is in accordance with the decrease in dipole moment upon optical excitation reported in literature for this kind of dipolar chromophore (g = 17.1 D, e = 12.6 D) S15 . Also for the dimer a negative but smaller solvatochromic effect can be observed.

Transition Dipole Moments
The electronic transition dipole moments (eg) were calculated according to equation (S1). IA is the integrated absorption, NA the Avogadro constant, 0 the vacuum permittivity, c0 the speed of light in vacuum and h the Planck constant.
With S defined as: (S2)

Time-dependent UV/Vis Measurements
The higher aggregate shows kinetic stability in MCH upon dilution from c0  1.1 × 10 3 M to c0  1.1 × 10 5 M at 298 K. Time-dependent UV/Vis measurements show the slow transition from the higher aggregates to the dimers (Fig. S2). After rapid cooling from 353 to 298 K, two processes can be observed in the UV/Vis spectra of a concentrated solution of merocyanine 1 in MCH (c0  8.9 × 10 4 M): 1) The dimer species that is formed at 353 K reassembles and 2) the absorption band of higher aggregate slowly returns from a more broad shape to its original room temperature shape of an H-aggregate with a maximum at  = 477 nm and a shoulder at  ~ 540 nm (Fig. S3).

Fig. S4
Calculated UV/Vis absorption spectra of the higher aggregate of merocyanine 1 from global fit analysis of the concentration-dependent spectra in MCH at 298 K (black), 323 K (blue) and 353 K (red) according to the pentamer model.

Global Fit Analysis
A global fit algorithm as introduced in previous literature examples S16-18 was used to evaluate the concentration-dependent UV/Vis studies according to different aggregation models. The dimer, trimer, tetramer, pentamer and isodesmic model S18, 19 were applied.
The dimer fit S19 (2 M ⇌ D) was applied to selected data from the dilution series at 353 K in MCH, showing the transition from dimer (D) to monomer (M) (Fig. S6a). The experimental apparent extinction data ( ̅ ) of samples with a total molecular concentration c0 from several dilution series is compared to the simulated curve for the dimer model according to with the extinction of the pure monomer (M) and dimer (D) species, as well as the dimerization constant KD = 4.5 × 10 6 M 1 obtained from global fit analysis (Fig. S6b,c).

S14
The trimer, tetramer and pentamer fit S18 (n D ⇌ H, n = 3,4,5) were applied to selected data of the dilution series at 298 (Fig. S7) Additionally, the isodesmic fit S19 was applied to selected data of the dilution series at 298 The binding constants for the formation of the higher aggregate out of dimers (n D ⇌ H) at different temperatures, obtained from global fit analysis according to the different aggregation models, are summarized in Table S1. S15 Fig. S7 a) Concentration-dependent UV/Vis absorption spectra (dashed grey lines) of merocyanine 1 in MCH at 298 K in comparison with calculated spectra (solid grey lines) from global fit analysis according to the pentamer model. Arrows indicate the spectral changes upon increasing the concentration from c0 = 4.0 × 10 6 -5.1 × 10 3 M. Colored spectra are calculated spectra for the individual dimer (D, red) and higher aggregate (H, orange) species. b, d) Concentration-dependent experimental extinction coefficients of several independent dilution experiments at 298 K at the maximum of the dimer (b, 503 nm) or higher aggregate (d, 477 nm) absorption band in comparison with simulated curves according to different aggregation models (eq. (S6) and (S7)). c, e) Comparison of calculated spectra of dimer (c) and higher aggregate (e) from global fit analysis according to different models.  (S7)). c, e) Comparison of calculated spectra of dimer (c) and higher aggregate (e) from global fit analysis according to different models.   Tetramer 5.5 × 10 4 6.6 × 10 3 1.6 × 10 3 Trimer 4.9 × 10 4 5.3 × 10 3 1.3 × 10 3 Isodesmic 3.2 × 10 4 3.0 × 10 3 0.8 × 10 3

Multiple Linear Regression (MLR) Analysis
To determine the concentration of molecules present as monomer (xM), dimer (xD) and higher aggregate (xH) in a sample with a total molecular concentration c0, multiple linear regression (MLR) analysis was performed. Based on Lambert-Beer's law (equation (S8) Additionally, the calculated total concentration c  xM  xD  xH matches the experimentally balanced concentration c0 well (Fig. S10, insets), which confirms the validity of the method.

Degree of Aggregated -faces
In general, the degree of aggregated -faces agg- is defined as the number N of occupied faces divided by the total number of -faces (equation (S9)).
For simplification it is assumed, that for H all 20 -faces are occupied.

Vapor Pressure Osmometry
The molecular weight of the higher aggregate of merocyanine 1 (Mmonomer = 1026.55 g/mol) was determined by vapor pressure osmometry (VPO). This method allows to determine the total osmolality of solutions. Solutions containing solutes have lower vapor pressure than the pure solvent, which leads to a vapor pressure difference, and thus to a temperature difference (ΔT) at the thermistors during the measurement. This ΔT is proportional to the number of particles dissolved in the solution. Benzil (M = 210.23 g/mol) and polystyrene PS5270 (Mn = 5270 g/mol) were used as standards to determine the relation of measurement value (MV) and osmolality for the device and the applied measurement conditions. MCH was used as solvent and the measurements were performed at 318 K to ensure saturation of the gas phase of the cell with solvent vapor.
To evaluate the data two different methods can be applied. The first one assumes a linear relation between MV and the number of particles. A linear fit was applied to the data of the two standards with a fixed intercept at y = 0. With the slope s of these fit curves, the concentration c in mol kg 1 (moles of particles per kg solution) of unknown samples can be directly determined from the measurement value MV according to equation (S16). In this way, the average number <Xn> of molecules incorporated in these particles can be calculated.
The slope obtained from the linear fit of benzil and PS5270 reference measurements was identical (s = 562 kg mol 1 ) and <Xn> was calculated for samples with different concentrations The value rises with concentration up to <Xn> = 6.6. This is in accordance with the fact, that in the lower concentrated samples at 318 K mixtures of dimer and higher aggregates are present according to the UV/Vis spectra (Fig. S11b).
The second way to evaluate the data from VPO measurements considers that for large molecules (M > 500 g/mol) the relation between MV and the number of particles is not necessarily linear.
It is recommended to use a standard with a similar molecular weight as expected for the analyte

Fig. S22
Illustration of the geometry-optimized decamer stack of merocyanine 1 (PM7, dodecyl chains replaced by methyl groups) as well as the hydrodynamic radius (transparent grey sphere) as obtained from the Stokes-Einstein equation deduced from b) the 1 H DOSY NMR (600 MHz, 295 K) spectrum of 1 in MCH-d14 (c0 = 1.0 × 10 3 M). Inset shows a representative attenuation curve for the integral from 6.86 -6.62 ppm with the respective fit.

Fig. S23
Illustration of the geometry-optimized dimer structure of merocyanine 1 (B97D3/def2-SVP) as well as the hydrodynamic radius (transparent grey sphere) as obtained from the Stokes-Einstein equation deduced from b) the 1 H DOSY NMR (600 MHz, 348 K) spectrum of 1 in MCH-d14 (c0 = 2.3 × 10 4 M). Inset shows a representative attenuation curve for the integral from 7.07 -6.80 ppm with the respective fit.

Relative Determination of the Higher Aggregate Fluorescence Quantum Yield
As the fluorescence quantum yield of the higher aggregate of merocyanine 1 could not be determined directly, it was estimated relative to the quantum yield of the dimer. For this, the spectral data of a dilution experiment were evaluated, which display the slow transition from higher aggregate to dimer after instant dilution from c0 = 1.0 × 10 3 M to c0 = 1.0 × 10 5 M over time (Fig. S26). Since the excitation wavelength was set at the isosbestic point at 489 nm, the sample always absorbs a constant amount of photons and the intensity of the emission can be directly correlated to the varying higher aggregate and dimer ratio. The fluorescence quantum yield Fl is proportional to the integral A of the emission spectrum (equation (S18)). A system of linear equations can be formulated as equations (S19) Accordingly, H and D are the integrals of the theoretical emission bands of 100 % higher aggregate or 100% dimer, respectively, which are proportional to the respective quantum yields of the species. Since the quantum yield of the dimer is known to be Fl ( ) = 2.3%, which is proportional to D , the quantum yield of the higher aggregate can be determined relative to this by solving the system of linear equations and comparing D and H . Fl(H) was determined by this procedure to be about 4.5%. Geometry-optimized P-helical decamer stack of merocyanine 1 (PM7). The dodecyl chains of the trialkoxyphenyl substituents were replaced by methyl groups for the calculation to reduce computational effort and added manually (in grey) to the optimized structure to illustrate the sterical demand of the solubilizing substituents.