Tuning symmetry breaking charge separation in perylene bichromophores by conformational control†

Understanding structure–property relationships in multichromophoric molecular architectures is a crucial step in establishing new design principles in organic electronics as well as to fully understand how nature exploits solar energy. Here, we study the excited state dynamics of three bichromophores consisting of two perylene chromophores linked to three different crown-ether backbones, using stationary and ultrafast electronic spectroscopy combined with molecular dynamics simulations. The conformational space available to the bichromophores depends on the structure and geometry of the crown-ether and can be significantly changed upon cation binding, strongly affecting the excited-state dynamics. We show that, depending on the conformational restrictions and the local environment, the nature of the excited state varies greatly, going from an excimer to a symmetry-broken charge separated state. These results can be rationalised in terms of a structure–property relationship that includes the effect of the local environment.


Absorption
Absorption spectra were measured on a Cary 50 spectrometer.

Emission
Emission spectra were measured on a Horiba Scientific FluoroMax-4 fluorometer and corrected using a set of secondary emissive standards. 1

General Remarks
The TA data presented in this work were recorded with three different experimental setups: a fs-ps visible (fs-VIS), a fs-ps near infrared (fs-NIR) and a ps-µs visible (ps-VIS) TA setup. A detailed description of the general principle of the fs-ps as well as ps-µs TA applying referenced detection using two spectrographs is presented elsewhere. 2 In the following, experimental details concerning the three setups and the data analysis are discussed. The fs-VIS and fs-NIR setups share the same pump path, whereas the design of the fs-VIS and the ps-VIS probe path are identical. The absorbance of the sample at the excitation wavelength was 0.2-0.4 (at 400 nm) on 1 mm. The absorption spectra of all samples before and after the transient absorption experiments showed no signs of degradation. The samples in DCM were measured in a 1 mm quartz cuvettes (Starna, model 1GS/Q/1) and bubbled with nitrogen during the measurements giving a wavelength dependent IRF of about 80-350 fs (fwhm of optical Kerr effect (OKE)). The samples in ACN were measured with a flow cell with 200 µm, C7980 5F windows and a spacer thickness of 400 µm giving a wavelength dependent IRF of about 80-120 fs (fwhm of OKE).

fs-ps Pump
Excitation is performed using 400 nm pulses generated by frequency doubling part of the output of a 1 kHz Ti:Sapphire amplified system (Spectra Physics, Solstice Ace). The TA signal was checked prior to the experiment to scale linearly with the pump energy. Samples were excited with ca. 0.2-0.5 mJ cm -2 pump intensity. The polarization of the pump pulses was set to magic angle relative to the white-light pulses. In order to check for pump beam divergence and/or delay-line misalignment, the fs-TA dynamics of a calibration sample (perylene in DMSO) were compared to the dynamics measured on a time-correlated single photon counting setup.

ps-µs Pump
The ps-µs pumping was described in detail in ref. 3. Excitation was performed at 355 nm using a passively Qswitched, frequency doubled Nd:YAG laser (Teem Photonics, Powerchip NanoUV) producing pulses with a 500 Hz repetition rate, approximately 20 µJ energy per pulse, and 300 ps duration.

Visible Probe
Probing was achieved using white light pulses generated by focusing the 800 nm pulses of the Ti:Sapphire amplified system in a CaF 2 plate. The experimental setup was the same as that described earlier, 4 except that all lenses, after white light generation, were replaced by spherical mirrors to prevent chromatic aberration.

Near Infrared Probe
The white light was generated by focusing the 800 nm pulses in a YAG crystal. To balance the intensity of the white light, the high intensity light at 800 nm was removed by a beam stop after generation as well as by a 1 mm cuvette containing IR140 in DMSO. The probe light was then separated into a reference and a sample beam by a reflective metallic neutral density filter. After passing the sample, the beam was dispersed by a home-built prism spectrometer and the intensity recorded with a InGaAs detector. To balance the white light spectrum, apodizing neutral density filters were placed directly before both detectors.

Data Treatment
The pixel to wavelength conversion was achieved using a standard containing rare earth metals (NIST 2065 for NIR and holmium oxide for VIS) which shows narrow bands from the UV to the NIR. All transient absorption spectra were corrected for background signals showing up before time zero (e. g. spontaneous emission). The fs-ps spectra were corrected for the dispersion due to the optical chirp using the optical Kerr effect. 5 For merging the fs-vis and the fs-nir spectra, the TA signal in the overlap region between 690 and 740 nm was compared and one of the two datasets was multiplied with a constant factor accounting for the difference in pump power. The comparison of the kinetics recorded in the overlap region additionally serves as quality control, directly revealing erroneous kinetics e.g. due to deviation from the magic angle or poor alignment of the delay stage.

Broadband Fluorescence Up-converison Spectroscopy (FLUPS)
Broadband FLUPS measurements were performed on a setup similar to that described in detail in refs. 6, 7. In brief, excitation was performed with a 100 fs pulses at 400 nm generated by frequency doubling part of the output of a standard 1 kHz Ti:Sapphire amplified system. The pump intensity on the sample was below 1 mJ/cm 2 . The gate pulses at 1340 nm were produced by an optical parametric amplifier (TOPAS Prime, Light Conversion). Detection of the up-converted spectra was performed with a home-built spectrograph coupled to a CCD camera (Andor, DV420A-BU). The full width at half-maximum (FWHM) of the cross correlation of the gate with the solvent Raman signal was approximately 170 fs. Time-resolved emission spectra were recorded in two sequential measurements with a linear time grid from -2 to 2 ps and with a logarithmic time grid extending up to 1.5 ns. The measurement consisted of 10 successive scans with 2 s collection time at each time step.The raw data were transferred into spectra vs. wavenumber and subsequently corrected by calibration with secondary emissive standards as described in ref. 6. The chirp due to group velocity dispersion was determined by measuring the instantaneous response of BBOT in the same solvent.

Molecular Dynamics Simulations
Classical molecular dynamics (MD) simulations were performed using NAMD2 8 in combination with the CHARMM36 force field. 9, 10 Topology files for all chromophores and ions were retrieved from the CHARMM General Force Field (CGenFF), 11 and those for solvents (acetonitrile, ACN, and dichloromethane, DCM) were obtained through the virtualchemistry.org website. 12 Since the default DCM parameters generated a solvent environment with a very low dielectric constant ( r = 2.7), partial charges of DCM were further refined to more closely reproduce the experimental r value of 8.9. For this optimisation, the restricted electrostatic potential (RESP) method 13  For all production simulations, non-bonded interactions were evaluated with a cutoff of 12Å, with a switching function starting at 10Å. Long-range electrostatic interactions were accounted for by the particle mesh Ewald (PME) 16 method with a grid density of 1Å −3 and a PME interpolation order of 6. All bonds involving hydrogen atoms were kept rigid using the SHAKE algorithm. 17 Simulations were performed in an NPT ensemble employing Nosé-Hoover thermostat and Langevin piston (1 atm, period: 50 fs, decay: 25 fs, damping coefficient 1.0 ps −1 ). 18,19 MD time-step was set to 2 fs and both non-bonded and PME forces were updated at every time-step. Trajectories were visualized and analysed by VMD. 20 The data were collected every 4 ps and were analysed by in-house scripts. In particular, four intramolecular parameters were extracted from the trajectories ( Figure S1): α: the angle between the two Pe long axes, i.e., the angle between the transition dipole moment of each individual Pe unit; β: the angle between the normals to the Pe planes; γ: the slip angle, defined as the angle between the long axis of one Pe moiety and the axis connecting the geometric centres of the two Pe moieties; and δ: the distance between the geometric centres of the two Pe groups. Each simulated system contained one of the bichromophores of interest (18c6, 18c4 or 16c4) immersed in a solvent box (ACN or DCM) with a side of 50Å. We also prepared a control simulation in which two monomers (Ref ) were placed in the same simulation box in order to reproduce the same number of Pe moieties as in the covalently-linked chromophores. The estimated concentration of the monomers in the box was 26 mM, which is higher than the experimental concentration used in this work (about 0.1 mM). Each MD system was replicated 5 times and each replica simulated in production mode for 21 ns. The first nanosecond of each replica was discarded. The remaining 20 ns of each replica per system were combined (overall 100 ns) and used for subsequent analysis. For the simulation containing barium cation, a sulfate ion was used to neutralize the overall charge of the system. In order to prevent a strong interaction between the sulfate and the chromophore due to the finite volume of the solvent box, the position of the sulfate was kept fixed in space and the partial charges of sulfur and oxygen in the sulfate were reduced to 0.288 and -0.572, respectively (initial values were 0.992 and -0.748). In all other simulations containing cations, we placed two Na + so to have the same amount of positive ions in solution as for the simulations α β γ δ Figure S1: Intramolecular parameters extracted from MD trajectories (blue rectangles represent Pe moieties): α) the angle between the two Pe long axes, β) the angle between the normals to the Pe planes, γ) the angle between the long axis of one Pe moiety and the axis connecting the geometric centres of the two Pe heads, and δ) the distance between the geometric centres of the two Pe groups.
containing barium. The two positive sodium ions were neutralized with a sulfate as for barium and the positions of sulfate and the sodium ion not involved in the interaction with the crown ether were fixed in space. Since the formation of the host:guest (HG) complex by equilibrium MD is computationally expensive, initial structures for such complexes were prepared by placing the cation at the geometric centre of each crown ether, minimizing for 500 steps and running short simulations with positional restraints of decreasing force constants on the complex at each simulation. This procedure provided HG complexes where the cation remained inside the crown-ether cavity stably during the simulations without applying any restraint.
For the simulations involving the 1:2 complex of 18C6 and Na + in DCM, both sodium cations and the sulfate were free to move. However, two NBFIX 21 parameter sets were added between 1) sulfate oxygens and sodium cations, and 2) sulfate oxygens and amide hydrogens, in order to avoid short-range interactions of the sulfate with the complex. Parameters for both NBFIX were = -0.1 kcal/mol and R min = 20Å. In the starting configuration for the 1:2 complex, both sodium cations were placed inside the crown ether on the same plane as its oxygen atoms. After few hundreds of picoseconds of equilibration, a single Na + occupied the crown ether cavity and the second sodium cation interacted with the amide groups adopting the conformation described in the main text.

Host-Guest Complexes
In order to keep the host concentration constant a stock solution (ST1) of the sample (bichromophore or Ref ) was prepared in the respective solvent . ST1 was then used to dissolve a certain amount of salt giving the stock solution (ST2). ST2 was then titrated to ST1 and the spectral changes were recorded after each addition. Due to a limited availability of the samples, the molar absorption coefficient and therefore the concentration of the host could not be accurately determined. To estimate the host concentration, a molar absorption coefficient of 34500 M -1 cm -1 , which is the literature value 24 for the parent Pe, was used for the maximum of Ref upon complexation of the cation. A molar absorption coefficient of 69000 M -1 cm -1 was assumed for the low coupling bichromophores, which showed the signatures of a local excited state in the transient absorption spectra. The association constants as well as the spectra obtained from the global analysis of the absorption spectra with varying guest (salt) concentration are depicted in Table S1 and Figure S4, respectively. It has to be noted that for very high association constants as observed for the bichromophores, the error on the host concentration has a strong influence on the association constant. Therefore, the reported constants should only be used to compare the complexation behaviour between the different crown-ethers and Ref.

Ref in DCM
The stationary absorption and emission spectra, before and after addition of NaBAr F are shown in Figure S5. Details on the complexation are discussed in section 4.1.

18c6 in ACN
The excitation/emission map of 18c6 in ACN before (A), and after complexation of Ba 2 + is shown in Figure S6. Slices of this dataset are illustrated in Figure S7 depicting the wavelength dependence of excitation/emission of the emission/excitation spectra. Figure

18c6 in DCM
A B 18C6⊂Na + 18C6 † 18C6 ⊂2Na + 18C6⊂Na + 18C6 † 18C6 ⊂2Na + Figure S8: (A) Normalized emission spectra of 18c6 (blue), 18C6⊂ ⊂ ⊂Na + (red) and † 18C6⊂ ⊂ ⊂2 Na + (green). The red shifted, featureless bands of † 18C6⊂ ⊂ ⊂2 Na + and 18C6⊂ ⊂ ⊂Na + resemble each other and are therefore attributed to the 18C6⊂ ⊂ ⊂Na + sub-population present in † 18C6⊂ ⊂ ⊂2 Na + . Furthermore, the excimer spectrum of 18c6 contains an even further red shifted part compared to the 18C6⊂ ⊂ ⊂Na + spectrum. This could be due to weaker geometrical restrictions in 18C6 enabling the approach to a stronger coupled conformer and therefore a lower energy excimer. The absolute intensity (B) illustrates that the intensity of the excimer band increases upon formation of 18C6⊂ ⊂ ⊂Na + . This observation along with the absence of the NIR band in transient absorption suggest that the excimer conformations for 18C6 and 18C6⊂ ⊂ ⊂Na + are substantially different whereas the coupling is stronger in the former and closer to the eclipsed D 2h .  Figure S10: Histograms of different structural coordinates for 18c6 in DCM. 18c6 (blue); 18C6⊂ ⊂ ⊂Na + (red); 18C6⊂ ⊂ ⊂2 Na + (green).

Molecular Dynamics Simulations
Figure S11: MD trajectories of the distances between the Na + cations (that bound to the crown-ether and that bound to the amide oxygens) and the centres of the Pe heads (left), and histograms of the difference between these distances (right).

Salt concentration
The concentration of salt was adjusted to obtain at least 95 % 1:1 complexes, corresponding to about 100 µM. To obtain † 18C6⊂ ⊂ ⊂2 Na + , the concentration of NaBAr F was about 1.5 mM.    Figure S18: Transient absorption spectra in the UV/Vis region measured at various time delays after excitation of 18c6 in DCM at different Na + concentrations. Upon further addition of salt to 18C6⊂ ⊂ ⊂Na + , the 18C6⊂ ⊂ ⊂2 Na + /18C6⊂ ⊂ ⊂Na + ratio increases leading to an increase of the I 0−0 /I 1−0 ratio of the ground state bleach, an increase of the stimulated emission as well as the charge-transfer (CT) feature. The broadband transient absorption data were analysed globally using a global lifetime analysis based on the approach discussed in detail in ref. 25, 26. We assumed a sequential model for all the datasets analysed. This corresponds to a single population that evolves as a series of n successive exponential steps into further species without any losses or back-reactions as illustrated in Figure S21. In this case, one can obtain the so-called Evolution Associated Difference Spectra (EADS), which represent the spectral evolution. However, the EADS do not systematically correspond to given species or states but only help to visualizing the timescales on which the spectral dynamics occur. The lifetimes and EADS are shown in Table S2 and Figure S22 . Figure S21: Sequential model used in the global analysis. The time constants are listed in Table S2 and the evolution associated spectra A-D are illustrated in Figure S22.