Modulation of charge transfer by N-alkylation to control photoluminescence energy and quantum yield†

Charge transfer in organic fluorophores is a fundamental photophysical process that can be either beneficial, e.g., facilitating thermally activated delayed fluorescence, or detrimental, e.g., mediating emission quenching. N-Alkylation is shown to provide straightforward synthetic control of the charge transfer, emission energy and quantum yield of amine chromophores. We demonstrate this concept using quinine as a model. N-Alkylation causes changes in its emission that mirror those caused by changes in pH (i.e., protonation). Unlike protonation, however, alkylation of quinine's two N sites is performed in a stepwise manner to give kinetically stable species. This kinetic stability allows us to isolate and characterize an N-alkylated analogue of an ‘unnatural’ protonation state that is quaternized selectively at the less basic site, which is inaccessible using acid. These materials expose (i) the through-space charge-transfer excited state of quinine and (ii) the associated loss pathway, while (iii) developing a simple salt that outperforms quinine sulfate as a quantum yield standard. This N-alkylation approach can be applied broadly in the discovery of emissive materials by tuning charge-transfer states.

This journal is © The Royal Society of Chemistry 2020 measured using the relative method 1 and comparing to known standards of quinine sulfate 2 and 2aminopyridine 3 in aqueous 0.1 M H2SO4. The absorption spectra and photoluminescence spectra for the PLQYs were measured on a Shimadzu UV-3600 UV-VIS-NIR spectrophotometer and a Jobin Yvon Fluoromax or Fluorolog. The low temperature phosphorescence spectrum of Qn in a zeonex film was acquired by placing the sample in a Janis Research VNF-100 cryostat, used in conjunction with a Lakeshore 332 temperature controller, and exciting the sample with 337nm light from an N2 laser (LTBMNL 100, Lasertechnik Berlin) at 10 Hz. Sample emission was the directed onto a spectrograph and gated iCCD camera (Stanford Computer Optics). The X-ray single crystal data were collected at a temperature of 120.0(2) K using λCuKα radiation (λ = 1.54178 Å) on a Bruker D8Venture (Photon100 CMOS detector, IμS-microsource, focusing mirrors) diffractometer equipped with a Cryostream (Oxford Cryosystems) open-flow nitrogen cryostat. Both structures were solved by direct method and refined by full-matrix least squares on F 2 for all data using Olex2 4 and SHELXTL 5 software. All non-disordered non-hydrogen atoms were refined anisotropically, closely located disordered atoms in structure MeQn·BF4 were refined isotropically. The disordered atoms were refined with fixed site occupation factors 0.6 and 0.4. Hydrogen atoms in structure Me2Qn·2BF4 and in OH-groups of structure MeQn·BF4 were refined isotropically, the remaining hydrogen atoms in structure MeQn·BF4 were placed in the calculated positions and refined in riding mode. The absolute configuration of studied compounds was determined from anomalous scattering by calculating the Flack 6 (x) and Hooft 7 (y) parameters which should equal 0 for the correct absolute structure and 1 for the inverted model. Crystal data and parameters of refinement are listed in Table S1.   The mixture was then purified by reverse phase column chromatography (Teledyne Isco CombiFlash Rf+ system, 40 g C18-capped SiO2, 0.05 M HCl(aq) / MeOH, 0 -100% elution). After evaporation, the product was isolated as a yellow film (57 mg, 0.14 mmol, quantitative

MeQn·BF4
Crystals of MeQn·BF4 suitable for X-ray diffraction were grown by slow cooling of a saturated MeCN solution.

Me2Qn·2BF4
Crystals of Me2Qn·2BF4 suitable for X-ray diffraction were grown by slow cooling of a saturated MeCN solution.

Fig. S27
The phosphorescence emission spectra of a 1 wt% film of Qn in zeonex. The film was prepared by dropcasting a PhMe solution and recorded at 40 ms delay and 80 K.

Theoretical study Computational details
Ground-state geometry optimizations of Qn and its derivatives were performed both in vacuo and using an implicit acetonitrile solvent. We used a protonation of N1 and N2 instead of a methylation, but Table S7 below shows that the effect of such substitution is minimal for emission energies. We  (2)), 14 used along with a cc-pVDZ basis set. 15 ADC(2) calculations were performed with frozen core and resolution of identity approximations. To justify the use of a small basis set, tests were conducted for comparison with the larger aug-cc-pVDZ basis set. 16 Vertical transitions were calculated based on the optimized ground-state geometries using LR-TDDFT, with nonequilibrium linear-response IEFPCM solvation (acetonitrile). As our main focus is to determine the emission properties of the studied compounds, we also located minimum energy S37 structures on the potential energy surface of the first excited state (and also the second one for some specific cases). For these excited-state geometry optimizations, we used LR-TDDFT combined with equilibrium linear-response IEFPCM solvation (acetonitrile). Harmonic vibrational frequencies were computed for excited-state minima. The linear-response approach for implicit solvation models is known to provide reliable excited-state geometries, but less accurate emission energies. 17 Therefore, the emission energies were also computed with the state-specific approach as implemented in Gaussian09. ADC(2) excited-state optimizations were performed only in vacuo, and harmonic frequencies were not computed due to the high computational cost. To estimate the emission energy for the solvated systems, single-point ADC (2) computations with the conductorlike screening model (COSMO) 18 for solvation were performed a posteriori on the optimized gas phase geometries. This protocol relies on the assumption that the calculated excited-state structures are similar in gas and solvent phases. We also assume the equilibrium solvation limit, in which the electronic and nuclear degrees of freedom of the solvent adapt to the excited-state (rather than the ground-state) charge distribution of the solute. 19 It is important to note, however, that we do not attempt here to closely reproduce the experimental values with our calculations, but rather to detect the main qualitative changes in absorption and emission upon protonation of quinine.
Excited state characters were analysed by computing natural transition orbitals (NTO), 20 both with TDDFT and ADC(2). All DFT/LR-TDDFT calculations were performed with the Gaussian09 software, 21 while ADC(2) calculations were conducted with Turbomole 7.3.1 program package. 22 Molecular representations were produced with VMD 1.9.2. 23

Supporting calculations
Quinine can have different conformations, and its stable conformers were analyzed in detail in earlier theoretical and experimental work. 24,25 Here, we have mainly focused our attention on the so called cis-γ-open(3) conformer, which was identified as the most stable conformer based on jetcooled spectroscopy, vibrational circular dichroism and theoretical calculations. 25 The same conformer is found for the crystal structure of quinine. 26 The absorption spectrum of quinine contains both locally excited (LE) bright ππ* transition localized on the quinoline moiety, and a higher-lying charge-transfer (CT) state in which the electron is transferred from a n(N) orbital of quinuclidine to a π* orbital of quinoline (nπ*). As shown in Table   S3, the LE(ππ*) state is the lowest excited singlet state both with ωB97X-D and ADC (2). At the ground-state (i.e. Franck-Condon) geometry, the CT(nπ*) state lies higher in energy, and is likely to be overestimated by ωB97X-D. Using a larger basis set does not alter these trends -the relevant transitions are only slightly down shifted. Upon protonation of the N atoms having an available lone pair, the excited states are either shifted in energy or removed from the spectrum (as they cannot be formed anymore). Table S4 compares the absorption energies of the LE(ππ*) and CT(nπ*) states of Qn and its protonated derivatives, HQn + (proton on N1), iHQn + (proton on N2) and H2Qn 2+ (protons on N1 and N2) in acetonitrile solvent. Protonation of N1 prevents the formation of the CT(nπ*) states, while protonation of N2 causes a large red-shift of the LE(ππ*) and CT(nπ*) vertical transition (when latter exists). H2Qn 2+ exhibits both effects. The calculated vertical emission energies (Table S5) indicate that protonation of N2 causes a significant red-shift of the emission as compared to the pristine Qn. However, the LE emission of iHQn + is expected to be quenched due to the presence of the CT state, i.e., the molecule is likely to undergo nonradiative decay. This is consistent with our experimental findings for the analogous methylated derivatives, which found barely detectable fluorescence for iMeQn + . Interestingly, even in Qn the CT state falls down in energy upon geometry relaxation, and the nonradiative population exchange between LE and CT states becomes more likely (keeping in mind that CT energies are overestimated by ωB97X-D). Therefore, LE emission in Qn has very low intensity, while CT emission becomes almost equally pronounced (see Figure 1 in the main text). Protonation of N1 causes only slight red-shift in emission. Again, H2Qn 2+ benefits from two effects; the colour of its fluorescence is altered, while deleterious nonradiative effects are less likely due to the lack of CT state. We have also evaluated emission energies based on a state-specific solvation schemes (Table S6), in which solvent is supposed to fully equilibrate with the excited state from the solute, while the ground state of the solute is treated in a non-equilibrium limit. The underlying assumption is that the excited-state lifetime is sufficiently long such that the slow and fast degrees of freedom of the solvent have sufficient time to adapt. 19 One possible explanation for the too low emission energies of the CT states could be related to a breakdown of this assumption, the CT states being too short lived to allow for a full relaxation of the solvent (leading to an overestimation of relaxation effects in the calculations). This effect is particularly striking for Qn, which exhibits experimentally an emission band centred at around 2.2 eV, see Figure 1). We note here that such underestimation of charge transfer bands with state-specific implicit solvent models were reported in the literature. 27,28 Explicit solvent effects might be required to improve the description of this band. The LE states of Qn and iHQn + may also not have sufficiently long lifetime due to the nonradiative effects mentioned above, but the variations in LE emission energy with and without state-specific solvation are generally not very large.

Table S6
Emission energies from state-specific equilibrium solvation (acetonitrile). The optimised geometries are the same as those in Table S5, but additional single-point computations were conducted. Emission energies are given in eV.
ADC (2) Table S7 shows that methylation has almost the same impact on the calculated fluorescence as protonation.

Table S7
Comparison of vertical emission energies for the protonated and methylated Qn derivatives. Excited states were optimised with ωB97X-D/6-31G* and equilibrium linear-response solvation (acetonitrile). Emission energies are given in eV, along with corresponding oscillator strengths (f).