Strong modification of the transport level alignment in organic materials after optical excitation

Organic photovoltaic devices operate by absorbing light and generating current. These two processes are governed by the optical and transport properties of the organic semiconductor. Despite their common microscopic origin—the electronic structure—disclosing their dynamical interplay is far from trivial. Here we address this issue by time-resolved photoemission to directly investigate the correlation between the optical and transport response in organic materials. We reveal that optical generation of non-interacting excitons in a fullerene film results in a substantial redistribution of all transport levels (within 0.4 eV) of the non-excited molecules. As all observed dynamics evolve on identical timescales, we conclude that optical and transport properties are completely interlinked. This finding paves the way for developing novel concepts for transport level engineering on ultrafast time scales that could lead to novel functional optoelectronic devices.

(a) Distribution functions ΔPhole of the polarization energies Phole calculated for different electrostatic dipole moments p. The dipole moments p model the optically excited excitons in the C60 film. The distribution functions ΔPhole were extracted from 2D polarization energy maps Phole(x,y) of the surface layer of a C60 face center cubic (fcc) (111) crystal (slap of 6 layers). These maps were calculated for a regular array of excitons located in the first three surface layers with changing electrostatic dipole moments p and a constant exciton density 0.010 excitons/C60. The distribution function ΔPhole obtained for the electrostatic dipole p=p0 corresponds to the polarization energy distribution function ΔPhole at t0, i.e., right after the optical excitation, in analogy to the distribution functions shown in Fig. 4 of the main manuscript (see also Method section of main manuscript). The decay process of the excitons in the molecular film was described by a reduction of the strength of the dipole p. It is clearly visible in panel supplementary figure 3(a) that decreasing the strength of the dipole moment p results in a continuous reduction of the FWHM of the distribution functions ΔPhole. but not in an energy shift of the distribution function. The FWHM decreases linear with decreasing strength of the dipole moment for all exciton densities as shown in panel (b) and becomes zero as soon as the dipole moments vanish, i.e., p=0. Panels (c)-(f) illustrate our approach to determine the temporal evolution of the transient linewidth broadening of all molecular levels upon optical excitation in the framework of the extended micro-electrostatic simulation. For a more detailed description of our calculations, please refer to the supplementary note 1.

Supplementary Figure 4. Pump spectrum
The pump beam spectrum after second harmonic generation has been measured using a spectrometer. The center of mass of this spectral distribution is c = 391.3 nm, corresponding to ph,pump = 3.17 eV and the width is FWHM = 4.8 nm, corresponding to ph,pump,FWHM = 0.04 eV.

Supplementary Notes
Supplementary Note 1 To describe the time-dependent broadening, we have selected three discrete dipole moments p and one exciton density ρex to model the charge defects caused by the LUMO+1*, the LUMO* as well as the singlet exciton level. For each dipole strength p, we calculated the distribution function ΔP hole of the polarization energy (inhomogeneous broadening) using our micro-electrostatic simulations (see also Method section of main manuscript). The overall broadening of the distribution function in the presence of three different types of excitons is calculated by a linear superposition of the three distribution function ΔPhole for three dipoles (LUMO+1*, LUMO*, and singlet) multiplied by their relative occupation fi(t): Note that the simulated curve was scaled by factor 1.5 to quantitatively match the experimental findings. Such small quantitative differences are frequently observed when comparing experimental data with micro-electrostatic simulations due to the simplicity of the model [2]. However, the overall shape of the simulated transient broadening ΔFWHMsim(t) can perfectly describe the temporal evolution of the experimental data and hence can qualitatively explain the observed phenomena. The excellent qualitative agreement between our micro-electrostatic simulations and our experimental findings allow us to gain insight into the charge character of the exited states in C60. The parameters of our simulations strongly suggest that the excited level LUMO+1* and the LUMO* exhibit a significant (microscopic) charge distribution and can hence modify the energies of the transport levels of the surrounding molecules. Such a charge distribution is characteristic for excitons with dominant charge transfer exciton character. In contrast, the vanishing dipole moment of the singlet level in our simulations indicates that the corresponding singlet exciton does not reveal a nonvanishing charge distribution. This is consistent with an exciton of dominant Frenkel exciton character.

Details on the Spectral Analysis of Excited States
To extract the transient occupation of the different excited states from the raw data acquired in the angle-resolved mode of our photoemission spectrometer, we generated energy distribution curves (EDCs) for each time delay by integrating over the angular information obtained in our experiment (14° acceptance angle) and by averaging these 1D spectra over all delay scans recorded for one pump laser fluence. For the 1D spectra, an exponential background was subtracted for every step in the time delay scan. The data obtained this way is plotted in Fig. 2b and c for distinct time steps. Subsequently, we analysed the spectral shape of the exited states by fitting each spectrum individually with three Gaussian curves. During the fitting procedure, the peak width (FWHM) and the peak positions were constrained to constant values. The peak positions were taken from [3], the FWHM was optimized by iteratively repeating the fitting procedure with different, but timeindependent FWHM values. The only free fitting parameter is the peak intensity. The best fitting results are shown in Fig. 2c for selected time steps, the transient intensity of all three excited states in Fig. 2d.

Details on the Analysis of the Population Decay constant of the excited states
The population decay constants of the excited states LUMO+1* and LUMO* were obtained by fitting their time dependent intensity evolution by the exponential fitting function FLUMO+1*(t) and FLUMO*(t):

Details on the Analysis of the Transient Broadening of the Occupied Valence Band States
The time constants of the temporal evolution of the transient broadening of the occupied valence levels was obtained by a very similar fitting procedure as discussed above for the excited states. The which has been convoluted with a Gaussian function with ΔτFWHM = 70 fs to consider the temporal broadening of the pump and probe pulse (crosscorrelation of pump and probe). The fit function describes a double exponential decay process to model the two subsequent decay steps observed in the excited states. Again, the best fitting results were obtained for τLUMO_R =τLUMO+1*. The best fitting result is shown in Fig. 3f of the main text.