Polaron pair mediated triplet generation in polymer/fullerene blends

Electron spin is a key consideration for the function of organic semiconductors in light-emitting diodes and solar cells, as well as spintronic applications relying on organic magnetoresistance. A mechanism for triplet excited state generation in such systems is by recombination of electron-hole pairs. However, the exact charge recombination mechanism, whether geminate or nongeminate and whether it involves spin-state mixing is not well understood. In this work, the dynamics of free charge separation competing with recombination to polymer triplet states is studied in two closely related polymer-fullerene blends with differing polymer fluorination and photovoltaic performance. Using time-resolved laser spectroscopic techniques and quantum chemical calculations, we show that lower charge separation in the fluorinated system is associated with the formation of bound electron-hole pairs, which undergo spin-state mixing on the nanosecond timescale and subsequent geminate recombination to triplet excitons. We find that these bound electron-hole pairs can be dissociated by electric fields.

(1:2) measured at 1000 nm with 635 nm, 3.1 µJ.cm -2 excitation. The kinetics were recorded under nitrogen (before and after oxygen measurements) and oxygen atmospheres. The signal amplitude and decay was found to be completely reversible when measured under a nitrogen atmosphere (green and red decays) after the oxygen quenching experiment (blue decay), thus indicating low sample degradation during the duration of our measurements.
The kinetic (measured under N 2 ) was fitted with a ( ) function to account for the mono-exponentially decaying polymer triplet exciton and the power law obeying polaron signal decay. The triplet contribution to the overall transient absorption signal was thus extracted. A comparison of the triplet absorption at 300 ns after light excitation in the neat polymer and blend film shows that triplet generation in the blend is 3 times more efficient than in the neat polymer film. S 1 (S1) and T 1 (T1) are the lowest-energy singlet and triplet excited states, respectively, calculated for an isolated oligomer. The 1 CT 1 (1CT1) energy is the energy of the Coulombically-bound electron-hole pair across the interface. The triplet 3 CT 1 state is not shown but its energy is calculated to be almost degenerate with the singlet 1 CT 1 .   spectra was estimated to be 0.14 eV consistent with our quantum chemical calculations.

Supplementary
Electroluminescence was measured using a spectrograph (Shamrock 303) combined with a InGaAs photodiode array (iDUS) cooled to -90 °C. Electroluminescence spectra from blend and pure polymer devices were measured at 11 mA/cm 2 .

Supplementary
).   The torsional potential for the SiIDT-DTBT similarly has a minimum for planar structures and is shallower than that for Si-IDT-2FBT. Both polymers can potentially form a variety of different conformations that are compatible with the optimum planar structure. We consider just two, (1) where the two thiophene units or fragments flanking the BT are both oriented so that their sulphur atoms point away from the thiodiazole unit, which we denote as 'wavy' and (2) where the two thiophene units or fragments flanking the BT are oriented in opposite directions, which we denote as 'linear'. The particular structures studied are shown in Supplementary Figure 7. Although for both systems the 'wavy' conformer is found to be more stable from the gas phase calculations, we choose to consider henceforth only the 'linear' conformers. We select these structures because the linear conformers are better able to organise into ordered domains, as required from the observed tendency of SiIDT-2FBT to crystallise and because the difference in the gas phase energies relative to the minimum energy 'wavy' conformers is only 0.013 and 0.026 eV per repeat unit for SiIDT-2FBT and SiIDT-DTBT, respectively. The higher tendency of SiIDT-2FBT to crystallise is probably influenced by the planarisation induced by fluorination.
In the next stage, the structure of the PC70BM molecule is optimised, using DFT with First, we notice that in both cases the driving force for charge separation via the lowest 1 CT 1 states, quantified as the difference between the first singlet and the lowest CT state, 1S-1 CT1, is small (0.21 eV for SiIDT-DTBT and 0.15 eV for SiIDT-2FBT) (note that for P3HT it is ~0.9 eV) 4 ; however this driving force is slightly higher (by ~0.06 eV) for the SiIDT-DTBT:PC70BM system. In addition, for both model systems the triplet energy (T 1 ) is significantly lower (by 0.35 eV in both systems) than the lowest CT state ( 1 CT 1 ). Both the small 1S-1 CT1, energy and the substantial 1 CT1-1T energy are detrimental as they tend to limit charge separation and favour recombination to triplets. However, there is one feature that differs in the two systems: for SiIDT-DTBT:PC70BM the first higher lying ('hot') CT-state (namely 1 CT 4 ) is almost resonant (~0.01 eV higher) with the lowest oligomer singlet (S 1 ), while for SiIDT-2FBT:PC70BM the 1 CT 4 state lies 0.1 eV higher and is therefore less accessible energetically from the S1. The energy alignment of the S 1 and 1 CT 1 remains almost unaltered when sliding the PC70BM by 2 Å away from the initial position along the Importantly, the hole wavefunction of the 1 CT 4 state is more delocalised than in the 1 CT 1 state in each case, so that the average electron-hole separation is higher when the system lies in the 'hot' rather than the 'cold' CT states. The delocalisation of the hole density along the oligomer backbone can be expected to reduce the net Coulomb interaction and so improve charge separation. This advantage would only apply in the SiIDT-DTBT:PC70BM system since the hot state is not accessible from the singlet of the SiIDT-2FBT. Transient photocurrent was performed at short-circuit under different illumination intensities, as described previously. Through the method of differential charging it is also possible to obtain the charge carrier density at open-circuit. 5 The J-V curve can be described as the competition between a generation flux ( ) and recombination current ( ) such that ( ) ( ). Assuming field independent generation and no non-geminate recombination at short circuit, we make the approximation . Assuming only non-geminate losses as measured by CE and TPV we calculate ( ) using: Supplementary equation 1: ( ) where e is the electronic charge, d is the active layer thickness and , and are experimentally derived constants defined previously. As shown in Figure 4b the resulting J-V reconstruction is a poor match to the experimental data. For the field dependence observed in SiIDT-2FBT/PC70BM, a field dependent geminate recombination is included in the generation term, such that ( ) ( ), where has the form of a quadratic. In order to convert directly between optical density measured in field dependent TAS and current density, the generation profile was referenced relative to at 0.5 sun illumination and scaled linearly with light intensity as previously described by Credgington et al. 6 Supplementary References: