Thermally activated delayed fluorescence with 7% external quantum efficiency from a light-emitting electrochemical cell

We report on light-emitting electrochemical cells, comprising a solution-processed single-layer active material and air-stabile electrodes, that exhibit efficient and bright thermally activated delayed fluorescence. Our optimized devices delivers a luminance of 120 cd m−2 at an external quantum efficiency of 7.0%. As such, it outperforms the combined luminance/efficiency state-of-the art for thermally activated delayed fluorescence light-emitting electrochemical cells by one order of magnitude. For this end, we employed a polymeric blend host for balanced electrochemical doping and electronic transport as well as uniform film formation, an optimized concentration (<1 mass%) of guest for complete host-to-guest energy transfer at minimized aggregation and efficient emission, and an appropriate concentration of an electrochemically stabile electrolyte for desired doping effects. The generic nature of our approach is manifested in the attainment of bright and efficient thermally activated delayed fluorescence emission from three different light-emitting electrochemical cells with invariant host:guest:electrolyte number ratio.


Supplementary Note 1
The singlet-triplet energy gap (ΔE ST ) was estimated as the difference in energy of the PL peak at 300 K ("the fluorescence peak") and the delayed PL peak at 77 K ("the phosphorescence peak"), as presented in Supplementary  Figure 6d-f) the corresponding data for the device-optimized active material, with the TADF guest identified above the graphs. The radiative lifetime of prompt fluorescence (PF) and the radiative lifetime of delayed fluorescence (DF), as well as the singlet radiative rate (kr S ), the intersystem crossing rate (kISC), the reverse intersystem crossing rate (kRISC), and the triplet non-radiative rate (knr T ), were calculated with the procedure detailed in Supplementary Reference 1 using the PL intensity transients at 300 K (see Figure 5 in main manuscript and Supplementary Figure 5; note that the calculation of PF was done with PL transient data with a higher temporal resolution). Supplementary Figure 9. Performance recovery as a function of pristine operation time. The ratio between the peak luminance for a device rested for 24 h and the turn-off (a) or peak (b) luminance for the same device during its pristine operation as a function of pristine operation time. All devices were driven by a constant current density of 100 A m -2 . The presented data are the average for N investigated devices, with the values for N identified to the right of the graphs, and with the error bars representing the standard deviation.

Supplementary Note 2
The reversibility of the device performance was investigated by first driving pristine devices at constant current for a set time (between 5 and 20 min), then resting the device for 24 h at open circuit, and finally recording the luminance and voltage transients a second time at the same constant current. Supplementary Figures 7-8 show that the drive voltage is consistently starting at a slightly lower value during the second run (compare dash-dotted line with solid line), which implies that either the first pristine operation has resulted in faster ion-redistribution kinetics or that some minor doping remains in the active material following one day of resting at open circuit. The situation for the luminance is more complex, since the performance of the rested devices depends strongly on the operational time of the pristine device, with a short pristine operational time of 5 min resulting in an essentially retained peak luminance while a longer operational lifetime of 15 min is concomitant with a drop in the peak luminance.
To distinguish between reversible and non-reversible changes induced by the initial pristine operation, we present in Supplementary Figure 9 the peak luminance during the second run normalized by the luminance of the pristine device during the first run at turn-off (Supplementary Figure 9a) and at peak luminance (Supplementary Figure 9b); these data are presented as a function of the pristine operational time. Supplementary Figure 9a thus provides a measure of the reversible recovery of the final luminance at the end of the first operational cycle, whereas Supplementary Figure 9b contributes with information on non-reversible effects on the peak luminance. We find that a 5 min initial operation of the pristine devices results in a complete reversibility of the peak luminance. With increasing pristine operational time, this situation changes and we observe a significant reversible recovery of the final luminance in the pristine run ( Supplementary Figure 9a) but also a permanent degradation in the form of a lowered peak luminance (Supplementary Figure 9b).
Our conclusion is thus that the observed drop in luminance during the operation of pristine TADF-LECs originate in a combination of reversible and non-reversible effects. Based on preceding studies in the field, we suggest that the reversible recovery is due to a spatial separation of the doping/polaron and exciton populations during the off period, which results in lowered exciton-polaron quenching and increased efficiency. We further speculate that the highenergy species that is the result of the exciton-polaron and exciton-exciton interactions is (at least partially) the cause of the non-reversible change of the peak luminance via a degradation of the luminance capacity of the TADF emitter.

Supplementary Note 3 -Estimating the steady-state doping structure
Supplementary Reference 2 derived Supplementary Equation (1) for the calculation of the number of ions ( ions ) in a host-guest LEC required for the doping of all of the guest molecules and % of the host molecules in the doped transport regions sandwiching the p-n junction, with the latter being undoped and occupying % of the activematerial thickness: guest and host represent the total number of guest and host molecules, respectively, in the active material. The doping of all of the guests in the n-type/p-type transport region is equivalent to the filling of all of the electron/hole traps.
By setting = 0.2 and inserting the above number values into Supplementary Equation (1), we obtain = 0.15 for all three systems. In other words, we find that the optimized LECs feature a steady-state doping structure in which all guest traps and 15 % of the host units are doped in the doped transport regions surrounding the light-emitting (and undoped) p-n junction.