Efficient Triplet Exciton Fusion in Molecularly Doped Polymer Light‐Emitting Diodes

Solution-processed polymer organic light-emitting diodes (OLEDs) doped with triplet–triplet annihilation (TTA)-upconversion molecules, including 9,10-diphenylanthracene, perylene, rubrene and TIPS-pentacene, are reported. The fraction of triplet-generated electroluminescence approaches the theoretical limit. Record-high efficiencies in solution-processed OLEDs based on these materials are achieved. Unprecedented solid-state TTA-upconversion quantum yield of 23% (TTA-upconversion reaction efficiency of 70%) at electrical excitation well below one-sun equivalent is observed. When an organic light-emitting diode (OLED) is in operation, 75% of electron–hole recombination events are in spin-triplet configurations and only 25% are in spin-singlet configuration. In conventional fluorescent OLEDs,1, 2 the generation of photons is achieved from the radiative recombination of singlets. Relaxation of triplet excitons is quantum mechanically forbidden. Phosphorescent OLEDs overcome this restriction by allowing the triplet excitons to radiatively decay to the ground state.3, 4 This effect was achieved using strong spin–orbit coupling due to the presence of heavy metal elements in the molecular emitters. More recently, highly efficient OLEDs based on thermally activated delayed fluorescence (TADF) molecules were realized. In TADF emitters, triplet-to-singlet upconversion at room temperature is possible due to a small singlet–triplet energy gap.5 Another strategy to utilize triplet excitons in fluorescent OLEDs is through the generation of singlets by triplet–triplet annihilation (TTA) or “triplet fusion.”6-8 At the same time, photon upconversion through TTA has been widely considered for next-generation photovoltaic applications.9-14 Effective upconverters have been demonstrated using combinations of a triplet sensitizer such as platinum octaethylporphyrin (PtOEP)3 and a triplet acceptor/annihilator such as 9,10-diphenylanthracene (DPA),15 perylene,16 or rubrene.17 TTA-upconversion (TTA-UC) quantum yield (ratio of upconverted photons to absorbed photons) of 30% was observed in solution phase.18 This suggests that the intrinsic efficiency of the TTA-UC reaction (i.e., the probability of a triplet pair forming a singlet) in solution is >60%, much higher than the spin-statistical prediction of 25% (one pair of triplets collides to form one of the four states: one singlet S1 and three triplets T1, assuming higher triplet and quintet states are inaccessible). Practical application of solar energy conversion requires the TTA-UC material to be in solid state rather than in solution phase. However, the TTA-UC quantum yield of solid-state systems remains low, typically below 5%,10 and is moderately high (about 10%) in only one example.19 To achieve high efficiencies, high excitation intensities (200 mW cm−2 or above) are commonly required. These imply that in solid state, the experimentally observed reaction efficiency of TTA-UC was up to about 20%, below the 25% spin-statistical limit. To achieve highly efficient triplet fusion (TTA-UC) in OLEDs and other TTA upconverters, we consider four criteria for the selection of emitters: (1) high fluorescence quantum yield, (2) short singlet lifetime, (3) long triplet lifetime, and (4) the energy of two triplet excitons, 2E(T1) lies slightly above that of the singlet exciton, E(S1), but below the second triplet state, E(T2) (and also the energies of any spin-quintet states) (E(S1) ≲ 2E(T1) > kT). Therefore, the apparent lifetime of the triplet is expected to be significantly shortened when the TTA-UC process is efficient, as the efficient generation of singlets rapidly reduces triplet population. To estimate the monoexcitonic lifetime of triplets, we investigate the EL kinetics of later times (t ≈ 10 µs) when the triplet density is reduced significantly and monomolecular decay of triplets becomes dominant (kT >> kTTAnT). We infer a lower bound of 25 µs for the triplet lifetime in the rubrene device. Besides, increasing current pulse width (from 1 to 500 µs) at a fixed current density (1 mA cm−2) leads to an increase of the delayed EL intensity, from 16% to 51% (Figure 2c). This result agrees with our interpretation that the delayed EL originates from triplet fusion, in which the intensity of delayed fluorescence is related to the population density of triplets. We measured the magnetic field dependence of EL intensity for the device. The EL increases first when the magnetic field strength reaches 20 mT and reduces monotonically as the field intensity increases further (Figure 2d). This observation is consistent with TTA-induced magneto-EL behavior of OLEDs,26 further confirming that TTA plays an essential role in the EL of our devices. Figure 3a shows the emission spectra and photographs of the working LEDs. The DPA device shows efficient deep-blue emission peaked at 440 nm. Perylene, rubrene, and TIPS-pentacene devices exhibit green, yellow, and red EL centered at 520, 570, and 670 nm, respectively. The luminance–voltage characteristics are shown in Figure 3b. Peak luminance of over 7000 cd m−2 have been achieved with both rubrene and perylene. Maximum external quantum efficiencies (EQE) of about 6% or above have been obtained for DPA, perylene, and rubrene (Figure 3c and Table S1, Supporting Information). These values exceed the EQE limit (5%) of conventional electrofluorescence (assuming an optical outcoupling factor of 0.2),6-8 suggesting a significant contribution from triplets toward the total EL. We found that a high emitter doping concentration (20%) results in optimum OLED performance (Figure S2, Supporting Information), consistent with the view that a small intermolecular separation is required for efficient bimolecular TTA process. Increased doping concentration also leads to more pronounced lower-energy spectral features in the EL, as a result of enhanced intermolecular interactions. However, we note that the efficiency roll-off is significant at low current densities, indicating unbalanced charge injection and transport in these unoptimized devices. To improve charge injection and transport, we developed a solution-processed, inverted multilayer OLED structure (inset of Figure 3e) using ZnO electron-injection layer deposited from solution by atmospheric pressure spatial atomic layer deposition (AP-SALD).27 Interfacial energy level modification of the ZnO was achieved by a spin-coated polyethylenimine layer.28 For the emissive layer, we selected rubrene as the emitter (for its superior triplet fusion properties) and a conjugated polymer poly (9,9′-dioctylfluorene)-co-benzothiadiazole (F8BT) as the host matrix.29 We observed that the EL emission was dominated by the emission of rubrene (Figure S3, Supporting Information), confirming the effectiveness of the host–guest energy transfer. We deposited a hole-transport/electron-blocking layer of N,N′-bis(3-methylphenyl)-N,N′-diphenylbenzidine (TPD)30 from solution for the first time. This simply prepared layer effectively reduces exciton quenching by the MoOx/Au anode. The novel device architecture results in a very low, sub-bandgap turn-on voltage of 1.8 V and a high maximum brightness of 6 × 104 cd m−2 (Figure 3e). The maximum EQE obtained from the inverted FuLED is 6.3%, corresponding to a current efficiency of 20.7 cd A−1 (Figure 3f), outperforming other solution-processed rubrene OLEDs in the literature. The efficiency roll-off is not apparent across a wide range of current densities (up to 100 mA cm−2), indicating excellent charge balance. We also measured the transient-EL characteristics of the F8BT:rubrene device and compared it with a pristine F8BT device prepared in the same way. A significant difference in the EL kinetics was observed. It showed that the contribution of the delayed EL for the rubrene-doped F8BT was 55% when held at 10 mA cm−2, much higher than the 9% delayed EL contribution observed in the F8BT-only device (Figure 3g). This result suggests that the triplet-fusion process in rubrene-doped F8BT is more efficient than in pure F8BT. In contrast to the standard PVK:rubrene device, the delayed EL contribution in the inverted F8BT:rubrene device is relatively constant, at 50%–60%, across a range of current densities (1–100 mA cm−2). This, together with the insignificant efficiency roll-off, agrees with our previous discussion that the reduced TTA-fusion-related EL component at higher currents is due to unbalanced charge injection and transport, which are minimized in the inverted device architecture. The TTA-upconversion efficiencies and quantum yields of various TTA-UC emitters investigated in this work are summarized in Figure 4b. The highest ΦTTA-UC and ηTTA-UC for rubrene are 23% and 70%, respectively (For DPA and perylene, the efficiencies are slightly lower). To the best of our knowledge, these values are higher than that of any solid-state TTA-upconverters reported to date9-14, 19 and are comparable to some of the best TTA-UC efficiencies observed in solution-phase systems.9-12, 18 A summary of the efficiencies of solution systems17, 18, 31-33 is shown in Table S2 (Supporting Information). Comparisons of our devices with other solid-state TTA-UC systems13, 14, 19, 31, 34 and other TTA-enhanced OLEDs7, 8, 35-37 are presented in Tables S3 and S4 (Supporting Information), respectively. The remarkably high TTA-UC efficiencies in our best devices suggest that only singlets are formed during the TTA process. Besides, it is striking to note that the high TTA-UC efficiencies occur at very low excitation power densities (≈30 mW cm−2 for inverted structure and ≈0.001 mW cm−2 for standard structure, significantly lower than typical excitation power densities required for optical TTA-upconverters), indicating extremely efficient triplet formation and accumulation in these electroluminescent devices. Our findings suggest that the processes (such as intersystem crossing and triplet–triplet energy transfer) associated with triplet sensitizer molecules, which are widely used in optically excited TTA-UC systems, are some of the key limiting factors for achieving highly efficient TTA upconversion. The use of a wide-bandgap polymer host provides additional benefit for triplet confinement, ensuring high density and long lifetime of triplets at the same time. For TIPS-pentacene, TTA-UC efficiencies are moderate (ηTTA-UC of 35.1% and ΦTTA-UC of 6.8%), likely due to the competing process of singlet fission. However, the observation of the energetically unfavorable triplet fusion process in this material is noteworthy by itself. It is possible that the energy gap between E(S1) and 2E(T1) in TIPS-pentacene is close to kT, allowing endothermic triplet fusion reaction to effectively take place at room temperature. Finally, we find that DPA-, rubrene-, and perylene-doped polymer FuLEDs clearly exceed the fluorescence IQE limit of 25% (corresponding to an EQE limit of 5%). The highest IQE value obtained is >30% using the yellow fluorophor, rubrene. The largest percentage contribution of triplet fusion-related electrofluorescence is 60% in rubrene-based FuLEDs. It is by far the highest value observed in TTA-based OLEDs, consistent with the high TTA-UC efficiencies of the emitter molecules. Besides, triplet fusion-related EL contribution of as high as 38% was found in the singlet fission material, TIPS-pentacene. The very high TTA-UC efficiencies obtained in our FuLEDs and the method we developed to evaluate them provide valuable information for the design of high-efficiency solid-state TTA upconverters for photovoltaics. Our results demonstrate that conventional spin-statistical limits do not apply to these TTA-UC molecules in efficient devices. It also reflects the reciprocity of singlet fission and triplet fusion processes in materials such as TIPS-pentacene. Besides, our efficient, multicolor (deep blue, green, yellow, and red) FuLEDs produced using low-cost solution-processing methods may lead to further technological development toward large-area display and lighting applications. D.D. and L.Y. contributed equally to this work. D.D. and L.Y. conceived the project with R.H.F., designed the experiments, and performed all data collection and analysis. D.D. and L.Y. carried out the transient-EL experiments with help from J.R., and performed the magnetic-field EL and TCSPC measurements. D.D. and L.Y. developed and characterized the FuLEDs. D.D., K.M., R.M.A., and A.Y.A. developed the ZnO deposition using AP-SALD. J.L.M.-D. provided useful discussions. L.M. assisted with some experiments. D.C. and D.D. cosupervised L.M.'s work. D.D. and L.Y. cowrote the paper, which was revised by R.H.F. D.D. acknowledges the Department of Physics (University of Cambridge) and the KACST–Cambridge University Joint Centre of Excellence for financial support. L.Y. thanks the Singapore Agency for Science, Technology and Research (A*STAR) for a PhD studentship. The authors thank the Engineering and Physical Sciences Research Council (EPSRC) for financial support. As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.