Optimize Hyperfluorescent and TADF OLEDs by Reducing Excitonic Loss Channels

This work has been carried out in collaboration with Cynora GmbH.

Introduction

The efficiencies of organic light-emitting diodes (OLEDs) have continuously improved over the last decades. A short overview of the history of OLED technologies and improvements is given in an earlier blog post on TADF and hyperfluorescent emitters. The simulation and characterization results discussed in this blog are done on devices that use thermally activated delayed fluorescence (TADF) molecules either directly as emitters or intermediate molecules for energy transfer to the final fluorescent emitter (hyperfluorescence).

Compared to previous designs, hyperfluorescent OLEDs involve an increased number of excitonic processes. The increased complexity makes it difficult to simply estimate loss channels and optimize external quantum efficiency without the help of an electro-optical simulation tool.

Here, we analyze the measurements and discuss the simulation of four TADF/hyperfluorescent OLEDs with varying degrees of doping density of the TADF and hyperfluorescent constituents. Our study shows how the different doping densities influence charge trapping / charge balance and therefore excitonic loss channels as well as the shape of the recombination zone. This detailed analysis allows us to further optimize the device stack, specifically the guest doping concentrations, in order to improve the external quantum efficiency without the need for time-consuming trial and error experiments.

Devices and Simulation Approach

Four different devices, 1-4, with varying guest doping concentrations have been produced by Cynora GmbH.

Two samples have only one guest dopant (GD-1) in the emitting layer with either 15% or 30% concentration and two devices have an additional guest dopant 2 (GD-2) with a fixed concentration of 1% (see also table on the right). GD-1 is a TADF molecule, whereas GD-2 is a fluorescent molecule onto which the singlet exciton of GD-1 is transferred via Förster resonant energy transfer.

Figure 1 shows the position of the energy levels for electrodes and semiconducting materials of all devices. The position of the HOMO/LUMO levels of all the layers is one of the information required to perform the device simulation.

Each device has 16 subpixels. The experiments were performed with the all-in-one measurement platform Paios. Measurements of the subpixels have shown perfect reproducibility of the devices. At the same time, no device degradation has been seen during the measurements and results have been reproducible over several weeks. Therefore, these measurements are a perfect basis for electro-optical simulations to yield consistent, accurate, and robust results. For the simulation, we use the simulation software Setfos. This tool allows fully coupled 1D electro-optical simulations of the different OLEDs, including all excitonic processes.

The performed measurements with Paios show that the two guest dopants have a profound effect on the electrical properties.

GD-1 is responsible for the electron transport in the emitting layer (EML), therefore increasing GD-1 content enhances the electron mobility. At the same time, it reduces the hole mobility as hole transport is mediated through the host concentration which is reduced with increasing GD-1 content. GD-2 acts as traps for holes whereas it affects the apparent charge mobility for both charge carrier types. Because of these findings, we have used the HOMO level of the host and the LUMO level of GD-1 as transport levels for holes and electrons, respectively, in Setfos. For devices with GD-2 content, a hole trap was enabled with an energetic depth given by the HOMO level difference of host and GD-1 material.

Figure 2 summarizes the EML parameters which were affected when going from one device to the next one. Otherwise, the simulation parameters have been kept the same for all simulated devices. This “global fit” approach is very important as it improves the reliability of established device model.

Figure 2. Required modifications of the EML simulation parameters to reproduce the respective measurements, when going from one device to the next one. All other simulation parameters were kept constant.

Steady State Measurements and Simulation

Measurements and simulation of JV curves show an excellent agreement over several orders of magnitude. The characterization and simulation are shown in Figure 3 (left-right) on linear and logarithmic scale.

Figure 3 (left-right). Simulation (squares) and measurement (lines) of the JV curves in a linear (left) and logarithmic plot (right). The excellent agreement exemplifies the reliability of the parameter set.

The agreement between experiments and simulations corroborates the reliability of the extracted parameter set and makes the conclusions significant and robust. Changing the GD-1 content from 15% to 30% changes the charge balance in the EML and leads to a split recombination zone. The simulated charge recombination profile is shown in Figure 4. This finding has been confirmed further by optical characterizations on the same samples (not shown here).

Figure 4. Simulated charge recombination profiles. Only the emitting layer is shown here. The left edge corresponds to the interface to the electron blocking layer, the right edge to the hole blocking layer (HBL). For devices with low GD-1 content (samples 1 and 2) charge recombination only occurs at the interface to the HBL whereas the devices with a high GD-1 content (samples 3 and 4) show a distributed recombination over the whole layer with peaks at the interfaces to the neighboring layers.

Doping the emitting layer with GD-2 (samples 2 and 4) results in a lower current density (see Figure 3). This current density reduction is due to the hole trap states in the emitting layer which are created by the GD-2 molecules. We assumed that each GD-2 molecule introduces a hole trap state resulting in 10^19 traps/cm3. The trap depth is assumed to be 0.36 eV, which corresponds to the energy difference between the HOMO levels of the host material in the EML and the GD-2 molecules. We found a good agreement between simulated and measured steady-state curves when we fixed these two parameters (Figure 3). Figure 5 shows the resulting hole, hole trap and electron profiles in the EML for all 4 devices. The trapped hole densities are much higher than the free charge carrier densities which explains the reduced current. Moreover, as we will see below, these trapped charges can also act as quenching sites for excitons which can affect the roll-off behaviour.

Figure 5: Charge density profiles of free holes (continuous lines) and trapped holes (dashed lines) in the emitting layer. In the simulation, we assumed hole traps with a density of 10^19 cm-3, which is equal to 1% of the available states and exactly corresponds to the GD-2 content for these samples. Introducing these traps results in a situation in which most holes in the emitting layer are trapped. These trapped charge carriers do not contribute to charge transport and effectively reduce the number of free holes by 1 to 2 orders of magnitude.

Capacitance-Voltage and Polar Emitting Layer

Measurements and simulation in the frequency domain give additional insight into the electric properties of the OLED stacks.

Changes in the GD-1 content results in a shift of the capacitance onset by 0.5 to 1 V (Figure 6). This large shift cannot be explained from insights from the steady-state measurement and simulation alone. We were not able to explain steady-state and frequency-dependent data at the same time, neither by increasing the electron mobility nor by modifying the electronic barrier. Thus, we had to extend the model.

Many organic molecules are known to have a microscopic dipole moment which can lead to a polar layer. In our case, samples 1 and 2 only differ by the content of GD-1. Therefore, we assume that GD-1 is partially oriented inside the EML and that this alignment is leading to a macroscopic polarity. The sheet charge densities that led to the best agreement between measurement and simulations are 0.2 and 0.4 mC/m^2 for devices with 15 and 30% GD-1 content, respectively. These values are in the range of other doped emitting layers but about 3-8 times smaller than pure Alq3 layers [1, 2]. DFT calculations on the GD-1 molecule have shown this to have a permanent dipole moment of around 5.5 debye. This permanent dipole moment in the GD-1 molecule itself can lead to the observed polar layer in the CV measurement.

Figure 6. Capacitance voltage measurement (filled circles) and simulation (empty circles) at 10 kHz. Devices with high GD-1 content (3 and 4) show a clear shift of the capacitance onset towards lower voltages. In the simulation this could only be reproduced by assuming a polar EML. The polarity increases with the GD-1 content. The sketch on the right shows how the polar EML, which scales with GD-1 concentration, was modeled.

Current Efficiency Roll-Off and Exciton Processes

As mentioned above, hyperfluorescent OLEDs involve an increased number of excitonic processes which occur between charge recombination and light emission. Understanding these complex processes and improving the resulting OLED efficiency requires a comprehensive model covering charge transport, excitonics, and photon emission.

With Setfos we fitted the simulated current efficiency roll-off to the measured curves (Figure 7). By including triplet-triplet annihilation (TTA) and triplet-polaron quenching (TPQ) we were able to explain the roll-off for the samples without GD-2 content. GD-2 introduces traps in the emitting layer and, therefore, an additional mechanism of triplet quenching with trapped holes has been considered for those two samples. The four samples have been simulated with the same annihilation/quenching parameters.

Figure 7. Current efficiency roll-off due to exciton annihilation and exciton quenching processes. Triplet-triplet annihilation and triplet-polaron quenching is able to explain the roll-off in the samples without GD-2, while an additional quenching of triplets at trap holes needs to be considered for samples with GD-2. All samples have been simulated with the same annihilation/quenching parameters.

Device Optimization by Simulation

To further improve the overall device efficiency vs. current density, we analyze the effect of different hypothetical GD-1 and GD-2 doping concentrations. In order to test the effect of the GD-1 concentration, we interpolate and extrapolated the related material parameters of the EML, i.e. electron mobility, the density of states, and emission layer polarity, and simulated the device performances with these varied EMLs. The simulated current efficiency vs. current density is shown in Figure 8.

Already at low GD-1 concentrations (~ 25%), the efficiency starts saturating at around 58 cd/A for low current densities. An increase of GD-1 content would further increase the efficiency at high current densities. A GD-1 content of 35% is found to be the optimum as it leads to the highest efficiency over the widest current density range. A similar simulation-based optimization can also be performed for the GD-2 concentration or even for both guest dopants together.

Figure 8: Simulated current efficiency vs. current density for varied GD-1 concentration in the EML. The 15 and 30% results (symbols) agree with the experimental data. The simulation suggests that a doping concentration of 35% GD-1 would be optimal, as it provides the highest efficiency and lowest roll-off.

Conclusion

We have shown that a combined experimental/simulation study on a set of slightly different OLED samples allowed us to investigate and understand the working principles of a TADF/hyperfluorescent OLED stack to a greater extent. For the simulation, we used the simulation software Setfos and for the measurement, we used the all-in-one measurement platform Paios.

Steady-state simulations revealed that increasing the TADF dopant in the emitting layer changes the emission zone from a localized to a split emission zone, which has been validated by optical measurements. Capacitance voltage measurements further showed that the emitting layer becomes more polar if guest doping (GD-1) concentrations are increased. From the analysis of the measured current efficiency roll-off via simulations we conclude that triplet-triplet annihilation, triplet-polaron quenching, and triplet quenching at traps are the main processes determining the roll-off characteristics. By combining all these insights we systematically varied the GD-1 doping concentrations in the simulations to maximize the OLED efficiency. The optimal doping concentration was found to be around 35% which would significantly increase the current efficiency at higher current densities. Similar optimization steps can also be performed to optimize GD-2 concentrations as well as to find a global optimum of GD-1 and GD-2 concentrations.

References:

1. Brütting, W. & Paasch, G. Interfacial charges and electric field distribution in organic hetero-layer light-emitting devices. Org. Electron. 1, 41–47 (2000). https://doi.org/10.1016/S1566-1199(00)00007-0

2. Morgenstern, T. et al. Correlating Optical and Electrical Dipole Moments to Pinpoint Phosphorescent Dye Alignment in Organic Light-Emitting Diodes. ACS Appl. Mater. Interfaces 10, 31541–31551 (2018). https://pubs.acs.org/doi/10.1021/acsami.8b08963

Back to the top