Analysis of TADF OLED degradation by combining advanced characterization and simulation

Despite the success of Organic Light-Emitting Diodes (OLED) displays in commercial products, there are still issues regarding the efficiency and lifetime of these devices. Especially the blue emission color is currently lagging behind the performance and reliability of the green and red pixels which affects the overall power consumption and pixel layout design. Therefore, it is of utmost importance to better understand the device physics in general and the origins of degradation in these OLED devices in particular. Organic light-emitting diodes are multilayer devices in which several different physical processes occur at the same time. Therefore, simple models and analytic formulas can only be used to a limited extend. This is one of the reasons why we make use of device simulations in combination with advanced characterization.

In this blog post, we will show you an example for which we used our tools Paios and Setfos to understand device limitations and analyze the degradation mechanisms in state-of-the-art sky blue TADF OLED devices. The same analysis can be carried out on the new perovskite LEDs, by following the protocol that is presented in this tutorial.

The concept that we employ to analyse the degradation phenomena is sketched in Figure 1.

Figure 1. Concept that is used to analyse and understand degradation mechanisms in full OLED devices.

As usual, we stress the OLED electrically. This can be done by either applying constant voltage or – more commonly – constant current. During this time, the decay in emission intensity and current density (or rise in device voltage) is continuously monitored. So far this is not really remarkable. What we do differently is that interrupt the constant current/voltage stressing regularly and perform various steady-state, impedance, and transient experiments. This allows one directly to identify or exclude several degradation pathways.

In order to analyze the layer degradation mechanism, an electro-optical device model is set up. We use Setfos to simulate the response of the fresh OLED and adapt the parameters until the agreement with the experiment is convincing. If you want to learn more about this fitting procedure, please visit our OLED case study page or check the corresponding open-access publication.

In the simulation, we can then easily vary individual material parameters in order to understand the effect on the IV, transient EL, impedance, and the CELIV data. This simulated data are then be compared with the experimental results of the degraded cell which allows us to analyze which parameters changed during OLED degradation. This approach is not only really versatile, most importantly it allows us to understand the degradation in the full OLED device!

The OLEDs investigated in this study have the following layer sequence: ITO/NPB/TCTA/ EML/NBPhen/LiQ/Al (see Figure 2). Most materials are well known. This structure is typically used by the company Cynora to test new emitter material in a full device. The EML consists of a mCBP host with a 20 vol% TADF emitter inside. Both the NPB and the NBPhen layer were varied systematically in thickness. In total 8 different OLEDs were characterized. Selected measurements of the fresh devices are shown in Figure 2.

All these measurements were performed with our all-in-one characterization platform PAIOS.

Figure 2. Left: OLED layer stack and legend for the different layer thickness combinations. Right: Selected DC, AC and transient measurements on the fresh device.

Thanks to the thickness variation, we could already learn something about those devices. The capacitance-frequency rise at 3 V, the CELIV peak as well as the TEL onset show all a systematic shift with decreasing NPB layer thickness. This is an indication that we see in these devices the effect of hole transport. This observation will also be important for the analysis of degraded cells.

These measurements have now been simulated with the simulation software Setfos. The fitting process is here not further discussed. The interested reader is referred to the corresponding tutorial. We have been able to reproduce all relevant features observed in the measurement by our device model. It can describe 8 devices and 4 different measurement techniques which give us confidence about the determined material and device parameters. This is also important for the following degradation analysis.

The final fit for device S42 is shown in Figure 3. The same device was now stressed with a constant current of 15 mA/cm^2 while the voltage rise and emission decay were recorded.

Figure 3. Final fit of device S42. Several experimental characteristics are fitted with a global analyis of the sample in stady-state and transient regimes.

At specific points in time - which are visible as luminance dips in Figure 4 - a full Paios measurement routine was performed. A few of these advanced characterization results at different points in time are shown in Figure 4 on the right-hand side.

Paios is allowing us to study the degradation of the OLED with several characterization techniques without touching the sample and, therefore, introducing experimental errors in the analysis.

Figure 4. Emission intensity reduction during constant current stressing (left) and measurement result of the advanced characterization during stress interruptions (right).

Degradation can occur due to various reasons. A few of them are listed in the following. One by one we will test whether they are also occurring in our TADF OLED.

•        External contact modifications

•        Formation of parasitic current pathways

•        Formation of exciton quenching sites

•        Charge transport/balance modification

•        Formation of non-radiative recombination centers (traps)

The first two can be excluded quickly as we do not see a modification of the series or shunt resistance in the IV, the C-f and the dark-CELIV curves (Figure 5).

Figure 5. Interrupted IV, C-f (V=0V) and dark-CELIV measurements. No signatures of a modified series or parallel resistance is observed.

A more commonly observed degradation phenomenon is the formation of exciton quenching sites. In order to evaluate this, we look at the transient EL decay as a function of stress time. The lifetime extracted from the first decay is 5 us which corresponds to the delayed recombination lifetime of the TADF emitter, determined from transient PL experiments. Importantly, we do not see a change in this time with increased stress time.

Figure 6. TEL decay for various stressing times. The inset shows the calculated lifetimes from the delayed component as a function of stressing time. The lifetime of 5 us is in agreement with the TrPL results (not shown) and does not change over time.

If exciton quenching sites would be formed during degradation, we would see a change in non-radiative rates of singlet and/or triplets. Such a change would in turn result in a modification of the delayed recombination lifetime as seen from the formula below or also when changing those parameters in the simulation. Therefore, we can also cancel out this degradation mechanism from our list.

Next, we consider charge transport modifications. Indeed we see a change in the transient EL and in the CELIV signal. Both those signals can be used to extract and apparent mobility and we see that this value is decreasing over time.

Figure 7. TEL rise and CELIV results for devices after different stressing time. The apparent mobility that can be extracted from those measurements is clearly decreasing over time.

Thanks to the thickness variation, we can assign the observed changes to hole injection/transport modifications. However, we cannot say which layer or barrier is involved when we only look at the experiments. Therefore, we will now make use of the device simulations, presented in Figure 5. In order to determine the material parameter which is causing the charge transport decrease during degradation, we will systematically vary all potential parameters in simulation and compare the trends with the experiments. We will show the concept just for two parameters, i.e. hole traps and mobility in NPB.

In Figure 7 can see that an increase in NPB hole trap density would lead to a higher overshoot in the TEL signal. This is not what we see in the experiment. Similarly, the simulated trend in injection CELIV is different from the experimental data. Only the impedance signal could somehow be explained by the increasing hole traps in the NPB layer. But overall, an increasing NPB trap density cannot be the reason for the observed changes during degradation. Again, we see that it is highly important to consider different measurement techniques.

Figure 8. Experimental modifications of TEL rise, injection CELIV and C-f data upon OLED degradation (centre). On the left and the right, simulation parameters (NPB traps and mobility) have been modified in order to check whether these parameters could be responsible for the observed changes during degradation.

Next, we look at the effect of hole mobility modifications in the NPB (Figure 7 right). Indeed a decrease of mobility could cause the same changes in the TEL rise, CELIV, and impedance data as seen in the experiment. Therefore, this parameter could be responsible for the charge transport modifications during stressing. The same analysis has been done for all parameters that could potentially explain the charge transport changes during degradation.

Overall, we can explain the observed changes during degradation by either an increased hole injection barrier or a decreased NPB hole mobility. These are only 2 parameters out of 10 that we still have to consider, thanks to the power of device simulations.

Although we clearly see signatures of charge modifications changes during degradation, this effect cannot explain a current efficiency reduction. The reason for this is that both current and luminance are proportional to mobility and therefore their ratio would not be modified. However, we see a current efficiency decrease in the experiment. So we need a further explanation for the OLED degradation.

Another potential explanation is the formation of non-radiative recombination centers inside the EML. We know that such trap states are predominantly formed at locations of high exciton concentrations.

In the Setfos simulation we have therefore used the exciton profile as an electron trap density profile in the EML and varied their concentration. Indeed, when we compare the current efficiency trends upon degradation (Figure 8, on the left) with the trend in simulation upon increasing trap density (Figure 8, on the right) we see a good qualitative agreement. Note that the usage of the trap profile is really essential. When a constant trap density is used, such a decrease in current efficiency could not be reproduced in the simulation.

Figure 9. Measured and simulated current efficiency decrease during OLED degradation. The use of the trap profile instead of a constant trap density was crucial for the Setfos simulation.

We conclude that electron traps are formed in the EML at locations of high exciton densities and that those are responsible for the decrease in current efficiency.

Here, the Paios system was used for stressing and advanced characterization. Paios has the limitation that it can only stress one device at a time. For more advanced systematic studies in which different current densities and/or temperatures are used, one would need to use our new stress test system Litos. This is what we are carrying out at the moment as well. The results of those studies will be shared with you in upcoming blog posts and conference contributions.

If we have sparked your interest get in contact with us or directly ask for a free of charge trial version of Setfos.