Probing polaron-induced exciton quenching in TADF based organic light-emitting diodes

Polaron-induced exciton quenching in thermally activated delayed fluorescence (TADF)-based organic light-emitting diodes (OLEDs) can lead to external quantum efficiency (EQE) roll-off and device degradation. In this study, singlet-polaron annihilation (SPA) and triplet-polaron annihilation (TPA) were investigated under steady-state conditions and their relative contributions to EQE roll-off were quantified, using experimentally obtained parameters. It is observed that both TPA and SPA can lead to efficiency roll-off in 2,4,5,6-tetra(9H-carbazol-9-yl)isophthalonitrile (4CzIPN) doped OLEDs. Charge imbalance and singlet-triplet annihilation (STA) were found to be the main contributing factors, whereas the device degradation process is mainly dominated by TPA. It is also shown that the impact of electric field-induced exciton dissociation is negligible under the DC operation regime (electric field < 0.5 MV cm−1). Through theoretical simulation, it is demonstrated that improvement to the charge recombination rate may reduce the effect of polaron-induced quenching, and thus significantly decrease the EQE roll-off.


Reviewer #1 (Remarks to the Author):
This work describes the mechanism of roll-off and device degradation related to polaroninduced exciton quenching, namely singlet-polaron annihilation (SPA) and tripletpolaron annihilation (TPA) in TADF-based OLEDs. The work has been done meticulously and, for the most part, described clearly, with detailed device data and solid photophysical data support. However, there are some points listed below that need further clarification; therefore, I recommend publication of the manuscript after minor revision.

Response:
We thank the Reviewer for the comments and careful evaluation of our work. We are delighted and grateful for the Reviewer's positive feedback. 1) For Figure 5, the authors claimed that SPA and TPA were the major factors that contribute to the device's efficiency roll-off and degradation. However, from Figure 5a, it seems that STA would be a major factor that contributes to the singlet dynamics when the voltage is larger than 6 V. The authors mentioned "the impact of STA is almost five time higher than SPA…, which emphasized the necessity of reduction of kST" in the original text. Later in the text, the authors seems to overlook the significance of STA, and to assign further contribution to roll-off effects to SPA rather than STA. It would be convincing to the reader if the authors can explain the boost in STA at higher voltage and interpret the role of STA in the singlet dynamics.

Response:
We thank the Reviewer for pointing out the importance of STA in the singlet dynamics and EQE roll-off in TADF OLEDs. Indeed, both SPA and STA directly reduces singlet density. From Fig. 5a (now Fig. 4a in the revised manuscript), it is clear that STA mostly dominates over SPA as the deactivation pathway. However, any comparison between the impact of STA and TPA had not been discussed in the manuscript. Hence, to further highlight the role of the STA process, we have now added a new part in the manuscript.  Fig. 6 as a function of current density and TPA rate constants. From the intrinsic of 4CzIPN (dash-dot horizontal straight line) and '0' contour line, it is clear that the loss due to TPA dominates over STA under current densities lower than ~10 2 mA cm −2 , while for higher current densities (> 10 2 mA cm −2 ) STA starts to dominate over TPA. As rapid EQE roll-off was observed in the steady-state device operation beyond the current density of ~10 2 mA cm −2 , thus it is appropriate to infer that the contribution of STA in EQE roll-off is higher than that of TPA".

Added line in the Conclusion
"Charge imbalance and STA were found to be the main contributors of the efficiency roll-off in the 4CzIPN-based OLEDs".
2) The formation of polaron would influence the carrier mobility in the system. The authors haven't clarified the role of carrier mobility in the device optimization. 3) There are many fittings in this work, both the main manuscript and supplementary information, but none of the fitting parameters has been given to support the goodnesses of fit.
The authors should give all the fitting parameters to convince the reader about the correctness of all fittings.  Figure below shows the current density−voltage response from the HOD and EOD devices along with their energy diagrams. It can be assumed from current density−voltage response that doped EML is more favourable for hole injection/transport rather than electron injection/transport. Furthermore, the electron transport can be slightly contributed by 4CzIPN due to its deeper LUMO (more negative) than mCP. For the above-mentioned reasons, we believe charge transport in the OLEDs was mainly dominated by holes. 5) The authors claim that the PL intensity drop with increase in voltage can be attributed to the SPA, TPA, and field-induced quenching processes. But in Supplementary Fig. 3, there is an apparent negligible PL intensity drop observed with an increase in applied voltages, which might suggest that the effect of SPA, TPA, and field-induced quenching processes were less serious than what the main manuscript discussion emphasizes.

Response:
We thank the Reviewer for raising a valid point. The exciton generation in the polaron-induced quenching experiment was carried out with a low power optical of ~ 50μW excited at 325 nm, which is approximately equivalent to ~ 8.2×10 13 photons per second. By considering photon absorption from different interlayers and the absorption of mCP at 325 nm, the actual number of generated excitons in the active layer might be even lower than the calculated photon numbers. Since the amount of loss due to annihilation process is exciton density dependent, in our experiment the PL intensity drop due to SPA and TPA processes was expected to be low considering the low photon density. In addition, the HOD devices were operated under low current densities to avoid any device breakdown, which provided low polaron density as well as low PL quenching. Though we believe the PL quenching was still

Figure | PL comparison a Normalized voltage dependent PL spectra collected from HOD. b
Normalized PL spectra from HOD without applied bias voltage and PL spectra from quartz substrate. Table 1, ∅_P has been denoted as ∅_D --although ∅_P and ∅_D were 41.94 % and 42.05%, respectively. However, these values are presented as 41% and 43%, respectively, instead in the main paper.

Response:
We thank the Reviewer for pointing out the mistake. The correction has now been made in the revised manuscript and Supplementary Information. 8) Equations inserted in the main paper should be numbered in order and referred to in the text.

Response:
We thank the Reviewer for pointing out the omission. We have now added the equation numbers properly in the revised manuscript and Supplementary Information.

Reviewer #2 (Remarks to the Author):
The authors studied the effect of polarons-induced exciton quenching to understand the efficiency roll-off in TADF based OLEDs. To investigate SPA and TPA, drift-diffusion numerical modeling, exciton dissociation theory, kinetic exciton dynamics were introduced and successfully allowed understanding of bimolecular exciton kinetics. The manuscript was well organized, and the results of experiment and modeling were convincing.

Response:
We are delighted and very grateful for the Reviewer's positive feedback and valuable comments.
However, some issues should be addressed before the publication.     6. The moleculer structure of BP4mPy was not included. Also, inclusion of device structure with frontier orbital energetics would be good to better understanding of the readers.  Supplementary Fig 2b, Fig 7a, Fig 10b and Fig 11a as

Reviewer #3 (Remarks to the Author):
Presented by Hasan et al., this paper managed to discuss polaron-induced exciton quenching in TADF-based organic light-emitting diodes (OLEDs). By utilizing steady-state photoluminescence (PL) and electroluminescence (EL) measurements, the authors attempted to quantify the quenching rates of singlet-polaron annihilation and triplet polaron annihilation then to analyze their influence on the OLEDs performance. As understanding quenching mechanisms in OLEDs plays crucial role in improving device efficiencies, this paper is of importance to the materials science community, thus qualified for being published in Nat.
Commun. Nevertheless, several issues should be addressed properly before further consideration for being published. Comments and suggestions are listed below: Response: We thank the Reviewer for careful evaluation of our work and positive comments.
We have now made the relevant changes to the main text and supporting information based upon the concerns raised. 2. The authors should comment on why a dual excitation setup was specifically used for this study.
Response: Exciton-exciton interaction (SSA, STA, and TTA) and exciton-polaron interaction (SPA and TPA) are spin-allowed processes in TADF and thus can result in the reduction of efficiency in OLEDs. It is quite challenging to separate these processes under steady-state OLED operation, specifically in TADF as they can all simultaneously occur in high current densities. Without separating annihilation processes appropriately, any measurements of annihilation rates can lead to overestimation of the value. The benefit of using a dual excitation setup for the investigation of polaron-induced quenching is that it can be used to separate exciton-polaron quenching from exciton-exciton interactions. Since this setup can constantly produce same number of excitons via optical pumping, while any change in the PL intensity collected from the device can be assigned due to exciton-polaron interactions. In addition, under reverse bias in OLEDs, it can be utilised to block any charge injection in order to collect change in PL intensity due to different applied electric fields. In this way, more accurate estimation of exciton-polaron quenching rate and field-induced dissociation rate can be obtained by using dual excitation setup.
3. In Supplementary Fig.4b, the prompt and delay lifetimes are obtained through tail fit instead of deconvolution fit. Since the decay profile (purple dots) is already a convolution between excitation source and population decay, it is recommended to perform deconvolution fit with instrumental response function to obtain an accurate prompt decay time, ФP and ФD.

Response:
We would like to thank the Reviewer for pointing out this important issue. TCSPC measurements to extract the lifetimes was performed with Jobin-Yvon Fluorolog-3, by exciting the samples at an excitation wavelength of 372 nm, generated by a pulsed nano-LED and an instrument response of about 1 ns. Figure below shows the instrument response function. In our work, IRF << prompt and delayed lifetime of 4CzIPN. As a consequence, we believe that the extracted lifetime will not significantly differ from the deconvolution fit. Figure   5. I recommend the authors to move Fig. 1 to the supplementary information, as it appears that the purpose of simulating current density-voltage response is only to obtain the hole mobility and characteristic field value, both of which are not further evaluated.

Response:
We appreciate the Reviewer's suggestion. In the revised manuscript All the calculations in the manuscript are now updated using this set of equations.
We also would like to point out that the term ′ mentioned in Phys. Rev. B 66, 035321 (2002) is represented as in our work. In Phys. Rev. B 66, 035321 (2002), they have taken as coefficient × ′ . The updated coefficient used in this work is very close to the value of the coefficient used in that paper.
9. The fitted kSP and kTP from PL (6.1 × 10-12 cm3s-1 and 5.8 × 10-13 cm3s-1, respectively) and EL measurements (2 × 10-11 cm3s-1 and 9.1 × 10-13 cm3s-1, respectively) for mCP:4CzIPN (also for ACRXTN OLED) is not reasonable to me. As the lifetime of singlet excitons is known to be much faster than that of triplet excitons, even disregarding the population gap between singlet and triplet excitons, the chance for polarons to annihilate singlet excitons should be much lower than to annihilate triplet excitons. How do the authors explain the results that kSP > kTP?
Response: The Reviewer has made a very important point here. Interestingly, the magnitude of and is independent of singlet and triplet lifetime, respectively. Though the amount of quenching caused by an individual annihilation process is greatly dependent on excited state lifetimes. For example, in case of TPA, the amount of quenching/loss due to TPA is proportional to , where lifetime of triplet influences the loss mechanism via triplet density ( ). High triplet lifetime can give rise to triplet density (triplet accumulation) over time, which can increase the amount of quenching with a low . However, the value of and depends on triplet and singlet diffusivity. The SPA and TPA predominantly occur via long-range Förster and short-range Dexter transfer processes, respectively. Triplet-polaron quenching requires a collision between the species, and hence the rate is much lower than singlet-polaron quenching. In general, Förster-transfer based annihilation process such as SSA, STA and SPA tends to have high-rate constant values compared to Dexter-transfer based annihilation constants such as TPA and TTA.
10. For the fitting processes to acquire kSP and kTP, the authors state that the STA and TTA rate constants were taken from the reference paper (Appl. Phys. Lett. 113, 083301 (2018)). Yet, this reference paper reported a series of simulations with different sets of STA and TTA values.
Which set of values did the authors adopted in this manuscript? Please specify all the fixed values in the fit process at least in the supplementary. information.

Response:
We thank the Reviewer for raising a very valid point here. In the manuscript we have now specified the values of STA and TTA, taken from Appl. Phys. Lett. 113, 083301 (2018).
Modified line: The STA and TTA rate constants were taken from the literature as 1 × 10 −11 cm 3 s −1 and 5 × 10 −18 cm 3 s −1 , respectively 18 and used to fit the EQE curve.
Minor points: