Substrates with four devices were fabricated as described in Methods. To follow and quantify the degradation, the electroluminescence (EL) as well as device characteristics were measured, and two devices were driven at a constant current with a luminance of 1000 cd m−2 until their 95% (T95) and 70% electroluminescence levels (T70) were reached. One device, PL55, was degraded under irradiation of 400-nm laser until a 55% photoluminescence level (PL) was reached for comparison. The fourth device, T100, was left pristine. The degradation is summarised in Supplementary Table 1. Performance characteristics of the T100, T95, and T70 devices including external quantum efficiency (EQE), current efficiency, EL spectrum, and current density-voltage plot are provided in Supplementary Figure 1. The pristine device showed a typical external quantum efficiency of 17.4% at 1000 cd m−2 with a maximum EQE of 19.1%, indicating that the blue OLED device with an Ir complex emitter utilized most of injected electrons and holes for phosphorescent emission and its internal quantum efficiency (IQE) was close to unity. In spite of the decent EQE of the blue OLED, its operational lifetime was limited to 1.5 - 2 hours for T70.
To understand the physical and chemical origin of the blue OLED device degradation, we investigated the device characteristics and chemical changes. The max. EQE of the blue OLED decreased gradually from 19.1–12.8% and the driving voltage at a constant luminescence condition increased from 5.83 V to 6.10V upon the device degradation from T100 to T70 (See Supplementary Table 1 and Supplementary Figure 1). The EL intensity showed a gradual decrease with device aging, but the normalized EL spectra were identical to each other regardless of the degradation. Although these data provided typical device characteristics and performance evaluation, the underlying origin of the short operational lifetime for the blue device is still elusive, requiring further physical and chemical analysis. High-resolution mass spectra were acquired from a 200 µm × 200 µm area successively deeper into the device using a 5 keV Ar2500+ gas cluster ion beam for sputtering with analysis using the Orbitrap HF mass spectrometer (Figure 1a). Each spectrum consumes a remarkably small amount of material, approximately 0.17 nm thickness of material (equivalent to 6.8 µm3). A reference device containing a 1 nm-thick blue Ir dopant layer in a host matrix with a total thickness of 50 nm was used (Figure 1b) to evaluate the depth resolution under these conditions (FWHM ~ 7 nm). Depth profiles were obtained from all four devices. A positive ion depth profile (Figure 1c) from T100 pristine device shows that the chemistry of the different layers can be resolved in agreement with the schematic of the device multilayer architecture shown in Figure 1c. The low fragmentation from the argon cluster sputtering and high-resolution Orbitrap MS results in high-resolution LC-MS quality spectra with ~ 7 nm depth resolution. The organic materials in each layer give rise to molecular or pseudo-molecular ion peaks that are readily identified based on their accurate mass with sub-ppm mass resolution (Supplementary Figures 2-3 and Supplementary Tables 2-3) and example spectra with peaks from all regions in the device are shown (Figure 1d).
To check the technical repeatability, three repeat depth profiles were recorded from the same device (Supplementary Figure 4). The profiles are essentially indistinguishable from each other demonstrating good measurement repeatability. On first inspection, the depth profile from a degraded device has only minor differences to the pristine device (T100), however, the spectra from the emissive layers differ in detail. To compare the spectra, each spectrum is first normalised to its total intensity to account for any variation related to the measurement, e.g. primary beam current variation. Five peaks at m/z 369.1038, m/z 577.1679, m/z 943.2475, m/z 1019.2789 and m/z 1127.2924 have significantly higher normalised intensities in the spectra from T70 and PL55 than in the spectrum from the pristine device. Using the accurate mass measurement and the known initial composition of the layers where the ions are detected, putative mass and structure assignments are given (Table 1, Figure 2a-b) They are all potential reaction products of DBFPO (derivatives of DBFPO:Liq) which is in the hole blocking and electron transport layers (HBL and ETL, respectively). DBFPO consisting of diplenylphosphine oxide and dibenzofuran groups served often as ETL and ambipolar host.18,23 The normalised intensities for these peaks are, on average, between 14 and 120 times higher in the spectrum from PL55, and between 5 and 23 times higher in the spectrum from T70 than the pristine device (Figure 2c). The effect of the degradation of T95 are not well pronounced. However, it can reasonably be postulated that changes observed at the EL 70% degradation level predominantly result from mechanisms that are also in play during the initial degradation. Chemical analysis of OLED degradation has in other studies required further device aging, even to the T10 level of degradation.13,19,26
Depth profiles for each device of the five degradation products (Figure 2b) reveal a localisation to the ETL/EML (emissive layer) interface, suggesting that the ions may be degradation products from ETL, HBL, and EML materials. There are common features as well as some differences in the depth profiles. Regardless of the aging method - current-driven or photo-irradiated, we found that the dominant degradation products were identical. Subtle differences were observed in the depth profiles. For example, the m/z 369.1038 is located deeper in the EML than the other degradation products. In PL55, the m/z 577.1679 is observed with much higher intensity and before the EML is reached, compared with the T70 device where it is distinctly in the EML. This suggests that the photo-irradiation of PL55 has caused the formation of the degradation product in the electron injection and electron transport layers. This degradation is less intense and is not formed in these layers in electrically-driven devices and it is pronounced when the device was irradiated by a 400-nm laser (bandwidth ~ 20 nm). Considering the wavelength of the laser and energy levels of layer materials13, HTL (NPB, 3.1 eV), blue Ir dopant (2.9 eV), and ETL (DBFPO:Liq, 3.4 eV) may absorb the intense light, which leads to material degradation through high energy exciton generation via exciton-exciton annihilation or photo-thermal effect.27
We also performed unsupervised multivariate analysis28 of a combined dataset with depth profiles for each device to identify thermal degradation products in the NPB layer of the PL55 device, which are likely induced by the laser irradiation29 (Supplementary note 1 and Supplementary Figures 5-6).
For comparison with previous studies, the devices were also analysed by traditional LC-MS methods for bulk analysis (Supplementary Figure 7). However, the subtle differences discovered in the OrbiSIMS depth resolved data were not observed, in particular, the reaction products from aged devices.
Table 1
Putative assignments for potential degradation products in the blue OLED device.
m/z
|
Assignment
|
Mass deviation (ppm)
|
369.1038
|
C24H18PO2+
|
[M – R]+, R=C12H8PO
|
0.1
|
577.1679
|
C36H26P2O3LiH2+
|
[DBFPO + LiH2]+
|
1.7
|
943.2475
|
C60H43LiO5P3+
|
LiQ + DBFPO
|
-1.4
|
1019.2789
|
C66H47LiO5P3+
|
LiQ + DBFPO + Benzene
|
-0.6
|
1127.2924
|
C72H52P4O5Li+
|
[(DBFPO)2 – O + Li]+
|
0.9
|
In order to reveal the origin of dominant degradation at the HBL/EML interface, we investigated the exciton distribution in the emission zone using the sensing layer method developed by Forrest et al.30 For this purpose, we fabricated a series of 5 blue OLED devices (Figure 3a) with a monolayer-thick sensing layer of red-emitting Ir dopants progressing from the EBL2/EML interface to the HBL/EML interface located at 0, 100, 200, 300 and 400 Å depths in the EML. The red emission intensity increased towards the HBL/EML layer (Figure 3a) indicating a large population of excitons close to the HBL. This exciton profile reveals that high energy excitons or polarons from exciton-exciton and exciton-polaron interactions generate most likely close to the HBL/EML interface, where unfavourable chemical reactions with the hole blocking layer DBFPO cause device degradation.11 The distribution of the degradation products also shows that mobile Li+ ions diffused from ETL into the HBL and deeply to the EML where the exciton density was maximized. The direct driving force of Li+ diffusion towards the HBL/EML interface is not clear at the moment, because it is expected that mobile Li+ ions may be attracted to the Al cathode due to the external electric field. The laser-irradiated device (PL55) showed the identical chemical identities for the degradation products to those in the current-aged devices T95 and T70. These results indicated that the interfacial region with higher exciton density or high energy excitons possibly caused the unfavourable chemical reactions and diffusion through direct bond cleavage and photothermal effects.
We also performed photoluminescence (PL) measurements for a quantitative analysis on blue emitting dopant degradation (Figure 3b). Whereas electroluminescence (EL) decreased to 95% and 70% in T95 and T70, respectively, the PL intensity was 98.2% and 92.9% from the corresponding devices. This indicates that the device degradation in the aged devices T95 and T70 does not mainly originate from blue Ir dopants, which contributes only a 7% to the total luminescence loss of the device T70.
A device with a slightly modified host material (mCBP-CN) presented approximately two orders of magnitude longer operational lifetime to T70 (Figure 4a). We applied our nanoscale depth profiling method to measure the DBPFO-related degradation products and observed that these are present in much less abundance, which indicates that a slight change in the host material may cause a shift of exciton distribution, resulting in less accumulation of degradation products and prolonged device operational lifetime. In fact, Sim et al.18showed that the blue OLED device containing mCBP-CN host with the identical device structure except the EML thickness has a long operational lifetime LT50 of 431 hours at 500 cd/m2 and the excitons within the EML distributed towards the HTL/EML interface. These results provide good evidence that our nanoscale chemical depth profiling method can successfully be used in the design and optimisation of new materials to achieve longer lifetimes.