We investigated the influence of vacuum chamber impurities on the lifetime of highly efficient TADF-based OLEDs. Batch-to-batch lifetime variations are clearly correlated with the results of contact angle measurements, which reflect the amount of impurities present in the chamber. Introduction of ozone gas can clean the impurities out of the vacuum chamber, reducing the contact angle to less than 10°. In the vacuum chamber of a new deposition system designed using resin-free vacuum components, various plasticizers and additive agents were initially detected by WTD-GC-MS analysis, but these impurities vanished after ozone gas cleaning. Devices fabricated in the new chamber exhibited lifetimes that are approximately twice those of OLEDs fabricated in a pre-existing chamber. These results suggest that impurities, particularly from plasticizers, in the vacuum chamber greatly influence the OLED lifetime.
Organic light-emitting diodes (OLEDs) have attractive features such as self-luminescence, wide viewing angles, high contrast ratios, and thin profiles, and they are expected to be the key to realizing flexible displays on plastic substrates, which are difficult with LCD and inorganic LED technologies. As such, in recent years, the development of next-generation 8K-HDR TVs, big-screen TVs, smart phone displays, and flexible lighting based on OLED technologies has been rapid.1–4 For practical OLED applications, utilization of highly efficient organic materials is indispensable, and state-of-the-art emitter systems, which include those based on phosphorescent materials,5 thermally activated delayed fluorescence (TADF) materials,6,7 or energy transfer from TADF materials to fluorescent materials,8 i.e., hyperfluorescence, already enable a 100 % internal quantum efficiency. To improve the external quantum efficiency, technologies to enhance light extraction efficiency are also being intensively studied, which will lead to lower power consumption and longer lifetimes.9–12 Another core performance aspect is OLED lifetime, and the study of degradation mechanisms has recently shed new light on ways toward further improvement.13–22 Device fabrication methods and process control, which are indispensable for the commercialization and mass production of OLEDs,23–26 are directly tied to all of these issues and have greatly advanced as of late, but ways to improve processing based on a deeper basic understanding of the origin of performance variations are still needed. For example, we recently reported that even extremely small amounts of impurities in the organic source materials and in the vacuum chamber can greatly affect the OLED lifetime,27,28 but ways to manage chamber conditions to minimize OLED degradation are still not well defined.
Experts are well aware that chamber contamination affects device characteristics. In our previous report, we identified many impurities, such as previously deposited OLED materials and plasticizers from vacuum components,27 present in the vacuum chamber even when the evaporation sources are not heated and the deposition chamber is under vacuum. Thus, cleaning the inner surfaces of the vacuum chamber and/or exchanging the inner deposition shields on a regular basis is a common practice, but better cleaning procedures are critical to further improving lifetime and reproducibility. Furthermore, removal of persistent sources of impurities from vacuum components should also be considered during chamber design.
In this study, we investigate ozone gas cleaning as an improved method to remove these impurities and clarify the relationship between chamber cleanliness and the operating lifetime of highly efficient TADF-based OLEDs. A similar process, UV-O3 treatment, is widely employed for cleaning ITO substrates. The treatment decomposes organic substances on the ITO surface, thereby improving device lifetime.29–31 Thus, we propose that ozone gas exposure can also be a powerful method for the cleaning of vacuum chambers. To verify this, we analyzed chamber impurities by liquid chromatography-mass spectrometry (LC-MS) and wafer thermal desorption gas chromatography mass spectrometry (WTD-GC-MS) for two cases: immediately after assembling a new deposition chamber and after intensive cleaning of said chamber. We then fabricated OLEDs with the new deposition chamber, which was designed to minimize plasticizers from vacuum components, and compared the device characteristics and lifetime with those from a pre-existing deposition chamber.
Devices were fabricated on ITO-coated glass substrates (Atsugi Micro Co., Ltd.) with pre-patterned polyimide banks to define active areas of 0.04 cm2. The materials used in these experiments were 1,4,5,8,9,11-hexaazatriphenylenehexacarbonitrile (HAT-CN) as the hole-injection layer (HIL), 9,9’,9”-triphenyl-9H,9’H,9”H-3,3’:6’,3”-tercarbazole (Tris-PCz) as the hole-transport layer, 3,3-di(9H-carbazol-9-yl)biphenyl (mCBP) doped with (4s,6s)-2,4,5,6-tetra(9H-carbazol-9-yl)isophthalonitrile (4CzIPN) as the emission layer, 2,4,6-tris(biphenyl-3-yl)-1,3,5-triazine (T2T) as the hole-blocking layer, 2,7-bis(2,2’-bipyridine-5-yl)triphenylene (Bpy-TP2) as the electron-transport layer, LiF as the electron-injection layer, and Al as the cathode. The device structure was ITO/ HAT-CN (10 nm)/ Tris-PCz (30 nm)/ 15% 4CzIPN: mCBP (30 nm)/ T2T (10 nm)/ Bpy-TP2 (40 nm)/ LiF (0.8 nm)/ Al (100 nm). The current density-voltage-luminance (J-V-L) characteristics and the external quantum efficiencies (ηext) of the OLEDs were measured with a light distribution measurement system (C9920-11, Hamamatsu Photonics K. K.) and an automatic JVL measurement system (ETS-170, System Giken Co., Ltd.). Device lifetimes were measured with a lifetime measurement system (System Giken Co., Ltd.) at a constant temperature of 30 °C and a constant current density for an initial luminance of 1,000 cd/m2.
The procedure to evaluate the chamber cleanliness was as follows. ITO substrates were cleaned and treated with a UV-O3 system (UV253, Filgen Inc.) before being loaded into the chamber. The substrates were then stored in the chamber evacuated to high vacuum for 30 minutes while the evaporation sources were not heated, i.e., at room temperature. Immediately after unloading the ITO substrates from the chamber, contact angles of water on the ITO surfaces were measured using a DropMaster DMe-201 contact angle meter (Kyowa Interface Science Co., Ltd.). For reference, the contact angles of ozone-treated substrates prior to loading into the chamber were typically less than 5°. Film thicknesses were measured using an n&k analyzer (model: 1280, n&k technology Inc.).
We also analyzed impurities unintentionally present in the chamber. Clean Si wafers were loaded and stored in the vacuum chamber for typically 15 hours. Impurities that adsorbed onto the wafers were then identified using chromatographic methods. LC-MS analyses were performed using a liquid chromatography system (LC-30AD, Shimadzu) equipped with a UV/Visible detector (SPD-20A, Shimadzu) and a mass detector (Exactivetm, Thermo Scientific). For quantitative analysis of an impurity from LC-MS data, a standard curve usually must be obtained.32 Since obtaining standard curves for all impurities is impractical, we focused on the ion counts, which indicate the relative change in the number of each impurity compound. WTD-GC-MS analyses were performed using a silicon wafer analyzer (SWA-256, GL Sciences) in combination with a gas chromatography system (7890A GC, Agilent Technologies) and a mass spectrometer (5975C inert XL MSD, Agilent Technologies). The WTD-GC-MS quantifies the masses of the adsorbed impurities.
Ozone gas was generated with a UV-O3 system (UV144, Filgen, Inc.) and a UV lamp (UVR-40T, SEN LIGHTS Co., Ltd.). The ozone gas concentrations were measured using a Kitagawa gas detector system.
RESULTS AND DISCUSSION
We discussed in our previous report that OLED lifetime decreased with an increase of device fabrication time, which is the length of time used for the organic and metal depositions.27 The results suggested that the lifetime is influenced by the cleanliness of the vacuum chamber. To investigate the correlation between chamber cleanliness and device lifetime, we performed contact angle measurements (see the experimental section for measurement details) to estimate chamber cleanliness each time we fabricated devices.
The contact angle of a water droplet on an ITO substrate can be used to probe chamber cleanliness because the surface of a clean ITO substrate treated with UV-O3 is hydrophilic, which leads to a low contact angle less than 5°. Most organic materials are known to be hydrophobic (e.g., the contact angle of water on 4,4'-bis (N-carbazolyl) biphenyl (CBP) is 96°),16 so the contact angle increases when the ITO surface is contaminated with organic impurities from the vacuum chamber. The contact angle of a liquid on a solid surface is expressed using Young’s equation:33
where γS is the surface tension of the solid, γL is the surface tension of the liquid, θ is the contact angle, and γLS is the solid/liquid interfacial tension. Generally, substances with a low surface tension induce small intermolecular forces and are volatile in vacuum. Contamination of the ITO surface with these substances will lead to a lower surface tension and, thus, an increased contact angle according to Eq. (1).
Figure 1 shows a clear correlation between the device lifetime and the contact angle. For example, the LT95 (the time for the brightness of a device operated under a constant current with an initial luminance of 1,000 cd/m2 to drop by 5%) were 332 hours and 75 hours, different by a factor of four, for devices fabricated in a chamber with contact angles before device fabrication of 14.2° and 33.7°, respectively. The change in chamber conditions during the fabrication of the series of devices was not intentional but was rather the result of various organic materials being deposited as a part of daily experiments, leading to variations in the impurities in the chamber and the contact angles. Note that all of the devices were prepared with exactly the same process and with almost identical fabrication times (100–120 minutes). The clear correlation in Fig. 1 implies that the chamber condition, i.e., the amount of impurities in the chamber, is the dominant factor deteriorating the device lifetime.
From our experience, when plasticizers from vacuum components exist in a chamber, a low contact angle below 10° is difficult to achieve only by ordinary chamber cleaning methods (for example, replacing deposition shields with clean shields and wiping chamber surfaces with acetone to remove deposited organic materials27). We therefore employed an ozone gas cleaning method to get rid of the impurities from the vacuum chamber. First, we confirmed the effect of ozone gas cleaning using contact angle measurements. For this purpose, a small test chamber (approximately ⌀100 mm, 150 mm tall) was set-up and pre-cleaned with approximately 20 ppm of ozone gas. A cycle of (1) ozone gas introduction into the empty chamber filled with air, (2) subsequent evacuation for 1 hour, and (3) contact angle measurement was repeatedly performed. Figure 2 shows the change in the contact angle as a function of the accumulated duration of ozone gas exposure. At the beginning, i.e., immediately after assembling the chamber, the contact angle was 35°, and it increased to 64° after the first cycle of ozone cleaning. The increase of the contact angle can be attributed to decomposition of impurities by ozone gas leading to impurity fragments, thereby increasing the total number of impurities. However, after 10 hours of ozone gas exposure, the contact angle decreased to less than 10°. Although the contact angle also slightly decreased just by keeping the chamber evacuated, it is a relatively slow process, with contact angle decreasing from 28.4° to 26.5° after continuous evacuation for 11 days. Thus, the large decrease of the contact angle with increasing ozone gas exposure time can be attributed primarily to the ozone gas aiding in the removal of impurities.
The need for nearly 10 hours of ozone exposure before the contact angle saturated implies that the ozone etching, i.e., decomposition of organic compounds by ozone gas, is a relatively slow process, presumably taking place within only a few monolayers each cycle. We examined the ability of ozone to etch an organic compound by measuring the thickness of a N,N′-di(1-naphthyl)-N,N′-diphenyl-(1,1′-biphenyl)-4,4′-diamine (NPD) film placed in a chamber with approximately 80 ppm ozone gas exposure. The thickness of the NPD film decreased at a rate of 1.6 nm per 1 hour of ozone exposure (Supplementary figure 1), and a simple extrapolation suggests that it would have taken approximately 125 hours to remove the entire NPD film (200-nm thick), making the procedure impractical for removing thick layers. Therefore, ozone gas cleaning should be considered as a finishing method to clean out very small amounts of impurities rather than an early step for removing thick organic layers, which should be removed from the chamber walls by conventional methods such as by wiping the surfaces inside the chamber with acetone and by replacing the deposition shields with clean ones.
Next, to quantify the effect of ozone gas cleaning on vacuum chamber impurities, we performed LC-MS and WTD-GC-MS. In the experiments, Si wafers were stored in a new deposition vacuum chamber for 15 hours under vacuum to probe impurities. The vacuum components of the new chamber system were designed to be resin-free. The LC-MS and WTD-GC-MS analyses of impurities adsorbed on the Si wafers were performed twice, immediately after assembling the chamber and after intensive cleaning by wiping and subsequent ozone gas cleaning, which resulted in contact angles less than 10°.
For LC-MS, the total ion count before cleaning was 9.5 × 107, which decreased drastically to below the reference background after cleaning. For WTD-GC-MS, the total adsorbed area density decreased from 60 ng/cm2 before cleaning to 3.3 ng/cm2 after cleaning. This reveals that the amount of impurities was significantly reduced by the ozone gas cleaning. Figure 3 displays a histogram of the mass to charge ratios (m/z) of the 215 materials detected by LC-MS before ozone gas cleaning. Even though the evaporation sources were not heated at all, impurities with relatively high molecular masses were detected. However, the amounts of such impurities were too small to be discerned after ozone gas cleaning.
Based on exact molecular masses and compositions calculated by WTD-GC-MS analysis of impurities adsorbed on the order of several ng/cm2, we can tentatively identify some of the impurities as cyclononasiloxane, eicosamethyl-cyclodecasiloxane, olenitrile, octadecanenitrile, hexadecanamide, 9-octadecenamide, octadecanamide, bis(2-ethylhexyl) adipate (DOA), bis(2-ethylhexyl) phthalate (DOP), and dibutylphthalate (DBP) (see Supplementary figure 2 for the chemical structures). Siloxane is known to be used as a silicone grease and a caulking compound for buildings. Aliphatic nitriles might be introduced in the chamber through nitrile rubber gloves used when assembling the chamber. Aliphatic amides are often used as an additive agent in lubrication oils used during machining. DOA, DOP, and DBP are well-known plasticizers that can be introduced during chamber cleaning if they leached out of the bottles containing the organic solvents used to wipe the chamber. While a variety of materials were thus initially detected in the new chamber, these impurities disappeared after ozone gas cleaning, and only a few materials, which are thought to be fragments of decomposed impurities, could be identified (Supplementary figure 3). These results also affirm the effectiveness and usefulness of ozone gas cleaning of vacuum chambers.
Finally, we fabricated OLEDs with the new deposition vacuum chamber after being in service for one year and compared the performance with devices from a pre-existing deposition chamber. Both chambers were cleaned with the same procedures on a regular basis with only the new chamber undergoing thorough ozone gas cleaning once after assembly. First, we compared the level of impurities present in the two chambers. Si wafers were stored in the two chambers for 15 hours under vacuum, and the samples were analyzed by WTD-GC-MS. The adsorbed area densities of DOA and DOP in the pre-existing chamber were detected to be 1.9 ng/cm2 and 0.15 ng/cm2, respectively. However, DOA and DOP could never be discerned in the new chamber, even after being in use for one year, because of the use of resin-free vacuum components. These plasticizers, which we also detected in our previous work,27 are likely to be continuously released to some extent from the vacuum components of the pre-existing chamber regardless of the extent of cleaning.
The initial characteristics (J-V-ηext characteristics and emission spectra) of the devices fabricated with the new chamber are identical to those from the pre-existing chamber (Supplementary figure 4). Prior to device fabrication, the contact angles of the new deposition chamber and the pre-existing deposition chamber were 8.1° and 14.2°, respectively. As seen in Fig. 4, LT95 = 629 hours from the new chamber was superior to that (332 hours) from the pre-existing chamber. We cannot actually ascribe the improved lifetime only to the level of impurities since other process conditions, such as the device fabrication time (typically 150 and 110 minutes for the new and the pre-existing chambers, respectively), the size of substrates, and the geometry of evaporation sources, were not exactly the same for the two cases. However, based on the fact that some plasticizers were detected from the pre-existing chamber while none could be discerned in the cleaned new chamber, we can reasonably attribute the nearly doubling of lifetime with the new chamber to the cleanliness of the chamber. These results demonstrate that the use of resin-free components coupled with initial chamber cleaning by ozone gas can significantly improve device lifetime.
We found a clear correlation between OLED lifetime and contact angle, which is one indicator of the chamber cleanliness. The effect of ozone gas cleaning was confirmed by experiments with a test chamber and could reduce contact angle to below 10°. Ozone gas cleaning is inadequate for removing thick organic layers deposited in a chamber but is a powerful method when used as a final step for cleaning newly assembled chambers to remove small amounts of remaining impurities. However, we sometimes observed in other experiments that, after ozone gas cleaning, the contact angle fluctuated because of the presence of a persistent source of impurities, such as vacuum grease on a transfer rod, in the chamber (Supplementary figure 5). Thus, such sources must be removed for the ozone gas cleaning to be effective.
We performed detailed chromatographic analyses of the impurities in a resin-free new deposition vacuum chamber system. While various plasticizers and additive agents were initially detected in the new chamber, these impurities disappeared after ozone gas cleaning. The lifetime of the devices fabricated with the new chamber was approximately twice that of the devices from a pre-existing chamber, revealing that employing resin-free vacuum components and initially cleaning the chamber with ozone gas can significantly improve device lifetime. We demonstrated that impurities present in a vacuum chamber have a great impact on OLED lifetime and, therefore, the vacuum chamber for OLEDs should be designed as to avoid releasing plasticizers.
See supplementary material for ozone etching rate data, chemical structures of identified impurities, initial device characteristics, and fluctuation in contact angle with a vacuum-grease-contaminated transfer rod.
This work was supported financially by the Program for Building Regional Innovation Ecosystems of the Japanese Ministry of Education, Culture, Sports, Science and Technology (MEXT) and also supported partly by the International Institute for Carbon Neutral Energy Research (WPI-I2CNER) program, sponsored by MEXT, and the Japan Science and Technology Agency (JST), ERATO, Adachi Molecular Exciton Engineering Project, under JST ERATO Grant Number JPMJER1305. We acknowledge the help from Dr. William J. Potscavage Jr. of Kyushu University for preparation of this paper. We also thank Mr. Takashi Suekane of Sumika Chemical Analysis Service Ltd. for performing the LC-MS and the WTD-GC-MS analyses.