Degradation of blue-phosphorescent organic light-emitting devices involves exciton-induced generation of polaron pair within emitting layers

Degradation of organic materials is responsible for the short operation lifetimes of organic light-emitting devices, but the mechanism by which such degradation is initiated has yet to be fully established. Here we report a new mechanism for degradation of emitting layers in blue-phosphorescent devices. We investigate binary mixtures of a wide bandgap host and a series of novel Ir(III) complex dopants having N-heterocyclocarbenic ligands. Our mechanistic study reveals the charge-neutral generation of polaron pairs (radical ion pairs) by electron transfer from the dopant to host excitons. Annihilation of the radical ion pair occurs by charge recombination, with such annihilation competing with bond scission. Device lifetime correlates linearly with the rate constant for the annihilation of the radical ion pair. Our findings demonstrate the importance of controlling exciton-induced electron transfer, and provide novel strategies to design materials for long-lifetime blue electrophosphorescence devices. The short lifetime of blue-phosphorescent organic light-emitting devices owing to material degradation impedes their practical potential. Here, Kim et al. study the molecular mechanism of the degradation that involves exciton-mediated electron transfer as a key step for the generation of radical ion pairs.

xtensive research has enabled organic light-emitting devices (OLEDs) to outperform conventional displays in various commercial applications. One widely recognized remaining challenge, however, is to improve device lifetime. In particular, the short operation time of devices that emit blue light are significant impediments to exploiting the full potential of OLEDs.
Despite these advances, there has been a limited understanding of the fundamental reason for the significant instability of devices that emit blue light. Note that an intermolecular pathway for radical generation has been overlooked. It is possible that the exciton is reductively quenched by electron transfer from a nearby molecule with a shallow oxidation potential to form a radical ion pair (i.e., a polaron pair), due to the positive driving force of the electron transfer (Fig. 1). Such radical ion pair would rapidly undergo charge recombination to restore the original neutral states, but its labile nature can facilitate degradation of both the host and dopant materials. This mechanism may explain the greater instability of devices that emit blue light than those that emit green and red light, because it predicts faster formation and slower annihilation of the radical ion pair in blue emission layers (vide infra). Therefore, it is envisioned that studies of intermolecular electron transfer can lead to an ability to control on the radical species responsible for intrinsic degradation of emitting layers.
In the current research, we investigate a charge-neutral, exciton-induced generation of radical ion pairs between a wide bandgap host and blue-phosphorescent Ir(III) complex dopants. The radical ions and their annihilation processes in the host-dopant binary system are directly monitored for the first time, with employing a variety of chemical techniques. The rate constants for charge recombination by back electron transfer (k BeT ) are determined through second-order kinetics analyses. Multilayer OLEDs are fabricated, and a linear correlation is found between k BeT and device lifetime. The study reveals the importance of controlling the electrochemical potentials of a host exciton and its dopant for achieving long device lifetimes.

Results
Synthesis of materials and thermodynamic analyses of electron transfer. A series of homoleptic triscyclometalated Ir(III) complexes having N-heterocyclocarbenic (NHC) ligands (Ir1-Ir4) were newly synthesized and employed as blue-phosphorescent dopants. 3,3′-Biscarbazolyl-5-cyanobiphenyl (H) was used as a wide-bandgap host material 37 . Chemical structures of the Ir dopants and H are depicted in Fig. 2a. Synthetic details and structural characterization data for the compounds are summarized in Methods. Structures of the benzimidazole-based (Ir1 and Ir2) and imidazopyrazine-based (Ir3 and Ir4) NHC ligands were systematically varied to tailor the electrochemical potentials of the complexes. This synthetic control was indeed demonstrated by an observation of a shift of the reversible Ir(IV/III) redox potentials (0.81−1.12 V vs. standard calomel electrode (SCE); Fig. 2b). H displayed irreversible one-electron oxidation and reduction processes at 1.50 V (E ox ) and −1.82 V (E red ) vs. SCE, respectively (Supplementary Fig. 1). An optical bandgap energy (ΔE g ) as large as 3.10 eV was determined, which enabled calculation of the excited-state oxidation (E* ox ) and reduction (E* red ) potentials of H according to the relationships E* ox = E ox − ΔE g and E* red = E red + ΔE g . In this way, E* ox and E* red of H were calculated to be −1.60 and 1.28 V vs. SCE, respectively. Comparison of the E* red value of H with the oxidation potentials of the Ir dopants revealed positive driving forces (−ΔG eT ) in the range 0. 16 (Fig. 3b). Photoinduced electron paramagnetic resonance (EPR) spectroscopy provided direct evidence for the formation of radical species. As shown in Fig. 3c, 1.0 mM H, 1.0 mM Ir3, and a mixture of 1.0 mM H and 1.0 mM Ir3 (Ar-saturated THF) did not display any apparent paramagnetic signals in the dark. Photoirradiation of the binary mixture of H and Ir3 produced an increase in paramagnetic signals at a g value of 2.012, typical of a free radical. This peak was accompanied by rhombic signals presumably due to an Ir(IV) species of Ir •+ . These peaks disappeared upon cessation of photoirradiation, which indicated   that charge recombination occurred. Taken together, the results revealed that a radical ion species formed through rapid electron transfer.
To monitor the generation and annihilation of the radical ion species, transient absorption spectra were acquired for deaerated THF solutions containing 150 μM Ir3 and 3.0 mM H. As shown in Fig. 4a, weak positive absorption signals were observed in the NIR region of the spectra, although the spectra were dominated by the emission of Ir3. The positive signals were not seen for solutions containing only Ir3 or only H, which indicated electrontransfer species (e.g., radical ion pairs) to be responsible for the NIR signals. Comparison of the transient absorption spectra with simulated electronic transition energies of H •− (the magenta curve in Fig. 4b) and Ir3 •+ (the blue curve in Fig. 4b) (TD-CAM-B3LYP/LANL2DZ:6-311G(d,p)//CAM-B3LYP /LANL2DZ:6-311G(d,p)) indicated that the NIR spectra were due to Ir3 •+ . This assignment was further corroborated by the spectroelectrochemical measurement taken for Ir3 (2.0 mM, deaerated THF) during oxidative electrolysis at 1.30 V vs. SCE (the black curve in Fig. 4b). Here, spectral signatures of H •− were not observed, presumably due to the instability of the oneelectron-reduced state, as suggested from the irreversible reduction process.
Charge recombination by back electron transfer from H •− to Ir3 •+ was monitored by recording the decay of the transient absorption signal of Ir3 •+ at a wavelength of 1100 nm. Its k BeT value was determined to be 1.5 × 10 11 M −1 s −1 according to a second-order kinetics analysis (Fig. 4c); this analysis employed a molar absorbance value of Ir3 •+ (ε = 2170 M −1 cm −1 ), which was determined from the spectroelectrochemical measurements. The k BeT values determined for the other Ir dopants were 4.5 × 10 10 M −1 s −1 (for Ir1), 3.0 × 10 10 M −1 s −1 (for Ir2), and 2.8 × 10 11 M −1 s −1 (for Ir4). The k BeT values corresponded well to the theoretical curves obtained from the classical Marcus theory for adiabatic outer-sphere electron transfer, with the reorganization energies in the ranges 1.7-1.9 eV and 2.2-2.3 eV for the benzimidazole-based NHC Ir(III) complexes (Ir1 and Ir2) and the imidazopyrazine-based NHC Ir(III) complexes (Ir3 and Ir4), respectively (Fig. 4d). Note that k BeT was observed to increase as the value of −ΔG BeT became smaller, which revealed that the charge recombination was located in the Marcus-inverted region of electron transfer 38 . Since more rapid charge recombination is beneficial for suppression of irreversible degradation from radical ion species, this result provided two valuable approaches for improving device lifetimes: decreasing −ΔG BeT and increasing the reorganization energy. The former approach may involve shifting E ox of a dopant cathodically by increasing the electron density of the NHC ligand. The latter approach requires further understanding about reorganization processes, but potentially involves the use of charge-delocalizing ligand frameworks.
Degradation from radical ion pairs. It was found during the laser flash photolysis experiments that the binary mixture of H and Ir4 underwent irreversible coloration, which suggested that these materials became degraded. UV-vis absorption changes were observed for a mixture of H and Ir4 during steady-state photolysis upon irradiation using a Xenon lamp (300 W). The UV-vis absorption spectrum of Ir4 showed a characteristic intense band at 390 nm due to an intraligand charge transfer (ILCT) transition localized within the NHC ligand. The absorbance of the ILCT transition band decreased during the photolysis ( Fig. 5a and Supplementary Fig. 3), which indicated that the NHC ligand fragmented. A similar bleaching behavior was observed for vacuum-evaporated films of H containing 15 wt % Ir4 (Fig. 5b). As expected, such an absorption change was absent for redox-innocent poly(methyl methacrylate) films doped with 15 wt % Ir4 (Fig. 5c).
Liquid chromatograms obtained for the photolyzed samples revealed the formation of byproducts (Fig. 5d, e). Electrospray ionization mass spectra of the byproducts displayed peaks corresponding to fragments due to cleavage of the bond connecting the imidazopyrazine and phenyl moieties in the NHC ligands in Ir4 (Fig. 5f). Such fragmentation products were found for Ir4 after its oxidative electrolysis at 1.25 V vs. SCE (Fig. 5g). Analyses of the degradation products of Ir1-Ir3 revealed fragmentation behaviors identical to Ir4 ( Supplementary Fig. 4). The degradation was also observed for vacuum-evaporated films of H doped with 15 wt % Ir during photoirradiation (Supplementary Fig. 5). The results collectively indicated that the Ir dopants degraded upon undergoing one-electron oxidation. Tang and co-workers suggested a radical cation of a bluephosphorescent Ir(III) complex to be unstable 8 . In addition to observing the degradation of dopants, we were able to observe byproducts of the host. These byproducts mainly originated from scission of C-N bonds between biphenyl and carbazole groups of Fig. 4 and 5). Such bond scission was also reported for host materials having similar biscarbazolylbiphenyl frameworks 14 . On the basis of the results, we conclude that unstable radical ion species are generated intermolecularly between a host and a dopant molecules, even without the charge carrier injection, and that these radical ions are responsible for the irreversible degradation.   Liq (lithium quinolinate) served as a hole-injection layer, a holetransporting layer, an electron-blocking layer, an electrontransporting and hole-blocking layer, and a buffer layer, respectively. The emitting layer consisted of a 30-nm-thick film of H that was molecularly doped with 10 or 20 wt % Ir dopant. The electroluminescence spectra (Fig. 6b) were identical to the photoluminescence spectra (Fig. 1c), excluding the contribution of any bimolecular species, such as exciplex, to the electroluminescence emission. An external quantum efficiency (EQE) as high as 17.8% was recorded at 500 cd m −2 for a 20 wt % Ir4 device, and the efficiency decreased in the order Ir3, Ir2, and Ir1 (Fig. 6d) To correlate device lifetime with the electron-transfer behavior, luminance decays of the H:Ir devices were recorded during operation in a constant current driving mode with an initial current value defined at a luminance of 500 cd m −2 . Figure 6e depicts the luminance profiles of the devices, each as a function of operation time. LT 70 , which corresponds to the operation time when the luminance decreases to 70% of its initial value, approached 93.05 h for the 20 wt % Ir4 device, and decreased in the order Ir4, Ir3 (19.70 h), Ir2 (4.11 h), and Ir1 (1.88 h). The dissimilarity between this LT 70 order and the order of the triplet state energy value of the Ir dopants was interpreted as excluding exciton-localized homolysis of the dopant from being a dominant degradation pathway. It is worth noting that the LT 70 value was observed to increase in proportion with k BeT (Fig. 6f). Photoluminescence stability of the films of H:Ir also followed this trend, indicating that the intermolecular electron-transfer reactions, not different current levels in devices, were responsible for device lifetime ( Supplementary Fig. 6). These results provided direct evidence for acceleration of charge recombination being crucial for device longevity. Finally, validity of our mechanism was further examined by comparing the LT 70 values with those of devices having mCBP (3,3′-bis(9-carbazolyl)biphenyl) in place of H as a host material. It is predicted that radical ion pairs are more accumulated in the mCBP:Ir layers than the H:Ir layers, due to the more positive E* red (1.  Table 3).

Discussion
We have investigated exciton-induced generation of radical ion pairs between a wide bandgap energy host and bluephosphorescent Ir(III) complexes having NHC ligands. Spectroscopic techniques were employed to detect charge-neutral production of a pair of the radical cation of the dopant and the radical anion of the host. In particular, photoinduced EPR experiments and nanosecond laser flash photolysis provided direct evidence for the formation of the radical cation of the dopant molecule. Charge recombination kinetics within the radical ion pair was analyzed by employing the second-order kinetics model. Charge recombination was found to occur in the Marcus-inverted region of electron transfer. Analyses of the radical ion-mediated degradation products revealed the occurrence of oxidative cleavage of the dopant and reductive degradation of the host. Finally, a strong linear proportionality was found between the device lifetime and the rate constant for charge recombination. This finding is valuable, because it provides reasonable explanations for the poor stability of OLEDs that emit specifically blue light. Here, large driving forces for electron transfer (−ΔG eT ) facilitate more rapid formation of radical ion pairs, but much larger driving forces for back electron transfer (−ΔG BeT ) impede charge recombination occurring in the Marcus-inverted region of electron transfer. The combined effect is a longer lifetime of radical ion pairs in an emitting layer of a blue-phosphorescent OLED than those of green-phosphorescent and red-phosphorescent OLEDs. It should be noted that accumulated levels of the radical ion species are minimal, because the forward electron transfer competes with energy transfer to a dopant, and, also, rapid charge recombination occurs. Overall, our mechanistic study revealed that exercising the electrochemical control to minimize the density and lifetime of radical ion pairs is crucial for realizing long operation lifetimes of OLEDs.    Steady-state UV-vis absorption measurements. UV-vis absorption spectra were collected on an Agilent, Cary 300 spectrophotometer. Solution samples were prepared in THF to a 10 μM concentration prior to the measurements, unless otherwise stated.
Steady-state photoluminescence measurements. Photoluminescence spectra were obtained using a PTI, Quanta Master 400 scanning spectrofluorimeter at 298 K. The 10 μM solutions or the films were used for the measurements. The Ir solutions were deaerated by bubbling Ar for >15 min. Photoexcitation wavelengths were 341 nm (Ir1), 340 nm (Ir2), 396 nm (Ir3), 400 nm (Ir4), and 340 nm (H). The PLQYs were determined employing an absolute PLQY measurement system (Hamamatsu, C11347-01).  Calculation methods. Geometry optimization was performed using Becke's threeparameter B3LYP exchange-correlation functional modified with the Coulombattenuated method (CAM-B3LYP), the double-ξ quality LANL2DZ basis set for the Ir atom, and the 6−311 + G(d, p) basis set for all the other atoms. A pseudo potential (LANL2DZ) was applied to replace the inner core electrons of the Ir atom, leaving the outer core [(5 s) 2 (5p) 6 ] electrons and the (5d) 6 valence electrons. Frequency calculations were subsequently performed to assess the stability of the convergence. Time-dependent density functional theory (TD-DFT) calculations were carried out for the optimized geometries using the same functional and basis sets. Geometry optimization and single-point calculations were performed using the Gaussian 09 program 39 . GaussSum was employed for simulation of the predicted electronic absorption spectra 40 .  Steady-state photolysis. An Ar-saturated THF solution (3.0 mL) containing 3.0 mM H and 100 μM Ir was photoirradiated under a broad-band light from a Xenon lamp (300 W, Asahi Spectra, Max 303) for 10 min. A color change from yellow to orange was observed during the photolysis, which was monitored through steadystate UV-vis absorption spectroscopy. Photolysis was also performed for vacuumevaporated films of H molecularly dispersed with 10 or 15 wt % Ir (glass substrate). The films were photoilluminated under 325 nm (3.5 mW, He-Cd laser, Kimmon Koha, IK3202R-D) or the broad-band light from the Xenon lamp, during which the changes in the photoluminescence (325 nm irradiated films) or UV-vis absorption (broad-band illuminated films) spectra were monitored. PMMA films molecularly doped with Ir dopants (15 wt %) were prepared and served as controls.
Degradation product analyses. HPLC experiments were performed on an Agilent, 6120 DW LC/MSD instrument equipped with a Poroshell, EC-C18 column.
The photolyzed solutions were diluted in HPLC grade CH 3 CN (1:9, v/v), and passed through a membrane filter (pore size = 8.0 μm) prior to injection. A 5 μL was injected and allowed to pass through the column at room temperature, using an eluent gradiently increased fractions of CH 3 CN in H 2 O. Chromatographic detection was performed with employing a UV detector (λ obs = 254 nm). Electrospray ionization mass analyses were subsequently performed at a positive ion detection mode (voltage = 70 V) in the range 200-1500 a.m.u. The photoirradiated films of H and Ir were dissolved in CH 3 CN (HPLC grade) for the HPLC analyses.
Device fabrication and characterization. The organic layers used were deposited consecutively on pre-cleaned ITO glass substrates by employing a thermal evaporation system at a pressure <1.0 × 10 −6 torr. A 1-nm-thick Liq layer and a 100nm-thick Al layer were deposited as a cathode through thermal evaporation. The deposition rates of the organic and metal layers were 0.1 and 0.5 nm s −1 , respectively. Deposition of Liq was carried out at a rate 0.01 nm s −1 . The active device area of 4 mm 2 was defined by the area of an overlap between the ITO and Al electrodes. Current, voltage, and luminance of the devices were measured with a system consisting of a Keithley, 2400 Source-Mete,r and a PR-650 spectroradiometer. Operational lifetime measurements of the devices were taken in a constant current mode. LT 70 values were determined from the decay traces of % luminance plotted as a function of operation time. Operation time at which the % luminance decreased to 70% corresponded to LT 70 .
Data availability. The data that support the findings of this study are available from the corresponding author upon reasonable request.