Reducing Efficiency Roll‐Off in Multi‐Resonant Thermally Activated Delayed Fluorescent OLEDs through Modulation of the Energy of the T2 State

The S1 state and high‐lying triplet excited states (ΔES1Tn) offer insight into clarifying the mechanism of efficiency roll‐off of organic light‐emitting diodes (OLEDs). However, experimental detection of the ΔES1Tn is challenging due to Kasha's rule. Here, two emitters, PhCz‐O‐DiKTa and PhCz‐DiKTa, showing multi‐resonant thermally activated delayed fluorescence (MR‐TADF) are reported. By modulating the conjugation between the MR‐TADF DiKTa emissive center and donor substituent, emission directly from the T2 state is for the first time observed in MR‐TADF emitters. Single crystal and reduced density gradient analyses reveal the origin of the reduced observed concentration‐quenching results from weak CH···π and slipped π···π stacking interactions, which suppress nonradiative transitions. Theoretical and photophysical investigations reveal that the ΔES1T2 difference influences the reverse intersystem crossing rate. The OLEDs employing PhCz‐O‐DiKTa and PhCz‐DiKTa as emitters show maximum external quantum efficiencies (EQEmax) of over 20%, but very different efficiency roll‐off behavior (54.5% vs 13.6% at 100 cd m−2). Thus, this design provides a possible solution to mitigating device efficiency roll‐off by designing MR‐TADF emitters with degenerate S1 and T2 states.


DOI: 10.1002/adom.202300114
success of these applications. Fluorescent OLEDs can only convert 25% of singlet excitons into emission, with the loss of 75% triplet excitons as heat. [2] In the past decade, thermally activated delayed fluorescence (TADF) materials have emerged as emitters for the third-generation OLED, able to harvest 100% of excitons to generate light, at comparable exciton utilization efficiencies to phosphorescent OLEDs. Generally, TADF emitters are based on a strongly twisted donor-acceptor (D-A) design motif to electronically separate the highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) that results in a small energy gap between S 1 and T 1 states (ΔE S1T1 ). [3][4][5] However, a relatively large nuclear reorganization and charge-transfer (CT) transition in the excited state unavoidably produces a broad emission band, reflected by a large full width at half maximum (FWHM) of around 100 nm. [6] These undesired properties in D-A TADF emitters restrict their practical application in high-resolution displays where the color point requirements demanded by industry are difficult to obtain using materials showing broad emission.
At present, commercialized high-resolution OLED displays mainly employ a color filter or microcavity to reach the color purity required by broadcasting standards, which is inevitably accompanied by losses in device efficiency and energy utilization. [7] Therefore, it would be of great utility to devise an emitter showing narrowband emission. Recently, Hatakeyama et al. introduced a subclass of TADF emitters coined multi-resonant TADF (MR-TADF), that shows the desired narrowband emission. This is due to their rigid polyaromatic hydrocarbon (PAH) framework, [8] where the rigid molecular skeleton and the shortrange CT (SRCT) nature of the emissive singlet excited state enable simultaneously narrowband emission and small ΔE S1T1 . In this design, the electron densities of the HOMO and LUMO are localized on the adjacent atoms. Typically, boron/nitrogen (B/N) doped PAHs have emerged as the dominant design. For example, Hatakeyama and co-workers designed a B/N-doped MR-TADF compound, ν-DABNA, that shows very narrowband emission (FWHM of 18 nm) at a photoluminescence (PL) maximum (λ PL ) of 467 nm, with a PL quantum yield (Φ PL ) of 90%, and where the device's maximum external quantum efficiency The S 1 state and high-lying triplet excited states (ΔE S1Tn ) offer insight into clarifying the mechanism of efficiency roll-off of organic light-emitting diodes (OLEDs). However, experimental detection of the ΔE S1Tn is challenging due to Kasha's rule. Here, two emitters, PhCz-O-DiKTa and PhCz-DiKTa, showing multi-resonant thermally activated delayed fluorescence (MR-TADF) are reported. By modulating the conjugation between the MR-TADF DiKTa emissive center and donor substituent, emission directly from the T 2 state is for the first time observed in MR-TADF emitters. Single crystal and reduced density gradient analyses reveal the origin of the reduced observed concentrationquenching results from weak CH···π and slipped π···π stacking interactions, which suppress nonradiative transitions. Theoretical and photophysical investigations reveal that the ΔE S1T2 difference influences the reverse intersystem crossing rate. The OLEDs employing PhCz-O-DiKTa and PhCz-DiKTa as emitters show maximum external quantum efficiencies (EQE max ) of over 20%, but very different efficiency roll-off behavior (54.5% vs 13.6% at 100 cd m −2 ). Thus, this design provides a possible solution to mitigating device efficiency roll-off by designing MR-TADF emitters with degenerate S 1 and T 2 states.

Introduction
Organic light-emitting diodes (OLEDs) are surging in use across multiple markets such as those of wearable electronics, smartphones, flat displays, and solid-state lighting. This is largely driven by their numerous promising advantages compared to conventional liquid crystal displays (LCDs) such as lower cost, greater flexibility, better picture quality, and fast response time. [1] Fabrication of high-performance OLEDs is paramount to the (EQE max ) was an impressive 34.4%. [9] A hyperfluorescence (HF) strategy has been extensively employed to mitigate efficiency roll-off in the device, wherein generally a TADF emitter serves as the assistant dopant to efficiently harvest excitons and then these are efficiently transferred to the terminal emitter, a compound showing narrowband emission and with a fast radiative rate constant, k r . [10] Benefitting from these pioneering works, Lee and co-workers reported an HF device that showed an EQE max of 33.5% and a long operational device lifetime (LT 50 , 50% of initial luminance) of 9500 h at 100 cd m −2 by employing ν-DABNA as the terminal emitter and a triazine derivative PPCzTrz as the assistant dopant; [11] the device with only ν-DABNA as the emitter was much less stable (LT 50 = 31 h). [9] A similarly themed report by Adachi and co-workers documented an HF device using ν-DABNA as the terminal emitter with an EQE max of 41% and an LT 50 > 10 000 h. [12] Besides HF, other strategies have been explored to improve the performance of MR-TADF OLEDs. [13][14][15][16][17][18] For instance, Yang and co-workers proposed an effective strategy for mitigating aggregation-caused quenching (ACQ) by introducing sterically bulky groups onto a B/N-doped emitter. [13] Carbonyl (CO)/N-doped PAHs are a second family of MR-TADF emitters. [19][20][21][22][23][24] Initial studies by Yuan et al. and us using DiKTa (Figure 1a) derivatives documented the potential of this design; [21,25] however, the devices typically suffered from significant efficiency roll-off (>30%) at 100 cd m −2 and low maximum luminance (L max ≤ 13 000 cd m −2 ). Recently, we demonstrated that by decorating DiKTa with various donors, the OLEDs showed an improved EQE max and reduced efficiency roll-off compared to the device with the parent emitter; [26] however, this improvement in efficiency roll-off was not present beyond 100 cd m −2 . Similarly, Zhang and co-workers reported three CO/N-doped PAHs, QAD-Cz, QAD-2Cz, and QAD-mTDPA. [27] Although having very similar molecular structures and near unity Φ PL , there is a stark dichotomy in the observed efficiency roll-off at 100 cd m −2 (QAD-Cz: 73.4%, QAD-2Cz: 12.5%, and QAD-mTDPA: 51.0%). These results provide insight that a high-lying triplet excited state (T n ) is likely responsible for the differences in reverse intersystem crossing (RISC) rate constants, k RISC , and thus the different efficiency roll-off profiles. Improved MR-TADF design can be realized if we were above to experimentally determine the energy of these close-lying triplet states to the S 1 state. The employment of transient absorption spectroscopy can be used to detect the existence of T n states, [28] but it lacks the capability of quantifying the energy of these states. Very recently, transient electron paramagnetic resonance spectroscopy (EPR) was employed by Evans and co-workers to reveal that multiple triplet excited states aid to mediate delayed fluorescence. [29] In our previous work, we utilized transient EPR to distinguish the T 2 and T 1 states of the anthracene derivative DPAAnCN and revealed that exciton harvesting proceeded by a hybridized local and charge-transfer (HLCT) fluorescence mechanism. Despite these advances, it remains challenging to map out the triplet-state landscape.
We recently demonstrated using time-resolved PL spectroscopy that higher-lying triplet states could be observed in multiple-phosphorescence emitters. [30,31] Capitalizing on this design paradigm, here we utilize conjugation modulation as a tool to brighten the T 2 state of MR-TADF emitters. We report two DiKTa derivatives PhCz-O-DiKTa and PhCz-DiKTa (Figure 1a), and explore the impact of conjugation from the decorated donor moiety on the efficiency roll-off of the OLED ( Figure 1b); in the conceived emitter design the degree of conjugation between 9-phenyl-9H-carbazole (PhCz) and DiKTa is employed to regulate the vibronic coupling between two triplet excited states and activate emission from the T 2 state; PhCz acts as a perturbator of the energy of the T 2 state to tune k RISC (Figure 1c). The Φ PL values as a function of doping level reveal that both emitters are reasonably resistant to concentration quenching. The existence of the T 2 state was demonstrated by time-resolved PL measurements and further confirmed by the observable T 2 afterglows. The OLED devices employing PhCz-O-DiKTa and PhCz-DiKTa as emitters indicate 100% exciton utilization assuming a 25% out-coupling efficiency. Similar EQE max (PhCz-O-DiKTa: 20%; PhCz-DiKTa: 22%) values were obtained; however, there is a stark contrast in efficiency roll-off at 100 cd m −2 (PhCz-O-DiKTa at 54.5% and PhCz-DiKTa at 13.6%).  PhCz and DiKTa as building blocks. The synthetic routes are outlined in Scheme S1, Supporting Information. All samples were purified via silica column chromatography followed by temperature-gradient vacuum sublimation. A combination of 1 H & 13 C nuclear magnetic resonance spectroscopy (NMR), high-resolution mass spectrometry (HRMS), melting point determination, high-performance liquid chromatography (HPLC), elemental analysis (EA), and single-crystal X-ray diffraction were used to confirm the molecular structures and assess their purity (Figures S1-S13 and Table S1, Supporting Information).

Theoretical Modeling
First, the photophysical properties of PhCz-O-DiKTa and PhCz-DiKTa were modeled using density functional theory at the PBE0/6-31G(d,p) level. [32,33] As shown in Figure 2a, PhCz-O-DiKTa and PhCz-DiKTa possess the LUMOs of similar energy and character, reflected by the DiKTa-localized density distribution. A slight energy difference between the HOMOs of the two compounds was observed, where the HOMO is localized on the PhCz subunit in PhCz-O-DiKTa but destabilized in PhCz-DiKTa due to the extended π-conjugation possible in this compound. Time-dependent (TD) DFT calculations within the Tamm-Dancoff approximation (TDA) [34] performed at the same level of theory predict that PhCz-O-DiKTa has a larger spin-orbit coupling constant between the S 1 and T 1 states at the optimized S 1 geometry than PhCz-DiKTa (0.43 vs 0.10 cm −1 ) but similar oscillator strength (0.11 vs 0.12), and both exhibit a relatively large energy gap between S 1 and T 1 , ΔE S1T1 > 0.4 eV, (Figure S14a, Supporting Information), which implies that neither compound will show TADF. Calculated vertical excitations to the S 1 state predict that the character of this state in both compounds is hybridized local and CT ( Figure S14b, Supporting Information). We have previously demonstrated the importance of using wavefunction-based methods to accurately model the excited state energies and ΔE S1T1 of MR-TADF emitters, [35,36] thus, we computed the energies of low-lying singlet and triplet excited states using spin-component scaling secondorder algebraic diagrammatic construction (SCS-ADC (2)) functional with the cc-pVDZ basis set (Figure 2b). [37] The calculation data are summarized in Table S2, Supporting Information. As anticipated, a somewhat reduced ΔE S1T1 (≈0.25 eV) exists in both PhCz-O-DiKTa and PhCz-DiKTa. The S 1 oscillator strength of both compounds is over 0.24, reflecting a highly emissive state. The calculated distance (d CT ) of the S 1 state of DiKTa is 0.803 Å (Table S1, Supporting Information). A similar d CT exists in PhCz-O-DiKTa (0.944 Å), but a larger value is recorded for PhCz-DiKTa (1.292 Å) due to the π-conjugated geometry. Nevertheless, difference density plots of the four lowest-lying excited states reveal a remarkably SRCT localized on the DiKTa core ( Figure 2b). Such a small reorganization of the electron density in the S 1 state is expected to translate into a narrowband emission. The difference density plots for PhCz-O-DiKTa and PhCz-DiKTa are similar to that of DiKTa ( Figure S15, Supporting Information), thus demonstrating that the incorporation of PhCz only causes a perturbation of the excited-state energies without remarkably influencing the nature of the lowlying excited-states. Notably, although an ΔE S1T1 of ≈0.25 eV has been predicted, it remains challenging to achieve fast k RISC at 298 K with this energy gap as according to the Boltzmann distribution, the approximate exciton ratio between the S 1 and T 1 is 6 × 10 −5 :1. Thus, to promote RISC in these compounds, a higher-lying triplet excited state T 2 of different orbital type to the S 1 state must be implicated to enable secondorder coupling. [28] The calculations predict that the S 1 and T 2 states of the two emitters are dominated by SRCT and n-π* Adv. Optical Mater. 2023, 11, 2300114 transitions (Figure 2b), respectively. PhCz-DiKTa possesses a smaller ΔE S1T2 than that of PhCz-O-DiKTa (0.21 vs −0.08 eV) ( Figure 2b). Therefore, we can reasonably expect that faster k RISC and shorter TADF lifetime (τ DF ) occur in PhCz-DiKTa than in PhCz-O-DiKTa via second-order spin-vibronic coupling. As a result, it is vital to experimentally detect the T 2 energy to clarify the RISC processes.

Optoelectronic Characterization in Solution
We next investigated the optoelectronic properties in dilute solution. The energies of FMOs were inferred from the electrochemistry, where oxidation and reduction potentials (E ox /E red ) were obtained using cyclic voltammetry (CV) and differential pulse voltammetry (DPV) in deaerated dichloromethane (DCM) with 0.1 m [ n Bu 4 N]PF 6 as the supporting electrolyte. The electrochemical data are presented in Table S3 [25] result from the weak electron-donation of the PhCz donor with the DiKTa core.
UV-vis absorption spectra of the two emitters in toluene reveal a relatively intense SRCT transition at ≈450 nm (Figure 3b). Compared to DiKTa, [25] PhCz-O-DiKTa possesses a similar absorption coefficient (ε < 1 × 10 4 M −1 cm −1 ) around 300-400 nm, which is highly enhanced in PhCz-DiKTa due to the π-conjugated geometry. At 298 K, narrowband emissions were evidenced in the steady-state PL spectra of PhCz-O-DikTa and PhCz-DiKTa, with λ PL at 478 and 490 nm and FWHMs of 39 and 55 nm (Figure 3b), respectively. The narrower and slightly blue-shifted emission of PhCz-O-DiKTa results from the π-conjugation-breaking geometry, which restricts the electronic coupling between PhCz and DiKTa. This is reflected by the larger interplanar angle between PhCz and DiKTa in PhCz  character; this is also substantiated by a larger calculated S 1 d CT (1.292 Å) compared with that of PhCz-O-DiKTa (d CT = 0.944 Å). The emission of PhCz-O-DiKTa at ≈400 nm in DMF is ascribed to LE emission (Figure 3d). In degassed toluene, steady-state PL intensities of the two emitters are both enhanced compared to those recorded in aerated toluene ( Figure S18, Supporting Information). Φ PL values of PhCz-O-DikTa and PhCz-DiKTa in aerated toluene are 62% and 45%, which improve to 71% and 61%, respectively, demonstrating the contribution from triplet excitons ( Figure S19, Supporting Information). The ΔE S1T1 values of PhCz-O-DikTa and PhCz-DiKTa were estimated to be 0.36 eV, from the difference in energy of the onset of prompt and delayed emission spectra in 2-methyltetrahydrofuran (2-MeTHF) glass at 77 K (Figure 3c,f). The photophysical data are summarized in Table S4, Supporting Information. The magnitude of this ΔE S1T1 implies that TADF is not operational in these compounds. Reminiscent of the design model ( Figure 1c) and theoretical calculations (Figure 2b), the higherenergy emission band probably originates from a higher degenerate triplet excited state, here defined as the T 2 state, which is expected to be the bridge for TADF occurrence in this design.

Analysis of Intermolecular Interactions
Prior to investigating the photophysical behaviour in the solid state, we first studied the X-ray single crystal structures of  (Figure 4a), although they do act in a mutually supportive manner. PhCz-O-DiKTa packs in a head-to-tail manner, but each end of the molecule interacts with a separate neighboring molecule, leading to the formation of weakly interacting chains. The more planar PhCz-DiKTa also displays a head-to-tail intermolecular stacking pattern, but with both head and tail of the molecule interacting with the same neighbor, and thus strengthens molecular electronic coupling between adjacent emitter molecules, and broadens the PL. The reduced density gradient (RDG) analysis using the Multiwfn program [39] corroborates the existence of intermolecular Van de Waals interactions, as well as weakly repulsion steric interactions (Figure 4b), which are beneficial to restricting molecular motions.

Photophysical Investigations in the Solid State
We next turned our attention to the photophysical characterization in an OLED-relevant host matrix, 1,3-bis(N-carbazolyl)benzene (mCP) as it has a suitably high triplet energy (E T = 3 eV). [40] The mCP matrix can also contribute to the suppression of nonradiative processes in the emitters, and host-guest interactions can also regulate triplet vibronic coupling and energies. [30,41] We first studied the steady-state PL behavior of the two emitters in mCP at various doping concentrations ( Figure S20  those at 1 (Φ PL < 55%) and 5 wt% (Φ PL < 70%) doping (Table  S5, Supporting Information). Thus, 10 wt% emitter-doped mCP films were selected for further photophysical measurements. The corresponding photophysical data are summarized in Tables S6 and S7, Supporting Information. Similar spectral profiles of the steady-state (SSPL), prompt, and delayed emissions were recorded in PhCz-O-DikTa and PhCz-DiKTa, with the SSPL maxima, λ PL , of ≈500 and ≈525 nm (Figure 5a,b) (Table S6, Supporting Information), respectively, which improve to 80% and 91% in N 2 , respectively. Based on the contribution of the delayed component to the average τ DF , the calculated Φ DF associated with the delayed fluorescence in N 2 is 59% and 78% for  (Table S7, Supporting Information), respectively, representing a 12-fold enhancement in the latter despite the similar ΔE S1T1 . However, according to Marcus electron transfer theory, [42] the k RISC of PhCz-DiKTa should be slower than that of PhCz-O-DikTa because the SOC constant of PhCz-DiKTa between S 1 and T 1 states is much smaller than that of PhCz-O-DiKTa (0.10 versus 0.43 cm −1 , Figure S14a, Supporting Information). Thus, a higher-lying triplet excited state T 2 probably serves as a bridge for boosting the spin vibronic coupling via second-order perturbation.
Temperature-dependent PL measurements were further performed to evidence the presence of the T 2 state. As temperature decreases from 298 to 200 K, the λ PL of the steady-state PL spectrum of PhCz-O-DiKTa gradually red-shifts from 504 to 512 nm (Figure 6a), together with a remarkable enhancement in emission intensity. This excludes TADF as the origin for this spectral change as the intensity should weaken with decreasing temperature if the origin were TADF. Thus, this highly strengthened PL must originate from emission from triplet excitons. As the temperature further decreases to 77 K, the emission intensity at 512 nm diminishes accordingly, and a new concomitant T 1 emission at 540 nm emerges. Evidently, the intensity ratio between 540 and 512 nm increases due to the  suppressed thermal activation of excitons from T 1 to T 2 states (Figure 6a). The picture of PhCz-DiKTa is similar as a function of temperature (Figure 6b). Next, temperature-dependent timegated measurements were conducted to isolate the T 1 and T 2 emissions. As the temperature decreases to 200 K, compared to the TADF emission at ≈500 nm, the enhanced T 2 phosphorescence centered at 515 nm appears (as a result of the combined suppressed RISC and nonradiative transitions), and then gradually disappears. Alongside this, there is the emergence and enhancement of T 1 phosphorescence at 550 nm as temperature decreases to 77 K, which cuts off thermally activated reverse internal conversion. T 1 and T 2 phosphorescence can also be evidenced by green and yellow phosphorescence afterglows at 200 and 77 K (Figure 6c, inset), respectively. Notably, the T 2 emission recorded in mCP is blue-shifted compared to T 1 recorded in 2-MeTHF glass at 77 K, which removes the possibility that T 2 phosphorescence originates from emission from low-energy aggregates ( Figure S23, Supporting Information). A similar phosphorescence evolution over temperature can be observed in PhCz-DikTa (Figure 6d). At 200 K, there is a strongly enhanced T 2 phosphorescence at 534 nm, while at 77 K T 1 phosphorescence dominates the time-gated PL spectrum. Similarly, two different phosphorescence afterglows were observed at 200 and 77 K (Figure 6d, inset). Normalized phosphorescence spectra of PhCz-DiKTa and PhCz-O-DiKTa gradually red-shift with decreasing temperature, reflecting the varying contributions from emissive T 2 and T 1 states ( Figure S24, Supporting Information). The increasing phosphorescence lifetimes measured at 515 and 550 nm as temperature decreases are consistent with the presence of emissive T 2 and T 1 states ( Figures S25 and S26, Supporting Information). The T 2 energies, estimated from the onsets of the phosphorescence spectra of PhCz-O-DiKTa and PhCz-DiKTa at 200 K, are 2.54 and 2.51 eV ( Figure S27, Supporting Information), respectively; thus, the estimated ΔE S1T2 values are 0.11 and 0.07 eV, respectively, which are close to predicted ΔE S1T2 values shown in Figure 2b. The smaller ΔE S1T2 of PhCz-DiKTa correlates with the recorded shorter τ d and faster k RISC compared to those of PhCz-DiKTa (Figure 5c and Table  S7, Supporting Information).
Shown in Figure 7c are the EL spectra of the two devices. EL maxima (λ EL ) of the devices with PhCz-O-DiKTa and PhCz-DiKTa are 497 and 524 nm and associated FWHM of 58 and 72 nm, respectively. The corresponding Commission Internationale de l'Éclairage (CIE) coordinates are (0.21, 0.55) and (0.31, 0.61) ( Figure S28, Supporting Information). The high coinci-dence in PL and EL spectra demonstrates the efficient energy transfer in the emissive layer between the host and the emitter. All the devices show relatively low turn-on voltages (V on ) of ≈3 eV at maximum EQE. The obtained high maximum luminance (L max ) is over 22 000 cd m −2 , benefitting from the suitable energy cascade between the functional layers ( Figure 7e). Given that the measured Φ PL value of PhCz-O-DiKTa and PhCz-DiKTa are 80% and 91%, respectively, the theoretical EQE max values are 20% and 22.75%, assuming an out-coupling efficiency of 25%. As depicted in Figure 7f   of 54.5% is observed in the device with PhCz-O-DiKTa, which is ameliorated considerably to 13.6% in the device with PhCz-DiKTa. The EQE reduces to 2.9% at a luminance of 1000 cd m −2 for the device with PhCz-O-DiKTa; however, for the device with PhCz-DiKTa, the EQE 1000 remains at 11.1%. A plausible explanation for this difference is that the smaller ΔE S1T2 in PhCz-DiKTa serves to mitigate the efficiency roll-off due to the faster k RISC and shorter delayed lifetime, thus reducing triplet exciton populations at high current densities. However, most of the previously reported ketone-based MR-TADF OLEDs showed a rather severe efficiency roll-off (>45%); [21][22][23][24][25][26]43,44] the origin of the large roll-off has not been explored in any detail. A summary of previously reported EL data related to ketone-based MR-TADF devices is presented in Figures S30 and S31, Supporting Information. This report experimentally demonstrates that the T 2 state is the key to regulating k RISC (Figure 7d) and provides insight into improving the efficiency roll-off of devices, at least with ketone-containing MR-TADF derivatives.

Conclusions
In summary, we demonstrate how it is possible to dramatically reduce efficiency roll-off in MR-TADF OLEDs by judiciously decorating the core emitter with a PhCz group. This serves to reduce the ΔE S1T2 , leading to an accelerated RISC. Theoretical modeling of two MR-TADF emitters, PhCz-O-DiKTa and PhCz-DiKTa, predicts that the T 2 state could act as a conduit in boosting k RISC . Temperature-dependent steady-state and time-gated PL measurements provided evidence of the presence and energies of the T 2 state, which explains why PhCz-DiKTa possesses a near 12-fold enhancement in k RISC compared to PhCz-O-DiKTa. The OLED devices with PhCz-O-DiKTa and PhCz-DiKTa exhibited EQE max values of 20% and 22%, respectively, indicating 100% exciton utilization. The remarkable divergence in efficiency roll-off of the two devices was ascribed to the smaller ΔE S1T2 in the device with PhCz-DiKTa. This study provides a rarely documented yet clear picture of the dynamics of intermediate triplet excited states and a possible solution to improve device performance without sacrificing color purity.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.