Highly Efficient Thermally Activated Delayed Fluorescence from an Excited-State Intramolecular Proton Transfer System.

Thermally activated delayed fluorescence (TADF) materials have shown great potential for highly efficient organic light-emitting diodes (OLEDs). While the current molecular design of TADF materials primarily focuses on combining donor and acceptor units, we present a novel system based on the use of excited-state intramolecular proton transfer (ESIPT) to achieve efficient TADF without relying on the well-established donor-acceptor scheme. In an appropriately designed acridone-based compound with intramolecular hydrogen bonding, ESIPT leads to separation of the highest occupied and lowest unoccupied molecular orbitals, resulting in TADF emission with a photoluminescence quantum yield of nearly 60%. High external electroluminescence quantum efficiencies of up to 14% in OLEDs using this emitter prove that efficient triplet harvesting is possible with ESIPT-based TADF materials. This work will expand and accelerate the development of a wide variety of TADF materials for high performance OLEDs.

Introduction

Thermally activated delayed fluorescence (TADF) has attracted considerable interest for practical use in organic light-emitting diodes (OLEDs) since the report of a high external electroluminescence (EL) quantum efficiency (ηext) of over 20% in 2012.[1−5] In accordance with spin statistics, the generation of singlet and triplet excitons in a 1:3 ratio upon charge carrier recombination in an OLED had limited the internal EL quantum efficiency (ηint) to a low level of ∼25% for standard fluorescent materials.[6] Although phosphorescent emitters based on heavy-metal complexes can harvest both singlet and triplet excitons to achieve ηint of nearly 100%,[7−9] phosphorescent materials containing rare metals such as iridium and platinum face potential cost issues.

Delayed fluorescence (DF) offers an alternative process for harvesting triplet excitons without using rare metals. Generally, DF materials can be classified based on the underlying process of either triplet–triplet annihilation (TTA, P-type DF) or thermal activation (TADF, E-type DF). Although TTA can improve OLED efficiency in fluorescence-based OLEDs through the formation of additional singlets from the annihilated triplets, at least half of the triplets undergo nonradiative energy transfer to S0, resulting in a maximum ηint of 62.5%.[10−12] On the other hand, the TADF process involves a thermally activated up-conversion via reverse intersystem crossing (RISC) from the lowest energy triplet excited state (T1) to the lowest energy singlet excited state (S1) in systems with a small singlet–triplet splitting energy (ΔEST). Therefore, 100% exciton utilization is possible.

Although the phenomenon has been known for more than 50 years,[13] TADF was considered to be too inefficient to act as an effective pathway for high efficiency EL until recently[14] because of the rather low photoluminescence (PL) quantum yield (ΦPL) and RISC efficiency of traditional TADF materials such as fullerene derivatives,[15−19] ketones,[20−23] and thiones.[24] Localized n-π* transitions are a potential route to efficient TADF because the lack of overlap between n and π* orbitals leads to a small ΔEST, but the small oscillator strength of n-π* transitions seems to result in a low ηext in OLEDs.[25] An alternative strategy to obtain small ΔEST (generally less than 0.2 eV for efficient TADF) is based on minimizing the exchange integral between the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) through careful separation of the HOMO and LUMO in a molecule or between molecules.

Currently, the most successful design strategy for TADF molecules is an intra-/intermolecular donor–acceptor (D–A) system exhibiting a charge-transfer (CT) transition, and such systems have achieved excellent OLED performance based on the simultaneous realization of efficient RISC and high ΦPL.[26−31] Although this system has the advantage of an availability of a large variety of π-conjugated aromatics, the molecular designs are limited to donor and acceptor units combined with either a twisted conformation or a nonconjugated bridge between them to achieve an effective HOMO–LUMO separation. In an exceptional case developed very recently by the Hatakeyama group, a series of emitters (DABNA) based on fused D–A units exhibited efficient TADF by achieving HOMO–LUMO separation through a distinctive system called the multiple resonance effect.[32] However, the potential versatility of this molecular design still needs to be investigated more deeply since the derivatives with oxygen atoms instead of nitrogen atoms as the electron-donating group in a polycyclic framework did not lead to this phenomenon.[33,34]

The design we report herein completely differs from those previously established approaches by forgoing the use of a D–A system and enables the creation of efficient TADF emitters with new features, such as a fully fused, planar, and rigid ring system, that were previously not possible. The spatial separation of the HOMO and LUMO is obtained by an excited-state intramolecular proton transfer (ESIPT) process, which is a classical molecular photoisomerization process that affords a large Stokes shift.[35,36] In this process, a hydrogen atom covalently bonded to one atom and hydrogen bonded to a second in the same molecule switches to being covalently bonded to the second atom and hydrogen bonded to the first upon excitation. While TADF from nonrigid ESIPT molecules in solution has been previously observed by Park et al.,[37] emission in OLEDs occurs in the solid state, which may restrict molecular conformational changes, and neither TADF from ESIPT molecules in the solid state nor a detailed emission mechanism have been reported. In this study, we computationally demonstrate a reduction of ΔEST resulting from redistribution of the HOMO and LUMO after ESIPT, which is possible in rigid structures without conformational isomers. The existence of the triplet excited state upon charge recombination is investigated by using several host materials with different T1 levels, and OLEDs containing an ESIPT-based TADF emitter in an appropriate host reach high ηext of up to 14%, which confirms a large contribution of the triplets to emission in the device. This is the first example of an OLED using an ESIPT-based TADF material and offers an additional design approach to realizing a small ΔEST for high-performance TADF molecules.

Results and Discussion

Synthesis and Characterization of TQB Analogues

The molecular structure of the compound under investigation, triquinolonobenzene (TQB), is depicted in Figure along with the structures of related compounds. The structure consists of an acridone moiety, which is also a substructure of quinacridone (QD), a well-known pigment having outstanding stability and weather resistance.[38] Therefore, TQB is expected to also have high chemical stability and photostability. The synthesis and UV–vis absorption properties of TQB were reported in 1990 with the aim of developing of a new pigment,[39] but the compound is colorless and, thus, useless as a pigment. However, the compound holds unexplored potential as a fluorescent material since some pigments show excellent fluorescence characteristics. In addition, ESIPT is made possible by intramolecular hydrogen bonding in TQB and can lead to unique changes in the electronic structures of the molecule with consequences for processes such as TADF. Therefore, we synthesized TQB according to the literature along with the newly developed angular-shaped quinacridone (-QD), as detailed in Methods S1 in the Supporting Information.[39]

Figure 1

Chemical structures of QD, -QD, and TQB and the proton transfer model for TQB. The energy diagram shows the computationally calculated energies in the ground state for TQB-TA and in the excited state for TQB-TA, TQB-TB, TQB-TC, and TQB-TD. S1-ver: vertical excitation. S1-adi: adiabatic excitation.

The synthesis is simple and possible on the gram-scale. The products were purified by conventional vacuum sublimation. Solids of the products are crystalline and thermally stable up to at least 400 °C (Figure S1). Additionally, the stability of the excited state was investigated by observing the photodegradation behavior upon UV irradiation (Figure S2). The PL intensity normalized to the initial decreased more slowly for TQB than the standard TADF material (4s,6s)-2,4,5,6-tetra(9H-carbazol-9-yl)isophthalonitrile (4CzIPN), which is one of the most stable TADF materials among those reported.[1,40,41]

X-ray crystal structure analysis revealed that TQB has a completely planar conformation (RMS = 0.036 Å; see Methods S2, Figure S3, and Table S1 in the Supporting Information) and a π-stacking (3.39 Å) structure with 180° flip disorder molecules of 50/50 occupancy arising from the pseudo symmetry. Unfortunately, discussing the hydrogen bonding as well as the ground state geometry of protons further based on the X-ray data is difficult because of the crystal disorder, but 1H NMR can also provide information on hydrogen bonding. The N–H protons were observed at 15.7 ppm for TQB, 14.4 and 12.6 ppm for -QD, and 11.9 ppm for QD. Since stronger hydrogen bonding leads to a downfield shift of the peak, these results indicate strong intramolecular hydrogen bonding for TQB and one of the N–H protons in -QD. As molecules exhibiting ESIPT necessarily have intramolecular hydrogen bonding in the ground state, TQB and -QD are likely to undergo ESIPT.

Computational Calculation of the Effect of ESIPT

The proton transfer reaction can give four different tautomers, TQB-TA, TQB-TB, TQB-TC, and TQB-TD in Figure , so the stationary points of both the ground S0 and the first excited S1 states were investigated for each tautomer by density functional theory (DFT) and time-dependent DFT (TD-DFT) methods, respectively, at the B3LYP/6-31++G(d,p) level (Methods S3 in the Supporting Information). The most stable tautomer in the ground S0 state (Table S2) is TQB-TA (3N–H, acridone form). A ground state of TQB-TA is experimentally supported by the Fourier transform infrared (FT-IR) spectroscopy and 1H NMR, which indicate an acridone-based structure with a C3h symmetry in the ground state like TQB-TA based on the similar IR spectra among TQB, -QD, and QD and the single peak in the NMR corresponding to the proton involved in hydrogen bonding. The photoinduced tautomerization begins with vertical excitation in the ground state geometry by UV absorption (S1-vertical), followed by the structural rearrangement to the global minimum of the first excited state (S1-adiabatic) of TQB-TA. The energies of the excited states of the tautomers (Figure and Table S3) reveal that the tautomer with a single transferred proton, TQB-TB, is the most stable isomer with a large decrease in energy (0.41 eV or 9.6 kcal/mol) compared to TQB-TA. The tautomer with two transferred protons, TQB-TC, is also found to be more stable relative to TQB-TA (energy decrease of 0.06 eV or 1.5 kcal/mol). On the other hand, TQB-TD with three transferred protons was calculated to have an excited state energy higher than those of the other tautomers, making it unlikely to be formed, so it was not investigated further.

Although the energy difference between TQB-TB and TQB-TC is relatively large (0.35 eV), there is a long-standing discussion regarding whether the excited-state double-proton transfer (ESDPT) process is a concerted (synchronous) or sequential (asynchronous) reaction.[42,43] Thus, the existence of TQB-TC in the excited state is an interesting matter to be studied in detail. Unfortunately, the S1 → S0 emission energies have similar values (2.3–2.4 eV) for both TQB-TB and TQB-TC, making experimental determination difficult since the emission from the two tautomers cannot be easily distinguished. Instead, we calculated the two-dimensional (2D) potential energy surface (PES) plots of S0 and S1 as shown in Figure and Figure S4.

Figure 2

Calculated 2D potential energy surface (PES) plots. (a) Chemical structure of TQB with atom labels. (b) PES of the first excited state S1. (c) PES of the ground state S0. Symbols: star = TQB-TA, circle = TQB-TB, triangle = TQB-TC. TQB-TD was not investigated because of its much higher energy at its stationary point.

The S1 map clearly indicates sequential proton transfer is expected because of the large energy barrier for the concerted transfer of two protons. This is the same conclusion reached in other investigated cases of ESDPT.[42,43] In addition, the first proton transfer moves through a very smooth potential surface, suggesting effective and ultrafast tautomerization toward TQB-TB in the excited state. The calculations also reveal the existence of a large barrier between TQB-TB and TQB-TC. Overall, tautomer TQB-TB is considered to be the main fluorescent species by taking into account a sequential proton transfer and the large increase in S1 energy from TQB-TB to TQB-TC. Calculations for -QD also confirm that -QD-TA (N–H, acridone form) is stable in the ground state and that the emission will come from -QD-TB (O–H form) after ESIPT (Tables S2–S3).

The calculated HOMOs and LUMOs of TQB are depicted in Figure . TQB-TA has degenerate HOMOs and LUMOs, with the HOMOs filling many of the gaps in the LUMOs and vice versa. Although this distribution seems to be similar to the multiple resonance effect of the DABNA series,[32] the calculated ΔEST of TQB-TA is rather large (>0.5 eV, see Table S3), and TQB-TA is actually not an emissive form. In contrast, the well-separated HOMO and LUMO of TQB-TB provide clear evidence that a smaller ΔEST might be achievable after ESIPT.

Figure 3

Distributions of electron density in HOMO–1, HOMO, LUMO, and LUMO+1. The computational calculations were performed at the B3LYP/6-31++G(d,p) level for TQB-TA, TQB-TB, -QD-TA, and -QD-TB.

In general, ESIPT tautomer species exhibit a drastic redistribution of electronic density that results in the proton donor becoming a stronger donor and the proton acceptor becoming a stronger acceptor. The separation of the orbitals was quantified by calculating the overlap integral of the hole and electron distributions, which roughly correspond to those of the natural transition orbital (NTOs), of TQB-TA and TQB-TB (Figure S5).[44] The values of the overlap integrals decreased after ESIPT, indicating better separation of the highest occupied NTO and lowest unoccupied NTO. Interestingly, TQB-TA showed relatively large values, suggesting that it may not exhibit the multiple resonance effect despite the substitution of alternating electron donating amino and electron withdrawing carbonyl groups.

Although the HOMO and LUMO distributions of -QD-TB are quite similar to those of TQB-TB, the calculated ΔEST of -QD-TB was still large (>0.4 eV). This might be explained by a conjugation area for -QD that is too small and further orbital delocalization over the additional acridone moiety in TQB. The triplet excitations are characterized as HOMO → LUMO (85%), HOMO → LUMO+1 (6%), and HOMO–1 → LUMO (2%) transitions in TQB-TB, whereas they are only a HOMO → LUMO transition in -QD-TB. The spreading of the LUMO+1 of TQB-TB over a larger volume that includes the relatively neutral acridone site, where the HOMO is less localized, should also be beneficial for reducing ΔEST.

Photophysical Properties

The UV–vis absorption and PL spectra of TQB in toluene, THF, and DMF are shown in Figure a along with those calculated by DFT for the absorption of TQB-TA and the emission of TQB-TB. The experimental data are in good agreement with the simulated data, supporting the existence of proton transfer in the excited state. Although emission from ESIPT materials before proton transfer is sometimes observed in polar solvents, emission from TQB-TA (expected to be around 380 nm based on the calculations) could not be observed even in DMF, indicating ultrafast ESIPT. The PL spectra slightly red shift with an increase in the polarity of the solvent. Films of TQB doped into various hosts exhibited the same tendency as shown in Figure b and Figure S6, although the difference is small. In addition, a neat film and bulk crystal showed quite similar emission maxima despite the strong π-stacking interactions.

Figure 4

PL and transient PL spectra in solution and solid state. (a) Ultraviolet–visible (UV–vis) absorption (dashed lines) and photoluminescence (PL) spectra (solid lines) for TQB measured in solution and calculated by DFT (for the calculations, the TQB-TA structure was used for absorption and the TQB-TB structure for the emission). (b) UV–vis absorption (dashed lines) and PL spectra (solid lines) for TQB-doped films (ca. 10 wt %), a neat TQB film, and a TQB crystal (excitation spectrum is displayed instead of UV–vis absorption spectrum for the crystal). (c) Temperature dependence of transient PL decay for a TQB-doped film of DPEPO. Inset: PL spectra at 6 K (no delayed component) and at 300 K resolved into prompt and delayed components. (d) PL transient decay spectra at room temperature of TQB-doped films (ca. 10 wt %) of DEPPO, PPT, mCBP, CBP, and CzSi that were encapsulated in a glovebox. IRF is the instrument response function.

The ΦPL of the neat film and crystal were measured to be 21% and 62% (Table S4). In general, the suppression of π-stacking is considered to be effective for overcoming concentration quenching such as in the design of systems for aggregation-induced emission (AIE) and crystallization-induced emission (CIE).[45,46] Since the ΦPL of the neat film is much higher than just a few percent and that of the crystal is higher compared to other conditions such as in solution (∼59%) and doped films (40–55%), a well-aligned π-stacking columnar structure does not cause concentration quenching in this system, which has also been reported for other ESIPT materials.[37,47] This might be because the ESIPT process creates a pseudo doped system of emissive TQB-TB species in a host of TQB-TA, thereby avoiding resonant energy transfer. In addition, the tight stacking in the crystal may suppress molecular motion leading to nonradiative relaxation.

In toluene solution, the ΦPL increased after N2 bubbling from 37% to 59%, indicating the contribution of triplets to the fluorescence process. Additionally, the transient PL clearly showed both prompt PL (6.9 ns) and DF (74 μs) (Figure S7). The DF component as well as the ΦPL of TQB decreased with an increase of the polarity of the solvent. In contrast, -QD exhibited ESIPT and a very small ΦPL of <5% in toluene and THF, while its ΦPL only slightly increased in DMF through suppression of the proton transfer (Figure S8 and Table S5). A DF component could not be observed for -QD under any conditions, as expected based on the DFT calculations (Figure S9).

The TADF process was further investigated in TQB-doped thin films. Figure c shows the temperature dependence of the transient PL decay in a TQB-doped bis[2-(diphenylphosphino)phenyl]ether oxide (DPEPO) matrix (see Table S6 and Figures S10–S11 for more detailed characterization). At low temperatures (<150 K), delayed components could not be observed with our measurement system, indicating a small radiative decay rate from the excited triplet state to the ground state. On the other hand, the DF increased with a rise in temperature in agreement with typical TADF characteristics. The same dependence was also observed in a neat film of TQB though with a much weaker delayed component (Figure S12). To exclude the possibility of the DF originating from TTA, the dependence of the DF on the intensity of excitation light from a laser was recorded (Figure S13).[48] An ideal slope of 1.00 in a log–log scale confirms that the delayed emission of TQB is due to E-type DF and not P-type DF.

The prompt (Φp) and DF (Φd) components of the ΦPL at 300 K were estimated to be 46% and 9.5%, respectively, from the decay curve. The relatively small contribution of TADF indicates a slow RISC process. Since phosphorescence could not be observed in either the doped or neat film using our experimental setup, the T1 energy and ΔEST could not be easily determined. Instead, the activation energy of the DF, ΔEaTADF, was estimated from an Arrhenius plot of kRISC vs 1/T based on the relationship kRISC = exp(−ΔEaTADF/kBT), where kRISC is the rate of reverse intersystem crossing, kB is Boltzann’s constant, and T is the temperature (Figure S14). The kRISC can be experimentally estimated from the rate constants of intersystem crossing (kISC), prompt emission (kp), and delayed emission (kd) and Φp and Φd (see Table S6) according to the equation kRISC = kpkdΦd/kISCΦp. The kRISC at 300 K was 3.5 × 103 s–1, and ΔEaTADF was calculated to be 195 meV, which is a relatively large value for a TADF molecule but smaller than that of fullerene C70.[15]

The HOMO and LUMO energies of TQB as estimated from electrochemical measurements are −6.10 and −2.58 eV, respectively (Figure S15). Similar values were also found using photoelectron yield spectroscopy (PYS) (Figure S16) for the HOMO (−5.99 eV) and adding the optical energy gap from the UV spectrum for the LUMO (−2.88 eV). From the emission spectrum, the S1 energy of TQB-TB is estimated to be 2.64 eV, so the T1 energy is expected to be slightly below that since most TADF materials exhibit a ΔEST of less than 0.2 eV with only a few exceptions.[2,49] Although the corresponding energies of TQB-TA are difficult to determine because the emission spectrum could not be measured, they would lie at higher energies. Since the ESIPT materials have a four-level photocycle scheme, host materials must be carefully chosen to ensure confinement of the triplet excitons, especially for OLEDs. Therefore, TQB-doped films (10 wt %) in various host materials were evaluated. However, the number of triplets created in TQB-TA are expected to be much lower under photoexcitation, which initially generates only singlets, than electrical excitation, which generates 75% triplets, because the ESIPT should be much faster than ISC. Thus, the triplet state of TQB-TA may not be as important in the case of photoexcitation.

The T1 of the host materials investigated here is as follows (also depicted in Figure S17 along with the chemical structures): 2.98 eV for DPEPO;[50] 2.96 eV for 2,8-bis(diphenylphosphoryl)dibenzo[b,d]thiophene (PPT);[51] 2.84 eV for 3,3′-bis(carbazol-9-yl)biphenyl (mCBP);[41,52] 2.66 eV for 4,4′-bis(carbazol-9-yl)biphenyl (CBP);[52] 3.02 eV for 9-(4-tert-butylphenyl)-3,6-bis(triphenylsilyl)-9H-carbazole (CzSi).[53] The PL transient decay spectra (Figure d) and ΦPL (Table S4) show a clear dependence on the T1 energy of the host material with DF decreasing as the host T1 energy decreases. The host CBP, which can be expected to have a slightly higher triplet energy than TQB based on an S1 energy of 2.64 eV in TQB-TB, quenches a considerable fraction of the triplet excitons of TQB, although DF can still be observed. The T1 energy difference between mCBP and TQB is >0.2 eV and might be expected to be large enough to confine the triplets, but a small amount of exciton loss still exists compared to the other hosts with higher triplet energies (ΔT1 > 0.3 eV), presumably because RISC is very slow in TQB. Thus, DPEPO, PPT, and CzSi are the best hosts for confining the triplet excitons.

Triplet Harvesting in OLED Devices

On the basis of these results, OLEDs were fabricated with a configuration of ITO/1-bis((di-4-tolylamino)phenyl)cyclohexane (TAPC) (30 nm)/tris(4-carbazoyl-9-ylphenyl)amine (TCTA) (20 nm)/CzSi (10 nm)/TQB (10 wt %):host (30 nm)/PPT (40 nm)/LiF (0.8 nm)/Al (100 nm), where the host was DPEPO (device A), PPT (device B), mCBP (device C), CBP (device D), or CzSi (device E). An electron blocking layer (EBL) of CzSi and an electron transporting layer (ETL) of PPT were selected to avoid exciton quenching at the interfaces of the TQB-doped emissive layer (see Figure S17 for all of the chemical structures).

Figure presents an energy level diagram, external EL quantum efficiency versus current-density (ηext–J) plots, and current-density versus voltage (J–V) characteristics for devices A–E. The EL spectra of the devices were similar to the corresponding PL spectra of the doped films, indicating emission solely from TQB-TB through the ESIPT process under electrical excitation (Figure S18). Device A with DPEPO as the host showed relatively poor J–V characteristics, which is attributed to the poor charge carrier transport capabilities and deep HOMO and shallow LUMO energy levels of DPEPO.[54] As a result, the maximum ηext was only 7.8%, although this already exceeds the theoretical limit for conventional fluorescent OLEDs. Meanwhile, devices B and E, with PPT and CzSi as host, respectively, exhibited high performance with maximum ηext of 14% and maximum current efficiency (ηc) of 46 cd A–1 at J = 0.01 mA cm–2 (Table S7). Such high ηext cannot be reached by only TTA, confirming the contribution of TADF. The relatively large roll-off of ηext at high current densities is attributed to the formation of excess triplet excitons because of a slow RISC in TQB, resulting in exciton quenching by triplet–triplet and/or single−triplet annihilation.[55]

Figure 5

Characteristics of OLEDs using TQB as the emitter. (a) Energy level diagram of the devices. Energy levels are in units of eV, and values in parentheses indicate layer thicknesses in nm. (b) External electroluminescence quantum efficiency as a function of current density for OLEDs with different host materials. (c) Current-density–voltage (J–V) characteristics of the OLED devices.

Under electrical excitation, triplet excitons are likely to form directly on TQB-TA by charge carrier recombination,[56] which is different from the situation for photoexcitation. Since the triplet level of TQB-TA should be higher than that of TQB-TB based on the results of the DFT calculations (Table S3), triplet excitons might be expected to be largely quenched by the host in an OLED, especially for devices C and D with lower triplet hosts. Nevertheless, these devices had ηext higher than would be expected for their ΦPL if only the 25% of excitons formed as singlets contributed to emission, confirming the presence of TADF. This implies an ultrafast ESIPT process from T1 of TQB-TA to T1 of TQB-TB and negligible back energy transfer of triplets to the host material regardless of the energy difference.

Another possibility is that intramolecular proton transfer occurs in the radical cation/anion state, followed by charge recombination and exciton formation on the TQB-TB structure. Indeed, the calculated energies of the radical cation and anion of TQB-TB were smaller than those of TQB-TA (Figure S19 and Table S8), although the difference is small (ΔE = ∼50 meV). In either case, the triplet level of the structure before proton transfer seems to have little influence on the OLED. Most importantly, we could demonstrate the successful harvesting of triplet excitons in an ESIPT-based TADF material.

Orientation of Transition Dipole Moment and Outcoupling

Using the ΦPL of ∼55% for the TQB-doped films and assuming an exciton formation efficiency of 100%, the light out-coupling efficiency (ηout) is expected to be greater than 25% to achieve ηext of 14%. This relatively high ηout is ascribed to a horizontal molecular orientation. To verify the impact of dipole orientation on ηout, the angular dependence of the PL was measured.[57] Optical mode analysis of the radiation patterns in Figure S20 yields the following orientation order parameters (S) for TQB doped in each host matrix: −0.11 (DPEPO), −0.12 (PPT), −0.15 (mCBP), −0.16 (CBP), and −0.18 (CzSi). Considering that S is defined as a parameter ranging from −0.5 to 1, where −0.5, 0, and 1 respectively indicate horizontal, random, and vertical orientation relative to the plane of the film surface, these S values suggest that TQB molecules are preferentially oriented to be lying flat on the film surface. The differences among the films were small but significant. We attribute the variations in S to a difference in kinetic or thermodynamic effects of the host molecules during the film fabrication.

Given that only long, linear-shaped guest molecules have displayed high orientational order with S < −0.4 thus far, such linear-shaped molecules are thought to have the most potential for achieving horizontal orientation.[58] However, we demonstrated here that horizontal orientation is feasible even in small, planar guest molecules, whereas the similarly sized and linear-shaped PXZ-TRZ exhibited random orientation in our previous work.[59] Considering the dipole orientation, the position of the dipole, and the refractive indices and thicknesses of each layer, ηout values for each device were simulated via an optical mode analysis. Using the simulated ηout, the ηext were estimated to be 12.4%, 10.3%, and 14.3% for the PPT, mCBP, and CzSi host matrices corresponding to the devices B, C, and E, respectively, from the relationship ηext = ηout × ηint. We used ΦPL as a substitute for ηint and assumed an emitter at the center of light-emitting layer for the simulations. These ηext are in good agreement with the experimentally obtained ηext (within 1.5 percentage points). On the other hand, the error for the DPEPO and CBP host matrices was much larger (over 5 percentage points), suggesting that improved assumptions and models are needed to reproduce ηint by accounting for poor charge carrier balance and/or large triplet quenching.

Conclusions

We demonstrated the first example of a high-performance OLED using an ESIPT-based TADF material. The ESIPT process paves the way for HOMO and LUMO separation for TADF without using a D–A system and facilitates the development of new classes of TADF materials such as fully fused, planar and rigid ring systems with high chemical stability. The ΔEaTADF of TQB (200 meV) was small enough for TADF to occur, and OLEDs exhibited ηext as high as 14%, indicating successful triplet harvesting even with the ESIPT process. This work introduces not only a new design strategy for building TADF molecules, but also a new perspective on ESIPT materials with applications in solid-state emitters.

Methods

General

Commercially available materials were used as received from the suppliers for the synthesis and purified by sublimation for the device fabrication (Table S9). Details of instruments and physical measurements are given in Table S10.

Materials Synthesis, Characterization, and X-ray Structure Analysis

The synthesis procedures and characterization data of materials used in this study are described in the Supporting Information (Methods 1–2), and NMR and IR spectra are given in Data S1–S2.

Quantum Calculations

Details of the calculations are described in the Supporting Information (Methods 3), and the coordinates for optimized geometries are given in Tables S11–S32.

Preparation of Organic Films for Photophysical and Electrochemical Measurements

Organic thin-films for optical measurements were fabricated on clean quartz and silicon substrates by thermal evaporation at a pressure lower than 5 × 10–4 Pa. The substrates were cleaned with acetone and isopropanol and then treated with UV/ozone to remove adsorbed organic species before deposition. Thin films for cyclic voltammetry measurements were fabricated on 100 nm-thick layers of tin-doped indium oxide (ITO) on glass substrates by thermal evaporation at a pressure lower than 7 × 10–4 Pa.

OLED Fabrication and Performance Characterization

Glass substrates with a prepatterned, 100 nm thick, 100 Ω sq–1 ITO coating were used as anodes. Substrates were washed by sequential ultrasonication in neutral detergent, distilled water, acetone, and isopropanol and then exposed to UV/ozone. After the substrates were precleaned, effective device areas of 4 mm2 were defined on the patterned-ITO substrates by a polyimide insulation layer using a conventional photolithography technique. Organic layers were formed by thermal evaporation at a pressure lower than 1 × 10–4 Pa. The current-density–voltage–luminance characteristics of the OLEDs were evaluated using a source meter (Keysight B2911A, Keysight Technologies) and a luminance meter (CS-2000, Konica Minolta, Japan) at a constant DC current at room temperature. The reproducibility of the device performance of the presented devices was confirmed by measuring at least four different samples.

Angular Dependent PL Measurements

Details of the measurements are described in the Supporting Information (Methods S4).