Highly efficient electroluminescence from purely organic donor–acceptor systems

Thermally activated delayed fluorescence and its application in organic light-emitting diodes

Gilbert N. Lewis, one of the most influential chemists in the 20^th century, first measured thermally activated delayed fluorescence (TADF) [1]. Before Lewis, most dyes were known to exhibit two phosphorescent bands associated with different photophysical processes. The band lying at higher energy was called the alpha emission band, while the other was called the beta emission band. Alpha and beta emissions are now called TADF and phosphorescence, respectively. Jabłoński proposed that alpha emission involved thermal activation from a phosphorescent state to a normal (fluorescent) state, while beta emission was spontaneous emission from a phosphorescent state. His interpretation is supported by several experimental observations: the alpha emission band is indistinguishable from the fluorescence band of a material, and lies at higher energy than the beta emission band; the rate of alpha emission increases with increasing temperature, while that of beta emission is much less dependent on temperature; and only beta emission is observed at low temperatures. Lewis showed that the alpha emission of fluorescein in boric acid obeyed the Arrhenius equation, and experimentally determined the thermal activation energy to be 8 ± 1 kcal mol^–1 [1]. This was the beginning of experimental study on TADF. Since the 1960s, TADF has been observed from various compounds such as eosin [2], benzophenone [3], aromatic thiones [4], and fullerenes [5, 6], suggesting that TADF emitters have diverse chemical structures.

In 2009, an attempt was made to use TADF emitters in organic light-emitting diodes (OLEDs) [7]. OLEDs have attracted considerable attention because of their potential application as flexible flat-panel displays and next-generation solid-state lighting sources [8–10]. In an OLED, excitons are generated by charge recombination, and singlet and triplet excitons are generated in a ratio of 1:3. Therefore, singlet and triplet excitons account for 25% and 75% of the total excitons, respectively. As a result, the electroluminescence (EL) efficiency of an OLED depends largely on the triplet-to-light conversion efficiency. TADF emitters can convert triplet states into light via thermal activation from triplet to singlet states, so they hold promise as potential emitters and singlet-exciton harvesters for OLEDs. In 2011, TADF-based OLEDs were shown to be able to outperform conventional fluorescent OLEDs [11]. Since this discovery, we have markedly improved the luminescence efficiency of TADF emitters and external quantum efficiency (η_EQE) of TADF-based OLEDs. In 2012, we developed the first purely organic TADF emitter that realized highly efficient OLEDs [12] with η_EQE of 19%, comparable to those of phosphorescent OLEDs. TADF is now a promising alternative to fluorescence (used in first-generation devices) and phosphorescence (used in second-generation devices), and is considered to be the third-generation technique to obtain highly efficient OLEDs [13]. To date, TADF emitters with wide ranges of both emission colors and chemical structures have been designed and synthesized [7, 11, 12, 14–55].

Recently, the use of TADF has greatly expanded, with developments including OLEDs with promising operational stability [56], white OLEDs for application as a solid-state lighting source [31, 57], assist dopants in highly efficient fluorescent OLEDs [58, 59], controlling emitter orientation to enhance light-outcoupling efficiency [60–62], solution-processed OLEDs [63, 64], a roll-to-roll process for low-cost OLEDs [65], highly efficient electrogenerated chemiluminescence [66], and time-resolved fluorescence cell imaging [67]. The recent progress in the application of TADF suggests that a paradigm shift from OLED dopants that exhibit normal fluorescence and phosphorescence to those that show TADF is currently occurring.

Photoluminescence and electroluminescence processes involving TADF

Figure 1 shows photoluminescence (PL) and EL processes involving TADF. We assume here that TADF involves the lowest excited singlet state (S₁), the lowest triplet state (T₁), and the ground state (S₀). TADF can be viewed as involving S₁ ← T₁ reverse intersystem crossing (RISC) and S₁ → S₀ radiative decay (Fig. 1c,f). In the PL process, S₁ excitons are generated after the photoexcitation of S₀ (Fig. 1a). Of these S₁ excitons, Φ_p is converted into light (S₁ → S₀ radiative decay), and 1 – Φ_p is converted to T₁ via S₁ → T₁ intersystem crossing (Fig. 1b). Of the T₁ excitons, Φ_d is finally converted into light as TADF (Fig. 1c). The ratio Φ_d/(1 – Φ_p) is the T₁-to-light conversion efficiency. PL quantum yield (PLQY, Φ_PL) is given by Φ_PL = Φ_p + Φ_d, where Φ_p and Φ_d are prompt and delayed components of Φ_PL, respectively, and can be determined from transient PL decay measurements [15]. For fluorescent compounds that do not emit TADF, Φ_d = 0.

Fig. 1:

Photoluminescence and electroluminescence processes involving TADF. S₁ is the lowest excited singlet state, T₁ is the lowest triplet state, and S₀ is the ground state. Φ_p and Φ_d are prompt and delayed components of PLQY (Φ_PL), respectively.

In the EL process, S₁ and T₁ are generated in a ratio of 1:3, so S₁ and T₁ account for 25% and 75% of the total excitons, respectively (Fig. 1d). Next, 0.25 Φ_p decays radiatively, and 0.25(1 – Φ_p) is converted to T₁, resulting in T₁ of

(Fig. 1e). Of the T₁ excitons,

is finally extracted as TADF. When Φ_d = 0, only 0.25 Φ_p can be extracted as light. This situation is true for conventional fluorescent OLEDs.

η _EQE of a TADF-based OLED depends on charge recombination factor (γ) and light outcoupling efficiency (η_out) as well as Φ_p and Φ_d. γ is between 0 and 1, while η_out is considered to be 0.2–0.3. η_EQE can be written as

The corresponding internal quantum efficiency (IQE) is expressed as η_EQE/η_out, so it depends on γ, Φ_p, and Φ_d:

When perfect charge balance is achieved (γ = 1) and nonradiative decay is completely suppressed (Φ_d = 1 – Φ_p), the IQE is 100% and η_EQE reaches 20–30%. In fact, η_EQE of 20–30% has been realized by optimizing device structure [61, 68–70]. For conventional fluorescent OLEDs, because Φ_d = 0, IQE and η_EQE are limited to 25% and 5–7.5%, respectively, even if Φ_p of nearly 1 is achieved using a highly efficient fluorescent emitter. Thus, TADF emitters can substantially improve IQE and η_EQE by effectively converting triplet excitons into light. Eqs. 3 and 4 are useful to evaluate the potential of TADF emitters in OLEDs. Once Φ_p and Φ_d of a TADF emitter are determined experimentally, by setting γ = 1 and using eqs. 3 and 4, we can calculate the maximum IQE and range of maximum η_EQE of TADF-based OLEDs. The contributions from prompt fluorescence (η_p) and TADF (η_d) to η_EQE can be calculated from

η _p can be interpreted as the EQE for conventional fluorescent OLEDs. For example, for a green TADF emitter with Φ_p = 0.34 and Φ_d = 0.34, the maximum IQE and η_EQE are calculated to be 56% and 11–17%, respectively [37]. η_p and η_d are calculated to be 2–3% and 9–14%, respectively. η_d is larger than η_p, suggesting that TADF contributes considerably to η_EQE.

Molecular design of TADF emitters

As stated above, TADF involves S₁ ← T₁ RISC and S₁ → S₀ radiative decay. Therefore, the efficiency of TADF depends largely on the rates of S₁ ← T₁ RISC (k_RISC) and S₁ → S₀ radiative decay (k_r). Increasing k_RISC and k_r is an effective approach to improve the luminescence efficiency of TADF emitters. Here, for simplicity, S₁ and T₁ are assumed to be dominated by one-electron excitation from the highest occupied molecular orbital (HOMO) to the lowest unoccupied molecular orbital (LUMO). In such cases, ΔE_ST is approximately expressed as 2K, where K is the exchange integral between two electrons occupying the HOMO and LUMO. K is small when the HOMO and LUMO are spatially separated. Because k_RISC increases as the energy difference between S₁ and T₁ (ΔE_ST) decreases [5], compounds with well-separated HOMO and LUMO are expected to exhibit fast k_RISC. In contrast, compounds with strongly overlapping HOMO and LUMO are expected to exhibit fast k_r [40, 41]. This is because k_r increases with increasing transition dipole moment between S₀ and S₁ (μ₁₀), and μ₁₀ increases with increasing HOMO-LUMO spatial overlap; more specifically, the overlap density between the HOMO and LUMO [40, 41]. Thus, there is a trade-off between fast k_RISC and fast k_r. However, by carefully controlling HOMO and LUMO distributions and the resultant spatial overlap between them, moderate k_RISC and k_r are compatible, so we can obtain highly efficient TADF emitters.

Most TADF emitters contain electron-donating and -accepting units. Intramolecular donor–acceptor systems have been reported to efficiently convert electricity into light and used in electrochemiluminescence systems [71–73]. In a donor–acceptor system, the HOMO is predominantly distributed on an electron-donating unit, while the LUMO is predominantly distributed on an electron-accepting unit. Therefore, a donor–acceptor system has spatially separated HOMO and LUMO, and consequently, a small ΔE_ST. When suitable electron-donating and -accepting units are combined, ΔE_ST is sufficiently small to facilitate efficient TADF. Simple and effective approaches to further separate the spatial overlap between HOMO and LUMO are to induce steric hindrance between electron-donating and -accepting units and increase the twist angle between them [11, 12, 21]. Both of these approaches can effectively decrease the HOMO-LUMO spatial overlap. Examples of electron-donating units that can induce large twist angles include phenoxazine [18–20, 30, 31, 34, 36, 38, 44], phenothiazine [35], phenazine [30, 37], and 9,9-dimethylacridane [30, 34]. Donor–acceptor systems bridged by a spiro structure also possess good HOMO-LUMO separation and small ΔE_ST [14, 15, 24, 33]. Recently, TADF processes involving RISC from higher excited triplet states to excited singlet states have been suggested [74–76]. This finding opens a new direction for molecular design of TADF emitters.

Sky-blue TADF emitter 9-(4-(benzo[d]thiazol-2-yl)phenyl)-N³,N³,N⁶,N⁶-tetraphenyl-9H-carbazole-3,6-diamine (DAC-BTZ)

Here, we report a novel sky-blue TADF emitter, 9-(4-(benzo[d]thiazol-2-yl)phenyl)-N³,N³,N⁶,N⁶-tetraphenyl-9H-carbazole-3,6-diamine (DAC-BTZ, Fig. 2a). The synthetic route to DAC-BTZ is outlined in scheme 1. DAC-BTZ is a purely organic donor–acceptor-type molecule, in which N³,N³,N⁶,N⁶-tetraphenyl-9H-carbazole-3,6-diamine (DAC-II) serves as an electron-donating unit and 2-phenylbenzo[d]thiazole (BTZ) acts as an electron-accepting unit. BTZ has good electron-accepting ability and appears in other TADF emitters [36, 37]. DAC-II has great potential as an electron-donating unit to realize highly efficient TADF emitters because it improves the compatibility between fast k_RISC and k_r, and enhances TADF [41]. By combining DAC-II and BTZ, we obtained the efficient sky-blue TADF emitter DAC-BTZ.

Fig. 2:

(a) Chemical structure, (b) optimized S₀ geometry, (c) HOMO, and (d) LUMO of DAC-BTZ. The S₀ geometry, HOMO, and LUMO were calculated at the M06-2X/cc-pVDZ level of theory.

Figure 2b–d show structures of DAC-BTZ obtained from density functional theory calculations. All calculations were performed using the Gaussian 09 program package [77]. The S₀ geometry of DAC-BTZ was optimized at the M06-2X/cc-pVDZ level of theory [78, 79]. BTZ is almost coplanar, while the outmost diphenylamino groups of DAC-II have a propeller-like structure (Fig. 2b). The calculated twist angle between DAC-II and BTZ is 50.6°, which is small compared with those of twisted donor–acceptor-type TADF emitters [14, 15, 18–20, 24, 30, 31, 33–38, 44]. Excited states of DAC-BTZ were computed by time-dependent density functional theory [80, 81]. S₁ of DAC-BTZ mainly consists of HOMO-LUMO excitation and has charge-transfer (CT) character. The HOMO and LUMO of DAC-BTZ are predominantly distributed on DAC-II and BTZ, respectively (Fig. 2c,d), and are spatially well separated. Consequently, ΔE_ST is sufficiently small to facilitate TADF. The outer N atoms of DAC-II both decrease ΔE_ST and increase μ₁₀, making fast k_RISC and k_r compatible. By shifting the HOMO in the opposite direction to BTZ as indicated by green arrows in Fig. 2c, DAC-II decreases the HOMO-LUMO spatial overlap particularly on the central phenyl ring, leading to small ΔE_ST. At the same time, the overlap density between the HOMO and LUMO extends over the entire molecule, leading to a large μ₁₀. The effects of DAC-II on the photophysical properties of TADF emitters have been discussed extensively elsewhere [41].

Synthesis and characterization of DAC-BTZ

9-(4-(Benzothiazol-2-yl)phenyl)-9H-carbazole

To a solution of 3.00 g of 2-(4-bromophenyl)-benzothiazole (10.3 mmol), 2.07 g of 9H-carbozole (12.4 mmol), 1.43 g of sodium tert-butoxide (14.9 mmol) and 0.19 g of tris(dibenzylideneacetone)dipalladium(0) (0.21 mmol) in 50 mL of toluene was added, with stirring, a solution of 2.0 M tri-tert-butylphosphine in 0.50 mL of toluene. The stirred mixture was heated under reflux for 7 h, and then 10 mL of water and 30 mL of chloroform were added to the cooled mixture. The resulting mixture was stirred for 30 min and then filtered through a pad of Celite. The filtrate was extracted with chloroform. The combined organic layers were washed with brine, dried over Mg₂SO₄, and concentrated in vacuo. Column chromatography of the residue solid (eluent: toluene/ethyl acetate = 5/1) afforded 3.45 g of the target compound (yield: 88%). ¹H NMR (CDCl₃, 500 MHz) δ = 7.32 (t, 2H, J = 7.50 Hz), 7.44 (t, 1H, J = 7.50 Hz), 7.45 (t, 2H, J = 7.75 Hz), 7.51 (d, 2H, J = 8.50 Hz), 7.54 (t, 1H, J = 7.00 Hz), 7.73 (d, 2H, J = 8.50 Hz), 7.95 (d, 1H, J = 8.00 Hz), 8.12 (d, 1H, J = 8.00 Hz), 8.15 (d, 2H, J = 8.00 Hz), 8.34 (d, 2H, J = 8.50 Hz). Matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) m/z Calcd for C₂₅H₁₆N₂S: 376; found: 376.

Scheme 1:

Synthesis of DAC-BTZ.

Experimental methods

NMR spectra were recorded on a Bruker Biospin Avance-III 500 NMR spectrometer at ambient temperature. All NMR spectra were referenced to residual solvent peaks. Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF-MS) data were recorded on a Bruker Daltonics Autoflex III spectrometer in positive mode. UV-vis spectra were obtained using a UV-vis spectrophotometer (UV-2550, Shimadzu, Japan). PL spectra were recorded using a spectrofluorometer (Fluoromax-4, Horiba Jobin Yvon). PLQYs were measured using an absolute PLQY spectrometer (C11347 Quantaurus-QY, Hamamatsu, Japan). Transient PL decays of solution were obtained using a fluorescence lifetime spectrometer (C11367 Quantaurus-tau, Hamamatsu Photonics, Japan). Transient PL decays of doped films were measured using a streak camera (C4334, Hamamatsu Photonics, Japan). A N₂ gas laser with a wavelength of 337 nm (Ken-X, Usho Optical Systems, Japan) was used as the excitation source. The HOMO energy level of DAC-BTZ was determined from photoelectron spectrometry (Riken Keiki, AC-3) measurements of a neat DAC-BTZ film fabricated by vacuum deposition. The LUMO energy level of DAC-BTZ was determined from its HOMO energy level and the UV-vis absorption spectrum of its neat film.

TADF behavior of DAC-BTZ in toluene solution

Figure 3a shows UV-vis absorption and PL spectra measured for a solution of DAC-BTZ (10^–5 M) in toluene at room temperature. The peak maxima of the first absorption and emission bands are at 383 and 496 nm, respectively. The broad absorption and emission bands reflect the CT character of S₁ of DAC-BTZ. In air, the PLQY of DAC-BTZ in toluene solution was 0.45. This value increased to 0.65 after bubbling N₂ gas through the solution for 10 min. The increase in PLQY probably originates from TADF. In air, TADF is suppressed because triplet states of DAC-BTZ are quenched by oxygen dissolved in the solution, so the PLQY is low. Bubbling N₂ gas through the solution removes oxygen from it, allowing triplet states to survive. As a result, TADF is observed from DAC-BTZ, causing the increase in PLQY. Figure 3b shows the transient PL decays for DAC-BTZ in toluene solution. After N₂ gas bubbling, a long-lived component was observed in the transient PL decay curve, reflecting the generation of TADF. This oxygen-dependent quenching of TADF provides a quick and convenient method to screen the potential of TADF emitters. As stated below, the long-lived component caused by TADF is more clearly observed when DAC-BTZ is doped into a solid-state host layer than in solution.

Fig. 3:

(a) UV-vis absorption and PL spectra of DAC-BTZ in toluene solution (10^–5 M). (b) Transient photoluminescence decays for DAC-BTZ in toluene solution measured in air and after bubbling N₂ gas through the solution.

TADF behavior of DAC-BTZ in a solid-state host layer

To evaluate DAC-BTZ as a potential emitter for OLEDs, we investigated its TADF behavior in a layer of bis(2-(diphenylphosphino)phenyl)ether oxide (DPEPO) [82]. DPEPO has high T₁ energy and has been widely used as a host material for blue TADF emitters [16, 17, 20, 21, 23, 24, 29–31, 43]. A 6 wt% DAC-BTZ-doped DPEPO thin film was fabricated by vacuum deposition. Φ_PL of the doped film was 0.82. Figure 4a shows transient PL decays of the film measured in the time range from 0 to 10 ms at temperatures of 100, 200, and 300 K. The decay curves consist of an intense peak and long tail. The intense peak corresponds to prompt fluorescence, while the long tail corresponds to delayed fluorescence. The delayed fluorescence originates from TADF because its intensity increases with increasing temperature. The decay curve at 300 K was fitted using a triexponential function, indicating superimposition of ^1*CT → S₀ and ^3*CT → S₀ emissions resulting from efficient spin-orbit coupling induced by the heavy sulfur atom. The decay times of the fitted function were 52 μs (prompt), 660 μs (delayed), and 4.6 ms (delayed). The respective amplitudes connected with the decay times were 1.900, 0.012, and 0.003. The decay time of 52 μs is long compared with conventional fluorescence lifetime. This is because when transient PL decay is measured in 0–10 milliseconds time range, time resolution is insufficient to measure decay time for prompt fluorescence. To achieve sufficiently high time resolution, PL decays should be measured in 0–100 ns as described below. From the fitted function, Φ_p and Φ_d were calculated to be 0.67 and 0.15, respectively. Therefore, the T₁-to-light conversion efficiency of DAC-BTZ is 45%. Figure 4b shows transient PL decays measured for the same film in the time range of 0–100 ns at temperatures of 100, 200, and 300 K. The decay curve at 300 K in Fig. 4b was fitted using a triexponential function with decay times of 5.5 ns, 660 μs, and 4.6 ms. Thus, the lifetime of the prompt fluorescence was determined to be 5.5 ns.

Fig. 4:

Transient photoluminescence decays for a 6 wt% DAC-BTZ-doped DPEPO thin film measured at temperatures of 100, 200, and 300 K for time ranges of (a) 0–10 ms and (b) 0–100 ns. (c) Photoluminescence spectrum of the doped film measured at 300 K. (d) Prompt fluorescence and phosphorescence spectra of the doped film measured at 9 K.

Figure 4c shows a PL spectrum of the doped film. The broad emission band is attributed to fluorescence from DAC-BTZ. The peak position of the emission band is consistent with that measured for the toluene solution (Fig. 3a). ΔE_ST of TADF emitters have been elucidated by various methods including the energy difference between the onsets of fluorescence and phosphorescence spectra [17, 18, 20, 23–25, 29–31, 34, 37, 38, 41, 43, 44], the difference between peak wavelengths of fluorescence and phosphorescence spectra [11, 40, 42, 58], the temperature dependence of k_RISC [7, 12, 21, 36], and Berberan-Santos plots [14, 15, 32]. The method based on shift between fluorescence and phosphorescence spectra is valid when the energies of the Franck-Condon states reached in ^1*CT → S₀ and ^3*CT → S₀ emissions are same. Otherwise, it underestimates ΔE_ST. Figure 4d shows prompt fluorescence and phosphorescence spectra of the doped film measured at 9 K. ΔE_ST of DAC-BTZ was estimated to be 0.18–0.22 eV from the energy difference between the onsets of these spectra.

Figure 5 illustrates the PL process of the doped film and proposed EL process of a DAC-BTZ-based OLED based on the above transient PL decay measurements. For the PL process, of S₁ excitons generated by photoexcitation, 67% decay radiatively, and the remaining 33% are converted to T₁ excitons (Fig. 5a,b). Because 45 % of the T₁ excitons are converted into light as TADF (Fig. 5c), 82% of the total excitons are extracted as light. For the EL process, out of the 25% of S₁ excitons generated by charge recombination, 67% (17% of the total excitons) decay radiatively, and the remaining 33% (8% of the total excitons) are converted to T₁ excitons. Out of the 83% (75% + 8% of the total excitons) of T₁ excitons, 45% are converted into light as TADF. As a result, 54% of the total excitons are extracted as light. From eq. 4, the maximum IQE (γ = 1) for the doped film is calculated to be 54%, and η_EQE of a DAC-BTZ-based OLED should be 11–16%. The calculated η_EQE is higher than those obtained using conventional fluorescent OLEDs (5–7.5%), suggesting that DAC-BTZ is promising as an EL material.

Fig. 5:

(a) Photoluminescence process for the 6 wt% DAC-BTZ:DPEPO thin film. (b) Electroluminescence process for a DAC-BTZ-based OLED.

Electroluminescence from a DAC-BTZ-based OLED

An OLED containing DAC-BTZ as a sky-blue emitter was fabricated by vacuum deposition. The structure of the OLED was indium tin oxide (ITO)/4,4′-bis[N-(1-naphthyl)-N-phenyl]biphenyl diamine (α-NPD) (30 nm)/ 9,9′-biphenyl-3,3′-diyl-bis-9H-carbazole (m-CBP) (10 nm)/6 wt% DAC-BTZ:DPEPO (15 nm)/DPEPO (10 nm)/ 1,3,5-tris(N-phenylbenzimidizol-2-yl)benzene (TPBi) (35 nm)/LiF/Al (Fig. 6a). The HOMO and LUMO energy levels of DAC-BTZ were determined to be 5.4 and 2.7 eV, respectively, which are inside the HOMO-LUMO energy gap of DPEPO. Figure 6b shows the EQE-current density characteristics and EL spectrum of the OLED. The OLED exhibits sky-blue emission with a peak wavelength of 493 nm. The maximum EQE of the OLED was 10.3% at a current density of 3 × 10^–3 mA/cm². Although the maximum EQE is high compared with those of conventional fluorescent OLEDs (5–7.5%), it is lower than that predicted theoretically from Φ_p and Φ_d (11–16%). This suggests that there is room to further increase the maximum EQE, e.g. by improving carrier balance and controlling emitter orientation [60–62]. Assuming that γη_out = 0.19, eq. 3 gives an experimental maximum EQE of 10.3%. By using γη_out = 0.19 in eqs. 5 and 6, η_p and η_d are calculated to be 3.2% and 7.1%, respectively. Therefore, without TADF, the EQE of the OLED would be 3.2%. η_d is more than twice as large as η_p, suggesting that TADF makes a considerable contribution to the high EQE of this device.

Fig. 6:

(a) Energy levels (in eV) of the components in the DAC-BTZ-based OLED. (b) External quantum efficiency-current density characteristics of the OLED. The inset shows the electroluminescence spectrum of the device at a current density of 100 mA/cm².

Summary

We briefly reviewed the recent progress in TADF research and reported the novel sky-blue TADF emitter, DAC-BTZ. DAC-BTZ is a purely organic donor–acceptor system exhibiting high luminescence efficiency. When doped into the host DPEPO, DAC-BTZ showed a high PLQY of 0.82. Transient PL decay measurements confirmed that DAC-BTZ emitted TADF in the DPEPO host layer. An OLED containing DAC-BTZ as a sky-blue emitter exhibited a high EQE of 10.3%, outperforming conventional fluorescent OLEDs. This high EQE results from the efficient generation of TADF in DAC-BTZ.