Purely Organic Thermally Activated Delayed Fluorescence Materials for Organic Light‐Emitting Diodes

The design of thermally activated delayed fluorescence (TADF) materials both as emitters and as hosts is an exploding area of research. The replacement of phosphorescent metal complexes with inexpensive organic compounds in electroluminescent (EL) devices that demonstrate comparable performance metrics is paradigm shifting, as these new materials offer the possibility of developing low‐cost lighting and displays. Here, a comprehensive review of TADF materials is presented, with a focus on linking their optoelectronic behavior with the performance of the organic light‐emitting diode (OLED) and related EL devices. TADF emitters are cross‐compared within specific color ranges, with a focus on blue, green–yellow, orange–red, and white OLEDs. Organic small‐molecule, dendrimer, polymer, and exciplex emitters are all discussed within this review, as is their use as host materials. Correlations are provided between the structure of the TADF materials and their optoelectronic properties. The success of TADF materials has ushered in the next generation of OLEDs.


Introduction
Organic light-emitting diodes (OLEDs) have been the object of intense research since their initial invention in 1987 by Tang and Van Slyke [1] as they represent an unprecedented advancement in both display and lighting technologies. Compared with existing liquid-crystal displays (LCDs), OLEDs provide improved image quality and contrast, faster response times/refresh rates, are viewable over wider viewing angles, and are thinner and lighter. Even more impressive is the capacity to fabricate these devices on flexible substrates, to the point where OLED displays can be rolled up like a poster, a characteristic unfathomable in older-generation displays. OLEDs are also more energy-efficient because they do not require a backlighting system. Given that lighting alone constitutes around 20% of the global electricity consumption, a significant amount of electricity can be saved if OLEDs are widely adopted as a lighting technology. [2] spin statistics. [6] Whilst the former contributes to light emission by fluorescence (k F , Figure 1) with short emission lifetimes (in the nanosecond regime), phosphorescence (k P , Figure 1) from the latter with emission lifetimes extending to milliseconds occurs due to the spin-forbidden nature of the emission. The very long emission lifetime makes the triplet excitons vulnerable to nonradiative deactivation as heat loss to the surroundings. Assuming Lambertian emission and a light outcoupling efficiency of 20%, the maximum external quantum efficiency (EQE) for an OLED with a fluorescent emitter is only 25% of 20% = 5%. [7] As a response to this shortcoming in device efficiency, Baldo et al. in 1998 reported an OLED device in which a red-emitting organometallic complex, 2,3,7,8,12,13,17,18-octaethyl-21H,23H-porphyrine platinum(II) (PtOEP), was doped into a fluorescent host. [6] Through efficient energy transfer from the host to the emitter, both singlets and triplets were harvested for light emission with reported external and internal quantum efficiencies (IQE) of 4% and 23%, respectively. This seminal contribution can be regarded as the dawn of phosphorescence-based OLEDs. The value of employing organometallic complexes as emitters is their capacity to access triplet states via intersystem crossing (ISC) from the singlet excited state through strong spin-orbit coupling mediated by the heavy metal (e.g., Ir and Pt) in the complexes. [2,6,8] The phosphorescence emission decays in these complexes within a useful microsecond regime. Since 2001, devices based on organometallic complexes with nearly 100% IQE have been reported. [9] Commercial OLED devices for displays presently rely on green-and red-emitting cyclometalated iridium complexes. [10] However, the rarity of the heavy-metal salt reagents is a major detracting feature and contributes to increased cost of the device; potential environmental contamination of these heavy metals also is an element of concern. Moreover, while organometallic complexes exhibit impressive performance metrics in red and green-emitting devices, the corresponding blue-emitting complexes are thus far not satisfactory in terms of their combined stability, their color purity and their brightness during device operation. [11] Possible reasons for their poor performance include triplet _ polaron annihilation (TPA), which results in the formation of: i) highly energetic polarons that cause device degradation, [11] and ii) unstable radical cations on the complex, which cause ligand dissociation or complex isomerization. [12] Bright, deep blue and stable emitters for OLEDs are urgently needed, and this is representative of the grand challenges in emitter design at present.
In response to this need, thermally activated delayed fluorescence (TADF) is the most promising exciton harvesting mechanism used in OLED devices. Since the first reported OLED based on an organic TADF emitter in 2011 [13] tremendous attention in recent years has been devoted to improving their performance (Figure 2). [14] Similar to phosphorescent organometallic emitters, purely organic TADF emitters can recruit both singlet and triplet excitions for light emission and hence achieve 100% IQE. [14b] One important advantage of TADF emitters is that they can be purely organic, thus avoiding the problems associated with the use of heavy-metal-based organometallic complexes. [15] TADF relies on a small singlet-triplet energy gap, ΔE ST , defined as the gap between the lowest energy triplet state (T 1 ) and the lowest energy singlet state (S 1 ). When ΔE ST is sufficiently small, taken usually as <0.1 eV, thermal upconversion from the triplet state to the singlet state by reverse intersystem-crossing (RISC) becomes possible. [15a] TADF emitters typically show two types of photoluminescence (PL): the prompt fluorescence in which the history of the singlet exciton does not involve communication with the triplet manifold, and delayed fluorescence, which is the result of an initial ISC to the triplet state followed by repopulation of the singlet state via RISC. [15a] In an organic TADF material, ΔE ST is critical to the success of singlet harvesting because it governs the rate of RISC, k RISC , according to the following Boltzmann distribution: [14a] where k RISC is the rate constant of RISC, k B is Boltzmann's constant and T is the temperature. A consequence of Equation (1) is that a large ΔE ST results in slow k RISC and a correspondingly longer delayed fluorescence (k DF = 1/τ d ). [16] The delayed fluorescence lifetime, τ d , has been found to decrease with either an increasing rate of ISC (k ISC ) [17] or of RISC (k RISC ). [16] In particular, τ d can be expressed mathematically as: [18] where k r s (k r s = k PF , Figure 1) and k nr s are the radiative and nonradiative decay rate constants of the S 1 state, respectively. In Cu(I) TADF compounds where k ISC , k RISC , and k P are relatively fast as a result of the increased spin-orbital coupling due to the heavy atom effect of Cu metal (ζ: 857 cm −1 ), which enhances the intersystem-crossing rate, an equilibrium between excited singlet and triplet states ensues, which is governed by a Boltzmann distribution, and the emission decay time τ is given by: [19] E k T E k T = 3 + exp In organic compounds, formally spin-forbidden processes are much slower. The key difference between Equation (2) and Equation (3) is the presence of "τ(T 1 )" such that in Cu(I) complexes phosphorescence from T 1 becomes kinetically competitive with both nonradiative quenching, k T nr , of the triplet state, and k RISC , whereas in purely organic materials both of these processes are retarded.
Importantly, ΔE ST is correlated with the structure of the emitter, as it is proportional to the exchange integral, J: [20] 2 ST S T J in turn depends on the electron density overlap between the HOMO and LUMO, assuming that the S 1 and T 1 states are dominated by HOMO to LUMO transitions: where φ HOMO and φ LUMO are the spatial distributions of the HOMO and the LUMO, respectively, and r 1 and r 2 are position vectors (for a more detailed treatise see ref. [21]). It follows that reducing the overlap between the HOMO and the LUMO decreases the exchange integral (J) and hence ΔE ST . The normally spin-forbidden ISC and RISC processes in purely organic TADF emitters can therefore become efficient thanks to a small ΔE ST and can be explained by Equation (6): [15a] H E SO ST λ ∝ ∆ (6) where λ and H SO are the first-order mixing coefficient between singlet and triplet states and the spin-orbital interaction, respectively. Thus, despite the small H SO in purely organic materials, ISC and RISC processes can still occur readily if the ΔE ST is sufficiently small. The nature of the S 1 and T 1 states is also an important factor when determining the efficiency of the RISC process. El-Sayed's rule [22] broadly states that for ISC (and RISC) to efficiently occur there must be a change in the symmetry of the excited state. Thus, k ISC and k RISC will be fast if, for instance, T 1 were a locally excited (LE) state while S 1 were a charge-transfer (CT) state. However, there are cases when RISC can be efficient without El-Sayed's rule being rigorously followed. Indeed, excited states involved in the TADF process are not typically either pure CT or LE states but mixed CT-LE with possibly different fractions of CT character. Therefore, El-Sayed's rule would not strictly apply. [23] If the ΔE ST is very small, hyperfine-coupling (HFC) comes into play and El-Sayed rule does not have to be obeyed in this case. [24] Some organic TADF emitters have elements that are relatively heavier than typical C, H, N elements, and these elements assist ISC or RISC by the heavy-atom effect. A typical example is sulfur, which is commonly used in deep-blue or sky-blue TADF emitters. [15c] Finally, regardless of the nature of the lowest triplet state, it cannot be ruled out that there are higher triplet levels that are themselves lower in energy than the lowest singlet excited state. This permits a route governed by both reverse internal conversion (RIC) and RISC to harvest triplet excitons.
Indeed, TADF, also known as E-type delayed fluorescence in the early literature, [25] is a photophysical mechanism that was first reported in 1961, when eosin was observed to emit delayed fluorescence in ethanol at 70 °C ( Figure 2). Other examples of organic molecules that have been shown to emit via TADF include benzophenone, [26] aromatic thiones, [27] thioketones [28] and 9,10-anthraquinone. [29] Although the vast majority of effort and attention is devoted to purely organic TADF emitters, the very first TADF emitters applied in OLED devices stemmed from more traditional organometallic complexes and, interestingly, the TADF-emitter development history showed a gradual transition from heavy metals (e.g., Ir and Pt) to lighter elements (e.g., C and N). [30] In 2008, Yersin and Monkowius filed a patent in which multinuclear complexes based on iridium, palladium, platinum, rhodium, and gold characterized by a small singlet-triplet energy gap (500-2000 cm −1 ) between the lowest triplet state and the first higher lying singlet state were applied in OLEDs for singlet harvesting. [31] In 2009, Adachi et al. reported the use of a Sn 4+ porphyrin as a TADF emitter in OLEDs, albeit in devices with very low efficiencies. [32] The following year, Deaton et al. reported the first example of a highly emissive TADF bis(phosphine)diarylamido dinuclear copper(I) complex, [24] work that has inspired much additional research into this class of emitters by the likes of Yersin, [33] Thompson, [33c,34] Bräse, [35] and others. [36] In 2011, the first purely organic TADF emitter, PIC-TRZ, was reported by Adachi et al. [13] The EQE of the OLED was 5.3%, which is still close to the theoretical limit of conventional fluorescent emitters [13] and thus not incontrovertible proof that the mechanism of emission in the device was TADF. Finally in 2012, a series of groundbreaking organic TADF emitters (CDCBs) were reported by the same group. [15a] These emitters are based on donor-bridgeacceptor architectures in which the donor carbazoles are in a highly twisted conformation relative to the phthalonitrile plane, resulting in a reduced overlap between the HOMO and the LUMO and a correspondingly small singlet-triplet energy gap, ΔE ST . The best performing OLED in this study achieved an astonishing EQE of 19.3%, which clearly surpasses the expected theoretical maximum of 5% for electroluminescent (EL) devices employing ordinary fluorescent emitters. Such a high efficiency was the definitive demonstration that both singlet and triplet excitons were being harvested within the device. Since this report, organic TADF emitters for use in EL devices have become a red-hot topic of academic and industrial research, evidenced by more than 400 reports in this field to date (Figure 3).
This review aims to provide a comprehensive summary of the development of organic TADF emitters, together with their monochromatic device performance and their use in white OLEDs. We shall then discuss host materials designed for TADF emitters, which themselves can be employed as

TADF Emitters
The first purely organic TADF emitter PIC-TRZ (Figure 4) was reported in 2011 by Adachi et al. [13] Limited by its moderate PLQY of 39% in doped thin film (6 wt% in 1,3-bis(Ncarbazolyl)benzene (mCP) and only 32% triplet utilization efficiency, the device (ITO/α-NPD/mCP/6 wt% emitter:mCP/ BP4mPy/LiF/Al) (ITO = indium tin oxide; α-NPD = N,N′di(1-naphthyl)-N,N′-diphenyl-(1,1′-biphenyl)-4,4′-diamine; BP4mPy = 3,3′,5,5′-tetra[(m-pyridyl)-phen-3-yl]biphenyl) shows an EQE of 5.3%. With the explosion of interest in the development of TADF emitters, there have been more than 100 new compounds reported over the past 3.5 years. As will be demonstrated below, the performance of OLEDs employing TADF emitters across the entire visible spectrum is comparable in terms of efficiency with those using organometallic phosphors. Device EQE can reach beyond 30% using TADF emitters. [37] Numerous other TADF devices show an EQE of greater than 20%. There have been several recently published reviews where there has been either a partial focus on TADF compounds or the reviews have been written to take a broad overview of the subject, including a discussion of organometallic TADF complexes. [14,19,38] Given the paradigm-shifting importance of TADF emitters in EL devices, now is the time to provide a comprehensive review of the field, organized herein as a function of EL emission color.

Blue TADF Emitters
In this section, we define blue emitters as those compounds whose electroluminescence peak wavelength (EL max ) is shorter than 500 nm. This definition is used as EL max data are always provided, whereas Commission Internationale De L'Éclairage (CIE) coordinates are only provided in about 50% of reports. We recognize that not all emitters in this section will produce blue light, but our stated threshold serves to aid in the categorization of the emitters within this review, one that we have successfully used in the past for phosphorescent iridium emitters. [39] The distribution of CIE coordinates for the OLEDs reported in this section is shown in Figure 5 (vide infra). Table 1 summarizes the photophysical properties of emitters in this section while Table 2 summarizes the OLED device performance metrics. As blue phosphorescent complexes are widely recognized for their instability in EL devices, [11] great hope exists that blue TADF devices can address this weakness in OLED technology. In 2012, Adachi et al. reported the very first class of deep-blue TADF emitters (1-3) (Figure 6) based on diphenylsulfone as the acceptor. [15c] The best device within the study (ITO/α-NPD/TCTA/CzSi/10 wt% emitter:DPEPO/ DPEPO/TPBI/LiF/Al) (TCTA = tris(4-carbazoyl-9-ylphenyl) amine; DPEPO = bis[2-(diphenylphosphino)phenyl] ether oxide; TPBI = 2,2′,2″-(1,3,5-benzinetriyl)-tris(1-phenyl-1-Hbenzimidazole)) shows an EQE of 9.9% and CIE coordinates of (0.15, 0.07) with 3 (λ max : 423 nm; PLQY: 80%; τ d : 540 µs, 2600 µs in 10 wt% DPEPO; ΔE ST : 0.32 eV) used as the emitter. The authors suggested that in order to achieve a small ΔE ST , the energy gap between the lower energy triplet donor-centered locally excited π-π* state ( 3 LE) and the higher-energy triplet charge-transfer state ( 3 CT) must be small. This hypothesis was verified by the appearance of delayed fluorescence when the medium of the emitter was changed from nonpolar hexane to polar methanol, where the 3 CT is significantly stabilized, evidenced by the positive solvatochromism. Under these conditions reverse internal conversion (RIC) occurs from 3 LE to 3 CT, followed by efficient RISC to 1 CT. Indeed, emitters 1 (λ max : 421 nm; PLQY: 60%; τ d : 850 µs in 10 wt% DPEPO; ΔE ST : 0.54 eV) and 2 (λ max : 430 nm; PLQY: 66%; τ d : 840 µs, 8200 µs in 10 wt% DPEPO; ΔE ST : 0.45 eV), while having a similar energy 1 CT state compared with 3 (λ max : 423 nm), each possess a diphenylamine 3 LE state that is lower in energy than that with carbazole, thereby translating into compounds with larger ΔE ST of 0.54 eV and 0.45 eV, respectively. Nevertheless, it is particularly challenging to design deep-blue TADF emitters that adhere to this ordering of excited states because of the high intrinsic energies of charge-transfer singlet and triplet states. Control of the conjugation length (e.g., through sterics [40] or substitution pattern [41] ) and the choice of donor are important in this regard. However, Dias et al.
proposed another plausible mechanism of RISC in this type of molecule. [42] They performed a detailed photophysical study on a series of emitters and found that, for emitters with ΔE ST greater than 0.3 eV, RISC is still possible and may even be very efficient. For example, the molecule DTC-DBT (cf. Figure 7) possesses a high ΔE ST of 0.35 eV, but 100% of the triplet excitons are harvested by the RISC process. The presence of heteroatom lone pairs form an important "hidden" 3 n-π* state sandwiched between the higher 3 CT and the lower 3 LE state. Thus, up-conversion happens in an even more complex cascade manner: 3 LE → 3 nπ* → 3 CT → 1 CT; however, the energy gap between the higher 3 CT and the 3 n-π* depicted in their work is around 0.38 eV (Figure 8), which we believe is probably too high for an efficient up-conversion.
Indeed, Chen et al. pointed out that, for the molecule DTC-DBT, the lowest 3 LE state already has some mixing with the 3 n-π* state. [43] Rather, they stressed the importance of non-adiabatic effects in butterfly-shaped donoracceptor-donor (D-A-D) molecules such as DTC-DBT. When the donor and acceptor groups rotate such that they are all mutually orthogonal, a conical intersection (CI) between the low-lying excited states (i.e., between S 1 and S 2 as well as between T 1 and T 2 ) occurs. At the CI point, the non-adiabatic coupling (NAC) matrix element between two excited states becomes infinite, which is proportional to the RISC rate. It is important to note that this situation does not hold for antisymmetric donor-acceptor (D-A) mole cules. A recent study by Etherington et al. [44] suggests that second order vibronic coupling between the 3 LE and 3 CT states facilitate the RISC process.
Among the original diphenylsulfone-based deep-blue TADF emitters (1-3, DMOC-DPS, and DMAC-DPS), DMAC-DPS is the most promising due its smallest ΔE ST (0.08 eV) and high solid state PLQY (80%) amongst this series of sulfone emitters. This suggests the use of acridan donors is preferable to using carbazole (3 and DMOC-DPS) or diphenylamine (1 and 2) donors. This can be attributed to the electron richness of the acridan donor, which promotes greater HOMO and LUMO separation, along with its highly rigid structure, which contributes to reducing nonradiative decay paths from the excited state. Compared with DMAC-DPS, the use of 9,9-dimethyl-9H-thioxanthene-10,10-dioxide as the acceptor (DMTDAc) results in an even higher solid-state PLQY (100%), probably due to a greater rigidity of the acceptor moiety, and a vanishing ΔE ST (0 eV). Invoking a dimerization strategy produces improved blue TADF emitters (compare DTC-mBPSB and DTC-pBPSB vs 3) because of their slightly redshifted emission energies (i.e., a lowerenergy S 1 state, which is the lowest-energy 1 45] A benzophenone-type acceptor can also be used in the design of blue TADF emitters. Lee et al. [49] fabricated two blue devices (ITO/α-NPD/mCP/6 wt% emitter:DPEPO/DPEPO/TPBi/LiF/ Al) using Cz2BP (λ max : 444 nm; PLQY: 55%; τ d : 710 µs in 6 wt% DPEPO; ΔE ST : 0.21 eV) and CC2BP (λ max : 475 nm; PLQY: 73%; τ d : 460 µs in 6 wt% DPEPO; ΔE ST : 0.14 eV) (Figure 9) as the emitters. The OLEDs obtained EQEs of 8.1% and 14.3% at CIE coordinates of (0.16, 0.14) and (0.17, 0.27), respectively. CC2BP, with an extended carbazole donor system, exhibits a smaller ΔE ST than Cz2BP by 0.07 eV. This also enhances the donor strength in CC2BP (HOMO: −5.65 eV) compared with Cz2BP (HOMO: −5.74 eV) and results in a significant redshifted emission (31 nm, 1470 cm −1 ). Rajamalli et al. [50] reported two novel TADF emitters based on a benzoylpyridine acceptor DCBPy (λ max : 514 nm; PLQY: 88%; τ d : 0.6 µs in 5 wt% 9,9′-(sulfonylbis(4,1-phenylene))bis(9H-carbazole) (CzPS); ΔE ST : 0.07 eV) and DTCBPy (λ max : 518 nm; PLQY: 91%; τ d : 1.0 µs in 5 wt% 4,4′-bis(N-carbazolyl)-1,1′-biphenyl (CBP); ΔE ST : 0.08 eV) ( Figure 9). The only difference between these two emitters is the presence of tert-butyl groups in the DTCBPy. Given their modest electron-donating nature, DTCBPy shows a small redshift in emission wavelength by only 4 nm (150 cm −1 ). The other photophysical properties seem to be unaffected. Intramolecular through-space interaction between the ortho-carbazole donor and the benzoylpyridine acceptor is believed to induce efficient TADF. The interaction also suppresses intermolecular aggregation in the solid state, and thus the PLQY in the solid state is greatly enhanced in doped film (up to 91.4%) compared with solution (14-36%). DCBPy (ITO/ NPB/mCP/5 wt% emitter:CzPS/DPEPO/TmPyPb/LiF/Al) and DTCBPy (ITO/NPB/TAPC/5 wt% emitter:CBP/PPT/TmPyPb/ LiF/Al) (TmPyPb = 1,3,5-tri(m-pyridin-3-ylphenyl)benzene; PPT = 2,8-bis(diphenylphosphoryl)dibenzo- [b,d]  the BPy series are from −2.87 eV to −2.88 eV. In addition, the latter family of emitters has smaller ΔE ST , most likely due to ortho-carbazole substitution that increases the torsion between the donor and acceptor moieties due to increased steric hindrance between carbazoles, resulting in greater localization of the HOMO and LUMO, a feature that is absent in the Cz2BP and CC2BP series. Apart from the popular diphenylsulfone moiety, the 1,3,5-triazine is one of the most common acceptors used for blue TADF emitters. A blue-greenish TADF emitter CzT (λ max : 502 nm; PLQY: 40%; τ d : 42.6 µs in 3 wt% DPEPO; ΔE ST : 0.09 eV) (Figure 10) was employed in an OLED device (ITO/α-NPD/ TCTA/CzSi/3 wt% emitter:DPEPO/DPEPO/TPBi/LiF/Al), which showed an EQE of 6% at CIE coordinates of (0.23, 0.40). [51] However, a structurally similar emitter PhCzTAZ (PhCzTAZ = 3-(2′-(4,6-diphenyl-1,3,5-triazin-2-yl)-[1,1′biphenyl]-2-yl)-9-phenyl-9H-carbazole) does not show TADF because of the absence of charge-transfer emission, probably due to limited HOMO and LUMO communication restricted by steric hindrance around the biphenyl bridge. To improve on the low PLQY of the CzT emitter, BCzT (λ max : 483 nm; PLQY: 96%; τ d : 33 µs in 6 wt% DPEPO; ΔE ST : 0.29-0.33 eV) was then developed, in which the overlap density between the excited state and the ground state (ρ 10 ) was increased. [52] The presence of an additional phenyl-ring bridge increases the overlap integral between the HOMO and the LUMO as its presence mediates increased conjugation. As a result, the ρ 10 in BCzT is more widely distributed than in CzT, and the transition dipole moment of the former is consequently larger than the latter, resulting in a higher radiative rate constant (k r ). It should be noted that the addition of the phenyl-ring bridge increases the ΔE ST , but triplet-to-light efficiency (defined as the ratio of PLQY contribution by delayed component (Φ d ) to triplet formation yield (Φ T )) in BCzT (76.2%) is much higher than CzT (25%). We believe this is due to a much higher k r in BCzT that decreases the cycling between singlet and triplet states and thus eliminates the probability of nonradiative decay in both singlet and triplet states. Therefore, for efficient TADF to occur, both high RISC rate (k RISC ) and radiative rate constants (k r ) are essential. The device with BCzT gives a sky-blue emission (ITO/α-NPD/ m-CBP/6 wt% emitter:DPEPO/TPBi/LiF/Al) with an EQE of 21.7% and EL max at 492 nm, which is far improved compared to the 6% EQE obtained by the OLED with CzT.  Figure 10) with the only difference being two additional carbazole moieties attached to the phenyl ring in a meta fashion in DDCzTrz. As "meta linkages" limit conjugation length, the two emitters have similar emission energies and ΔE ST. The OLED (ITO/PEDOT:PSS/TAPC/mCP/25 wt% emitter:DPEPO/TSPO1/TPBI/LiF/Al) using the latter has an impressive EQE of 18.9% at CIE coordinates of (0.16, 0.22). [53] In particular, this device shows an LT 80 (the time required for the luminance to drop to 80% of its initial value) of 52 h, which is approximately three times longer than the blue phosphorescent analog using tris[1-(2,4-diisopropyldibenzo [b,d] furan-3-yl)-2-phenyl-1H-imidazole]-iridium(III), [Ir(dbi) 3 ]. The authors attributed the stability to three main factors: firstly, the carbazole and triazine moieties are robust and the nearly planar structure of the molecule gives the peripheral carbazole moiety further stabilization through conjugation with the triazine; Secondly, the nature of charge transfer makes the excited state resemble a pair of positive carbazole and negative triazine polarons, which are known to be stable; lastly, the excellent thermal stability of the emitters (glass-transition temperature (T g ) for DCzTrz and DDCzTrz: 160 °C and 218 °C, respectively) contributes positively to the device stability. However, based on the above reasoning, one would expect the device stability to be similar for DCzTrz and DDC-zTrz. Indeed, while DDCzTrz shows an LT 80 of 52 h, DCzTrz has an LT 80 of only 5 h. The authors asserted that the higher emission energy of DCzTrz was responsible for the poorer stability. We believe that the poorer stability is due to the intrinsic structure of the emitters, which are designed using the well-known "meta-linkage" interconnection mode. This approach effectively limits the conjugation length of the whole molecule as the number of π-conjugated systems keeps increasing in order to avoid redshifts in the emission and lowering of the triplet energy level. [41,54] Therefore, the emission energy of the emitters is expected to be similar, which, experimentally, is the case where the PL and EL spectra of these two emitters are essentially the same. The authors later modified DCzTrz through addition of more carbazole donors to the emitter to generate three new compounds: TCzTrz (λ max : ≈450 nm; PLQY: 100%; τ d : 13.5 µs in 30 wt% DPEPO; ΔE ST : 0.16 eV), TmCzTrz (λ max : ≈470 nm; PLQY: 100%; τ d : 13.3 µs in 30 wt% DPEPO; ΔE ST : 0.07 eV) and DCzmCzTrz (λ max : ≈490 nm; PLQY: 98%; τ d : 9.7 µs in 30 wt% DPEPO; ΔE ST : 0.20 eV) ( Figure 10). [55] By comparing these emitters with the parent DCzTrz, the authors suggested that having more carbazole donors present in the emitter helps to reduce the ΔE ST (e.g., one additional carbazole in TCzTrz lowers the ΔE ST by 0.09 eV compared with DCzTrz), while uneven distribution of the electron density in the HOMO increases ΔE ST (e.g., the HOMO is localized on the dimethylcarbazole moieties in DCzmCzTrz due to the electron-donating nature of methyl groups, conferring DCzm-CzTrz with a larger ΔE ST than that of TCzTrz and TmCzTrz by 0.04 eV and 0.13 eV, respectively). Among these emitters, the OLED (ITO/PEDOT:PSS/TAPC/mCP/40 wt% emitter:DPEPO/ TSPO1/TPBI/LiF/Al) with TCzTrz is sky-blue with an impressive EQE of 25.0% at CIE coordinates of (0. 18 at ≈490 nm. [56] Acridan, being a stronger donor than carbazole, redshifts the DMAC-TRZ compared with carbazole-based triazine TADF emitters. Interestingly, the emitter also demonstrates an excellent EQE of 20% in the absence of a host. Indeed, the PLQYs of the emitter in doped thin film (90%) and neat film (83%) are very similar, which is attributed to the methyl groups on the acridan unit that serve via sterics to suppress intermolecular interactions, similar to that observed for DMAC-DPS ( Figure 6). [46] Sun et al. [57] reported a potent deep-blue TADF emitter DTPDDA (λ max : 444 nm; PLQY: 74%; τ d : 0.1, 2.3, 25.4 µs in 16 wt% mCP:TSPO1; ΔE ST : 0.14 eV) whose device (ITO/4 wt% ReO 3 :mCP/mCP/16 wt% emitter:mCP:TSPO1/ TSPO1/4 wt% Rb 2 CO 3 :TSPO1/Al) performance reaches an outstanding EQE of 22.3% with CIE coordinates of (0.15, 0.20). Silicon was chosen for its rigid tetrahedral configuration, which, due to its size, effectively suppresses nonradiative decay pathways and enhances the morphological stability of the molecule. Beneficially, the HOMO of the azasiline (−5.57 eV) is lowered compared with the carbon analog DMAC-TRZ (HOMO: −5.30 eV) because of the longer Si-C bond that limits the antibonding interactions between the two azasiline carbons bonded to the silicon atom. [58] The lowering of the HOMO energy level increases the bandgap and contributes to the deep-blue emission of the device. Aside from designing emitters that can attain 100% IQE, the efficiency of the device can be improved by increasing the light-outcoupling efficiency, which is generally around only 20% for typical emitters. Mayr et al. [59] designed a largely planar sky-blue TADF emitter CC2TA (λ max : ≈490 nm; PLQY: 62%; τ d : 22 µs in 6 wt% DPEPO; ΔE ST : 0.06 eV). It has previously been demonstrated that planar, long, linear molecules have preferential horizontal orientations on the substrate due to favourable intermolecular interactions with the host during film deposition. The horizontal placement of emitters in the film orients their transition dipole more optimally, thereby enhancing light outcoupling. [60] The outcoupling efficiency of the CC2TA emitting layer was found to be 31.3% and the device (ITO/α-NPD/mCP/6 wt% emitter:DPEPO/ DPEPO/TPBi/LiF/Al) achieved an EQE of 11% with EL max at 490 nm. Very recently, Lin et al. [61] reported a novel triazine-based blue TADF emitter, spiroAC-TRZ (λ max : 480 nm; PLQY: 100%; τ d : 2.1 µs in 12 wt% mCPCN; ΔE ST : 0.07 eV), which shows a strong horizontal dipole ratio, Θ // , (where Θ // is defined as the horizontal emitting dipole/total emitting dipoles) as high as 83%. In comparison with structurally similar DMAC-TRZ, the phenyl rings attached on the acridan moiety in spiroAC-TRZ weaken the electrondonating capacity of the donor inductively, thus producing a modest blueshift of 15 nm in the emission. The PLQY of spiroAC-TRZ (100%) was also higher than that of DMAC-TRZ (90%), which was attributed to the more planar conformation of the acridan moiety, as confirmed by single-crystal X-ray diffraction, which showed that the dihedral angles between the acridan phenyl planes were 0° and 11°, respectively. [61] Together with both Adv. Mater. 2017, 29,1605444 www.advancedsciencenews.com www.advmat.de  100% PLQY and IQE, a high-performance blue device (ITO/ MoO 3 /TAPC/mCP/12 wt% spiroAC-TRZ:mCPCN/3TPYMB/ LiF/Al) with an EQE of 37% and CIE coordinates of (0.18, 0.43) was obtained, making it the best blue TADF device reported so far, despite the large y-ordinate CIE value.
Park et al. [73] designed a series of three related compounds (DCN1-3) ( Figure 13) as potential TADF emitters for OLED applications. While TADF is observed in DCN3 (λ max : 482 nm; PLQY: 49%; τ d : 3.26 µs in toluene; ΔE ST : 0.13 eV), DCN1 and DCN2 do not demonstrate any TADF properties, which according to the authors, is based on their large ΔE ST values. Though a reasonable assertion, no experimental ΔE ST values are provided in the study. As the molecular scaffold of these molecules is basically the same, the increased spacing of the HOMO to the peripheral carbazoles in DCN3 with respect to the isophthalonitrile-localized LUMO should be responsible for the realization of TADF. This is in line with the theoretical studies from, independently, Adachi et al. [74] and Sancho-Gracíer et al. [21] The former group compared TADF emitters DACQ and CZQ ( Figure 13) and found that the installation of diphenylamino groups in DACQ promotes further localization of the HOMO density toward the periphery of the molecule, thereby further spatially separating the HOMO and LUMO, resulting in a smaller ΔE ST . The latter team performed a theoretical study comparing three non-TADF against three known TADF emitters and proposed an inverse relationship between Δr(NTO) and ΔE ST , where Δr(NTO) is a measure of the electron-hole separation after excitation (NTO = natural transition orbital) (Figure 14). The OLED (ITO/HAT-CN/ NPB/TAPC/10 wt% emitter:PPT/PPT/TPBi/ LiF/Al) (HAT-CN = 1,4,5,8,9,11-hexaazatriphenylenehexacarbonitrile) with DCN3 gave sky-blue emission with an EQE of 13.3% at CIE coordinates of (0.20, 0.37). Very recently, Cho et al. [75] reported two biphenyl-based blue TADF emitters CNBPCz (λ max : 458 nm; PLQY: 46%; , which was attributed to the "donor interlock" molecular design of CzBPCN that relates to a much larger rotational barrier of the biphenyl as a result of increased ortho-substitution. This resulted in a deep blue device with CIE coordinates of (0.14, 0.12) with CzBPCN as the emitter.
Cho et al. [76] reported two solution-processable blue TADF emitters 3CzFCN (λ max : ≈440 nm; PLQY: 74%; τ d : 28 µs in 10 wt% diphenyldi(4- The use of more rigid and bulkier donors in the 2CzPN scaffold results in a boost in the PLQY by nearly a factor of two in toluene (compare BTCz-2CN and BFCz-2CN with 2CzPN), which is probably due to the suppression of nonradiative decay modes conferred by these rigid structures. Additionally, the ΔE ST decreases significantly (0.13 eV and 0.17 eV for BFCz-2CN and BTCz-2CN, respectively, while it is 0.31 eV for 2CzPN in PhMe) [21] which results in more efficient RISC. Steric hindrance about the donor in these emitters is important, which is evidenced by the turn-off of TADF in DCN1 and DCN2 where, in these emitters, the carbazole donor has no adjacent groups to provide the steric bulk required to generate sufficient HOMO and LUMO separation. However, this analysis does not apply for 35IPNDCz and 26IPNDCz. A possible reason for this incoherence is due to their A-D-A symmetric molecular scaffold, which may promote more efficient TADF emitters, similar to that which is observed for D-A-D analogs. Steric hindrance about the donor is not necessary for TADF.  (Figure 15). [77] The much redshifted emission in PXZ-PXB is due to the phenoxazine donor (HOMO: −5.  Figure 15). [78] Similar to that observed between DMAC-PXB and PXZ-PXB, the phenoxazine analog PXZ-Mes 3 B demonstrated the most redshifted emission of the three emitters in the study due to its strong electrondonating power. 2DAC-Mes 3 B shows slightly redshifted emission by 10 nm (431 cm −1 ) compared with DAC-Mes 3 B due to the enhanced conjugation, which results from the additional diphenylamine in the former. Regardless of the donor, all three compounds are highly emissive in doped films and exhibit similarly small ΔE ST . Devices using these emitters give bluish-green emission with EQEs ranging from 14.0% to 22.8% and CIE coordinates from (0.17, 0.30) to (0.22, 0.55), respectively. Numata et al. [79] reported a series of blue TADF emitters (ACRPOB, SFD-PAPOB, SXDPAPOB, and TMCPOB, Figure 15) based on a 10H-phenoxaborin acceptor and either an acridan or carbazole donor. There is effective suppression of intermolecular interactions that contribute to quenching of their emission because of the large dihedral angle between the 10H-phenoxaborin and the acridan moieties. The best device using emitter ACRPOB (λ max : 475 nm; PLQY: 100%; τ d : 1.6 µs in toluene; ΔE ST : 0.06-0.12 eV) demonstrates an EQE of 21.7% (ITO/HAT-CN/α-NPD/ CCP/50 wt% emitter:PPF/PPF/TPBi/LiF/Al) (CCP = 9-phenyl-9H-3,9′-bicarbazole; PPF = 2,8-bis(diphenylphosphoryl)dibenzo [b,d]furan) at CIE coordinates of (0.14, 0.23).
In general, most of the above boronbased blue TADF emitters show small ΔE ST (<0.18 eV), probably because of the strong LUMO localization effect induced by the boron atom. Additionally, the majority of these emitters exhibit excellent PLQYs in the solid state (87-100%). These results suggest that boron-based TADF scaffold is a potent avenue for blue TADF emitters. The DABNA series is of particular interest, as the emission profiles are considerably sharper. This is due to the distinct strategy of localizing the HOMO and the LUMO as a function of the regiochemistry of the donor and acceptor.

Green-Yellow TADF Emitters
In this section, we define green-to-yellow emitters as those whose electroluminescence peak wavelength (EL max ) lies between 500 nm and 580 nm. The distribution of CIE coordinates for the OLEDs reported in this section is shown in Figure 19 (vide infra). Table 3 summarizes the photophysical properties of emitters in this section while Table 4 summarizes the OLED device performance metrics. A vast majority of green-to-yellow TADF emitters contain cyano-based acceptors. For example, in the pioneering report by Adachi et al. [15a] the emitters consisted of carbazole donors and a dicyanobenzene (phthalonitrile) acceptor. The molecular design is based on the presence of a twisted conformation of the donor carbazoles with respect to the phthalonitrile plane, which confers a well-separated HOMO and LUMO, and thus a small ΔE ST . Depending on the regiochemistry of the donor carbazoles and the acceptor nitriles about the central benzene ring, this family of molecules can emit across the visible spectrum. The best-performing device in the report used the green emitter 4CzIPN (λ max : ≈510 nm; PLQY: 82%; τ d : 3370 µs in 6 wt% CBP; ΔE ST : 0.08 eV) (Figure 20), with an EQE as high as 19.3% and EL max at ≈510 nm (ITO/α-NPD/5 wt% emitter:CBP/TPBi/LiF/Al). Since their initial report, 4CzIPN has become the representative green TADF emitter and has been frequently used in photo physical, [96] mechanistic, [24,97] host [98] and device-optimization studies. [70,99] Lee et al. [100] modified 4CzIPN by adding solubilizing tert-butyl groups in t4CzIPN (λ max : ≈520 nm; PLQY: 78%; τ d : 2.9 µs in 3 wt% SiCz) ( Figure 20), an emitter that is now much more easily solution processable and exhibits less aggregation in the doped thin film. The solution-processed device (ITO/PEDOT:PSS/PVK/1 wt% emitter:SiCz/TSPO1/LiF/Al) using t4CzIPN (EQE: 18.3%) shows a markedly enhanced performance compared with the parent 4CzIPN (EQE: 8.1%).
Kawasumi et al. [104] reported two TADF emitters TPA-QNX(CN) 2 (λ max : 487 nm; PLQY: 44%; τ d : 2.4 µs in cyclohexane; ΔE ST : 0.11 eV) and TPA-PRZ(CN) 2 (λ max : 475 nm; PLQY: 55%; τ d : 6.5 µs in cyclohexane; ΔE ST : 0.08 eV) (Figure 20) based on an unusual triptycene scaffold in which a small ΔE ST is realized by installing the diphenylamine donor and dicyanoquinoxaline or dicyanopyrazine acceptors on different arms of the propeller triptycene, which communicate with each other electronically by homoconjugation (i.e., a through-space interaction). Indeed, homoconjugation has already been employed in several previously reported spiro-based TADF emitters, although the term "homoconjugation" is not invoked explicitly. [82,102,103] Although the choice of dicyanoquinoxaline or dicyanopyrazine acceptor results in little impact on the photophysics of the emitters, it is interesting to note that their device performance differs greatly. OLEDs The cyano group on 4CzCNPy is important, without which there is no TADF observed in this molecule. The cyano group is also believed to be essential in boosting the PLQY compared with its analog 4CzPy whose n-π* state quenches the emission, whereas for 4CzCNPy, the emission is dominated by ICT states.
Cyano-based green TADF emitters can essentially be divided into three classes: i). a monomeric series with ortho steric hindrance to realize HOMO and LUMO separation (4CzIPN, Adv. Mater. 2017, 29, 1605444 www.advancedsciencenews.com www.advmat.de     Monomeric emitters are already very potent green TADF emitters with small ΔE ST (<0.1 eV) and high solid-state PLQYs (81.8-100%). In particular, 4CzIPN is frequently employed to evaluate new bespoke host materials and is arguably the most widely studied TADF emitter and the highest recorded EQE using this emitter is 31.2%. On the other hand, HOMO and LUMO separation can be elegantly achieved by homoconjugation, but those emitters in general suffer from lower PLQYs compared with monomeric and the dimeric emitters. This reflects that homoconjugation limits orbital overlap too substantially, such that the transition dipole moment is significantly weakened, negatively impacting emission efficiency. This is especially true for Spiro-CN, whose solid-state PLQY is only 26%, which seriously limits the device performance. The dimeric emitters show significantly higher ΔE ST (0.11-0.21 eV) than emitters from the other two classes. It is worth noting that PLQYs in toluene solution (61-87%) of these dimeric emitters are lower than the monomeric analogs (e.g., 93% for 4CzIPN), which actually runs contradictory to the design paradigm of these emitters (i.e., an increase in molar absorptivity to boost the PLQY).
Adv. Mater. 2017, 29, 1605444 www.advancedsciencenews.com www.advmat.de  from the better orbital overlap between the donor and acceptor moieties, and thus a stronger transition dipole moment that thereby enhances the k r . The device (ITO/α-NPD/6 wt% emitter:mCBP/TPBI/LiF/Al) with DPA-TRZ gives a green emission with an EQE of 13.8% and EL max at 548 nm. A close analog, DACT-II (λ max : ≈520 nm; PLQY: 100% in 10 wt% DPEPO; ΔE ST : 0.01 eV) (Figure 21), realizes a simultaneous 100% emission and RISC efficiency, thus giving a state-of-art green device (EL max ≈ 520 nm) (ITO/TAPC/9 wt% emitter:CBP/ BAlq/Liq/Al) (BAlq = bis(8-hydroxy-2-methylquinolinato)-(4-phenylphenoxy)aluminum; Liq = 8-quinolinolato lithium) with an impressive EQE of 29.6%. [37] The excellent photophysical properties of DACT-II are a consequence of the fine tuning of the dihedral angle (α) between the central carbazole and the bridging phenyl ring so that the oscillator strength (f) and the ΔE ST can be both optimized (f and ΔE ST both increase with increasing conjugation). The former is responsible for the large radiative decay rate (k r ), and hence boosts the PLQY, while the latter is linked to the efficiency of the RISC process. Therefore, while PXZ-TRZ, with a large dihedral angle of 74.9° between the donor and the bridge, shows a PLQY of 66%, the PLQY of DACT-II reaches unity. However, DAC-BTZ (λ max : 496 nm; PLQY: 82%; τ d : 52 µs in 6 wt% DPEPO; ΔE ST : 0.18-0.22 eV), a close analog of DACT-II in which the triazine acceptor is changed to benzothiazole, gives a blue device (EL max at 493 nm) with a much lower EQE of 10.3% (ITO/α-NPD/m-CBP/6 wt% emitter:DPEPO/DPEPO/TPBi/LiF/Al) due to its larger ΔE ST compared with that of DACT-II (0.009 eV). [94] This suggests that the acceptor choice is very important in the design of high-efficiency TADF emitters. Tanaka et al. [109] Adv. Mater. 2017, 29, 1605444 www.advancedsciencenews.com www.advmat.de  (Figure 21) using this scaffold in order to color-tune the emission. Despite the evident redshift in their emission energies compared with the parent PXZ-TRZ, bis-PXZ-TRZ and tri-PXZ-TRZ show almost identical emissions in both their PL and EL spectra. This observation can be attributed to the fact that the additional PXZ arms are metadisposed where the conjugation length is not strongly affected. Nevertheless, the device (ITO/α-NPD/6 wt% emitter:mCBP/ TPBi/LiF/Al) using tri-PXZ-TRZ shows a green emission with an EQE of 13.3% and EL max at 553 nm. A solution-processable green emitter, 3ACR-TRZ (λ max : 504 nm; PLQY: 98%; τ d : 6.7 µs in 16 wt% CBP; ΔE ST : 0.02 eV), was reported by Adachi et al. (Figure 20), [110] which shows a very efficient triplet utilization of 96% together with a PLQY close to unity in doped CBP. It is worth noting that although the dihedral angle between the acridan donor and the phenyl bridge is nearly 90°, the PLQY of the emitter remains exceptionally high. This result may seem to contradict the design paradigm illustrated in PXZ-TRZ and DACT-II, where the torsional angle between the donor and the phenyl bridge plays a crucial role in the balance between PLQY and ΔE ST . The solution-processed device (ITO/PEDOT:PSS/16 wt% emitter:CBP/BmPyPhB/Liq/Al) (BmPyPhB = 1,3-bis[3,5-di(pyridin-3-yl)phenyl]benzene) shows a very high EQE of 18.6% with a EL max at ≈520 nm.
Sagara et al. [117] prepared a series of green emitters (BT, BT2, BOX, cis-BOX2, and trans-BOX2) (Figure 23) based on phenoxazine donors and benzoxazole or benzothiazole accepters in order to contrast the properties and device performance between D-A and D-A-D molecular design. They found that D-A-D molecules have faster radiative rate constants (k r ), better PLQYs, and smaller ΔE ST . All emitters gave reasonably good device performance (EQEs of 9.1% to 16.6%) with the best OLED (ITO/α-NPD/6 wt% emitter:mCBP/TPBi/LiF/Al) using cis-BOX2 as the emitter (λ max : ≈510 nm; PLQY: 98% in 6 wt% mCBP; ΔE ST : 0.03 eV). The study concluded that D-A-D molecules clearly outperform their D-A analogs as emitters in OLEDs. Very recently, cis-BOX2 was found to exhibit a completely horizontal orientation when doped into a CBP matrix at 200 K. It was important to keep the deposition temperature below the glass-transition temperature of the CBP matrix; otherwise, the preferred horizontal orientation was no longer present. Thanks to the greatly enhanced light-outcoupling, the device (ITO/TAPC/15 wt% cis-BOX2:CBP/6 wt% cis-BOX2:CBP)/PPT/LiF/Al) gave an EQE of 33.4%. [118] Though not explicitly discussed by the authors, trans-BOX2 should possess a very similar molecular length, planarity, and shape compared with cis-BOX2, and thus it would be expected that trans-BOX2 should also give a preferential horizontal arrangement, as was observed for cis-BOX2.
The emitter HAP-3MF (λ max : 520 nm; PLQY: 26% in 6 wt% DPEPO; ΔE ST : 0.17 eV) (Figure 27), employing an interesting heptaazaphenalene, was reported by Adachi et al. [121] Among all the TADF emitters reported so far, this molecule is special for two reasons: Firstly, the lowest singlet (S 1 ) and triplet (T 1 ) involved in the RISC process are all highly n-π* excited states, which is highly unusual. Secondly, to the best of our knowledge, HAP-3MF is arguably the only TADF emitter for OLED applications that does not involve a nitrogen-based donor (the heptaazaphenalene here acts as the acceptor). However, because emission results from an n-π* excited state, k r is small (on the order of 10 −6 s −1 ), resulting in a relatively low PLQY of 26% in degassed toluene for the molecule. The OLED (ITO/α-NPD/ mCP/6 wt% emitter:DPEPO/DPEPO/TPBI/ LiF/Al) using this emitter gave a green emission with an EQE of 6.0% with EL max at ≈520 nm.
Wu et al. [122] prepared a series of three pyrimidine-based green TADF emitters ( Figure 28). As expected, having only two nitrogen atoms in the pyrimidine renders it a weaker acceptor than the threenitrogen-atom-containing triazine analog (bis-PXZ-TRZ). [109] The λ EL of these pyrimidine emitters are therefore blueshifted by ≈20 nm. The two classes of emitters share very similar ΔE ST (≈0.05 eV), while the triazine emitter has significantly lower PLQY (64%) than the pyrimidine emitters Xiang et al. [123] employed an asymmetric 1,2,4-triazine acceptor, a regioisomeric analog of the commonly used 1,3,5-triazine acceptor in the construction of TADF emitters. With phenoxazine as the donor, the authors fabricated yellow emitters TPXZ-as-TAZ (λ max : 555 nm; PLQY: 53%; τ d : 1.  Remarkably, the device underwent a small efficiency roll-off of 11.5% at high luminance of 1000 cd m −2 , thanks to a short decay lifetime of the delayed component (1.10 µs) that contributes to minimizing triplet quenching processes.
Wang et al. [128] reported the first near-infrared (NIR) TADF emitter, TPA-DCPP (λ max : 645 nm; PLQY: 50%; τ d : 86.2 µs in 10 wt% TPBI; ΔE ST : 0.13 eV) (Figure 33), which was based on a dicyanodiazatriphenylene acceptor moiety. The best device performance was achieved for a doped device (ITO/NPB/ TCTA/20 wt% emitter:TPBi/TPBi/LiF/Al), which exhibited an EQE of 9.8% with CIE coordinates of (0.68, 0.32). It is important to note that the DPA groups are not significantly twisted compared to the 2,3-dicyanopyrazinophenanthrene (DCPP) plane, with a dihedral angle between the phenyl and the DCPP of only 35°. Nevertheless, RISC is efficient in this molecule, most likely due to the extended distance between the donor and the acceptor. This is in line with the conclusions from Adachi et al. [74] about the requirements for TADF molecular design, where the twist angle between the donor and acceptor can be reduced if the distance between donor and acceptor units is increased. A smaller twist angle results in better orbital overlap, which is important for increasing the radiative rate constant, k r . Indeed, TPA-DCPP in degassed toluene shows an excellent PLQY of 84%.
Data et al. [125] reported a series of three TADF emitters based on dibenzo-  Figure 37) as the emitting layer, strongly exciplex-driven NIR emission at 741 nm was observed with an excellent EQE of about 5%. No corroborating evidence was provided to support the TADF mechanism of this exciplex system.

Application of TADF in Light-Emitting Electrochemical Cells
Light-emitting electrochemical cells (LEECs) are an alternative type of EL device, which possesses a much simpler device architecture compared to multilayered OLEDs. In LEECs, the emitter also serves as a charge transporter, facilitated by the presence of charged groups. These groups also facilitate charge injection from the electrodes, resulting in charge injection being insensitive to electrode work function, and thus an airstable cathode (e.g., Al) can be used. LEECs are fabricated via spin-coating techniques, and the device performance is relatively insensitive to the emitting layer thickness. [130] We recently reported the first charged organic TADF emitters (TL-1 and TL-2, Figure 38) for use in LEECs. [67] Emission in MeCN is uniquely fluorescence due to preferential stabilization of the triplet excited state with respect to the singlet excited state, which increases ΔE ST and turns off TADF. Photophysical studies in both neat and 10 wt%-doped PMMA film on the other hand confirm TADF, with dramatically enhanced PLQY that is O 2 -sensitive. The emitter TL-2 (λ max : 536 nm; PLQY: 21%; τ d : 2.73 µs in neat film) in the LEEC (ITO/PEDOT:PSS/emitter/ Al) gave an EQE of 0.39%, significantly lower than when the same emitter was used in an OLED device configuration (EQE of 5.1%, vide supra).

TADF Macromolecules for OLED Applications
Macromolecules have a profound place in OLED history, and they form an important class of emitters. [131] The chief merits for the use of macromolecules are the capability of solutionprocessing the materials and their supreme morphological stability (i.e., high T g ). Given these important benefits, it is surprising that there are so few reports of TADF macromolecules (e.g., dendrimers and polymers).
Nikolaenko et al. [126] reported the first TADF polymer TP-AEN (λ max : ≈540 nm; PLQY: 44% in neat film; ΔE ST : 0.22 eV) (Figure 40) for OLED applications. TP-AEN is made of 5 mol% triphenylamine donor, 50 mol% triazine acceptor, and 45 mol% backbone unit. Low doping of the donor helps to prevent its concentration quenching. Together with 50% high doping of the acceptor, the electron mobility of the polymer is preferentially enhanced to drive the recombination zone away from the cathode. The backbone unit, which consists of electronically insulating alkyl groups, limits an increase in conjugation length and thus avoids generating triplet traps in the polymer. The polymer was applied neat by solution-processing, and the device (ITO/PEDOT:PSS/interlayer/neat polymer/NaF/Al/Ag) achieved an EQE of 10.0% at CIE coordinates of (0.32, 0.58). However, it is worth noting that no physical characterization (e.g., NMR, IR, gel-permeation chromatography (GPC), etc.) was included in the report.

White Organic Light-Emitting Diodes using TADF Emitters
White OLEDs (WOLEDs) can be fabricated by stacking multicomponent emitters of red, green, and blue colors. One common way of achieving this is to employ a high-energy fluorophore used together with low-energy phosphor, the so-called hybrid WOLED. [135] A single emitter with a broad emission spectrum can also be utilized for WOLEDs. [136] Adv. Mater. 2017, 29, 1605444 www.advancedsciencenews.com www.advmat.de   Figure 13). When used as a lighting source, WOLEDs are more energyefficient than compact fluorescent lighting technologies, and for this reason, much research effort has been devoted in this area. [135b,137] The following section will review the contribution of TADF in the area of WOLED development.
Adachi et al. [138] prepared an all-TADF WOLED that achieved an EQE of 17.1% with CIE coordinates of (0.30, 0.38) operating at 3.6 V. The emitters 3CzTRZ, 4CzPN, and 4CzTPN-Ph were used as blue, green, and red TADF emitters, respectively, which form three adjacent stacked layers in the device. So far, this is the only all-TADF WOLED. However, it was discovered that each layer would be subjected to different degrees of exciton annihilation (e.g., singlet-triplet annihilation) because each TADF emitter has its own characteristic triplet exciton lifetime, mainly due to different k RISC values based on their intrinsic ΔE ST . As a result, a gradual EL spectral shift was observed upon increasing current density. More recently, the same group proposed the use of a blue TADF emitter (DMAC-DPS) stacked with an emitting layer of green and red conventional fluorescent emitters (TTPA and DBP, Figure 45) to fabricate a WOLED whose EQE reached 12.1% with CIE coordinates of (0.24, 0.31). The device unfortunately suffers from low operational stability (LT 50 < 1 h) because of the poor electrochemical stability of the DMAC-DPS emitter. [139] Zhang et al. [140] studied the quenching effect in mCP of the heavy-metal phosphors such as Ir(ppz) 3 , FIrpic, PO-01, and Pt(COD)Cl 2 , ranging from 0-25 wt% concentration on the emission of TADF fluorophores (2CzPN and t-Bu 4 CzTPN) in WOLEDs (Figure 46). It was found that, while the conventional fluorophore (bis(10hydroxybenzo[h] quinolinato)beryllium (Bebq 2 )) emission was strongly suppressed due to enhanced ISC induced by the external heavy-atom effect, TADF emission was barely affected due to its unchanged k ISC and enhanced k RISC . Given the intrinsically high triplet energy of blue TADF fluorophores, phosphor emission is likewise not quenched by the blue TADF fluorophore. Therefore, mutual quenching between fluorophore and phosphor in TADF-based WOLEDs is effectively blocked. The authors used 2CzPN and PO-01 as a blue TADF fluorophore and a yellow phosphor (Figure 46), respectively, and achieved a WOLED of maximum forward EQE of 19.6% with CIE coordinates of (0.42, 0.48), far from pure white of (0.33, 0.33). The authors also prepared a singleemitting-layer hybrid WOLED in which the blue-emitting DMAC-DPS was used as a host for the PO-01 phosphor dopant. [141] According to the authors, the rationale of this design was to minimize the short-range Dexter energy transfer from the host to the dopant, which requires triplet diffusion, a process associated with significant triplet loss. Instead, triplets on the TADF host can be upconverted to singlets via RISC followed by long-range Förster energy transfer to the dopant. Conventional fluorophore hosts are not able to fulfil this simply because they are unable to carry out RISC. The WOLED thus fabricated achieved a forward EQE of 20.8% with CIE coordinates of (0.40, 0.46). The efficiency roll-off of the device was small as well. Song and Lee [142] co-doped the same combination of emitters (DMAC-DPS and PO-01) into a DPEPO host, which formed the emitting layer in a WOLED device that achieved an EQE of 22.4% with CIE coordinates of (0.30, 0.37). The EL spectrum of the device was stable up to 5000 cd m −2 , which was  attributed to the energy transfer from DMAC-DPS to PO-01 being implicated in the dominant emission mechanism. The group also prepared another WOLED by replacing the PO-01 phosphor with a yellow TBRb fluorophore ( Figure 46). [143] The device showed an EQE of 15.5% with CIE coordinates of (0.28, 0.35). It was important to maintain TBRb at very low doping concentration (0.05%) in order to minimize both its chargetrapping effect and Dexter energy transfer from the DMAC-DPS to TBRb as triplet excitons on TBRb cannot be upconverted to singlets for light emission. Efficient Förster energy transfer between DMAC-DPS and TBRb is an important condition for the operation of this device. Song et al. [134] prepared a singleemitting-layer WOLED in which a blue TADF emitter CzAcSF was used as the host for conventional blue (TBPe) and yellow (TBRb) fluorophores ( Figure 46). Efficient energy transfer from the host to the dopant resulted in a WOLED with an EQE of 14.0% with CIE coordinates of (0.31, 0.37). Zhao et al. applied the blue TADF emitter DMAC-DPS as both emitter and host for the orange fluorophore rubrene to achieve a two-component WOLED ( Figure 46). [144] The device showed an EQE of 7.48% with CIE coordinates of (0.36, 0.44). Zhang et al. [145] prepared a WOLED by stacking blue (2CzPN:mCP) and yellow (PO-01:TAZ) emitting layers. Given that the exciton formation region lies at the interface between the two emitting layers, the TADF emitter 2CzPN solves the traditional problem that the low triplet energy of the blue fluorophore (large ΔE ST in conventional fluorophores) can quench the emission of the PO-01 phosphor. The device demonstrated a forward EQE of 22.6% with CIE coordinates of (0. 45, 0.48). Similarly, Meng et al. [146] fabricated a WOLED by stacking a yellow TADF emitting layer (TXO-TPA:mCP) [113] and blue fluorescent emitting layer (4P-NPB:mCP) to afford a WOLED with an EQE of 4.7% at (0.34, 0.34), very close to pure white emission. Cho et al. [87] reported an interesting device structure to achieve white light. They prepared a bespoke blue TADF emitter DCzIPN (Figure 20), which was used as both emitter and host for the yellow phosphor PO-01 in their study. The blue-emitting layer (DCzIPN:mCP) and yellowemitting layer (PO-01:DCzIPN) were stacked in the fashion of blue-yellow-blue, absent of any interlayer between them for whitelight generation. By controlling the thickness of the central yellow-emitting layer, warm and cool white-light devices could be fabricated. The former displayed an EQE of 22.9% with CIE coordinates of (0.39, 0.43) and the latter showed an EQE of 21.0% with CIE coordinates of (0.31, 0.33). Kim and Lee [147] fabricated a WOLED using a green TADF emitter (4CzIPN) together with blue FIrpic and red [Ir(pq) 2 (acac)] phosphors (Figure 47)   respectively. In another report by the same group, 4CzIPN and FIrpic constituted one emitting layer stacked on a [Ir(pq) 2 acac] layer. [148] The hybrid WOLED showed an EQE of 20.2% with CIE coordinates of (0.49, 0.41). TADF exciplexes can also play an important role in WOLED development, where the donors and acceptors used are shown in Figure 48 (see Section 6 for discussion of TADF exciplexes). In 2014, Hung et al. [149] reported the first all-exciplex TADF WOLED. With mCP:PO-T2T and DTAF:PO-T2T (PO-T2T = 2,4,6-tris[3-(diphenylphosphinyl)phenyl]-1,3,5-triazine; DTAF = 9,9-di[4-(di-p-tolyl)aminophenyl]fluorene) as blue-and yellowemitting layers, respectively, stacked against each other in the device, white light was achieved with an EQE of 11.6% with CIE coordinates of (0.29, 0.35) and a color rendering index (CRI) of 70.6, device metrics of which are nearly independent of EL intensity. Zhao et al. [150] fabricated a WOLED based on blue (TCTA:Bphen) and orange (TAPC:3P-T2T) TADF exciplexes (3P-T2T = 2,4,6-tris(3-(1H-pyrazol-1-yl)phenyl)-1,3,5-triazine). The device showed an EQE of 9.17% with CIE coordinates of (0.41, 0.44) with only negligible shifting of the EL spectrum upon increasing voltage from 6 V to 14 V, which was attributed to balanced charge injection and transport of the exciplex layers. Liu et al. [151] employed a TADF exciplex formed between CDBP and PO-T2T as both blue emitter and host for green [Ir(ppy) 2 (acac)] and red [Ir(MDQ) 2 (acac)] phosphors to achieve a WOLED operating at 2.5 V that exhibited a forward-viewing EQE of 25.5% with CIE coordinates of (0.41, 0.45). The exciplex host again was shown to be beneficial in terms of balanced charge injection and transport.

Development of Host Materials for TADF Devices
In addition to the emitters for light emission, the development of host materials is also of prime importance. [152] In general, emitters are doped in a host material in the device to avoid self-quenching. There are a number of requirements for an effective host material. Firstly, the host should have a high triplet energy level to avoid back energy transfer from the dopant to the host so that the triplet excitons can be confined on the dopant for light emission. Secondly, the host should be chemically and thermally robust. Host molecules must show a large degree of electrochemical reversibility. They must not degrade when they are vacuum deposited during device fabrication. They should have a high glass-transition temperature (T g ) so that they stay amorphous during device operation when Joule heat is produced. Thirdly, the energy levels of the frontier molecular orbitals of the host should be in close alignment with the adjacent layers to ease charge injection. Fourthly, the charge mobility of the host materials has to be high and balanced so that the recombination zone can be widely dispersed within the emitting layer. [153] Finally, a host material of enhanced solubility can be exploited for solution-processed devices. [154] Therefore, the design of host materials for highly efficient OLED devices is, in many ways, as challenging as the TADF emitters themselves. In this section, we summarize the most commonly used host materials and review the recent developments of bespoke host materials for TADF OLEDs.
Currently, in the vast majority of cases, the OLEDs fabricated with a TADF dopant utilize traditional host materials. Figure 49 shows the acronyms and structures of hosts material most commonly employed in the fabrication of TADF OLEDs while Table 7 summarizes their electronic properties. The preference for these host materials is certainly based on their low cost, availability, and well-documented physical properties, with many precedents of the fabrication of high-performance phosphorescent and fluorescent OLEDs.
For blue TADF OLEDs, one of the most commonly used host materials is DPEPO. [15c,45,48,49,86,159] DPEPO benefits from a high triplet level (E T : 2.995 eV), [156] which prevents undesirable energy transfer from the dopant to the host and, as asserted by Monkman and co-workers, [159] has high polarity, which contributes to a lowering of the energy of the 1 CT state so as to minimize the ΔE ST (ΔE ST in this compound is the gap between the 1 CT state and the lowest local 3 LE state). Representative examples of other hosts used for blue TADF OLEDs include mCP, [84,87] PPT, [15a,73] and SiCz. [76] For green TADF OLEDs, the most commonly used hosts are mCP [63,77,113] and mCBP. [107,117,119] CBP [15a,110] has also been used, suggesting back energy transfer from the dopant to the host is less prominent in green TADF devices, given that the triplet level of CBP (T 1 : 2.6 eV) is significantly lower than mCP (T 1 : 2.9 eV) and mCBP (T 1 : 2.8 eV). [160] DPEPO has also been used in some green TADF devices. [72,86] CBP [15b,127] seems to be the most Adv. Mater. 2017, 29, 1605444 www.advancedsciencenews.com www.advmat.de  commonly used host for red TADF devices, while TPBI [128] and mCBP [16a] have also been employed as host materials.
The electronic properties of bespoke TADF compounds used as host materials in OLEDs are summarized in Table 8. Kim et al. [98c] employed 2,7-bis(diphenyl phospho ryl)-9-phenyl-9Hcarbazole (PPO27) as a bipolar host for 4CzIPN, where the device achieved a maximum EQE of 24.2% (Figure 50). The TADF emitter interacted strongly with the PPO27 host due to the latter's strong polarity. As a result, at high doping concentrations, a significant redshift in the EL spectrum was observed. Due to the poor stability of the PPO27 host, the device lifetime was poor. Im et al. [98d] reported two bipolar hosts, 3TPAPFP and 4TPAPFP, which are based on a triphenylamine donor and a furodipyridine acceptor for balanced charge mobility. These functional groups have high triplet energy levels, which mitigates back energy transfer from the dopant to the host. However, 4TPAPFP showed more pronounced intermolecular interactions with the 4CzIPN dopant, which was attributed to the elongated shape of the host as a result of the para configuration of the ambipolar functional groups. [161] Energy transfer to 4CzIPN was poor as a result. Thus, while a maximum EQE of 21.2% was achieved using 3TPAPFP as the host, a much lower EQE of 6.6% was achieved in the case of 4TPAPFP as the host. The same group prepared a carbazole analog, 3CzPFP, as the host and achieved a significant enhancement in EQE up to 31.2% ( Figure 50). [162] The enhancement in performance is due to the absence of exciplex formation in the emitting layer 3CzPFP:4CzIPN, which is probably due to the deeper HOMO of the carbazole moiety in 3CzPFP compared with triphenylamine in 3TPAPFP. Gaj et al. [98e] employed an ambipolar host based on carbazole and diphenyl sulfone (mCPSOB) with 4CzIPN as the emitter, and achieved an excellent EQE of 26.5% (Figure 50). The T g of mCPSOB is 110 °C and the host forms a morphologically stable thin film to give high operational stability. Nishimoto et al. prepared an ambipolar PzCz host with a cyclophosphazene core decorated with six carbazole moieties ( Figure 50). [160b] PzCz has a high triplet energy of 3.00 eV and thanks to its rigid, planar core it demonstrates excellent chemical and thermal stability. The 5% weight-loss decomposition temperature (T d ) is as high as 474 °C. No signal is observed for the glass transition, crystallization, or melting points by differential scanning calorimetry (DSC) measurements, yet the authors claimed the host material to be intrinsically amorphous. Devices using emitter layers (PzCz:CzTPN, blue-green) and (PzCN:4CzIPN, green) gave EQEs of 15.0% and 18.2%, respectively. Cho et al. [163] reported a universal ambipolar host DCzDCN for both TADF and phosphorescent devices. Using 4CzIPN as the TADF dopant, they achieved a maximum EQE of 26.7% with much longer device lifetime (LT 90   that using CBP as the host (LT 90 : 10.2 h). The enhanced stability was attributed to improved charge confinement, morphological stability, low singlet energy, and little charge accumulation at the interface. Suzuki et al. [164] prepared a solution-processable host CPCB and used it in conjunction with 4CzIPN as the emitter. The device achieved a maximum EQE of 9.9%. CPCB solutionprocessed devices showed longer device half-lives (LT 50 up to 184 h) compared with that using CBP as the host (LT 50 : 56 h). This was attributed to the poor morphological stability of CBP, which crystallizes easily during device operation. Indeed, the glass-transition temperature (T g ) of CPCB (165 °C) is much higher than that of CBP (62 °C), [165] justifying the longer device lifetime. Cui et al. [98a] reported the first pure hydrocarbon (PHC) hosts, SF33 and SF34, for TADF devices (Figure 50). PHC hosts are believed to have higher chemical stability than heteroatom-based analogs. By tuning the interconnection pattern, a greater twisting of the two spirofluorenes is obtained in SF34, resulting in greater ambipolar charge transport.  Figure 50). It was found that DBTTP1 conferred higher device stability (LT 80 : 250 h) than DBTTP2 (LT 80 : 100 h), and the authors attributed this to higher chemical stability of the host in which the planar triphenylene moiety has a great conjugation compared with the terphenyl group in the DBTTP2 host. Li et al. [166] reported a highly twisted spirocyclic phosphine oxide host, SFXSPO (Figure 50), which prevents intermolecular interaction between host molecules and allows uniform dispersion of TADF dopants. Notably, SFXSPO could be used as a host for blue to red full-color TADF dopants, including a white device made of blue and yellow TADF dopants. The EQEs obtained using this host ranged from 13.9% to 22.5%. Notably, using 4CzPNPh [15a] as the yellow TADF dopant, a record high EQE of 22.5% was obtained, much higher than the 9.9% EQE obtained using mCP as the host. Cui et al. [167] reported two benzimidazobenzothiazole-based hosts, 29Cz-BID-BT and 39Cz-BID-BT, (Figure 50). These hosts possessed high triplet energies (>3.0 eV) and favor ambipolar charge transport. Using DMAC-TPZ as a blue TADF dopant (10 wt%), the two hosts gave practically the same device EQEs (20.2-20.4%) (ITO/ HAT-CN/TAPC/10 wt% DMAC-TRZ: host/TSPO1/TPBi/LiF/Al). Lee et al. [98b] prepared a series of three novel ambipolar hosts (oCzB-2CN, mCzB-2CN, and pCzB-2CN) based on carbazole and isophthalonitrile in order to probe the relationship between the donor-group regiochemistry and the device performance (Figure 51). Importantly, the triplet energies (E T ) are strongly affected by the regiochemistry. The E T of ortho, meta, and para isomers were found to be 2.99 eV, 2.73 eV, and 2.58 eV, respectively, which is due to: i) the steric hindrance between the carbazole and isophthalonitrile moieties for the o-isomer, and ii) the relative degree of conjugation in the hosts. The maximal current density was found in the device based on the para analog due to its better intermolecular orbital overlap with the dopant. However, the highest EQE of 26.0% in the study was obtained using mCzB-2CN because the PLQY of the 4CzIPN:mCzP-2CN film is the highest. A similar study was carried out by Kim et al. [168] who studied the effect of interconnection pattern on the device performance by comparing two hosts, 3CN34BCz and 4CN34BCz. [169] The latter, in which the cyano group is installed para to the carbazole, is more capable of accommodating electron density than the meta analog. Thus, the charge transport is more balanced in the 4CN34BCz:4CzIPN device, which achieved a maximum EQE of 20.8% compared with only 13.8% using 3CN34BCz as the host. Li et al. [170] prepared four novel ambipolar hosts based on carbazole and pyrazole (o-CzDPz, m-CzDPz, 3-CzDPz, and mCPDPz) (Figure 51). The different interconnection patterns result in varying degrees of twisting of the conformations of the hosts, which thereby affect important properties such as triplet energies and charge mobility. In addition, the increase in n-type pyrazole to p-type carbazole ratio in mCPDPz enhances the host electron density in the device, resulting in improved charge balance. The best device performance in this study was achieved using 2CzPN:3-CzDPz and 4CzIPN:o-CPDPz as the emitting layers, demonstrating EQEs of 15.8% and 13.7% respectively. Zhang et al. [171] reported an impressive true-blue TADF device by doping DMAC-DPS into a bespoke host DPETPO as the emitting layer, which showed excellent color purity at EL max of 464 nm and CIE coordinates of (0.16, 0.21) (Figure 51). The EQE and power efficiency were 23.0% and 44.4 lm W −1 , respectively, with a small efficiency rolloff of 15% at 1000 cd m −2 and a small onset voltage of 2.8 V. Indeed, DPETPO is sophisticatedly designed such that the asymmetry of the molecule reduces the intermolecular packing. The central p-type diphenyl ether (DPE) core is partially exposed to achieve the best hole and electron mobility compared with analogs DPEPO and DPEQPO, which show maximal and minimal DPE exposure, respectively.
The primary merit of the mixed host approach is balanced charge mobility due to the presence of both p-type and n-type components. Kim and Lee [172] used a mixed host mCP:BmPyPb for 4CzIPN and achieved an excellent EQE max of 28.6% (Figure 52). The authors pointed out the importance of a deep HOMO of the p-type component to prevent exciplex formation in the mixed host. Efficient mixed host systems require that the singlet energy of the mixed host should be high and emission should strongly overlap with the absorption spectrum of the TADF dopant to realize efficient Förster energy transfer. Later, the same group doped 4CzIPN emitter in mixed hosts of mCP:TSPO1 and mCP:SPPO1 and achieved an EQE max of 27.5% ( Figure 52). [173] The latter host system gave a slightly more redshifted emission spectrum because of stronger intermolecular interaction between the emitter in the excited state and the host molecules. Table 9 summarizes the device performances using 4CzIPN as the emitter in different hosts. The highest EQE reported is 31.2% using the bespoke host 3CzPFP. Indeed, due to different device architectures and doping concentrations, a precise ranking of the hosts is not possible. However, there are some clear conclusions that can be drawn.  Firstly, a mixed host approach gives excellent device EQEs (27.5% and 28.6%), which are comparable to that of the device employing 3CzPFP. [172,173] Secondly, in most cases, devices using bespoke hosts with tailored energy levels or mixed hosts display improved performance than devices that use a single traditional host (e.g., CBP). [15a] Finally, CPCB is the only solution-processed host among all those listed that contributes to a device with a decent EQE of 14.5%. Thus, solution-processable bespoke hosts for TADF emitters are currently still underexplored and deserve greater attention.
Cho et al. [174] carried out a study with the goal of designing the ideal combination of hosts and dopants in order to maximize device efficiency. They classified hosts and dopants each into two groups: common ones and TADF ones. They fabricated four types of devices: a common host with a common dopant; a common host with a TADF dopant; a TADF host and a common dopant; and a TADF host and a TADF dopant. It was found that the combination of a common host with a TADF dopant resulted in devices with the highest EQE, which was attributed to efficient energy transfer from the host to the dopant. In particular, singlet and triplet energies on the common host can be transferred to the TADF dopant by Förster and Dexter mechanisms, respectively. Together with efficient RISC occurring in the TADF dopant, very efficient emission was realized using this approach. It was also determined that, for all combinations, the best doping concentration was 1%, with significant drop in EQE even when the concentration was increased to 5%. On the other hand, Fan et al. [175] revealed the importance of compatible molecular configuration and polarity between the host and dopant. Both the blue TADF emitter DMAC-DPS and the host DBTDPO have similar V-shape molecular configurations and polarities (Figure 53). As a result, the thin film of the emitting layer (DMAC-DPS:DBTDPO) has a small root-mean-square (RMS) surface roughness of 0.25 nm. The authors found that device performance was improved by matching the molecular configuration and polarity of the electron-transporting layer with those of the host material in the emitting layer. Thus, 46DBSODPO performs better than the 28DBSODPO and 37DBSODPO analogs for its similar shape compared to DMTDPO (Figure 53). The best blue TADF device reported in this work showed an EQE of 16.1%.
Komino et al. [176] studied the relationship between efficiency roll-off and the molecular orientation of the host. Deposition of the CBP host at high temperature (350 K) results in a random orientation of molecules with a corresponding 1.8-fold-lower hole mobility compared with the film deposited at 200 K where the CBP molecules are more ordered. Due to the lower hole mobility, the recombination zone of the emitting layer CBP:4CzIPN is shifted away from its interface with the electron-transporting layer, resulting in a decrease in the efficiency roll-off of 30%. The authors then reported the preferential horizontal orientation of a linearly shaped PXZ-TRZ ( Figure 21) TADF emitter in an mCBP host by lowering the deposition temperature (T deposition ) at which the anisotropic molecular configuration is slower than the deposition rate. [177] The device using the more horizontally orientated PXZ-TRZ in mCBP (T deposition = 200 K) achieved an EQE of 11.9%, higher than that using a more vertically oriented PXZ-TRZ (T deposition = 300 K), where the EQE was 9.6%, this the result of an enhancement in the outcoupling efficiency.
Ding et al. [178] studied the effect of intergroup distance in host molecules (9CzFSPO, 9CzFDPEPO, and 9CzFDPESPO) on device performance (Figure 54). Despite their identical optical and charge-injection properties, the authors showed that a larger distance between the phosphine oxide and the carbazole moieties (9CzFDPESPO) results in a higher hole mobility. The device with the blue TADF DMAC-DPS in 9CzFDPESPO gave the best EQE of 16.7%, compared with 13.2% and 12.2% using 9CzFDPEPO and 9CzFSPO, respectively.
Méhes et al. [179] carried out photophysical studies of the TADF emitter ACRFLCN (see Section 2.2) doped in a variety of traditional hosts such as DPEPO and CzSi, as well as T2T, and found that the choice of host can have a dramatic effect on the PLQY (15% to 70%) and the magnitude of the delayed component of the emission lifetime (2-70 ms) of the TADF emitter ACRFLCN. The authors suggested avoiding exciplex formation between the host and dopant due to its more pronounced nonradiative decay. Interestingly, the study demonstrated that ΔE ST is a time-dependent parameter after excitation, which is due to the dipole interaction between the host and the ACR-FLCN emitter in the excited state. According to the authors, when ACRFLCN is excited, the population of 3 CT states increases such that the polar host molecules will orient to stabilize both highly polar 1 CT and 3 CT states: the former, being more polar, is stabilized to a greater extent. This results in an initial decrease in ΔE ST , which is, by definition, the energy gap between the lowest singlet state (e.g., 1 CT state) and the lowest triplet state (e.g., 3 LE state that is a π-π* state in nature). As the T 1 state is not a charge-transfer state, its energy will not be affected to the same degree as the CT states. As radiative decays occurs, the populations of both the 1 CT and 3 CT states will decrease, followed by a loss of host molecule orientation. Thus, the energy of both the 1 CT and 3 CT states will increase again, resulting in a subsequent blueshift in the emission spectrum and an increase in ΔE ST .

TADF Emitters employed as Host Materials
Qiu et al. [180] employed a TADF material, PIC-TRZ (a.k.a. PBICT), [181] as the host for the well-known phosphor, fac-[Ir(ppy) 3 ], and contrasted its device performances with those using the traditional host CBP. It was found that the TADF host permits a much lower doping concentration (3 wt%) of phosphor while achieving excellent device efficiency (EQE: 23.9%). Triplet excitons on the TADF host can be up-converted to singlets, which can then be transferred to [Ir(ppy) 3 ] by long-range Förster energy transfer, and thus, a low doping concentration (i.e., longer separation between host and dopant) can be tolerated. On the other hand, triplets on CBP cannot carry out RISC and can only transfer to the dopant via short-range Dexter energy transfer, and, therefore, under low doping concentration, the energy transfer is incomplete and loss occurs in the form of triplet-triplet annihilation. The device lifetime using PBICT as the host also showed less dependence on doping concentration, whereas the CBP device becomes less stable at low doping concentration because of the more pronounced interactions between CBP excitons and its positive polarons, given the incomplete energy transfer to the dopant. The same group also used this TADF-sensitized phosphorescence strategy to address the commonly observed problem of efficiency rolloff in TADF devices. [182] They noted that the roll-off is mainly caused by singlet-triplet annihilation (STA) and triplet-triplet annihilation (TTA), and therefore, the key to reducing the rolloff is to minimize the triplet density in the emitting layer. In devices where TADF dopants are used as the end-emitters, the triplet density is inevitably high due to the intrinsically low k r of the singlet states (i.e., cycling between S 1 and T 1 occurs as k isc competes with k r ). Their strategy involved making use of a high Förster resonance energy transfer rate (k FRET ) and efficient emission from the triplet state of the phosphor dopant to effectively remove the singlet excitons of the TADF host. The elegance of their design is that while k r of the TADF emitter is an intrinsic parameter (i.e., cannot be altered), k FRET can be easily tuned by varying the doping concentration. Zhang et al. [181] attempted to construct a fluorescent OLED using a TADF material as the host. They noted that while triplet excitons can be upconverted to singlets via RISC, the small exchange integral present in the TADF materials intrinsically results in a low k r and thus a low PLQY. They used both PIC-TRZ and DIC-TRZ as the TADF host, which acted as a sensitizer for the fluorescent dopant DDAF (Figure 55). In these devices, the host is responsible for singlet harvesting via RISC, while the dopant is responsible for light emission. The host transfers its energy to the dopant via Förster energy transfer; Dexter energy transfer should be avoided because triplets on DDAF cannot undergo RISC. Thus, a doping concentration as low as 1 wt% was used in these devices. Interestingly, the device efficiencies achieved using PIC-TRZ and DIC-TRZ are very different (EQE: 4.5% and 11.7%,   www.advancedsciencenews.com www.advmat.de respectively). This is, in part, due to the smaller ΔE ST of DIC-TRZ (0.06 eV). More importantly, as a result of the energy levels of the frontier molecular orbitals and charge-transporting abilities of the TADF hosts, charges are trapped on DDAF in the case of the device with PIC-TRZ as the host, whereas they reside on the host in the case of DIC-TRZ. Wang et al. [183] employed 4CzIPN as the host for C 4 -DFQA and C 4 -TCF 3 QA fluorophores and achieved a maximum EQE of 14.6%. Lee et al. (Figure 56). [90] used a co-host system consisting of a TADF emitter (CzAcSF) and DPEPO and combined it with the fluorophore TBPe: a maximum EQE of 18.1% was obtained.
TADF materials can also be used as pseudo-hosts, a strategy that Adachi et al. coined as "hyperfluorescence", in which a combination of a host matrix (an ordinary host material) containing TADF materials (termed an "assistant dopant") and ordinary fluorescent materials (termed an "emitter dopant") are used as the emitting layer. [185] The assistant dopant is responsible for upconverting electrically generated triplets into singlets by RISC and then transferring this energy to the emitter dopant via an FRET mechanism. For example, an OLED device of structure ITO/α-NPD/1 wt% TBRb: 25 wt% PXZ-TRZ: mCBP/T2T/Alq 3 /LiF/Al demonstrated an EQE of 18.0%, with emission originating from TBRb (Figure 46), which undoubtedly surpasses the theoretical limit of traditional fluorescent devices. Obviously, one important merit of a "hyperfluorescence" device is the emission color purity, as it is not of charge-transfer in nature. Apart from the high device efficiency and improved color purity, hyperfluorescent devices have also been demonstrated to have improved operational lifetimes, which is attributed to a more optimized position of the recombination zone within the device. Liu et al. [186] developed the TADF compound PrDPhAc (ΔE ST : 0.07 eV) and applied it as a host for red phosphors such as [Ir(MDQ) 2 (acac)] (Figure 58). The OLED achieved an EQE of 25.8% with CIE coordinates of (0.61, 0.39), much higher than the 12.4% EQE reported [187] when using CBP as host. Unfortunately, the use of PrDPhAc as the TADF emitter in an OLED was not investigated.

TADF Exciplexes
While emitters of a single chemical entity are the "norm" for light generation, exciplexes, which are based on the electronic coupling of two distinct donor and acceptor molecules, can also be an important class of emitting materials employed in OLED devices.   Holes and electrons reside on the donor and acceptor, respectively, followed by radiative relaxation, which involves the two molecules. The advantages of using exciplexes for light emission include: i) facile control of emission colors based on the energy levels of the frontier molecular orbitals of the donor and acceptor, [188] and ii) the intrinsic ambipolar charge-transporting characteristics, which can contribute to a simplified device architecture. [135c,188b] In this section, the recent developments of exciplex systems that demonstrate upconversion of triplet states by TADF is reviewed.
The first TADF exciplex was reported by Adachi et al. in 2012. [189] They compared two types of TADF exciplexes m-MTDATA:tBu-PBD (tBu-PBD = 2-(4-tert-butylphenyl)-5-(4biphenylyl)-1,3,4-oxadiazole) and m-MTDATA:3TPYMB ( Figure 37 and Figure 59). The latter showed a higher device EQE of 5.4% with a EL max at ≈550 nm in the green region, although the PLQY of the m-MTDATA:3TPYMB film was only 26%. The same group studied a green TADF exciplex consisting of m-MTDATA and PPT and achieved an EQE of 10.0% with EL max at ≈520 nm. [190] The authors demonstrated that, when the concentration of the donor (m-MTDATA) is sufficiently high (70 mol%), the triplet excitons on the exciplex can be lost to the donor, resulting in lower device efficiency. The device, based on an mCP:HAP-3MF exciplex, showed green emission with an EQE of 11.3%, with EL max at ≈540 nm. [191] This work is the first report of an exciplex based on a heptaazaphenalene-derived acceptor for OLED devices. Zhang et al. [192] studied TADF exciplex systems between m-MTDATA and BPhen or TPBi. The best result was obtained using the m-MTDATA:BPhen exciplex system, where the green device showed an EQE of 7.79% with EL max at ≈550 nm. Such a high efficiency is due to the near-zero ΔE ST of the exciplex system, where there is very efficient RISC. However, the efficiency rolloff was found to be as high as 41% at 100 mA cm −2 , which was attributed to imbalanced charge in the exciplex-emitting layer leading to singlet and triplet excitons in the exciplex being quenched by polarons. The group then compared a series of three TADF exciplexes between mCP as the donor and three different acceptors (TBPi, BPhen, and 3PT2T). [193] However, the efficiencies of these violet-to-sky-blue devices are poor (EQEs: 0.57-2.23%). The authors proposed, based on theoretical calculations, that the HOMO of mCP is mainly located on the peripheral carbazole moiety whose planar structure inhibits the spin flip required for RISC to occur. When the blue exciplex mCP:3PT2T was used as a host for yellow phosphor [Ir(bt) 2 (acac)] with a high doping concentration (4.0 wt%), enhanced charge trapping by the dopant occurred and no emission from the exciplex was observed. Zhang et al. [194] employed a TADF exciplex, TCTA:Tm3PyBPZ (Tm3PyBPZ = 2,4,6-tris(3′-(pyridin-3-yl)biphenyl-3-yl)-1,3,5-triazine), as the emitting layer to fabricate a green device with EL max at 528 nm and EQE of 13.1% with a very low operating voltage of 2.4 V. Park et al. [195] employed a related TADF exciplex TCTA:B3PYMPM in a bluish-green device (EL max at ≈500 nm) whose EQE increased from 3.1% to 10% when the temperature decreased from room temperature to 35 K. This is because the PLQY of the exciplex is maximized (99%) at 35 K. Interestingly, the delayed emission becomes more dominant at 35 K compared with room temperature, due to the exceedingly low nonradiative losses of exciplex singlets and triplets at this lower temperature, while maintaining efficient RISC. Liu et al. [196] fabricated a TADF OLED based on the exciplex TAPC:DPTPCz, which has a very low ΔE ST of 47 meV and a high PLQY of 68%. The green device showed an EQE of 15.4% with CIE coordinates of (0.27, 0.52). The authors pointed out that the emission energy of the exciplex can be conveniently estimated from solution electrochemical measurements. For efficient RISC to occur, it is also of prime importance that the E T of individual components be higher than the exciplex emission energy.
The same group also reported the first example of using a TADF exciplex (TAPC:DPTPCz) as a host for sensitizing a conventional fluorescent emitter (C545T). [197] Given that RISC is not prevalent in C545T, the HOMO and LUMO of the TAPC:DPTPCz exciplex system must be comparable to those of C545T in order to avoid charge trapping by the dopant. In the study, the doping concentration was kept at very low level (1 wt%) to avoid undesirable Dexter energy transfer to C545T   photophysics study on a 1:1 blend of m-MTDATA and photoluminescence of the exciplex, they observed there was no MFE on a short time-scale (<100 ns), while there was a strong MFE observed at longer time scales (positive for 100 ns to 1 µs and negative for 5-30 µs). By establishing a quantum-mechanical rate model, the authors concluded that the TADF process observed in this exciplex system is dynamic in nature. While, initially, the ΔE ST is so large that MFE is unable to influence the rate of singlet-triplet interconversion, the hole and electron on the exciplex soon separate beyond the exchange radius (≈1 nm), resulting in a decrease in ΔE ST and efficient RISC by hyperfine coupling. The hole and electron finally recombine to form an emissive singlet. It was estimated that about 30% of RISC is carried out by this "indirect" path.
Hung et al. [201] reported a bilayer interfacial TADF exciplex employed in an OLED device. Yellow emission was generated between two adjacent layers of TCTA and 3P-T2T. Thanks to the excellent and balanced charge mobility of the exciplex components, together with the large energy-level offset (0.8 eV) of the TCTA/3P-T2T interface, the device showed an EQE of 7.7% with CIE coordinates of (0.40, 0.55). There have been no other reports of interfacial TADF exciplex applications in OLED devices apart from this work.

Conclusions and Outlook
It is exciting and gratifying to witness OLED devices based on purely organic TADF emitters that show performance metrics comparable, and in some cases improved upon, compared to the current state-of-the-art organometallic complexes. What is particularly surprising is the rate of development of and increased interest in TADF emitters since Adachi's seminal Nature paper in 2012. However, we believe that there is still much to learn with respect to understanding the intricacies of the TADF mechanism and to developing TADF materials that exhibit improved performance in EL devices, particularly in the following areas: i) Mechanism of TADF While it is generally agreed that ΔE ST is an important factor that governs the efficiency of TADF, it presupposes that the singlet-harvesting process is the result of a RISC from T 1 to S 1 . Other pathways have been proposed in the literature. For example, the presence of "hidden" 3 nπ* states [42] and non-adiabatic effects between excited states [43] have been proposed by different research groups to explain the efficient RISC process in diphenylsulfone-based blue TADF emitters with rather large ΔE ST (>0.3 eV). The TADF process has been proposed to be induced by spin-orbit coupling (SOC) or hyperfine coupling (HFC) in certain cases. [24] Some researchers have suggested that ΔE ST is a dynamic property, which can change during the timeframe of the emission decay. [179] Access to higher energy triplet states through reverse internal conversion (RIC) followed by RISC renders ΔE ST a less accurate indicator of singlet-harvesting efficiency as access to the singlet excited state does not occur directly through a T 1 -S 1 RISC pathway. In each of these frameworks, the photophysics of the emitter are discussed without taking the conformational flexibility within the molecule into account. Each conformer will possess an associated set of HOMO and LUMO energy levels, as well as different singlet and triplet excited-state energies. The experimental spectroscopic picture produced is an aggregate of the photophysics of each of these conformers, which complicates analysis. Clearly, there is still much research to be done to comprehend the detailed nature of the TADF mechanism.
ii) Lifetime of TADF EL devices We note that TADF EL device lifetime data are seriously underreported. We can find only six cases out of more than 140 TADF emitters where their device lifetimes have been dis closed. [46,53,99d,120,202] Indeed, it is probably fair to state that the reported device architectures have been designed to optimize efficiency and not stability, despite the fact that device stability is an essential factor for the marketability of this technology. Gratifyingly, a recent report by Tsang and Adachi demonstrated a device lifetime (LT 95 ) of 1315 h under 1000 cd m −2 operation, which is comparable to existing green phosphorescent OLEDs. [99d] Designing similarly stable blue and red (and white) TADF OLEDs remains a challenge.

iii) Fabrication cost of TADF OLEDs
The fabrication cost of TADF OLEDs also plays a critical role in their marketability, particularly in the context of their use in lighting. Simplified device architecture and solutionprocessing are two major approaches to cut fabrication costs. A focus on the development of suitable small-molecule, polymeric, and dendrimeric TADF emitters for solution-processable devices possessing comparable performance metrics is strongly desired.

iv) Emission energy of TADF emitters
There are now ample examples of high performance green and yellow emitters, and a large number of blue emitters. There remains a dearth of deep-red TADF emitters and, for other applications such as telecommunications, a need for near-IR emitters. v) Consistency in the reporting of the optoelectronic properties of emitters There needs to be a commonly accepted protocol for determining the optoelectronic properties of TADF emitters. As an example, ΔE ST is determined by a number of different methods that may not necessarily provide the same value. These include: extrapolation from an Arrhenius plot of variable-temperature emission data; estimating the singlet and triplet energies from room and low-temperature measurements (either as their λ max values or by determining their emission onset); and determining the ΔE ST from time-resolved emission spectroscopy (TRES). Likewise, the reporting of methodology for evaluating emission lifetimes, particularly the delayed fluorescence lifetime, is frequently absent. It is mechanistically important to know, for instance, whether the delayed fluorescence is monoexponential or polyexponential. It is usually the case that a single delayed emission lifetime value is reported, without substantiation for whether this is an average value or whether the delayed emission fits to a monoexponential decay.