Main

Commercial organic light-emitting diodes (OLEDs) still require better blue emitters1,2. Even today, fluorescent and/or triplet fusion emitters are used for commercial blue OLED pixels, as these materials alone provide acceptable device lifetimes3, despite lower efficiency compared with blue phosphorescent or thermally activated delayed fluorescence (TADF) emitters4,5. The potential benefits of stable higher-performance blue emitters are even more pressing because of newly envisioned OLED display architectures, which exclusively utilize blue pixels with external fluorescent colour conversion layers6,7. These offer ~30% reduction in power consumption, and in the future could save more than 150 TWh in annual global residential electricity consumption (Supplementary Information).

Hyperfluorescent (HF) OLEDs offer ways to achieve this, circumvent the limitations of both TADF or phosphorescent blue emitters, and using them as sensitizers for separate terminal emitters8, but this approach is still relatively immature and not well understood9,10,11. A high-performance deep-blue ‘triplet-harvesting’ sensitizer with good spectral overlap to a terminal emitter is required, enabling efficient Förster resonance energy transfer (FRET)10,12,13. The broad emission spectra and low radiative rates of a TADF sensitizer can, thus, be compensated for by the fast and narrowband emission from the terminal emitter and 100% triplet excitons are utilized by the former14,15,16. Concomitantly, we have shown that some green-emitting sensitizers can support blue HF-OLEDs17. This translates to a lowering of the singlet and triplet energies of all the materials in the emission layer, improving the device stability.

However, many papers report HF-TADF OLEDs with external quantum efficiencies (EQEs) lower than their TADF counterparts12,18. Where EQEs are higher10,19, no clear pattern or identifiable underlying cause is given. Multiresonance TADF terminal emitters14 that also harvest triplet excitons yield HF-OLEDs that typically outperform non-HF devices20,21,22,23,24. Along with EQE, in many cases, the HF-OLED efficiency roll-off performance is better, but in some cases, it is worse15,24. In almost all the reports, the ‘best’ TADF emitters are used in pre-optimized OLED stacks with a small concentration of the terminal emitter added to make an HF-OLED—prima facie, a sensible strategy12,25,26,27. However, the TADF emitter may already be performing at its maximum potential, and the introduction of a terminal emitter opens up additional loss pathways, for example, charge or triplet trapping on the terminal emitter. The terminal emitter can contribute to higher performance through higher photoluminescence quantum yield (PLQY) or improved light outcoupling, but these do not always outweigh the additional losses16,28.

The performance of such best TADF emitters also relies on a delicately balanced highest occupied molecular orbital–lowest unoccupied molecular orbital overlap29. A high overlap will increase the oscillator strength and PLQY (the ratio of radiative and non-radiative decay rates), but is detrimental to triplet-harvesting reverse intersystem crossing rates (krISC) and vice versa30. Thus, donor–acceptor TADF emitters with the fastest reverse intersystem crossing (rISC) rates but limited PLQYs do not produce high-performance OLEDs and are typically not investigated in HF-OLEDs. Counterintuitively, in HF-OLEDs, the PLQY of the sensitizer is not crucial if the FRET rate (kFRET) outcompetes the non-radiative decay, and instead, the higher krISC is the most beneficial.

Here we demonstrate extraordinary HF-OLED performance enhancement using TADF sensitizers, which themselves only give moderate non-HF device performance. Starting from our recent work using the high-performance TADF emitter DMAC-TRZ as a benchmark17,31, we demonstrate that ACRSA (moderate OLED performance) gives HF-OLEDs with an almost tripled EQEmax (refs. 10,32). Similar results are also demonstrated with the deeper-blue TADF emitter AZB-TRZ (ref. 33), which has fast rISC but low PLQY. Surprisingly, despite the photoluminescence (PL) spectra of ACRSA and DMAC-TRZ being energetically similar (comparable J factors), ACRSA exhibits notably higher FRET efficiency with ν-DABNA, even compared with the deeper-blue AZB-TRZ sensitizer, which has better spectral overlap. We demonstrate that this arises because of ACRSA’s rigid structure and long-lived excited states, normally a detriment in OLEDs. These findings establish new and unexpected design rules for high-performance HF sensitizers.

Results

Photophysics

The steady-state PL spectra of ACRSA and DMAC-TRZ at 10 wt% in the 3,3′-di(carbazol-9-yl)-5-cyano-1,1′-biphenyl (mCBPCN) host are shown in Fig. 1. The intrinsic broadness of the DMAC-TRZ charge transfer (CT) emission, centred at 500 nm, is due to contributions from different molecular configurations with different donor–acceptor dihedral angles in the film, together with some aggregate emission (Supplementary Fig. 2b)34,35. These contributions are resolved in time-resolved spectra even when FRET to a terminal emitter is active (in this case, ν-DABNA)14,27 (Supplementary Fig. 1c). In ACRSA, the absence of dihedral-angle inhomogeneity results in a narrower intrinsic PL spectrum, as evident in the deconvoluted spectra of ACRSA (Supplementary Fig. 2a) and as reported previously35,36.

Fig. 1: Sensitizer terminal emitter spectral overlap for FRET and the corresponding energy pathways.
figure 1

a, Steady-state absorption of ν-DABNA in a toluene solution, alongside the PL spectra of ACRSA:mCBPCN (10 wt%) and DMAC-TRZ:mCBPCN (10 wt%) films, excited at 355 and 340 nm, respectively. DMAC-TRZ:mCBPCN films exhibit a strong emission band centred at 500 nm, which we have shown is the 1CT emission from the highly twisted quasi-equatorial conformer distribution17. In ACRSA:mCBPCN, the broad emission band at ~500 nm is a characteristic of ACRSA 1CT emission in solid-state hosts36; the emission at 400 nm is from the mCBPCN host whose absorption at 340 nm is much stronger compared with that of ACRSA. The filled areas are the spectral overlaps between the curves used to calculate the FRET radius. b, FRET mechanism in a guest:sensitizer:terminal emitter system. The emission from the sensitizer (yellow arrow) occurs when FRET is less than 100% efficient.

Full size image

Excluding the mCBPCN contribution to the ACRSA PL spectrum, comparing the spectral overlap in the two TADF emitters with ν-DABNA absorption27, ACRSA displays slightly increased spectral overlap because of its higher energy PL onset (Fig. 1, red-filled area), compared to the DMAC-TRZ overlap which is highlighted in green. For both emitters, only the blue edge of their emission overlaps with the narrow S0–S1 absorption of ν-DABNA, but efficient FRET surprisingly still ensues for both materials (although for different reasons, as discussed below). The calculated FRET radii between ν-DABNA and ACRSA or DMAC-TRZ are estimated to be 2.8 and 3.2 nm, respectively (Supplementary Information). ACRSA:ν-DABNA has slightly better spectral overlap but the calculated FRET radius is lower than DMAC-TRZ:ν-DABNA because the ACRSA PLQY is three times smaller.

The FRET dynamics in the two HF systems were investigated using time-resolved emission spectroscopy. For the ACRSA:mCBPCN (10 wt%) film (Fig. 2a), three species are identified; mCBPCN, ACRSA singlet CT (1CT) and ACRSA excimers (Table 1 lists their lifetimes). The very long 1CT decay lifetime arises from the highly decoupled spiro-linked donor–acceptor in ACRSA (ref. 37). The FRET-accelerated prompt fluorescence (PF) lifetime of ACRSA in the HF film reduces from 322.03 to 39.27 ns (Table 1). Comparably, the PF lifetime of DMAC-TRZ in mCBPCN reduces from 21.66 to 15.55 ns with 1 wt% ν-DABNA. This substantial lifetime quenching has an enormous impact on the FRET efficiency (ηFRET) and subsequent HF-OLED performance of ACRSA:ν-DABNA (Supplementary Table 1).

Fig. 2: Time-resolved emission measurements following the exciton dynamics.
figure 2

a–c, Time-resolved PL spectra of ACRSA:mCBPCN (10%) (a), ν-DABNA:mCBPCN (1%) (b) and ACRSA:ν-DABNA:mCBPCN (10:1:89 wt%) (c). The DF lifetime of 10 wt% ACRSA:mCBPCN is 7.9 μs because of the excimer states. For comparison, ACRSA at 1 wt% loading in zeonex or bis[2-(diphenylphosphino)phenyl]ether oxide (DPEPO) films have DF lifetimes of 6.3 and 7.3 µs, respectively, with no excimer emission observed. At 10 wt% in DPEPO, similar to the mCBPCN host, the ACRSA DF lifetime is 8.7 µs with a clear excimer emission observed36. The excimer emission is not observed in the ACRSA:ν-DABNA:mCBPCN (10:1:89 wt%) delayed emission. d, PL decay traces of the evaporated films at room temperature. With the additional presence of 1 wt% ν-DABNA (in the ACRSA:mCBPCN system), the PF lifetime has two components; directly excited ν-DABNA with a lifetime of 3.5 ns, which is the same as ν-DABNA in mCBPCN, and ACRSA excitation followed by FRET to v-DABNA with an average lifetime of 39.27 ns. Exciting the HF film at 420 nm to selectively excite ν-DABNA produces the same emission decay kinetics as a film without the ACRSA sensitizer (Supplementary Fig. 3). Here λexc = 355 nm.

Full size image
Table 1 Key photophysical characteristics of ACRSA:ν-DABNA:mCBPCN, ACRSA:mCBPCN and ν-DABNA:mCBPCN films at room temperature (λexc = 355 nm)
Full size table

The kFRET and ηFRET values of the HF systems can be estimated from the changes in the sensitizer PF lifetime in the presence of the terminal emitter (Supplementary Information)9. For the ACRSA HF system, ηFRET is calculated as 87.8% (Supplementary Table 6), which is very high given such limited spectral overlap (Fig. 1). In the DMAC-TRZ HF system, ηFRET is calculated as 28.2%. This value is obviously at odds with our previously reported PL and electroluminescence (EL) spectra of HF-OLEDs, because we only considered the quasi-equatorial conformer emission17. Additional FRET from the quasi-axial conformer to the terminal emitter means that the FRET radius, kFRET and ηFRET are larger in the actual DMAC-TRZ:ν-DABNA HF17.

The very long ACRSA 1CT emission lifetime enables relatively slow FRET to achieve high ηFRET, despite ACRSA:ν-DABNA’s smaller R0 and r values. Indeed, considering the rate constants derived from the emission decays38, we calculate a kFRET value of 2.24 × 107 s−1 for ACRSA:ν-DABNA, very similar to the DMAC-TRZ HF system (1.81 × 107 s−1; Supplementary Table 6). In DMAC-TRZ:ν-DABNA, kFRET is in direct competition with the intersystem crossing rate (kISC = 3.37 × 107 s−1) and fluorescence rate (kF = 4.62 × 107 s−1; Supplementary Table 4). As the PLQY of DMAC-TRZ is high (~0.98), the non-radiative rates must be small39. In ACRSA, kISC and kF are both much slower, namely, 1.0 × 106 and 3.1 × 106 s−1, respectively. Non-radiative rates must also be slow to support such a long singlet exciton lifetime, thereby allowing FRET to dominate far beyond what it can achieve with DMAC-TRZ:ν-DABNA, despite the similar kFRET values (Supplementary Table 5). To generalize, sensitizers with slow radiative and slow non-radiative rates provide enough time for kFRET to accumulate and dominate the exciton dynamics. These same properties, however, limit the performance of the sensitizer as a TADF emitter in its own right.

Although FRET properties can be estimated with great confidence40, this is not the case for Dexter rates9. We estimated the average distance between the sensitizer and terminal emitter based on the kFRET and R0 calculations (Supplementary Table 6). In both cases, the average intermolecular distance is greater than a reasonable estimate of the effective Dexter radius (~1 nm, required for direct orbital overlap)9, and thus, we assume that Dexter transfer is negligible in these films.

We also note that the calculated ηFRET values represent the probability for singlet energy transfer during the singlet exciton lifetime. Although appropriate for standard fluorescent materials, instead, for excitons cycling (potentially multiple times) between singlet and triplet states (as in a TADF sensitizer), the overall/cumulative probability of FRET substantially increases with each cycle. Here we estimate ηFRET using only the PF lifetimes; thus, TADF materials with a fast rISC rate and high delayed fluorescence (DF)/PF ratio will have a much higher cumulative FRET efficiency. This can be observed in HF systems where experimental ηFRET and PL spectra are in contrast, indicating a far more complete FRET than the calculated ηFRET values otherwise imply (as discussed later).

An additional factor contributing to the small calculated ηFRET in the DMAC-TRZ HF system is the notable emission from lower-energy configurations, evident in late PF and early DF (time delay, 20–5,000 ns; Supplementary Fig. 1c,d34). Although not noticeably detrimental in DMAC-TRZ OLEDs, these low-energy states have vanishing spectra overlap with ν-DABNA and effectively zero ηFRET. The ACRSA HF system does not have this broad distribution of states (Fig. 2c and Supplementary Fig. 6), such that all the excited states have similar ηFRET values. This comparison highlights how homogeneous emitter conformations in TADF materials are a key benefit in HF-OLEDs supporting complete FRET.

The sensitizer rISC controls triplet exciton harvesting and lifetimes that support high efficiency in devices. ACRSA:ν-DABNA has a DF lifetime of 4.5 μs, which matches ν-DABNA:mCBPCN (Table 1), whereas it is 7.9 μs for 10 wt% ACRSA:mCBPCN, being influenced by the excimer states (Fig. 2). Thus, the ACRSA sensitization step to ν-DABNA prevents ACRSA excimer formation entirely (supported by the time-resolved spectra; Fig. 2a,c), even at 10 wt% loading, such that all ACRSA excitons are transferred to ν-DABNA, giving near-perfect sensitization.

DMAC-TRZ:ν-DABNA, in contrast, clearly has different DF contributions (Supplementary Fig. 1). A fast component exists in the early DF regime (~1 μs) with high contribution to the time-resolved PL emission, from the highly twisted DMAC-TRZ conformers (emission at 510 nm), which have weak overlap with the terminal emitter and thus low ηFRET (Supplementary Fig. 1d). The second regime is similar to the ν-DABNA:mCBPCN film DF, yielding efficient sensitization and a long DF tail from the slow sensitization of unfavourable DMAC-TRZ species (less twisted conformers with slow krISC). The average DF lifetime is 11 μs (Supplementary Table 1), indicating that these slow species greatly increase the overall DF lifetime of the HF system, an undesirable phenomenon in OLEDs.

The ACRSA emission and ν-DABNA absorption overlap is small (Fig. 1), but this does not limit ηFRET. This crucial realization allows us to maintain efficient cumulative FRET as the excited-state energy of the sensitizer is minimized. This is not unique to ACRSA, but arises from the long excited-state lifetime so that the singlet excited-state population primarily resides in the S1 zeroth vibrational level (Fig. 3). In ν-DABNA, the S0–S1 transition has very high oscillator strength and narrow bandwidth. Thus, the overlap between the sensitizer populated states and the spectrally narrow ν-DABNA remains suitably strong, making the ACRSA:ν-DABNA HF system ideal for maximal ηFRET despite the minimal spectral overlap. From this, surprisingly low sensitizer singlet and triplet excited-state energies can be used to sensitize blue terminal emitters, considerably increasing the range of usable low-energy hosts and emitter materials. This realization redefines the photophysical design requirements for sensitizer and terminal emitters to achieve optimal FRET alongside minimal sensitizer excited-state energies in HF-OLEDs.

Fig. 3: Schematic of the energy-level diagram for FRET.
figure 3

FRET between a hypothetical TADF sensitizer and ν-DABNA terminal emitter, showing how an apparent very small spectral overlap can still give rise to fast and efficient energy transfer between them. FRET is relatively slow because the sensitizer (1CT state) transition dipole moment is very small. Thus, the dipole–dipole interaction is weak, such that we are in the weak-coupling (FRET) regime. The energy donor, thus, has a long excited-state lifetime so that the vast majority of singlet excited-state population resides in the S1 zeroth vibrational level; consequently, the overlap of just the red edge of the acceptor absorption with the blue edge of the donor emission still ensures that all of the donor population can be transferred to the acceptor. S0 and S1 represent the ground and first excited states, respectively, of the sensitizer (left) and terminal emitter (right). It is noteworthy that the sensitizer’s PL spectrum is vibronic for illustrative purposes, to allow an easier correlation of the vibronic levels with the energy diagram. Typically, and indeed in the examples examined here, the PL spectrum of a TADF sensitizer is broad and Gaussian shaped without a clear separation of vibronic states.

Full size image

Devices

OLEDs were fabricated using the same device architecture for all the devices as the sensitizers used have similar highest occupied molecular orbital and lowest unoccupied molecular orbital values. (Fig. 4a). Although device performance could improve in devices individually optimized for each emitter, comparing the sensitizer, terminal emitter and HF combinations in the same device structure provides an insight into the sensitization processes in each emitter.

Fig. 4: OLEDs.
figure 4

a–c, Device architecture (a) and EL spectra of TADF devices (b) and HF-OLEDs (c). A virtually complete FRET is observed in the ACRSA HF-OLEDs that have the EL spectra identical to those of ν-DABNA. Conversely, residual sensitizer emission is observed (on the red side of the EL emission) in the DMAC-TRZ and AZB-TRZ HF-OLEDs. d–f, EQE versus luminance of DMAC-TRZ (d), AZB-TRZ (e) and ACRSA (f) OLEDs with and without the terminal emitter (blue curve for ν-DABNA without the presence of the sensitizer is shown for comparison).

Full size image

It is well established that the slow rISC of multiresonance TADF materials such as ν-DABNA leads to unsatisfactory efficiency roll-off at higher current densities in non-triplet-harvesting host devices20,21 (Fig. 4d–f, blue curve). We note that ν-DABNA performs exceptionally well in a DOBNA-OAr host, but this is not well understood41. ACRSA does not perform as well in OLEDs as DMAC-TRZ (refs. 10,32). In our device stack, the EQEmax value is limited to ~11% compared with ~25% for DMAC-TRZ (10 wt% sensitizer in the mCBPCN host). The roll-off for ACRSA is also worse than DMAC-TRZ OLEDs, a result of its slower rISC and longer exciton lifetimes (Supplementary Table 4). However, introducing 1 wt% ν-DABNA to the ACRSA devices, the EQEmax almost triples to 28.5%, whereas DMAC-TRZ:ν-DABNA HF-OLEDs show little change (Fig. 4). Furthermore, FRET is virtually complete in the ACRSA HF-OLEDs (Fig. 4c).

We propose that this remarkablet performance enhancement in ACRSA HF-OLEDs arises from the same properties that make ACRSA a less effective TADF emitter. In particular, the slow radiative and non-radiative rates of ACRSA are extensively outcompeted by FRET to ν-DABNA, overcoming its low PLQY. This long energy transfer window as well as the absence of any unfavourable conformers with poor ηFRET supports complete ACRSA:ν-DABNA FRET. In contrast, kISC and kF are comparably faster in DMAC-TRZ, leading to less complete FRET despite similar spectral overlap (Supplementary Table 3). The high PLQY in DMAC-TRZ means that radiative decay already outcompetes non-radiative decay, and therefore, the FRET processes cannot further improve the device performance. There also exists a sub-population of DMAC-TRZ molecules in unfavourable dihedral-angle configurations with redshifted emission inaccessible for FRET (Fig. 4c).

The similar EQE roll-off in the ACRSA TADF and HF devices indicate that triplet harvesting remains the exclusive role of the sensitizer, not the ν-DABNA terminal emitter (Supplementary Table 7). The possibility of device degradation during the J–V–L measurements is excluded by the identical consecutive sweeps of the same pixel (Supplementary Fig. 11). Nonetheless, the roll-off slope is improved at high luminance, indicating that the HF-OLED probably reduces the accumulation of ACRSA excitons through rapid FRET. The demonstrated devices open a new path for blue OLEDs, through which deep-blue sensitizers and associated high triplet hosts are not necessary4,5,15. Green-emitting TADF sensitizers with lower-triplet hosts can instead be used (offering substantially better stability in devices); combined with a highly absorbing red edge terminal emitter, these surprisingly effective HF systems with seemingly inadequate FRET overlap demonstrate an unexpected path to highly efficient blue OLEDs.

Supplementary Fig. 12 provides the stability of these devices. Surprisingly, the LT50 value for DMAC-TRZ OLEDs (at 1,500 cd m−2) is four times longer than the corresponding HF-OLEDs (34.0 to 8.5 h). However, the opposite trend occurs in ACRSA-based devices, in which the lifetime improves by 2.5 times in HF-OLEDs (28 to 69 min at 2,500 cd m−2), following efficient FRET and EQE improvements.

To explore this strategy further, we investigated an additional non-rigid TADF molecule, with blueshifted emission compared with DMAC-TRZ. AZB-TRZ (Fig. 4e) has fast non-radiative decay rate, fast krISC and low PLQY compared with DMAC-TRZ (Supplementary Table 2)33. Similar to DMAC-TRZ, a clear distribution of dihedral angles is observed through the time-dependent emission spectra (Supplementary Fig. 5a), giving redshifted emission with limited spectral overlap to ν-DABNA. Nonetheless, the spectral overlap is higher compared with ACRSA and DMAC-TRZ sensitizers (Supplementary Fig. 4), giving the highest J factor and FRET radius (Supplementary Tables 3 and 6) in this series of sensitizers.

Unsurprisingly, the CT energy dispersion together with fast kF and kISC of AZB-TRZ compete with kFRET (Supplementary Table 6), leading to a seemingly low ηFRET of 37%, but a high triplet population. The latter indicates why ηFRET appears to be underestimated; the HF-OLED spectra indicate near-complete energy transfer, that is, a small contribution from the sensitizer in the EL spectrum (~510 nm; Fig. 4c). Intersystem crossing/rISC cycling can efficiently harvest the high triplet population and FRET occurs throughout the DF/triplet regime, giving a larger cumulative ηFRET.

The combination of FRET and fast rISC, therefore, helps to greatly enhance the AZB-TRZ HF device efficiency, but the energy transfer is not as complete as in ACRSA. The fast radiative and non-radiative singlet decay rates do not allow FRET to completely dominate the exciton dynamics in AZB-TRZ HF devices, and their colour coordinates do not shift as much as that in ACRSA HF-OLEDs, despite AZB-TRZ having a higher energy emission and spectral overlap (Fig. 5). The device EQEmax, nonetheless, quadruples from 5% to 20%, and the already excellent roll-off from rapid rISC in AZB-TRZ devices is enhanced in HF devices.

Fig. 5: Comparison of CIE coordinates of ACRSA and AZB-TRZ OLEDs and HF-OLEDs.
figure 5

The ACRSA sensitizer giving the most complete energy transfer to the terminal emitter shows the largest blueshift in the colour coordinates to virtually that of the ν-DABNA:mCBPCN OLED. The EL spectra are collected at 1,000 cd m−2.

Full size image

The power and current efficiencies of both ACRSA and AZB-TRZ HF-OLEDs show enhancement (Supplementary Figs. 7 and 8), only partially explained by the improved quantum efficiencies. In ACRSA, they rise from 31 cd A−1 and 28 lm W−1 in the greenish TADF OLEDs to 37 cd A−1 and 36 lm W−1 in the deeper-blue HF-OLEDs (Supplementary Table 7). The effect is even greater for AZB-TRZ HF-OLEDs, rising from 12 cd A−1 and 9 lm W−1 in TADF OLEDs to 30 cd A−1 and 27 lm W−1 in HF-OLEDs (Supplementary Fig. 8). In contrast, DMAC-TRZ HF-OLEDs show a decrease in power efficiency as EQE remains unchanged and only the spectrum changes (Supplementary Fig. 9). Considering the eye spectral response, a green emitter with similar radiance to a deeper-blue emitter will always have higher power efficiency—a trend that the overwhelming efficiency enhancements of our HF-OLEDs disrupts.

Discussion

We have elucidated key photophysical parameters that yield optimal triplet upconversion sensitization in HF-OLEDs. We show that these parameters considerably differ from those required for an efficient non-HF-TADF OLED emitter. Importantly, rigid TADF molecules with homogeneous CT emission energy, exemplified by the spiro-linked TADF molecule ACRSA, allow all the molecules in the film to undergo efficient, complete FRET. Long radiative lifetimes and low intersystem crossing rates are ideal for sensitization, allowing small spectral overlaps to achieve efficient cumulative FRET to the terminal emitter. High rISC (and consequently high intersystem crossing) rates remain desirable for fast triplet-harvesting and minimal triplet annihilation/quenching phenomena in TADF OLEDs. In HF-OLEDs, a much slower kISC and kr can still support efficient sensitization if they can strike a delicate balance between being fast enough to ensure device stability and slow enough to allow FRET to outcompete the sensitizer’s radiative and non-radiative processes, removing the need for high sensitizer PLQY. This strategy also enables the sensitizer excitation energies to be pushed lower and retaining deep-blue terminal emission, with substantial benefits on the choice of host material and device longevity.

These results reveal an interesting blind spot in TADF HF research. By limiting the scope to only the best-performing TADF materials as sensitizers, that is, DMAC-TRZ, many excellent HF pairs using ‘less ideal’ TADF sensitizers have probably been overlooked, and would probably never be considered or discovered. Because of this ability to turn intrinsically ‘poor’ TADF emitters into blue HF-OLEDs with exceptional performance (Supplementary Fig. 10), a re-evaluation of many previously reported and unreported TADF materials should be stimulated.

Methods

Photophysical measurements

Solid-state samples were evaporated on quartz substrates into a Kurt J. Lesker Super Spectros deposition chamber, at pressures below 10–7 mbar. Steady-state absorption and emission spectra were measured using a double-beam Shimadzu UV-3600 UV/VIS/NIR spectrophotometer and a Horiba Jobin Yvon Fluorolog-3 spectrofluorometer, respectively. Time-resolved measurements were performed using a spectrograph (Horiba Triax) and a Stanford Computer Optics 4 Picos intensified charge-coupled device camera, where samples were excited with a Nd:YAG laser (EKSPLA, 10 Hz, 355 nm) under a vacuum.

Device fabrication

OLEDs were fabricated on patterned indium-tin-oxide-coated glass (VisionTek Systems) with a sheet resistance of 15 Ω sq–1. After sonicating in acetone and isopropanol, oxygen-plasma-cleaned substrates were loaded into a Kurt J. Lesker Super Spectros deposition chamber, and both small-molecule and cathode layers were thermally evaporated at pressures below 10–7 mbar. The materials used for the production of the OLEDs were N,N-bis(naphthalene-1-yl)-N,-bis(phenyl)benzidine (NPB) as the hole transport layer; 3,3′-di(9H-carbazol-9-yl)-1,1′-biphenyl (mCBP) as the electron blocking layer; the emission layer had mCBPCN or 1,3-bis(N-carbazolyl)benzene (mCP) as the host, DMAC-TRZ, ACRSA and AZB-TRZ as the sensitizer, ν-DABNA as the terminal emitter, 2,4,6-tris(biphenyl-3-yl)-1,3,5-triazine (T2T) as the hole blocking layer, T2T and 8-hydroxyquinolinolato-lithium (Liq) as the electron transport/injection layer and an aluminium (Al) cathode. NPB, mCP, mCBP and T2T were purchased from Sigma-Aldrich, and mCBPCN and ACRSA were purchased from Ossila; they were sublimed before use. Sublimed DMAC-TRZ was purchased from Lumtec. ν-DABNA was synthesized and purified by T. Hatakeyama’s group. AZB-TRZ was synthesized and purified by E. Zysman-Colman’s group.

Device characterization

Freshly evaporated devices were transferred into a calibrated six-inch integrating sphere (Labsphere) in a glovebox, and their electrical properties were measured using a source meter (Keithley 2400). The emission spectra were simultaneously measured using both calibrated fibre-coupled spectrometer (Ocean Optics USB4000) and photodiode, for low luminance. All the devices were evaluated at 293 K and under a N2 atmosphere.

Device lifetime

The OLED lifetime measurements were measured in a glovebox under a N2 atmosphere using an Ossila LED Measurement System.