Multiple resonant thermally activated delayed fluorescent (MR-TADF) compounds based on boron and nitrogen (B,N)-doped nanographene have attracted significant interest for their commercial potential in enhancing the performance of organic light-emitting diodes (OLEDs)1-5. These compounds are distinguished by their intense emission and narrowband feature, which are crucial for achieving satisfactory color purity necessary for ultrahigh definition displays (UHD). In addition, their ability to utilize both singlet and triplet excitons enables them to achieve an internal quantum efficiency (IQE) of up to 100%. This character is critical in the development of blue emitters to supplant existing triplet-triplet annihilation materials with restricted IQE and suboptimal color purity. Nonetheless, MR-TADF emitters face challenges in converting triplet excitons into singlet excitons due to their moderately large singlet-triplet energy gap (ΔEST)1,6-8, resulting in slower reverse intersystem crossing (RISC) process compared to those of leading donor–acceptor (D–A) type TADF materials9-11, which in turn causes significant efficiency roll-off in devices.
Fortunately, the triplet harvesting capability and color purity of MR-TADF frameworks scale favorably with their size12. On one hand, recent theoretical investigations have demonstrated that the π-extension of MR framework can concurrently reduce ΔEST and enlarge the oscillator strength (fosc), owing to enhanced charge transfer (CT) delocalization volume13,14. On the other hand, π-extension is considered beneficial for enhanced molecular rigidity and reduced reorganization energy, thus reducing full width at half-maximum (FWHM)15-17. Adhering to this strategy, numerous multiple boron-embedded compounds have been constructed, demonstrating superior performance in OLEDs compared to their single boron-embedded counterparts18. One notable example in the blue region is ν-DABNA developed by Hatakeyama et al., which achieved high external quantum efficiencies (EQE) of 34.4% and mitigated efficiency roll-off, underscoring the viability of π-extension in creating narrowband organic materials with high efficiency and expedited RISC processes3.
Despite the significant advancements, there exists considerable potential for further exploration in this field. Firstly, the investigation of π-extended MR-TADF molecules containing more than three boron atoms has been relatively limited due to constraints in molecular design and synthesis19. Secondly, as delineated in Scheme S1, current examples primarily concentrate on linearly extended skeletons15,20-22, with a scarcity of research on molecules featuring diverse topologies, where V-DABNA, ω-DABNA and some derivatives stand as rare exceptions23-25. Consequently, elucidating the structure-property relationship among higher-order B,N-motifs remains challenging, thereby hindering the development of advanced emitters. From the perspective of device applications, it also remains a formidable quest at the current stage to achieve deep-blue dopants concurrently featuring ultra-narrowband emission (i.e. FWHM below 0.10 eV) and fast RISC rate constant (i.e. kRISC on the order of 106 s-1) to guarantee high color purity and desirable efficiency at practical brightness26.
In this context, we wish to elucidate the isolated effect of topological structure on photophysical outcomes for B,N-doped nanographenes and shed light on the design tactic for blue MR-TADF emitters. To this end, we constructed a series of quadruple borylated MR-TADF emitters featuring isomeric skeletons. Within these frameworks, the boron and nitrogen atoms are arranged ortho-/para- to each other, and the number of ortho-B-π-N and para-B-π-N pairs is kept the same. This design therefore promises a comparable electron push-pull effect within these emitters and assures their emission falls into the deep-blue region with CIEy below 0.08, permitting a fair comparison of their excited-state properties. Our investigation uncovered that the orientation of the heteroatoms markedly affects the molecular conformation, leading to substantial variations ranging from negatively curved to quasi-planar geometries. Subsequent theoretical and experimental analysis elucidated that the isomer exhibiting greater planarity confers multiple advantages over its more curved counterparts. These benefits include enhanced short-range charge transfer (SR-CT), reduced reorganization energy, and increased horizontal dipole ratio (Θ//)27,28. The superior structural rigidity, optimized exciton dynamics, and improved light out-coupling efficiency of the quasi-planar compound culminated in its outstanding electroluminescent (EL) performance. The related device demonstrated a narrow FWHM of 15 nm (0.08 eV) and a high EQE of 30.4% at a practical luminance of 1000 cd m-2 without employing any additional sensitizers, marking one of the highest efficiencies ever reported for deep-blue narrowband OLEDs. Our work highlighted the importance of the topological structure in modulating the photophysical properties of novel heteroatom-doped nanographenes. These in-depth understandings of the structure-property relationship should shed light on the design of MR-TADF materials.
Molecular Design, Synthesis and Characterization
Our molecular design strategy and associated chemical structures of the emitters are illustrated in Fig. 1A. The DB with two B and three N atoms is selected as a basic structural element for constructing π-extended B,N-doped nanographenes for two reasons. First, it presents emission in the deep-blue region with fast radiative decay rate constant (kr) and moderate kRISC, and has recently been shown to be a promising motif in the construction of efficient deep-blue MR-TADF emitters29,30. Second, it can be afforded via lithium-free bora-Friedel-Crafts reaction in satisfactory yield, which ensures synthetic feasibility31. The three derived quadruple-borylated emitters, denoted as QB-U, QB-J and QB-I, are structural isomers, all of which are composed of two DB subunits fused through a phenyl linker. The nomenclature of the compounds essentially follows their molecular conformations (see discussions vide infra).
The corresponding synthetic procedures are detailed in Supplementary Information. The key precursors were prepared via consecutive Buchwald-Hartwig coupling reactions from commercially available sources (Scheme S2). The steric protection of the mesityl groups suppressed borylation at undesired reactive sites and avoided insolubilization of the products. Borylation using 56 equiv. of BBr3 at 220 °C in an autoclave afforded QB-U, QB-J and QB-I in moderate yields of 43%, 24%, 17%, respectively19. The identity and purity of the synthesized molecules were established from a combination of nuclear magnetic resonance spectroscopy (NMR), high-resolution mass spectrometry (HR-MS) and elemental analysis (EA) (Supplementary Fig. 1-39). The thermogravimetric analysis (TGA) and differential scanning calorimetry (DSC) (Supplementary Fig. 40) confirmed the thermal stability for these compounds, with 5% mass loss (Td) occurring above 511 °C, 524 °C, 556 °C, respectively, as well as glass transition temperatures (Tgs) all above 300 °C. The good thermal properties were beneficial for the vacuum-deposited OLED fabrication process, and enabled to prevent phase separation as well as crystallization to improve device stability.
Single crystals of QB-U could be cultured from chlorobenzene (Fig. 1, Supplementary Fig. 40 and Supplementary Table 1). Unfortunately, obtaining single crystals of QB-J and QB-I was unfeasible after multiple trials. The X-ray crystal structure of QB-U revealed a highly distorted, U-shaped conjugated framework, with the absence of any other conformer. Due to the two fused B,N,B-[4]helicene subunits with opposite chirality, QB-U was negatively curved into an achiral, cisoid saddle shape with the pendant mesitylene units situated on the same side of the central phenyl unit. Individual molecules of QB-U packed as dimers with π-systems separated by 3.37 Å at their closest point, slightly less than the sum of van der Waals radii (3.40 Å). In addition, multiple close C-H/π contacts were observed, which made the 3D network together with the weak π−π interactions.
To delve deeper into the geometric features and distortion elements of these compounds, we conducted density functional theory (DFT) calculations at the B3LYP/6-31G(d,p) level (Supplementary Fig. 43-45). These calculations revealed a pair of conformers for each of the three emitters, depending on the helicity of the two B,N,B-[4]helicenes. In the gas phase, the energy differences (ΔE) between conformers were found to be very small for QB-J (0.22 kcal mol-1) and QB-I (0.14 kcal mol-1), explaining the presence of two sets of signals in the corresponding 1H NMR spectrum. In contrast, for QB-U, the cisoid conformer was predicted to be favored over the transoid conformer by a higher ΔE of 1.30 kcal mol-1. This preference was supported by X-ray crystallography and only one set of signals in 1H NMR. The associated transition-state barriers for the conversion of the least-energy conformer were in the range of 23.7-25.3 kcal mol-1, comparable to carbo[5]helicene (23.9 kcal mol-1)32, indicating that each of the two conformers was interconvertible under ambient conditions. These findings suggest the necessity of considering both conformers, specifically in the case of QB-J and QB-I, where the energy discrepancy is minimal. However, we observed remarkable similarity in the overall conformations and electronic structures of the conformers despite their diastereomeric structures. Therefore, the least-energy conformer for each emitter was selected for further discussion and analysis.
As indicated in Fig. 1A, the simulated results revealed gradual planarized geometries in the sequence of QB-U, QB-J and QB-I, as suggested by a decreasing trend of the dihedral angles between the central phenyl ring C1 and C2 in the two DB fragments (∠C1-C2 = 59.6°, 32.6° and 0.5°, respectively). Such difference can be associated to the intrinsic non-planarity of the DB subunit. The mismatched bond lengths between C-N and C-B bonds caused a nearly coplanar orientation between ring C and ring A/A’ in DB, while ring B/B’ is distorted out of this plane. This conformation underlined the molecular curvature observed in QB-U. Conversely, the fusion through ring B(B’) and ring A(A’) in QB-J induced a J-shaped architecture, and the fusion through two ring A(A’) in QB-I generated a quasi-planar geometry (I-shaped). Of note, the distinctive aromatic fusion mode dictates not only the planarity but also the dimensional parameters of each molecule. The data revealed that the molecular length/width of QB-U, QB-J and QB-I were approximately 2.34/1.41, 2.61/1.60 and 2.73/1.71 times their height, respectively, with the latter achieving a configuration favorable for horizontal dipole orientation within an OLED context.
The scanning tunneling microscopy (STM) analysis of emitters thermally deposited onto Au(111)/Mica surface enabled precise visualization of their molecular structures (Fig.1B). At low submonolayer coverage, both QB-J and QB-I exhibited quartets of bright lobes, aligning with the positions of the mesitylene units. In contrast, QB-U was characterized by a central bright protrusion alongside four lobes, attributed to an upward, out-of-plane distortion of the terminal benzene rings in the B,N,B-[4]helicene subunits, indicative of a markedly nonplanar structure. The dimensions and configurations of these features concurred with those predicted by the simulations. At higher submonolayer coverage, STM imaging revealed the formation of extensive self-assembled supramolecular networks amidst irregular assemblies (Supplementary Fig. 46-48). Within these networks, QB-U formed a chevron-shaped pattern, QB-J displayed a striped pattern with bilateral symmetry, and QB-I developed a porous, honeycomb-like array. Notably, QB-I exhibited enantiomorphism with discernible chiral domains on the surface. These observations suggest that molecular topological structure profoundly influences the self-assembly process.
Electronic Structures
We subsequently engaged in theoretical analysis to elucidate the electronic structures of the emitters (Fig. 2, Supplementary Fig. 49 and Supplementary Table 2). The frontier molecular orbitals displayed an alternating pattern between the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) across different atoms, characteristic of MR-TADF molecules (Fig. 2B). Utilizing time-dependent density functional theory (TD-DFT) for both the parent molecule and its derived isomers, a slight reduction in the S1 energy level was observed across the sequence: DB (2.90 eV), QB-U (2.85 eV), QB-J (2.83 eV), to QB-I (2.80 eV), affirming the effectiveness of our design in maintaining deep-blue emission. Importantly, we noted a decrease in ΔEST from QB-U to QB-I and an increase in fosc correlated with conformational changes (Fig. 2C). The trend was also confirmed by more precise calculations at the mPW2PLYP/def2-SVP level of theory to account for electron correlation effects in double excitations33, where the ΔEST values were progressively decreased from 0.15 eV for QB-U to 0.11 eV for QB-J and 0.00 eV for QB-I (Supplementary Table 3). Given recent theoretical insights into the critical role of the high-lying triplet excited state in enabling energy upconversion in MR-TADF emitters, we also examined the close-lying T2 state34. The minimal T2-S1 energy gap and significant spin-orbit coupling (SOC) matrix element values (〈S1| SOC|T2〉) imply that RISC could be a higher-order process. Although the SOC trend was not straightforward, the cumulative values of the SOC constants for T2-S1 and T1-S1 were considerably large in all isomers. Consequently, the proximity of the S1, T1, and T2 states, along with significant SOC constants in QB-I, is anticipated to lead to a high kRISC based on the relationship: kRISC ∝ <S1|ĤSOC|Tn>²/ΔEST, which is critical for the efficient utilization of triplet excitons in EL devices.
The above observation aligns closely with the findings of Pershin et al., who reported that more delocalized SR-CT in MR-TADF frameworks should lead to smaller ΔEST and higher fosc13. Although Pershin's conclusions are based on a series of gradually extended chromophores, our findings similarly indicate increased CT delocalization from QB-U to QB-J, despite their isomeric structures. To unveil this, we calculated electronic spatial extent (<r2>) of the frontier molecular orbitals for each molecule, which describes the electron density volume as a physical property (Fig. 2D). The results revealed an increase of <r2> value for both HOMO and LUMO from QB-U to QB-I, indicating a more diffused electron cloud with a gradually planarized skeleton.
The underlying mechanism of this phenomenon is linked to the electronic structure of DB. Analysis of the frontier molecular orbital pattern reveals that the HOMO and the LUMO exhibit reduced distribution on phenyl ring B compared to ring A, attributed to steric hindrance induced out-of-plane distortion of ring B. Fig. 2A presents the isosurface map of the localized orbital locator (LOL-π) for DB that demonstrates the π-electron delocalization pathways. An isovalue of 0.37 was chosen to effectively differentiate the extent of π-electron delocalization across the rings. This visualization suggests that the interconnecting 1,4-azaborine between rings B and A shows six-center conjugation, as evidenced by the continuous isosurfaces. Conversely, the 1,4-azaborine between rings B and C displays truncated isosurfaces on the convex side, indicating restricted π-electron delocalization due to the bent configurations of these segments. As another indicator to evaluate π-electron delocalization, the π Mayer bond orders were evaluated to quantify the average number of shared π-electron pairs between bonded atoms35. The result showed that the 1,4-azaborine unit between rings B and C displayed conspicuously weakened π Mayer bond order as well as elongated bond lengths for all the C-N and C-B bonds (except for the shared one) compared to that between ring B and ring A. Such observation is in line with the FMO and LOL-π analysis, and can be understood by a simplified molecular model shown in Fig. 2A, that is, the basic MR motif of dibenzoazaborine is less prone to generate the corresponding zwitterionic resonance state upon bending. In this scenario, the central 1,4-azaborine is less aromatic, the bond orders are decreased accordingly, and the multiple resonance effect is weakened. Therefore, the pronounced curvature at the central site in QB-U and QB-J disrupts π-delocalization more than the edge distortions in QB-I, eventually leading to larger ΔEST and smaller fosc.
Photophysical Properties
The UV-vis absorption and PL spectra of the quadruple borylated emitters were recorded in dilute solutions (2 × 10−6 M) at room temperature. As presented in Fig. 3A, all the emitters manifested characteristic MR-TADF-type absorption profiles with intense absorption bands below 400 nm in toluene, denoting localized π-π* and n-π* transitions, alongside with distinctive long-wavelength absorption band attributed to SR-CT transitions. The absorption maxima (λabs) of the lowest energy band corresponding to S0→S1 transition exhibited a gradual red-shift, and the optical bandgaps progressively shrank from 2.72 to 2.68 and 2.63 eV, respectively. We note that the extinction coefficient of QB-I was measured to be 3.05 × 105 M-1 cm-1 at 455 nm (Supplementary Fig. 50), corresponding to 2-fold and 1-fold increase compared to that of QB-U and QB-J, in line with drastically increased oscillator strengths predicted theoretically.
The three isomers all displayed strong deep-blue emissions with weak positive solvatochromatic effect on the emission wavelength, reflective of the SR-CT character of their emissive excited state (Fig. 3A, Supplementary Fig. 51 and Supplementary Table 4). In toluene, the photoluminescence wavelength (λPL) was red-shifted in the sequence of QB-U (450 nm/2.76 eV), QB-J (454 nm/2.73 eV) to QB-I (464 nm/2.67 eV). The bathochromic shifts in the absorption and emission were similarly induced by promoted SR-CT within the skeleton. Nonetheless, the emission spectrum became narrower from QB-U to QB-I (FWHM of 21 nm/0.13 eV, 15 nm/0.09 eV, 13 nm/0.07 eV, respectively) and the vibronic shoulder at the long-wavelength region became weaker. The discernable enhancement in color purity was a clear consequence of the structural planarization, which minimized the change in structural displacement between S0 and S1 states and reduced molecular vibration in each of the electronic states. This point could be further elucidated by the apparent reduction of theoretically calculated reorganization energy (Λ) from QB-U (0.11 eV), QB-J (0.10 eV) to QB-I (0.06 eV) (Fig. 3B). Eventually, the net effect of these spectra changes rendered similar CIE coordinates between QB-U (0.146, 0.046) and QB-J (0.144, 0.046), and only a slight red-shift in QB-I (0.130, 0.065) (Supplementary Fig. 52). These CIE values are considered ideal for deep-blue OLEDs.
To further determine the key energy levels of both compounds, we measured their fluorescence and phosphorescence spectra in frozen toluene at 77 K (Fig. 3C, Supplementary Fig. 53). Based on the fluorescence and phosphorescence maxima, the energies of their S1 states were estimated to be 2.77, 2.74 and 2.67 eV, and those of their T1 states were estimated to be 2.58, 2.59 eV and 2.67 eV, respectively. The concurrent stabilization of S1 and destabilization of T1 from QB-U to QB-I reflected a gradual decrease in the exchange energy. Correspondingly, their ΔEST values were further computed to be 0.19, 0.15 and ~0 eV, respectively. This trend aligned with the theoretical estimations, indicating the beneficial effect of structural planarization on reducing ΔEST and thus promoting the RISC process.
We next investigated the thin film photophysical behaviors of these emitter doped in a 7-((2'-methyl-[1,1'-biphenyl]-4-yl)oxy)-3,11-di-o-tolyl-5,9-dioxa-13b-boranaphtho[3,2,1-de]anthracene (DOBNA-OAr) with 1 wt% doping ratio to alleviate the concentration-induced quenching (Fig. 4). Compared to the emission profiles in dilute solution, the solid-state PL spectra were accompanied with moderate red-shift and spectral broadening, where the FWHM values were 32 nm/0.19 eV, 19 nm/0.11 eV, 16 nm/0.09 eV for QB-U, QB-J and QB-I, respectively (Fig. 4A). This phenomenon could be tentatively attributed to the solid-state solvation effect and the presence of intermolecular interactions. The emitters presented exceptional ΦPL values of 84% for QB-U, 95% for QB-J, and 99% for QB-I. The increasing ΦPL may be attributed to enhanced molecular rigidity upon planarization as suggested by decreased reorganization energy. Fig. 4B depicts the transient PL (TR-PL) decays at room temperature. Exponential fitting of the room temperature decay revealed two different components that can be assigned as prompt fluorescence and delayed fluorescence (Supplementary Fig. 54). Importantly, despite similar prompt lifetimes (τPF) of these emitters in the range of 2.0-2.8 ns, QB-I displayed much shortened delayed lifetime (τDF) of only 0.8 μs compared to those of QB-U (21.9 μs) and QB-J (4.5 μs). We further evaluated photophysical dynamics based on these data using a previously reported method36 (Table 1). In consistency with the largest molar excitation coefficient and oscillator strength of QB-I, it presented the larger kr value (2.4×108 s-1) than that of QB-U (8.7×107 s-1) and QB-J (1.5×108 s-1). In addition, benefiting from the sub-microsecond-scale τDF, the kRISC of QB-I reached an impressive value of 2.7×106 s−1, representing one of the highest values for heavy-element-free MR-TADF emitters (Supplementary Table 6). The kRISC corresponded to a roughly 17-fold and 4-fold increase compared with that of QB-U (1.6×105 s-1) and QB-J (6.9×105 s-1). This observation can be well explained by its near-zero ΔEST value as well as the sufficiently large SOC value between the S1 and the close-lying T2 state, which contributed to the overall RISC process. As a further support, the TR-PL decay profiles at various temperatures (from 160 to 300 K) were measured (Fig. 4C). In all films, the delayed emission component decreased along with decreased temperatures, but remained apparent at 160 K, reminiscent of typical TADF phenomenon. The temperature dependency of kRISC was next analyzed using Arrhenius plots, which also pointed to a decreasing trend of activation energies (ΔEa) from the QB-U (0.12 eV), QB-J (0.09 eV) to QB-I (0.05 eV). The disparity between ΔEST and ΔEa reflected a significant contribution of the higher triplet states (such as T2) to the RISC process37,38.
Table 1. Photophysical data of the emitters.
|
Emitter
|
λabsa
[nm]
|
λema
[nm]
|
FWHMa
[nm/eV]
|
ΔESTb
[eV]
|
ΦPLc
[%]
|
τPFd
[ns]
|
τDFd
[μs]
|
kre
[107 s-1]
|
knre
[106 s-1]
|
kISCe
[107 s-1]
|
kRISCe
[105 s-1)
|
|
QB-U
|
433
|
450
|
21/0.13
|
0.19
|
84
|
2.8
|
21.9
|
8.7
|
16.6
|
2.6
|
1.6
|
|
QB-J
|
445
|
454
|
15/0.09
|
0.15
|
95
|
2.0
|
4.5
|
15.0
|
7.9
|
3.4
|
6.9
|
|
QB-I
|
455
|
464
|
13/0.07
|
~0
|
99
|
2.0
|
0.8
|
23.8
|
2.4
|
2.7
|
27.2
|
a) Peak of absorption (λabs) and fluorescence (λem, 300K) spectra, as well as color coordinates in the CIE 1931 chromaticity diagram and full-width at half-maximum (FWHM) of fluorescence; b) S1-T1 energy gap (ΔEST) determined from the onset of low-temperature fluorescence and phosphorescence spectra measured in 2×10-6 M toluene solution. c) absolute photoluminescence quantum yield (ΦPL); d) lifetimes of prompt fluorescence (τPF) and delayed fluorescence (τDF), percentage value given in parentheses refer to the fractional intensity; e) rate constants of singlet radiative decay (kr), non-radiative decay (knr), intersystem crossing (kISC), reverse intersystem crossing (kRISC) measured in 1 wt% DOBNA-OAr-doped film.
To illustrate the topological influence on transition dipole orientations, a critical factor contributing to the outcoupling efficiency in device, angle-dependent p-polarized PL spectra were obtained for the doped films (Fig. 4D). QB-J and QB-I exhibited high Θ// values of 91% and 96%, respectively, while QB-U displayed a significantly lower Θ// of 63%. This discrepancy in dipole orientation anisotropy corresponds well with the order of aspect ratio (vide supra) and simulated transition dipole moments (TDMs), which showed a remarkable decrease in the z-axis component from QB-U to QB-I, suggesting that enhanced molecular planarity promotes horizontal alignment of molecules (Supplementary Fig. 55)39.
A comprehensive comparison of five key photophysical properties in 1 wt% DOBNA-OAr-doped films relevant to EL performance is presented in the radar diagram in Fig. 4E. In this diagram, a larger outer area corresponds to improved performance across the measured properties, which include FWHM, ΦPL, kr, kRISC, and Θ//. It is evident that QB-I surpasses QB-J and QB-U in all aspects measured, emphasizing the importance of topological optimization in the design of high-performance MR-TADF materials.
Electroluminescence
Motivated by the intriguing structure-property relationship manifested by these deep-blue emitters, their potential EL performance was further assessed in devices. The OLEDs were fabricated with the following optimized configuration: indium tin oxide ITO/dipyrazino[2,3-f:2′,3′-h]quinoxaline-2,3,6,7,10,11-hexacarbonitrile (HAT-CN, 5 nm)/1,1-bis((di-4-tolylamino)phenyl)-cyclohexane (TAPC, 30 nm)/tris(4-carbazolyl-9-ylphenyl)amine (TCTA, 15 nm)/1,3-di(9H-carbazol-9-yl)benzene (mCBP, 10 nm)/99 wt% DOBNA-OAr: 1 wt% emitter (20 nm)/2,8-bis(diphenylphosphoryl)dibenzo[b,d]furan (PPF, 15 nm)/1-(4-(10-([1,1′-biphenyl]-4-yl)anthracen-9-yl)phenyl)-2-ethyl-1H-benzo[d]-imidazole (ANT-BIZ, 30 nm)/8-hydroxyquinolinolato-lithium (Liq, 2 nm)/aluminum (Al, 100 nm). HAT-CN was used as a hole injection layer, TAPC and TCTA functioned as the hole transporting layers, and ANT-BIZ was employed as electron-transport layers, respectively. mCBP and PPF served as exciton-blocking layers, and Liq and Al acted as the electron injection layer and cathode, respectively. The chemical structures of the organic materials, the energy levels diagram and the EL performance are depicted in Fig. 5, Scheme S3 and detailed parameters are provided in Table 2.
All the fabricated devices displayed nearly identical turn-on voltages (Von) of 3.7-3.8 V and high maximum luminescence (Lmax) over 10000 cd m-2 (Fig. 5D). The EL spectra remained stable at different voltages (from the turn-on voltage to 8 V) for all devices measured, indicating that the exciton recombination zone was almost consistent at different voltages (Supplementary Fig. 56). The maximum emission peaks were recorded at 453, 457 and 466 nm for QB-U, QB-J and QB-I, with the associated FWHM values of 23, 17, and 15 nm (Fig. 5B), respectively, which were consistent with the trend observed in toluene solution and doped film. Of note, such narrow FWHM value of device based on QB-I represents one of the smallest values reported for OLEDs to date. The corresponding CIE coordinates of the three devices were (0.142, 0.064), (0.141, 0.060) and (0.127, 0.078), respectively, better than the blue point (0.14, 0.08) defined by National Television System Committee (Fig. 5C). Specifically, the area ratio of harmful blue light below 450 nm in the EL spectrum of QB-I was significantly reduced to 1.5%, which was barely achievable in conventional deep-blue OLEDs with CIEy < 0.08, clearly highlighting the benefit of its ultranarrow linewidth. Furthermore, the device based on QB-I also exhibited maximum EQE and current efficiency (EQEmax and CEmax) of 32.5% and 22.2 cd A-1, respectively, without any optical out-coupling technologies, outstripping those based on QB-U (24.7%, 15.5 cd A-1) and QB-J (30.8%, 18.7 cd A-1) (Fig. 5E, Supplementary Fig. 57-58). The state-of-the-art EQE value of QB-I should stem from its superior ФPL and Θ// values.
Aside from the ultranarrow bandwidth and high EQEmax, the device incorporating QB-I also demonstrated mitigated efficiency declines at practical brightness, with EQE1000/5000 values (EQE at 1000/5000 cd m–2) of 30.4% and 26.5%, respectively. We highlight these values as record-setting for deep-blue OLEDs employing binary emitting layer (Fig. 5F, Supplementary Table 7). This performance translated to a reduction in efficiency of 6.4% and 18.5%, contrasting sharply with the 53.4%/74.9% and 32.1%/54.5% reductions observed in devices with QB-U and QB-J at the same luminance level. Analysis of the EQE-current density (EQE-J) curves revealed that efficiency roll-off in these devices was primarily driven by either triplet-triplet annihilation (TTA) or triplet-polaron annihilation (TPA), as illustrated in Supplementary Fig. 59. The reduced efficiency roll-off can thus be directly linked to the rapid depletion of triplet excitons. Acknowledging additional factors besides kRISC that influence efficiency roll-off, we introduced a recently proposed figure of merit (FOM) for efficiency roll-off, defined as 40. The estimated FOMs inclined from 4.3×104, 2.3×105 and 1.4×106 s-1 for QB-U, QB-J and QB-I, respectively. A quasi-linear relationship between the FOM values and the current density at which EQE decreases to 90% of its peak (J90) was observed in a double logarithmic plot (Supplementary Fig. 59), suggesting a strong correlation between the dynamic equilibrium of singlets and triplets and the roll-off characteristics. Importantly, the improvement in FOM from QB-U to QB-I exceeded the increase in kRISC, attributed to an elevation in kr and a marginal change in intersystem crossing rate constant (kISC). Consequently, triplet excitons in QB-I are efficiently converted to light rather than dissipating energy through multiple spin-flip cycles, minimizing energy losses.
Table 2. Summary of the key device data.
|
Device
(emitter)
|
Vona
[V]
|
λELb
[nm]
|
FWHMc
[nm/eV]
|
Lmaxd
[cd m-2]
|
CEmaxe
[cd A-1]
|
PEmaxf
[lm W-1]
|
EQEmax/1000/5000g
[%]
|
CIEh
[x, y]
|
|
QB-U
|
3.8
|
453
|
23/0.14
|
11853
|
15.5
|
12.2
|
24.7/11.5/6.2
|
(0.142, 0.064)
|
|
QB-J
|
3.7
|
457
|
17/0.10
|
16281
|
18.7
|
14.7
|
30.8/20.9/14.0
|
(0.141, 0.060)
|
|
QB-I
|
3.7
|
466
|
15/0.08
|
29526
|
22.2
|
17.0
|
32.5/30.4/26.5
|
(0.127, 0.078)
|
a) Turn-on voltage recorded at the luminance of 1 cd m-2; b) maximum EL wavelength; c) FWHM of EL spectra; d) maximum brightness; e) maximum current efficiency; f) maximum power efficiency; g) efficiency at maximum and1000 cd m-2; h) EL color coordinates in the CIE 1931 chromaticity diagram recorded at 6 V.
In assessing the operational stability, the devices based on QB-U, QB-J and QB-I exhibited half-lifetimes (LT50) of 3.0, 6.3, and 8.4 hours, respectively, at an initial luminance (L0) of 500 cd m-² (Supplementary Fig. 60). When these LT50 values are adjusted for a lower initial luminance of 100 cd m-² by employing a degradation accelerating factor (n) of 1.75, they extend to 50.6, 105.2, and 140.5 hours. The sequential enhancement in stability from QB-U to QB-I devices can be tentatively attributed to a decrease in emission lifetime coupled with a stabilization in onset emission energy. This outcome highlights the beneficial impact of structural planarization on device longevity. We posit that further optimizations of the device, including the integration of a stable bipolar host with a high T1, a reliable hole-blocking layer, and an efficient sensitizer, will significantly enhance their suitability for practical applications41-43.