Theoretical calculations and h-BNCO molecular-design strategy
To first provide a fundamental understanding of the electronic structure of our designed h-BNCO-1 and the conventional BNCZ, quantum-chemistry calculations were carried out. Previous theoretical investigations have demonstrated that the electron correlation effect is important in MR molecules, and popular DFT approaches cannot provide an exact description of the energies and characteristics of their excited states.21-23 Here, the high-level Coupled-Cluster quantum-chemistry approach was exploited to examine the electronic-structure properties of the excited states (for details, see the “Computational Details” section).
As shown in Figure 1, the conventional BNCZ with a small distorted angle (~ 8.4°) between carbazole (CZ) and DABNA fragments has a ΔES1T1 of 0.2 eV and a ΔET2T1 of 0.26 eV. The SOC between the S1 and T1/T2 states is small, ~0.07/0.29 cm-1. The combination of small SOC values, large ΔES1T1, and large ΔET2T1 thus leads to a slow kRISC. Compared with BNCZ, h-BNCO-1 shows a much more distorted geometry, with a dihedral angle of ~ 25° between the acridone and DABNA groups. Due to the electron-withdrawing effect of the carbonyl groups and the electronic delocalization on the carbonyl groups, the S1- and T1-state energies both decrease, with a reduced ΔES1T1 of 0.12 eV. Moreover, ΔET2T1 decreases to 0.18 eV, which is important to accelerating the RISC process. The decrease in ΔET2T1 (i.e., also the energy splitting between the triplet states of the two same fragments in h-BNCO-1) could be induced by its highly distorted geometry since the electronic communication between the two same fragments is blocked. The SOC(S1-T1) is negligible, while the SOC(S1-T2) is very large, ca. 1.26 cm-1. This difference in their SOC values is mainly due to the difference in the excited-state characteristics, since the total angular momentum for the electrons must be conserved in any spin-flip event. The T1 and S1 states both mainly show a similar ππ* excitation character, although on the carbonyl groups the T1 state shows a mixing of ππ* and nπ* (see the region circled by the blue-dashed line in Figure 1d and Supplementary Fig. 1). In contrast to the T1 state, the ππ* excitation character of the T2 state on the BN main body shows an apparent difference with that of the S1 state, as highlighted by the black-dashed line in Figure 1d. Moreover, the T2 state on the carbonyls shows only nπ* excitation (see the region circled by the green-dashed line in Figure 1d), which is different from the ππ* excitation character of the S1 state on the groups. Therefore, such a combination of small ΔES1T1, small ΔET2T1, and large SOC(S1-T2) gives rise to a fast Boltzmann-averaged kRISC value of ~ 1.4×106 s-1.
To further demonstrate the advantage of our newly proposed h-BNCO molecular-design strategy, we replaced the two carbonyls of h-BNCO-1 with the electron-withdrawing phenyl-boron groups and thus designed a new BN-MR molecule (named BNBNB) without the carbonyls. The results of our theoretical calculations on BNBNB are also shown in Figure 1 and the corresponding data are summarized in Supplementary Table 1. Clearly, compared with BNCZ, the introduction of phenyl-boron groups also leads to a smaller ΔES1T1 (0.08 eV) and ΔET2T1 (0.18 eV), similar to h-BNCO-1. However, its SOC value (0.69 cm-1) between the S1 and T2 states is smaller than that in h-BNCO-1, due to the absence of the nπ* excitation in the T2 state. Eventually, the Boltzmann-averaged kRISC value (~ 2.62×104 s-1) in BNBNB is two orders of magnitude slower than that in h-BNCO-1. In addition, a few examples of nitrogen-carbonyl-conjugated frameworks were reported to show MR-TADF properties. However, their device performances were less ideal, showing broader spectral bands and poor efficiencies with serious efficiency roll-offs.24-33 In short, our calculation results have thus demonstrated that the newly proposed h-BNCO strategy not only reduces ΔES1T1 and ΔET2T1 but also increases the SOC values, eventually substantially accelerating the RISC process.
Photophysics
UV–Vis absorption and PL spectra of BNCZ and h-BNCO-1 were recorded in dilute toluene (1.0×10-5 M) at room temperature. As shown in Figure 2a, BNCZ and h-BNCO-1 both display strong and sharp absorption bands at 458 and 489 nm, respectively. The PL spectrum of BNCZ shows a maximum at 477 nm with an FWHM of 0.13 eV (23 nm); in contrast, that of h-BNCO-1 shows an evident bathochromic shift to 516 nm due to the incorporation of carbonyl groups, while importantly, its FWHM is well maintained at 0.13 eV (corresponding to 28 nm).
To further determine the key energy levels of both compounds, we measured their fluorescence and phosphorescence spectra in frozen toluene at 77 K (shown in Supplementary Fig. 2 and the data are summarized in Table 1). Based on the fluorescence and phosphorescence maxima of BNCZ and h-BNCO-1, the energies of their S1 states are estimated to be 2.567 and 2.385 eV, and those of their T1 states are estimated to be 2.460 and 2.362 eV, respectively. Their ΔES1T1 values were further computed to be 107 and 23 meV, respectively. This indicates that the introduction of carbonyl groups substantially reduces ΔES1T1, thus benefitting the RISC process, which is consistent with our theoretical calculation results.
We then prepared 1 wt% BNCZ / h-BNCO-1 doped in 5-(3-(4,6-diphenyl-1,3,5-triazin-2-yl)phenyl)-7,7-dimethyl-5,7-dihydroindeno[2,1-b]carbazole (DMIC-TRZ) films. The ΦPLs of the BNCZ- and h-BNCO-1-based films were evaluated to be 84 and 100%, respectively. Figure 2b and Supplementary Fig. 3 depict their transient PL decays at room temperature. Both films show obvious delayed components due to the TADF process, while the delayed contributions are obviously different. With the incorporation of carbonyl groups, the delayed fluorescence quantum yields are greatly enhanced from 20% for BNCZ to 94% for h-BNCO-1. We further evaluated photophysical dynamics based on these data (summarized in Supplementary Table 2). Importantly, the kRISC of h-BNCO-1 reaches an impressive value of 1.08 × 106 s-1, which is much faster than that of BNCZ (3.12 × 104 s-1). The kRISC of h-BNCO-1 is among the fastest values in MR-TADF emitters and comparable to those in the Se-integrated MR compounds. The substantially accelerated RISC can be well explained by our theoretical estimation results. The higher-lying T2 state in h-BNCO-1 shows a larger SOC with the S1 state due to carbonyl incorporation, which contributes to the overall RISC process. For OLEDs, the substantially enhanced RISC process can not only enhance triplet utilization but also reduce triplet-polaron/triplet annihilation at high exciton densities, which can give rise to high peak efficiency, suppressed efficiency roll-off at high current densities as well as operational stability.
Table 1. Photophysical properties of BNCZ and h-BNCO-1.
Compound
|
λabsa
[nm]
|
λema
[nm]
|
FWHMb
[nm]/[eV]
|
ES1c
[eV]
|
ET1c
[eV]
|
ΔES1T1d
[meV]
|
BNCZ
|
458
|
477
|
23/0.13
|
2.567
|
2.460
|
107
|
h-BNCO-1
|
489
|
516
|
28/0.13
|
2.385
|
2.362
|
23
|
a Peak wavelengths of absorption and fluorescence spectra in toluene. b Full-widths at half-maxima. c Lowest excited singlet (S1) and triplet (T1) energy levels, respectively, estimated from fluorescence and phosphorescence spectra in dilute toluene at 77 K. d ΔES1T1 = S1 – T1.
OLED performances and stability
Finally, to compare their respective electroluminescence (EL) performances, OLED devices with a binary EML composed of only host and emitters were fabricated by employing BNCZ / h-BNCO-1 as emitter, with an optimized device configuration of ITO (indium tin oxide)/HAT-CN (1,4,5,8,9,11-hexaazatriphenylene hexacarbonitrile, 7 nm)/TAPC (N,N-bis(p-tolyl)aniline, 30 nm)/TCTA (tris(4-carbazoyl-9-ylphenyl)amine, 10 nm)/mCBP (3,3-di(9H-carbazol-9-yl)biphenyl, 10 nm)/DMIC-TRZ: 1 wt% emitter (20 nm)/TmPyPB (3,3′-[5′-[3-(3-pyridinyl)phenyl] [1,1′:3′,1′′-terphenyl]3,3′′diyl]bispyridine, 40 nm)/LiF (1 nm)/Al. The corresponding energy level diagram of the devices is displayed in Figure 3a.
The device performances are summarized in Supplementary Table 3 and shown in Supplementary Fig. 4. As shown in Figure 3b, the BNCZ-based device shows a sky-blue EL spectrum with a peak at 484 nm and a FWHM of 34 nm (178 meV), and the h-BNCO-1-based device shows a green EL spectrum with a peak at 528 nm, an FWHM of 39 nm (173 meV) and a nearly Lambertian emission pattern (Supplementary Fig. 5a). With an appropriate EL peak as well as a narrow spectral bandwidth, the h-BNCO-1-based device finally operates at high-quality green CIE coordinates of (0.24, 0.71), which is very close to the CIE coordinates of (0.21, 0.71) in the green standard defined by the National Television System Committee (NTSC) (see Supplementary Fig. 5b). Moreover, the device based on h-BNCO-1 exhibits a maximum EQE of 40.1%, which is evidently superior to the BNCZ-based device (12.1%) and among the highest EQEs in OLEDs. Importantly, at an initial brightness of 1000 cd m-2, the h-BNCO-1-based device still maintains a decent EQE of 34.6%, corresponding to a relative roll-off of 14%. Such efficiency roll-off behaviour in the h-BNCO-based device is superior to not only the BNCZ-based device but also most of the ever-reported OLEDs exploiting binary EMLs with MR-emitters (see Supplementary Table 4). These device results further demonstrate the importance of h-BNCO in accelerating the RISC process and relieving exciton quenching at high current densities. The h-BNCO-1-based device simultaneously achieves high EQE, low efficiency roll-off, and high colour purity, which apparently surpasses the reported OLEDs.
Traditionally, the severe efficiency roll-off and poor operational stability of TADF-based devices could be mainly ascribed to triplet accumulation, which originates from the slow RISC process from the triplet to the singlet excited state. To further validate the benefits of the accelerated RISC process, operational EL stability based on h-BNCO-1 was subsequently evaluated. Herein, organic functional materials TAPC, TCTA, and TmPyPB with relatively fragile glass transition temperatures (79, 155, and 79 °C, respectively) were replaced by their more stable analogous N,N′-di(1-naphthyl)-N,N′-diphenyl-(1,1′-biphenyl)-4,4′-diamine (NPD, 95℃), 9-phenyl-3,6-bis(9-phenyl-9Hcarbazol-3-yl)-9H-carbazole (TrisPCz, 163℃), and 2-(9,9'-spirobi[fluoren]-3-yl)-4,6-diphenyl-1,3,5-triazine (SF3-TRZ, 135 °C).34 As shown in Supplementary Figure 6, the optimized device with a binary EML exhibited a low turn-on voltage of 2.5 V at a luminance of 1 cd m−2, a high EQE of 25.0%, and pure-green narrowband emission with CIE coordinates of (0.26, 0.70). Significantly, benefitting from the high kRISC (>1×106 s−1) of h-BNCO-1 and the device structure, the lifetime reaching 95% of the initial luminance (LT95, starting from 1000 cd m−2) was measured around 140 hours (Figure 4a). To our knowledge, this operational lifetime is one of the best performances in the reported MR-OLED based on a binary EML (Figure 4b). Additionally, our device stability is also comparable to that of its ternary-EML analogous via extra assistance by TADF molecules or phosphors (see Supplementary Figure 7 and Supplementary Table 5). Therefore, the outstanding performance and stability obtained via a simple binary-EML structure composed of only pure organics indicates the potential for future low-cost commercialization.