Molecular asymmetry and rigidification as strategies to activate and enhance thermally activated delayed fluorescence in deep-blue MR-TADF emitters

Abstract

Two novel deep-blue multi-resonance thermally activated delayed fluorescence (MR-TADF) emitters, 1B-CzCrs and 2B-CzCrs, containing a fused carbazole unit were synthesized. The carbazole contributed to the emergence of TADF in these small molecules. Particularly, organic light-emitting diodes with 1B-CzCrs doped in the mCP host achieve a maximum external quantum efficiency of 12.8% at CIE coordinates of (0.146, 0.062).

---

Organic thermally activated delayed fluorescence (TADF) materials have been steadily developed over the past decade for organic light-emitting diodes (OLEDs) due to their ability to convert 100% of the electrically generated excitons to photons without using scarce metals, garnering great interest within the display community.^1,2 Typical TADF emitters consist of donor (D) and acceptor (A) moieties connected in a twisted conformation to result in a minimal orbital overlap between the highest occupied and lowest unoccupied molecular orbitals (HOMO and LUMO, respectively). This provides a small energy splitting (ΔE_ST) between the lowest singlet (S₁) and triplet (T₁) excited states.³ However, such a long-range charge transfer (LRCT) character of these D–A TADF materials leads to a broad emission spectrum (full width of half maximum, FWHM, >45 nm) even in compounds where conformational motion is restricted.⁴ Broad emission is undesired for colour saturation. In 2016, Hatakeyama et al. reported a new class of TADF materials, so-called multi-resonant TADF (MR-TADF) compounds that typically have a narrowband emission (FWHM of <30 nm).⁵ These compounds contain electron-rich and electron-poor dopants disposed meta or para with respect to each other within a polycyclic aromatic compound. Such compounds emit from a short-range charge transfer (SRCT) state associated with a characteristically small ΔE_ST.^6,7 MR-TADF materials not only exhibit narrowband emission but also have high photoluminescence (PL) quantum yields (Φ_PL) linked to fast radiative decay rate constants.

On the other hand, examples of deep-blue MR-TADF emitters satisfying the industry standard for high-definition displays i.e., the BT.2020 standard are still limited; indeed, the blue devices must emit at Commission Internationale de l’Éclairage (CIE) coordinates of CIE_x,y = (0.131, 0.046).^8–10 In the literature, the strategy of fusing aromatic tertiary amines with boron atoms has been widely used to construct deep-blue MR-TADF emitters. This approach has also been modified by replacing the nitrogen atoms with other heteroatoms, and/or by replacing the often used diphenylamine (DPA) moieties with carbazole (Cz) to achieve color and emissivity control.^6,7 Previously, we reported the deep blue emitters 1B-DTACrs and 2B-DTACrs (Table S1, ESI†) wherein the devices emit with CIE_y ≤ 0.05.¹¹1B-DTACrs showed near unity Φ_PL of 99% but was not TADF, while 2B-DTACrs was TADF, with a fast RISC rate constant (k_RISC ∼ 10⁵); however, the Φ_PL was lower at 64%. In this study, we replaced the DTA with Cz and designed and synthesized two deep-blue MR-TADF emitters, 1B-CzCrs and 2B-CzCrs (Table S1, ESI†). The optoelectronic properties of the two deep-blue MR-TADF emitters were studied in detail, supported by computations. Interestingly, compared to the previously reported DPA-based 1B-DTACrs, the Cz-containing 1B-CzCrs is TADF, as is 2B-CzCrs. 1B-CzCrs and 2B-CzCrs emit at λ_PL of 446 and 441 nm, with FWHM of 25 and 19 nm in toluene, respectively and the asymmetric 1B-CzCrs showed better TADF properties.

The detailed synthesis procedure, characterization, and electrochemical data of 1B-CzCrs and 2B-CzCrs are provided in ESI† (Fig. S1–S19 and Table S2). Thermal gravimetric analysis (TGA) has been carried out for 1B-CzCrs and 2B-CzCrs under a nitrogen atmosphere to determine the decomposition temperature (T_d). The T_d values of both emitters were found to be higher than 400 °C, suggesting their high thermal stability (Fig. S20, ESI†).

The ground-state and excited-state geometries of 1B-CzCrs and 2B-CzCrs were calculated using Density Functional Theory (DFT) and the post Hartree–Fock methods, respectively. The HOMO–LUMO energy gaps for 1B-CzCrs and 2B-CzCrs were calculated as 3.88 eV and 3.77 eV, respectively (Fig. 1). These values are smaller than those reported for 1B-DTACrs and 2B-DTACrs (4.03 and 3.91 eV) due to the increased conjugation that results from the incorporation of the fused carbazole moiety (Table S3, ESI†). This trend supports the experimentally obtained values, which are 3.13 and 3.10 eV for 1B-CzCrs and 2B-CzCrs, respectively, and 3.17 and 3.05 eV for 1B-DTACrs and 2B-DTACrs, respectively (Table S2, ESI†), especially for the monoborylated counterparts. The oscillator strength (f) of the S₀–S₁ transition of 1B-CzCrs is larger (0.27) than that of 2B-CzCrs (0.16), while the ΔE_ST are essentially the same (0.16 and 0.17 eV) (Table S4, ESI†). The HOMO–LUMO transition in 1B-CzCrs is purely SRCT, while in 2B-CzCrs it exhibits more of LRCT character, which is reflected in the lower f in the latter. Analysis of the excited-state difference density plots confirmed that the SRCT S₁ state in 1B-CzCrs is extended over more of the compound compared to 2B-CzCrs. The nature of the T₁ state in both molecules is essentially the same as the respective S₁ states, yielding comparable ΔE_ST in both molecules. More notably, there are significant differences in the nature of the T₂ state, with T₂ of 1B-CzCrs having more of a LE character on the Cz moiety, while that of 2B-CzCrs still retains SRCT character, similar to that of S₁ state. This results in a considerably higher spin–orbit coupling (SOC) value in 1B-CzCrs compared to that in 2B-CzCrs (Fig. 1).

Fig. 1 Theoretical calculation results for 1B-CzCrs (a) and (c) and 2B-CzCrs (b) and (d); (a) and (b) HOMO and LUMO distributions at the PBE0/6-31G** level of theory (gas phase), (b) and (d) excited-state difference density plots of S₁, T₁, and T₂ and those energies at the excited-state geometries at the SCS-ADC(2)/cc-pVDZ level of theory. f values denote the oscillator strength.

Fig. 2 shows the absorption and emission spectra of 1B-CzCrs and 2B-CzCrs in toluene (1.0 × 10⁻⁵ mol dm⁻³). There are two strong absorption bands, one below 350 nm and the other at around 400 nm that can be attributed to π–π* locally-excited (LE) and SRCT transitions, respectively. The structured absorption bands at 405 and 431 nm of 2B-CzCrs can be attributed to S₀–S₂ and S₀–S₁ transition, respectively (Fig. 1a). As predicted by the calculated f values, there is a stronger S₀–S₁ absorption in 1B-CzCrs than in 2B-CzCrs. 1B-CzCrs and 2B-CzCrs emit narrowly in the deep-blue region with λ_PL of 446 and 441 nm (FWHM of 24 and 19 nm), respectively (Table 1). The corresponding CIE_y were <0.04, indicating a high potential for realizing deep blue OLEDs having high colour purity. Because the MR-TADF core has been rigidified by introducing the Cz moiety, the emissions of 1B-CzCrs and 2B-CzCrs are narrower than those of 1B-DTACrs and 2B-DTACrs (27 and 21 nm).¹¹ The ΔE_ST was estimated from the difference in the onsets of the fluorescence and phosphorescence spectra after applying the Jacobian conversion at 0.21 and 0.24 eV for 1B-CzCrs and 2B-CzCrs, respectively (Fig. S21, ESI†).¹² These values agree well with the theoretical ones calculated at the SCS-ADC(2)/cc-pVDZ level of theory (0.16 and 0.17 eV, respectively).¹³ The Φ_PL values of 1B-CzCrs and 2B-CzCrs in the air are 63 and 55%, respectively, and these significantly increased to 96 and 94% under N₂. This indicates the large contribution of triplets to their light-emitting process. The high Φ_PL values can be attributed to their highly rigid structure, resulting in reduced nonradiative decay. However, only a short decay lifetime component was observed in the transient PL for both 1B-CzCrs and 2B-CzCrs in toluene (Fig. S22 and S23, ESI†). This is because, despite the high Φ_PL, non-radiative decay competes with RISC. The lifetime slightly increases under N₂, indicating that oxygen may affect the singlet radiative decay rates.¹⁴

Fig. 2 Absorption (black line), Steady-state (SS) PL at room temperature (RT, blue line), and at 77 K (green dashed line), and phosphorescence (Phos.) at 77 K (red line, delay: 115 ms, gate: 150 ms) for (a) 1B-CzCrs and 2B-CzCrs (b) in toluene. λ_exc = 310 nm.

Table 1 Photophysical data of 1B-CzCrs and 2B-CzCrs in toluene and mCBP

In toluene									5 wt% in mCBP
	Φ PL (%)	τ p (ns)	λ PL (nm)	FWHM (nm)	S1^c (eV)	T1^d (eV)	ΔEST (eV)	CIEx,y	Φ PL (%)	τ p (ns)	τ d (ms)	λ PL (nm)	FWHM (nm)	k r (10^7 S^−1)	k nr (10^7 s^−1)	k ISC (10^7 s^−1)	k RISC (10^3 s^−1)
a Φ PL were measured using an integrating sphere at λexc = 310 nm under N2-saturated and aerated (in parentheses) conditions. b τ p was measured with a streak camera under N2-saturated and aerated (in parentheses) conditions at λexc = 355 nm. c Estimated from the onset wavelength of the steady-state PL spectrum at 77 K. d Estimated from the onset wavelength of the delayed emission spectrum (115–150 ms) at 77 K (λexc = 310 nm). e Rate constants were calculated using the reported method in the literature^15 under the approximation of no phosphorescence contributed to the delayed emission.
1B-CzCrs	96 (63)	6.57(5.37)	446	24	2.89	2.68	0.21	0.150, 0.037	67	5.08	0.33	451	28	11.90	5.86	1.88	3.36
2B-CzCrs	94 (55)	8.70(6.52)	441	19	2.91	2.67	0.24	0.142, 0.032	59	8.63	2.98	449	28	5.80	4.03	1.75	0.41

---

On the other hand, a clear delayed decay was observed for both 1B-CzCrs and 2B-CzCrs in a 3,3′-di(9H-carbazol-9-yl)-1,1′-biphenyl (mCBP) host at 5 wt% doping concentration (Fig. 3). It is worth noting that 1B-DTACrs exhibited no TADF behavior, while 1B-CzCrs showed delayed emission in host matrices. The S₁ and T₁ levels of the 5 wt% doped film in mCBP are the same as those in toluene, but the PL spectra in mCBP are slightly broader, with FWHM of 28 nm for each (Table S5, ESI†). The Φ_PL under Ar of 67 and 59% are lower than those in toluene. This behavior was also observed in 1B-DTACrs and 2B-DTACrs, but the degree of spectral broadening and the Φ_PL drop in 1B-CzCrs and 2B-CzCrs are more pronounced. These data suggest that these planar Cz-containing emitters more easily form either aggregates and/or intermolecular interactions with the host in their film state, despite the introduction of tert-butyl groups in the skeleton of these compounds in addition to the bulky o-tolyl group. This was also confirmed by the narrowing of the emission spectra and the higher Φ_PL at lower doping concentrations (Fig. S24, S25, and Table S6, ESI†).

Fig. 3 Photophysical properties of 5 wt% 1B-CzCrs and 2B-CzCrs doped in mCBP; (a) normalized absorption (dashed lines), SS PL at RT (solid lines, λ_exc = 310 nm), and phosphorescence at 77 K (dotted lines, λ_exc = 310 nm), (b) transient PL decays (λ_exc = 355 nm) at RT.

Notably, the delayed emission lifetimes of 1B-CzCrs and 2B-CzCrs (0.33 and 2.98 ms, respectively) were significantly longer than that of the 2B-DTACrs (13.1 μs).¹¹ The estimated rate constants of radiative and nonradiative decays from S₁ (k_r, and k_nr, respectively), intersystem crossing (k_ISC), and k_RISC are summarized in Table 1.¹⁵ The k_RISC values for 1B-CzCrs and 2B-CzCrs are 3.36 and 0.41 × 10³ s⁻¹, respectively; these are a two order difference with that of 2B-DTACrs (1.3 × 10⁵ s⁻¹). There is also a one-order difference in k_RISC between 1B-CzCrs and 2B-CzCrs. To help explain this difference, SOC magnitude was investigated with PBE0/6-31g(d,p) level of theory in the gas phase (Table S7, ESI†). The SOC value between T₁ and S₁ for 1B-CzCrs is 0.1187 cm⁻¹, a two-fold higher value compared with that of 2B-CzCrs (0.0592 cm⁻¹). The SOC values between S₁ and T₂ also reflect this same trend at 0.0995 cm⁻¹ and 0.060 cm⁻¹ for 1B-CzCrs and 2B-CzCrs, respectively. This in itself explains the faster k_RISC in 1B-CzCrs compared with 2B-CzCrs. Further, considering the energy splitting between S₁–T₁ and S₁–T₂ (0.33 and 0.12 eV for 1B-CzCrs, 0.40 and 0.08 eV for 2B-CzCrs, respectively), it is obvious why the asymmetric 1B-CzCrs has a faster k_RISC than the symmetric 2B-CzCrs. From these results, the replacement of the DPA by Cz in this emitter design has both pros and cons, i.e., turning on TADF yet slowing k_RISC. It is however noteworthy that the asymmetry in 1B-CzCrs leads to faster k_RISC than in 2B-CzCrs while this was not the case with the 1B-DTACrs/2B-DTACrs family.

Temperature-dependent transient emission decay profiles for 5 wt% doped films of 1B-CzCrs and 2B-CzCrs under vacuum demonstrate that the delayed emission components of both emitters are thermally activated yet become shorter with increasing temperature (Fig. S26, ESI†). The activation energy for the ISC (E^ISC_a) and RISC (E^RISC_a) processes and the effective SOCME was estimated from the intercept (Fig. S27 and Table S8, ESI†).¹⁰ The experimentally estimated ΔE_ST, that is E^RISC_a − E^ISC_a, was larger for 1B-CzCrs (65.9 meV) than for 2B-CzCrs (19.9 meV). However, the significantly larger effective SOC strength of 1B-CzCrs (0.0181 and 0.0173 cm⁻¹ for k^S_nr = 0 and k^T_nr = 0, respectively) compared to 2B-CzCrs (0.0029 and 0.0027 cm⁻¹ for k^S_nr = 0 and k^T_nr = 0, respectively) explains the overall enhanced TADF in the former compared to the latter. This provides clear evidence of the contribution of the T₂ state to the RISC process of 1B-CzCrs.

We also investigated the film state photophysics of 1B-CzCrs and 2B-CzCrs in a 1,3-di(9H-carbazol-9-yl)benzene (mCP) host matrix. Unlike in mCBP, both emitters exhibited slightly broadened and red-shifted emission in mCP, even at 1 wt% doping concentrations (Fig. S28, S29, and Table S9, ESI†). This can be ascribed to the more polar nature of mCP, which has a higher dipole moment (1.35 D), compared to mCBP (0.8 D).^16,17 Nonetheless, suppressed aggregation formation was observed upon increasing the doping concentration for both emitters in mCP compared to the emitters in mCBP (Fig. S24 and S28, ESI†). Stabilizing the excited states of individual MR-TADF dopant molecule within the mCP host matrix by relatively stronger dipole–dipole interactions and limiting their aggregation can explain this observation. However, a reduction in Φ_PL was observed for the 1 wt% doped film of 1B-CzCrs in both hosts, due to an insufficient energy transfer between the hosts and the emitter (Fig. S30, ESI†).

To investigate the electroluminescence (EL) properties of 1B-CzCrs and 2B-CzCrs, TADF-OLEDs were fabricated by vacuum deposition; the OLED with 1B-DTACrs was also fabricated as a reference. The following device structure using mCP as the host provided the best maximum external quantum efficiency (EQE_max): indium-tin-oxide (ITO, 100 nm)/HAT-CN (10 nm)/TAPC (30 nm)/mCP (10 nm)/mCP: 5 wt% of 1B-CzCrs, 2B-CzCrs or 1B-DTACrs (20 nm)/PPT (10 nm)/TmPyPB (30 nm)/Liq (2 nm)/Al (100 nm). All the device data are depicted in Fig. 4 and Fig. S31, S32 (ESI†), Table 2 and Table S10 (ESI†). The OLED with 1B-CzCrs exhibited a higher EQE_max of 12.8% than the device with 2B-CzCrs (EQE_max of 8.4%). The EL of 1B-CzCrs and 2B-CzCrs was observed at λ_EL of 451 and 450 nm, similar to the corresponding PL emission in mCP. Both devices demonstrated a relatively low turn-on voltage below 4.0 V. Furthermore, the device with 1B-CzCrs showed a much higher EQE_max of 12.8% than that with 1B-DTACrs (2.1%). This, again, confirms that TADF is operational in the former and not the latter. To investigate the molecular orientation of both emitters, angle-dependent PL measurements were performed on the 5 wt% doped emitters in mCP films (Fig. S33, ESI†). Given that the horizontal-dipole ratios (Θ) values of 0, 0.33, and 1 correspond to the perfectly horizontal, isotropic, and vertical orientations, respectively, the obtained Θ values for 1B-CzCrs (0.43) and 2B-CzCrs (0.50) indicate that the transition dipole moment of both emitters adopts a slightly vertical orientation in mCP, reducing the light-outcoupling efficiencies of both emitters below 20% (Table S11, ESI†). The theoretical EQE values estimated for the devices with 1B-CzCrs and 2B-CzCrs were 13.3 and 9.5%, respectively, and were in agreement with the experimental data.

Fig. 4 Device performance with 5 wt% 1B-CzCrs and 2B-CzCrs in mCP; (a) EL spectra at 100 cd m⁻², and (b) EQE-L profile.

Table 2 Device performance of 5 wt% 1B-CzCrs and 2B-CzCrs in mCP

	V on (V)	EQEmax^b (%)	EQE100/500 (%)	λ EL (nm)	CIEx,y
a Turn-on voltages were estimated as 1 cd m^−2. b Maximum external quantum efficiencies were estimated at 1 cd m^−2.
1B-CzCrs	3.8	12.8	3.3/0.9	451	0.146, 0.062
2B-CzCrs	3.7	8.4	2.9/1.7	450	0.153, 0.143

---

The PL emissions of 1 wt% doped film of 1B-CzCrs and 2B-CzCrs in mCBP host have CIE coordinates closer to the BT.2020 standard for blue (Table S6, ESI†). Although the OLEDs with 1 wt% doped 1B-CzCrs and 2B-CzCrs in mCBP host gave CIE_y < 0.05, the EQE_max values were inferior compared to the devices using mCP as the host (Fig. S34 and Table S12, ESI†). This can be ascribed to a charge carrier imbalance that is reflected in the higher turn-on voltage of these devices compared to those using mCP as the host.

In conclusion, two deep-blue emitters, 1B-CzCrs and 2B-CzCrs, were developed to assess the impact of incorporating a Cz moiety within this small MR-TADF core. The Cz moiety contributed to turning on TADF in both emitters, but slowed the k_RISC, especially in the case of the symmetric 2B-CzCrs. Nonetheless, we present here a comparative example that shows superior photophysical and device performance for the asymmetric design of 1B-CzCrs, which we believe is a promising strategy that can be applied in future designs of MR-TADF emitters. Despite the introduction of several large steric hindrance units, these two emitters were prone to aggregate. The poorer device efficiencies in mCP could be rationalized due to the slightly vertical dipole orientation of the emitters in this host. Nonetheless, the device with 1B-CzCrs in mCP could almost realize the BT.2020 for blue with CIE of (0.146, 0.06) along with a moderate EQE of 12.8%, a significant improvement over the device with 1B-DTACrs.

The authors acknowledge Ms N. Nakamura and Ms K. Kusuhara for their technical assistance. This work was supported financially by Kyulux Inc, JSPS Grant-in-Aid for Specially Promoted Research (23H05406), JST CREST (JPMJCR22B3), JSPS KAKENHI (23K20039), ESPRC (EP/R035164/1, EP/W007517/1), City University of Hong Kong (Project No. 9610637 and 9231531).

Data availability

This research data supporting this publication can be accessed at https://doi.org/10.17630/2fd8e197-e110-4fe2-ba94-2471629aaa09.

Conflicts of interest

There are no conflicts to declare.

References

1. H. Uoyama, K. Goushi, K. Shizu, H. Nomura and C. Adachi, Nature, 2012, 492, 234–238.
2. M. Y. Wong and E. Zysman-Colman, Adv. Mater., 2017, 29, 1605444.
3. A. Endo, K. Sato, K. Yoshimura, T. Kai, A. Kawada, H. Miyazaki and C. Adachi, Appl. Phys. Lett., 2011, 98, 083302.
4. Y. J. Cho, S. K. Jeon, S.-S. Lee, E. Yu and J. Y. Lee, Chem. Mater., 2016, 28, 5400–5405.
5. T. Hatakeyama, K. Shiren, K. Nakajima, S. Nomura, S. Nakatsuka, K. Kinoshita, J. Ni, Y. Ono and T. Ikuta, Adv. Mater., 2016, 28, 2777–2781.
6. H. J. Kim and T. Yasuda, Adv. Opt. Mater., 2022, 16, 2201714.
7. M. Mamada, M. Hayakawa, J. Ochi and T. Hatakeyama, Chem. Soc. Rev., 2024, 53, 1624–1692.
8. R. W. Weerasinghe, S. Madayanad Suresh, D. Hall, T. Matulaitis, A. M. Z. Slawin, S. Warriner, Y. Lee, C. Chan, Y. Tsuchiya, E. Zysman-Colman and C. Adachi, Adv. Mater., 2024, 36, 2402289.
9. J. Ochi, Y. Yamasaki, K. Tanaka, Y. Kondo, K. Isayama, S. Oda, M. Kondo and T. Hatakeyama, Nat. Commun., 2024, 15, 2361.
10. J. Bae, M. Sakai, Y. Tsuchiya, N. Ando, X.-K. Chen, T. B. Nguyen, C.-Y. Chan, Y.-T. Lee, M. Auffray, H. Nakanotani, S. Yamaguchi and C. Adachi, Front. Chem., 2022, 10, 990918.
11. C.-Y. Chan, S. Madayanad Suresh, Y.-T. Lee, Y. Tsuchiya, T. Matulaitis, D. Hall, A. M. Z. Slawin, S. Warriner, D. Beljonne, Y. Olivier, C. Adachi and E. Zysman-Colman, Chem. Commun., 2022, 58, 9377–9380.
12. J. Mooney and P. Kambhampati, J. Phys. Chem. Lett., 2013, 4, 3316–3318.
13. D. Hall, J. C. Sancho-García, A. Pershin, G. Ricci, D. Beljonne, E. Zysman-Colman and Y. Olivier, J. Chem. Theory Comput., 2022, 18, 4903–4918.
14. N. Notsuka, H. Nakanotani, H. Noda, K. Goushi and C. Adachi, J. Phys. Chem. Lett., 2020, 11, 562–566.
15. Y. Tsuchiya, S. Diesing, F. Bencheikh, Y. Wada, P. L. dos Santos, H. Kaji, E. Zysman-Colman, I. D. W. Samuel and C. Adachi, J. Phys. Chem. A, 2021, 125, 8074–8089.
16. F. Rodella, S. Bagnich, E. Duda, T. Meier, J. Kahle, S. Athanasopoulos, A. Köhler and P. Strohriegl, Front. Chem., 2020, 8, 657.
17. S. Ihn, D. Jeong, E. S. Kwon, S. Kim, Y. S. Chung, M. Sim, J. Chwae, Y. Koishikawa, S. O. Jeon, J. S. Kim, J. Kim, S. Nam, I. Kim, S. Park, D. S. Kim, H. Choi and S. Kim, Adv. Sci., 2022, 9, 2102141.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4cp02664k

---