Rational Design of π-Conjugated Tricoordinated Organoboron Derivatives With Thermally Activated Delayed Fluorescent Properties for Application in Organic Light-Emitting Diodes

A series of donor–acceptor (D–A) tricoordinated organoboron derivatives (1–10) have been systematically investigated for thermally activated delayed fluorescent (TADF)-based organic light-emitting diode (OLED) materials. The calculated results show that the designed molecules exhibit small singlet-triplet energy gap (ΔEST) values. Density functional theory (DFT) analysis indicated that the designed molecules display an efficient separation between donor and acceptor fragments because of a small overlap between donor and acceptor fragments on HOMOs and LUMOs. Furthermore, the delayed fluorescence emission color can be tuned effectively by introduction of different polycyclic aromatic fragments in parent molecule 1. The calculated results show that molecules 2, 3, and 4 possess more significant Stokes shifts and red emission with small ΔEST values. Nevertheless, other molecules exhibit green (1, 7, and 8), light green (6 and 10), and blue (5 and 9) emissions. Meanwhile, they are potential ambipolar charge transport materials except that 4 and 10 can serve as electron and hole transport materials only, respectively. Therefore, we proposed a rational way for the design of efficient TADF materials as well as charge transport materials for OLEDs simultaneously.


INTRODUCTION
Organic light-emitting diodes (OLEDs) have drawn considerable attention for applications in displaying and lighting fields owing to their outstanding advantages nowadays (Choy et al., 2014;Zhang et al., 2015;Im et al., 2017a;Liu et al., 2017bLiu et al., , 2018Pal et al., 2018;Zhu et al., 2018). Unfortunately, their commercialization applications are still limited by the low device performance at high luminance and low external quantum efficiency (EQE) of the emitters. It is noteworthy that charge recombination results in 25% singlet and 75% triplet excitons in the traditional fluorescence process (Tang and Vanslyke, 1987;Burroughes et al., 1990). The triplet excitons cannot be directly utilized by emitters because they are spin forbidden and decay is non-emissive in producing undesired heat. As a consequence, the internal quantum efficiency (IQE) of OLEDs does not exceed 25% using singlet excitons for traditional fluorescent materials. Remarkably, it is challenging to develop highly efficient emitters utilizing triplet excitons. The phosphorescent emitters incorporating heavy metals can reach an IQE of nearly 100% by enhancing electron spin-orbit coupling (Baldo et al., 1998;Adachi et al., 2001;Xia et al., 2014). However, the applications of phosphorescent materials containing heavy metals are limited owing to their high cost and environmental contamination (Li et al., 2013;Liu et al., 2017a;Wang et al., 2018). Accordingly, it is critically important to develop novel metal-free materials, which can be functionalized as efficient emitters and exhibit environment friendliness. To address this issue, thermally activated delayed fluorescence (TADF) materials have been considered recently as promising candidates because of their potential in achieving 100% IQEs by harvesting all the triplet excitons (Uoyama et al., 2012;Wu et al., 2017;Yang et al., 2017;Chatterjee and Wong, 2019;Wang et al., 2019). As for TADF materials, a small energy gap ( E ST ) between the lowest singlet (S 1 ) and triplet (T 1 ) states promotes efficient spin conversion from the triplet to singlet manifold by thermal activation through reverse intersystem crossing (RISC), subsequently resulting in fluorescence from the converted S 1 to the ground (S 0 ) states (Evans et al., 2018;Li et al., 2018b;Zhu et al., 2019). It is noteworthy that a small E ST is favorable for the RISC process and may obtain TADF. Furthermore, it is noticeable that the interplay between theory and experiment is capable of providing a deeper insight into the understanding of the optical and electronic properties of molecules in ground as well as excited states. Many theoretical research efforts have been made in this regard. It is critically important to get the relationship between topologic structure and optical as well as electronic properties for designing novel materials with improved properties. Density functional theory (DFT) and time-dependent DFT (TD-DFT) approaches have been remarkably successful in accurately evaluating a variety of ground and excited-state properties, in particular for TADF materials (Lu et al., 2015a,b;Wang et al., 2017;Hussain et al., 2019). In this regard, the rational design of a twisted donoracceptor (D-A)-type structure can endow the molecules with TADF characteristics. These systems can efficiently reduce E ST in virtue of their separating the highest occupied molecular orbitals (HOMOs) and the lowest unoccupied molecular orbitals (LUMOs) (Geng et al., 2017;Li et al., 2018a). Additionally, D-A architectures are beneficial for the intramolecular charge transfer (ICT) in the excited states (Santos et al., 2016;Deng et al., 2019;Zhang et al., 2020). In addition, TADF materials with different donor moieties such as 9,9-dimethyl-9,10-dihydroacridine, 10H-phenoxazine, 5,10-dihydrophenazine, 10H-phenothiazine, 9H-carbazole, 3,6-di-tert-butyl-9H-carbazole, 9,9-diphenyl-9,10dihydroacridine 9H-spiro[4,5]fluorene-9,10-dihydroacridine, 9-(3-(9H-carbazol-9-yl)phenyl)-9H-carbazole, and diphenylamine possess excellent electroluminescence performance (Bezvikonnyi et al., 2019;Hosokai et al., 2019;Jin et al., 2019;Joo et al., 2019;Kim et al., 2019;Sharma et al., 2019;Zhong et al., 2019). Meanwhile, there are many TADF materials with different acceptor moieties such as benzonitrile, triazines, sulfones, benzophenone, quinoxaline, and naphthalimide reported in literature (Li et al., 2016;Nobuyasu et al., 2016;Tsujimoto et al., 2017;Sommer et al., 2018;Wu et al., 2018;Yu et al., 2018). The HOMO/LUMO distribution and energy levels of the typical donor and acceptor moieties have been found in the relevant literature (Im et al., 2017b). Among the numerous reported TADF materials, organoboron compounds have drawn significant interest recently owing to their appealing optical properties in OLEDs. Because of the lack of electrons in the P z orbital, the central boron atoms in the organoboron compounds exhibit strong electron-accepting ability through pπ conjugations, which is favorable for the ICT (Ji et al., 2017;Giustra and Liu, 2018;Liang et al., 2018;Meng et al., 2019).
Considering the merits and characteristics mentioned above, in this work, we design a series of novel D-A tricoordinated organoboron compounds with 5,9-dioxa-13b-boranaphtho[3,2,1-de]anthracene (DOBA) as electron acceptors and different polycyclic aromatic fragments as electron donors for TADF molecules (Scheme 1). Applying DFT and TD-DFT approaches, optoelectronic properties including frontier molecular orbital (FMO) energies (E HOMO and E LUMO ), HOMO-LUMO gaps (E g ), adiabatic ionization potentials (AIP), adiabatic electron affinities (AEA), reorganization energy (λ), and E ST were systematically investigated. The absorption, fluorescence, and phosphorescence spectra of the designed molecules were predicted using the TD-DFT method. This provides a useful insight for designing novel TADF materials as well as charge transport materials for OLEDs.

COMPUTATIONAL DETAILS
Geometry optimizations for all the designed molecules in the S 0 states were accomplished using the DFT method in the Gaussian 09 package (Frisch et al., 2009). The corresponding frequency calculations were carried out at the same level to prove the nature of each stationary point. Using the TD-DFT approach, the structures in S 1 and T 1 states were optimized. Under the optimized structures in S 0 , S 1 , and T 1 states, their absorption, fluorescence, and phosphorescence spectra were predicted by using the TD-DFT method. Meanwhile, the E ST values were evaluated by adiabatic excitation energy. The 6-31G (d,p) basis set was selected for all the calculations in this work. With the aim to select reasonable exchange correlation functionals, six functionals including B3LYP (Becke, 1993), PBE0 (Adamo and Barone, 1999), ωB97XD (Chai and Head-Gordon, 2008), M062X (Zhao and Truhlar, 2008), and CAM-B3LYP (Yanai et al., 2004) were employed to optimize the geometries of molecule 1 in S 0 , S 1 , and T 1 states. On the basis of the optimized geometries in S 0 and S 1 states, the absorption and fluorescence spectra were predicted using the corresponding TD-DFT method. The E ST values were also calculated using the different methods. The corresponding results as well as the experimental data are shown in Figure 1 and Supplementary Table 1. As visualized in Figure 1, obviously, the fluorescence emission wavelength (λ fl ) obtained using the B3LYP functional (567 nm) is quite close to the experimental value (557 nm) (Meng et al., 2019), with the deviation of 10 nm. Furthermore, both the E ST values SCHEME 1 | Molecule models of 1-10 investigated in this work.
of molecule 1 obtained using B3LYP and PBE0 functionals (0.0068 and 0.0107 eV) are close to the experimental value (0.0091 eV). Additionally, comparing the optimized geometries of 1 (for the atom numbering, see Supplementary Figure 1) in S 0 states with its crystal structure data [CCDC 1887610], one can find that the main geometrical parameters obtained using the B3LYP/6-31G(d,p) method are in better agreement with crystal structure data than those obtained with other methods (Supplementary Table 2). Therefore, geometry optimizations in S 0 , S 1 , and T 1 states, band gaps E g , E ST , absorption, fluorescence, and phosphorescence spectra of molecules under investigation were performed by the B3LYP/6-31G(d,p) and TD-B3LYP/6-31G(d,p) methods.
The reorganization energy λ is composed of external and internal reorganization energies (λ ext and λ int ). We focus only on the λ int because λ ext is quite complicated to predict at this stage. λ int can be expressed as (Köse et al., 2007;Sancho-García, 2007): Here, E ± and E stand for the energies of the charged and neutral states in the ground states, respectively; E * ± corresponds to the energy of the charged state with the optimized neutral molecule structure. E * represents the energy of neutral molecule with the optimized charged geometry. Furthermore, AIP and AEA were calculated by the adiabatic potential-energy surfaces of neutral/charged species. AIP can be obtained with the energy difference between cation and neutral specie. AEA was determined by the energy difference between neutral and anion specie. The AIP, AEA, and λ for the electron (λ e ) and hole (λ h ) of the designed molecules were predicted at the B3LYP/6-31G(d,p) level.

Molecular Geometries in the Ground and Excited States
The main geometrical parameters of the designed molecules in S 0 , S 1 , and T 1 states are given in Supplementary Table 3. The results displayed in Supplementary Table 3 reveal that there are no significant changes in the bond lengths of acceptor DOBA fragments for the designed molecules in S 0 states. Compared with the parent molecule 1, the bond length change mainly appears on the bonds which connect the DOBA acceptor and different donor fragments. Obviously, the lengths of C 5 -N 22 bonds for molecules 2, 3, 5-8, and 10 have been decreased by 0.005, 0.003, 0.02, 0.022, 0.001, and 0.024 Å, while the corresponding values of C 5 -N 22 for 4 and C 5 -C 22 for 9 have lengthened by 0.002 and 0.05 Å with respect to that of molecules 1, respectively. Similar phenomena are found for C 18 -N 24 bonds for 2-8 and 10 and C 18 -C 24 bond for 9. Furthermore, the designed molecules display a large twist angle (β) between their acceptor and donor fragments in the S 0 states owing to their large steric hindrance. The β values of 1-4, 7, and 8 are about 81-100 • , while the corresponding values of 5, 6, 9, and 10 are about 34-51 • , respectively. The large β values facilitate for disrupting the electronic communication between D and A fragments. Comparing the geometrical parameters of the designed molecules in S 1 states with those in the S 0 states, the bond lengths of B 1 −C 2 , C 4 −C 5 , C 6 −C 7 , C 9 −C 10 , C 13 −C 14 , B 1 −C 15 , C 17 −C 18 , and C 19 −C 20 (0.002-0.025 Å) are shortened, while other bond lengths are enlarged (0.001-0.025 Å), respectively. The more obvious bond length variations are found for the connecting bonds between acceptor and donor fragments for the designed molecules. The C 5 −N 12 and C 18 −N 24 bonds for 1-8 and 10 are stretched (0.005-0.038 Å and 0.005-0.038 Å) in the S 1 states as compared to those in the S 0 states, respectively. However, compared with the C 5 −C 12 and C 18 −C 24 bonds for 9 in the S 0 state, obviously, they are shortened by 0.036 and 0.003 Å in the S 1 state, respectively. Additionally, it can be seen that the twist angle β is closer to 90 • in the S 1 state as compared to those in the S 0 states for the designed molecules except for 9, whose two β values in the S 1 state are smaller than those in S 0 states, respectively. Apparently, inspection of Supplementary Table 3 reveals that the geometrical parameters in T 1 states are similar to those in S 1 states for 1-4, 7, and 8. On the contrary, the molecular geometries in T 1 states for 5, 6, 9, and 10 are closer to those in S 0 states, respectively.

Frontier Molecular Orbitals
It is worth noting that the FMO energies (E HOMO and E LUMO ) and E g play dominant roles for the optical and electronic properties. The molecular geometry is affected by electron density redistribution that is caused by an electronic excitation (Forés et al., 1999;Helal et al., 2011). The distributions of HOMO and LUMO in S 0 states for the designed molecules are plotted in Supplementary Figure 2. Additionally, we investigated the contributions of A and D fragments to the FMOs and the overlap value (ρ) between D and A fragments on HOMOs and LUMOs, as displayed in Supplementary Table 4. An inspection in Supplementary Figure 2 reveals clearly that the HOMOs of 1-10 are mainly localized on the D fragments, whereas the corresponding LUMOs predominantly reside at the A fragments. The contributions of D fragments for HOMOs in 1-4 and 7-9 are 93. 1, 94.4, 95.6, 95.4, 93.1, 92.8, and 99.4%, while the corresponding contributions for 5, 6, and 10 are 83.1, 85.8, and 66.5%, respectively. For the LUMOs, the contributions of A fragments in 1-8 are larger than 90%, while the corresponding contributions for 9 and 10 are 75.9 and 84.5%, respectively. Apparently, the HOMO to LUMO transitions lead the electronic density to flow from D to A fragments. The percentages of charge transfer are in the order of 4 (89.3%) > 3 (88.9%) > 2 (87.5%) > 7 (86.4%) > 8 (85.9%) > 1 (85.7%) > 6 (76.9%)> 9 (75.3%)> 5 (74.8%) > 10 (51%). Moreover, the overlap (ρ) between D and A fragments on HOMOs is in the range of 0.001-0.040, while the corresponding ρ values for LUMOs are 0.013-0.026, for the designed molecules except for 10, whose ρ values on its HOMO and LUMO are 0.082 and 0.037, respectively. It indicates that the HOMO to LUMO transitions exhibit a strong charge transfer nature and slight electron exchange energy, resulting in small E ST values.
The qualitative HOMOs and LUMOs in S 1 states for 1-10 are shown in Figure 2. The calculated E HOMO , E LUMO , and E g , the overlap (ρ) between D and A fragments, and the contributions of A and D fragments (in %) to the FMOs of 1-10 are summarized in Table 1. Comparing the HOMOs and LUMOs in S 1 states with those in S 0 states, we find that the electron density plots of both HOMOs and LUMOs in S 1 states are similar to those in S 0 states for all the investigated molecules, respectively. The HOMOs are mainly centralized on the D fragments, whereas the corresponding LUMOs predominantly distributed on the A fragments. The contributions of D fragments for HOMOs are about 91.2-99.9%. The contributions of A fragments for the LUMOs are about 90.5-93.8% except that the contribution of A fragment for 9 is 68.7%. The percentages of charge transfer from D to A fragments are in the order of 3 (89.5%) > 4 (89.1%) > 2 (88.6%) > 6 (88.1%) > 5 (87.8%) > 7 (87.3%) > 8 (86.6%)> 1 (86.3%)> 10 (81.7%) > 9 (68.6%). Clearly, the charge transfer character in S 1 states is more significant than that in S 0 states. Furthermore, the ρ values for HOMOs are in the range of 0.001-0.019, while the corresponding ρ values for LUMOs are about 0.014-0.037. It suggests that the HOMOs and LUMOs are also separated efficiently in the S 1 state. This implies the efficient separation between D and A fragments, which display the potential TADF features. The sequence of E HOMO and E LUMO is 3 > 4 > 2 > 1> 8 > 10 > 7 > 6 > 5 > 9 and 10 > 3 > 6 > 1> 7 > 8 > 5 > 2 > 4 > 9, respectively. Thus, the E g values are in the order of 5 > 10 > 9 > 6> 7 > 8 > 1 > 2 > 4 > 3. It suggests that the E g values of 2-4 decrease, while the corresponding value of 5-10 increases compared with that of 1. Therefore, one can conclude that 2-4 may possess a red shift, whereas 5-10 could exhibit a blue shift in their emission spectra in comparison with the parent molecule 1. It implies that the introduction of different aromatic heterocyclic donors can tune effectively the E HOMO , E LUMO , and E g values for the designed molecules. Furthermore, we investigated the qualitative HOMO and LUMO plots in T 1 states for 1-10 (Supplementary Figure 3). One can find that the qualitative HOMOs and LUMOs in T 1 states are similar to those in S 1 states for all the investigated molecules. It is also favorable for the RISC process from T 1 to S 1 states. Table 2 collected the vertical excitation energy (E S1 and E T1 ) in S 1 and T 1 states and E ST of 1-10 at the TD-B3LYP/6-31G (d,p)  level. It is noticeable that the lower the E ST values, the easier the RISC process from the T 1 to S 1 states. The E S1 and E T1 values  Meng et al. (2019). are in the orders 5 > 9 > 10 > 6 > 7 > 8 > 1 > 2 > 4 > 3 and 5 > 9 > 6 > 10 > 7 > 8 > 1 > 2 > 4 > 3, respectively. As expected, the order of E S1 values is similar to those E g values except that a sequential interchange is found between 9 and 10. Moreover, the sequence of E T1 values is similar to E S1 values except that a sequential interchange is found between 6 and 10.

Charge Transport Properties
It is quite clear that the charge injection and charge transfer play dominant roles in the device performance of OLEDs. λ can be used to estimate the charge transfer rate. The AIP and AEA are used to evaluate the energy barrier for the injection of holes and electrons. Generally, the low λ value corresponds to the big charge transfer rate (Marcus, 1964(Marcus, , 1993. The smaller the AIP value and the larger the AEA value, the easier the injection of holes and electrons, respectively. The calculated λ e , λ h , AIP, and AEA of 1-10 are listed in Table 4. As presented in Table 4, we find that 3 and 5 have the smallest and largest AIP values (5.151 and 6.387 eV), respectively. The AIP values of 1-10 are observed in the following: 5 > 9 > 6 > 7 > 10 > 8 > 1 > 4 > 2 > 3, suggesting that the hole injection and transportations of 2-4 are expected to be easier than the others with respect to parent molecule 1. On the other hand, the AEA values of 3, 5, 6, and 10 are lower, while the corresponding values of 2, 4, and 7-9 are higher than that of 1 (0.878 eV). The order of the AEA of 1-10 is as follows: 9 > 4 > 2 > 8 > 7 > 1 > 5 > 3 > 6 > 10; this implies that the abilities to accept electrons in 2, 4, and 7-9 are improved with respect to parent molecule 1. So, we can deduce that the introduction of 10H-phenoxazine (2) and 10H-phenothiazine (4) donor fragments will enhance both the hole and electron injection abilities, whereas the introduction of 9,9-diphenyl-9,10-dihydroacridine (7), 9H-spiro[4,5]fluorene-9,10-dihydroacridine (8), and 9-(3-(9H-carbazol-9-yl)phenyl)-9H-carbazole (9) donor fragments can improve the electron injection ability only compared with that of parent molecule 1.
For charge transport materials in OLEDs, generally, TPD (λ h = 0.290 eV) and Alq3 (λ e = 0.276 eV) are used as typical hole and electron transport materials, respectively (Gruhn et al., 2002;Lin et al., 2005). According to Table 4, the calculated λ h values of 1-3 and 5-10 are smaller than those of TPD. It suggests that the hole transfer rates of the designed molecules except for 4 may be higher than that of TPD. The λ h values are in the order 4 > 3 > 1 > 2 > 7 > 10 > 8 > 6 > 5 > 9. On the other hand, the λ e values of 1-9 are smaller than that of Alq3. It implies that the electron transfer rates of 1-9 may be higher than that of Alq3, suggesting that their electron transfer rates might be lower than that of Alq3. The order of the predicted λ e is as follows: 10 > 9 > 6 > 5 > 7 > 2 > 3 > 6 > 1 > 8. Remarkably, the difference between λ h and λ e for 1-3, 5-9 (0.029-0.216 eV) is small enough, suggesting that they are potential ambipolar charge transport materials. Nevertheless, 4 and 10 can serve as electron and hole transport materials only, respectively.

CONCLUSION
In summary, a series of novel D-A tricoordinated organoboron derivatives have been systematically investigated for TADF-based OLED materials. The calculated results show that the designed molecules exhibit small E ST values, suggesting that they can be used as excellent TADF candidates. DFT analysis indicated that the designed molecules display efficient separation between donor and acceptor fragments because of the small overlap between donor and acceptor fragments (ρ) on HOMOs and LUMOs. Furthermore, the delayed fluorescence emission color can be tuned effectively by introduction of different polycyclic aromatic fragments in the parent molecule. Molecules 2, 3, and 4 possess more significant Stokes shifts (326.6, 320.0, and 330.5 nm) and red emission with small E ST values (0.0065, 0.0057, and 0.0062 eV). Nevertheless, other molecules exhibit green (1, 7, and 8), light green (6 and 10), and blue (5 and 9) emissions. Meanwhile, they are potential good ambipolar charge transport materials except that 4 and 10 can serve as electron and hole transport materials only, respectively. Therefore, we proposed a rational way for the design of efficient TADF materials as well as charge transport materials for OLEDs simultaneously.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.

AUTHOR CONTRIBUTIONS
RJ conceived and designed the research, and headed, wrote, and revised the manuscript. JX contributed to the performance and analysis of the frontier molecular orbitals, absorption fluorescence, and phosphorescence spectra, and the reorganization energies. Both authors contributed to the manuscript revision and read and approved the submitted version.