1. Introduction

Since Tang's innovative work in 1987 ^[1], organic light emitting diodes (OLEDs) have experienced nearly three-decades development and become one of the most promising technologies for flat- panel display and solid-state lightening sources due to the advantages of light weight, flexibility, and energy saving ^[2]. During the three-decade evolution of OLEDs, there are so many materials and devices that were reported. Although the diversity of materials and devices can be nearly infinite, their emission manners can be by far summarized as only three: (i) emission from the radiative transition of singlet excitons, corresponding to the fluorescent OLEDs ^[1]; (ii) emission from the radiative transition of triplet excitons, corresponding to the phosphorescent OLEDs ^[3]; (iii) emission from the radiative transition of singlet excitons converted from triplet excitons, corresponding to the OLEDs based on the mechanisms of thermally activated delayed fluorescence (TADF) ^[4], triplet-triplet annihilation (TTA) ^[5], hybridized local and charge-transfer (HLCT) emission ^[6] and triplet-polaron-interaction (TPI) induced upconversion ^[7]. It should be noted that all the light-emitting materials used in the OLEDs mentioned above are closed-shell molecules although their emission manners are different. Recently, a new kind of OLEDs using open-shell molecules such as neutral π-radicals as emitters wherein the emission comes from the radiative decay of doublet excitons was reported ^[8]. Because the transition back to the groud state of doublet excitons is spin-allowed, the transition problem of triplet excitons of fluorescent materials is thus circumvented.

2. Summarization of all kinds of OLEDs classified by the emission manners

It is known that 25% singlet excitons and 75% triplet excitons will be created when the injected hole and electron meet each other in the light-emitting region according to the simple spin statistics ^[9]. As shown in Fig. 1a, in OLED using traditional fluorescent materials as emitters, only singlet excitons can radiatively decay, and the triplets which account for 75% of the total excitons are wasted ^[9]. Thus the upper limit of internal quantum efficiency (IQE) which is defined as the ratio of the number of emitted photons to the number of injected charge carriers is 25%, given that the PL quantum yield of fluorescent materials is 100%.

To improve the IQE of OLEDs, efficient harvesting and utilizing the triplet excitons is always the most important issue in the following research of OLEDs. In 1998, the EL emission from triplet excited states of metal-to-ligand complexes made it possible to efficiently utilize triplet excitons via heavy metal effect ^{[3a, 3b]}. In OLEDs based on metal-to-ligand complexes (phosphorescent materials), 100% IQE can be achieved due to that the spin-orbit

coupling effect of heavy metals enhances the transition possibility from T₁ to S₀ (Fig. 1b). Thus phosphorescent OLEDs containing heavy metals can overcome the problem of IQE upper limit of the conventional fluorescent OLEDs. However, phosphorescent OLEDs have some disadvantages, e.g., room-temperature deep blue phosphorescent molecules are scarce, and the sky blue phosphorescent molecules always show instability during the vacuum evaporation and device operation ^[10]. Moreover, most of the useful heavy metals for phosphorescent materials are confined to iridium (Ir) and platinum (Pt), which are expensive and limited nature resources. As a result, production cost of the phosphorescent OLEDs is rather high.

In 1999, Cao et al.reported that the spin statistical limit of singlet formation ratio of 25% can be broken in the electroluminescence of some conjugated polymers ^[2a]. Shuai et al. proposed that the reverse intersystem crossing (RISC) can happen from triplet interchain polaron pair to singlet interchain polaron pair which are the precursors of triplet and singlet exciton, respectively, due to the strong interchain coupling ^[2b]. Other approaches to harvesting triplet excitons are TTA, TADF, HLCT and TPI induced upconversion that convert non-emissive triplet excitons to emissive singlet excitons. As shown in Fig. 1c, TTA is that two triplet excitons collide each other to form a ground state (S₀) and a singlet exciton (S1) (2T₁ ＞ S₁), which causes the π-type delayed fluorescence. Ideally, the upper limit of IQE of the devices via TTA process can be enhanced to 62.5% ^[11].

The first application of TADF in OLEDs was reported by Adachi et al. in 2009 ^[4a]. In TADF process, as shown in Fig. 1d, extremely small energy gap between Si and Ti (ΔE_ST) is required to benefit RISC from Ti to Si through thermal activation. In this way 75% nonemissive triplet excitons can be efficiently utilized, resulting in 100% IQE of TADF-based OLEDs. To design TADF molecules, spatial separated distribution between highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) is needed, which is matched by the donor-acceptor (D-A) type structure ^{[4c, 12]}. However, D-A architecture often leads to intramolecular charge transfer (ICT) property that is not beneficial to achieve blue TADF materials and high PL quantum yield. Another problem is that TADF-based OLEDs have serious efficiency roll off as operation voltage increases ^[13].

Another feasible method to harvest triplet excitons is HLCT emission (Fig. 1e). By using moderate donor (D) and accepter (A) groups and controlling the steric hindrance between D and A, a new state can be obtained in which charge transfer (CT) state and locally excited (LE) state simultaneously exist in a molecule. LE part has the large transition moment and CT part can enhance the S1 component through RISC between high-lying triplet and singlet excited states ^{[6a, 6b, 14]}.

Recently, we reported a new possible way to harvest triplet excitons through TPI-induced upconversion ^[7]. Different from that TADF and HLCT benefit from the RISC processes at the lowest- and higher-level excited states, respectively, and the spin angular momentum is not conserved, TPI induced upconversion is a intermolecular process. It has two distinct characteristics: (i) TPI is a spin conserving process, i.e. the spin-flip does not need; (ii) TPI can occur circularly, that is TPI can continuously convert triplet to singlet, which may induce much higher singlet ratio in OLEDs.

It should be noted that all the emitters used in the OLEDs based on the mechanisms of phosphorescence, TTA, TADF, HLCT and TPI- induced upconversion are all closed-shell molecules. Last year, we demonstrated the possibility using neutral open-shell molecules as the emitter of OLEDs in which the emission comes from the doublet excitons ^[8].

3. Neutral π-radical-based OLEDs

For conventional luminescent materials (closed-shell molecules), their HOMO is filled with two electrons with opposite spin directions. If these molecules are excited, one electron leaps into the lowest unoccupied molecular orbital (LUMO) and the other electron leaves in the HOMO. The spin configurations of two electrons can be classified as singlet and triplet, respectively. According to Paul's exclusion principle, the transition of triplet excitons back to the ground state is spin-forbidden, only the transition of singlet excitons back to the ground state is spin- allowed and emits light.

However, neutral conjugated radicals (open-shell molecules) have only one electron in the outer molecular orbital, i.e., the singly occupied molecular orbital (SOMO). The spin configuration of one electron is doublet (spin-up and spin-down). In the EL process of an OLED based on neutral conjugated radicals (as shown in Fig. 2), electron and holes are injected from cathode and anode respectively. Then holes are transmitted into the HOMO of the hole-transporting layer (HTL) or doping-host molecules, finally hop into the SOMO of the radical. And the SOMO of the radical becomes empty due to only one electron occupies the SOMO. Similarly, electrons are transmitted into the LUMO of the electrontransporting layer (ETL) or doping-host molecules, then hop into the singly unoccupied molecular orbital (SUMO) of the radical. Thus the doublet excitons at radical molecules are formed. In addition, the doublet excitons can also be created through the energy transfer from the host. The transition back of the electron in SUMO to SOMO will generate a photon. Theoretically, the transition of doublet excitons of neutral radicals is always spin- allowed due to the vacancy of SOMO.

Up to date, the reported room-temperature stable emissive radicals are rare. However, rational molecular design not only endows the radicals stable enough in air and at room temperature but also makes them emissive. Most of the ambient-condition- emissive and stable radicals are concentrated on triphenylmethyl radical derivatives (for details please see Section 4).

Aiming to confirm our idea, we chose a stable deeπ-redemissive radical (TTM-1Cz) whose molecular structure is given in Fig. 3 ^[15] as an emitter to successfully fabricate an OLED. Because TTM-1Cz compound is severely self-quenched in sold state, we dispersed it into the 4, 4'-N, N'-dicalbazolylbiphenyl (CBP) host. The configuration of the OLED is ITO/MoO₃ (6 nm)/NPB (30 nm)/TTM- 1Cz: CBP (5 wt%, 40 nm)/TPBi (35 nm)/LiF (0.8 nm)/Al (100 nm). The device gives a deeπ-red emission at the lower voltages, whereas the blue component appears at the higher voltages. Maximum external quantum efficiency (EQE) of 2.4% was achieved, which is relatively high among the deep red OLEDs. To further identify whether the red emission originates from the doublet, magnetic field effect (MFE) testing was performed. In general, Zeeman Effect of the triplet is thought to be closely related with the MFE of OLEDs ^[16]. Thus, the existent magneto-electroluminescence (MEL) can verify the emission from singlet or triplet excitons (singlet and triplet always accompany each other). On the contrary, the non-existent MEL will confirm that the emission does not come from singlet or triplet excitons but from the doublet excitons. We used filters to separate the blue emission (from closed shell molecule) and red emission (from TTM-1Cz) and checked their MEL, respectively. As shown in Fig. 4, red emission shows no response to the magnetic field, whereas the blue emission has the obvious MEL. The result suggests that the red emission indeed comes from the doublet excitons of the radical molecule.

4. Triphenylmethyl-based radicals

The first organic free radical, triphenylmethyl radical, was discovered by Gomberg in 1900 ^[17]. Until now, the study of organic radicals has experienced more than a century and different kinds of radicals have been developed. Although organic radicals are generally believed to be short-lived and extremely active chemical substance, the long-lived radicals still can be obtained through proper molecular design, such as, organic nitroxides ^[18], (general formula R₂NO ), nitronyl nitroxides (NNOs ) ^[19], and π-conjugated radicals ^[20]. Among the π-conjugated radicals, only triphenylmethyl derivatives, i.e., perchlorophenylmethyl (PTM) and tris(2, 4, 6-trichlorophenyl)methyl (TTM) derivative radicals are emissive and more stable. It was reported that the half-lives of the PTM radicals are more than 100 years ^[21]. Thus, we focused our attention on PTM and TTM series radicals as the emitters of OLEDs.

Triphenylmethyl radicals were found to hardly dissociate in solvent or solid state, and tend to dimerize a century ago ^[20]. The significant progress happened in 1970 when Ballester et al. successfully synthesized the first extremely stable and room- temperature-emissive PTM radical ^[22]. The PTM radical is monomeric even in the crystalline state ^[23]. The stability of the PTM radical is ascribed to essentially steric shielding effect of the tricovalent central carbon by three ortho and para-chloro- substituted phenyl rings. It is noteworthy that when ortho- and para-positions of the phenyl ring is substituted by some atoms (Cl, F, Br) or some substituent groups (OMe, OH, NO₂), the stability is more increased and dimerization can also be avoided ^[23]. After Ballester's work on PTM radical, (Fig. 5a), numerous efforts have been focused on synthesizing the PTM and TTM radical derivatives and exploiting their physical and chemical properties ^[24].

Since 1994, Julia and co-workers have been contributing their outstanding works on synthesizing and tuning the optical properties of PTM- and TTM-cored radical derivatives ^[25]. In 2006, Julia et al. introduced carbazole unit into TTM core with C-N coupling reaction in the presence of cesium carbonate and obtained a strong red-emissive carbazole-pendant radical, TTM- 1Cz ^{[15, 26]}. In the following year, they also reported series of emissive TTM-cored carbazole derivative radicals, as shown in Fig. 5b ^[27]. Lambert et al. have made another excellent series works on synthesizing and studying the intramolecular electron transfer PTM-cored radicals (Fig. 5c) in which PTM is linked to different substituted-amine donors by various spacers ^[28]. Recently Nishihara et al. reported a more stable radical, PyBTM (Fig. 5a). The photostability of PyBTM is over 115 times higher than TTM and the fluorescence quantum yield is up to 0.81 in EPA matrix (diethyl ether:isopentane:ethanol) at 77 K ^[29]. They also prepared a PyBTM-based metal complex radical. After PyBTM was cooperated with metal (Au) complex stability was more enhanced than PyBTM ^[30].

5. Conclusion and outlooks

In this review, we briefly introduced the working principles and existing problems of different emission manners of OLEDs. In the traditional fluorescent OLEDs, only 25% singlet excitons can be utilized, but in other type OLEDs, the exciton utilizing ratio approaches a unit or more than a half because of harvesting nonemissive triplet excitons. In particular, in radical-based OLEDs the transition problem of triplet excitons is circumvented because the emission comes from the radiative decay of doublet excitons which is spin-allowed. Nevertheless, there are some problems that need to be solved in radical-based OLEDs including: (i) solid state selfquenching should be overcome and PL quantum yield should be increased in term of the material designing; (ii) In order to obtain the full-color display, emission spectra of light-emitting radicals have to be tuned to green and blue region; (iii) the stability of light- emitting radicals is also crucial for devices; (iv) in term of the devices, the working principle of the radical-based OLEDs need to be further investigated, e.g., underlying energy transfer mechanism between closed-shell and open-shell molecules, relation between device structure and efficiency.

Although the study of radical-based OLEDs is staying on its young age, we believe it would give another optional choice in OLEDs development as the emergence of new kinds of light- emitting radicals.

Acknowledgments The authors are grateful for financial support from the Ministry of Science and Technology of China (No. 2015CB655003), National Natural Science Foundation of China (No. 61275036).