Efficient and Bright Organic Radical Light‐Emitting Diodes with Low Efficiency Roll‐Off

Organic radicals have been of interest due to their potential to replace nonradical‐based organic emitters, especially for deep‐red/near‐infrared (NIR) electroluminescence (EL), based on the spin‐allowed doublet fluorescence. However, the performance of the radical‐based EL devices is limited by low carrier mobility which causes a large efficiency roll‐off at high current densities. Here, highly efficient and bright doublet EL devices are reported by combining a thermally activated delayed fluorescence (TADF) host that supports both electron and hole transport and a tris(2,4,6‐trichlorophenyl)methyl‐based radical emitter. Steady‐state and transient photophysical studies reveal the optical signatures of doublet luminescence mechanisms arising from both host and guest photoexcitation. The host system presented here allows balanced hole and electron currents, and a high maximum external quantum efficiency (EQE) of 17.4% at 707 nm peak emission with substantially improved efficiency roll‐off is reported: over 70% of the maximum EQE (12.2%) is recorded at 10 mA cm−2, and even at 100 mA cm−2, nearly 50% of the maximum EQE (8.4%) is maintained. This is an important step in the practical application of organic radicals to NIR light‐emitting devices.


Introduction
Near-infrared (NIR) organic light-emitting diodes (OLEDs) have been extensively investigated due to their potential applications in a wide range of technological areas, such as biosensing, optical imaging, photodynamic therapy, surveillance, security, and communications.The exploitation of thermally activated delayed fluorescence (TADF) in NIR electroluminescence (EL) has been DOI: 10.1002/adma.202303666[15][16][17][18][19][20][21][22][23][24][25][26] Organic radicals, incorporating one unpaired electron on the outermost molecular orbital, show a doublet spin manifold, which differs from the singlet and triplet spin states observed in typical organic semiconductors.As the total spin quantum number (S) of both the ground and excited states is maintained at S = ½, the D 1 (the first electronically excited doublet state) → D 0 (the ground electronic doublet state) transition is a spin-allowed process enabling an internal quantum efficiency (IQE) of 100% with nanosecond decay lifetime. [27,28]However, although an IQE of almost 100% in doublet EL devices was reported in 2018, the maximum radiance was only ≈700 mW sr −1 m −2 due to large efficiency roll-off and unoptimized device structures. [24]In our recent report, a system comprising a radical guest in a host of the hole-transport material shows severe trapping of electrons on the radical guest causing a large efficiency roll-off due to the narrow emission zone at the interface of the emissive layer (EML) and electron-transport layer (ETL). [25]Enhancement of electron transport in the EML is required to balance hole and electron transport and exciton formation.
For this purpose, we have selected 3′,5′-di(carbazol-9-yl)-[1,1′biphenyl]-3,5-dicarbonitrile (DCzDCN) as a host for organic radicals since it can support both electron and hole transport, which was reported to contribute to a considerable improvement of device performance in TADF and phosphorescent OLEDs. [29]To explore how DCzDCN works in a doublet fluorescence emissive system, we used a tris(2,4,6-trichlorophenyl)methyl (TTM)-based radical emitter, TTM-3PCz (see ref. [24] for synthetic details), as a doublet emitter and compared this system with devices made with 4,4-bis(carbazol-9-yl)biphenyl (CBP) as a control host.Their chemical structures are displayed in Figure 1a.In this work, we have investigated these two systems in terms of photophysics and device physics, showing how the different host characteristics affect the doublet emission mechanism and the importance of host selection for high performance in doublet EL devices.

Results and Discussion
Firstly, the optical characteristics of the CBP-and DCzDCN-based doublet emissive systems were explored by steady-state and transient photophysical measurements.These insights give optical signatures for later characterization of the EL mechanism and mode of host or guest excitation in devices.Figure 1b shows the steady-state absorption and photoluminescence (PL) spectra of these systems.As the TTM-3PCz absorption spectrum broadly covers between 350 to 500 nm, the PL-absorption overlap between the two hosts and TTM-3PCz is sufficient for effective Förster resonance energy transfer (FRET) despite the slightly redshifted PL spectrum of DCzDCN compared to CBP.The PL spectra of the TTM-3PCz 3% doped in CBP and DCzDCN films excited at 330 nm (see Figure S1, Supporting Information, for absorption spectra of CBP and DCzDCN neat films) are shown in Figure S2 (Supporting Information).Strong NIR emission from TTM-3PCz with an emission peak of 709 nm is observed in both films, indicating efficient energy transfer from the host to the radical emitter.There is some residual UV/blue host emission (CBP blend: 7% and DCzDCN blend: 6% in total emission), in line with our previous work. [25]ansient photophysical measurements show different exciton dynamics in the two hosts.Figure 2a shows the comparison of the transient PL kinetics for the CBP:TTM-3PCz 3% and DCzDCN:TTM-3PCz 3% films when the hosts are excited.Rapid doublet fluorescent emission with a prompt decay lifetime ( p ) of ≈20 ns is measured for both blends at early times (<200 ns).However, there is a distinct delayed emission for the DCzDCN blend beyond 200 ns in contrast to the CBP blend.There are two routes for the delayed doublet exciton formation: 1) intersystem crossing (ISC) → reverse intersystem crossing (RISC) → singletto-doublet FRET and 2) ISC → triplet-to-doublet Dexter energy transfer (DET).Detailed exciton decay rates and time constants are summarized in Tables S1-S4 (Supporting Information).Transient kinetics were obtained by the integration of the emission spectrum at each time.The transient PL of pristine DCzDCN shows a slower prompt decay time of 8.8 ns than CBP (3.4 ns) with weak delayed components (Figure S3, Supporting Information).The rate constants of ISC and RISC between singlet and triplet excited states in DCzDCN were calculated as ISC rate: 1.39 × 10 6 and RISC rate: 4.21 × 10 3 s −1 (see the calculation method in the Supporting Information).The RISC rate for DCzDCN is much slower than other more efficient TADF materials with a value of >10 6 s −1 since the singlet-triplet energy gap of DCzDCN is quite large, 0.27 eV. [29]Considering the rather slow RISC rate for DCzDCN, route 2) is expected to be the dominant route in this system.Transient PL characteristics were investigated at different temperatures to confirm this assumption.Figure 2b,c shows the temperature-dependent transient PL plots for the TTM-3PCzdoped and nondoped films for host excitation at 330 nm.There is little temperature dependence for kinetics from 0 to 200 ns, but the fitted monomolecular lifetime beyond 200 ns shows a strong temperature dependence, as shown in Figure 2d, and from which activation energies, E a of 0.204 ± 0.003 eV for DCzDCN:TTM-3PCz 3% and 0.228 ± 0.002 eV for pristine DCzDCN are extracted from the Arrhenius plot (Figure S4, Supporting Information).
If the delayed emission in DCzDCN:TTM-3PCz 3% originates from delayed singlet-to-doublet FRET (ISC and RISC followed by FRET), which is the same mechanism as a hyperfluorescent emissive system, [30,31] the activation energy of DCzDCN:TTM-3PCz 3% should be identical to pristine DCzDCN.However, the delayed lifetime ( d ) in DCzDCN:TTM-3PCz 3% is considerably shorter than in pristine DCzDCN at all temperatures (Figure 2d) and suggests additional pathways, in particular triplet-to-doublet DET. [26]Figure 2e shows the temperature dependence of transient PL for the DCzDCN:TTM-3PCz 3% film when only the radical is excited (at 530 nm).In contrast to the host-excited decay kinetics in Figure 2b, transient PL decay shows negligible dependence on temperature.This result shows that as the radical is excited, directly photoexcited doublet excitons decay with ≈20 ns lifetime.From the comparison of the normalized integrated PL intensity for host and radical excitation (Figure 2f), we observe a 24% increase of the normalized integrated PL intensity following photoexcitation of the host, arising beyond 200 ns.We propose this is due to triplet-to-doublet DET. Figure 2g  Turning to LED devices, the role of triplet excitons is very consequential when electron-hole capture is in the host since spin statistics give 75% triplet generation, and the generated triplet excitons can be transferred to the radical guest via DET, which could contribute to the doublet EL process. [26]Our previous report suggests that most doublet excitons are directly formed at the radical under electrical excitation, as charges are easily captured at radical sites. [25]Accordingly, both bulk recombination and direct recombination at the radical should be considered in doublet EL devices based on the photophysical studies above.
EL devices were engineered to optimize device performance.In summary of device engineering, to enhance electron injection and transport to the EML, 4,6-bis[3,5-(dipyrid-4yl)phenyl]-2-methylpyrimidine (B4PYMPM) was chosen as the ETL after comparing with bis-4,6-(3,5-di-3-pyridylphenyl)-2methylpyrimidine (B3PYMPM).B4PYMPM has a deeper LUMO energy level and higher electron mobility, which leads to reducing driving voltage and improving efficiency roll-off and radiance. [32]lso, an mCP layer contributes to improving hole injection to the EML by cascade energy level alignment, as well as preventing the exciton leakage to the hole-transport layer (HTL) by high singlet and triplet energy levels of 3.6 and 2.9 eV (Figure S5, Supporting Information), [33] which leads to the further improvement of efficiency roll-off and radiance for the DCzDCN-based device (see the details of device engineering in Supporting information).The best-performing device structure is shown in Figure 3a.This shows the optimized device structure along with the chemical structures of the charge transporting materials: B4PYMPM as ETL, and 1,1-bis[(di-4-tolylamino)phenyl]cyclohexane (TAPC), 4′,4″-tris(carbazol-9-yl)triphenylamine (TCTA), and 1,3-bis(Ncarbazolyl) benzene (mCP) as HTL.
The optoelectronic performance of these devices is shown in Figure 3b-d.Figure 3b shows the current density-voltage (J-V) plots for the devices.The DCzDCN-based device shows slightly higher current densities than the CBP-based device, which we consider is due to enhanced electron transport. [29]Figure 3c shows the EQE-current density plots.Regarding the CBP-based device, efficiency is high up to 1 mA cm −2 , but falls rapidly at higher current densities.In contrast, the DCzDCN-based device shows high EQEs at high current densities (12.2% at 10 mA cm −2 and 8.5% at 100 mA cm −2 ).As we discuss below, we consider this to be due to better electron and hole current balance at higher current densities.Figure 3d shows the radiance-current density plots to illustrate the substantial enhancement of light intensity from the DCzDCN-based device (110 000 mW sr −1 m −2 ) compared to the CBP-based device (3900 mW sr −1 m −2 ) (see also the EQE-radiance plots in Figure S6, Supporting Information).The inset photo in Figure 3d shows emission with a peak wavelength of 707 nm from the DCzDCN-based device (see the EL spectra of the devices in Figure S7, Supporting Information).This simultaneous accomplishment of high EQE and radiance can be ranked as one of the highest performance characteristics in NIR OLEDs with over 700 nm peak wavelengths.In addition, we measured the operational lifetime (EL intensity decay to 50%) of these devices under constant drive at 0.1 mA cm −2 .The DCzDCNbased device presents a significantly improved lifetime (more than a factor of 10) compared to the CBP-based device and our previous results for doublet EL devices (Figure S8, Supporting Information). [16,26]ransient EL measurements were performed to explore the doublet EL mechanism in devices and compare them with the transient photophysical studies above.Figure 4a shows the transient EL plot for the CBP-based device.A prominent EL overshoot after turn-off is recorded, and we consider it results from the remaining trapped and accumulated charges at the radical sites and the interface of the EML and ETL, [34,35] considering CBP only supports hole transport, [25] which will be further demonstrated in single-carrier device analysis below.In addition, Figure S9 (Supporting Information) shows the comparison of the J-V characteristics for the full devices with and without doping, indicating more prominent electron trapping in the CBP-based device than the DCzDCN-based device.In contrast, Figure 5b presents the transient EL profile for the DCzDCN-based device.The DCzDCN-based device does not show an EL overshoot or additional emission after turn-off but only exhibits rapid doublet fluorescence.Moreover, no delayed EL is detected in the DCzDCN-based device, which is not associated with the host excitation-based transient PL kinetics (Figure 2b) but the radical excitation-based transient PL kinetics (Figure 2e).As discussed above, effective energy transfer is observed when excitons are generated on host molecules under photoexcitation.On the other hand, when electrons and holes are directly captured by radicals under electrical excitation, energy transfer between host and guest molecules is less relevant. [25]Therefore, we consider the transient EL profiles suggest that the primary doublet EL mechanism in these combinations is direct charge recombination at radical sites, rather than energy transfer from host molecules.Nevertheless, we have investigated whether bulk recombination can also contribute to generating excitons at the host.38][39][40] MEL is defined as MEL(%) = (EL(B) − EL(0))∕EL(B) where EL(B) and EL(0) are the EL intensity in the presence and absence of a magnetic field, B, respectively.Figure 4c shows the MEL profiles for the CBP-and DCzDCN-based devices measured at 5 V.The CBP-based device shows negligible MEL dependence, indicating the dominant direct formation of doublet excitons at the radical.However, the DCzDCN-based device shows a positive dependence (≈0.8%) of the doublet emission with the application of the magnetic field, indicating that there is some contribution from host excitons to the doublet EL due to energy transfer, and we consider this indicates triplet-to-doublet DET in line with the host excitation-based transient PL (Figure 2b) and our previous report. [26]Figure 4d shows the schematic illustrations of the PL and EL mechanisms for the DCzDCN-based emissive system.We consider the PL is mainly attributed to energy transfer from the host to the radical guest with the small contribution of directly excited doublet excitons.In contrast, direct recombination at radical sites is dominant in the EL, but the small proportion of excitons formed on host molecules can transfer to the radical, contributing to the EL.
Single-carrier devices were investigated to confirm the dominant direct charge recombination-based doublet EL mechanism and the different electrical properties in these two systems.Hole-only and electron-only devices (HODs and EODs) (detailed device layouts are shown in Figure S10, Supporting Information) were designed based on the full device structure shown in Figure 3a.For HODs, TAPC was used instead of B4PYMPM between the EML and cathode to block electron injection, while the B4PYMPM layer was deposited on indium tin oxide (ITO) to exclude the hole current in EODs.Also, to explore the effect of the radical doping in the EML on the overall current flow, the HODs and EODs with the presence and absence of the radical doping in the two hosts were investigated, and their J-V characteristics are plotted in Figure 5a,b.For the CBP-based devices, the hole trapping is almost negligible below 5 V, evidenced by the almost identical J-V characteristics with and without TTM-3PCz doping, and at higher voltages, only a slight trapping effect is detected, with very good hole-transporting properties even in the radical-doped device.In contrast, as CBP has poor electrontransporting properties, the J-V profiles of the EODs are much shallower than those of the HODs, and severe electron trapping is recorded.These unbalanced hole and electron-transporting properties support the transient EL profile for the CBP-based device showing the EL overshoot after turn-off.However, the DCzDCN-based HODs and EODs show different J-V features.In summary, the HOD results indicate that TTM-3PCz doping causes little trapping in the CBP host (since the HOMO levels are similar) but causes significant hole trapping in the DCzDCN host.For the EOD characteristics, we note that TTM-3PCz doping causes significant electron trapping, but this is less severe for the DCzDCN host than for the CBP host.Accordingly, the much more balanced hole and electron-transporting properties support the substantial enhancement of device performance for the DCzDCN-based device.Also, it is shown that electrons are easily trapped by radical sites, which suggests that the main doublet EL mechanism is charge trapping-based direct recombination in line with the transient EL and magneto EL studies above.Note that host emission is not detected in the EL (Figure S7, Supporting Information), as generally observed in radicalbased OLEDs, [24,25] also indicating additional evidence for this demonstration.
Figure 5c,d shows the schematic illustrations of the charge transport and EL mechanisms studied above.For the CBPbased device, most of the holes and electrons recombine near the EML:ETL interface.For the DCzDCN-based device, the emission zone is more extended away from the EML:ETL interface, which reduces bimolecular quenching processes.The voltage-dependent EL spectra of the full devices displayed in Figure S11 (Supporting Information) also support these representations of charge transport and trapping and doublet exciton formation and degradation mechanism in devices.Note that for the CBP host, we observe some evidence for blueshifted emission at 500 nm, which we consider may arise from CBP to B4PYMPM exciplex emission, but this is not observed with the DCzDCN host.
To further support the narrow recombination zone for the CBP-based device and extended recombination zone for the DCzDCN-based device, the so-called "zone doping" experiment was performed (see details and schematic illustration in Figure S12, Supporting Information).To distinguish the NIR emission from TTM-3PCz, one of the conventional blue emitters, 2,5,8,11-tetra-tert-butylperylene (TBPe), was employed as the secondary dopant in the four different zones in the EML and compared their EL spectra in these two systems (Figures S13 and S14, Supporting Information).The high intensity of TBPe blue emission indicates that exciton generation is dominant in that zone.Accordingly, the recombination zone for the devices can be estimated based on the integrated intensity of the blue emission from TBPe, as shown in Figure 6a,b.The narrow recombination zone close to the EML:ETL interface is anticipated for the CBP-based device, considering the much higher integrated intensity at Zone 4 than other zones.In contrast, the substantially extended emission zone to the EML:HTL interface is expected in the DCzDCN-based device, given that the high intensity of blue emission is observed at Zone 1 compared to Zone 4, and its intensity further increases as voltage increases.
We consider the main loss mechanism causing large efficiency roll-off at high current densities to be bimolecular quenching: doublet-polaron quenching (DPQ) in doublet EL devices. [41]herefore, we modeled the efficiency roll-off characteristics by exploiting the exciton diffusion equation with DPQ rates (see details in Supporting Information), and the simulated efficiency curves are compared with the experimental values in Figure 6c.

Conclusion
We have demonstrated high-performing doublet EL devices with the maximum EQE and radiance of >17% and >100 000 mW sr −1 cm −2 .By exploiting the new doublet emissive system using the TADF host, DCzDCN, supporting both hole and electron abilities, its exciton dynamics have been successfully investigated in terms of both photophysics and device physics, depending on how and where excitons are generated.Our new design of the nonradical host:radical guest emissive system shows the disruptive potential of organic radicals for NIR light-emitting technologies.

Experimental Section
Sample Preparation: 50 nm of TTM-3PCz 3% doped in CBP and DCzDCN films were deposited on glass substrates by a thermal evaporation process in a vacuum chamber (<5 × 10 −7 Torr) for steady-state and transient photophysical measurements.For the measurements of the temperature-dependent transient PL, 150 nm of DCzDCN:TTM-3PCz 3% and DCzDCN neat films were deposited on quartz substrates by a thermal evaporation process under high vacuum (<5 × 10 −7 Torr).For the fabrication of OLEDs and single-carrier devices, ITO-coated substrates were cleaned with acetone and isopropyl alcohol, and then O 2 plasma treatment was applied to align the energy level with a hole-transporting layer.All layers, including organic layers and a LiF/aluminum cathode, were thermally deposited under a high vacuum (<5 × 10 −7 Torr).The doping concentrations stated in this study denoted weight percentages.
Photophysical Measurements: Steady-state PL spectra were measured by an Edinburgh Instruments fluorescence spectrometer (FLS980) with a monochromated xenon arc lamp at  Ex = 330 nm under a nitrogen flow.A Shimadzu UV-3600 Plus spectrophotometer was employed for the measurement of absorption spectra.FLS980 with an integrating sphere under a nitrogen flow was used to measure PLQE, and the films were excited at 330 nm.Transient PL was recorded by using an Andor electrically gated intensified charge-coupled device (ICCD) (Andor iStar DH740 CCI-010) with 330 nm laser excitation; the decay kinetics were obtained from the integration of the total spectrum at each time.An optical cryostat (Oxford Instruments) was utilized to measure the temperature-dependent transient PL.
Device Characterization: The J-V characteristics of single-carrier devices were recorded by a Keithley 2635 sourcemeter.The performance of the OLED devices was measured by a Keithley 2635 sourcemeter and a calibrated Si photodiode.The EL spectra were recorded by an Ocean Optics Flame spectrometer.MEL measurements were performed with magnet cores (GMW 3470 electromagnet).A Keithley 2635 sourcemeter was utilized to apply the voltage to the device, and EL spectra were recorded by an Andor spectrometer (Shamrock 303i and iDus camera) with and without a magnetic field.Transient EL was measured by a Keithley 2401 function generator using square voltage pulses with a frequency of 20 000 Hz and a pulse width of 5 μs for the on-cycles (forward bias).The voltage pulse corresponds to 1 mA cm −2 , and the off-voltage is −5 V.The decay kinetics were obtained from the integration of the total spectrum at each time, detected by an Andor ICCD camera (Andor iStar DH740 CCI-010).

Figure 1 .
Figure 1.a) Chemical structures of TTM-3PCz, CBP, and DCzDCN.b) The absorption of TTM-3PCz solution and the PL spectra of CBP and DCzDCN neat films.
shows a schematic diagram of possible doublet emission mechanisms for the DCzDCN:TTM-3PCz system depending on how doublet excitons are formed at radical sites based on the detailed transient PL analysis.If excitons are predominantly formed on host molecules, most doublet excitons are created by singlet-todoublet FRET and triplet-to-doublet DET.On the other hand, if direct exciton formation dominates the emission process, energy transfer from/to host molecules can be excluded.Photoexcitation of the TADF host gives triplet generation via ISC, and this has allowed the identification of direct host triplet to doublet energy transfer.

Figure 3 .
Figure 3. Device architecture and optoelectronic performance.a) Device layout with energy level diagram and chemical structures of the charge transporting materials used in devices.b) Current density-voltage plots for the CBP-and DCzDCN-based devices.c) EQE-current density plots for the CBPand DCzDCN-based devices.d) Radiance-current density plots with the inset photo showing the operation of the DCzDCN-based device.

Figure 4 .
Figure 4. a,b) Transient EL plots for CBP-based (a) and DCzDCN-based (b) devices.For the CBP-based device, an initial EL spike is recorded after the turn-off due to the remaining trapped and accumulated charges.In contrast, the DCzDCN-based device does not show the additional emission or EL overshoot but exhibits rapid doublet fluorescence, indicating that holes and electrons rapidly recombine at radical sites.c) Comparison of magneto-EL profiles for the devices at 5 V. d) Schematic diagram of the emission mechanism under photoexcitation and electrical excitation.Energy transfer from the host is dominant in the PL, whereas the EL mainly results from direct recombination at the radical.

Figure 6 .
Figure 6.a,b) Integrated intensity of TBPe emission from 5 to 10 V at zones 1-4 for CBP-and DCzDCN-based devices.c) Normalized EQE vs current density plots with fitting curves (solid lines) for CBP-and DCzDCN-based devices based on doublet-polaron quenching.d) EQE vs radiance for the reported NIR fluorescent OLEDs showing the peak wavelengths between 700 and 800 nm.The numbers in brackets denote references.