Triplet-Triplet Upconversion in Organic Light- Emitting Diodes: Implications to Injection Lasing

While significant progress has been made over last few years in the field of organic solid-state lasers, achieving lasing action from organic semiconductors under electrical excitation still remains a big challenge. One of the major barriers towards electrically pumped organic lasers are optical losses due to triplet excitons. In this work, we report both experimental and theoretical results that confirm positive contribution of triplet excitons for the scope of electrically driven organic laser. We studied a model fluorescence material, 9-(9-phenylcarbazole-3-yl)-10(naphthalene-1-yl) anthracene and reveals that the threshold current densities required to achieve lasing emission under electrical injection can be significantly reduced with the aid of triplet-triplet exciton upconversion processes. Furthermore, we demonstrate that, to achieve the best performance, the singlet-triplet exciton annihilation must be minimized. These results are unprecedented and provide a pathway towards development of new class of triplet-triplet upconversion materials for injection laser.


Introduction
Light-emitting optoelectronic devices based on organic semiconductors such as organic lightemitting diodes (OLEDs) 1-3 and light-emitting field-effect transistors (LEFETs) [4][5][6] have shown promising advancements in last two decades towards applications in areas such as high-end TVs, flexible display technology, low-cost solid-state lighting and data communications. 4,7 Organic solid state-state lasers (OSSLs) on the other hand have received significant attention in recent years because of their ability to provide tunable, coherent light emission while allowing compatibility with a variety of substrates such as silicon, ITO/glass or plastic at low cost processing techniques. At the moment, most of the OSSLs are optically pumped using a pulsed excitation source with pulse widths ranging from femtoseconds (fs) to few nanoseconds (ns). 8 15 However, the devices demonstrated in the study were found to be unstable due to the high driving voltages required to achieve lasing emission. Hence, further studies towards development of new materials for electrically pumped organic lasers are much required.
Achieving lasing emission under electrical excitation is a difficult task. This is primarily because of the trade off in photoluminescence quantum yields (PLQYs) and charge carrier mobility of organic materials. 16 Obtaining high mobility in organic compounds essentially requires higher degree of π-π stacking between the chromophores in films, this in turn leads to drastic reduction in PLQYs due to aggregate induced luminescence quenching. 17 To overcome this, molecular doping approach has been employed to achieve high film PLQYs, low lasing threshold (Eth) and high efficiency in case of optically pumped laser and OLEDs. 18 However, doping approach leads to reduction in charge carrier mobility of these systems. Apart from high mobility and high PLQY, organic laser gain medium should also possess high radiative rate constant (kr) and high electroluminescence (EL) quantum yield in order to achieve lasing action at low optical pump intensities and hence at low injection current densities. To address these issues, continuous efforts are made to obtain high EL efficiency from non-doped fluorescent emitters with high radiative rate constants. 19 Unfortunately, the maximum internal quantum efficiency (IQE) of OLEDs based on fluorescent materials is limited only to 25%. On the other hand, OLEDs based on phosphorescent and recently developed materials utilizing thermally activated delayed fluorescence (TADF) 20 provide 100% IQEs. However, these materials do not give stimulated emission due to long excited-state lifetimes and low radiative rates. Hence, it is crucial to discover new classes of lightemitting materials that can possess high kr and high EL quantum yield in order to achieve lasing action from organic semiconductors at low current densities. Anthracene derivatives have been one of the most extensively studied materials, which show efficient EL and charge transport properties in non-doped conditions in OLEDs. [21][22][23][24][25][26][27] Anthracene derivatives in OLEDs have continuously showed high external quantum efficiencies (EQEs) exceeding the theoretical limit of traditional fluorescent emitters for past several years. [22][23][24][28][29][30] This is mostly attributed to the triplet-triplet upconversion (TTU) in these anthracene derivatives. Furthermore, recently Adachi et al. reported enhancement in TTU efficiency of anthracene derivatives close to 50% through spin-orbit coupling leading to high IQEs of close to 60%. 31 The TTU enhanced efficiency of the anthracene derivatives is promising for application in injection lasing.
However, the influence of TTU in OSSLs has not yet been explored.
We found that high EQEs of this material in non-doped conditions can be attributed to TTU at low current densities (J) while at higher current densities singlet-triplet annihilation (STA) becomes a detrimental factor leading to sharp EQE roll-off. We performed time-resolved photoluminescence spectroscopy (TRPL), fluence dependence study for prompt (PF) and delayed fluorescence (DF) and nanosecond-transient absorption spectroscopy (ns-TAS) to extract rate constants and to understand the various bimolecular recombination processes.
Furthermore, we performed theoretical modelling to extract rate STA and TTU rate constants.
To gain information on lasing, amplified spontaneous emission (ASE) characteristics were measured for doped and neat films of PCAN under nanosecond-optical excitation. Using the exciton-exciton annihilation parameters, the current density required to achieve the lasing threshold were then simulated. We show that the high TTU rate constant can lead to significant reduction in threshold current density required to achieve lasing emission under electrical excitation. This study also highlights, STA rate constants need to be reduced significantly alongside increment in TTU rate constants in future material design. To the best of our knowledge, this is the first comprehensive study on TTU material that sheds light into interplay between triplet up-conversion, singlet-triplet annihilation and lasing threshold.

Steady-state photophysical properties
Photophysical properties of PCAN were studied in toluene solution, neat and doped films (10wt% PCAN in 4,4′-bis(N-carbazolyl)-1,1′-biphenyl (CBP)). As shown in previous theoretical study, the highest occupied molecular orbital (HOMO) and lowest occupied molecular orbital (LUMO) are primarily delocalized on the anthracene unit along non-planar (tilted by close to 90º) orientation of the 9, 10 substituents, which allows to maintain the electronic structure of anthracene while the non-planar structure helps in reducing the aggregate induced photoluminescence quenching as compared to other anthracene derivatives. 32 Figure 1b shows the normalised absorption and photoluminescence spectra of PCAN in dilute toluene solution, neat films and doped films. The absorption spectra of PCAN in solution and neat films show typical vibronic structures arising from the anthracene unit while absorption spectra in the doped films is majorly dominated by the absorption of CBP host. The PL spectrum of PCAN in solution shows major emission wavelength peak at 428 nm along with a slight shoulder at 445 nm. On the other hand, PL spectrum for the neat film was found to be red-shifted with emission peak at 452 nm, compared to that of solution. The red shift in emission of neat film can be attributed to the intermolecular interaction in neat films, which is absent in dilute solutions. 16 The emission spectra in doped films also showed major peak at 452 nm, though the contribution of shoulder at 435 nm was found to be enhanced in doped film, suggesting that the intermolecular interaction in doped films are significantly reduced as compared to the neat films. The above features of the PL spectra are further supported by the PLQY values. The PLQY of dilute toluene solution was found to be 64±7%, while for neat films this value drops down to 37±6%. However, doping in CBP host helps in retaining the high PLQY value of 54±6% for this system.

OLED characteristics
To understand the EL properties, we fabricated OLEDs with non-doped and 10wt% PCAN   Table S1 shows the performance comparison of PCAN based doped and non-doped devices. Even though the peek EQE of 2.8% at 1600 cd/m 2 achieved in this study for non-doped devices was found to be slightly lower than the reported EQE for PCAN, 32 it is interesting to note that peak EQE still exceeds the theoretical EQE limit of ≈1.9% for typical fluorescent OLEDs (calculated from the neat-film PLQY of 37%, assuming singlet: triplet generation of 1:3 and light out-coupling of 20%). Devices with doped PCAN EMLs on the other hand showed a peak EQE of ≈2.3% at 1 cd/m 2 , which is lower than that achieved with non-doped devices and also lower than the theoretical fluorescent limit (≈2.75%) even despite the doped EML by having a higher PLQY (≈55%). These results suggest a vital contribution of triplet to singlet up-conversion in non-doped devices, which is otherwise absent or significantly reduced in doped devices, leading to superior performance of non-doped PCAN OLEDs. Furthermore, the absence of any strong donor-acceptor moieties in the molecular structure dismisses the presence of TADF or "hot exciton" channel, which can be mediated via hybrid charge transfer-locally excited (HLCT) character of singlet and higher lying triplet states. [36][37][38]  between luminance and current density is the typical characteristics of up-conversion of triplet excitons to singlet excited-states via TTU process.

TRPL and fluence dependent measurements
To obtain insight into the non-linear relationship (slope >1) between luminance and current density in PCAN neat-film devices, we studied the time resolved characteristics of PCAN under optical excitation. Figure S2 shows the time correlated single photon counting (TCSPC) decay curves at peak PL wavelengths for PCAN in dilute toluene solution, neat and 10wt% doped in CBP films. In dilute toluene solution (deoxygenated), the fluorescence intensity decays mono-exponentially with lifetime of 3.5 ns. However, the neat and doped films showed bi-exponential decay in prompt fluorescence (PF) along with small contributions from delayed fluorescence (DF) in the longer times. The details of the lifetime and their respective contributions are listed in Table S2. For neat films, the fast components of 1.1 ns (81%) and 3.5 ns (14%) in the PF can be attributed to aggregates and non-aggregate state respectively.
While the presence of long lived (~ 85 ns) DF is a characteristic feature observed in case of up-conversion of triplet excitons to singlet excited-states via TTU process 40 .
To further study the long-lived DF in neat and doped films, we used variable gatedelays (ranging from picosecond to microsecond range) based TRPL technique. Figure 3a shows the complete PL decay kinetics of neat and doped films excited at 355 nm (840 ps pulse width) laser source with excitation fluence of 80 μJ cm -2 . The neat films show distinct longlived DF, compared to the doped films. Interestingly, neat films showed two DF regions with lifetime in order of several hundreds of ns (black solid line) and 10 µs (blue solid line), respectively. We attribute the origin of the two DFs in neat films to different triplet diffusivity due to different orientations of chromophores in the bulk. Similar impact of molecular packing on DF has been previously observed for several systems. 41 Figure 3b shows the PL spectra for neat films at different gate delays. The similarity in spectra for PF and DF confirms that both emissions arise from the same excited-state species.
To confirm the TTU under optical excitation, we have carried out time resolved fluence dependence PL and carefully monitor the relative contribution of DF and PF. Fig. 3c shows normalised DF intensities as a function of normalised PF intensities at same pump fluence.
Herein, the absolute PF and DF intensities were obtained by integrating the respective spectra over all wavelengths. We note that under lower fluence region, both PF and DF follow almost linear relationship and then start to shift towards almost quadratic relation at higher pump fluence. This is due to fact that increment of triplet excited-state population with increasing fluence leads enhancement in contribution due to TTU. The near quadratic relation in fluence dependent studies confirm the presence of TTU, which is consistent with non-linear relation of luminance and current density under electrical excitation.

Nano second Transient electroluminescence studies
To obtain a complete picture of TTU which contribute to EQE "roll-up" processes, we studied the transient electroluminescence at high current densities. The OLEDs were subjected to 100 ns electrical pulse at high current density. The prime benefits of applying a short pulse to the OLED is suppression of the joule heating. The schematic of the transient EL setup is shown in Figure S3. The device areas of the OLEDs were reduced to 0.3 mm 2 to minimise the RC response distortion by reducing the geometrical capacitance. Figure S4a  leading to reduction in singlet density. STA dominates primarily via Förster-type energy transfer mechanism, whose efficacy is determined by the spectral overlap between S1 emission and T1→Tn absorption bands.
To confirm STA in PCAN, we performed nano-second transient optical pump-probe absorption spectroscopy. Figure S5 shows the triplet and singlet excited-state absorptions for neat PCAN films measured under vacuum. The long-lived triplet excited-state absorption band positioned between 400-500 nm was found to overlap strongly with the emission confirming STA in PCAN. Figure S6 further shows the normalised EL turn-off characteristics for doped and non-doped devices at 20 V, showing prompt and delayed components.
Next, we quantitatively extracted the STA and TTU rate constants from singlet and triplet population using rate equations designed for anthracene derivatives (details in the Supporting Information Section S1). Figure 5a and b shows the theoretical fits and the corresponding triplet population density for non-doped and doped devices, respectively at 20 and 40 V. From EL transients, STA and TTU rate constants for doped devices are obtained as 1.44 × 10 −9 cm 3 s −1 , and 6.67 ×1 0 −11 cm 3 s −1 , respectively ( Table S4). The extracted annihilation rate constants for non-doped PCAN devices are significantly higher than the doped PCAN devices with a value of 7.37 × 10 −9 cm 3 s −1 for STA and 2.32 × 10 −10 cm 3 s −1 for TTU (Table S3). These rate constant are well within the typical range of STA and TTU rate constants reported for organic semiconductors in the literature. 42 Clearly, the STA rate constant for the non-doped devices is much higher than typical fluorescent OLED emitters but this is still lower than those reported for anthracene single crystals. 43 The higher value of TTU rate constant in non-doped devices explains the higher peak EQEs, compared to doped devices at steady-state condition as contribution from delayed fluorescence is significantly enhanced.
However, the higher STA rate constant in non-doped devices implies a trade-off between TTU and STA. 41,44-47

Light amplification properties
To study the light amplification properties of PCAN, we explored the ASE characteristics for PCAN neat and doped films. ASE characteristics were obtained by optically exciting the samples using nitrogen laser with pulse width of 3.5 ns at 337 nm. The emission from the edge of the samples was collected using a spectrometer. The ASE threshold (Eth) was obtained from the abrupt change of slope in plots of output intensity from the edge of the sample vs input excitation fluence along with drop in full width at half maxima (FWHM) of the output spectra. The ASE threshold measured above can be translated into threshold singlet exciton density by employing optical model similar to the one described in Section S1 (details in the Supporting Information Section S2). The threshold singlet exciton density was found to be 3 × 10 19 cm -3 and 1 × 10 18 cm -3 for neat and doped films, respectively. We then extract the threshold current density (Jth) required to reach this singlet exciton density under electrical excitation (details in the Supporting Information Section S2). For comparative analysis, we considered three conditions: 1) an ideal case with no annihilation of singlet-singlet (SSA), singlet-triplet (STA) and triplet-triplet (TTU) excitons; 2) when only STA and 3) when both STA and TTU are taken in account. Using the kSTA and kTTU values extracted in Section 2.4, we calculated the singlet exciton density as the function of current density in all the three conditions. Figure 7a and b shows the calculated singlet exciton density as the function of current density for doped and neat systems, respectively, with threshold singlet exciton density for lasing shown by horizontal blue line. It is interesting to note that when only STA is considered, the singlet population do not reach the threshold required to achieve lasing in both doped as well as non-doped system. This suggest that the losses due STA are extremely detrimental to lasing. When both STA and TTU rate constants are considered, the singlet population can reach the threshold of current density, Jth ~ 40 kA cm -2 for doped devices. These results highlight the positive impact of TTU on reducing the Jth required to achieve lasing in organic semiconductors.
To further elucidate the impact of STA and TTU on lasing threshold, we varied kTTU in range 10 -11 -10 -10 and kSTA in range 10 -9 -10 -8 and simulated the Jth in doped PCAN devices while keeping the other rate constants unchanged.  Figure S7. Clearly, the STA has much severe detrimental impact on Jth, whereas TTU facilitates in reducing the lasing threshold. Thus, TTU plays a crucial role in reducing the injection lasing threshold. However, STA must also be simultaneously minimized to obtain the best performance. The way to reduce the STA is by designing TTU material that has minimum spectral overlap between singlet emission and triplet absorption spectrum. Other probable techniques include, the use of dual core TTU chromophores 48 with high up-conversion yield and utilizing materials with high TTU efficiency as a host matrix that can lead to enhancement in EL properties of existing highperformance organic laser dyes. [49][50][51][52]

Conclusion
In summary, we report both experimental and theoretical results that confirm positive contribution of TTU for the scope of electrically driven organic laser. We selected, PCAN as a model material exhibiting efficient triplet-triplet up-conversion and showed light amplification properties under optical pulse excitations. We studied the transient EL response of PCAN OLEDs and demonstrated that triplet population plays a critical role in controlling the device efficiency as a function of current density through TTU and STA processes.
Furthermore, by theoretical modelling we showed that the threshold current densities required to achieve light amplification can be significantly reduced with the aid of TTU. We believe that these results are of significant importance for developing a new class of TTU based laser material for injection lasing.

Experimental Section
Photophysical measurements: UV-visible absorption spectra for thin films and solutions were measured using Cary-5000 UV-Vis spectrometer. Horiba Jobin Yvon Fluoromax was used to measure PL spectra. Solution PLQYs were determined using Quinine sulphate in 0.1 M H2SO4 as the standard (PLQY = 55%) as the standard. Optical density of standard and the sample were measured to be 0.1 at excitation wavelength of 360 nm. 53 Solid-state PLQY measurements were performed using a calibrated integrating sphere. 54 Nano-second transient absorption spectroscopy (TAS) were performed as described in literature. 10 Pump wavelength of 400 nm was used to excite the sample. TCSPC measurements were performed using Halcyone fluorescence spectrometer with IRF (instrument response function) of 150 picoseconds (ps) with excitation wavelength 337 nm.
Thin film preparation: Thin films for photophysical and ASE measurements were prepared using same procedure. PCAN was purchased from LUMTEC and used as purchased. Nondoped and CBP doped films were deposited by thermal evaporation under high vacuum (~10 -6 mbar) on pre-cleaned fused silica substrates to obtained thin films with 200 nm thickness as measure by Dektak 150 profilometer (Bruker). All the substrates were cleaned using acetone, isopropanol followed by UV-ozone.