Suppression of roll-off characteristics of organic light-emitting diodes by narrowing current injection/transport area to 50 nm

Organic light-emitting diodes (OLEDs) have been extensively studied over the past few decades for next-generation light-emitting applications such as flat-panel displays and lighting sources because of their attractive features, including flexibility, light weight, and low manufacturing cost.^1–3 Although current OLEDs show excellent light-emitting performance, major challenges still remain unsolved particularly at high current density. Under these conditions, exciton densities are high, inducing a variety of exciton annihilations, which lead to a marked decrease in efficiency.⁴ Such roll-off performance becomes a serious problem for obtaining high luminance in lighting and passive matrix driving that require high brightness of over several tens of thousands of cd m⁻². This is also a fundamental issue for realizing organic semiconductor laser diodes (OSLDs), which requires extremely high exciton density to achieve light amplification.^5–7 To produce an OSLD, there are four principle requirements: (i) injection and transport of high current density of over a few kA cm⁻², (ii) suppression of various exciton annihilation processes, (iii) minimization of waveguide loss such as light absorption by metal electrodes, and (iv) minimization of absorption loss by triplet excitons and polarons. Exciton–polaron annihilation is the most critical issue to overcome because all polarons have a broad absorption spectrum.^8–11 In our previous study, we demonstrated high current injection of over 1000 A cm⁻² without device breakdown into OLEDs with an unusual device configuration of a small active area (200 × 200 nm) on a high thermally conductive substrate by application of a short pulse voltage of less than 1 μs.^12,13 However, very serious efficiency roll-off was observed at such high current density.¹³ 

Roll-off characteristics originate from four main mechanisms: singlet-singlet annihilation (SSA),^5,14 singlet-triplet annihilation (STA),^15,16 singlet-polaron annihilation (SPA),^5,7,17,18 and singlet-heat annihilation (SHA).¹⁴ These annihilation processes can be described as

where $kss$⁠, $kST$⁠, $kSP$⁠, and $kheat$ are the annihilation rate constants for SSA, STA, SPA, and SHA, respectively.^4–19 To avoid these exciton annihilation processes at high current density, particularly exciton formation at organic/organic heterointerfaces that tend to allow accumulation of charge carriers, it has been recognized that expansion of the carrier recombination width is a useful method.^20,21

In this study, we demonstrate suppression of roll-off by introducing the idea of spatial separation of the “charge carrier flow, recombination, and exciton formation area” and the “exciton decay area” in fluorescence based OLEDs. For this purpose, we designed a narrow linear light-emitting structure to separate charge carriers and molecular excitons spatially. Various line structures with a length of 2 mm and width of 50 nm to 100 μm were formed using e-beam lithography,^9,22 as shown in Fig. 1. First, an insulator layer was prepared by spin-coating a photoresist solution onto an indium-tin-oxide (ITO)-coated glass substrate. Here, the photoresist resin (ZEP520A-7, ZEON Co.) and thinner solvent (ZEP-A, anisole, ZEON Co.) (in volume) were used. Then, the line pattern was engraved on the photoresist film by e-beam lithography. The exposed area of the photoresist was removed. Finally, organic layers were deposited, followed by a metal cathode. In this device architecture, current flow and successive exciton formation are limited inside the line pattern created by e-beam lithography, and the singlet excitons generated randomly diffuse in the emitter layer (EML) during their lifetime. Thus, we expect that some of the excitons can escape from the narrow carrier flow region in these OLEDs.

In this study, we used a neat 4,4′-bis[(N-carbazole)styryl]biphenyl (BSB-Cz) layer as an EML because it shows very low threshold for amplified spontaneous emission.^{7,20,21,23,24} Figure 2 shows the chemical structure of BSB-Cz and the energy diagram of the device consisting of a hole transport layer (HTL) of 4,4′-bis[N-(1-napthyl)-N-phenyl-amino]biphenyl (α-NPD), EML, electron transport layer (ETL) of phenyldipyrenylphosphine oxide (POPy₂), hole-blocking layer (HBL) of 2,9-dimethyl-4,7-diphenyl-1,10-phenanthroline (BCP), and electron-blocking layer (EBL) of N,N-dicarbazolyl-3,5-benzene (mCP). All layers were vacuum deposited. For the measurement of electroluminescence (EL) performance, we employed the same procedure as described in Ref. 7.

Figure 3(a) shows a cross-sectional transmission electron microscopy (TEM) image of a fabricated line-patterned device composed of an ultrathin photoresist layer with a linear channel having a depth of 15 nm, width (X) of 100 nm, and length of 2 mm. From the photos of the cross-sectional view of fabricated devices, we estimate that the error bar of the width is ±5%. The EL from the equivalent device with X = 200 nm (Fig. 3(b)) shows a clear linear emission pattern with nearly the same width of 200 nm. From microscope observation, we confirmed no EL from areas other than the patterned lines, ensuring that the 15 nm-thick photoresist layer well functions as an insulating layer throughout the device area. Because the 15 nm-thick photoresist layer has sufficient insulating characteristics, the current flow is completely confined to the channel. The current density (J)−voltage (V) characteristics of the unpatterned reference device with dimensions of 2 × 2 mm and the linear devices with X = 50 nm and 1 μm under both direct current (DC) and pulse operation are presented in Fig. 4(a). The reference device showed a maximum current density just before device breakdown (J_max) of 0.56 A cm⁻², while the linear device with X = 50 nm showed a fairly high J_max of 96 A cm⁻² even under DC operation. The breakdown of OLEDs under high current density can be primarily attributed to the melting of organic layers caused by Joule heating from the resistance to current injection and transport in organic semiconducting layers.⁹ Thus, Joule heating is suppressed in devices with a small active area, as we confirmed previously.^9–13,25 Furthermore, under pulse operation with a pulse width of 5 μs, a very high J_max of 2.8 kA cm⁻² was achieved in the linear device with X = 50 nm because of further suppression of device heating.

Figure 4(b) depicts external EL quantum efficiency (η_ext) − J characteristics of the devices with X = 50 nm, 100 nm, 200 nm, 500 nm, and 1 μm. While the reference device showed strong roll-off characteristics, relaxation of roll-off was observed as X decreased, and the narrowest device with X = 50 nm showed the most relaxed η_ext − J behavior. We confirmed the reliability of our J-V-η_ext experiments by measuring the characteristics of devices with a large number of narrow lines patterned in a 2 × 2 mm area (see the supplementary material, Fig. S1).²⁶ 

Here, we discuss exciton quenching in the devices based on SHA and SPA models.^7,14 Figure 4(c) shows fitting of the roll-off characteristics of the reference device under DC and pulse operation, demonstrating good agreement with the SHA model for DC operation and SPA model for pulse operation. (See the supplementary material, Fig. S2, for details of fitting parameters.²⁶ Also, we note that J is defined as the ratio between the current and the stripe area.) First, we consider the effect of device size on roll-off performance based on SPA. SPA can be described by Eq. (3) and has a rate equation of

where $[ S1* ]$ is the singlet exciton density, $τ$ is the lifetime of fluorescence, $kP$ is the SPA rate coefficient, $[ nC ]$ is the charge density, $q$ is the elementary charge, and $d$ is the thickness of the recombination area.⁷ $[ S1*¯ ]$ is the exciton density confined in the active area interacting with polarons, namely, exclusion of diffused excitons outside the channel. Thus, the term $kP[ nC ][ S1*¯ ]$ describes the net exciton quenching by polarons. Therefore, when excitons diffuse from the current injection area, $[ S1*¯ ]$ can be decreased, resulting in the relaxation of roll-off. Here, we defined $[ S1*¯ ]$ as $(1−z)[ S1* ]$⁠, where $z$ is the unconfinement factor of excitons from the active area. Therefore, Eq. (5) can be rewritten as

As a first approximation, we can assume that J–V characteristics can be described by the trap charge limited current model: $J∝Vl+1$(Refs. 27 and 28). Thus, the η_ext − J characteristics of the model under steady-state conditions are given by

where $η0$ is the external quantum efficiency without SPA, and the characteristic current density when the initial efficiency drops by half (J₀) is

where $NC$ is the density of state, $Nt$ is the trap density, $Et$ is the trap depth, $d$ is the thickness of the exciton generation zone, and $μ$ is the carrier mobility. Equation (8) reveals that J₀ increases with z. Based on this equation, we discuss the effect of exciton diffusion on the roll-off characteristics depending on X in the linear devices. As depicted in Fig. 5(a), excitons can diffuse from the active area to the areas outside the channel where the current density can be assumed to be almost zero. Here, the singlet exciton diffusion process can be described by

where $[ S1* ]$ is the singlet exciton density, $[ T1* ]$ is the triplet exciton density, $Ds$ is the singlet exciton diffusivity, $LD$ is the exciton diffusion length in an EML, $τ$ is the lifetime of an exciton, $R$ is the exciton generation rate, $q$ is the electric charge, $d$ is the thickness of the recombination area, $kr$ is the radiative decay rate, $knr$ is the non-radiative decay rate, $kisc$ is the intersystem crossing rate, $krisc$ is the reverse intersystem crossing rate, and $ktta$ is the triplet-triplet annihilation rate.^29–31 Figure 5(b) simulates exciton diffusion for the reference and linear devices with X = 2 mm, 200 nm and 50 nm under constant current density of 1 A cm⁻² with $LD$ of 13 nm in a neat film of BSB-Cz, which was measured in a separate experiment (see the supplementary material, Fig. S3).²⁶ We note that exciton upconversion and triplet-triplet annihilation processes were omitted for the calculation as the first approximation. The simulation of exciton diffusion indicates that decreasing X increases separation between polarons and excitons because of exciton diffusion. In particular, in the device with the smallest active area of X = 50 nm, it can be estimated that 26% of excitons diffused out from the carrier flow region.

Finally, we analyzed the roll-off characteristics of the devices based on the SHA and SPA models. Figure 5(c) shows an X − J₀ plot obtained from the experimental data. The green, blue, and red lines correspond to the theoretical curves of the SPA model with L_D = 8, 13, and 18 nm, respectively. The experimental data for the devices with X under 200 nm lie between the theoretical curves corresponding to L_D = 13 and 18 nm, which are slightly higher than the independently measured value of L_D = 13 nm. It has been shown that exciton diffusion length increases with temperature.³² Therefore, we anticipate that, because of heating, exciton diffusion length may be increased near the recombination zone. In our device, Joule heating creates a temperature gradient that promotes exciton drift unidirectionally away from the recombination zone, which effectively increases the exciton diffusion length. Because a device with smaller X provides stronger temperature gradient, a larger effective exciton diffusion length is expected and the experimental data falls onto a theoretical curve of higher L_D. In addition, for X larger than 200 nm, the experimental plots deviate from the theoretical curve, which can be ascribed to the decrease of J₀ caused by Joule heating at high current injection.

It is important to note that the lateral diffusion of charge carriers is undesirable and has to be suppressed. By considering the time of flight of injected carriers through the device, we estimate the characteristic diffusion length of charges in a lateral direction of $Lch=dkTVq$⁠, where d is the device thickness, k is the Boltzmann constant, T is the temperature, V is the applied voltage, and q is the elementary charge (see the supplementary material, Fig. S4).²⁶ Thus, thinner devices and higher operating voltages are needed for shorter L_ch. For instance, in a 100 nm-thick device at room temperature, L_ch is estimated to be about 5 nm at 5 V, and 3 nm at 30 V. These numbers are 2–4 times shorter than the exciton diffusion length of 13 nm; therefore, excitons are able to diffuse away from charges, resulting in the observed suppression of exciton–polaron annihilation. The effects of space charge on lateral charge displacement should also be considered to achieve deeper understanding of device operation.

In summary, we prepared a finely patterned OLED structure that allows partial spatial separation of charge carriers and singlet excitons. It was clarified that the roll-off based on the SHA was primarily suppressed with the small device areas (>200 nm) and further suppression based on SPA was realized with the smaller device areas (<200 nm). In this OLED architecture, injection of a fairly high current density of over 1000 A cm⁻² and relaxation of roll-off characteristics were observed. Line-patterned structures would therefore be a useful architecture for high-brightness applications. However, even the OLED with the narrowest width of 50 nm still showed marked roll-off behavior, which was ascribed to the rather short exciton diffusion length of the BSB-Cz layer. We expect that two-dimensional confinement will further suppress exciton annihilation by introducing tiny dot structures and highly ordered materials with longer exciton diffusion lengths of a few hundred nanometers. Future work will include further clarification of the exciton deactivation mechanism of SHA. Since Joule heating would dissociate excitons to produce holes and electrons, SPA would be accelerated. Thus, we would like to formulate SHA and successive SPA processes based on the exciton dissociation by Joule heating taking into account the non-uniformity of electrical field, lateral charge diffusion, exciton blocking by transport layers, and shorting effect of exciton diffusion length by SPA.

We gratefully acknowledge KOBELCO Research Institute, Inc., for observation of the cross-sectional TEM image. This work was supported in part by the Funding Program for World-Leading Innovative R&D on Science and Technology (FIRST). O.V.M. and T.Q.N. thank the support from the National Science Foundation (No. DMR-1411349).