1. Introduction

In response to recent demands for flexible display and lighting devices using organic light-emitting diodes (OLEDs), demands for flexible transparent electrodes in OLEDs have also increased. Conventional oxide-based transparent electrodes, however, cannot easily be used on flexible substrates because they are brittle. To realize flexible OLEDs, therefore, it is necessary to find an alternative electrode material bearing a sufficient mechanical compliance. In this respect, graphene is a strong candidate. With its outstanding electrical and optical properties that are comparable to those of conventional transparent electrodes [1–7]. Graphene has been extensively studied in the past decade as a next-generation transparent electrode. Furthermore, graphene has an advantage as a flexible electrode material owing to its high mechanical strain rate [8–12]. These advantages well surpass those of other candidate materials such as metal nanowires, metal meshes, conducting polymers, and carbon nanotubes. This is because they have a lower compatibility with existing display manufacturing processes, especially due to their lack of processing stability [13–18]. Despite these merits, however, the out-coupling efficiencies of the OLEDs with graphene electrodes (graphene-OLEDs) in previous studies have been lower than those of OLEDs with conventional transparent electrodes. A most likely reason for this is the weaker microcavity effect encountered in graphene-OLEDs. In conventional OLEDs, the microcavity effect has been an efficient means for improving the light extraction efficiency. In graphene-OLEDs, however, this microcavity effect is weak because it depends on the reflectance of the transparent electrode and the graphene’s reflectance is low [19–21]. In order for graphene to be used in flexible OLEDs, therefore, it is necessary to develop a technology that compensates the weak microcavity effect. To this goal, we have employed two technologies in this study. The first one is the introduction of a light extraction layer. A typical OLED is composed of a stack of layers with different refractive indices, which optically confines the generated light within the device, lowering the out-coupling efficiency of the OLED [22–24]. This optical loss occurs in the graphene-OLED as well and hinders the extraction of light from the device. The light extraction layer is an efficient method for overcoming this. It is a layer to alter the light paths to facilitate the escape of the emitted light from the device, which significantly reduces the internal reflection and thereby improves the light extraction efficiency [25–30]. The second strategy is to minimize the optical absorption in the graphene electrode. Since graphene has an optical absorption of about 2.3% per layer, the absorption should increase in proportion to the number of layers [31,32]. In this work, we investigated the impacts of varying the number of graphene layers on the optical properties of the graphene-OLEDs. Finally, optical simulations were conducted to optimize the OLED performance, and the optical and electrical properties of the optimized graphene-OLEDs were compared to those of OLEDs with conventional electrodes.

2. Experiments

To form the nanostructure, a thin metal film was evaporated onto the SiO_x surface and then heated to induce agglomeration of the metal. Technical details regarding the scattering layer formation can be found elsewhere [33]. The planarization layer had a higher refractive index (n = 1.81 at a wavelength of 550 nm) than the nanostructure (n = 1.48). This refractive index difference was sufficient to produce the scattering effect for light extraction. To planarize the scattering structure, a 1.5-µm-thick layer of SiN_x was deposited on the scattering structure and then polished using cerium as a slurry until it reached a total thickness of 1 µm, including the scattering structures, to remove the residual surface roughness. Single-layer graphene was separately grown on a thin Cu film via thermal chemical vapor deposition. Graphene was then isolated by removing the Cu film using an etching solution. During the etching process, graphene was heavily p-doped with a benzimidazole solution. The graphene was finally transferred onto the target substrate using a thermal release tape method [34,35]. The active luminous area was 2 mm × 2 mm. All materials were electronic grade and were used without further purification. All organic layers were deposited in a high vacuum chamber below 6.67 × 10⁻⁵ Pa using a thermal evaporation method. Each device contained a stack structure of graphene (one, two, and four layers)/an alternating HTL structure of 1,4,5,8,9,11-hexaazatriphenylene hexacarbonitrile [HAT-CN] (10 nm) and 1,1-bis[(di-4-tolylamino)phenyl]cyclohexane [TAPC] (40 nm) (t_HTL = 250 nm in total)/2,6-bis[30-(N-carbazole)phenyl]pyridine:Tris[2-phenylpyridinato-C2,N]Iridium(III) DCzPPy:Ir(ppy)₃] (20 nm) as the emission layer (EML)/1,3-bis(3,5-dipyrid-3-yl-phenyl)benzene (BmPyPB) (60 nm) as the electron transport layer. The EML produced a green color with a main peak wavelength of λ = 517 nm. The surface morphologies of the nanostructure were investigated by means of scanning electron microscopy (SEM, Model: Sirion 400, Philips). The direct transmittance of the graphene films on the glass and on the scattering layer was measured using an UV visible spectrophotometer (U-3501, Hitachi). The J–V characteristics were measured using a current/voltage source unit (Keithley 238). The angular-dependent luminances and EL spectra and the device efficiencies were measured using a goniometer-equipped spectroradiometer (Minolta CS-2000) and an integration half-sphere system (Otsuka electronics) at a constant current density level of 80 µA, respectively. The optical simulations were conducted using SETFOS (Fluxim), which is an OLED-specialized simulator. To obtain realistic results, we extracted the optical constants (n: k) of the organic materials, which were measured using an ellipsometer (M-2000D, J.A. Woolam Co.).

3. Results and discussion

The nanostructure was composed of nano-pillars with their height of ~500 nm and diameters of 300–500 nm, with a proper random distribution as shown in Figs. 1(b) and 1(c). Such a structure can effectively diffuse the incident light from the organic light-emitting layer and reduce the optical confinement. In addition, unlike periodic structures with fixed feature sizes, it enables wavelength-independent optical scattering of all visible lights [36,37]. Surface planarization is essential when using a nanostructure in an OLED because a rough surface can easily degrade the operational stability of the OLED. The planarization layer should reduce the surface roughness without causing micro-faults and transfer the light generated in the light-emitting layer to the nanostructure without substantial absorption. The planarization layer material and its fabrication process should not degrade the graphene properties. In a previous study, we used an organic/TiO₂ sol-based hybrid material for the planarization layer [38]. However, this material was inadequate as a supporting layer for graphene because of its friable surface with a roughness (R_a) of approximately 2.1 nm, which is greater than the thickness of monolayer graphene. In this study, silicon nitride (SiN_x) was chosen as the planarization layer material to satisfy the aforementioned requirements. This planarization process significantly decreased the surface roughness, from 500 nm to 0.4 nm, as shown in Figs. 1(d) and 1(e).

 

Fig. 1 Schematic of graphene OLED with the scattering layer (a), scanning electron microscope image of the nanostructure (b) and (c), and AFM measurement of the nanostructure (d) without and with the planarization layer.

Download Full Size | PDF

Figure 2 presents the simulation results. The simulation was made assuming the multiple interference, with the layer thickness and the optical constants being chosen as the main parameters. The thickness of the graphene layer was varied for one, two, and four atomic layers while the thickness of the monolayer graphene was fixed at 0.34 nm. To consider the microcavity effect, which changes as a function of the OLED thickness, the hole transport layer (HTL) thickness (t_HTL) was varied from 250 nm to 400 nm. The HTL plays an important role in facilitating the hole charge transport in OLED structures [39–41]. For comparison, we also conducted a simulation on the OLED with an oxide anode (indium zinc oxide: IZO). The radiance of the graphene-OLEDs increases as the number of graphene layers decreases for a fixed t_HTL value. Its dependence on t_HTL, however, is not so sensitive as that of the IZO-OLED. Although the graphene-OLEDs exhibit higher radiances than that of the IZO-OLED for t_HTL ~350 nm, the order is inverted for t_HTL values greater than 360 nm. The radiance of the IZO-OLED significantly oscillates in a sinusoidal fashion due to its strong microcavity effect, and its highest value at t_HTL = 380 nm is greater than any values of graphene-OLEDs. Because graphene has a lower reflectance than IZO, the microcavity effect is weak in graphene-OLEDs and their efficiency is rather insensitive to the HTL thickness [19].

 

Fig. 2 The simulation results; (a) the effect of the microcavity and optical absorption of graphene electrodes on OLEDs and (b) The absorption effect of graphene electrodes on OLEDs with and without the scattering layer. (Inset ; the experiment results of the reflectance of the glass substrate and the scattering layer on the glass)

Download Full Size | PDF

The increase of the radiance with the decrease of the number of graphene layers is ascribed to the reduced optical absorption in the graphene electrode. The extinction coefficient (k) of monolayer graphene was reportedly larger than 1 [42,43] which obviously attenuates the light passing through the graphene. To explore the maximum attainable radiance, we conducted a simulation on a monolayer graphene-OLED with a virtual k of zero (open squares in Fig. 2(a)). Even in this case, however, the radiance could not exceed the maximum value of the IZO-OLED at t_HTL~380 nm. This result clearly indicates that it is not possible for a graphene-OLED, without additional light extraction methods, to exceed the efficiency of OLEDs with conventional oxide electrodes. To explore alternative possibilities, we examined introduction of a scattering layer between graphene and the substrate using the simulation (Fig. 2(b)). The structure was designed to have a scattering layer with a total thickness of 1 µm with a root-mean-square roughness of 500 nm and a planarization layer. The actual refractive index (n) of silicon oxide (SiO_x; n = 1.48 at a wavelength of 550 nm) was used for the nanostructure layer and that of SiNx (n = 1.81) was used for the planarization layer. The number of graphene layers was varied from one to 10 while the t_HTL value was held constant at 250 nm. We hereafter refer to the OLEDs with and without the nanostructure layer as a scattering device and a planar device, respectively. As the number of graphene layers decreases, the radiances of both the planar and the scattering devices increase. The impact of introducing the scattering layer is evident. At a graphene film thickness of 1.5 nm, the scattering device shows a radiance which is higher than that of the planar device by 35.77%. This enhancement becomes even stronger for thinner graphene layers.

The radiance of the scattering device appears to be more sensitively varying with the graphene thickness as compared with the planar device. To understand this trend, we measured the reflectance of the glass substrate itself and compared it with that of the scattering layer on the glass (Fig. 2(b) inset). The reflectance of the scattering layer fluctuates rapidly, which is compared to the monotonic behavior of the glass. This difference implies a presence of complicated optics in the scattering device than in the planar device. In the planar device, the generated light travels toward the transparent electrode and the reflective cathode surface. The light components interfere and create internal optics analogous to those of a Fabry–Pérot interferometer. The reflectance measurement on the scattering layer suggests that some of the generated light is back-reflected from the scattering layer surface in random directions. Because the back reflection occurs repeatedly, the absorption effect is expected to play a dominant role in scattering devices, which accounts for the rapidly decreasing radiance as a function of the graphene number in this device.

The sheet resistance (Fig. 3(a)) and the direct transmittance (DT) (Fig. 3(b)) of the graphene films on the glass and on the scattering layer were measured. As the number of graphene layers increases, both the sheet resistance and DT decrease. The sheet resistances of monolayer and four-layer graphene film are approximately 270 Ω/sq and 70 Ω/sq, respectively. The graphene films on the scattering layer exhibit sheet resistances slightly lower than those of the films on glass. This is attributed to the SiN_x in contact with the graphene. Previous studies also suggest that the graphene’s sheet resistance decrease if the dielectric constant of the layer supporting the graphene increases [44], but no significant impact is evident in the present results. DT decreased with the number of graphene layers for both “graphene only” and “graphene on scattering layer” samples (Fig. 3(b)). The DT of the graphene/glass structure was higher than that of the graphene/scattering layer/glass structure. Since the glass substrate used in the experiment has a DT value that is higher than 95% throughout the visible range, the transmittance is primarily affected by the layers on the glass. For the graphene-only samples, DT was always larger than 70% while it never exceeded 50% in the “graphene on scattering layer” samples. The scattering layer with a nano-structure diffuses the incident light in forward directions. This result demonstrates the scattering layer’s capacity of widely diffusing the incident light. This light diffusion, expressed as a diffuse transmittance, amounts 31% in our scattering layer, which was higher than that of glass substrate (less than 1%).

 

Fig. 3 The sheet resistances (a) and the direct transmittances (b) of the graphene films on the glass and on the scattering layer with the variation of the graphene layer as one, two, and four.

Download Full Size | PDF

Figure 4 presents the current density–voltage (J–V) characteristics and the electroluminescence (EL) spectra in directions normal to the graphene-OLEDs with and without the scattering layer. Devices with one, two, and four graphene layers were used. All of them exhibit similar J–V characteristics with low leakage current levels and on/off current ratios higher than 10⁸, as shown in Figs. 4(a) and 4(b), respectively. This high electrical stability is ascribed to the low surface roughness (R_a ~0.4 nm) of the SiN_x planarization layer. The graphene used in this experiment had a sheet resistance of about 270, 148, and 80 Ω/sq for one, two, and four layers, respectively. Such a substantial difference in the sheet resistance could have affected the driving voltage of large OLEDs, but no significant difference was observable for the present 2 × 2 mm² devices. The EL main peak remained at a same position for all the OLED devices, as shown in Figs. 4(c) and 4(d). The shape of the EL spectrum however slightly changed by introducing the scattering layer. The planar graphene-OLEDs showed a small shoulder at around 560 nm, which is related to the microcavity effect. Without the scattering layer, the reflectance of one, two, and four graphene layers on a glass substrate were obtained as 0.42%, 0.85%, and 2.38% at 550 nm, respectively, which accounts for the presence of this shoulder as well as its graphene-thickness dependence through the microcavity effect. Introducing a scattering layer beneath the graphene might have hindered this effect via random scattering.

 

Fig. 4 J-V and EL spectra characteristics of graphene-OLEDs; planar devices ((a) and (c)) and scattering devices ((b) and (d)) the scattering layer. The number of the graphene layer was varied for one, two, and four and EL spectra of those OLEDs were measured at the normal direction.

Download Full Size | PDF

Figure 5 plot the external quantum efficiency (EQE) (Fig. 5(a)) and luminous efficacy (LE) (Fig. 5(b)) of the graphene-OLEDs with and without the scattering layer, as a function of the number of graphene layers. They also show the enhancement ratio. For the same number of graphene layers, the scattering devices exhibit much higher efficiencies than the planar devices. This improvement can be understood by considering the scattering effect, which effectively compensates the optical losses that occur during the internal reflections. In general, the absorption of a transparent electrode lowers the efficiency of the OLED device. Decreasing the number of graphene layers from four to one thus improves the EQE and LE. The amount of the improvement however differs between the two devices; while EQE and LE increase only by 8.22% and 7.96% in the planar devices those for the scattering device amounts as large as 16.14% and 13.24%, respectively. This result is consistent with the simulation result depicted in Fig. 2(b). The backscattering and the reflection caused by the scattering layer may create complicated optical paths inside the OLEDs [45], thereby increasing the absorption rate of the graphene. Since this absorption increases with the number of graphene layers, the impact of the number of graphene layers on the efficiencies should be more pronounced in the scattering devices than in the planar devices, which accounts for the present result. Specifically, decreasing the number of graphene layers from four to one in the scattering device increases the EQE and LE enhancement from 39.6% to 52.78% and from 40.95% to 49.52%, respectively.

 

Fig. 5 The EQE (a) and LE (b) of graphene-OLEDs as planar devices and scattering devices. the number of graphene layers was varied as one, two, and four

Download Full Size | PDF

The angular-dependent luminance (L) of the devices was measured as a function of the number of graphene layers (Figs. 6(a) and 6(b)). For both devices, the one with fewer graphene layer numbers exhibits a higher L in all directions. As was expected, L is greater for the scattering devices and increases with decreasing number of graphene layers. For devices with fewer layers of graphene in particular, the impact of the light absorption by graphene is more significantly reduced in the scattering devices than in the planar devices. Furthermore, the scattering structure diffuses the incident light to a wider range of angles.

 

Fig. 6 The angular-dependent luminance (L) of the graphene-OLEDs as planar devices (a) and scattering devices (b) with the variation of the graphene layer as one, two, and four.

Download Full Size | PDF

Based on the above results, we compared the property of the single-layer graphene (SLG)-OLED with that of IZO-OLEDs (as a typical OLED). Except for the transparent electrodes, the IZO-OLED (thickness: 100 nm) had the same organic stack as that of the SLG-OLEDs. Figure 7(a) summarizes the EQE and the LE of the devices. The planar SLG- and IZO-OLEDs show similar EQE and LE values when t_HTL = 250 nm is shared by both of the devices. Introduction of a scattering layer minimizes the microcavity effect, which on the other hand alters the optical traveling path and thereby contributes to the elimination of internal reflection loss [33]. Thus the use of the scattering layer increased the EQE and the LE values by almost the same amount for both types of the devices. The scattering layer made the OLEDs less dependent on the cavity length, and, hence, the organic thickness [38]. These experimental results demonstrate the possibility of achieving graphene-OLEDs with efficiencies similar to those of conventional OLEDs with oxide anodes. The 1931 International Commission on Illumination (CIE) color coordinates of the SLG and IZO devices were extracted from the EL spectra (Fig. 7(b)). Compared to the IZO-OLEDs, the CIE coordinates of the SLG-OLEDs remained almost constant, less than ± 0.001 and ± 0.002 for x and y coordinates, respectively, over a wide range of viewing angles, which corresponds to a negligible color shift. This is presumably due to the very low reflectance of the graphene surface. The proposed device is also very useful for stabilizing the angular EL spectra (Figs. 7(c) and 7(d)). Although the scattering layer cannot totally eliminate the microcavity because its SiN_x surface acts as a weak mirror, the distortion of the EL spectrum of the graphene-OLEDs caused by the residual microcavity effect is negligible compared to that exhibited by IZO-OLEDs.

 

Fig. 7 The Comparison of (a) EQE and LE, (b) 1931 Commission internationale del’éclairage (CIE) color coordinates and (c), (d) EL spectra of SLG-OLEDs and IZO-OLEDs with and without the scattering layer. EL spectra of those OLEDs were measured at the normal direction.

Download Full Size | PDF

4. Conclusion

To improve the efficiencies of graphene-OLEDs, we evaluated the optical absorption of graphene and explored the use of a random scattering layer containing an array of aperiodic SiO_x nano-pillars and a SiN_x planarization layer. A comparison between simulations and device characterizations have shown that the optical absorption and the substantial reduction of the microcavity effect limit the efficiencies of the OLED with graphene anodes. Introduction of a scattering layer between the substrate and the graphene anode was found to enhance EQE and LE by more than 50%. As a result, we succeeded in the fabrication of graphene-OLEDs having efficiencies comparable to those of conventional OLEDs with oxide anodes. Furthermore, the angular EL spectrum variations were stabilized, which is difficult to achieve using conventional OLEDs with microcavity designs.