1. Introduction

Organic light-emitting diodes (OLEDs) have raised much interest during the last decade. There have been OLED based displays being used in consumer electronics for a few years and the first steps towards efficient, long lifetime general lighting devices have been made [1]. The advantages of OLED technology lie in the low power consumption, the wide variety of colors, the possibility to use flexible substrates and the cost-efficient production. In 2009, Reineke et al. showed a complex and optimized OLED stack emitting white light with an efficiency comparable to fluorescent tubes [2]. However, commercially available OLEDs still lack these efficiencies. While the conversion efficiency of injected electrons into photons (internal quantum efficiency) has reached almost 100 % over the last years [3], there is still a lot of room for improving the light extraction efficiency. The loss mechanisms in OLEDs are well understood [4–6]. The major part (about 80 %) of the photons being generated in the emission layer cannot exit the device. About 50 % are trapped as guided modes in the indium tin oxide (ITO) anode and the organic layers or as surface plasmon polaritons at the metal/organic interface, where they are finally absorbed. Another 30 % of the photons are lost due to total internal reflection at the substrate/air interface. Thus, only about 20 % of the generated photons can leave the device as useful light [4,7,8]. There have been great efforts to extract the guided modes from the device. A well-known method is to corrugate the metal/organic or ITO/organic interface to extract the waveguide modes [9–13]. Other groups have reported on devices containing low-index aerogel layers between the emissive layer and the glass substrate [14] or low index grids within the device [15] to increase the extraction efficiency. There have also been approaches to embed nanoparticles inside the organic layers to increase the internal quantum efficiency [16] or to scatter out guided modes and substrate modes [17]. By modifying the substrate/air interface with microlenses [18], meshed surfaces [19] or sandblasting [20] the substrate modes may be extracted efficiently [21].

Here, we report on a device stack design and method, which may extract light trapped in the waveguided modes and the substrate modes by scattering [22]. We textured the glass substrate surface with three different roughening methods. On top of the rough surface a highly transparent Ta₂O₅-layer, which exhibits a refractive index of n = 2.1, is evaporated. This layer acts as the waveguide layer, confining the light due to its high refractive index. In order to smooth the rough surface of this layer, a transparent photoresist layer with a smaller index of refraction is brought on top. The ITO-free OLEDs that are fabricated upon these modified substrates possess an enhanced efficiency compared to OLEDs fabricated on flat substrates (but also featuring the Ta₂O₅- and photoresist-layer) (Fig. 1).

 

Fig. 1 Schematic diagram of a structured OLED. Please note, that the lateral and vertical lengths are not true to scale.

Download Full Size | PDF

This article is structured as follows. In section 2 we describe the three different fabrication methods that lead to rough substrate surfaces. In section 3 we show how the rough surfaces are finished and characterized via atomic force microscopy (AFM). Section 4 deals with the fabrication and characterization of the ITO-free OLEDs.

2. Fabrication of rough interfaces

We use standard 1 mm thick soda-lime glass substrates that are cut into pieces of 25 mm x 25 mm. Our structuring processes are sandblasting, abrasion and etching. Those three different methods result in rather different surface topographies. Sandblasting is performed with an aluminum oxide abrasive (grain size 50 μm, Girrbach Dental, Germany), which is shot onto the glass surface with a pressure of 3 bar. The emerging surface is shown in the atomic force microscope (AFM) image in Fig. 2(a). The root mean squared roughness is R_q = 250 nm and the peak-to-valley roughness is about R_t = 1.4 μm.

 

Fig. 2 AFM images of roughened glass substrates textured via sandblasting (a), grinding paste (b) and glass etching cream (c).

Download Full Size | PDF

Roughening by abrasion is done with a grinding paste (METADI Diamond Polishing Compound, grain size 3 μm, BUEHLER, Germany), which is spread onto a large glass support plate together with a droplet of dish washing detergent and a little bit of water. Then the glass substrate is put on top of the paste and lapped in the mixture using a ”figure-eight” pattern for 15 minutes. The resulting surface topography is shown in Fig. 2(b). It features a R_q of 67 nm and a R_t of about 400 nm.

The third roughening method utilizes glass etching cream (Glasotan, CREARTEC trend-design-gmbh, Germany), which is applied onto the glass substrate for 10 seconds. The corresponding R_q and R_t are 150 nm and 550 nm, respectively (Fig. 2(c)). All roughness values are also given in Table 1. After the roughening treatment the substrates are cleaned with acetone and then with isopropanol in an ultrasonic bath.

Table 1. Roughnesses R_q and R_t of all samples for each fabrication step.

View Table

3. Finishing and characterization of the rough interfaces

Since the roughnesses of the samples being structured either with glass etching cream or by sandblasting are too large and possess too many pronounced peaks, we polish those textured substrates with a lapping disk on a lapping maschine (PHOENIX 4000 Sample Preparation System, diamond grain size 3 μm, BUEHLER, Germany ) for 10 minutes. The sample, which was roughened with grinding paste, is not polished, as its surface did not exhibit sharp peaks. The resulting surface of the sandblasted sample is shown in Fig. 3(a) and reveals a R_q of 200 nm and a R_t of 900 nm.

 

Fig. 3 AFM images of the polished glass substrates textured via sandblasting (a) and glass etching cream (b).

Download Full Size | PDF

The polished surfaces of samples structured with glass etching cream exhibit a R_q of 24 nm and a R_t of 96 nm (see Fig. 3(b)).

Now the roughness is in a suitable range; the peaks have vanished and mainly dips are visible (see Table 1). Next, the samples are coated with a 50 nm thick layer of Ta₂O₅. This is done in high vacuum by electron beam evaporation. Ta₂O₅ is transparent in the visible region and features a high refractive index of n = 2.1. It acts therefore as a waveguide layer, confining and separating light from the cathode where a major part of light is usually lost. The evaporated Ta₂O₅ layer forms a very good replica of the glass substrate topography. The layer thickness was controlled during the evaporation process by an oscillating quartz crystal. SEM measurements made afterwards confirmed this thickness. Even after polishing, the roughness is still too high to fabricate an OLED on top of it. In order to smooth the surface, a 200 nm thick layer of the negative tone photoresist SU-8 (Microchem SU-8 2000.1) is spin coated on top of it. After a short prebake, the photoresist is illuminated with a uv-exposure unit (proMa Technologie GmbH) at 365 nm wavelength. The resulting SU-8 layer is transparent and robust. Although the resulting surfaces are not completely flat (see Fig. 4), they are all suitable for OLED fabrication.

 

Fig. 4 AFM images of the photoresist-smoothed glass substrates textured via sandblasting (a), grinding paste (b) and glass etching cream (c).

Download Full Size | PDF

4. Fabrication and measurement of OLEDs

The finished substrates are cleaned with acetone and isopropanol in an ultrasonic bath for 10 minutes, respectively. The surface of the photoresist is then exposed to an oxygen plasma to make it hydrophilic and to remove organic residues. Afterwards, the polymer anode which is a mixture of poly(3,4-ethylenedioxythiophene): poly(styrenesulfonate) (PEDOT:PSS; CLEVIOS PH 750 purchased from H.C. Starck) and 5 % of dimethyl sulfoxide (DMSO) is prepared. It is processed by spincoating the solution at a speed of 3000 rpm onto the substrate. The thickness of the resulting layer is 80 nm and it has a conductivity of about 650 S/cm. It is then partly covered with a stainless steel mask and exposed to an oxygen plasma for 5 minutes to create the anode pattern in the PEDOT:PSS layer [23]. All following fabrication steps are conducted in a glove box under nitrogen atmosphere. In order to eliminate the residues of water in the polymer anode, the substrates are baked in a vacuum oven at 130° C for 30 minutes. On top of the PEDOT:PSS anode the emission layer is applied. It consists of phenylene substituted poly(para-phenylenevinylene) (Ph-PPV; also known as Super Yellow from Merck OLED Materials GmbH). The Super Yellow is solubilized in toluene with a concentration of 3 mg/ml and then spincoated at 1000 rpm resulting in a layer thickness of about 70 nm. The OLEDs are then metallized with 50 nm of calcium as the cathode and 200 nm of aluminum as a protective layer (see Fig. 1).

Since the optical measurement setup is located in normal ambient conditions, the OLEDs are encapsulated with an epoxy resin adhesive and a glass cover. J-V curves and the electroluminescence of all OLEDs are measured with a source-measure unit (Keithley SMU 236) and an integrating sphere (Gigahertz-Optik, UMBB-210) being connected via a multimode fiber to a spectrometer (Acton Research Corporation SpectraPro-300i) with an intensified charge-coupled device (Princeton Instruments PiMax:512).

In Fig. 5 the J-V-characteristics of all structured OLEDs and an unstructured reference OLED are shown. The onset voltage for all devices is the same (typically 2.5 V). Compared to OLEDs containing an ITO anode this is about 1.5 V smaller, which is due to the smaller difference of the HOMOs (highest occupied molecular orbital) of PEDOT:PSS and Super Yellow. Since the lateral resistance of the PEDOT:PSS anode is higher than that of ITO, the slope of the curves in the J-V-characteristics are smaller. There is a rather large variation in the J-V-characteristics visible which is due to the not exactly equally large active OLED areas and particularly due to the different cross sections of the polymer anode. A possible reason for that might be that the surface is partially too rough for an efficient PEDOT:PSS anode resulting in a high voltage drop along the anode.

 

Fig. 5 J-V-characteristics of the structured OLEDs and an unstructured reference device. The OLEDs reveal almost the same onset voltage. The reason for the difference of the curves for higher voltages is attributed to the different thicknesses of the PEDOT:PSS anodes.

Download Full Size | PDF

To investigate whether this affects the efficiency one has to examine the dependence of the luminous flux on the current density, which is given in Fig. 6.

 

Fig. 6 Luminous flux as a function of the current density. Those samples which have been modified by sandblasting and grinding paste perform better than the reference sample. The sample roughened with glass etching cream could not reach the efficiency of the reference device.

Download Full Size | PDF

From this graph, the improvement in the performance of the samples roughened by sand-blasting and grinding paste can be deduced to 37.4 % and 15.5 % at 20 mA/cm², respectively. Evidently, the voltage drop along the PEDOT:PSS anode cannot account for the poor performance of the sample treated with glass etching cream in Fig. 5. The same argument is applicable for the other, better performing modified samples, too. The origin of the increase in the luminous flux must either lie in an improved internal quantum efficiency or in a better outcoupling efficiency due to light scattering within the substrate and the high index layer. Furthermore, there was no change visible in the emission spectra of the different samples.

The overall performance of the samples depending on the applied power is shown in Fig. 7. The behavior of the curves is comparable to Fig. 6. The sample which was sandblasted reveals the highest light output, owing this to some extent to the best current injection (see Fig. 5). In order to find out more about the outcoupling efficiencies of the structured samples photoluminescence (PL) measurements were performed. The devices were mounted on a goniometer which was rotated around the vertical axis. A UV laser (Newport, Explorer Scientific All Solid State UV Laser, EXPL-349-120-CDRH) was coupled into a UV optical fiber being attached to the goniometer to maintain a constant spot size on the OLED while rotating the goniometer. A fixed multimode fiber was connected to the spectrometer described before. The structured interfaces could also increase the amount of exciting laser light coupled into the modified devices and thus, the number of emitted photons. Therefore, we measured the specular and diffuse reflectance of all samples in a spectrometer. The difference in the reflectance of the samples was found to be small. The results of the PL measurements are shown in Fig. 8.

 

Fig. 7 Luminous flux as a function of the electrical power.

Download Full Size | PDF

 

Fig. 8 Angle resolved photoluminescence (PL) emission of the structured OLEDs and the unstructured reference device. The structured samples exhibit a higher intensity compared to the reference sample. This proves that roughening inner surfaces in OLEDs increases their outcoupling efficiency.

Download Full Size | PDF

Apparently, the samples with roughened interfaces exhibit higher outcoupling efficiencies compared to the reference device. Furthermore, there are small differences in the angular emission profile. The sample which was treated by sandblasting reveals an almost Lambertian profile closely followed by the sample being treated with grinding paste. The sample which was structured with glass etching cream shows the best light output in forward direction, whereas at higher angles its performance forfeits as compared to the other structured samples. Since in the PL measurement all structured samples revealed a higher outcoupling efficiency, the poor performance of the sample roughened with glass etching cream is attributed to a drop in the internal quantum efficiency. The reason for that might lie in the surface of the SU-8 smoothing layer, which still might possess local areas where the roughness is too high to guarantee smooth organic layers.

We attribute the improved light outcoupling to both, waveguided modes and substrate modes [22].

5. Conclusion

We demonstrated three different methods for roughening glass substrates for ITO-free OLEDs: abrasion by sandblasting, abrasion by the use of grinding paste and wet etching with glass etching cream. We showed that all three methods can produce substrate roughnesses, which are suitable for improving the outcoupling efficiency of OLEDs by scattering waveguide modes and substrate modes. When driven electrically, OLEDs with rough interfaces made by sand-blasting and grinding paste revealed an enhancement of 37.4 % and 15.5 % (at 20 mA/cm²), respectively. This enhancement is mainly attributed to a better outcoupling efficiency, whereas the surface roughness might also have a small impact on the internal quantum efficiency.