1. Introduction

One of the most important issues in advancing solid-state lighting is to improve the light extraction efficiency of GaN-based light-emitting diodes (LEDs). Photons emitted from LED devices tend to be trapped within LED chips by the total internal reflection (TIR) at the interface between the indium tin oxide (ITO) transparent electrode and air/resin, which leads to a low light extraction efficiency [1]. The TIR at this interface can be effectively decreased by creating nano-scale patterns on the ITO electrode surface that have pattern dimensions smaller than the wavelengths of emitting light to allow more light emission out of the LED. Many ITO surface nano-patterning techniques, such as fabrication of photonic crystals [2,3], growth of nanowires [4], and surface texturing by photolithography or nano-imprinting [5–8], have been adopted to date, with apparent enhancement of light extraction efficiency [9,10]. However, current ITO nano-patterning methods either require expensive and time-consuming processes (such as photolithography and nano-imprinting) or involve harsh processing environments (such as high-temperature growth of nanowires or dry plasma etching). As a result, the underlying thin film components in the LED are quite often damaged after nano-patterning, and the overall device efficiency is therefore decreased.

Recently, a simple maskless wet dip-etching technique for surface nano-patterning was introduced [11,12]. The LED with an as-deposited (before annealing) ITO layer was dipped into an etching solution for less than 10 seconds at room temperature. The short etching time and the room temperature process caused virtually no damage to other material layers of the LED. As compared to LEDs with a smooth ITO surface (referred to as “reference LEDs”), LEDs with ITO surfaces patterned with this method demonstrated a significant 30% increase in output light efficiency [10]. Although this low-cost process is promising for batch production of such highly efficient LEDs, several problems persist. First, the dip-etching time to obtain such a nano-patterned ITO surface is too short (<10 seconds) to make the process controllable. Second, after etching, the effective thickness of ITO for electronic conduction is decreased, which results in a significant increase in both the sheet resistance of ITO and the operating voltage of the LED, and consequently the device efficiency is sacrificed.

In this work, we present a novel design of a “dual-layer” ITO electrode, featuring a controllable process for a nano-patterned ITO surface with simple wet dip-etching, and at the same time demonstrate a large improvement in the light extraction efficiency of GaN-based blue-light LEDs.

2. Experimental

To compare the light extraction efficiency in this study, we prepared LEDs with three type ITO electrode as follows:

The dual-layer nano-structured ITO electrode (DL400/400) was fabricated by first depositing a 400 nm-thick bottom ITO layer and then annealing at 600 °C for 1 minute in air so that the layer would be resistant to the subsequent wet etching process. A second ITO layer with a thickness of 400 nm was deposited on the annealed ITO layer. The nano-structured ITO layer was created by dipping the as-deposited second ITO layer in a buffered oxide etchant (BOE) solution diluted (1:6) with de-ionized (DI) water. The surface nano-structure was controlled by etching in the BOE solution for 0, 20, 40, 60, 90, and 120 seconds for observation of the evolution of electrode morphology. After etching, the ITO layer was annealed at 600 °C for 1 minute in air. For the single-layer nano-structured ITO electrodes (SL400 and SL800), the as-deposited layer and the nano-structured surface were patterned with the same conditions as the dual-layer ITO top electrode.

LED devices with ITO electrodes of SL400, SL800, and DL400/400 and a non-patterned reference electrode were fabricated by the following procedure. Epitaxial layers for a blue-LED (wavelength = 405 nm) consisting of undoped GaN, Si-doped n-GaN, an InGaN/GaN multiple-quantum-well active layer, and Mg-doped p-GaN were grown on sapphire substrates by a metal-organic chemical vapor deposition system. The blue-LED devices of 300 × 300 µm chip size were fabricated with mesa etching by inductively coupled plasma. A Ti/Au (50/200 nm in thickness) layer was deposited by e-beam evaporation as the n-type electrode. The Ohmic contact of the p-GaN layer was made by depositing the p-contact transparent copper indium oxide (3 nm) /ITO electrode onto p-GaN by e-beam evaporation, and the film was subsequently annealed at 600 °C in pure oxygen ambient for 1 minute. The LED structure was completed by deposition of the bonding pad electrode of Cr/Au (50/200 nm) on top of the ITO transparent electrode by e-beam evaporation.

The surface morphology of ITO electrodes was observed by scanning electron microscopy (SEM) (Hitachi, S-4700) and atomic microscopy (AFM) (NanoFous Inc.; non-contact mode). Current–voltage curves and light output–current measurements were carried out by a parameter analyzer (HP 4155A) and a Si photodiode (818-UV/CM) connected to an optical power meter, respectively. The pictures of LED lighting were taken at input current of 100 μA by the optical microscope of a probe station (MS Tech, MST-6000C).

3. Results and discussion

We have successfully demonstrated an ultra-high brightness and stable turn-on voltage ITO/GaN thin film LED with such a dual-layer ITO electrode. Figure 1 illustrates the schematics of LEDs with a single-layer (Fig. 1(a)) and a dual-layer (Fig. 1(b)) ITO electrode. Both the single-layer and the dual-layer design have a nano-structured top layer patterned by the dip wet-etching process to minimize the light-trapping by TIR. For the dual-layer ITO, an additional dense ITO layer is pre-deposited and annealed underneath the top patterned layer to maintain a good overall in-plane electronic conduction.

 

Fig. 1 Schematic of LEDs with (a) single-layer ITO electrode, and (b) dual-layer ITO electrode. The dual layer ITO has an additional current spreading layer to maintain good electronic conduction during device operation.

Download Full Size | PDF

To optimize the light extraction efficiency, the evolution of ITO surface morphology with dip-etching time was first investigated. We compared the surface morphology between the following ITO electrode configurations: SL400, SL800, and DL400(top)/400(bottom), as schematically shown in Fig. 2.

 

Fig. 2 The SEM images of (a) SL400, (b) SL800, and (c) DL400/400 ITO surface morphologies with various dip wet-etching time, from as-deposited to 120 seconds of etching time.

Download Full Size | PDF

The SEM images of etched surface morphologies are shown in Fig. 2. For SL400 and DL400/400, the patterns were uniformly distributed for etching times of 20 and 40 seconds, while the size and inter-distance of patterns increased and became rather random as etching time increased. For SL400 and DL400/400, the morphology evolutions with etching time were very similar, as expected, considering the same initial thickness of the etched ITO layer (400 nm). For SL800, the morphology evolution was also similar to those of SL400 and DL400/400, except that after 60 seconds, fine nano-bumps were still visible at the bottom of the surface, possibly due to the thicker initial ITO layer.

The sheet resistances of SL400, SL800, and DL400/400 electrodes with different dip-etching times are shown in Fig. 3(a). The sheet resistance of SL400 increased significantly from 20 to 250 Ohm/sq at 60 seconds of etching time. Such a huge jump in sheet resistance was not observed in SL800, even after 120 seconds of etching; the sheet resistance remained as low as 50 Ohm/sq. This indicated that most of the 400 nm ITO layer in SL400 was etched out after 60 seconds, while in SL800, a finite thickness of ITO remained un-etched and thus a good current-spreading layer. On the other hand, the sheet resistance of the DL400/400 ITO electrode had only a slight increase from 20 to 22 Ohms/sq, even after the longest etching of 120 seconds. These results justified that the sheet resistance of ITO is largely affected by the etching and can be successfully kept low by the additional pre-deposited and annealed bottom layer, which serves as an effective current spreading layer.

 

Fig. 3 (a) The evolution of sheet resistance of SL400, SL800, and DL400/400 ITO electrode as a function of dip wet-etching time; (b) The evolution of sheet resistance vs. etching time for 200, 300, and 400 nm-thick of the dense bottom ITO layer. (c) The current-voltage-current curve for LEDs with DL400/400 ITO electrode etched for various times.

Download Full Size | PDF

To determine the optimized thickness for effective current spreading, we also investigated the effect of the bottom ITO thickness on the overall sheet resistance of the dual-layer electrode. We examined the overall sheet resistance of samples with three bottom ITO thicknesses: 200 nm (DL400/200), 300 nm (DL400/300), and 400 nm (DL400/400). The sheet resistance behavior trend with dip-etching time is shown in Fig. 3(b). For all three electrode configurations, the sheet resistance still increased with etching time, but not as significantly as in the cases of single layer electrodes. Considering that the initial top ITO thickness is the same, the differences in overall sheet resistance should be essentially due to the different bottom layer thicknesses, and as can be expected, a thicker bottom layer lowered the sheet resistance. Further increasing the bottom ITO thickness to over 400 nm may cause only trivial decreases in sheet resistance. For economy, minimizing the layer thickness is preferred in order to reduce the material consumption as well as to reduce the annealing time needed to fully oxidize the as-deposited ITO layer.

Since the BOE solution has very good etching selectivity between the as-deposited (amorphous) top ITO layer and the annealed bottom ITO (polycrystalline) [10], the bottom layer was expected to remain nearly intact after the wet dip-etching. We verified this by fabricating LED devices and increasing the wet dip-etching times on DL400/400 electrodes, after which we measured their I-V characteristics (Fig. 3(c)). The turn-on voltage and the forward voltage at 20 mA were 2.7 V and 3.4 V, respectively. The I-V curves of DL400/400 LEDs with dip-etched ITO electrodes remained consistent with LEDs with un-etched ITO, even after the longest etching time of 120 seconds. This shows that the annealed bottom ITO is resistive to long wet dip-etching without sacrificing device performance. Therefore, adding a pre-annealed bottom ITO layer is an effective way to stabilize the current spreading and helps to prevent electrical degradation during operation.

Regarding the simplicity of the process, if the same initial thickness of ITO electrodes (SL800 and DL400/400, for example) is to be fabricated, it appears that the single-layer ITO is preferable, since it requires only one deposition and one annealing step, while the dual-layer ITO requires two separate deposition and annealing steps. However, to obtain good optical transmittance, the annealing time must be sufficient to fully oxidize the as-deposited ITO into completely transparent ITO. Since the oxidation is a diffusion process, a significantly longer annealing time is required to fully oxidize an 800 nm metal film in one single annealing step than to fully oxidize a 400 nm metal film. Indeed, relatively higher optical transmittances in LEDs with SL400 and DL400/400 than in those with SL800 for etching times beyond 60 seconds were observed (results not shown in this paper). Also, if annealing time is too long, more energy will be consumed during manufacturing, and the material and device properties of other components in the LED, such as p-GaN ohmic contact, could be damaged.

The variation of light intensity between DL400/400 ITO-electrode LEDs with different ITO surface morphologies, achieved by increasing etching time, is shown in Fig. 4(a). The light intensity increases with dip wet-etching time, reaching a maximum at 60 seconds, and decreases thereafter. We take this DL400/400 electrode with 60 seconds etching time as our most optimized electrode fabrication condition. Compared to the 10 seconds of etching time previously reported in [9], 60 seconds is practically long enough to be controllable for mass production. The output light intensity of an LED with this optimized dual-layer ITO electrode was compared to a reference LED with no pattern on its ITO electrode. As shown in Fig. 4(b), the light intensity was much higher for the dual-layer LED than for the reference LED at all wavelengths. At 100 mA of input current, the light intensity is 2.18 times higher than that of the reference LED. The optical microscope image in the inset of Fig. 4(b) shows that the brightness of the DL400/400 LED is much higher than that of the reference LED at an input current of 100 µA. This enhancement resulted from the etching induced surface nano-patterns, which have dimensions smaller than the wavelength of light. The nano-patterns serve as scattering centers to allow photons to escape through the ITO surface, in contrast to surface nano-patterns with dimensions larger than the wavelength of light, which only reflect photons backwards [12].

 

Fig. 4 (a) The variation of light intensity between DL400/400 ITO-electrode LEDs with different ITO surface morphologies achieved by increasing etching time; (b) The comparison of light intensity between a reference LED with un-etched ITO electrode and an LET with DL400/400 electrode. The light intensity was much higher for the dual-layer LED than for the reference LED at all wavelengths; (c) The AFM surface morphology of the top ITO layer after wet dip-etching for 60 seconds with the average size of nano-bumps to be 63.7 nm.

Download Full Size | PDF

Considering the wavelength of 450 nm in our LED devices, DL400/400 electrode etched for 60 seconds may give the optimized size distribution of nano-patterns that scatters the most output light. The AFM surface morphology of the top ITO layer after wet dip-etching for 60 seconds is shown in Fig. 4(c), where the average size of nano-bumps is 63.7 nm. Longer etching time generates larger patterns (>500 nm) that reflect a larger portion of the light, instead of assisting light extraction, which results in a comparably lower output power. This result is in agreement with previous observations [11]. Therefore, to optimize the light extraction efficiency of practical LED devices, it is essential to control the structural dimensions of surface patterns.

4. Summary and conclusion

In summary, we have demonstrated the fabrication of a dual-layer nano-structured ITO transparent electrode patterned by a simple maskless dip wet-etching process that effectively increases the light extraction efficiency of GaN-based blue-LEDs. The dual-layer ITO electrode features a nano-structured top layer with optimized surface nano-patterns to largely reduce the entrapment of light in the LED device, and a dense bottom layer that acts as a stable current-spreading layer. The light output power at 100 mA of the DL400/400 ITO electrode LED is 2.18 times higher than that of a reference LED. The existence of the dense bottom current-spreading layer under the patterned layer stabilizes the dual-layer electrode electrical conduction, and no significant degradation in sheet resistance has been observed. The dual-layer ITO design and fabrication process can effectively improve the light extraction efficiency of a GaN blue-LED without sacrificing its electrical properties. The nano-patterning is done by a controllable wet dip-etching time of 1 to 2 minutes, and the annealing time (1 minute) in a high temperature furnace is much shorter than the time needed for growing nanowires, which takes hours of processing at high temperatures. The simple, low-cost, and controllable process is readily applicable in mass production to high brightness blue light LEDs and is likely to advance the spread of blue-light LEDs in the market.