1. Introduction

Transparent conductive electrodes (TCEs) with high optical transmittance and good electrical conductivity are a key component for improving the quantum efficiency of diverse optoelectronic devices such as light-emitting diodes (LEDs) [1,2], photovoltaics [3], sensors [4], transparent electronic memory [5], touch screens [6], and flat-panel displays [7]. When TCEs are applied to various types of semiconductor devices, one of the most important problems is to obtain an ohmic contact between the TCEs and semiconductors for effective current injection and distribution. Traditionally, the ohmic contact was achieved by matching the work functions of the electrodes and semiconductors, together with surface state modifications of the semiconductors using intentional doping with impurities, surface, and/or thermal treatments.

Indium tin oxide (ITO) is currently the most widely used p-type TCEs in gallium nitride (GaN)-based LEDs. However, ITO has some drawbacks such as increasing cost of indium due to its limited supply, and high-temperature annealing for crystallization and/or ohmic contacts to semiconductors [8–11]. Moreover, a large work function (Φ) difference between the ITO (Φ = ~4.8 eV) and p-type (Al)GaN (Φ > 7.5 eV) leads to a large Schottky barrier height at the interface, making it difficult to form ohmic contacts [12].

Accordingly, various materials such as thin metal films [13], metal nanowires [14], carbon nanotubes [15], graphene layers [16], and conductive polymers [17] have been developed to replace ITO. However, the performance of LEDs using these materials as the TCEs has been still below that of the ITO-based LEDs due to a trade-off between electrical conductivity and optical transmittance [18,19].

Recently, we have reported a novel ohmic contact method using electrical breakdown (EBD) of wide-bandgap TCEs, such as silicon nitride (SiN_x) [20,21], and successfully applied the TCE to GaN-based vertical LEDs (VLEDs) as an n-type electrode [22]. However, some questions related to the reliability problems that might occur during the EBD process have been raised.

In this study, we investigate the effect of conducting filament (CF) density on the device performance using three GaN-based VLEDs with three different (i.e., low, medium, and high) CF densities. Then, the optimum density of the CFs in SiN_x TCEs is discussed by comparing the performance of the three proposed VLEDs with the conventional one with metal electrodes.

2. Experiment

2.1 Fabrication of GaN-on-Si-based VLEDs

The GaN-on-Si-based LED wafers emitting at 455 nm were grown by metal-organic chemical vapor deposition on the silicon (111) substrates, which consist of a 0.2-μm-thick AlN buffer layer, a 2-μm-thick undoped GaN layer, a 3.5-μm-thick Si-doped n-type GaN layer, six pairs of InGaN/GaN multiple-quantum-well layers, and a 150-nm-thick Mg-doped p-type GaN layer. Before wafer bonding, ITO/Ni/Ag multilayers were deposited for p-type ohmic contact reflection layers, and Au/In layers were used as the bonding metal. The GaN-based LEDs grown on the Si (111) substrate was bonded to the Ag/Au/Cr layers deposited on (100)-oriented Si substrates using thermal pressure bonding for 1 h at 220 °C. Then, the Si (111) substrate, AlN buffer, and undoped GaN layers were removed using chemical lift-off and inductively coupled plasma dry etching process. Standard photolithography and inductively coupled plasma reactive ion etching (ICP-RIE) were employed to define isolated mesa structures. Then, the 20-nm-thick SiN_x TCE films, with isolated mesa structures, were deposited on the n-GaN layers using an radio frequency (RF) magnetron sputtering system in Ar–O₂ (or Ar–N₂) gas environments at a base pressure of ~2 × 10⁻⁷ Torr and working pressure of ~3 × 10⁻³ Torr. After that, the EBD process involving voltage application was conducted on the surface of SiN_x/n-GaN in VLEDs to provide current paths between the SiN_x TCEs and n-GaN. More details on the EBD process are given in Fig. 1. Finally, Cr/Al/Ni/Au was deposited on the n-GaN surface as an n-type electrode by electron beam evaporation. For comparison, VLEDs with metals (denoted as conventional VLEDs) were also fabricated using the same LED epitaxial wafer under the same fabrication conditions. The size of the LED chip was 1.5 mm × 1.5 mm.

 

Fig. 1 Schematic illustration and electrical conduction mechanism of the GaN-based VLED with CF-based SiN_x TCEs. (a) Schematic view of the GaN-based VLED with CF-based SiN_x TCEs after EBD; a magnified figure shows that the CFs consist of chains of nitrogen vacancies within the crystal structures of SiN_x. (b) I–V characteristic curves measured for the 20-nm-thick SiN_x TCE/n-GaN before and after EBD. (c) Current levels of LRS at 1 V as a function of retention time, indicting long-term stability. Inset of (c) shows the current levels within 1 V measured for the 20-nm-thick SiN_x TCE/n-GaN before and after EBD. (d) CAFM images of the SiN_x TCEs before (left panel) and after (right panel) EBD measured at a reading voltage of 1 V and 5 μm × 5 μm contact area.

Download Full Size | PDF

2.2 Electrical and Optical Characterizations

First, transmission line model (TLM) patterns were fabricated on the n-GaN layer to evaluate the contact resistance of the proposed SiN_x-based TCEs. The TLM patterns with varying spacing from 5 to 25 μm were fabricated using a photolithographic process. Then, the 20-nm-thick SiN_x TCEs were deposited on top of the n-GaN layer using an RF magnetron sputtering system in Ar–N₂ gas environments at a base pressure of ~2 × 10⁻⁷ Torr and working pressure of ~3 × 10⁻³ Torr. Then, an Ag top-metal electrode was deposited using an electron beam evaporator and a lift-off process in acetone was performed to define the TLM patterns of 100 μm × 100 μm. The EBD process and electrical characterization were performed using Keithley 4200 semiconductor parameter analyzer (Keithley 4200 SPA). The compliance current was limited to 10⁻² A and the voltage was swept until the reversible breakdown occurred (near 20 V) with a margin of ± 5 V.

Next, the 20-nm-thick SiN_x TCE films were deposited on quartz substrates using an RF magnetron sputtering system for optical characterization (i.e., transmittance) of the proposed SiN_x TCEs. After deposition, optical transmittance was measured as a function of wavelength using a Lambda 35 UV/VIS Spectrometer in the range from 280 to 700 nm.

Finally, light output power and operating voltage were measured for full-structure LED chips using a wafer-level LED measurement system (OPI-150, With Light Co., Ltd.). More specifically, the light output power of each LED was measured from the top side of the LED using a Si photodiode connected to an optical power meter. Moreover, light emission images from the LED chip surface were acquired using a photoemission microscope.

3. Results and discussion

Prior to device fabrication, electrical properties of the CF-based SiN_x TCEs on n-GaN were investigated because the SiN_x TCE is in direct ohmic contact with the N-face n-GaN (referred to hereafter as n-GaN) layer in this study. Figure 1(a) depicts the structures of GaN-based VLEDs using CF-based SiN_x TCEs, where the magnified figure shows that CFs can be formed by generating nitrogen vacancies in the SiN_x TCEs to provide carrier injection between the SiN_x TCEs and n-GaN layers. The EBD process (DC bias voltage sweep) in terms of two-point probe contacts (along the metal lines on SiN_x/n-GaN using the Keithley 4200 SPA) was performed to form the CFs in the SiN_x TCEs. When the first sweep voltage (black line) from 0 to 20 V was applied on SiN_x/n-GaN, a sudden increase in current was observed at 15.8 V, as shown in Fig. 1(b). A current compliance was set at 10 mA during the EBD process to avoid any damage to the device. Then, the current linearly increased with an increase of voltage (red line) when the applied voltage was swept again, as is the case of memristors [23]; where the current level at 0.5 V increased from ~3.1 pA to over 10 mA after EBD, as shown in the inset of Fig. 1(c). This sudden transition from a high resistance state (HRS) to low resistance state (LRS) is attributed to the CFs generated in the SiN_x TCEs after EBD. To evaluate the long-term stability of the CFs, we investigated the variation in resistance of the LRS at 1 V using the extrapolation of the best-linear-fitted graph as shown in Fig. 1(c), and found that the LRS is maintained for more than 3 × 10⁸ s (~10 years). This implies that the CF-based SiN_x TCEs are good enough to be used for LED electrodes. We also measured the current mapping images of the CF-based SiN_x TCEs using a conductive atomic force microscopy (CAFM) to verify the formation of CFs in the film after EBD [24–27] as shown in Fig. 1(d). Before EBD, no conductive spot (current level < 1 pA) was observed, whereas a number of nanosize conductive spots related to the current signals of the CFs were observed across a 5 μm × 5 μm scanned area after EBD. The current mapping images were obtained at 1 V, with a compliance current of 10 nA, using a cantilever moving across the film surface at a scan speed of 27 μm/s. These CAFM images showing the existence of conducting spots are well consistent with the widely accepted filament model reported in the literatures [28,29].

Then, we measured the specific contact resistances (ρ_c) of the conventional metal contacts and CF-based SiN_x TCEs on n-GaN layers using the TLM. Figures 2(a), 2(b), and 2(c) show the current–voltage (I–V) characteristic curves and the corresponding total resistances measured for the Cr/Al/Ni/Au metal contacts and CF-based SiN_x TCEs on n-GaN layers, respectively, for pad gap spacings of 5, 10, 15, 20, and 25 μm. In comparison, the conventional metal contacts on n-GaN exhibited quasi-ohmic behavior as shown in Fig. 2(a). On the other hand, the SiN_x TCE contacts on n-GaN exhibited no current flow before EBD, whereas those exhibited perfect ohmic behavior after EBD, as shown in Fig. 2(b). Furthermore, the specific contact resistance of the CF-based SiN_x TCE contacts (ρ_c = 3.8 × 10⁻⁵ Ω·cm²) was approximately one order of magnitude lower than that of the conventional metal contacts (ρ_c = 2.7 × 10⁻⁴ Ω·cm²). This improvement is attributed to the reduced Schottky barrier height between the TCEs and n-GaN because of the locally generated energy levels of the CFs within the SiN_x TCEs, which allows effective carrier injection and transport from the metal contacts to n-GaN across the TCEs via the CFs [20–22,30–33]. Then, we measured the optical transmittance of the proposed SiN_x TCEs as a function of wavelength using a quartz substrate as shown in Fig. 2(d). It was found that the transmittance of the 20-nm-thick SiN_x TCE is 99% at 455 and 96.2% at 300 nm. In particular, the high optical transmittance is stably maintained in the ultraviolet region as well. Based on these properties, the CF-based SiN_x TCEs are desirable for n-GaN transparent ohmic electrodes.

 

Fig. 2 Electrical and optical properties of the proposed CF-based SiN_x TCEs and conventional metal contacts on n-GaN layers. (a) I–V characteristic curves for different pad gap spacings measured for Cr/Al/Ni/Au metal contacts on n-GaN layers (TLM patterns with pad gap spacings of 5, 10, 15, 20, and 25 μm). (b) I–V characteristic curves for different pad gap spacings measured for the 20-nm-thick SiN_x TCEs after EBD. The inset shows the I–V characteristic curves measured for the 20-nm-thick SiN_x TCEs before EBD. (c) Total resistance for different pad gap spacings of 5, 10, 15, 20, and 25 μm measured for the Cr/Al/Ni/Au metal contacts (upper panel) and CF-based SiN_x TCEs on n-GaN layers (lower panel). (d) Optical transmission spectra of the CF-based SiN_x TCEs on quartz substrates in the wavelength range of 280–700 nm. The inset shows a photograph of the sample placed on a background logo of Korea University. The upper figure of the inset shows the sample structure of the photograph.

Download Full Size | PDF

To examine the effect of the CF density on the performance of LEDs, we fabricated three types of VLEDs, each having a different number of CF spots: 4 spots (low density), 15 spots (medium density), and 25 spots (high density) on the SiN_x/n-GaN layers, which are referred to as L-VLED, M-VLED, and H-VLED, respectively, as shown in Fig. 3(a). For comparison, the VLEDs with conventional metal electrodes (referred to as C-VLED) were also prepared using the same epitaxial wafer under the same fabrication conditions. Figures 3(b) and 3(c) show a comparison of the I–V and leakage current characteristics measured for the four types of VLEDs. The operating voltages of the C-VLED, L-VLED, M-VLED, and H-VLED were 3.23, 3.24, 3.13, and 3.19 V at 350 mA, respectively, as shown in Fig. 3(b). Lower operating voltages of the H-VLED and M-VLED, when compared to that of the C-VLED, may result from the relatively lower contact resistance via the CFs, as discussed in Fig. 2(c). Interestingly, as the number of the CF spots (CF densities) on the CF-based SiN_x TCEs increased, except for the H-VLED, the electrical properties of the VLEDs including operating voltage were improved. In this case, the increase in the number of interfacial CFs within the SiN_x TCEs indicates the increment of the conducting paths and electrical bridges between the metal contacts and n-GaN, thus leading to the increased current injection and transport efficiency in the LED devices. On the other hand, the reverse-bias leakage current characteristics of the four types of VLEDs were measured to examine the influence of the CF densities on device reliability as shown in Fig. 3(c). The leakage currents were approximately −1.58 × 10⁻⁴, −2.01 × 10⁻⁴, −6.27 × 10⁻⁵, and −0.029 A at a reverse voltage of −10 V for the C-VLED, L-VLED, M-VLED, and H-VLED, respectively. All VLEDs except for the H-VLED exhibited similar leakage currents ranging from −10⁻⁴ to −10⁻⁵ A, which are low enough for stable device operation. However, the H-VELD showed a rapid increase in the leakage current due to device damage that might be caused by the EBD-induced electrical shock accumulation because of excessive CF densities. Therefore, we conclude that the EBD with 15 CF spots is desired for our LED structure in terms of device reliability.

 

Fig. 3 Device configuration and electrical performance of the VLEDs with the CF-based SiN_x TCEs and conventional metal contacts. (a) Schematic configuration of the VLEDs with the CF-based SiN_x TCEs for three different CF densities. (b) Current–forward voltage characteristics of the four types of VLEDs. (c) Reverse leakage current characteristics measured for the four types of VLEDs.

Download Full Size | PDF

Figure 4(a) shows a comparison of the light output power for the four types of VLEDs. For all VLEDs, the light output power shows a linear increase as the injection current increases to 350 mA.

 

Fig. 4 Optical performance of the VLEDs with the CF-based SiN_x TCEs and conventional metal contacts. (a) Light output power–current characteristic curves of the VLEDs for different CF densities. (b) Electroluminescence spectra of the VLEDs for different CF densities at 350 mA

Download Full Size | PDF

The M-VLED exhibited a light output power of 7.71 mW, approximately 9.2% higher than that of the C-VELD (7.06 mW), whereas the L-VLED and H-VLED exhibited a relatively lower light output power of 6.76 mW and 6.89 mW, respectively, compared with the C-VELD. The light output power enhancement in the M-VLED can be explained by the increased current injection and efficient transport to the multiple-quantum-well layers because of the optimal CF densities. On the other hand, the reduction in light output power of the L-VLED is attributed to the insufficient current injection and transport because of the lower CF densities, whereas that of the H-VLED is attributed to some optical losses via large leakage paths because of the excessive CF densities. The large current leakage paths in the H-VLED could be regarded as the tunneling current caused by defect generation, thereby reducing the light output efficiency of the LEDs [34]. In the case of electroluminescence spectra, all VLEDs exhibits a similar trend in that the light output power exhibited a dominant peak in the range of 458–460 nm at 350 nm, as shown in Fig. 4(b).

To investigate the effect of the CF densities on the light output enhancement and light intensity distribution of the LEDs, we measured the micrographic light emission images from the surface of the LED chips under 50, 200, and 350 mA current injections, respectively, as shown in Fig. 5. Note that the n-type electrode patterns used in this study have been carefully designed to achieve efficient current spreading and high light output power via optimization between the electrode area and shadowing effect. At a low injection current of 50 mA, the M-VLED exhibited a uniform light distribution and slightly brighter light intensity compared with that of the C-VLED. As the injection current is increased above 200 mA (under a high injection current above 200 mA), the light emission was partially nonuniform and slight current crowding occurred around the electrical contact pad in both LEDs; however, the area of higher light intensity in the M-VLED was larger than that in the C-VLED, indicating that the current spreading of the M-VLED is superior to that of the C-VLED (i.e., the injection current spreads out more effectively in the M-VLED). It should be noted that the device performance was measured for a single n-contact before packaging; therefore, the current crowding effect can be further alleviated by an additional n-contact (i.e., double n-contacts) [35]. This implies that the optimal CF density plays a crucial role in the efficient current injection from the metal to n-GaN via the CFs, and in effective current spreading to reduce the discontinuity between the CFs formed in pyramid (or branchlike) shape across the SiN_x TCE [36,37].

 

Fig. 5 Light emission images of the VLEDs for different CF densities. Comparison of the light intensity distribution of the four VLEDs at 50, 200, and 350 mA. The linear color scale for intensity distribution is given on the right-hand side.

Download Full Size | PDF

Finally, photographs of the cleaved wafer-level VLEDs and the magnified optical microscope image of the fabricated VLED chip with the CF-based SiN_x TCEs are given in Fig. 6(a) to exhibit the wafer-level processing. Figures 6(b) shows the light emission from one of the VLED chips on the wafer. Note that the 6-inch wafer used in this experiment was cleaved to an approximately 75 mm × 75 mm size, which contains a number of the VLED chips.

 

Fig. 6 (a) Photograph of the cleaved wafer-level VLEDs and the magnified optical microscope image of the fabricated VLED chip with the CF-based SiN_x TCEs. (b) Photograph of light emission from the cleaved wafer-level VLEDs with the CF-based SiN_x TCEs.

Download Full Size | PDF

Based on these observations, we concluded that the CF density can influence not only the current spreading and light output characteristics, but also the operating voltage and reliable device operation. Therefore, these improvements can be attributed to the introduction of an optimal CF density within the SiN_x TCEs that led to a minimal optical loss (or high transmittance), better current injection, and superior current distribution over the entire n-GaN surface.

4. Conclusion

In this study, we proposed the CF-based SiN_x TCEs that can inject current effectively into the semiconductors with high transmittance, and investigated the effect of the CF density in the TCEs on the electrical and optical properties of the GaN-on-Si-based VLEDs for device reliability. For this experiment, we prepared three VLEDs, each with a different number of CF spots in the SiN_x TCEs (4, 15, and 25 CF spots), together with a reference VLED with metal contacts for comparison. Among the samples, the SiN_x TCEs with 15 CF spots exhibited the lowest specific contact resistance of 3.8 × 10⁻⁵ Ω·cm², which is approximately one order of magnitude lower than that of the metal contacts (ρ_c = 2.7 × 10⁻⁴ Ω·cm²), and the highest optical transmittance of 99% at 455 nm and 96.2% at 300 nm. The VLEDs with the optimal CF density (15 spots), referred to as M-VLED, exhibited the best performance among the CF samples. The operating voltage of the M-VLED decreased from 3.23 V to 3.13 V at 350 mA (for a single n-contact) whereas its output power increased from 7.06 mW to 7.71 mW at 350 mA, compared with the reference VLED with metal contacts. Furthermore, at this CF density, no electrical degradation was observed with the lowest reverse leakage current level of −6.27 × 10⁻⁵ at −10V. This result indicates that an optimal CF density that led to a minimal optical loss, and an effective current injection and distribution, over the entire n-GaN surface should be carefully considered to guarantee the best performance and reliability when these CF-based SiN_x TCEs are used for various optoelectronic devices.