Introduction
High-performance GaN-based light emitting diodes (LEDs) with high efficiency and excellent reliability have been of technological importance for applications in full color display, automotive lighting, and solid state lighting [1]. In particular, due to fundamental drawbacks such as low Mg-activated carrier concentration (by low activation efficiency of Mg-H complex) and low p-ohmic contact resistance [2-4], GaN-based LEDs having a lateral current path suffer from the current crowding, high turn-on voltage, and high leakage current. To resolve these problems, InGaN-GaN and AlGaN-GaN superlattices (SLs) incorporated into GaN-based LED structures [5-9] have been extensively investigated. It was found that Mg-doped (Al, In)GaN-GaN SLs lead to a considerable improvement in the overall device properties such as output power and turn-on voltage by an improvement of high conductivity of these SLs, originating from an increased acceptor concentration by polarization effect in the strain-induced layers on p-GaN [7].
Recently, according to Jang [10], the use of the electrically reverse-connected Schottky diode (SD) as well as the SLs is very effective to reduce defect-related influence and hence improve the device reliability characteristics of GaN-based LEDs if the SD on p-GaN is electrically reverse-operated. In fact, the GaN-based LEDs suffer from various defects such as threading dislocations and various point defects [4,11-13]. These defects could affect forward/reverse leakage current and device reliability characteristics during the device operation. Thus, to develop high-efficiency and high-reliability GaN-based LEDs, how to reduce defect-related influence as well as how to improve p-conductivity should be taken into account.
Mg delta (δ)-doping in p-GaN has been considerably attractive because of the fact that the δ-doping is very effective in obtaining high hole concentration (≥1×1018 cm-3) of p-GaN [14-17]. In other words, the δ-doping can be a good candidate for replacing the SLs in p-GaN regions of GaN-based LEDs. The improvement of the hole concentration is due to the reduction in the Mg activation energy and self-compensation effects in δ-doped GaN [16]. Therefore, in this paper, we have investigated the SD and δ-doping influence on the electrical and optical characteristics of GaN-based LEDs. Possible carrier transport mechanism at the interface between transparent conducting electrode (TCE) and p-GaN containing the d-doped layer is also discussed using specific contact resistance-temperature (Rsc-T) and current-voltage-temperature (I-V-T) data. It is found that the integration of the SD and the Mg δ-doping in the p-GaN layer produce the high device performance characteristics including that there is no occurrence in the current crowding even at high current density of 380 mA/cm2.
Operational mechanism for SD-integrated GaN-based LEDs with Mg δ-doped GaN layer
Figure 1 shows schematic cross-sectional diagram, electrically equivalent circuit, and possible current path of the normal LED and the SD-integrated LED having the δ-doped layer. For simplicity, a LED without/with the SD and δ-doped layer is named here as a normal LED and a δ-SD LED, respectively. As for the normal LED (Fig. 1(a)), Ni/Au bonding pad is directly formed on indium tin oxide/nickel oxide (ITO/NiOx) TCE. On the contrary, for the δ-SD LED as shown in Fig. 1(b), the Schottky electrode on p-GaN (like n+-p diode) is electrically reversed to the GaN-based LED. Unlike the normal LED operation (Fig. 1(c)), if a reverse differential resistance (dV/dI) of the SD is high enough to act as a high-resistance resistor, it could be expected that most of the current is injected into the δ-SD LED through the ITO/δ-doped p-GaN TCE window during the forward LED operation (Fig. 1(d)). In addition, the use of the δ-doped layer in the LED could lead to an improvement of conductivity of p-contact layer and p-ohmic contact as described previously [14–17], thus, for this δ-SD LED structure, current injection density becomes larger, and hence current injection efficiency becomes better, as compared to the normal LED. Furthermore, an influence of parasitic defects existing in the p-current injected area of the LED could be effectively reduced by means of a current blocking-area of the SD, resulting in a decrease of leakage current and considerably improving the overall reliability characteristics of the LED.
Fig.1 (a) and (b) schematic cross-sectional diagram and electrically equivalent circuit of normal LED and δ-SD LED; (c) possible current injection path (IT = IA + IB) of normal LED; and (d) that (IT ≈ IB) of δ-SD LED if the resistance of p- SD is assumed to be much higher, during the forward operation |
Experiments
Fabrication of C-TLM patterns and measurements
In order to investigate the carrier transport mechanism at the interface between metals and GaN, circular-transmission line model (C-TLM) patterns were fabricated. Metal-organic chemical vapor deposition system was used to grow a 2 µm thick unintentionally (u)-doped GaN layer on a 30 nm thick GaN nucleation layer/(0001) sapphire substrate. For Mg-doped GaN, the 100 nm thick p-GaN layer was grown on the u-doped GaN. As for Mg δ-doped GaN, the 0.5 µm thick layers consisting of 35 periods was grown on the u-doped GaN layer. All samples were then rapid-thermal-annealed at 650°C for 10 min under nitrogen ambient to activate Mg dopants. Prior to the fabrication of C-TLM patterns, the surfaces of the all samples were ultrasonically degreased with acetone, methanol, and ethanol for 5 min in each step, and then rinsed with deionized water followed by N2 blowing. Buffered oxide etch (BOE) surface treatment was ultrasonically carried out for 5 min at room temperature. C-TLM patterns were defined using a photolithographic technique. The inner radius of the C-TLM pad was 100 µm and the spacing between the inner and outer pads was 10, 15, 20, and 30 µm. Prior to the deposition of metal films, the C-TLM patterned layers were dipped into a BOE solution for 30 s. The Ni/Au (5/5 nm) metals were then deposited by an electron-beam evaporation and rapid-thermally annealed at 550°C to form NiO/p-GaN. After the formation of the NiO, the ITO (220 nm) film was deposited on the NiO by an electron-beam evaporation. I-V and I-V-T characteristics were measured using an on-wafer probing system having a hot chuck and a semiconductor analyzer (HP4155A).
Fabrication of LEDs and Measurements
Metal organic chemical vapour deposition (MOCVD) system was used to grow InGaN-GaN multiple quantum wells (MQWs) LED wafers with a peak wavelength of around 450 nm on c-face sapphire substrates. The LED epitaxial layers consist of a 30 nm thick GaN nucleation layer on sapphire substrate, a 2 µm thick unintentionally doped GaN layer, a 1.5 µm thick Si-doped GaN n-contact layer (Nd about 4×1018 cm-3), an active region with five periods of InGaN-GaN MQWs, a 0.1-µm-thick Mg-doped GaN layer, and the 0.5-µm-thick Mg-δ doped p-contact layer consisting of 35 periods. The activated Mg concentration was determined to be 3.2×1017 and 0.93×1017 cm-3 for p-GaN and δ-doped p-GaN by means of Hall effect measurement. The LEDs (900 µm × 900 µm) were fabricated using photolithography patterning and inductive coupled plasma (ICP) etching to a depth of 1.1 µm. Unlike the normal LED, the second ICP etching (with a depth of 22 nm) was carried out to expose the p-GaN surface in the δ-SD LED. For normal LEDs, sputtered ITO films (220 nm)/electron-beam-evaporated Ni/Au (5/5 nm) layers were deposited on p-GaN as a transparent ohmic electrode and then were annealed at 550°C for 30 s in a N2 ambience. Electron-beam-evaporated Ni/Au (30/300 nm) bonding-pad electrodes were deposited on the ITO/NiO ohmic electrode. However, for the δ-SD LED, the sputtered ITO layer on the δ-doped p-GaN were annealed at 550°C for 30 s in a N2 ambience, and the electron-beam-evaporated Al (300 nm) scheme was formed on p-GaN as a p-SD. Cr/Ni/Au (20/30/300 nm) films as an n-ohmic electrode were deposited on n+-GaN layer for all the LEDs. All the p- and n-electrodes were annealed at 500°C for 30 s in a flowing N2 ambient. I-Vdata were obtained using a semiconductor parameter analyzer (HP4155A). Optical power-current (L-I) data were obtained using an optical spectrometer and a photodiode detector.
Carrier transport mechanism at the interface between TCE and δ-doped p-GaN
Figure 2 shows the I-V characteristics for the p-GaN and the δ-doped GaN contacts. The p-GaN contact exhibits a nonlinear I-V behavior, while the δ-doped p-GaN contact displays a linear I-V characteristic. The specific contact resistance (Rsc) was determined from a plot of the measured resistances vs. the spacing between C-TLM patterns. The C-TLM method was used to fit a log-scaled line to the experimental data [18]. The Rsc was obtained to be 1.5×10-2 and 2.3×10-5 Ω·cm2 for the p-GaN and δ-doped p-GaN contacts, respectively. It should be stressed that the use of the δ-doping yields a dramatic improvement of the Rsc, revealing that it is very promising to achieve high-quality ohmic electrodes. Unlike the p-GaN contact, the improvement in the I-V curves for the δ-doped p-GaN contact can be closely associated with a high-accumulated acceptor concentration in the δ-doped GaN surface [16]. Thus, high hole concentrations (about 1018 cm-3) increase a hole-tunneling probability at the interface between ITO/NiO and δ-doped p-GaN, and hence result in the linear I-V behavior.
To interpret the carrier transport mechanism for the p-GaN and δ-doped GaN contacts in detail, a metal/p+-/p-GaN model was employed [19]. In this model, the high-accumulated hole region in the δ-doped p-GaN is assumed to have p+-GaN. A carrier distribution function of the p+-region is also assumed to be Na_p+-region(x) = N0·exp(-x/d). More detailed model information can be referred in Ref. [19]. In this work, we used two carrier transport models such as thermionic emission (TE) and field emission (FE) [19,20]. Figure 3 exhibits plots of the experimental and theoretical values of the Rsc as a function of 1/T using a metal/p+-/p-GaN model. Comparisons show that the theoretical results are in good agreement with experimental data, indicating that our modeling is valid. The TE conduction is considerably dominant for the p-GaN contact, while the FE conduction is sensitively dominant for the δ-doped GaN contact. The results mean that the δ-doping in the p-GaN is very effective in increasing holes near the GaN surface as compared to the p-GaN contact. Larger hole-accumulation gives rise to the reduced barrier width and energy band-bending (EF-Ev), and consequently results in tunneling of carriers at the interface. More detailed results will be published elsewhere [21].
Device characteristics for p-SD and δ-SD LED and discussion
Figure 4 shows the I-V characteristic of the Al (300 nm) Schottky contact on p-GaN as a p-SD. In this study, Pt/Ni/Au (20/30/100 nm) scheme was employed to form ohmic contact [22]. The rectifying I-V characteristic of the non-alloyed Al SD is indicative of the formation of Schottky contact at the interface between Al and GaN. A dynamic resistance of dV/dI for the forward-biased I-V curve was calculated to be more than 1 MΩ, indicating that the SD is good enough to act as a high-resistance resistor as described from Fig. 1.
Table 1 shows the summary of electrical data for both the normal and the δ-SD LED. It is shown that the mean turn-on voltage (Vth) of the normal LED is 4.0 V at 200 mA, while that of the δ-SD LED is 3.3 V. It is noted that the use of p-SD and δ-doping in the p-GaN produces a considerable reduction of the Vth. Series resistances of the LEDs were calculated using the relation, I(dV/dI) = I·Rs + nkT/q, where Rs is series resistance and n is ideality factor. Calculations show that the mean Rs of the normal LEDs is 7.1 Ω, while that of δ-SD LED is 2.7 Ω. A considerable improvement of the Rs and Vth for the δ-SD LED (as compared to that of the normal LED) could be due to an improvement of p-conduction by an increased acceptor concentration and an increased current injection density (by using the high-resistance SD). As for the ideality factor, the mean n of the normal LED is calculated to be 7.7, whereas that of the δ-SD LED is 2.5. This implies that the SD is very effective to reduce the parasitic defect-related effect during the forward operation as illustrated from Fig. 1. In addition, these behaviours can also give appropriate explanations why the reverse leakage (at - 5 V) and forward leakage (at 2 V) currents of the δ-SD LED are much lower than those of the normal LEDs in Table 1.
Tab.1 Summary of electrical data obtained from the normal and δ-SD LED |
| Vth/V at 200 mA | Rs/Ω | n | IF/nA at 2 V | IR/nA at -5 V | |
|---|---|---|---|---|---|
| normal LED | 4.0±0.06 | 7.1±0.2 | 7.7±0.4 | 32±5 | -42±6 |
| δ-SD LED | 3.3±0.07 | 2.7±0.3 | 2.5±0.5 | 3±2 | -6±4 |
The output L-I characteristics of the LEDs are shown in Fig. 5. A comparison of the L-I data shows that the output power of the δ-SD LED (at 200 mA) is higher than that of both the normal LED by about 20 %. In this work, light-transmittance (at 450 nm) of the ITO/NiO/p-GaN and ITO/δ-doped p-GaN was comparable and the value was around 91%. Based on I-V characteristics, L-I data, and light-transmittance results, it can be concluded that higher output power could be attributed to an increased current injection density as well as high-conduction δ-doped GaN.
Figure 6 shows the optical emission photographs obtained from the normal and the δ-SD LED. Unlike the normal LED suffering from severe current crowding effect, the δ-SD LEDs do not experience any current crowding effect even at high current density of 380 mA/cm2. These behaviours indicate that the SD-integration and the Mg δ-doping in the p-GaN lead to very stable device-operation characteristics as compared to the normal LEDs without them. In this work, we also investigated the SD area dependence on the current crowding effect of the SD-integrated LEDs with/without δ-doped GaN layer. The results (not shown here) showed that the current crowding phenomena of the SD-integrated LEDs without a δ-doped GaN layer are sensitively affected by the ratio of the SD and TCE window area. On the other hand, there is no occurrence for current crowding effect in the SD-integrated LEDs having the d-doped GaN layer irrespective of current density range up to 400 mA/cm2. Thus, it can be concluded that the current crowding phenomena is absolutely affected by an electrically p-conduction in LEDs rather than others including the SD integration.
Conclusions
In summary, this paper successfully demonstrated the high-performance and current crowding-free InGaN/GaN LEDs using an electrically-reverse-connected SD and an Mg δ-doped GaN layer. Possible carrier transport mechanism at the interface between TCE and p-GaN with/without the δ-doped layer was also described. From the electrical and optical investigation results, the use of the SD and δ-doping in the p-GaN region is very practical to realize current crowding-free and high-efficiency GaN-based LEDs.