Introduction
The III-V nitride semiconductor materials system has become increasingly popular among the applications of UV/blue light emitting diodes (LEDs), laser diodes (LDs) and high-temperature electronic devices due to its large direct bandgap energy (3.4 eV for GaN at room temperature). Even though the potentials of III-V nitrides were well known for more than three decade, it was only until 1992 with the realization of Mg doped p-GaN film after the thermal annealing [Reference Nakamura, Iwasa, Senoh and Mukai1] could the realization of high brightness blue/green InGaN/GaN quantum well(s) LED [Reference Nakamura, Mukai and Senoh2] and the recent demonstrations of long-lifetime pulsed and CW blue LDs [Reference Nakamura, Senoh, Nagahama, Naruhito, Yamada, Matsuchita, Sugimoto and Kiyoku3] be achieved. Despite the commercialization of LEDs, the low p-type doping concentration (and hence high p-type resistivity) in GaN still limits the device performance in two ways: first it reduces the current spreading in the LED. This causes most of the light to emit in the active region directly underneath the p-type contact, and is therefore blocked by the p contact. Secondly, low p-type doping results in a rather high and non-linear contact resistance. While the current spreading problem has been effectively resolved by depositing a thin semi-transparent metal layer on top of the p-GaN cap layer, high contact resistance is still a problem to be tackled. In the present work, we report the successful growth of single quantum well InGaN/GaN LEDs on multi-wafer MOVPE reactor with high thickness uniformity and the development of a self-aligned LED fabrication process that incorporates a thin current spreading layer. We also present detailed studies on the variation of contact resistance rc of Ni/Cr/Au on p-doped GaN layer versus annealing temperature.
Experiment
The InGaN/GaN LED samples were first grown on the Aixtron production model multi-wafer MOVPE reactor. The structures consist of 1.5 μm n-GaN layer, a thin InGaN quantum well (4-8 nm), and a 0.5 μm p-GaN cap layer, all grown on 2-inch (0001) sapphire substrates. The growth uses standard precursors including TMGa and TMIn for Group III and NH3 for Group V elements, and silane and MgCp2 are used for n-type and p-type dopant sources, respectively. The un-intentionally doped GaN is n-type with carrier concentration ∼ 1-3×1017/cm3. The p-type GaN with hole concentration of ∼5×1017/cm3 are routinely achieved after post-growth annealing. The thickness uniformity across the diameter of a 2-inch wafer was evaluated using spectral reflectometry measurement in the range of 670 to 1100 nm.
The process flow for fabricating a LED with a structure shown in Fig. 1(a) , is described as follows. First, a thin current spreading layer consists of Ni/Au (40 Å/40 Å) was deposited on the entire wafer using electron beam evaporation. The estimated optical absorption due to the current spreading layer is ∼ 38% [Reference Palik4]. The sample is thermally annealed at 500°C under N2 ambient for 100 seconds. The LED mesa is then defined by a photoresist pattern. This is followed by ion milling to remove all except the current spreading layer, and Reactive Ion Etching (RIE) in a mixture of Cl2 and BCl3 to remove the GaN all the way to the n-layer. In this way, the current spreading layer and the mesa are formed in a self-aligned way. Figure 1(b) shows the SEM micrographs of the RIE etched surface around the LED mesa. Both the bottom and lateral surfaces of the etched mesa appear to be very smooth. Finally, Ni/Cr/Au (150 Å/150 Å/1200 Å) were e-beam evaporated to form both n and p-type contact pads. LEDs without current spreading layer are also fabricated for comparison. By varying the In composition and the thickness of the InGaN quantum well, LED emission from 420 to 490 nm were obtained.
Fig. 1(a) The schematics of self-aligned InGaN/GaN LED structure. Note that a thin Ni/Au layer is deposited on top of the LED mesa to improve the current spreading.
Fig. 1(b) SEM of a typical LED structure with a magnified view of the edges of mesa after RIE etching.
To investigate the contact resistance to p-type GaN, a rectangular shaped, mesa isolated transmission line (TL) structure was fabricated. The width of the TL pattern is 300 μm while the spacing between adjacent Ni/Cr/Au contacts are varied from 5 to 50 μm at increments of 5 μm. The TL pattern is defined using RIE etching similar to the LED mesa etching. The wafer was then diced into several pieces each containing 4 TL patterns and was subsequently annealed at temperatures from 300°C to 600°C for 100 seconds. The TL structures are measured at constant current and the specific contact resistance rc ’s were extracted using standard technique [Reference Murrmann and Widmann5]. These experimental results will be discussed in the next section.
Results and Discussions
Figure 2 shows a typical reflectance spectrum and the thickness nonuniformity across the diameter of a 2-inch LED sample. The thickness of n-GaN layer for this particular LED sample is ∼ 3 μm. The epilayer thickness, d is calculated from d = [λ2/(2nΔλ)], where Δλ is the wavelength separation between successive reflectance peak, and n is the refractive index of the GaN epilayer. The thickness shown in Fig. 2 are averaged using Δλ from 5 successive reflectance peaks. Thickness nonuniformity of less than < 1% is observed across the entire 2-inch wafer. The uniformity result we obtained on a typical LED device layer agrees reasonably well with Aixtron’s own evaluation on 2-inch single-layer GaN wafer [Reference Beccard, Schoen, Schineller, Schmitz, Heuken and Juergensen6]. Figure 3(a) shows the current vs. voltage of the same LED wafer with and without the addition of the current spreading layer. It can be seen that the incorporation of current spreading layer has drastically reduced the spreading series resistance of the LED, which is caused by the high resistive p-GaN cap layer. With the current spreading layer, the operating voltage of the LED at 20 mA is reduced to 5 V or below. The use of self-alignment process for forming both the mesa and current spreading layer ensures maximum current spreading of a LED. Microscope observation shows that uniform emission from the entire LED mesa except the p-contact pad is achieved. The measured series resistance of 9.19×104 μm2 LED is ∼26Ω. Figure 3(b) shows the emission spectrum of a typical LED we fabricated. For this particular device, the emission peak is ∼ 490 nm and the FWHM is ∼ 50 nm.
Fig. 2 Thickness uniformity across the diameter of a 2-inch LED wafer. A typical reflectance spectrum is also shown here.
Fig 3 (a) The I-V of the LED with and without the current spreading layer; and (b) the emission spectrum of the LED.
Owing to the low carrier concentration of the p-type GaN and the relative large barrier height between the metal and p-GaN, the I-V characteristics of the TL are non-linear. That is, the contact resistance depends on the measurement current. Figure 4 shows a typical resistance vs. contact separation at two current levels. Figure 5 shows the extracted rc for Ni/Cr/Au system at different measurement currents for samples annealed at different temperatures. In all cases, the value of rc decreases monotonically with increasing current density (current/unit length) in the TL pattern and then approaches a saturation value. The specific contact resistance, rc is seen to decrease monotonically with increasing current. The optimal annealing temperature is observed to be 300-500°C. Further increase in the annealing temperature at 600°C and above results in an increase of rc . This phenomenon is likely caused by the formation of nitrogen vacancy that occurs when samples are subjected to high temperature treatments. The nitrogen vacancy in turn compensates the acceptor dopants at the metal/p-GaN interface. Several groups have already reported similar findings on these issues [Reference Lester, Brown, Ramer, Zhang, Hersee and Zolper7,Reference Zolper, Rieger, Baca, Pearton, Lee and Stall8,Reference Venugopalan, Mohney, Luther, Wolter and Redwing9]. In addition, the presence of the high resistive Ni3N and Ni4N phases at the metal-semiconductor interface at the annealing temperature greater than 500 °C can also result in higher contact resistance [Reference Sheu, Su, Chi, Chen, Chen, Huang, Hong, Yu, Wang and Lin10]. The rc measurement results are summarized in Table I . The lowest rc we obtained is about 0.016 Ω−cm2 (the resistivity of the TL sample extracted is on the average of 4.05 ± 0.15 Ω-cm). The values rc we obtained are similar to previously reported data on Ni/p-GaN system [Reference Mori, Kozawa, Ohwaki, Taga, Nagai, Yamasaki, Asami, Shibata and Koike11,Reference Ishikawa, Kobayaxhi, Kiode, Yamasaki, Nagai, Umesaki, Koike and Murakami12]. The lowering of rc due to thermal annealing below 600 °C is not clear at this stage. Earlier studies had suggested that thermal annealing of Ni/p-GaN helps to removes the contaminants at the interface resulting in lowering of rc [Reference Ishikawa, Kobayaxhi, Kiode, Yamasaki, Nagai, Umesaki, Koike and Murakami12]. Recent results on the lowering of rc by aqua regia etching of p-GaN has strongly suggested that surface oxide plays a paramount role on the specific contact resistance to p-type GaN [Reference Kim, Lee, Lee, Shin, Park and Kim13].
Fig. 4 The measured resistance versus spacing of the TL test pattern at two different measurement currents.
Fig. 5 The extracted rc measured at different current levels and annealing temperatures. The width of the rectangular TL pattern is 300 μm.
Table I The extracted average rc and the corresponding standard deviations from the TL test pattern at different current levels for samples under different annealing temperatures.
| Specific Contact Resistance (Ω- Sqr. cm) | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Current (μA) | No Annealing | Std. Dev. | 300 Deg. C | Std. Dev. | 400 Deg. C | Std. Dev. | 500 Deg. C | Std. Dev | 600 Deg. C | Std. Dev |
| 50 | 2.13 | 0.38 | 0.18 | 0.06 | 0.17 | 0.05 | 0.18 | 0.012 | 1.98 | 0.47 |
| 250 | 0.28 | 0.01 | 0.051 | 0.01 | 0.049 | 0.007 | 0.058 | 0.008 | 0.34 | 0.024 |
| 1000 | 0.06 | 0.002 | 0.019 | 0.004 | 0.017 | 0.004 | 0.021 | 0.002 | 0.078 | 0.023 |
| 2000 | 0.016 | 0.006 | 0.016 | 0.004 | 0.018 | 0.004 | 0.043 | 0.017 | ||
From the LED I-V, the extracted series resistance is 26 Ω measured between 40 to 80 mA. The current density of the LED operating at 60 mA corresponds to an average current density of ∼ 31.3 mA/cm for the TL measurement. From Fig. 5 , the corresponding rc extracted from the TL measurement is 0.021 Ω−cm2. The estimated contact resistance of the LED is therefore ∼ 22.9Ω. Based on this estimate, over 88 % of the series resistance of the LED originates from the p-GaN contact resistance.
Conclusion
We report highly uniform growth (thickness nonuniformity < 1%) of InGaN/GaN LED on 2-inch sapphire substrate using a multi-wafer MOVPE reactor. A self-aligned LED fabrication process is developed to incorporate a thin Ni/Au metal for improved current spreading of the LED device. It is also found that the annealing temperatures between 300-500 °C significantly reduce the contact resistance of as-deposited Ni/Cr/Au metals on p-type GaN. Specific contact resistance in the range of 0.016 to 0.021 Ω−cm2 is achieved at annealing temperatures between 300°C to 500°C.
