Deep-UV LEDs were grown using a metalorganic vapor phase epitaxy on c-plane sapphire substrates with a miscut angle of 0.35° toward the sapphire [11
0] direction. Trimethylaluminium, trimethylgallium, triethylgallium, Bis(cyclopentadienyl)magnesium, monosilane gas, and ammonia gases were used as Al, Ga, Mg, Si, and N sources under hydrogen gas, respectively. The sapphire substrates were thermally cleaned in the H
2 ambient, and then, a 3-μm-thick AlN was grown using a two-step growth technique [
27]. Threading dislocation densities of screw and edge dislocations including mixed components in the AlN underlayer were estimated using an X-ray rocking curve at 9 × 10
7 cm
−2 and 1 × 10
9 cm
−2, respectively [
28]. The 1.3-μm-thick n-type Al
0.62Ga
0.38N doped with a Si concentration of 3 × 10
19 cm
−3 was grown on an AlN template [
21,
22]. Multiple-quantum wells, an Al
0.85Ga
0.15N electron blocking layer (EBL), a p-type AlGaN, and a p
+-type AlGaN were grown on the n-type AlGaN underlayer. The p
+-type AlGaN was doped with Mg at a concentration of 1.7 × 10
20 cm
−3. Subsequently, n
+-type and n-type AlGaN were grown under the same conditions as those of the n-type AlGaN underlayer, as indicated in
Figure 1b. The mesa was formed by dry etching using HCl gas. Thereafter, we formed 20/150/50/100/240-nm-thick V/Al/Ti/Pt/Au electrodes as both n-type AlGaN electrodes, and they were simultaneously annealed under a nitrogen (N
2) ambient at 720 °C for 30 s. Further, the annealing process contributes Mg activation under lateral hydrogen diffusion from the exposed mesa-parts of the p-type layers [
29,
30,
31,
32,
33]. For comparison, we prepared a conventional pn-diode-based LED with a thin p-type GaN contact layer grown on a p-type AlGaN shown in
Figure 1a. We adopted indium zinc oxide (IZO) for the anode. The emitted UV light was fully absorbed at the IZO electrode; the sizes of the LED and the anode, and the thickness of the sapphire substrate were 1 mm
2, 0.56 mm
2, and 200 μm, respectively. The light output power was measured using an integrating sphere. For the former, we prepared an AlGaN homoepitaxial TJ LED (TJ#1 to TJ#5) with various Si concentrations and C incorporations in the n
+-type AlGaN layer, as summarized in
Table 1. The carbon concentration was approximately 3.0 × 10
18 cm
−3 (TJ#1 and TJ#2), and it was reduced to 6.5 × 10
17 cm
−3 (TJ#3 to TJ#5) by changing the growth pressure from 50 mbar to 100 mbar. In case of the latter, we prepared MgZnO/Al electrodes for the TJ LED with TJ#5. We deposited a 50-nm-thick MgZnO electrode by RF magnetron sputtering at a substrate temperature of 200 °C, and a typical lift-off process was employed. The sputtering target for MgZnO was prepared as a 2-inch MgZnO sintered material of purity 4N, which is the MgO:ZnO mixing atomic ratio of 1:2. The RF power, sputtering gas, and gas pressure were 100 W, Ar, and approximately 3.4–3.5 × 10
−1 Pa, respectively. After forming the MgZnO electrode, the conductivity was improved by annealing at 850 °C for 5 min under N
2 ambient. In the cathode, Ti/Al electrodes were deposited by the electron beam (EB) method and alloyed at 450 °C under N
2 ambient. Al/Ti/Pt/Au electrodes with 300/50/100/240 nm were formed on the MgZnO electrode via the EB method to obtain a high-reflective electrode. The reflectance of the electrodes for TJ LEDs was measured using a UV-visible spectrophotometer (UV-VIS). For comparison, Ti/Al electrodes for the TJ LED anode were prepared via the same process as the cathode.