Current injection efficiency
One of the most important and serious issues in the realization of high-efficiency UV and DUV light emitters is conductivity control—particularly the control of p-type conductivity. Highly conductive p-type layers are necessary to improve the performance of UV/DUV emitters.
However, the control of p-type conductivity in AlxGa1_xN has proven to be extremely difficult. With increasing Al content in AlxGa1-xN, the activation energy of Mg dopants increases as well as the resistivity of the alloys. In addition, although high-crystalline-quality AlxGa1-xN films are necessary, it is difficult to grow them owing to the high concentration of dislocations in AlxGa1-xN alloys. For this reason, it is very important to grow high-crystalline-quality AlxGa1-xN and control the p-type conductivity of AlxGa1-xN. There have been several reports on the analysis of Mg-doped p-type AlxGa1-xN by variable-temperature Hall measurements, PL and CL [38-41]. The analysis of Mg-doped p-type AlN is also one of the main challenges for the realization of high-efficiency UV and DUV devices.
To determine the activation energy of Mg in AlGaN, it is necessary to avoid other unwanted effects such as residual impurities and extended defects. High- temperature growth has been found to be effective for reducing the concentration of residual impurities. Figure 1.11 shows the concentrations of hydrogen, oxygen, silicon, and carbon in nominally undoped AlN grown at 1,100°C and 1,600°C . As shown, the concentration of residual impurities is less than the detection limit of SIMS in HT-grown AlN.
Fig. 1.11. SIMS depth profiles of AlN grown at 1,100°C (left) and 1,600°C (right) on 6H SiC.
Fig. 1.12. Effective activation energy of Mg acceptor in Al0.5Ga0.5N measured by temperature-dependent Hall effect measurement as a function of Mg concentration to the power of —1/3.
For Mg doping, Et-Cp2Mg or Cp2Mg is transported into the growth chamber during growth. The as-grown layer becomes highly resistive upon hydrogen passivation . Post-growth annealing is usually conducted to activate the hydrogen-passivated Mg .
Figure 1.12 shows the effective acceptor activation energy of Mg in Alo.5Gao.sN as a function of the Mg concentration to the power of —1/3. The effective activation energy decreases with increasing Mg concentration up to 4 x 1019 cm~3. Above this concentration, the activation energy increases again. The reason for the decrease in activation energy with increasing Mg concentration is the Coulomb interaction between negatively charged Mg ions and positively charged valence band holes [44, 45].
In contrast, the reason for the increase in activation energy above a Mg concentration of 4 x 1019 cm~3 has not yet been clarified. Microscopic observation of highly Mg-doped GaN and AlN shows the existence of inversion domain boundaries. Figure 1.13 shows a cross-sectional TEM image of Mg-doped/undoped AlN on a sapphire (0001) substrate. Black dots can only be observed in the Mg-doped layer, which are thought to be inversion domain boundaries caused by Mg segregation at the surface. Similar dots have been reported to exist in heavily Mg-doped GaN [46, 47].
Therefore, the optimum Mg concentration can be deduced and the compositional dependence of the maximum hole concentration at room temperature can be estimated. As shown in Fig. 1.14, the expected hole concentration in AlN is only 3 x 1012 cm~3 at room temperature.
In a p-type AlGaN with such a low hole concentration, current injection efficiency (pie) should be low. At present, there is no method of directly measuring Pie. In this chapter, our attempts to estimate pIE in a UV LD is explained. The sample discussed was reported in ref.  and its structure is depicted in
Fig. 1.13. Cross sectional TEM image of AlN:Mg/AlN. Black dots in the AlN:Mg show the inversion domain boundaries caused by Mg precipitation.
Fig. 1.15. LD characteristics are as follows: Jth: 8.0 kA/cm2, nth: 2.6 x 1019 cm-3, A = 1.2 x 109 s-1, B = 2.4 x 10-11 cm3 s-1, niQE slightly below Jth: 34%, Ald: 354.2 nm. The Auger coefficient is thought to be negligible.
In this LD, a 20 nm p-type Alo.5Gao.5N electron blocking layer was grown on a p-type Al0.1Ga0.9N waveguide layer to avoid waveguide loss. SRH measurement by current injection was conducted. In the case of PL measurement, the same MQW structure without the p-type layers was measured. Figure 1.16 shows a comparison of the efficiency measured by excitation-density-dependent PL and electroluminescence (EL) current injection. If we assume that niQE measured by PL gives the true niQE and that measured by EL gives the product of niQE and niE, we can estimate niE, which is about 30-35%.
Fig. 1.14. Estimated maximum hole concentration at room temperature as a function of Al content. In this calculation, ND is assumed to be one tenth of Na.
Fig. 1.15. UV LD structure used in this study.
In the case of DUV LEDs, the situation is more serious because the hole concentration in the p-type cladding decreases with increasing Al content. To overcome this problem, several approaches have been considered. One is the use of a multiple quantum barrier structure . This idea was first investigated by Iga et al.  in 1.5-pm-range InGaAsP/InP and visible AlGaInP lasers. The maximum EQE achieved was 1.2% at an emission wavelength of 250 nm.
Another approach is the use of a polarization field. By using a compositionally graded p-type AlGaN layer on AlN with an Al face, polarization-induced hole doping is possible [51, 52].
Current state-of-the-art DUV LEDs have an EQE of 14.3% at 2 mA with an emission wavelength of 285 nm. In these DUV LEDs, niE has been estimated to be higher than 70%.
Fig. 1.16. SRH measurement of UV LD characterized by PL and EL.
Fig. 1.17. Schematic structure of a DUV LED having a partial reflective electrode.
Light extraction efficiency
Compared with aiqe and aie, Alee is still limited to a low value, as explained in the introduction. One solution is to use a partially reflective electrode in the face-down configuration . Figure 1.17 shows schematically the structure. The output power was increased by 55% without a serious increase in the operating voltage.