Comparison of MOVPE grown layers to HVPE grown layers

range suggesting quite pure material dopable down to 10¹⁴cm⁻³.

6.6 Comparison of MOVPE grown layers to HVPE grown layers

In this section, I am going to list the advantages and disadvantages of layers grown either by MOVPE or HVPE. For this purpose, I will focus on the comparison of the electrical properties since these are crucial for the desired application in high power devices.

Figure 6.32: Comparison of the electron Hall mobility as a function of the electron Hall con-centration at300 Kforβ-Ga2O3 homoepitaxially grown by MOVPE and HVPE on various substrates. The literature values for the HVPE layers are from K. Gotoet al. ([150]: orange octagons).

First, I compare the electrical transport within these layers. Fig. 6.32 shows the electron Hall mobility as a function of the charge carrier density of MOVPE and HVPE layers. It is visible, that HVPE layers show lower doping densities down to mid 10¹⁵cm⁻³ leading to higher overall mobilities due to reduced scattering of electrons on ionized impurities. However, these values scatter a lot due to inho-mogeneity in the layers. The MOVPE grown layers will not scatter so much if incoherent twin boundaries are absent like for the layers grown on (010)-oriented substrates. Comparing the mobilities in the same doping range (5×10¹⁶cm⁻³ to 2×10¹⁸cm⁻³), the MOVPE grown layers show higher mobilities than the HVPE grown layers, probably, due to less structural defects. To check the homogeneity of

6.6 Comparison of MOVPE grown layers to HVPE grown layers

the layers, net doping mappings of the layers grown by MOVPE or HVPE are carried out using C-V measurements. Such mappings are shown in Fig. 6.33(a)&(b). For MOVPE grown layers on (100)4^◦ miscut substrates a mean deviation of 4 % from the mean value ND−NA = 2.3×10¹⁸cm⁻³ is visible for a representative excerpt of 3 mm×3 mm. This is higher than the deviations desired for optimal device reliabil-ity, which is lower than 1 %. However, the HVPE grown layers show an even higher mean deviation of 15 % from the mean value ND−NA= 1.2×10¹⁶cm⁻³.

Figure 6.33: Net doping mapping of homoepitaxially grown layers on conductive substrates using C-V measurements.

(a)ND−NA mapping over 3 mm×3 mm of a MOVPE grown layer on a (100)4^◦ miscut substrate is shown. A mean deviation of4 %from the mean value ND−NA = 2.3×10¹⁸cm⁻³ is visible.

(b)ND−NA mapping over3 mm×3 mmof a HVPE grown layer on a (001)-oriented substrate is shown. A mean deviation of 15 % from the mean value ND−NA = 1.2×10¹⁶cm⁻³ is visible.

Fig. 6.34 shows a comparison of the DLTS spectra of MOVPE and HVPE grown layers. The HVPE layers show overall less density of deep electron trap states in the measurable region of the bandgap from 0.1 eV to 1.4 eV below the conduction band.

However, the MOVPE layers show also a low enough acceptor density of about 10¹⁵cm⁻³ to reach the shallow donor region of HVPE grown layers. The occurrence in the respective layers of the electron traps and their parameters is summarized in Tab. 6.1. The E2 electron trap related to the iron acceptor level is present in all layers independent on the growth method. E1 and E4 is only present in the tin doped layer. E3 is present in the MOVPE layer grown on (010)-oriented substrate and in the HVPE grown layer. The possible origin of this trap is speculated to be the gallium vacancy. If this hypothesis is correct, the MOVPE growth on (100)-oriented will be gallium rich while the MOVPE growth on (010)-oriented substrates will be oxygen rich under the same growth conditions. This would show, that the chemical reactions with the used precursors (TEGa and O2) on the respective surfaces is much more complex than expected and that the formation of point defects in MOVPE growth can not simply be explained by chemical potentials.

6.6 Comparison of MOVPE grown layers to HVPE grown layers

Figure 6.34: Comparison of the DLTS spectra of an HVPE layer on an (001)-oriented, and MOVPE layers on either (010)-oriented or (100)6^◦ miscut, conductive β-Ga₂O₃ substrates measured at the emission rate in the peak maximum of en,max= 114 s⁻¹ The electron traps E2and E3have all samples in common.

NT (cm⁻³)

Ec−ET (eV) σ (cm²) MOVPE(100)6^◦ MOVPE(010) HVPE(001)

E1 0.3 2×10⁻¹⁷ - 1.3×10¹⁵

-E2 0.7 - 0.8 1 - 4×10⁻¹⁵ 2.3×10¹⁴ 4.3×10¹⁴ 3.5×10¹³

E^∗∗2 0.8 7×10⁻¹⁷ 2.9×10¹⁴ -

-E3 1.1 - 1.2 1 - 6×10⁻¹⁴ - 1.8×10¹⁵ 7.3×10¹²

E4 1.4 2×10⁻¹³ - 1.1×10¹⁵

-Table 6.1: Parameters of the deep electron traps of MOVPE and HVPE grown layers compared to Cz grown bulk crystals. The given ranges of variation in trap energy and extrapolated capture cross section cover the spread between different samples. The electron traps E₂ and E₃ have all samples in common. But, it was not possible to determine the parameters of E₃ in the MOVPE layer grown on (100)6^◦ miscut substrate due to too less data points. While the MOVPE grown layer on (010)-oriented substrate is tin doped, the HVPE grown layer and the MOVPE grown layer on (100)6^◦ miscut substrate are silicon doped.

All in all, the HVPE grown layers show better overall electrical properties than the MOVPE grown layers. However, these values scatter a lot due to issues with the homogeneity of the layers. The MOVPE grown layers show a better homogeneity than the HVPE grown layers and higher mobilities comparing the same doping region. However, it is not possible to dope the MOVPE grown layers down to the mid 10¹⁵cm⁻³ due to unintentional doping issues. Also the unintentional concentration of compensating acceptors has to be decreased further by one order of magnitude from 10¹⁵cm⁻³ to 10¹⁴cm⁻³. Nevertheless, if the MOVPE grown layers prevent the unintentional doping and reach lower charge carrier densities, the mobility will

6.7 Comparison of measured mobilities to literature and adjustment of the