Growth of InGaN layers and quantum wells on m-plane and different semipolar surfaces, i.e. (1012), (1011), (2021), (1122)
InGaN quantum wells are the most important part of visible emitting LEDs and lasers. Recombination dynamics in non- and semipolar InGaN are significantly different, not only due to the reduced polarization fields (see also Section 8.1) but also due to the reduced symmetry of the surfaces leading to anisotropic and shear strain. Also, there are major differences in the growth of InGaN due to different surfaces and reconstructions leading to different adatom bonding strengths, adatom incorporation, and adatom diffusion barriers (see Section 8.3.1). This leads to a different indium incorporation for the different non- and semipolar surfaces.
Indium incorporation efficiency for different surface orientations
The indium incorporation depends strongly on the surface orientation (Wernicke et al., 2012). However, a conclusive picture cannot yet be drawn, since growth parameters have an impact on the relative indium incorporation on different planes (Browne et al., 2012). Thus, the reported relative indium incorporation of the (1122) varies from higher (Tanikawa et al., 2011; Wernicke et al., 2012) to similar (Jonen et al., 2012) to lower incorporation (Browne et al., 2012) in comparison to the (0001) c-plane. Because the presence of morphological defects, e.g. caused by dislocations or BSF, typically increases the indium incorporation (Chakraborty et al., 2005b), we will focus only on QW grown on high-quality bulk substrates. Firstly, the indium content of a semipolar InGaN layer is not easily accessible by X-ray diffraction (XRD) due to a triclinic distortion of the unit cell caused by anisotropic strain and shear strain (Romanov et al., 2006). Different methods of analyzing the XRD data were proposed for pseudomorph- ically strained InGaN layers on GaN (Young et al., 2011) as well as for layers with an arbitrary strain state (Jonen et al., 2012). Additionally, measurements on quantum wells are more difficult, since the material volume is usually small and evaluation requires the knowledge of the exact barrier and quantum-well thickness. The comparison of the emission wavelength, e.g. from PL or EL measurements, is much easier and more reliable than the comparison of the indium content. However, extracting the relative indium incorporation requires complex calculations that imposes rather large uncertainties. Finally, it must be kept in mind that indium content from luminescence experiments relate to regions with the highest indium content (Kuokstis et al., 2002), whereas XRD provides the average indium content. Larger localization energies therefore cause a deviation of local PL indium content and average XRD indium contents. The emission energy for a set of QWs grown on polar, semipolar, and non-polar GaN with different growth temperatures at a gas phase composition xgas of 25% is shown in Fig. 8.11 (Wernicke et al., 2012). The following order-of-emission energies can
Fig. 8.11. Emission wavelength of polar, semipolar and non-polar QWs. Reprinted with permission from Wernicke et al. (2012), © 2012 IOP Publishing Ltd.
be extracted: (1011) < (1122) = (0001) < (2021) < (1010) = (1012). This ratio can be found for the whole temperature range. This means that long-wavelength emitters on (1011) can be grown at substantially higher temperatures than on the c-plane, and even more so on (10l0) GaN. From this set of data the corresponding indium content was calculated using k.p theory and Schrdinger-Poisson calculations (Schade et al., 20116). Again, a hierarchy for the indium incorporation efficency can be found throughout the entire temperature range: (1011) > (1122) > (0001) = (2021) = (1020) = (1012) (Wernicke et al., 2012). Thus the (10П) surface shows the highest indium incorporation. Also, the (1122) exhibits increased indium incorporation, whereas incorporation of indium on the other planes is relatively similar. We also compared the indium incorporation into thick (>30nm) InGaN layers on (0001) and (2021) GaN, which can be accurately measured by XRD. The indium incorporation for (2021) InGaN was lower by a factor of ~ 2 at 725°C, and xgas up to 41% was measured (Wernicke et al., 2012). This is in contradiction to the results of the QWs, and indicates a much higher localization energy in (2021) InGaN, a strong (vertical) segregation a c-plane or a higher polarization field in (2021). Overall, final conclusions cannot be drawn yet, and additional optical and structural investigations are necessary. Although the indium incorporation is highest on (1011), the best device results are obtained on other planes (see Sections 8.4 and 8.5). Other factors such as point-defect incorporation and critical layer thickness must be considered as well.