Growth of GaN on planar heteroepitaxial substrates
In order to reduce the LED device fabrication costs, large-area low-cost substrates are in demand. Polar c-plane LEDs have been demonstrated on (0001) sapphire and (111) silicon with up to a 6-inch diameter (Dadgar et al., 2006; Lu et al., 2011). Therefore the easiest approach is to grow heteroepitaxial GaN on foreign substrates, e.g. sapphire and silicon. The resultant growth plane depends on the local atomic arrangement and lattice mismatch, i.e., for every substrate material and orientation an epitaxial relationship is defined. An overview of different orientations that have been realized is given in Fig. 8.8. Non-polar (1120) a-plane GaN was first grown on (1012) r-plane sapphire (Sano and Aoki, 1976), but can also be grown on (1120) a-plane SiC (Craven et al., 2004b). (1010) m-plane GaN was realized on (100) y-LiAlO2 (Hellman et al., 1997), (1010) m-plane SiC (Gardner et al., 2005), and (1010) m-plane sapphire. Also, semipolar
Fig. 8.8. Overview of realized surface orientations for planar heteroepitaxial growth.
GaN can be grown heteroepitaxially. (10П) GaN was realized on (100) spinel (Baker et al., 2005), (1013) GaN on (110) spinel (Baker et al., 2005) and (10l0) m-plane sapphire (Matsuoka and Hagiwara, 2001; Baker et al., 2006), and (1122) GaN on (10l0) m-plane sapphire as well. Semipolar (1126) GaN was grown on (10l2) r-plane sapphire by using an AlInN nucleation layer (Bastek et al., 2010). Semipolar GaN of (101l) type with relatively small inclination angles of 18°, 26°, 29°, 31°, 34°, and 47° can be grown on high-indexed Si substrates by nucleation of (0001) GaN on Si (111) microfacets (Ravash et al., 2010). Obviously, the epitaxial relationship for all these substrates is not unique. For example, on m-plane sapphire, (10l0), (1013), and (1122), GaN was grown. Actually, many more orientations can occur (Ploch et al., 2010), but none of the others were ever observed to be dominant. A selection of the orientation can be reached by ex situ pretreatment to achieve (10l0) orientation (Armitage and Hirayama, 2008), nitridation of the surface to achieve (1122) orientation (Baker et al., 2006; Ploch et al., 2010), and vice versa, to achieve (10l3) orientation without nitridation. Also, composition and growth conditions of the nucleation layer, e.g. AlN (Bougrioua et al., 2007) or high-temperature GaN nucleation layers to achieve (1122) GaN on m-plane sapphire (Lee et al., 2010; Zhu et al., 2010), as well as the subsequent growth process to select a single domain (Ploch et al., 2011) can lead to a layer with a single surface orientation. Another issue with heteroepitaxial growth is twinning (formation of domains with different in-plane relationship) due to an ambiguous epitaxial relationship, as observed for (1011) GaN on spinel (Kaeding et al., 2006) or (1013) GaN on sapphire (Frentrup et al., 2011). This can be solved by using the proper substrate miscut (Kaeding et al., 2006).
But even for phase-pure semipolar or non-polar GaN, basal-plane stacking faults (BSF) and associated defects are typically observed for all heteroepitaxial layers grown on planar substrates with a very high density (BSF 105-106 cm-1, threading dislocation density ~ 1010 cm-2 (Craven et al., 2002a; Baker et al., 2005; Venngus et al., 2007)). BSF are faults in the wurtzite ABAB stacking order, e.g. ABABCBCB, and represent a two-dimensional cubic inclusion within a hexagonal matrix. Different types of stacking faults are possible (Stampfl and Van de Walle, 1998), classified by the number of cubically coordinated layers and the displacement vector. Associated defects are partial dislocations as well as prismatic stacking faults (Li et al., 2005). BSF can be identified by transmission electron microscopy as well as by their luminescence (Li et al., 2005), especially the I1 BSF at 3.42eV. In terms of defect density, m-plane GaN is exceptional. Here the BSF density and threading dislocation density are surprisingly low (104 cm-1 and 109 cm-2), possibly due to a very small lattice mismatch (Neumann et al., 2009). However, due to an ambiguity in the nucleation site, stacking mismatch domains occur (Waltereit et al., 2000; Wernicke, 2010) which might cause the formation of BSF.
Stacking faults do not contribute to non-radiative recombination, but they deteriorate the surface morphology and thereby heterointerfaces. Also, the associated defects contribute to non-radiative recombination.
Several reasons for stacking-fault formation were observed: (1) growth of (000!) at the early coalescence (Wu et al., 2003), (2) ambiguous epitaxial relationship relating to the stacking sequence (Vennegues et al., 2008), (3) formation at monoatomic steps in the substrate (Vanfleet et al., 2003), and (4) relaxation of strain (Cho et al., 2008; Wu et al., 2010; Wu et al., 2011).