GaAs, InP, GaSb, and InAs were investigated for layer transfer to other III-V substrates using hydrogen ion exfoliation and wafer bonding to develop III-V based wafer bonded templates for subsequent epitaxial growth of device structures. High-resolution X-ray diffraction was proven to be particularly helpful in this investigation enabling nondestructive testing of the initial implantation profile as well as the strain relief ͑associated with the diffusion of hydrogen and other point defects͒ after various annealing sequences. The kinetics of exfoliation for many different III-V materials ͑including GaAs, InP, InAs, and GaSb͒ showed similar dependence on the processing temperature relative to the material melting temperature ͑and other materials parameters͒, i.e., lower melting temperature materials required lower temperature processing to retain the implanted hydrogen for exfoliation. In all of these cases, a multiple annealing sequence was shown to produce the most efficient exfoliation.

No abstract

No abstract

We investigate the surface roughness and dislocation distribution of compositionally graded relaxed SiGe buffer layers by inserting two tensile-strained Si layers. The 20nm thick strained Si layers, less than the critical thickness for dislocation formation, are inserted at 10 and 20% Ge content regions of the 1μm thick graded SiGe layer with a final Ge content of 30%. The surface immediately after growing the second strained Si layer on SiGe with 20% Ge content is found to be flat with about 1.1nm root-mean-square roughness. However, the crosshatched surface with the 7.8nm roughness develops during subsequent SiGe growths, which is slightly less than the 10.3nm value for SiGe without inserted Si layers. Another important issue of consideration is that inserting the strained Si layers leads to increased interaction among dislocations as shown by cross-sectional transmission electron microscopy. This study explores the possibility of using strained layers for achieving flat surfaces and illustrates the need for optimization when using this approach.

No abstract