December 1-6, 2013 | Boston
Meeting Chairs: Charles Black, Elisabetta Comini, Gitti Frey, Kristi Kiick, Loucas Tsakalakos
The interplay of plastic and elastic relaxation mechanisms in strained epitaxial films is a complex balance of the energetics and kinetics of the evolving system. We will review the use of in-situ electron microscopy methods to quantify the kinetics of misfit dislocation generation and to map defect microstructures as functions of growth and processing conditions in such thin film systems, specifically in Ge(x)Si(1-x)/Si. This work has led to enormous volumes of data (e.g. hundreds of hours of video, thousands of photographic negatives). In the spirit of the US “Materials Genome Initiative” we seek to develop new methodologies to collate, organize and analyze this data to develop new understanding and avenues for process control through consideration of the whole body of data. We present a new methodology for organizing and illuminating complex interdependent dislocation kinetic mechanisms, that is related to the phylogenetic methods used in bio-informatics and evolutionary biology. Within our analogous “materials cladogram”, different branches indicate the evolution of different structures from a common original structure, based upon different kinetic pathways. This approach can capture both the points of divergence at which different structures emerge during a growth or processing sequence, and the relevant kinetic parameters that define the structure of divergent branches. The development of such pathways can be mapped using simulations that capture and integrate the essential quantitative kinetic descriptions derived from the experiments (e.g. activation parameters for dislocation glide, dislocation nucleation rates, dislocation interaction processeshellip;). Comparing the generated maps with large numbers of specific experimental observations then allows refinement of the simulation structure and increased accuracy in the determination of the relevant kinetic parameters, eventually enabling generation of new cladogram branches by simulation alone. Ultimately, this can provide the structure for a processing map that captures the set of different kinetic pathways and resulting structures for a given system, and helps define the key experimental parameters required for extension to new systems. We acknowledge the contributions of David Sandler (RPI). Original experimental work in collaboration with J. Bean, J. Floro (U. Virginia); F. Ross (IBM); and E. Stach (BNL).
Metamorphic buffer layers (MBLs) are of great interest for the development of new semiconductor devices with alloy compositions that are not typically feasible due to the high defect density resulting from the mismatched epitaxy. MBLs grown by hydride vapor phase epitaxy (HVPE) are especially promising because they can achieve a high degree of strain relaxation while depositing thick layers that enable the use of chemical mechanical planarization. While MBLs have been in use for quite some time, the mechanisms which govern dislocation generation and propagation, strain relaxation, and tilting are still unclear. HVPE-grown MBLs provide a unique tool for understanding these processes, as a wide range of layer thicknesses, beyond what is typically employed in MOVPE and MBE, can be employed allowing relaxation to be observed at many different stages of growth. A combination of TEM, high resolution reciprocal space mapping (RSM), and electron microprobe was employed to gain a clearer picture of the compositional and strain states of the various layers in a series of HVPE-grown InxGa1-xAs MBLs. It was found that there are dislocations lying perpendicular to the growth direction in the constant-composition capping layer of the MBL that lie above the final compositional interface. These dislocations were correlated with RSM data that indicate that the capping layer in these step-graded MBLs is partially relaxed. Since it was observed that the majority of the capping layer closest to the surface is defect free, it appears that these dislocations have climbed from sources present at or near the last compositional interface, relaxing the lower portion of the cap. The upper portion of the cap remains nearly fully strained with respect to the previous composition step. This is in contrast to the commonly assumed mechanism, in which dislocation loops are thought to nucleate at the surface and then propagate down towards the nearest compositional interface. Tilting behavior in these layers was measured by x-ray diffraction omega-phi mapping. It was found that tilt magnitude typically increased with xInAs and did not depend on grading style (linear vs. step-grading). The direction of the tilt was initially random on nominally (100) oriented substrates and changed as grading continued, appearing to ‘twist&’ around the growth direction. MBLs grown on 4° miscut substrates tilted in the opposite direction of the miscut, and the tilt magnitude for a given composition was greater.
Group III-Sb materials show the best hole transport among III-V materials and are considered as promising candidates for future p-type MOSFETs in all III-V CMOS technology. Silicon as a universal microelectronics platform is especially attractive substrate, and III-Sb films with low-defect density and smooth surface are critical for achieving commercial viability of devices made of these materials. However, large lattice mismatch and non-polar nature of Si substrate present a challenge for growth of high quality III-V materials. Molecular beam epitaxy of GaSb and strained InGaSb quantum wells were employed using metamorphic buffers, GaSb/AlSb superlattice or AlGaSb layers. In both cases, the growth was