Summary of the Workshop on Bonded and Compliant Substrates

Summary of the Workshop on Bonded and Compliant Substrates

D. Kurt Gaskill

The Workshop on Bonded and Compliant Substrates, first of a new format of ONR sponsored one-shot meetings on topical subjects took place 31 January to 5 February 1998 at the Radisson Normandie Hotel in San Juan, Puerto Rico. The meeting was organized and chaired by T. F. Kuech and U. Goesele, sponsored by ONR (C. Wood) and TMS, attended by 42 people, and conducted in an informal atmosphere. The Workshop consisted of three themes: Semiconductor Bonding, Compliant Substrate Concepts, and Lateral Epitaxial Overgrowth (LEO). For the first two themes, Si, SiC, C and III/V materials were the focus whereas LEO focused primarily on GaN. This summary mainly covers the GaN aspects.

The concept of bonding two different materials into a single, more utilitarian component was first explored by Galileo in the 16th century; this line of thinking has lead to significant progress in bonding semiconductor materials and to compliant substrate demonstrations (SiGe epi) in just the last few years. It is thought that using a compliant substrate approach for GaN epitaxy would greatly reduce dislocations that are now widely observed.

A compliant substrate is one whereby the dislocations generated by the epitaxial mismatch are not propagated through the growing film. There are two ways of implementing a compliant approach: the first makes use of a viscous buffer layer that allows the substrate to be elastically compliant with the epilayer and the second employs a thin twist-bonded layer which plastically deforms to adopt the epilayer lattice. In both cases, bonding technology is employed. The key issue for a compliant substrate is that, in addition to the issues associated with bonding, the time scale of compliance is less than or equal to the layer deposition time. The key issues in bonding are: surfaces need to be carefully prepared to be atomically clean and free of particulates and atomically smooth over large areas and the bonding process (layer orientation, annealing parameters, ambients, pressure, etc.) must be tailored to the material under consideration. Technological hurdles that have not been completely overcome are: dealing with different coefficients of thermal expansion (for dissimilar materials), controllable thinning below 100 nm, lowering the bonding temperature, identification of impurities and source of impurities, and source of the conduction barrier that arises when two different semiconductors are bonded (such as from impurities/traps, local grain boundary effect or dipolar field).

Several presentations modeling the compliant process were made focusing on details of the compliance mechanism and developing a characteristic relaxation time (when using viscous buffer layers). [Freund, Brown; Tan, Duke; Scheerschmidt, Max Planck, Halle; Englestad, Wisconsin] In the latter case, values of various materials parameters were used to estimate the relaxation time for borosilicate-silicate glasses that demonstrated the potential effacy of such a compliant approach [Freund]. Some very early experimental results of the borosilcate-silicate glass approach were presented [Kuech, Babcock, Moran, Matyi,Wisconsin] showing bonding to the glass was possible but the morphology of subsequent epitaxy (InGaAs) was not smooth. Semiconductor layer thinning (part of the twist-bonding process) after bonding is a critical technological hurdle for compliant technology to overcome as these essential layers are often less than 10 nm thick [Lo, Cornell; Meyer, Penn State]. Many attendees felt that the rough morphology often observed after selective chemical etching pointed to problems with the selectivity of the etchant being used for the particular layer composition being etched. Though the twist-bonding approach has demonstrated great promise for GaAs and InP-based epi, some groups have had limited success - indicating that additional work on this approach is needed - and no work in the nitrides was reported. In addition to the technological hurdles described previously, the role of charge traps generated at the twist-bonded interface [Eastman and Greenberg, Cornell] and potential impurity diffusion in the viscous buffer layer approaches need to be investigated.

Some variations on the bonding/compliant substrate theme were AlN/GaN growth on wet oxidized AlAs on Si which displayed CL intensities comparable to GaN grown on sapphire [Dapkus, USC], GaN growth on SOI was described where 350 cm²V^-1s^-1 300K mobilities were obtained for an electron concentration of about 2-3 x 10¹⁷ cm^-3 and Si diffusion into the films was measured by SIMS [Pavlidis, Michigan], and GaN growth on LGO substrates by MBE showed good promise mainly resulting from the significantly improved quality of the substrates - subsequent bonding to GaAs and other substrates was envisioned [Brown, Doolittle, Georgia Tech].

A second approach to achieving high quality epitaxy of a semiconductor when a suitable lattice-matching substrate is unavailable is using lateral epitaxial overgrowth and successful applications of this technique for the growth of GaN were presented [Davis, NCState (6H-SiC substrates); Mishra, UCSB (Al₂O₃ substrates)]. The success was due in part to the nature of dislocations in GaN which lie mainly parallel to the c-axis, which is the growth direction. Then, dislocations originating from the GaN grown in SiO₂ defined trenches do not propagate into the GaN film laterally overgrown over the SiO₂ masked regions. Plan-view TEM images show little or no dislocations in the overgrown material, though some data suggested that dislocations were bending in the vicinity of the SiO₂ mask. AFM investigations show the lateral overgrowth to take place by a step-flow mechanism as no dislocations with a partial screw character were located. Electron concentrations in these LEO films are estimated to be high, probably from n-type impurity diffusion. Several outstanding technical issues remain to be investigated: the gas phase and surface kinetics of the lateral overgrowth mechanism especially with respect to enhancing the lateral to vertical growth rates, the presence or extent of mosaic structure in the overgrown regions and effect of the substrate/buffer layer on mosaicity, the nature of strain in the vicinity of the masks and effect on dislocation propagation, the extent of impurity diffusion (Si, O) from the mask into the lateral layers, and the effect of LEO growth parameters on the yellow band PL emission from the lateral layers. An innovative approach to a nearly dislocation free GaN substrate was proposed that employs two successive LEO steps with the mask of step 2 positioned over the openings defined by the mask of step 1.

 MIJ-NSR    News