Evolution of stress and microstructure in Si-doped aluminum gallium nitride thin films

Open Access
Manning, Ian
Graduate Program:
Materials Science and Engineering
Doctor of Philosophy
Document Type:
Date of Defense:
December 17, 2010
Committee Members:
  • Joan Marie Redwing, Dissertation Advisor
  • Joan Marie Redwing, Committee Chair
  • Professor Christopher Muhlstein, Committee Member
  • Elizabeth C Dickey, Committee Member
  • David Snyder, Committee Member
  • wide bandgap semiconductors
  • doping
  • strain
  • stress
  • thin films
  • AlGaN
Al<sub>x</sub>Ga<sub>1-x</sub>N thin films are foreseen to be of central importance to photoemitters and detectors operating within the ultraviolet spectral range, due to the range of achievable bandgap energies spanned by the endpoint values corresponding to GaN and AlN. Such devices are typically based on the formation of junctions between layers of a given semiconductor that have been doped n- and p-type. The dopant that is most commonly used to obtain n-type conductivity in Al<sub>x</sub>Ga<sub>1-x</sub>N is Si. However, the incorporation of Si has been found to reduce the critical thickness for crack formation through the generation of tensile stress. The present work examines the effects of the Si incorporation on the stress evolution of Al<sub>x</sub>Ga<sub>1-x</sub>N thin films deposited using metalorganic chemical vapor deposition. Specifically, tensile stress generation was evaluated using an in situ wafer curvature measurement technique, and correlated with the inclination of edge-type threading dislocations observed with transmission electron microscopy (TEM). This microstructural process had been theorized to relax compressive strain with increasing film thickness by expanding the missing plane of atoms associated with the dislocations. Prior work regarded dislocation bending as being the result of an effective climb mechanism. In a preliminary investigation, the accuracy of the model derived to quantify the strain induced by dislocation inclination was tested. The relevant parameters were measured to calculate a theoretical stress gradient, which was compared with the gradient as extract from experimental stress data. The predicted value was found to overestimate the measured value. It was also confirmed during the preliminary investigation that Si incorporation alone was sufficient to initiate dislocation bending. The overestimation of the stress gradient yielded by the prediction of the model was then addressed by exploring the effects of dislocation annihilation and fusion reactions occurring during film growth. Si-doped Al<sub>0.42</sub>Ga<sub>0.58</sub>N layers exhibiting inclined threading dislocations were grown to different thicknesses. The dislocation density at the surface of each sample was then measured using plan-view TEM, and was found to be inversely proportional to the thickness. As the original model assumed a constant dislocation density, applying the correction for its reduction yielded a better prediction of the stress evolution. In an attempt to extend the predictive capabilities of the model beyond the single composition examined above, and to better understand the interaction of Si with the host Al<sub>x</sub>Ga<sub>1-x</sub>N lattice, several sets of Al<sub>x</sub>Ga<sub>1-x</sub>N films were grown, each with a unique composition. The Si-doping level was varied within each set. It was determined that the dominant influence on tensile strain generation is in fact the initial dislocation density, which increased with increasing Al content as observed with plan-view TEM. This was expounded in a series of modeling examples. In addition, threading dislocation inclination was studied in nominally undoped and Si-doped Al<sub>x</sub>Ga<sub>1-x</sub>N grown under conditions of tensile stress to isolate the influence of Si from that of compressive stress, which had also been found to induce dislocation bending. The effects due to Si and compressive stress were found not to combine as expected, based on a stochastic model of dislocation jog formation that had been developed in prior work to describe the inclination mechanism. Having confirmed the strong, direct relationship between the initial dislocation density and the degree of tensile stress generated in the Al<sub>x</sub>Ga<sub>1-x</sub>N epilayers during growth, an effort was made to demonstrate the advantage that might be gained by using AlN substrates rather than SiC. In principle, AlN provides a growth surface that inhibits defect formation due to its close similarity to Al<sub>x</sub>Ga<sub>1-x</sub>N lattice structure and chemistry, particularly at high Al mole fractions. Threading dislocation densities were reduced by an order of magnitude in comparison with samples grown on SiC, with a corresponding reduction in the stress gradient arising from dislocation inclination.