The influence of stress concentration and grain boundary diffusion on hillock growth in aluminum film

Several related projects concerned with the growth of hillocks on aluminum films and interconnects are described. Hillock growth is caused by two phenomena, electromigration and the compressive stress due to the thermal expansion mismatch between aluminum and silicon, on which aluminum is commonly deposited in the manufacture of microelectronic devices. Grain boundary diffusion is generally accepted as the principal transport mechanism for hillock growth, but a few researchers have argued for the primacy of surface diffusion.;A series of experiments were performed with aluminum films containing impurities in selected regions. The films were annealed to produce hillocks, after which secondary ion mass spectroscopy and Auger electron spectroscopy were used to characterize the hillocks. It was found that hillocks match the composition of the bulk of the original film much more closely than that of the pre-annealing surface, indicating that they grow by grain boundary diffusion rather than a surface effect.;Hillocks grow at the borders of misoriented grains, and it is inferred that this is a consequence of slip within such grains producing local stress relaxation. The question of why hillocks form at some triple points of such grains, but not others, remains. An effort was made to simulate the distribution of microstresses within the grains of an actual aluminum film, using a finite element model, and to correlate the results of the model with the actual hillock sites. This did not prove successful, probably due to factors not included in the model. However, a correlation was found between hillock sites and the presence of grain boundary misorientations whose type and magnitude had been found by an earlier researcher to produce high grain boundary diffusion coefficients.;Attempts were made to deposit a film with crystallites sufficiently large that their orientations could readily be measured before annealing. This would have made it possible to predict hillock sites from grain misorientations, and compare these predictions to the sites of actual hillock growth. However, these attempts proved unsuccessful, although it is hoped that current developments in microscopy will make it possible to carry out such experiments in the future.