ABSTRACTDifficulty during plasma etching and post-etch corrosion are major drawbacks of Al-Si-Cu alloy films, when used for integrated circuit interconnect. Moreover, the relatively large solute mobility of Cu in Al may lead to void formation by precipitate coarsening. As integrated circuit dimensions decrease reliability issues, such as electromigration and mechanical stress voiding, are becoming increasingly important. At present several types of Al alloys are considered as possible alternatives for Al-Si-Cu: Al-Pd, Al-Sc, Al-Pd-Cu, Al-Si-Pd, Al-Si-V, Al-Si-Sc, Al-Si-Pd-Nb, and Al-Si-V-Pd. The latter quaternary alloy has been designed such as to combine the positive aspects of both Pd and V. In comparison with Cu in Al, a) the (low temperature) solid solubility is negligible for Pd and small for V, and b) the mobility is similar for Pd, but very small for V.With transmission-electron microscopy, passivated Al-Si-Cu alloy films have been studied after thermal stressing at 200 °C: ө-Al2Cu coarsening was observed together with void condensation. Lifetests on unpassivated Al-Si-V-Pd alloys at 180 °C and 2xl06A/cm2 have shown an extremely high resistance to electromigration. Electromigration and microstructural data on these quaternary alloys will be presented. These findings suggest how the microstructure is stabilized by the combined action of the V and Pd solute atoms, a) by nm-scale (A1,V) precipitates within the Al grains and b) by small (Al,Pd) particles at the Al grain boundaries. Furthermore, the key issues in terms of reliability related microstructural phenomena are both solute and solvent mobilities in grain interiors as well as along interfaces and grain boundaries. Arguments will be given showing that at low solute concentrations the metals (V and Pd) each by themselves are not effective enough to influence the solvent motion of aluminium along interfaces and grain boundaries significantly. The combination of the two metals, however, was found to be very effective.
Recycled integrated circuit detection using reliability analysis and machine learning algorithms