Electromigration measuring techniques for grain boundary diffusion activation energy in aluminum

The activation energy E_g for electromigration along aluminum grain boundaries has been examined intensively by many authors and different methods. However, the common techniques yield very contradictory results, with E_g ranging between 0.3 and 1.2 eV. For clarifying these contradictions, the present work examines the validity of the different test methods and their basic assumptions. It is shown that for most of these techniques the test quantity being measured is given by time dependent divergences which cause local stripe variations from scarcely detectable voids up to stripe interruptions. This means that constant conditions are not maintained during the test period, thus yielding more or less incorrect activation energies. A well defined measurement of the mass transport and the activation energy is practicable only if the divergence of the electromigration is both constant for a fixed test condition and proportional to the flux for different test conditions. Additionally, all other diffusion terms except electromigration must be negligible, especially the diffusion term based on pressure gradients. These preconditions are satisfied for the direct measurement of the drift velocity of an aluminum front on a TiN stripe, proper test conditions provided. That is why this technique is applied here extensively for clarifying several contradictions. Other techniques were performed for comparison yielding different results. The here obtained valid values of E_g and their variation between 0.43 and 0.47 eV are smaller than generally reported, even for extremely different evaporation and sputtering conditions. In contrast to former results obtained by improper test methods, the activation energy is temperature independent in the examined temperature range between 140 and 280°C, and E_g is not dependent on grain size even for most different fabrication conditions, yielding grain sizes from 0.3 up to 3 μm.