The passing of current through thin metal lines results in the formation of voids which will eventually cause the line to fail electrically. Gathering of experimental data on electromigration has been limited by two facts: The dimensions of the metal lines on modern integrated circuits are on the order of the resolution limit of light microscopy; and voids in these lines behave quite differently when they are surrounded by a protective film of SiO2 or other dielectric. This dielectric protects the metal surface and constrains the line, pushing back on the metal displaced by void formation. Several additional factors further complicate experimental analysis: the sample test current must be precisely controlled during observation; the sample must be accurately heated to several hundred degrees; the metal lines must be long and narrow (typically 0.2-3.0 (am wide by 300 μm long); testing is often lengthy, sometimes extending over several days even under conditions selected to minimize the time to failure; and electromigration failures are characterized by long incubation periods during which no changes are observed followed by rapid and complex behavior.