A study of the thermal behavior of different test patterns used in differential high resolution electromigration measurements

The differential high-resolution electromigration (EM) measurement techniqu e requires special test patterns since for the implementation of this techn ique the presence of two metal lines is necessary, one metal line to be str essed with a high current density ('test' or 'stressed' resistor) and a sec ond line ('monitor' or 'reference' resistor) to compensate for the thermal instabilities of the stressed one. It is accepted that due to Joule heating the stressed line always acquires higher temperature than the reference li ne and, therefore, additional resistance changes due to thermally induced p henomena, like precipitation of additional elements, will affect the measur ement. In order to minimize these unwanted effects, an optimal high-resolut ion EM test structure should show a minimum temperature difference between the stressed and the reference lines. Based on this requirement, in this wo rk we simulated three different test structures being used by various resea rch groups for high-resolution measurements and subsequently compared these test structures on account of their thermal behavior. Each test structure was examined for the case of two different widths of metal lines, 4 and 0.5 mu m. The results obtained from the simulation of these test patterns demo nstrated that the test pattern comprised of two parallel stress and referen ce lines shows better thermal behavior than the 'lined-up' (continuous) met al lines where the stress and reference lines are actually part of the same line. In particular, the test pattern comprised of two parallel straight l ines has slightly better behavior than the one with meandered lines in term s of minimization of the temperature difference between the stressed and th e reference resistors. The difference between the thermal behavior of the t wo structures though is very small. For that reason, the parallel lines sho uld be preferred from the meandered ones only if layout restrictions do not require the choice of a more compact (i.e., meandered) solution. Both test structures have shown a better thermal behavior than the 'lined-up' metal lines. (C) 1999 Elsevier Science Ltd. All rights reserved.