Electromigration in thin-film interconnection lines: models, methods and results

Abstract

Electromigration (EM) in thin-film interconnection lines is one of the major concerns for the development of ULSI devices, employing advanced design rules. Starting from the early sixties, several techniques have been used to characterize this phenomenon, producing a large, but frequently contradictory, amount of data. Different models have been proposed, but the complete comprehension of the basic physical mechanisms leading to EM is still unsatisfactory. In this work, well-established results and unsolved problems are reviewed. The physical model based on the general diffusion theory is used to describe the EM failure mechanism; the influence of the different stress parameters (temperature, current density, mechanical stress), of material properties (structural inhomogeneities, chemical composition) and line topography are taken into account. The accelerated methods employed to evaluate the EM resistance of the lines are classified into destructive and non-destructive, according to their effects on the samples under test. In the first group, a core position is occupied by the so-called median time to failure (MTF) technique, that has been extensively employed to gather results on many different materials and structures. Within the same group a survey is given of resistometric methods and faster techniques, based on a further acceleration of EM by means of high current densities and related Joule heating. The parameters extracted with these techniques are discussed in relation with the MTF results. The choice of a suitable statistical distribution, related to both the times to failure (TTFs) and the parameters used to estimate the EM performance with alternative methods, is also reviewed. More recently, an increasing importance has been achieved by non-destructive techniques able to give reliable information about the phenomenon without irreversibly damaging the samples. An important section of this work is devoted to the discussion of these techniques, which are mainly based on the accurate measurement of the resistance drift or low-frequency noise induced by the damage occurring at microscopic level. In the course of the discussion, particular emphasis is given to the comparison of the results obtained with the different techniques and to the improvement achievable by employing new materials and structures, including different aluminum alloys and Al/refractory metal sandwiches.