MOS VLSI reliability and yield trends

Abstract

The yield and reliability capability of an MOS technology are shown to be the product of at least six different technological trends; namely those towards: 1) more complex device structures, 2) scaled down feature sizes, 3) increased wafer sizes, 4) factory automation, 5) increased die size and package lead counts, and 6) increasingly sophisticated computer-aided design tools. The capabilities of a specific technology are a function of the equipment and processes by which it is manufactured and these are often the rate-limiting factors for evolution to the next generation of technology. However, because of the impact of scaling trends on MOS IC failure mechanisms, reliability concerns are starting to dominate the rate of technology change. This is evidenced, even at the present, by the fact that technology decisions must be made by trading off one reliability failure mechanism against another. For example, the high storage charge density needed to provide strong signals for alpha-particle-induced soft error immunity, produces high electric fields and oxide breakdown problems in thin-oxide MOS storage capacitors. Failure distribution for both failure mechanisms are shown to be exponentially dependent (in an inverse manner) on the scale factor for oxide thickness. Hot-electron degradation is also exponentially dependent on the scale factor for channel length. Metal and contact electromigration lifetimes can be reduced by the seventh and nineth powers of scale factor, respectively. The implications are that the dominant reliability mechanisms may change in the future, and that wearout will start to impinge on reliability life.