On the Pattern Dependency and Substrate Effects During Chemical-Mechanical Planarization for Ulsi Manufacturing
- PDF / 754,467 Bytes
- 6 Pages / 417.6 x 639 pts Page_size
- 92 Downloads / 147 Views
decreases with time until planarization is achieved. As the polish approaches the patterns underneath, the interaction between substrate effects and pattern dependency results in the resurgence of step height. The implication of this newly found phenomenon is discussed. INTRODUCTION The utilization of chemical-mechanical planarization (CMP) process during ULSI manufacturing has been a standard practice in most IC fabs worldwide to planarize the uneven surface topography and to delineate circuit patterns. The popularity of CMP, however, has not been accompanied adequately by a clear understanding of the operating mechanisms involved, in order for a more systematic control of this process. To planarize the wafer surface topography, the polish action has to remove and clear the step height resulting from the deposition process, before the specified planarity can be achieved. This action is complicated by the differential polish rates between the "up" and "down" features, and their dependence on the density of the patterns buried beneath the layer being polished (i.e., pattern dependency).'1 3 Further complexity arises when the polish rates of the same material vary with different substrate materials underneath (i.e., substrate effects). 4 '5 Lack of control over the issues described above would lead to loss in planarization efficiency, over or under polish, and poor process reliability. Some previous modeling and experimental studies have shed some light on the origins of pattern dependency and substrate effects. However, there are contradictions and ambiguities, and further work is still needed before a clear picture can be drawn. Most people attributed the pattern dependency to the uneven partitioning of the down pressure among circuit patterns with different densities and areas, leading to local pressure variation and hence different local removal rate. As a consequence, the step height reduction rate would be a function of local removal rate, which depends on the pattern density. Stine et al. found that the amount of step height reduction is a linear function of time.l However, Grillaert et al. considered local pressure resulting from pad deformation and found that, the step height reduction is a linear function of time for an incompressible pad, and an exponential function of time for a compressible pad.3 Wang et al. reported that the removal rate of blanket oxide wafers decreases or increases depending on the type substrate material beneath (nitride or Al, respectively), when the oxide being polished is thinned down below a certain thickness. 4 To what extent can this substrate effect interact with the pattern dependency to affect the planarization behavior is not clear yet. In this work, we perform a series of CMP experiments on wafers with different pattern densities and width. The evolution of step height reduction is monitored continuously. The interaction between pattern dependency and substrate effects is observed, and the impacts of these effects on CMP process control are discussed. 217 Mat. Res. Soc. Sy
Data Loading...
