Modeling Pattern Effects in Oxide CMP
- PDF / 7,378,805 Bytes
- 6 Pages / 612 x 792 pts (letter) Page_size
- 60 Downloads / 252 Views
W5.10.1
Modeling Pattern Effects in Oxide CMP R. Rzehak* Infineon Technologies SC300 GmbH & Co.OHG, Königsbrücker Str. 180, 01099 Dresden, Germany ABSTRACT We present an extension of the density-stepheight model for pattern effects in oxide CMP which accounts for the roughness of the polishing pad's surface. The model is compared to polishing data for processes using different pressure and speed. Agreement with the data is improved especially in the initial regime of polishing before the pad contacts the down areas. I. INTRODUCTION As the sizes of modern integrated circuits (IC's) continue to shrink, requirements on the planarity of structures after chemical mechanical planarization (CMP) become more and more stringent so that a better understanding of the relevant mechanisms effecting planarization is necessary to meet future process specifications. One particular problem that has attracted a large amount of research activity is the within-chip non-uniformity of CMP'ed oxide films at various stages during IC fabrication which may affect the proper functioning of the circuit. This nonuniformity arises from a pattern-dependence of the planarization. To predict its severity, optimize the CMP process and eventually modify the chip design e.g. by insertion of dummy structures, chip-scale CMP models have been developed [1-3]. Parameters appearing in these partly phenomenological models are determined by a process characterization methodology matching the model to polishing data obtained for simple test structures, commonly lines of varying density and pitch [4]. Since in oxide CMP non-uniformities due to density variations are typically several times as large as those due to pitch variations focus has been more on the former [5]. In essence, most successful chip-scale CMP models attribute pattern-dependent planarization chiefly to the different distribution of pressure between up and down regions for structures of different densities. To calculate the local pressures, the pad is modeled as a collection of independent Hookean springs as sketched in Fig. 1. Two regimes of polishing are distinguished. Before the so-called contact time, the pad is assumed to touch only the protruding (up) regions and the pressure exerted on these is simply the nominal pressure corrected for the reduced area of contact while no pressure is exerted on the recessed (down) regions. After the contact time, the pad also touches the down regions. Due to Hookes`s law, the pressure difference between up and down regions then is proportional to the step height while the density-weighted sum of the pressures exerted on both up and down regions still must equal the nominal pressure. These two relations uniquely determine the pressures exerted on the up and down regions after the contact time. Relations for removal rates are finally obtained by invoking Preston`s law. The bare-bones concept just described may be considered as a kind of minimal model for pattern-dependent planarization which represents the common ground of many more elaborate *
This wor
Data Loading...
