Modelling the influence of pad bending on the planarization performance during CMP
- PDF / 396,820 Bytes
- 6 Pages / 420.48 x 639 pts Page_size
- 60 Downloads / 198 Views
Abstract One of the major problems with oxide-CMP is the oxide thickness variation after CMP within one die, the socalled Within Die Non-Uniformity (WIDNU). The variations in pattern density of the design layout causes different local removal rates across the die resulting in the WIDNU. In this paper we shown that this is influenced by the pad stack. Depending on the thickness of the top pad the WIDNUJ can be reduced from about 470 nm until almost zero. This will be related to the bending of the top pad. The modelling will focuss on the two extreme cases of perfect pad bending and no pad bending.
Introduction Chemical mechanical polishing planarizes efficiently structures up to 1 mm. On one hand this short range planarity results in small step heights approaching zero within one structure on a wafer polished for a long time. On the other hand global planarity requires a flat surface within one die. This long range planarity up to the die size (2-3 cm) is often referred to as the Within Die Non Uniformity (WIDNU) for the oxide, i.e. the range of oxide thickness within one die. Typical values range between 200 to 400 nm. This is to large for the subsequent lithography steps. In ref. 1-6 it is concluded that the pattern density is the dominant layout factor contributing to the W1DNU. A model is presented for the dependence of the local
removal rate on the pattern density. Based on this model the oxide thickness as function of the polishing time is calculated and the WIDNU is deduced. Generally a stacked pad is used during CMP: a hard top pad is glued on a soft bottom pad. Two extreme cases are considered: the top pad bends perfectly or there is no bending at all. To our knowledge up to now only perfect pad bending was assumed in the literature (see e.g. ref. 3 and 4). In the literature several models are presented that try to predict the oxide thickness after CMP at each point starting from the layout of the design (ref. 3 and 6). This is of course the ultimate target, but in order to obtain a reasonable agreement between the model and the experiment several fit parameters without physical interpretation are introduced. The purpose of the model presented in this paper is not to calculated the oxide thickness after CMP at each location of the die, but to give an understanding of the importance of the pad bending and to identify the key parameters. With the obtained insight further optimization of the CMP-process is possible.
Experimental setup The test wafers had the following process flow. After the deposition of 200 nm of oxide the metal stack with a total thickness of 2200 nm was sputtered. The wafers 45 Mat. Res. Soc. Symp. Proc. Vol. 566 ©2000 Materials Research Society
were patterned with a special test mask with structures of various pattern density (0 %,25 %, 50 % and 75 %). The pitch of these structures was 400 or 800 ptm allowing optical measurements on a KLA-Tencor FT-700. After metal etching a thick oxide of 4000 nm was deposited. The wafers were polished on an Ipec Avantgaard 472 and cleaned on an Ont
Data Loading...
