Elsevier

CIRP Annals

Volume 48, Issue 1, 1999, Pages 143-146
CIRP Annals

The Effect of Pad Wear on the Chemical Mechanical Polishing of Silicon Wafers

https://doi.org/10.1016/S0007-8506(07)63151-5Get rights and content

Abstract

The Chemical Mechanical Polishing process planarises wafers with a high degree of success, however wear on the polishing pad causes the planarisation rate and the post-process planarity to deteriorate. To date, there has been no method of predicting the effect of this wear on the wafer planarity. Using finite element models of the process for new and worn pads the wafer stress distribution on the wafer surface can be predicted. Equating high stresses to high material removal rates these models predict that the process should become ‘centre slow’ as the pad wears. This correlates well with experimental data.

References (11)

There are more references available in the full text version of this article.

Cited by (42)

  • Polishing pad for reducing edge roll-off while maintaining good global flatness of silicon wafer full names, addresses and affiliations of authors

    2020, Precision Engineering
    Citation Excerpt :

    Because the thickness uniformity of a wafer, especially, is critical to achieving high yield in semiconductor manufacturing [3,4], there is a continuing demand for further improvement in the surface flatness of a wafer. However, the surface flatness significantly deteriorates near the wafer edge because edge roll-off (ERO) occurs in the polishing process as the final stage of wafer manufacturing [5]. Reduction in the amount of ERO in the polishing process, therefore, is one of the most crucial technical challenges for every wafer manufacturer.

  • A virtual metrology method with prediction uncertainty based on Gaussian process for chemical mechanical planarization

    2020, Computers in Industry
    Citation Excerpt :

    Considering the chemical interactions during CMP, MRR is collectively affected by various dynamic process variables, including the usage variables of the replaceable materials, which are usually difficult to be illustrated in the physical-based models, but can be modeled by data-driven models. However, in the general practice, these process variables changing with time due to machine degradation makes the MRR predictions inaccurate (Byrne et al., 1999). Although most of the mentioned data-driven models consider the change of MRR with time, they can only involve limited past MRR values.

View all citing articles on Scopus
View full text