Chip scale modeling of chemical mechanical planarization (CMP) for layout dependent variation

Empirical pattern density-based models have been successfully used for the characterization of layout dependant variation in CMP. However, current models are empirical and do not incorporate physical parameters representing consumable properties and process parameters. This work aims to model the layout dependant chip scale variation in CMP as a function of the physical properties of consumables and process parameters. This is a critical part in the design for manufacturing (DFM) of an integrated chip, and for CMP process and consumable optimization for a given chip layout.; The main source of pattern dependency in CMP is the non uniform contact pressure between the polishing pad and the micro topographies on the wafer. In this work, a different approach from previous chip scale models is taken, and the contact pressure distribution between the pad and the wafer topography is evaluated at the local real contact area between pad asperities and the wafer. The pad asperity height distribution is then used to evaluate the mean contact pressure at the asperity tip. The local material removal rate is expressed as a function of asperity height distribution, asperity size, abrasive size, hardness of the polishing material, and the pattern density. This model effectively captures the effects of pad surface roughness and polishing down force on the CMP step height evolution characteristics. Model predictions for a specially designed test chip fit well with the experimental data with an RMS error of 31nm for 100 measurement points over the test chip, which is approximately in the same range as the error range of purely empirical models.; The CMP model was then applied to a shallow trench isolation (STI) process. To address the effect of highly non-uniform initial topographies generated by the high density plasma CVD (HDP-CVD) deposition, topography dependant stress concentration factors to capture the edge effect were applied to the CMP model to modify the pressure distribution. The relationship between underlying trench layout and the HDP deposition thickness was investigated by measuring the thickness variation as a function of pattern density, line width, and line space of the test pattern.; The variation of the pad surface height distribution, which is one of the key modeling parameters, was examined by white light interferometer measurement in a copper CMP test. The result showed that the aggressiveness of the in-situ conditioning does not significantly affect the final pad surface height distribution. However, the pad surface degradation effect during the pad life time was clearly observed.