Broadband optical end-point detection for linear chemical-mechanical planarization (CMP) processes using an image matching technique

In this paper we discuss an end-point detection (EPD) method for the dielectric linear chemical-mechanical planarization (CMP) processes. The proposed EPD algorithms utilize the interferometry optical signals to determine the post-CMP film thickness. A set of collected broadband spectral signals are formed as a spectral image. An image-matching technique is then used to match the processed signal image to the reference image template obtained at the target film thickness. Several matching criteria are discussed and compared. We find that the image correlation coefficient is a good indicator to determine the process end-point. We also consider the impact of the material removal rate variations on the interferometry spectral signals. An analytical calculation is carried out to find an extraction and compression searching range of the spectral image to compensate for the removal rate uncertainties. The correctness and effectiveness of the proposed algorithms have been demonstrated through applications to an inter-metal dielectric (IMD) device CMP process. Compared with other optical EPD methods, the proposed image-matching method is robust to the CMP process variations.