In-situ CMP Endpoint Detection Using Acoustic Emission

Abstract

Chemical mechanical planarization (CMP) is a highly complex process where material removal may vary between wafers and within a wafer surface. Optimal performance and high yields in CMP require process monitoring, particularly endpoint detection, which aims to determine the exact polish time needed to avoid defects created by overpolishing. This paper shows that endpoint detection using acoustic emission (AE) measurements determines the endpoint of a simulated shallow trench isolation (STI) CMP process ∼10 s faster than a traditional friction force approach, which prevents a 5% overpolish. A so-called endpoint frequency method based on Fast Fourier Transform (FFT) analysis is also described for CMP endpoint detection. This method predicts the endpoint a priori and is more accurate than the traditional box method found in industry. Furthermore, the use of FFT analysis in the endpoint frequency method inherently provides process parameter information that enables in-situ fault detection.