Investigation of CMP Performance According to Pad Surface Condition and Abrasive Types

Abstract

Chemical mechanical planarization (CMP) is a planarization technique that combines mechanical and chemical factor for enhanced polishing and finishing. CMP process reduces the surface roughness to meet the requirements of semiconductors. The characteristics of CMP, such as material removal rate (RR), thickness uniformity and surface quality, are determined by various mechanical and chemical factors. CMP has many consumables. Slurry, conditioner, and pad are typically used in CMP process. These consumables’ interaction is very important in CMP process. The reason why, the other can affect each other. These factors are known to significantly affect the results of CMP. During the process time, the lack of conditioner will block the pore and glazing on the pad surface. The longer process time, the lower pad cut rate is occurred because of poor conditioner performance [1]. And then pad glazing is occurred. When the pad glazing occurred, RR is changed [2]. At this point, the RR has different value according to abrasive. The number of particles that can act on the wafer depends on the pad surface changed. This affects the local pressure acting on the wafer. As the amount of abrasive acting on the wafer increases, the local pressure on the wafer is decreased. As briefly think about the situation, reduction of local pressure should result in decreasing removal rate. However, when the pad glazing situation, the RR has difference phenomenon according to abrasive types. Normally, Fig.1 (a) shows RR using ceria abrasive. It has higher RR after pad glazing. Fig.1 (b) shows RR using silica abrasive. It has lower RR after pad glazing. In this research, the difference in RR according to abrasive types generated during pad glazing was investigated. Based on these phenomena, I present the optimum pad land area each abrasive. Reference [1] C. Shin, H. Qin, S. Hong, S. Jeon, A. Kulkarni, and T. Kim, Journal of Mechanical Science and Technology, pp. 5659-5665, (2016).[2] H. Kim et al., Wear, pp. 93-98, (2017). Figure 1