2020 年 86 巻 1 号 p. 80-86
With the background of spreading the small and high performance mobile computers and developing the electric vehicles, the establishment of various high performance devices is demanded. Since chemical mechanical polishing (CMP) process has played an important role for these development, there are several unclear points in the CMP mechanism. For example, the removal rate of CMP increases when the polishing pad which has large number of actual contact points between polishing pad and substrate was used. On the other hand, the excessive number of contact points adversely affects the removal rate, however, this mechanism has not been elucidated. Currently, development of artificial intelligence (AI) technologies remarkably makes progress, so that the introducing method of AI to the CMP apparatus is expected. In this paper, we propose the polishing system by neural networks as well as try to predict and operation among the desired removal rate, the pad surface asperities and conditioning parameters. As a result, we succeed in presenting the example of proposed system to the CMP apparatus. By using the proposed system, we can predict relationship between the removal rate and the polishing pad surface asperities precisely. Furthermore, the proposed system can achieve the desired removal rate within a few percent error by controlling the conditioning parameters.
