化學機械研磨製程中物理現象之動態特性

Abstract

化學機械研磨（CMP）已被認為在製造高性能元件技術上扮演極重要的地位。雖然傳統化學機械研磨能處理非均勻性、平坦度、移除率及表面缺點，可是在缺少即時動態的限制條件下，傳統化學機械研磨無法提昇平坦化的尺度及晶片全面性的均勻度。因為傳統化學機械研磨忽略了製程中晶片表面的現象，而所使用的方法對要求平坦化及移除率的提昇是有點力不從心的。因此，一種嶄新的參數式控制運算法被引進，以達到工業上所要求的全面性高品質均勻度。動態限制條件所考量的是如何去達成晶片表面的均勻度及提昇其效率，模擬出來的結果再與傳統機械研磨技術所獲得的結果作一比較，希望能因此而提昇均勻度及高移除率。這也是目前大家所投入及追求的目標。本文中試圖提出一種新的控制技術，共分為二部份：一為控制系統旋轉速度及運動方式，透過最佳化來達到晶片全面性的平坦；其次為晶片表面壓力的控制，如何透過表面壓力的測量及數位影像處理的技術，了解晶片表面的真正輪廓，均勻度及壓力分佈的狀況，並提供研製程中的即時控制機制。

CMP (Chemical mechanical polishing) is recognized to be of critical importance to high performance interconnect technology. In the CMP process, although conventional CMP technology can process characteristics such as removal rate, nonuniformity, degree of planarization, and surface defects, they still fail to meet the flatness to upgrade and yield a plannarized wafer with high global thickness uniformity due to the absence of instant dynamic process constraints involved. The conventional approach is oversimplified for planarization as well as removal rate to advance since it has ignored the configuration of a wafer in action. However, a novel parametric control algorithm is presented to meet the industrial specifications with high global thickness uniformity, The concern of the dynamic constraints nust be implemented efficiently to achieve an unprecedented degree of thin film smoothness. The simulation results are then compared to those with respect to the conventional technique. It is expected to be increased the uniformity and removal rate. The product characteristics of concern are the removal rate (corresponding to a controlled amount of oxide polished during the step) and the within-wafer uniformity of that removal rate across the wafer. This paper attempts to advance the state of the practice to CMP process control by applying a new algorithmic control technology. In this work, we outline the CMP process and the control scheme used for simulations. The basis idea behind the process is set into two parts. Firstly, system speed, and its kinematics are detected and optimized to a predicted extednd of non-uniformity. Secondly, wafer surface pressure is imaged on prescale films. And then wafer surface images in environments are developed and transferred into 3-D data patterns. This output helps understanding the distribution of pressure over a wafer and improving the conventional model based recipes.