Advanced Etch-to-Depth Process Control Using Adaptive Interferometric Endpoint and Endpoint/Host Parameter Exchange

摘要

Accurate endpoint control solutions for fabrication of trench capacitor structures can be complicated by two things: etch rates which vary significantly with increasing etch depth, and etch mask thickness variations that are present wafer-to-wafer and lot-to-lot. This work reports on two phases of effort, and includes relative measures of success of each phase, which have addressed these process control challenges. The first phase is the implementation of an reactive/adaptive approach to the etch-to-depth control problem that accommodates I) etch rate variation within a wafer run and ii) wafer to wafer mask thickness variations. The second phase includes further refinement of the adaptive control algorithm with feed-forward information related to the lateral critical dimension (CD) of wafer trench capacitor structures. The feed-forward CD information allows the control algorithm to more accurately "track" the changing ER earlier within the wafer process. The etch target of a recess etch process is ideally the difference in height of the top of the capacitor electrode being etched and the bottom of the mask, it is not the total etch depth from the start of the process. Thus, the capability to measure and compensate the mask thickness on a run-to-run basis is the key to achieving optimal process control. The process control algorithm must also account for pre recess, or "dishing" of the capacitor structures from two modes: over-etching in the blanket etch back of the poly trench fill (in recess 2, or R2), and CMP selectivity (in recess 3, or R3). In the course of this work, an advanced endpoint system was introduced into the production environment and the performance improvements evaluated. The recess process of record (POR) relied upon daily characterization of chamber etch rate, with the ER result then used to adjust chamber etch times so as to hit the etch target. Etch time adjustments were made to compensate variations in incoming CD and mask thickness and were unique to each chamber. In line metrology used for compensation included mask thickness measurement by Optiprobe and atomic force microscopy for recess depth determinations. A summary of typical variation of the POR is shown normalized about the mean value in Figure 1. The production variation of the etch depth around the normalized target depth for the POR is 0.08 (1σ). An advanced feature endpoint system was implemented to provide closed loop control of the recess process. This endpoint system (Applied Materials EyeD-IEP) performed in-situ mask thickness measurements prior to the recess etch, and then during the recess etch, the endpoint algorithm adapted to changing etch rate with depth to arrive at the depth under mask measurement and the endpoint determination. The results across a number of production lots for the variation in etch depth is 0.044 (1σ). The results are presented in Figure 2. The second phase of the work added run-to-run process control to the endpoint approach: lateral CD values were downloaded from the fab host to the endpoint system. This parameter exchange capability is bidirectional, so the host could both send and receive parametric information to the endpoint system on a run-to-run basis. In this work, the CD values were used to compensate the endpoint algorithm, providing a measure of correction for the etch rate dependency on CD. The results across a sampling of production lots for the variation in etch depth is 0.038 (1σ). These results are presented in Figure 3. Details regarding the results and approaches will be presented in the full paper.