Predictive Filter Subsystem for CMP Endpoint Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Chemical mechanical polishing (CMP) processes face challenges in achieving consistent material removal rates across substrates due to variations in initial thickness, slurry composition, polishing pad conditions, and pressure, making it difficult to determine the polishing endpoint and achieve desired substrate profiles.
Innovation Solution
A computer program product and system that utilize predictive filters, such as Kalman filters, to process in-situ measurements of substrate thickness and polishing rates, adjusting pressure and polishing rates in real-time to ensure uniformity and precision in achieving target thickness across multiple zones on a substrate.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a constant pressure is applied during polishing, then the polishing process is simple to control, but variations in material removal rate occur due to substrate thickness variations, slurry composition changes, polishing pad condition, and relative speed variations
Solution Approach 1:
The system transitions from static constant pressure control to dynamic pressure control by continuously adjusting the carrier head downforce based on real-time thickness measurements from optical monitoring. The control system modifies polishing parameters during the process to compensate for variations in material removal rate, ensuring consistent endpoint achievement across different substrate regions.
Solution Approach 2:
The system implements closed-loop feedback control by using in-situ optical monitoring to measure substrate thickness in real-time during polishing. The measured thickness data is fed back to the control system, which then adjusts the polishing pressure and/or speed to maintain the desired material removal rate and achieve the target endpoint thickness uniformly across all zones.
2Ease of operation
If the polishing endpoint is determined solely as a function of polishing time, then the process is easy to manage, but it is impossible to achieve the desired profile due to variations in material removal rate
Solution Approach 1:
The system replaces time-based endpoint determination with optical monitoring-based endpoint detection. Instead of relying on predetermined polishing durations, the system uses in-situ optical measurements to directly monitor substrate thickness and determine the actual endpoint based on measured thickness values, eliminating the need for time-based estimation.
Solution Approach 2:
The control system continuously monitors substrate thickness in real-time during polishing and uses this feedback information to determine the precise endpoint. The system compares measured thickness values against target thickness specifications to dynamically determine when polishing should stop, ensuring accurate profile achievement regardless of variations in material removal rate.
3Measurement precision
If in-situ optical monitoring is used during polishing, then real-time thickness measurement is achieved, but existing optical monitoring techniques do not satisfy increasing demands for endpoint consistency and polishing precision
Solution Approach 1:
The system performs preliminary characterization of the substrate and polishing process by measuring thickness at multiple zones before and during polishing. This preliminary data is used to establish zone-specific polishing parameters and predictions, enabling the system to proactively compensate for expected variations in material removal rate across different substrate regions.
Solution Approach 2:
The system implements zone-specific control by dividing the substrate into multiple zones and applying different polishing parameters to each zone based on local thickness measurements and characteristics. The optical monitoring system tracks thickness variations in each zone independently, and the control system adjusts pressure and/or speed for each zone to achieve uniform endpoint thickness across the entire substrate surface.
Data Source
AI summary
A measured characterizing value dependent on a thickness of a region of a substrate is input into a first predictive filter. The first predictive filter generates a filtered characterizing value. A measured characterizing rate at which the measured characterizing value changes is input into a second predictive filter. The second predictive filter generates a filtered characterizing rate of the region of the substrate. The measured characterizing value and the measured characterizing rate are determined based on in-situ measurements made at or before a first time during a polishing process of the substrate. A desired characterizing rate is determined to be used for polishing the region of the substrate after the first time and before a second, later time based on the filtered characterizing value and the filtered characterizing rate.