ARCHIMEDES - Automatic and Reliable Classification of Highly Inline Measured Wafer Edge Defects using Embedded Screeners

Project: Research project

Project Details

Description

During the last years is developed and patented a groundbreaking new technology for optical measurement of object edges with the aid of FFG funding, and created a functional laboratory demonstrator, which produces a continuous stream of edge geometry data. The so-called „Wafer Edge Screener“ system is particularly intended for inspection of the bevel of semiconductor wafers, and enables completely new applications in process optimisation and quality control, in order to improve early fault detection and to increase production yield. Existing wafer edge inspection tools typically use multi-camera vision systems and image processing, which produces data that is very difficult to interpret accurately, which are very costly (typically six- to seven-figure $-sums), and which usually are large and bulky installations. For that reason, they are only deployed in very small numbers at key locations across a production facility, for inspection of small sample sets of wafers. The Wafer Edge Screener has a very compact form factor, and measures the geometrical wafer bevel thickness directly, which can be utilised for automatic defect detection and characterisation. It also operates largely independent of optical surface properties such as transparency, reflectivity or absorption of the wafer material. It therefore can be integrated and deployed in many process facilities and equipment in order to monitor wafer edge integrity and quality throughout the entire factory, which is impossible with existing technology on the market. The goal of this R&D project is primarily to conduct research for new solutions and methods in order to autonomously analyse and interprete the measurement data of the Wafer Edge Screener, such that wafer edge features can be characterised automatically, and further handling decisions can be derived in a robust manner. Natural features shall be discriminated from any defects or contamination present, classified into categories and parametrised. A pilot installation shall be implemented and integrated, in order to provide comprehensive measurement data from a production environment. That data shall be complemented with additional laboratory measurement data to create a sufficiently large database, which can be used as a basis for training machine learning algorithms. Also, already ongoing research work with the goal of drastically improving the measurement accuracy and rate shall be continued, which is necessary to reliably detect microscopic defects. This will ultimately improve early detection of defects that are in the process of formation. A redesigned prototype batch shall be built, and equipped with the required interfaces such that the Wafer Edge Screener can be effectively integrated and utilised in the production environment. First basic results of the research that focuses on the machine learning methods shall be integrated into the measurement system, and tested in a laboratory environment as well as the production environment.
StatusFinished
Effective start/end date1/08/1730/09/19

Fingerprint

Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.