Abstract : In situ metrology and real time process control are fundamental challenges for the future of the microelectronics industry. For this purpose, scatterometry, which is an optical characterisation technique based on scattered light seems to have a great potential. It is a non destructive method that allows indirect measurements of periodic patterns.
For solving this inverse problem, one use to compare an acquired optical signature (from an ellipsometer, for example) with a large set of pre-computed signatures: the library. In this thesis, this principle, which is commonly applied in static conditions (on line sample measurements), is expanded to dynamic applications (real time process control) for which acquired signatures have a low resolution but a high acquisition frequency.
These developments are based, on the one hand, on a new shape reconstruction algorithm (inspired by the Tikhonov regularization) and on the other hand, on a innovative computing architecture: the graphics processing units (GPU).
For adjustment and validation purpose, we have dealt with microelectronics processes for which real time monitoring is key for the future: plasma etch process (namely resist trimming) and resist reflow for nano-imprint.