Optical scatterometry with neural network model for nondestructive measurement of submicron features

Characterization of the geometrical parameters of microstructures in electronics and photonics is an important problem from the point of view of fabrication methods. With UV and especially electron-beam lithography feature sizes of the order of hundreds of nanometers are attainable. However, there is a lack of methods for a fast, reliable, and quantitative inspection of these structures. Scanning electron microscopy and atomic force microscopy have a good resolution but they require expensive equipment and are not suitable as on-line methods. Optical scatterometry is a nondestructive technique which predicts the structure geometry from a scattered intensity distribution. We utilize optical scatterometry with a hierarchical system of neural networks for the characterization of diffractive gratings with submicrometer features. It is shown that five geometrical parameters may be predicted simultaneously from the grating with an accuracy of less than 5 nm for the depths of the grooves and the line widths. Furthermore, the hierarchical system reduces the requirements for a prior information of the grating structure.