Abstract: Provided is a method for determining one or more profile parameters of a structure using an optical metrology model, the optical metrology model comprising a profile model, an approximation diffraction model, and a fine diffraction model. A simulated approximation diffraction signal is generated based on an approximation diffraction model of the structure. A set of difference diffraction signals is obtained by subtracting the simulated approximation diffraction signal from each of simulated fine diffraction signals and paired with the corresponding profile parameters and used to generate a library of difference diffraction signals. A measured diffraction signal adjusted by the simulated approximation diffraction signal is matched against the library to determine at least one profile parameter of the structure.

Abstract: Provided is a method of controlling a fabrication cluster using a machine learning system, the machine learning system trained developed using an optical metrology model. A simulated approximation diffraction signal is generated based on an approximation diffraction model of the structure. A set of difference diffraction signal is obtained by subtracting the simulated approximation diffraction signal from each of simulated fine diffraction signals and paired with the corresponding profile parameters. A first machine learning system is trained using the pairs of difference diffraction signal and corresponding profile parameters. A library of simulated fine diffraction signals and profile parameters is generated using the trained first machine learning system and using ranges and corresponding resolutions of the profile parameters. A measured diffraction signal is input into the trained second machine learning system to determine at least one profile parameter.

Abstract: Provided is a method of controlling a fabrication cluster using a machine learning system, the machine learning system trained developed using an optical metrology model, the optical metrology model comprising a profile model, an approximation diffraction model, and a fine diffraction model. A simulated approximation diffraction signal is generated based on an approximation diffraction model of the structure. A set of difference diffraction signal is obtained by subtracting the simulated approximation diffraction signal from each of simulated fine diffraction signals and paired with the corresponding profile parameters. A first machine learning system is trained using the pairs of difference diffraction signal and corresponding profile parameters. A library of simulated fine diffraction signals and profile parameters is generated using the trained first machine learning system and using ranges and corresponding resolutions of the profile parameters.