Research On Method Of Model-Based Infrared Reflectometry For Micro/Nano Deep Trench Structures

Deep trench structures with high-aspect ratio have been widely used in the design and manufacture of microelectronic and microelectromechanical system (MEMS) devices. As devices with deep trench structures have been packed more tightly, the critical dimension (CD) and depth dependent features must now be monitored to ensure process control. The endpoint method is commonly used in the control of etching depth which is assumed to be dependent on etching time multiplied by etching rate, however, the etching depth error using this method is usally very large due to the fact that the etching rate relies on a variety of factors and is difficult to be kept as constant. Therefore, it is important to perform the on line monitoring of etching process of the micro/nano deep trench structures accurately and nondestructively so as to ensure the effective process control. On the basis of such a research background, this dissertation intends to study systematically on the theory and method of the model-based infrared reflectometry (MBIR) for the micro/nano deep trench structures, and to develop successfully a prototype system of the MBIR metrology.At first, the effective optical model of the micro/nano deep trench structures with high-aspect ratio has been established. A modeling method named corrected effective medium approximation (CEMA) has been proposed to calculate the reflectivity of the one-dimensional (1-D) and two-dimensional (2-D) deep trench structures, and has been demonstrated to be not only fast in calculation but also accurate enough in comparison with the rigorous coupled wave analysis (RCWA) method for both 1-D and 2-D trench structures, hence is especially suitable for the MBIR metrology. Based on such a novel modeling method, extensive simulations have been carried out on typical types of micro/nano deep trench structures, and thus discuss the relationship between the the reflectance spectrum and the geometric parameters of the trench structures.Then, the extraction of the geometric parameters of trench structures in the MBIR metrology has been investigated as a typical spectral inverse problem. With the help of the above proposed modeling method, two approaches to fast parameter extracting have been put forward, namely adaptive simulated annealing and Levenberg-Marquardt combined method (ASA-LM) and artificial neural neural network and Levenberg-Marquardt combined method (ANN-LM). Simulations performed on a 1-D deep trench structure has verified that both of the proposed ASA-LM and ANN-LM methods are accuarate and fast in computation for the parameter extraction. Finally, a prototype system of the MBIR metrology for the measuring of micro/nano deep trench structures is described in detail. Several key technologies involved in developing the system has been investigated, including design of the infrared detecting optical path with high sensitivity, suppression of the noisy reflection from the wafer backside, high-speed signal acquision and processing for the spectral measument, software programming for the signal analysis and automatic parameter extraction, etc. The prototype system has been applied to some multilayer thin films structures and a typical deep trench structure with experimental resutls demonstrated that the trench geometric parameters can be extracted with a nanometer scale accuracy.The proposed MBIR technique is hopeful to provide a non-contact, non-destructive, time effective, low cost and high resolution tool for the measurement of deep trench structures. It is expected that this technique will have potential applications in the on-line monitoring and process control for microelectronics and MEMS manufacturing, and it will be widely used in the trench etching fault diagnosis, in-line monitoring, wafer mapping and other aspects.