Mechanism of fluoride-based etch and clean processes
Pande, Ashish Arunkumar
MetadataShow full item record
Fluoride-containing solutions are widely used to etch silicon dioxide-based films. A critical issue in integrated circuit (IC) and microelectromechanical systems (MEMS) fabrication is achievement of adequate selectivity during the etching of different film materials when they are present in different areas on a device or in a stack. The use of organic fluoride-based salts in aqueous/organic solvent solutions can yield etch selectivities <1.9 for thermally-grown silicon dioxide relative to borophosphosilicate glass films, and thus may also obviate the need to add surfactants to the etch solutions to realize uniform etching. Etch studies with aqueous-organic fluoride salt-based solutions also offer insight into the etch mechanism of these materials. Specifically, the importance of water content in the solutions and of ion solvation in controlling the etch chemistry is described. With respect to fluoride-containing solutions, etching of SiO₂ films using aqueous HF-based chemistries is widely used in IC and MEMS industries. To precisely control film loss during cleaning or etching processes, good control over the contact time between the liquid (wet) chemistry and the substrate is necessary. An integrated wet etch and dry reactor system has been designed and fabricated by studying various geometrical configurations using computational fluid dynamics (CFD) simulations incorporating reaction kinetics from laboratory data and previously published information. The effect of various process parameters such as HF concentration, flow rate, and flow velocity on the etch rates and uniformity of thermally-grown silicon dioxide and borophosphosilicate glass films was studied. Simulations agree with experiments within experimental error. This reactor can also be used to wet etch/clean and dry other films in addition to SiO₂-based films using aggressive chemistries as well as aqueous HF under widely different process conditions. A spectroscopic reflectometry technique has been implemented in-situ in this custom fabricated reactor to monitor the thickness and etch rate in wet etching environments. The advantages of this technique over spectroscopic ellipsometry in specific situations are discussed. A first principles model has been developed to analyze the reflectometry data. The model has been validated on a large number of previously published studies. The match between experimental and simulated thickness is good, with the difference ~ 5 nm. In-situ thickness and etch rate have been estimated using Recursive Least Squares (RLS), Extended Kalman Filter (EKF) and modified Moving Horizon Estimator (mMHE) analyses applied to spectroscopic reflectometry using multiple wavelengths with ZnO employed as a model film. The initial guess for EKF and mMHE has been obtained from a CFD model. It has been shown that both EKF and mMHE are less oscillatory than RLS/LS in the prediction of thickness and ER and more robust when a smaller number of wavelengths are used, in addition, the computational time for EKF is less than that of mMHE/RLS. For no restrictions on computational requirements, LS should be the method of choice whereas in the case of faster etching systems, with the availability of a better process model, EKF should be starting point. The choice of algorithm is thus based on sampling rate for data collection, process model uncertainty and the number of wavelengths required.