High-accuracy, high-speed measurement of deep submicron and nano-structure gratings using specular reflected light techniques.

Abstract

Specular reflected light techniques have been proven to be successful for monitoring vacuum processes on unpatterned substrates. However, applications of these methods in semiconductor manufacturing have been limited by the diffraction problem of periodic patterns on the wafers and the sensitivity of measurement techniques. In this dissertation, feasible solutions for monitoring deep sub-micron and nano-structure surface relief gratings by these techniques coupled with a vector diffraction theory (using RCWA) are presented and demonstrated. The RCWA method has been implemented with efficient numerical techniques. The numerical stability has been investigated. A nonlinear regression procedure has been introduced to extract pattern profiles from the optical measurements and to statistically study the measurement sensitivity. Applying this technique to reactive ion etch (RIE) monitoring, we have reported the first <italic>in situ</italic> monitoring of deep submicron topography evolution using real-time two-channel spectroscopic reflectometry (2CSR) measurement coupled with off-line RCWA modeling. The grating samples have a period of 700nm. The profile parameters, critical dimension, feature height, and sidewall angle, have been obtained with the average 95.4% confidence limits of 1.8nm, 1.8nm, and 0.32&deg;, respectively. The narrowest measured top linewidth is 76nm. To accurately monitor the submicron topography evolution in real-time, we have introduced a non-linear filtering (NLF) algorithm, which on-line estimates the grating shapes based on pre-generated RCWA simulations. Using 2CSR, RCWA, and NLF, we have reported the first real-time <italic>in situ</italic> monitoring and endpoint detection of deep submicron structures during the RIE process. A follow-up experiment has exploited this technology to a real application level using a real-time SE on an industrial production chamber. We propose that a normal-incidence (NI) SE can be used for high-accuracy topography measurements. Experimental measurements of 350nm linewidth structures and theoretical simulations of 10nm linewidth structures show that SE at near NI is capable of grating measurements in extreme submicron or nanometer regime. We have also demonstrated that the novel near-NI approach improves the ability to separately extract topography parameters more than the traditional off-normal SE. Finally, I demonstrate the importance of deep UV light for detailed shape modeling. All experimental measurements are in strong agreement with cross-sectional SEM photos.