In-Situ Optical Metrology of Polycrystalline Silicon

Accurate characterization of polycrystalline silicon (polysilicon) is not only a critical monitoring technique for chemical vapor deposition (CVD) process control, but also a necessity for gate line-width control in i-line/DUV lithography and dry etching. It requires that the thickness, refractive index and extinction coefficients (from DUV to near red) in polysilicon films be precisely and rapidly determined. However, crystallinity, grain size and void fraction all have an effect on the optical constants. So optical constants and films thickness must be measured simultaneously. This paper presents a versatile reflectometry method for determination of the polysilicon optical constants as well as its thickness. With aid of sophisticated dispersion equations, a simulated annealing algorithm is used to solve the global optimization problem. It is expected that this methodology will greatly facilitate in-situ thin-film characterization, which is essential to supervisory lithography control. 1.0 Introduction Polysilicon characterization is an important metrology issue in semiconductor manufacturing. To continually reduce the variability of gate critical dimension (CD) over larger chip areas, wafer to wafer, lot to lot, we need to know the optical properties of the poly layer. However, polysilicon is considerably variable in its structure, which consists of crystalline silicon, amorphous silicon and voids. The proportion of each component depends on the process condition of polysilicon. These changes in polysilicon are reflected in its refractive index values. In-situ monitors can provide the optical information fast and accurately. The information can be fed into lithography tool controllers, models or simulators in real-time, so that complex processes can be controlled effectively to reduce the CD variation. It is well-known that the optical constants of polysilicon are strong functions of both the deposition and annealing conditions [4]. For example, the refractive index of polysilicon which is deposited temperature over the range 545 to 605 °C can change 30%. To get accurate characterization, refractive index, extinction coefficients and film thickness must be measured at the same time. Optical film metrology that includes ellipsometry, reflectometry and interferometry, is very attractive for the multiple-layer thin-film characterization. Because of its experimental simplicity, normal incidence reflectometry is often integrated into the real-time process control paradigm [5]. The reflectometry is based on the Kramers-Kronig relation or on curve fitting with the aid of dispersion equations. The fact that reflectance is measured over a broad spectral region reduces the likelihood of multiple solutions, which is usually one of the most challenging task in reflectometry. Usually the number of variable parameters of the dispersion model increases dramatically with the sophistication of the model. The nonlinearity of the dispersion model lead to local minima in the reflectance curve fitting optimization. Traditional optimization algorithms are not appropriate here, because they need many starting points. In this paper, first we briefly discuss the different dispersion equations. Sophisticated dispersion models derived from Kramers-Kronig analysis are used over wide wavelength range. Then we describe the