 |
Figure 1. Typical reflectance spectrum. |
Harland
G. Tompkins, Jeffrey H. Baker, Steven Smith, Diana Convey
Motorola Labs,
2100 E. Elliot Rd.
Tempe AZ 85284
We discuss how analytical tools in a characterization lab can be used to enhance metrology tools in a fab. The emphasis is on the interaction between a characterization lab and a wafer fab belonging to the same industrial company (as opposed to labs in academia, government labs, original equipment manufacturers, or the manufacturers of optical analytical or metrology tools). Specifically this work will deal with ex situ analysis rather than in situ metrology.
A metrology tool in a fab must be clean, fast, simple to operate, and must be able to measure small features. A classical example is a reflectometry tool.1 On the other hand, a tool in a characterization lab can handle a much more complex sample, can take a significantly longer time for the analysis, and often has the luxury of using a blanket wafer. The analyst is also available for consulting with regard to optimum samples configuration in order to generate the information required. Our emphasis will be on using spectroscopic ellipsometry in a characterization lab to develop optical constants of unusual material so that this information can be used in a reflectometry tool in the fab.
|
Figure 1. Typical reflectance spectrum. |
As we begin, let us consider the reflectometry tool in its mode of operation where the lab interaction is not needed. For this example, shown in Figure 1, we consider the measurement of the thickness of well known materials such as thermal oxide, LPCVD nitride, or photoresist on a silicon wafer. In this case, the spectrum has lots of structure, the index of refraction is well understood and the extinction coefficient is zero. There are no surprises here and the reflectometry tool must simply determine the thickness. No involvement from outside the fab is required. One simply pushes the button and gets the answer.
 |
Figure 2. Optical constant spectra of a silicon-rich nitride. |
The next example, shown in Figure 2, is a situation where a small amount of interaction with a characterization lab is needed. Consider silicon-rich nitride. This might be deposited by either PECVD or LPCVD and the excess silicon is to be added either to control stress properties or to make an anti-reflective coating. These are not fixed, as is the case for thermal oxide, but depend on how the material is fabricated. The function of the characterization lab is to determine the optical constants, usually with a variable-angle spectroscopic ellipsometer. A table of the optical constants are then generated and fed into the reflectometry tool in the fab. At that point, as long as the material is fabricated in the same manner, the metrology tool can be used in its normal mode, i.e. fast, simple. We would expect that the characterization lab would model the film using a dispersion equation such as the Cauchy equation, or with the slightly more complex Tauc-Lorentz equation.2
For the next example, we consider a more complex issue. Optical techniques are not usually used to measure metal films. Historically, metal films have been thick enough to be opaque and if the light does not reach the lower interface, having more material does not change the measured quantity. Metal films with thicknesses less than 300Å are used in several applications, however, and in this thickness range, the films are not totally opaque and can be measured with an analytical or metrology tool. Examples of the use of thin metal films are for MRAM,3 Flat Panel Displays, Advanced Lithography, etc.
 |
Figure 3. Ellipsometric parameters for several thin Cr films on Si. |
The optical constants of metals depend on the fabrication methods; hence, one must determine the optical constants for the methods used rather than to use handbook values. The dispersion relationship for metal films often cannot be described easily with an equation and the values must be determined for each wavelength used. The difficulty is that for a single film, the problem is underdetermined and there are many combinations of thickness and optical constants with will give equally good fits to the measured data. One of the most powerful analysis methods in ellipsometry is to have several films of the same material, differing only in thickness. We asked the fabricator to deposit four films with target thicknesses in the range of 50 - 300Å. The ellipsometric spectra and resulting optical constants are shown in Figure 3 and 4 respectively. Calculated values are also shown for zero thickness and opaque films.
 |
Fig. 4. Optical constants determined from data in Fig. 3 |
Following the analysis, the optical constants are formatted for the metrology tool and then used in the reflectometry mode. Note that whereas a dielectric film with thickness less than 100Å does not produce a response significantly different from no film, a metal film a few tens of angstroms will change the reflectometry response significantly.
 |
Fig. 5. Optical constants for amorphous carbon, after anneal. |
As a final example, we show a possibility for quality control in a fab. Let us consider an amorphous carbon film deposited by PECVD with methane, a small amount of nitrogen, and argon. This material has very little hydrogen and no crystalline structure. As deposited, the carbon has primarily sp3 bonding. When it is annealed, some of the bonding changes to sp2. The resulting optical constants are shown in Figure 5 after annealing at temperatures ranging from 400 - 1000°C. The material was modeled using the Cauchy/Urbach dispersion equations. It is expected that in the fab, this measurement would be made with a spectroscopic ellipsometer with a fixed number of wavelengths using a CCD array. The function of the characterization lab would be to provide the seed parameters for a Cauchy analysis. The optical constants determined in the fab would be used for quality control of the annealed material.
1. “Spectroscopic Ellipsometry and Reflectometry: A User’s Guide”, H. G. Tompkins and W. A. McGahan, John Wiley & Sons, New York (1999).
2. G. E. Jellison, Jr. and F. A. Modine, Appl. Phys. Lett., 69, 371 (1966), Erratum, Appl. Pys. Lett, 69, 2137 (1996).
3. H. G. Tompkins, T. Zhu, and E. Chen, J. Vac. Sci. & Technol. A, 16, 1297 (1998).