Hi,

the problem with ellipsometry is that it is an indirect measurement technique. In general, you cannot directly deduce the optical constants and the film thickness from the measured Psi and Delta data. To overcome this problem, you have to create an dispersion model, consisting of one or more oscillator functions, which is able to describe your data in a sufficient way. The important thing is that your model has to be physically resonable, i.e., your oscillator functions must be allocatable to optical transistions such as the band-band transition. These model functions typically consist of many paramters which have to be fitted to your experimental data. To make the fit more reliable, several ellipsometry spectra should be acquired under different angles of incidence. Your model should be able to describe all these data appropriately.  

However, there are some tricks to make life more easy. For example, you can analyze your data in a wavelength region where your sample is almost transparent (k=0). Hence, you have only two free paramters in that case, the thickness (d) and the real part of the refractive index (n). You can then apply a simple Cauchy dispersion model to your data, which will give you reasonable results for n and d. However, in "real world", things are often more complicated and surface roughness or anistropy of your film can play a role, which makes the fitting procedure much more difficult.

Some tutorials about spectral ellipsometry can be found here:

http://www.jawoollam.com/tutorial_1.html

http://grover.mirc.gatech.edu/data/gatech_ellipsometry_seminar-dec_2010.pdf

Hi,

the problem with ellipsometry is that it is an indirect measurement technique. In general, you cannot directly deduce the optical constants and the film thickness from the measured Psi and Delta data. To overcome this problem, you have to create an dispersion model, consisting of one or more oscillator functions, which is able to describe your data in a sufficient way. The important thing is that your model has to be physically resonable, i.e., your oscillator functions must be allocatable to optical transistions such as the band-band transition. These model functions typically consist of many paramters which have to be fitted to your experimental data. To make the fit more reliable, several ellipsometry spectra should be acquired under different angles of incidence. Your model should be able to describe all these data appropriately.  

However, there are some tricks to make life more easy. For example, you can analyze your data in a wavelength region where your sample is almost transparent (k=0). Hence, you have only two free paramters in that case, the thickness (d) and the real part of the refractive index (n). You can then apply a simple Cauchy dispersion model to your data, which will give you reasonable results for n and d. However, in "real world", things are often more complicated and surface roughness or anistropy of your film can play a role, which makes the fitting procedure much more difficult.

Some tutorials about spectral ellipsometry can be found here:

http://www.jawoollam.com/tutorial_1.html

http://grover.mirc.gatech.edu/data/gatech_ellipsometry_seminar-dec_2010.pdf

Thank you Sebastian, what a great answer!  

I will check out the tutorials.  Do you know of any really good tools for constructing the material layers and modeling for ellispometry?  Do people use tools that come with ellipsometry machines?  Are the good tools free or commercial?  I actually did read the tutorial in the first link that you sent me also, but your explanation cleared it up a lot.

Hi Adam,

It's a pleasure for we that I could help you a bit!

I think most people use the software that comes along with the ellipsometry. One well-known manufacturer is J. A. Woollam. We're also running a Woollam machine in our labs, together with the software WVASE32. I think the software is quite good and very common.

Kind regards,

Sebastian

Is that software free or did you purchase it?  

It came along with the machine. I think you have to pay for it:

http://www.jawoollam.com/WVASE.html

Hi Adam

There are some free software that extract optical parameters of single/multi thin film like n, k, d from T, R, or both. PUMA is one of them. To run the program, you need C compiler to get the EXE file. You may check this site, it has all documents you need.

http://www.ime.usp.br/~egbirgin/puma/#license

I hope it helps, Best wishes

As previously mentioned J.A. Wollam can be very helpful if you contact them.