I would suggest X-ray reflectivity measurements, because you are not only able to measure the outermost surface, but also buried layers, multilayers etc may be analyzed. Also diffuse X-ray scattering may help, because in a non-specular geometry, the intensity of the reflected X-rays also contain information about the lateral structure (correlation length) and vertical correlations between different layers.
See e.g. the following articles
V. Holy, J. Kubena, I. Ohlidal, K. Lischka, W. Plotz, Phys. Rev. B 47 (1993) 15896
V. Holy, U. Pietsch, T. Baumbach, Springer Tracts in Modern Physics, vol. 149, Springer, Berlin, 1999
S.K. Sinha, E.B. Sirota, S. Garoff, H.B. Stanley, Phys. Rev. B 38 (1988) 2297
M. Tolan, Springer Tracts in Modern Physics, vol. 148, Springer, Berlin, 1999.
L.G. Parratt, Phys. Rev. 95 (1954) 359.
S.M. Heald, H. Chen, J.M. Tranquada, Phys. Rev. B 38 (1988) 1016.
@ Mohammedreza: For (spectroscopic) Ellipsometry you need a contrast between the different layers involved - i.e. different optical constants in the investigated spectral range. This is sometimes not given, and the investigated films must be (to some extent) optically transparent, which causes problems especially in the case of metallic layers.
So it depends on the sample system whether or not ellipsometry is the tool of choice!