characterising surfaces

[baffledMatt]

Does anybody know any neat measurements one can do to characterise a surface? I'm thinking generally of rough surfaces such as the ones you come across in epitaxial systems.

In other words:
If I have two ensembles of surfaces, what is the best method to use to determine whether the same growth process was used for each ensemble?

The ideas I have so far are:
Power spectrum
Wavelet analysis

but these can be unreliable as it is quite easy to make two different processes create very similar looking power spectra, etc. This makes life especially difficult with experimental data because you generally don't have the statistics you would like.

Also, does anybody know anything about what resampling methods exist for this type of scenario? I have seen people who scan lines across the surface and take separate measurements for each one, then do exactly the same thing again but with the lines going at right angles to the first ones. Due to the huge amount of correlation, this method seems dodgy as hell to me.

Matt

[Gokul43201]

I'm not sure exactly what you are trying to achieve here but you could look at the following :

auger spectroscopy(AES), STM, XPS, SIMS - these are commonly used for surface characterization

[baffledMatt]

Thanks, but I'm more looking for theoretical techniques to analyse my data.

I have measurements of some rough surfaces in the form h(x,y), which is the height at a given (x,y) coordinate (x,y is in the plane of the surface) and I want to compare them to similar surfaces generated on a computer.

Matt

[Simon666]

Mandelbrot used fractals to characterize roughness. Don't ask me how that one works.

[baffledMatt]

Mandelbrot used fractals to characterize roughness. Don't ask me how that one works.

Yeah, I discussed this problem with him last week and he thinks that the best thing is to use Hurst analysis. I'm just not so sure because this sort of thing amounts to trying to decide if one straight line on a log-log plot is the same as another straight line on a log-log plot (which is how we determine the Hurst exponent) It's all just a little unsatisfying really.

Matt