Guys above have provided enough information. I believe stylus
profilometers are easiest way to do the job.

Just to add one more potential way to measure the metal thickness:

I know someone in my lab test the sheet resistance of the metal layer,
and then with the resistivity, you can calculate the thickness. I am
not very sure about the accuracy.

2009/2/17, Albert Henning :
> Just to clarify about optical means:
>
> Ellipsometers and reflectometers measure blanket (unpatterned) films.
> They cannot provide useful thickness or composition information for
> thicknesses above about 50 nm.
>
> Optical interferometers (Wyko, or Zygo) can be used to measure patterned
> films, either single steps (using e.g. the already-suggest Kapton-tape
> approach), or multiple steps/trenches (patterned lithographically, or
> with a shadow mask).  Calibration is essential, and unless you have
> access to a well-maintained system, with a trained operator, it will
> take quite a bit of time to get a useful measurement.  But, the
> technique is sound and successful, whether you're measuring a single
> step in a field of perhaps only a few microns square, or many steps
> simultaneously across an entire wafer.
>
> In my experience, stylus profilometers are faster and simpler, and just
> as accurate, as compared to optical interferometry, if all you want is a
> couple of measurements on a wafer.
>
> Albert K. Henning, PhD