Since thickness is about 40 nm, I would suggest to go with AFM. Please remember, you need step to measure the thickness. You can put one mask during the growth of the gold film and then do line scan in AFM from bare substrate to gold film through step.

An AFM-based technique will be better since you are dealing with thickness above several nm. Ellipsometry will be difficult. Profilometry might lack the resolution you need, unless you need the cross-section profile over several mm.

Thanks all for your kind answers.

Please see my attached paper, this will give you some idea to measure the thickness using AFM.

Thanks Dr. Gupta

I made a step on my gold layer with lithography and wet etching, then measured the height of the step with AFM. But I am afraid that the gold layer was not removed completely or a layer of acetone, which was used as the resist etchant, remained on sample surface. So my measurements wont be accurate.

In another way, I used a blade to remove gold layer and I think it worked and I had a sharp step. But I doubt whether these methods are reliable or I should think of another methods?

What kind of mask should I use, if I want to create a sharp step during the gold layer sputtering?

I thought of putting a blade on parts of the substrate which I do not want to be sputtered, during the growth. but this method may not create a sharp step.

You can put some kind of sticky tape (3M) during growth and you can take out after deposition. This will give you sharp edge. If edge is not sharp, please so the scan for longer length and take the difference from base to step (not near the edge, but away from the edge).

Good idea, but the residue left by the sticky tape will reduce the accuracy.

You can buy the residual free tapes from 3M company.

Thanks so much Dr. Gupta.

Sharpness of the  step is necessary however one could try to measure barring some area near step during analysis by masking. Again another problem is uniformity of the fim. I get request for measuring film thickness, but often i face problems in quoting a value because of the fact that nonuniformity becomes comparable to such thin films. If you make a step you can use optical profiler / stylus profilers also

I think you use any technique, but if the film is so thin (< 20 nm) ellipsometer is perfect. Otherwise, profiler is the best once the step height is sharp as Rina mentioned