Sputter coating always results in roughened layers. You can try to anneal the sputtered substrate at a few hhndred degrees to obtain a smoother layer with facets of single crystal regions. However, it requires a bit of experience and possibly a few tries to get the result right for you. You can find information in several papers such as

https://www.researchgate.net/publication/225726721_Annealing_of_gold_nanostructures_sputtered_on_glass_substrate

Article Annealing of gold nanostructures sputtered on glass substrate

Did you tried to reduce process pressure and power applied? Disadvantage is that probably you will reduce rate deposition but if this is not an inconvenience for your experiments, it may be one solution.

I think if you could apply uniaxial tension to your substrate during formation of your thin film you may have a chance to get smooth surface according to our recent computer simulation studies. You can get the uniaxial tension by  bending the substrate and film to create convex surface having desired curvature.The application of the electrostatic field along the surface of the thin film might help also to smoothen out the surface undulations through the electromigration at elevated temperatures.

 After the formation of thin film you may give annealing treatment as  Dear Jan PHİLİP commented PROPERLY while decreasing  the bending curvature  steadily to nil.. 

I think annealing the film or growing at higher temperature will reduce the roughness. I hope your substrate during deposition rotates. This rotation gives uniform film.

If you want to use sputter coating, you need make careful control of the glass temperature. If its inappropriate you will get strong roughness and annealing will not help you strongly! Intensity of gold evaporation (temperature dependent) is also influence on than. Good luck

I agree with R.K. Gupta, rotation of the substrate while deposition could be an interesting aspect to increase the homogeneity.And also use a Unbalanced Magnetron.

On my experience with argon sputtering, the best way to reduce the layer roughness is using low energy sputtering parameters. 

On the other hand, you will need more time for your experiments. 

The basic principle is to enhance the surface diffusion of Au adatoms.   Lower the process prssure, higher deposition temp, or ion-assisted may be useful methods.

To get  low coating roughness you may have try few combinations:

increase the surface to substrate distance

reduce sputtering power

reduce sputtering pressure

All these will reduce the deposition rate but will give smooth film.

I hope you are not using any bias to substrate during deposition it can cause roughness. Also the substrate surface matters if it has some inherent roughness.

Thanks all for your kind answers.

Mohsen,

I agree with A. Kale. Longer substrate to target distance, lower power, and higher substrate temperature will help to get smoother films. If you have substrate rotation, please always use it. 

Agree with RK Gupta but also consider underlayers  (Ti, Ta, Cr etc) if possible.

Hey,

you should sputter the 40 nm gold on a thin mics sheet, than glue it on the prism (gold between prism and mica) and peel the mica off afterwards.

Its the so called templating method..

Hegner, M., Wagner, P. & Semenza, G. ULTRALARGE ATOMICALLY FLAT TEMPLATE-STRIPPED AU SURFACES FOR SCANNING PROBE MICROSCOPY. Surface Science 291, 39-46, doi:10.1016/0039-6028(93)91474-4 (1993).

Best regards

Sputtering off centered from the gun can improve the roughness, too (especially if the gun is very close). Also fixing the substrate with at least two clamps on each side is important for the homogeneous temperature distribution which effects the roughness of the deposited film. Everything else has been mentioned already. You probably also know that the roughness increases with film thickness, so if you reduce it, films are going to be less rough...

Two points to consider (i) if your glass substrate is rough (nm variations) perhaps a pre- mechanical polish might help, Or sputter cleaning however you might have problems with uneven electric field potential and space charge. (ii) Supposing we have a very smooth substrate and that the roughness of Au is intrinsic to the deposition process, then I would recommend that you increase the atom diffusivity by  in situ annealing of the substrate at elavated T. One could also perform a post plasma etching on the deposited Au to diminish the roughness of the film.

Dear all.

There are some important correlations that are important and need to be considered in some more detail: Of course the gold diffusion is enhanced for an increased temperature, so increasing the temperature might assumed to be favourable in order to obtain a smoother film. However, depending on the substrate material, the interaction of the deposited gold atoms with the substrate material is weak, for example in the case of float glass, quartz glass, or silicon wafers. Therefore, according to the fundamentals of growth processes, the formation of three-dimensional crystals is energetically favoured because of the thermodynamics (the 3D-structures minimize the total energy because of the strong Au-Au interaction and the low Au-substrate interaction). If the amount of deposited material is small and the temperature high, isolated gold clusters may grow. For example, heating a nominally 20 nm thick layer to about 200°C or more leads to such structures. If the nominal thickness is larger, heating to more elevated temperatures generate the same structure, which of course has a higher surface roughness compared to a film deposited at lower temperature, with less diffusion, but of course smaller individual crystallites. It should also be mentioned that both, the temperature during deposition as well as a post deposition annealing process influence the resulting roughness. Some facts are collected in Surface and Interface Analysis 38(2006) 715 - 718 (see DOI: 10.1002/sia.2258) where we studied the influence of temperature during and after deposition y means of atomic force microscopy and electrical conductivity measurements. Hope you find this useful, cheers, Dirk

The easiest way is to use Zalar rotation, i.e. to rotate the sample while sputtering. If you will combine it with moderate sample heating, the result will be very good.