In Gaussian 09, I think DFT b3lyp/6-31+g(d) is a good selection.

But to be honest, Gaussian software package is too old and too slow, VASP code package is better.

Thanks Lei.

But I think 6-31+g(d) is not sufficient for optical properties, in case of generally for metals and transition metals. And I have only gaussian.

Thanks Bruhn Sir. I 'll try wit triple-zeta basis sets..

def2 basis sets are good enough.

and the functionals you can try cam-b3, m06 series, and wb97xd, etc.

Thank You Fu-Quan Sir...

Have a look at the paper below. 

Article Choosing a density functional for static molecular polarizabilities

 Thank you Ajit Sir. Thank you very much..

Dear Smruti Sahoo

You can go through this article through the link below, maybe it can help.

http://dx.doi.org/10.1063/1.442552

 Thank You Jacob...