There's no way to do this. Once you define a periodic system, Gaussian will always treat everything as part of that periodic system.

And if you think about this, that makes sense: the whole idea of an PBC calculation is that you only have to define a single cell of the periodic system and the calculation results apply to the whole, technically infinite system.

Once you add a modification to a single unit cell, it's no longer really a periodic system.

What's usually done in situations like yours is that you prepare a large part of your normally peridioc system (in your case your graphene) as basically single molecule as model. You then place the metal atoms (or single molecules) at the center of this and run your calculation.

Dear Daniil,

As Simon said, it is not possible. There are two alternatives:

-The use of a "cluster model". The idea is to use a huge system such its electronic structure is almost the same than the one of the periodic system (border effects and quantization due to finite size can be neglected). Then, you study whatever you are interested in in the center of your system. See e.g. (also an example with graphene):

Article Exploring the catalytic efficiency of X-doped (X=B, N, P) gr...

-There is a different approach, the use of a "super cell". In this case you use a periodic system, but again very big and your adsorbed molecule, defect... Is also periodic. Use this approach if you can't neglect border effects but you can neglect interaction between the defects/whatever. See e.g.

Article Theoretical Insights Into Vinyl Derivatives Adsorption on a ...

If you want to use Gaussian, I would recommend you to use the first approach.

Hope it helps,

Fernando