Question about "Whether Surface Slab thickness need to be larger than Rcut"

[roger13231]

Dear Developer and colleague,

I have a question about Rcut and surface model thickness selection during surface potential training.

A reviewer had an opinion on an article that my senior participated in that, physically speaking, if the thickness of the interface slab is less than Rcut, then the interaction of atoms at the top surface will be affected by the vacuum above and a small portion of the vacuum layer below. This "small portion of the vacuum layer below" could cause a slight different influence on predicting the interaction of those top surface atoms.

I am wondering what this suggestion sould like or whether I could continue use my current training data, which slab is smaller than Rcut. I don't physcially understand this influence and what the difference on considering this small portion of vaccum or not.
If the thickness needs to be increased, it seems like it would require many layers, possibly up to 7A. I see some current colleague in DP discussion group mention they use 4 or 5 metal layer, which is maybe 4-6 A.