If I will modify the silica surface and a small molecule gets adsorbed on the surface, then how much shift in wavenumber (peak) of characteristics peak of the molecule is acceptable before and after adsorption?
If there is a change in the peak before and after adsorption process then you can explain the adsorption property of the synthesized particle and after the regeneration process compare if the peak matched with the peak (synthesized particle before adsorption) then you can predict the stability of the synthesized particle/adsorbent.
I am not acquainted with the particular material you are using as an adsorbent. But, generally for adsorption or similar surface-chemical properties, Raman peak-shift should not be more than 15 cm-1.
Besides in most Raman instruments, shifts smaller than 2 cm-1 are difficult to detect. So, you can consider this range.
However, if doping, composite or other processes are involved at the bulk level, that can modify the nature of lattice vibrations, then higher shifts are possible. In fact, new peaks may also appear, if the associated selection rules are broken or modified.
If you are examining the silica related Raman lines and if the silica filler diameter is greater than 1 micrometer, you should not observe any Raman shift. You simply do not have interface related sensitivity. Only if you are using a nanometer-level high surface area silica, and if your analysis is very careful, you might be able to observe small amount of shift, though I doubt it. Silica is usually not sensitive in Raman due to the high polarity. FTIR is much more sensitive in siliceous materials. There, you can observe surface silanol groups again if you use high surface area silica with the surface area greater than few tens of m2/g.