You have to specify solvent in Gaussian, and you have a number of options to do that:
- use: SCRF(Solvent=XXX) and as XXX you put the the name of your solvent. Gaussian recognized some of the most popular solvents just by name (like water, methanol and so on): see their website for the full list. This method works for all SCRF methods.
- use: SCRF(Solvent=generic) and then after the geometry specification make a blank line and then define the solvent. You define the solvent by providing at least four parameters: eps (static dielectric constant), epsinf (dynamic dieletric constant), rsolv (radium of the solvent) and density (density, in particles/anngtrom^3). You have to look them up in some papers. Beware, that if you don't specify them Gaussian will take these parameters from water. This method works for all PCM methods.
- use: SCRF(dipole, a0=XXX, dielectric=YYY). You define the solvent by specifying its radius (XXX) and dielectric constant (YYY). This works only for the simple Onsager (Dipole) method.
The first step is to first optimize your geometry in vaccu (no solvation model just opt freq using any chemistry model you choose) and then run your CPCM calculations on that structure using the same chemistry model you used for your optimizations.
The keyword for single point calculation with CPCM solvation model is:
scrf=cpcm
use it in command line. you can also add sp or freq as well (not necessary).
Note that Gaussian by default assumes that the solvent is highly polar and the cavity surface is conductor for CPCM model ( the limit is close to water dielectric constant which is around 80).
you dont have to include a radius in your calculations unless you want to work with a solvent very different than water. Then you should specify its properties otherwise the Gaussian uses the values for water as Bartosz just mensioned.
Make sure that the solvent you are working with is polar, otherwise CPCM is not suitable.