Gaussian itself doesn't have any way of computing sigma-profiles on its own. As you noticed, it has the scrf=COSMORS
keyword, but this just produces a file that is used by a dedicated program (e.g. COSMO-RS, COSMOtherm).
On the Gaussian side, it really is as straightforward as using that keyword. Here is a minimal example:
#p B88-VWN-P86/TZVP scrf=cosmors
peroxide
0 1
O -0.22216884 -0.6214829 -0.08709297
H 0.09625091 -1.07514452 0.69674375
O 0.22216884 0.6214829 -0.08709297
H -0.09625091 1.07514452 0.69674375
h2o2.cosmo
This produces the necessary .cosmo file. Note the choice of functional and basis set is important, as COSMOtherm only ships parameter sets for certain combinations. In this case, you need to be sure the geometry is converged at this level of theory to use the BP_TZVP_C30_1701.ctd parameter file. If you can't converge it at that level of theory, you can use the SVP basis set and a different parameter file (BP_SVP_AM1_C30_1701.ctd).
I don't have access to either of the dedicated COSMO programs, but at least the input for COSMOtherm seems pretty straightforward. In an input file (name.inp), you would at minimum write:
ctd=name.ctd
#Identifier
f=name.cosmo
This would be run using cosmotherm name.inp
from the commandline.
SCRF(COSMORS)
(gaussian.com/scrf/?tabid=2) will produce some kind of interface file. Then the manual of Cosmotherm details how to use these files (pdf). Related also: ccl.net/chemistry/resources/messages/2006/02/21.005-dir/… $\endgroup$