I want to compare the electronic properties (HOMO-LUMO, etc.) of an aromatic heterocycle in vacuum and in solvent at DFT level with PCM model (using G09 package). Do you think that this kind of comparison makes sense?
Yes, it makes sense... In PCM model you will adjust the environment dieletric constant to you system! Just try to be carefull with your conclusions, because you can get results that are dependent on the model (in this case, the PCM) For some systems and depending on what you want to describe, the PCM will be enough! But in some cases, you will have to explicitly put some solvent molecules! In that case, you cannot compare molecular orbitals of the free solute molecule with the solute plus explicit solvent molecules plus PCM!
As Renaldo wrote, yes, this is entirely reasonable. See, for example, a nice little article about solvatochromism - how solvent changes the optical spectrum of molecules
I am more familiar with COMSO than PCM, but they should yield comparable results. Validate the compution by comparing to experiment. If the results are not as good as you'd like, you may need to add a solvent shell or 2 of explicit solvent molecules, then soak the enire thing in PCM (as Renaldo wrote).
Thank you for all your valuable responses. I have experimental IR spectrum of solvated sample. After a detailed inspection I have decided that the best model for me is "PCM + explicit solvent molecules". Now, I'm planning to investigate theoretical electronic properties by that way.
As is well known, except for the HOMO (which describes the true HOMO *provided that* the exchange-correlation is exact), all other Kohn-Sham "orbitals" (KS-MOs) are mathematical objects rather than true molecular orbitals. This is the case even in vacuo, and, as it should be obviously, the presence of a solvent does add new complications. To see that KS-MOs in solvents are both qualitatively and quantitatively inadequate, you may visit http://dx.doi.org/10.1016/j.elecom.2013.08.027, where a workaround is indicated.
Hm, I just wonder what is a "true molecular orbital"? An orbital is usually the eigenfunction of some one-particle operator, e.g. effective ones like a Fock operator, a KS operator, a Hartree-type operator, or even a density operator, etc.. Don't forget the spin orbitals :). In chemistry, we are used to LCAO-type molecular orbitals build from atom-centered basis functions but one may use other types of basis function, eg. plane waves, as is still quite often used in molecular and solid state physics groups.
But one must not forget that and how the MO coefficients are calculated. Every orbital is a mathematical object stemming from some theoretical model of nature. Depending on that model and its relevance in describing measurable facts, the mathematical object "orbital" may or may not be useful or adequate.
The modeling of the effects of a solvent is always a little tricky if you do it by some coupling to an "external continuum", since one has to balance the description of effective one- and more-particle interactions. You also have to avoid "double counting" of effects, e.g., when you additionally use some solvent molecules in the calculations.
So, it is not really surprising me that concepts that have been developed for a pure ab initio context, and mostly on the basis of the Born-Oppenheimer approximation, are not easily transferable to models for molecules in solvents.