CCL: Advice on sterol/steroid QC calculations wanted.



 Sent to CCL by: "David Anthony Mannock" [dmannock[#]ualberta.ca]
 Greetings fellow molecular chemists! We have recently completed a series of
 experiments on a range of 30+ sterol structures in model lipid bilayers using
 calorimetric and spectroscopic techniques with considerable help from a student.
 I have already performed a number of DFT calculations using the opt/freq option
 in G09 and B3LYP/6.31 or 311++ (d,p). [I hope I got that right!] This has
 generated a number of conformations for each sterol/steroid structure in the
 ground state and I am not unhappy with the results. However, I need advice on a
 number of issues. [I do have the Handbook of computational chemistry (Cook) and
 Electronic Structure Modeling (Shillady)for back up].
 1) Is this approach sufficient to examine the electronic structures of the
 various sterol/steroid conformations that are evident? Would similar
 calculations with AIM be beneficial either in providing more accurate/realistic
 orbital energies or would a Hirschfeld type of analysis be more appropriate?
 2) I have seen various messages here on conformations in excited states and
 energies. What is a good approach in such calculations in G09?
 3) Sterols exhibit definite anomeric electronic effects that interact with
 nearby orbitals originating from double bonds. There are also possible inter-
 and intra-molecular CH-pi interactions present in some molecules. What is the
 best calculation solution?
 4) Similarly in some of the steroids, resonance effects exist. I am not clear
 whether these would be covered in calculations of the excited states.
 5) With 8-cores some calculations take a week or more. Upgrading to 12 and now
 64-cores helps for those subroutines that can run in parallel in G09, but I am
 wondering if there is a way to break down this extensive computation into parts
 to decrease the computation time? I am thinking that I could perform some
 calculations on my 12-core server and others on the 64-core machine. I am
 thinking that this will be possible by using a line such as "(guess =
 read)" to access a checkpoint file.
 6) I have an Nvidia M2075 card and I know there are mp software solutions which
 use CUDA that are free for academic use. Any suggestions would be welcome, as I
 am not familiar with using that software.
 Finally, I should add that while I have learned a lot, I know just enough to be
 dangerous! Guidance from experts would be most helpful in getting this done
 correctly.
 Thanks in advance, Dr David Mannock
 dmannock*o*ualberta.ca