Re: CCL:UV spectrum calculation (new question)

I'm curious how well INDO/S-RPA works against INDO/S-CIS for large systems. 
 I only remember one article by M. Zerner quite a long time ago..
A comparison of ZINDO/RPA vs. ZINDO/CIS on various conjugated oligomers is included as part of my larger paper (which also compares HF/CIS, HF/RPA, TDDFT/CIS and TDDFT/RPA): 
 J. Phys. Chem. A (2002) 106(44) p. 10596-10605.
At the time, the only ZINDO/RPA I could put my hands on was contributed by Sergei Tretiak and Shaul Mukamel, which didn't let me set an active space. (In fact, it uses the Lanczos method for a full active space.) A parameterization for RPA calculations or a smaller active space would probably help.
Probably the calculation takes a lot longer (Hamiltonian matrix is four times as big since mixing of particle-hole and hole-particle excitations occurs).
I think with an efficient implementation, it's not really an issue. (It's a semiempirical calc. after all!) I never did timing tests, but I can relate that up to 60+ heavy atoms, ZINDO/RPA calculations still only took something on the order of minutes (on an 800MHz Pentium III, so faster times nowadays). 
 
 Are there any problems with numerical instabilities with the
 implemented RPA scheme in a INDO/S-RPA calculation ?
I'll second Dr. Thompson's comments. I haven't run any extremely "strange" molecules, but haven't seen any instabilities. 
 Cheers,
 -Geoff
 --
 -Geoff Hutchison		<hutchisn=at=chem.northwestern.edu>
 Ratner/Marks Groups	(847) 491-3295
 Northwestern Chemistry	http://www.chem.northwestern.edu/