From chemistry-request@server.ccl.net Mon Oct 22 23:25:14 2001 Received: from relay.it.northwestern.edu ([129.105.16.56]) by server.ccl.net (8.11.6/8.11.0) with ESMTP id f9N3PEB22720 for ; Mon, 22 Oct 2001 23:25:14 -0400 Received: (from mailnull@localhost) by relay.it.northwestern.edu (8.9.3/8.9.3) id WAA02266; Mon, 22 Oct 2001 22:28:05 -0500 (CDT) Received: from [129.105.186.87] (quandong-67-186087.nuts.nwu.edu [129.105.186.87]) by relay.acns.nwu.edu via smap (V2.0) id xma001889; Mon, 22 Oct 01 22:27:54 -0500 Mime-Version: 1.0 X-Sender: hutchisn@mercury.chem.northwestern.edu (Unverified) Message-Id: In-Reply-To: <157A51F55AAAD3119CD70008C7B1629D01C15666@lvlxch01.unitedcatalysts.com> References: <157A51F55AAAD3119CD70008C7B1629D01C15666@lvlxch01.unitedcatalysts.com> Date: Mon, 22 Oct 2001 22:27:44 -0500 To: "Shobe, Dave" From: Geoff Hutchison Subject: Re: CCL:Is TD or Zindo better than CIS? Cc: "'CCL'" Content-Type: text/plain; charset="us-ascii" ; format="flowed" At 1:25 PM -0400 10/22/01, Shobe, Dave wrote: >spectrum. Gaussian has two other methods (neither of which I've used) for >calculating excited states: Zindo and TD. Are these significatnly better >than CIS? Zindo being a semiempirical method, is it only good for certain >kinds of molecules? Should I be looking at programs other than Gaussian for >this type of calculation? It's strange this should come up as I'm working on a manuscript along these lines for a certain set of molecules. But first there's a nomenclature issue. Normally when people say "Zindo," they mean "INDO/S CIS," and so you can't compare "Zindo" to "CIS." Also, I'm not quite sure I know what you mean when you say "TD" since you usually refer to TD-HF or TD-DFT for time-dependent HF or time dependent DFT. If you compare Zindo CIS to HF-CIS for most neutral organic molecules, you'll find Zindo is more accurate. If you compare TD-HF (say the RPA) to HF-CIS, the HF-RPA is better--no surprise since the RPA is one step past the Tamm-Dancoff approximation (CIS). But HF-RPA is not as good as Zindo for most neutral organic molecules. TD-DFT seems to be slightly worse than Zindo for neutral organic molecules, but not by much. Also keep in mind that Zindo CIS as a semiempirical method sometimes incorrectly identifies the composition of certain excitations even if the energy is pretty good. Can't say how Gaussian does on these sorts of things--we can't get it here. As Mark Thompson pointed out, you can also go beyond CIS to CISD, etc. The original ZINDO program from Zerner does let you do things like this, though YMMV. Other programs as well have CISD, CISDT and other higher-order methods. (Dalton springs to mind.) -- -- -Geoff Hutchison Ratner/Marks Groups (847) 491-3295 Northwestern Chemistry http://www.chem.nwu.edu/