CCL: W:GTO and STO

 Sent to CCL by: "Ahmed" [ahmed.bouferguene{:}ualberta.ca]
 Hi all,
 Thanks a lot for the links you've provided. In fact, ADF is a density
 functional code and in this case it is not strange to have high efficient
 codes since the most difficult integrals needed for such code are the so
 called three-center nuclear attraction integrals (these are one-electron
 integrals).
 However, it is in the ab-initio that the damage is visible since in this
 case one needs ALL the integrals (including the infamous 4-center 2 electron
 integrals).
 Cheers
 -----Original Message-----
 > From: owner-chemistry{:}ccl.net [mailto:owner-chemistry{:}ccl.net]
 Sent: Thursday, September 08, 2005 2:51 AM
 To: Bouferguene, Ahmed
 Subject: CCL: W:GTO and STO
 Sent to CCL by: Stan van Gisbergen [vangisbergen~~scm.com]
 Dear Dr. McKelvey and others interested in STO code performance,
 Thank you for your compliments regarding ADF's capabilities.
 As there seems to be some doubt regarding the efficiency of STO-based
 codes,
 such as ADF, I would like to give one timing example to make the
 discussion less theoretical.
 An 8-CPU job on a Linux cluster
 (http://www.sara.nl/userinfo/lisa/description/index.html)
 for a geometry step at the GGA level of a Pt-complex with 105 atoms in
 a DZP basis set
 (1021 STO's) takes 17.5 minutes elapsed time. My conclusion is that
 this is fast enough
 to do a lot of useful chemistry with ADF during one coffee break. I can
 provide the input
 file upon request so you can check if other DFT codes are similarly
 efficient.
 Further information on the efficiency of ADF is available in our
 brochure
 (http://www.scm.com/SCMForms/BrochureRequest.jsp)
 and can be tested through a free trial.
 Best regards,
 Stan van Gisbergen,
 Scientific Computing & Modelling (www.scm.com), provider of ADF
 On Sep 7, 2005, at 4:17 PM, CCL wrote:
 >
 > Sent to CCL by: John McKelvey [jmmckel;;attglobal.net]
 > All,
 >
 > I have not seen any timing benchmarks of ADF and Slater functions
 > versus a gaussian based code.  With intent to compare accuracy I
 > suppose one could start by comparing a D-Z Slater calculation in ADF
 > with an equivalent D-Z/STO-6G basis in a gaussian based code. The cusp
 > conditions and hence energies would be off some, but I would guess
 > that  geometries might be very similar....  Not sure of codes that
 > have STO-NG for d functions..
 > I have to say that after sitting through the ADF seminar at the recent
 > ACS meeting I am quite impressed with it's capabilities.   CPU times
 > may not be the only issue; after all understanding chemistry is the
 > main point. :-)
 >
 > Best regards,
 >
 > John McKelvey
 >
 >
 > CCL wrote:
 >
 >> Sent to CCL by: Laurence Cuffe [Laurence.Cuffe^ucd.ie]
 >>
 >>
 >> ----- Original Message -----
 >>
 >>> From: CCL <owner-chemistry{:}ccl.net>
 >>>
 >> Date: Tuesday, September 6, 2005 6:31 pm
 >> Subject: CCL: W:GTO and STO
 >>
 >>
 >>> Sent to CCL by: Serguei Patchkovskii [ps:+:ned.sims.nrc.ca]
 >>>
 >>>> Sent to CCL by: Laurence Cuffe [Laurence.Cuffe]*[ucd.ie]
 >>>> The short answer is that calculating the overlap between two
 >>>> Gaussiantype -Orbitals can be done in closed form. That is
 given two
 >>>> GTO's A and B you can write an relatively simple algebraic
 >>>> expression
 >>>> for the size of the overlap between them.  This is not possible
 with
 >>>> Slater type orbitals.
 >>>>
 >>> This statement, of course, is false. Closed-form expressions for
 >>> overlap integrals in terms of exponential integral-type functions
 >>> are very well known. For example:?(cut)
 >>> Dr. Serguei Patchkovskii
 >>>
 >> A fair point Dr Patchkovskii. In hindsight I should, perhaps, have put
 >> more emphasis on ?relatively simple?  As Ahmed. Bouferguene wrote:
 >> The "flip side of the coin" is, multi-center integrals (which
 is the
 >> heart
 >> of ab initio calculations) over STOs is much much more difficult than
 >> with
 >> GTOs. I think I?d trust his judgment in this area, as its one where
 >> he?s
 >> published a number of papers e.g.
 >> Ahmed Bouferguene 2005 J. Phys. A: Math. Gen. 38 2899-2916 ?Addition
 >> theorem of Slater type orbitals: a numerical evaluation of
 >> Barnett?Coulson/Löwdin functions?
 >> Historically evaluating GTO?s was faster, and while there are now
 >> claims
 >> that highly optimised STO codes can beat GTO codes, I remain to be
 >> convinced that highly optimised STO codes can beat highly optimised
 >> GTO
 >> codes.
 >> In the wider sense original question was about the popularity of
 >> Gaussian type orbital codes over STO based ones. Here I think the
 >> reason
 >> is also historical, and is based on the early popularity of the
 >> gaussian
 >> program as a collaborative venture among a number of theoretical
 >> chemists.  The use of GTO?s was, I think, first suggested by S.F.
 >> Boys, in Proc.
 >> Roy. Soc. A200, 542 (1950)
 >>
 >> ADF (using STO?s) has been around for a long time, and I know that the
 >> Zeigler group has made many substantial contributions to it.  It has
 >> not
 >> (yet) overtaken Gaussian in popularity, but maybe if( or when) it does
 >> we?ll be wondering what the advantages of GTO?s are. I?m not holding
 >> my
 >> breath.
 >> All the best
 >> Dr Laurence Cuffe> To send e-mail to subscribers of CCL put the
 string
 CCL: on your
 > Subject: line>
 > Send your subscription/unsubscription requests to:
 > CHEMISTRY-REQUEST{:}ccl.net HOME Page: http://www.ccl.net   | Jobs Page:
 > http://www.ccl.net/jobs
 > If your is mail bouncing from ccl.net domain due to spam filters,
 > please>
 -+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-
 > +-+
 >
 >
 >
 >
 Dr. S.J.A. van Gisbergen      
 Scientific Computing & Modelling NV
 Theoretical Chemistry, Vrije Universiteit
 De Boelelaan 1083
 1081 HV Amsterdam
 The Netherlands                
                
 vangisbergen{:}scm.com  
 http://www.scm.com
 Please note NEW FAX AND TELEPHONE NUMBERS:
 T: +31-20-5987626    
 F: +31-20-5987629