From owner-chemistry@ccl.net Sun Jan 21 13:56:01 2007 From: "Jozsef Csontos jozsefcsontos^^^creighton.edu" To: CCL Subject: CCL:G: CCL question regarding comparison of Jaguar and Gaussian03 programs Message-Id: <-33418-070121130124-23998-dK6ZOgXrKFYtM/HQwwWu6Q%x%server.ccl.net> X-Original-From: Jozsef Csontos Content-Transfer-Encoding: 7bit Content-Type: text/plain Date: Sun, 21 Jan 2007 11:25:22 -0600 Mime-Version: 1.0 Sent to CCL by: Jozsef Csontos [jozsefcsontos]*[creighton.edu] Another issues are that Jaguar uses the pseudospectral method (PS) as default (Friesner, Ann. Rev. Phys. Chem. 1991,42,341; Friesner, Chem. Phys. Lett. 1985, 116, 39) and some dft functionals defined differently (for example, BHandH and BHandHLYP are such functionals). However, you can turn the PS off using the nops=1 keyword and define functionals which are equivalent to those in Gaussian (look for the idft keyword in the manual). I don't know what significantly different means in your Lab :)) but you can reproduce the Gaussian results within a few tenth of kcal/mol. I know that it goes without saying, but be aware that the above is valid for the gas-phase HF and DFT calculations only. The solvation models are considerably different and the other Jaguar methods are not available in Gaussian. The folks at the Jaguar help desk are very helpful and fast, indeed. Best wishes, Jozsef On Fri, 2007-01-19 at 11:02 -0500, david.giesen__kodak.com wrote: > Sent to CCL by: david.giesen],[kodak.com > Ken - > > You'll probably never get complete agreement between the two programs as > they use different grids and methods for their DFT calculations. Also, be > aware that Jaguar uses 6D functions while Gaussian sometimes uses 5. > However, there is another issue that causes differences on the order of a > few kcals/mol. This explanation was sent to me by folks at > help[]schrodinger.com many years ago (they are very helpful on matters like > this, BTW). > > Before solving the SCF equations, Jaguar removes eigenvectors of the > overlap matrix that are smaller than a certain tolerance. Doing this > increases the stability of the SCF convergence, but alters the final > energy obtained. You can tell if this happens during a Jaguar job by > comparing two lines in the Jaguar output file. > > Look for "number of basis functions" near the top of the file, and compare > that number to "number of canonical orbitals" printed throughout the > output file. If the two numbers are the same, then something else is > afoot and I can't help you. If the two numbers are different, then Jaguar > has removed some of those small eigenvectors. In that case, look for the > line that says "smallest eigenvalue of S". There will be a number listed > there that is something like 1.0E-3, or whatever it is for your system. > You'll need that number. > > You'll want to add two keywords to the input file: cut10=x and cut20=x, > where x < the smallest eigenvalue of S. cut10 and cut20 set the tolerance > for removing eigenvalues of the overlap matrix. Once you've set them, you > should see the number of canonical orbitals match up with the number of > basis functions. > > In practice, it's a pain to run a job, change the keywords and restart the > job all the time. To avoid this, I always set cut10 and cut20 to 5.0e-4 > > from the start and rarely do I have SCF convergence or eigenvector removal > problems. > > Dave Giesen > > > > > "Hunter, Ken ken.hunter^_^uleth.ca" > Sent by: owner-chemistry[]ccl.net > 01/18/2007 03:07 PM > Please respond to > "CCL Subscribers" > > > To > "Giesen, David J " > cc > > Subject > CCL:G: CCL question regarding comparison of Jaguar and Gaussian03 programs > > > > > > > We have been testing the applications of the Jaguar program and would like > to incorporate the versatility of Jaguar to our computational chemistry > group. However, one hesitation remains. When testing energies from our > standard test molecules the energies we get from Jaguar are significantly > different from our test calculations in Gaussian. Looking at the > convergence criteria appears to be the same between the two programs and > they have defined the basis sets the same. Differences seem to arise in > optimizations and single points. We were curious if there any insights on > how to reproduce Gaussian energies with Jaguar. Any assistance would be > greatly appreciated. > > Ken Hunter> > > -- Jozsef Csontos, Ph.D. Department of Biomedical Sciences Creighton University, Omaha, NE From owner-chemistry@ccl.net Sun Jan 21 21:14:00 2007 From: "Sengen Sun sengensun++yahoo.com" To: CCL Subject: CCL:G: question regarding comparison of Jaguar and Gaussian03 programs Message-Id: <-33419-070121201334-31644-N7CdpzYbA2BowA1QP3ZusQ=server.ccl.net> X-Original-From: Sengen Sun Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset=iso-8859-1 Date: Sun, 21 Jan 2007 16:13:23 -0800 (PST) MIME-Version: 1.0 Sent to CCL by: Sengen Sun [sengensun++yahoo.com] I would emphasize that you can only use relative energies (e.g., rotational barrier, activation energy...) for the comparison of different programs. > It is meaningless to compare the total energies of a molecule from different programs as we all know that absolute value of potential energy does not exist. > I have some experience with Titan that uses the Jaguar code. It repeats well the data (e.g., activation energies of Diels-Alder reactions) from other major programs with errors less than 0.1 kcal/mol. > Sengen --- Jozsef Csontos wrote: > > > Another issues are that Jaguar uses the > pseudospectral method (PS) as > default (Friesner, Ann. Rev. Phys. Chem. > 1991,42,341; > Friesner, Chem. > Phys. Lett. 1985, 116, 39) and some dft functionals > defined differently > (for example, BHandH and BHandHLYP are such > functionals). > However, you can turn the PS off using the nops=1 > keyword and define > functionals which are equivalent to those in > Gaussian > (look for the idft > keyword in the manual). > I don't know what significantly different means in > your Lab :)) but you > can reproduce the Gaussian results within a few > tenth > of kcal/mol. I > know that it goes without saying, but be aware that > the above is valid > for the gas-phase HF and DFT calculations only. The > solvation models are > considerably different and the other Jaguar methods > are not available in > Gaussian. > The folks at the Jaguar help desk are very helpful > and fast, indeed. > Best wishes, > Jozsef > On Fri, 2007-01-19 at 11:02 -0500, > david.giesen__kodak.com wrote: > > Sent to CCL by: david.giesen],[kodak.com > > Ken - > > > > You'll probably never get complete agreement > between the two programs as > > they use different grids and methods for their > DFT > calculations. Also, be > > aware that Jaguar uses 6D functions while > Gaussian > sometimes uses 5. > > However, there is another issue that causes > differences on the order of a > > few kcals/mol. This explanation was sent to me > by > folks at > > help[]schrodinger.com many years ago (they are > very > helpful on matters like > > this, BTW). > > > > Before solving the SCF equations, Jaguar removes > eigenvectors of the > > overlap matrix that are smaller than a certain > tolerance. Doing this > > increases the stability of the SCF convergence, > but > alters the final > > energy obtained. You can tell if this happens > during a Jaguar job by > > comparing two lines in the Jaguar output file. > > > > Look for "number of basis functions" near the top > of the file, > and compare > > that number to "number of canonical orbitals" > printed throughout > the > > output file. If the two numbers are the same, > then > something else is > > afoot and I can't help you. If the two numbers > are > different, then Jaguar > > has removed some of those small eigenvectors. In > that case, look for the > > line that says "smallest eigenvalue of S". There > will be a > number listed > > there that is something like 1.0E-3, or whatever > it > is for your system. > > You'll need that number. > > > > You'll want to add two keywords to the input > file: > cut10=x and cut20=x, > > where x < the smallest eigenvalue of S. cut10 > and > cut20 set the > tolerance > > for removing eigenvalues of the overlap matrix. > Once you've set them, you > > should see the number of canonical orbitals match > up with the number of > > basis functions. > > > > In practice, it's a pain to run a job, change the > keywords and restart the > > job all the time. To avoid this, I always set > cut10 and cut20 to 5.0e-4 > > > from the start and rarely do I have SCF > convergence or eigenvector > removal > > problems. > > > > Dave Giesen > > > > > > > > > > "Hunter, Ken ken.hunter^_^uleth.ca" > > > Sent by: owner-chemistry[]ccl.net > > 01/18/2007 03:07 PM > > Please respond to > > "CCL Subscribers" > > > > > > To > > "Giesen, David J " > > cc > > > > Subject > > CCL:G: CCL question regarding comparison of > Jaguar > and Gaussian03 programs > > > > > > > > > > > > > > We have been testing the applications of the > Jaguar > program and would like > > to incorporate the versatility of Jaguar to our > computational chemistry > > group. However, one hesitation remains. When > testing energies from our > > standard test molecules the energies we get from > Jaguar are significantly > > different from our test calculations in Gaussian. > > Looking at the > > convergence criteria appears to be the same > between > the two programs and > > they have defined the basis sets the same. > Differences seem to arise in > > optimizations and single points. We were curious > if there any insights on > > how to reproduce Gaussian energies with Jaguar. > Any assistance would be > > greatly appreciated. > > > > Ken Hunter> > > > > > -- > Jozsef Csontos, Ph.D. > Department of Biomedical Sciences > Creighton University, > Omaha, NE > > > > > > ____________________________________________________________________________________ > Never miss an email again! > Yahoo! Toolbar alerts you the instant new Mail > arrives. > http://tools.search.yahoo.com/toolbar/features/mail/ > ____________________________________________________________________________________ Bored stiff? Loosen up... Download and play hundreds of games for free on Yahoo! Games. http://games.yahoo.com/games/front