From owner-chemistry@ccl.net Mon Mar 23 14:14:01 2009 From: "Johannes Hachmann jh388{}cornell.edu" To: CCL Subject: CCL:G: (correction) Energy change upon swapping two orbitals in Gaussian 98 Message-Id: <-38909-090323110805-15657-NH8x1V7EVdyEWGFmzTqLiw#%#server.ccl.net> X-Original-From: "Johannes Hachmann" Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset="us-ascii" Date: Mon, 23 Mar 2009 11:07:47 -0400 MIME-Version: 1.0 Sent to CCL by: "Johannes Hachmann" [jh388() cornell.edu] Dear Yutao, > It seems that this SCF is extremely sensitive, any minor > change is sufficient to lead to a different solution. But I'm > very curious what could be the reason? The iron? At least the > frontier orbital energies don't seem to be very close: Transition metal calcs are a bit tricky - you could argue that the problems one finds with multiple electronic minima is a reflection of the great chemical diversity of transition metal compounds. That's why it is in my eyes particularly important to double check every output one gets with a lot of scrutiny for potential errors - they can just occur so easily... > I didn't use Guess=Mix and the symmetry doesn't seem to be a > problem, since the structure is highly asymmetric and the > point group is actually C1 through all calculations. Be careful here: The Guess=Mix not only breaks spatial symmetry but in particular also spin symmetry, which is important if you want to get a proper unrestricted solution! Best Johannes --------------------------------------------------------------- Johannes Hachmann Chan Research Group 140D Baker Laboratory Department of Chemistry and Chemical Biology Cornell University Ithaca, NY 14853-1301 USA --------------------------------------------------------------- > -----Original Message----- > From: owner-chemistry+jh388==cornell.edu]~[ccl.net > [mailto:owner-chemistry+jh388==cornell.edu]~[ccl.net] On Behalf > Of Yutao Yue yutao.yue%x%gmail.com > Sent: Sunday, March 22, 2009 7:22 PM > To: Hachmann, Johannes > Subject: CCL:G: (correction) Energy change upon swapping two > orbitals in Gaussian 98 > > Thanks dear friend, I've done what you suggested. The G98 job > just got executed and finished, and arrived at a result > that's almost identical to the swapped result (with lower energy)! > > I didn't use Guess=Mix and the symmetry doesn't seem to be a > problem, since the structure is highly asymmetric and the > point group is actually C1 through all calculations. > > What also happened is that, if I use the swapped result as > initial guess, to do exactly the same SCF calculation using > Guess=Read (NOT swapping back the orbitals), I arrived at > another solution with an energy higher than the "swapped > solution" but lower than the original solution. Ideally this > should not happen because it starts with a good initial guess > (with lower energy) and finishes at a bad solution (higher energy). > > It seems that this SCF is extremely sensitive, any minor > change is sufficient to lead to a different solution. But I'm > very curious what could be the reason? The iron? At least the > frontier orbital energies don't seem to be very close: > > Energy E=-3295.07657288 > NBO Fe Q=1.46819 S=3.67069 > S**2 guess= 6.0137 6.0163 > > 166 -0.02694 > 167 0.00934 > Alpha HOMO is MO # 167 > 168 0.01592 > > 161 -0.02639 > 162 0.00532 > 163 0.01207 > Beta HOMO is MO # 163 > 164 0.01835 > > > Best, > Yutao > > > On Fri, Mar 13, 2009 at 3:47 AM, Johannes Hachmann jh388 a > cornell.edu wrote: > > > > Sent to CCL by: "Johannes Hachmann" [jh388/a\cornell.edu] > Dear Yutao, > > it's hard to say what exactly is going on without a > closer look, but if the > calc doesn't take too long, try something like > > # UBLYP/6-31g(d) POP=FULL GFPRINT GFOLDPRINT SYMM=LOOSE > GUESS=READ > INT=ULTRAFINE STABLE=OPT > > That will take care of potential threshold and > instability issues and you > can compare the result to your swap result. Did you use > GUESS=MIX in your > very first calc? If not, a high symmetry guess might > have led you to an > instability. I am however a bit surprised that a swap > within the occupied > orbs helped... > > Best > > Johannes > > --------------------------------------------------------------- > Johannes Hachmann > > Chan Research Group > 140D Baker Laboratory > Department of Chemistry and Chemical Biology > Cornell University > Ithaca, NY 14853-1301 > USA > --------------------------------------------------------------- > > > -----Original Message----- > > From: owner-chemistry+jh388==cornell.edu-,-ccl.net > > [mailto:owner-chemistry+jh388 > ==cornell.edu-,-ccl.net] On Behalf > > Of Yutao Yue yutao.yue:-:gmail.com > > Sent: Thursday, March 12, 2009 3:16 PM > > To: Hachmann, Johannes > > Subject: CCL:G: (correction) Energy change upon swapping two > > orbitals in Gaussian 98 > > > > Thank you very much Johannes, I did swap the two highest > > OCCUPIED beta orbitals. I was also doubting that the original > > solution was not stable. I compared the solutions before and > > after swapping, the MO compositions are almost identical for > > frontier orbitals, and only some minor differences for some > > deeper orbitals. Plus, the S2 values, the NBO charges and > > spins of some characteristic atoms (metal, etc) are almost > > exactly the same, up to quite a few digits. > > > > On the other hand, although with highly similar MO > > compositions, the MO energy themselves are fairly different > > through all MOs including the deepest, e.g., the difference > > for a single MO is on the order of 0.01 a.u.. > > > > By the way, I used the default SCF convergence level (which > > is equivalent to conver=4). But I guess that's far from > > introducing errors of this magnitude right? > > > > I wonder is it possible that Gaussian does something > > artificial (and arbitrary? because in one case energy went up > > and the other case it went down) to the MO energies, once the > > initial guess swapping is introduced? > > > > Regards, > > Yutao > > > > > > On Thu, Mar 12, 2009 at 12:50 PM, Johannes Hachmann > > jh388~!~cornell.edu wrote: > > > > > > > > Sent to CCL by: "Johannes Hachmann" > [jh388]*[cornell.edu] > > Dear Yutao, > > > > sounds to me like you might have landed on an > > electronic instability the > > first time. SCF for (transition-)metal compounds can be > > very hard to > > converge. Try STABLE=OPT starting from your original > > calc without the > > orbital swap and see whether that lowers the energy (in > > that case, your swap > > has introduced some mixing/symmetry breaking and led to > > a more stable > > state). You can check the S2 value of your two results > > to see, whether your > > state has changed. You might also want to check your > > SCF convergence > > threshold and DFT grid (although it should not make a > > difference of 20mH). > > > > Generally, swapping the highest two HOMOs should not > > change the energy at > > all, but maybe you ment the HOMO-LUMO pair? A vanishing > > HOMO-LUMO gap > > indicates a system with nondynamic correlation, and you > > might not want to > > use a single reference method like DFT in the > first place. > > > > Hope that helps. Best > > > > Johannes > > > > > --------------------------------------------------------------- > > Johannes Hachmann > > > > Chan Research Group > > 140D Baker Laboratory > > Department of Chemistry and Chemical Biology > > Cornell University > > Ithaca, NY 14853-1301 > > USA > > > --------------------------------------------------------------- > > > > > > > > > > > -----Original Message----- > > > From: owner-chemistry+jh388==cornell.edu-.-ccl.net > > > [mailto:owner-chemistry+jh388 > > > > ==cornell.edu-.-ccl.net] > On Behalf > > > Of Yutao Yue Yutao.Yue#,#gmail.com > > > Sent: Thursday, March 12, 2009 10:55 AM > > > To: Hachmann, Johannes > > > Subject: CCL:G: (correction) Energy change > upon swapping two > > > orbitals in Gaussian 98 > > > > > > > > > Sent to CCL by: "Yutao Yue" [Yutao.Yue- > -gmail.com] (sorry > > > the energy changes are 0.02 a.u., NOT 0.2 a.u.) > > > > > > Dear All, > > > > > > I have a successful G98 single point calculation > > > (UBLYP/6-31G*) on a metal-containing system > of around 70 > > > atoms. I use the result MOs as initial guess, > but swapped the > > > two highest occupied beta orbitals (which are > supposed to be > > > highly competing with each other) upon > reading in the initial > > > guess, to do another G98 single point > calculation on the same > > > system. I then arrived at a solution that, > the frontier > > > orbitals are almost identical to the original solution > > > including the ordering of the beta orbitals, > while the single > > > point energy is 0.02 a.u. lower! > > > > > > Similar calculations was done with > UBPW91/6-31G*, and again > > > the orbitals go back to original ordering, > but the single > > > point energy is 0.02 a.u. higher now! > > > > > > Anybody have any idea about those energy changes? > > > > > > Thanks a lot and wish everybody a nice day! > > > > > > -Yutao Yue > > > > > > > > > > > > -= This is automatically added to each > message by the mailing > > > script =- To recover the email address of the > author of the > > > message, please change the strange characters > on the top line > > > to the -.- sign. You can also look up the > > X-Original-From: line > > > > > in the mail header.> Conferences: > > > > > > http://server.ccl.net/chemistry/announcements/conferences/> > > > > > > > > > > > -= This is automatically added to each message by the > > mailing script =- > > > > > > > > E-mail to subscribers: CHEMISTRY ~ ccl.net or use:> > > E-mail to administrators: CHEMISTRY-REQUEST ~ > ccl.net or use> Conferences: > > http://server.ccl.net/chemistry/announcements/conferences/ > > > > Search Messages: > > http://www.ccl.net/chemistry/searchccl/index.shtml> > > > > > > > > > > > > > > -= This is automatically added to each message by the > mailing script =- > > > > E-mail to subscribers: CHEMISTRY!=cl.net or use:> > E-mail to administrators: CHEMISTRY-REQUEST!=cl.net or use> Conferences: > http://server.ccl.net/chemistry/announcements/conferences/ > > Search Messages: > http://www.ccl.net/chemistry/searchccl/index.shtml> > > > > >