From oesterei@hrz.ba-freiberg.de Sat Jun 4 10:43:50 1994 Received: from orion.hrz.ba-freiberg.de for oesterei@hrz.ba-freiberg.de by www.ccl.net (8.6.4/930601.1506) id KAA05870; Sat, 4 Jun 1994 10:33:40 -0400 Message-Id: <199406041433.KAA05870@www.ccl.net> Received: by orion.hrz.ba-freiberg.de (16.6/15.6) id AA22291; Sat, 4 Jun 94 16:31:27 +0200 From: Ralf Oestereich Subject: Summary of GAMESS gradient problem! To: chemistry@ccl.net (Computer Chemistry List) Date: Sat, 4 Jun 94 16:31:27 MESZ Mailer: Elm [revision: 66.25] Dear netters, This is a summary of the responses to my question concerning GAMESS. To remind you, my question was why GAMESS terminates the calculation with an error like: Gradient vector to big. Thanks to everyone who helped us and spend time with our problem. You really help us! From evanseck@tammy.harvard.edu Tue May 31 14:04 MES 1994 Ralf, One large improvement would be to add d-functions on Si and Cl (3-21G*). Such a small and inflexible basis set needs the polarization functions. GAMESS may do this for you, but I can't tell from your message. Good Luck, J. D. Evanseck From steinke@sc.ZIB-Berlin.DE Tue May 31 14:59 MES 1994 Sehr geehrter Herr Oestereich, >...my problem ... bedeutet, dass der Gradient der Totalenergie bezueglich der Kernkoordnaten waehrend der Optimierung zu gross geworden ist (FMAXT-Parameter). > > As input with use coordinates which are optimzed with ALCHEMEY > (PC-Windows Version), and then simply translate them, so that > the first Silicon lies at 0/0/1.52. At least out input-file: > ... das scheint mit sehr problematisch zu sein: Kommen den vernueftige Strukturen fuer Si-Verbindungen mit ALCHEMY heraus? Mit welchen Kraftfeld rechnet ALCHEMY? Vielleicht ist es besser, mit einem semi-empirischen Hamilton-Operator (AM1, PM3) vorzuoptimieren (?!). Haben Sie es auch schon 'mal mit einer Dichtfunktionalrechnung (Programme wie DGauss oder DMol, haben wir auf unserer Cray) versucht? Falls es Sie interessiert, kann ich Ihnen Kontaktaddressen von Kollegen nennen, die auf diesem Gebiet auch Rechnungen durchfuehren. > > What must we change, to calculate the molecule? > Is our basis-set okay? > Why does GAMESS terminate, what does it mean, if the > gradient vector is to big? ... - neue und/oder "bessere" (im Sinne der qm-Niveaus) Startgeometrie - Basissatz fuer das erste Scannen der Potentialhyperflaeche OK From frj@dou.dk Tue May 31 15:01 MES 1994 Maybe your input is in bohr and the program expect them as angstrom? Otherwise, to force the calculation to proceed, change the FMAX in routine statpt.src and recompile this link. Frank From suter@physik.unizh.ch Tue May 31 15:20 MES 1994 Lieber Herr Oesterreicher, max. Gradient out of range bedeutet ganz einfach, dass die Geometrie so schlecht ist, dass man mit den so berechneten Gradienten keine Geometrieoptimierung machen kann. Was Sie einmal anschauen sollten im Output, welche Koordinaten da so schlecht sind. Vor der Linie die Sie uns geschickt haben ist der Output mit 3 Gradienten per Atom. Vielleicht schaffen sie es damit, die Geometrieoptimierung zu starten in dem Sie schon einmal das oder diese Atome in die angezeigte Richtung schieben. Versuchen Sie das vielleicht zuerst semiemprisch, PM3 sollte mit den Si-Dingern noch einigermassen zurechtkommen. mit freundlichen Gruessen Hans Ulrich Suter From djh@ccl.net Tue May 31 15:25 MES 1994 First, I would suggest defining a z-matrix for your system. I think that GAMESS will guess a reasonable hessian, based on some empirical rules, if internal coordinates are available. When only using cartesians GAMESS may be forced to start with a strictly diagonal hessian, so that the first optimization steps are merely steepest descent. If that fails, you might try an STO-3G calculation with an analytical hessian, and use the resulting geometry and hessian as input to the 3-21G run. It might also help to pick a few symmetric geometries and optimize them, then distort them just slightly and re-run without symmetry constraints. From chris@chep118.uni-bielefeld.de Tue May 31 15:32 MES 1994 Hallo Ralf, nach meiner Erfahrung mit GAMESS bei der Optimierung von Strukturen sollte man die mit einem Molekuel-Editor (wir benutzen HyperChem) erzeugten Strukturen zuerst mit einer kleinen Basis, oder besser mit einer semiempirischen Methode vorminimieren. Ich minimiere zuerst mit dem standard MM2+-Kraftfeld (HyperChem), danach mit der PM3-Methode (MOPAC, oder auch GAMESS). Erst danach benutze ich die abinitio Basissaetze (3-21G, dann 6-31G) zur Minimierung. Dieses Vorgehen ermoeglicht nicht nur eine fehlerfreie Minimierung (keine Aussetzer mit "Gradient to large"), sondern spart auch Rechenzeit. Man rechnet erst dann mit einer aufwendigere Methode weiter, wenn schon eine gut minimierte Struktur vorliegt. Ich hoffe ein wenig geholfen zu habe. ------------------------------------------------------------- From katzer@bkfug.kfunigraz.ac.at Tue May 31 15:45 MES 1994 > MethylChlor2-Sil-Sil-MethylChlor2 RHF/3-21G, Symmetrie C1 > C1 > SILICON 14.0 0.0000 0.0000 0.8094 > SILICON 14.0 2.2068 0.2512 -0.7219 > CHLOR 17.0 -0.4898 -0.7698 -0.6029 > CARBON 6.0 -0.0041 -0.4822 1.8261 > CARBON 6.0 2.2029 0.9573 -2.4745 > CHLOR 17.0 2.7256 0.9298 0.0391 > HYDROGEN 1.0 1.6782 1.9252 -2.4824 > HYDROGEN 1.0 1.6753 0.2673 -3.1496 > HYDROGEN 1.0 3.2307 1.105 -2.843 > HYDROGEN 1.0 -1.0186 -0.6726 2.1871 > HYDROGEN 1.0 0.4482 0.3304 2.421 > HYDROGEN 1.0 0.6092 -1.3903 1.9685 > CHLOR 17.0 -0.5346 0.9436 -0.1505 > CHLOR 17.0 2.7156 -0.7913 -0.6104 Hallo Ralf! Ich habe grosze Zweifel an diesen Koordinaten. Schau sie Dir einmal mit dem MOLPLT an! Abgesehen vom Si-Si-Abstand, sind fast alle Bindungen viel zu kurz: Die kurzeste Si-Cl Bindung ist gar nur 1.14 A (2-6) lang! Auch die Si-C Bindung mit 1.13 A (1-4) ist rekordverdaechtig. Letztlich ist auch der Cl-Cl-Abstand von 1.77 A (3-13) bemerkenswert (im Cl2 ist es, glaub' ich, knapp unter 2 A). Auszerdem stimmen die Winkel am Si nicht so recht. Ich wuerde mal die Herkunft dieser Koordinaten genau ueberpruefen! Viel Erfolg, Gernot From windus@mercury.chem.nwu.edu Tue May 31 18:40 MES 1994 Dear Ralf, I am responding to your question on CCL about GAMESS. Since the gradient is that large (>3!), it means that the starting geometry is VERY bad (at least for the level of theory tried). I ran an EXETYP=CHECK run and looked at the internuclear distances. The Si-Si distance is almost 3 Angstroms (a "normal" Si-Si single bond is about 2.4). The Cl's seem to be very close together. I used MOLPLT (the graphics package that comes with GAMESS) and found that the molecule is very "strange". The geometries around the Si's are rather flat and one of the methyl groups looks like it is trying to leave the molecule. I would suggest bringing the Si's closer together, pyramidalizing the other substituents around the Si's and giving "reasonable" distances for the substituents to Si. In other words, I think the program you used must have given you a bad geometry. It might be easier to just put in a guess by hand. Hope this helps. From hideaki@si.fi.ameslab.gov Tue May 31 18:51 MES 1994 > > Dear netters, > > we have the following problem with GAMESS (US): > > While calculating the molecule: > > It means the initial geometry is too bad to optimize. Particulary you should be careful about bond length. In this case, I found the Si-Cl bond lengths are too short. It's standard length is somewhere around 2.1 angstrom. So, the following input will work. $data 1,1-dimethyltetrachlorodisilane c1 Si Si 1 2.4 C 1 1.88 2 110.0 C 2 1.88 1 110.0 3 180.0 Cl 1 2.1 2 110.0 3 120.0 Cl 1 2.1 2 110.0 3 -120.0 Cl 2 2.1 1 110.0 4 120.0 Cl 2 2.1 1 110.0 4 -120.0 H 3 1.1 1 110.0 2 180.0 H 3 1.1 1 110.0 9 120.0 H 3 1.1 1 110.0 9 -120.0 H 4 1.1 2 110.0 1 180.0 H 4 1.1 2 110.0 12 120.0 H 4 1.1 2 110.0 12 -120.0 $end Of course you need to put COORD=ZMT in $contrl group and I think use NZVAR=36 is a good idea. Also if you are seeking global minimum, starting another optimization with torsion of C-Si-Si-C set 60 is necessary. Hideaki Shimizu From jan@si.fi.ameslab.gov Tue May 31 21:06 MES 1994 Dear Ralf: A very large gradient is usually indicative of a bad starting geometry. I recommend that you re-check your input structure -- either graphically or by looking at the distance matrix in the GAMESS output. Best regards, Jan Jensen From frits@rulglj.LeidenUniv.nl Wed Jun 1 10:46 MES 1994 Why don't you use a good Z-matrix? Then you can see which bonds/angles/torsions run away from the optimum. The molecule doesn't seem too complicated. It means, that your optimization diverges; because the target of a gradient optimization is to find the stable point (or one of the stable extrema) where the gradient vector is zero. So the program notices that something is very wrong and terminates, when it sees the gradient increases abnormally. BTW what version of GAMESS do you use??? I've never seen such a weird input file :-) Is it the GAMESS-US version? I haven't looked at your problem more closely because I don't have the time but I wish you good luck with it. Greetings, Frits From raji@yu1.yu.edu Thu Jun 2 19:37 MES 1994 I read your message on the computaional network. I seem to encounter similar problems with gamess when I try to optimize geometries of some clusters. I was wondering if you received any useful responses. Please reply to raji@yu1.yu.edu. Thank you for your help. Raji From oesterei@hrz.ba-freiberg.de Sat Jun 4 11:43:59 1994 Received: from orion.hrz.ba-freiberg.de for oesterei@hrz.ba-freiberg.de by www.ccl.net (8.6.4/930601.1506) id KAA05909; Sat, 4 Jun 1994 10:49:17 -0400 Message-Id: <199406041449.KAA05909@www.ccl.net> Received: by orion.hrz.ba-freiberg.de (16.6/15.6) id AA22385; Sat, 4 Jun 94 16:47:01 +0200 From: Ralf Oestereich Subject: PLEASE READ! LOOKING FOR HELP! To: chemistry@ccl.net (Computer Chemistry List) Date: Sat, 4 Jun 94 16:47:01 MESZ Mailer: Elm [revision: 66.25] Dear netters! Now for something completely different: I am looking for a chance to study for a while outside Germany. I am now in the 6th semestre, so that i want to leave at the end (September or October) of 1995. The problem is, that i dont have a DAAD or Fullbright stipendium right now. So my question is, if there is anybody out there, who can help me? This is of course not related to computer chemistry. I know that this sounds a little bit strange, but please remember when you are young (:-))))))) ), and try to help me! Please send help to oesterei@orion.hrz.ba-freiberg.de From toukie@zui.unizh.ch Sat Jun 4 16:43:59 1994 Received: from rzusuntk.unizh.ch for toukie@zui.unizh.ch by www.ccl.net (8.6.4/930601.1506) id PAA07537; Sat, 4 Jun 1994 15:47:10 -0400 Received: from rzurs5.unizh.ch by rzusuntk.unizh.ch (4.1/SMI-4.1.9) id AA02963; Sat, 4 Jun 94 21:47:06 +0200 Received: by rzurs5.unizh.ch (AIX 3.2/UCB 5.64/4.03) id AA28258; Sat, 4 Jun 1994 21:47:04 +0200 From: toukie@zui.unizh.ch (Hr Dr. S. Shapiro) Message-Id: <9406041947.AA28258@rzurs5.unizh.ch> Subject: Use of MOPAC 6.0 for isotopically-substituted molecules To: chemistry@ccl.net Date: Sat, 4 Jun 1994 21:47:04 +0100 (MEST) Cc: toukie@zui.unizh.ch X-Mailer: ELM [version 2.4 PL13] Content-Type: text Content-Length: 1725 Dear colleagues; I am a user of semi-empirical quantum mechaical techniques rather than a quantum mechanist, so perhaps someone will be kind enough to respond to this request for information. (i) I would like to use the AM1 and PM3 modules of MOPAC 6.0 in order to compare the properties of hydrogen isotope-substituted molecules. I construct- ed Z-matrices for three small molecular weight organic molecules which differ- ed only in that one contained only protium, another contained only deuterium, and the third contained only tritium. I ran these in MOPAC under the following conditions: t=72000 uhf precise am1 nointer ef The final outputs for all three molecules were identical EXCEPT FOR THEIR MOLE- CULAR WEIGHTS (which were as they should have been). Are MOPAC calculations too insensitive to the impact of deuterio- and tritio-substitution to be of value, or are there some "tricks" which I must use in order to extract the re- quired information? (ii) In NOT TOO TOO technical terms, what are the definitions of total energy electronic energy core-core repulsion sz s**2 Why is it that for some molecules the ionisation potential (-1 x homo) corres- ponds to the alpha eigenvalues and for other molecules it corresponds to the beta eigenvalues? Why are the alpha and beta eignevalues identical for some molecules and different for other molecules? Thanks in advance to all respondents. Sincerely, S. Shapiro Internet: toukie@zui.unizh.ch