From owner-chemistry@ccl.net Mon Nov 30 16:40:01 2009 From: "Ronald C Bakus rbakus(-)chem.ucsb.edu" To: CCL Subject: CCL: Questions about linear dependence and basis set Message-Id: <-40812-091130155308-20786-QbKvKjnjDuqGouNayjWc8g^^server.ccl.net> X-Original-From: "Ronald C Bakus" Date: Mon, 30 Nov 2009 15:53:04 -0500 Sent to CCL by: "Ronald C Bakus" [rbakus%chem.ucsb.edu] One suggestion for solving the issues I was having with SCF convergence (as stated before, I have already tried QC without avail) in a distyrylbenzene type molecule was to check if there was linear dependence, and if there was a linear dependence issue, to switch to a 'small center diffuse basis set'. I actually have no idea how to check if there is a linear dependence issue (though during SCF I do see line in the log about 'Large coefficient', see below), and was hoping for suggestions to check if that is the case. I have been using 6-31+g(d,p), could someone suggest a 'small center diffuse basis set' with similar computational efficiency (I would like to avoid more than doubling my wall clock time) that I might use that wont suffer from this issue. In addition, would changing to a different hybrid functional (I have been using B3LYP) have any effect on the linear dependence? Finally, would changing the value of IOP(3/59) (threshold for throwing away eigenvectors of S) possibly be a solution to this, or is there a way to adjust the overall accuracy of the integrals to deal with linear dependence? Sample line from log file ---- Cycle 17 Pass 1 IDiag 1: RMSU= 1.43D-07 CP: 1.01D+00 1.03D+00 7.95D-01 7.10D-01 6.27D-01 CP: 5.16D-01 2.87D-01 -1.25D+00 -1.11D+00 3.00D+00 CP: 3.00D+00 1.48D+00 8.38D-01 1.67D+00 9.92D-01 CP: 3.00D+00 E= -1117.39107591657 Delta-E= -0.000000006455 Rises=F Damp=F DIIS: error= 6.19D-08 at cycle 17 NSaved= 17. NSaved=17 IEnMin=17 EnMin= -1117.39107591657 IErMin=16 ErrMin= 4.33D-08 ErrMax= 6.19D-08 EMaxC= 1.00D-01 BMatC= 2.71D-13 BMatP= 1.39D-13 IDIUse=1 WtCom= 1.00D+00 WtEn= 0.00D+00 Large coefficients: NSaved= 17 BigCof= 0.00 CofMax= 10.00 Det=-1.85D-15 Inversion failed. Reducing to 16 matrices. Large coefficients: NSaved= 16 BigCof= 0.00 CofMax= 10.00 Det=-1.86D-15 Inversion failed. Reducing to 15 matrices. Coeff-Com: -0.707D-04-0.137D-03-0.287D-03-0.283D-03 0.128D-02 0.144D-01 Coeff-Com: 0.486D-01 0.119D+00 0.387D+00-0.637D+00 0.182D-02 0.708D+00 Coeff-Com: 0.180D+01-0.301D+01 0.156D+01 Coeff: -0.707D-04-0.137D-03-0.287D-03-0.283D-03 0.128D-02 0.144D-01 Coeff: 0.486D-01 0.119D+00 0.387D+00-0.637D+00 0.182D-02 0.708D+00 Coeff: 0.180D+01-0.301D+01 0.156D+01 Gap= 0.111 Goal= None Shift= 0.000 RMSDP=8.11D-07 MaxDP=5.36D-05 DE=-6.45D-09 OVMax= 1.30D-07 ---- Thanks in advance for any suggestions. Ron Bakus rbakus*at*chem.ucsb.edu University of California-Santa Barbara From owner-chemistry@ccl.net Mon Nov 30 18:13:00 2009 From: "Ol Ga eurisco1:-:pochta.ru" To: CCL Subject: CCL:G: Questions about linear dependence and basis set Message-Id: <-40813-091130181202-29950-sp8qHizWoyj0ae9XX6Nb4g|-|server.ccl.net> X-Original-From: "Ol Ga" Date: Mon, 30 Nov 2009 18:11:58 -0500 Sent to CCL by: "Ol Ga" [eurisco1===pochta.ru] Dear Ronald C. Bakus, I used geometry posted by you earlier but I changed slightly route - I changed method to unrestricted (Ub3lyp). Also, I recommend to add NoSymm. I observed smooth SCF-convergence before XQC began to performe. Cycle 15 Pass 1 IDiag 1: RMSU= 6.17D-09 CP: 6.72D-01 7.68D-01 9.51D-01 5.80D-01 7.61D-01 CP: 7.40D-01 9.08D-01 9.12D-01 8.16D-01 8.60D-01 CP: 8.93D-01 8.18D-01 7.10D-01 8.90D-01 E= -1117.18461578981 Delta-E= 0.000000000079 Rises=F Damp=F DIIS: error= 1.15D-08 at cycle 15 NSaved= 15. NSaved=15 IEnMin=13 EnMin= -1117.18461578993 IErMin=15 ErrMin= 1.15D-08 ErrMax= 1.15D-08 EMaxC= 1.00D-01 BMatC= 2.74D-13 BMatP= 2.88D-12 IDIUse=1 WtCom= 1.00D+00 WtEn= 0.00D+00 Coeff-Com: -0.140D-06 0.330D-06 0.303D-05 0.392D-05 0.829D-05 0.102D-04 Coeff-Com: -0.102D-03-0.495D-03-0.180D-02-0.238D-02 0.363D-02 0.175D-01 Coeff-Com: 0.164D+00 0.285D+00 0.535D+00 Coeff: -0.140D-06 0.330D-06 0.303D-05 0.392D-05 0.829D-05 0.102D-04 Coeff: -0.102D-03-0.495D-03-0.180D-02-0.238D-02 0.363D-02 0.175D-01 Coeff: 0.164D+00 0.285D+00 0.535D+00 Gap= 0.111 Goal= None Shift= 0.000 Gap= 0.111 Goal= None Shift= 0.000 RMSDP=4.49D-09 MaxDP=2.30D-07 DE= 7.87D-11 OVMax= 1.03D-07 SCF Done: E(UB3LYP) = -1117.18461579 A.U. after 15 cycles Convg = 0.4487D-08 -V/T = 2.0098 = 0.0000 = 0.0000 = 0.0000 = 0.0000 S= 0.0000 = 0.000000000000E+00 KE= 1.106300942892D+03 PE=-6.771451539404D+03 EE= 2.455058951308D+03 Annihilation of the first spin contaminant: S**2 before annihilation 0.0000, after 0.0000 DiagDN has N= 672 LTot= 2048 but NE2= 3 cannot use DSYEVD. Leave Link 502 at Tue Dec 1 08:49:35 2009, MaxMem= 131072000 cpu: 7983.5 QCSCF skips out because SCF is already converged. What revision of Gaussian do you use? It is smart software. Thus, linear dependence can be observed very rarely. Sincerely, Ol Ga From owner-chemistry@ccl.net Mon Nov 30 18:52:01 2009 From: "Soren Eustis soren.eustis~~env.ethz.ch" To: CCL Subject: CCL: Global Minima Message-Id: <-40814-091130181613-31471-sF/Z4oayBIqknzUNeK1vzg#%#server.ccl.net> X-Original-From: "Soren Eustis" Date: Mon, 30 Nov 2009 18:16:09 -0500 Sent to CCL by: "Soren Eustis" [soren.eustis() env.ethz.ch] I am curious about CCL users and their experience with global minima programs. I would like to find a program that would provide simulated annealing or a similar method to provide ab initio, MM, or semi-empirical global minimum structures. I have found a lot of really interesting information on the subject, but it seems to be a lot of code, with few binaries available. If I start monkeying with too much code, I will get lost in the details. Ideally, a program would be able to interface with G09, or GAMESS - but I could imagine simply using a MM or semi-emp. program to find candidate global minimum structures that I would then use as a starting point for DFT or MP2 optimizations. Looking forward to hearing everyone's ideas. Regards, Soren Eustis soren]_[env.ethz.ch