From owner-chemistry@ccl.net Sat Nov 10 15:45:00 2012 From: "Ivo Jacobs jacobs.ivo^gmail.com" To: CCL Subject: CCL:G: Strange energies for Gaussian IRC calculation Message-Id: <-47873-121110154238-27729-u3Hc3EtKhDVj11qsMcg+DA[a]server.ccl.net> X-Original-From: "Ivo Jacobs" Date: Sat, 10 Nov 2012 15:42:37 -0500 Sent to CCL by: "Ivo Jacobs" [jacobs.ivo[a]gmail.com] Hello, I have started to do some IRC calculations with Gaussian '03, but it gives me very strange results. I have previously located a transition state, done the frequency analysis, and it's a true saddle point, energy is -1706 Hartree. For the IRC calculation, my input looks like this: %NProcShared=4 %NProcLinda=1 %mem=2000MB %chk=checkpnt.chk #P TEST IRC=(FORWARD,RCFC) B3LYP CHKBASIS GEOM=CHECKPOINT GUESS=READ Title 0 1 This initially seems to work very slowly (SCF needs more than 40 cycles to converge, which is much more than I have typically seen before), but more problematically, the energies are wildly different (-3127 Hartree on the first cycle, same order of magnitude for the later ones). If I do a single point energy starting from the original transition state checkpoint file, the SCF finishes in one cycle with -1706 Hartree, so the problem is not in the checkpoint file. I've tried to repeat the IRC calculation with CalcAll or with CalcFC, and this gives the same problem. Also without using "Guess=Read" the same result is obtained. It looks to me like the initial reading of the guess fails for some reason. Anybody any idea what is going on here? Thanks in advance, Ivo Jacobs